nature biotechnology

Review article https://doi.org/10.1038/s41587-024-02320-1

Epigenome editing technologies for

discovery and medicine

Received: 14 November 2023 Sean R. McCutcheon1,2, Dahlia Rohm1,2, Nahid Iglesias1,2 &
Charles A. Gersbach 1,2
Accepted: 19 June 2024

Published online: 29 July 2024

Epigenome editing has rapidly evolved in recent years, with diverse
Check for updates applications that include elucidating gene regulation mechanisms,
annotating coding and noncoding genome functions and programming
cell state and lineage specification. Importantly, given the ubiquitous role
of epigenetics in complex phenotypes, epigenome editing has unique
potential to impact a broad spectrum of diseases. By leveraging powerful
DNA-targeting technologies, such as CRISPR, epigenome editing exploits
the heritable and reversible mechanisms of epigenetics to alter gene
expression without introducing DNA breaks, inducing DNA damage or
relying on DNA repair pathways.

Although all cells in the human body share nearly identical genetic Synthetically recapitulating natural mechanisms
codes, there is immense diversity in cellular phenotypes and functions of regulation
as a result of highly coordinated, cell type-specific gene regulation. Epigenome-editing technologies have largely emerged from funda-
Furthermore, our susceptibility and resilience to disease, age-related mental advances in engineering molecules that target and bind specific
cellular dysfunction, responses to environmental and pharmacologic DNA sequences. While early work focused on DNA-binding proteins
stimuli and nearly all other complex phenotypes are largely driven by such as zinc finger proteins (ZFPs), meganucleases and transcription
alterations in gene regulation rather than the sequences of genes them- activator-like effectors (TALEs)4, the landmark studies describing Cas9
selves. These critical differences in gene regulation are controlled by as an RNA-guided DNA nuclease with genome-editing activity in human
heritable yet reversible epigenetic modifications to DNA chemistry, cells dramatically expanded interest in programmable DNA targeting5–7.
chromatin structure and the transcriptional landscape. This ensures a The transformative advantage of CRISPR–Cas systems is the RNA-based
proper gene expression program throughout the lifetime of each cell programmability, which follows simple Watson–Crick base-pairing
and allows for the transgenerational inheritance of a programmable rules, unlike protein-based DNA-binding platforms that require exten-
genomic code. Large-scale genome-annotation projects such as the sive, cumbersome protein engineering for each genomic target. Com-
National Institutes of Health ENCODE Consortium1,2 and thousands of bined with other technological advances such as high-throughput DNA
genome-wide association studies (GWAS) have highlighted the impor- synthesis for guide RNA (gRNA) production and DNA sequencing for
tance of genetic variation and gene regulation on phenotypic traits. designing and monitoring editing outcomes, CRISPR–Cas systems have
The concept of epigenetics, which has substantially contributed to our been widely adopted across academia and industry. Over the last dec-
understanding of gene regulation, can be traced back to the work of ade, the field has progressed from discovering the molecular features
early pioneers such as Conrad Waddington, who was among the first to of CRISPR systems to using CRISPR gene editing in multiple clinical
propose an epigenetic landscape as a determining factor of cell identity3. trials to treat severe diseases. Notably, the inactivation of one or both
Given the broad importance of epigenetic control, the ability to nuclease domains of Cas9 to create Cas9 nickase (nCas9)6 or ‘dead’ Cas9
precisely manipulate the epigenome at one or more targeted sites (dCas9)5,8 has expanded the CRISPR–Cas9 DNA-targeting platform to
in living cells has profound implications for science and medicine. include base editors, prime editors and epigenome editors, each suited
This Review details how epigenome editing builds upon and extends for specific applications (Fig. 1). This Review focuses on epigenome
beyond other related technologies, summarizes current and future editing, but we recommend other recent reviews on nuclease-based
applications of epigenome editing in biotechnology and provides an gene editing, base editing and prime editing for a deeper understand-
outlook of challenges and future possibilities. ing of the complete genome-engineering toolbox9–11.

1
Department of Biomedical Engineering, Duke University, Durham, NC, USA. 2Center for Advanced Genomic Technologies, Duke University, Durham,
NC, USA. e-mail: [email protected]

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1199

Review article https://doi.org/10.1038/s41587-024-02320-1

Cas9 nuclease Adenine base editor (ABE) Cytosine base editor (CBE) Prime editor dCas9 epigenome editor

Uracil
Cas9 nCas9 glycosylase nCas9 nCas9 dCas9
Spacer inhibitor
(on sgRNA) Reverse
transcriptase

A C

PAM Adenosine Cytidine Prime editing Effector

deaminase deaminase gRNA (pegRNA) protein
Primer
binding RT template
sgRNA Complementary strand Target sequence (on non-complementary strand) site (containing edit)

dCas9 epigenome editor SAM SunTag Casilio

HSF1
p65 Effector protein
MCP
MS2 PUF
dCas9 dCas9 dCas9 PBS dCas9
GCN4
scFV
Effector protein
Effector VP64
protein

Chemical epigenetic CRISPR-DREAM CRISPR-Display FIRE-Cas9

modifiers (CEMs) Effector protein
Recruitment of proteins,
CEM ncRNA, IncRNA
FKBP eNRF2 Frb
STAT1 RAP
MRTF-A dCas9
Fkbp
Endogenous MCP
dCas9 dCas9 dCas9 MCP
epigenetic MS2 MS2
factor

Fig. 1 | Schematic of CRISPR-based nucleases, base editors, prime editors noncoding RNA; ncRNA, noncoding RNA; PAM, protospacer adjacent motif;
and epigenome editors with a detailed panel of various epigenome-editing RT, reverse transcription; scFv, single-chain variable fragment; STAT1, signal
platforms. Epigenome-editing platforms recruit epigenetic effectors or transducer and activator of transcription 1; FKBP, FK506-binding protein; eNRF2,
transcriptional machinery to the target DNA sequence via direct fusions with engineered nuclear factor erythroid 2-related factor 2; MRTF-A, myocardin-
dCas9, RNA aptamer systems, protein-recruitment systems or chemical ligation related transcription factor A; PUF, Pumilio/fem-3 binding factor; PBS, Pumilio
strategies. Once recruited, CRISPR-based epigenetic effectors can mediate binding sites; GCN4, peptide from Saccharomyces cerevisiae transcription factor
physical or chemical changes to histones and DNA sequences to silence or Gcn4; FRB, FKBP-rapamycin-binding domain of mTOR; RAP, rapamycin; MS2,
activate gene expression. CEM, chemical epigenetic modifier; lncRNA, long MS2 RNA aptamer; MCP, MS2 coat protein.

Epigenetic modifications are changes to genome structure transcriptional machinery, altering chromatin architecture or modify-
that affect gene expression without altering the underlying DNA ing histone residues or DNA methylation. Initially used to silence gene
sequence. These changes, including DNA methylation, histone expression in bacteria by sterically blocking DNA motifs8, this tech-
modifications, chromatin accessibility, looping and interactions nology was extended to human cells by fusing dCas9 with repressive
with noncoding RNA12 (Fig. 2), can activate or suppress genes and chromatin modifiers like the Krüppel-associated box (KRAB) domain
lead to phenotypic changes within an organism. This complex net- for CRISPR interference (CRISPRi)23 or transactivation domains like
work of diverse biochemical modifications collaboratively regulates p65 or VP64 (a tetramer of the minimal VP16 activation domain from
gene expression and chromatin structure13, playing a pivotal role herpes simplex virus) for CRISPR activation (CRISPRa)18,23,24. These
in governing various biological processes, such as development, technologies have evolved from initial applications of regulating gene
learning, memory, aging and the onset and progression of disease. expression by targeting promoters to modulating distal noncoding
Perturbations in epigenetic processes can have profound implica- regulatory elements19,20,25. Manipulating expression of genes within
tions for both health and behavior, contributing to conditions such their endogenous chromosomal context enables powerful applications
as addiction, neurodegenerative diseases, autoimmune diseases that were not possible with earlier genetic engineering technologies.
and cancer14–16. For example, epigenetic repression can be stable and maintained
Editing epigenetic modifications and/or chromatin structure through cell division, in contrast to gene repression by RNA interfer-
enables precise, tunable control of gene expression by repurposing ence, which is dependent on the continuous activity of small RNAs.
natural mechanisms of gene control, circumventing DNA damage- and Moreover, endogenous gene activation can produce multiple iso-
repair-related concerns associated with nucleases and nickases typi- forms, physiological expression levels, and noncoding elements such
cally used for gene editing of DNA sequences. The CRISPR–dCas9-based as the untranslated regions that are not captured by transgene over-
epigenome-editing system offers a versatile array of transcriptional or expression. Consequently, CRISPR–dCas9-based epigenome editing
epigenetic effector domains that can be fused to dCas9 and precisely has become a powerful tool for biomedical research and therapeutic
targeted to virtually any genomic locus using a gRNA complementary to applications, offering precise control of gene expression in human
the target sequence17–22. These effector domains function by recruiting cells (Table 1).

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1200

Review article https://doi.org/10.1038/s41587-024-02320-1

a Nuclear organization b Noncoding RNA: recruitment, scaffold, decoy, chromatin looping, precursor of miRNA

Nucleus
Recruitment Scaffold Decoy
Chromatin Ribonucleoprotein
modifiers complex
Chromosomal
territories ncRNA

TADs Active TAD compartments

Repressed TAD c Chromatin modification
TAD
boundary compartments
Repressive chromatin modifications Activating chromatin modifications
Nuclear
lamina
TFs
Chromatin
TAD interactions
loops
Inaccessible Repressed Accessible Expressed
chromatin gene chromatin gene
Regulatory element Gene
d
DNA methylation H3K27 deacetylation H3K9me2/3 H3K4me2 demethylation DNA demethylation H3K27 acetylation

DNMT1 SUV39H1/2 GADD45A

DNMT3A/L SETDB1/2 TET1 TDG p300
HDAC3 LSD1 TET3
MECP2 HP1 ZNF– CBP
KRAB
me me Ac me me me me Ac

G9a/GLP PRC2 SMYD3 SWI/SNF

NurD
HP1 ZNF– EZH2 HP1 ZNF– complex HP1 ZNF– PRDM9 DOT1L
KRAB KRAB KRAB SS18
me me me me me me me Ac me
FOG1

H3K9me1/2 H3K27 methylation Repressor factors H3K4 methylation H3K79 methylation SWI/SNF complex

Effector Effector
domain domain

Activated locus Repressed locus

dCas9 dCas9

1 1

Repressed locus Activated locus

2 2

Epigenetic editing Epigenetic editing

Fig. 2 | Schematic depicting the epigenetic regulation of gene expression. silencing 1-like; FOG1, friend of GATA protein 1 or zinc finger protein member
a, In the nucleus, chromosomes occupy defined territories, where the chromatin 1 (ZFPM1); G9A, euchromatic histone lysine methyltransferase 2 (EHMT2);
is either accessible or inaccessible. Within topologically associating domains GADD45A, growth arrest and DNA damage-inducible-α; GLP, G9a-related protein
(TADs), chromatin regions have greater interaction frequency, forming or euchromatic histone-lysine N-methyltransferase 1 (EHMT2); H3K, histone
chromatin loops connecting regulatory elements to their target genes. H3 lysine; HDAC, histone deacetylase; me, methyl; me2, dimethylation; me3,
b, Noncoding RNAs can recruit chromatin modifiers to gene promoters, either trimethylation; HP1, heterochromatin protein 1 homolog-α or chromobox
enhancing or repressing transcription in cis or trans. They can also act as decoys homolog 5 (CBX5); LSD1, lysine demethylase 1A (KDM1A); MXD1, MAX
for these modifiers, sequestering them from gene promoters. Additionally, dimerization protein 1; NuRD, nucleosome remodeling deacetylase; SUV39H1,
they form RNA–DNA hybrids during transcription, recognized by modifiers or histone H3K9 methyltransferase 1 or suppressor of variegation 3-9 homolog 1;
TFs to activate or inhibit gene transcription. miRNA, microRNA. c, Chemical PRC2, polycomb repressive complex 2; PRDM9, PR–SET domain 9; SETDB, SET
modifications to DNA or histone tails can alter chromatin accessibility and domain bifurcated histone lysine methyltransferase 1; SMYD3, SET and MYND
affect gene expression levels. d, Repressive (red) and active (green) epigenetic domain-containing 3; SUV39H1, suppressor of variegation 3-9 homolog 1 or
modifications to DNA or histone tails. Ac, acetyl; CBP, cAMP response element- histone H3K9 methyltransferase 1; TDG, thymine DNA glycosylase.
binding protein binding protein (CREBBP); DOT1L, disruptor of telomeric

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1201

Review article https://doi.org/10.1038/s41587-024-02320-1

Table 1 | Catalog of notable epigenetic activators and repressors

System (effect) Effector domain(s) Effector size Pros Cons Notes Refs.
(bp)

VP64 ~150 Small, versatile with activity Not always as potent as other Two copies of VP64 23,35
across cell types and contexts activation systems leads to more potent
activation
VPR (VP64 + p65 + Rta) ~1,600 Recruits multiple activation Challenging to stably deliver; Toxicity has been 36,218,219
domains, which can lead to toxicity reported in both observed with VPR in
better gene activation Drosophila and mice human cells in some
settings
miniVPR ~900 ~40% smaller than VPR yet Relatively large, cannot In vivo activity when 219,220
retains activity package into AAV with fused to dCas12
SpCas9
p300 ~1,900 Enables targeting of proximal Challenging to stably Conditional dCas9– 19,179
and distal enhancers deliver. p300; can be toxic if p300 mouse model
expressed at high levels. available; some
gRNA-independent
activity
TET1 ~2,200 Can lead to stable and heritable Requires Varying efficiency 52,175,221
activation with transient methylation-sensitive depending on
delivery promoter or regulatory construct design and
Direct fusion element; challenging to linker length
(activation) deliver due to size
TET3CD ~2,800 Activates gene expression via Requires Less frequently 222
targeted DNA demethylation methylation-sensitive used than TET1
promoter or regulatory for targeted DNA
element; challenging to demethylation
deliver due to size applications
PRDM9 (SET domain) ~900 Activates gene expression via Does not activate all silenced Context dependent: 58
H3K4 methylation genes; typically not as potent stability of
as VP64 reactivation depends
on DNA methylation
DOT1L ~1,300 Activates gene expression via Does not activate all silenced Context dependent: 58
H3K79 methylation genes; typically not as potent stability of
as VP64 reactivation depends
on DNA methylation
SMYD3 ~1,400 Activates gene expression Relatively weak activation; PC4 stabilizes 223
through H3K4 methylation and does not activate all silenced SMYD3; requires a
other chromatin-remodeling genes pool of gRNAs for
functions, in conjunction with efficient activation
PC4
CRISPR-DREAM MRTF-A, STAT1, eNRF2, ~900 Robust gene activation at Different Cas9 or fusion Can activate 39
(activation) others with MCP or promoters and enhancers; can systems may require microRNA and
direct fusion to dCas be miniaturized; deliverable different arrangements of the lncRNA expression;
with AAV as an all-in-one vector transactivation domains for also used with
optimal activation orthogonal dCas9,
dCas12 and type I
CRISPR systems
SunTag VP64, TET1b a
Recruits multiple copies of Optimal number and spacing 5- and 22-amino acid 37,53
(activation or effector to the targeted site for of GCN4 peptide sequences linkers were optimal
repressiona) signal amplification is dependent on effector for VP64 and TET1,
respectivelyc
SAM Direct VP64 ~150 + ~1,400 Robust gene activation, Requires gRNA with an MS2 Functional in primary 21,38
(activation) fusion + MCP–p65– flexibility in plasmid design hairpin human T cells
HSF1
Chemical Direct dCas9–FKBP a
Recruits endogenous Limited modularity; custom Requires a direct 224
epigenetic fusion + chemical chromatin-activating machinery chemical compounds need dCas9–FKBP fusion
modifiers compounds that bind to target genes for activation to be synthesized for each for programmability
(activation) BRD4, BRPF1 or CBP/ unique effector
p300
Casilio TET1, TET1 + GADD45Ab a
Recruits multiple copies of Configuration of effector Compatible with 225
(activation) effector for signal amplification; proteins requires plasmid transfection
programmable recruitment of optimization or PiggyBac
different effectors to same Cas9 integration and
transfection
CRISPRon TET1 + p65–Rta ~2,200 + ~700 Potent, stable activation with Requires Demonstrated to 47
(activation) (on gRNA) transient delivery; faster kinetics methylation-sensitive reactivate previously
and activation than TET1 alone promoter; components split CRISPRoff-silenced
across plasmids due to size genes
constraints; challenging to
deliver

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1202

Review article https://doi.org/10.1038/s41587-024-02320-1

Table 1 (continued) | Catalog of notable epigenetic activators and repressors

System (effect) Effector domain(s) Effector size Pros Cons Notes Refs.
(bp)

CRISPR-Display RNA cargo (for a

Potent reporter activation Uncertain efficacy at diverse Used to recruit Fos 226,227
(activation or example, lncRNA) or repression; allows for endogenous loci enhancer RNA to
repression) and/or RNA-binding recruitment of endogenous the Fos enhancer to
proteins or artificial RNA, RNA-binding weakly activate Fos
proteins or fusion proteins (for expression
example, PP7–VP64) to target
FIRE–Cas9 dCas9 + MS2–FKBP a
Inducible and reversible Relatively subtle effects Transient BAF 228
(activation or protein + FRB tag recruitment of chromatin on gene expression; recruitment does not
repressiona) chromatin regulator regulators to target genes multiple-part system maintain long-term
(BAF for activation, activation of bivalent
HP1cs for repression) genes
KRAB ~150–300 Small, versatile repressor; Effects are transient; does not Conditional dCas9– 23,179
rapidly silences gene program epigenetic memory KRAB mouse model
expression available
EZH2 ~800 Superior repression compared Larger than KRAB; steric EZH2 requires 45,46,228
(truncated) or to KRAB in specific contexts interference may contribute co-delivery of
~2,200 (full to repression full-length DNMT3L
length) for maximal activity
LSD1 ~2,500 Potent enhancer repression Large effector; LSD1 is LSD1 demethylates 25,229
implicated in demethylation H3K4me2, reduces
of both repressive and active H3K27ac and
histone methylation may interact with
other histone
demethylases
KRAB–MECP2 ~1,100 Superior repression to KOX1 Larger effector; effects have Can improve genetic 230
KRAB domain alone but inferior not been examined across interaction mapping
to ZIM3 KRAB domain many cell types
HDAC3 ~1,300 Targeted histone (H3K27) Weak effect on target gene Useful for assessing 231
Direct fusion deacetylase expression role of histone
(repression) acetylation but
not established for
robust target gene
silencing
FOG1 ~150 per Small effector; recruits Few examples of targeted FOG1 binds NuRD 57
domain endogenous machinery to epigenome editing using complex, which
deposit H3K27 trimethylation FOG1 recruits PRC2 for
H3K27 trimethylation
SUV39H1 ~1,000 Can achieve similar levels of Few examples of targeted Repressed HER2 57
repression as G9A without epigenome editing using expression
depositing H3K9me3 SUV39H1 without depositing
H3K9me3,
H3K27me3 or
H3K9me2
MQ1/M.Sss1 ~1,200 Prokaryotic DNA Repression by dCas9–MQ1 Frequently observed 54,55
methyltransferase for targeted was primarily due to steric a ~20-bp gap in
DNA methylation of specific hindrance, not methylation methylation at
CpG sites without affecting deposition gRNA-binding site
global methylation levels
MS2/MCP MXD1 (SID domain) 90 Effective repression of Has not been extensively Interacts with SIN3 232
recruitment noncoding RNA characterized for repression proteins to recruit
(repression) of protein-coding genes HDAC1 and HDAC2
CRISPRoff DNMT3A-3L + KRAB DNMT3A-3L, Potent and stable repression May require Stability of silencing 47
(repression) 1,500; with transient delivery; methylation-sensitive in non-CpG island
KRAB, ~150 reversible with demethylation promoter promoter context
appears to vary
a
Dependent on recruited effector. bModular (other effectors could be used with the system). cOf tested linker lengths. AAV, adeno-associated virus; BAF, BRG1/BRM-associated factor;
BRD4, bromodomain-containing protein 4; BRPF1, bromodomain- and PHD finger-containing protein 1; HER2, human epidermal growth factor receptor 2; HP1cs, chromo-shadow domain of
heterochromatin protein 1; MECP2, methyl-CpG-binding protein 2; MQ1, DNA methyltransferase from Mollicutes spiroplasma (M.Sss1) strain MQ1; PC4, positive cofactor 4; PP7, PP7 phage coat
protein; SIN3, paired amphipathic helix protein Sin3a.

Recent technology innovations and developments Next-generation CRISPR activators

The engineering of synthetic transcription factors (TFs) has a Early work in generating CRISPR-based activators relied on fusions
decades-long history, and a variety of epigenome editors based on to concatemers of the minimal transactivation domain of the VP16
ZFPs, TALEs and other DNA-binding proteins established an early foun- protein from herpes simplex virus17,18,23,24,31, which promotes transcrip-
dation for the epigenome-editing field26–30. Here, we describe the most tion by recruiting co-activators and transcriptional machinery to the
recent innovations that have catalyzed the broad use and interest in target gene. However, in many cases, these first-generation activa-
these technologies across science and medicine. tors did not achieve the desired levels of activation. To enhance gene

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1203

Review article https://doi.org/10.1038/s41587-024-02320-1

a Screens of coding genes c Single-locus screen

–300 0 –50 +300 Low High Low High

gRNA 1 500 500
gRNA 2 100 800
Gene 1 Gene 2 gRNA 3 480 500
gRNA 4 800 150
TSS … … …
gRNA n 400 440

b Screens of noncoding regions Target gene expression

Saturation
d scRNA-seq
Gene 1
gRNA 1 gRNA 2 gRNA n
Gene A Gene B … Gene C
Gene C Gene C … Gene H
Targeted Gene E Gene K … Gene S

UMAP-2
Gene C Gene T … Gene X
gRNA 1
gRNA 2
Gene 1 gRNA 3
gRNA 4
NT
Region 1 2 3 UMAP-1

e Cell fitness
Compare gRNA counts
gRNA library
lentivirus 200 200

150 150
Counts

Counts
100 100

50 50

0 0
1 2 3 4 1 2 3 4
gRNA gRNA
CRISPRi/a cell line Transduced
Initial population Final population
or primary cells cell population

Fig. 3 | Schematic displaying gRNA library design for screens of genes and sorting or magnetically activated cell sorting to separate cells based on target
noncoding elements with various screening readouts. Cells expressing gene expression, followed by gRNA library amplification and deep sequencing
CRISPR epigenome-editing machinery are transduced with lentivirus encoding to compare gRNA abundances between sorted populations. d, In scRNA-seq
the gRNA library. a, Gene-focused screens target regions both upstream and screens, both mRNA and gRNA transcripts within individual cells are captured,
downstream of the promoter of each target gene, with CRISPRa gRNAs optimally and computational workflows assign gRNAs to cells and group cells by gRNA
positioned within 300 bp upstream of the transcriptional start site (TSS) and assignment, and gene expression profiles for each gRNA are compared relative
CRISPRi gRNAs positioned between 50 bp upstream and 300 bp downstream to control cells. NT, nontargeting; UMAP, uniform manifold approximation
of the TSS64. b, Noncoding screens use gRNA libraries tiling a specific genomic and projection. e, Cellular fitness-based screens measure the effects of genes
region or targeted gRNA libraries for specific genomic features (represented or regulatory elements on cell survival and proliferation. Changes in gRNA
as blue triangles demarcating, for example, chromatin accessibility, histone abundances between the initial and final cell populations determine whether
marks, TF binding). The readout of a pooled CRISPR screen is determined by the perturbations have a positive or negative impact on cellular fitness.
study’s objectives. c, Single-marker screens often use fluorescence-activated cell

activation, one strategy has been to simultaneously target multiple heterochromatin-forming proteins to rapidly silence gene expression.
adjacent regions on a promoter or a regulatory element to achieve Nevertheless, KRAB domains can differ significantly in their potency of
synergistic effects18,24,32,33 or introduce VP16 concatemers on both transcriptional repression, in part due to the strength of their interac-
termini of the dCas9 protein34,35. For more potent and/or lasting gene tions with tripartite motif-containing protein 28 (TRIM28 or KAP1)43,44.
activation, next-generation CRISPR activators use novel fusions and While most CRISPRi approaches use the zinc finger protein 10 (KOX1)
multi-effector recruitment strategies. Examples include the tripartite KRAB domain, the zinc finger-imprinted 3 (ZIM3) KRAB domain was
activator VPR36 (VP64, p65 and Rta) as well as peptide- or RNA-based shown to outperform the standard CRISPRi platform43. Other classes
recruitment systems like SunTag37, synergistic activation mediator of repressive chromatin regulators such as de novo DNA methyltrans-
(SAM)38 and CRISPR-DREAM39. These advancements harness multiple ferases DNMT3A and DNMT3B exhibit slower kinetics and achieve less
activation domains to better emulate natural transcription processes, potent gene silencing than KRAB but can install more durable epige-
which rely on the coordination of various factors (Table 1). An in-depth netic memory42.
comparison of dCas9-based gene activation methods in multiple spe- A longstanding goal of epigenome editing has been to take advan-
cies, including human, mouse and insect cell lines, showed that these tage of heritable mechanisms of native epigenetic regulation to induce
systems consistently outperformed VP64 for gene activation40. None- stable gene repression or activation following only transient activity
theless, the optimal approach for gene activation is likely to vary based of epigenetic effectors. By harnessing the unique properties of mul-
on the target gene, cell type and chromatin context. tiple epigenetic effectors, methods have been developed to dura-
bly silence gene expression. For example, stable gene silencing with
Stable gene repression transient epigenome editor treatment was achieved by co-delivery
Early studies revealed that different repressive chromatin regulators, of three DNA-targeting proteins fused separately to KRAB, DNMT3A
even within the same family, have distinct kinetics, strength and sta- and DNMT3L41. Another study found that persistent gene silencing
bility of gene silencing41,42. For example, the human genome encodes was dependent on both the target gene and recruited effector pro-
over 350 KRAB domains, which generally recruit co-repressors and teins45. While co-delivery of dCas9 fused to DNMT3A and the histone

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1204

Review article https://doi.org/10.1038/s41587-024-02320-1

Table 2 | Comparing CRISPR-based screening readouts editors can facilitate interrogation of the roles of specific epigenetic
modifications in gene regulation and genome structure51. For instance,
Screening Advantages Disadvantages dCas9-tethered DNA methyltransferases precisely deposit CpG methyla-
readout
tion at target loci with minimal off-target effects52–55. With these tools,
Single marker • Relatively inexpensive • Requires optimization of researchers have shown that DNA methylation at CCCTC-binding fac-
• Easily scalable target gene detection tor (CTCF) binding sites disrupts CTCF binding52. Furthermore, dCas9
• Simple analyses • Unidimensional readout
• Compatible with protein only captures how each alone sterically blocks DNA methylation, allowing for investigation of
(antibody staining) and RNA perturbation affects target the role of DNA methylation at specific genomic sites without potential
(HCR–FlowFISH) detection gene or protein levels off-target enzymatic activity56. Other studies have used epigenome
scRNA-seq • High-dimensional • Expensive editing to directly link DNA methylation to transcriptional repression
data capture the full • Limited scalability at CpG-rich promoters42,47,52. Although targeted deposition of particular
transcriptomic effects of • Sparse data can limit histone modifications can result in transcriptional changes, attempts
each perturbation resolution of lowly
• Permits investigation of cell expressed genes to dissect the roles of these marks suggest that their functions depend
type-specific effects • Complex computational on the chromatin context and/or DNA methylation19,45,57–59. Additional
• Does not require analyses research is needed to establish defined rules linking individual and com-
hypotheses a priori
binations of chromatin modifications to transcriptional output. Nota-
Cell fitness • Relatively inexpensive • Can only detect bly, epigenome-modifying enzymes tethered to DNA-binding proteins
• Easily scalable regulatory elements or (dCas9, dCas12a (dCpf1), etc.) are powerful tools to explore these links;
• Simple analyses genes that impact cell
• No advanced fitness however, new protein fusions often require substantial optimization to
instrumentation (for achieve robust expression and activity19,47,57. Furthermore, if a full-length
example, cell sorter or protein is used, domains aside from the core enzymatic domain may
single-cell device) required recruit other proteins, contributing to transcriptional modification
HCR–FlowFISH, hybridization chain reaction fluorescence in situ hybridization coupled with through additional functions. In some cases, the core enzymatic domain
flow cytometry.
may be sufficient to induce changes in gene expression57,60,61.

methyltransferases EZH2 or KRAB led to long-term repression of some CRISPR-based epigenetic screens explore gene
genes, only co-delivery of DNMT3A–dCas9 and EZH2–dCas9 facili- regulation mechanisms
tated stable HER2 repression45, prompting further exploration of the The robustness of CRISPR technology and the ease of designing and
molecular determinants of heritable gene silencing46. Recently, the synthesizing large gRNA libraries has led to widespread use of CRISPR
CRISPRoff system was developed to simplify delivery by fusing a single screens for discovery of gene functions and regulatory element activity.
dCas9 protein to KRAB, DNMT3A and DNMT3L effector domains47. While many applications have used CRISPR nucleases for gene-knockout
CRISPRoff demonstrated highly specific and heritable gene silencing screens, CRISPR-based epigenome-editing tools have uniquely enabled
across numerous endogenous genes by establishing DNA methyla- gain-of-function screens via gene activation and high-resolution anno-
tion and depositing repressive histone modifications. This approach tation of the noncoding genome62. All CRISPR epigenetic screens share
has also been expanded to ZFP- and TALE-based DNA targeting and three key components: (1) a CRISPR-based epigenetic modifier (for
was similarly effective in stably silencing target genes in vivo when example, dCas9 fused to VP64 or KRAB) to manipulate gene expression,
delivered to the liver encoded by messenger RNA (mRNA) carried by (2) a gRNA library to target specific genomic regions and (3) a readout
lipid nanoparticles48. for gRNA activity (Fig. 3 and Table 2). These screens can be broadly
classified into gene-targeting screens (Fig. 3a) and noncoding-targeting
Expanding the toolkit of transcriptional effector protein screens (Fig. 3b), each with specific applications.
domains
High-throughput screening strategies have been developed using syn- Gene-discovery screens for diverse cellular phenotypes and
thetic reporters to quantitatively evaluate thousands of 80-amino acid functions
sequences from candidate human transcriptional effector domains44 CRISPRi and CRISPRa screens have advanced functional genomics by
and viral proteins49,50. These screens identified many KRAB repressive discovering genes that regulate general processes, such as survival63,64,
domains and even a divergent KRAB activator domain as well as hun- differentiation65–67 and protein trafficking68, and cell type-specific pro-
dreds of viral transcriptional effectors. Because these studies were lim- cesses, such as monocyte activation69, microglial activation and phago-
ited to screening 80-amino acid domains as a result of size constraints cytosis70 and neuronal response to chronic oxidative stress71. These
of DNA-synthesis technologies for pooled libraries, opportunities to screens can either target all ~20,000 annotated protein-coding genes
explore larger domains remain. Additionally, the activity of certain (genome-wide) or a curated subset based on gene function (for exam-
domains may rely on their native protein context. Another unbiased ple, genes encoding TFs, epigenetic regulators or metabolic factors),
proteome-scale screen assayed the activity of full-length open reading noncoding RNA72 or genomic profiling data22,73. However, a tradeoff
frames (ORFs) encoding 13,571 human genes and identified over 200 exists between the scale of a screen and the necessary resources, time
transcriptional activators50. Future work could involve screening the and cost. Genome-wide screens may not be feasible in post-mitotic cells
activity of full-length effector domains and epigenome-modifying or primary cells of limited quantity. Fortunately, the development of
enzymes targeted to endogenous genes within different chromatin optimized, compact gRNA libraries with only three to six gRNAs per
contexts and cell types to expand the epigenome-editing toolbox and gene has made larger-scale gene-discovery screens more manage-
improve our understanding of native chromatin regulation. able63,64. Many of the most impactful discovery screens for genes that
control complex functions have been in the context of cellular repro-
Deciphering the epigenetic code with targeted gramming and ex vivo cell engineering for therapeutic applications,
epigenome engineering as discussed below.
Although dozens of epigenetic chemical modifications to DNA and
histones have been reported, knowledge of the function of each modi- Gene screens to identify cell fate regulators
fication and how that changes across genomic locations, cell types Seminal work identified key TFs that reprogram somatic cells into
and environmental conditions is often lacking12. Targeted epigenome induced pluripotent stem cells (iPSCs)74. iPSCs offer infinite self-renewal

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1205

Review article https://doi.org/10.1038/s41587-024-02320-1

a Cell fate reprogramming b T cell applications

Pluripotent Engineer T cell phenotypes

stem cell Counter or reverse exhaustion Enhance effector function
TF
Gene PRF1
Functional GZMB
T cell
IFN-γ
TNF

Proliferation or persistence

Dir
My g
4, min
c)

ec (TF
2, gram

ted 1,
Exhausted

diff TF2
Klf
, S pro

T cell

ere )
ct4 re
ox

nt i
TF
(O lular

ionat
l
Ce

TF TF Define noncoding elements

Cell type specific Conserved enhancer
High expression
TF1 TF2
Gene
Enhancer Enhancer Promoter
1 2
Moderate expression
TF2
Transdifferentiation Gene
Somatic (TF3, TF4, TF5) Somatic Enhancer Enhancer Promoter
cell type 1 cell type 2 1 2

c Epigenome editing for disease

Healthy Diseased Rescued
Maternal

Gene A Gene A Gene A

Paternal

Mutation

Gene A Stop Gene A Stop Gene A

Dosage-sensitive gene

Fig. 4 | Schematic of ex vivo and in vivo applications of epigenome to define conserved and cell type-specific noncoding elements across T cell
engineering. a, Epigenome engineering has been used to reprogram cells to a subsets. GZMB, granzyme B; IFN-γ, interferon γ; TNF, tumor necrosis factor; PRF1,
pluripotent state, direct iPSC differentiation and transdifferentiate somatic cells, perforin-1. c, CRISPRa can rescue haploinsufficient or aberrantly silenced genes.
both ex vivo and in vivo. b, Epigenome engineering can enhance the therapeutic In the case of loss of expression of one allele of a gene, CRISPRa can increase
potential of T cell-based immunotherapy by countering or reversing T cell expression from the second intact allele by depositing activating chromatin
exhaustion and improving cell-intrinsic properties such as effector function, marks, removing DNA methylation or recruiting transcriptional machinery.
proliferation and persistence. Additionally, epigenome editing could be applied

and the ability to differentiate into various cell lineages, enabling (ESCs) at single-cell resolution87. The authors constructed expression
in-depth exploration of disease mechanisms, drug testing and the profiles downstream of overexpression of each TF and identified cell
development of personalized and allogeneic cell therapies. This work fate-specifying TFs, building upon previous TF ORF screens88,89.
prompted widespread efforts to find the optimal combination of TFs Gene-targeting CRISPRa screens are also a powerful method for
to direct the differentiation of iPSCs to specific lineages (for exam- systematically discovering cell fate-regulating genes. Unlike ORF over-
ple, iPSCs to neurons) or coordinate the direct transdifferentiation expression, CRISPRa enables programmable control of endogenous
of somatic cells (for example, fibroblasts to neurons) (Fig. 4a). While gene expression in its native context, more closely mimicking natural
early studies primarily focused on constitutive overexpression of ORFs gene regulation. Early studies established that endogenous activation
of specific candidate TFs75–80, the success of reprogramming based on of genes encoding master TFs can efficiently drive cell fate specification
testing candidate factors hinges on gene selection and thus is biased and offer certain advantages over exogenous overexpression34–36,90–92.
by existing literature81–86. To expand capabilities for cell reprogram- For example, endogenous activation of Oct4 or OCT4 (Pou5f1 or
ming beyond this limited knowledge set, notable efforts have focused POU5F1) using TALEs or dCas9 could replace transgenic Oct4 in repro-
on developing high-throughput technologies to assess all genes and gramming fibroblasts to iPSCs in both mouse and human cells91,93,94.
isoforms that can direct any cell conversion. A key advantage of endogenous activation is its potential to initiate
ORF libraries have expanded empirical gene evaluation for cell fate and maintain autonomous gene networks independently of CRISPRa
specification. A library of 3,548 ORFs encoding all annotated human machinery. For example, transient expression of dCas9-based activa-
TF isoforms was recently screened in human embryonic stem cells tors or gRNAs targeting endogenous loci encoding master regulators

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1206

Review article https://doi.org/10.1038/s41587-024-02320-1

of myogenic or neurogenic lineages led to cellular reprogramming CREs and have led to enhanced predictive models for linking enhanc-
and sustained expression of target genes, which was not observed with ers to genes107. Notably, this work confirms that single genes can be
transient exogenous overexpression of these same factors34,35. These regulated by several distal elements, and a distal enhancer can control
feedback loops are likely established via chromatin remodeling, as multiple genes or skip over one or more nearby genes to regulate a more
only CRISPRa deposited histone marks positively associated with gene distant target107. However, single-locus screens are limited in scope:
expression at the targeted loci35. gRNA design is restricted to a predefined genomic region, potentially
These studies established a foundation for unbiased, missing long-range interactions. Additionally, these screens typically
high-throughput CRISPRa screens in the context of cell fate specifica- assess the effects of CRISPRi and CRISPRa perturbations on the expres-
tion. Such screens often use an engineered cell population with con- sion of a single gene, thus limiting the identification of regulatory
stitutive expression of a CRISPR activator and a tagged gene of interest elements impacting other genes or phenotypes (Table 2).
for specific cellular reprogramming. Alternative cell-sorting-based Noncoding screens can also be coupled to more complex readouts,
readouts like RNA hybridization, immunostaining or single-cell profil- such as single-cell RNA sequencing (scRNA-seq) or specific cell pheno-
ing can reduce upfront cell engineering. In these screens, the cell popu- types. Single-cell gene expression readout enables causal inference of
lation is typically transduced with a gRNA library such that each cell the effect of perturbations on all transcripts by simultaneously captur-
receives a single gRNA, and cells are cultured under specific reprogram- ing mRNA and gRNA transcripts within individual cells (Fig. 3d)111–114.
ming conditions. For cell-sorting screens, cells are separated based on Using this method, 470 cis enhancer–gene pairs were identified in the
reporter gene expression into high- and low-expression populations. human K562 cell line115. This approach is particularly powerful when an
Target genes of gRNAs enriched in these high- and low-expression unbiased method to mapping regulatory connections is desired but is
populations represent candidate positive and negative regulators of limited by scalability, as a result of the high costs of single-cell technolo-
the desired cell type, respectively (Fig. 3c). gies (Table 2). Innovative scaling strategies116 as well as decreasing costs
CRISPRa screens have discovered transcriptional regulators and will continue to catalyze larger-scale noncoding screens.
epigenetic factors driving neuronal fate in mouse ESCs65 and human While single-locus and scRNA-seq screens facilitate the discovery
iPSCs66. This led to enhanced neuronal differentiation using orthogonal of gene-regulatory mechanisms, it is often not immediately obvious
dCas9-based epigenetic tools to simultaneously activate positive regu- how these interactions coordinate cell phenotypes. Consequently,
lators and repress negative modulators66. Another study conducted a screens linked to specific phenotypes amenable to high-throughput
genome-wide CRISPRa screen to discover factors that reprogrammed analysis have been pursued to directly link cell functions to genes and
primed epiblast stem cells into a pluripotent ESC-like state67, demon- associated regulatory elements. Cell fitness, for instance, serves as
strating the general applicability of the approach to any cell conversion. a straightforward indicator, where gRNAs that direct perturbations
affecting cell survival or proliferation become enriched or depleted
Noncoding CRISPR screens over time (Fig. 3e). Cell fitness-based screens have been implemented
CRISPR screens also enable the identification of noncoding regula- to discover essential genes in fast-growing cancer cell lines98,106. Their
tory elements62. Large-scale, multi-center research initiatives such as scalability, leveraging rapidly dividing cell lines and a simple phe-
ENCODE1,2 and the Roadmap Epigenomics project48,49 have provided notypic readout that does not require cell sorting or transcriptional
comprehensive genome-wide maps of chromatin accessibility, DNA profiling, allows the interrogation of millions of gRNAs to discover non-
methylation, histone modifications, TF binding and mRNA expression coding regions impacting cell fitness117. Moving forward, we expect the
across diverse cell types, identifying millions of candidate regulatory inclusion of even more complex phenotype traits to serve as selection
elements. In addition, over 45,000 GWAS across more than 5,000 criteria in these screens, such as cell migration118 and morphology119.
human traits have associated genetic variants with disease traits and
specific phenotypes95. However, most candidate regulatory elements Ex vivo cell engineering with epigenome editing
and genetic variants remain uncharacterized. Ongoing efforts aim to Ex vivo cell engineering holds enormous therapeutic potential for
identify functional regulatory elements, elucidate how they affect treating hematological disorders, infectious disease and cancer, as well
chromatin structure and gene expression and determine the causal as numerous possibilities in regenerative medicine. To date, there are
genetic variants associated with each condition. CRISPR screens have more than 20 cell therapies approved by the Food and Drug Adminis-
been at the forefront of the ambitious effort to catalog and characterize tration, with only one incorporating genome editing. This balance is
all functional regulatory elements in the noncoding genome62. Unlike likely to change soon, with dozens of active clinical trials evaluating cell
the conventional massively parallel reporter assays employed for func- therapies engineered with CRISPR technologies. Two leading clinical
tional genomics, such as STARR-seq (self-transcribing active regulatory trials (NCT03745287, NCT03655678) are evaluating exagamglogene
sequencing)96 and MPRA (massively parallel reporter assays)97, CRISPR autotemcel (exa-cel) for the treatment of transfusion-dependent
screens investigate candidate cis-regulatory elements (cCREs) within β-thalassemia and sickle cell disease (SCD)120. Whereas SCD results from
their native genomic context. a recurrent specific mutation, transfusion-dependent β-thalassemia is
Single-locus, gRNA tiling CRISPR screens are a simple form of caused by a spectrum of loss-of-function mutations in the same hemo-
noncoding screen that can identify cCREs regulating the expression globin subunit B (HBB) gene encoding β-globin. Rather than tailoring
of target gene(s) in a specific genomic region (Fig. 3c). While earlier gene-correction strategies for each patient, exa-cel’s gene-editing strat-
screens employed Cas9 nuclease to disrupt or delete specific DNA egy targets BCL11A, leveraging a GWAS result that associated reduced
regions98–105, CRISPRi and CRISPRa screens offer several advantages expression of the BCL11A transcriptional repressor with milder symp-
for cCRE discovery. Epigenetic effectors have a larger footprint of toms in patients with β-thalassemia and SCD121. Reduced BCL11A repres-
local epigenome-modifying activity (~1 kb), enabling greater genomic sor levels lead to upregulation of fetal γ-globin, which compensates for
coverage with fewer gRNAs20,22,106. Additionally, while genetic screens lost β-globin function. To this end, an erythroid-specific enhancer that
are limited to loss-of-function interrogation, CRISPRi and CRISPRa controls expression of the BCL11A gene in red blood cells was identified,
screens allow both loss-of-function and gain-of-function perturba- along with gene-editing strategies that only disrupt BCL11A expres-
tions. For instance, CRISPRa screens can reveal CREs that are epi- sion in this lineage122. Exa-cel disrupts the BCL11A enhancer in hemat-
genetically silenced in a particular cell type and thus complement opoietic stem cells using CRISPR–Cas9, which reduces expression of
CRISPRi findings22. CRISPR epigenome screens have discovered CREs BCL11A and increases fetal hemoglobin specifically in erythrocytes.
at many genomic regions, including those with complex genetic regu- Importantly, restoration of fetal hemoglobin can compensate for loss
lation22,98,107–110. These screens have revealed fundamental features of of adult hemoglobin, as patients enrolled in the exa-cel clinical trials

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1207

Review article https://doi.org/10.1038/s41587-024-02320-1

have maintained high levels of allelic editing in the bone marrow and Schmidt et al. overcame these challenges by optimizing lentiviral titers
blood, increased levels of fetal hemoglobin and transfusion independ- of nuclease-deactivated Streptococcus pyogenes Cas9 (dSpCas9)-based
ence for more than a year after treatment120. This approach highlights epigenome editors, which enabled them to conduct genome-wide
both the utility of CRISPR technologies for cell therapy and the power CRISPRi and CRISPRa screens to decode cytokine regulation21. In addi-
of developing therapeutics targeting specific gene-regulatory sites tion, a smaller, orthogonal nuclease-deactivated Staphylococcus aureus
in the noncoding genome, as ubiquitous BCL11A ablation can cause (dSaCas9)-based epigenome-editing toolkit was developed for CRISPRi
other pathologies. While exa-cel uses the more traditional DNA-editing and CRISPRa screens and used to identify transcriptional and epigenetic
technology, the success of the approach provides strong support for regulators of CD8+ T cell states139. These advances are poised to drive
use of epigenome-editing technologies that similarly reprogram cel- gene-discovery screens for various T cell phenotypes, such as T cell
lular gene expression but do not require DNA sequence changes. In exhaustion, effector function, proliferation and persistence (Fig. 4b).
fact, activation of fetal γ-globin was an early application of ZFP-based Moreover, CRISPRi and CRISPRa technologies are uniquely equipped
epigenome editors123,124. to resolve the role of noncoding regulatory elements in primary human
T cells. For instance, the vast majority (77%) of the 24,000 putative
Engineering T cells for cancer treatment enhancers across CD8+ T cell subsets are specific to a particular cell
Chimeric antigen receptor (CAR) T cell therapies account for a sub- subset, suggesting distinct mechanisms of gene regulation (Fig. 4b)153,154.
stantial fraction of all cell therapies approved by the Food and Drug To empirically evaluate these enhancers, CRISPRi and CRISPRa screens
Administration, with six CAR T cell products approved for treatment of could be designed to both measure their activity and identify gene
various blood cancers. Nevertheless, less than half of patients achieve targets across CD8+ T cell subsets110. These insights could propel the
long-term tumor control, with many factors influencing these divergent development of the next generation of T cell-based therapies.
outcomes125–127. Several retroactive studies have identified specific
transcriptional signatures of the infused T cell product associated with In vivo therapeutic applications of epigenome
clinical response, highlighting the importance of epigenetic control editing
of T cell fitness in treatment efficacy128–130. The unintentional disrup- Hijacking epigenetic mechanisms of gene control for targeted in vivo
tion of the gene encoding the DNA demethylase TET2 within a single manipulation of gene expression has tremendous potential for treat-
clone of a CAR T cell product led to an altered differentiation status ing diseases and catalyzing regenerative medicine (Table 3). Unlike
and complete tumor remission in a patient with chronic lymphocytic small-molecule drugs that globally inhibit specific epigenetic modi-
leukemia130. While this discovery stemmed from a single patient, it fiers, CRISPR-based epigenome editing offers a more precise approach
underscores the potential of leveraging epigenetic mechanisms to to treat a wider range of disorders stemming from diverse forms of
engineer more potent, durable and safe T cells. Strategies to achieve epigenetic dysregulation. These include diseases resulting from gene
this goal include constructing logic-controlled CARs127,131–134, arming overexpression, duplication155,156 or loss of expression157 as well as com-
T cells with specific transgenes135–139 and manipulating the genome140–145 plex alterations16,158, such as haploinsufficiency, X-linked inheritance,
or epigenome21,139,146. imprinting disorders and promoter and enhancer mutations.

CRISPR epigenome editing of primary human T cells Haploinsufficiency diseases

Arrayed and screening-based CRISPR knockout studies have expanded Haploinsufficiency, in which partial gene expression resulting from a
the catalog of gene targets for enhancing T cell potency141,143–145,147–149. loss of one autosomal copy is insufficient for proper function, can have
A phase I clinical trial demonstrated the feasibility and safety of mul- profound phenotypic effects151. Approximately 300 genes are known to
tiplexed CRISPR gene-edited autologous T cells, although aberrant cause haploinsufficiency-related diseases159,160 and can produce serious
genome-editing outcomes were detected in the infused T cell prod- neurological161 and metabolic phenotypes162. Some of these disorders
uct150. Subsequent studies discovered that the order of operations are treatable by introducing an extra gene copy or increasing the expres-
in T cell manufacturing influences genome-editing outcomes151,152, sion of the functional endogenous gene in relevant tissues. This latter
providing insights into mitigating adverse editing outcomes. approach is particularly valuable for genes exceeding the delivery
Adapting epigenome-editing technologies for primary human capacity (~4.7 kb) of conventional AAV vectors used in gene therapy.
T cells has been more challenging until recently. Fine-tuning gene Compact activating epigenome editors, like dCas9 from S. aureus163 or
expression levels in cell therapy products is critical, as overexpres- dCas12f164–166, can be packaged into a single AAV and targeted to activate
sion, underexpression or loss of expression of certain genes can have the unaffected gene copy (Fig. 4c). Another advantage of CRISPRa is its
deleterious consequences. For example, B2M, which encodes the light ability to increase expression of multiple gene isoforms167–170, whereas
invariant chain of major histocompatibility complex I (MHC-I), plays a complementary DNA delivery is typically limited to a single isoform.
central role in immune response initiation. Allogeneic T cell products Furthermore, while cell type-specific promoters can restrict transgene
often use B2M knockout to prevent transplantation rejection, but com- expression to a particular cell or tissue type, CRISPRa can target cell
plete ablation of B2M can make T cells vulnerable to attack by natural type-specific enhancers, providing an additional layer of safety against
killer cells. Thus, an ideal allogeneic T cell product should have an unwanted gene expression changes in nontarget tissues, even with
optimal level of reduction of MHC-I molecules. An ongoing clinical trial widespread effector and gRNA expression.
(NCT04649112) with a candidate allogeneic CD19 CAR T cell product A landmark study demonstrated the potential of CRISPRa to treat
(PBCAR19B) uses a short hairpin RNA targeting B2M mRNA to avoid haploinsufficiency diseases by rescuing a monogenic obesity pheno-
immune rejection and improve T cell persistence. Epigenome editing type caused by deficiency of the transcription factor SIM1 (ref. 171).
could be another viable strategy for fine-tuning expression-sensitive Targeting the Sim1 promoter or enhancer with dCas9–VP64 and a
genes such as B2M and offers the potential advantage of not requiring single gRNA in Sim1 heterozygous knockout (Sim1+/−) mice increased
constitutive expression like RNA-targeting modalities. Sim1 mRNA levels and attenuated weight gain to levels comparable
CRISPRi-mediated silencing of the PDCD1 gene, encoding the to that of the wild type. Interestingly, targeting the hypothalamic
PD-1 immune inhibitory receptor, demonstrated the feasibility of enhancer, located 270 kb downstream of the Sim1 gene, has efficacy
epigenome-editing technologies in primary human T cells146. How- comparable to that of targeting the Sim1 promoter. When targeting
ever, scaling up for high-throughput screens to discover new gene the enhancer in Sim1+/− transgenic mice expressing dCas9–VP64 and
targets posed technical challenges due to low lentiviral transduction the gRNA, upregulation of Sim1 occurred only in the hypothalamus,
rates of Cas9 and limited culture duration of primary human T cells. despite expression of dCas9–VP64 in other tissues. This highlights the

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1208

Review article https://doi.org/10.1038/s41587-024-02320-1

Table 3 | Examples of epigenome-editing applications for treating human diseases and disorders

Class Disorder or Gene(s) Effect Strategy Notes Refs.

disease (effector)

Obesity Sim1 rAAV encoding CRISPRa machinery Obesity phenotype rescued with either 171
targeting Sim1 to the PVN to rescue the dSpCas9–VP64 or dSaCas9–VP64 targeted to the
obesity phenotype Sim1 promoter or enhancer
Haplo­ Activation
insufficiency Dravet Scn1a (VP64) Dual-AAV9-based system encoding Scn1a activation with one gRNA targeting the 173
syndrome CRISPRa machinery targeting Scn1a to proximal promoter
the brain via ICV injections to attenuate
seizures
FXS Fmr1 Activation dCas9–TET1 to activate Fmr1 in iPSCs In vitro methylation-edited neurons remained 175
(TET1) and neurons stable after engraftment into the mouse brain
RTT Mecp2 Activation dCas9–TET1 to activate Mecp2 in hESCs Increased in vitro MECP2 activation in neurons 176
X-linked (TET1 ± CTCF) and neurons by targeting flanking CTCF anchor sites with
dCas12–CTCF in combination with dCas9–TET1
Infantile Cdkl5 Activation dCas9–TET1 or dCas9–VP64 to activate Co-delivery of TET1 and VP64 more robustly 177
epilepsy (TET1 ± VP64) Cdkl5 in a neuronal-like cell line activated Cdkl5 than either TET1 or VP64 alone
Hepatitis B HBV viral Repression Repression of viral promoters or Transcriptional silencing in vitro and in vivo 183–186
virus (HBV) genome enhancers for functional cure in HBV transgenic mice by introduction of
Infectious heterochromatin or DNA methylation
disease
HIV HIV viral Activation Activation of latent virus to facilitate Hotspots identified for activation of latent virus in 187–189
genome clearance by antiretrovirals infected human T cells
Regeneration Vegf ZFP–VP64 to activate multiple Endogenous activation of all VEGF isoforms; 168
and repair isoforms encoding VEGF in models of believed to be more efficacious than
angiogenesis and wound healing administering a single isoform
Activation
SCD and Hbg (ZFP–VP64a) ZFP–VP64 to activate fetal γ-globin to Epigenetic activation of Hbg recapitulates the 123,124
β-thalassemia compensate for lost adult β-globin protective effect of natural genetic variants
that create hereditary persistence of fetal
hemoglobin
Type 1 Pdx1 AAV9 encoding CRISPRa machinery Transgenic Cas9 nuclease mouse, but designed 182
diabetes targeting Pdx1 to reprogram liver cells short gRNAs (14 or 15 bp) to prevent Cas9 from
into insulin-secreting cells for treatment creating a double-stranded break
Activation of type 1 diabetes
Duchenne Dmd, (SAM) Dual-AAV9-based system encoding Also ameliorated the DMD phenotype by 182
muscular mutated; CRISPRa machinery targeting Utrn to activating klotho or Fst
dystrophy Utrn, partially rescue the DMD phenotype
(DMD) targeted
Activation AAV9 encoding TALE–VP64 to activate Increased Fxn and FXN mRNA and protein levels 192
(TALE–VP64a) Fxn in heart, muscle and liver 1 month after treatment

Friedreich’s Activation Sequence-specific synthetic Activated FXN and Fxn in primary patient cells 194
Fxn (synthetic transcription elongation factors for and a mouse xenograft model
ataxia
transcription transcription across Fxn’s intronic GAA
elongation repeats
factorsa)
Angelman Ube3a Indirect AAV-PHP.eB encoding ZFP–KRAB Restored UBE3A protein in neurons to ~25% of 197
Other syndrome activation targeting the Ube3a antisense wild-type levels
(ZFP–KRABa) transcript to activate Ube3a expression
and partially rescue the Angelman
syndrome phenotype
Retinitis Nrl Dual-AAV2-Y444F-based system Intein-mediated split-Cas9 system 180
pigmentosa encoding CRISPRi machinery to
repress Nrl and reprogram rod cells
into cone-like cells that resist the
effects of retinitis pigmentosa-specific
mutations
Repression
High LDL Pcsk9 Dual-AAV8-based system encoding Pcsk9 repression for 24 weeks after a single 178
(KRAB)
cholesterol CRISPRi machinery targeting Pcsk9 to treatment despite moderate immune response to
reduce cholesterol levels dSaCas9–KRAB
Chronic pain Scn9a Single AAV9 encoding ZFP–KRAB or Long-term efficacy in two independent pain 181
dual-AAV9 system encoding CRISPRi models
machinery to repress Scn9a for chronic
pain alleviation
Huntington’s HTT ZFP–KRAB targeting CAG repeat Molecular and functional improvement in three 195
disease expansion, selectively repressing the mouse disease models
disease allele
Tau-related MAPT Tau repression by ZFP–KRAB following Rescue of neuronal damage around amyloid 196
brain local AAV delivery to the hippocampus plaques in a mouse model of Alzheimer’s disease
disorders or systemic intravenous delivery for
whole-brain transduction
FXN, frataxin; Hbg, hemoglobin subunit γ1 (HBG1); hESCs, human ESCs; ICV, intracerebroventricular; LDL, low-density lipoprotein; PVN, paraventricular nuclei of the
hypothalamus; rAAV, recombinant AAV; VEGF, vascular endothelial growth factor (VEGFA). aDNA-binding platform other than CRISPR.

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1209

Review article https://doi.org/10.1038/s41587-024-02320-1

utility of tissue-specific enhancers for tissue- or cell type-specific gene When the AAV encoding the CRISPRa gRNA library was administered to
activation. A similar approach was applied to a mouse model of Dravet mice with tumors, it stimulated a potent immune response, increasing
syndrome, an epileptic disorder resulting from Scn1a haploinsuffi- tumor-reactive lymphocytes, reducing tumor growth and improv-
ciency172. In Scn1a+/− mice, intracerebroventricular injections of AAV ing remission rates. Although this study demonstrates the potential
encoding dCas9–VP64 and single-guide RNA (sgRNA) targeting the of large-scale cancer antigen targeting with pooled gRNA libraries,
Scn1a promoter improved seizure-related phenotypes173. A subsequent building from the concept of other neoantigen vaccines191, long-term
study targeted activation of SCN1A with AAV delivery of transactiva- safety considerations, especially potential aberrant gene activation
tors based on ZFPs174. In addition to ameliorating the phenotype in the in non-tumor tissues like the liver, spleen, heart and lung, warrant
mouse model, safety and biodistribution were examined in non-human further investigation.
primates to further support clinical development. Other DNA-binding platforms such as TALEs and ZFPs have also
proven useful for therapeutic applications. For example, custom
X-linked disorders TALE-based activators192,193 and synthetic transcription elongation
Epigenome editing is a promising therapeutic approach to rescue dis- factors194 activated expression of the epigenetically silenced FXN gene
orders resulting from the insufficient expression of genetically intact (as a result of intronic trinucleotide GAA repeat expansion) in Friedre-
but epigenetically silenced genes on the X chromosome. Epigenome ich’s ataxia mouse models and patient cells. One of the first epigenome
editing has been applied to activate endogenous expression of target editors to target endogenous genes was ZFP–VP64, used for activa-
genes implicated in fragile X syndrome (FXS)175, Rett syndrome (RTT)176 tion of Vegf to stimulate tissue regeneration and repair168, as well as
and infantile epilepsy177. FXS is caused by FMR1 gene silencing due to activation of γ-globin expression to compensate for lost β-globin in
hypermethylation of the CGG expansion mutation in the 5′ untranslated models of SCD and β-thalassemia123,124. Recently, artificial ZFP–KRAB
region of FMR1. dCas9-mediated targeting of the methylcytosine dioxy- fusions have been used for HTT repression for Huntington’s disease195
genase TET1 to these CGG repeats reversed hypermethylation, leading and MAPT repression for tau-related human brain diseases196. Moreo-
to FMR1 expression in FXS-derived iPSCs and neurons, and rescued elec- ver, ZFP–KRAB targeting the long noncoding ubiquitin protein ligase
trophysiological abnormalities of FXS neurons175. FMR1 expression in E3A (UBE3A) antisense transcript restored UBE3A protein to 25% of
edited neurons remained stable after engraftment into the mouse brain. wild-type levels in an Angelman syndrome mouse model and partially
RTT is a rare genetic neurological disorder that arises from het- rescued exploratory motor abilities197.
erozygous mutations in the MECP2 gene, offering a potential therapeutic
avenue through epigenome editing for female patients with RTT, who Developing a toolkit for combinatorial epigenome-editing
carry one functional and one mutated copy of MECP2. During develop- screens
ment, random X chromosome inactivation leads to roughly 50% of neu- A longstanding goal in biology is to unravel complex gene networks
rons lacking functional methyl-CpG-binding protein 2 (MECP2) protein. governing cellular functions in vivo. Mapping gene interactions in vivo
Using an approach similar to the one in the FXS study, demethylation of requires both relevant mouse models and robust editors for combi-
the MECP2 promoter with dCas9–TET1 activated MECP2 in human ESCs, natorial perturbations. Fortunately, progress has been made on both
restoring levels to those of wild-type cells176. However, in RTT neurons, fronts. Transgenic mice have been generated, featuring Cre-dependent
stable restoration reached only 30% of wild-type MECP2 levels. This dCas9–effectors or active Cas9, enabling proof-of-concept studies
efficiency was increased to 59% through a multiplex epigenome-editing for therapeutic purposes179,198 and gene-discovery screens. In fact, an
strategy employing dCas9–TET1 targeting the MECP2 promoter and in vivo CRISPRa screen successfully identified factors associated with
targeted CTCF insulation of edited MECP2 with dCas12–CTCF at flank- mouse liver regeneration, highlighting the potential of these technolo-
ing CTCF anchor sites. Future work could assess the applicability of gies to explore gene function within live animals199. To facilitate multi-
multiplexed epigenome editing to endogenous genes beyond MECP2. plex epigenome editing in vivo, the activity of dCas12a was optimized.
Finally, a study reducing DNA methylation of Cdkl5, a causative gene This versatile platform can target multiple genomic loci by processing
for infantile epilepsy, demonstrated that co-delivery of TET1 and VP64 numerous CRISPR RNAs from a single transcript, as demonstrated by
led to more robust gene activation, eclipsing 60% of wild-type Cdkl5 activating three endogenous genes in the retinas of post-natal mice200.
levels177. These studies highlight the utility of TET1 to activate DNA Thus, dCas12a could potentially be leveraged for combinatorial gene
methylation-sensitive genes and underscore the potential synergistic screens in vivo. However, in vivo screens are often limited to organs like
effects of delivering multiple, mechanistically distinct effector proteins. the liver, where gene delivery is sufficient to yield enough transduced
cells for adequate library coverage and avoid bottlenecking issues201,202.
Other classes of disease Achieving the required coverage, especially with factorial growth of
CRISPR epigenome editors have found success in various other disease combinatorial libraries, poses challenges. For example, targeting all
applications. Examples of gene silencing include Pcsk9 repression in the combinations of three genes from only 100 candidate genes (that is,
mouse liver to lower low-density lipoprotein cholesterol levels48,178,179, 100 choose three) would require a 161,700-member gRNA library with
Nrl repression to reprogram rod cells into cone-like cells that resist only one gRNA per gene, already surpassing the size of genome-wide
retinitis pigmentosa-specific mutations180 and Scn9a repression to libraries targeting all human genes. Thus, computational methods are
alleviate chronic pain181. Examples of gene activation include Pdx1 gene essential for gene prioritization and predicting combinatorial effects
activation to transdifferentiate liver cells into insulin-producing cells between genes. In fact, an algorithm has been developed to predict
for type 1 diabetes179,182 and utrophin activation to compensate for the the transcriptional response of combinatorial perturbations based on
lost dystrophin gene in an mdx mouse model of Duchenne muscular single-perturbation scRNA-seq data203. However, a substantial knowl-
dystrophy (DMD)182. Both repression and activation strategies have edge gap remains between transcriptional profiles and phenotypic
demonstrated potential in the area of infectious disease, including effects at the cellular and organism levels.
silencing of hepatitis B viral DNA transcription183–186 and reactivation
of latent human immunodeficiency virus (HIV) to enable clearance by Challenges of epigenome engineering
antiretroviral treatment187–189. Causality
A unique CRISPRa application was devised to craft personal- Decoupling the associative presence of epigenetic modifications
ized treatments for solid mammary tumors in mice190. This approach from the cause of transcriptional changes remains an important area
involved creating a gRNA library that targeted mutated endogenous of research57. To define causality, the field can build from existing
genes, aiming to increase the expression of tumor-associated antigens. epigenome-editing tools like dCas9 fusions with epigenetic modifiers

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1210

Review article https://doi.org/10.1038/s41587-024-02320-1

(for example, TET1, DNMT3A and p300) to expand the toolbox of editors Timing
that write or erase specific epigenetic marks51. Synthetic DNA-binding The timing of treatment can be critical, as, in some cases, therapeutic
modules that precisely write orthogonal epigenetic marks, which can efficacy is limited to early developmental stages. For instance, deliver-
only be read by synthetic readers, provide an additional toolkit for investi- ing CRISPRa components before the onset of abnormal weight gain in
gating and encoding epigenetic regulation204. Finally, orthogonal CRISPR Sim1+/− mice resulted in normal growth171. However, it is unclear whether
systems using dCas9 proteins, dCas12 proteins and gRNA or CRISPR RNA restoring Sim1 gene expression after the obesity phenotype manifests
molecules from various bacterial species allow for simultaneous recruit- can attenuate or reduce weight gain. As SIM1 deficiency in humans is
ment of distinct epigenetic effectors to specific genomic sites, enabling often identified only after manifestation of severe obesity213,214, a treat-
installation of more complex networks of epigenetic modifications205,206. ment would need to be effective after symptom onset. These consid-
erations apply to other disorders affecting various tissues at various
Context developmental stages.
The activity of epigenome editors is dependent on the specific chroma-
tin context and cell type. For example, epigenome editors that deposit Dose
or erase DNA methylation are dependent on CpG content at the target The level of gene activation or repression relative to wild-type levels is
site. In addition, cell type-specific epigenetic signatures can influence also critical for therapeutic applications that require precise control
CRISPR–Cas-based DNA-binding dynamics207, making efficacy incon- over gene dosage. Although supraphysiological expression may not
sistent across cell types. Epigenetic context should therefore be con- be problematic for some genes, others, such as MECP2, require precise
sidered during the design and interpretation of screens. To this end, dosage control because physiological abnormalities can occur with
CRISPR-based screens should use multiple orthogonal epigenome edi- both haploinsufficiency and duplication155. Conversely, inadequate
tors (for example, VP64 and TET1) across several cell types. Compiling levels of gene activation or repression might not rescue specific pheno-
these types of datasets should help to construct a general framework for types. The selection of delivery vector, promoter, Cas protein, effector
selecting the optimal epigenome-editing tool for a specific epigenetic domain and gRNA warrants careful consideration, as these components
context, as demonstrated by recent advances51. collectively determine gene expression levels. These technical chal-
lenges are further exacerbated by our limited understanding of the
Delivery optimal range or threshold of gene expression required to rescue some
A critical barrier to the clinical translation of epigenome-editing tech- specific disease phenotypes. Animal models that recapitulate disease
nologies is in vivo delivery. In vivo delivery can be accomplished using phenotypes will be essential to provide the biological framework to
either viral (for example, AAV, adenovirus, lentivirus) or nonviral (for develop effective epigenome-editing therapies. In the case of diseases
example, electroporation, lipid nanoparticles, polymeric vehicles) vec- lacking relevant preclinical models, first-in-human clinical trials might
tors. Important considerations when selecting a delivery vector include be necessary to answer these questions.
potential for genomic integration, timing and duration of transgene
expression, cargo capacity, immunogenicity, biodistribution and deliv- Specificity
ery efficiency and cell type selectivity. These specific details are outside The specificity of epigenome-editing tools, especially for therapeutic
the scope of this Review, but delivery methods have been extensively applications, is an important safety parameter. While gRNA design
reviewed elsewhere163–165. A shared challenge across delivery platforms is can mitigate off-target Cas9 binding, the extent of promiscuous
selectively targeting disease-relevant tissues. Continued efforts to evolve gRNA-independent effects of various epigenome-editing effector
capsid tropisms for viral vectors and integrate specific moieties such as domains has been underexplored. Ideal assays to define specificity
antibodies208, glycoproteins209 and lipids210 into nonviral vectors should may need to be catered to the epigenome editor. The most strin-
reshape our ability to target payloads to specific cell types and tissues. gent measure of effector specificity would be directly assaying the
deposited or erased epigenetic mark with comprehensive, unbiased
Durability genome-wide methods. However, genomic assays alone may provide
Another key consideration is the durability of gene expression changes. a limited view of the safety profile of an epigenome-editing therapy
Depending on the application, sustained or temporary changes in if nonspecific chromatin modifications are transient or insufficient
gene expression may be desired outcomes. For example, treatment to alter gene expression. To address this limitation, genomic assays
of haploinsufficiency diseases likely requires permanent upregulation should be complemented with functional assays to determine the phe-
of the missing gene product, whereas cellular reprogramming, tissue notypic effects of nonspecific chromatin modifications and identify
regeneration or elimination of infectious disease may only necessitate any potential consequences. Select studies have evaluated specific-
transient changes in gene expression211. Permanent gene silencing or ity with such assays, including RNA-seq for transcriptional changes,
activation requires either sustained expression of an epigenome editor chromatin immunoprecipitation followed by sequencing (ChIP-seq)
that can transiently modulate gene expression or a transient pulse of an for histone marks, assay for transposase-accessible chromatin with
epigenome editor that can impart stable, heritable epigenetic changes. sequencing (ATAC-seq) for chromatin structure and genome-wide
While the latter strategy offers substantial flexibility in delivery and bisulfite sequencing for DNA methylation. While some studies have
could pave the way for one-time, curative epigenome therapies48, it is showed remarkable specificity19,20,24,171,215, others have reported
limited by the small toolkit of editors capable of installing epigenetic gRNA-independent off-target activities of certain effector domains179,216
memory, especially for gene activation. The combination of protein that likely are dependent on delivery method, expression level and
engineering to optimize effector function, large-scale screens to duration, and other protein-engineering aspects that dictate activity.
discover novel epigenetic effectors and improved understanding of Similar efforts using ZFP-based epigenome editors have demonstrated
epigenetic rules should expand the toolkit. Alternatively, a transient the ability to minimize off-target editing by careful selection of the
pulse of an epigenome editor with temporary effects is an option if DNA-targeting domain48.
redosing is feasible. For example, a recent study used lipid nanopar-
ticles to deliver mRNA of the potent viral activator VPH–dCas9–SS18, Non-CRISPR systems
composed of a VP64–p65–heat shock TF 1 (HSF1) fusion and the SWI– While CRISPR-based systems can rapidly nominate therapeutic targets,
SNF chromatin-remodeling subunit SS18, to the mouse liver, leading they also present challenges for specific applications as a result of their
to a transient increase in expression of the target gene erythropoietin relatively large size and potential immunogenicity. Epigenome editors
that was repeatable with redosing212. built from compact, human-derived ZFP domains could circumvent

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1211

Review article https://doi.org/10.1038/s41587-024-02320-1

these challenges48,168,195,196. However, the development of epigenome 10. Porto, E. M., Komor, A. C., Slaymaker, I. M. & Yeo, G. W. Base
editors with synthetic ZFPs has lagged behind that of CRISPR-based editing: advances and therapeutic opportunities. Nat. Rev. Drug
tools because they require complex protein engineering. Fortunately, a Discov. 19, 839–859 (2020).
deep learning model was recently developed to solve zinc finger design 11. Zhao, Z., Shang, P., Mohanraju, P. & Geijsen, N. Prime editing:
for any genomic target217. To capitalize on the distinct advantages of advances and therapeutic applications. Trends Biotechnol. 41,
each platform, a robust pipeline might leverage CRISPR-based systems 1000–1012 (2023).
for target discovery, while diligently exploring non-CRISPR systems 12. Holtzman, L. & Gersbach, C. A. Editing the epigenome: reshaping
for product development. the genomic landscape. Annu. Rev. Genomics Hum. Genet. 19,
43–71 (2018).
Outlook 13. Fischle, W., Wang, Y. & Allis, C. D. Histone and chromatin
Epigenome editing has the potential to define mechanisms of gene cross-talk. Curr. Opin. Cell Biol. 15, 172–183 (2003).
regulation, program cell phenotypes, develop disease models and 14. Esteller, M. Epigenetics in cancer. N. Engl. J. Med. 358, 1148–1159
establish new therapies for genetic disorders. While many of the first (2008).
therapeutic applications have focused on specific classes of diseases 15. Feinberg, A. P. Phenotypic plasticity and the epigenetics of
(for example, haploinsufficiency, X-linked inheritance), epigenome human disease. Nature 447, 433–440 (2007).
editing is not limited to diseases that manifest from dysregulated 16. Zoghbi, H. Y. & Beaudet, A. L. Epigenetics and human disease.
expression of a single gene. Widespread epigenetic changes are impli- Cold Spring Harb. Perspect. Biol. 8, a019497 (2016).
cated in cancer, neurological diseases and aging, and we anticipate that 17. Cheng, A. W. et al. Multiplexed activation of endogenous genes
epigenome editing will impact these therapeutic areas in the coming by CRISPR-on, an RNA-guided transcriptional activator system.
years. In fact, many efforts are underway with the objective to slow or Cell Res. 23, 1163–1171 (2013).
reverse the aging process through cellular reprogramming or engineer 18. Maeder, M. L. et al. CRISPR RNA-guided activation of endogenous
resilience to frailty or degeneration into biological systems. Epigenome human genes. Nat. Methods 10, 977–979 (2013).
editing will likely play an instrumental role in both target discovery and 19. Hilton, I. B. et al. Epigenome editing by a CRISPR–Cas9-based
therapeutic intervention for these efforts. Although epigenome editing acetyltransferase activates genes from promoters and enhancers.
alone may not suffice for complex diseases caused by combinations Nat. Biotechnol. 33, 510–517 (2015).
of genetic and environmental factors, understanding the interplay 20. Thakore, P. I. et al. Highly specific epigenome editing by CRISPR–
between these variables may enable combination therapies or pairing Cas9 repressors for silencing of distal regulatory elements.
of epigenetic modulation with behavioral adjustments. Nat. Methods 12, 1143–1149 (2015).
More broadly, we expect that the emergence of artificial intel- 21. Schmidt, R. et al. CRISPR activation and interference screens
ligence and machine learning algorithms will profoundly impact criti- decode stimulation responses in primary human T cells. Science
cal challenges facing epigenome editing such as deconvoluting the 375, eabj4008 (2022).
contribution of specific epigenetic marks on gene expression across 22. Klann, T. S. et al. CRISPR–Cas9 epigenome editing enables
genetic backgrounds, cell types and environmental conditions, in high-throughput screening for functional regulatory elements in
addition to optimizing gRNA design across CRISPR enzymes, engineer- the human genome. Nat. Biotechnol. 35, 561–568 (2017).
ing new DNA-targeting proteins and effector domains and designing 23. Gilbert, L. A. et al. CRISPR-mediated modular RNA-guided
and refining delivery vehicles. In summary, parallel efforts to define regulation of transcription in eukaryotes. Cell 154, 442–451
epigenetic rules, improve targeted epigenome editors and develop (2013).
safe and efficient delivery vehicles could transform biological research 24. Perez-Pinera, P. et al. RNA-guided gene activation by CRISPR–
and medical practice. Cas9-based transcription factors. Nat. Methods 10, 973–976 (2013).
25. Kearns, N. A. et al. Functional annotation of native enhancers with
References a Cas9–histone demethylase fusion. Nat. Methods 12, 401–403
1. ENCODE Project Consortium. An integrated encyclopedia of DNA (2015).
elements in the human genome. Nature 489, 57–74 (2012). 26. Thakore, P. I., Black, J. B., Hilton, I. B. & Gersbach, C. A. Editing the
2. Luo, Y. et al. New developments on the Encyclopedia of DNA epigenome: technologies for programmable transcription and
Elements (ENCODE) data portal. Nucleic Acids Res. 48, epigenetic modulation. Nat. Methods 13, 127–137 (2016).
D882–D889 (2020). 27. Nakamura, M., Gao, Y., Dominguez, A. A. & Qi, L. S. CRISPR
3. Waddington, C. H. The Strategy of the Genes. A Discussion of technologies for precise epigenome editing. Nat. Cell Biol. 23,
Some Aspect of Theoretical Biology (Allen & Unwin, 1957). 11–22 (2021).
4. Gersbach, C. A., Gaj, T. & Barbas, C. F. 3rd Synthetic zinc finger 28. Villiger, L. et al. CRISPR technologies for genome, epigenome
proteins: the advent of targeted gene regulation and genome and transcriptome editing. Nat. Rev. Mol. Cell Biol. 25, 464–487
modification technologies. Acc. Chem. Res. 47, 2309–2318 (2024).
(2014). 29. Gjaltema, R. A. F. & Rots, M. G. Advances of epigenetic editing.
5. Jinek, M. et al. A programmable dual-RNA-guided DNA Curr. Opin. Chem. Biol. 57, 75–81 (2020).
endonuclease in adaptive bacterial immunity. Science 337, 30. Sgro, A. & Blancafort, P. Epigenome engineering: new
816–821 (2012). technologies for precision medicine. Nucleic Acids Res. 48,
6. Cong, L. et al. Multiplex genome engineering using CRISPR/Cas 12453–12482 (2020).
systems. Science 339, 819–823 (2013). 31. Farzadfard, F., Perli, S. D. & Lu, T. K. Tunable and multifunctional
7. Mali, P. et al. RNA-guided human genome engineering via Cas9. eukaryotic transcription factors based on CRISPR/Cas. ACS Synth.
Science 339, 823–826 (2013). Biol. 2, 604–613 (2013).
8. Qi, L. S. et al. Repurposing CRISPR as an RNA-guided platform for 32. Perez-Pinera, P. et al. Synergistic and tunable human gene
sequence-specific control of gene expression. Cell 152, 1173–1183 activation by combinations of synthetic transcription factors.
(2013). Nat. Methods 10, 239–242 (2013).
9. Anzalone, A. V., Koblan, L. W. & Liu, D. R. Genome editing with 33. Maeder, M. L. et al. Robust, synergistic regulation of human gene
CRISPR–Cas nucleases, base editors, transposases and prime expression using TALE activators. Nat. Methods 10, 243–245
editors. Nat. Biotechnol. 38, 824–844 (2020). (2013).

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1212

Review article https://doi.org/10.1038/s41587-024-02320-1

34. Chakraborty, S. et al. A CRISPR/Cas9-based system for 57. O’Geen, H. et al. dCas9-based epigenome editing suggests
reprogramming cell lineage specification. Stem Cell Rep. 3, acquisition of histone methylation is not sufficient for target gene
940–947 (2014). repression. Nucleic Acids Res. 45, 9901–9916 (2017).
35. Black, J. B. et al. Targeted epigenetic remodeling of endogenous 58. Cano-Rodriguez, D. et al. Writing of H3K4me3 overcomes
loci by CRISPR/Cas9-based transcriptional activators directly epigenetic silencing in a sustained but context-dependent
converts fibroblasts to neuronal cells. Cell Stem Cell 19, 406–414 manner. Nat. Commun. 7, 12284 (2016).
(2016). 59. Adhikari, A. et al. Functional rescue in an Angelman syndrome
36. Chavez, A. et al. Highly efficient Cas9-mediated transcriptional model following treatment with lentivector transduced
programming. Nat. Methods 12, 326–328 (2015). hematopoietic stem cells. Hum. Mol. Genet. 30, 1067–1083
37. Tanenbaum, M. E., Gilbert, L. A., Qi, L. S., Weissman, J. S. & Vale, (2021).
R. D. A protein-tagging system for signal amplification in gene 60. Choudhury, S. R., Cui, Y., Lubecka, K., Stefanska, B. &
expression and fluorescence imaging. Cell 159, 635–646 (2014). Irudayaraj, J. CRISPR–dCas9 mediated TET1 targeting for selective
38. Konermann, S. et al. Genome-scale transcriptional activation DNA demethylation at BRCA1 promoter. Oncotarget 7,
by an engineered CRISPR–Cas9 complex. Nature 517, 583–588 46545–46556 (2016).
(2015). 61. Ford, E. et al. A modular dCas9–SunTag DNMT3A epigenome
39. Mahata, B. et al. Compact engineered human mechanosensitive editing system overcomes pervasive off-target activity of direct
transactivation modules enable potent and versatile synthetic fusion dCas9–DNMT3A constructs. Genome Res. 28, 1193–1206
transcriptional control. Nat. Methods 20, 1716–1728 (2023). (2018).
40. Chavez, A. et al. Comparison of Cas9 activators in multiple 62. Yao, D. et al. Multicenter integrated analysis of noncoding
species. Nat. Methods 13, 563–567 (2016). CRISPRi screens. Nat. Methods 21, 723–734 (2024).
41. Amabile, A. et al. Inheritable silencing of endogenous genes by 63. Horlbeck, M. A. et al. Compact and highly active next-generation
hit-and-run targeted epigenetic editing. Cell 167, 219–232 (2016). libraries for CRISPR-mediated gene repression and activation.
42. Bintu, L. et al. Dynamics of epigenetic regulation at the single-cell eLife 5, e19760 (2016).
level. Science 351, 720–724 (2016). 64. Sanson, K. R. et al. Optimized libraries for CRISPR–Cas9
43. Alerasool, N., Segal, D., Lee, H. & Taipale, M. An efficient KRAB genetic screens with multiple modalities. Nat. Commun. 9, 5416
domain for CRISPRi applications in human cells. Nat. Methods 17, (2018).
1093–1096 (2020). 65. Liu, Y. et al. CRISPR activation screens systematically identify
44. Tycko, J. et al. High-throughput discovery and characterization of factors that drive neuronal fate and reprogramming. Cell Stem
human transcriptional effectors. Cell 183, 2020–2035 (2020). Cell 23, 758–771 (2018).
45. O’Geen, H. et al. Ezh2–dCas9 and KRAB–dCas9 enable 66. Black, J. B. et al. Master regulators and cofactors of human
engineering of epigenetic memory in a context-dependent neuronal cell fate specification identified by CRISPR gene
manner. Epigenetics Chromatin 12, 26 (2019). activation screens. Cell Rep. 33, 108460 (2020).
46. O’Geen, H., Tomkova, M., Combs, J. A., Tilley, E. K. & Segal, 67. Yang, J. et al. Genome-scale CRISPRa screen identifies novel
D. J. Determinants of heritable gene silencing for KRAB– factors for cellular reprogramming. Stem Cell Rep. 12, 757–771
dCas9 + DNMT3 and Ezh2–dCas9 + DNMT3 hit-and-run epigenome (2019).
editing. Nucleic Acids Res. 50, 3239–3253 (2022). 68. Coukos, R. et al. An engineered transcriptional reporter of
47. Nunez, J. K. et al. Genome-wide programmable transcriptional protein localization identifies regulators of mitochondrial and
memory by CRISPR-based epigenome editing. Cell 184, ER membrane protein trafficking in high-throughput CRISPRi
2503–2519 (2021). screens. eLife 10, e69142 (2021).
48. Cappelluti, M. A. et al. Durable and efficient gene silencing 69. Luteijn, R. D. et al. SLC19A1 transports immunoreactive cyclic
in vivo by hit-and-run epigenome editing. Nature 627, 416–423 dinucleotides. Nature 573, 434–438 (2019).
(2024). 70. Drager, N. M. et al. A CRISPRi/a platform in human iPSC-derived
49. Ludwig, C. H. et al. High-throughput discovery and microglia uncovers regulators of disease states. Nat. Neurosci. 25,
characterization of viral transcriptional effectors in human cells. 1149–1162 (2022).
Cell Syst. 14, 482–500 (2023). 71. Tian, R. et al. Genome-wide CRISPRi/a screens in human neurons
50. Alerasool, N., Leng, H., Lin, Z. Y., Gingras, A. C. & Taipale, M. link lysosomal failure to ferroptosis. Nat. Neurosci. 24, 1020–1034
Identification and functional characterization of transcriptional (2021).
activators in human cells. Mol. Cell 82, 677–695 (2022). 72. Liu, S. J. et al. CRISPRi-based genome-scale identification of
51. Policarpi, C., Munafo, M., Tsagkris, S., Carlini, V. & Hackett, J. A. functional long noncoding RNA loci in human cells. Science 355,
Systematic epigenome editing captures the context-dependent aah7111 (2017).
instructive function of chromatin modifications. Nat. Genet. 56, 73. Klann, T. S., Black, J. B. & Gersbach, C. A. CRISPR-based methods
1168–1180 (2024). for high-throughput annotation of regulatory DNA. Curr. Opin.
52. Liu, X. S. et al. Editing DNA methylation in the mammalian Biotechnol. 52, 32–41 (2018).
genome. Cell 167, 233–247 (2016). 74. Takahashi, K. & Yamanaka, S. Induction of pluripotent stem cells
53. Morita, S. et al. Targeted DNA demethylation in vivo using dCas9– from mouse embryonic and adult fibroblast cultures by defined
peptide repeat and scFv–TET1 catalytic domain fusions. factors. Cell 126, 663–676 (2006).
Nat. Biotechnol. 34, 1060–1065 (2016). 75. Davis, R. L., Weintraub, H. & Lassar, A. B. Expression of a single
54. Lei, Y. et al. Targeted DNA methylation in vivo using an engineered transfected cDNA converts fibroblasts to myoblasts. Cell 51,
dCas9–MQ1 fusion protein. Nat. Commun. 8, 16026 (2017). 987–1000 (1987).
55. Xiong, T. et al. Targeted DNA methylation in human cells using 76. Vierbuchen, T. & Wernig, M. Molecular roadblocks for cellular
engineered dCas9–methyltransferases. Sci. Rep. 7, 6732 (2017). reprogramming. Mol. Cell 47, 827–838 (2012).
56. Sapozhnikov, D. M. & Szyf, M. Unraveling the functional role of 77. Weintraub, H. et al. Activation of muscle-specific genes in
DNA demethylation at specific promoters by targeted steric pigment, nerve, fat, liver, and fibroblast cell lines by forced
blockage of DNA methyltransferase with CRISPR/dCas9. Nat. expression of MyoD. Proc. Natl Acad. Sci. USA 86, 5434–5438
Commun. 12, 5711 (2021). (1989).

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1213

Review article https://doi.org/10.1038/s41587-024-02320-1

78. Zhang, Y. et al. Rapid single-step induction of functional neurons 102. Gasperini, M. et al. CRISPR/Cas9-mediated scanning for
from human pluripotent stem cells. Neuron 78, 785–798 (2013). regulatory elements required for HPRT1 expression via thousands
79. Rao, L., Qian, Y., Khodabukus, A., Ribar, T. & Bursac, N. Engineering of large, programmed genomic deletions. Am. J. Hum. Genet. 101,
human pluripotent stem cells into a functional skeletal muscle 192–205 (2017).
tissue. Nat. Commun. 9, 126 (2018). 103. Korkmaz, G. et al. Functional genetic screens for enhancer
80. Vierbuchen, T. et al. Direct conversion of fibroblasts to functional elements in the human genome using CRISPR–Cas9. Nat.
neurons by defined factors. Nature 463, 1035–1041 (2010). Biotechnol. 34, 192–198 (2016).
81. D’Alessio, A. C. et al. A systematic approach to identify candidate 104. Rajagopal, N. et al. High-throughput mapping of regulatory DNA.
transcription factors that control cell identity. Stem Cell Rep. 5, Nat. Biotechnol. 34, 167–174 (2016).
763–775 (2015). 105. Sanjana, N. E. et al. High-resolution interrogation of functional
82. Morris, S. A. et al. Dissecting engineered cell types and enhancing elements in the noncoding genome. Science 353, 1545–1549
cell fate conversion via CellNet. Cell 158, 889–902 (2014). (2016).
83. Rackham, O. J. et al. A predictive computational framework for 106. Chen, P. B. et al. Systematic discovery and functional dissection
direct reprogramming between human cell types. Nat. Genet. 48, of enhancers needed for cancer cell fitness and proliferation. Cell
331–335 (2016). Rep. 41, 111630 (2022).
84. Xu, Q. et al. ANANSE: an enhancer network-based computational 107. Fulco, C. P. et al. Activity-by-contact model of enhancer–promoter
approach for predicting key transcription factors in cell fate regulation from thousands of CRISPR perturbations. Nat. Genet.
determination. Nucleic Acids Res. 49, 7966–7985 (2021). 51, 1664–1669 (2019).
85. Jung, S., Appleton, E., Ali, M., Church, G. M. & Del Sol, A. A 108. Reilly, S. K. et al. Direct characterization of cis-regulatory
computer-guided design tool to increase the efficiency of cellular elements and functional dissection of complex genetic
conversions. Nat. Commun. 12, 1659 (2021). associations using HCR–FlowFISH. Nat. Genet. 53, 1166–1176
86. Marazzi, L., Shah, M., Balakrishnan, S., Patil, A. & Vera-Licona, P. (2021).
NETISCE: a network-based tool for cell fate reprogramming. 109. Simeonov, D. R. et al. Discovery of stimulation-responsive immune
NPJ Syst. Biol. Appl. 8, 21 (2022). enhancers with CRISPR activation. Nature 549, 111–115 (2017).
87. Joung, J. et al. A transcription factor atlas of directed 110. Mowery, C. T. et al. Systematic decoding of cis gene regulation
differentiation. Cell 186, 209–229 (2023). defines context-dependent control of the multi-gene
88. Ng, A. H. M. et al. A comprehensive library of human transcription costimulatory receptor locus in human T cells. Nat. Genet. 56,
factors for cell fate engineering. Nat. Biotechnol. 39, 510–519 (2021). 1156–1167 (2024).
89. Parekh, U. et al. Mapping cellular reprogramming via pooled 111. Dixit, A. et al. Perturb-seq: dissecting molecular circuits with
overexpression screens with paired fitness and single-cell scalable single-cell RNA profiling of pooled genetic screens. Cell
RNA-sequencing readout. Cell Syst. 7, 548–555 (2018). 167, 1853–1866 (2016).
90. Liu, P., Chen, M., Liu, Y., Qi, L. S. & Ding, S. CRISPR-based 112. Adamson, B. et al. A multiplexed single-cell CRISPR screening
chromatin remodeling of the endogenous Oct4 or Sox2 locus platform enables systematic dissection of the unfolded protein
enables reprogramming to pluripotency. Cell Stem Cell 22, response. Cell 167, 1867–1882 (2016).
252–261 (2018). 113. Datlinger, P. et al. Pooled CRISPR screening with single-cell
91. Balboa, D. et al. Conditionally stabilized dCas9 activator for transcriptome readout. Nat. Methods 14, 297–301 (2017).
controlling gene expression in human cell reprogramming and 114. Jaitin, D. A. et al. Dissecting immune circuits by linking
differentiation. Stem Cell Rep. 5, 448–459 (2015). CRISPR-pooled screens with single-cell RNA-seq. Cell 167,
92. Wei, S. et al. Conversion of embryonic stem cells into 1883–1896 (2016).
extraembryonic lineages by CRISPR-mediated activators. Sci. 115. Gasperini, M. et al. A genome-wide framework for mapping gene
Rep. 6, 19648 (2016). regulation via cellular genetic screens. Cell 176, 377–390 (2019).
93. Gao, X. et al. Reprogramming to pluripotency using designer 116. Yao, D. et al. Scalable genetic screening for regulatory circuits
TALE transcription factors targeting enhancers. Stem Cell Rep. 1, using compressed Perturb-seq. Nat. Biotechnol. https://doi.org/
183–197 (2013). 10.1038/s41587-023-01964-9 (2023).
94. Gao, X. et al. Comparison of TALE designer transcription factors 117. Klann, T. S. et al. Genome-wide annotation of gene regulatory
and the CRISPR/dCas9 in regulation of gene expression by elements linked to cell fitness. Preprint at bioRxiv https://doi.
targeting enhancers. Nucleic Acids Res. 42, e155 (2014). org/10.1101/2021.03.08.434470 (2021).
95. Sollis, E. et al. The NHGRI-EBI GWAS Catalog: knowledgebase and 118. Cosgrove, B. D. et al. Mechanosensitive genomic enhancers
deposition resource. Nucleic Acids Res. 51, D977–D985 (2023). potentiate the cellular response to matrix stiffness. Preprint at
96. Arnold, C. D. et al. Genome-wide quantitative enhancer activity bioRxiv https://doi.org/10.1101/2024.01.10.574997 (2024).
maps identified by STARR-seq. Science 339, 1074–1077 (2013). 119. Feldman, D. et al. Optical pooled screens in human cells. Cell 179,
97. Melnikov, A. et al. Systematic dissection and optimization of 787–799 (2019).
inducible enhancers in human cells using a massively parallel 120. Frangoul, H. et al. CRISPR–Cas9 gene editing for sickle cell
reporter assay. Nat. Biotechnol. 30, 271–277 (2012). disease and β-thalassemia. N. Engl. J. Med. 384, 252–260 (2021).
98. Fulco, C. P. et al. Systematic mapping of functional enhancer– 121. Uda, M. et al. Genome-wide association study shows BCL11A
promoter connections with CRISPR interference. Science 354, associated with persistent fetal hemoglobin and amelioration of
769–773 (2016). the phenotype of β-thalassemia. Proc. Natl Acad. Sci. USA 105,
99. Canver, M. C. et al. BCL11A enhancer dissection by Cas9-mediated 1620–1625 (2008).
in situ saturating mutagenesis. Nature 527, 192–197 (2015). 122. Vierstra, J. et al. Functional footprinting of regulatory DNA. Nat.
100. Diao, Y. et al. A new class of temporarily phenotypic enhancers Methods 12, 927–930 (2015).
identified by CRISPR/Cas9-mediated genetic screening. Genome 123. Graslund, T., Li, X., Magnenat, L., Popkov, M. & Barbas, C. F. 3rd
Res. 26, 397–405 (2016). Exploring strategies for the design of artificial transcription
101. Diao, Y. et al. A tiling-deletion-based genetic screen for factors: targeting sites proximal to known regulatory regions for
cis-regulatory element identification in mammalian cells. Nat. the induction of γ-globin expression and the treatment of sickle
Methods 14, 629–635 (2017). cell disease. J. Biol. Chem. 280, 3707–3714 (2005).

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1214

Review article https://doi.org/10.1038/s41587-024-02320-1

124. Wilber, A. et al. A zinc-finger transcriptional activator designed 150. Stadtmauer, E. A. et al. CRISPR-engineered T cells in patients with
to interact with the γ-globin gene promoters enhances fetal refractory cancer. Science 367, eaba7365 (2020).
hemoglobin production in primary human adult erythroblasts. 151. Nahmad, A. D. et al. Frequent aneuploidy in primary human T cells
Blood 115, 3033–3041 (2010). after CRISPR–Cas9 cleavage. Nat. Biotechnol. 40, 1807–1813
125. Lim, W. A. & June, C. H. The principles of engineering immune (2022).
cells to treat cancer. Cell 168, 724–740 (2017). 152. Tsuchida, C. A. et al. Mitigation of chromosome loss in clinical
126. Majzner, R. G. & Mackall, C. L. Clinical lessons learned from the CRISPR–Cas9-engineered T cells. Cell 186, 4567–4582 (2023).
first leg of the CAR T cell journey. Nat. Med. 25, 1341–1355 (2019). 153. He, B. et al. CD8+ T cells utilize highly dynamic enhancer
127. Labanieh, L. & Mackall, C. L. CAR immune cells: design principles, repertoires and regulatory circuitry in response to infections.
resistance and the next generation. Nature 614, 635–648 (2023). Immunity 45, 1341–1354 (2016).
128. Krishna, S. et al. Stem-like CD8 T cells mediate response of 154. Henning, A. N., Roychoudhuri, R. & Restifo, N. P. Epigenetic
adoptive cell immunotherapy against human cancer. Science control of CD8+ T cell differentiation. Nat. Rev. Immunol. 18,
370, 1328–1334 (2020). 340–356 (2018).
129. Fraietta, J. A. et al. Determinants of response and resistance to 155. Na, E. S., Nelson, E. D., Kavalali, E. T. & Monteggia, L. M. The
CD19 chimeric antigen receptor (CAR) T cell therapy of chronic impact of MeCP2 loss- or gain-of-function on synaptic plasticity.
lymphocytic leukemia. Nat. Med. 24, 563–571 (2018). Neuropsychopharmacology 38, 212–219 (2013).
130. Haradhvala, N. J. et al. Distinct cellular dynamics associated with 156. Thivierge, C. et al. Overexpression of PKD1 causes polycystic
response to CAR-T therapy for refractory B cell lymphoma. Nat. kidney disease. Mol. Cell. Biol. 26, 1538–1548 (2006).
Med. 28, 1848–1859 (2022). 157. Rice, A. M. & McLysaght, A. Dosage sensitivity is a major
131. Roybal, K. T. et al. Precision tumor recognition by T cells with determinant of human copy number variant pathogenicity. Nat.
combinatorial antigen-sensing circuits. Cell 164, 770–779 (2016). Commun. 8, 14366 (2017).
132. Zhu, I. et al. Modular design of synthetic receptors for 158. Moss, T. J. & Wallrath, L. L. Connections between epigenetic gene
programmed gene regulation in cell therapies. Cell 185, silencing and human disease. Mutat. Res. 618, 163–174 (2007).
1431–1443 (2022). 159. Huang, N., Lee, I., Marcotte, E. M. & Hurles, M. E. Characterising
133. Kloss, C. C., Condomines, M., Cartellieri, M., Bachmann, M. & and predicting haploinsufficiency in the human genome. PLoS
Sadelain, M. Combinatorial antigen recognition with balanced Genet. 6, e1001154 (2010).
signaling promotes selective tumor eradication by engineered 160. Dang, V. T., Kassahn, K. S., Marcos, A. E. & Ragan, M. A.
T cells. Nat. Biotechnol. 31, 71–75 (2013). Identification of human haploinsufficient genes and their
134. Tousley, A. M. et al. Co-opting signalling molecules enables genomic proximity to segmental duplications. Eur. J. Hum. Genet.
logic-gated control of CAR T cells. Nature 615, 507–516 (2023). 16, 1350–1357 (2008).
135. Lynn, R. C. et al. c-Jun overexpression in CAR T cells induces 161. Freischmidt, A. et al. Haploinsufficiency of TBK1 causes familial
exhaustion resistance. Nature 576, 293–300 (2019). ALS and fronto-temporal dementia. Nat. Neurosci. 18, 631–636
136. Seo, H. et al. BATF and IRF4 cooperate to counter exhaustion in (2015).
tumor-infiltrating CAR T cells. Nat. Immunol. 22, 983–995 (2021). 162. Seidner, G. et al. GLUT-1 deficiency syndrome caused by
137. Legut, M. et al. A genome-scale screen for synthetic drivers of haploinsufficiency of the blood–brain barrier hexose carrier.
T cell proliferation. Nature 603, 728–735 (2022). Nat. Genet. 18, 188–191 (1998).
138. Blaeschke, F. et al. Modular pooled discovery of synthetic 163. Ran, F. A. et al. In vivo genome editing using Staphylococcus
knockin sequences to program durable cell therapies. Cell 186, aureus Cas9. Nature 520, 186–191 (2015).
4216–4234 (2023). 164. Xu, X. et al. Engineered miniature CRISPR–Cas system for
139. McCutcheon, S. R. et al. Transcriptional and epigenetic regulators mammalian genome regulation and editing. Mol. Cell 81,
of human CD8+ T cell function identified through orthogonal 4333–4345 (2021).
CRISPR screens. Nat. Genet. 55, 2211–2223 (2023). 165. Wu, T. et al. An engineered hypercompact CRISPR–Cas12f system
140. Shifrut, E. et al. Genome-wide CRISPR screens in primary human with boosted gene-editing activity. Nat. Chem. Biol. 19, 1384–1393
T cells reveal key regulators of immune function. Cell 175, (2023).
1958–1971 (2018). 166. Kim, D. Y. et al. Hypercompact adenine base editors based on
141. Carnevale, J. et al. RASA2 ablation in T cells boosts antigen transposase B guided by engineered RNA. Nat. Chem. Biol. 18,
sensitivity and long-term function. Nature 609, 174–182 (2022). 1005–1013 (2022).
142. Freitas, K. A. et al. Enhanced T cell effector activity by targeting 167. Kwon, J. B., Vankara, A., Ettyreddy, A. R., Bohning, J. D. &
the mediator kinase module. Science 378, eabn5647 (2022). Gersbach, C. A. Myogenic progenitor cell lineage specification by
143. Prinzing, B. et al. Deleting DNMT3A in CAR T cells prevents CRISPR/Cas9-based transcriptional activators. Stem Cell Rep. 14,
exhaustion and enhances antitumor activity. Sci. Transl. Med. 13, 755–769 (2020).
eabh0272 (2021). 168. Rebar, E. J. et al. Induction of angiogenesis in a mouse model using
144. Chen, J. et al. NR4A transcription factors limit CAR T cell function engineered transcription factors. Nat. Med. 8, 1427–1432 (2002).
in solid tumours. Nature 567, 530–534 (2019). 169. Sakowski, S. A. et al. Neuroprotection using gene therapy to
145. Chen, Z. et al. In vivo CD8+ T cell CRISPR screening reveals control induce vascular endothelial growth factor-A expression. Gene
by Fli1 in infection and cancer. Cell 184, 1262–1280 (2021). Ther. 16, 1292–1299 (2009).
146. Yang, Z. et al. Contextual reprogramming of CAR-T cells for 170. Dai, Q. et al. Engineered zinc finger-activating vascular
treatment of HER2+ cancers. J. Transl. Med. 19, 459 (2021). endothelial growth factor transcription factor plasmid DNA
147. Belk, J. A. et al. Genome-wide CRISPR screens of T cell induces therapeutic angiogenesis in rabbits with hindlimb
exhaustion identify chromatin remodeling factors that limit T cell ischemia. Circulation 110, 2467–2475 (2004).
persistence. Cancer Cell 40, 768–786 (2022). 171. Matharu, N. et al. CRISPR-mediated activation of a promoter or
148. Guo, A. et al. cBAF complex components and MYC cooperate enhancer rescues obesity caused by haploinsufficiency. Science
early in CD8+ T cell fate. Nature 607, 135–141 (2022). 363, eaau0629 (2019).
149. Fraietta, J. A. et al. Disruption of TET2 promotes the therapeutic 172. De Jonghe, P. Molecular genetics of Dravet syndrome. Dev. Med.
efficacy of CD19-targeted T cells. Nature 558, 307–312 (2018). Child Neurol. 53, 7–10 (2011).

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1215

Review article https://doi.org/10.1038/s41587-024-02320-1

173. Colasante, G. et al. dCas9-based Scn1a gene activation restores 195. Zeitler, B. et al. Allele-selective transcriptional repression of
inhibitory interneuron excitability and attenuates seizures in mutant HTT for the treatment of Huntington’s disease. Nat. Med.
Dravet syndrome mice. Mol. Ther. 28, 235–253 (2020). 25, 1131–1142 (2019).
174. Tanenhaus, A. et al. Cell-selective adeno-associated 196. Wegmann, S. et al. Persistent repression of tau in the brain using
virus-mediated SCN1A gene regulation therapy rescues mortality engineered zinc finger protein transcription factors. Sci. Adv. 7,
and seizure phenotypes in a Dravet syndrome mouse model and eabe1611 (2021).
is well tolerated in nonhuman primates. Hum. Gene Ther. 33, 197. O’Geen, H. et al. Transcriptional reprogramming restores UBE3A
579–597 (2022). brain-wide and rescues behavioral phenotypes in an Angelman
175. Liu, X. S. et al. Rescue of fragile X syndrome neurons by DNA syndrome mouse model. Mol. Ther. 31, 1088–1105 (2023).
methylation editing of the FMR1 gene. Cell 172, 979–992 (2018). 198. Zhou, H. et al. In vivo simultaneous transcriptional activation
176. Qian, J. et al. Multiplex epigenome editing of MECP2 to rescue of multiple genes in the brain using CRISPR–dCas9–activator
Rett syndrome neurons. Sci. Transl. Med. 15, eadd4666 (2023). transgenic mice. Nat. Neurosci. 21, 440–446 (2018).
177. Halmai, J. et al. Artificial escape from XCI by DNA methylation 199. Jia, Y. et al. In vivo CRISPR screening identifies BAZ2 chromatin
editing of the CDKL5 gene. Nucleic Acids Res. 48, 2372–2387 remodelers as druggable regulators of mammalian liver
(2020). regeneration. Cell Stem Cell 29, 372–385 (2022).
178. Thakore, P. I. et al. RNA-guided transcriptional silencing in vivo with 200.Guo, L. Y. et al. Multiplexed genome regulation in vivo with
S. aureus CRISPR–Cas9 repressors. Nat. Commun. 9, 1674 (2018). hyper-efficient Cas12a. Nat. Cell Biol. 24, 590–600 (2022).
179. Gemberling, M. P. et al. Transgenic mice for in vivo epigenome 201. Bock, C. et al. High-content CRISPR screening. Nat. Rev. Methods
editing with CRISPR-based systems. Nat. Methods 18, 965–974 Primers 2, 9 (2022).
(2021). 202. Braun, C. J., Adames, A. C., Saur, D. & Rad, R. Tutorial: design and
180. Moreno, A. M. et al. In situ gene therapy via AAV–CRISPR– execution of CRISPR in vivo screens. Nat. Protoc. 17, 1903–1925
Cas9-mediated targeted gene regulation. Mol. Ther. 26, 1818–1827 (2022).
(2018). 203. Roohani, Y., Huang, K. & Leskovec, J. Predicting transcriptional
181. Moreno, A. M. et al. Long-lasting analgesia via targeted in situ outcomes of novel multigene perturbations with GEARS.
repression of NaV1.7 in mice. Sci. Transl. Med. 13, eaay9056 (2021). Nat. Biotechnol. 42, 927–935 (2023).
182. Liao, H. K. et al. In vivo target gene activation via CRISPR/ 204. Park, M., Patel, N., Keung, A. J. & Khalil, A. S. Engineering
Cas9-mediated trans-epigenetic modulation. Cell 171, 1495–1507 epigenetic regulation using synthetic read–write modules. Cell
(2017). 176, 227–238 (2019).
183. Zhao, X. et al. Creation of a six-fingered artificial transcription 205. Gao, Y. et al. Complex transcriptional modulation with orthogonal
factor that represses the hepatitis B virus HBx gene integrated and inducible dCas9 regulators. Nat. Methods 13, 1043–1049 (2016).
into a human hepatocellular carcinoma cell line. J. Biomol. 206. Esvelt, K. M. et al. Orthogonal Cas9 proteins for RNA-guided gene
Screen. 18, 378–387 (2013). regulation and editing. Nat. Methods 10, 1116–1121 (2013).
184. Luo, W. et al. Engineered zinc-finger transcription factors inhibit 207. Wu, H. & Zhang, Y. Reversing DNA methylation: mechanisms,
the replication and transcription of HBV in vitro and in vivo. Int. J. genomics, and biological functions. Cell 156, 45–68 (2014).
Mol. Med. 41, 2169–2176 (2018). 208. Hamilton, J. R. et al. In vivo human T cell engineering with
185. Bloom, K. et al. Inhibition of replication of hepatitis B virus using enveloped delivery vehicles. Nat. Biotechnol. https://doi.org/
transcriptional repressors that target the viral DNA. BMC Infect. 10.1038/s41587-023-02085-z (2024).
Dis. 19, 802 (2019). 209. Banskota, S. et al. Engineered virus-like particles for efficient
186. Xirong, L. et al. Hepatitis B virus can be inhibited by DNA in vivo delivery of therapeutic proteins. Cell 185, 250–265 (2022).
methyltransferase 3a via specific zinc-finger-induced methylation 210. Cheng, Q. et al. Selective organ targeting (SORT) nanoparticles
of the X promoter. Biochemistry 79, 111–123 (2014). for tissue-specific mRNA delivery and CRISPR–Cas gene editing.
187. Bialek, J. K. et al. Targeted HIV-1 latency reversal using CRISPR/ Nat. Nanotechnol. 15, 313–320 (2020).
Cas9-derived transcriptional activator systems. PLoS ONE 11, 211. Yilmazer, A., de Lazaro, I., Bussy, C. & Kostarelos, K. In vivo
e0158294 (2016). cell reprogramming towards pluripotency by virus-free
188. Saayman, S. M. et al. Potent and targeted activation of latent overexpression of defined factors. PLoS ONE 8, e54754 (2013).
HIV-1 using the CRISPR/dCas9 activator complex. Mol. Ther. 24, 212. Beyersdorf, J. P. et al. Robust, durable gene activation in vivo via
488–498 (2016). mRNA-encoded activators. ACS Nano 16, 5660–5671 (2022).
189. Ji, H. et al. Specific reactivation of latent HIV-1 by dCas9–SunTag– 213. Gonsalves, R. et al. Severe early onset obesity and
VP64-mediated guide RNA targeting the HIV-1 promoter. Mol. hypopituitarism in a child with a novel SIM1 gene mutation.
Ther. 24, 508–521 (2016). Endocrinol. Diabetes Metab. Case Rep. 2020, 20-0042 (2020).
190. Wang, G. et al. Multiplexed activation of endogenous genes by 214. Ramachandrappa, S. et al. Rare variants in single-minded 1 (SIM1)
CRISPRa elicits potent antitumor immunity. Nat. Immunol. 20, are associated with severe obesity. J. Clin. Invest. 123, 3042–3050
1494–1505 (2019). (2013).
191. Peng, M. et al. Neoantigen vaccine: an emerging tumor 215. Polstein, L. R. et al. Genome-wide specificity of DNA binding,
immunotherapy. Mol. Cancer 18, 128 (2019). gene regulation, and chromatin remodeling by TALE- and CRISPR/
192. Chapdelaine, P. et al. Development of an AAV9 coding for a Cas9-based transcriptional activators. Genome Res. 25, 1158–1169
3XFLAG–TALEfrat#8–VP64 able to increase in vivo the human (2015).
frataxin in YG8R mice. Gene Ther. 23, 606–614 (2016). 216. Galonska, C. et al. Genome-wide tracking of dCas9–
193. Tremblay, J. P., Chapdelaine, P., Coulombe, Z. & Rousseau, methyltransferase footprints. Nat. Commun. 9, 597 (2018).
J. Transcription activator-like effector proteins induce the 217. Ichikawa, D. M. et al. A universal deep-learning model for zinc
expression of the frataxin gene. Hum. Gene Ther. 23, 883–890 finger design enables transcription factor reprogramming. Nat.
(2012). Biotechnol. 41, 1117–1129 (2023).
194. Erwin, G. S. et al. Synthetic transcription elongation factors 218. Ewen-Campen, B. et al. Optimized strategy for in vivo
license transcription across repressive chromatin. Science 358, Cas9-activation in Drosophila. Proc. Natl Acad. Sci. USA 114,
1617–1622 (2017). 9409–9414 (2017).

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1216

Review article https://doi.org/10.1038/s41587-024-02320-1

219. Yamagata, T. et al. CRISPR/dCas9-based Scn1a gene activation 232. Raffeiner, P. et al. An MXD1-derived repressor peptide identifies
in inhibitory neurons ameliorates epileptic and behavioral noncoding mediators of MYC-driven cell proliferation. Proc. Natl
phenotypes of Dravet syndrome model mice. Neurobiol. Dis. 141, Acad. Sci. USA 117, 6571–6579 (2020).
104954 (2020).
220. Vora, S. et al. Rational design of a compact CRISPR–Cas9 Acknowledgements
activator for AAV-mediated delivery. Preprint at bioRxiv We thank Gersbach laboratory members and collaborators for
https://doi.org/10.1101/298620 (2018). feedback and helpful discussions. This work was supported by
221. Kojima, S. et al. Epigenome editing reveals core DNA methylation National Institutes of Health grants U01AI146356, UM1HG012053,
for imprinting control in the Dlk1-Dio3 imprinted domain. Nucleic R01MH125236, R01CA289574 and RM1HG011123, National Science
Acids Res. 50, 5080–5094 (2022). Foundation grant EFMA-1830957, DARPA grant HR0011-19-2-0008,
222. Xu, X. et al. High-fidelity CRISPR/Cas9- based gene-specific the Foundation for Prader–Willi Research, an Allen Distinguished
hydroxymethylation rescues gene expression and attenuates Investigator Award from the Paul G. Allen Frontiers Group to C.A.G.,
renal fibrosis. Nat. Commun. 9, 3509 (2018). the Open Philanthropy Project and the Duke–Coulter Translational
223. Kim, J. M. et al. Cooperation between SMYD3 and PC4 drives a Partnership.
distinct transcriptional program in cancer cells. Nucleic Acids Res.
43, 8868–8883 (2015). Competing interests
224. Chiarella, A. M. et al. Dose-dependent activation of gene S.R.M., D.R., N.I. and C.A.G. are named inventors on patent applications
expression is achieved using CRISPR and small molecules that related to epigenome-engineering technologies. C.A.G. is a cofounder
recruit endogenous chromatin machinery. Nat. Biotechnol. 38, of Tune Therapeutics and Locus Biosciences and is an advisor to
50–55 (2020). Sarepta Therapeutics.
225. Cheng, A. W. et al. Casilio: a versatile CRISPR–Cas9–Pumilio
hybrid for gene regulation and genomic labeling. Cell Res. 26, Additional information
254–257 (2016). Correspondence and requests for materials should be addressed to
226. Shechner, D. M., Hacisuleyman, E., Younger, S. T. & Rinn, J. Charles A. Gersbach.
L. Multiplexable, locus-specific targeting of long RNAs with
CRISPR-Display. Nat. Methods 12, 664–670 (2015). Reprints and permissions information is available at
227. Carullo, N. V. N. et al. Enhancer RNAs predict enhancer–gene www.nature.com/reprints.
regulatory links and are critical for enhancer function in neuronal
systems. Nucleic Acids Res. 48, 9550–9570 (2020). Publisher’s note Springer Nature remains neutral with regard
228. Braun, S. M. G. et al. Rapid and reversible epigenome editing by to jurisdictional claims in published maps and institutional
endogenous chromatin regulators. Nat. Commun. 8, 560 (2017). affiliations.
229. Perillo, B., Tramontano, A., Pezone, A. & Migliaccio, A. LSD1: more
than demethylation of histone lysine residues. Exp. Mol. Med. 52, Springer Nature or its licensor (e.g. a society or other partner) holds
1936–1947 (2020). exclusive rights to this article under a publishing agreement with
230. Yeo, N. C. et al. An enhanced CRISPR repressor for targeted the author(s) or other rightsholder(s); author self-archiving of the
mammalian gene regulation. Nat. Methods 15, 611–616 (2018). accepted manuscript version of this article is solely governed by the
231. Kwon, D. Y., Zhao, Y. T., Lamonica, J. M. & Zhou, Z. Locus-specific terms of such publishing agreement and applicable law.
histone deacetylation using a synthetic CRISPR–Cas9-based
HDAC. Nat. Commun. 8, 15315 (2017). © Springer Nature America, Inc. 2024

Nature Biotechnology | Volume 42 | August 2024 | 1199–1217 1217