PERSPECTIVES

TECHNOLOGIES AND TECHNIQUES

Activating enhancer elements.
I N N O VAT I O N
CRISPR-mediated gene activation
(CRISPRa) uses nuclease-deactivated Cas9
Applying CRISPR–Cas9 tools (known as dCas9) fused to transcription
activating domains (BOX 2). This approach
to identify and characterize was first tested on promoters using the
transcription activation domains of the
transcriptional enhancers VP64 (a tetrameric repeat of herpes simplex
virus protein VP16) and p65 (nuclear
factor‑κB p65 subunit) proteins. Indeed,
Rui Lopes, Gozde Korkmaz and Reuven Agami dCas9–VP64 and dCas9–p65 could activate
endogenous genes when targeted by one
Abstract | The development of the CRISPR–Cas9 system triggered a revolution in sgRNA to their promoter region16,20,21.
the field of genome engineering. Initially, the use of this system was focused on the Yet, each protein fusion required the
study of protein-coding genes but, recently, a number of CRISPR–Cas9‑based tools combination of several sgRNAs per gene
have been developed to study non-coding transcriptional regulatory elements. to achieve high expression levels of the
target gene21,22. To overcome this limitation,
These technological advances offer unprecedented opportunities for elucidating several groups developed CRISPRa tools
the functions of enhancers in their endogenous context. Here, we discuss the containing multiple activation domains23,24.
application, current limitations and future development of CRISPR–Cas9 systems These improvements made the simultaneous
to identify and characterize enhancer elements in a high-throughput manner. activation of dozens of genes as well as
genome-wide activation screens possible.
The activating capacity of VP64 relies on
Enhancers were initially identified as opportunities for targeted genome editing the sequential recruitment of cofactors, such
distal regulatory elements that increase in human cells12–15. The RNA-guided as the histone acetyltransferase p300 and
transcription independently of their endonuclease, Cas9 can be targeted to TFIID25,26, suggesting that it can potentially
orientation, position and distance relative any genomic locus by sequence-specific be used to modulate enhancer activity.
to a target promoter 1. These elements single guide RNA (sgRNA) to initiate Indeed, a dCas9–VP64 fusion was shown
are essential for precise spatiotemporal double-stranded DNA cleavage (BOX 2). to activate endogenous genes by recruiting
regulation of gene expression, which is Importantly, Cas9 can be directed to p300 to acetylate H3K27 (histone H3 Lys27)
required for proper cell development, non-coding transcriptional regulatory at their cognate enhancer regions27. However,
differentiation and homeostasis2. A key elements, such as promoters16,17 and targeting dCas9–VP64 to the locus control
feature of enhancers is their ability to enhancers18,19. This feature led to the region (LCR) of the haemoglobin genes
function as transcription factor binding development of several Cas9‑based failed to activate some of them; dCas9–VP64
platforms, and genetic alterations methods and tools to study non-coding was able to induce H3K27 acetylation at the
at these regions are associated with regulatory elements in their native targeted enhancers but not at the cognate
pathological states3–8. chromatin environment. Here, we haemoglobin promoters, and this correlated
Enhancers are usually defined by summarize recent advances in CRISPR– with lack of activation of gene expression28.
criteria unrelated to their endogenous Cas9 systems that allow the modulation This suggests that the recruitment of
biological function, such as the makeup of of the activity of enhancers by altering transcription pre-initiation complex (PIC)
specific histone modifications, increased their transcriptional, epigenetic or genetic components by VP64 (REFS 25,26) might
chromatin accessibility and the occurrence features, and discuss the current limitations not be sufficient to elicit enhancer activity.
of bidirectional transcription in the loci of these technologies in the study of It is therefore not clear to what extent
in question9–11. Based on these features, enhancers. We also provide a perspective dCas9–VP64 can activate enhancer elements.
several techniques have been used to on future developments of CRISPR–Cas9
identify putative enhancers and validate tools and their application in identifying Repressing enhancer elements. The utility of
their activity in a high-throughput fashion and characterizing enhancers in an CRISPR–Cas9‑based tools in repressing gene
(BOX 1). Nonetheless, these methods unbiased fashion. expression (termed CRISPR interference
cannot fully determine whether a putative (CRISPRi)) was first demonstrated by
enhancer is required for transcription in Transcriptional modulation of enhancers showing that a dCas9–sgRNA complex
its native context or which gene (or genes) An attractive approach to studying could block RNA polymerase elongation29.
it regulates. enhancers in their endogenous context is This strategy was also successfully applied to
The emergence of the CRISPR–Cas9 the use of CRISPR–Cas9 tools to enforce interfere with transcription factor binding,
technology has opened unprecedented enhancer activation or repression directly. thereby disrupting enhancer activity 27.

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 17 | SEPTEMBER 2016 | 597

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVES

The use of dCas9 alone achieved only modest elements of composite enhancers and still Epigenetic modulation of enhancers
repression of gene expression in mammalian achieve highly specific repression of genes32,33. As noted above, enhancer activity is
cells. To improve CRISPRi, dCas9 was This effect seems to be mediated by decreased associated with dynamic epigenetic states,
fused with repressive effectors, such as the chromatin accessibility at both enhancer including acetylation and methylation of
Krüppel-associated box (KRAB) domain and target promoters32. It was also noted histone tails36,37. The ability to modulate these
of KOX1 (REFS 16,30). A dCas9–KRAB that directing dCas9–KRAB to enhancer epigenetic modifications is essential to gain
fusion was shown to efficiently repress the regions resulted in higher H3K9me3 (histone better understanding of their importance
expression of protein-coding and non-coding H3 Lys9 trimethylation) levels at the target for enhancer function. This prompted the
genes on a genome-wide scale17. KRAB promoter 32,34. Therefore, it is possible that development of several epigenome-editing
fusions were later shown to inactivate the KRAB fusions generate off-target effects tools that are based on fusions of epigenetic
expression of endogenous genes by targeting through heterochromatin spreading 35, and regulators (writers and erasers) to dCas9 and
their distal enhancer elements27,31. Indeed, that they silence promoter activity rather allow direct manipulation of epigenetic states
dCas9–KRAB can be directed to single than inactivating the target enhancer. of gene regulatory elements.
Lys-specific histone demethylase 1 (LSD1)
catalyses the removal of H3K4 methylation38.
Box 1 | Methods for genome-wide identification of putative gene regulatory elements Recently, a dCas9–LSD1 fusion was targeted
to a distal enhancer of the pluripotency
The comprehensive identification of transcriptional regulatory elements is a major challenge in gene Oct4 (also known as Pou5f1) and
genomic research. The emergence of next-generation sequencing propelled the development of a achieved specific gene repression and loss
number of high-throughput methods that were used to identify putative enhancers based on their
of pluripotency in embryonic stem (ES)
features and activity.
cells34. It was also shown that enhancer
Transcription factor binding sites are the core building blocks of regulatory elements. ChIP–seq
(chromatin immunoprecipitation followed by sequencing) is the most common technique to targeting by dCas9–LSD1 caused a marked
determine transcription-factor occupancy on a genome-wide scale72. This method is particularly decrease in H3K4me2 and H3K27Ac
suitable for identifying stimulus-induced changes or occupancy across different cell types, as well marks34,38, consistent with reduced enhancer
as for identifying binding sites of transcription cofactors (such as the histone acetyltransferase activity. However, dCas9–LSD1 was
p300), which are frequently associated with enhancers73. Genome-wide mapping of histone unable to repress genes when targeted
post-translational modifications by ChIP–seq revealed that H3K4me1 (histone H3 Lys4 to promoter regions34, a surprising
monomethylation) and H3K27ac (H3K27 acetylation) are enriched at active-enhancer regions74, result given the well-documented role of
whereas active-promoter regions are marked by H3K4me3 and H3K27ac. This ‘histone code’ is LSD1 in the repression of endogenous
widely used to annotate regulatory elements and corresponds well with heterologous reporter
promoters38. A possible explanation for
assays60, which have been commonly used to interrogate the activity of any genomic region.
these results is that LSD1 alone is ineffective
Cloning of a putative enhancer downstream or upstream of a reporter gene can reflect its
functionality by inducing the expression of the reporter gene. and may require additional cofactors
Active-enhancer regions are depleted of nucleosomes, leaving the DNA accessible to enzymatic to inactivate some regulatory elements,
cleavage. This has been exploited to identify regulatory regions across the genome, for example by indicating that current LSD1 fusions may
coupling digestion by DNase I with DNase-seq (DNase I hypersensitive sites sequencing)75. have low efficacy, compromising their
Distal enhancers are brought into spatial proximity with their target promoters through DNA application in high-throughput screens of
looping. Chromosome conformation capture (3C)76 and its variants77 are useful for predicting enhancer elements.
putative enhancers and their target genes. Chromatin interaction analysis with paired-end tag A fusion protein of dCas9 with the
(ChIA–PET) sequencing78 is a combination of 3C‑based methods with ChIP, which enables the catalytic core domain of p300 (dCas9–p300)
identification of proteins involved in the formation of specific chromosomal contacts.
was created recently 28, and this allowed the
Active enhancers support divergent transcription of their own loci79. The expression of these
manipulation of H3K27ac levels at both
enhancer-associated RNAs (eRNAs) has been used to identify active enhancers in a cell type-specific
manner, because eRNA expression correlates well with enhancer activity. eRNA expression can be proximal and distal regulatory elements.
detected by methods that measure nascent RNAs, such as global run‑on sequencing80 (GRO-seq). Interestingly, dCas9–p300 was capable of
The ability to increase transcription from minimal promoters in heterologous reporter vectors is activating genes with high specificity using
a hallmark of enhancer activity1, and has been used to identify enhancer elements both in cells one sgRNA per target gene28 and with higher
and in animals. Initially, these methods had low throughput. The recent development of massive transactivation capacity than dCas9‑VP64
parallel reporter assays81,82 (MPRA) and self-transcribing active regulatory region sequencing83 (REF. 28). This was particularly evident at
(STARR-seq) enables the evaluation of enhancer activity of thousands or even millions of DNA distal enhancer elements, where dCas9–
sequences simultaneously. VP64 displayed little capacity to activate
The combined application of the techniques mentioned above led to a tremendous increase
target genes28. These results establish dCas9–
in our understanding of transcriptional regulatory elements. DNase-seq in combination with
p300 as a robust tool to modulate histone
ChIP–seq of epigenetic modifications was applied to predict putative enhancer regions. Yet, the
epigenetic status of a region is not identical to enhancer activity, mostly owing to the use of acetylation and activate gene expression.
arbitrary cut-offs of histone modifications ratios (for example, H3K4me1/H3K4me3) as a measure However, it is unclear whether dCas9–p300
of activity84–86, which resulted in the estimation that the human genome harbours approximately is suitable for use in genome-wide functional
one million enhancers84–86. Instead, when methods that directly measure transcriptional activity screens. The higher activity of dCas9–p300
were used87, only 40,000 to 65,000 transcriptionally active putative enhancers were predicted seems to be related to the direct acetylation
(though any active but not transcribed enhancer would have been missed)9,88. of downstream target promoters, as
Altogether, these techniques cannot assess specific enhancer functionality because they provide opposed to the indirect recruitment of PIC
only circumstantial evidence that the identified elements are actively engaged in transcription by VP64 (REFS 25,26). Still, it remains to be
regulation. Thus, other methods are required for the identification of enhancers that drive gene
determined whether acetylation of H3K27
expression in certain biological settings.
alone is sufficient for gene activation.

598 | SEPTEMBER 2016 | VOLUME 17 www.nature.com/nrm

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 PERSPECTIVES

The consequences of manipulating Box 2 | Genome editing and regulation using CRISPR–Cas9 systems
epigenetic states on transcription regulation
are not completely understood, and the The CRISPR–Cas9 system is based on the Cas9 nuclease sgRNA
utilization of epigenome-editing tools might molecular machinery of type II CRISPR bacterial
immune system and has been repurposed for
be associated with potential off-target effects.
targeted gene editing, providing major advances RuvC
Nevertheless, the expansion of the current in efficiency, speed and scale of genome
dCas9 epigenetic toolbox will be much engineering 12,52,55,57
. The endonuclease Cas9 has PAM
needed to shed light on the connections two nuclease domains, RuvC and HNH, and is NGG
between the epigenome and gene expression. directed to specific DNA sequences by pairing of
Moreover, the precise manipulation of a chimeric single guide RNA (sgRNA) with target
epigenetic modifications holds tremendous DNA (see the figure)13,29. This requires the
presence of a 5ʹ protospacer adjacent motif HNH
therapeutic potential.
(PAM) in the DNA, which in
Genetic manipulation of enhancers Streptococcus pyogenes is usually NGG.
Genome engineering with Cas9 exploits DNA
The genomic features that define a
double-strand break (DSB) repair pathways to
regulatory element are poorly understood. create mutations at specific genomic locations. In mammalian cells, DSBs are usually repaired either
For example, some enhancers consist of by homology-directed repair (HDR), for which an exogenous Nature Reviews
template | Molecular
is provided for theCell Biology
conversion
a single unit, whereas others — known as of the endogenous sequence following the formation of DSBs by Cas9, or by non-homologous end
super-enhancers — are composed of multiple joining (NHEJ), which is an error-prone mechanism that generates random insertion and deletion
clusters of enhancers39. Genetic perturbation mutations at the targeted location. The continuous expression of the Cas9–sgRNA complex leads
is a powerful approach to draw causal links to near-complete allelic modification in a short time frame58. However, the use of Cas9 nuclease
between genetic information and cellular is potentially associated with the formation of off-target DSBs, which may cause undesired
functions. The advent of the CRISPR–Cas9 effects (BOX 3).
system facilitated genome editing and, The enzymatic activity of Cas9 can be abolished by mutations in the RuvC and HNH domains,
generating nuclease-deactivated Cas9 (dCas9), which can still be specifically targeted to DNA89.
recently, it has been applied to dissect the
dCas9 can modulate transcription when fused to an effector domain (for example, a transcription
DNA sequences required for appropriate activator or repressor) and expressed in concert with a sgRNA targeted to a regulatory element of
activity of enhancer elements in their interest16,29. An advantage of using dCas9 effectors is that their effect is reversible; moreover, dCas9
native context. fusions can be multiplexed and set to perform simultaneous transcription activation and repression
The Cas9 nuclease can be directed by of different loci71. Potential disadvantages of dCas9 systems include disruption of native chromatin
one sgRNA to induce a double-strand architecture, steric hindrance with transcription factors24, and blocking of RNA polymerases29.
break (DSB) at a specific genomic region. In addition, the efficiency of activation or repression can fluctuate dramatically32,34. This can be
These DSBs are generally repaired by the explained by features of the target region, such as chromatin structure, and by the availability of
error-prone repair pathway non-homologous additional cofactors32. Figure adapted from REF. 90, Nature Publishing Group.
end joining (NHEJ), resulting in the
formation of insertion and deletion (indel) and decreased expression of the TAL1 (T-cell (GATA binding protein 2) expression40. The
mutations (usually smaller than 10 bp). acute lymphocytic leukemia protein 1) authors showed that this regulatory element
The initial application of CRISPR–Cas9 proto-oncogene. is frequently translocated in leukaemia cells,
for genome editing was centred on We note several limitations when using leading to haploinsufficiency of GATA2 and
protein-coding genes, but it was soon a single Cas9–sgRNA complex to target activation of EVI1 (ecotropic viral integration
adopted to target non-coding regulatory enhancer elements. First, one sgRNA might site 1) oncogene. Cas9 nuclease can also be
elements. Recently, it was found that not fully disrupt enhancer activity owing to used to generate monoallelic deletions of
CRISPR–Cas9 targeting by one sgRNA can the small size of indels generated. The use of enhancer elements. When targeting essential
produce comparable genetic effects to the paired sgRNAs to delete genomic elements regulatory elements, a monoallelic deletion
deletion of a whole enhancer unit18 and can address this issue (see bellow). Second, is performed to circumvent the cellular
cause phenotypes that are robust enough when working with a population of cells, lethality associated with biallelic deletions.
for screening 19 (see below). The strong the effect of a sgRNA might be diluted, The combination of this targeted approach
effects produced by one sgRNA go against as targeting by Cas9 results in different with gene expression profiling identified a
the well-established idea that regulatory (heterozygous and homozygous) lesions in distal enhancer cluster that is required for
elements are robust and redundant. Indeed, different cells. The isolation of individual Sox2 (SRY (sex determining region Y)-box 2)
functionally important sequences within the mutant clones is a laborious process, but it transcription41,42 and ES cell differentiation41.
enhancer are known to be highly sensitive can potentially solve this problem. Finally, Interestingly, deletion of this cluster did not
as well: even single-nucleotide alterations some sgRNAs may suffer from intrinsic low affect other nearby genes, indicating that the
in these regulatory sequences can have DNA-editing efficiency (BOX 3). regulation of gene expression by enhancers
substantial effects on gene expression and A regulatory element can be fully is highly specific41,42. The high precision of
cause pathological conditions3,4,6–8. In line inactivated by Cas9 nuclease through the Cas9‑mediated cleavage allows the deletion
with this, it was recently proposed that small generation of a specific deletion. This of individual constituents of composite
mutations can activate oncogenes in cancer can be easily achieved by targeting two enhancers18,33,43. Recent evidence suggests
cells by creating a new super-enhancer 5. sgRNAs to the borders of the target region. that deletion of a single element can have
In this study, a single Cas9–sgRNA complex An example of this strategy comes from a comparable effects to the deletion of the
was used to delete the mutant sequence, report in which CRISPR–Cas9 was used to entire composite enhancer 18. More studies
resulting in the collapse of enhancer activity excise an enhancer that regulates GATA2 are needed in order to understand whether

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 17 | SEPTEMBER 2016 | 599

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVES

Box 3 | Targeting specificity of CRISPR–Cas9 systems specific locus by a short (20‑nucleotide)

sgRNA enables the rapid generation
Targeting accuracy of CRISPR–Cas9 systems is defined as the ratio between generation of DNA of sgRNA libraries for screening. The
double-strand breaks (DSBs) at the intended locations (on‑target effects) and at unintended sgRNAs can be delivered by a lentiviral
locations (off-target effects) and is assessed for each single guide RNA (sgRNA) in terms of their vector and stably integrated in the target
consistency in targeting the same locus54,56–58. Nuclease-deactivated Cas9 (dCas9)-based screening
cells, facilitating the screen readout using
(CRISPR-mediated gene activation (CRISPRa) and CRISPR interference (CRISPRi)) can modulate
gene expression, and recent dCas9‑based tiling screens improved their on‑target effects and next-generation sequencing (FIG. 1). Recently,
sgRNA consistency (that is, different sgRNAs efficiently targeting same region)17,23. The use of several research groups have taken advantage
dCas9 for functional enhancer screens has not yet been evaluated, but the enzymatically active of these features to perform high-throughput
CRISPR–Cas9 system was recently demonstrated to be a powerful enhancer screening tool18,19,46. functional screens of endogenous enhancers.
Nevertheless, sgRNAs display large variations in efficiency (a result of chromatin accessibility
and the composition of the targeted sequence54,91,92), which may influence screening results. Dissecting enhancers using CRISPR–Cas9
The continuous expression of Cas9 leads to near-complete allelic modification owing to the tiling screens. Single-nucleotide changes
irreversible modification of target DNA54,56,57. However, not every mutation will abolish gene can substantially affect gene regulation44,
function as DSB repair by non-homologous end joining generates different mutations in different and this finding led to the hypothesis
cells. Moreover, the size of insertion and deletion (indel) mutations varies substantially between
that Cas9‑mediated mutagenesis could
the targeted loci54,56,57,93. Generating a DSB in a constitutively spliced exon usually results in the
formation of either a premature stop codon or frameshift mutations44,92. By contrast, the critical uncover elements required for the function
regulatory elements of enhancers are poorly defined, and therefore it is challenging to target them of enhancers. The basic strategy is to tile
with one sgRNA. Cas9‑induced mutations in regulatory elements can result in more variable hundreds of sgRNAs across a putative
effects on gene expression compared to mutations in protein-coding regions45,46. This is probably enhancer to disrupt nearly its entire sequence
caused by a wide spectrum of mutations with different likelihood to disrupt critical DNA elements, (FIG. 2). The screen is then used to infer
such as transcription factor binding sites. which enhancer sequences are potentially
The specificity of CRISPR–Cas9 reagents is a major concern for their applicability in both important for its function. However, the
research and clinical settings. Cas9 nucleases can tolerate up to five mismatches between the full spectrum of the driver mutations and
sgRNA and a genomic sequence94,95. Reassuringly, the majority of these off-target binding events potential regulatory DNA elements can
do not lead to the generation of indels91,96, suggesting that imperfect matching of the sgRNA to the
only be examined at the validation stage
target sequence is insufficient for DNA cleavage97. dCas9‑based tools were shown to be sensitive
to even fewer mismatches17, and global gene expression analysis revealed highly specific activity of using individual sgRNA transductions, as
both transcriptional and epigenetic modulators16,17,28. Nonetheless, CRISPR reagents with sgRNA enrichment scores might differ from
improved specificity were developed, such as paired Cas9 nucleases that nick DNA at two adjacent expectations19. Importantly, the resolution
sequences53,98, truncated sgRNAs99, and modified Cas9 proteins with high precision100,101. The D10A of such tiling screens approaches saturation
mutant of Cas9 nicks one strand of DNA, and directing paired Cas9(D10A) nucleases to a specific mutagenesis in situ, with average cleavage
location results in fewer off-target effects compared with Cas9 (REFS 53,98). Truncated sgRNAs, frequency ranging from 4 to 10 bp18,19. The
which are designed according to the finding that PAM-distal regions of sgRNA usually have less application of this method enables the study
effect on on‑target cleavage efficiency, have reduced off-target effects (up to 5,000 times) while of the regulation of single genes by multiple
conserving on‑target cleavage efficiency99. enhancer elements18,19,45,46.
Increasing CRISPR–Cas9 fidelity may also reduce false-negative rates in a screen (true hits that
Tiling screens were used successfully
were missed in the screen). To our knowledge, false-negative rates have not been reported so far
for CRISPR–Cas9 enhancer screens. It is possible that the effects of sgRNAs with low activity are to dissect the human and mouse BCL11A
missed, resulting in higher false-negative rates; if this is the case improving target specificity and (B‑cell lymphoma/leukaemia 11A)
sgRNA consistency should improve genetic screening approaches by reducing false-positive composite enhancer 18. Interestingly, the
as well as false-negative rates. critical components of the composite
enhancer were found to be different
between human and mouse18, despite
this is an exception or a rule. The use of the identification of enhancers and their sharing sequence homology and chromatin
paired sgRNAs enables control of the size target genes, which is a major advantage signatures. As BCL11A regulates the
of the deletion, in contrast with the random compared to previous technologies (BOX 1). switching from fetal to adult haemoglobin
indels induced by using a single sgRNA. This combination is suitable for matching gene expression by repressing the expression
However, this method is far less efficient an individual enhancer with its target gene, levels of fetal heamoglobin, the results
in inducing deletions44: the efficiency and cannot yet be used on a genome-wide from this study suggest that genetic
of targeted deletions by paired sgRNAs scale. It is also worth noting that the function perturbation of the BCL11A enhancer might
is approximately 25%, but it decreases of some cis regulatory elements might be be a therapeutic option for β‑haemoglobin
substantially with increasing deletion redundant (see below). This problem can disorders. Similarly, a CRISPR–Cas9
sizes44. In addition, the introduction of two be tackled by targeting Cas9 nuclease to tiling screen to identify critical sequences
sgRNAs theoretically carries twice the risk multiple redundant enhancers and abolishing that contribute to the function of a distal
of off-target effects (BOX 3) compared with their activity in a combinatorial manner. enhancer of the p53 target gene CDKN1A19
one sgRNA. revealed that additional domains, besides the
These studies emphasize the importance High-throughput screening of enhancers p53 binding site, are required for the activity
of perturbing the endogenous state of The power of the CRISPR–Cas9 system of the enhancer and activation of CDKN1A
enhancers to obtain greater insight into relies on two main factors: the ease of expression upon oncogene-induced
their biological function. The combination RNA-mediated gene targeting and its senescence. High-throughput CRISPR–
of CRISPR–Cas9‑induced mutagenesis and efficient induction of mutations12,14. Cas9 mutagenesis was also employed to
genome-wide expression profiling allows The capacity of targeting Cas9 to a systematically interrogate putative regulatory

600 | SEPTEMBER 2016 | VOLUME 17 www.nature.com/nrm

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 PERSPECTIVES

elements of POU5F1 in human ES cells46. Another high-throughput mapping known epigenetic markers associated with
In this case, a POU5F1–GFP reporter of regulatory sequences is multiplexed regulatory elements. These results expose the
was used to determine the function of the editing regulatory assay (MERA)45. It is shortcomings of correlative annotations to
candidate enhancer regions. The authors based on the integration of a pooled sgRNA predict regulatory sequences and highlight
identified classical regulatory elements, but library into a specific locus by homologous the importance of direct perturbation to
also a class of non-canonical enhancers, recombination and relies on the expression definitively characterize genetic elements.
termed temporarily phenotypic (Temp) of a GFP-tagged gene locus to determine Considering these successful
enhancers. Although they harbour the effect of the mutations45. MERA led experiments, high-throughput
common enhancer features, loss of these to the identification of expected promoters, CRISPR–Cas9 tiling screens should be
elements caused temporary and reversible enhancers and transcription factor binding readily implemented to interrogate at
transcriptional impairment owing to sites, but also what might be a new class near-nucleotide resolution the function of
disruption of the local chromatin structure. of elements, designated as unmarked numerous non-coding regulatory elements.
Further work is required in order to regulatory elements (UREs). UREs were The Cas9 protein scans for a short DNA
understand the biological function of Temp shown to regulate endogenous gene sequence known as protospacer adjacent
enhancers and how they are regulated. expression although they do not contain any motif (PAM; in Streptococcus pyogenes it

a Library design b Library construction CRISPR enhLibrary

sgRNA
Enhancer

eRNA

sgRNA Pool of annealed

oligonucleotides oligonucleotides Cloning into viral vector Virus production

c Screening and data analysis

Screen readout
Step 1 Step 2 Step 3
Positive Positive selection

Enrichment score
Active sgRNA selection (enrichment)
enhancers
(enrichment) Next-generation
sequencing
sgRNA
endcoding
sequence
Inactive
enhancers
Genomic Viral
Negative selection

Enrichment score
DNA vector
(dropout)

Negative
selection
(dropout)

Figure 1 | Functional genetic screens of active enhancers using the a star) results in a final list of enhancers to be targeted by CRISPR–Cas9.
CRISPR–Cas system. a | Bioinformatic design of a single guide RNA The third step is identification of suitable Cas9 cleavage sites within the
(sgRNA) library for global targeting of putative enhancer elements consists transcription factor binding sites, according
Nature to the
Reviews availability
| Molecular Cellof proto-
Biology
of three steps19. The first step is identification of the enhancers active in the spacer adjacent motif (PAM) sequences. It is essential to target precisely
specific experimental condition. The chromatin at active enhancers is typ- the transcription factor binding site within the enhancer region; yet, the
ically marked with specific epigenetic modifications (high level of availability of such targetable sites varies according the availability of PAMs
H3K4me1 (histone H3 Lys4 monomethylation) and H3K27ac (histone H3 in the transcription factor consensus sequence. For instance, the consen-
Lys27 acetylation), and low level of H3K4me3 (histone H3 Lys4 trimethyl­ sus motif of p53 has high NGG PAM content compared to ERα (resulting in
ation)) and occupied by RNA polymerase II. Thus, publicly available ChIP– ~90% versus ~60% target availability, respectively), and some regions can
seq (chromatin-immunoprecipitation followed by sequencing) data sets of be targeted with more than one sgRNA. b | The sgRNA oligonucleotides
chromatin modifications and of RNA polymerase II occupancy can be used targeting the suitable Cas9 cleavage sites are synthesized, annealed with
to identify putative enhancer regions. Furthermore, enhancer RNA (eRNA) 3ʹ and 5ʹ cloning primers, pooled and cloned into viral constructs to pro-
expression correlates well with enhancer activity and allows active regu- duce an enhancer-targeting sgRNA expression library (CRISPR enhLibrary),
latory elements to be distinguished; eRNA expression data sets are also from which viruses are produced to confer stable expression of sgRNAs in
publicly available. The second step is identification of active enhancer cells. c | Virus transduction of cells should ideally be performed such that
regions harbouring transcription factor binding sites (transcription factor each cell expresses only one sgRNA, but that all the sgRNAs are expressed
binding sites (for example, p53 and oestrogen receptor-α (ERα)) are shown in the transduced cell population, to maintain the complexity of the library.
by yellow star and triangle symbols). Enhancer regulation depends on bind- The transduced cells are subjected to a proliferation-based screening
ing of specific transcription factor combinations, and each enhancer is selection to identify sgRNAs that confer cell growth advantage or
responsive to a different set of transcription factors. Therefore, the inter- disadvan­tage according to the designed assay, and next-generation
section of active enhancer regions (as defined in step 1) with transcription sequencing is used to assess which sgRNAs were enriched or depleted
factor activity relevant to the specific experimental condition (marked with (shown in red) in the selected cell population.

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 17 | SEPTEMBER 2016 | 601

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVES

Transcription factors Functionally important DNA elements and resistance to drugs and toxins41,56,57.
(for example, for binding of
accessory proteins, DNA looping and Likewise, high-throughput CRISPR–Cas9
chromatin structure) strategies are suitable for investigating the
role of multiple enhancers in the regulation
H3K4me1 H3K27Ac of multiple genes and their contribution to
a specific phenotype for large-scale genetic
?
screening of regulatory elements19.
Indeed, CRISPR–Cas9 was used in both
Enhancer region positive and negative selection screens
to identify functional enhancers19 by
sgRNA inactivating putative enhancers through
the disruption of transcription factor
binding sites and screening for phenotypic
changes (FIG. 1). As a result, several novel
Enrichment and/or dropout screening enhancers were identified, which have key
roles in p53‑dependent oncogene-­induced
Assay type senescence and oestrogen receptor 1
(for example, for proliferation,
differentiation or metabolism) (ESR1)-mediated cell proliferation19.
This screening method was found to
be efficient and robust, generating high
levels of mutagenesis, sgRNA consistency
Functionally important Transcription factor (that is, different sgRNAs with similar
enhancer element binding sites efficiency in targeting the same region)
and strong phenotypic effects. In particular,
Figure 2 | High-resolution dissection of active enhancerNature
sequences (marked
Reviews by H3K4me1
| Molecular and
Cell Biology
H3K27ac) by CRISPR–Cas9 tiling screens to uncover novel regulatory elements. To gain detailed the performance of Cas9 in the negative
understanding of functionally important DNA elements (represented by the question mark) within selection setting, as determined by high
enhancers, such as those important for binding of transcription factors or mediating DNA looping, all validation rates, was impressive19, as this
protospacer adjacent motif sites within a selected enhancer are identified by bioinformatics, and type of screen requires higher sensitivity
single guide RNAs (sgRNAs) densely covering (tiling) the region are produced and used for screening to detect changes in the representation of
by an assay of interest. Following screening, the abundance of sgRNAs in the final cell population can individual sgRNAs58.
be used to infer key elements that are required for the function of the enhancer, such as novel Studies discussed above18,19,45,46 have
transcription binding sites. H3K4me1, histone H3 Lys4 monomethylation; H3K27ac, histone H3 produced some common results, despite
Lys27 acetylation. being performed in completely distinct
biological settings. First, both enhancers and
is usually NGG) at the 3ʹ end of the target sgRNAs yielding the same phenotype super-enhancers can be susceptible to small
sequence to achieve a successful binding to are considered. Also, rational design of mutations generated by single DSBs, which
the target region, and generates DSBs 3 bp sgRNAs can minimize false positives owing enabled these high-throughput functional
upstream of the PAM. Therefore, the use to off-target effects52,53. It was shown that screens18,19,45,46. However, we cannot rule
of Cas9 orthologues and variants, which the PAM-proximal 10–12 bp of sgRNA out that some enhancers are robust and not
have different PAM requirements, could sequence is the main determinant of affected by small indels. The development
increase the resolution of CRISPR–Cas9 specificity; thus, restraining the similarity of methods to generate pooled libraries of
in situ mutagenesis47–50. Alternatively, between the target and off-target sites and paired sgRNAs can potentially address this
activating homology-directed repair (HDR) allowing no more than three mismatches limitation59. Second, the vast majority of
concomitantly with CRISPR–Cas9 could can reduce the off-target effects. However, sgRNAs were not enriched or depleted,
allow DNA editing at nucleotide resolution. when targeting regulatory elements with whereas the ones that were, colocalized to
The coupling of CRISPR–Cas9 mutagenesis CRISPR–Cas9, the availability of PAM discrete genomic regions18,19. This suggests
and multiplex HDR has already enabled sequences is a limiting factor in the design that enhancer elements are composed of
saturation editing of protein-coding genes51, of multiple unique sgRNAs targeting the many redundant and only a few critical
suggesting that it can also be applied to same region. In this case, the performance sequences. Finally, the pattern of mutations
dissect the function of regulatory elements of single sgRNAs should be evaluated in generated by a sgRNA can be used for motif
with high resolution. However, we note that several independent experiments to identify discovery without a priori knowledge18,19,45.
HDR in conjugation with CRISPR–Cas9 has a consistent effect. Finally, the inclusion This can be done by mutagenizing enhancers
low DNA-editing efficiency, that it is time of many control sgRNAs will facilitate the in a population of cells using multiple,
consuming and might not be possible for distinction between true- and false-positive unique sgRNAs, applying selective conditions
every cell type. screen results. and comparing the initial and final
In high-throughput screens of abundance of mutations along the enhancer
protein-coding genes, false identification Genetic screens of enhancers. In recent regions. In principle, the relative abundance
of screen hits (false positives) can be years, CRISPR–Cas9‑based genetic of mutations at specific sequences should
avoided by designing a library containing screens have been applied successfully to reflect their necessity for the function of an
multiple unique sgRNAs targeting the identify protein-coding genes involved enhancer element. Altogether, these studies
same gene. Only screen hits with distinct in fundamental cellular processes54,55 establish CRISPR–Cas9 as a powerful tool

602 | SEPTEMBER 2016 | VOLUME 17 www.nature.com/nrm

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 PERSPECTIVES

to dissect the functions of the non-coding genetic alterations is unknown as it is not Rui Lopes, Gozde Korkmaz and Reuven Agami are at
the Division of Biological Stress Response,
genome in an unbiased fashion18,19,45,46. clear whether they affect gene expression.
The Netherlands Cancer Institute, Plesmanlaan 121,
A caveat of the sgRNA libraries used The combination of CRISPR–Cas9‑mediated 1066 CX Amsterdam, The Netherlands.
to date is that they target a limited set of mutagenesis and multiplex HDR promises Reuven Agami is also at the Department of Genetics,
enhancers and therefore cannot be used to facilitate the interpretation of sequence Erasmus University Medical Center, Wytemaweg 80,
for unbiased screening in any biological variants of uncertain significance reported 3015 CN Rotterdam, The Netherlands.
setting. We envision that, in the near by clinical sequencing 51. R.L. and G.K. contributed equally to this work.
future, it will be possible to interrogate the Correspondence to G.K. and R.A.
function of all genomic regulatory elements Manipulation of high-order chromatin [email protected]; [email protected]
using CRISPR–Cas9 tools. This will be a organization. Transcriptional regulation doi:10.1038/nrm.2016.79
monumental task, given that the estimated can also be achieved by customized editing Published online 6 July 2016;
corrected online 22 July 2016
number of enhancers in the human of the genome architecture. It was recently
genome is estimated to be approximately demonstrated that disruption of topological 1. Banerji, J., Rusconi, S. & Schaffner, W. Expression
of a β-globin gene is enhanced by remote SV40 DNA
one million (REFS 10,60). The identification boundaries by Cas9 nuclease resulted in sequences. Cell 27, 299–308 (1981).
of functional enhancers by CRISPR–Cas9 loss of facultative heterochromatin and 2. Bulger, M. & Groudine, M. Enhancers: the abundance
and function of regulatory sequences beyond
on a genome-wide scale will require further activation of gene expression67. Mutations promoters. Dev. Biol. 339, 250–257 (2010).
technological advances to improve the affecting topological boundaries were found 3. Bauer, D. E. et al. An erythroid enhancer of BCL11A
subject to genetic variation determines fetal
coverage of libraries and speed of screening. in different types of cancer, and this was hemoglobin level. Science 342, 253–257 (2013).
recently proposed to be a mechanism of 4. Harismendy, O. et al. 9p21 DNA variants associated
with coronary artery disease impair interferon-γ
Future perspective activation of proto-oncogenes by enhancer signalling response. Nature 470, 264–268 (2011).
Since it was first used in genome engineering elements68. Forced chromatin looping using 5. Mansour, M. R. et al. An oncogenic super-enhancer
formed through somatic mutation of a noncoding
in 2013 (REFS 12–15), the CRISPR–Cas9 a synthetic zinc finger–LDB1 fusion protein intergenic element. Science 346, 1373–1377
system and its applications have continued resulted in the activation of developmentally (2014).
6. Maurano, M. T. et al. Systematic localization of
to improve at an unprecedented rate. Here, silenced genes in mouse cells69,70. This common disease-associated variation in regulatory
we pinpoint future directions that can be suggests that programmable DNA looping DNA. Science 337, 1190–1195 (2012).
7. Musunuru, K. et al. From noncoding variant to
applied to study transcriptional regulatory could be a novel therapeutic approach phenotype via SORT1 at the 1p13 cholesterol locus.
regions with this system. for human diseases. Although there are Nature 466, 714–719 (2010).
8. Oldridge, D. A. et al. Genetic predisposition to
no CRISPR–Cas-based tools available to neuroblastoma mediated by a LMO1 super-enhancer
Light-inducible systems. The development induce changes in chromatin conformation, polymorphism. Nature 528, 418–421 (2015).
9. Andersson, R. et al. An atlas of active enhancers
of light-inducible DNA binding proteins dCas9 has the potential to bring together across human cell types and tissues. Nature 507,
represents a breakthrough for genome two or more genomic loci through the use 455–461 (2014).
10. Thurman, R. E. et al. The accessible chromatin
engineering. Initially, this technology was of sgRNAs that scaffold multiple domains, landscape of the human genome. Nature 489, 75–82
applied to transcription activator-like which are capable of recruiting Cas9 and (2012).
11. Visel, A., Rubin, E. M. & Pennacchio, L. A. Genomic
effector proteins (TALEs)61 but it has other sgRNAs71. The spread of this strategy views of distant-acting enhancers. Nature 461,
recently been introduced to CRISPR–Cas9 to CRISPR–Cas9 systems could help to 199–205 (2009).
12. Cong, L. et al. Multiplex genome engineering using
systems. These tools are incredibly versatile characterize the role of three-dimensional CRISPR/Cas systems. Science 339, 819–823 (2013).
as they can be activated within minutes and genome conformations in the function of 13. Jinek, M. et al. RNA-programmed genome editing in
human cells. eLife 2, e00471 (2013).
do not require additional chemical cofactors. regulatory elements and gene expression. 14. Mali, P. et al. RNA-guided human genome engineering
Light-inducible CRISPR–Cas9 systems have via Cas9. Science 339, 823–826 (2013).
15. Cho, S. W., Kim, S., Kim, J. M. & Kim, J. S. Targeted
been used to control genome editing 62 and Concluding remarks. The future application genome engineering in human cells with the Cas9
modulate gene expression63,64 in a reversible of CRISPR–Cas9 systems in genetic screens RNA‑guided endonuclease. Nat. Biotechnol. 31,
230–232 (2013).
manner. In principle, this technology can of enhancers will certainly accelerate the 16. Gilbert, L. A. et al. CRISPR-mediated modular
be adapted to modulate the activity of identification of regulatory networks by RNA‑guided regulation of transcription in eukaryotes.
Cell 154, 442–451 (2013).
endogenous enhancers in space and time connecting non-coding elements to their 17. Gilbert, L. A. et al. Genome-scale CRISPR-mediated
with very good resolution. target genes and to a specific phenotype. control of gene repression and activation. Cell 159,
647–661 (2014).
We believe that the most comprehensive 18. Canver, M. C. et al. BCL11A enhancer dissection by
Modelling human diseases. The CRISPR– definition of the function of regulatory Cas9‑mediated in situ saturating mutagenesis. Nature
527, 192–197 (2015).
Cas9 system simplified and accelerated elements will arise from combining different 19. Korkmaz, G. et al. Functional genetic screens for
the modelling of complex human diseases. strategies, including disruption of the DNA enhancer elements in the human genome using
CRISPR‑Cas9. Nat. Biotechnol. 34, 192–198 (2016).
For example, paired sgRNAs were used to sequence and modulation of its activity. 20. Maeder, M. L. et al. Targeted DNA demethylation and
engineer a chromosomal inversion and However, the majority of CRISPR–Cas9 activation of endogenous genes using programmable
TALE–TET1 fusion proteins. Nat. Biotechnol. 31,
generate mouse models of human non-small tools have only been applied to study a 1137–1142 (2013).
cell lung cancers65. This approach can also limited number of endogenous enhancers 21. Perez-Pinera, P. et al. RNA-guided gene activation
by CRISPR–Cas9‑based transcription factors.
be used to investigate human diseases to date. It is paramount to evaluate these Nat. Methods 10, 973–976 (2013).
associated with repositioning of enhancers tools across many regulatory elements 22. Maeder, M. L. et al. CRISPR RNA-guided activation
of endogenous human genes. Nat. Methods 10,
by chromosomal rearrangements40. In recent and in different biological settings to fully 977–979 (2013).
years, it has become clear that the vast appreciate their potential. Together, we are 23. Konermann, S. et al. Genome-scale transcriptional
activation by an engineered CRISPR‑Cas9 complex.
majority of disease-associated mutations privileged to live in the ‘CRISPR–Cas9 era’, Nature 517, 583–588 (2015).
and single-nucleotide polymorphisms a time in which our imagination seems to be 24. Tanenbaum, M. E., Gilbert, L. A., Qi, L. S.,
Weissman, J. S. & Vale, R. D. A protein-tagging system
are present in non-coding regions of the the only limiting factor in discovering the for signal amplification in gene expression and
genome9,66. The clinical significance of these secrets of the human genome. fluorescence imaging. Cell 159, 635–646 (2014).

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 17 | SEPTEMBER 2016 | 603

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVES

25. Choy, B. & Green, M. R. Eukaryotic activators function 51. Findlay, G. M., Boyle, E. A., Hause, R. J., Klein, J. C. 79. Kim, T. K. et al. Widespread transcription at neuronal
during multiple steps of preinitiation complex & Shendure, J. Saturation editing of genomic regions activity-regulated enhancers. Nature 465, 182–187
assembly. Nature 366, 531–536 (1993). by multiplex homology-directed repair. Nature 513, (2010).
26. Memedula, S. & Belmont, A. S. Sequential recruitment 120–123 (2014). 80. Core, L. J., Waterfall, J. J. & Lis, J. T. Nascent RNA
of HAT and SWI/SNF components to condensed 52. Hsu, P. D. et al. DNA targeting specificity of sequencing reveals widespread pausing and divergent
chromatin by VP16. Curr. Biol. 13, 241–246 (2003). RNA‑guided Cas9 nucleases. Nat. Biotechnol. 31, initiation at human promoters. Science 322,
27. Gao, X. et al. Comparison of TALE designer 827–832 (2013). 1845–1848 (2008).
transcription factors and the CRISPR/dCas9 in 53. Mali, P. et al. CAS9 transcriptional activators for target 81. Melnikov, A. et al. Systematic dissection and
regulation of gene expression by targeting enhancers. specificity screening and paired nickases for cooperative optimization of inducible enhancers in human cells
Nucleic Acids Res. 42, e155 (2014). genome engineering. Nat. Biotechnol. 31, 833–838 using a massively parallel reporter assay.
28. Hilton, I. B. et al. Epigenome editing by a (2013). Nat. Biotechnol. 30, 271–277 (2012).
CRISPR‑Cas9‑based acetyltransferase activates genes 54. Wang, T., Wei, J. J., Sabatini, D. M. & Lander, E. S. 82. Patwardhan, R. P. et al. Massively parallel functional
from promoters and enhancers. Nat. Biotechnol. 33, Genetic screens in human cells using the CRISPR‑Cas9 dissection of mammalian enhancers in vivo.
510–517 (2015). system. Science 343, 80–84 (2014). Nat. Biotechnol. 30, 265–270 (2012).
29. Qi, L. S. et al. Repurposing CRISPR as an RNA-guided 55. Wang, H. et al. One-step generation of mice carrying 83. Arnold, C. D. et al. Genome-wide quantitative enhancer
platform for sequence-specific control of gene mutations in multiple genes by CRISPR/Cas-mediated activity maps identified by STARR-seq. Science 339,
expression. Cell 152, 1173–1183 (2013). genome engineering. Cell 153, 910–918 (2013). 1074–1077 (2013).
30. Chen, B. et al. Dynamic imaging of genomic loci in living 56. Koike-Yusa, H., Li, Y., Tan, E. P., Velasco‑Herrera Mdel, C. 84. Calo, E. & Wysocka, J. Modification of enhancer
human cells by an optimized CRISPR/Cas system. Cell & Yusa, K. Genome-wide recessive genetic screening in chromatin: what, how, and why? Mol. Cell 49,
155, 1479–1491 (2013). mammalian cells with a lentiviral CRISPR-guide RNA 825–837 (2013).
31. Crocker, J. & Stern, D. L. TALE-mediated modulation library. Nat. Biotechnol. 32, 267–273 (2014). 85. Shlyueva, D., Stampfel, G. & Stark, A. Transcriptional
of≈transcriptional enhancers in vivo. Nat. Methods 10, 57. Shalem, O. et al. Genome-scale CRISPR‑Cas9 knockout enhancers: from properties to genome-wide predictions.
762–767 (2013). screening in human cells. Science 343, 84–87 (2014). Nat. Rev. Genet. 15, 272–286 (2014).
32. Thakore, P. I. et al. Highly specific epigenome editing 58. Shalem, O., Sanjana, N. E. & Zhang, F. High-throughput 86. Rivera, C. M. & Ren, B. Mapping human epigenomes.
by CRISPR‑Cas9 repressors for silencing of distal functional genomics using CRISPR–Cas9. Nat. Rev. Cell 155, 39–55 (2013).
regulatory elements. Nat. Methods 12, 1143–1149 Genet. 16, 299–311 (2015). 87. Hah, N., Murakami, S., Nagari, A., Danko, C. G.
(2015). 59. Vidigal, J. A. & Ventura, A. Rapid and efficient one-step & Kraus, W. L. Enhancer transcripts mark active
33. Zhang, X. et al. Identification of focally amplified generation of paired gRNA CRISPR‑Cas9 libraries. estrogen receptor binding sites. Genome Res. 23,
lineage-specific super-enhancers in human epithelial Nat. Commun. 6, 8083 (2015). 1210–1223 (2013).
cancers. Nat. Genet. 48, 176–182 (2016). 60. Consortium, E. P. An integrated encyclopedia of DNA 88. Arner, E. et al. Transcribed enhancers lead waves of
34. Kearns, N. A. et al. Functional annotation of native elements in the human genome. Nature 489, 57–74 coordinated transcription in transitioning mammalian
enhancers with a Cas9–histone demethylase fusion. (2012). cells. Science 347, 1010–1014 (2015).
Nat. Methods 12, 401–403 (2015). 61. Konermann, S. et al. Optical control of mammalian 89. Jinek, M. et al. A programmable dual-RNA-guided DNA
35. Groner, A. C. et al. KRAB–zinc finger proteins and endogenous transcription and epigenetic states. Nature endonuclease in adaptive bacterial immunity. Science
KAP1 can mediate long-range transcriptional 500, 472–476 (2013). 337, 816–821 (2012).
repression through heterochromatin spreading. 62. Nihongaki, Y., Kawano, F., Nakajima, T. & Sato, M. 90. Dominguez, A. A., Lim, W. A. & Qi, L. S. Beyond editing:
PLoS Genet. 6, e1000869 (2010). Photoactivatable CRISPR‑Cas9 for optogenetic genome repurposing CRISPR–Cas9 for precision genome
36. Heinz, S., Romanoski, C. E., Benner, C. & Glass, C. K. editing. Nat. Biotechnol. 33, 755–760 (2015). regulation and interrogation. Nat. Rev. Mol. Cell Biol.
The selection and function of cell type-specific 63. Nihongaki, Y., Yamamoto, S., Kawano, F., Suzuki, H. 17, 5–15 (2016).
enhancers. Nat. Rev. Mol. Cell Biol. 16, 144–154 & Sato, M. CRISPR‑Cas9‑based photoactivatable 91. Wu, X. et al. Genome-wide binding of the CRISPR
(2015). transcription system. Chem. Biol. 22, 169–174 endonuclease Cas9 in mammalian cells.
37. Shlyueva, D. et al. Hormone-responsive enhancer- (2015). Nat. Biotechnol. 32, 670–676 (2014).
activity maps reveal predictive motifs, indirect 64. Polstein, L. R. & Gersbach, C. A. A light-inducible 92. Doench, J. G. et al. Rational design of highly active
repression, and targeting of closed chromatin. Mol. Cell CRISPR‑Cas9 system for control of endogenous gene sgRNAs for CRISPR–Cas9‑mediated gene inactivation.
54, 180–192 (2014). activation. Nat. Chem. Biol. 11, 198–200 (2015). Nat. Biotechnol. 32, 1262–1267 (2014).
38. Shi, Y. et al. Histone demethylation mediated by the 65. Maddalo, D. et al. In vivo engineering of oncogenic 93. Bae, S., Kweon, J., Kim, H. S. & Kim, J. S.
nuclear amine oxidase homolog LSD1. Cell 119, chromosomal rearrangements with the CRISPR/Cas9 Microhomology-based choice of Cas9 nuclease target
941–953 (2004). system. Nature 516, 423–427 (2014). sites. Nat. Methods 11, 705–706 (2014).
39. Whyte, W. A. et al. Master transcription factors and 66. Lawrence, M. S. et al. Mutational heterogeneity in 94. Fu, Y. et al. High-frequency off-target mutagenesis
mediator establish super-enhancers at key cell identity cancer and the search for new cancer-associated genes. induced by CRISPR-Cas nucleases in human cells.
genes. Cell 153, 307–319 (2013). Nature 499, 214–218 (2013). Nat. Biotechnol. 31, 822–826 (2013).
40. Groschel, S. et al. A single oncogenic enhancer 67. Narendra, V. et al. CTCF establishes discrete functional 95. Pattanayak, V. et al. High-throughput profiling of
rearrangement causes concomitant EVI1 and GATA2 chromatin domains at the Hox clusters during off‑target DNA cleavage reveals RNA-programmed
deregulation in leukemia. Cell 157, 369–381 (2014). differentiation. Science 347, 1017–1021 (2015). Cas9 nuclease specificity. Nat. Biotechnol. 31,
41. Zhou, H. Y. et al. A Sox2 distal enhancer cluster 68. Hnisz, D. et al. Activation of proto-oncogenes by 839–843 (2013).
regulates embryonic stem cell differentiation potential. disruption of chromosome neighborhoods. Science 96. Kuscu, C., Arslan, S., Singh, R., Thorpe, J. & Adli, M.
Genes Dev. 28, 2699–2711 (2014). 351, 1454–1458 (2016). Genome-wide analysis reveals characteristics of
42. Li, Y. et al. CRISPR reveals a distal super-enhancer 69. Deng, W. et al. Controlling long-range genomic off‑target sites bound by the Cas9 endonuclease.
required for Sox2 expression in mouse embryonic stem interactions at a native locus by targeted tethering of a Nat. Biotechnol. 32, 677–683 (2014).
cells. PLoS ONE 9, e114485 (2014). looping factor. Cell 149, 1233–1244 (2012). 97. Sternberg, S. H., Redding, S., Jinek, M., Greene, E. C.
43. Hnisz, D. et al. Convergence of developmental and 70. Deng, W. et al. Reactivation of developmentally silenced & Doudna, J. A. DNA interrogation by the CRISPR
oncogenic signaling pathways at transcriptional globin genes by forced chromatin looping. Cell 158, RNA-guided endonuclease Cas9. Nature 507, 62–67
super‑enhancers. Mol. Cell 58, 362–370 (2015). 849–860 (2014). (2014).
44. Canver, M. C. et al. Characterization of genomic 71. Zalatan, J. G. et al. Engineering complex synthetic 98. Ran, F. A. et al. Double nicking by RNA-guided CRISPR
deletion efficiency mediated by clustered regularly transcriptional programs with CRISPR RNA scaffolds. Cas9 for enhanced genome editing specificity. Cell 154,
interspaced palindromic repeats (CRISPR)/Cas9 Cell 160, 339–350 (2015). 1380–1389 (2013).
nuclease system in mammalian cells. J. Biol. Chem. 72. Johnson, D. S., Mortazavi, A., Myers, R. M. & Wold, B. 99. Fu, Y., Sander, J. D., Reyon, D., Cascio, V. M.
289, 21312–21324 (2014). Genome-wide mapping of in vivo protein-DNA & Joung, J. K. Improving CRISPR-Cas nuclease
45. Rajagopal, N. et al. High-throughput mapping of interactions. Science 316, 1497–1502 (2007). specificity using truncated guide RNAs. Nat. Biotechnol.
regulatory DNA. Nat. Biotechnol. 34, 167–174 73. Visel, A. et al. ChIP-seq accurately predicts tissue- 32, 279–284 (2014).
(2016). specific activity of enhancers. Nature 457, 854–858 100. Kleinstiver, B. P. et al. High-fidelity CRISPR–Cas9
46. Diao, Y. et al. A new class of temporarily phenotypic (2009). nucleases with no detectable genome-wide off-target
enhancers identified by CRISPR/Cas9‑mediated genetic 74. Ernst, J. et al. Mapping and analysis of chromatin state effects. Nature 529, 490–495 (2016).
screening. Genome Res. 26, 397–405 (2016). dynamics in nine human cell types. Nature 473, 43–49 101. Slaymaker, I. M. et al. Rationally engineered Cas9
47. Esvelt, K. M. et al. Orthogonal Cas9 proteins for (2011). nucleases with improved specificity. Science 351,
RNA‑guided gene regulation and editing. Nat. Methods 75. Boyle, A. P. et al. High-resolution mapping and 84–88 (2016).
10, 1116–1121 (2013). characterization of open chromatin across the genome.
48. Kleinstiver, B. P. et al. Broadening the targeting range Cell 132, 311–322 (2008). Acknowledgements
of Staphylococcus aureus CRISPR‑Cas9 by modifying 76. Dekker, J., Rippe, K., Dekker, M. & Kleckner, N. The authors apologize to those whose work they were not
PAM recognition. Nat. Biotechnol. 33, 1293–1298 Capturing chromosome conformation. Science 295, able to mention or cite owing to space limitations. R.L. is a
(2015). 1306–1311 (2002). fellow of the Fundação para a Ciência e Tecnologia, Portugal
49. Kleinstiver, B. P. et al. Engineered CRISPR‑Cas9 77. Lieberman-Aiden, E. et al. Comprehensive mapping of (SFRH/BD/74476/2010; POPH/FSE). This work was
nucleases with altered PAM specificities. Nature 523, long-range interactions reveals folding principles of the supported by funds of enhReg ERC-AdV (ERC‑2012‑ADG
481–485 (2015). human genome. Science 326, 289–293 (2009). - N°322493) and NWO (NGI 93512001/2012) to R.A.
50. Ran, F. A. et al. In vivo genome editing using 78. Fullwood, M. J. et al. An oestrogen-receptor-α-bound
Staphylococcus aureus Cas9. Nature 520, 186–191 human chromatin interactome. Nature 462, 58–64 Competing interests statement
(2015). (2009). The authors declare no competing interests.

604 | SEPTEMBER 2016 | VOLUME 17 www.nature.com/nrm

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.