REVIEWS

The chemistry of Cas9 and its CRISPR

colleagues
Janice S. Chen1 and Jennifer A. Doudna1–5
Abstract | RNA-guided binding and cleavage of nucleic acids by CRISPR–Cas systems is a
defining feature of bacterial and archaeal adaptive immunity against viruses and plasmids.
As a result of their programmable ability to cut specific DNA and RNA sequences, Cas9 and
related single-subunit effector proteins from CRISPR–Cas systems have been widely adopted
for research and therapeutic genome engineering applications. In this Review, we discuss the
chemistry of macromolecules involved in the multistep interference pathway used by CRISPR–
Cas systems that mediate accurate nucleic acid target recognition and cutting. Although this
Review mainly focuses on DNA interference by Cas9, we briefly explore nucleic acid targeting
by the single-effector proteins Cas12 and Cas13 to emphasize the conserved themes of precision
DNA and RNA cleavage within class 2 CRISPR–Cas systems. We further highlight the unique
mechanisms of surveillance complex formation, substrate recognition and target cleavage in
molecular detail across diverse single-subunit CRISPR–Cas interference proteins.

Mobile genetic elements

DNA sequences that are
capable of moving around a
genome, including transposons,
plasmids and bacteriophages.
Protospacers
Spacer precursors that are
captured from foreign DNA
and that are complementary
to the CRISPR RNA (crRNA)
spacer sequence.
1
Department of Molecular
and Cell Biology, University of
California.
2
Howard Hughes Medical
Institute, University of
California.
3
Innovative Genomics Institute,
University of California.
4
Department of Chemistry,
University of California,
Berkeley, California 94720,
USA.
5
Molecular Biophysics and
Integrated Bioimaging
Division, Lawrence Berkeley
National Laboratory, Berkeley,
California 94720, USA.
Correspondence to J.A.D. 
doudna@berkeley.edu
doi:10.1038/s41570-017-0078
Published online 4 Oct 2017
CRISPR–Cas systems involve complex RNA-guided
adaptive immune pathways in bacteria and archaea to
combat infection by mobile genetic elements (MGEs)1–3.
Immunological memory begins with the acquisition
of foreign DNA sequences derived from invading
bacterio­phages or plasmids4–7, known as protospacers,
the selection of which requires a protospacer adjacent
motif sequence (PAM sequence). These protospacers are
integrated into the leader proximal end of a host CRISPR
array 8,9, which yields new spacers that are flanked by
~20–50 base pair (bp) direct repeats10,11. Transcription of
the CRISPR array generates a non-coding RNA transcript
— the ‘precursor CRISPR RNA’ (pre-crRNA) — which is
further processed by specific Cas proteins or host factors
to produce short mature crRNAs that contain a single
spacer and at least one repeat 12–16. Finally, assembly of
the mature crRNA with the single-effector Cas protein
or multiple Cas subunits forms a surveillance ribonucleo-
protein (RNP) complex. This RNP complex recognizes
foreign nucleic acids containing a PAM sequence and
destroys sequences complementary to the spacer segment
of the crRNA through endonucleolytic cleavage by Cas
nucleases17–20 (FIG. 1).
The programmable, RNA-guided DNA-targeting
and RNA-targeting functions of CRISPR–Cas systems
have been repurposed for genome- and transcriptome-
engineering applications21–28, which has led to extensive
study and interest in CRISPR–Cas biology and enzyme
mechanisms. In particular, the facile ability to direct the
widely used Cas9 protein to any sequence of interest
by manipulating its guide RNA has inspired a genome-
editing revolution29–34. Recent efforts in mining bacterial
and archaeal genomes for single-effector proteins have
uncovered rare and divergent CRISPR–Cas systems, which
offer new RNA-guided endonucleases with functional
diversity 35–37. To fully use the nucleic acid-targeting
capabilities of Cas9 and additional class  2 effector
Cas proteins, biochemical and structural studies have
guided our understanding of the molecular mechanisms
underpinning target search and destruction.
In this Review, we briefly introduce the evolutionary
relationships between Cas9 and other single-effector
proteins of class 2 CRISPR–Cas systems and we detail
the multistep mechanisms for RNP complex formation,
substrate recognition and endonucleolytic cleavage.
Although the Cas9 interference pathway remains the focus
of this Review, we highlight the mechanistic similarities
and differences in target recognition and cleavage across
the evolutionarily distinct class 2 single-effector proteins.
CRISPR–Cas systems
The ongoing competition between interference by
CRISPR–Cas systems and infection by MGEs has trig-
gered the rapid evolution and extensive diversification of
CRISPR loci and Cas proteins38,39. Although CRISPR–Cas
systems are functionally modular, the selective pressures
inflicted by invading genetic elements have led to the
emergence of different Cas proteins that carry out both
specialized and combined roles in spacer acquisition,
crRNA biogenesis, DNA and RNA interference, and

NATURE REVIEWS | CHEMISTRY VOLUME 1 | ARTICLE NUMBER 0078 | 1

©
2
0
1
7
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
.
REVIEWS

Bacteriophage

Adaptation
Class 2 interference genes Protospacer
Type II
RuvC I RuvC II RuvC III S1 S2 S3
cas effector cas1 cas2 R R R R
HNH CRISPR
cas9 array Transcription

Type V RuvC I, RuvC II and RuvC III

Effector Pre-crRNA
protein
cas12a, cas12b, cas12c, cas12d and cas12e RNP crRNA
formation processing
Type VI Mature
HEPN HEPN Interference crRNA
cas13a, cas13b and cas13c

Protospacer adjacent motif
sequence
(PAM sequence). A short
sequence adjacent to the
protospacer within foreign
DNA. Recognition of the PAM
sequence by effector Cas
proteins triggers target
interference.
Endonucleolytic cleavage
Achieved by Cas proteins that
hydrolyse internal
phosphodiester bonds within a
nucleotide chain.
Guide RNA
An RNA molecule that includes
the CRISPR RNA (crRNA) and
directs the Cas interference
protein to a target site that is
complementary to the spacer
sequence.
Interference
The final stage of CRISPR
immunity that involves
RNA-directed cleavage of
target nucleic acids by Cas
proteins.
Nuclease
An enzyme that catalyses the
cleavage of phosphodiester
bonds between nucleic acids.
Helicase
An enzyme that unwinds
double-stranded DNA using
the energy from ATP
hydrolysis.
Figure 1 | Overview of CRISPR–Cas adaptive immunity and interference genes from class 2 systems.
Nature CRISPR–Cas
Reviews | Chemistry
immunity is encoded in the host genome at loci comprising the CRISPR array and neighbouring cas genes. During
infection, adaptation machinery acquires ‘protospacers’ from the invading genetic element (for example, a
bacteriophage) and integrates them into the leader end of the CRISPR array to encode immunological memory.
Transcription of the CRISPR array generates a precursor CRISPR RNA (pre-crRNA) transcript, which is further processed by
specific Cas proteins or host factors into mature crRNAs containing the unique protospacer sequence (S1–3; red, yellow
and purple, respectively). Assembly of interference Cas proteins with crRNAs forms the ribonucleoprotein (RNP)
surveillance complex, which scans invading nucleic acids and targets foreign complementary sequences for degradation.
In class 2 CRISPR–Cas systems, the interference module comprises a type II (Cas9), type V (Cas12a, Cas12b, Cas12c,
Cas12d and Cas12e) or type VI (Cas13a, Cas13b and Cas13c) single-effector protein encoded by the cas operon. HEPN,
higher eukaryotes and prokaryotes nucleotide-binding RNase domain; HNH, histidine–asparagine–histidine nuclease
domain; R, direct repeat; RuvC, RNase H‑like fold nuclease domain.
ancillary functions3. These ancillary functions include On the basis of the available genomic data, type I and
the putative role for the 5ʹ to 3ʹ exonuclease activity of III systems seem to be distantly related and account for
Cas4 during protospacer selection40, the possible involve- ~50% and ~25% of all known CRISPR loci in bacteria
ment of Csn2 in chromosomal protection during spacer and archaea, respectively 3,53,54.
integration41,42 and a novel role for Cas10 in the synthesis
of cyclic oligonucleotides to allosterically activate Csm6 Class 2 CRISPR–Cas systems. Class 2 CRISPR–Cas
ribonuclease activity 43,44. In this section, we briefly dis- systems are characterized by RNA-guided effector
cuss the two major classes of CRISPR–Cas systems complexes that require only a single multidomain pro-
to introduce the evolutionary relationships of genes tein for interference3. The most comprehensively studied
encoding the interference module. For a detailed overview example in this class is type II, which is characterized
of CRISPR–Cas classification, we recommend several by the signature gene cas9 (REF. 55). Cas9 is capable of
comprehensive reviews3,37,45,46. programmable RNA-guided DNA interference without
the need for additional proteins21, a property that has
Class 1 CRISPR–Cas systems. Class 1 CRISPR–Cas been used to repurpose Cas9 for genome editing. This
systems comprise the multisubunit crRNA–effector finding has since attracted intensive investigation into
complexes that are required for target interference and its function and mechanism using biochemical recon-
are further classified into types I, III and IV on the basis stitution and in cellular contexts29–32. Although type II
of the presence of signature cas genes3. Type I systems systems represent a small proportion (~5%) of all known
are characterized by Cas3, which has dual nuclease and CRISPR–Cas systems in bacteria, the Cas9 phylogeny
helicase activity and is recruited by the multiprotein com- is diverse and further classified into subtypes II-A,
plex Cascade (CRISPR-associated complex for antiviral II-B and II-C3,54. Subclassification of type II systems is
defence) for DNA target degradation47. Type III systems based on the presence of auxiliary cas genes csn2 (sub-
include the large subunit Cas10, which is required for type II-A) or cas4 (subtype II-B), which are involved in
Csm (subtype III-A) and Cmr (subtype III-B) assembly adaptation but not in the interference stage40,56. Although
and degradation of both RNA and transcriptionally type II systems were previously thought to be restricted
active DNA48–52. Finally, the putative type IV systems con- to bacteria, a divergent Cas9 protein has been discovered
tain the uncharacterized gene csf1, which encodes a large in nanoarchaea; however, no functional data have so far
subunit that is related to the Cas8 protein in Cascade3. showed that this system is active35.

2 | ARTICLE NUMBER 0078 | VOLUME 1 www.nature.com/natrevchem

©
2
0
1
7
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
.
 REVIEWS

RuvC nuclease domain
Contains an RNase H‑like fold
and cleaves single-stranded
DNA through a two metal
mechanism.
HEPN domain
(Higher eukaryotes and
prokaryotes nucleotide-binding
RNase domain). Contains
conserved motifs and functions
as an RNase or a non-catalytic
RNA-binding domain.
Ribonuclease
An enzyme that hydrolyses the
phosphodiester bonds of an
RNA backbone.
Exonuclease
An enzyme that hydrolyses
phosphodiester bonds, one at
a time, from the ends of a
nucleic acid chain.
Single-guide RNA
A chimeric RNA molecule in
which the CRISPR RNA
(crRNA), which contains a
sequence complementary to
the target DNA, is covalently
linked to a trans-activating
crRNA (tracrRNA).
The interest in discovering new facets of CRISPR
biology and harnessing the natural diversity of class 2
single-effector proteins for genome-engineering applica-
tions has led to the identification of type V and type VI
CRISPR–Cas systems36. To date, the experimentally tested
type V systems (the effector proteins of which have been
redesignated as Cas12a–e (REF. 37)) include the single-
effector proteins Cas12a (also known as Cpf1; subtype
V-A), Cas12b (also known as C2c1; subtype V-B), Cas12c
(also known as C2c3; subtype V-C), Cas12d (also known
as CasY; subtype V-D) and Cas12e (also known as CasX;
subtype V-E)46, all of which are evolutionarily distinct
from Cas9 (REFS 35–37). Similarly to Cas9, Cas12 pro-
teins contain a conserved RuvC nuclease domain that is
known to hydrolyse single-stranded DNA (ssDNA)57–59.
Notably, Cas12a contains a second catalytic domain
that is independently responsible for processing its own
pre-crRNA59,60.
Type VI CRISPR–Cas systems consist of an effector
protein (reclassified as Cas13 (REF. 37)) containing the
signature HEPN domain (higher eukaryotes and pro­
karyotes nucleotide-binding RNase domain)61, of which
the Cas13a (also known as C2c2; subtype VI-A) and
Cas13b (subtype VI-B) proteins have been experimen-
tally shown to function as RNA-guided RNases36,62.
Although Cas13a effectors have been shown to indis-
criminately cleave single-stranded RNAs (ssRNAs) on
binding to a complementary target RNA63–66, they also
possess pre-crRNA processing activity 62–64, similar to that
of Cas12a proteins36,60. Compared with type II systems,
types V and IV are rare and represent only ~2% of
sequenced CRISPR–Cas systems in bacteria and archaea
combined3,37; however, the streamlined and diverse abil-
ities of type V and IV systems to target nucleic acids
holds potential for various research and therapeutic
applications.
crRNA processing and RNP formation
The molecular memory of infection encoded in the
CRISPR array is transcribed to generate a long, non-
coding pre-crRNA transcript that requires processing
and loading into Cas nucleases to authorize RNA-
guided CRISPR immunity 12. In this section, we briefly
review two major classes for crRNA processing in
class 2 CRISPR systems — trans-activating crRNA
(tracrRNA)-dependent pathways, which require a host
factor RNase III (REF. 13), and self-processing pathways,
which are accomplished by the interference protein
alone60,64. We also discuss the structural determinants for
assembling the guide RNAs into their effector proteins
to generate the RNP surveillance complex.
tracrRNA-dependent pathways. Nucleic acid targeting
by RNP surveillance complexes is dependent on mature
crRNAs, which facilitate complex activation and molecular
recognition of the cognate target 12,16. However, for
type II systems, crRNA processing and Cas9 activation
requires an additional small, non-coding tracrRNA13.
Bioinformatic analyses followed by experimental valida-
tion using RNA sequencing (RNA-seq) and biochemical
reconstitution have revealed that tracrRNAs, which
are typically ~75–110 nucleotides in length, contain
a sequence complementary to CRISPR repeats that is
required for RNP assembly and crRNA processing and
are abundantly expressed within the intergenic regions
of the CRISPR–Cas loci13,67,68. Although tracrRNAs can
be diverse in sequence and secondary structure, they
contain a conserved ‘anti-repeat’ sequence that base
pairs with each direct repeat of the pre-crRNA to form
an ~30 bp double-stranded RNA (dsRNA) duplex 68. In
the presence of Cas9, this dsRNA substrate is cleaved
by a non-Cas ribonuclease (RNase III), which generates
individual Cas9‑bound crRNA–tracrRNA complexes
from the pre-crRNA transcript 13. Further processing by
an unknown exonuclease trims the 5ʹ end of the crRNA,
generating the Cas9 surveillance complex that is primed
to cleave a complementary DNA target13 (FIG. 2a).
To simplify dual-guided proteins such as Cas9,
Cas12b and Cas12e for programmable targeting in
research applications, the crRNA–tracrRNA molecules
required for their activity have been fused by a tetraloop
to generate a chimeric single-guide RNA (sgRNA)21,69,70.
Since the first generation of Cas9 sgRNAs, efforts to
optimize their on‑target efficiency in cultured cells
have involved maintaining the 3ʹ hairpins, extending
the repeat–anti-repeat duplex and adding synthetic
modifications to improve the stability of the sgRNA71–73.
Although many sgRNA variants have been developed for
diverse applications74–78, a highly efficient sgRNA design
for genome editing and genomic loci imaging comprises
a long tracrRNA tail containing the native 3ʹ hairpins and
a 5 bp extension of the crRNA–tracrRNA duplex 71,72,79.
Activation of dual-guided interference proteins requires
competent RNP complex formation, which depends on
RNA folding, stability and sequence, as well as structural
recognition of the crRNA–tracrRNA duplex. Therefore,
in the case of type II-A Cas9 proteins, minor alterations
to the sgRNA scaffold have been shown to reduce or to
abolish endonuclease activity 80–82.
Secondary structure predictions and sgRNA-bound
and target-bound Cas9 crystal structures have revealed
conserved RNA folds that are necessary to define their
functional roles83–86. The sgRNA from the widely used
type II-A Streptococcus pyogenes Cas9 (SpCas9) includes
a 5ʹ spacer and repeat sequence from the crRNA, followed
by the tracrRNA-derived anti-repeat sequence (the
repeat–anti-repeat duplex of which consists of an upper
stem, bulge and lower stem), nexus and hairpins 1 and 2
at the 3ʹ end84–87 (FIG. 2b). Functional studies have shown
that, irrespective of the spacer sequence, specific nucleo­
tides in the bulge and nexus sequences of the SpCas9
sgRNA are most crucial for cleavage87. With respect to
binding energy, SpCas9 binds hairpins 1–2 alone with an
equilibrium dissociation constant (Kd) in the picomolar
range, which is almost indistinguishable from that of the
full-length sgRNA88. In comparison, SpCas9 binds to the
spacer–nexus region with nanomolar affinity, which sug-
gests that the structured 3ʹ end of the sgRNA contributes
most of the binding energy for the Cas9–sgRNA com-
plex 88. Förster resonance energy transfer (FRET)-based
experiments have also shown that the 20‑nucleotide
spacer sequence at the 5ʹ end of the sgRNA is crucial

NATURE REVIEWS | CHEMISTRY VOLUME 1 | ARTICLE NUMBER 0078 | 3

©
2
0
1
7
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
.
REVIEWS

a tracrRNA-dependent d Self-processing

S1 S2 S3
cas effector R R R R cas effector

Transcription
Transcription

Effector pre-crRNA Effector

protein protein pre-crRNA
tracrRNA

RNase III

crRNA–tracrRNA crRNA

RNP RNP

b SpCas9 e AsCas12a
Upper stem
A
A A
G UA Lower stem
Spacer A Spacer
U G
C C AA 5′
G
Tetra-loop
A G Bulge G A CU
linker G U
A A A U A U U
Repeat A U G A G C C G
U A U
U A U A U A C G Repeat A U
U A G G U A CA U
G U A
tracrRNA U A U C G C G C G
G U G C U A
G C A U G C U U
5′ AA G UU A U C A G UUUU 3′ 3′
Spacer Nexus Hairpins 1 and 2 Pseudoknot Spacer

c AaCas12b f LshCas13a
R–AR duplex 1 and 2 Spacer
C GGCAC
U UGAGAAGU 3′
A ACUCUUCG A CCGUG
A
A
Spacer
U Spacer AUC
G G C C A C U UU U G
C Stem 3 A A
A C G G U G G AC A A
A C G
C G
A G U U G C C C GA Stem 2 C G
tracrRNA C GA
U U C A A C G GG U Repeat
Repeat U C Bulge Spacer
U U A A G A C A G GA A U
G Stem 1 5′ GG CC A AAA C 3′
5′ G G U C U A
Figure 2 | TracrRNA-dependent and self-processing pathways for crRNA biogenesis and RNP Nature Reviews
complex | Chemistry
formation.
The CRISPR array is transcribed to form the precursor CRISPR RNA (pre-crRNA) transcript containing multiple spacers
(S1–3; red, yellow and purple, respectively) flanked by direct repeats (R; blue), which requires further processing to produce
mature crRNAs containing a single spacer sequence (yellow) and at least one direct repeat (blue). In part a, trans-activating
crRNA (tracrRNA)-dependent crRNA biogenesis is shown in which transcription of an intergenic region of the CRISPR array
generates the small, non-coding tracrRNA molecule (red), which contains a region of complementarity to the direct repeat
(anti-repeat) of the pre-crRNA. The anti-repeat sequence base pairs with each direct repeat, thus forming a pre-crRNA–
tracrRNA double-stranded RNA (dsRNA) duplex. In the presence of Cas9 (Protein Data Bank identifier: 4UN3; part b) or
Cas12b (PDB ID: 5U30; part c), RNase III recognizes and cleaves the repeat–anti-repeat (R–AR) RNA duplex. The dual
crRNA–tracrRNA molecule (which can be fused to generate a chimeric single-guide RNA (sgRNA)) remains tightly
associated with the interference protein to form the ribonucleoprotein (RNP) complex and stabilizes a conformation that
is competent for target surveillance. In part b, hybridization of the crRNA repeat (blue) and tracrRNA anti-repeat (red)
generates the crRNA–tracrRNA duplex containing an upper stem, bulge and lower stem, and the remaining tracrRNA
sequence folds into the nexus and hairpins 1 and 2. In part c, the crRNA–tracrRNA duplex forms the R–AR duplex, and the
remaining tracrRNA topology consists of a pseudoknot structure and four helical stems. In part d, self-processing crRNA
biogenesis is shown in which an additional RNA molecule or nuclease is not required for pre-crRNA processing beyond
the interference protein itself. The single-effector protein Cas12a (PDB ID: 5B43; part e) or Cas13a (PDB ID: 5WTK;
part f) recognizes and binds to the structured repeat within the pre-crRNA transcript, which triggers endoribonuclease
activity that is distinct from that of the active site involved in DNA target cleavage. The resulting RNP complex is competent
for downstream target recognition and degradation. In part e, the crRNA repeat (blue) folds into a pseudoknot structure
upstream of the spacer sequence (yellow). In part f, the crRNA repeat folds into a stem–loop containing a 2‑nucleotide
bulge at the base of the stem–loop. AaCas12b, Alicyclobacillus acidoterrestris Cas12b; AsCas12a, Acidaminococcus spp.
Cas12a; LshCas13a, Leptotrichia shahii Cas13a; SpCas9, Streptococcus pyogenes Cas9.

4 | ARTICLE NUMBER 0078 | VOLUME 1 www.nature.com/natrevchem

©
2
0
1
7
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
.
 REVIEWS

Scissile phosphate
The phosphate within the
nucleic acid backbone that is
cleaved by a nuclease.
for triggering the ‘lobe closure’ conformation that is
required to generate the target search complex, whereas
the 3ʹ end hairpins mainly contribute to the stability of
the closed state89.
Crystal structures of the sgRNA-bound and target-
bound Cas9 complexes have validated these experimental
results and showed that the Cas9 protein makes extensive
hydrogen bond contacts and aromatic stacking interac-
tions with the repeat–anti-repeat duplex, the nexus
sequence and the base of hairpins 1 and 2 (REFS 83–86).
Furthermore, the first ten nucleotides from the 5ʹ end
of the spacer are disordered in the crystal structure,
whereas the remaining ten nucleotides are pre-organized
in a pseudo-A form conformation, which is thermo­
dynamically favourable for initiating R‑loop formation83.
Variations in sgRNA hairpin structure were observed
in crystal structures of type II-A Staphylococcus aureus
Cas9 (SaCas9)90 and type II-B Francisella novicida Cas9
(FnCas9)91 proteins, which provided the structural basis
for orthogonality between cognate Cas9–sgRNA pairs92.
Notably, the crystal structure of the sgRNA–target-bound
type II-C orthologue Campylobacter jejuni Cas9 (CjCas9)
features a triple-helix within a pseudoknot that involves
extensive intramolecular interactions necessary for
activating endonucleolytic cleavage93.
The general pathway for dual crRNA–tracrRNA-
guided processing and assembly extends beyond type II
systems and has also been observed and experimentally
shown for the evolutionarily distant Cas12 interference
proteins, Cas12b and Cas12e (REFS 35,36). In contrast
to the type II sgRNA architecture, secondary structure
predictions for the Cas12e and Cas12b sgRNAs have
shown an opposite polarity to that of Cas9 sgRNAs,
comprising a 5ʹ end hairpin and anti-repeat sequence
from the tracrRNA, followed by the crRNA repeat and
spacer sequence at the 3ʹ end35,36,94,95. For Alicyclobacillus
acidoterrestris Cas12b (AaCas12b), sgRNA–target-
bound crystal structures have shown an unexpected
sgRNA topology consisting of a pseudoknot structure and
four helical stems, compared with the simple stem-loop
fold that is predicted to be its minimum free-energy
conformation94,95 (FIG. 2c). A disordered RNA linker that
was previously thought to form part of the repeat–anti-
repeat duplex was further shown to be dispensable for
AaCas12b activity 94. On the basis of in silico co‑folding,
these structural features also seem to be possible with
the crRNA–tracrRNA sequence from the orthologous
Bacillus thermoamylovorans Cas12b (BtCas12b), which
suggests that these RNA folds might be conserved in
type VB systems94.
Self-processing pathways. Unlike the dual crRNA–
tracrRNA-guided examples, the finding that the Cas12a
effector does not require additional RNA or protein com-
ponents for crRNA processing and DNA target cleavage
offered a new paradigm for single crRNA-guided class 2
interference proteins58–60 (FIG. 2d). Initial studies showed
that a minimal type V-A expression plasmid consisting
of only cas12a from F. novicida and a CRISPR array was
capable of pre-crRNA maturation and plasmid DNA
interference58; the dual RNase and DNase activity by
Cas12a was validated by small RNA sequencing and
biochemical reconstitution58–60. The mature crRNA
comprises a structured repeat followed by the spacer
sequence, and the 5ʹ stem–loop provides the majority of
the crRNA binding energy for Cas12a endoribonuclease
activity 60,96,97 (FIG. 2e). Although Cas12a has ~50‑fold
higher affinity for crRNA in the presence of a magnesium
ion (Mg 2+)96, one study established that pre-crRNA
processing is metal independent and requires the
2ʹ‑hydroxyl (2ʹ‑OH) group of the ribonucleotide
upstream of the scissile phosphate for acid–base catalysis59.
On the basis of the crRNA-bound Cas12a crystal struc-
ture from Lachnospiraceae spp. bacterium ND2006
(LbCas12a) and crRNA–target–Cas12a structures from
Acidaminococcus spp. Cas12a (AsCas12a) and F. novi-
cida Cas12a (FnCas12a), the RNA processing and RuvC
catalytic centres are spatially distant and functionally
independent 59,96–99.
Similarly to Cas12a, the Cas13a and Cas13b proteins
also evolved the streamlined functions of crRNA process-
ing and target interference in a single-effector protein62–65.
Unexpectedly, the crRNA-bound structure of Leptotrichia
shahii Cas13a (LshCas13a) revealed a two-nucleotide
bulge at the base of the 5ʹ stem–loop in the crRNA,
which seemed to be evolutionarily conserved and func-
tionally required65 (FIG. 2f). Biochemical and structural
studies have further validated the functional and spatial
separation of the crRNA-processing and RNA-targeting
active sites in Cas13a (REFS 63–66,176), offering another
example of convergent evolution and underscoring the
functional modularity across CRISPR–Cas systems.
PAM recognition
In most CRISPR–Cas systems, target DNA search and
recognition requires both complementary base pairing
between the guide RNA and the target DNA and the
presence of a conserved PAM sequence on the target
DNA adjacent to the target site100–102. PAM recognition
by Cas proteins activates interference as a strategy to
distinguish self from non-self sequences2, and single
mutations in the PAM can prevent CRISPR–Cas cleavage
activity in vitro and in vivo21,102–106.
The native PAM sequence for the commonly used
SpCas9 is 5ʹ‑NGG‑3ʹ, in which N can be any of the four
DNA nucleotides100. Single-molecule experiments have
shown that the Cas9–sgRNA complex initiates its search
for target DNA by binding to a PAM sequence before
interrogating the flanking DNA for potential comple-
mentarity to the guide RNA103. Target recognition occurs
through three-dimensional collisions; Cas9 rapidly dis-
sociates from DNA that does not contain the appropriate
PAM sequence, whereas it binds for a longer duration at
sites containing a PAM sequence, with the dwell time
depending on the degree of complementarity between
the guide RNA and the adjacent DNA107,108. Once Cas9
has found a target site with the appropriate PAM, it trig-
gers DNA unwinding from the PAM-proximal end to
the PAM-distal end of the target site103.
The molecular mechanism of PAM recognition
involves binding of the PAM DNA to a positively
charged groove in Cas9, in which the first base in the

NATURE REVIEWS | CHEMISTRY VOLUME 1 | ARTICLE NUMBER 0078 | 5

©
2
0
1
7
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
.
REVIEWS

DNA melting PAM sequence (which can be any nucleotide) remains interactions via the major groove with two respective
The process of DNA strand base paired with its nucleobase counterpart but does arginine residues (Arg1333 and Arg1335) located in a
separation that does not not interact with Cas9 (REFS  85,86,109) . The adja- β‑hairpin of the PAM-interacting domain of SpCas9
require an external energy cent GG dinucleotide on the non-target strand of the (REFS 85,86,109) (FIG. 3a). Although naturally occurring
source such as ATP.
PAM duplex forms base-specific hydrogen-bonding differences in sequence and location of PAM-interacting
residues explain how diverse PAM sequences are rec-
ognized by Cas9 orthologues90,91,93, engineered residue
a SpCas9 modifications in the PAM-interacting domain have been
Guide RNA Phosphate
lock-loop shown to alter PAM specificities109,110. Structural studies of
5′ P P P P P 3′ S1109 engineered SpCas9 PAM variants support an ‘induced fit’
S1109 K1107
+5 +4 +3 +2 +1 –1 –2 –3
–3* –2* mechanism, whereby recognition of non-canonical PAM
Target –1* K1107 sequences requires subtle DNA backbone distortion of the
strand 3′ P P P P P P P P 5′
N PAM C C R1333 PAM duplex without changing the conformation of
Non- N G G
the PAM-interacting β-hairpin85,109. In particular, mutation
target 5′ P P P P P P P P 3′
strand +5* +4* +3* +2* +1* –1* –2* –3* R1335 of a threonine to an arginine residue (Thr1337Arg) ena-
Target bled recognition of a guanine base in the fourth position
strand Non-target
R1333 R1335 strand of the varied PAM (5ʹ‑NGNG‑3ʹ)109.
For SpCas9, both the PAM duplex-bound structure
b AsCas12a and the double-stranded DNA (dsDNA)-bound struc-
ture suggest that PAM–Cas9 interactions trigger local
Guide RNA
structural changes that destabilize the adjacent DNA
K548 5′ P P P P P
3′
M604 –2*
–1*
duplex base pairing and facilitate hybridization of the
–4 –3 –2 –1 +1 +2 +3 +4
T539
guide RNA to the target DNA85,86. In the PAM duplex-
3′ P P P P P P P P 5′ K607
–3* bound SpCas9 structure, a sharp kink is observed in the
A A A N
T T T N target DNA strand immediately upstream of the PAM,
T167
5′ P P P P P P P P 3′ with the connecting phosphodiester group (referred to
–4* –3* –2* –1* +1* +2* +3* +4* –4* as +1 phosphate) stabilized by a phosphate lock-loop
Target (facilitated by Lys1107 and Ser1109) located in the
T539 T167 K607 Non-target
strand
strand PAM-interacting domain85 (FIG. 3a). Superimposition
of the guide RNA-bound SpCas9 structure with the
c AaCas12b
PAM duplex-bound SpCas9 structure shows that this
Guide RNA phosphate lock-loop moves inwards towards the central
N400 5′ P P P P P 3′ –1* nucleic acid recognition channel on binding to target
–3 –2 –1 +1 +2 +3 +4 +5 –2* DNA83,85. Interactions between the phosphate lock-loop
3′ P P P P P P P P 5′
A A N
N144 R122 of SpCas9 and the +1 phosphate on PAM recognition
T T N –3* have been proposed to aid DNA melting and stabilization
5′ P P P P P P P P 3′ G143 N400 of the guide RNA–target DNA hybrid85,86. However,
–3* –2* –1* +1* +2* +3* +4* +5*
mutation of the phosphate lock-loop residues in SpCas9
G143 N144
R122 Target Non-target did not reduce endonucleolytic cleavage activities on a
strand strand complementary target DNA85. Although the phosphate
Figure 3 | PAM recognition and sequence-specific target identification. lock-loop is probably not involved in DNA duplex dest-
a | Streptococcus pyogenes Cas9 (SpCas9; Protein Data Bank identifier: 4UN3) | Chemistry
Nature Reviews abilization, a recent study from 2017 suggested that these
recognizes a 5ʹ‑NGG‑3ʹ protospacer adjacent motif (PAM) sequence, where N signifies residues instead promote initial pairing of the guide
any nucleotide, located downstream of the DNA target sequence. Hydrogen bonding RNA with the target strand111. Furthermore, the lack of
interactions between the crucial arginine (Arg1333 and Arg1335) residues and the sequence similarity of the phosphate lock-loop among
guanine dinucleotide (GG) on the non-target strand occur through the major groove of
Cas9 orthologues suggests that initial DNA unwinding
the PAM duplex, which induces local structural changes that enable DNA unwinding.
Stabilization of the upstream (+1) phosphate on the target strand by the Ser1109 and might require additional elements beyond the phosphate
Lys1107 residues on the PAM-interacting domain of SpCas9 make up a phosphate lock-loop residues85,90,91,93,111.
lock-loop, which has been suggested to facilitate guide RNA–target DNA hybridization. The dsDNA-bound SpCas9 structure also suggests
b | Acidaminococcus spp. Cas12a (AsCas12a; PDB ID: 5B43) recognizes a 5ʹ‑TTTN‑3ʹ PAM how PAM recognition triggers local DNA unwinding. As
located upstream of the DNA target sequence. The crucial Lys607 residue of AsCas12a observed in the PAM duplex-bound SpCas9 structure,
forms hydrogen bonds with the thymine (–2*) and adenine (–3) on the minor groove of the target DNA strand kinks at the +1 phosphodiester
the PAM duplex, whereas the conserved Lys548, Met604, Thr167 and Thr539 residues linkage and then pairs with the guide RNA segment to
form hydrogen bonds or van der Waals interactions with nucleotides on both strands of form a pseudo-A form RNA–DNA hybrid85,86. By con-
the PAM duplex. c | Alicyclobacillus acidoterrestris Cas12b (AaCas12b; PDB ID: 5U30) trast, the non-target DNA strand threads into a tight
recognizes a 5ʹ‑TTN‑3ʹ PAM located upstream of the DNA target sequence. The key
tunnel located within the nuclease lobe towards the
arginine (Arg122) and asparagine (Asn144 and Asn400) residues also reach into the minor
groove of the PAM to form hydrogen bonds with A–T base pairs within the PAM duplex. RuvC active site of SpCas9. The PAM-proximal non-
The target strand and the non-target strand (*) of the DNA duplex are labelled, target DNA strand is stabilized by extensive hydropho-
phosphate groups within the target sequence are labelled with (+) and those outside the bic and van der Waals interactions, in which the first
target sequence are labelled with (–). Note that the polarity of the target sequence for nucleotide upstream of the PAM (referred to as the +1*
Cas12a in part b and Cas12b in part c is opposite to that of Cas9. position) stacks onto the PAM duplex 86. This intra-strand

6 | ARTICLE NUMBER 0078 | VOLUME 1 www.nature.com/natrevchem

©
2
0
1
7
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
.
 REVIEWS

RNA strand invasion
A process in which the guide
RNA segment interrogates a
double-stranded DNA (dsDNA)
target to initiate DNA
unwinding and to form an
RNA–DNA heteroduplex.
R‑loop
An RNA–DNA heteroduplex
and a displaced DNA strand,
which is the end result of RNA
strand invasion by Cas
ribonucleoprotein (RNP)
complexes.
base stacking might help to stabilize the PAM duplex
and thereby to facilitate PAM recognition by SpCas9
through base-specific interactions with the GG dinu-
cleotide. The sharp kinks and flipped bases include
nucleotides at positions +1 on the target strand and
+2* and +3* on the non-target strand, which are mainly
exposed to bulk solvent 86 (FIG. 3a). The orientation of
these nucleotides, which function as the nucleation site
to initiate target binding, might explain how Cas9 sam-
ples the DNA adjacent to the PAM sequence for guide
RNA complementarity and RNA strand invasion following
PAM recognition.
PAM recognition in type V systems. In contrast to
SpCas9, the PAM for Cas12 proteins contains a thymine
(T)-rich sequence and is located at the 5ʹ end of the target
sequence36,58 (FIG. 3b,c). The crRNA–target–AsCas12a
crystal structure revealed a distorted PAM duplex with
a narrow minor groove, which functions as the location
for base-specific contacts with amino acid side chains
via van der Waals and hydrogen bond interactions97,98. In
particular, the critical Lys607 residue of AsCas12a forms
a hydrogen bond with thymine at the −2* and adenine
at the −3 position of the 5ʹ‑TTTN‑3ʹ PAM, whereas the
neighbouring Met604 and Pro599 residues interact with
the −2 adenine base pair 97,98 (FIG. 3b). Hydrogen bonding
between amino acid side chains and the PAM duplex
also occurs through the minor groove in the crRNA–
target–AaCas12b crystal structure, which recognizes the
5ʹ‑TTN‑3ʹ PAM sequence upstream of the complemen-
tary DNA target 95 (FIG. 3c). For both Cas12a and Cas12b,
mutation of residues involved in base-specific contacts
with the PAM duplex markedly reduced or abolished
endonucleolytic cleavage activity, consistent with the
requirement for base specificity in PAM recognition95,98.
However, a structure-guided mutagenesis screen suc-
cessfully identified amino acid substitutions in the
PAM-interacting domain of LbCas12a, which conferred
cleavage activities at non-canonical PAM sequences, thus
showing that Cas12 proteins can be engineered to alter
PAM specificity 112–114.
R‑loop formation and DNA unwinding
Following PAM verification, the RNP surveillance
complex interrogates a complementary dsDNA target
by RNA strand invasion. Base pairing of the crRNA
spacer sequence with the DNA target strand displaces
the non-target strand and forms an R‑loop structure,
the stability of which is crucial for permitting target
cleavage86,115–117 (FIG. 4). The observation that mismatches
between the guide RNA and the target DNA directly
upstream of the PAM diminished Cas9 binding and
cleavage of the target DNA provided initial evidence for
a ‘seed region’ that functioned as the nucleation site for
R‑loop formation21,22,103,105. Substrate competition experi-
ments with SpCas9 also showed that extending the guide
RNA–target DNA sequence complementarity beyond
the first 10–12 bp of the target site promoted stepwise
RNA–DNA heteroduplex formation and disfavoured
re‑annealing of the dsDNA duplex 103. Additional studies
have since confirmed the importance of the seed region,
thus supporting a model for unidirectional R‑loop
propagation towards the PAM-distal end of the target
sequence103,118–120 (FIG. 4).

a Cas9
Target scanning PAM recognition Local DNA unwinding R-loop formation

Effector
protein

5′ 3′
sgRNA

Cas12 Cas9
PAM PAM
dsDNA
target

b Cas12

5′ 3′

Figure 4 | Model for unidirectional DNA unwinding and R‑loop formation in Cas9 and Cas12 interference
Nature proteins.
Reviews | Chemistry
Protospacer adjacent motif (PAM) sequence recognition induces guide RNA–double-stranded (dsDNA) target
hybridization and unidirectional DNA unwinding from the 3ʹ−5ʹ direction in Cas9 proteins (part a) or from the 5ʹ−3ʹ
direction in Cas12 (part b) to form the R‑loop structure. In the presence of mismatches in the seed region (the region
adjacent to the PAM), RNA strand invasion is non-productive and the ribonucleoprotein (RNP) surveillance complex rapidly
dissociates from the dsDNA substrate. PAM-distal mutations still enable rapid dsDNA–RNP complex association rates and
productive RNA strand invasion, but the lack of dsDNA–guide RNA complementarity at the PAM-distal end alters R‑loop
stability, which favours re‑annealing of the DNA strands and thus decreases the cleavage rate. Target recognition and
cleavage therefore depends on stable R‑loop formation, which probably involves key protein interactions with the entire
length of the RNA–DNA heteroduplex to activate Cas9 or Cas12 nuclease activity. sgRNA, single-guide RNA.

NATURE REVIEWS | CHEMISTRY VOLUME 1 | ARTICLE NUMBER 0078 | 7

©
2
0
1
7
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
.
REVIEWS

Kinetic inhibition
A model in the context of RNA
strand invasion that explains
how impeding the rate of
R‑loop formation reduces Cas9
cleavage activity.
Kinetic Monte Carlo
analyses
Reveal the time evolution of a
given process (that is, stability
of the R‑loop over time with
mismatched substrates or
single-guide RNA (sgRNA)
variants).
HNH nuclease domain
Contains a histidine–
asparagine–histidine (HNH)
motif and is the nuclease within
Cas9 that hydrolyses the target
DNA strand through a one
metal ion mechanism.
As dsDNA unwinding by Cas9 is ATP independent21,
RNA strand invasion also depends on duplex distortion
on PAM binding to induce local DNA melting 86,111,121,
which is reminiscent of the DNA bending process
observed in type I-E and I-C Cascade complexes106,122–124.
A cryo-electron microscopy (cryo‑EM)-based re­consti-
tution of SpCas9 containing a complete R‑loop showed a
~30° bend angle between the protruding dsDNA ends86,
which is thought to stabilize the R‑loop structure by
enabling strand separation and reducing torsional
stress121,125,126. Although this static model provided a
snapshot of the full R‑loop before target cleavage, ques-
tions regarding R‑loop initiation could be more directly
addressed by capturing the PAM-bound (pre-strand
invasion) Cas9–sgRNA–dsDNA complex, the structure
of which has yet to be elucidated. Notably, the structurally
compact Cas9 orthologues in type  II-C systems
have been shown to lack robust DNA-unwinding
capabilities, which suggests that kinetic inhibition of
the initial RNA strand invasion step occurs in these
smaller variants82.
In silico analyses based on thermodynamic param-
eters for dsDNA duplex and DNA–RNA heteroduplex
stabilities have offered insight into the propensity for
Cas9 R‑loop formation. On the basis of kinetic Monte
Carlo analyses (KMC analyses) using nearest neighbour
free-energy parameters, simulations showed that the
guide RNA completely and stably invades a comple-
mentary 20‑nucleotide DNA target, independently of
Cas9 (REF. 127). By comparing KMC simulations with
mismatches between the guide RNA and target DNA to
reported SpCas9 activity at off-target sites, KMC analysis
showed a correlation between guide RNA stability
around the 14–17 nucleotide position and SpCas9 off-
target cleavage, which emphasizes that the R‑loop is capa-
ble of traversing mismatches given transient stability at
this crucial position71,127. As R‑loop stability was a strong
predictor of cleavage rate compared with DNA–RNA
binding energies or the position of mismatches alone,
understanding the kinetics of RNA strand invasion
proved valuable for anticipating off-target cleavage127. It
is worth noting that all KMC experiments were carried
out with an initial R‑loop length of 10 bp, which assumed
strand invasion beyond the seed region, and were there-
fore unable to test de novo R‑loop formation127. A sepa-
rate biophysical model predicted that SpCas9 was much
more likely to spontaneously dissociate than to form a
stable R‑loop when mutations in the target sequence
were present, which suggests that target search involves
many rounds of PAM recognition, local unwinding and
initial strand displacement and dissociation before suffi-
cient target DNA–guide RNA complementarity is found,
which drives R‑loop formation108,128 (FIG. 4a).
To experimentally understand the energetics of DNA
unwinding and the effects of mismatches between the
guide RNA and target DNA on RNA strand invasion,
a few single-molecule studies have also probed the
real-time kinetics of SpCas9 R‑loop formation. Single-
molecule DNA supercoiling experiments have shown
that PAM mutations regulate R‑loop formation through
kinetic inhibition, whereas PAM-distal mutations
only affect R‑loop stability 116. Consistent with the tar-
get search model that involves many rounds of R‑loop
formation and dissociation before interrogation of a
complementary target, R‑loops were not observed with
targets containing <11 bp of PAM-proximal comple-
mentarity, which suggests that Cas9 is capable of sensing
along the entire RNA–DNA heteroduplex 116. A separate
single-molecule FRET (smFRET) study measuring
R‑loop conformations with a complementary DNA tar-
get observed two distinct structural states at the PAM-
distal end, which shows more heterogeneity than was
previously expected81. These studies have contributed
greater complexity to the simple model of unidirectional
R‑loop formation.
The observation that PAM-distal mutations have a
marked effect on R‑loop stability and observed cleav-
age rates without altering the rate of R‑loop formation
emphasizes the crucial role of target recognition by the
effector protein116. Although Cas9 is more tolerant of
PAM-distal mismatches than seed region mutations129,
several lines of evidence support a model for DNA–
RNA heteroduplex sensing at the PAM-distal end.
Cas9 does not use an R‑loop ‘locking’ mechanism,
during which protein domains reorganize to stabilize
the R‑loop structure116; this would require additional
proofreading steps in addition to simple base pairing
to ensure fidelity of target recognition and cleavage.
Small fluctuations in R‑loop stability at the PAM-distal
end induced by target mismatches have a considerable
effect on cleavage activity 129, and increasing the number
of PAM-distal mismatches impedes conformational
activation of the HNH nuclease domain (histidine–arginine–
histidine nuclease domain)89. Furthermore, comparison
of the pre-target-bound and ssDNA-bound RNP complex
structures showed a dramatic conformational change
when SpCas9–sgRNA binds to a target DNA strand, even
without the PAM-containing non-target strand83,84. On
the basis of these crystal structures, R‑loop sensing prob-
ably involves key residues in the α-helical lobe of SpCas9
that interact with the DNA–RNA hetero­duplex 84–86,111
to enable further Cas9 conformational rearrangements
to trigger target cleavage.
R-loop formation in type V systems. Although the deter-
minants of R‑loop formation and target recognition for
Cas12 effector proteins are not as well understood as for
SpCas9, existing experimental and structural evidence
suggests that Cas12a and Cas12b also initiate RNA strand
invasion at a 5ʹ PAM-proximal seed region (FIG. 4b).
Single mismatches between the FnCas12a crRNA and
target DNA within positions +1 to +5 reduced in vitro
target cleavage activity, whereas single mismatches
between the AaCas12b sgRNA and target DNA within
positions +1 to +18 abrogated cleavage58,94. Cell-based
activity assays also showed that AsCas12a and LbCas12a
were highly intolerant of single mismatches between the
crRNA and target DNA within nucleotide positions +2
to +6 and +11 to +18 (REF. 130). Interestingly, PAM-distal
mismatches at the 3ʹ end beyond position +20 had no
effect on cleavage for AsCas12a and LbCas12a130, which
was consistent with target-bound crystal structures in

8 | ARTICLE NUMBER 0078 | VOLUME 1 www.nature.com/natrevchem

©
2
0
1
7
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
.
 REVIEWS

Apoprotein
The inactive, unbound state of
the protein.
Molecular dynamics
simulation
(MD simulation). Computer
simulations that capture the
time evolution of atomic and/or
molecular systems by
numerically solving Newton’s
equations of motion.
Cas9‑digested
whole-genome sequencing
An in vitro method for
detecting Cas9 cleavage sites
within genomic DNA using
whole-genome sequencing.
which the last four PAM-distal nucleotides on the target
strand were not duplexed with the crRNA97,98. However,
single mismatches at positions before +20 markedly
reduced cleavage activity, which suggests that Cas12a
target recognition might also require PAM-distal sens-
ing through α-helical rearrangements to achieve high
fidelity cleavage58,130,131.
Nucleic acid target cleavage
Cleavage of the nucleic acid target marks the final step in
the CRISPR–Cas interference pathway. Precise cutting
requires tight coupling between the processes of target
recognition and catalytic activation, which is typically
achieved through conformational rearrangements in
the effector proteins in order to properly orient one or
more active sites with the desired substrate132.
Cas9 nuclease domains and activity. For Cas9, gener-
ating a blunt dsDNA break requires coordination of the
two catalytic nuclease domains, the HNH domain and
the RuvC domain, which cleave the target strand and the
non-target strand, respectively 21 (FIG. 5a,d). Notably, Cas9
is considered a single-turnover enzyme; after the DNA
substrate has been cleaved, the target strand remains
base paired to the guide RNA and the protein is unable to
bind additional targets for subsequent reactions103. Cas9
hydrolyses the scissile phosphates of the target strand
and the non-target strand via a one metal ion mecha-
nism in the HNH nuclease domain and a two metal ion
mechanism in the RuvC nuclease domain133–136.
The HNH domain is characterized by the conserved
ββα-fold and its catalytic pocket consists of three active
site residues (Asp839, His840 and Asn863 in SpCas9)83–
85,137
. As the HNH domain structure has yet to be eluci-
dated in its active conformation, details of the cleavage
process remain elusive and are inferred based on homol-
ogy to other well-studied systems, such as the Holliday
junction resolvase phage T4 endonuclease VII and the
nonspecific periplasmic Vibrio vulnificus nuclease84,138,139.
By homology modelling, the active site residues Asp839
and Asn863, together with the oxygen atoms of the target
strand scissile phosphate, coordinate a magnesium ion
(Mg 2+) within the catalytic centre of the HNH domain.
The His840 residue functions as a general base to activate
a water molecule, which acts as a nucleophile to attack
the electrophilic +3 phosphate on the target strand and
generates the 5ʹ phosphate and 3ʹ hydroxyl (3ʹ‑OH) prod-
ucts21,84,134 (FIG. 5b). Although the HNH active site requires
Mg 2+ for catalysis21,133–135, divalent cations (including
Mg 2+, Ca2+ and Co2+) have also been shown to stabilize
the HNH domain in its active conformation140,141.
The RuvC nuclease domain, which shares an RNase
H‑like fold with other retroviral integrases142–144, consists
of four active site residues (Asp10, Glu762, His983 and
Asp986 in SpCas9)83–85. A crystal structure containing
the non-target strand revealed the presence of the −3
phosphate near the RuvC catalytic centre86. Although
no metals were present in this structure, the non-target
strand-containing RuvC domain aligned almost per-
fectly with a separate manganese(ii) ion (Mn2+)-bound
RuvC apoprotein structure86,137. Superimposition of
the Mn2+ ions into the RuvC active site predicted a
metal coordination distance of ~5.5 Å between the
Mn2+ ion and the non-bridging oxygen atom on the
−3 phosphate, which was slightly longer than the 2.1 Å
Mg–O coordination distance typically observed for
catalysis86,137. However, the putative cleavage mecha-
nism involves two metal ion coordination by Asp10,
Glu762 and Asp986 and the oxygen atoms of the scis-
sile phosphate, which makes the phosphate susceptible
to nucleophilic attack by the His983‑activated water
molecule21,84,142–144 (FIG. 5c). To model how metal ions bridge
the gap between the RuvC active site and the non-target
strand scissile phosphate, a molecular dynamics simulation
(MD simulation) suggested that introducing two Mg 2+
ions at the −4 phosphate led to the largest decrease in
distance between the scissile phosphate and the Asp10
and His983 active site residues145. The MD simulations
also showed that Mg 2+ binding was more stable at the −4
phosphate, providing in silico evidence that Cas9 might
instead generate a 1 bp 5ʹ staggered end145. Although Cas9
has been biochemically shown to cleave the non-target
strand at the −3 phosphate18,21, in vitro Cas9‑digested whole-
genome sequencing (Digenome-seq) has shown cutting at
both the −3 and the −4 phosphates146. Notably, previous
biochemical studies have also showed that the non-target
strand is trimmed by 3ʹ−5ʹ exonuclease activity of the Cas9
RuvC nuclease following the initial cleavage event 21,86,
which might contribute to the observed staggered cut.
Although the MD simulations provide a putative mech-
anism for RuvC-dependent trimming of the non-target
strand, the elucidation of the preferred −4 phosphate
cleavage site requires further experimental evidence.
Allosteric recognition and targeting specificity. Crystal
structures of SpCas9 in various substrate-bound
states have also shown a large, rigid-body translation
and rotation by the HNH domain towards the target
strand83–86,137, the direct interaction of which with the
scissile phosphate has yet to be captured structurally
(FIG.  2a) . Biochemical and bulk FRET experiments
first showed that the conformational state of the
HNH domain controls DNA cleavage activity 89, and
a smFRET study later showed that the HNH domain
occupies a stably ‘docked’ conformation when bound
to a complementary DNA substrate140. However, when
PAM-distal mismatches are present, smFRET exper-
iments revealed that the HNH domain transitions
towards a catalytically inactive ‘conformational check-
point’ (REF.  140), thus explaining how Cas9 cleaves
only a subset of sequences to which it binds104,119,147.
Although the HNH domain is sensitive to mismatches
along the RNA–DNA heteroduplex 89,140, the domain
itself does not directly interact with PAM-distal nucleic
acids on the RNA–DNA heteroduplex 83–86, alluding to a
mechanism of allosteric recognition that controls HNH
conformational activation148.
The role of allosteric effects on Cas9 conformational
activation has been shown in several steps of target
recognition and cleavage86,89,145,148–150. MD simulations
showed that PAM binding induces an ‘open to closed’
conformational transition that reorients the domains

NATURE REVIEWS | CHEMISTRY VOLUME 1 | ARTICLE NUMBER 0078 | 9

©
2
0
1
7
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
.
REVIEWS

REC3
(sensor)
REC2
Recognition (regulator)
lobe

HNH (inactive HNH (active

conformation) conformation)
REC1

sgRNA
RuvC

Nuclease
lobe
PAM-interacting

Target strand
Non-target strand

b HNH active site c RuvC active site

N854

H840 D839 E762

N863
D10
D986 H983

H840
O Base H983
O O Base
H H N O
O H O H H
NH2 O H O NH O N
E762 H
N863 O– O
– H O H NH
O P O
O Mn2+ H
Mg2+ O Base O– P O
O O O Base
O– O Mn2+ O
D839 O NH2
O O H D10
O O– O H O H
O– P O
H
N854 O O– P O
D986 O

d RNA
Cas9 Cas12 Cas13
trans
dsDNA dsDNA
PAM PAM cis
RNA PFS

Nature Reviews | Chemistry

10 | ARTICLE NUMBER 0078 | VOLUME 1 www.nature.com/natrevchem

©
2
0
1
7
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
.
 REVIEWS

◂ Figure 5 | Nuclease domain structure and metal-dependent nucleic acid target cleavage. a | Crystal structure of
Streptococcus pyogenes Cas9 (SpCas9; Protein Data Bank identifier: 5F9R) bound to its single-guide RNA (sgRNA) and
target double-stranded (dsDNA), depicting the major domains of the recognition (REC) and nuclease lobes. SpCas9
undergoes a large conformational change on dsDNA target binding in order to position the histidine–asparagine–
histidine (HNH) nuclease domain in the cleavage-competent, active conformation and simultaneously to trigger RuvC
(RNase H‑like fold) nuclease domain activity. The HNH and the RuvC domains of the nuclease lobe cleave the target strand
and the non-target strand of the dsDNA target, respectively. Although existing crystal structures have failed to capture
the active site residues that are engaged with the DNA substrate, details of the Cas9 cleavage mechanism have been
inferred on the basis of homology to known nucleases that contain the conserved HNH or RuvC domains. b | A structural
alignment of the SpCas9 HNH domain (PDB ID: 5F9R) with the phage T4 endonuclease VII HNH domain (PDB ID: 2QNC)
bound to its target strand scissile phosphate supports a one metal ion mechanism involving magnesium ion (Mg2+)
coordination at the active site (top panel). In the chemical depiction, the polar residues (Asn863, Asp839 and Asn854) help
to coordinate the Mg2+ ion in the catalytic site. The His840 residue functions as a general base to activate a water molecule
for nucleophilic attack on the Mg2+-coordinated scissile phosphate, generating the 5ʹ phosphate and 3ʹ hydroxyl (3ʹ-OH)
ends on the cleaved DNA backbone (bottom panel). c | A structural alignment of the non-target strand-bound SpCas9
RuvC domain (PDB ID: 5F9R) with the manganese(ii) ion (Mn2+)-bound apoprotein (apo) SpCas9 RuvC domain (PDB ID:
4CMQ) shows a putative two metal ion cleavage mechanism (top panel). In the chemical depiction, the acidic residues
(Glu762, Asp10 and Asp986) help to coordinate the two Mn2+ ions in the active site. The His963 residue activates a water
molecule to attack the electrophilic scissile phosphate, generating the 5ʹ phosphate and 3ʹ-OH ends on the cleaved DNA
products (bottom panel). d | Although the general mechanism of phosphodiester hydrolysis is consistent across class 2
interference proteins, cleavage patterns are distinct among Cas9 (blunt end), Cas12 (5ʹ staggered overhang) and Cas13
(promiscuous cis- and trans-cleavage) systems. For Cas9, dsDNA cleavage is accomplished by the concerted activation of
the HNH and RuvC nucleases, which respectively cut the target strand and the non-target strand upstream of the
protospacer adjacent motif (PAM) sequence. For Cas12, the RuvC nuclease is independently responsible for cleaving both
the target strand and the non-target strand of the dsDNA target downstream of the PAM, but the precise cleavage
mechanism remains unknown. For Cas13, binding to a complementary single-stranded RNA (ssRNA) sequence adjacent
to a protospacer flanking sequence (PFS; analogous to PAM) activates the higher eukaryotes and prokaryotes
nucleotide-binding RNase (HEPN) domain nuclease to indiscriminately cleave the bound RNA target (cis-cleavage) and
other ssRNAs in solution (trans-cleavage).

of SpCas9 for target recognition150. On formation of
the R‑loop structure, recognition of the RNA–DNA
heteroduplex by the REC3 domain (within the recog-
nition lobe) was shown to allosterically regulate global
Cas9 conformational changes by communicating with
the REC2 domain, which sterically restricts or permits
HNH domain access to the scissile phosphate148. The
displacement of the non-target strand was also crucial
for activating and stabilizing HNH docking in the
active conformation140,145. In addition, conformational
activation of the HNH domain exerts allosteric control
over the RuvC nuclease via two α-helical linkers (L1
and L2) that connect the nuclease domains; these were
proposed to function as allosteric transducers to medi-
ate concerted DNA cleavage of the target strand and
non-target strand89. A dsDNA-bound crystal structure
of SpCas9 also showed major structural rearrangements
of L1 and L2 to facilitate HNH conformational activa-
tion in the pre-cleavage state86. Furthermore, MD sim-
ulations indicated how the structural plasticity of Cas9
mediates allosteric crosstalk between the REC lobe,
HNH and RuvC domains to achieve the catalytically
competent structural conformation145,149,150. However,
the precise mode of communication from REC3 to the
HNH nuclease domain remains unknown and warrants
further investigation.
Despite conformational regulation, SpCas9 is still
capable of cleaving sequences that resemble the comple-
mentary target, as detected by several unbiased genome-
wide off-target methods104,147,151–153. The mechanism of
off-target cleavage correlates with transient R‑loop sta-
bility and HNH docking 89,127, but off-target interrogation
also depends on the number of PAM sequences within the
genome and the epigenetic context in mammalian cells71,
107,154,155
. Avoiding off-target cleavage is an ongoing chal-
lenge for using Cas9 in target-specific applications and
existing strategies for reducing off-target effects include
direct RNP delivery (by titrating the Cas9:sgRNA con-
centration)156,157, sgRNA modifications (by lengthening
and/or truncating the 5ʹ end)74,158 or protein engineering
(by introducing paired nickases, dimerization or point
mutations)148,159–164, some of which have been extensively
reviewed165–167. With respect to truncating the sgRNA
and introducing alanine substitutions at the protein–
nucleic acid interface, the underlying rationale for
improving specificity focused on weakening Cas9–target
interactions 74,163,164. Although these manipulations
reduced the number of detectable off-target events
compared with the wild-type SpCas9–sgRNA complex
in cultured cells74,163,164, it was biochemically shown that
the cleavage specificity of high-fidelity Cas9 variants
does not depend on target binding affinity 148. Rather,
smFRET experiments have shown that these Cas9 var-
iants have an altered threshold for HNH conformational
activation when bound to DNA substrates and that muta-
tion of residues within REC3 that are involved in target
recognition can alter the specificity of target cleavage148.
A kinetic model was also proposed to maximize RNP
target discrimination by simultaneously increasing the
RNP dissociation rate and decreasing the cleavage rate
of off-target dsDNA sequences168; however, the missing
emphasis on maintaining and improving on‑target effi-
ciency remains an important caveat of these predictions.
Understanding how these high-fidelity strategies affect
R‑loop stability also remains an important unanswered
question that could improve targeting specificity.

NATURE REVIEWS | CHEMISTRY VOLUME 1 | ARTICLE NUMBER 0078 | 11

©
2
0
1
7
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
.
REVIEWS

cis-cleavage
Occurs when the Cas13–
CRISPR RNA (crRNA) complex
binds a complementary
single-stranded RNA (ssRNA)
target, which activates the
external higher eukaryotes and
prokaryotes nucleotide-binding
RNase (HEPN) domain
catalytic pocket to cleave the
bound ssRNA target.
trans-cleavage
Occurs when the Cas13–
CRISPR RNA (crRNA) complex
binds a complementary
single-stranded RNA (ssRNA)
target, which activates the
external higher eukaryotes and
prokaryotes nucleotide-binding
RNase (HEPN) domain
catalytic pocket to cleave
nonspecific ssRNAs in solution.
Target cleavage in type V and type VI systems. Crystal
structures of Cas12a and Cas12b bound to both the crRNA
and dsDNA target have offered greater insight into the tar-
get cleavage mechanism of Cas12 effector proteins59,94–98.
These RNA-guided, dsDNA-targeting proteins contain
the conserved RuvC nuclease domain36, but several
features indicate a DNA-targeting mechanism that is
distinct from that of Cas9. The cleavage reaction for
both Cas12a and Cas12b generates a 2–8 nucleotide
5ʹ overhang staggered cut 58,59,94,95 (FIG. 5d). A crRNA–
target-bound structure of FnCas12a showed that
the target strand scissile phosphate is located in a DNA
duplex downstream of the 3ʹ end of the crRNA, which
was speculated to undergo further unwinding through
an unknown mechanism to facilitate target strand
cleavage59. Notably, two genome-wide studies of Cas12a
specificity reported substantially fewer off-target cleavage
events and slightly lower on‑target efficiency compared
with SpCas9 (REFS 130,146), which suggests different
mechanisms for target recognition or conformational
activation. Furthermore, Cas12a and Cas12b lack an
HNH nuclease domain or additional protein domain
with detectable structural or sequence homology to
known nucleases63,94–98. Nonetheless, a putative nuclease
(Nuc) domain that is responsible for cleaving the target
strand has been assigned to both Cas12a and Cas12b
on the basis of its proximity to the target strand cleav-
age site; this remains controversial owing to a paucity of
evidence for a bona fide active site and a lack of struc-
tural similarity with other nucleases (and even between
Cas12 effector proteins)94–98. Only a single point mutation
(Arg1226Ala) selectively inhibited target strand cleavage
by AsCas12a98; however, this residue is located within a
linker between the RuvC and Nuc domain and there-
fore does not conclusively represent a catalytic residue
within the Nuc domain. No catalytic residue specific for
target strand cleavage has yet been identified for Cas12b.
However, for both Cas12a and Cas12b, a single RuvC
point mutation was shown to abrogate cleavage of both
dsDNA strands58,59,94,95,98.
Following RuvC mutational analysis and the obser-
vation that an extended single-stranded segment of the
target strand folds back to the RuvC active site in target-
bound AaCas12b structures95, the RuvC catalytic pocket
was suggested to be responsible for cleaving both the
target strand and the non-target strand via separate
cleavage events 59,95. A subsequent FnCas12a study
was unable to identify any mutations within the Nuc
domain that selectively inhibited target strand cleavage,
thereby eliminating the nuclease domain as a bona fide
nuclease59. However, the model of target strand and
non-target strand cleavage by a single active site in Cas12
effectors still warrants further investigation, as questions
regarding the kinetics of strand scission and structures
of the cleavage-competent conformational state remain
unanswered.
The functionally validated Cas13a and Cas13b pro-
teins recognize and cleave RNA by a mechanism that is
distinct from the previously discussed class 2 effectors.
Initial studies with Cas13a showed that crRNA-guided
recognition of a complementary ssRNA activates two
catalytic HEPN domains that promiscuously cleave the
target at a region beyond the crRNA–ssRNA duplex,
known as cis-cleavage63,64. Binding of Cas13a to the
cognate ssRNA also triggers indiscriminate cleavage
of ssRNAs in solution, known as trans-cleavage63,64. This
finding has uncovered a mechanism for nonspecific
RNase activity by RNA target recognition, which has
since been used for RNA detection applications64,169. It
was later shown that the Cas13a family is functionally
divided into two groups on the basis of crRNA and
ssRNA substrate preferences 170. In addition to tar-
geted and collateral RNA cleavage, the related Cas13b
protein is further regulated by the accessory proteins
Csx27 and Csx28, which are separately encoded in the
CRISPR–Cas locus and repress and enhance RNase
activity, respectively 62. A recent crystal structure of the
crRNA-bound LshCas13a showed an external HEPN
catalytic pocket comprising two active site residues from
each HEPN domain65, consistent with the mechanism of
cis- and trans-cleavage by a single catalytic site (FIG. 5d).
A subsequent target-bound Cas13a structure has also
shown that HEPN dimerization is a prerequisite for
complex activation63–66. Finally, it has been speculated,
but remains unknown, whether type VI CRISPR–Cas
systems have an ‘off switch’ against collateral RNA damage
by Cas13 effectors to prevent off-target activation
of RNase activity.
Conclusions
The unique features of RNA-guided adaptive immu-
nity in CRISPR–Cas systems have motivated the next
generation of genome-editing technologies, with Cas9
paving the way for sequence-specific DNA-targeting
applications. Although numerous biochemical and
structural studies on Cas9 have provided key insights
into each step of the DNA interference pathway, interest
in enhancing targeting specificity with the widely used
SpCas9 still requires a comprehensive understanding of
the thermodynamic and kinetic parameters that control
target recognition and complex activation. Furthermore,
there is growing momentum in discovering and adopt-
ing smaller Cas9 orthologues such as SaCas9 and CjCas9
for genome editing; these are compact molecules that
could aid in vivo delivery and have distinct specificity
profiles147,171,172.
Finally, the discovery of type V and type VI effector
proteins is a reminder that the Cas9 family represents
only a fraction of the diversity in class 2 CRISPR–Cas
systems. Emerging mechanistic studies on these distinct
CRISPR–Cas systems have shown both common themes
and functional differences, from RNP complex forma-
tion to substrate preference and nucleic acid degradation.
However, Cas12a is the only interference protein in
the type V and type VI systems that has been success-
fully repurposed for genome editing in mammalian
cells63,130,146,173–175. As more systems become optimized for
in vivo applications, our understanding of DNA inter-
ference at a molecular level becomes increasingly more
important in order to maximize the potential of diverse
programmable endonucleases for novel research tools
and innovative therapies.

12 | ARTICLE NUMBER 0078 | VOLUME 1 www.nature.com/natrevchem

©
2
0
1
7
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
.
 REVIEWS

1. Sorek, R., Kunin, V. & Hugenholtz, P. CRISPR — a
widespread system that provides acquired resistance
against phages in bacteria and archaea. Nat. Rev.
Microbiol. 6, 181–186 (2008).
2. van der Oost, J., Westra, E. R., Jackson, R. N. &
Wiedenheft, B. Unravelling the structural and
mechanistic basis of CRISPR-Cas systems. Nat. Rev.
Microbiol. 12, 479–492 (2014).
3. Makarova, K. S. et al. An updated evolutionary
classification of CRISPR−Cas systems. Nat. Rev.
Microbiol. 13, 722–736 (2015).
4. Mojica, F. J., Diez-Villasenor, C., Garcia-Martinez, J. &
Soria, E. Intervening sequences of regularly spaced
prokaryotic repeats derive from foreign genetic
elements. J. Mol. Evol. 60, 174–182 (2005).
This study uncovers the origin of spacers on the
basis of homology to foreign DNA sequences and
proposes that CRISPR is involved in specific
immunity against MGEs.
5. Bolotin, A., Quinquis, B., Sorokin, A. & Ehrlich, S. D.
Clustered regularly interspaced short palindrome
repeats (CRISPRs) have spacers of extrachromosomal
origin. Microbiology 151, 2551–2561 (2005).
6. Pourcel, C., Salvignol, G. & Vergnaud, G. CRISPR
elements in Yersinia pestis acquire new repeats by
preferential uptake of bacteriophage DNA, and
provide additional tools for evolutionary studies.
Microbiology 151, 653–663 (2005).
7. Barrangou, R. et al. CRISPR provides acquired
resistance against viruses in prokaryotes. Science
315, 1709–1712 (2007).
This is the first study to experimentally show the
role of CRISPR systems in bacterial adaptive
immunity.
8. Arslan, Z., Hermanns, V., Wurm, R., Wagner, R. &
Pul, U. Detection and characterization of spacer
integration intermediates in type IE CRISPR−Cas
system. Nucleic Acids Res. 42, 7884–7893 (2014).
9. Nunez, J. K., Lee, A. S., Engelman, A. & Doudna, J. A.
Integrase-mediated spacer acquisition during CRISPR−
Cas adaptive immunity. Nature 519, 193–198
(2015).
10. Jansen, R., Embden, J. D., Gaastra, W. &
Schouls, L. M. Identification of genes that are
associated with DNA repeats in prokaryotes. Mol.
Microbiol. 43, 1565–1575 (2002).
11. Kunin, V., Sorek, R. & Hugenholtz, P. Evolutionary
conservation of sequence and secondary structures in
CRISPR repeats. Genome Biol. 8, R61 (2007).
12. Brouns, S. J. et al. Small CRISPR RNAs guide antiviral
defense in prokaryotes. Science 321, 960–964
(2008).
This study shows that the formation of mature
crRNAs containing the spacer and repeat
sequences is essential for mediating an antiviral
response.
13. Deltcheva, E. et al. CRISPR RNA maturation by trans-
encoded small RNA and host factor RNase III. Nature
471, 602–607 (2011).
This study identifies the tracrRNA molecule and
shows its role in directing crRNA maturation and
antiviral immunity.
14. Carte, J., Wang, R., Li, H., Terns, R. M. & Terns, M. P.
Cas6 is an endoribonuclease that generates guide
RNAs for invader defense in prokaryotes. Genes Dev.
22, 3489–3496 (2008).
15. Gesner, E. M., Schellenberg, M. J., Garside, E. L.,
George, M. M. & Macmillan, A. M. Recognition and
maturation of effector RNAs in a CRISPR interference
pathway. Nat. Struct. Mol. Biol. 18, 688–692 (2011).
16. Hochstrasser, M. L. & Doudna, J. A. Cutting it close:
CRISPR-associated endoribonuclease structure and
function. Trends Biochem. Sci. 40, 58–66 (2015).
17. Garneau, J. E. et al. The CRISPRCas bacterial immune
system cleaves bacteriophage and plasmid DNA.
Nature 468, 67–71 (2010).
18. Gasiunas, G., Barrangou, R., Horvath, P. & Siksnys, V.
Cas9−crRNA ribonucleoprotein complex mediates
specific DNA cleavage for adaptive immunity in
bacteria. Proc. Natl Acad. Sci. USA 109,
E2579–E2586 (2012).
19. Wiedenheft, B., Sternberg, S. H. & Doudna, J. A. RNA-
guided genetic silencing systems in bacteria and
archaea. Nature 482, 331–338 (2012).
20. Marraffini, L. A. & Sontheimer, E. J. CRISPR
interference: RNA-directed adaptive immunity in
bacteria and archaea. Nat. Rev. Genet. 11, 181–190
(2010).
21. Jinek, M. et al. A programmable dual-RNA-guided
DNA endonuclease in adaptive bacterial immunity.
Science 337, 816–821 (2012).
This study characterizes the determinants of DNA
cleavage for SpCas9 and establishes programmable
targeting using the crRNA–tracrRNA molecule or a
chimeric sgRNA.
22. Cong, L. et al. Multiplex genome engineering using
CRISPR−Cas systems. Science 339, 819–823 (2013).
23. Jinek, M. et al. RNA-programmed genome editing in
human cells. eLife 2, e00471 (2013).
24. Mali, P. et al. RNA-guided human genome engineering
via Cas9. Science 339, 823–826 (2013).
25. Rath, D., Amlinger, L., Hoekzema, M., Devulapally, P. R.
& Lundgren, M. Efficient programmable gene silencing
by Cascade. Nucleic Acids Res. 43, 237–246 (2015).
26. O’Connell, M. R. et al. Programmable RNA recognition
and cleavage by CRISPR−Cas9. Nature 516,
263–266 (2014).
27. Gilbert, L. A. et al. Genome-scale CRISPR-mediated
control of gene repression and activation. Cell 159,
647–661 (2014).
28. Larson, M. H. et al. CRISPR interference (CRISPRi) for
sequence-specific control of gene expression. Nat.
Protoc. 8, 2180–2196 (2013).
29. Doudna, J. A. & Charpentier, E. Genome editing. The
new frontier of genome engineering with
CRISPR‑Cas9. Science 346, 1258096 (2014).
30. Hsu, P. D., Lander, E. S. & Zhang, F. Development and
applications of CRISPR‑Cas9 for genome engineering.
Cell 157, 1262–1278 (2014).
31. Mali, P., Esvelt, K. M. & Church, G. M. Cas9 as a
versatile tool for engineering biology. Nat. Methods
10, 957–963 (2013).
32. Sander, J. D. & Joung, J. K. CRISPR−Cas systems for
editing, regulating and targeting genomes. Nat.
Biotechnol. 32, 347–355 (2014).
33. Fellmann, C., Gowen, B. G., Lin, P. C., Doudna, J. A. &
Corn, J. E. Cornerstones of CRISPR−Cas in drug
discovery and therapy. Nat. Rev. Drug Discov. 16,
89–100 (2017).
34. Barrangou, R. & Horvath, P. A decade of discovery:
CRISPR functions and applications. Nat. Microbiol. 2,
17092 (2017).
35. Burstein, D. et al. New CRISPRCas systems from
uncultivated microbes. Nature 542, 237–241 (2017).
36. Shmakov, S. et al. Discovery and functional
characterization of diverse class 2 CRISPR−Cas
systems. Mol. Cell 60, 385–397 (2015).
This study reports the discovery of three distinct
class 2 CRISPR–Cas systems (C2c1, C2c2 and
C2c3) and shows that these CRISPR loci contain
functional interference proteins.
37. Shmakov, S. et al. Diversity and evolution of class 2
CRISPR−Cas systems. Nat. Rev. Microbiol. 15,
169–182 (2017).
This review provides a comprehensive overview of
all class 2 CRISPR–Cas systems discovered to date
and discusses their evolutionary origins and
relationships.
38. Stern, A. & Sorek, R. The phage-host arms race:
shaping the evolution of microbes. Bioessays 33,
43–51 (2011).
39. Koonin, E. V. & Makarova, K. S. CRISPR−Cas:
evolution of an RNA-based adaptive immunity system
in prokaryotes. RNA Biol. 10, 679–686 (2013).
40. Zhang, J., Kasciukovic, T. & White, M. F. The CRISPR
associated protein Cas4 is a 5ʹ to 3’ DNA exonuclease
with an iron-sulfur cluster. PLoS ONE 7, e47232
(2012).
41. Nam, K. H., Kurinov, I. & Ke, A. Crystal structure of
clustered regularly interspaced short palindromic
repeats (CRISPR)-associated Csn2 protein revealed
Ca2+-dependent double-stranded DNA binding
activity. J. Biol. Chem. 286, 30759–30768 (2011).
42. Arslan, Z. et al. Double-strand DNA end-binding and
sliding of the toroidal CRISPR-associated protein
Csn2. Nucleic Acids Res. 41, 6347–6359 (2013).
43. Kazlauskiene, M., Kostiuk, G., Venclovas, C.,
Tamulaitis, G. & Siksnys, V. A cyclic oligonucleotide
signaling pathway in type III CRISPR−Cas systems.
Science 357, 605–609 (2017).
44. Niewoehner, O. et al. Type III CRISPR−Cas systems
produce cyclic oligoadenylate second messengers.
Nature 548, 543–548 (2017).
45. Mohanraju, P. et al. Diverse evolutionary roots and
mechanistic variations of the CRISPR−Cas systems.
Science 353, aad5147 (2016).
46. Koonin, E. V., Makarova, K. S. & Zhang, F. Diversity,
classification and evolution of CRISPR−Cas systems.
Curr. Opin. Microbiol. 37, 67–78 (2017).
47. Sinkunas, T. et al. Cas3 is a single-stranded DNA
nuclease and ATP-dependent helicase in the CRISPR−
Cas immune system. EMBO J. 30, 1335–1342 (2011).
48. Tamulaitis, G., Venclovas, C. & Siksnys, V. Type III
CRISPR−Cas immunity: major differences brushed
aside. Trends Microbiol. 25, 49–61 (2017).
49. Samai, P. et al. Co‑transcriptional DNA and RNA
cleavage during type III CRISPR−Cas immunity. Cell
161, 1164–1174 (2015).
50. Liu, T. Y., Iavarone, A. T. & Doudna, J. A. RNA and
DNA targeting by a reconstituted thermus
thermophilus type III‑A CRISPR−Cas system. PLoS
ONE 12, e0170552 (2017).
51. Elmore, J. R. et al. Bipartite recognition of target
RNAs activates DNA cleavage by the Type IIIB
CRISPR−Cas system. Genes Dev. 30, 447–459
(2016).
52. Estrella, M. A., Kuo, F. T. & Bailey, S. RNA-activated
DNA cleavage by the Type IIIB CRISPR−Cas effector
complex. Genes Dev. 30, 460–470 (2016).
53. Makarova, K. S., Wolf, Y. I. & Koonin, E. V.
Comparative genomics of defense systems in archaea
and bacteria. Nucleic Acids Res. 41, 4360–4377
(2013).
54. Makarova, K. S. et al. Evolution and classification of
the CRISPR−Cas systems. Nat. Rev. Microbiol. 9,
467–477 (2011).
55. Chylinski, K., Makarova, K. S., Charpentier, E. &
Koonin, E. V. Classification and evolution of type II
CRISPRCas systems. Nucleic Acids Res. 42,
6091–6105 (2014).
56. Heler, R. et al. Cas9 specifies functional viral targets
during CRISPR−Cas adaptation. Nature 519,
199–202 (2015).
57. Aravind, L., Makarova, K. S. & Koonin, E. V. Survey
and summary: Holliday junction resolvases and related
nucleases: identification of new families, phyletic
distribution and evolutionary trajectories. Nucleic
Acids Res. 28, 3417–3432 (2000).
58. Zetsche, B. et al. Cpf1 is a single RNA-guided
endonuclease of a class 2 CRISPR−Cas system. Cell
163, 759–771 (2015).
59. Swarts, D. C., van der Oost, J. & Jinek, M. Structural
basis for guide RNA processing and seed-dependent
DNA targeting by CRISPR−Cas12a. Mol. Cell 66,
221–233.e4 (2017).
60. Fonfara, I., Richter, H., Bratovic, M., Le Rhun, A. &
Charpentier, E. The CRISPR-associated DNA-cleaving
enzyme Cpf1 also processes precursor CRISPR RNA.
Nature 532, 517–521 (2016).
This study shows the most streamlined CRISPR–
Cas system by demonstrating that Cpf1 is
independently capable of both crRNA processing
and DNA target interference.
61. Anantharaman, V., Makarova, K. S., Burroughs, A. M.,
Koonin, E. V. & Aravind, L. Comprehensive analysis of
the HEPN superfamily: identification of novel roles in
intra-genomic conflicts, defense, pathogenesis and
RNA processing. Biol. Direct 8, 15 (2013).
62. Smargon, A. A. et al. Cas13b is a type VIB CRISPR-
associated RNA-guided RNase differentially regulated
by accessory proteins Csx27 and Csx28. Mol. Cell 65,
618–630.e7 (2017).
63. Abudayyeh, O. O. et al. C2c2 is a single-component
programmable RNA-guided RNA-targeting CRISPR
effector. Science 353, aaf5573 (2016).
This study validates a novel class of CRISPR
interference proteins capable of programmable,
RNA-guided RNA cleavage and uncovers its
nonspecific RNase activity.
64. East-Seletsky, A. et al. Two distinct RNase activities of
CRISPR−C2c2 enable guide-RNA processing and RNA
detection. Nature 538, 270–273 (2016).
65. Liu, L. et al. Two distant catalytic sites are responsible
for C2c2 RNase activities. Cell 168, 121–134.e12
(2017).
66. Liu, L. et al. The molecular architecture for RNA-guided
RNA cleavage by Cas13a. Cell 170, 714–726.e10
(2017).
67. Karvelis, T. et al. crRNA and tracrRNA guide
Cas9‑mediated DNA interference in Streptococcus
thermophilus. RNA Biol. 10, 841–851 (2013).
68. Chylinski, K., Le Rhun, A. & Charpentier, E. The
tracrRNA and Cas9 families of type II CRISPR−Cas
immunity systems. RNA Biol. 10, 726–737 (2013).
69. Yang, H., Gao, P., Rajashankar, K. R. & Patel, D. J.
PAM-dependent target DNA recognition and cleavage
by C2c1 CRISPR–Cas endonuclease. Cell 167,
1814–1828.e12 (2016).
70. Burstein, D. et al. New CRISPR–Cas systems from
uncultivated microbes. Nature 542, 237–241 (2017).
71. Hsu, P. D. et al. DNA targeting specificity of RNA-
guided Cas9 nucleases. Nat. Biotechnol. 31,
827–832 (2013).

NATURE REVIEWS | CHEMISTRY VOLUME 1 | ARTICLE NUMBER 0078 | 13

©
2
0
1
7
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
.
REVIEWS

72. Dang, Y. et al. Optimizing sgRNA structure to improve
CRISPR−Cas9 knockout efficiency. Genome Biol. 16,
280 (2015).
73. Hendel, A. et al. Chemically modified guide RNAs
enhance CRISPR-Cas genome editing in human
primary cells. Nat. Biotechnol. 33, 985–989 (2015).
74. Fu, Y., Sander, J. D., Reyon, D., Cascio, V. M. &
Joung, J. K. Improving CRISPR−Cas nuclease
specificity using truncated guide RNAs. Nat.
Biotechnol. 32, 279–284 (2014).
75. Nissim, L., Perli, S. D., Fridkin, A., Perez-Pinera, P. &
Lu, T. K. Multiplexed and programmable regulation of
gene networks with an integrated RNA and CRISPR−
Cas toolkit in human cells. Mol. Cell 54, 698–710
(2014).
76. Doench, J. G. et al. Rational design of highly active
sgRNAs for CRISPR−Cas9‑mediated gene inactivation.
Nat. Biotechnol. 32, 1262–1267 (2014).
77. Konermann, S. et al. Genome-scale transcriptional
activation by an engineered CRISPR−Cas9 complex.
Nature 517, 583–588 (2015).
78. Nowak, C. M., Lawson, S., Zerez, M. & Bleris, L. Guide
RNA engineering for versatile Cas9 functionality.
Nucleic Acids Res. 44, 9555–9564 (2016).
79. Chen, B. et al. Dynamic imaging of genomic loci in
living human cells by an optimized CRISPR−Cas
system. Cell 155, 1479–1491 (2013).
80. Thyme, S. B., Akhmetova, L., Montague, T. G., Valen, E.
& Schier, A. F. Internal guide RNA interactions
interfere with Cas9‑mediated cleavage. Nat. Commun.
7, 11750 (2016).
81. Lim, Y. et al. Structural roles of guide RNAs in the
nuclease activity of Cas9 endonuclease. Nat. Commun.
7, 13350 (2016).
82. Ma, E., Harrington, L. B., O’Connell, M. R., Zhou, K. &
Doudna, J. A. Single-stranded DNA cleavage by
divergent CRISPR−Cas9 enzymes. Mol. Cell 60,
398–407 (2015).
83. Jiang, F., Zhou, K., Ma, L., Gressel, S. & Doudna, J. A.
Structural Biology. A Cas9‑guide RNA complex
preorganized for target DNA recognition. Science
348, 1477–1481 (2015).
84. Nishimasu, H. et al. Crystal structure of Cas9 in
complex with guide RNA and target DNA. Cell 156,
935–949 (2014).
85. Anders, C., Niewoehner, O., Duerst, A. & Jinek, M.
Structural basis of PAM-dependent target DNA
recognition by the Cas9 endonuclease. Nature 513,
569–573 (2014).
86. Jiang, F. et al. Structures of a CRISPR−Cas9 R‑loop
complex primed for DNA cleavage. Science 351,
867–871 (2016).
87. Briner, A. E. et al. Guide RNA functional modules
direct Cas9 activity and orthogonality. Mol. Cell 56,
333–339 (2014).
88. Wright, A. V. et al. Rational design of a split‑Cas9
enzyme complex. Proc. Natl Acad. Sci. USA 112,
2984–2989 (2015).
89. Sternberg, S. H., LaFrance, B., Kaplan, M. &
Doudna, J. A. Conformational control of DNA target
cleavage by CRISPR−Cas9. Nature 527, 110–113
(2015).
This study shows that the conformation of the HNH
nuclease domain within SpCas9 regulates RuvC
nuclease activity to ensure accurate and concerted
cleavage of both DNA strands.
90. Nishimasu, H. et al. Crystal structure of
Staphylococcus aureus Cas9. Cell 162, 1113–1126
(2015).
91. Hirano, H. et al. Structure and engineering of
Francisella novicida Cas9. Cell 164, 950–961 (2016).
92. Esvelt, K. M. et al. Orthogonal Cas9 proteins for RNA-
guided gene regulation and editing. Nat. Methods 10,
1116–1121 (2013).
93. Yamada, M. et al. Crystal structure of the minimal
Cas9 from Campylobacter jejuni reveals the molecular
diversity in the CRISPR‑Cas9 systems. Mol. Cell 65,
1109–1121.e3 (2017).
94. Liu, L. et al. C2c1−sgRNA complex structure reveals
RNA-guided DNA cleavage mechanism. Mol. Cell 65,
310–322 (2017).
95. Yang, H., Gao, P., Rajashankar, K. R. & Patel, D. J.
PAM-dependent target DNA recognition and cleavage
by C2c1 CRISPR−Cas endonuclease. Cell 167,
1814–1828.e12 (2016).
96. Dong, D. et al. The crystal structure of Cpf1 in
complex with CRISPR RNA. Nature 532, 522–526
(2016).
97. Gao, P., Yang, H., Rajashankar, K. R., Huang, Z. &
Patel, D. J. Type V CRISPR−Cas Cpf1 endonuclease
employs a unique mechanism for crRNA-mediated
target DNA recognition. Cell Res. 26, 901–913
(2016).
98. Yamano, T. et al. Crystal structure of Cpf1 in complex
with guide RNA and target DNA. Cell 165, 949–962
(2016).
99. Stella, S., Alcon, P. & Montoya, G. Structure of the
Cpf1 endonuclease R‑loop complex after target DNA
cleavage. Nature 546, 559–563 (2017).
100. Mojica, F. J., Diez-Villasenor, C., Garcia-Martinez, J. &
Almendros, C. Short motif sequences determine the
targets of the prokaryotic CRISPR defence system.
Microbiology 155, 733–740 (2009).
101. Shah, S. A., Erdmann, S., Mojica, F. J. & Garrett, R. A.
Protospacer recognition motifs: mixed identities and
functional diversity. RNA Biol. 10, 891–899 (2013).
102. Deveau, H. et al. Phage response to CRISPR-encoded
resistance in Streptococcus thermophilus. J. Bacteriol.
190, 1390–1400 (2008).
103. Sternberg, S. H., Redding, S., Jinek, M., Greene, E. C. &
Doudna, J. A. DNA interrogation by the CRISPR RNA-
guided endonuclease Cas9. Nature 507, 62–67 (2014).
104. Tsai, S. Q. et al. GUIDE-seq enables genome-wide
profiling of off-target cleavage by CRISPR−Cas
nucleases. Nat. Biotechnol. 33, 187–197 (2015).
105. Jiang, W., Bikard, D., Cox, D., Zhang, F. &
Marraffini, L. A. RNA-guided editing of bacterial
genomes using CRISPR−Cas systems. Nat. Biotechnol.
31, 233–239 (2013).
106. Westra, E. R. et al. CRISPR immunity relies on the
consecutive binding and degradation of negatively
supercoiled invader DNA by Cascade and Cas3. Mol.
Cell 46, 595–605 (2012).
107. Knight, S. C. et al. Dynamics of CRISPR−Cas9 genome
interrogation in living cells. Science 350, 823–826
(2015).
108. Singh, D., Sternberg, S. H., Fei, J., Doudna, J. A. &
Ha, T. Real-time observation of DNA recognition and
rejection by the RNA-guided endonuclease Cas9. Nat.
Commun. 7, 12778 (2016).
109. Hirano, S., Nishimasu, H., Ishitani, R. & Nureki, O.
Structural basis for the altered PAM specificities of
engineered CRISPR−Cas9. Mol. Cell 61, 886–894
(2016).
110. Kleinstiver, B. P. et al. Engineered CRISPR−Cas9
nucleases with altered PAM specificities. Nature 523,
481–485 (2015).
111. Mekler, V., Minakhin, L. & Severinov, K. Mechanism of
duplex DNA destabilization by RNA-guided Cas9
nuclease during target interrogation. Proc. Natl Acad.
Sci. USA 114, 5443–5448 (2017).
112. Gao, L. et al. Engineered Cpf1 variants with altered
PAM specificities. Nat. Biotechnol. 35, 789–792
(2017).
113. Nishimasu, H. et al. Structural basis for the altered
PAM recognition by engineered CRISPR−Cpf1. Mol.
Cell 67, 139–147.e2 (2017).
114. Yamano, T. et al. Structural basis for the canonical and
non-canonical PAM recognition by CRISPR−Cpf1. Mol.
Cell. 67, 633–645.e3 (2017).
115. Jore, M. M. et al. Structural basis for CRISPR RNA-
guided DNA recognition by Cascade. Nat. Struct. Mol.
Biol. 18, 529–536 (2011).
116. Szczelkun, M. D. et al. Direct observation of R‑loop
formation by single RNA-guided Cas9 and Cascade
effector complexes. Proc. Natl Acad. Sci. USA 111,
9798–9803 (2014).
117. Rutkauskas, M. et al. Directional R‑loop formation by
the CRISPR−Cas surveillance complex cascade
provides efficient off-target site rejection. Cell Rep. 10,
1534–1543 (2015).
118. Pattanayak, V. et al. High-throughput profiling of off-
target DNA cleavage reveals RNA-programmed Cas9
nuclease specificity. Nat. Biotechnol. 31, 839–843
(2013).
119. Wu, X. et al. Genome-wide binding of the CRISPR
endonuclease Cas9 in mammalian cells. Nat.
Biotechnol. 32, 670–676 (2014).
120. Zheng, T. et al. Profiling single-guide RNA specificity
reveals a mismatch sensitive core sequence. Sci. Rep.
7, 40638 (2017).
121. Coulombe, B. & Burton, Z. F. DNA bending and
wrapping around RNA polymerase: a “revolutionary”
model describing transcriptional mechanisms.
Microbiol. Mol. Biol. Rev. 63, 457–478 (1999).
122. Westra, E. R. et al. Cascade-mediated binding and
bending of negatively supercoiled DNA. RNA Biol. 9,
1134–1138 (2012).
123. Hochstrasser, M. L., Taylor, D. W., Kornfeld, J. E.,
Nogales, E. & Doudna, J. A. DNA targeting by a
minimal CRISPR RNA-guided cascade. Mol. Cell 63,
840–851 (2016).
124. Xiao, Y. et al. Structure basis for directional R‑loop
formation and substrate handover mechanisms in
type I CRISPR−Cas system. Cell 170, 48–60.e11
(2017).
125. Nelson, P. Transport of torsional stress in DNA.
Proc. Natl Acad. Sci. USA 96, 14342–14347
(1999).
126. Vassylyev, D. G. et al. Crystal structure of a bacterial
RNA polymerase holoenzyme at 2.6 Å resolution.
Nature 417, 712–719 (2002).
127. Josephs, E. A. et al. Structure and specificity of the
RNA-guided endonuclease Cas9 during DNA
interrogation, target binding and cleavage. Nucleic
Acids Res. 43, 8924–8941 (2015).
128. Farasat, I. & Salis, H. M. A biophysical model of
CRISPR−Cas9 activity for rational design of genome
editing and gene regulation. PLoS Comput. Biol. 12,
e1004724 (2016).
129. Cencic, R. et al. Protospacer adjacent motif (PAM)-
distal sequences engage CRISPR−Cas9 DNA target
cleavage. PLoS ONE 9, e109213 (2014).
130. Kleinstiver, B. P. et al. Genome-wide specificities of
CRISPR−Cas Cpf1 nucleases in human cells. Nat.
Biotechnol. 34, 869–874 (2016).
131. Kim, H. K. et al. In vivo high-throughput profiling of
CRISPR−Cpf1 activity. Nat. Methods 14, 153–159
(2017).
132. Jiang, F. & Doudna, J. A. CRISPR−Cas9 structures and
mechanisms. Annu. Rev. Biophys. 46, 505–529
(2017).
133. Yang, W., Lee, J. Y. & Nowotny, M. Making and
breaking nucleic acids: two‑Mg2+-ion catalysis and
substrate specificity. Mol. Cell 22, 5–13 (2006).
134. Yang, W. An equivalent metal ion in one- and two-
metal-ion catalysis. Nat. Struct. Mol. Biol. 15,
1228–1231 (2008).
135. Yang, W. Nucleases: diversity of structure, function
and mechanism. Q. Rev. Biophys. 44, 1–93
(2011).
136. Steitz, T. A. & Steitz, J. A. A general two-metal-ion
mechanism for catalytic RNA. Proc. Natl Acad. Sci.
USA 90, 6498–6502 (1993).
137. Jinek, M. et al. Structures of Cas9 endonucleases
reveal RNA-mediated conformational activation.
Science 343, 1247997 (2014).
138. Biertumpfel, C., Yang, W. & Suck, D. Crystal structure
of T4 endonuclease VII resolving a Holliday junction.
Nature 449, 616–620 (2007).
139. Li, C. L. et al. DNA binding and cleavage by the
periplasmic nuclease Vvn: a novel structure with a
known active site. EMBO J. 22, 4014–4025
(2003).
140. Dagdas, Y. S., Chen, J. S., Sternberg, S. H.,
Doudna, J. A. & Yildiz, A. A conformational checkpoint
between DNA binding and cleavage by CRISPR‑Cas9.
Sci. Adv. 3, eaao0027 (2017).
141. Osuka, S. et al. Real-time observation of flexible
domain movements in Cas9. Preprint at http://www.
biorxiv.org/content/early/2017/03/29/122069
(2017).
142. Ariyoshi, M. et al. Atomic structure of the RuvC
resolvase: a Holliday junction-specific endonuclease
from E. coli. Cell 78, 1063–1072 (1994).
143. Chen, L., Shi, K., Yin, Z. & Aihara, H. Structural
asymmetry in the Thermus thermophilus RuvC dimer
suggests a basis for sequential strand cleavages
during Holliday junction resolution. Nucleic Acids Res.
41, 648–656 (2013).
144. Gorecka, K. M., Komorowska, W. & Nowotny, M.
Crystal structure of RuvC resolvase in complex with
Holliday junction substrate. Nucleic Acids Res. 41,
9945–9955 (2013).
145. Palermo, G., Miao, Y., Walker, R. C., Jinek, M. &
McCammon, J. A. Striking plasticity of CRISPR−Cas9
and key role of non-target DNA, as revealed by
molecular simulations. ACS Cent. Sci. 2, 756–763
(2016).
146. Kim, D. et al. Genome-wide analysis reveals
specificities of Cpf1 endonucleases in human cells.
Nat. Biotechnol. 34, 863–868 (2016).
147. Ran, F. A. et al. In vivo genome editing using
Staphylococcus aureus Cas9. Nature 520, 186–191
(2015).
148. Chen, J. S. et al. Enhanced proofreading governs
CRISPR‑Cas9 targeting accuracy. Nature http://dx.doi.
org/10.1038/nature24268 (2017).
149. Palermo, G., Miao, Y., Walker, R. C., Jinek, M. &
McCammon, J. A. CRISPR‑Cas9 conformational
activation as elucidated from enhanced molecular
simulations. Proc. Natl Acad. Sci. USA 114,
7260–7265 (2017).

14 | ARTICLE NUMBER 0078 | VOLUME 1 www.nature.com/natrevchem

©
2
0
1
7
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
.
 REVIEWS

150. Palermo, G. et al. Protospacer adjacent motif-induced
allostery activates CRISPR‑Cas9. J. Am. Chem. Soc.
http://dx.doi.org/10.1021/jacs.7b05313 (2017).
151. Frock, R. L. et al. Genome-wide detection of DNA
double-stranded breaks induced by engineered
nucleases. Nat. Biotechnol. 33, 179–186 (2015).
152. Wang, X. et al. Unbiased detection of off-target
cleavage by CRISPR−Cas9 and TALENs using integrase-
defective lentiviral vectors. Nat. Biotechnol. 33,
175–178 (2015).
153. Kim, D. et al. Digenome-seq: genome-wide profiling of
CRISPR−Cas9 off-target effects in human cells. Nat.
Methods 12, 237–243 (2015).
154. O’Geen, H., Henry, I. M., Bhakta, M. S., Meckler, J. F. &
Segal, D. J. A genome-wide analysis of Cas9 binding
specificity using ChIP-seq and targeted sequence
capture. Nucleic Acids Res. 43,
3389–3404 (2015).
155. Horlbeck, M. A. et al. Nucleosomes impede Cas9 access
to DNA in vivo and in vitro. eLife 5, e12677 (2016).
156. Kim, S., Kim, D., Cho, S. W., Kim, J. & Kim, J. S.
Highly efficient RNA-guided genome editing in human
cells via delivery of purified Cas9 ribonucleoproteins.
Genome Res. 24, 1012–1019 (2014).
157. Lin, S., Staahl, B. T., Alla, R. K. & Doudna, J. A.
Enhanced homology-directed human genome
engineering by controlled timing of CRISPR−Cas9
delivery. eLife 3, e04766 (2014).
158. Cho, S. W. et al. Analysis of off-target effects of CRISPR−
Cas-derived RNA-guided endonucleases and nickases.
Genome Res. 24, 132–141 (2014).
159. Guilinger, J. P., Thompson, D. B. & Liu, D. R. Fusion of
catalytically inactive Cas9 to FokI nuclease improves the
specificity of genome modification. Nat. Biotechnol. 32,
577–582 (2014).
160. Ran, F. A. et al. Double nicking by RNA-guided CRISPR−
Cas9 for enhanced genome editing specificity. Cell 154,
1380–1389 (2013).
161. Mali, P. et al. CAS9 transcriptional activators for target
specificity screening and paired nickases for cooperative
genome engineering. Nat. Biotechnol. 31, 833–838
(2013).
162. Tsai, S. Q. et al. Dimeric CRISPR RNA-guided FokI
nucleases for highly specific genome editing. Nat.
Biotechnol. 32, 569–576 (2014).
163. Slaymaker, I. M. et al. Rationally engineered Cas9
nucleases with improved specificity. Science 351,
84–88 (2016).
164. Kleinstiver, B. P. et al. High-fidelity CRISPR−Cas9
nucleases with no detectable genome-wide off-target
effects. Nature 529, 490–495 (2016).
165. Tsai, S. Q. & Joung, J. K. Defining and improving the
genome-wide specificities of CRISPR−Cas9 nucleases.
Nat. Rev. Genet. 17, 300–312 (2016).
166. Tycko, J., Myer, V. E. & Hsu, P. D. Methods for
optimizing CRISPR−Cas9 genome editing specificity.
Mol. Cell 63, 355–370 (2016).
167. Yee, J. K. Off-target effects of engineered nucleases.
FEBS J. 283, 3239–3248 (2016).
168. Bisaria, N., Jarmoskaite, I. & Herschlag, D. Lessons
from enzyme kinetics reveal specificity principles for
RNA-guided nucleases in RNA interference and
CRISPR-based genome editing. Cell Syst. 4, 21–29
(2017).
169. Gootenberg, J. S. et al. Nucleic acid detection with
CRISPR−Cas13a/C2c2. Science 356, 438–442 (2017).
170. East-Seletsky, A., O’Connell, M. R., Burstein, D.,
Knott, G. J. & Doudna, J. A. RNA targeting by
functionally orthogonal type VIA CRISPR−Cas enzymes.
Mol. Cell 66, 373–383.e3 (2017).
171. Friedland, A. E. et al. Characterization of
Staphylococcus aureus Cas9: a smaller Cas9 for
all‑in‑one adeno-associated virus delivery and
paired nickase applications. Genome Biol. 16, 257
(2015).
172. Kim, E. et al. In vivo genome editing with a small Cas9
orthologue derived from Campylobacter jejuni. Nat.
Commun. 8, 14500 (2017).
173. Hur, J. K. et al. Targeted mutagenesis in mice by
electroporation of Cpf1 ribonucleoproteins. Nat.
Biotechnol. 34, 807–808 (2016).
174. Kim, Y. et al. Generation of knockout mice by
Cpf1‑mediated gene targeting. Nat. Biotechnol. 34,
808–810 (2016).
175. Tu, M. et al. A ‘new lease of life’: FnCpf1 possesses
DNA cleavage activity for genome editing in human
cells. Nucleic Acids Res. https://doi.org/10.1093/nar/
gkx783 (2017).
176. Knott, G. J. et al. Guide-bound structures of an RNA-
targeting A-cleaving CRISPR-Cas13a enzyme. Nat.
Struct. Mol. Biol. https://doi.org/10.1038/nsmb.3466
(2017).
Acknowledgements
The authors thank M. L. Hochstrasser, L. B. Harrington and
A. V. Wright for critical reading and valuable input on the
manuscript. J.S.C. is a National Science Foundation
Graduate Research Fellow and J.A.D. is a Howard Hughes
Medical Institute Investigator.
Author contributions
The authors contributed equally to this manuscript.
Competing interests statement
J.A.D. is a co‑founder of Caribou Biosciences, Editas Medicine
and Intellia Therapeutics; a scientific adviser to Caribou
Biosciences, Intellia Therapeutics, eFFECTOR Therapeutics
and Driver; and executive director of the Innovative Genomics
Institute at the University of California Berkeley (UC Berkeley)
and the University of California San Francisco (UCSF). J.S.C.
and J.A.D. are inventors on UC Berkeley and Howard Hughes
Medical Institute patents for clustered regularly interspaced
short palindromic repeats (CRISPR) technologies.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
How to cite this article
Chen, J. S. & Doudna, J. A. The chemistry of Cas9 and its
CRISPR colleagues. Nat. Rev. Chem. 1, 0078 (2017).
DATABASES
Protein Data Bank: http://www.rcsb.org/pdb/home/home.do

NATURE REVIEWS | CHEMISTRY VOLUME 1 | ARTICLE NUMBER 0078 | 15

©
2
0
1
7
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
.