Subheadi

ngst
ates
concl
usi
on

Sectioncont
entcor
re-
spondstoFig.1
(above)
 Ifyouthinkyourdat ar eal
lyprove
somethingdefini
tively,andit’
si mpor-
tantf
oryourr eadert oknow it,you
canusesubj ecti
vet ermslike“ cl
earl
y.”

Concl
usi
on

Transi
ti
on+
rat
ional
e+
methods
Fi
ndi
ngs

Concl
usi
on

Sect i
oncontent
correspondsto
Fig.2( andso
onf orfol
lowi
ng
sections)

Experimental
rat
ional
e+
methodsdescri
p-
ti
onwi t
houttoo
muchdetail

Quickdescript
ion
ofhow assaywas
inter
preted

Statement
offi
ndings

I
nthi
spaper,compari
sonwit
hprevi
ouslydiscover
edsyst
ems
i
simpor
tantf
orestabl
ishi
ngwhatt
hefindi
ngsmean.
A Figure 2. Heterologous Expression of
FnCpf1 and CRISPR Array in E. coli Is Suffi-
cient to Mediate Plasmid DNA Interference
and crRNA Maturation
(A) Small RNA-seq of Francisella novicida U112
reveals transcription and processing of the FnCpf1
CRISPR array. The mature crRNA begins with a
19-nt partial direct repeat followed by 23–25 nt of
spacer sequence.
(B) Small RNA-seq of E. coli transformed with a
plasmid-carrying synthetic promoter-driven
FnCpf1 and CRISPR array shows crRNA pro-
cessing independent of Cas genes and other
sequence elements in the FnCpf1 locus.
(C) E. coli harboring different truncations of the
FnCpf1 CRISPR locus shows that only FnCpf1 and
the CRISPR array are required for plasmid DNA
interference (n = 3; error bars show mean ± SEM).

we also found that FnCpf1 requires the 50 -

TTN PAM to be in a duplex form in order
to cleave the target DNA (Figure 3E).

The RuvC-like Domain of Cpf1

Mediates RNA-Guided DNA
C
Cleavage
The RuvC-like domain of Cpf1 retains all
of the catalytic residues of this family of
endonucleases (Figures 4A and S4) and
is thus predicted to be an active nuclease.
Therefore, we generated three mutants—
FnCpf1(D917A), FnCpf1(E1006A), and
FnCpf1(D1225A) (Figure 4A)—to test
whether the conserved catalytic residues
are essential for the nuclease activity of
FnCpf1. We found that the D917A and
E1006A mutations completely inactivated
the DNA cleavage activity of FnCpf1, and
D1255A significantly reduced nucleolytic
FnCpf1 (Figure S2) and assayed its ability to cleave the same activity (Figure 4B). These results are in contrast to the mutagen-
protospacer-1-containing plasmid used in the bacterial DNA esis results for Streptococcus pyogenes Cas9 (SpCas9), where
interference experiments (Figure 3A). We found that FnCpf1 mutation of the RuvC (D10A) and HNH (N863A) nuclease do-
along with an in-vitro-transcribed mature crRNA-targeting proto- mains converts SpCas9 into a DNA nickase (i.e., inactivation of
spacer 1 was able to efficiently cleave the target plasmid in a each of the two nuclease domains abolished the cleavage of
Mg2+- and crRNA-dependent manner (Figure 3B). Moreover, one of the DNA strands) (Jinek et al., 2012; Gasiunas et al.,
FnCpf1 was able to cleave both supercoiled and linear target 2012) (Figure 4B). These findings suggest that the RuvC-like
DNA (Figure 3C). These results clearly demonstrate the suffi- domain of FnCpf1 cleaves both strands of the target DNA,
ciency of FnCpf1 and crRNA for RNA-guided DNA cleavage. perhaps in a dimeric configuration. Interestingly, size-exclusion
We also mapped the cleavage site of FnCpf1 using Sanger gel filtration of FnCpf1 shows that the protein is eluted at a size
sequencing of the cleaved DNA ends. We found that FnCpf1- of
300 kD, twice the molecular weight of a FnCpf1 monomer
mediated cleavage results in a 5-nt 50 overhang (Figures 3A, (Figure S2B).
3D, and S3A–S3D), which is different from the blunt cleavage
product generated by Cas9 (Garneau et al., 2010; Jinek et al., Sequence and Structural Requirements for the
2012; Gasiunas et al., 2012). The staggered cleavage site of Cpf1 crRNA
FnCpf1 is distant from the PAM: cleavage occurs after the 18th Compared with the guide RNA for Cas9, which has elaborate
base on the non-targeted (+) strand and after the 23rd base on RNA secondary structure features that interact with Cas9 (Nish-
the targeted (–) strand (Figures 3A, 3D, and S3A–S3D). Using imasu et al., 2014), the guide RNA for FnCpf1 is notably simpler
double-stranded oligo substrates with different PAM sequences, and only consists of a single stem loop in the direct repeat

762 Cell 163, 759–771, October 22, 2015 ª2015 Elsevier Inc.
A Figure 3. FnCpf1 Is Guided by crRNA to
Cleave DNA In Vitro
(A) Schematic of the FnCpf1 crRNA-DNA-targeting
complex. Cleavage sites are indicated by red
arrows.
(B) FnCpf1 and crRNA alone mediated RNA-
guided cleavage of target DNA in a crRNA- and
Mg2+-dependent manner.
B C D (C) FnCpf1 cleaves both linear and supercoiled
DNA.
(D) Sanger-sequencing traces from FnCpf1-
digested target show staggered overhangs. The
non-templated addition of an additional adenine,
denoted as N, is an artifact of the polymerase used
in sequencing (Clark, 1988). Reverse primer read
represented as reverse complement to aid visual-
ization. See also Figure S3.
(E) Dependency of cleavage on base-pairing at the
50 PAM. FnCpf1 can only recognize the PAM in
E correctly Watson-Crick-paired DNA.
See also Figures S2 and S3.

Therefore, in order to assess the feasi-

bility of harnessing Cpf1 as a genome-
editing tool, we exploited the diversity of
Cpf1-family proteins available in the pub-
sequence (Figure 3A). We explored the sequence and structural lic sequences databases. A BLAST search of the WGS database
requirements of crRNA for mediating DNA cleavage with FnCpf1. at the NCBI revealed 46 non-redundant Cpf1-family proteins
We first examined the length requirement for the guide (Figure S5A), from which we chose 16 candidates that, based
sequence and found that FnCpf1 requires at least 16 nt of guide on our phylogenetic reconstruction (Figure S5A), represented
sequence to achieve detectable DNA cleavage and a minimum the entire Cpf1 diversity (Figures 6A and S5). These Cpf1-family
of 18 nt of guide sequence to achieve efficient DNA cleavage proteins span a range of lengths between
1,200 and
1,500
in vitro (Figure 5A). These requirements are similar to those amino acids.
demonstrated for SpCas9, in which a minimum of 16–17 nt of The direct repeat sequences for each of these Cpf1-family
spacer sequence is required for DNA cleavage (Cencic et al., proteins show strong conservation in the 19 nt at the 30 of the
2014; Fu et al., 2014). We also found that the seed region of direct repeat, the portion of the repeat that is included in the
the FnCpf1 guide RNA is approximately within the first 5 nt on processed crRNA (Figure 6B). The 50 sequence of the direct
the 50 end of the spacer sequence (Figures 5B and S3E). repeat is much more diverse. Of the 16 Cpf1-family proteins cho-
Next, we studied the effect of direct repeat mutations on the sen for analysis, three (2, Lachnospiraceae bacterium MC2017,
RNA-guided DNA cleavage activity. The direct repeat portion Lb3Cpf1; 3, Butyrivibrio proteoclasticus, BpCpf1; and 6,
of mature crRNA is 19 nt long (Figure 2A). Truncation of the direct Smithella sp. SC_K08D17, SsCpf1) were associated with direct
repeat revealed that at least 16, but optimally more than 17 nt, of repeat sequences that are notably divergent from the FnCpf1
the direct repeat is required for cleavage. Mutations in the stem direct repeat (Figure 6B). However, even these direct repeat
loop that preserved the RNA duplex did not affect the cleavage sequences preserved stem-loop structures that were identical
activity, whereas mutations that disrupted the stem loop duplex or nearly identical to the FnCpf1 direct repeat (Figure 6C).
structure completely abolished cleavage (Figure 5D). Finally, Given the strong structural conservation of the direct repeats
base substitutions in the loop region did not affect nuclease that are associated with many of the Cpf1-family proteins, we
activity, whereas the uracil base immediately proceeding the first tested whether the orthologous direct repeat sequences
spacer sequence could not be substituted (Figure 5E). Collec- are able to support FnCpf1 nuclease activity in vitro. As ex-
tively, these results suggest that FnCpf1 recognizes the crRNA pected, the direct repeats that contained conserved stem
through a combination of sequence-specific and structural fea- sequences were able to function interchangeably with FnCpf1.
tures of the stem loop. By contrast, the direct repeats from candidates 2 (Lb3Cpf1)
and 6 (SsCpf1) were unable to support FnCpf1 cleavage activity
Cpf1-Family Proteins from Diverse Bacteria Share (Figure 6D). The direct repeat from candidate 3 (BpCpf1) sup-
Common crRNA Structures and PAMs ported only a low level of FnCpf1 nuclease activity (Figure 6D),
Based on our previous experience in harnessing Cas9 for possibly due to the conservation of the 30 -most U.
genome editing in mammalian cells, only a small fraction of bac- Next, we applied the in vitro PAM identification assay (Fig-
terial nucleases can function efficiently when heterologously ex- ure S6A) to determine the PAM sequence for each Cpf1-family
pressed in mammalian cells (Cong et al., 2013; Ran et al., 2015). protein. We were able to identify the PAM sequence for seven

Cell 163, 759–771, October 22, 2015 ª2015 Elsevier Inc. 763
 A FnCpf1 helical zinc finger- signed to target DNMT1 was able to cleave a PCR amplicon
region like domain of the DNMT1 genomic region in vitro (Figure 7C). However,
RuvC I RuvC II III when tested in human embryonic kidney 293FT (HEK293FT)
D917 E1006 D1255
cells, only two out of the eight Cpf1-family proteins (7, AsCpf1
and 13, LbCpf1) exhibited detectable levels of nuclease-induced
catalytic residues indels (Figures 7C and 7D). This result is consistent with previous
B experiments with Cas9 in which only a small number of Cas9
FnCpf1 SpCas9 orthologs were successfully harnessed for genome editing in
mammalian cells (Ran et al., 2015).
n

We further tested each Cpf1-family protein with additional

ei

A
A
7A
ot

55
06

A
pr

10
91

12
T

T
0
genomic targets and found that AsCpf1 and LbCpf1 consistently
E1
W

W
no

D
D

D
mediated robust genome editing in HEK293FT cells, whereas the
606 bp remaining Cpf1 proteins showed either no detectable activity or
native TBE PAGE

only sporadic activity (Figures 7E and S7) despite robust expres-

424 bp sion (Figure S6D). The only Cpf1 candidate that expressed
poorly was PdCpf1 (Figure S6D). When compared to Cas9,
AsCpf1 and LbCpf1 mediated comparable levels of indel forma-
tion (Figure 7E). Additionally, we used in vitro cleavage followed
by Sanger sequencing of the cleaved DNA ends and found that
182 bp
7, AsCpf1 and 13, LbCpf1 also generated staggered cleavage
sites (Figures S6E and S6F, respectively).
606 nt
denaturing TBE-Urea

424 nt DISCUSSION

In this work, we characterize Cpf1-containing class 2 CRISPR

systems, classified as type V, and show that its effector protein,
182 nt
Cpf1, is a single RNA-guided endonuclease. Cpf1 substantially
differs from Cas9—to date, the only other experimentally charac-
terized class 2 effector—in terms of structure and function and
Figure 4. Catalytic Residues in the C-Terminal RuvC Domain of might provide important advantages for genome-editing appli-
FnCpf1 Are Required for DNA Cleavage cations. Specifically, Cpf1 contains a single identified nuclease
(A) Domain structure of FnCpf1 with RuvC catalytic residues highlighted. The
domain, in contrast to the two nuclease domains present in
catalytic residues were identified based on sequence homology to Thermus
thermophilus RuvC (PDB: 4EP5). Cas9. The results presented here show that, in FnCpf1, inactiva-
(B) Native TBE PAGE gel showing that mutation of the RuvC catalytic residues tion of RuvC-like domain abolishes cleavage of both DNA
of FnCpf1 (D917A and E1006A) and mutation of the RuvC (D10A) catalytic strands. Conceivably, FnCpf1 forms a homodimer (Figure S2B),
residue of SpCas9 prevents double-stranded DNA cleavage. Denaturing TBE- with the RuvC-like domains of each of the two subunits cleaving
Urea PAGE gel showing that mutation of the RuvC catalytic residues of FnCpf1 one DNA strand. However, we cannot rule out that FnCpf1 con-
(D917A and E1006A) prevents DNA-nicking activity, whereas mutation of the
tains a second yet-to-be-identified nuclease domain. Structural
RuvC (D10A) catalytic residue of SpCas9 results in nicking of the target site.
See also Figure S4.
characterization of Cpf1-RNA-DNA complexes will allow testing
of these hypotheses and elucidation of the cleavage mechanism.
Perhaps the most notable feature of Cpf1 is that it is a single
new Cpf1-family proteins (Figures 6E, S6B, and S6C), and the crRNA-guided endonuclease. Unlike Cas9, which requires
screen confirmed the PAM for FnCpf1 as 50 -TTN. The remaining tracrRNA to process crRNA arrays and both crRNA and
eight tested Cpf1 proteins did not show efficient cleavage during tracrRNA to mediate interference (Deltcheva et al., 2011), Cpf1
in vitro reconstitution. The PAM sequences for the Cpf1-family processes crRNA arrays independent of tracrRNA, and Cpf1-
proteins were predominantly T rich, only varying in the number crRNA complexes alone cleave target DNA molecules, without
of Ts constituting each PAM (Figures 6E, S6B, and S6C). the requirement for any additional RNA species. This feature
could simplify the design and delivery of genome-editing tools.
Cpf1 Can Be Harnessed to Facilitate Genome Editing in For example, the shorter (
42 nt) crRNA employed by Cpf1
Human Cells has practical advantages over the long (
100 nt) guide RNA in
We tested each Cpf1-family protein for which we were able to Cas9-based systems because shorter RNA oligos are signifi-
identify a PAM for nuclease activity in mammalian cells. We cantly easier and cheaper to synthesize. In addition, these find-
codon optimized each of these genes and attached a C-terminal ings raise more fundamental questions regarding the guide
nuclear localization signal (NLS) for optimal expression and nu- processing mechanism of the type V CRISPR-Cas systems. In
clear targeting in human cells (Figure 7A). To test the activity of the case of type II, processing of the pre-crRNA is catalyzed
each Cpf1-family protein, we selected a guide RNA target site by the bacterial RNase III, which recognizes the long duplex
within the DNMT1 gene (Figure 7B). We first found that each of formed by the tracrRNA and the complementary portion of the
the Cpf1-family proteins along with its respective crRNA de- direct repeat (Deltcheva et al., 2011). Such long duplexes

764 Cell 163, 759–771, October 22, 2015 ª2015 Elsevier Inc. All rights reserved. Reproduced here for educational purposes only.