Conditional Guide RNAs: Programmable Conditional Regulation of
CRISPR/Cas Function in Bacterial and Mammalian Cells via Dynamic
RNA Nanotechnology
Mikhail H. Hanewich-Hollatz,
†
,
∥
Zhewei Chen,
†
,
∥
Lisa M. Hochrein,
†
Jining Huang,
†
and Niles A. Pierce
*
,
†
,
‡
,
§
†
Division of Biology & Biological Engineering, California Institute of Technology, Pasadena, California 91125, United States
‡
Division of Engineering & Applied Science, California Institute of Technology, Pasadena, California 91125, United States
§
Weatherall Institute of Molecular Medicine, University of Oxford, Oxford OX3 9DS, United Kingdom
*
S
Supporting Information
ABSTRACT:
A guide RNA (gRNA) directs the function of a CRISPR protein e
ff
ector to a
target gene of choice, providing a versatile programmable platform for engineering diverse
modes of synthetic regulation (edit, silence, induce, bind). However, the fact that gRNAs are
constitutively active places limitations on the ability to con
fi
ne gRNA activity to a desired
location and time. To achieve programmable control over the scope of gRNA activity, here
we apply principles from dynamic RNA nanotechnology to engineer conditional guide RNAs
(cgRNAs) whose activity is dependent on the presence or absence of an RNA trigger. These
cgRNAs are programmable at two levels, with the trigger-binding sequence controlling the
scope of the e
ff
ector activity and the target-binding sequence determining the subject of the
e
ff
ector activity. We demonstrate molecular mechanisms for both constitutively active
cgRNAs that are conditionally inactivated by an RNA trigger (ON
→
OFF logic) and
constitutively inactive cgRNAs that are conditionally activated by an RNA trigger (OFF
→
ON logic). For each mechanism, automated sequence design is performed using the reaction
pathway designer within NUPACK to design an orthogonal library of three cgRNAs that respond to di
ff
erent RNA triggers. In
E. coli
expressing cgRNAs, triggers, and silencing dCas9 as the protein e
ff
ector, we observe a median conditional response of
≈
4-fold for an ON
→
OFF
“
terminator switch
”
mechanism,
≈
15-fold for an ON
→
OFF
“
splinted switch
”
mechanism, and
≈
3-
fold for an OFF
→
ON
“
toehold switch
”
mechanism; the median crosstalk within each cgRNA/trigger library is <2%,
≈
2%, and
≈
20% for the three mechanisms. To test the portability of cgRNA mechanisms prototyped in bacteria to mammalian cells, as
well as to test generalizability to di
ff
erent e
ff
ector functions, we implemented the terminator switch in HEK 293T cells
expressing inducing dCas9 as the protein e
ff
ector, observing a median ON
→
OFF conditional response of
≈
4-fold with median
crosstalk of
≈
30% for three orthogonal cgRNA/trigger pairs. By providing programmable control over both the scope and target
of protein e
ff
ector function, cgRNA regulators o
ff
er a promising platform for synthetic biology.
■
INTRODUCTION
Dynamic RNA nanotechnology holds great promise as a
paradigm for introducing synthetic regulatory links into living
cells and organisms. We envision small conditional RNAs
(scRNAs) that, upon detection of a programmable nucleic acid
input, change conformation to produce a programmable
output that up-regulates or down-regulates the activity of a
biological pathway. In this scenario, the input controls the
scope of regulation, and the output controls the target of
regulation, with the scRNA performing signal transduction to
create a logical link between the two.
1
,
2
Any pathway that
recognizes RNA is a potential candidate for conditional
regulation by scRNAs (e.g., RNA interference, RNase H,
PKR, RIG-1); the CRISPR/Cas pathway is a particularly
attractive candidate because of its functional versatility, high
regulatory dynamic range, and portability between species.
3
−
5
The repurposing of RNA-guided CRISPR e
ff
ectors through
development of modi
fi
ed guide RNAs (gRNAs) and CRISPR-
associated (Cas) proteins has yielded a suite of powerful tools
for biological research and synthetic biology. Precision genome
editing has been achieved in a variety of organisms using
gRNAs to direct the nuclease activity of Cas9 and Cas12a
(Cpf1) to a target gene of choice.
3
,
6
−
8
Mutation of the
nuclease domains to produce a catalytically dead Cas9 (dCas9)
has enabled silencing of genetic expression via inhibition of
transcriptional elongation,
4
,
9
or induction (or silencing) of
genetic expression using dCas9 fusions that incorporate
transcriptional regulatory domains.
5
Other dCas9 fusions
have mediated target-binding to enable visualization of
genomic loci,
10
,
11
epigenetic modi
fi
cation,
12
and single-base
editing at a speci
fi
c genomic locus.
3
,
13
Hence, gRNA:e
ff
ector
complexes combine the bene
fi
ts of the rich functional
vocabulary of the protein e
ff
ector (edit, silence, induce,
Received:
April 3, 2019
Published:
June 4, 2019
Research Article
http://pubs.acs.org/journal/acscii
Cite This:
ACS Cent. Sci.
2019, 5, 1241
−
1249
© 2019 American Chemical Society
1241
DOI:
10.1021/acscentsci.9b00340
ACS Cent. Sci.
2019, 5, 1241
−
1249
This is an open access article published under an ACS AuthorChoice License, which permits
copying and redistribution of the article or any adaptations for non-commercial purposes.
Downloaded via CALIFORNIA INST OF TECHNOLOGY on July 26, 2019 at 15:22:23 (UTC).
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
bind) and the programmability of the gRNA in targeting
e
ff
ector activity to a gene of choice.
Because gRNAs are constitutively active, additional meas-
ures are needed to restrict e
ff
ector activity to a desired location
and time. Temporal control can be achieved by small-molecule
induction of gRNAs
14
,
15
or Cas9,
16
but this comes with
limitations in terms of multiplexing and spatial control.
Spatiotemporal control has been achieved by regulation of
Cas9 via photoactivation
17
or via tissue-speci
fi
c promoters
18
,
19
or microRNAs,
20
which comes with the unwelcome restriction
that all gRNAs are subject to the same regulatory scope.
Systematic mapping of the structure and sequence properties
of functional gRNAs has revealed that Cas9 activity is tolerant
to signi
fi
cant modi
fi
cations to the standard gRNA struc-
ture,
21
,
22
facilitating introduction of auxiliary domains that
enable conditional control of gRNA activity via structural
changes induced by small-molecules,
23
−
25
protein-bound
RNAs,
26
nucleases,
27
or nuclease-recruiting DNAs.
27
Alter-
natively, the activity of standard gRNAs has been modulated
by antisense RNAs
28
or by photolysis of antisense DNAs
incorporating photocleavable groups.
29
For generality, it is
highly desirable to control the regulatory scope in a manner
that is both conditional and programmable, a tantalizing
prospect central to the proposed scRNA paradigm based on
dynamic RNA nanotechnology.
With this paradigm in mind, we set out to engineer
conditional guide RNAs (cgRNAs) that change conformation
in response to an RNA trigger X to conditionally direct the
function of dCas9 to a target gene Y. Unlike a standard gRNA,
a cgRNA is programmable at two levels, with the trigger-
binding sequence controlling the scope of cgRNA activity and
the target-binding sequence determining the subject of e
ff
ector
activity. Functionally, the cgRNA must perform sequence
transduction between X and Y as well as shape transduction
between active/inactive conformations. In principle, cgRNA
activity can be engineered to toggle either OFF
→
ON (as was
recently demonstrated by Siu and Chen
30
)orON
→
OFF in
response to a cognate RNA trigger X; this conditional control
can be exerted over dCas9 variants that either edit, silence,
induce, or bind the target Y, emphasizing the broad functional
potential available via interplay between cgRNA logic and
protein e
ff
ector function (
Figure 1
a). For example, by selecting
an endogenous transcript X with a desired spatiotemporal
expression pro
fi
le during development, the downstream
regulatory e
ff
ect on target Y could be restricted to a desired
tissue and developmental stage within a model organism
(
Figure 1
b). Alternatively, in a therapeutic context, X could be
a disease marker and Y an independent therapeutic target,
enabling selective treatment of diseased cells leaving healthy
cells untouched.
■
RESULTS AND DISCUSSION
Constitutively Active Terminator Switch cgRNAs (ON
→
OFF Logic) with Silencing dCas9 in Bacteria.
As a
starting point, consider the constitutively active
“
terminator
switch
”
cgRNA mechanism of
Figure 2
b that is conditionally
inactivated by RNA trigger X (ON
→
OFF logic). Compared
to a standard gRNA (
Figure 1
c), the cgRNA has a modi
fi
ed
terminator region with an extended loop and rationally
designed sequence domains
“
d
−
e
−
f
”
. Hybridization of the
RNA trigger X to these modi
fi
ed domains is intended to form a
structure incompatible with cgRNA mediation of dCas9
function. We validated the cgRNA mechanism
in vivo
in
E.
coli
expressing silencing dCas9
4
as the protein e
ff
ector and a
fl
uorescent protein reporter (mRFP) as the target gene Y
(conditional logic:
“
if not X then not Y
”
;
Figure 2
a). An
E. coli
strain expressing the cgRNA exhibits low
fl
uorescence (ON
state) while a strain expressing both the cgRNA and the
cognate RNA trigger exhibit high
fl
uorescence (OFF state),
achieving a conditional ON
→
OFF response (
Figure 2
c).
Figure 1.
Programmable regulators. (a) A conditional guide RNA
(cgRNA) changes conformation in response to a programmable
trigger X to conditionally direct the activity of a protein e
ff
ector to a
programmable target Y. Top: a constitutively active cgRNA is
conditionally inactivated by X (ON
→
OFF logic). Bottom: a
constitutively inactive cgRNA is conditionally activated by X (OFF
→
ON logic). (b) Molecular logic of programmable regulation using a
standard gRNA (
“
not Y
”
) vs programmable conditional regulation
using a cgRNA (
“
if X then not Y
”
). In this conceptual illustration, the
standard gRNA silences Y in all tissues, while the cgRNA silences Y
only in tissues where and when X is expressed, exerting
spatiotemporal control over regulation. (c) A standard guide RNA
(gRNA) is constitutively active, directing the function of protein
e
ff
ector dCas9 to a target gene Y; di
ff
erent dCas9 variants implement
di
ff
erent functions (edit, silence, induce, bind). From 5
′
to 3
′
,a
standard gRNA comprises a target-binding region, a Cas9 handle
recognized by the protein e
ff
ector, and a terminator region.
ACS Central Science
Research Article
DOI:
10.1021/acscentsci.9b00340
ACS Cent. Sci.
2019, 5, 1241
−
1249
1242
With the terminator switch mechanism, the sequences of the
RNA trigger X and the silencing target Y are fully independent,
with the cgRNA mediating allosteric regulation
the trigger
down-regulates cgRNA:dCas9 function not by sequestering the
target-binding region (orange in
Figure 2
b) but by hybridizing
to the distal trigger-binding region (blue). To test programm-
ability, we used NUPACK
31
,
32
to design a library of three
orthogonal cgRNA/trigger pairs (
Figure 2
d), achieving a
median
≈
4-fold conditional ON
→
OFF response to
expression of the cognate trigger (left) and median crosstalk
below 2% between noncognate cgRNA/trigger combinations
(right). Ideally, a cgRNA would have a strong ON state with
activity equivalent to a standard gRNA (ideal ON state) and a
clean OFF state with minimal activity equivalent to a no-target
gRNA lacking the target-binding region (ideal OFF state). For
this cgRNA mechanism, there is room for improvement in
both the ON and OFF states (
Figure 2
c and
Table S13a
). The
bimodality of the
fl
uorescence distributions observed for both
the standard gRNA control strain and the cgRNA-only strain
(
Figure 2
c) is a property of the assay and not of the terminator
switch mechanism; the same gRNA and cgRNA sequences
yield unimodal
fl
uorescence distributions in
E. coli
strains
created using a di
ff
erent plasmid layout (
Figure S34a
).
Single and Double Sequence Inserts for Construction
of Allosteric cgRNAs in Bacteria.
Seeking to improve
cgRNA performance for ON
→
OFF conditional logic, we
undertook a systematic study of single-stranded sequence
inserts into the standard gRNA structure, seeking to identify
inserts that satis
fi
ed two key properties: (1) strong ON state
−
inserts well-tolerated by dCas9; (2) clean OFF state
−
cgRNA
inactivated by hybridization of complementary trigger to
inserted domains. We created a total of 71
E. coli
strains to test
designed sequence inserts for each of three lengths (15, 25, 35
nt) at each of four insert sites (5
′
-extension, Cas9 handle loop,
terminator loop 1, terminator loop 2;
Figure S39
)orat
pairwise combinations of insert sites. Each of these modi
fi
ed
gRNAs represented a candidate allosteric cgRNA mechanism,
as the trigger sequence X is fully independent of the target
gene Y. Interestingly, all of the single and double inserts were
well-tolerated by dCas9 with a strong ON state comparable to
the standard gRNA, but most inserts did not mediate e
ff
ective
silencing when the cognate trigger was expressed (
Figure S39
and Table S16
). A notable exception was the modi
fi
ed gRNA
with 35 nt inserts in both the dCas9 handle loop and one of
the terminator loops, providing the basis for the
“
splinted
switch
”
cgRNA mechanism presented next.
Constitutively Active Splinted Switch cgRNAs (ON
→
OFF Logic) with Silencing dCas9 in Bacteria.
The
constitutively active
“
splinted switch
”
cgRNA mechanism
(
Figure 3
b) has extended loops in both the Cas9 handle
(domain
“
d
”
) and terminator (domain
“
e
”
). Hybridization of
RNA trigger X to both loops is intended to form a splint that is
structurally incompatible with cgRNA mediation of dCas9
function. In
E. coli
expressing silencing dCas9 and a
fl
uorescent
protein reporter (sfGFP) as the target gene Y (conditional
logic:
“
if not X then not Y
”
;
Figure 3
a), the splinted switch
exhibits a conditional ON
→
OFF response to expression of
RNA trigger X (
Figure 3
c). Examining a library of three
orthogonal splinted switch cgRNA/trigger pairs designed using
NUPACK (
Figure 3
d), we observe a median
≈
15-fold ON
→
OFF conditional response to expression of the cognate trigger
and median crosstalk of
≈
2% between noncognate cgRNA/
trigger combinations. As expected from our insert studies
(
Figure S39 and Table S16
), splinted switch cgRNAs exhibit a
strong ON state comparable to the ideal ON state of a
standard gRNA, and the OFF state could still be improved
relative to the ideal OFF state of a no-target gRNA lacking the
target-binding region (
Figure 3
c and
Table S13b
). As with the
terminator switch mechanism, splinted switch cgRNAs are
allosteric regulators
the trigger down-regulates cgRNA:d-
Figure 2.
Constitutively active terminator switch cgRNAs (ON
→
OFF logic) with silencing dCas9 in bacteria. (a) Conditional logic: if
not X then not Y. (b) cgRNA mechanism: the constitutively active
cgRNA is inactivated by hybridization of RNA trigger X. Rational
sequence design of cgRNA terminator region (domains
“
d
−
e
−
f
”
comprising 6 nt linker, 4 nt stem, 30 nt loop) and complementary
trigger region (domains
“
f
*
−
e
*
−
d
*
”
). (c) Expression of RNA trigger
X (40 nt unstructured + synthetic terminator hairpin) toggles the
cgRNA from ON
→
OFF, leading to an increase in
fl
uorescence.
Single-cell
fl
uorescence intensities via
fl
ow cytometry. Induced
expression (aTc) of silencing dCas9 and constitutive expression of
mRFP target gene Y and either: standard gRNA (ideal ON state),
cgRNA (ON state), cgRNA + RNA trigger X (OFF state; trigger
expression is IPTG-induced), no-target gRNA that lacks target-
binding region (ideal OFF state). Auto
fl
uorescence (AF): cells with
no mRFP. (d) Programmable conditional regulation using 3
orthogonal cgRNAs (A, B, C). Left: raw
fl
uorescence depicting ON
→
OFF conditional response to cognate trigger (fold change = OFF/
ON = [cognate trigger
−
AF]/[no trigger
−
AF]). Right: normalized
fl
uorescence depicting orthogonality between noncognate cgRNA/
trigger pairs (crosstalk = [noncognate trigger
−
no trigger]/[cognate
trigger
−
no trigger]). Bar graphs depict mean
±
estimated standard
error calculated based on the mean single-cell
fl
uorescence over
20 000 cells for each of
N
= 3 replicate wells (fold change and
crosstalk calculated with uncertainty propagation).
ACS Central Science
Research Article
DOI:
10.1021/acscentsci.9b00340
ACS Cent. Sci.
2019, 5, 1241
−
1249
1243
Cas9 function by hybridizing to extended loops (blue in
Figure
3
b) distal to the target-binding region (orange). The resulting
full sequence independence between RNA trigger X and target
gene Y provides the
fl
exibility for X to control regulatory scope
independent of the choice of Y.
Constitutively Inactive Toehold Switch cgRNAs (OFF
→
ON Logic) with Silencing dCas9 in Bacteria.
To reverse
the conditional logic, we then tested a constitutively inactive
“
toehold switch
”
cgRNA mechanism (
Figure 4
b) that is
conditionally activated by RNA trigger X (OFF
→
ON logic).
The target-binding region of the cgRNA (domain
“
u
”
)is
initially sequestered by a 5
′
extension to inhibit recognition of
target gene Y; hybridization of trigger X to this extension is
intended to desequester the target-binding region and enable
cgRNA direction of dCas9 function to target gene Y. In
E. coli
expressing silencing dCas9 and a
fl
uorescent protein reporter
Figure 3.
Constitutively active splinted switch cgRNAs (ON
→
OFF
logic) with silencing dCas9 in bacteria. (a) Conditional logic: if not X
then not Y. (b) cgRNA mechanism: the constitutively active cgRNA is
inactivated by hybridization of RNA trigger X. Rational sequence
design of the 35 nt Cas9 handle loop (domain
“
d
”
) and an extended
35 nt terminator hairpin loop (domain
“
e
”
). (c) Expression of RNA
trigger X (70 nt unstructured + synthetic terminator hairpin) toggles
the cgRNA from ON
→
OFF, leading to an increase in
fl
uorescence.
Single-cell
fl
uorescence intensities via
fl
ow cytometry. Induced
expression (aTc) of silencing dCas9 and constitutive expression of
sfGFP target gene Y and either: standard gRNA (ideal ON state),
cgRNA (ON state), cgRNA + RNA trigger X (OFF state), or no-
target gRNA that lacks target-binding region (ideal OFF state).
Auto
fl
uorescence (AF): cells with no sfGFP. (d) Programmable
conditional regulation using 3 orthogonal cgRNAs (A, B, C). Left:
raw
fl
uorescence depicting ON
→
OFF conditional response to
cognate trigger (fold change = OFF/ON = [cognate trigger
−
AF]/[no
trigger
−
AF]). Right: normalized
fl
uorescence depicting orthogonality
between noncognate cgRNA/trigger pairs (crosstalk = [noncognate
trigger
−
no trigger]/[cognate trigger
−
no trigger]). Bar graphs depict
mean
±
estimated standard error calculated based on the mean single-
cell
fl
uorescence over 20 000 cells for each of
N
= 3 replicate wells
(fold change and crosstalk calculated with uncertainty propagation).
Figure 4.
Constitutively inactive toehold switch cgRNAs (OFF
→
ON logic) with silencing dCas9 in bacteria. (a) Conditional logic: if X
then not Y. (b) cgRNA mechanism: the constitutively inactive cgRNA
is activated by hybridization of RNA trigger X. Rational sequence
design of the toehold (domain
“
d
”
; 15 nt) and loop (domain
“
e
”
;8
nt)
fl
anking the sequestration domain
“
u
*
”
(20 nt). (c) Expression of
RNA trigger X (35 nt unstructured + synthetic terminator hairpin)
toggles the cgRNA from OFF
→
ON, leading to a decrease in
fl
uorescence. Single-cell
fl
uorescence intensities via
fl
ow cytometry.
Induced expression (aTc) of silencing dCas9 and constitutive
expression of mRFP target gene Y and either: no-target gRNA that
lacks target-binding region (ideal OFF state), cgRNA (OFF state),
cgRNA + RNA trigger X (ON state), or standard gRNA (ideal ON
state). Auto
fl
uorescence (AF): cells with no mRFP. (d) Program-
mable conditional regulation using 3 orthogonal cgRNAs (A, B, C).
Left: raw
fl
uorescence depicting OFF
→
ON conditional response to
cognate trigger (fold change = OFF/ON = [no trigger
−
AF]/[cognate
trigger
−
AF]). Right: normalized
fl
uorescence depicting orthogonality
between noncognate cgRNA/trigger pairs (crosstalk = [noncognate
trigger
−
no trigger]/[cognate trigger
−
no trigger]). Bar graphs depict
mean
±
estimated standard error calculated based on the mean single-
cell
fl
uorescence over 20 000 cells for each of
N
= 3 replicate wells
(fold change and crosstalk calculated with uncertainty propagation).
ACS Central Science
Research Article
DOI:
10.1021/acscentsci.9b00340
ACS Cent. Sci.
2019, 5, 1241
−
1249
1244