A sensitive switch for visualizing natural gene silencing in single cells

Karmella A. Haynes1, Francesca Ceroni2, Daniel Flicker3, Andrew Younger3, and Pamela A.
Silver3,4

1. School of Biological and Health Systems Engineering, Arizona State University, Tempe, AZ
85287
2. Laboratory of Cellular and Molecular Engineering, University of Bologna, I-47521 Cesena,
Italy
3. Department of Systems Biology, Harvard Medical School, Boston, MA 02115
4. The Wyss Institute for Biologically Inspired Engineering, Harvard Medical School, Boston,
MA 02115

ABSTRACT

RNA interference is a natural gene expression silencing system that appears throughout the
tree of life. As the list of cellular processes linked to RNAi grows, so does the demand for
tools to accurately measure RNAi dynamics in living cells. We engineered a synthetic RNAi
sensor that converts this negative regulatory signal into a positive output in living
mammalian cells thereby allowing increased sensitivity and activation. Furthermore, the
circuit’s modular design allows potentially any microRNA of interest to be detected. We
demonstrated that the circuit responds to an artificial microRNA and becomes activated
when the RNAi target is replaced by a natural microRNA target (miR-34) in U2OS
osteosarcoma cells. Our studies extend the application of rationally designed synthetic
switches to RNAi, providing a sensitive way to visualize the dynamics of RNAi activity rather
than just the presence of miRNA molecules.

KEYWORDS: genetic switch, synthetic repressor, RNA interference, miR-34

1
Translational application of synthetic devices for medicine and basic research is enabled by
enhancing the synthetic biology toolkit with parts from mammalian cells. 1,2 Recently,
functional synthetic gene switches have been successfully designed for applications in
human health.3,4 Similar artificial networks could be designed to report important dynamic
cellular mechanisms that play an important role in normal cell development and disease.

RNA interference (RNAi) is linked to essential processes including cell cycle

progression{Boyerinas:2010gt, Lin:2003uu, Nimmo:2009di}, cellular differentiation
{Merkerova:2008dr}, apoptosis{Sun:2008fy}, and cancer development{Lynam-Lennon N. et
al, 2009} in myriad organisms. One class of small non-coding RNAs acts at the post-
transcriptional level to inhibit translation from target messenger RNAs (mRNAs). microRNAs
(miRNAs) are transcribed in the cell nucleus as long precursor molecules (pri-miRNA),
processed into hairpin RNAs (pre-miRNA) by the protein Drosha{202 Cullen,B.R. 2004}, and
exported into the cytoplasm. Pre-miRNAs are cleaved by the protein Dicer into single-
stranded 22-nucleotide RNAs that are incorporated into the RISC protein complex. The RISC-
miRNA complex can then bind target mRNAs and inhibit translation.

Recent studies are beginning to connect the dynamics of miRNA expression with cellular
and tissue phenotypes, advancing our knowledge of RNAi beyond a collection of data
showing which miRNAs are present in a tissue at a fixed point in time. So far, time-course
miRNA expression profiles have been created for different human developmental
processes. The accumulation of a set of miRNAs including miR-1, miR-133a, miR-133b, and
miR-206, has been observed during natural human fetal muscle development, as well as
during artificial induction of muscle cell differentiation in situ {Koutsoulidou:2011is}. A key
step in neural development is also linked to a dynamic miRNA expression profile, where
differentiation-associated miRNAs accumulate and proliferation-regulating miRNAs
decrease upon Schwann cell maturation.{Gokey:2011gx} Periodic expression of some
miRNAs appears to rely upon the circadian clock machinery{Yang:2008bm} and in some
cases may regulate the clock itself.{Cheng:2007wx} Rapid reporter systems that track

2
closely with miRNA activity in real time will enable us to discover to what extent miRNA
timing is linked to a biological purpose.

In the studies cited above, populations of cells or tissues were collected at specific time
points and lysed for RNA analyses. Other recent studies have aimed to generate expression
profiles in vivo by placing a fluorescence-producing gene (i.e. GFP) under the control of
microRNA (miRNA) or short interfering RNA (siRNA) so that the loss of expression indicates
miRNA and siRNA activity.{DePietriTonelli:2006wl, Zhang:2011cy, Varghese:2010hw} Using
loss of signal as a proxy for RNAi has several disadvantages. Signal depletion due to photo-
bleaching or cell division can confound the results of time-course experiments. Moreover,
loss-of-signal detection can be delayed by the slow degradation of stable fluorescent
proteins.{Milo:2010cz} Lastly, natural miRNAs with low activity levels may not efficiently
knock down highly-expressed reporters. Ideally, RNAi reporters should be fast, sensitive,
and produce a positive output.

We have engineered a synthetic reporter that converts RNAi-mediated gene silencing into a
positive, visible signal on the order of hours in single living cells. We built a new genetic
circuit based in part on the double repression scheme previously used by others
{Rinaudo:2007jq}. Thus far, such circuits have used strong synthetic gene promoters to
create cell-type classifiers that detect high levels of miRNA.{Xie:2011fi} Our device extends
the range of RNAi detection for synthetic circuits. We engineered a constitutive human
promoter that can be held in the initial “Off” state without relying an overwhelmingly
strong repressor, giving the device the sensitivity to detect low levels of miRNA near the
onset of RNAi. The double repression scheme places an output gene under the control of a
repressor protein, which is targeted by a specific miRNA. This double repression approach
links a gain of signal to RNAi activity. Thus, the observer can look for accumulation of signal
above a negative background level, which can be seen more readily than loss of signal
relative to positive background (Fig. 1A). The sooner the RNAi signal is seen, the more
closely the reporter tracks with RNAi in real time, and the more useful the reporter is for

3
live cell analysis.

Our genetic circuit has two states, which we qualitatively describe as “Off” and “On.” In the
Off state, a cyan fluorescent signal is repressed. In the On state, the cyan fluorescent signal
becomes expressed. The circuit has a modular design (Fig. 1B), which allows us to simply
choose a corresponding target site for different miRNAs of interest. The off state is
maintained by a large double RFP-tagged protein (RFP-Repressor) that binds downstream of
the CFP-Reporter promoter. The RFP-Repressor is placed under the control of miRNA such
that miRNA expression leads to repressor silencing and activation of the reporter (Fig. 1C).
Gene circuit components were optimized to achieve a stable “Off” state that could be
readily turned on in presence of a miRNA of interest. A synthetic RNAi system (miR-luc) was
used to test the performance of the device in living cells. We then added natural miRNA
sites to detect miR-34 in U2OS osteosarcoma cells. Our sensor system produced the first
evidence of cell-cycle-arrest associated miR-34 in unstressed U2OS cells.

RESULTS AND DISCUSSION

Regulating transcription by steric hindrance from a fluorescent protein. The RFP-Repressor

is a novel synthetic transcriptional repressor that binds near a promoter to occlude the
transcriptional machinery through steric hindrance. Most of its bulk comes from visible
flurophores; it does not rely upon a transcriptional repression domain to maintain gene
repression. Thus, its promoter-silencing activity is essentially a function of repressor protein
production and degradation, not epigenetic silencing effects. For instance, previously
reported synthetic genetic switches utilize the Krüppel associated box (KRAB) repression
domain.{Kramer:2004kq, Rinaudo:2007jq} KRAB interacts with heterochromatin
components such as HP1, which recruit chromatin modifying enzymes that form mitotically
stable gene repression.{Ryan:1999wj} We aimed to engineer fast reporter activation, where
repression is reversed on the order of minutes or hours, before cell division occurs.
Therefore we used steric hinderance, reminiscent of the TetR and LacI regulation

4
systems{Ramos:2005bt, Lewis:2005cg}, to compete with PolII at the promoter and block
transcriptional activation. Instead of tagging a full-length repressor with a fluorescent
protein, we fused two mCherry monomers to the Gal4 DNA-binding domain from yeast to
create a visible repressor. We included a PEST sequence motif to decrease protein half-life
and promote the loss of the repressor when RNAi is active.

We optimized the Off state of the circuit by testing increasingly active constitutive
promoters to drive RFP-Repressor expression. The relative strengths of the promoters we
used are Ubiquitin C (Ubc) < human phosphoglycerate kinase (HPK) < cytomegalovirus
(CMV), based on fluorescence microscopy and flow cytometry of cells transfected with CFP
driven by each promoter (Fig. 2A). We observed greater CFP repression as RFP-Repressor
expression was increased (Fig. 2B). Repression well below the fully active state was
achieved by using the Ubc-Gal4 or HPK-Gal4 promoter to regulate CFP. (something about
SNR here) Interestingly, CMV-driven RFP-Repressor appears to decrease CFP levels when
there are no Repressor binding sites at the promoter of the CFP-Reporter (where Px = CMV
and Gal4 = 0). This suggests that CMV on one plasmid may indirectly affect expression from
the other plasmid, perhaps through competing for a fixed pool of transcription factors in the
cell.

We also considered the possibility that CMV might produce too much RFP-Repressor, the
direct target of RNAi in this circuit, and overwhelm the natural silencing machinery. Thus,
we altered the binding sites to increase the local concentration of RFP-Repressor at the CFP-
Reporter promoter without increasing RFP-Repressor expression. We found that doubling
the number of repressor binding sites (from 5 to 10 copies of Gal4) was sufficient to achieve
greater repression in cells that express RFP-Repressor with weak and medium strength
promoters (Ubc and HPK, respectively) (Fig. 2B). We ruled out the Ubc-CFP-Reporter
constructs because expression of the fully active promoter is difficult to detect. These
results allowed us to identify an optimal design (i.e., HPK-RFP-Repressor/ HPK-CFP-
Reporter), where the Off state can be maintained without high CMV-driven expression of

5
the repressor.

The sensor is activated upon miRNA induction. We demonstrated that the miR-sensor
produces increased CFP signal in the presence of miRNA by using a drug-inducible
orthogonal RNAi system. The JDS33 cell line{Shih:2011cg} carries a transgene that expresses
luciferase miRNA (Luc-1601{Chung:2006hj}, referred to here as “miR-luc”) when doxycycline
(dox) is added to the cell culture medium. miR-luc has imperfect complementarity with the
target sequences at the 3’ end of the RFP-Repressor gene, which is characteristic of natural
miRNA regulation. Co-expressed visible YFP indicated the amount of miR-luc present, as
confirmed by real time PCR analysis and flow cytometry (Fig. 3A). Therefore, the miR sensor
should switch to the On state when YFP is expressed.

We observed that the circuit becomes activated in the presence of miR-luc. Two different
circuit configurations were compared. Circuit 1 had 10 Gal4 sites at the CFP-Reporter and an
HPK (medium strength) promoter driving the RFP-Repressor. Circuit 2 had 5 Gal4 sites at the
CFP-Reporter and a CMV (strong) promoter driving the RFP-Repressor (Fig. 3B). We
transfected JDS33 cells with each circuit and treated the cells with 1 μg/ml dox for 48 hours.
About 1.7x104 molecules of miR-luc per cell is present under these conditions (Fig. 3A).
Shifts from low to high CFP signal showed that both sensors became activated after dox-
induced miR-luc expression (Fig. 3). Activation of Circuit 1 was comparable to Circuit 2, thus
a clear switch from Off to On does not require an over-expressed repressor.

A sensor designed to detect miR-34 is activated in human osteosarcoma cells over a short
time scale. miRNA sensors carrying target sequences for the natural miRNA miR-34 become
activated in the human bone cancer-derived cell line U2OS. miR-34 plays an important role
in cell physiology. miR-34 silences cyclinD1 (CCDN1), cyclin dependent kinase 6 (CDK6), and
B-cell lymphoma 2 (BCL2){Sun:2008fy}, which promote progression into mitosis and prevent
apoptosis. Thus, miR-34 activity has a negative impact on cell proliferation and is expected
to be diminished or absent in dividing cancer cells. miR-34 has been detected in U2OS at

6
low or high levels, depending upon a proliferating or arrested growth status, respectively.
{He:2009hm} We used our system to look at miR-34 activity over a time interval of
approximately one cell division (20 hours) in proliferating U2OS cells with possibly low levels
of miR-34.{He:2009hm}

Sensors carrying miR-34 target sites from CCDN1 and CDK6, but not BCL2 showed switch-
like activation in U2OS cells. All three sensors were constructed based on the configuration
for Circuit 1 (Fig. 3B), transfected into U2OS cells, and analyzed six hours later via live cell
time-lapse microscopy. Circuit 1, without miR-34 target sites, was used as a negative control
since it is not expected to become activated in U2OS cells that do not express artificial miR-
luc. The negative control showed a gradual linear accumulation of both signals over the
time frame analyzed. The Bcl2 sensor behaved the same as the negative control construct
(data not shown). In contrast, the CCDN1 and CDK6 sensors showed a rise and fall in RFP
signal, accompanied by a roughly sigmoidal increase in CFP production. These results
indicate that for time-course single cell analysis performed in under 20 hours, activated
sensors can be distinguished by more rapid accumulation of CFP signal (Fig. 4A).

Computational analysis of RNA hybridization suggest that the affinity of miR-34 for each of
the targets may determine the differences in the behavior of the sensors we tested. We
calculated the free energy of binding between CCND1, CDK6, and BCL2 target sites and each
member of the miR-34 family (miR-34a, b, and c). Unresponsiveness of the BCL2 sensor
might be a consequence of this target’s low affinity for miR-34, as indicated by relatively
high free energies of the predicted miR-34-target hybrids (Fig. S1). Other target site
sequence variants may need to be tested to create an effective BCL2 sensor.

We also observed activation of a redesigned miR-34 sensor in which CFP was destabilized to
reduce signal noise in the Off state. The accumulation of both RFP and CFP expression from
the construct that was designed to be constitutively off (lacking miR-34 target sites) was
unexpected, since our previous analyses indicated an inverse relationship between RFP and

7
CFP (Fig. 4A). The time-course results may be due to some instability of the off-state before
proteins reach steady state levels. We reduced accumulation of CFP signal in the Off state
by destabilizing CFP with a PEST protein degradation tag. Cells carrying the miR-34 sensor
with destabilized CFP showed greater CFP signal than a sensor that lacked the miR-34 target
sites (Fig. 4B).

Conclusion. Our work presents two important advances for synthetic device engineering
and applications. First, the RFP-Repressor may serve as a generally useful model for
designing synthetic transcriptional regulators de novo. Other commonly used repressors
(i.e., LacI and Tet) are derived from naturally occurring gene repression systems. To our
knowledge, the RFP-Repressor is the first reported functional synthetic repressor that has
been built entirely from protein modules that are not typically associated with
transcriptional repression. Our results directly demonstrate that a functional repressor may
be built from a protein designed simply to occupy sites immediately downstream of a
constitutive promoter, and that its function depends upon the local concentration of
repressors at the promoter.

Second, we have extended the application of the double repression circuit design to sense
miRNAs previously detected at low levels in the proliferating cell state{He:2009hm}, over a
short time scale. miR sensor activation may be sensitive to low levels of miR-34, or activated
sensors might signify the onset of miR-34 activity prior to subsequent cell cycle arrest. For
medical applications in cancer therapy, our circuit design could be engineered to activate a
cancer-killing gene near the onset of metastasis-associated miRNA activity. So far, we have
expressed the switches from extra-chromosomal plasmids. Future applications that stably
integrate the circuit into the genome will be useful for observing the stochasticity of miR-34
activity at the individual cell level, which is important for determining whether single cancer
cells alter states to become more aggressive. Our work demonstrates the power and
potential of modular rational design in extending the utility synthetic devices for research
and medical applications.

8
METHODS

Constructs. DNA fragments with universal cloning sites (EcoRI, NotI, XbaI, SpeI, and PstI)
were generated by PCR and assembled according to a modified BioBrick DNA assembly
method {Phillips:2006wn}, using E. coli DH5-alpha maintained under standard conditions.
All DNA fragment and annotated full-length construct sequences are freely available in the
MIT Registry of Standard Biological Parts {Endy:2005cy} (See Table S1). The RFP-Repressor
was constructed from a Gal4 DNA-binding domain followed by two copies of mCherry. The
constitutive promoters Ubc, HPK and CMV were cloned upstream of five (5xGal4) or ten
(10xGal4) copies of the Gal4 DNA binding sites to generate Gal4-regulated promoters.
Complete constructs were excised from the V0120 high-copy cloning vector and inserted
into modified pcDNA 3.1+ vectors. All plasmids were verified by restriction digests and DNA
sequencing (Genewiz, Inc., Cambridge, MA) from the vector into the 5’ and 3’ end of the
inserts prior to transfection.

Cell culture and transfections. Cell line JDS33 is described by Shih et al. {Shih:2011cg}. U2OS
and JDS33 cells were grown in McCoy’s 5A medium supplemented with 10% tetracycline-
free fetal bovine serum and 1% penicillin and streptomycin. Cells were grown at 37°C in a
humidified CO2 incubator. For transfections, plasmid DNA was extracted and purified from
E. coli using a QIAGEN miniprep kit. 1 μg plasmid DNA, 3 μl Lipofectamine LTX (Invitrogen),
and 3 μl PLUS reagent in 190 μl Opti-MEM was added to ~2.5x10 5 cells per well (in a 12-well
plate) in antibiotic-free growth medium.

Flow cytometry.
Cells were grown in a 12-well plate (~5x10 5 per sample/ well), trypsinized, collected in
growth medium, and washed and resuspended in 1x Dulbecco’s PBS. Flow cytometry was
performed 18 hours after transfection using a BD Biosciences LSRII HTS-3 Laser high
throughput sampler platform. For flow cytometry analyses to detect mCherry, YFP (Venus),
and AmCyan, the following excitation lasers and filters were used: 594 nm Yellow/ DsRed

9
(620/22), 488 nm Blue/ FITC (520/50), and 405 nm Violet/ AmCyan (525/50). For every
experiment, non-fluorescent U2OS cells were used to adjust the voltage to minimize
autofluorescence signal in all three channels, and to create a 1:1 forward and side scatter
ratio. Background was subtracted from signal by setting the threshold gate above the signal
from non-fluorescent U2OS cells. Data was visualized using FACS Diva software, and
statistically analyzed using FlowJo software.

Quantitative real-time reverse transcription PCR of miR-luc and YFP detection

JDS33 cells were cultured in 6-well plate (~1x10 6 per sample/ well) and treated with 0, 0.01,
0.03, 0.1, 0.3 or 1 μg/ml doxycycline to induce YFP and miR-luc expression for 48 hours. One
set of samples was harvested for YFP detection via flow cytometry. A duplicate set was
processed for RNA isolation. RNA was extracted with TRIzol Reagent (Invitrogen 15596-018)
and purified according to the Invitrogen protocol. cDNA synthesis and PCR were designed as
described in {Raymond:2005hw}. cDNA synthesis (SuperScript III, Invitrogen 18080-051) was
performed according to the manufacturer’s protocol with either 5 μg template RNA from
the dox-treated JDS33 cells (Experimental) or 2 pmol synthesized “Input Control” RNA oligos
(IDT DNA), plus 50 pmol miR-luc “tailed” primers (5’-catgatcagctgggccaagaaaatcagagag)
instead of oligo dT in a final reaction volume of 20 μl. To normalize for Experimental RNA
loading, oligo dT primers were used in separate reactions to generate Total cDNA for
measuring GAPDH transcripts. Completed reactions were treated with 1 μl of RNaseH.

qRT-PCR was performed on Experimental cDNA, Input Control cDNA, or Total cDNA.
Experimental and Input Control cDNA reactions contained 0.8 μl cDNA, 36.0 pmol primers
(miR-luc forward 5’-tttatgaggatctctct; miR-luc reverse 5’-catgatcagctgggcca), and 7.5 μl SYBR
Green Master Mix (Applied Biosystems) in a final volume of 15 μl. GAPDH reactions
contained 0.8 μl Total cDNA and 2.25x10 -3 pmol primers (GAPDH forward 5’-
ccgcatcttcttttgcgtcgcc; GAPDH reverse 5’-accaggcgcccaatacgacc). For standardization, we
used a synthesized DNA oligo template Standard that carries the full miR-luc PCR template
sequence. qRT-PCR was performed as described above using 0.8 ul of 1.0x10 -5, x10-4, x10-3,

10
or x10-2 uM synthesized DNA solution and 36.0 pmol primers (miR-luc forward and miR-luc
reverse).

Fold change relative to the least abundant standard template was caluclated as 2^(Ct Standard 1
- Ctsample) for all samples. DNA oligo molecules per reaction for Standards 1 through 4
(4.816x106, x107, x108, and x109) was plotted against fold change to generate a standard
curve. The line of best fit equation was used to calculate miR-luc molecules per reaction for
the Experimental cDNA samples; these values were divided by Total cDNA GAPDH
normalization values (2^(CtGAPDH sample / CtGAPDH 0 dox)) and adjusted to take into account 73%
cDNA synthesis eficiency (as indicated by the Input Control values). “miR-luc molecules per
cell” (Fig. 3A) was calculated as normalized Experimental miR-luc molecules per reaction
divided by 8000 cells per reaction.

Single cell fluorescence assays

U2OS cells were cultured in 12-well glass-bottom plates and transiently transfected with
miRNA sensor circuit constructs for 6 hrs. Phase contrast, red fluorescent, and cyan
fluorescent images were collected at 1 hour intervals using a Nikon inverted microscope,
controlled by Metamorph acquisition software. Mean fluorescence intensities of individual
interphase nuclei were manually measured using ImageJ.

AUTHOR CONTRIBUTIONS
KAH and PAS conceptualized the project. KAH, FC, DF, and AY built and transfected
constructs and performed flow cytometry. KAH and DF performed time-course microscopy.
FC and AY designed and performed miR-luc RT-PCR. DF performed the RNAhybrid analysis.
All authors drafted the manuscript.

The authors declare no conflicts of interest.

11
ACKNOWLEDGMENTS
We thank J. Moore for assistance with flow cytometry, Z. Xie and J. Lohmeuller for general
advice, and J. Shih for JDS33 cells and the miR-luc sequences. This work was supported, in
whole or in part, by National Institutes of Health Grants GM36373 (to PAS) and
1F32GM087860 (to KAH). FC was supported by University of Bologna. DF and AY were
supported by the Harvard CSB (P50GM068763).

SUPPORTING INFORMATION AVAILABLE

Supplemental material includes a detailed list of constructs as entries into the MIT Registry
of Standard Biological Parts (Table S1), Figure S1, Figure S2, supplemental methods, and
references. This information is available free of charge via the Internet at
http://pubs.acs.org/.

REFERENCES

Boyerinas, B., Park, S.-M., Hau, A., Murmann, A. E., & Peter, M. E. (2010). The role of let-7 in
cell differentiation and cancer Endocrine-related cancer, 17(1), F19–36. doi:10.1677/ERC-09-
0184
Cheng, H.-Y. M., & Obrietan, K. (2007). Revealing a role of microRNAs in the regulation of
the biological clock Cell cycle (Georgetown, Tex.), 6(24), 3034–3035.
De Pietri Tonelli, D., Calegari, F., Fei, J.-F., Nomura, T., Osumi, N., Heisenberg, C.-P., &
Huttner, W. B. (2006). Single-cell detection of microRNAs in developing vertebrate embryos
after acute administration of a dual-fluorescence reporter/sensor plasmid BioTechniques,
41(6), 727–732.
Gokey, N. G., Srinivasan, R., Lopez-Anido, C., Krueger, C., & Svaren, J. (2011). Developmental
Regulation of MicroRNA Expression in Schwann Cells Molecular and cellular biology.
doi:10.1128/MCB.06270-11

12
Greber, D., & Fussenegger, M. (2007). Mammalian synthetic biology: engineering of
sophisticated gene networks Journal of biotechnology, 130(4), 329–345.
doi:10.1016/j.jbiotec.2007.05.014
Haynes, K. A., & Silver, P. A. (2009). Eukaryotic systems broaden the scope of synthetic
biology The Journal of cell biology, 187(5), 589–596. doi:10.1083/jcb.200908138
He, C., Xiong, J., Xu, X., Lu, W., Liu, L., Xiao, D., & Wang, D. (2009). Functional elucidation of
MiR-34 in osteosarcoma cells and primary tumor samples Biochemical and Biophysical
Research Communications, 388(1), 35–40. doi:10.1016/j.bbrc.2009.07.101
Koutsoulidou, A., Mastroyiannopoulos, N. P., Furling, D., Uney, J. B., & Phylactou, L. A.
(2011). Expression of miR-1, miR-133a, miR-133b and miR-206 increases during
development of human skeletal muscle BMC developmental biology, 11, 34.
doi:10.1186/1471-213X-11-34
Kramer, B. P., Viretta, A. U., Baba, M. D.-E., Aubel, D., Weber, W., & Fussenegger, M. (2004).
An engineered epigenetic transgene switch in mammalian cells. Nature Biotechnology,
22(7), 867–870. doi:10.1038/nbt980
Krüger, J., & Rehmsmeier, M. (2006). RNAhybrid: microRNA target prediction easy, fast and
flexible Nucleic Acids Research, 34(Web Server issue), W451–4. doi:10.1093/nar/gkl243
Lewis, M. (2005). The lac repressor Comptes rendus biologies, 328(6), 521–548.
doi:10.1016/j.crvi.2005.04.004
Lin, S.-Y., Johnson, S. M., Abraham, M., Vella, M. C., Pasquinelli, A., Gamberi, C., Gottlieb, E.,
et al. (2003). The C elegans hunchback homolog, hbl-1, controls temporal patterning and is
a probable microRNA target Developmental Cell, 4(5), 639–650.
Milo, R., Jorgensen, P., Moran, U., Weber, G., & Springer, M. (2010). BioNumbers--the
database of key numbers in molecular and cell biology Nucleic Acids Research, 38(Database
issue), D750–3. doi:10.1093/nar/gkp889
Nimmo, R. A., & Slack, F. J. (2009). An elegant miRror: microRNAs in stem cells,
developmental timing and cancer. Chromosoma, 118(4), 405–418. doi:10.1007/s00412-009-
0210-z
Ramos, J. L., Martínez-Bueno, M., Molina-Henares, A. J., Terán, W., Watanabe, K., Zhang, X.,

13
Gallegos, M. T., et al. (2005). The TetR family of transcriptional repressors Microbiology and
molecular biology reviews : MMBR, 69(2), 326–356. doi:10.1128/MMBR.69.2.326-356.2005
Raymond, C. K., Roberts, B. S., Garrett-Engele, P., Lim, L. P., & Johnson, J. M. (2005). Simple,
quantitative primer-extension PCR assay for direct monitoring of microRNAs and short-
interfering RNAs RNA, 11(11), 1737–1744. doi:10.1261/rna.2148705
Rehmsmeier, M., Steffen, P., Hochsmann, M., & Giegerich, R. (2004). Fast and effective
prediction of microRNA/target duplexes RNA, 10(10), 1507–1517. doi:10.1261/rna.5248604
Rinaudo, K., Bleris, L., Maddamsetti, R., Subramanian, S., Weiss, R., & Benenson, Y. (2007). A
universal RNAi-based logic evaluator that operates in mammalian cells. Nature
Biotechnology, 25(7), 795–801. doi:10.1038/nbt1307
Ryan, R. F., Schultz, D. C., Ayyanathan, K., Singh, P. B., Friedman, J. R., Fredericks, W. J., &
Rauscher, F. J. (1999). KAP-1 corepressor protein interacts and colocalizes with
heterochromatic and euchromatic HP1 proteins: a potential role for Krüppel-associated box-
zinc finger proteins in heterochromatin-mediated gene silencing Molecular and cellular
biology, 19(6), 4366–4378.
Shih, J. D., Waks, Z., Kedersha, N., & Silver, P. A. (2011). Visualization of single mRNAs
reveals temporal association of proteins with microRNA-regulated mRNA. Nucleic Acids
Research, 39(17), 7740–7749. doi:10.1093/nar/gkr456
Chung, K. H., Hart, C.C., Al-Bassam, S., Avery, A., Taylor, J., Patel, P.D., Vojtek, A.B., Turner,
D.L. (2006). Polycistronic RNA polymerase II expression vectors for RNA interference based
on BIC/miR-155. Nucleic Acids Research, 34(7), e53–e53. doi:10.1093/nar/gkl143
Sun, F., Fu, H., Liu, Q., Tie, Y., Zhu, J., Xing, R., Sun, Z., et al. (2008). Downregulation of
CCND1 and CDK6 by miR-34a induces cell cycle arrest FEBS letters, 582(10), 1564–1568.
doi:10.1016/j.febslet.2008.03.057
Varghese, J., Lim, S. F., & Cohen, S. M. (2010). Drosophila miR-14 regulates insulin
production and metabolism through its target, sugarbabe Genes & Development, 24(24),
2748–2753. doi:10.1101/gad.1995910
Xie, Z., Wroblewska, L., Prochazka, L., Weiss, R., & Benenson, Y. (2011). Multi-Input RNAi-
Based Logic Circuit for Identification of Specific Cancer Cells. Science, 333(6047), 1307–1311.

14
doi:10.1126/science.1205527
Yang, M., Lee, J.-E., Padgett, R. W., & Edery, I. (2008). Circadian regulation of a limited set of
conserved microRNAs in Drosophila BMC Genomics, 9, 83. doi:10.1186/1471-2164-9-83
Ye, H., Daoud-El Baba, M., Peng, R.-W., & Fussenegger, M. (2011). A synthetic optogenetic
transcription device enhances blood-glucose homeostasis in mice Science, 332(6037), 1565–
1568. doi:10.1126/science.1203535
Zhang, X., Zabinsky, R., Teng, Y., Cui, M., & Han, M. (2011). microRNAs play critical roles in
the survival and recovery of Caenorhabditis elegans from starvation-induced L1 diapause
Proceedings of the National Academy of Sciences of the United States of America, 108(44),
17997–18002. doi:10.1073/pnas.1105982108

Figure 1. Schematic diagram of the mammalian miRNA reporter circuit. (A) Qualitative

15
comparison of two approaches for detecting RNAi, loss of signal vs. accumulation of signal.
Signal 1 is directly repressed by RNAi. Signal 2 is repressed by a protein that is deactivated
by RNAi. Once RNAi activity reaches an arbitrary threshold “a,” target protein degradation
exceeds production. When Repressor concentration is insufficient to repress expression,
Signal 2 accumulates (“c”) to a visible level sooner than Signal 1 is depleted by 50% (“b”). (B)
General construct map. For all experiments, both genes were placed on the same vector. Py
and Px = constitutive promoters driving expression of the CFP-Reporter and RFP-Repressor
genes, respectively; NLS = nuclear localization sequence; CFP = AmCyan; RFP = mCherry;
polyA = bovine growth hormone polyadenylation signal; PEST = protein degradation signal.
(C) In the absence of miRNA, RFP-Repressor is expressed and inhibits CFP (Off). When a
miRNA recognizes its specific target site within the 3’-UTR of the RFP-Repressor mRNA
transcript, the Repressor protein is degraded and CFP is expressed (On).

16
Figure 2. The off state can be tuned by increasing the concentration of RFP-Repressor at
the output gene promoter. (A) Fluorescence microscopy of cells carrying CFP driven by a
Ubc, HPK, or CMV promoter. Numbers show total CFP intensity (mean CFP fluorescence in
CFP-positive cells, multiplied by the frequency of CFP-positive cells) from flow cytometry.
(B) RFP-Repressor expression levels were regulated by Ubc, HPK, and CMV promoters (Px),
respectively. The local concentration of RFP-Repressor at the CFP promoter (Py) was
modulated by inserting zero, 5, or 10 copies of RFP-Repressor binding sites (Gal4)
downstream of the Py transcription start site. “Normalized CFP output” is the CFP intensity
(mean CFP fluorescence in CFP/RFP-positive cells, multiplied by the frequency of CFP-
positive cells) divided by the corresponding values shown in Fig. 2A (i.e., Py = Ubc values /
Ubc 4.2x105; Py = HPK values / HPK 1.09x106; Py = CMV values / CMV 2.6x106). The
arithmetic mean was used for all calculations.

17
Figure 3. The miRNA sensor responds to an artificial miRNA. (A) YFP signal indicates the
level of doxycycline-induced miR-luc in JDS33 cells. MiR-luc molecules per cell was
measured by real time quantitiave RT-PCR. YFP signal was detected by flow cytometry
analysis. YFP total signal = mean fluorescence x frequency of fluorescence signal. (B)
Comparison of two miRNA sensor designs, one with ten binding sites for the RFP-Repressor
(Circuit 1) vs. one with higher RFP-Repressor expression (Circuit 2). YFP, CFP, and RFP signal
(left to right) was measured by flow cytometry. Top graphs: error bars show the standard
error of the mean signal intensity. Bottom graphs: total signal (mean signal intensity x

18
frequency of fluorescence signal). The arithmetic mean was used for all calculations.

Figure 4. The behavior of natural microRNA sensors indicate miR-34 activity in U2OS cells.
(A) Live cells carrying one of three miR sensor circuits with different miR target sites
(CCND1, CDK6, or luc) were imaged via fluorescence microscopy over a ~30 hour period.
Each thick, thin, or dashed line corresponds to the same single cell in the RFP (top) and CFP
(bottom) channel. RFP intensity and CFP output were measured as the mean signal of
individual (within a single cell cycle). In cells where RFP-Repressor is targeted by the miR-34
silencer (CCND1 and CDK6), a rise and then fall in RFP signal is associated with faster
accumulation of CFP output, compared to a negative control where RFP-Repressor is not

19
targeted by miR-34 (luc). (B) The addition of a PEST tag to CFP reduces accumulation of CFP
expression from a sensor, whereas the sensor that carries the CDK6 target is activated.
Phase contrast, RFP, and CFP channels are shown from left to right.

20