Design of synthetic epigenetic circuits featuring memory effects and reversible switching based on DNA methylation

Epigenetic systems store information in DNA methylation patterns in a durable but reversible manner, but have not been regularly used in synthetic biology. Here, we designed synthetic epigenetic memory systems using DNA methylation sensitive engineered zinc finger proteins to repress a memory operon comprising the CcrM methyltransferase and a reporter. Triggering by heat, nutrients, ultraviolet irradiation or DNA damaging compounds induces CcrM expression and DNA methylation. In the induced on-state, methylation in the operator of the memory operon prevents zinc finger protein binding leading to positive feedback and permanent activation. Using an mf-Lon protease degradable CcrM variant enables reversible switching. Epigenetic memory systems have numerous potential applications in synthetic biology, including life biosensors, death switches or induction systems for industrial protein production. The large variety of bacterial DNA methyltransferases potentially allows for massive multiplexing of signal storage and logical operations depending on more than one input signal.

activates the lacZ reporter gene in a quantifiable manner. Gene activation of ZnF_Gal11P fusion proteins (dark blue bars) was tested in comparison to background signals measured for Gal11p in absence of ZnF (light blue bar). The DNA binding behavior of ZnF proteins was also tested under conditions in which the DNA-(adenine N6)-methyltransferase CcrM was expressed. The dark green bars show the ß-galactosidase activity in the presence of ZnF_Gal11p and CcrM, and the light green bars show the background ß-galactosidase activity in the presence of CcrM. The error bars indicate the s.d. of two biological replicates.
As a control, a ZnFinger_Gal11p fusion protein and its corresponding binding site was used, which does not include a CcrM recognition site. The control was provided as part of the used Addgene Kit #1000000010 to test ZnF protein binding. The tested ZnF binding sites (ZFBS) are underlined, CcrM target sequences are overlined and methylatable adenines are shown in bold. The amino acid sequences of the ZnF protein's alpha helixes forming the sequence specific DNA contacts are indicated. The Asn and Gln residues designed to be involved in binding the unmethylated adenine residue are marked in blue.
Supplementary Figure 3. Insertion of the CcrM DNA methyltransferase into the pAC-Kan-alphaGal4 plasmid of the bacterial two-hybrid system. A) Plasmid map of pAC-Kan-alphaGal4_CcrM with indicated restriction sites. B) Restriction digest of pAC-Kan-alphaGal4 and pAC-Kan-alphaGal4_CcrM plasmid DNA with the endonucleases HinfI and NdeI. HinfI only cleaves GANTC sites in an unmethylated state. The results show that the introduced CcrM gene is expressed and active in bacterial cells, because GANTC sites in pAC-Kan-alphaGal4_CcrM are protected from HinfI cleavage. Supplementary Figure 4. EMSA substrates and quality control. A) EMSA DNA substrate sequence with zinc finger binding sites (ZFBS). Indicated are HinfI restriction sites, which are protected from cleavage by CcrM methylation of the underlined adenine base. EMSA substrates were generated by PCR using 5'-Cy5-labelled oligonucleotides as primers. This substrate was used in gel shift experiments ( Supplementary Fig. 5) in unmethylated form (BS, "for binding site") and methylated form (BS-M, for "binding site methylated"). Methylation of the substrate was performed in vitro using purified CcrM enzyme 7,8 . B) EMSA DNA substrate sequence without ZFBS (c, for "control") in gel shift experiments ( Supplementary Fig. 5). C) Digestion of EMSA substrates containing ZFBSs. In vitro CcrM-methylated and unmethylated DNA substrates were digested with HinfI and the products separated by 8% non-denaturing polyacrylamide gelelectrophoresis in 0.5 x TB buffer. DNA fragments were visualized by GelRed nucleic acid stain and by fluorescence detection of the Cy5 dye. Expected product sizes visible with GelRed staining are 22 bp, 25 bp, and 105 bp for unmethylated substrate, and 156 bp for the methylated substrate.  Figure 6. Schematic representation and validation of the artificial methylation sensitive gene expression system. A) Plasmid map of the ZnF2 GFP reporter plasmid. B) Circuit designs of the artificial methylation sensitive gene expression system. The artificial ZnF repressor is constitutively expressed. The monomeric ZnF protein does not repress GFP gene expression, hence, a fluorescence signal is detected (orange trace). A dimerized version of the zinc finger protein (ZnF2), however, is able to repress GFP expression (black trace). The green trace shows GFP expression in the absence of ZnF proteins. C) Induction of the artificial methylation sensitive gene expression system. Upon arabinose supplementation CcrM is expressed and it methylates the binding sites for the ZnF repressor, weakening ZnF2 binding and leading to GFP expression (green trace). GFP expression is not detectable under repressive glucose conditions (black trace). Figure 7. Plasmid maps of the trigger plasmids for arabinose induction of CcrM used in memory system II. A) First generation of the trigger plasmid based on pBAD24 plasmid containing ccrM gene. B) Second generation of the arabinose trigger plasmid. The pBAD24 plasmid containing a ccrM gene was modified by replacing the pBR322 origin of replication with a p15A origin of replication in order to propagate the plasmid in E. coli cells containing a second plasmid with a pBR322 origin of replication. The gene for the red fluorescent protein mCherry was introduced downstream of the ccrM gene with its own ribosome binding site generating a polycistronic operon.

Supplementary
Supplementary Figure 8. Physical induction of the memory system I by heat. A) Scheme of the memory system in the off-state during cultivation at 30°C. The ZnF4 repressor binds the promoter region of the reporter-maintenance operon and represses transcription because the ZnF binding sites are in an unmethylated state (empty lollipops). The ZnF4 repressor negatively regulates the transcription of its own gene (dashed lines). This autoinhibition is unaffected by CcrM expression, because the ZnF binding sites in the ZnF4 promoter do not overlap with a GANTC site. B) Switching into the on-state. Upon a temperature shift to 37°C, binding of the ZnF4 repressor to its binding sites is weakened, which leads to expression of the reporter-maintenance operon. C) Stable on-state propagation. The CcrM methyltransferase methylates the ZnF binding sites and hinders DNA binding of the ZnF4 protein. DNA methylation lost by DNA replication is immediately restored by CcrM, resulting in a positive feedback loop (memory function). D) Annotated DNA sequences of the promoter regions of the ZnF protein and the memory operon. Zinc finger binding sites are indicated with blue arrows, -35 and -10 regions as dark grey boxes and the genes for ZnF4, EGFP and CcrM are represented as colored boxes, similarily as drawn in panels A-C.
Supplementary Figure 9. Efficiency of different ZnF binding site arrangements for off-state maintenance. A) Scheme of the memory system with indicated palindromic double binding sites. O1 contains one double binding site, similarly to the methylation sensitive gene expression system in Supplementary Fig. 6. The O2 construct has an additional double binding site 100 bps upstream. O3 harbors the double binding site O1 and additionally one double binding site in the intergenic region just upstream of the CcrM gene. The O2_O3 construct includes all three binding sites. B) Fluorescence measurements of 30°C overnight cultures containing the different memory plasmids. All three palindromic double binding sites are necessary for stable and efficient repression of the memory system allowing to maintain an offstate with no detectable fluorescence.
Supplementary Figure 10. Autoregulation of the ZnF repressor protein expression. In order to achieve a regulated expression of the ZnF repressor protein (ZnF4), two binding sites for ZnF_1012 not overlapping with GANTC sites were introduced into the promoter region of ZnF4. A) We studied the effect of the autoregulation of ZnF4 by comparison of the protein expression level of FLAG tagged ZnF4 with (lanes labelled with +) and without autoregulation (lanes labelled with -) by Western Blot. XL1-Blue cells transformed with the memory-maintenance plasmid with and without autoregulation of ZnF4 were grown overnight at 30°C in LB medium supplemented with kanamycin and cells were harvested by centrifugation. Protein extracts were prepared by suspending 0.5 OD600nm cells in 2 x SDS-PAGE loading buffer followed by incubation at 95°C for 10 minutes. Total cell lysates were separated by polyacrylamide gel electrophoresis on a 15% SDS-Polyacrylamide gel. Proteins were transferred onto a nitrocellulose membrane and the membrane was blocked by incubation in PBST containing 5% skim milk for 2 h at room temperature. After washing, the blot was incubated with primary anti FLAG antibody (Sigma, SLBN5629V) for 1 h and washed again before incubating with secondary anti mouse antibody coupled to horseradish peroxidase (GE Healthcare, 370149). Detection was performed with enhanced chemiluminescent substrate (Pierce® ECL Western Blotting Substrate) following the manufacturer's recommendations and a FUSION Solo (VWR International) system was used for signal recording. The western blot of the synthetic FLAG tagged ZnF4 repressor (theoretical size: 16.6 kDa) shows a much lower expression of ZnF4 under autoregulation conditions. B) Coomassie stained SDS-PAGE serving as a loading control for panel A. C) Fluorescence measurements of memory system I with and without autoregulation of the ZnF4 repressor (BI, before induction; AI, after induction; 10 h AI, 10 h after induction). 10 h after heat induction of the memory system, the system returned to the off-state without autoregulation of the ZnF repressor, presumably because of high concentrations of the repressor protein blocking access of CcrM. In contrast, a stable on-state was observed with autoregulation of the ZnF repressor.
Supplementary Figure 11. Stability of the on-state of memory systems I and II. A) XL1-Blue cells with the memory system I were switched to the on-state by heat induction. After 48 h cultivation in liquid culture at 30°C, cells were plated to single colonies on LB agar plates containing appropriate antibiotics. After incubation for additional 48 h on agar plates at 30°C, individual colonies were used to inoculate overnight cultures and EGFP fluorescence intensity was measured in cell extracts after OD260nm normalization (green traces). The results indicate that the on-state was maintained in all analyzed colonies. Grey traces are not induced overnight cultures propagated in the off-state. B) Average fluorescence signal of the experiment shown in panel A (green bar: on-state cultures, n=10, error bars indicate s.d.) and off-state for comparison (grey bar: off-state cultures, n=9, error bars indicate s.d.). C) XL1-Blue cells with memory system II were induced with arabinose and 24 h AI competent cells were prepared from this culture. Cells were mock transformed with 5 ng of PCR product and plated on LB agar plates containing kanamycin, ampicillin, and glucose at 30°C. Overnight cultures were inoculated from individual colonies and EGFP expression analysed as described in A). All analysed cultures had maintained the on-state (green traces). Grey traces represent fluorescent signals recorded with off-state cultures containing memory system II. D) Average fluorescence signal of the experiment shown in panel C (green bar: on-state cultures, n=12, error bars indicate s.d.) and off-state negative control (grey bar: off-state cultures, n=6, error bars indicate s.d.).
Supplementary Figure 12. Principle of the methylation analysis of the reporter-maintenance operon promoter. Schematic view of the plasmid region analyzed by quantitative PCR with indicated primers. Prior to qPCR, the plasmid DNA was digested with HinfI, which cleaves DNA at unmethylated GANTC sites. "Methylation" refers to DNA molecules which are not cleaved by HinfI. As an internal standard, a reference amplicon in the kanamycin resistance gene without HinfI restriction site was used.
Supplementary Figure 13. Chemical induction of the memory system II by arabinose addition. A) In the off-state, AraC represses the trigger operon including the trigger MTase (orange CcrM) and mCherry reporter. The reporter-maintenance operon encoded on the memory plasmid is repressed by the ZnF4 repressor, which binds to its unmethylated binding sites (empty lollipops). B) Induction of the on-state by arabinose induction. AraC activates the PBAD promoter, which subsequently leads to the expression of the trigger MTase and mCherry. The trigger MTase methylates the ZnF binding sites on the reporter-maintenance operon (filled lollipops) and hinders ZnF4 binding. The EGFP reporter and the maintenance MTase (yellow CcrM) are expressed. C) Stable perpetuation of the on-state. After switching the medium to glucose containing medium, the trigger operon is repressed again. However, the reporter/maintenance operon is still being expressed due to constant remethylation of the operator sites by the maintenance MTase.
Supplementary Figure 14. Characterization of memory system II response at different arabinose concentrations in the culturing medium after overnight induction. A) EGFP fluorescence measurements of cells, which were cultured overnight in media containing 0.2% (samples 1 and 2), 0.01% (samples 3 and 4), or 0.0005% arabinose (samples 5 and 6), no additional sugar (samples 7 and 8), or 0.2% glucose (samples 9 and 10). B) mCherry measurements of the same samples from panel A. C) Bacterial cultures analyzed in panels A and B were transferred to medium containing 0.2% glucose but no arabinose. Cells were cultured for 6 h and fluorescent measurements for EGFP fluorescence were performed. D) mCherry fluorescence measurements of the same samples as in panel C. The data show that cells cultivated in ≥ 0.01% arabinose switched to the stable on-state. The low mCherry signal indicates that the on-state was maintained by the memory system.
Supplementary Figure 15. Influence of different induction durations on memory system II onswitching. Experiments were conducted with 0.2% arabinose in the medium. A) EGFP fluorescence measurements of cells harboring memory system II which have been cultured with 0.2% glucose (samples 1 and 2). Samples 3-6 were cultivated in medium containing 0.2% arabinose for 2 h (samples 3 and 4) or 4 h (samples 5 and 6). B) mCherry measurements of the same samples as in panel A. C) EGFP measurements of the samples from panel A after cultivation for 6 h in 0.2% glucose. The data reveal switching to the on-state for samples 3-6. D) mCherry measurements of the same samples from panel C. The data show that cells induced for more than 2 h with 0.2% arabinose switched to the stable on-state. The low mCherry signal indicates that the on-state was maintained by the memory system.
Supplementary Figure 16. Growth rates and fluorescence signal of XL1-Blue cells harboring memory system II in the on-state and the off-state. A) Growth rates do not differ between cells in the on-state and cells in the off state and at different culturing times. Cells were cultured in medium under not inducing conditions (0.2% glucose) with appropriate antibiotics in exponential growth phase. Doubling times were calculated from OD600nm measurements. B) EGFP fluorescence signals of the cultures used to determine the growth rates which are shown to confirm their on-and off-state (error bars indicate s.d., n=2).
Supplementary Figure 17. Plasmid map of the DNA damage sensor plasmid used in memory system III. The plasmid is derived from the pBAD24 vector carrying the Caulobacter crescentus CcrM gene. The pBR322 origin of replication of the pBAD24 vector was replaced by the p15A origin of replication. The pBAD promoter and the araC gene were removed and replaced by the colD promoter, which is modified from the promoter region of the cda gene in the E. coli plasmid pColD-157 (Genbank: Y10412.1). As E. coli XL1-Blue cells are recA -, a copy of wild-type recA was introduced downstream of the AmpR gene with its own ribosome binding site creating a polycistronic operon.  Figure 19. Reset of memory system IV by protein degradation. A) Schematic drawing of the states of the reversible memory system. : Initial or induced off-state. : The memory system is switched in the on-state by overnight heat-induction (37°C). After transferring the cells back to 30°C and letting them reach a stable on-state, mf-Lon is induced by arabinose supplementation. : Subsequently, the off-state is reached, leading to a drop in EGFP fluorescence levels. B) EGFP fluorescence data of switching cycles of the reversible memory system using the maintenance CcrM MTase carrying a specific mf-Lon degradation tag. The memory system was induced by heat. The mf-Lon protease (induced by arabinose) was used to degrade tagged CcrM (CcrM-deg) and reset the system. The off-state was also maintained after switching back to glucose containing media, no longer expressing mf-Lon. It was possible to switch the system to the on-state again and subsequently switch it off again by mf-Lon expression in one continuous culture (error bars indicate s.d., n=3). For controls see Supplementary Fig. 20.
Supplementary Figure 20. Signal reset control experiments. A) Switched on memory system with untagged CcrM. The system remains in the on-state during the entire observation time. B) Switched on memory system with tagged CcrM but without mf-Lon induction. The system stays in the on-state as observed in panel A (error bars indicate s.d., n=3). C) EGFP fluorescence data with untagged CcrM maintenance MTase, which does not respond expression of mf-Lon and stays in on-state permanently in cyclic phase switching experiment conducted as described in Supplementary Fig. 19B (error bars indicate s.d., n=3).

Supplementary Tables
Statistical analysis of the data shown in Figs. 1C, 1D, 2B, 2C, 3B and 3C. The indicated p-values refer to the probability of an increase in the fluorescence or DNA methylation signals calculated using a onesided Ttest assuming equal variance. P-values smaller than 0.05 are shaded in grey. n.a. not applicable (because the experimental signal was decreasing).
Supplementary Table 1. Statistical analysis of the data shown in Figure 1.   Supplementary Table 3. Statistical analysis of the data shown in Figure 3.