Engineering | Heidelberg

Processing

Characterization of the PHOENICS Circuit Components

We directly validated phosphorylation of the SynSubstrate and confirmed that interaction between SynSubstrate and SynPhosphobinder depends on phosphorylation and increases with SynKinase levels. Finally, we seeked to connect the SH2 domain to a transcriptional reporter and identified promising constructs for future phosphorylation assays.

Iteration 1

Iteration 2

Iteration 3

Figure 1: Schematic representation of SynSubstrate and SynKinase constructs designed the Western Blot.

DESIGN

The PHOENICS circuit relies on the phosphorylation of a tyrosine residue within a synthetic substrate (SynSubstrate) composed of three CD3ζ-derived ITAM repeats. Both the SynSubstrate and the truncated ABL kinase (SynKinase) catalyzing its phosphorylation were adapted from Yang et al. (2025). In this system, phosphorylation occurs through proximity-induced interaction mediated by complementary coiled-coils. SynKinase is a truncated ABL kinase fused to one half of the coiled-coil (ABL-bZipEE) and the SynSubstrate is fused to the other coiled-coil (CD3ζ-bZipRR). This kinase truncation prevents intrinsic binding between the enzyme and the substrate, ensuring that phosphorylation occurs exclusively upon their coiled-coil-mediated proximity. To validate these literature-derived components, we first designed a Western blot experiment to confirm protein expression in the cells and phosphorylation. We fused the SynSubstrate to triple Myc tag (3×Myc) to enable the detection of both phosphorylation and expression via immunostaining in a single experiment.

BUILD

For experimental validation, cells in a 12-well plate were transfected either with the substrate fused to a triple Myc tag (3×Myc) or with both the substrate and truncated ABL kinase carrying the complementary coiled-coil (Fig. 1). 48 hours post transfection, cells were lysed and total protein content was quantified. Equal amounts of protein (17 µg) from each sample were loaded into the wells of an SDS-PAGE gel and transferred on a membrane after running the PAGE. The blot was first probed with primary antibodies specific to phosphorylated tyrosine (pY) within the ITAM motif of the substrate, followed by secondary staining. Subsequently, an anti-Myc antibody was used to confirm the presence of the substrate in the samples.

Figure 2: Western blot verification of SynSubstrate phosphorylation by SynKinase. HEK293T cells were transfected with 60 ng of SynSubstrate either alone or together with 55 ng of SynKinase. Phosphorylation was detected using a phospho-tyrosine-specific antibody. A strong phospho-tyrosine signal was observed only in the presence of SynKinase, confirming kinase-dependent phosphorylation of the SynSubstrate.

TEST

The immunoblot demonstrated the phosphorylation of the SynSubstrate only when SynKinase is co-expressed. In the α-pY staining (top panel), no phosphorylation was observed for the substrate expressed alone (lane 1), whereas a vivid phospho-tyrosine signal was detected only when SynKinase was present (lane 2). This suggests that the tyrosine on the substrate was phosphorylated by the SynKinase. In the α-Myc staining (bottom panel), it is not possible to directly compare the levels of expressed substrate between conditions, as the signal is influenced by residual anti-pY signal due to the reuse of the same secondary antibody without stripping. However, the presence of a distinct Myc band in both lanes demonstrates that SynSubstrate is present in both samples. Importantly, the sensitive anti-pY staining detects phosphorylation only in the SynKinase co-transfection, confirming that the substrate is not phosphorylated when SynKinase is absent (Fig. 2). This confirms that SynKinase specifically mediates tyrosine phosphorylation of the substrate.

LEARN

We were able to confirm that SynSubstrate phosphorylation only occurs when SynKinase is present. This indicates that native cellular kinases do not phosphorylate the substrate, confirming the orthogonality of phosphorylation.

Figure 3: Design of the SpliFAST assay for the fluorescence readout of phosphorylation-mediated SynSub and SH2 interaction. The SynSubstrate, fused to the N-terminal fragment of SplitFAST (RspA[N]), is phosphorylated in the cytosol by SynKinase. The SH2 domain, acting as SynPhosphobinder, is fused to the complementary C-terminal fragment (RspA[C]) of SplitFAST. Upon phosphorylation, SynSubstrate interacts with the SH2 domain, promoting reconstitution of SplitFAST. Addition of the HMBR fluorogen induces fluorescence, which visualizes the phosphorylation-dependent protein–protein interaction and can be quantified by flow cytometry.

DESIGN

After successful validation of the SynSubstrate’s phosphorylation we aimed to investigate the semi-physiological interaction between the phosphorylated substrate and the signal-propagating SH2 domain, also referred to as SynPhosphobinder. It is considered semi-physiological because the pY-SH2 interaction is native, but the phosphorylation is driven by a different cellular kinase, not our SynKinase. We selected a recently published switchable fluorescence complementation reporter, SplitFAST (Tebo and Gautier, 2019). SplitFAST allows rapid and reversible detection of protein-protein interactions. The N-RspA(N) half of the split protein was fused to the SynSubstrate, and the complementary C-RspA(N) to the SH2 (Fig. 3). Only when the phosphorylated SynSubstrate interacts with the SH2 domain, the reporter protein is reassembled. Addition of the fluorogen HMBR triggers fluorescence in the reconstituted complex, visualizing the phosphorylation-mediated interaction.

BUILD

To evaluate the dynamic range of the SynSubstrate and SH2 interaction, cells were transfected with SplitFAST constructs alongside increasing amounts of ABL kinase. 100 ng of both RspA(N)-CD3𝜁-bZipRR and RspA(C)-SH2 were co-transfected. After 48 hours of expression, cells were trypsinized to generate a single-cell suspension, and the HMBR fluorogen was added. The fluorescence from reconstituted SplitFAST complexes was then quantified by flow cytometry, enabling detection of phosphorylation-dependent SynSubstrate and SH2 interactions.

TEST

Flow cytometry analysis revealed a clear shift toward higher fluorescence intensity in cells co-transfected with SynKinase, indicating successful reconstitution of the fluorescence reporter (Fig. 4). This shift demonstrates that phosphorylation of the SynSubstrate enables interaction with SH2, leading to SplitFAST complementation and fluorescence. Without SynKinase, fluorescence remained at background levels, confirming the specificity of the interaction. Additionally, increasing SynKinase expression produced a corresponding increase in fluorescence, indicative of a dose-dependent effect on phosphorylation and dimerization. At the highest SynKinase amounts, fluorescence plateaued, consistent with saturation of SynSubstrate phosphorylation and SH2 binding.

Figure 4: Validation of the phosphorylation-dependent dimerization using SplitFAST.Y-axis description corresponds to the ng amounts of SynKinase (0-250 ng). 48 hours post-transfection, dimerization-dependent SplitFAST fluorescence was quantified by flow cytometry, employing excitation lasers at 488 nm and 561 nm. Emission was recorded at 525 nm for splitFAST (Ligand: TF Lime) and at 610 nm for constitutively expressed mCherry, serving as a transfection control. Data is depicted as the geometric mean of fluorescence +/- SEM with ~30000 single cells.

LEARN

These results demonstrate that phosphorylation induces a specific interaction between SynSubstrate and the SH2 domain (SynPhosphobinder). This provides a strong foundation for extending the SH2 module with additional signal transduction functionalities in future designs.

DESIGN

Figure 5: Improved design of the phosphorylation-dependent transcription activator cleavage assay. Panel A: Final system design; Panel B: Rapalog-induced. Panel A: Upon phosphorylation-induced dimerization SynSubstrate, and SH2 fused to nTEV and cTEV fragments are brought together, restoring TEV protease activity. This enables cleavage of a membrane-tethered transcriptional activator at a TEV site, thereby releasing the activator. The released activator translocates to the nucleus to bind the 5x UAS sequence and drive NanoLuc expression.Panel B: Design of validation of TEV proteolytic activity upon rapalog-induced reconstitution.

Having successfully validated SynSubstrate phosphorylation and the phosphorylation-dependent interaction with the SH2 domain, we sought to translate a phosphorylation event into a quantifiable, high-throughput reporter output. To achieve a cost-effective and scalable readout, we engineered a system in which reporter gene expression is conditionally activated upon the interaction between the phosphorylated substrate and SH2.

We designed a membrane-anchored transcriptional activator coupled to a phosphorylation-dependent cleavage system. Specifically, the C-terminus of the split TEV protease (cTEV) was fused to the Synthetic Substrate (3xCD3ζ), and the N-terminus (nTEV) was fused to the SH2 domain. Upon phosphorylation, the substrate binds SH2, bringing the two parts of the splitTEV protease into proximity and reconstituting its enzymatic activity. The active TEV protease then cleaves the Gal4 DNA-binding domain (Gal4DBD) fused to a VP64 transcription factor (Gal4DBD-VP64) from the membrane tether. The released Gal4DBD-VP64 translocates to the nucleus and activates expression of the nano luciferase (NanoLuc) reporter gene under the 5x UAS binding site with minimal CMV promoter. This system would convert substrate phosphorylation into a measurable luminescent signal (Fig. 5A).

To validate rapalog-dependent reconstitution of TEV protease activity, we first tested whether rapalog-induced dimerization could restore proteolytic function. For this purpose, we fused the cTEV to FRB and the nTEV to FKBP, following designs reported in a recent publication (Vlahos et al., 2022). This construct enables chemically induced proximity between FKBP and FRB upon addition of rapalog, which is expected to bring together the splitTEV fragments and reconstitute enzymatic activity (Fig. 5B).

BUILD

After confirming successful cloning by sequencing, HEK293T cells were transfected with the constructs. 24 hours post-transfection, cells were induced with 200 nM AP21967 (Rapalog) or vehicle control (ethanol). 48 hours after transfection, cells were lysed and subjected to the Nano-Glo assay to measure luminescence. The NanoLuc luminescence readout was normalized to constitutively expressed firefly luciferase (FFLuc) levels present in all conditions to control for variation in transfection efficiency and cell number.

Figure 6: Quantification of reporter expression activation using the Nano-Glo luminescence readout of membrane tethered transcription activator assay. For the Rapa-inducible Split TEV condition, HEK293T cells were transfected 24 hours post-seeding, with FRB-cTEV, FKBP-nTEV, and 5xUAS-NanoLuc plasmids, each at 20 ng per well. The positive control consisted of a construct encoding a fused Gal4DBD-VP64 transcription factor to provide constitutive NanoLuc expression. Additional control groups included a negative control lacking the Gal4DBD-VP64 membrane-tethered construct, a No TEV condition, and a Full TEV construct. All transfection conditions incorporated a FFLuc plasmid to normalize results. Luminescence was measured 48 hours post-transfection with Tecan plate reader. Data are presented as the mean ± SEM from n = 3 technical replicates. Statistical significance was determined with Ordinary one-way ANOVA with Tukey’s multiple comparison test, with a single pooled variance. *p < 0.05, ****p < 0.001.

TEST

The transcriptional reporter assay demonstrated robust performance of the controls, with high luminescence observed in the positive control and no detectable signal in the negative control, confirming the absence of transcriptional leakage. Comparison of the ‘No TEV’ and ‘Full TEV’ conditions revealed no significant difference in reporter activity, indicating that the TEV protease system as configured could not detect phosphorylation-dependent events. Furthermore, induction with rapalog did not produce a significant increase in luminescence in the splitTEV group compared to the uninduced condition, suggesting that rapalog-induced dimerization failed to enable the desired transcriptional activation under these experimental parameters (Fig. 7).

LEARN

Overall, we were satisfied that the NanoLuc reporter showed high expression in the positive control and demonstrated no leakage in the negative control, confirming the reliability of the transcriptional readout. However, to circumvent the potential issues related to membrane embedding and rapid degradation observed with the initial design, we decided to focus on establishing an alternative system as discussed in a second cycle.

Optimization of Phosphorylation Readout Assay

We engineered a robust cytosolic phosphorylation reporter using split Gal4DBD/VP64 transcriptional activators. Signal sensitivity and specificity were enhanced through iterative optimization of SynKinase/SynPhosphatase concentration, nuclear-cytoplasmic localization of reporters, and fine-tuning of component stoichiometry, enabling simple and reliable detection of the substrate phosphorylation level.

Iteration 1

Iteration 2

Iteration 3

Iteration 4

Iteration 5

Iteration 6

Figure 1: Improved design of the phosphorylation-dependent transcriptional reporter circuit. SynSubstrate undergoes phosphorylation in the cytosol by SynKinase and dephosphorylation by SynPhosphatase. Upon phosphorylation, the substrate translocates into the nucleus where it interacts with the nuclear-resident SynPhosphobinder. This interaction facilitates UAS-NanoLuc binding, producing the luminescence signal proportional to substrate phosphorylation.

DESIGN

Building on lessons from the previous engineering cycle, we moved forward with a new design focused on a fully cytosolic system to improve signal robustness and avoid issues related to membrane localization. We continued using the SynSubstrate, consisting of three CD3ζ-derived ITAM repeats, alongside the truncated ABL kinase (SynKinase). Additionally, a truncated form of PTPN1 (SynPhosphatase) was adapted from Yang et al., (2025). In this system, phosphorylation and dephosphorylation are controlled through proximity-induced interactions mediated by complementary coiled-coils. SynKinase is fused to one half of the coiled-coil (bZipEE), and SynPhosphatase is fused to the same bZipEE module. Both enzymes compete for the SynSubstrate, which is fused to the complementary coiled-coil (bZipRR), allowing proximity-dependent phosphorylation and dephosphorylation.

The new transcriptional readout is based on the proximity-driven assembly of a transcriptional activator complex composed of VP64 and the Gal4DBD. In this design, Gal4DBD is fused to the substrate (CD3z-Gal4DBD), while VP64 is fused to the SH2 domain (SH2-VP64, referred to as SynPhosphobinder). Upon phosphorylation of the substrate by the kinase, the SH2 domain binds the phosphorylated substrate, bringing VP64 and Gal4DBD together. This assembled complex translocates into the nucleus, where it activates transcription of the Nanoluciferase (NanoLuc) reporter gene under the 5xUAS repeats (UAS-NanoLuc), producing a luminescent readout of phosphorylation-dependent signaling (Fig. 1).

For both modification 1 and modification 2, we engineered new substrate constructs by inserting a NES to promote nucleocytoplasmic shuttling. In modification 1, the NES was added to the 5' end of the NLS-Gal4DBD-CD3ζ-bZipRR substrate, while in modification 2, the NES was positioned internally between the Gal4DBD and 3xCD3ζ domains. For both designs, we removed the NES sequence from the SH2-VP64 construct, ensuring that SH2-VP64 remains resident in the nucleus to interact with the phosphorylated substrate when it enters (Fig. 5). Instead of normalizing the NanoLuc against the FFLuc their values were plotted separately to independently assess the signal strength of transcriptional readout modifications (NanoLuc signal) and to evaluate cell viability (Firefly signal). This allowed clear comparison of reporter expression and viability across tested conditions.

Figure 6: Quantification of phosphorylation-dependent transcriptional activation using the Nano-Glo luminescence readout. Panel A: Raw NanoLuc; Panel B: Raw FFLuc. HEK293T cells were transfected were transfected 24 hours post-seeding, with the 5xUAS-NanoLuc reporter plasmid and either 15 ng SH2-VP64 plus 15 ng Gal4DBD-CD3ζ, or 25 ng SH2-VP64 plus 5 ng Gal4DBD-CD3ζ per well, as indicated for each condition. In all cases, 25 ng of kinase plasmid was included to drive substrate phosphorylation. All samples were also co-transfected with 10 ng of FFLuc plasmid for normalization of luminescence signals. Luminescence was measured 48 hours post-transfection with Tecan plate reader. Data are presented as the mean ± SEM from n = 3 technical replicates. Statistical significance was determined with Ordinary one-way ANOVA with Tukey’s multiple comparison test, with a single pooled variance. *p < 0.05, ****p < 0.001.

TEST

Analysis of raw NanoLuc values revealed that both construct modifications outperformed the original assay, with modification 1 yielding the highest reporter signal when 25 ng SH2-VP64 and 5 ng Gal4DBD-CD3ζ were co-transfected (Fig. 6). However, the Firefly luciferase cell viability assay indicated increased cellular stress for both modifications relative to the original construct.

LEARN

We learned that, as predicted, redesigning the system so the substrate shuttles between the cytosol and nucleus while SH2-VP64 remains nuclear significantly improved transcriptional readout performance. This spatial arrangement enhanced the efficiency of phosphorylation-dependent reporter activation. To balance the maximized signal and cell viability, we proceeded with modification 1 for future experiments.

Figure 7: Kinase titration experiment for assay modification 1. Quantification of phosphorylation-dependent transcriptional activation using the Nano-Glo luminescence readout. HEK293T cells were transfected 24 hours post-seeding, with increasing kinase plasmid amounts (0.4 to 60 ng), together with fixed amounts of modified reporter assay plasmids: 25 ng Gal4DBD-CD3ζ, 5 ng SH2-VP64, and 10 ng 5xUAS-NanoLuc., along with 10 ng FFLuc. Luminescence was measured 48 hours post-transfection with Tecan plate reader. Data are presented as the mean ± SEM from n = 3 technical replicates.

DESIGN

Having optimized all aspects of the transcriptional readout for substrate phosphorylation, we repeated the SynKinase and SynPhosphatase titration experiment for the best-performing modification 1 of the assay.

BUILD

In the kinase titration experiment, we again transfected the cells with increasing amounts of kinase. In SynPhosphatase titration, each condition contained 20 ng kinase to saturate phosphorylation, with varying SynPhosphatase levels. Amounts of the reporter components were constant: 25 ng SH2-VP64, 5 ng Gal4DBD-CD3ζ, and 10 ng 5xUAS-NanoLuc. 48 hours after transfection, cells were lysed and NanoLuc expression measured by the Nano-Glo assay.

Figure 8: Phosphatase titration experiment for assay modification 1. HEK293T cells were transfected 24 hours post-seeding, with a fixed 20 ng of kinase to saturate the phosphorylation of the substrate, and increasing SynPhosphatase amounts ranging from 1 ng to 20 ng. The same reporter amounts were used: 25 ng of Gla4DBD-Cd3Z, 5 ngSH2-VP64, 10 ng of UAS-NanoLuc, along with 10 ng FFLuc. Luminescence was measured 48 hours post-transfection with Tecan plate reader. Data are presented as the mean ± SEM from n = 3 technical replicates.

TEST

Phosphorylation-dependent reporter activation was already detectable with 0.4 ng of kinase plasmid, and NanoLuc expression reached a plateau at 15 ng of kinase, proving high sensitivity and a broad dynamic range of the readout system (Fig. 8).

In the SynPhosphatase titration, dephosphorylation was visible at a concentration of 3 ng SynPhosphatase , and complete suppression of the phosphorylation signal was achieved with 15 ng SynPhosphatase (Fig. 9). These results confirm that the assay is capable of resolving kinase-mediated 'on' states and SynPhosphatase-induced 'off' states with well-defined thresholds.

LEARN

From this iteration, we learned that the transcriptional readout assay was sensitive even at low kinase levels, with a detectable signal starting at 0.4 ng kinase and plateauing at 20 ng, which we adopted for subsequent experiments to minimize cellular stress. Similarly, phosphatase titration revealed that dephosphorylation was evident at 3 ng SynPhosphatase and complete substrate dephosphorylation occurred beyond 10 ng. This defined the functional dynamic range of our assay for both phosphorylation and dephosphorylation activities.

Figure 9: Test of the dPhosphatase (Panel A) and the dKinase (Panel B) catalytic activity. Panel A: Transfection of 20 ng of kinase to saturate the phosphorylation, and either dead or active phosphatase variant (as stated on figure). Additionally, fixed amounts of modified reporter assay plasmids were transfected: 25 ng Gal4DBD-CD3ζ, 5 ng SH2-VP64, and 10 ng UAS-NanoLuc, along with 10 ng Firefly Luciferase. Panel B: Transfection of 25 ng Gal4DBD-CD3ζ, 5 ng SH2-VP64, 10 ng UAS-NanoLuc, and with 10 ng Firefly Luciferase. On Y axis the increasing amounts (20 ng, 30 ng) of either active SynKinase or catalytically dead SynKinase are labeled.HEK293T cells were transfected 24 hours post-seeding. Luminescence was measured 48 hours post-transfection with Tecan plate reader. Data are presented as the mean ± SEM from n = 3 technical replicates. Statistical significance was determined with Ordinary one-way ANOVA with Tukey’s multiple comparison test, with a single pooled variance. *p < 0.05, ****p < 0.001.

DESIGN

Having our assay fully optimized and characterized, we introduced the final touchups. We observed differences in cell viability, reflected by FFLuc signal, between transfection conditions. Conditions with higher total amounts of coding constructs tended to have reduced viability. Although DNA amounts were equalized using pBlueScript filler plasmid, this did not fully normalize cellular stress due to differences in expressed protein load. To control for this, we cloned catalytically inactive variants of both SynKinase and SynPhosphatase enzymes and tested their impact in the optimized luminescence readout assay, checking if they produced any non-specific signal independent of catalytic activity. In parallel, we cloned enzyme versions without the coiled-coils to see if there is any affinity to the substrate that is not proximity-induced through complementary coiled-coils dimerization.

BUILD

We performed PCR deletions of the catalytic residues of the kinase, followed by KLD circularization to create inactive enzyme variants. To eliminate kinase catalytic activity while preserving the structural integrity of the ABL kinase domain, we introduced two point mutations: K271R and Y393A. Similarly, we generated a catalytically inactive SynPhosphatase variant by PCR deletion and KLD cloning. As a stringent control, we introduced D181A and R221M substitutions in the phosphatase domain. Similarly, we deleted the coiled-coil domains (bZipEE) from both enzymes, resulting in truncated SynKinase and SynPhosphatase variants. This allowed us to evaluate the impact of coiled-coil-mediated proximity on catalytic activity within the assay.

Figure 10: Test of the role of the bZipEE coiled-coil in SynKinase-mediated phosphorylation. HEK293T cells were transfected 24 hours post-seeding, with increasing amounts (0, 10, 20, or 30 ng) of SynKinase either with or without the bZipEE coiled-coil domain, as indicated. Fixed amounts of modified reporter assay plasmids were also included: 25 ng Gal4DBD-CD3ζ, 5 ng SH2-VP64, and 10 ng 5xUAS-NanoLuc along with 10 ng FFLuc for normalization. Luminescence was measured 48 hours post-transfection with Tecan plate reader. Data are presented as the mean ± SEM from n = 3 technical replicates. Statistical significance was determined with Ordinary one-way ANOVA with Tukey’s multiple comparison test, with a single pooled variance. *p < 0.05, ****p < 0.001.

TEST

As expected, conditions with catalytically dead phosphatase (dSynPhosphatase) exhibited significantly higher luminescence compared to those with active phosphatase. This indicates effective dephosphorylation only occurs in the presence of functional SynPhosphatase, confirming that the phosphatase activity was successfully abolished (Fig. 10A). Conversely, conditions containing catalytically inactive SynKinase (dSynKinase) showed negligible signal compared to robust reporter activation with active SynKinase (Fig. 10B) also confirming successful deletion of catalytic activity.

In all tested transfection amounts, the SynKinase lacking the bZipEE coiled-coil domain produced a significantly lower luminescence signal compared to the SynKinase containing the coiled-coil (Fig. 11). This confirms that the great majority of kinase activity in the system requires coiled-coil-mediated proximity for efficient substrate phosphorylation. Notably, a significant increase in signal from 0 to 30 ng of the kinase without coiled-coil was observed, corresponding to a residual background enzymatic activity, but this level remained far below that seen with the fully functional, coiled-coil-containing SynKinase. These results demonstrate that loss of the coiled-coil domain effectively disables the engineered interaction and substrate targeting for the kinase, suppressing most of its activity in the assay.

LEARN

This iteration demonstrated that dSynPhosphatase and dSynKinase, as catalytically dead variants, can now serve as optimal filler constructs to balance metabolic load across conditions, ensuring more consistent cell viability measurements across experimental conditions. Their lack of luminescent signal confirms they do not contribute to phosphorylation or dephosphorylation activity in the assay. Furthermore, the significantly abolished activity of kinase without coiled-coil variants confirms that efficient substrate phosphorylation is strictly proximity-based, mediated by engineered coiled-coil interactions.

Responding

Optimization of PhosphoTEV

We designed and developed a phosphorylation inducible TEV protease variant, based on a split TEV system. The first two iterations focused on validating existing TEV systems working with the NanoLuc-RELEASE reporter. The following four iterations include the engineering of the initial split PhosphoTEV design and the following adaptation of a single chain version. The final iteration shows an application of PhosphoTEV-RELEASE for the secretion of IL-12 as a proof-of-concept for the integration of the system into circuits including receptors and therapeutically relevant output proteins.

Iteration 1

Iteration 2

Iteration 3

To test the IL-12-RELEASE construct, we co-transfected it with the 7AP-PhosphoTEV and either 50 or 70 ng of KORD-ABL, which could induce phosphorylation and thus secretion in dependence of SalB in previous experiments. Cells were induced with SalB 24 hours post-transfection and secreted IL-12 in 100 µL of supernatant was measured through ELISA with a chemiluminescent readout. There was a significant increase with SalB induction for both the 50 and 70 ng ABL-KORD conditions, with a better induction at 70 ng.

LEARN

With this experiment, we could successfully show that we can secrete a therapeutically relevant cytokine in IL-12 and even use it in a ligand-inducible system with KORD-ABL. However, we faced several limitations in the methodology as we did not have the exact material recommended by the manufacturer, for example the horse radish peroxidase substrate. For this reason, and because we could not get a proper standard curve to work, we were unable to exactly quantify the IL-12 concentration in mg/mL.

Sensing — GPCR

Controlling Phosphorylation in a Switch-like Manner Using a GPCR-based Synthetic Receptor

Starting from the PAGER_TF, we engineered switch like synthetic receptors that is able to regulate cellular behaviour by sensitively and specifically phosphorylating the CD3𝜁 of our proessing core module. We established two receptor classes, one sensitive to a small molecule and the other sensing both a small molecule and a peptide ligand in an AND logic gate like manner.

Iteration 1

Figure (4A) shows the SalB titration with constant GFP on GFP-PAGER-ABL. Noticeably, at intermediate SalB concentrations (50 nM; 100 nM), full binary switch-like behaviour can be observed with nonsignificant expression above background when induced with just SalB, but 13.8-fold and 11.6-fold increase in expression upon both SalB and GFP induction, respectively. GFP titration showed significant inducibility down to a concentration of 10 nM GFP (Figure 4B). Expression upon only SalB induction is significantly higher than background for all conditions.

LEARN

Titrating both ligands showed that GFP-PAGER-ABL is able to show full AND logic gate functionality at low SalB concentration. This also highlights that the inducibility of the GFP-PAGER-ABL construct is mainly dependent on SalB, whereas GFP concentration can determine total output but not affect leakiness. Engineering the construct further using an anti-VEGF nanobody changed the Arodyn gating to an extent, where the receptor does not show significant induction in the presence of just SalB until 500 nM (shown in Results and BBa_250PWRGF).

Engineering a GPCR Based Synthetic Receptor Able to Act as a Repressor on a Post-translational Level

By adapting the GPCR scaffold, effectively utilized for inducing phosphorylation upon ligand sensing, we engineered a receptor that is able to directly interact with, and quench our phosphorylation circuit on a fully post-translational level using our SynPhosphatase.

Iteration 1

Iteration 2

Figure 1: Titration of KORD-PTPN1 against KORD-ABL HEK293 T cells were transfected 24 hours after seeding. NanoLuc and Firefly Luciferase expression was measured using the Dual Luciferase assay 24 hours after induction. Induction occurred 24 h after transfection, using either DMSO or SalB [500 nM]. Each condition was transfected with 30 ng of KORD-ABL and 5 ng of 𝛽-Arrestin2-bZipEE. Data is depicted as the mean +/- SEM with n = 3 technical replicates. Two-way ANOVA with Sidak's multiple comparison test. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001; NS, not significant.

DESIGN

To expand our modular receptor system and enable it to directly transduce an inhibitory signal on a post-translational level, we wanted to engineer the KORD-PTPN1: a synthetic GPCR fused with our SynPhosphatase, a truncated version of PTPN1. Similar to the rationale behind the engineering of KORD-ABL, we consciously selected synthetic GPCRs as one of our architectures due to their ability to recruit an intracellular accessory protein for signal transduction instead of relying on dimerization. To develop a modular, but most importantly tightly regulated receptor, we also wanted to adapt the programmable antigen-gated G protein coupled engineered receptor (PAGER) as our scaffold for our inhibitory receptor.

BUILD

To validate the functionality of the SynPhosphatase on a membrane bound receptor, we initially set out to engineer the KORD-PTPN1 variant as a simplified receptor model. The construct was assembled by Golden Gate Assembly from the cytosolic SynPhosphatase and the Lag2-PAGER_TF. After fusion of the PTPN1 phosphatase to intracellular C-terminal tail of KORD, connected by a 22 amino acid linker, KLD was done to excise the extracellular GFP sensing domain. To test the construct for inducible phosphatase activity, it was first titrated against the KORD-ABL construct, as the cytosolic enzyme variants exhibit significantly stronger activity.

TEST

The titration showed that the Synphosphatase is active at the membrane and able to effectively quench KORD-ABL phosphorylation in a dose dependent manner (Figure 1). Equal ratios of KORD-ABL and KORD-PTPN1 result in a 3.0-fold reduction of expressed NanoLuciferase compared to only KORD-ABL.

LEARN

The transfection successfully showed the functionality of our SynPhosphatase when membrane bound, and more specifically the functionality of the KORD-PTPN1 construct. To expand upon this, the truncated PTPN1 can be fused to the GFP-PAGER construct, allowing for a peptide ligand gated, inducible receptor phosphatase.

Figure 2: Validation of GFP-PAGER-PTPN1 functionality on a single cell level HEK293T cells were transfected 24 hours after seeding. Induction occurred 24 hours after transfection. mCherry expression was measured 24 hours after induction using flow cytometry. Excitation lasers at 488 nm and 561 nm were used, and fluorescence was measured at 525 nm and 610 nm for GFP and mCherry, respectively. Induction occurred 24 h after transfection with either DMSO, GFP [1 µM] + SalB [500 nM], or SalB [500 nM] or GFP [1 µM]. Data is depicted as the geometric mean of fluorescence +/- geometric SEM with >20000 single cells.

DESIGN

Expanding on the previous KORD-PTPN1 construct we wanted to facilitate extracellular antigen sensing in addition to SalB. Thus, we wanted to adapt the original PAGER_TF system in a similar manner to KORD. We selected the PAGER platform for its high modularity, which allows for the detection of diverse extracellular antigens through nanobody interfaces. Moreover, given the thorough characterization by Kalogriopoulos et al. (2025), we anticipate that other computationally designed binders could also be integrated into this system.

BUILD

The construct was cloned using Golden Gate Assembly from the original Lag2-PAGER_TF construct. The intracellularly fused transcription factor and AsLOV2 domain were replaced by the truncated PTPN1 also utilized on the KORD-PTPN1 with an identical linker length and make-up. We used a GFP sensing nanobody in order to retain cost-effectiveness and improve sample handling.

TEST

Analyzing the ability of the GFp-PAGER-PTPN1 to dephosphorylate upon SalB + GFP induction was measured at single cell resolution using flow cytometry (Figure 2). Phosphatase activity can be observed by a reduction in mean fluorescence intensity between the negative control (only KORD-ABL) and KORD-ABL with GFP-PAGER-PTPN1. It also appears to be dose dependent, with fluorescence decreasing further upon addition of more GFP-PAGER-PTPN1. However, AND logic gate functionality of the GFP-PAGER-PTPN1 does not appear to be present, with 1.1-fold reduction between SalB and SalB + GFP induction.

LEARN

While we assumed full functionality, and AND logic gate behaviour the construct displayed phosphatase activity without further inducibility from GFP. This could indicate a reduced activation dependent activity or an error with the construct, leading to reduced or abolished GFP sensitivity. Future engineering will focus on changing linker length and rigidity, as well as using differently truncated versions of PTPN1.

Sensing — MESA

Optimization of Kinase-based Gene Expression

We developed rapalog-sensing MESA receptors using kinase-mediated phosphorylation to drive gene expression. Iterative engineering of (i) intracellular circuit architecture, (ii) receptor stoichiometry, (iii) transmembrane domains selection (iv) extracellular/intracellular linker lengths, and (v) signal peptide selection achieved 3.4-fold ligand-dependent induction with improved cell viability.

Iteration 1

Iteration 2

Iteration 3

Iteration 4

Iteration 5

Figure 1: Test of rapalog-induced transcription-factor cleavage. Quantification of reporter activation using the Nano-Glo assay. Cells were transfected with the complete expression-coupled PHOENICS circuit, varying only the receptor configuration as indicated. Neg. ctrl: reporter only. Receptor conditions contained 9:1, 3:1, and 1:1 ratios of TEV protease to transcription factor halves (20 ng total receptor plasmid). NanoLuciferase luminescence values (AU) are normalized by Firefly luminescence (AU) and display mean ± SEM. Fluorescence was analyzed by two-way ANOVA. Medium: DMEM (10% FCS). P values: ns > 0.05; * ≤ 0.05; ** ≤ 0.01; *** ≤ 0.001; **** ≤ 0.0001.

DESIGN

Building on insights from Yang et al., (2025) and direct recommendations from Prof. Dr. Rebecca Wade, we initiated testing of our receptor architecture using the well-established MESA receptor system for the rapalog AP21967. We adapted a construct described by Zhou et al., (2023) in which the transcription factor Gal4-VP64, linked to a TEV cleavage site, was attached to the FKBP receptor half, while a full TEV protease was fused to the FRB receptor half. Upon ligand-induced dimerization, we hypothesized that the TEV protease would cleave the transcription factor, releasing it to activate NanoLuciferase expression, thereby providing a quantifiable readout of receptor dimerization.

BUILD

HEK293T cells were transfected with varying ratios of the FRB and FKBP receptor constructs (9:1, 3:1, and 1:1, with the transcription factor-containing construct in excess). This design was based on the hypothesis that a single TEV protease could cleave multiple transcription factors through cycles of receptor dissociation and re-dimerization, making transcription factor excess potentially advantageous. Twenty-four hours post-transfection, cells were treated with either 200 nM rapalog or vehicle control (ethanol). After an additional 24 hours, NanoLuciferase expression was measured via luminometry and normalized by Firefly luciferase, which was co-transfected on a separate plasmid under a constitutive promoter to control for transfection efficiency.

TEST

Normalized luminescence values revealed that both induced and uninduced signals increased as the transcription factor ratio decreased relative to the TEV protease construct. The highest signal was achieved with equimolar amounts of both receptor halves. A negative control containing only the reporter plasmid (no transcription factor) showed negligible luminescence, confirming assay sensitivity. The positive control, consisting of cytosolic TEV protease (5 ng) and transcription factor-FKBP construct (10 ng), produced the highest luminescence levels, which were comparable to those achieved with equal amounts of receptor constructs.

LEARN

We observed that different ratios of receptor halves strongly influenced overall NanoLuciferase expression and luminescence intensity. However, high background prevented detection of significant fold-change between induced and uninduced conditions, challenging the functionality of this synthetic receptor architecture. Additionally, our hypothesis that excess transcription factor would enhance signal output was disproved, as maximal luminescence occurred with equimolar amounts of both receptor constructs. These results parallel recent findings by Teng et al., (2020) who reported similar inducibility issues with MESA-based synthetic receptors. Their work demonstrated that MESA receptors failed to respond to membrane-permeable rapalog despite the ligand's ability to cross cellular membranes, suggesting that proper receptor localization and membrane environment may be critical for functionality. Multiple studies have indicated potential ER retention of MESA receptors rather than cell surface expression, which could compromise receptor dimerization even with membrane-permeable ligands due to differences in membrane composition, molecular crowding, or trafficking machinery in the ER versus plasma membrane Edelstein et al., (2020). Given the variability in reported MESA receptor localization and the ambiguity of our results, we decided to investigate the subcellular localization of our receptor constructs before proceeding further.

Figure 2: Subcellular localization of sfGFP–Rapalog MESA fusion receptors. Fluorescence microscopy of HEK293T cells expressing sfGFP–FKBP (Rapalog) MESA constructs. Left panel A: DAPI (nuclei); right panel B: sfGFP signal. Cells were transfected with 60 ng FKBP-sfGFP. The fusion construct carries sfGFP at the C terminus, separated from the transmembrane domain by a 12-aa GS linker on the intracellular side. Medium: DMEM (10% FCS). Images acquired at 100× magnification. Scale bars, 10 µm.

DESIGN

To investigate whether faulty ER retention or impaired receptor trafficking explained the absence of ligand-induced response in Iteration 1, we designed a localization experiment. The intracellular domains were replaced with superfolder GFP (sfGFP), creating fluorescent fusion proteins that would reveal subcellular distribution in HEK293T cells.

BUILD

HEK293T cells were transfected with the sfGFP-tagged receptor constructs. Thirty-six hours post-transfection, cells were fixed with paraformaldehyde and nuclei were counterstained with DAPI. Fluorescence imaging was performed at 100x magnification using immersion oil.

TEST

Fluorescence microscopy showed sfGFP signal at the cell periphery in addition to intracellular accumulation, indicating that receptor fusion proteins reached the plasma membrane and were not retained entirely in the ER. While image quality limited precise quantification of membrane versus intracellular distribution, the observed peripheral localization was sufficient to rule out complete trafficking failure.

LEARN

These results confirmed plasma membrane localization of our receptor constructs, ruling out ER retention or other trafficking defects as the cause of failed ligand induction in Iteration 1. Given that receptor localization was not the issue, we hypothesized that the high background signal stemmed from the TEV protease-transcription factor readout architecture itself. Specifically, potential constitutive TEV activity or spontaneous transcription factor release may have obscured ligand-dependent changes. Consequently, we transitioned to our phosphorylation-dependent intracellular signaling system for subsequent iterations.

DESIGN

Figure 3: Rapalog-induced reporter activation by MESA-coupled kinase receptor (CD28/CD28). (A) Quantification of SH2-mediated reporter activation using the Nano-Glo assay. Cells were transfected with the complete expression-coupled PHOENICS circuit, varying only the receptor configuration as indicatead. Neg. ctrl: RAPA–ABL receptor half lacking the bZipEE half. Receptor conditions contained 20, 40, 60, or 80 ng total receptor plasmid. (B) Firefly luciferase values indicating cell viability under the same conditions. Cells were co-transfected with equimolar amounts of a Firefly construct under HSV promoter control. NanoLuciferase luminescence values (AU) are normalized by Firefly luminescence (AU), both are displayed as mean ± SEM. Fluorescence was analyzed by two-way ANOVA. Medium: DMEM (10% FCS). P values: ns > 0.05; * ≤ 0.05; ** ≤ 0.01; *** ≤ 0.001; **** ≤ 0.0001.

Having confirmed membrane localization in Iteration 2, we hypothesized that the absence of ligand-induced response in Iteration 1 resulted from high background signal caused by basal TEV protease activity or constitutive transcription factor release. To address this, we implemented a phosphorylation-dependent signaling circuit requiring multiple conditional interactions for transcriptional activation.
We coupled a truncated, synthetic kinase domain (ABL) to the FRB receptor half and a coiled-coil domain (bZipEE) to the FKBP receptor half. The complementary coiled-coil (bZipRR) was fused to a synthetic ABL substrate (synSub) along with the Gal4 DNA-binding domain (Gal4DBD). The bZipEE-bZipRR pair exhibits moderate affinity, creating a dynamic equilibrium wherein the synSub-Gal4DBD construct shuttles between a membrane-proximal state (bound to receptor-associated bZipEE, enabling ABL phosphorylation) and a cytosolic state (dissociated, allowing interaction with SH2-VP64). Upon ligand-induced receptor dimerization, this proximity-driven phosphorylation creates a docking site for an SH2 domain fused to the VP64 activation domain. The phosphotyrosine-SH2 interaction reconstitutes full Gal4DBD-VP64 transcriptional activity. Nuclear localization signals (NLS) and nuclear export signals (NES) on circuit components facilitate shuttling between the cytosol and nucleus, ensuring that only the fully assembled, phosphorylation-activated transcription factor accumulates in the nucleus to drive gene expression. This multi-layered architecture was designed to minimize background signal through several orthogonal control mechanisms: (i) spatial separation of the kinase and substrate until ligand-induced dimerization, (ii) conditional transcription factor assembly dependent on phosphorylation state, and (iii) dynamic protein localization ensuring signal amplification occurs only when all components are properly activated.

BUILD

HEK293T cells were transfected with varying amounts of FRB-ABL and FKBP-bZipEE receptor constructs (20, 40, 60, 80ng; 1:1 ratio), along with bZipRR-synSub-Gal4DBD and SH2-VP64 constructs (20 ng each, 1:1 ratio). The induction protocol remained the same with rapalog or vehicle control (ethanol) being added twenty-four hours post transfection and NanoLuciferase expression being measured after subsequent twenty-four hours. NanoLuciferase expression was measured and normalized by constitutively expressed Firefly luciferase (HSV promoter). Cells transfected with the full circuit but with the RAPA–ABL receptor half lacking the bZipEE-domain served as a negative control.

TEST

A clear dose-dependent increase in luminescence was observed across conditions: rapalog-induced samples always displayed elevated expression, resulting in fold changes between 1.5x and 1.8x All fold changes with the exception of the 80ng condition, were not significant due to error bars. With increasing transfected plasmid amounts both uninduced and induced luminescence scaled roughly linearly. Admittedly the cell viability decreased to extremely low levels as indicated by Firefly luciferase values, with the exception of the negative control.

LEARN

The phosphorylation-dependent circuit successfully demonstrated ligand-responsive behavior with a clear increase from uninduced to induced conditions, confirming that engineered cells can sense and respond to environmental AP21967 through our synthetic signaling cascade. The negative control validated that transcriptional activation depends on receptor-mediated proximity and phosphorylation. The titration proved that by increasing plasmid amounts signal strength can be enhanced, while also increasing background. This kept the achieved fold changes consistent at 1.5x These results validated our circuit architecture but revealed the need for improved signal-to-noise ratio to achieve robust, significant fold changes after induction. Furthermore the dwindling cell viability challenged the functionality of our circuit. To address this, we focused on optimizing the receptor components themselves, specifically investigating the effects of linker length and transmembrane domain selection on receptor dimerization efficiency and signal transduction in subsequent iterations.

Figure 4: Rapalog-induced reporter activation by MESA-like kinase receptor (CD28/CD28M3). (A) Quantification of SH2-mediated reporter activation using the Nano-Glo assay (Fig. 2). Cells were transfected with the complete expression-coupled PHOENICS circuit, varying only the receptor configuration as indicated. Neg. ctrl: RAPA–ABL receptor half lacking the bZipEE half. Receptor conditions contained 1:9, 1:4, and 1:1 ratios of ABL to bZipEE halves and either 10 or 20 ng total receptor plasmid. (B) Firefly luciferase values indicating cell viability under the same conditions. Cells were co-transfected with equimolar amounts of a Firefly construct under HSV promoter control. NanoLuciferase luminescence values (AU) are normalized by Firefly luminescence (AU), both are displayed as mean ± SEM. Fluorescence was analyzed by two-way ANOVA. Medium: DMEM (10% FCS). P values: ns > 0.05; * ≤ 0.05; ** ≤ 0.01; *** ≤ 0.001; **** ≤ 0.0001.

DESIGN

Building on the validated circuit architecture from Iteration 3, we sought to improve signal-to-noise ratio and reduce variability by optimizing the MESA receptor components themselves. Guided by strategies described by (Yang et al., 2025). and (Edelstein et al., (2020)we systematically implemented three receptor engineering approaches: (i) receptor ratio optimization, (ii) transmembrane domain (TMD) engineering, and (iii) linker length variation.
We replaced the original TMDs with a CD28/CD28M3 heterodimer pair to suppress ligand-independent dimerization and reduce background signal. The receptor ratio was explored based on the hypothesis that a single FRB-ABL kinase could sequentially phosphorylate multiple FKBP-bZipEE-bound substrates through cycles of coiled-coil association and dissociation, suggesting that bZipEE excess might enhance signal output. Finally, we optimized linker lengths between extracellular and transmembrane domains, and between transmembrane and intracellular domains, to balance structural flexibility (enabling coiled-coil dimerization) with proximity (ensuring efficient substrate phosphorylation). Specifically, we implemented 10×GS linkers flanking the TMD for the FKBP-bZipEE construct, and 10×GS (extracellular) and 5×GS (intracellular) linkers for the FRB-ABL construct, compared to the original 12×GS linkers used previously for both receptor halves.

BUILD

HEK293T cells were transfected with varying ratios of FRB-ABL and FKBP-bZipEE receptor constructs: 1:4, 1:9 (bZipEE in excess), and 1:1, at total plasmid amounts of 10 ng or 20 ng per receptor pair. The cytosolic signaling components (bZipRR-synSub-Gal4DBD and SH2-VP64) were co-transfected as in Iteration 3. Twenty-four hours post-transfection, cells were treated with 200 nM AP21967 or vehicle control (ethanol). After an additional 24 hours, NanoLuciferase expression was measured and normalized by Firefly luciferase. As a negative control the full signaling circuit with FRB-ABL lacking the Rapalog receptor bZipEE domain (preventing substrate recruitment) was included.

TEST

Receptor optimization yielded significant improvements in assay performance. Statistically significant fold-changes were observed with reduced variability across replicates. The 20 ng transfection conditions produced approximately 2-fold induction with rapalog treatment across all tested receptor ratios. The 10 ng condition showed lower absolute signal (approximately half the luminescence of the 20 ng conditions) but also reduced background, resulting in fold-changes of 3.4x and 2.7x. The negative control (ABL without bZipEE) showed minimal activation, validating that substrate recruitment to the membrane is essential for circuit function.

LEARN

Systematic optimization of TMD selection and linker lengths successfully improved both signal-to-noise ratio and experimental reproducibility. The CD28/CD28M3 heterodimer effectively reduced ligand-independent background, while optimized linker lengths enhanced kinase-substrate proximity and coiled-coil-mediated recruitment dynamics. The achievement of statistically significant 3.4x fold induction represents a substantial improvement over Iteration 3, validating our receptor engineering strategy. However, cell viability inconsistencies indicated that further optimization of experimental conditions or receptor expression levels could yield additional improvements in subsequent iterations.

Figure 5: Rapalog-induced reporter activation by MESA-like kinase receptors with different signal peptides (Iglam2SP/IgKSP). (A) Quantification of SH2-mediated reporter activation using the Nano-Glo assay (Fig. 2). Cells were transfected with the complete expression-coupled PHOENICS circuit, varying only the receptor configuration as indicated. Neg. ctrl: RAPA–ABL receptor half lacking the bZipEE half. Receptor conditions contained a 1:5 ratio of ABL to bZipEE halves and 30 ng total receptor plasmid. (B) Firefly luciferase values indicating cell viability under the same conditions. Cells were co-transfected with equimolar amounts of a Firefly construct under HSV promoter control. NanoLuciferase luminescence values (AU) are normalized by Firefly luminescence (AU), both are displayed as mean ± SEM. Fluorescence was analyzed by two-way ANOVA. Medium: DMEM (10% FCS). P values: ns > 0.05; * ≤ 0.05; ** ≤ 0.01; *** ≤ 0.001; **** ≤ 0.0001.

DESIGN

Despite achieving statistically significant ligand-induced responses in Iteration 4, persistent variability in absolute signal intensity and cell viability across replicates prompted investigation of additional receptor optimization parameters. We hypothesized that inefficient receptor trafficking or protein folding stress might contribute to experimental inconsistency and reduced cell health. Signal peptides, which direct nascent proteins to the secretory pathway, critically influence membrane protein expression levels and cellular stress. Based on literature indicating that signal peptide selection can profoundly impact receptor surface expression and cell viability, we tested an alternative signal peptide (IgKSP) known for low cellular toxicity. We compared receptor constructs using the original signal peptide with those incorporating IGKSP, maintaining the optimized CD28/CD28M3 TMDs and linker configurations from Iteration 4.

BUILD

HEK293T cells were transfected with FRB-ABL and FKBP-bZipEE receptor constructs at a 5:25 ng ratio (1:5, bZipEE in excess). Receptor variants were tested with either the original signal peptide (Iglam2SP) or the IgKSP signal peptide. Cytosolic signaling components (bZipRR-synSub-Gal4DBD and SH2-VP64) were co-transfected as in previous iterations. Twenty-four hours post-transfection, cells were treated with 200 nM AP21967 or vehicle control (ethanol). After an additional 24 hours, both NanoLuciferase and Firefly luciferase were measured. Firefly luciferase expression served as an internal control for transfection efficiency and cell viability. Parallel assay planning and execution precluded the use of the optimal ratio and transfected amount (1:9 ratio, 10ng total plasmid amount) determined in Iteration 4 of this engineering cycle.

TEST

Signal peptide optimization did not yield substantial improvements in regards to the observed fold changes: We detected a 1.3x Fold change for the old signal peptide (Iglam2SP) and a 2.7x fold change for the newly adapted IgKSP signal peptide. However it must be noted that these fold changes were achieved with the suboptimal ratio and 30ng receptor plasmids transfected instead of the 10ng. The IgKSP-modified receptors demonstrated markedly improved Firefly luciferase values compared to the original constructs, indicating enhanced cell viability and more consistent transfection (right panel). Correspondingly, normalized NanoLuciferase expression showed robust ligand-induced responses with reduced error bars across replicates (left panel).

LEARN

The systematic optimization of signal peptide selection successfully addressed the persistent cell viability and variability issues that limited previous iterations. IGKSP incorporation not only improved cell health—as evidenced by elevated Firefly luciferase expression—but also enhanced overall assay robustness and reproducibility. These results underscore the importance of signal peptide selection in synthetic receptor engineering, particularly for applications requiring consistent, quantitative readouts. The combination of optimized TMDs (CD28/CD28M3), linker lengths, receptor ratios, and signal peptide (IgKSP) established a fully functional, reproducible MESA receptor platform capable of reliable ligand-dependent transcriptional control. This optimized system provides a robust foundation for subsequent applications, including testing with alternative ligands, integration of phosphatase-mediated signal termination, and adaptation to therapeutic contexts requiring precise environmental sensing and cellular response.

Phosphatase-Mediated Quenching of Gene Expression

We engineered phosphatase-fused MESA receptors for ligand-dependent signal termination. Systematic optimization of transmembrane domain configurations across three iterations—testing CD28-CD28, CD28-FGFR1, and CD28-CD28M3 pairs—achieved 2.8-fold rapalog-induced reporter quenching, establishing proof-of-concept for bidirectional cellular control.

Iteration 1

Iteration 2

Iteration 3

Figure 1: Rapalog-induced reporter quenching by MESA-coupled phosphatase receptor. Quantification of reporter quenching using the Nano-Glo assay. Cells were transfected with the complete expression-coupled PHOENICS circuit, varying only the receptor configuration as indicated. Receptors contained CD28 homodimerizing transmembrane domains. Neg. ctrl: reporter only. Receptor conditions contained 1:9, 1:4, and 1:1 ratios of PTPN1 phosphatase to bZipEE coiled-coil construct (20 ng total receptor plasmid). NanoLuciferase luminescence values (AU) are normalized by Firefly luminescence (AU) and display mean ± SEM. Fluorescence was analyzed by two-way ANOVA. Medium: DMEM (10% FCS). P values: ns > 0.05; * ≤ 0.05; ** ≤ 0.01; *** ≤ 0.001; **** ≤ 0.0001.

DESIGN

LEARN

TMD optimization substantially improved assay performance, with the FGFR1-FGFR1 homotypic pair outperforming the CD28-CD28M3 heterotypic pair in both fold-change (3.7× vs. 1.5×) and background suppression. The elevated basal activity of CD28-CD28M3 receptors likely reflects increased constitutive TMD association, consistent with the heterotypic interaction promoting spontaneous clustering independent of ligand. Conversely, the FGFR1-FGFR1 configuration maintained low background while preserving TNF-α responsiveness, suggesting that homotypic FGFR1 TMDs balance weak basal interaction with ligand-induced proximity. Based on these results, we selected the FGFR1-FGFR1 TMD pair at a 1:9 ratio for subsequent assays. Future work would focus on further reducing variability and improving statistical significance through additional parameter refinement, following the multi-dimensional optimization strategy validated in rapalog receptor development.

Additional Modules

Mutational Separation of EnvZ/OmpR Enzymatic Functions

We engineered a bacterial two-component system (TCS) as an orthogonal signaling module. Based on the EnvZ/OmpR pair, separate kinase and phosphatase variants were created to fine-tune OmpR phosphorylation in mammalian cells. Reporter assays confirmed specific activation by the kinase and revealed background binding of unphosphorylated OmpR, guiding further optimization of reversibility and signal strength.

Iteration 1

Iteration 2

Figure 1: Reporter assay of the synthetic EnvZ/OmpR system. HEK293T cells were transfected with 10 ng of OmpR together with the indicated amounts of kinase (EnvZK) and phosphatase (EnvZP) variants. Reporter activity was measured 48 h post transfection using a Dual Luciferase assay, with Firefly luciferase normalized to constitutively expressed Renilla luciferase. Bars represent mean ± SEM, n = 3 technical replicates.

DESIGN

To establish an alternative signaling network alongside our PHOENICS circuit, we explored the use of two-component systems (TCS) as an orthogonal mechanism for intracellular communication. The major advantage of TCSs is their complete orthogonality to endogenous mammalian signaling pathways, making them ideal for modular and interference-free circuit design.
TCSs are widespread bacterial signal transduction systems that enable cells to sense and respond to environmental cues such as osmolarity, pH, or nutrients. They typically consist of a histidine kinase (HK) that detects a stimulus and autophosphorylates on a conserved histidine residue. The phosphate group is then transferred to an aspartate residue on a response regulator (RR), which subsequently activates or represses gene expression by binding specific promoter regions.
To build a synthetic and modular TCS in mammalian cells, we adapted the EnvZ/OmpR system described by Jones et al. (2022). In this design, the natural membrane-bound EnvZ, normally a dimeric enzyme with both kinase and phosphatase activity, was truncated and converted into cytosolic monomers with separate catalytic functions. Two variants were engineered: a kinase variant (EnvZK) that phosphorylates OmpR, and a phosphatase variant (EnvZP) that removes the phosphate group.
Both constructs consist of two Dimerization and Histidine Phosphotransfer domains (DHp; "A") and one Catalytic and ATP-binding domain (CA; "B") to enable modular interactions, a configuration that can be described as an AAB-version. Upon phosphorylation, OmpR acts as a transcriptional activator, inducing expression of a Firefly luciferase reporter under control of an OmpR-responsive promoter, providing a quantitative luminescent readout of signaling activity.

BUILD

Both enzyme variants were cloned following the design of Jones et (2022) as truncated cytosolic monomers carrying the respective catalytic mutations. Cloning was performed using Gibson Assembly to generate the AB-version constructs (residues 222-450; UniProt-ID: P0AEJ4), consisting of one DHp domain and one CA domain.
For both the kinase and phosphatase variants, the duplication and insertion of the DHp domain were achieved through PCR amplification followed by Golden Gate Assembly. The kinase construct additionally required the introduction of two point mutations via PCR. The specific point mutations introduced were T402K and Y287D for the kinase, and D244A and F390L for the phosphatase.

TEST

The reporter assay revealed measurable background activity in the absence of enzymes, indicating that unphosphorylated OmpR contributes to basal expression. Increasing amounts of the kinase variant (EnvZK) led to a dose-dependent rise in reporter activity, consistent with OmpR phosphorylation. Surprisingly, the phosphatase variant (EnvZP) also increased reporter output compared to the no-enzyme condition, suggesting residual kinase activity. Moreover, EnvZP was unable to quench the kinase-induced signal, indicating limited or inefficient dephosphorylation capacity in this variant.

LEARN

From these experiments, we learned that OmpR exhibits measurable background activity even in the absence of upstream enzymes (Kenney & Anand, 2020). This is likely due to its ability to bind DNA in an unphosphorylated state, leading to basal transcriptional activation; a phenomenon also confirmed during our interview with Dr. Jones.
In the same discussion, we received valuable feedback that the AAB-type phosphatase variant can retain residual kinase activity, which aligns with our observation of elevated reporter levels in the EnvZP condition. To address this issue, we decided to adopt an alternative phosphatase design (GCN4-EnvZP), in which two monomers are dimerized via N-terminal coiled-coils to maintain a certain configuration.

Figure 2: Quantification of OmpR-dependent reporter activation by cytosolic EnvZ variants. HEK293T cells were transfected with 30 ng of OmpR together with the indicated amounts of kinase (EnvZK) and phosphatase (EnvZP) variants. The reporter plasmid contains an OmpR-binding site upstream of a Firefly luciferase gene, enabling transcriptional activation upon OmpR phosphorylation. Reporter activity was measured 48 hours post transfection using a Dual Luciferase assay, with Firefly luminescence normalized to constitutively expressed Renilla luciferase. Data are shown as mean ± SEM (n = 3 technical replicates). Statistical significance was determined using one-way ANOVA with Tukey’s multiple comparison test (***p < 0.001, ****p < 0.0001, ns = not significant).

DESIGN

Based on our previous findings, we redesigned the phosphatase component of the EnvZ/OmpR system to overcome the residual kinase activity observed in the AAB-type variant. In this improved version, the phosphatase is expressed as a dimer, in which two monomers are fused via N-terminal GCN4 coiled-coil domains, forcing the DHp domains into a fixed rotational state.

BUILD

Starting from the AB phosphatase backbone, an N-terminal GCN4 leucine-zipper was cloned to enforce dimerization via Golden Gate assembly. To create a single-chain dimer, we tandem-fused two phosphatase monomers by connecting the C-terminus of the first monomer to the N-terminus of the second via a long, flexible GS-linker. The resulting construct encodes an N-terminal GCN4-Phos-Linker-Phos architecture. All phosphatase-point mutations from the previous design were retained in each monomer to preserve the intended catalytic profile.

TEST

Reporter activity was measured using 30 ng of OmpR under varying kinase and phosphatase conditions. In the absence of enzymes, a basal signal confirmed OmpR background activity. Increasing amounts of EnvZK led to a dose-dependent rise in reporter output, with strong activation at 20 ng and maximal signal at 40 ng. Co-expression of the GCN4-phosphatase (EnvZP) with 20 ng kinase barely reduced activity, indicating inefficient dephosphorylation. EnvZP alone had minimal impact on basal levels, confirming that the cytosolic kinase strongly activates.

LEARN

From these experiments, we learned that the redesigned GCN4-Phos variant no longer exhibits residual kinase activity, confirming the effectiveness of the dimeric architecture. However, it still fails to efficiently dephosphorylate OmpR, as reporter activity remained largely unchanged when co-expressed with the kinase. Interestingly, co-expression of GCN4-Phos with OmpR alone resulted in a slightly lower signal than OmpR without enzymes. This effect may arise from the formation of a non-productive complex between OmpR and the phosphatase, transiently sequestering OmpR from the promoter, a behavior consistent with known EnvZ-OmpR interactions reported by Qin et al. (2001).

A Tunable Third Input for the PHOENICS Circuit

Our long-term design vision for the PHOENICS circuit is to integrate Melt40 as a third, tunable input. By coupling Melt40 to an SH2 domain, localization can be modulated not only by phosphotyrosine signaling but also by temperature, enabling transcriptional output only when the positive signal from SH2-mediated recruitment outweighs inhibitory input. Temperature tuning adds an environmental control layer to engineered cell circuits.

Iteration 1

Iteration 2

Iteration 3

Figure 1: Temperature-gated transcription with fused Melt40 and transcription factor (Melt40_Gal4_VP64). At cooler temperatures, Melt40 oligomerizes and retains the fusion at the plasma membrane, preventing DNA binding. Heating to ~40 °C disrupts membrane association; the SV40 NLS drives nuclear import, Gal4DBD engages UAS sites, and VP64 activates mCherry expression. Cooling reverses the cycle. Elements are illustrative and not to scale.

DESIGN

Our long-term design vision for the PHOENICS circuit is to integrate Melt40 as a third, tunable input. By coupling Melt40 to an SH2 domain, localization can be modulated not only by phosphotyrosine signaling but also by temperature, enabling transcriptional output only when the positive signal from SH2-mediated recruitment outweighs inhibitory input. Temperature tuning adds an environmental control layer to engineered cell circuits.

To prototype this concept, we engineered a heat-gated transcription factor that is sequestered at the plasma membrane at baseline temperature, and released to the nucleus upon heating (~40 °C), applicable, for example, in a tumour-associated fever context. We used Melt40, a BcLOV4-derived thermal localization module reported to switch near 40 °C, and fused it to a Gal4DBD–VP64 transcription factor flanked by an SV40 NLS and an NES to promote reversible nuclear shuttling. The transcriptional activation is to be quantified by mCherry flow cytometry, but the system is readily adaptable to alternative reporters (Figure 1).

Figure 2: Fluorescence microscopy of Melt40. HEK293T cells were transfected with Melt40 constructs at 200 ng and treated with three different temperatures for 3 hours: 25°C (left), 37°C (middle), and 40°C (right). For localization visualization of Melt40, we utilized an mCherry fusion construct (red). Nuclei were stained using Hoechst (grey). Images were captured using a Leica fluorescence microscope at 20x magnification—scale bars: 20 µm.

BUILD

Module validation. Before adding the transcription factor (TF), we validated Melt40 localization using fixed-cell fluorescence microscopy at temperatures of 25 °C, 37 °C, and 40 °C. As expected, Melt40 remained membrane-bound at 25 °C, beginning to dissociate and partly enriching in the nucleus at 37 °C, and completely nuclear at 40 °C - confirming the intended thermal translocation behavior used as the actuation basis for the TF.

Fusion assembly. In this configuration, temperature acts as a gating input: below the activation threshold, Melt40 sequesters Gal4–VP64 at the plasma membrane; at elevated temperature (~40 °C), the fusion protein is released to the nucleus, where Gal4, as a DNA binding domain (DBD), binds UAS elements and VP64 drives transcription. We quantified output by mCherry flow cytometry, but the system is readily adaptable to alternative reporters. The constructed architecture resulted in the first Version: NLS_Gal4_BcLOV4_STIM_VP64_NES_NLS, shortly Gal4_Melt40_VP64_NES_NLS.

Figure 3: Flow Cytometry Analysis of Gal4_Melt40_VP64_NES_NLS Construct. Flow cytometry was used to assess gene expression and subcellular localization of the Melt40_TF construct under different conditions. No Melt control groups at 37°C and 40°C show similar geometric mean red fluorescence (RFU), indicating no positive expression and only background signal in the absence of Melt. The fluorescence levels for the control (no Melt) conditions are indistinguishable, suggesting that the construct does not induce transcriptional activity in the absence of Melt. Error bars represent the standard error of the mean (SEM).

TEST

The flow cytometry analysis was performed to evaluate gene expression and subcellular localization of the Melt40_TF construct at two different concentrations (40 ng and 160 ng) and at two temperatures (37°C and 40°C). The geometric mean red fluorescence (RFU) values were measured to quantify the level of reporter gene expression under these conditions

Results. The construct did not activate transcription above background under any of the tested conditions. At 40 ng, geometric means were ~83 RFU (37 °C) vs ~91 RFU (40 °C), comparable to the negative control whiteout Melt. At 160 ng, signals remained low (≈49 RFU at 37 °C vs ≈85 RFU at 40 °C) and again tracked background, indicating no temperature-separated activation. Density plots corroborated the lack of a responding subpopulation (Figure 3).

Notably, when comparing mCherry expression with the expression induced by a fused Gal4DBD-VP64 construct at 37°C and 40°C, only a minor (1.1-fold) increase of reporter expression was detected (Figure 4). Temperature differences alone thus do not induce a relevant increase in reporter expression.

Figure 4: Minimal temperature-induced variation of reporter expression without Melt. Quantification of mCherry reporter expression via Gal4DBD-VP64 transcription factor. Cells were transfected with an mCherry reporter construct under UAS-promoter control and the Gal4DBD-VP64 transcription factor. Transfected cells were incubated for 16h at 37°C or 40°C before measurement. mCherry fluorescence was quantified with flow cytometry. Fluorescence profiles of viable single cell populations are displayed in a ridgeline plot on the left, geometric mean fluorescence (RFU) ± geometric SEM are shown in a barplot to the right. Medium: DMEM (10% FCS).

LEARN

A plausible explanation would be an insufficient nuclear accumulation upon heating due to continuous NES-driven export, despite Melt40 release from the membrane. In other words, our shuttling balance likely over-favored export, preventing Gal4DBD–VP64 from reaching an effective nuclear concentration at 40 °C. A second explanation could be dose-dependent background or toxicity at higher DNA input, which can flatten fold changes.

Figure 5: Flow cytometry analysis and representative fluorescence microscopy image of Gal4_Melt40_VP64. Left, flow cytometry with two transfection concentrations (40 ng, 160 ng) and two temperatures (37 °C, 40 °C). Density plots (log RFU) show complete distributions; the bar chart reports geometric mean red fluorescence (RFU) with SEM. Negative Controls (Without Melt) at 37 °C and 40 °C are comparable and low , indicating background signal. At 40 ng, the construct shows a temperature response: mean RFU at 40 °C is ~1.9× higher than at 37 °C. At 160 ng, the effect is lower(~1.1×), consistent with dose dependent background. Right, a representative fluorescence microscopy image corresponds to the condition 40 °C, with 160 ng Gal4_Melt40_VP64. It shows a bright red signal (mCherry) with nuclear accumulation consistent with elevated fluorescence at 40 °C. Error bars represent the standard error of the mean (SEM) for triplicate samples.

DESIGN

Objective. Improve heat-ON transcription by re-ordering domains so that BcLOV4 (Melt40) sits N-terminal, followed by STIM polybasic, SV40 NLS, Gal4DBD, and VP64. This layout was chosen to firstly maximize membrane sequestration via the Melt40 module at baseline, secondly place the NLS immediately upstream of Gal-4DBD to favor rapid nuclear import on heat release, and lastly to separate Gal4DBD from the actuator to reduce steric hindrance during DNA binding, thereby enhancing activation at physiologically relevant temperatures (37–40 °C).

BUILD

To increase heat-induced nuclear retention, the NES from the first version of the construct was removed. This way, once Melt40 releases from the membrane at ~40 °C, the SV40 NLS can drive efficient nuclear import and retention of Gal4DBD–VP64 in the nucleus. The second version’s Gal4_Melt40_VP64 architectures are therefore, from N-terminal to C-terminal: NLS_Gal4DBD_BcLOV4_STIM_VP64.

TEST

Flow cytometry experiments were performed to evaluate gene expression and subcellular localization of the Melt40_TF construct at two different concentrations (40 ng and 160 ng) and at two temperatures (37°C and 40°C). The geometric mean red fluorescence (RFU) values were measured to quantify the level of reporter gene expression under these conditions.

Result: For the higher concentration (160 ng), a high background at both temperatures with minimal fold change (≈1.1×; 37 °C ≈1502 RFU, 40 °C ≈1666 RFU), indicating low temperature-induced separation at high doses. For the lower concentration (40 ng), clear temperature responsiveness is detected. The mean RFU at 40 °C showed a nearly 2-fold change, higher than at 37 °C (Figure 5).

LEARN

Interpretation: Removing the NES successfully uncovered a heat-ON response at lower DNA input (40 ng), consistent with improved nuclear retention on heating. However, overexpression (160 ng) produced a high background that compresses fold change, likely from a dose-related cellular stress that raises baseline reporter levels.

DESIGN

Objective. Improve heat-ON transcription by re-ordering domains so that BcLOV4 (Melt40) sits N-terminal, followed by STIM polybasic, SV40 NLS, Gal4DBD, and VP64. This layout was chosen to firstly maximize membrane sequestration via the Melt40 module at baseline, secondly place the NLS immediately upstream of Gal-4DBD to favor rapid nuclear import on heat release, and lastly to separate Gal4DBD from the actuator to reduce steric hindrance during DNA binding, thereby enhancing activation at physiologically relevant temperatures (37–40 °C).

BUILD

Optimization. We incorporated the transcription factor with an N-terminal, yielding the NES-free Melt40_Gal4_VP64 . The detailed composition of the final construct resulting in: BcLOV_STIM_NLS_Gal4DBD_VP64. The finalized plasmid encodes the actuator, import cue, and TF in the optimized order to couple heat-driven membrane release to rapid nuclear import and UAS-dependent activation.

Figure 6: Flow cytometry analysis and representative fluorescence microscopy image of optimized Melt40_Gal4_VP64. Left, flow cytometry at two DNA inputs (40 ng, 160 ng) and two temperatures (37 °C, 40 °C). Density plots (log RFU) show full distributions; the bar chart reports geometric mean red fluorescence (RFU) with SEM. The negative control whiteout Melt at 37 °C and 40 °C remains low and comparable, indicating background signal. . The optimized construct displays strong temperature separation: at 40 ng, the mean RFU at 40 °C is ~6.6× higher than at 37 °C; at 160 ng, the 40 °C signal is ~4.9× higher than 37 °C. Right, a representative fluorescence microscopy image (condition 40 °C, with 160 ng Melt40_Gal4_VP64) shows a bright nuclear red signal (mCherry) consistent with the elevated 40 °C fluorescence measured by flow cytometry. Error bars represent the standard error of the mean (SEM) for triplicate samples.

TEST

Results. The optimized construct exhibited strong temperature separation: at 40 ng, the mean RFU at 40 °C was ~6.6× higher than at 37 °C; at 160 ng, the 40 °C signal was ~4.9× higher than 37 °C. The distributional shifts indicate a broad population response rather than a small responding subfraction. A representative fluorescence microscopy image (condition: 40 °C, 160 ng Melt40_Gal4_VP64) shows bright nuclear mCherry, consistent with the elevated 40 °C fluorescence measured by flow cytometry, which supports temperature-gated nuclear accumulation and transcriptional activation (Figure 6).

Figure 7: Mechanism of temperature-gated transcription with Melt40. Schematic of the working model. At cooler temperatures, Melt40 (purple) oligomerizes and is recruited to the plasma membrane via the STIM polybasic segment, sequestering the fused Gal4DBD–VP64 away from DNA. Upon heating (red arrow), Melt disengages from the membrane, and the SV40 NLS drives nuclear import of Gal4DBD–VP64. In the nucleus, Gal4DBD binds UAS sites and VP64 activates transcription of the mCherry reporter, which accumulates in the cytosol (red). Cooling reverses the process (blue arrow), returning the fusion to the membrane and reducing transcriptional output. Elements are illustrative and not to scale.

LEARN

Further optimization, achieved by redesigning the order of the functional parts, resulted in the Melt40_Gal4_VP64 construct, which demonstrated significantly enhanced performance and a clear dose-dependent effect. This dose-dependent profile confirms that the system's dynamic range and output can be finely tuned by plasmid concentration, offering greater flexibility for application-specific needs.

These results collectively demonstrate that the optimized Melt40_Gal4_VP64 construct functions as an effective and tunable temperature-inducible gene expression system, with potential applications in synthetic biology, thermogenetic controls, and precision gene regulation under mild hyperthermic conditions.

Our results establish Melt40 as a reliable heat-ON actuator for nuclear translocation and transcriptional control, and the engineering cycles clarified how expression level, tag balance, and domain order shape performance. To realize our long-term vision, using Melt40 as a third, tunable input in signaling circuits via SH2-guided recruitment, we still need to do further cloning. Specifically, we will construct SH2–Melt40 fusion variants (testing orientation and linkers) and integrate these modules with our Gal4/UAS transcriptional output, allowing temperature to tune response strength only when SH2-mediated positive signals outweigh inhibitory inputs. These additional constructs will enable us to characterize the full temperature-by-phosphotyrosine input matrix and finalize an environmentally tunable, multi-input control layer for the PHOENICS circuits (Figure 7).

Dry Lab

Binder Design and Characterization Pipeline for Novel Targets

This engineering cycle describes how we progressively built a robust computational pipeline for designing, testing, and ranking de novo protein binders for our MESA and GPCR receptors. From initial BindCraft generation and molecular dynamics stability tests to optimized umbrella sampling and thermodynamic corrections, we established a reliable workflow for identifying the most promising binders based on binding affinities for further wet lab validation.

Iteration 1

Iteration 2

Iteration 3

Iteration 4

We calculated the final binding free energies (ΔG°) from the PMF and applied corrections for both volume (ΔG_V) and restraints (ΔG_R), effectively restoring lost degrees of freedom.

TEST

After correction, the results became biologically meaningful and thermodynamically realistic. We could now rank binders by their corrected ΔG° and K_D The corrected affinities were slightly weaker (more positive ΔG° as expected).

LEARN

We learned that strong restraints suppress entropy and can bias the results. Applying corrections gives back the degrees of freedom and ensures realistic thermodynamics, although the corrections can become large when the restraints are too strong, leading to higher, less favorable ΔG°.

Predictive Mathematical Model for Phosphorylation Circuit Design

This engineering cycle outlines our step-by-step process of constructing and refining a mathematical framework to simulate receptor dimerization and circuit behavior in PHOENICS cells. Through five design–build–test–learn iterations, we evolved from an unsolvable analytical model to a high-throughput, prediction pipeline capable of predicting circuit performance and guiding experimental design.

Iteration 1

Iteration 2

Initial tests with dummy parameters confirmed that the pipeline functions as intended. Subsequently, we ran simulations using biologically relevant parameters that reflect the components employed in the wet lab, ensuring that the output is directly applicable to experimental design.

LEARN

This pipeline provides a robust tool to systematically guide the design of PHOENICS cells. By mapping the effects of component combinations and concentrations, it enables informed decisions for tailoring cell architectures to specific experimental needs, streamlining the design process for future wet lab implementation.

Conducting Functional Membrane-bound MESA Receptor Simulations

This engineering cycle outlines our step-by-step process to develop a reliable computational workflow for modeling MESA receptor dimerization. Through multiple design–build–test–learn iterations, we moved from failed AlphaFold predictions to alignment-based structure generation, leading to refined, membrane-embedded coarse-grained simulations of CD28/CD28 and CD28/CD28(M3) transmembrane domains.

Iteration 1

Iteration 2

Iteration 3

Iteration 4

Using PyMOL we translated and rotated the coarse-grained coordinate sets to align the TM helices with the membrane normal and corrected any tilt or rotational offsets (Schrödinger, LLC, 2015). The two receptor halves were positioned with an initial center-of-mass separation of 10 nm to avoid artifactual interactions at t = 0, and these manually oriented systems were simulated for 2 μs under the Martini 3 force field to allow sampling of membrane-mediated approach and association.

TEST

Manual insertion produced well embedded and stably oriented receptor halves throughout the 2 μs trajectories and the simulations ran without the insertion artefacts observed previously. Analysis of the time-dependent center-of-mass distances between receptor halves revealed reproducible differences in baseline membrane-mediated dimerization between the CD28/CD28 and CD28/CD28(M3) pairs, supporting the hypothesis that TM sequence variation modulates ligand-independent association.

LEARN

This final iteration confirmed that manual orientation and placement are required when dealing with complex, asymmetric coarse-grained proteins and that Martini3 CG simulations provide a computationally efficient, robust platform to quantify relative background dimerization tendencies. Moving forward, these results motivate systematic replicate simulations and free-energy calculations to quantify association energetics and support rational design of low-background MESA constructs.

Human Practices

Empowering Young Scientists to Think Translationally

From an expert panel to a regional roadmap. First, we hosted a successful panel discussion that brought together leaders in translation, regulation, and patient advocacy, opening a vital dialogue and educating students on the challenges of bringing research to patients. The success of this event and the audience's call for "what's next?" directly led to our second initiative: the creation of a regional roadmap, a practical guide designed to help students navigate the local innovation ecosystem and take concrete steps with their ideas.

The roadmap will be tested throughout the current academic semester. We will measure the roadmap's success by tracking key metrics:

Quantitative: We will monitor download numbers and see if related innovation workshops are fuller this year compared to previous years.
Qualitative: We will actively collect feedback through a survey.

LEARN

This phase is currently in progress, with the results unfolding over the coming months. Our hypothesis is that by providing a single, comprehensive resource, we will significantly lower the barrier for students to engage with the translational ecosystem. We’ve already received requests from additional institutions to be included in a possible second edition of the guide. The feedback and data we gather will show us whether the roadmap is effective as is, or if a future iteration is needed to improve its content, design, or distribution strategy. The final outcome remains to be seen, and we are eager to learn from our community’s experience.

Shaping Future Innovators

We developed a series of educational formats to make complex synthetic biology concepts accessible to high school students. Beginning with interactive classroom workshops on signaling pathways, we expanded to a Summer School where students gained hands-on laboratory experience. To ensure lasting impact and broader accessibility, we transformed these materials into an interactive online learning platform for students and educators worldwide.

Furthermore, we consulted multiple high school teachers and university educators with our work to get their insights. We received only positive feedback, and some of the high school teachers even plan to integrate parts into their own lesson curricula. Our learning platform is open for students and educators worldwide and provides self-learning material for engaged students. Also, future iGEM teams can contribute to the platform, further expanding it with additional modules.

LEARN

Students responded positively to the learning platform and provided feedback requesting more pictures and illustrations. We incorporated this feedback into the current version of the platform, which is now available on our wiki. Nevertheless, it is important to acknowledge that the learning platform alone will not reach the students effectively. It must be supported by motivated educators who can spark students’ interest in synthetic biology.

Abramson, J., Adler, J., Dunger, J., Evans, R., Green, T., Pritzel, A., Ronneberger, O., Willmore, L., Ballard, A. J., Bambrick, J., Bodenstein, S. W., Evans, D. A., Hung, C.-C., O’Neill, M., Reiman, D., Tunyasuvunakool, K., Wu, Z., Žemgulytė, A., Arvaniti, E., … Jumper, J. M. (2024). Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature, 630(8016), 493–500. https://doi.org/10.1038/s41586-024-07487-w
Aho, N., Groenhof, G., & Buslaev, P. (2024). Do all paths lead to Rome? How reliable is umbrella sampling along a single path? Journal of Chemical Theory and Computation, 20(15), 6674–6686. https://doi.org/10.1021/acs.jctc.4c00134
Daringer, N. M., Dudek, R. M., Schwarz, K. A., & Leonard, J. N. (2014). Modular extracellular sensor architecture for engineering mammalian cell-based devices. ACS Synthetic Biology, 3(12), 892–902. https://doi.org/10.1021/sb400128g
Edelstein, H. I., Donahue, P. S., Muldoon, J. J., Kang, A. K., Dolberg, T. B., Battaglia, L. M., Allchin, E. R., Hong, M., & Leonard, J. N. (2020). Elucidation and refinement of synthetic receptor mechanisms. Synthetic Biology, 5(1), ysaa017. https://doi.org/10.1093/synbio/ysaa017
Govind Kumar, V., Polasa, A., Agrawal, S., Kumar, T. K. S., & Moradi, M. (2023). Binding affinity estimation from restrained umbrella sampling simulations. Nature Computational Science, 3(1), 59–70. https://doi.org/10.1038/s43588-022-00389-9
Jones, R. D., Qian, Y., Ilia, K., Wang, B., Laub, M. T., Del Vecchio, D., & Weiss, R. (2022). Robust and tunable signal processing in mammalian cells via engineered covalent modification cycles. Nature Communications, 13(1), 1720. https://doi.org/10.1038/s41467-022-29338-w
Kalogriopoulos, N. A., Tei, R., Yan, Y., Klein, P. M., Ravalin, M., Cai, B., Soltesz, I., Li, Y., & Ting, A. Y. (2025). Synthetic GPCRs for programmable sensing and control of cell behaviour. Nature, 637(8044), 230–239. https://doi.org/10.1038/s41586-024-08282-3
Kenney, L. J., & Anand, G. S. (2020). EnvZ/OmpR two-component signaling: An archetype system that can function noncanonically. EcoSal Plus, 9(1), ESP-0001-2019. https://doi.org/10.1128/ecosalplus.esp-0001-2019
Kroon, P. C., Grünewald, F., Barnoud, J., van Tilburg, M., Brasnett, C., Souza, P. C. T., Wassenaar, T. A., & Marrink, S.-J. (2025). Martinize2 and Vermouth: Unified framework for topology generation (No. arXiv:2212.01191). arXiv. https://doi.org/10.48550/arXiv.2212.01191
Levintov, L., Gorai, B., & Vashisth, H. (2024). Spontaneous dimerization and distinct packing modes of transmembrane domains in receptor tyrosine kinases. bioRxiv, 2024.05.09.593448. https://doi.org/10.1101/2024.05.09.593448
Lindorff-Larsen, K., Piana, S., Palmo, K., Maragakis, P., Klepeis, J. L., Dror, R. O., & Shaw, D. E. (2010). Improved side-chain torsion potentials for the Amber ff99SB protein force field. Proteins, 78(8), 1950–1958. https://doi.org/10.1002/prot.22711
Lu, C., & Wang, Z.-X. (2017). Quantitative analysis of ligand-induced heterodimerization of two distinct receptors. Analytical Chemistry, 89(13), 6926–6930. https://doi.org/10.1021/acs.analchem.7b01274
Pacesa, M., Nickel, L., Schellhaas, C., Schmidt, J., Pyatova, E., Kissling, L., Barendse, P., Choudhury, J., Kapoor, S., Alcaraz-Serna, A., Cho, Y., Ghamary, K. H., Vinué, L., Yachnin, B. J., Wollacott, A. M., Buckley, S., Westphal, A. H., Lindhoud, S., Georgeon, S., … Correia, B. E. (2025). One-shot design of functional protein binders with BindCraft. Nature, 1–10. https://doi.org/10.1038/s41586-025-09429-6
Qin, L., Yoshida, T., & Inouye, M. (2001). The critical role of DNA in the equilibrium between OmpR and phosphorylated OmpR mediated by EnvZ in Escherichia coli. Proceedings of the National Academy of Sciences of the United States of America, 98(3), 908. https://doi.org/10.1073/pnas.031383098
Schrödinger, LLC. (2015). The PyMOL Molecular Graphics System, Version 1.8.
Souza, P. C. T., Alessandri, R., Barnoud, J., Thallmair, S., Faustino, I., Grünewald, F., Patmanidis, I., Abdizadeh, H., Bruininks, B. M. H., Wassenaar, T. A., Kroon, P. C., Melcr, J., Nieto, V., Corradi, V., Khan, H. M., Domański, J., Javanainen, M., Martinez-Seara, H., Reuter, N., … Marrink, S. J. (2021). Martini 3: A general purpose force field for coarse-grained molecular dynamics. Nature Methods, 18(4), 382–388. https://doi.org/10.1038/s41592-021-01098-3
Teng, F., Cui, T., Zhou, L., Gao, Q., Zhou, Q., & Li, W. (2024). Programmable synthetic receptors: The next generation of cell and gene therapies. Signal Transduction and Targeted Therapy, 9(1), 7. https://doi.org/10.1038/s41392-023-01680-5
Vlahos, A. E., Kang, J., Aldrete, C. A., Zhu, R., Chong, L. S., Elowitz, M. B., & Gao, X. J. (2022). Protease-controlled secretion and display of intercellular signals. Nature Communications, 13(1), 912. https://doi.org/10.1038/s41467-022-28623-y
Wassenaar, T. A., Ingólfsson, H. I., Böckmann, R. A., Tieleman, D. P., & Marrink, S. J. (2015). Computational lipidomics with INSANE: A versatile tool for generating custom membranes for molecular simulations. Journal of Chemical Theory and Computation, 11(5), 2144–2155. https://doi.org/10.1021/acs.jctc.5b00209
Webb, B., & Sali, A. (2016). Comparative protein structure modeling using MODELLER. Current Protocols in Bioinformatics, 54, 5.6.1–5.6.37. https://doi.org/10.1002/cpbi.3
Yang, X., Rocks, J. W., Jiang, K., Walters, A. J., Rai, K., Liu, J., Nguyen, J., Olson, S. D., Mehta, P., Collins, J. J., Daringer, N. M., & Bashor, C. J. (2025). Engineering synthetic phosphorylation signaling networks in human cells. Science, 387(6729), 74–81. https://doi.org/10.1126/science.adm8485
Zhou, J., Ge, Q., Wang, D., Guo, Q., & Tao, Y. (2023). Engineering a modular double-transmembrane synthetic receptor system for customizing cellular programs. Cell Reports, 42(4), 112385. https://doi.org/10.1016/j.celrep.2023.112385

Wet Lab

Dry Lab

HP

Processing

Characterization of the PHOENICS Circuit Components

DESIGN

BUILD

TEST

LEARN

DESIGN

BUILD

TEST

LEARN

DESIGN

BUILD

TEST

LEARN

Optimization of Phosphorylation Readout Assay

DESIGN

BUILD

TEST

LEARN

DESIGN

BUILD

TEST

LEARN

DESIGN

BUILD

TEST

LEARN

DESIGN

BUILD

TEST

LEARN

DESIGN

BUILD

TEST

LEARN

DESIGN

BUILD

TEST

LEARN

Responding

Optimization of PhosphoTEV

DESIGN

BUILD

TEST

LEARN

DESIGN

BUILD

TEST

LEARN

DESIGN

BUILD

TEST

LEARN

DESIGN

BUILD

TEST

LEARN

DESIGN

BUILD

TEST

LEARN

DESIGN

BUILD

TEST

LEARN

Sensing — GPCR

Controlling Phosphorylation in a Switch-like Manner Using a GPCR-based Synthetic Receptor

DESIGN

BUILD

TEST

LEARN

DESIGN

BUILD

TEST

LEARN

DESIGN

BUILD