Logic

New antiparallel and destabilized coiled coil pairs were designed and functionally characterized in mammalian cells.
Coiled coils were combined with split luciferase fragments to design functions with logical negation.

Light and chemically inducible proteases were used as mediators in a functional proof of concept for fast regulated logic gates.

As the main challenge of our project was to create fast responsive synthetic circuits in cells, we sought to implement logic operations based on protein post-translational modification, rather than slower transcriptional activation. The developed set of orthogonal proteases that could additionally be split , provided the modules to implement logic functions, for which we had to design the appropriate framework. An inspiration was provided by the study by Shekhawat et al. in which they presented an in vitro protease sensor using autoinhibited coiled coil Shekhawat2009. The principle of their approach was that the two segments of a split reporter are linked to the coiled coil dimer forming peptides. Dimerization of the two chains is prevented by the presence of an antiparallel coiled coil segment that inhibits the binding of its partner to other CC peptides. Reconstitution is enabled by the proteolytic cleavage of the linker between the coiled coil fused to the split reporter and the autoinhibitory segment, which dissociates and can therefore be replaced by a second coiled coil forming peptide with the second segment of the split reporter (4.12.0).

**Principle of the protease sensor based on autoinhibited coiled coil interactions**

Coiled coil segments can reconstitute the active split reporter after cleavage of the autoinhibitory segment.

We realized that the same design could be adapted for our orthogonal proteases by replacing the cleavage sites with appropriate protease target motif, such as for the orthogonal proteases PPVp and TEVp.

Results

Initial testing

The constructs B:nLuc, cLuc:A, A’:TEVs:B:nLuc and cLuc:A:PPVs:B’2A, which do not possess autoinhibitory segments, were tested for CC binding by measuring luciferase reconstitution. Constructs without protease cleavage sites (B:nLuc, cLuc:A ) were used as a control (4.12.1.). A’:TEVs:B:nLuc and cLuc:A:PPVs:B’2A were tested in the presence of TEVp and PPVp, which cleave off the autoinhibitory CC, resulting in split luciferase reconstitution. Additionally, different ratios of constructs were tested in order to obtain the best luciferase activity (4.12.1.).

**Interactions and protease activated AB coiled coil formation.**

HEK293T cells were transfected with appropriate plasmids, 24 h after transfection cells were lysed and double luciferase assay was performed. (A) B:nLuc and cLuc:A coiled coils constructs fused with split firefly luciferase system spontaneously interact and reconstitute firefly luciferase. (B) A’:TEVs:B:nLuc and cLuc:A:PPVs:B’2A autoinhibitory CCs reconstitute activity of firefly luciferase upon cleavage by TEVp and PPV. Successive luciferase reconstitution is observed only when high amounts of both A’:TEVs:B:nLuc and cLuc:A:PPVs:B’2A.

Results showed that very high amounts of the constructs based on same coiled coil sequences used by Shekhawat et al. Shekhawat2009 (i.e. 50 ng of each) were needed to detect the firefly luciferase signal in mammalian cells (4.12.1.). Therefore, we decided to engineer designed coiled coils from a toolbox, used by the 2009 Slovenian iGEM team Gradisar2011a. In order to design an antiparallel coiled coil-based system applicable for logic operation in living cells we took into consideration the rules that establish the orientation and strength of the affinity of the CCs and designed new coiled coils, expanding on the available collection of orthogonal CC and modeled the ratio of the affinities that are required to obtain the optimal response at low leakage (Coiled coil interaction model).

Further explanation ...

Coiled coils

Alpha-helical segment interaction is a common feature in protein tertiary and quaternary structures, where helices form complexes of two or more coils (4.12.1.2). The most frequent interaction is between two alpha-helices, which form a dimeric coiled coil. Interactions can occur both in the two parallel or antiparallel orientation of the coil pairs Hadley2006. The interaction strength of different coiled coil pairs depends on their amino acid sequence and their structure, which determine the underlying noncovalent forces of attraction and repulsion the helices exert on each other. Understanding the rules that govern the interactions between coiled coils is thus inherently linked to understanding their amino acid sequences Woolfson2005.

Sequences of coiled coils that form interactions have a characteristic seven amino acid repeat, called heptad repeat. The position of each amino acid within a heptad is presented in a unified nomenclature (a,b,c,d,e,f,g). Interaction between two coils occurs on a continuous patch along the side of each alpha-helix with each patch facing the core of the dimer’s interface (4.12.1.2B). The amino acid residues which occupy this strip correspond to the a and d positions of the heptad; they are generally hydrophobic and represent the driving force behind dimerization Woolfson2005. Coiled coils are additionally stabilized by ionic interactions between polar amino acids (Asp, Glu, His, Lys, Asn, Gln, Arg, Ser or Thr) in positions e and g Woolfson2005, Gradisar2011a; while amino acids in positions b, c and f, which are less important to interactions between the helices; contribute to helix stability and solubility.

**Coiled coil structure and schematic representation of heptad repeats**

(A) Structure of coiled coil. Helical projection representing specific coiled coil interactions in (B) parallel and (C) antiparallel orientation

Two alpha-helices that form a coiled coil can interact either in a parallel or in an antiparallel orientation Oakley1998 (4.12.1.2B and C). The orientation of coiled coils is largely determined through interactions between amino acid residues in positions e and g Woolfson2005, Oakley1998. In coiled coils with a parallel orientation, electrostatic interactions form between position g on the first and position e on the second alpha-helix. In coiled coils with an antiparallel orientation, electrostatic interactions occur between g:g’ and e:e’ positions of the two helices Litowski2001. The repeating and predicable nature of these interactions can be used for the rational design of coiled coils Gradisar2011a.
Antiparallel CC orientation allows for fusion of C-termini of N-part of split protein to N-termini of CC via a shorter linker, thereby likely resulting in more efficient reconstitution upon binding with appropriate CC partner. As represented in the wheel helical projection in 4.12.1.2 parallel CC are stabilized by electrostatic interactions g:e’ and e:g’, while interactions between g:g’ and e:e’ positions stabilize antiparallel CC. While CC orientation is mainly influenced by electrostatic interactions specific amino acid residues such as Asn inside CC core can contribute to the orientation as well. Due to polarity of the Asn residue two asparagines prefer interaction with each other rather than with other hydrophobic residues in vicinity such as Leu and Ile. These interactions stabilize the core of intended CC orientation and destabilize the core of CC in the opposite orientation.

Antiparallel coiled coils

In order to compare the reconstitution efficiency of split protein dictated by parallel or antiparallel coiled coil interaction, we prepared fusion proteins with split firefly luciferase where we designed a new antiparallel peptide (AP4) and tested their activity in cells. Antiparallel coiled coils (AP4:P3) worked significantly better than parallel coiled coils (P4:P3) (4.12.4), thus demonstrating that a shorter linker between reporters and dimerizing units helps in the reconstitution of the split protein.

To investigate whether the newly designed antiparallel CC is suited for implementation as logic unit into our system, the constructs nLuc:AP4 and P3:cLuc were compared to the coiled coil cLuc:A and B:nLuc from Shekhawat et al. Shekhawat2009. Measurement of the reconstituted firefly luciferase activity showed that our designed coiled coils provided far higher (~50 fold) signal (4.12.5.), thus proposing the use of this new coiled coils for more sensitive logic gates that functions well in the cellular milieu.

**Comparison of the split protein reconstitution based on two different sets of CCs.**

HEK293T cells were transfected with different amounts of constructs. (A) Previously reported CCs were tested in different B:nLuc to cLuc:A ratios. Luciferase reconstitution can be observed at higher plasmid amounts. (B) nLuc:AP4 to P3:cLuc CCs were tested in different ratios. Luciferase activity was detected even with lower plasmid amounts used. Overall, the comparison between pairs of CCs B:nLuc and cLuc:A to nLuc:AP4 and P3:cLuc showed that our own CCs give a much higher signal, so lower amounts can be used for integration into our whole system.

The system presented by Shekhawat is able to process AND or OR logic functions but not those including negation (such as NOR, NAND etc.) We realized that this type of logic functions could be accomplished by introducing an additional cleavage site between the split reporter and coiled coil segment (4.12.6.1).

Cleavable constructs

Constructs were therefore modified by the addition of TEVp cleavage site (TEVs) between nLuc and AP4 and PPVp cleavage site (PPVs) between P3 and cLuc. This represents a logic NOR gate based on the input signals, represented by TEVp and PPVp.

**Introduction of protease cleavage site between the reporter (effector) and coiled coil segment(s)**

Cleavage sites in between CCs and reporter protein introduces logical negation.

**Optimization of protease and substrate plasmid amounts.**

HEK293T cells were transfected with plasmids for constructs with introduced TEVs and PPVs (substrates) and different plasmid amounts of either PPVp (left) or TEVp (right) protease. Results show that 1:5 ratio of substrate and protease, respectively, was needed to achieve adequate cleavage followed by the decrease in protease activity.

Indeed the system performed nicely (4.12.6.). Using this type of cleavage sites enabled us to design protease-based logic gates NOR, NOT A and NOT B (4.12.7.).

**Design of protease based logic operations NOT, NOT A and NOT B in HEK203 cells.**

HEK293 cells were transfected plasmids for nLuc:TEVs:AP4, P3:PPVs:cLuc, nLuc:AP4, P3:cLuc, TEVp and/or PPV as indicated in graphs. 24 h after transfection cells were lysed and double luciferase assay was performed. (A) Logic gate NOR, where the output signal is active only when none of the input signals are present. (B) Logic gate NOT A, in which output signal is active when none or just B input signals (TEVp) is present.(C) Logic gate NOT B, in which the output signal is active when none or just A input signal (PPV) is present.

Destabilized coiled coils

For implementation of the system with additional logic operations further modifications on our own CCs collection were needed. Analysis of the equilibrium model reveals that the affinity of the autoinhibitory segment should not be too strong, otherwise the inhibition will remain; but should also not be too weak, otherwise the system would be leaky and active already without cleavage. Stability of the coiled coil interaction can be tuned by introduction of non-favorable interactions e.g. by introducing Ala residues at a and d positions Acharya2002. We designed four different destabilized P3 coils by substituting b and c position with polar amino acids and a and d positions of different heptads with Ala residues (4.12.8.).

**Sequence alignment of different coils used to tune the affinity of antiparallel coiled coils.**

We designed different destabilized coils from our coiled coil pair AP4 and P3; Ile and Leu were substituted with Ala at the a and/or d position of the first and/or second heptad of P3mS (a more soluble variant of the original P3).

**P3mS-2A and P3mS were the best autoinhibitory coiled coil constructs.**

(A) In the presence of TEVp the auto inhibitory coil is cleaved off, allowing P3 to dimerize with AP4 and reconstitute the split luciferase. (B) Normalized luciferase activity was compared between samples with and without added TEVp to calculate the fold change of luciferase activity. Out of the four different constructs, the constructs which contained the inhibitory coils P3mS and P3mS-2A worked best, where we observed up to 15 times fold increase with the addition of TEVp.

Those variably destabilized peptides were used as autoinhibitory coiled coil forming segments to test the difference in activity between the uncleaved and TEVp cleaved forms.

To test which one of our four destabilized CCs worked best, all constructs were tested in vivo with and without the presence of TEVp (4.12.9.). We concluded that P3mS and P3mS-2A demonstrated the highest fold increase in the luciferase activity upon the addition of TEVp. The other two constructs showed little to no increase in luciferase activity upon the addition of TEVp, suggesting that the peptides were destabilized too much leading to the leakage in the uninduced form.

Inducible logic

The final test was to investigate if the system could indeed be controlled by two signals at the same time. In order to test this we constructed NOR gate with logic processing (nLuc:TEVs:AP4 and P3:PPVs:cLuc) and inducible components (split PPVp and TEVp inducible by the rapamycin and light, respectively) (4.12.13.).

**Protease-based NOR logic gate regulated by light and rapamycin.**

HEK293 cells were transfected with appropriate plasmids as indicated in the graph. 24 hours after the transfection the cells were induced with light and rapamycin for 15min and after 4 hours, lysed and double luciferase assay was performed.

**Logic gates A and A nimply B regulated by light (input A) and rapamycin (input B) based on protease processing produce correct and measurable output after 15 minutes.**

HEK293 cells were transfected with appropriate plasmids as indicated in graph. 24 hours after the transfection the cells were induced with light and rapamycin for 15 minutes, lysed and double luciferase assay was performed. (A) Logic gate A, in which output signal has a value of 1 when A input signal (TEVp induced with light) is present. (B) Logic gate A, in which output signal has a value of 1 when only A input signal (TEVp induced with light) is present.

The types building blocks that we developed made possible to construct all of the possible 16 two-input logic function based on the proteolysis. To demonstrate this we tested additional logic operations (A and A nimply B), where the response was measured directly after 15 minutes of induction to demonstrate the increased speed of protein-based processing, as shown on 4.12.11..

In both cases the system performed very well, producing clear difference between the active and inactive output states within 15 min after the stimulation by combinations of two signals. Logic function A nimply B is relatively difficult to implement but on the other hand it can be quite useful as for example the signal B may identify the cell type and trigger activation by an external signal only in a selected cell types or cells in a selected state.

The availability of a set of orthogonal proteases as well as an orthogonal coiled coil dimers toolset enabled the construction of a fast complex logic processing circuits. The previous CC toolbox has been further expanded with a strategy of generating antiparallel and destabilized CC. Furthermore, designed system based on split proteases can also be linked to many other input signals such as intracellular calcium increase. An important advance is the adaptation of the system to function in vivo in mammalian cells. Further, reporter as the output signal could be substituted by a split protease, enabling multi-layered processing, or used as an trigger for other cellular processes, such as the release of therapeutics . Therefore we believe that those results represent a valuable foundational advance in synthetic biology.

Team:Slovenia/Protease signaling/Logic