In our project, we created a plant-based detection system for degradation products of chemical weapons, for which a functional and specific receptor is crucial. We computationally designed a receptor based on a ribose binding protein to bind a given chemical. For the design we utilized Rosetta in combination with EvoDock, a Python script that adds an evolutionary approach, and increased the efficiency of the design process. We experimentally demonstrated the binding of Benzenetricarboxylic acid (BTCA) by the computationally engineered receptor and thereby verified the functionality of our model. This was performed in two approaches, in vitro binding analysis and in vivo by the activation of a signaling cascade resulting in the induced expression of GFP upon specific ligand binding in bacteria. Our design pipeline can be re-applied to engineer receptors for countless applications. Following our detailed workflow descriptions, future iGEM teams are able to design their own specific receptors as well as binding proteins


Wiki Tour

Computational Design

Proof of Protein design

Computational Design


An interchangeable receptor for the detection of compounds in soil and air is essential for our system. By using a modular system, the receptor can be changed easily to create a adaptable plant for detection of chemicals. This allows us to adapt our system for further chemical weapon degradation products, but also it’s adaption according to the needs for further applications. This can be achieved by computational protein design.

We created a adaptable protein design pipeline which enables the reliable design of receptors for a chemical of interest. We decided to use Rosetta as a versatile and powerful tool to simulate docking and to design improved structures. Rosetta is popular, well maintained and open-source tool for protein structure prediction. Its range of functions was expanded from structure prediction to many different areas. The predictions are based on the energy profile of the structures.[1]

To understand the procedure, it is important to note that protein design and most of its associated tasks such as structure prediction are very complex assignments from the computational perspective. Therefore it is not guaranteed the the algorithim produces the optimal structures. However, many algorithms include heuristic steps that often produce good results. To increase the probability to receive good results, most of the runs are repeated several times and the best results are selected.

An advantage of Rosetta is the possibikity of usage usage through many different frameworks. We used “RosettaScripts”, an XML-based format that allows for flexible tool usage[2] as well as the compatibility with EvoDock, a script that adds an evolutionary approach to speed up the design process.

Our receptor engineering approach can be split into the following procedures:

  • Docking
  • Relaxing
  • Design (EvoDock)
  • Docking
  • Relaxing
  • Test Structure
  • Repeat

The individual steps will be described in the following.

Figure 1:Aus[4]


We redesigned one of two binding proteins(coming from E.Coli) that are reported to interact with the published transmembrane protein Trg[3]. We decided to focus on the ribose binding protein, due to its well-known crystallographic structure we knew about the functional relevance of different amino acids.

We started our design process with a systematic in-silico testing of selected chemical weapon related chemicals wwith the wild-type binding protein to check for initial affinities. For its implementation, we simulated the ligand binding. In terms of computational software this is often referred to as docking. We used well established scripts( namely RosettaLigand docking with flexible XML protocol)[5] which proved successful over the past years.

For the docking of each ligand so called “rotamers” are needed. These were generated using the application bcl. Rosetta uses them to figure out how a small molecule can rotate around its bonds, and includes these information in its predictions to transform proteins.

In detail, these rotamers are discrete possibilities of side-chain orientation. A so-called ‘packer’ then chooses the rotamer combinations with the lowest energies. However, through combination, the number of possibilities can quickly overgrow any capability of discrete computation. To solve this problem Rosetta uses Monte-Carlo algorithms, which solve the problem through stochastic predictions. This means that the computations do not have to be exact, but rather they approximate the answer. Rosetta uses the Metropolis-Hastings algorithm.[6]

The Metropolis Criterion states that if the move results in a decrease in energy, it is always accepted, while if it results in an increase in energy, it is accepted with probability equal to e^(-ΔE/(kBT)). Practically speaking, the larger the increase in energy, the lower the likelihood of accepting the move, while at the same time allowing small increases in energy.[7]

In this process, it is possible that difficulties like side chain interactions (like long range interactions) which are not accounted for by the packer or Incorrect movement of the protein backbone. We resolved this problem with the application of the Rosetta “Relax” function. The Relax function repacks the protein structure using a so-called repulsive force field which “wiggles” the atoms into energetic minimal positions.[8]

We selected the amino acid positions which were replaced based on: binding studies with the natural ligand, proximity to the ligand and the protein function. This process was performed using a “packer” defined in the corresponding RosettaScript. Then the design step using EvoDock followed.


To accelerate the design process, we used a tool called EvoDock[9]. It utilizes an evolutionary algorithm that represents amino acids as genes and binding scores as a fitness function. Through a rational combination of possibilities, this allows the program to consider less variants with still good results. An evolutionary algorithm models real evolution by letting different mutants compete for the better result and selecting the winners.

In detail, EvoDock consists of four modules: The Evolution module for keeping track of all mutational variants and their respective scores, and the Mutation, Docking, and Evaluation modules, which perform the modifications on the structure and the energy calculation. As a first step, the Evolution module creates a list of different possible amino acids at the given positions, according to the settings given to the program. The other three modules are then used to generate the fitness scores for those amino acid combinations. Then EvoDock creates new mutants, recombines them and selects the best ones. To determine the fitness, the Mutation module applies the desired mutations to the 3D structure via PyRosetta. The Docking Module adapts the orientation and position of the ligand to the new structure during the docking, followed by relaxing. As a last step, the evaluation module then assigns every mutant a fitness score to be used for the next selection by calculating the binding energy.

To rank the results, the best score for every round will be picked, which results in an iterative improvement in the progress of the evolutionary algorithm. Thereby, scores are based on the free energy of the protein. The lower the score with the bound ligand, the more stable and therefore better is the thermodynamic state of the receptor. However, these scores are not real energies.[10]

The most promising structures were tested in the lab on its functionality.

Figure 2:Thiodiglycol as a pdb file after transforming using the mol to params script.Depiction in PyMol.

Receptor Design Pipeline

To gain insight into our procedure, the process of receptor design is exemplary explained in the following on the design of a receptor for thiodiglycol (Link zu Chemikalien-Seite). All following visualizations were recorded using pymol

First, for a new ligand the structure is obtained from a database. Then conformers need to be generated. The two .sdf files containing the ligand and its rotamers are transferred to a .params and a .pdb file using the tool included in Rosetta.

Figure 3:Dynamic presentation of the Ribose-binding Protein(PDB:2DRI)[]

Receptor Design Pipeline

The wild-type receptor structure is obtained from a database as a pdb file. The compatibility of the file is acchieved by using the script

This ligand structure is then manually joined with the according pdb file of the protein, by concatenating the contents of both files.

The Docking RosettaScript needs to be adjusted so that the docking begins at the binding site of the protein. This position is obtained from crystal structures of the receptor with their natural ligands. In addition, the params and pdb files need to be specified in an option file

The docking is perfomed multiple times and the highest scoring structure is used for the relaxing step. The most important value for evaluating a structure is the overall energy score. In some cases, it can be helpful to have a look at the different scores like bond energies but this only played a minor role for us. Some example scores are listed in a table below.

The relaxing step is repeated multiple times, due to its probabilistic algorithm. That means each run will have a different output, even if the starting point is the same.

After that, the protein structure can be designed using EvoDock. To be available for computation, the ligand has to be added to the pyrosetta collection of chemical residues. This is a database to provide the programm with information about the structure of the residues.

The amino acids to mutate need to be declared in a settings-file with the associated residue number and protein chain identifier. Further parameters can be given on top in the Evodock settings file:

-evo dock settings file: -init-pop-run-mode create-and-evolve 400 -task LIGAND-BINDING -targetscore INFINITE -prot-path input_dock -res-id A 16 -substitutions -res-id A 90 -substitutions -res-id A 164 -substitutions -cpu 24 -mutate MutateByPyRosetta -apply DockByPyRosetta -score ScoreByPyRosetta ...

EvoDock can be varied in population size, number of loops and further parameters in a run file. An overview of the generated files can be seen in the Evodock result file. Here, the different mutants will be ranked.

-evodock results ausschnitt, needs formating: Final Population Size: 100 Best score: 18.17784043092479 | with difference to target score: -18.17784043092479 Number of mutants sharing best score: 1 Number of accepted mutations: 2000 Number of accepted recombination products: 0 Best Score over Time: [17.6691249156097, 17.6691249156097, 18.17784043092479] Average Score over Time: [17.6691249156097, 13.238525017286563, 15.410423328892492] ['C', 'Q', 'W'], 18.17784043092479 ['I', 'I', 'M'], 18.01397636784054 ['L', 'C', 'W'], 17.57748331186076 ['F', 'W', 'W'], 17.463886106465225

Taking the best scoring mutant, an improved receptor version is created as seen in the Figure

The Difference can also be estimated by looking at how the enzyme closes and his binding pocket changes, which is not shown here.

When docked with the previous docking procedures all scores improved greatly.

Figure 4: Docking of Thiodiglycol before and after multiple successive runs of EvoDock. Depiction in PyMol.
Figure 5:Alignment of the Alphafold 2.0 and Rosetta predicted structures. Depiction in PyMol.

Designing a protein can change its three-dimensional structure. Rosetta predicts these changes during its relaxing step. To asses wether the prediction using rosetta can be replicated we compared the Rosetta calculated structure with a Alphafold 2.0 Structure (the slightly simplified Colab Notebook version[11]). Both Structures related very closely to each other with a RMSD of 0,831 Angström.

To verify the results of ligand docking to the designed receptor, we used different tools. The dockings of the mutated receptors with their ligands have been repeated with Autodock Vina[12]. Following through with the previous example, the docking of thiodiglycol with Autodock Vina also showed four out of five amino acids bound The same way the Rosetta docking predicted.

total_score description
-939.530 2DRI_Thiodi_0001
-935.904 2DRI_Thiodi_0002
-937.257 2DRI_Thiodi_0003
-937.679 2DRI_Thiodi_0004
-928.210 2DRI_Thiodi_0005
-936.212 2DRI_Thiodi_0006
-929.683 2DRI_Thiodi_0007
-934.657 2DRI_Thiodi_0008
-925.456 2DRI_Thiodi_0009
-936.127 2DRI_Thiodi_0010
-1.075.210 Mutant_C_Q_W_apl_Thiodi_0001
-1.076.973 Mutant_C_Q_W_apl_Thiodi_0002
-1.074.853 Mutant_C_Q_W_apl_Thiodi_0003
-1.073.510 Mutant_C_Q_W_apl_Thiodi_0004
-1.070.692 Mutant_C_Q_W_apl_Thiodi_0005
-1.068.652 Mutant_C_Q_W_apl_Thiodi_0006
-1.071.511 Mutant_C_Q_W_apl_Thiodi_0007
-1.071.948 Mutant_C_Q_W_apl_Thiodi_0008
-1.073.134 Mutant_C_Q_W_apl_Thiodi_0009
-1.069.683 Mutant_C_Q_W_apl_Thiodi_0010

Proof of functionality of the engineered receptor

We were able to show in vitro the binding of BTCA and in vivo the activation of our signalling cascade trough BTCA. The measured response of our BTCA receptor was higher than the response and thereby binding of the wild-type RBP. This Demonstrates the ability of designing a fully functional and improved receptor using our modeling approach.

Library-creation via Darwin-Assembly

Figure 1: Binding pocket of the Ribose-Binding-Protein (RBP) with Benzenetricarboxylic acid (BTCA) bound. Interactions to surrounding amino acids are marked with dotted lines. Depiction in PyMol.


In addition to our computational design approach towards increasing the affinity of the Ribose-Binding-Protein (RBP) (Part:BBa_K3900001) to our different chemicals, we decided to create a library of in-vitro mutated RBP-sequences. By combining experimental library creation with computational protein design, we want to overcome the limitations of computational engineering and create a RBP with even better binding pocket for our ligand of interest.

For the library creation we decided to use the Darwin-Assembly [13]. This protocol allows the mutation of single base-triplets and therefore is perfectly suited for our aim to mutate the most relevant amino-acids which affect the binding affinity to the ligand within the binding-pocket of the receptor.

We decided to mutate six amino acids involved in ligand binding and formation of the binding pocket to achieve an increased specificity for different chemicals. The most relevant amino acids were determined by analyzing structure predictions in PyMol [14] after the chemicals Diisopropyl methylphosphonate (DIMP) and Benzenetricarboxylic acid (BTCA) were docked seperately using Rosetta. We found amino acids suitable for mutation by analyzing the interaction between ligand and the RBP site focusing on the distance between ligand and binding site as well as the interaction strength towards the most relevant amino acids (Figure 1).

Basic Principle

The Darwin Assembly allows the mutation of single amino acids by introduction of randomized bases. We took the RBP sequence, which we also used as input in the computational protein design and for which we showed that it can indeed bind our ligands of interest strongly after designing it, as basis for the library creation.

Therefore, we cloned the gene into the plasmid pJOE5751.1 (promoter system: Part:BBa_K3900060) by Gibson assembly and created ssPlasmids by nicking one strand with Nt.BspQI and digesting it with Exonuclease III as shown in Figure 2. The RBP-sequence is flanked by a 5´-biotin-marked and a 3´-nuclease and polymerase-protected oligonucleotid that bind the RBP gene on both ends in the same orientation. NNK-oligonucleotids bind at the positions on the RBP sequence where the amino-acids of interest need to be mutated. Using a Phusion polymerase and Taq ligase, the primers are ligated together and many different RBPs with randomly exchanged target triplets are created. Those can be purified using paramagnetic streptavidin beads that bind the biotin at the 5´-end of the fragments [15].

While still being bound to the streptavidin beads the fragments are amplified using PCR. The resulting PCR fragments are then inserted into the plasmid backbone pJOE5751.1 via Gibson Assembly. This plasmid is then transformed into cells that contain the signaling cascade plasmid with an antibiotic-resistance instead of GFP. In the last step positive mutated clones are selected on agar that contains an antibiotic and one of the chemicals we want to test for. Thereby, only mutants with a functional receptor, which is able to bind the given chemical, survive (Figure 2).

Figure 2: The plasmid pJOE5751.1 (backbone black, RBP-sequence red) is nicked by the nicking endonuclease Nt.BspQI (at the cyan dot) and then digested by Exonuclease III, creating a single strand plasmid.(1) Different oligonucleotides of the same orientation bind at the ssplasmid (2.1), complementary fragments are created by the Phusion polymerase and ligated by the Taq ligase (2.2). The 5´-biotin labeled end of the created single strand binds streptavidin coated paramagnetic beads (3) and the fragments are amplified using PCR (4). The double stranded fragments are then cloned into a pJOE5751.1 backbone via Gibson-Assembly (5.1). The library plasmid is transformed into E. coli that contain the signaling cascade plasmid with an antibiotic resistance instead of GFP, harboring a selection system to receive only functional mutants (5.2).

Detailed Description

Oligonucleotide Design

For the Darwin-Assembly we designed different sets of oligonucleotides for different processes of the assembly:

1. RBP-amplification oligonucleotids flanking the sequence directly. Those oligonucleotids are used for amplification of the RBP sequence as well as cloning of the randomized sequence into pJOE5751.1 by Gibson-Assemblies with RBP and the plasmid pJOE. 2. Outer oligonucleotids laying outside of the RBP sequence and anneal in the same orientation. One oligonucleotid is 5´-biotin marked and the second oligonucleotid has a 3´-overhang adenosine and is double phosphorylated as well as phosophorothioat modified. This prevents the whole plasmid from being elongated by the polymerase or digested by nuclease activity[13]. Thereby, the polymerase is prevented from extending the 3´-end by the 3´-end overhang and double phosphorylation. A phosophorothioat modification prevents digestion by exonucleases or the 3´-5´-exonuclease activity of the Phusion-Polymerase. 3. Six mutagenic NNK oligonucleotides for the randomization at specific positions, where structure predictions determined the amino-acids of interest (Figure 1). The NNK-frame codes are N for a random base and K for G/T, to reduce the chance for stop-codon incorporation.

All oligonucleotides were designed to bind the same strand of the plasmid. The mutagenic oligonucleotides as well as the second outer oligonucleotid are phosphorylated at their 3´-end using T4 polynucleotide kinase (PNK) to enable ligation in later steps. Enzymes were inactivated at 80°C and the further steps were performed without purification.

Assembling the initial pJOE5751.1 with RBP for the Darwin Assembly

After amplifying the plasmid pJOE5751.1 with EGFP and the RBP sequence, the RBP sequence was inserted into pJOE5751.1 via Gibson-Assembly replacing EGFP (pJOE5751.1_RBP). The successful cloning step was verified by colony PCR and sanger sequencing.

1. Creation of a single stranded plasmid

One strand of pJOE5751.1_RBP was digested using the enzymes Nt.BspQ1 and Exonuclease III. Nt.BSpQ1 nicks the 5´-3´ strand and Exonuclease III digests it, enabling the binding of the oligonucleotides to the single strand in the following step.

2. Creation of mutated RBP sequences

In this step, the outer and the mutagenic oligonucleotides are added. Altered RBP sequences are created when the randomized oligonucleotides anneal. Each oligonucleotid has a NNK base-triplet at one position resulting in the insertion of a randomized base-triplet. By introducing six randomized codon triples, a library comprise of up to 6^20 different RBP sequence can be created.

After the annealing-process, the gaps in the sequence are closed by a Taq ligase and the fragments amplified using a polymerase.

3. Assembly purification using streptavidin-coated paramagnetic beads

The assembled fragments are purified using streptavidin-coated paramagnetic beads that bind the 5´-biotin of the created library fragment. For purification optimization, two different approaches were chosen:

1. The according Thermo Fisher Scientific protocol for the “Dynabeads™ M-280 Streptavidin” [16]. Instead of E&B buffer, 60 mM NaOH to singularize the DNA-strands was used to meet the purpose of the Darwin Assembly. 2. Since the authors changed the according Thermo Fisher Scientific protocol for use of the “Dynabeads™ MyOne™ Streptavidin C1” [17], this protocol was also tested with the “Dynabeads™ M-280 Streptavidin”.

There was no significant difference in DNA concentration after PCR amplification depending on the protocol used. However, sample concentrations between 18 ng/μL and 30 ng/μL showed that either a major share of the sample was lost during purification or the amplification was not efficient due to other influences.

4. PCR amplification of assembled DNA

While the library fragment is still bound to the paramagnetic beads, the fragments are amplified via PCR using the amplification oligonucleotides. The used polymerase was Phusion, in contrast to KOD Xtreme which was used by the authors of the paper [13].

The complexity of the library is analyzed using sanger sequencing and Next Generation Sequencing. If the library creation would have been successful, an equal bases distribution at the mutated positions of sample would have been expected (Figure 3). In our case, the creation was not successful and the bases can be clearly determined using sanger sequencing (Figure 4).

Figure 3: Alignment of a NNK-mutated library sequences from a sanger sequencing experiment with original sequences [18].
Figure 4: Sequence of our RBP with alignments of sanger sequencing results of DNA library fragments after the amplification step of the Darwin-Assembly.

We assume that step 2 (binding of the oligonucleotides and elongation) was the reason for not getting the expecting sequencing result (Figure 3) since we were able to amplify and sequence the sequence in a later step of the protocol. The reason why the creation of the RBP sequence library failed can only be speculated. The binding of the single stranded fragment to the paramagnetic beads proved, that the biotin-oligonucleotid was annealed, respectively. The protected oligonucleotide also bound as shown by sequencing results that only aligned to the RBP gene. The main reason for the failure of the library creation is suspected to be caused by the not binding, randomized oligonucleotides, resulting in the amplification of the input sequence (RBP without inserted base-triplet mutations).

If the library creation would have been successful, we would have analyzed the variability of the library using Illumina or Oxford Nanopore Sequencing.

5.1 Inserting the Library into pJOE5751.1

For further screening of positive mutants regarding the affinity to the ligand, the library fragments should have been cloned into the plasmid pJOE5751.1 via Gibson-Assembly and transformed into E. coli, already carrying the signal-cascade plasmid. This version of the signal-cascade plasmid codes for an antibiotic resistance gene instead of gfp. In the cloning process, egfp is replaced by the library and therefore an optical screening is facilitated. Clones showing a green phenotype carry the origin vector.

5.2 Library selection using the signaling cascade

If the receptor library protein binds the chemical, the antibiotic resistance is expressed and the bacteria would survive on medium, supplemented with antibiotics. Sequencing of those mutants would then determine which mutations are favorable for the binding of precise chemicals. These results would have been used for further docking simulations and subsequent design cycles:

1. As validation and evaluation of the relevance of our in-silico dockings.

2. As basis for a new design-test-cycle, improved by in silico docking and testing again in vitro and in vivo.

Proof of Protein design

For assessment of our computationally designed PBPs decoupled form our signaling cascade, we overexpressed them in E.coli BL21 DE3 and determined the binding capacity to the respective chemicals via surface plasmon resonance (SPR) spectroscopy. In vitro tesing of computationally designed periplasmic binding proteins.


We chose SPR to analyze the binding capacity of the designed receptor, because it enables the characterization of non-covalent bindings of proteins with a large variety of analytes including proteins, cells, and small molecules without the need for labels, tags or dyes. Additionally, information can be gained regarding the specificity, affinity association and dissociation rate constants[19].
SPR utilizes surface plasmons, that result from the collective oscillation of free electrons of the metals surface upon interaction with an electromagnetic wave. The surface plasmons can be excited with polarized light, with their expansion being dependent on the diffraction index of the surrounding medium. If a molecule is bound on the other side of the metals surface, the diffraction index changes proportional to the concentration of bound molecules. The light, that is reflected on the metals surface is influenced by the changing surface plasmons and thus the change in concentration of bound molecules can be detected by the change in the light’s diffraction angle and is expressed as resonance units (RU)[2,3]. If a protein is immobilized on the metal surface and an analyte, in our case a small molecule, is bound by the molecule, the interaction can be detected and characterized.
Protein expression and purification For protein expression ordered the modeled receptors from IDT and cloned them into pRSETb using Gibson assembly. These expression vectors were then transformed into E. coli BL21 DE3, allowing for an inducible expression of the designed receptors and purification via His-tag affinity chromatography. We inoculated overnight cultures in 10 mL LB, supplemented with 100 μg/mL ampicillin, in a shake flask at 180 rpm and 37°C. The main cultures of 450 mL were inoculated at OD600 of 0.1 and at 180 rpm and 37°C. Upon reaching an OD600 of 0.6, the culture was induced with 0.5mM IPTG and cultivated for 4 hours at 37°C. Harvesting and purification After protein expression the cultures were harvested through centrifugation at 8000x g for 3 min at 4°C. The supernatant was discarded and the pellet either stored at -20°C overnight or resuspended in 1x LEW buffer for purification. For purification, the cells were disrupted in the Precellys 24 Cell homogenizer during three cycles of 30 seconds and 6500 rpm, with the samples being stored on ice for 5 minutes in between the cycles. Subsequently the lysate was centrifuged at 20000x g at 4°C for 20 minutes to separate cell debris from the supernatant. Purification of the proteins from the supernatant was then conducted via IMAC in the Protino Ni-TED gravity flow columns. In this step the instructions of the manual for native protein purification from E.coli were followed. Afterwards, the eluted fractions were rebuffered and concentrated in PBS with Amicon centrifuge filters with a MWCO at 10 kDA. The pellet of the lysate was incubated with 1x LEW Buffer supplemented with 8 M urea to denaturate and solubilize inclusion bodies. The samples were also purified via IMAC and samples of supernatant, flowthrough, wash and elution step were stored for further analysis using SDS-PAGE.


To characterize the expression of the designed receptor, an SDS-PAGE was performed for samples from the purification process for both native and denaturated purification. The samples were mixed with 4x loading buffer and heated to 95°C for 5 minutes. 5 μL of sample volume and marker were then loaded onto the gel, the stacking gel was run at 100 V for 15 minutes and the resolving gel at 200 V for 60 minutes.

Protein quantification

To quantify protein yield for the fluorescence measurements, a Bradford assay was performed for both purified proteins. As reference a BSA standard curve was measured with concentrations from 1 g/L to 0.2 g/L BSA. The samples were prepared in a cuvette with 980 µL RotiQuant solution and 20 µL protein sample diluted 1:2, 1:5 and undiluted. In a photospectrometer the samples were then measured at a wavelength of 595 nm.


To characterize the receptor expression, samples of the supernatant of the centrifuged lysate (S), the flowthrough of the sample applied to the column (F), the wash fraction (W) and all three elution fractions (E1-3) were analyzed using SDS-PAGE. These samples were analyzed for the native purification of the DIMP and BTCA receptor (BTCA-R) and denatured purification (den) of the BTCA receptor.
Figure 1: SDS-gels with samples from native and denatured purification of BTCA-R and DIMP-R.

For BTCA the most prominent band in the elution samples can be seen at approximately 35 kDA which is about the size of our overexpressed receptor, indicating a successful purification through the His-tag affinity chromatography. Comparing the native with the denatured samples of BTCA-R, it becomes apparent, that a sizable amount of protein was insoluble during native purification and remained in the pellet after centrifugation of the cell lysate. The overexpression BTCA-R at 37°C for 4 hours lead to the formation of inclusion bodies. Hence, for further expressions of BTCA-R a cultivation at lower temperatures after induction is recommended, to improve the yield of soluble protein. Native expression of the DIMP receptor did not result in the desired outcome, as the most prominent band in the elution samples is between 55 kDA and 70 kDA with the band at the height of the BTCA-R is more faint. Expression of DIMP receptor will need to be optimized for sufficient results. Protein yield was measured using Bradford assay and resulted in 36,85 μm / L for the BTCA receptor protein 17,48 μm / L for the DIMP receptor protein.

Surface plasmon resonance spectroscopy

Under normal circumstances the detection of a ligand-analyte interaction with a SPR spectroscope needs to be adjusted through preliminary measurements for optimal results, however we only had a limited amount of time to dedicate to this measurement. Owed to that, we were only able to perform one measurement. We conducted our experiment with a Biacore 3000 SPR spectroscope and a CM5 integrated flow cell (IFC) with a gold surface and dextran layer for covalent protein immobilization. The IFC has four microfluidic channels in which samples can be injected and detected independently. For our analysis we used two channels, one for our protein of interest and one as reference. Before sample application, the flow cell was primed with HBS-EP pH 7.4. Before the protein was covalently bound, the ideal pH conditions are explored through applying samples diluted 1:10 in sodium acetate with pH 3; 3.5 and 4. For all pH conditions, measured attachment was sufficient so we chose pH 4, for slightly milder binding conditions. After pH scouting was performed, the surface of the protein channel was activated with 70 μL EDC-NHS, following that 70 μL of 4 μM BTCA receptor in Na-acetate with pH 4.0 were applied. As final binding step, the dextran surface was inactivated by 70 μL of 1M ethanolamin.
The change in RU indicates a sufficient concentration of bound protein, so that our chemical can be applied in the next step. We applied BTCA in concentrations of 0.1 mM, 1 mM and 10 mM for a first assessment of affinity and applied BCTA with concentrations of 2 mM, 4mM, 6mM, 8 mM and 10 mM in the following measurement.
For our measurement we chose to immobilize a high amount of protein because we suspected a weak binding of BTCA to our designed protein and a low change in signal in case of binding. The maximal gain in RU (Rmax) for a 1:1 ligand analyte binding, can be estimated by the ratio of analyte to ligand, which in our case would be 0,006% of the base signal. The signal of the reference channel was distracted from the channel with the bound protein and plotted against the time in figure 2.
Figure 2: Resonance units for injection of BTCA on flow cell with bound BTCA-R, with the reference signal distracted.
In our measurement a considerable drift of the baseline can be seen, indicating a loss of protein during the experiment. The thin peaks before and after are artefacts from the injection and originate from the moving of the pumps whenever another solution is injected. The results of our measurement show a rise in RU at every injection of protein, that is far higher than the anticipated change in RU for a 1:1 binding. To correct this factor while calculating our affinity, we set the signal upon injecting with our highest analyte concentration (10 μL) as Rmax. Since our results can not be fitted for calculating results from rate constants, the affinity is calculated from steady state data with the following equation: Kd = (c*Rmax-c*∆RU)/(∆RU)
∆RU was calculated by substraction of the measuring value of the baseline before injection from the measuring value from the plateau. With this calculation we receive a mean kD of 1,48 μmol / L for our measurement from injection with 4 μM to 10 μM and thereby proved the binding of our modeled BTCA receptor to BTCA. For more reliable results the experiment will have to be optimized and repeated. For further measurement the baseline drift must be minimized. Since the RU signal upon injection with the 2 μM analyte solution does not reach the same heights as for higher concentrations, it suggests, that further experiments should be performed for lower concentrations to avoid high noise signals.


[1]Leman, J. K. et al. Macromolecular modeling and design in Rosetta:recent methods and frameworks. Nature methods 17, 665–680;10.1038/s41592-020-0848-2 (2020).

[2]Fleishman, S. J. et al. RosettaScripts: a scripting language interface to the Rosetta macromolecular modeling suite. PloS one 6, e20161;10.1371/journal.pone.0020161 (2011)

[3]Baumgartner, J. W. et al. Transmembrane signalling by a hybrid protein: communication from the domain of chemoreceptor Trg that recognizes sugar-binding proteins to the kinase/phosphatase domain of osmosensor EnvZ. Journal of bacteriology 176,1157–1163;10.1128/jb.176.4.1157-1163.1994 (1994)

[4]Björkman, A. J. & Mowbray, S. L. Multiple open forms of ribose-binding protein trace the path of its conformational change. Journal of molecular biology 279, 651–664; 10.1006/jmbi.1998.1785 (1998)

[5]Lemmon, G. & Meiler, J. Rosetta Ligand docking with flexible XML protocols. Methods in molecular biology (Clifton, N.J.) 819, 143–155; 10.1007/978-1-61779-465-0_10. (2012).





[11]Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589; 10.1038/s41586-021-03819-2 (2021).

[12]Trott, O. & Olson, A. J. AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. Journal of computational chemistry 31, 455–461; 10.1002/jcc.21334 (2010).

[13] Cosenz, C. & Pinheiro, V. B. Darwin Assembly: fast, efficient, multi-site bespoke mutagenesis. Nucleic Acids Research 48, 51ff (2018).

[14] RCSB, structure of RBP (1URP)

[15] Lambert, K. N. & Williamson, V. M. cDNA Library Construction Using Streptavidin-Paramagnetic Beads and PCR. The Nucleic Acid Protocols Handbook 41, 289-294 (2000).

[16] ThermoFisher Protocol "Dynabeads™ M-280 Streptavidin"

[17] ThermoFisher Protocol "Dynabeads™ MyOne™ Streptavidin C1"

[18] iGEM Bielefeld-CeBiTec 2017

[19] Bakhtiar, R. Surface Plasmon Resonance Spectroscopy: A Versatile Technique in a Biochemist’s Toolbox. J. Chem. Educ. 90, 203–209; 10.1021/ed200549g (2013).

[20]Madriz, L. & Vargas, R. Key aspects of surface plasmon resonance spectroscopy for analytical chemistry applications. JAPLR 7; 10.15406/japlr.2018.07.00259 (2018).

[21] Boken, J., Khurana, P., Thatai, S., Kumar, D. & Prasad, S. Plasmonic nanoparticles and their analytical applications: A review. Applied Spectroscopy Reviews 52, 774–820; 10.1080/05704928.2017.1312427 (2017).

Rosetta Resources

We mostly used scripts based on the tutorials given by the Meiler lab research group (