Team:Freiburg/Results
Mirror-Image Phage Display
As an extension of classical phage display, mirror-image phage display (MIPD) was developed to identify not only L- but also D-peptide ligands. MIPD is based on the principle that an occurring interaction between a D-peptide ligand and an L-target protein, would also exist between the mirror-image L-peptide ligand and the D-version of the target protein.3
We aimed to utilize the power of this innovative method to identify therapeutic D-peptide ligands to tackle a current crisis of utmost importance to global health: antibiotic resistance.
Choosing the right phage display system and library depends on various factors. For our MIPD we used the M13 phage that displays the random peptide on its N-terminus of the pIII coat protein which provides a relatively low avidity resulting in high selectivity. Following recommendations regarding our short target PSMα3, we used a 12-mer phage display library kit from New England Biolabs (NEB).2 This kit contains a library of 1011 filamentous, non lytic M13 phages with 109 different clones displaying a random 12-mer peptide fused to the N-terminus of the pIII coat protein.
Fig. 1: Schematic illustration of the filamentous bacteriophage M13 with a circular single-stranded DNA. Gene VIII (pVIII) codes for the major structural protein of bacteriophage particles. Gene III (pIII) codes for the minor coat protein. Our phage library is characterized by an N-terminal fusion of a random 12-mer peptide to the pIII coat protein, represented as red star.
Establishment of phage display
Having obtained a D-target for our experiment we proceeded to establish MIPD in our lab. We decided to begin with a regular phage display, therefore we used biotinylated phytochrome B (kindly provided by Vincent Idstein) in its L-form as target protein which is a known photoreceptor in plants.14
We performed a first round of panning consisting of three parallel approaches: one is surface panning, the next is in solution panning and lastly a negative control without the target. After the first round of selection and amplification we titered the purified phages. For titering we made sequential 1:10 dilutions of the purified phages, infected E. coli and streaked them out on LB agar plates with IPTG and X-Gal. Phage-infected E. coli were visible as blue plaques within the bacterial lawn, the blue color resulted from the lacZ reporter system. From the number of blue plaques we could calculate the titer of the phage solution.
Fig. 2 Dilution row of phages visible as blue plaques on bacterial lawn
Titering is essential to determine the right amount of amplified phage eluate for the next round of panning. According to the manufacturer, it is advised to use at least an input titer of 109 phages for each round of panning. However, the first panning against phytochrome B (phyB) only yielded a maximum of 105 phages, which was not sufficient for a second round of panning.
Fig. 3 Comparison of the amount of phages after first round of selection and amplification. Red shows the required amount of phages for another round of selection
To increase the amount of phages we amplified the phages again according to the same protocol. The titer was significantly higher and exceeded 1012 phages per mL. So we performed a second round of selection with this eluate. However, after the selection there were no phages detectable in titering experiments, not even after repetition of the second panning.
In further repetitions of this experiment we always encountered the same problem of too low phage yields after amplification. To solve this problem we decided to optimize the amplification step before continuing with MIPD experiments.
Optimization of phage amplification
To optimize the amplification of the eluted phages we varied different parameters. For the optimization process, we used 104 phages from the original library without target incubation and panning procedures.
At first, we analyzed the effect of a longer amplification time and the effect of the E. coli OD600 when adding the phages for infection (Fig. 4). For further experiments we adjusted the protocol to an OD600 = 0.5 when infecting E. coli. Additionally, loss of phages by the two consecutive polyethylene glycol (PEG) precipitation step, for phage purification, was investigated (Fig. 5).
Fig. 4 Different amplification times of phages and different ODs of E.coli cultures were tested
Fig. 5 The total amount of phages did not increase with longer amplification time. On the other hand, the purification yield of phages could be improved by using only a single PEG precipitation instead of two.
During the process of establishment and optimization of our phage display we received lots of detailed advice from Dr. Hanna Wagner, our local expert on phage display techniques. Especially she recommended a short period of preincubation without movement, to improve infection of E.coli by the phages. By implementing all her recommendations as well as our own experiences, we successfully improved our phage display protocol. To avoid artificial enrichment of single phage clones we reduced the amplification time to 4.5 h.
Fig. 6 Amplification of 104 phages at 37°C with and without 30 min preincubation of phages with E. coli show a difference of a factor 10. Red line shows the required amount of phages for another round of selection
|
Manufacturer's protocol |
Optimized protocol |
|
|
OD600 starting culture |
0.01-0.05 |
0.5 |
|
Preincubation of phages with E. coli without shaking |
|
30 min |
|
amplification time |
4.5 h |
4.5 h |
|
PEG precipitation |
2x |
1x |
Table 1: Comparison of phage amplification protocols
As an alternative to the classical MIPD approach relying on a surface bound target protein we also applied in solution panning in direct comparison. We decided to use in-solution panning to increase the diversity of potential phage-target interactions by relocating the site of interaction.
Instead of limiting the binding to a fixed target protein, tightly coated and orderly aligned to a surface, we incubated the target protein with the phage library in solution (Fig. 8). To remove unbound phages, this solution was subsequently applied on a streptavidin-coated surface. Relying on the highly affine biotin-streptavidin interaction all bound phages stay attached while all the unbound phages are removed in several washing steps. After this additional incubation step the panning is continued. This in-solution approach allows a more diverse spectrum of interactions and mimics the natural state of the target protein in regard of accessible binding sites.
Fig. 7: The first step of biopanning in phage display is the addition of phages to the immobilized protein of interest.
Fig. 8: Two alternative panning strategies: surface and in solution panning. Left: For surface panning the phage library is added to a target-coated plate and the binding occurs on the surface of a well. For in solution panning the phages are preincubated with the target and pulled down via biotin/streptavidin interaction as a second step.
MIPD against synthetic peptide IZN24
After deciding to work with D-peptides we got in contact with Michael Kay from the University of Utah, who is an expert on mirror-image peptides and proteins. He kindly provided us the peptide IZN24 in L- and D-form. IZN24 is a synthetic peptide that mimics the presentation of specific sequences within gp4113. IZN24 is a part of the Ebola virus coat protein which mediates the entry of the virus into cells. Because the structure containing IZN24 plays a major part in the membrane fusion, it is of greatest interest to find an inhibitor against it.
Three rounds of panning against the D-form of IZN24 were performed and a subset of the phage clones was sequenced. Upon analysis of the sequences we identified the peptide sequence PPTIIRIKKHRM to be enriched in the eluate of the surface panning approach (Fig. 8). Therefore, it might be a promising binder for IZN24. Furthermore, we looked for off-target peptide motifs that are known from literature to bind to streptavidin or to the polystyrene of the well. None of the identified sequences showed a correlation to a potential negative motif.
Fig. 9 Alignment of peptide ligand sequences from the MIPD against IZN24. The plasmids within the phage clones isolated from the second MIPD against IZN24 were sequenced. The 36 bp coding for the random library peptide within the ORF of the coat protein pIII were translated to the one letter amino acid code. The resulting 12-mer peptide sequences of the potential ligands are coloured by hydrophobicity, alignment done with Geneious.
Given that we have optimized the phage amplification protocol and successfully performed an MIPD of IZN24, we moved on to our target PSMα3.
MIPD against PSMα3
Because the hydrophobic side of PSMα3 is important for its biological function, we aimed to attach our target protein PSMα3 on the surface with a more exposed positioning. For this purpose, a Lys-PEG4-biotin linker was attached to the C-terminus of PSMα3 for immobilization on a streptavidin coated surface. We chose a C-terminal biotinylation because the formylated N-terminus of PSMα3 is also functionally important7. With this experimental setup PSMα3 is more accessible to phages without steric hindrances.
Two independent MIPDs against D-PSMα3 were performed, each testing the two different panning approaches, surface (surf) and in solution (sol).
In the first MIPD we performed three rounds of panning against D-PSMα3. We could enrich several phage clones that were selected in every round qualifying as potential binders for PSMα3.
From this screen, we sequenced 10 clones for each approach (surf and sol). Moreover, 5 phage clones of our control panning without PSMα3 were sequenced to identify nonspecific sequence motifs. We found a phage clone that was enriched in both the solution and surface panning approach which suggests that it might be a potential binder for D-PSMα3.
Fig. 10 Alignment of peptide ligand sequences from the first MIPD against PSMα3. The plasmids within the phage clones isolated from the second MIPD against PSMα3 were sequenced. The 36 bp coding for the random library peptide within the ORF of the coat protein pIII were translated to the one letter amino acid code. The resulting 12-mer peptide sequences of the potential ligands are coloured by hydrophobicity, alignment done with Geneious.
None of the sequences revealed from sol and surf panning overlap those obtained from the control panning. In addition to this we analyzed the sequences with regard to motifs that are known to be streptavidin or polystyrene binders. Even with our negative selections in place we couldn’t completely avoid off-target binding in our panning approach. Some of the phage displayed peptides that were enriched in our first MIPD contained the polystyrene binding motifs WxxWxxxxW.15,16
To improve target selectivity we applied a more extensive negative selection of off-target binders and increased the detergent concentration in our washing steps. After three rounds of panning with this optimized protocol we sequenced 30 phage clones from the final enriched phage eluate.
Fig. 11 Blue phage plaques on an E.coli bacterial lawn containing a single phage clone each. Left: phage clones from the in-solution panning approach, right phage clones form the surface approach.
There was significantly less enrichment of phage clones displaying polystyrene binding motifs, pointing to an increased target selectivity of the second MIPD. Sequence analysis of the second MIPD showed an enrichment of phages expressing peptides sequence MPMFKHRMFHTH indicating that this peptide might be a potential binder to D-PSMα3 (Fig. 11). Moreover, there was one promising phage clone that was selected independently in both our MIPDs against D-PSMα3: VQVRDNLPTTTG.
Fig. 12Alignment of peptide ligand sequences from the second MIPD against PSMα3. The plasmids within the phage clones isolated from the second MIPD against PSMα3 were sequenced. The 36 bp coding for the random library peptide within the ORF of the coat protein pIII were translated to the one letter amino acid code. The resulting 12-mer peptide sequences of the potential ligands are coloured by hydrophobicity, alignment done with Geneious.
Some sequences can be excluded, such as those containing the motif YEYEYP as it appears unspecifically in all panning approaches, including the negative control. Having identified potential peptide sequences, we next had to determine the actual binding qualities of the identified binders, our next steps focused on the verification and evaluation of the results from our MIPD.
Verification of the Phage Display results
To analyze the affinity and selectivity of the peptide ligands displayed by the different phage clones we performed a phage ELISA. We amplified all sequenced phage clones and individually incubated them in separate wells of a microtiter plate. We compared the phage binding to a streptavidin coated, BSA blocked well with the binding to a similar well that was additionally coated with biotinylated D-PSMα3. Binding of phage particles was detected using a horseradish coupled anti-M13 antibody binding to the pVIII coat protein of M13.
Fig. 13 Phage ELISA: Left: D-PSMα3 coated well is incubated with single phage clones obtained from the MIPD. Right: negative control well without D-PSMα3 for detection of unspecific binding.
In sequential rounds of phage ELISA experiments we screened a total of 40 phage clones that were obtained from both MIPD experiments. After a preselection (data not shown) two dilutions of the amplified phage solutions were tested to make predictions regarding their affinity towards D-PSMα3 (Fig. 14).
Fig. 14 Relative binding affinity of Phage clones to PSMα3. Isolated phage clones from both MIPDs (#1 and #2) against PSMα3 were amplified and approximately 108 - 109 phages were incubated with target-coated ELISA plates. surf = surface panning sol = in solution panning n.d. = non detectable. Arrows mark the phage clones that were chosen for further experiments because they showed affinities above the cutoff, marked by a red line.
Based in the phage ELISA shown in Fig. 14, we defined a threshold and selected 5 phage clones that showed a high selectivity for D-PSMα3 and one negative control (ctrl2) for further analysis (Table 2):
|
Phage Display # |
panning approach |
clone number |
sequence |
|
1 |
in solution |
9 |
WDMWPSMDWKAE |
|
2 |
in solution |
17 |
DHAMHQNQNISN |
|
2 |
surface |
27 |
MPMFKHRMFHTH |
|
2 |
surface |
11 |
KHVQITSTFGVI |
|
2 |
surface |
8 |
KHLHYHSSVRYG |
|
2 |
negative control |
2 |
DMHGRYMMTTRE |
Table 2: Shown are the best binders to PSMα3
The selected phage clones were amplified again and purified by PEG precipitation. Then 3.5*109 phage particles of each clone were used in a quantitative phage ELISA (Fig. 14). The phage clones 2 surf 8 [MIPD 8 ] (TIPNTRMIMKML), 2 surf 11 [MIPD 11] (NNPKRRSNLHML) and 2 surf27 [MIPD 27] (MSMKHPMLKHRH) showed a significant binding affinity for the D-PSMα3. The negative control on the other hand showed equal affinity for both the control and the D-PSMα3-coated plate.
Fig. 15 Relative binding affinity of Phage clones to PSMα3. Isolated phage clones from both MIPDs (#1 and#2) against PSMα3 were amplified and 3.5*10p phages of each individual clone were incubated with target-coated ELISA plates. surf = surface panning sol = in solution panning.
The L- and D- Ligands were synthesized by SPPS and purified by HPLC. Different fractions of the crude ligands, here MIPD #1 ligand 9 sol were collected (Fig. 16 A) and analyzed by LC-MS. The purest fraction (Fig. 16 B) is then lyophilized and used in following experiments.
Fig. 16 Purification and analysis of MIPD #1 ligand 9 sol by HPLC and LC-MS. preparative HPLC with C18 column. Solvent A: H20 + 0.05% TFA (green) Solvent B: Acetonitril + 0.05% TFA (red), Gradient: 10-90% Acetonitril (A) Analytical LC-MS. mass: 1580,6 g/mol [M+2H]2+: 791,3 (B)
Evaluation of the binding affinity of isolated peptides
After this preselection by phage ELISA we verified the binding of these peptides to PSMα3 in a more natural setting. We chemically synthesized the ligands 8, 11 and 27 by SPPS to analyze their behaviour as peptides rather than the fusion proteins displayed on the phage particles. To examine the binding rates and affinities of the L-peptide ligands to D-PSMα3, we used the biolayer interferometry (BLI) method on an Octet RED96 system. BLI based on optical measurements of the refractive index of a target-coated sensor. Depending on ligand binding to this sensor the biological layer becomes thicker causing a change in the refractive index which can be measured by the Octet RED. We chose this method because it is one of the most accurate ways to determine protein-protein interactions and assay conditions have very little influence on the measurement.
Fig. 17 Binding of the ligands from MIPD (8, 11, 27) to D-PSMα3. Real time BLI measurement of the association of each ligand at 50 µM concentration to a D-PSMα3 coated sensor. First a baseline of the D-PSMα3 coated sensor is measured, then the association of the ligands and the resulting shift in refractive index. The last phase shows the dissociation of the ligands.
Biotinylated D-PSMα3 was immobilized on streptavidin biosensor tips and the binding of the L-peptide ligands obtained from phage display (MIPD 8, 11 and 27) was measured (Fig. 17). A binding was detected for MIPD 8 and MIPD 27 with affinity values (Kd) of 4.76 +/- 2.04. However, for MIPD 11 no binding to D-PSMα3 was detected. This result was surprising and contracting to our finding of the phage ELISA where we could measure a significant binding (Fig. 15).
We therefore chose another more sensitive method to prove the interaction of the L-peptide ligands with D-PSMα3. In addition to the MIPD L-peptide ligands we included those we identified in our modelling approach. With our software finDr we performed in silico MIPD and optimized the results using a genetic algorithm (GA). With this strategy we identified two peptides.
The second method we used for evaluating the binding affinity based on single colour reflectometry (SCORE) and implemented in an electrically controlled fluidic system. Microarrays of our L-ligands with a concentration of 1 mg/mL were spotted onto N-Hydroxysuccinimide (NHS) glass slides. The NHS functionalized slides react with the NH2 group of peptides thus forming a covalent bond with the surface leading to an immobilisation of the ligands. PSMα3 (10 µg/mL) in L- and D-form was then injected via a continuous flow over the slide. A camera detects the interference signal of all binding steps in real-time which is then visually enhanced. For the quantification of different binding kinetics, binding curves were calculated.
An advantage of this method is that not the ligand association is measured but rather the association of the larger PSMα3. This slight but significant size difference contributes to the improved sensitivity of this method. Since both D-PSMα3 and the ligands are still comparably small in size we decided on a multi-step strategy to achieve the best visualisation of the PSMα3 binding to our ligands. Between all steps there is a mild washing step (PBS pH7.5, 0.5 % BSA)
- Addition of biotinylated D-PSMα3
- Addition of L-PSMα3 (binding to D-PSMα3)
- Addition of complexes of biotinylated D-PSMα3 coupled to Streptavidin-Cy5 conjugate
This protocol relies on sequential association of proteins to the ligands (D-PSMα3 in step 1, L-PSMα3 in step 2 and finally D-PSMα3-Streptavidin-Cy5 in Step 3). This is meant to increase the size of the biolayer and therefore wavelength shift for detection.
Fig. 18 NHS slide spotted with ligands is flushed with D-PSM-biot + Strep-Cy5 MIPD11, MIPD27, GA2 show the highest binding to PSMα3
Figure 19 shows the binding curves of the different ligands from MIPD and software finDr. From the binding curves it can be seen that MIPD 11 shows the highest binding but with similar affinity to L- and D-PSMα3. This leads to the assumption that MIPD 11 to D-PSMα3 binding is independent from secondary structure and could rely more on charge differences. The peptide GA2, which is obtained from finDr behaves comparable to MIPD 11 with lower binding. MIPD 27 in contrast showed specific binding to D-PSMα3. After blocking a strong detachment could be observed in non-mixed spots before pipetting, this lead to the assumption that MIPD 27 forms oligomers after incubation. PSMα3 did not bind the negative control, obtained from phage display and the BSA spot which validates our findings.
Fig. 19 Binding of ligands to PSMα3. Real time single based on color reflectometry measurement. 10 µg/mL PSMα3 and 1 mg/mL ligand concentration was used.
After the measurements all free biotin on ligand-bound PSMα3 was saturated with Streptavidin-Cy5 conjugate (Fig. 21) We could visualize the total amount of D-PSMα3 which was associated to each ligand by fluorescence as a second readout for the binding affinity. In this additional step it can be seen that GA1 and GA2 show a high binding of Strep-Cy5.
Fig. 20 Fluorescence signal after step3 and step 4 A) D-PSMα3-Strep-Cy5 B) additional step 4 +Strep-Cy5. Also previously bound D-PSMα3 is detected
Fig. 21 Binding of ligands to PSMα3. Real time single based on color reflectometry measurement. 10 µg/mL PSMα3 and 1 mg/mL ligand concentration was used.
For ratiometric analysis the scans were coloured in green (step 3) and red (step 4) by imageJ even though both signals occur from 635nm channel. (Fig. 21) This makes it possible to measure the ratio of beforehand streptavidin in comparison to added streptavidin in step 4. If more Streptavidin is bound from Step 3 the intensity of the green signal increases. A high fluorescence in the red channel indicates a lot of Streptavidin was bound from step 4. Yellow indicates an equivalent binding from step 3 and 4. The ratiometric analysis highlighted that MIPD 27 and MIPD 11 bound a high amount of D-PSMα3-Strep-cy5 in step 3 and less after adding free Strep-Cy5. This indicates that biotinylated D-PSMα3 may loosen itself from the spot or that D-PSMα3 is so densely packed that no additional binding is possible. Interestingly, there was a higher signal after adding free Strep-cy5 to GA1 and GA2. This may originate from the ligands being saturated in the initial step with biotinylated D-PSMα3 or that they gathered biotinylated-PSMα3 from neighbour spots. However, this phenomenon would need further investigation. Taken together, we have managed to select and create binders against PSMα3. We demonstrated that both MIPD and software finDr lead to high affinity sequences in comparison to our internal negative control and BSA.
Evaluation of biological effects of ligands/ligand functionality
To reveal the binding mechanism of these ligands, we modelled their interaction with PSMα3 in silico and made predictions on the residues which are involved in this binding. How we generated this model is explained in more detail on our modelling page.
|
Ligand |
Residues of PSMα3 that are involved in ligand binding |
|
MIPD 8 |
MEFVA KLFKF FKDLL GKFLG NN |
|
MIPD 11 |
MEFVA KLFKF FKDLL GKFLG NN |
|
MIPD 27 |
MEFVA KLFKF FKDLL GKFLG NN |
Table 3: Residues of PSMα3 that interact with the ligands identified from MIPD. D-PSMα3 was docked in silico to the ligands that from MIPD and the residues involved in the interaction were analyzed using PyMOL.
The in silico model of the ligands’ interaction with PSMα3 showed that all three ligands interact with several of the functionally important residues. It has been shown by Cheung et al 2014 that the lysines (one letter code K) of PSMα3, especially those in position 6 and 12 are essential to its lytic properties. Therefore the ligands MIPD 8 and 11 but also 27 might have an impact on toxicity by masking these residues. Furthermore all ligands interact with either the formylated N-terminus or the C-terminus which are both important for FPR2 receptor activation.
Based on these promising in silico predictions, we tested the effect of the ligands on the lytic properties of PSMα3 in wetlab.
Biological effects of the PSMα3 ligands
Firstly we performed Lactate dehydrogenase (LDH) assays to quantify the cytotoxicity of PSMα3 in combination with its L-peptide ligands from MIPD. Using concentrations of 2 µM D-PSMα3 and 20 µM L-ligand we measured a reduction in cytotoxicity for all MIPD L-peptide ligands as predicted by our model.
Fig. 22Reduction of D-PSMα3 cytotoxicity by its MIPD ligands in biological triplicates. Relative cytotoxicity on Jurkat T-cells was quantified using the LDH CyQuant Kit described above after 1h incubation of PSMα3 with 200.000 cells at 37°C. D-PSMα3 and its ligands were preincubated for approximately 2 minutes at room temperature.
The Ligands MIPD 8 and 27 reduced cytotoxicity the most, but also MIPD 11 and the ligand obtained from in silico MIPD (cMIPD1) seem to have an effect on the PSMα3. Unfortunately, it was not possible to analyze the effect of GA2 in this experiment, even if this peptide was a good binder in our BLI experiments. In the toxicity assays the samples with GA2 showed a very high signal regardless of the presence of PSMα3. Therefore, we analyzed the properties of GA2 further and could show that the peptide or residues from the purification interacted with the reaction buffer of the LDH assay (Fig. 23). Consequently, the LDH assay we used is not suitable to assess the functional properties of GA2 and another setup needs to be established for this purpose.
Fig. 23: GA2 interacts with the LDH reaction buffer
Outlook: As a next step it would be interesting to analyze if the ligands also have an effect on the inflammatory effect of PSMα3 and might thereby reduce the symptoms. This could be done by performing signalling analyses using the FPR2-GFP construct (BBa_K3009002) we provide for the transfection of cells.
[1] Smith, G. Filamentous fusion phage: novel expression vectors that display cloned antigens on the virion surface (1985). Science 228, 1315-1317.
[2] https://international.neb.com/products/e8110-phd-12-phage-display-peptide-library-kit visited on 05.05.2019.
[3] Schumacher, T., et al. Identification of D-Peptide Ligands Through Mirror-Image Phage Display (1996). Science 271, 1854-1857.
[5] Mäde, V., et al. Automated solid-phase peptide synthesis to obtain therapeutic peptides (2014). Beilstein Journal of Organic Chemistry 10, 1197-1212.
[6] Amblard, M., et al. Methods and protocols of modern solid phase peptide synthesis (2016). Mol Biotechnol. 33: 239.
[7] Tornesello, A., et al. F. Solid Phase Formylation of N-Terminus Peptides (2016). Molecules 21, 736.
[8] Cheung, G., et al. Insight into structure-function relationship in phenol-soluble modulins using an alanine screen of the phenol-soluble modulin (PSM) α3 peptide (2014). The FASEB Journal 28, 153-161.
[9] Wang, R., et al. Identification of novel cytolytic peptides as key virulence determinants for community-associated MRSA (2007). Nature Medicine 13, 1510-1514.
[10] Tayeb-Fligelman, E., et al. The cytotoxic Staphylococcus aureus PSMα3 reveals a cross-α amyloid-like fibri (2017)l. Science 355, 831-833.
[11] Yao, Z., Cary, B., Bingman, C. & Gellman, S. Use of a Stereochemical Strategy to Probe the Mechanism of Phenol-Soluble Modulin A3 Toxicity (2019). Biophysical Journal 116, 87a.
[12] Kretschmer, D., et al. Human Formyl Peptide Receptor 2 Senses Highly Pathogenic Staphylococcus aureus (2010). Cell Host & Microbe 7, 463-473.
[13] Lind, S. et al., Functional and signaling characterization of the neutrophil FPR2 selective agonist Act-389949 (2019). Biochemical Pharmacology 166, 163-173.
[14] Casal, J. J. Photoreceptor signaling networks in plant responses to shade (2013). Annu. Rev. Plant. Biol. 64, 403–427.
[15] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6259150/, visited on 10.10.2019
[16] https://www.nature.com/articles/s41598-017-02891-x visited on 10.10.2019
In vivo incorporation of D-amino acids
In the course of evolution, all organisms have adopted the 22 canonical amino acids as the sole building blocks available for the synthesis of the vast diversity of proteins and peptides that make up the complexity of life. What differentiates proteinogenic amino acids from non-canonical ones is their ability to be recognized by aminoacyl-tRNA synthetases and to be subsequently loaded onto tRNA molecules. Each tRNA molecule bears an anticodon of three nucleotides, each being complementary to a codon found on a transcribed mRNA1.
The connection between these nucleobases and amino acids is completed by the ribosome. The ribosome is a macromolecular complex which can bind both an mRNA and a loaded tRNA, bringing together one codon/anticodon pair at a time. This is followed by the ribosome mediated catalysis of a peptide bond between two adjacent amino acids in a protein. To start or stop the translation of an mRNA, special codons exist, one start (AUG) and three stop codons: amber (UAG), opal (UAA) and ochre (UGA).2
Alongside the ribosome, an array of proteins called initiation, elongation and release factors mediate the correct translation of an mRNA into protein.
Fig. 1: Translation of an mRNA sequence is terminated at the UAG (amber) stop codon by Release Factor 1 (RF1).
How to introduce new amino acids
While the genetic code is preserved in all organisms, the mechanisms and structures of the components involved in translation are not. An aminoacyl-tRNA synthetase for instance, when taken from one organism and transferred into another, will not be compatible with the endogenous translation machinery. It can, however, perform a different function. 3 This feature built the basis for the incorporation of non-canonical amino acids (ncAAs) into proteins, which is called genetic code expansion
Specifically, aminoacyl-tRNA synthetases (aaRS), taken mostly from archaeal species, are expressed in a host organism, often E. coli, along with their cognate tRNAs. These synthetases are orthogonal to the translation system of the host organism: they “suppress” the stop codon, which in the host organism would be interpreted as a STOP, and incorporate a certain amino acid instead. In order to use the same aaRS/tRNA pair for a diverse set of ncAAs, some of the most commonly used have been subjected to mutagenesis REF3,4. Sometimes, the opposite property is desired, that is, an aaRS/tRNA that specifically recognizes a desired amino acid. In this case, protein evolution is used to obtain these new variants. 5 The most important requirement for genetic code expansion systems is complete orthogonality of the tRNA synthetase/tRNA pair. If the tRNA recognizes other codons than the amber stop codon or if it interacts with any other aaRS, there would be incorporation of other AAs at that position. 4
Although special codons consisting of nucleic acid analogues and quadruplet codons have been created, the amber stop codon remains the most common way to incorporate ncAAs into proteins. This is the rarest codon in all organisms, especially E. coli, where it makes up only 7 % of all termination codons 1. As aaRS, the polyspecific tRNA synthetase from Methanocaldococcus jannaschii is often employed 5.
This orthogonal translation system has been proven to incorporate a multitude of ncAAs into various proteins 6, but is still accompanied by major limitations. Most of all, the presence of endogenous mRNAs carrying the amber stop codon means that incorporation of the ncAA could happen also on endogenous proteins. This not only can be deleterious to the cell, but it also leads to low incorporation rates at the desired site on the target mRNA. The presence of RF1, moreover, also means that the amber stop codon on the target mRNA be read as STOP.
This issue has been independently solved in two different ways in two model systems: E. coli and mammalian cells. Specifically, Lajoie and colleagues engineered an E. coli strain where all amber stop codons were replaced with opal codons, allowing the deletion of release factor 1 1. In this strain, called C321.DeltaA, only the target mRNA will carry the amber stop codon, thus the orthogonal synthetase will not be titrated away at other unwanted sites. Moreover, the absence of RF1 ensures that the amber stop codon won’t be read as STOP. Since mutations of all amber stop codons is not feasible in cells with larger genomes, Reinkemeier and colleagues recently proposed an ingenious strategy to increase incorporation rates of ncAAs into proteins in mammalian cells: the aaRS is brought in close proximity to the target mRNA by use of the MCP-MS2 loops system, typically used to visualize mRNA in vivo 6. To further encapsulate the orthogonal translation system from the rest of the cell, the authors employed liquid droplets-forming proteins. These membraneless designer organelles allow ribosomes and all other necessary components to come in contact with the aaRS/tRNA pair and the target mRNA, all the while isolating them from all other cellular components.
Our idea: combine both strategies
As our goal was to increase the efficiency of D-AAs incorporation into proteins in E. coli, we thought of introducing the compartmentalization strategy of Reinkemeier and colleagues into the optimized strain of Lajoie and colleagues.
Since the system was developed for mammalian cells, we first had to examine the individual components in E. coli.
The MCP-MS2 loops system
To bring the target mRNA in close proximity to the orthogonal aaRS, it is necessary to use a molecule that binds specifically mRNA. This molecule is the Major Capsid Protein (MCP) of the MS2 bacteriophage, which has a strong affinity to stem-loop structures from the phage genome. These loops can be inserted in the untranslated region (5’ or 3’ UTR) of any mRNA construct. MCP, along with any protein fused to it, will be recruited to this specific mRNA. This system is mostly used for in vivo mRNA visualization through the recruitment of GFP or other FPs 6. The usage of the system to recruit the orthogonal synthetase to the mRNA bearing the amber suppression mutation is a new, outside-of-the-box approach for optimizing the translation system.
Fig. 2: Aminoacyl tRNA synthetase can be connected to an mRNA through the strong interaction d between MCP and stem-loops.
Being derived from a bacteriophage, this system is easy to implement in E. coli. e fused MCP to the M. jannashii tyrosyl-tRNA synthetase N-terminus, and inserted the MS2 loops into the 3’ UTR of sfGFP mRNA.
Creating membraneless organelles in E. coli
We tested several proteins for their ability to form membraneless organelles (liquid droplets) in E. coli. Unfortunately some could not be expressed. However, we were very happy to see the formation of these structures by the Spindle-Deficient Protein 5 (SPD5) from C. Elegans .
Fig. 3:Schematic representation of the liquid droplets formed by the SPD5 assembler proteins
After establishing the membraneless organelles, we only had to combine the two systems by fusing the synthetase to MCP and SDP5. This fusion protein should assemble into one droplet where the target mRNA will be recruited by interaction with MCP.
Fig. 4:Target mRNA and synthetase can be brought in spatial proximity by the MS2 system and then assembled into droplets by SPD5
Incorporation of D-Phe into sfGFP
In our research for a suitable model protein into which to incorporate a D-AA, we came upon the work of Ma and colleagues (Ma et al. RSC Adv., 2015), who have proven the ability of D-Phenylalanine (D-Phe) to shift the absorption and emission spectra of the GFP. This is, to the best of our knowledge, the only publication showing incorporation of a D-AA in E. coli cells.
The shifting of spectral properties is conventionally only achieved by mutation of canonical amino acids. For this reason, the ability of D-amino acids to shift spectra, and expand the possibilities of creating valuable variants of widely-used reporter proteins, was greatly enticing to us.
The spectral shifting works by the creation of a different isomer of the tripeptide chromophore of GFP, which is bound within its beta-barrel structure
To recreate this spectral shift, we created several variations of superfolder GFP in a low-copy pBAD33 plasmid, which is compatible with the pULTRA plasmid expressing the synthetase along with fusion proteins.
First, a control had to be created where the key amino acid enabling fluorescence, tyrosine was exchanged for L-phenylalanine. At this point, it is important to note that only aromatic amino acids can maintain the fluorescence of the GFP chromophore.
The fluorescence of the altered chromophore should be decreased in intensity, and also show altered spectral properties. 8
Secondly the sfGFP variants with amber stop codon mutations had created. We decided that in addition to having an amber stop codon inside the chromophore, we would also incorporate one at a permissive site inside the protein. This would make the study of D-Phenylalanine incorporation easier, since problems could be traced back to either a non-functional chromophore or translation system. Therefore, we chose to focus on two mutants, sfGFPY66* and sfGFPF27*.
This way, we hoped to better understand the amber suppression reaction, since we would have one protein restored to its original function, and one with new spectral properties for comparison.
Into these constructs, we inserted 3 repeats of the MS2 stem loops, enabling the recruiting of the mRNA by the MCP. This would then complete the optimization of the translational systems.
Fig. 5:Plasmid constructs used in the in vivo project. MCP, SPD56 and TyrRS are expressed as a trimolecular fusion from the pULTRA backbone, sfGFPmut) with ms2-loop tagging is expressed from the pBAD33 plasmid.
To recreate this spectral shift, we created several variations of superfolder GFP in a low-copy pBAD33 plasmid, which is compatible with the pULTRA plasmid expressing the synthetase along with fusion proteins.
First, a control had to be created where the key amino acid enabling fluorescence, tyrosine was exchanged for L-phenylalanine. At this point, it is important to note that only aromatic amino acids can maintain the fluorescence of the GFP chromophore. The fluorescence of the altered chromophore should be decreased in intensity, and also show altered spectral properties.8
After that, we cloned our two amber suppression mutants: sfGFPY66* and sfGFPF27*. D-Phenylalanine will be incorporated into the chromophore of the first, and into a permissive site of the second construct. This way, we hoped to better understand the amber suppression reaction, since we would have one protein restored to its original function, and one with new spectral properties for comparison.
Into these constructs, we inserted 3 repeats of the MS2 stem loops, enabling the recruiting of the mRNA by the MCP. This would then complete the optimization of the translational systems.
We began the implementation and testing of our optimized translational system by expressing different combinations of our sfGFP(mut) and TyrRS constructs and quantifying the resulting fluorescence signal intensity via Microscopy and Flow Cytometry. Our focus on the intensity stems from the initial assumption that only a noncanonical, aromatic amino acid is able to restore fluorescence in the otherwise truncated sfGFP mutants which we have constructed.
First, we looked at the expression of wildtype sfGFP and sfGFPY66F, to gain an understanding how the substitution of tyrosine with L-phenylalanine in the chromophore affects the fluorescent properties of the protein. All constructs were expressed and induced in bacterial liquid cultures of the C321.delta A bacterial strain.
Fig. 6:sfGFP and sfGFPY66F fluorescence intensity quantification. Untransformed C321.deltaA cells serve as negative control. Signal intensities were acquired by Fluorescence Microscopy and quantified by ImageJ.Graph shows values normalized to the sfGFP control. n(cells) = 50, error bars show standard error
sfGFPY66F shows a greatly decreased fluorescence intensity compared to the wildtype protein. Still, the intensity is still sufficiently strong for accurate quantification in our experiments.
Fig. 7: Left, C321.deltaA cells expressing sfGFP wt at 475 nm, merged fluorescence and DIC channels. Right, C321.deltaA cells expressing sfGFPY66Ft at 475 nm, merged fluorescence and DIC channels.
Second, we characterized our assembler proteins of choice SPD5, NICD and FUS/EWSR1 to find out which of them are able to form droplet structures in E.coli cells. Out of these three, successful cloning and expression could only be achieved for the SPD 5 proteins. An SPD5-GFP fusion seemed to form droplet-like structures of sub-micrometer size.
Fig. 8: C321.deltaA E.coli cells expressing SPD5::GFP assembler-reporter fusions. Droplet-like structures are present in most of the cells. Images taken with 100X magnification, DIC and fluorescence channels merged in ImageJ.
To ensure that the condensates we were seeing were actual, dynamic organelles and not dead protein aggregates, we looked at their fluorescence recovery after photobleaching (FRAP) behavior:
Fig. 9: Droplet formed by SPD5:GFP fusion under a confocal microscope using 100X magnification before and right after photobleaching
The fluorescing droplet structure was bleached for a short period of time using a high laser intensity. A droplet structure would recover its fluorescence due to dynamic protein diffusion inside the organelle, bringing unbleached fluorescent reporters to the surface, while an immobile aggregate would remain dark:
Fig.10 Recovery of fluorescence after photobleaching in SPD5::sfGFP expressing C321.deltaA cells.
The recovery of fluorescence is clearly visible in the region of interest. In the end, fluorescence intensity was recovered completely:
Fig. 11: Droplet formed by SPD5:GFP fusion under a confocal microscope using 100X magnification 1 s and 10 min after photobleaching
Next, we co-expressed our pBAD33 and pULTRA constructs in the C321.delta A and MG1655 bacterial strains, with MG1655 serving as a comparison for C321.deltaA, as the latter was derived from the former. Cells were double-transformed with different variations of the relevant constructs and induced with IPTG and Arabinose.
To test the general functionality of the TyrRS, the fluorescence intensities of these double-transformed cells were compared to the intensities of cells solely expressing sfGFPY66* and sfGFPF27*. Without the synthetase, protein synthesis should be interrupted at the stop codon, resulting in a truncated protein.
Fig. 12: sfGFPY66* and sfGFPF27* fluorescence intensity quantification with and without TyrRS in the cytosol. Signal intensities were acquired by Fluorescence Microscopy and quantified by ImageJ.Graph shows values normalized to the sfGFP control
Clearly, the expression of the orthogonal tyrosyl-tRNA synthetase increases the signal intensity of sfGFP(mut) in C321.A cells, but not in MG1655. This indicates that the Synthetase is indeed functional, but its activity is only fully unfolded in cells without other amber stop codons and Release Factor 1.
Fig. 13.1: C321.deltaA E.coli cells expressing sfGFPY66* in D-Phenylalanine supplemented medium Images taken with 100X magnification, DIC and fluorescence channels merged in ImageJ.
Fig. 13.2: C321.deltaA E.coli cells expressing sfGFPY66* in medium without D-Phenylalanine Images taken with 100X magnification, DIC and fluorescence channels merged in ImageJ.
We decided to focus on expressing sfGFPY66*, the mutant which should have D-Phenylalanine incorporated into the chromophore by the tyrosyl-tRNA synthetase /tRNA pair in both available cell lines and sfGFP27* solely in the C321.deltaA cell line, given the fact it gave more promising results for comparison. sfGFPY66* in C321.deltaA cells was additionally tested with the trimolecular, droplet-forming pULTRA constructs.
Our first attempts at quantifying fluorescence intensities, however, have turned out quite unexpected:
Although fluorescence intensities could be recovered compared to constructs not expressing the synthetase, we found that control samples without D-phenylalanine had an even stronger fluorescence signal than the samples with supposed D-phenylalanine incorporation, especially in the C321.A strain, which in general always showed a higher fluorescence intensity of all sfGFP variants than in MG1655.
This made us question our experimental setup and our ability to achieve specificity for the incorporation of D-amino acids using this optimized orthogonal translation system. We therefore decided to change our experimental setup, to find out the optimal induction protocol for our constructs. First, we decided to elongate expression time, increasing it from only 2h to 6h and 10h.
Fig. 14: sfGFPY66* and sfGFPF27* fluorescence intensity quantification after 2, 6 and 10 hours. Signal intensities were acquired by Fluorescence Microscopy and quantified by ImageJ. Graph shows values normalized to the sfGFP control.
As a next step, we tried out different induction times for tóur constructs. Since the folding time of the synthetase was unknown, we decided to induce the synthetase 2 hours earlier than the gene coding for our mutated sfGFP. To affirm our results, we analysed samples acquired using the 6 hour induction protocol through Fluorescence Assisted Cell Sorting (FACS).
Fig. 15: E.coli C321.deltaA cells were transformed with a pULTRA plasmid containing superfolder GFP(Y66*) (sfGFP(Y66*)) and analysed via fluorescence activated cell sorting. Furthermore, a second pULTRA plasmid was added containing spindle-defective protein 5 (SPD) fused to the Methanocaldococcus jannaschii tyrosyl-tRNA synthetase (MJTyrRs).(A) Uninduced E.coli C321.deltaA cells containing sfGFP(Y66*), sfGFP(Y66*) supplemented with D-amino acids (D-aas) and sfGFP(Y66*) supplemented with D-aas and SPD-MJTyrRs showed similar mean fluorescent intensities than the negative control. The strain containing sfGFP showed a strong fluorescent signal. (B) Induced E.coli C321.deltaA cells containing sfGFP(Y66*), sfGFP(Y66*) supplemented with D-amino acids (D-aas) and sfGFP(Y66*) supplemented with D-aas and SPD-MJTyrRs again showed similar mean fluorescent intensities than the negative control. The strain containing sfGFP showed a strong fluorescent signal.
Fig. 14: sfGFPY66* and sfGFPF27* fluorescence intensity quantification after a 2 hours space between both inductions. Signal intensities were acquired by Fluorescence Microscopy and quantified by ImageJ. Graph shows values normalized to the sfGFP control.
We then compared the signal intensities of synthetase constructs only fused to MCP and trimolecular fusions of MCP-synthetase and SPD5.
Fig. 16: sfGFPY66* with and without dynamic droplet structure assembled by SPD5. Fluorescence intensity quantification after 6 hours induction. Signal intensities were acquired by Fluorescence Microscopy and quantified by ImageJ. Graph shows values normalized to the sfGFP control.
Finally, we managed to consistently measure a higher fluorescence signal intensity for sfGFPY66* expressed in the C321.deltaA and MG 1655 strains. Moreover, the tripeptide fusion constructs containing the SPD5 assembler protein have also consistently shown a higher signal than constructs with a bimolecular MCP-ms2 fusion, indicating that the translational system is, at least, functional, and D-Phenylalanine incorporation can be presumed due to the consistent differences between cells supplemented with D-Phenylalanine and cells grown in unsupplemented media.
To further investigate whether D-Phenylalanine incorporation could truly be improved by well-coordinated incorporation times, we tested a protocol where the synthetase was induced first, and sfGFPY66* and sfGFPF27* two hours afterwards. This should improve D-Phenylalanine incorporation even further, since enough synthetase molecules should be present in the cytosol at the point of sfGFP induction, to dominate the endogenous unspecific incorporation.
All in all, we have constructed as well as optimized an orthogonal translation system that is shown to be functional in E.coli, and gained valuable insights for further investigations on this topic. This way, even iGEM teams can profit from the power of D-Peptide already today.
We hope that our optimized translation system can be further improved to one day functionally assist the incorporation of D-amino acids into proteins, and increasing yield and specificity the way it was intended to do.
To achieve this, however, further studies have to be conducted to investigate the cause and mechanisms of nonspecific incorporation, and which measures need to be taken to prevent it.
[1] Berg, Jeremy M.; Tymoczko, John L.; Stryer, Lubert; Hacker, Barbel (2011): Biochemistry 6. Ed. Heidelberg: Spektrum Akad. Verl. Online
[2] Gan, Rui; Perez, Jessica G.; Carlson, Erik D.; Ntai, Ioanna; Isaacs, Farren J.; Kelleher, Neil L.; Jewett, Michael C. (2017): Translation System Engineering in Escherichia coli Enhances Non-Canonical Amino Acid Incorporation into Proteins. In: Biotechnology and bioengineering 114 (5), S. 1074–1086
[3]Kodama, Koichiro; Nakayama, Hiroshi; Sakamoto, Kensaku; Fukuzawa, Seketsu; Kigawa, Takanori; Yabuki, Takashi et al. (2010): Site-specific incorporation of 4-iodo-L-phenylalanine through opal suppression. In: Journal of biochemistry 148 (2), S. 179–187
[4] Schultz, Kathryn C.; Supekova, Lubica; Ryu, Youngha; Xie, Jianming; Perera, Roshan; Schultz, Peter G. (2006): A genetically encoded infrared probe. In: Journal of the American Chemical Society 128 (43), S. 13984–13985
[5]Wang, L.; Brock, A.; Herberich, B.; Schultz, P. G. (2001): Expanding the genetic code of Escherichia coli. In: Science (New York, N.Y.) 292 (5516), S. 498–500
[6] Tharp, Jeffery M.; Wang, Yane-Shih; Lee, Yan-Jiun; Yang, Yanyan; Liu, Wenshe R. (2014): Genetic Incorporation of Seven ortho-Substituted Phenylalanine Derivatives. In: ACS chemical biology 9 (4), S. 884–890
[7] Reinkemeier, Christopher D.; Girona, Gemma Estrada; Lemke, Edward A. (2019): Designer membraneless organelles enable codon reassignment of selected mRNAs in eukaryotes. In: Science (New York, N.Y.) 363 (6434)
[8] Ma, Hairong; Liu, Nan; Shi, Shaobo; Wang, Shuzhen; Chen, Yijun (2015): Genetic incorporation of d -amino acids into green fluorescent protein based on polysubstrate specificity. In: RSC Adv. 5 (49), S. 39580–39586. .
Intein Expression System
When we decided to work with PSMα3, not just the synthesis of the D-form posed a problem but also the expression of the natural L-form. Because of its lytic properties an organism expressing it would pose a potential safety risk. Additionally, due to the antimicrobial activity of PSMα3, the expression in E.coli would be difficult1. These problems necessitated the chemical synthesis of L-PSMα3.
To enable future iGEM-Teams to safely express toxic proteins in E.coli and produce them in sufficient yields so that they can work with them, we thought about an expression system based on split inteins. Split inteins are naturally occurring proteins and when they bind together they reconstitute to a full intein that autocatalytically splices out of the precursor proteins and attaches the flanking regions together by forming a new peptide bond2.
Fig. 1: Intein-based expression system for unsafe products.
By splitting the toxic protein in two safe and non-toxic fragments and then fusing them to the N- and C-Intein of gp41.1 (BBa_K1362160 and BBa_K1362161) one can now efficiently express both constructs in separate cultures (Figure 1). After the expression the cells of both cultures are lysed and mixed so that the desired protein can now be assembled under safe in vitro conditions. By the expression of separated fragments one could also implement the incorporation system for D-amino acids, with the advantage that different D-amino acids (D-AAs) can be incorporated in each culture, leading to the possibility of having different D-AAs incorporated in the same protein.
To demonstrate the advantage of our parts over BBa_K1362160 and BBa_K1362161 we tested the expression of all four constructs. Therefore, we had four 4 mL cultures of BL21(DE3) cells, each containing one construct, and induced them, following an OD600 of 0.5, with 1 mM IPTG, then incubated them at 37°C, 200 rpm. After 3 h the cultures were pelleted and lysed. Denatured samples of the lysates were loaded to a SDS-PAGE gel (12%, 100 V, 1h 10 min) followed by Coomassie staining of the gel (Figure 2).
Expression system:
Backbone: pET302
Promoter: T7 promoter
Terminator: T7 terminator
RBS: T7 RBS (BBa_K1362090)
E.coli strain: BL21(DE3)
Fig. 2: Coomassie staining of SDS-PAGE with lysates from original and optimized gp41.1 split intein constructs expressed in BL21(DE3)
The SDS-PAGE shows that the improved BBa_K3009015 and BBa_K3009016 have been expressed in high amounts compared to the original parts and can be used together for analyzing different splicing conditions. To demonstrate that the optimized parts are qualified for this application, we tested the splicing reaction.
We cultured BL21(DE3) cells with gp41.1 C-Int + TRX and gp41.1 N-Int + MBP, induced them at an OD600 of 0.5 with 1 mM IPTG, then incubated them at 37°C, 200 rpm. After 3 h, the cultures were pelleted and lysed. The lysates were centrifuged at 21000 g, 4°C and the supernatants mixed und incubated at 42°C, with and without 4 mM Dithiothreitol (DTT). Samples were taken at different time points and a Western Blot assay was performed with them (Figure 3).
Fig. 3: Western Blot of lysates containing gp41.1 C-Int + TRX (BBa_K3009015) and gp41.1 N-Int + MBP (BBa_K3009016), mixed and incubated with and without 4 mM DTT at 42°C, stained against TRX. * spillover from lane 3
The WB analysis shows that our constructs conduct the splicing reaction rapidly and under different reducing conditions.
Overall, we managed to improve BBa_K1362160 and BBa_K1362161 by adding the maltose-binding protein or thioredoxin to it and thereby enabling strong expression, protein splicing at desired levels and facile WB detection. This makes BBa_K3009016 and BBa_K3009015 ideal parts for establishing the optimal splicing conditions for every iGEM team that wants to work with the gp41.1 split inteins and it demonstrates that the Intein-based expression systems can be used by any iGEM team struggling with expression of dangerous or toxic proteins.
A further optimization by this system could be achieved by secreting the intein-extein constructs into the medium. This would lead to simpler purification, avoidance of proteolytic digestion and a better chance of correct protein folding3. Due to its fluorescent properties, sfGFP is a secretion tag were the extracellular signal validation and therefore the monitoring of the secretion, would be simple.
To demonstrate the auto-secretion of sfGFP described by Zhang et al.4 and its ability to carry an intein-extein construct into the medium we designed two constructs. Each contains the Npu DnE C-Intein (BBa_K1362101) with thioredoxin as an insert (Figure 4). In one construct, the C-intein was fused to the carboxyl end of sfGFP (BBa_I746916).
Fig. 4: constructs for sfGFP secretion
Expression system:
Backbone: pET302
Promoter: T7 promoter
Terminator: T7 terminator
RBS: T7 RBS (BBa_K1362090)
E.coli strain: BL21(DE3)
We induced the sfGFP-CInt-TRX construct in 20 mL E. coli cultures under different temperatures (37, 30, 25°C), IPTG concentrations (1, 0.5, 0.1 mM, ctrl.), media (LB-medium, auto-induction medium) and took samples of each culture after 16, 24 or 48 hours. The samples were subsequently centrifuged at 21000 g and the fluorescence of the supernatant was measured with a plate reader (excitation: 495 nm; emission: 515 nm) (Figure 5).
Fig. 5: Fluorescence of the medium after different inductions
The induction with 0.1 mM IPTG at 37°C in LB-medium showed the highest relative fluorescence. This condition was used for a 50 mL culture and the secretion was analyzed over time. We measured the OD600 and obtained samples of the culture every 2 hours. Samples were centrifuged at 21000 g and the fluorescence of the supernatant was measured with the plate reader (excitation: 495 nm; emission: 515 nm)(Figure 6).
Fig. 6: Fluorescence of the medium and OD600 of the induced and uninduced 50 mL cultures containing E. coli BL21(DE3) with the sfGFP-CInt-TRX construct.
After 6 hours, the fluorescence in the medium increases, while the cells are still in the growth phase. These results were confirmed by Western Blot analyses of the supernatant, using an antibody against TRX (Figure 7). Expression of CInt-TRX in another 50 mL culture of E. coli BL21(DE3) under the same conditions did not show secretion of TRX (Figure 7: 4, 8, 10 h). This indicates the increase of fluorescence in the medium is based on the secretion of sfGFP-CInt-TRX and not on lysis of cells and sfGFP is carrying CInt-TRX into the medium.
Fig. 7: Western Blot of samples from different timepoints from the medium of cultures containing BL21(DE3) expressing CInt-TRX or sfGFP-CInt-TRX, all samples adjusted to an OD of 4, stained against TRX.
Altogether we demonstrated the auto-secretion of sfGFP and its ability to carry an intein-extein construct into the medium, analyzed different conditions for the secretion and measured the secretion over time.
[1] Towle K. M. et al., Solution Structures of Phenol-Soluble Modulins α1, α3, and β2, Virulence Factors from Staphylococcus aureus (2016), Biochemistry 55 (34), 4798-4806
[2] Li Y. et al., Split-inteins and their bioapplications (2015). Biotechnology Lett. 37 (11), 2121-2137
[3] Choi J. H. et al., Secretory and extracellular production of recombinant proteins using Escherichia coli (2004). Appl. Microbiol. Biotechnol. 64, 625–635
[4] Zhang Z. et al., Non-peptide guided auto-secretion of recombinant proteins by superfolder green fluorescent protein in Escherichia coli (2017). Scientific Reports. 7, 6990
In vitro incorporation of D-amino acids
PURExpress protein synthesis
Our initial aim was to establish the in vitro protein expression protocol with the PURExpress kit. Thus, DHFR, which was provided by NEB as a positive control and minC were expressed. Figures 1A-D show the instant blue stainings of 12% TRIS-glycine precast gels containing expressions of the positive control plasmid DHFR from NEB and of minC with PURExpress from NEB. In Figure 1A and Figure 1B the gel was loaded with 15 μl of each sample. No specific bands are visible in these gels because of blue smear. In Figure 1C and Figure 1D the gels were loaded with 2.5 μl of each sample and reactions were carried out with RNase-free tips and tubes and under a hub, which was cleaned with an anti RNase spray (RNase ZAP). Here, specific bands are visible at 21 kDA level in the well containing the DHFR expression and at 28 kDa level in the well containing the minC expression (red arrows). These results show that the first gels were overloaded with the sample and therefore no specific bands were visible.
Fig. 1: Expression of DHFR and minC with PURExpress: A - B) 15 µl of each sample was analyzed by SDS-PAGE using a 12% TRIS-glycine gel. C-D) 2.5 µl of each sample was analyzed by SDS-PAGE using a 12% TRIS-glycine gel
To further improve separation quality, a 16.5% TRIS-tricine gel was loaded with the same expression samples (Figure 2). The Instant Blue staining of the gel showed no specific bands. Therefore, further gel separations were carried out on 12% TRIS-tricine gels.
Fig. 2: Expression of DHFR and minC with PURExpress ΔRF123: 15 µl of each sample was analyzed by SDS-PAGE using a 16.5% TRIS-tricine gel. The first line contains the ladder, the second the negative control, the third the DHFR expression and the fourth line contains the minC expression.
In the light of establishing in vitro synthesis of peptides, the expression of a fluorescent protein was carried out. Figure 3 shows the expression of eYFP with PURExpress ΔRF123 on a 12% TRIS-glycine precast gel. The gel was loaded with 2.5 μl of each sample. A specific band at approximately 28 kDa height is visible (marked with a red arrow).
Fig. 3: Expression of eYFP and PURExpress ΔRF123: 2.5 µl of each sample was analysed by SDS-PAGE using a 12% TRIS-glycine gel. The first line contains the ladder, the second the negative control without a plasmid and the third the eYFP expression.
To demonstrate the functional state of the fluorescent protein, the samples were loaded on a native gel and the fluorescence was visualized with the Typhoon Imager (Figure 4). In a native gel the protein is not denatured, therefore its fluorescence properties will be maintained. The wells containing the eYFP reactions showed bands on the top of the well, which indicates the fluorescence of the protein.
Fig. 4: Expression of eYFP with PURExpress ΔRF123: 10 µl of each sample was by a 6% native polyacryl amide gel. A) The eYFP reaction is in the first well. B) The eYFP reaction is in the second well. In both gels the samples are run together with a negative control reaction sample (control).
Crude Extract-Based Cell-Free System (CEBCFS) Reactions
Next, the incorporation of D-Phenylalanine into sfGFP with PURExpress was attempted. However, to enable the incorporation of D-Phe, the addition of a polyspecific tRNA and its orthogonal tRNA-synthetase was required. Aiming this, the genes for MJ tyrosyl synthetase and its cognate tRNA should be cloned from pULTRA plasmid into a pET28a plasmid and expressed with PURExpress ∆RF123. However, the current literature did not give any examples about the feasibility of tRNA-synthesis with PURExpress ∆RF123. Special in vitro techniques are described by Korencic et al. [1] for in vitro production of tRNA transcripts. Moreover, it is mentioned that purified tRNA and tRNA synthetase should be added separately to the aimed reaction. Due to these reasons, the tRNA could not be synthetized easily with PURExpress. A cell free protein synthesis reaction with a cell lysate instead, would allow to avoid this problem, by transforming the cells with the pULTRA plasmid containing the genes for the polyspecific tRNA and preparing a lysate from these cells.
In this context a CFPS reaction with cell lysate seemed a more promising, simple and cost-efficient system for the incorporation of D-amino acids into proteins. Therefore, the expression system was shifted from PURExpress to cell lysates and no further experiments were performed with the PURExpress kit. To establish this technique, we first ran a control experiment by expressing wt sfGFP. The first sfGFP expressions with cell lysates derived from MG1655 and C321.∆A strains contained ATP, plasmid, dithiothreitol, RNase-Inhibitor and the cell lysate (Figure 8) Reactions with MG1655 grown in TB and C321.ΔA grown in TB with PURE elements show lower fluorescence intensities than water. The other expressions have nearly twice as high intensities than water. The highest intensity was obtained from C321.ΔA lysates without PURE elements. So these results implicate that the expression of sfGFP did occur.
Fig. 5:Expression of sfGFP with different cell lysates: expressions were carried out with cell lysates derived from MG1655 and C321∆A strains. Both strains were grown in LB and TB medium. “w/o PURE” bars show fluorescence intensity in samples which did not contain PURExpress elements and “with PURE” bars show those which contain PURExpress elements.
To validate the results, the expressions were repeated with lysates derived from MG1655 grown in TB medium and C321.ΔA grown in TB. The sfGFP gene was expressed with a photoinducible expression system which was kindly provided by Edoardo Romano (unpublished). Positive reactions of both lysates show lower fluorescence intensities than the negative controls without plasmids. Reactions with C321.ΔA lysates generally have lower intensities than those with MG1655 lysates. The reaction with PURExpress® elements and MG1655 lysate has a lower intensity than the other two MG1655 expressions. Also these results implicate that the expression of sfGFP did not work as the fluorescence intensities are higher for the negative control.
Fig. 6: Induced CFPS reaction for sfGFP expression: grey bars show intensities for CFPS reactions with MG1655 lysates grown in TB and light blue bars show intensities for CFPS reactions with C321∆A lysates grown in TB. Dark blue bar displays the intensity for water as a blank control.
As a negative control the expressions were carried out under the same conditions without induction. Again, reactions of both lysates containing plasmid show lower fluorescence intensities than the reactions without plasmids. Also here, reactions with C321.ΔA lysates have lower intensities than those with MG1655 lysates and the reaction with PURExpress elements and MG1655 lysate again has a lower intensity than the other two MG1655 expressions.
Fig. 7: Uninduced CFPS reaction for sfGFP expression induction: grey bars show intensities for reactions with MG1655 lysates and light blue bars show intensities for reactions with C321∆A lysates. Dark blue bar displays intensity for water.
In a next step we tried to express sfGFP under an AraC promoter to try another induction system, as the expressions of sfGFP under the pBAD promoter did not give measurable yields of protein. Figure 10 shows the intensities of sfGFP expressions with lysates derived from C321.ΔA grown in TB medium. Here we could obtain higher fluorescence intensities for expressions from cell lysates without PURExpress components. However, intensity of the negative control is lower than the blank, which gave doubts to trust the measurements. Therefore, a dilution row with fluorescein was measured to test the plate reader (Figure 8).
Fig. 8: CFPS reaction for sfGFP expression under pBAD promotor with arabinose induction: grey bars indicate intensities for reactions with C321∆A lysates without PURE elements and dark blue bars indicate intensities for reactions with PURE elements. Light blue bar shows the intensity of water.
Fig. 9: Dilution row with fluorescein provided by iGEM.
The expressions were then repeated twice (Figure 12). Here we obtained higher fluorescence intensities for expressions with PURExpress elements, which indicates that the expression works, when enough resources for transcription and translation are available. However, the error bars are so high that no it is not possible to drive a concrete conclusion.
Fig. 10: CFPS reaction with arabinose induction (n=2).
Given to the results produced in this work it is not possible to drive a definite conclusion about the success of sfGFP expression with the cell lysates. PURExpress contains all necessary components for transcription and translation. Therefore, it was suggested that addition of these elements would enhance the reaction and give higher protein yields. Experiments performed by Javin P. Oza et al. reinforced this suggestion as they also added Solution A and Solution B with major transcription and translation components to the cell free protein synthesis reactions. However, we could not obtain results confirming these suggestions. To drive a definite conclusion on this topic further experiments are needed.
[1] Korencic D. et al. A one-step method for in vitro production of tRNA transcripts (2002). Nucleic Acids Research. Volume: 30 No. 20 e105Proteinase K digestion assays confirm protease-resistance of D-peptides
To demonstrate the fact that D-proteins are resistant to proteolytic breakdown by naturally occurring proteases(REF) the natural L-form and the mirror-image D-form of the peptides Phenol soluble modulin alpha 3 (PSMα3) and IZN (a fragment of the Ebola virus coat protein, kindly provided by Dr.Michael Kay) were preincubated with the polyspecific endoprotease Proteinase K for 24 hours. As controls served identical aliquots without Proteinase K treatment. For visualization, the digest was analyzed on an SDS Polyacrylamide gel by Coomassie stain.
After 24 hour incubation with Proteinase K, the L-Versions versions of the peptides were not detectable anymore via Coomassie stain, demonstrating the functionality of the protease to degrade L-Peptides. In contrast, the samples containing their D-counterparts showed specific bands even with Proteinase K present, illustrating their resistance to proteolytic digest. (Fig 11) Taken together this proof of principle experiment demonstrates the resistance of D-peptides towards degradation by proteases.
Fig. 11 Visualization of proteinase K digest of the L- and D enantiomers of PSMα3 and IZN. 50 µg of the 4 peptides were incubated with 100 µg / mL Proteinase K at 37 °C for 24 hours (24) and compared to a sample without this enzyme (-). Separated on a 12% SDS PAGE gel and run at 80V specific bands for full-length PSMα3 were detected at 2.6 kDa in a Coomassie stain, for full-length IZN specific bands were detected at a height of 6.5 kDa. Proteinase K runs at 28.9 kDA.
Chemical Synthesis of L- and D-PSMα3
Solid phase peptide synthesis
To identify D-peptide ligands against PSMα3 (Fig. 2) using phage display, we had to chemically synthesize PSMα3 in D-Form. Additionally, we formylated the uncleaved product using formic acid and N,N′-dicyclohexylcarbodiimide (DCC). However, our formylation protocol did not only lead to formylation but also to N-terminal carboxylation as a side reaction. The formylated product was purified by high performance liquid chromatography (HPLC) and lyophilized. The products were then analyzed by liquid chromatography–mass spectrometry (LC-MS).
Fig. 1: Synthesis of PSMα3. Formylation lead to N-terminal formylation (A lower spectrum) and Carboxylation (A above). B) respective mass found at a retention time of 13.09 min (N-carboxylated) and 15.25 min (N-formylated)
Fig. 2: N-terminal formylated L-PSMα3
The respective mass of carboxylated PSMα3 was found at a retention time of 13.09 min, whereas the formylated PSMα3 at 15.25 min. The masses were separated by HPLC and formylated PSMα3 used in further experiments.
Yield:
L-PSMα3: 6.46 mg
L-PSMα3 form: 1.2 mg
D-PSMα3: 1.7 mg
As a next step, we tested the functionality of our self-synthesized peptides in various assays.
Verification of biological function
Toxicity:
In order to analyze the toxicity of PSMα3 to human immune cells, we chose the Jurkat T-cell line. A lactate dehydrogenase (LDH) assay was performed to determine the cytotoxicity effect of PSMα3 on cells. The assay based on the activity measurement of the cytoplasmic enzyme LDH that is rapidly released by damaged cells. The LDH activity was quantified by using the CyQUANT Kit from Thermo Fisher Scientific.
Fig. 3: Cytotoxicity assay using the CyQUANT kit from thermo fisher
To evaluate the success of our synthesis we compared the toxicity of PSMα3 that was chemically synthesized by ourselves to the commercially obtained PSMα3 from Genecust. A concentration of 8 µM PSMα3 was chosen relying on Wang R. et al Nature Medicine 2007 to test its cytotoxicity effect (Fig. 3).
The commercially obtained PSMα3 in L-form, as well as in D-form, induced cell death of Jurkat T-cells. Around 50% cytotoxicity compared to control lysed cells was determined. The same degree of cell death was measured with our own chemically synthesized L- and D-PSMα3. These data demonstrate that the toxicity of PSMα3 does not depend on its stereochemical orientation and supports its achiral mechanism of action by membrane disruption.
Fig. 4: Relative cytotoxicity of commercial and self-synthesized D-PSMa3 at 8µM.
Furthermore, we tested if the formylation of the N-terminal methionine has an impact on the cytotoxicity effect of PSMα3. Jurkat T-cells were treated with the chemically synthesized L- and D-PSMα3 with and without formylation. There was no significant difference detectable at the test concentration of 8 µM. A cytotoxicity of around 50% was detected for all forms of PSMα3 (Fig. 5) demonstrating that the formylation does not have any impact on the toxicity.
Fig. 5: Relative cytotxicity of PSMa3 formylated and non formylated isomers.
In further experiments we could demonstrate that L-PSMα3 is toxic in a concentration-dependent manner in the range between 2 to 10 µM, as described by Wang et al.
Fig. 6: Concentration dependent cytotoxicity of PSMa3.
Interestingly, we observed a decrease in the toxicity of PSMα3 at concentrations above 50 µM. This could potentially result from the formation of alpha helical amyloid sheets at high concentrations.
To gain a deeper understanding of this phenomenon, we additionally performed in silico analysis of the self-self association of PSMα3. While we observed a high binding affinity for homochiral L-L PSMα3 complexes we found an even higher affinity between L- and D-PSMα3. (For details on how we established this model see Modelling)
Fig. 7: Computational model of high affinity binding between L- and D-PSMa3
We decided to explore the association of L- and D-PSMα3 by conducting further experiments. Firstly, we wanted to verify the predictions we made in silico by wetlab binding analyses. We measured the binding of L-PSMα3 to D-PSMα3 via Bio Layer Interferometry (BLI) using the Octet RED96 system (FortéBio, Molecular Devices). BLI is based on optical measurements of the refractive index of a target-coated sensor. Depending on ligand binding to this sensor the biological layer becomes thicker causing a change in the refractive index which can be measured by the Octet RED system.
Fig. 8: Binding interaction of L- and D-PSMa3 measured by BLI
For this assay, biotinylated D-PSMα3 was bound to streptavidin sensors (Fig 8, loading) and the association and dissociation of L-PSMα3 to its D-form was measured. By using the data analysis software of the system, a binding constant kd of 13.1 +/- 2.5 µM was calculated using the software ForteBIO.
Upon further research in the literature we found a paper mentioning the fibril formation of racemic PSMα3 in the context of protein crystal formation11.
We could show that the heterochiral association of L- and D-PSMα3 also has an effect on cytotoxicity, even at non-amyloidogenic concentrations. Upon further research we found also a cutting-edge paper mentioning this phenomenon in the context of protein crystallization 11.
T-cell activation by FPR2 - receptor stimulation
PSMα3 is known to activate immune cells, especially T-cells and neutrophils by stimulating the FPR2 receptor (BBa_K3009001). The FPR2 receptor is a G-protein-coupled receptor which mediates inflammatory activation of T-cells. Activation of the receptor results in transient elevation of intracellular Ca2+ concentrations mediated by release and influx of Ca2+ through various cation channels12. To evaluate the inflammatory properties of our chemically synthesized PSMα3 we studied the calcium flux in Jurkat T-cells upon exposure to nanomolar concentrations of the L-PSMα3.
First, Jurkat T-cells were analyzed for the surface expression of FPR2 by using the fluorescent-coupled anti-FPR2-APC antibody that is directed against the extracellular domain of the receptor. Flow cytometry analysis confirmed the presence of endogenous FPR2 on Jurkat T-cells (Fig. 15).
Fig. 9: Jurkat T-cells express endogenous FPR2 on the surface. A) Mouse anti-Human FPR2 - Alexa647 is used for detection in flow cytometry
Given that Jurkat T-cells express FPR2 on their surface, we next examined the intracellular signalling in the cells by measuring the shifts of the intracellular calcium level as a highly dynamic real time readout for FPR2 activation. (see methods) Treatment of Jurkat T-cells with x 100 nM L-PSMα3 did not result in a significant calcium influx (Fig. 10). As a control we stimulated cells with the ionophore ionomycin to artificially increase the intracellular calcium level independent of calcium channels. While cells respond to ionomycin but not to L-PSMα3, it suggests that the endogenous expression of FPR2 on Jurkat T-cell is not sufficient to induce calcium influx.
Fig. 10: Calcium release of non-transfected Jurkat cellline.A) PSMα3 toxin [700nM] and B) activator WKYMVm [100nM] lead to no calcium release
Aiming to improve the detectability of FPR2 receptor activation we further generated a GFP-tagged FPR2 fusion construct (BBa_K3009002) for transfection of HEK293 cells and overexpression of the receptor. This construct was designed with a C-terminal HA-tag to be used for antibody staining and GFP for analyzing the cellular localization of FPR2 via microscopy. Transfected HEK293 cells were analyzed by flow cytometry using anti-XYZ antibody to detect membrane localization of FPR2 (Fig. 11).
Fig. 11: Transfection of HEK239 cells with eGFP-tagged FPR2 fusion construct in different concentrations. Mouse anti-Human FPR2, Alexa647 is used for detection in flow cytometry
Having verified that FPR2-GFP is expressed on the surface of HEK293 cells, we proceeded with calcium flux assays using our chemically synthesized L-PSMα3. Treating HEK293_FPR2-GFP cells with 100 nM L-PSMα3 an increase in intracellular calcium levels was detected. The reaction to PSMα3 was comparable to the positive control for FPR2 activation, a potent peptide activator with the sequence WKYMVm13. WKYMVm was obtained from Alomone labs and used at a concentration of 100 nM as the manufacturer specifies. (Fig. 12).
Taken together, these data demonstrate that our chemically synthesized L-PSMα3 can be recognized by its natural receptor FPR2 and thereby has a functional effect on transiently transfected FPR2-GFP in HEK293 cells.
Fig. 12: Calcium release of non-transfected Jurkat cellline.A) PSMα3 toxin [700 nM] and B) activator WKYMVm [100 nM] lead to calcium release.
