Bacterial device for generation and selection of interactors for any bait protein

We developed a method of selecting the most efficiently interacting protein (let us call the protein "hunter"). It may seem similar to the two-hybrid system, except that it doesn't require putting both "hunter" and "prey" in the same cell of "one-armed bandit" and therefore only "hunter" protein is mutated in the process of its specificity maximization. Moreover, cells expressing different variants of "hunter" protein compete for the same bait. Interaction between "hunter" and "prey" outside the cell is the basis of selection.

That is why we decided to direct the "hunter" protein to the bacterial cell surface by translational fusion of its gene with with a fragment of ompA (bacterial outer membrane protein). In this way "hunter" is presented on E. coli outer membrane, which allows it to interact with "prey" added to liquid culture medium (fig. 1).

The goal of the selective pressure is to eliminate the cells not producing strong interactors for the bait, and enable the survival of cells with the best interacting proteins. For the selection we have used TEM-1 β-lactamase (protein responsible for resistance to β-lactam antibiotics such as ampicillin) split into two complementing fragments: alpha and omega. Those fragments do not form active complexes spontaneously. Antibiotic resistance is achieved only when alpha and omega are in close proximity – i.e. when they are fused with two strongly interacting proteins. So "hunter" protein fused with one β-lactamase fragment will catch "prey" protein connected to another and will allow survival of the cell in ampicillin containing medium. Cell line carrying the best hunter protein will have selection advantage over others.

In order to confirm the effectiveness of this system we have chosen two small strongly interacting proteins B domain of Staphylococcal protein A and Z_SPA1 (formerly denoted as A and Z respectively). The A protein is the famous protein from Staphylococcus aureus which binds to constant fragments of IgG antibodies and has many applications in molecular biology. The Z protein is its artificially created close relative [PMID:15238637]. Apart from interacting with the A protein the Z protein tends to aggregate (this self-interaction is much weaker than with A though). We have created many variants of construct based on pACYC177 vector (low copy – 10 copies per cell) with IPTG-induced promoter. They contain OmpA fragment, A and Z proteins and β-lactamase fragments in various combinations. Most of our constructs contain duplicated B domain (we call it A), but for some purposes we used single A protein (A delta). Lists of pACYC177 constructs:

OmpA_alpha	OmpA_omega
OmpA_A_alpha	OmpA_Z_alpha
OmpA_A_omega	OmpA_Z_omega
OmpA_omega_Adelta_alpha	OmpA_omega_Adelta
OmpA_Adelta_alpha	OmpA_Adelta_omega

Expresion of fusion protein Omp_omega_deltaA_alpha in TOP10 strain was confirmed by Western blotting using anti-A antibody . It reacted with bands of different molecular weights, probably due to degradation of fusion protein by intrinsic E. coli proteases. Therefore we used the same construct to transform Rosetta strain, which lacks a set of proteases and repeated Western blot . However, there were no important differences between both strains. We tested also for presence of Omp_omega_deltaA_alpha on bacterial surface .

We developed a group of constructs used for overexpression and purification of "prey". For this purpose we used pET15b vector with N-terminally His-tagged prey proteins. We have created:

• His+Z+omega
• His+Z+alpha
• His+A+alpha

The first two constructs were successfully used for overexpression and purification of prey proteins. In case of fusion with A protein we used His-tag and NiNTA beads for purification and in case of fusion with Z protein it wasn't necessary because of Z protein aggregation resulting in the fact that 90% of the post-sonication debris constituted this protein.

1. We tried to find out what conditions lead to most efficient expression of OmpA from lactose promoter on pACYC177 is most efficient. We created OmpA_omega_Adelta_alpha construct to check if omega fragment of β-lactamase can be placed in the middle of a protein fusion and still interact with its alpha half. We supposed that ampicillin resistance should increase proportionally to the amount of construct, with the restriction that inductor overdose can decrease expression efficiency due to excess protein synthesis or due to toxicity of OmpA fragment.
   Fig. 3. Evaluation of optimal conditions for expression of the "hunter" protein fusion with OmpA
   Conclusion: conditions of most efficient expression: for IPTG: 0.25-0.50 mmol/mL and for ampicillin: 50-75 μg/mL.
   


2. We tested what is the minimum amount of protein added to the medium, necessary for survival of cells expressing interacting "hunter" proteins (fig. 4).
   Z_omega was added to bacterial culture expressing pACYC177+OmpA_A_alpha.
   Z_alpha was added to bacterial culture expressing pACYC177+OmpA_A_omega.
   A_alpha was added to bacterial culture expressing pACYC177+OmpA_Z_omega. Fig. 4. Bacterial growth depending on the amount of prey protein added to the medium. The amount is expressed in picomoles per mL of medium.


3. We tested which of the produced "prey" proteins allow the survival of which "hunter" strains in presence of ampicillin (tab. 1). We concluded that interaction of any "hunter" strain expressing alpha protein with "prey" omega, as well as interaction of any "hunter" expressing omega protein with "prey" alpha provides resistance to the antibiotic. Yet, in case of the minimal amount of prey (fig. 4), growth of bacteria lacking its interactor decreases twofold.

Table 1. Results of testing various hunter/prey combinations
(ampicillin concentration 50-75 μg/mL).

4. We tested competition between strains expressing interacting "hunter" proteins and those lacking interactors. In one test probe we placed the following combinations:
   
   • hunter OmpA_omega, OmpA_A_omega and OmpA_Z_omega with prey A_alpha
   • hunter OmpA_omega, OmpA_A_omega and OmpA_Z_omega with prey Z_alpha
   • hunter OmpA_alpha, OmpA_A_alpha and OmpA_Z_alpha with prey Z_omega
   We cultured these combinations overnight and the following day we isolated plasmids from these strains, performed restriction digest and checked which of the "hunter" won the competition. In each of these combinations "hunter" strain expressing interactor was represented by only one band on a gen, meaning that this strain dominated over the remaining strains. Fig. 5. Result of competition test:
   1 and 5 - DNA ladder,
   2 - Insert from isolated plasmid refers to OmpA_A_Alpha
   (Z_Omega protein added to the medium),
   3 - Insert from isolated plasmid refers to OmpA_A_Omega
   (Z_Alpha protein added to the medium),
   4 - Insert from isolated plasmid refers to OmpA_Z_Omega
   (A_Alpha protein added to the medium).

We intended to create "one-armed bandit" strain, which would randomize target protein sequence – nucleotides would be shuffled randomly in the same way it happens in popular hazard game. The simplest "one-armed bandit" strain is one of standard mutator E. coli strains without polymerase error correction activity or strain without DNA repair systems. It’s not optimal though - high mutation frequency in the whole bacterial genome would introduce some variance into our sequence, but would also cause problems with selection. Instead of screening for protein interactions we would most likely obtain selection-resistant strain. So we wanted to narrow scope of mutations to a small, well defined DNA fragment preferably carried on plasmid. Since the beginning of our project we focused on antibodies, the idea of using AID (Activation Induced Deaminase) protein came right away.

The AID protein is active in mammal lymphocytes, where it causes somatic hypermutation – an increase of mutation level in antibody coding sequences. Moreover, there is a publication [PMID:12097915] demonstrating AID activity in E. coli cells. As AID mutates all highly-transcribed E. coli genes (it prefers single-stranded DNA that appears in highly-transcribed loci) and we needed to target mutagenesis to a specific DNA region, we needed to make our DNA sequence a highly transcribed one (preferably achieve the highest transcription level in the cell). For this purpose we wanted to test mutation rate in a sequence transcribed from T7 promoter using transcriptional fusion between AID and T7 RNA polymerase. We went a step forward and created translational fusion between AID and T7 RNA polymerase. T7 polymerase traverses the DNA fragment containing T7 promoter and carries AID, which introduces mutations.

AID is a small protein and its closest homologues form oligomers. So we needed to consider such possibility and we created molecular device containing both free AID and AID-T7 fusion. We hoped that AID-T7 fusion would recruit free AID to DNA sequence containing T7 promoter.

To sum up we have created following molecular devices on pMPMT5 plasmid under arabinose promoter:

1. AID

2. AID in translational fusion with T7 RNA polymerase

3. AID in transcriptional fusion with T7 RNA polymerase

4. AID in transcriptional fusion with AID-T7 translational fusion

To test various variants of AID we needed proper reporter system. We have used alpha-complementing β-galactosidase fragment under control of T7 promoter. We used one-copy plasmid pZC320 (minireplicon of plasmid F), which contained this fragment. After obtaining and induction of cotransformants carrying one of AID devices and reporter plasmid, we hoped to get some white colonies on X-gal plates, indicating mutated clones.

Simultaneously we carried out the rifampicin test (plated liquid cultures of tested strains on plates containing 300 μg/mL rifampicin) to check mutation level in whole genome of tested strains.

We have obtained various numbers of white clones using different AID encoding devices but sequencing of β-galactosidase gene from those clones revealed no mutations. It has to be a flaw in our reporter system. We suppose that expressed β-galactosidase fragment encoded on pZC320 plasmid is somehow switched off. We have sequenced large fragments of many white clones of pZC320 – no mutations, no clues. We experimented with GFP and RFP reporter system but this turned out to be ineffective . The test was carried out in two E. coli strains: Top10 and GM2163. The latter has damaged Dam and Dcm methyltransferases so DNA repair systems relying on their activity are unable to repair mismatches created by AID. The results obtained so far are presented in table below:

Table 2. Results of rifampicin test evaluating mutation level in E. coli TOP10 and GM2163 strains.

Conclusion: AID increases mutation rate, generating strains resistant to rifampicin. AID in transcriptional fusion with T7 RNA polymerase has the same efficiency as AID alone, which is what we expected. On the contrary, translational fusion apparently doesn't increase mutation rate.

Interesting thing is that although results of β-galactosidase activity test (blue and white colonies) didn't give any information on mutation rate, they differed between strains bearing various AID variants and these results were almost identical between TOP10 and GM2163.

We have created a device which is able to screen a pool of protein coding sequences and promote clonal selection of bacterial strain expressing the strongest interactor for any given bait. The next step will be linking it with a mutagenesis device to obtain directed iterative protein evolution system. In this system protein library would be expressed as a fusion with our hunter expression device. The bait protein would be fused with prey expression device, purified and added to culture medium supplemented with ampicillin. Thanks to the antibiotic resistance given by interacting beta-lactamase domains, clones expressing proteins interacting with prey protein would survive, while the rest would die. In following iterations of the experiment, interaction between hunter and prey would be gradually optimized due to activity of the mutagenesis device and competition between resulting subclones in limiting prey concentrations. The whole selection would take place in cells requiring no further user input and resulting in selection of the strongest interaction possible.

We have tried to conduct "proof of concept" experiment involving reversal of mutations in A protein sequence. Lack of mutagenesis device led us to switch to mutD5 mutator strain, but it didn't work as we hoped it to . Although we confirmed AID activity in E. coli cells, our system involving its fusion with T7 RNA polymerase turned out to be ineffective. Despite of lack of proof that our system works as a whole we want to stress out that its fundamental elements have been constructed and proved to work. Both prey and hunter generators were tested in fusion with A and Z proteins. Interaction between A and Z used as hunter and prey allowed hunter expressing strain to gain enough selection advantage to win competition with other strains, which express hunter fusions incompatible with prey.