Engineering Success | SJTU-BioX-Shanghai

Thoughts makes difference – this is what the Monkey King always said when he encountered difficulties on the Journey to the West, and it is also the most important point in the engineering process. Read more to find out about our engineering cycles, and see how we grew from mini monkeys to (let’s hope) another Great Sage Equaling to Heaven.

ClyA: a promising scaffold for OMV surface display

ClyA, also named HylE, is a pore-forming cytotoxin expressed by E. coli. It is displayed on the surface of the outer membrane and also exported from the cells. The outer membrane vesicles (OMVs) have been reported to contain ClyA, where they form oligomeric pore assemblies [1-3]. That is to say, ClyA can be sufficiently expressed and accumulated in the OMVs. Therefore, we are adopting ClyA as the scaffold for protein surface display in our project.

PbrR: a receptor protein with characterized lead-binding function

The pbr resistance operon from Cupriavidus metallidurans CH34 is regulated by the Pb(II) responsive regulator PbrR. In the presence of Pb (II), PbrR is activated by the metal ion through cysteine coordination and subsequently activates the lead-resistance promoter, PpbrA. Considering the relatively high sensitivity to Pb (II) compared with other Pb (II)-binding proteins coded by other genes in the pbr operon, we are utilizing PbrR as the heavy-metal-binding module.

His6-tag: an efficient tag for protein immunoblot analysis and purification

We have fused a His6-tag at the C terminal of PbrR to better characterize the expression level of the protein.

The full DNA sequence coding ClyA-PbrR-His6 (CPH) was synthesized by Beijing Qingke Biotechnology Co., Ltd. and amplified by PCR. The CPH was successfully expressed in the OMVs generated by engineered strains of E. coli BL21(DE3) (Fig 2).

Ctrl, empty OMV vehicles. CPH1-7, engineered OMVs collected by ultracentrifugation.

To test the Pb (II)-binding capability, we designed an experiment workflow (Fig 3A). The samples were sent to the Shanghai Agricultural Products Quality Security Center for testing of Pb (II) concentration. The result indicated that CPH in OMVs significantly improved their capability of capturing and transporting Pb (II). However, the overall Pb (II) removal efficiency was unexpectedly low (Fig 3B).

Figure 3. Recovery ratio calculated from Nanoparticle Tracking Analysis counting in the first engineering cycle.

We looked back on the process of our experiment and suspected that the accidental OMV loss during the separation of supernatant and sediment after ultracentrifugation could explain the unsatisfying Pb (II) removal rate. On the other hand, the relatively low expression level of CPH could also contribute to the result. In conclusion, the Pb(II)-binding function of the CPH protein was verified, but improvements should be made to protein expression and experiment design.

At project designing stage, we first set out to seek a way to recover the OMVs. We didn’t restrict ourselves to traditional tags; from all possible means we could find, we made a list to compare the pros and cons:

(A) Size distribution graph of nanoparticles collected from ultracentrifugation. The peak value is within the range of 100-150 nm. (B) CPH is detected both in the crude extraction of bacterial cell total protein and in the OMVs. CPH, ClyA-PbrR-His6 protein; Ctrl, CPH^- OMVs; Test, CPH⁺ OMVs.

After comparing the pros and cons, we decided that we should first stick to the reliable polyhistidine tag. Our major concern was that the recovery module should have been easy to replace, in case the OMVs over-saturate. This led us to decide that the recovery should be completed by a nickel or zinc web, which also allows for faster water flow and higher clearance rate.

The PKS docking domains were also a luring choice, yet considering the possible instability within the design, we first arranged it into the now eliminated periplasmic space remediation module (see how we gave it up with Modelling’s advice). Since that module is no longer existent, we started to ponder on the possibility to utilize this unique, interesting mechanism on the recovery module. This may be a further direction to investigate.

We were not hesitant to use Lpp’OmpA as the scaffold, yet we also wondered how we could implement this ancient and streamlined design to better display our protein of interest – sometimes perhaps oligopeptides. A thorough research and investigation on protein linkers led to a decision that we may make use of the rigid linker, 1x~3x EAAAAK, to restrict the swaying of displayed subjects, thus increasing the possibility for displayed entities to mate and interact. This formed one of our basic parts, Lpp-OmpA-2xEAAAAK BBa_K4760008.

We build the construct Lpp’OmpA-2xEAAAAK-His10 instantly. This construct utilizes an intentionally elongated His-Tag, with flexible linker GGGGS on N-terminus and extended the conventional His6 to His10, which formed our basic part BBa_K4760002. The first cycle of ClyA-PbrR-His6 testing was already in progress, and we hope this recovery part to be providing a possibility of remedying the failure encountered there. Construction was performed using standard RE cleavage and ligation, which introduced a 12 bp scar CGCGGATCCGCG caused by BamHI; another scar, caused by EcoRI, is not in the ORF.

We expressed the protein with pACYCDuet in E.coli BL21 strain.

To test the recovery rate through Lpp’OmpA-His10, we utilized both unquantified and quantified methods.

First, after 14h of IPTG induction, we clear out the bacteria bodies with 7000g centrifugation. We verified the expression within total protein and ultracentrifugation with western blot.

We then incubated the supernatant with the nickel beads at 37 ℃ for 1 h, with a mild condition of 1.5 ml Ni beads in 50 ml supernatant. Then we eluted the OMVs with 25 mM and 500 mM imidazole. Western Blotting verified the significance of Lpp’OmpA-His10 in imidazole eluted OMVs over flow-through OMVs. The OMV recovery rate were calculated based on the Nanoparticle Tracking Analysis’ counting results. The total recovery rate was not satisfactory, yet did observe significance over control.

A further test of rising the Ni bead quantity and incubation time didn’t give better recovery rate, suggesting that current conditions are already reaching the binding maximum of OMVs. Meanwhile, the WB results simply indicate that unrecovered OMVs are lacking in Lpp’OmpA-His10 protein density.

For the Western Blotting image in (A), IMZ-25 means elution with 25 mM imidazole, and IMZ-500 means elution with 500 mM imidazole.

Overall, these results demonstrated that the displayed His-Tag can partially help recovering the OMVs with nickel beads, and the binding reaction is completed within one hour.

The polyhistidine tag, as we were once concerned about, did not show enough affinity to hook large OMVs on the matrix well. The binding does not seem strong, since significant ratio of OMVs are eluted by 25mM imidazole. This led us to again ponder on the possibility of utilizing PKS-originated Docking Domains in this module, where a reactive pair is displayed on the matrix and the OMVs.

Since the reaction efficiency is hard to further optimize with larger quantity of nickel beads addition, we propose that the efficiency is actually restricted by the maximum extent of protein expression. This, we suspect, may further be restricted by peptide exportation rate of E.coli BL21. An alteration of chassis should be on the schedule of further implementation.

(1) We re-designed the experiment to use 10 times the original OMV concentration so that the inevitable loss of OMV during collection would have less impact on the result of the Pb (II) removing test.

(2) We designed an elongated version of CPH, ClyA-PbrR-(linker)-His10 (Elo), to enhance the probability of Ni-beads capturing Elo⁺ OMVs (Fig7).

The CPH⁺ OMVs collected for Pb (II) binding were resuspended in half the former volume of Tris-HCl buffer. The OMV concentration in the reacting system was 10 times that in the previous experiment. Due to the time limit, the experiment had only been repeated with CPH, not with Elo.

Elo was constructed through reverse PCR and expressed in E.coli BL21(DE3) (Fig. 7) .

(A) Colony PCR result of Elo BL21 constructs, (B) Western Blotting verification.

We performed the Pb(II) binding test for CPH⁺ OMVs. The Pb(II) removal rate was significantly enhanced by the increase in OMV concentration, yet the overall Pb (II) removal rate was still below 25%.

Fig. 9 Recovery ratio calculated from Nanoparticle Tracking Analysis counting in the second engineering cycle. Compare with Fig. 3.

In the OMV recovery test, the Ni binding performance of Elo⁺ OMVs was similar to that of Lpp-OmpA-His10⁺ OMVs, which was statistically higher than that of empty OMVs vesicles (Fig. 9B). Similarly, Elo⁺ OMVs captured by Ni beads showed varied affinity (Fig. 9A).

Both the Pb (II)-binding module and the OMV recovery module demonstrated detectable but relatively low working efficiency. There were a few possible causes for this problem, including:

Low OMV loading rate: The proteins (Elo especially) might not have been sufficiently expressed, which resulted in a lot of unloaded OMV vesicles. The empty vesicles would, of course, lack both Pb (II)-binding and Ni-binding functions.

Low OMV concentration: the OMV production rate of ^E.coli BL21(DE3) had not been modified, which could limit the overall efficiency of the system.

The SpyTag and SpyCatcher, though unsuitable for OMV recovery, are considered quite useful for the cross-linking challenge, given their infinite affinity feature. We utilized the SpyTag003 and SpyCatcher003 pair, for their reportedly higher efficiency. ⁶

The part Lpp’OmpA-2xEAAAAK we introduced is expected to further function here, where we worry that the short peptide SpyTag003 might be hard to meet SpyCatcher003 displayed on the other type of OMV. The rigid linker EAAAAKEAAAAK is expected to stabilize the SpyTag003 peptide’s orientation.

We built the two parts similarly as how we built Lpp’OmpA-2xEAAAAK-His10, with RE cleavage and ligation. We consider the 12 bp scar, translated into RGSA, should be of minimal influence to both peptides’ function. In order that we can detect the binding of the proteins, we added two antibody detection tags behind both peptides.

Fig. 11 Composition of Lpp’OmpA-2xEAAAAK-SpyTag003-3xFlag and Lpp’OmpA-2xEAAAAK-SpyCatcher003-HA

The two parts were separately used to transform two E.coli. BL21 strains, and the expression was verified via Western Blot.

(A) SDS-PAGE clearly showing the expression of SpyTag. (B)(C) Western blotting of SpyTag003 and SpyCatcher003.

We then tested whether the displayed SpyTag003 and SpyCatcher003 can still react and help cross-link the OMVs. We prepared the protein-of-interest-containing OMVs through ultracentrifugation, and mixed them at different dilution multiples. Since the production of SpyCatcher003 was more unstable and overall short in supply, we decided to reduce the influence of expression instability through another group of mixing with higher concentration of SpyTag003 OMVs used. The reaction continued overnight, until we tested the OMVs’ diameter distribution with NTA and the protein reaction with Western blotting.

Fig. 13 Verification of SpyTag003/SpyCatcher003 containing OMVs’ binding affinity.

i. It is a completely novel approach to try to create vesicle aggregates. The cross-linking might have been a completely failure because of the membrane fusion; yet as far as we detected, perhaps thanks to the addition of rigid linker 2xEAAAAK, no significant sign of \sqrt{2} times of the original diameter peak was detected, while we did detect high-diameter peaks, meaning that the cross-linking led to aggregation instead of vesicle fusion. ii. We didn’t have enough time to create a double transformant to verify whether cross-linking did help OMV recovery or not, and we highly doubt the expression extent of our current construction does not allow for sufficient expression for up to two surface display parts. Optimization on either chassis or expression regulatory elements must be done to allow for further development of our project. iii. The OMVs cross-linking efficiency is lower than we expected. Though it is largely due to the lowly effective expression, we further suspected that the 3xflag tag shielded the SpyTag, like what we suspected to happen on ClyA-PbrR-His6. Further implementation is necessary here.

Throughout the project, a significant obstacle before us is the rather low display efficiency. This could be due to multiple factors like restriction on peptide export efficiency. This can be implemented by introducing excessive peptide exportation system, like heterologous Sec and Tat protein families.

During our interview with Prof Tang Hongzhi (See Human Practice), he gave precious advice on chassis alteration. This will release us from great burden of cultivating the E.coli at 37 Celsius degrees.

The chassis engineering could be further extended to topics of utilizing strains higher in vesicle secretion yield, which is also frequently reported to be concerned with higher protein secretion activity.

His-tags are reliable and light, but not always perfect. In our project, 25 mM of imidazole – regular washing concentration – is enough to elute half of the OMVs binding to the nickel beads. Thus, we have come up with the idea of improving it with the newly reported mPKSeal strategy⁶, which adopts the Polyketone Synthase docking domains to increase binding specificity and tightness. It is possible that we stick displayed N-Docking domain on the Nitrocellulose membrane, with C-terminus on the OMVs.

We used to worry that SpyTag003 is too small to be recognized; yet adding a 3xflag C-terminus on it should have also been too much. A better solution for displaying small tags like SpyTag should be multivalent displaying, which increases exposure rather than decreasing it.

        [1]. Sun Nyunt Wai et al., ‘Vesicle-Mediated Export and Assembly of Pore-Forming Oligomers of the Enterobacterial ClyA Cytotoxin’, Cell> 115, no. 1 (October 2003): 25–35.
        [2]. Daniel Roderer and Rudi Glockshuber, ‘Assembly Mechanism of the α-Pore–Forming Toxin Cytolysin A from Escherichia Coli’, Philosophical Transactions of the Royal Society B: Biological Sciences 372, no. 1726 (5 August 2017): 20160211.
        [3].Marcus Mueller et al., ‘The Structure of a Cytolytic α-Helical Toxin Pore Reveals Its Assembly Mechanism’, Nature> 459, no. 7247 (June 2009): 726–30.
        [4].Kang, W., Ma, T., Liu, M. et al. Modular enzyme assembly for enhanced cascade biocatalysis and metabolic flux. Nat Commun 10, 4248 (2019).
        [5].Sun, X., Yuan, Y., Chen, Q. et al. Metabolic pathway assembly using docking domains from type I cis-AT polyketide synthases. Nat Commun 13, 5541 (2022).
        [6].Keeble AH, Turkki P, Stokes S, et al. Approaching infinite affinity through engineering of peptide-protein interaction. Proc Natl Acad Sci USA. 2019;116(52):26523-26533.>