Team:Edinburgh/Design

The SuperGrinder


The SuperGrinder: Design Engineering Tools Enzymes Enzyme immobilisation Structural modelling Protein expression and secretion




The SuperGrinder: Design

Synthetic biology has long been promoted as a pivotal technology in the fight against climate change through its ability to engineer biological processes to increase efficiency and sustainability [1]. We chose to use the flexibility of a Golden Gate style DNA assembly (JUMP assembly) to create and express fusion enzymes with different N/C-terminal tags to facilitate their immobilisation to silica surfaces. Select “Enzyme Immobilisation” to find out more details about our immobilisation strategy, and “Engineering tools” to learn more about our methods.

Select “Enzymes” to find out which enzymes we selected for expression. We also tested the addition of multiple protein secretion tags to the N-terminus of some enzymes, where secretion was expected to be required to facilitate over-expression whilst limiting host toxicity. Select “Protein expression and secretion” to find out more details.


JUMP plasmid architecture

Figure 1 Edinburgh 2021 iGEM’s ‘SuperGrinder’. The mechanical force of the system shears the material, allowing cellulases (red ovals) immobilised to silica beads (grey spheres) to come into contact with the cellulose (green molecules) within the lignocellulosic biomass. The SuperGrinder combines both enzymatic and mechanical treatment in a single step.

The resulting immobilised enzymes can then be applied to the SuperGrinder reactor (Figure 1). In our design, enzyme-laden silica beads are used to grind up insoluble waste polymers, whilst also enzymatically degrading the waste materials to soluble products which can be used for various up-cycling/re-cycling purposes. Immobilisation generally increases enzyme stability and re-usability, as well as enabling separation and recovery of enzymes from the soluble product, allowing a continuous/semi-continuous industrial process to become more viable. Check out the Implementation Page page to see how this could be implemented in real-life.


Engineering Tools

Explanation of the novel JUMP assembly method that we used for molecular cloning

Enzymes

Here we present the enzymes that we selected to work with, and the reasons for our choices

Enzyme immobilisation

Here we present the immobilisation options available and describe the silica-tags chosen for our project.

Structural modelling

Here we show the results from modelling our fusion proteins for tag selection

Protein expression and secretion

Here we explain our choices of parts used to optimise protein production


Engineering Tools

We used a Golden Gate modular cloning method to enable us to combine multiple parts, without requiring multiple PCR reactions which could introduce mutations. Parts are combined in the order defined by the specific 4 bp overhangs flanking each part. This modular system allows us to test multiple variations of the same part in high-throughput assays, helping us to re-iterate the design-build-test-learn engineering process. This iterative testing process is important in the absence of proper prediction tools to predict the behaviour of each part in combination with other parts and due to the “context dependency” that adds complexity to models of biological systems.

We used ‘Joint Universal Modular Plasmids’ (JUMP) as our vector system which are a series of vectors designed as a platform to increase flexibility of Modular cloning [1]. The plasmids are based on Standard European Vector Architecture (SEVA) and are compatible with PhytoBrick and BioBrick standards (and the Registry of Standard Biological Parts) (Figure 2). Assemblies can be inserted to the primary module (using BsaI or BsmBI, type IIS restriction enzymes), or secondary upstream/downstream modules using different restriction enzymes (AarI/BbsI), or BioBrick prefix/suffixes, allowing easy modification of the backbone. This is useful to add features, for example flanking homologous regions or alternative selection markers to enable expression in different hosts. Standard backbones containing various different origins of replication or selection markers are also available from the Addgene repository.


JUMP plasmid architecture

Figure 2 Basic architecture of a JUMP vector, showing the multiple cloning sites and SEVA origin of replication and selection markers. Source [1]

In the JUMP assembly scheme, basic parts are contained in level 0 vectors, which are assembled to form a single Transcription Unit (TU) in a level 1 vector (either 1A, 1B, 1C or 1D) using BsaI (a type IIS restriction enzyme). To simplify cloning and screening, a reporter gene is replaced by the assembled TU and the destination vector contains a different selection marker than insert-donor plasmids. Using a second type IIS restriction enzyme (BsmBI) and selection marker, multiple level 1 assemblies (1A, 1B, 1C and 1D) can be combined in a level 2 vector (2A, 2B, 2C or 2D). In JUMP, level 1 plasmids can be used as level 3 assembly destination vectors. The alternating use of restriction enzymes allows simple incorporation of one assembly into another assembly to enable multi-level hierarchical assemblies, for example to combine multiple TU’s [2].

As Type IIS restriction enzymes recognise a site away from the cut site, parts can be designed to leave custom overhangs which determine the order of assembly. Linear DNA parts can be designed to be flanked by overlapping BsaI and BsmBI sites. BsaI sites enable direct use in level 1 assemblies, and BsmBI sites allow for use in a level 0 assembly to create a Level 0 vector for part storage and replication [2]. Coding sequences were optimised for expression in either E. coli or B. subtilis using the Benchling codon optimisation tool, and internal restriction sites for BsaI, BsmBI, AarI, BbsI and the BioBrick enzymes EcoRI, SpeI, XbaI and PstI were removed during the part domestication process, so that parts were compatible with both cloning into secondary modules of JUMP plasmids and standard BioBrick cloning for submission to the iGEM repository. Finally, when designing parts for assembly, care was taken to ensure the reading frame was maintained between the start codon, N, O and C parts, as a 4 bp overhang can cause a frameshift. Correction for this creates a one or two amino acid joining scar in most cases.

We designed parts for Level 1 assemblies following one of the two schemes below (Figure 3), depending on whether the tag was being tested as an N- or C-terminal tag. For proteins that we also wanted to add a secretion signal to, we used C-terminal tags only, enabling us to assemble the construct using an N-terminal secretion signal in the place of the N-terminal tag. We used a single “composite” CT part, instead of separate C and T parts, as we did not want to test different terminators, thus the parts could be combined to reduce the number of parts required for the assembly reaction and increase the efficiency of assembly.


JUMP assembly scheme

Figure 3 A. C-terminal tag assembly scheme, showing 4 bp overhangs on each part, with part type code in the box. To assemble an untagged protein, one can use a CT part containing only a terminator B. N-terminal tag assembly scheme.


References
  1. 1. Valenzuela-Ortega M, French C. Synth Biol (Oxf). 2021 Feb 2;6(1):ysab003. doi:10.1093/synbio/ysab003. eCollection 2021
  2. 2. Valenzuela-Ortega M., French C.E. (2020) Joint Universal Modular Plasmids: A Flexible Platform for Golden Gate Assembly in Any Microbial Host. In: Chandran S., George K. (eds) DNA Cloning and Assembly. Methods in Molecular Biology, vol 2205. Humana, New York, NY. https://doi.org/10.1007/978-1-0716-0908-8_15

Enzymes

Cellulases

A variety of cellulases were investigated in our study for cellulose degradation. Exoglucanases hydrolyse β-1,4-glucosidic bonds from the non-reducing end of cellulose and endoglucanases cleave internal β-1,4-glucosidic bonds within the polymeric strand, exposing more ends for exoglucanase action. For optimal cellulose degradation, a combination of exoglucanase, endoglucanase and additional complementary activities such as beta-glucosidase and endoxylanase activity (for endohydrolysis of β-D-xylosidic linkages of xylan, another important polysaccharide found in agricultural waste) should be used.

We studied the enzymes detailed below:

Enzyme

Enzymatic action

Source organism

Cex

Exoglucanase, endoxylanase

Cellulomonas fimi

Chu2268

Exoglucanase

Cytophaga Hutchinsonii

CenA

Endoglucanase

Cellulomonas fimi

Keratinases

We focussed on two keratinases: KerA from Bacillus licheniformis, and MtaKer from Meiothermus taiwanensis.

Proteinase and chitin deacetylase

Extraction of chitin from chitinous waste (crustacean shells) involves three main steps: demineralisation, deproteination, and deacetylation. Enzymatic steps are involved in the deproteination and deacetylation. To deproteinate crustacean waste, Subtilisin E was used from Bacillus subtilis encoded by the aprE gene. Subtilisin E is an alkaline serine protease which can non-specifically hydrolyse the variety of protein in crustacean waste. Due to being well secreted from B. subtilis, it was chosen as the ideal proteinase for the project. After demineralisation and deproteination, the chitin product needs to be deacetylated into its more industrially favourable derivative chitosan. To do this, chitin deacetylase was used from the fungus Colletotrichum lindemuthianum encoded by the cda gene. Chitin deacetylase hydrolses the acetamido group from the chitin monomers producing acetic acid and the more soluble chitosan.

Enzyme Gene Enzymatic action Source organism
Subtilisin E aprE Alkaline serine protease Bacillus subtilis
Chitin Deacetylase cda Carbohydrate esterase Colletotrichum lindemuthianum

PET degrading enzymes

Dr Joanna Sadler, The University of Edinburgh, kindly offered the genes for double mutant PETase (W159H/ S238F) cloned in pET28a plasmid (Table 1) and the LCC-WCCG variant (F243W/ D238C/ S283C/ Y127G) gene cloned in pET22b plasmid (Table 2).


Mutant PETase sequences

  PETase (S238F/ W159H)

Austin et al., 2018 [1] described an improved PET hydrolase capable of 90% depolymerisation of PET into monomers over 10 hours. They used in silico docking and molecular dynamics (MD) simulations, revealing new information on substrate binding. Surprisingly, this double mutant (S238F/ W159H), inspired by cutinase architecture, had a greater capability for PET breakdown than wild-type PETase. The S238 mutation introduced new π-stacking and hydrophobic contacts between neighbouring terephthalate moieties, while the modification to His159 from a bulkier Trp allowed the PET polymer to sit deeper inside the active-site channel.

  Leaf-branch compost cutinase (LCC)-WCCG

Tournier et al., 2020 [2] utilised amorphous PET to compare the activity of various enzymes previously reported to hydrolyse PET under optimum circumstances. They discovered that leaf-branch compost cutinase (LCC) outperformed all other enzymes at 65°C with a melting temperature of 84.7°C. Unfortunately, when bottle-grade PET was used as a substrate, this depolymerisation performance of LCC at 65°C was significantly worse. After 2 hours, the reaction kinetics decreased rapidly, reaching just 53% conversion after 20 hours. In order to optimise the depolymerisation yield of the enzyme, they identified amino acids for mutagenesis with the help of molecular docking and contact-surface analysis. They found a D238C/S283C variant that allowed the thermal stability of the LCC enzyme at a melting temperature of 94.5°C, with a decrease in the activity of 28%. They then added a new variant to the new thermostable variant resulting in the F243W/D238C/S283C (WCC) variant that restored the activity of the D238C/S283C variant to at least wild-type LCC levels (98%) with a melting temperature that was 10.1 °C higher. Subsequently, added Y127G mutation, resulting in F243W/ D238C/ S283C/ Y127G (WCCG) that retained the specific activity of the LCC enzyme, similar to or higher than that of wild-type LCC. WCCG variant achieved the best conversion level of 85% conversion in 15 hours as compared to the wild-type LCC reaching only 53% conversion of PET in 20 hours.


References
  1. 1. Austin, H. P. et al. (2018) “Characterization and engineering of a plastic-degrading aromatic polyesterase,” Proceedings of the National Academy of Sciences, 115(19), pp. E4350–E4357. doi: 10.1073/PNAS.1718804115
  2. 2. Tournier, V. et al. (2020) “An engineered PET depolymerase to break down and recycle plastic bottles,” Nature 2020 580:7802, 580(7802), pp. 216–219. doi: 10.1038/s41586-020-2149-4.

Enzyme Immobilisation

The use of free enzymes in industry is limited by low stability under industrial conditions, poor recovery and slow reaction rates [1]. Enzyme immobilisation can provide desirable savings in cost and often in performance. Increased enzyme stability after immobilisation enables reactions to be performed at higher temperatures, or further from the optimal pH, and increases activity retention over time, allowing faster rates of reaction and more re-uses before replacement. Immobilisation can also make downstream enzyme-product separation much easier. However, different methods for immobilisation each provide different advantages and disadvantages. In some cases, immobilisation can compromise activity by impairing substrate diffusion, impacting active site orientation or causing protein distortion and rigidity.

The main immobilisation options are physical entrapment, adsorption methods, and chemical bonding. Chemical bonding tends to provide the highest specificity and affinity of attachment to the support, but often requires modification of the target protein, unlike physical entrapment and adsorption methods [2]. Due to the higher specificity of chemical bonding, this method can be used for affinity purification of tagged proteins as well as immobilisation. In addition, immobilisation using fusion tags is considered beneficial for maintaining orientation and accessibility of the enzyme to the substrate compared to physical adsorption and entrapment methods, thus it was our method of choice [3].

Immobilisation surface

We considered three materials as suitable for physical grinding of the waste materials: hydroxyapatite, functionalised magnetite nanoparticles (MNP) and silica (Figure 4).

Immobilisation particles

Figure 4 Summary schematic diagram of different options for immobilisation supports for the SuperGrinder.

Hydroxyapatite, Ca10(PO4)6(OH)2, is a porous non-toxic calcium phosphate salt which binds non-specifically to a variety of proteins [4]. The highly negative charge created by the many surface phosphate groups enables binding via a cation exchange mechanism, in synergy with metal chelation via carboxyl groups [5]. This mechanism does not require any protein modification, however the lack of specificity confers additional costs for upstream purification of the enzymes prior to immobilisation.

Functionalised MNP’s exhibit superparamagnetism, indicating that the nanoparticles are only magnetic in the presence of a magnetic field [6]. The simple use of magnets for enzyme recovery is seen as an important functionality for industry and continues to be an active area of interest. Functionalisation (typically coating in a support polymer) protects the particles from oxidation and allows enzyme attachment via different mechanisms. Lima et al. immobilised an industrial cellulase onto MNP coated with poly(methyl methacrylate) (PMMA), functionalised with glutaraldehyde. The immobilised cellulase showed lower activity than the free enzyme and less stability across its pH profile, however greater stability at higher temperatures [7]. When discussing the recyclability, the researchers noted that the leaching of PMMA-MNP is possibly an environmental hazard in upscaling to industrial use, an undesirable quality for the Super Grinder. However Konwarh et al. recorded a surprising fourfold increase in the activity of PEG-functionalised MNP immobilised keratinase in comparison with the free enzyme, suggesting that the effect can vary depending on the enzyme and/or functionalisation type [8].

Silica is a cheap and robust material, already commonly used in laboratories for bead beating. In 2006, Taniguchi and colleagues discovered that the conserved E. coli L2 ribosomal protein binds very strongly to the silanol groups of mesoporous silica, offering opportunities for L2-fusion proteins to be immobilised to unmodified silica surfaces [9]. L2-GFP was found to have a dissociation constant (Kd) of 0.7nM at pH 7.5. This represents a very strong binding affinity in comparison to other commonly used protein tags, such as the polyHis-tag which has a Kd of 1μM (1400-fold lower). However, in assessing L2-GFP, the researchers found that the recombinant protein had a fluorescence of only 70% of the free protein. Similarly, in the same study L2-luciferase was analysed for activity with respect to its free counterpart, in which only 30-40% of activity was retained after immobilisation. This indicates that the tag reduces the activity of the associated enzyme. One potential explanation may be the large size of the tag interfering with the protein function. Later studies have discovered a range of smaller tags. We chose to test 3 of these in addition to L2, as summarised below.


Silica binding tags

Tag Description Size (amino acids) Dissociation constant Reference
L2 Ribosomal protein from E. coli with high affinity for silica 273aa/30 kDa 0.7 nM [9]
L2NC N-terminal 1-60 and C-terminal 203-273 amino acids of L2 (silica-binding regions of L2 only) 130aa/14.23 kDa 1.6 nM [10]
Car9 Synthetic dodecapeptide originally identified as a carbon-binding peptide but shown to also have high affinity for silica 12aa/1.35 kDa 1 uM [11]
Sb7 Heptapeptide derived from the C-terminal sequence of spore coat protein CotB1 of Bacillus cereus 7aa/0.85 kDa 2.5 uM* [12]

*Dissociation constant of SB7 is in fact for another Si-tag which is a double repeat of SB7 (rqssrgrrqs srgr).

Optimising immobilisation

To test the performance of the different silica tags and optimise the immobilisation protocol, we assembled and expressed superfolder GFP with each of the available tags, according to the assembly scheme shown below (Figure 5).


GFP tagging assembly scheme

Figure 5 Schematic for the design of our tagged GFP constructs for use in testing immobilisation.


References
  1. 1. Chapman J, Ismail A, Dinu C. Industrial Applications of Enzymes: Recent Advances, Techniques, and Outlooks. Catalysts [Internet]. 2018 Jun 5 [cited 2021 Apr 23];8(6):238. Available from: http://www.mdpi.com/2073-4344/8/6/238
  2. 2. Garcia-Galan C, Berenguer-Murcia Á, Fernandez-Lafuente R & Rodrigues RC (2011) Potential of Different Enzyme Immobilization Strategies to Improve Enzyme Performance. Adv. Synth. Catal. 353: 2885–2904
  3. 3. Ikeda T, Hata Y, Ninomiya K ichi, Ikura Y, Takeguchi K, Aoyagi S, Hirota R & Kuroda A (2009) Oriented immobilization of antibodies on a silicon wafer using Si-tagged protein A. Anal. Biochem. 385: 132–137
  4. 4. Qi D, Gao M, Li X, Lin J. Immobilization of pectinase onto porous hydroxyapatite/calcium alginate composite beads for improved performance of recycle. ACS Omega [Internet]. 2020 Aug 18 [cited 2021 Apr 23];5(32):20062–9. Available from: /pmc/articles/PMC7439264/
  5. 5. Hou Y, Morrison CJ, Cramer SM. Classification of Protein Binding in Hydroxyapatite Chromatography: Synergistic Interactions on the Molecular Scale. Anal Chem [Internet]. 2011 [cited 2021 Apr 23];83:19. Available from: https://pubs.acs.org/sharingguidelines
  6. 6. Vaghari H, Jafarizadeh-Malmiri H, Mohammadlou M, Berenjian A, Anarjan N, Jafari N,et al. Application of magnetic nanoparticles in smart enzyme immobilization [Internet]. Vol. 38, Biotechnology Letters. Springer Netherlands; 2016. p. 223–33. Available from: https://link.springer.com/article/10.1007/s10529-015-1977-z
  7. 7. Lima JS, Araújo PHH, Sayer C, Souza AAU, Viegas AC, de Oliveira D. Cellulase immobilization on magnetic nanoparticles encapsulated in polymer nanospheres. Bioprocess Biosyst Eng [Internet]. 2017 Apr 1;40(4):511–8. Available from: https://link-springer-com.ezproxy.is.ed.ac.uk/article/10.1007/s00449-016-1716-4
  8. 8. Konwarh R, Karak N, Rai SK, Mukherjee AK. Polymer-assisted iron oxide magnetic nanoparticle immobilized keratinase. Nanotechnology [Internet]. 2009 May 12;20(22):225107. Available from: https://iopscience.iop.org/article/10.1088/0957-4484/20/22/225107
  9. 9. Taniguchi K, Nomura K, Hata Y, Nishimura T, Asami Y, Kuroda A. The Si-tag for immobilizing proteins on a silica surface. Biotechnol Bioeng [Internet]. 2007 Apr 15;96(6):1023–9. Available from: http://doi.wiley.com/10.1002/bit.21208
  10. 10. Kim S, Joo K Il, Jo BH, Cha HJ. Stability-Controllable Self-Immobilization of Carbonic Anhydrase Fused with a Silica-Binding Tag onto Diatom Biosilica for Enzymatic CO2 Capture and Utilization. ACS Appl Mater Interfaces [Internet]. 2020 Jun 17;12(24):27055–63. Available from: https://dx.doi.org/10.1021/acsami.0c03804
  11. 11. Coyle BL, Baneyx F. A cleavable silica-binding affinity tag for rapid and inexpensive protein purification. Biotechnol Bioeng [Internet]. 2014 Oct 1;111(10):2019–26. Available from: http://doi.wiley.com/10.1002/bit.25257
  12. 12. Abdelhamid MAA, Ikeda T, Motomura K, Tanaka T, Ishida T, Hirota R & Kuroda A (2016) Application of volcanic ash particles for protein affinity purification with a minimized silica-binding tag. J. Biosci. Bioeng. 122:

Structural modelling

To help predict properties of enzymes after the addition of a silica-binding fusion tag, we performed structural modelling using AlphaFold. We were interested in whether the tag would facilitate immobilisation in the right orientation so that the active site was accessible to the substrate, and also wanted to check if the tag itself might obstruct or distort the active site of the enzyme. Our results are reported below.

Electrostatic interactions between Cex and the silica surface

A fundamental property of protein adsorption is electrostatic interactions between the protein of interest and the silica surface. The surface of the silica support is negatively charged at pH greater than three due to the deprotonated silanol groups. As such, the protein of interest must have a positively charged face to adsorb1.

As illustrated in Figure 6, the surface charge of the Cex CD contains two negatively charged faces and thus will not stably interact with the silica surface as the repulsion will increase the free energy within the system. The association of the CBD to the silica surface will only contribute minimally to the overall stability of adsorption due to the weak electrostatic interactions between the mostly uncharged surface of the CBD (Figure 7) and silanol groups. It is thus expected only a minority of free Cex enzymes will naturally adsorb onto silica beads in solution and these interactions will be weak and easy to disrupt in industrial and practical settings.

The conjugation of silica binding tags to the N or C-terminus of the Cex is suitable to strengthen the required electrostatic interactions2,3. Three Si-tags were modelled: the truncated Escherichia coli ribosomal protein L2NC, L2NC N-terminally conjugated to a linker peptide (L2NCL)2, and the Car9 dodecapeptide3. The only modelled conjugated protein constructs were those in which the N-terminus of the tag was fused to the C-terminus of Cex. This reflects the synthesised constructs that were experimentally tested. All conjugated tags exhibit sufficiently positive surface charges for high-affinity silica binding whereby 29%, 28%, and 31% of residues in the L2NC, L2NCL, and Car9 tags respectively are positively charged2,3.

All Si-binding tags exhibit an unstructured and flexible conformation2,3, enabling each residue’s maximum surface area to make electrostatic interactions with the silica surface. It is important to note the strong positive charges of these flexible Si-tags may work to reduce the activity of Cex. As noted previously, the catalytic domain contains two negatively charged faces. These faces coincide with the active site residues of the glycanase. It is possible Cex may adopt an autoinhibitory conformation in which the flexible Si-tags may interact with the active site to disrupt its catalytic mechanism. This hypothesis is not demonstrated by this descriptive structural analysis and requires experimental verification.


beta-1,4-glycanase Cex catalytic domains surface charge

Figure 6 An image representing the surface charge (top) and cartoon tertiary structure (bottom in cyan) of two faces (the left and right columns) of the beta-1,4-glycanase Cex catalytic domains. The faces are related by a 180-degree rotation about the plane of the image. The colour in the surface charge representation indicates the charge where red indicates a negative charge, blue a positive charge, and white a neutral charge.


cellulose-binding domain of the beta-1,4-glycanase Cex surface charge

Figure 7 An image representing the surface charge (left) and cartoon tertiary structure (right in green) of the cellulose-binding domain of the beta-1,4-glycanase Cex. The colour in the surface charge representation indicates the charge where red indicates a negative charge, blue a positive charge, and white a neutral charge.

Silica binding improvements

The immobilisation of enzymes onto silica surfaces result in conformation changes that may disrupt the catalytic ability of the protein. One influencing variable to the susceptibility to these structural changes is the rigidity of the protein4. As stated previously, the Cex enzyme consists of two primary domains connected by a flexible linker, with the addition of a flexible and unstructured Si-binding tag. As a result of this confirmation, the enzyme is not rigid. To improve upon this, I think the Si-binding tag should be conjugated to the C-terminus of the CD. This removes the unneeded flexible linker and CBD that have no functional roles in the context of silica immobilisation. Additionally, due to the decreased surface area of these truncated Cex proteins, it is theoretically possible a greater number of the proteins can immobilise onto the silica surface, increasing each bead’s catalytic ability.

However, as illustrated in Figure 8, the positively charged Si-binding tags are even closer to the active site of the catalytic domain and thus an autoinhibitory conformation is probably even more likely. Having said this, there may be a conformational difference between the protein in solution and when immobilised on the silica surface. This can be investigated via molecular dynamic modelling and experimentally verified through experimental procedures like cryo-electron microscopy or FRET.


Molecular structure of Si-tagged Cex

Figure 8 An image representing the cartoon tertiary structure of the Cex catalytic domain (CD) in magenta conjugated to the specified silica-binding tag (green). The active site glutamate residues are highlighted in yellow. The N and C-terminals are labelled in each Cex construct.

Methodology

For each sequence specified below, the tertiary structure atomic model was predicted using Alphafold 2 (msa_mode: MMseqs2, num_models: 5, homooligomer: 1, use amber: no, use templates: yes)5 and visualised with Pymol. To determine the accuracy of the predicted model, the Cex CD and CBD were aligned to the experimentally determined CD (PBD: 1EXP, 1.80 Å) and CBD (PDB: 1EXG) respectively. If successful, the surface charge of the complex was visualised using the Adaptive Poisson-Boltzmann Solver Electrostatics plugin in Pymol.

Cex construct sequences

Cex-CD+L2NC (NC-NC):

AATTLKEAADGAGRDFGFALDPNRLSEAQYKAIADSEFNLVVAENAMKWDATEPSQNSFSFGAGDRVASYAADTGKELYGHTLVWHSQLPDWAKNLNGSAFESAMVNHVTKVADHFEGKVASWDVVNEAFADGDGPPQDSAFQQKLGNGYIETAFRAARAADPTAKLCINDYNVEGINAKSNSLYDLVKDFKARGVPLDCVGFQSHLIVGQVPGDFRQNLQRFADLGVDVRITELDIRMRTPSDATKLATQAADYKKVVQACMQVTRCQGVTVWGITDKYSWVPDVFPGEGAALVWDASYAKKPAYAAVMEAFMAVVKCKPTSPGRRHVVKVVNPELHKGKPFAPLLEKNSKSGGRNNNGRITTRHIGGGHKQVLGKAGAARWRGVRPTVRGTAMNPVDHPHGGGEGRNFGKHPVTPWGVQTKGKTRSNKRTDKFIVRRRSK

Cex-CD+L2NC Linker (NC-NC):

AATTLKEAADGAGRDFGFALDPNRLSEAQYKAIADSEFNLVVAENAMKWDATEPSQNSFSFGAGDRVASYAADTGKELYGHTLVWHSQLPDWAKNLNGSAFESAMVNHVTKVADHFEGKVASWDVVNEAFADGDGPPQDSAFQQKLGNGYIETAFRAARAADPTAKLCINDYNVEGINAKSNSLYDLVKDFKARGVPLDCVGFQSHLIVGQVPGDFRQNLQRFADLGVDVRITELDIRMRTPSDATKLATQAADYKKVVQACMQVTRCQGVTVWGITDKYSWVPDVFPGEGAALVWDASYAKKPAYAAVMEAFEGKSSGSGSESKSTMAVVKCKPTSPGRRHVVKVVNPELHKGKPFAPLLEKNSKSGGRNNNGRITTRHIGGGHKQVLGKAGAARWRGVRPTVRGTAMNPVDHPHGGGEGRNFGKHPVTPWGVQTKGKKTRSNKRTDKFIVRRRSK

Cex-CD+Car9 (NC-NC):

AATTLKEAADGAGRDFGFALDPNRLSEAQYKAIADSEFNLVVAENAMKWDATEPSQNSFSFGAGDRVASYAADTGKELYGHTLVWHSQLPDWAKNLNGSAFESAMVNHVTKVADHFEGKVASWDVVNEAFADGDGPPQDSAFQQKLGNGYIETAFRAARAADPTAKLCINDYNVEGINAKSNSLYDLVKDFKARGVPLDCVGFQSHLIVGQVPGDFRQNLQRFADLGVDVRITELDIRMRTPSDATKLATQAADYKKVVQACMQVTRCQGVTVWGITDKYSWVPDVFPGEGAALVWDASYAKKPAYAAVMEAFGGGSDSARGFKKPGKR

Cex cellulose-binding domain:

SGPAGCQVLWGVNQWNTGFTANVTVKNTSSAPVDGWTLTFSFPSGQQVTQAWSSTVTQSGSAVTVRNAPWNGSIPAGGTAQFGFNGSHTGTNAAPTAFSLNGTPCTVG

Cex catalytic domain:

AATTLKEAADGAGRDFGFALDPNRLSEAQYKAIADSEFNLVVAENAMKWDATEPSQNSFSFGAGDRVASYAADTGKELYGHTLVWHSQLPDWAKNLNGSAFESAMVNHVTKVADHFEGKVASWDVVNEAFADGDGPPQDSAFQQKLGNGYIETAFRAARAADPTAKLCINDYNVEGINAKSNSLYDLVKDFKARGVPLDCVGFQSHLIVGQVPGDFRQNLQRFADLGVDVRITELDIRMRTPSDATKLATQAADYKKVVQACMQVTRCQGVTVWGITDKYSWVPDVFPGEGAALVWDASYAKKPAYAAVMEAF


References
  1. 1. Meissner, J., Prause, A., Bharti, B. and Findenegg, G., 2015. Characterization of protein adsorption onto silica nanoparticles: influence of pH and ionic strength. Colloid and Polymer Science, 293(11), pp.3381-3391.
  2. 2. Soto-Rodríguez, J., Coyle, B., Samuelson, A., Aravagiri, K. and Baneyx, F., 2017. Affinity purification of Car9-tagged proteins on silica matrices: Optimization of a rapid and inexpensive protein purification technology. Protein Expression and Purification, 135, pp.70-77.
  3. 3. Kim, S., Joo, K., Jo, B. and Cha, H., 2020. Stability-Controllable Self-Immobilization of Carbonic Anhydrase Fused with a Silica-Binding Tag onto Diatom Biosilica for Enzymatic CO2 Capture and Utilization. ACS Applied Materials & Interfaces, 12(24), pp.27055-27063.
  4. 4. Secundo, F., 2013. Conformational changes of enzymes upon immobilisation. Chemical Society Reviews, 42(15), p.6250.

Protein expression and secretion

Host chassis selection:

We selected E.coli BL21 (DE3) and B. subtilis IIG-Bs27 as host strains. The E. coli BL21 (DE3) strain lacks specific cytoplasmic and outer membrane proteases to increase efficiency of recombinant protein expression. It contains the gene for T7 RNA polymerase under the lacUV5 promoter, allowing IPTG-inducible expression from the T7 promoter. It is straightforward to modify and has a developed genetic toolkit of well-characterised parts (including promoters and ribosome binding sites) enabling precise control of expression levels.

Bacillus subtilis is a gram-positive bacterium that has many advantages as a protease production platform. It is efficient at protein secretion, directly exporting proteins to the extracellular medium without needing to cross a second membrane, unlike gram-negative bacteria. This can allow production yields above 20g protein per litre culture, far exceeding intracellular protein production capacities [1]. We chose to test this alongside E. coli BL21 (DE3) for extracellular protease production. However, despite possessing many beneficial characteristics for industrial applications, B. subtilis can be limited by poor stability of plasmids after uptake. This contrasts with its high efficiency of DNA uptake and integration by homologous recombination. Therefore, we designed a strategy to convert the standard JUMP plasmid into an integrative plasmid, containing flanking 5’ and 3’ homologous regions to our selected insertion site, and a selective chloramphenicol resistance cassette, so that the construct would be stably integrated into the chromosome. More details can be found under “Engineering tools”.

Expression constructs:

E. coli and B. subtilis work optimally with different genetic toolkits, therefore we chose different promoters, ribosome binding sites and terminators for expression in each host.

Secretion:

The only known example of functional expression of keratinase in E. coli utilised a “ZZ secretion signal” to secrete soluble protein [2]. If secretion is not achieved, keratinase is generally found in insoluble inclusion bodies and requires extensive downstream processing/re-folding procedures to recover activity [3].

Due to repetitive sequence content the “ZZ signal” could not be synthesised by IDT, therefore we chose to screen a variety of other previously characterised signal peptides for functionality. Despite many studies, it is still not generally possible to predict which signal peptides will result in better secretion and therefore improved yields of any given extracellular protein. Chosen signal peptides needed to be compatible with the host’s export machinery to target the protein for secretion and maintain the target in a translocation active state. Therefore, screening was more likely to find a suitable option than rational design [4].

We screened for clones with efficient protease secretion by comparing zones of clearance on milk agar plates. Whilst hydrolysis of milk does not test specificity toward keratinous substrates, the screen is a quick and easy measure of secreted proteolytic activity.

We tested eight signal peptides in E. coli, assembled in JUMP Level 0 plasmids by Marcos Valenzuela-Ortega (Chris French lab, University of Edinburgh), and also tested secretion of KerA using the native KerA signal peptide from B. licheniformis.



References
  1. 1. Harwood CR & Cranenburgh R (2008) Bacillus protein secretion: an unfolding story. Trends Microbiol. 16: 73–79
  2. 2. Tiwary E & Gupta R (2010) Extracellular expression of Keratinase from Bacillus licheniformis ER-15 in Escherichia coli. J. Agric. Food Chem. 58: 8380–8385
  3. 3. Wang JJ, Swaisgood HE & Shih JCH (2003) Bioimmobilization of keratinase using Bacillus subtilis and Escherichia coli systems. Biotechnol. Bioeng. 81:
  4. 4. Degering C, Eggert T, Puls M, Bongaerts J, Evers S, Maurer K & Jaeger K (2010) Optimization of protease secretion in Bacillus subtilis and Bacillus licheniformis by screening of homologous and heterologous signal peptides. Appl. Environ. Microbiol. 76: 6370–6376