The invention relates to methods of purifying or isolating proteins or polypeptides using a magnetite binding domain attached to one or more magnetic nanoparticles, and additionally to constructs for use in such methods.

Naturally occurring organisms produce a wide range of different inorganic materials, such as calcium phosphate to form bones and teeth, calcium carbonate to make shells and spines, and the production of other materials. Magnetotactic bacteria take up iron from solution and biomineralize magnetite nanoparticles within organelles called magnetosomes. Mms6 is a small protein embedded on the interior of the membrane associated with such particles, and is known to be tightly associated with the formed material (see Staniland S.S. and Rawlings A.E., Biochem.Soc. Trans. (2016) 44, 883-890). Because of its association with magnetite, Mms6 has been used to study the binding of the protein to magnetite and additionally the role in the formation of magnetite nanoparticles.

Mms6 is produced by a number of different bacterial species within an initial protein size of approximately 12-15 kDa, with a truncated 6 kDa version found in magnetosomes. The truncated version is approximately 60 amino acids long and comprises a highly conserved glycine-lysine repeating sequence, together with a hydrophilic, acid rich, C-terminal amino acid region.

The Authors of the paper subsequently tried to identify further peptide sequences that might be used to assist in the production of magnetite nanoparticles (MMPs).

Rawlings et al. (Chem. Soc. - 2015, 6, 5586-5594) describes using phage selected magnetite interacting Adhiron. These are peptide display scaffold proteins that we used for molecular recognition applications. A phage display library of Adhirons each expressing two variable binding loops was screened to identify specific peptides that are able to interact with [100] faces of cubic magnetite nanoparticles. The selected variable regions identified displayed a strong preference for basic residues, with a particular preference for lysine residues Adhiron scaffolds that are described in detail in WO 2014/125290A. The scaffold, holding the two variable regions are based on plant cystatins. This reference is incorporated, herein in its entirety. These were stated to be needed to stabilise peptide aptamers. A heterologous peptide sequence is inserted into a loop region of the scaffold, the site of the peptide being between 3 and 30 amino acids. The system described in that patent application and used in Rawlings et al. Supra, uses two B sheet regions to stabilise a library of peptides at one end of each of two loops of the B sheet regions. The system is used to produce libraries of peptide, which were then screened for activity, such as for binding proteins. Two separate variable peptide regions each in a different loop are demonstrated in the patent application. Binding to a variety of different proteins is described using strep II tagged peptides.

The paper by Rawlings shows a number of peptide sequences based on two loops of the Adhiron scaffold. The authors concluded that the binding loops of the Adhiron functioned in a similar binding manner as polylysine which is a known binder of magnetite. This was suggested as allowing the production of cubic magnetite, thereby improving the production of such nanomaterials.

The production of nanoparticles of magnetite is also explored in Rawlings et al. (Nature Commun. 2019, 10, 2873 . This describes the use of magnetite binding peptides on coiled- coil scaffold proteins to control the production of magnetite nanoparticles. Such coiled coils typically comprise two or more alpha-helices which wrap around each other in a supercoiled assembly. These were used to mediate production of such nanoparticles. Moreover the binding of other peptides to other magnetic nanoparticles has also been demonstrated for materials such as CoPt and FePt (Bioconjugate Chem., 2020, 31, 1981- 1994) and cobalt-ferrite (J. Mater. Chem., 2011, 21, 15244) and has been used to improve production of such nanoparticles.

The inventors realised that the polypeptides bound to magnetic nanoparticles such as magnetite give the polypeptides the property of being attracted to a magnetic source. They realise that this could be used to separate or purify proteins, from, for example, a host organism that is used to produce them, or alternatively separate and purify proteins that have been used in other processes, such as the food industry, to make commercially useful food products.

A number of different microbial enzymes, for example, are currently used in the food industry, o-amylase is used in baking, brewing, starch liquefaction and the clarification of fruit juice. Proteases, are often involved in brewing, meat tenderising, coagulation of milk and bread quality improvement. Lactase are used in the production of pre-biotic food ingredients and in lactose intolerance reduction in people. Lipases are used in the production of cheese flavour development and Cheddar cheese production, as is phospholipase. Other enzymes include esterase in the production of fruit juices, dietary fibre and the production of short chain flavour esters, cellulose, xylanase, pectinase, glucose oxidase, catalase, peroxidase and asparaginase, which are all used in food and drink production. Additionally, for example, invertase catalyses the breakdown of sucrose into fructose and glucose to produce inverted sugar syrup which is used in a wide range of products and processes.

Such enzymes are often very expensive to produce. There is therefore a need to remove them from the food product once they are finished with and reuse them. It is also often desirable to remove them to prevent them continuing to react with the substrate. The inventors realised that if you could attach magnetite or other magnetic nanoparticles to such proteins, then it would be relatively easy to remove the enzymes by attracting the enzymes to a suitable magnetic source, allowing the rest of the food product to be removed or washed away.

Magnetic beads are generally known in the art. However, they are typically complex and expensive materials and are usually used in relatively small amounts in the production of specialist biological materials within the laboratory. Typically, these use a complex multilayer magnetic bead coated with the binding material for a biological of interest. Such beads are available from a number of commercial sources, but typically use a binding coating such as carboxylate, amine, neuravidin, streptavidin or protein A/G coatings to allow different materials to be adsorbed onto the surface of the magnetic bead.

The inventors realised that if the sequence of the protein of interest could be modified so that the binding component for magnetite or other magnetic nanoparticles could be incorporated into the sequence of the protein, then an improved system to allow the removal of such proteins, or indeed the production of other proteins or polypeptides of interest, could be produced.

A first aspect of the invention provides: a method of purifying or isolating a heterologous protein or polypeptide, comprising providing a sample containing the protein or polypeptide attached to a magnetic nanoparticle binding protein, the magnetic nanoparticle binding protein comprising a magnetic binding domain attached to one or more magnetic nanoparticles; exposing the sample to a magnetic source to which the magnetic nanoparticles bound to the magnetic binding domain attract; and washing non-attracted material from the magnetic source to leave the purified or isolated protein or polypeptide.

Magnetic nanoparticles are typically cobalt, iron, or nickel containing and include magnetite of the formula Fe3O4, doped magnetite comprising at least 85% magnetite plus additional inorganic material, ferrites such as cobalt-ferrites, CoPt, FePt or other magnetic nanoparticles. Most typically magnetite is the magnetite is the preferred compound. They are typically in the range 1-500 nm, most typically 5-200nm or 15-100 nm diameter and are typically cuboid in shape. The magnetic binding domain typically binds to the [100] surface of magnetite nanoparticles.

The magnetic source may be an electromagnetic source or, for example, a permanent ferromagnetic source. The magnetic nanoparticle binding domain is attached directly to the magnetic nanoparticles' surface allowing the protein or polypeptide to be attracted to the magnetic source via the magnetic nanoparticles. The protein or polypeptide can then be immobilised directly where the magnetic source is placed, for example at the side of a vessel, or the material is attracted towards the wall of the vessel where the magnetic source is placed, thereby attaching to the wall of the vessel until this magnetic source is removed.

The inventors have found that binding of the magnetic nanoparticle binding domain to the magnetic nanoparticles is often improved by attaching the magnetic binding domain to a scaffold protein. This may, for example, be in the form of a single or double β sheet loop, an o-helix, a coiled coil, a monobody scaffold, or indeed a free polypeptide sequence. Typically a scaffold in having a secondary structure, such as a single or double β sheet loop, an o-helix, a coiled coil, a monobody scaffold. These have been found to improve the presentation of the magnetic binding domain to the magnetic nanoparticles.

The term "protein" and "polypeptides" is often used interchangeably in the art. However, polypeptides tend to be considered to be a chain of amino acids linked together by peptide bonds and have a molecular weight of up to about 10,000 Da. Proteins tend to be considered to be larger than 10,000 Da molecular weight.

The protein or polypeptide may be covalently attached to the magnetic binding protein via a selectively cleavable linkage. A selectively cleavable linkage is one that typically allows the protein or polypeptide to be released from the magnetic binding protein at a time selected by the user of the method. A selectively cleavable link may simply be a disulphide bond, for example, through a cys-cys linkage, with the disulphide bond being reduced, for example, by application of dithiothreitol (DTT) a protease cleavage site (such as a TEV protease cleavage site) or a self-cleaving peptide. Self-cleaving peptides include, for example, intein sequences. This allows the heterologous protein or polypeptide to be released from the magnetic binding protein. The magnetic binding protein can then be subsequently removed from the heterologous protein or polypeptide by application of a magnetic source as described above.

The heterologous protein or polypeptide may be attached to the magnetic binding protein via a non-covalent linkage. This is typically between two moieties having high affinity for each other, such as via a (strept)avidin-biotin linkage.

The magnetic binding domain may comprise, for example, the binding domain of a naturally occurring bacterial magnetic nanoparticle binding protein such as Mms6, MmsF, Mmsl3, and MamC. However, the inventors recognised that using the techniques described in Rawlings et al (Supra) it is possible to produce peptides with greater affinity to magnetic nanoparticles such as magnetite than the naturally occurring proteins. Typically, the magnetic binding domain is 6-14 amino acids long, more typically 7-12, 8- 10 or 9 or 12 amino acids long.

The inventors have also found that such peptides in combination with the scaffold often bind to nucleic acids, such as RNA or DNA. This is a potential problem where, for example, the heterologous protein or polypeptide is being produced within a cell system, or indeed a cell-free system, where nucleic acids are present. They have unexpectedly found that some magnetic nanoparticle binding proteins are able bind magnetic nanoparticles, such as magnetite, in the presence of nucleic acids such as DNA or RNA. Accordingly, the magnetic nanoparticle binding protein may be pre-selected so that the presence of nucleic acids does not substantially inhibit binding of the magnetic binding domain to the magnetic nanoparticles, thereby allowing the production of proteins in systems where there are nucleic acids present. Typically, the magnetic nanoparticle binding domains have < 50% basic amino acids. They typically do not require the presence of two p-sheet loops. They have been unexpectedly found to only require a single p-sheet loop, or indeed only need the presence of an o-helix, coiled coil, monobody scaffold.

Such binding may be determined, for example using an ELISA to investigate binding between the proteins and magnetite or other magnetic nanoparticles in the presence of DNA, non-nucleic acid binders can identified as giving a positive signal in the ELISA for magnetite binding. A signal four fold above the baseline non-magnetite binding protein control may be used as a threshold. The magnetic nanoparticle binding domain may be Mms6, MmsF, Mmsl3, or MamC. However, more preferably the magnetic nanoparticle binding domain is selected from a magnetic nanoparticle binding domain having 6-14 amino acids, more typically 7-12, 8- 10 or 9 or 12 amino acids. More typically

A3 HNHKSKKHK

E8 AHMYTKAQT

F3 YVKSTKKK

G1 NWEKMVTYK

G4 QRAQSVSKK

H6 IHHIHHKHT

A8 GSYKKKKHN

B8 PAGKKHKPA

B9 ISKKKTFYV

C3 QMKKSKMWK

G3 IYKKKTLRL

G4 QRAQSVSKK

The following were found to bind less well to magnetite:

1 FHMPLTDPGQVQ

2 ASHTGTAQISRQ

3 HSLPHQPPYTTT

4 DPPIKSVRPSTT

5 KAALTSTAIPSL 6 YQQASDNELSYL

7 ASAVTDSISQYP

Additionally the following peptides have been found to be particularly advantageous as they have reduced, or substantially no binding to nucleic acids, compared to the control as defined above.

A8 GSYKKKKHN

B8 PAGKKHKPA

B9 ISKKKTFYV

C3 QMKKSKMWK

G3 IYKKKTLRL

G4 QRAQSVSKK And a two loop Adhiron binding peptide (with two domains, one on each loop):

H4 Loopl: QKFVPKSTN Loop2: PKKSKIELK

1 FHMPLTDPGQVQ

2 ASHTGTAQISRQ

3 HSLPHQPPYTTT

4 DPPIKSVRPSTT

5 KAALTSTAIPSL 6 YQQASDNELSYL

7 ASAVTDS I SQYP

The immediately above peptides (1-7) have been found to bind magnetite when bound to free peptide sequences, but not as well at those peptides bound to more constrained scaffold structures

Lio FePt binding peptide sequences include:-

VYNHMRQ,

HARMPWT,

YQGASEN,

EMSHFIA

CoPt binding peptides include:

KTHEIHSPLLHK (LSI)

HNKHLPSTQPLA (LS2)

KSLSRHDHIHHH (LS3)

SVSVGMKPSPRP (LS4)

VISNHRESSRPL (LS5)

The heterologous protein or polypeptide may be selected from a fluorescent protein (such as green fluorescent protein), a tag such as a strept II or HIS tag, an enzyme, a phage particle, an antibody or an antigen specific fragment of an antibody. Most typically, the polypeptide is an enzyme, especially an industrially used enzyme, including: o-amylases, proteases, lactase, lipases, phospholipases, esterases, cellulases, xylanase, pectinases, glucose oxidases, catalases, peroxidases, asparaginases, and invertases.

An antibody or an antigen-specific fragment of an antibody may also be especially used. The antibody may be a light chain-heavy chain dimer or a light chain-heavy chain-heavy chain-light chain tetramer. The fragment may be an F(ab), F(ab')2 or F(ab') fragment, a single domain antigen antibody fragment (SDAF) or indeed an antibody mimic, such as an affibody.

Typically, the heterologous protein or polypeptide is linked to the magnetic binding protein by virtue of being expressed as a single fusion protein. However, alternatively, the magnetic binding protein and heterologous protein or polypeptide may be expressed separately and subsequently joined, for example by using a suitable covalent linker, such as a bismaleimide crosslinker, such as bis(maleimido)ethane or l,4-bis(maleimido)butane, which are generally known in the art to crosslink sulfhydryl residues, for example, as presented on cysteine residues of proteins and polypeptides.

The protein or polypeptide may be an enzyme and the enzyme may be used to convert one or more substrates to one or more products and is subsequently removed from the products and substrates by a method according to the invention.

A further aspect of the invention provides a heterologous protein or polypeptide covalently attached to a magnetic binding protein, the magnetic binding protein comprising a magnetic binding domain. The magnetic binding domain may be attached to one or more magnetic nanoparticles as described above.

The magnetic binding protein and magnetic binding domains may be as defined above.

The protein or polypeptide may be as defined above, but are typically selected from a fluorescent protein, an enzyme or an antibody or antigen-specific fragment of an antibody as defined above.

Optionally, a selectively-cleavable link may be used to covalently attach the heterologous protein or polypeptide to the magnetic binding protein as defined above. Other aspects of the invention provides a nucleic acid encoding a fusion protein comprising a heterologous protein or polypeptide covalently bound to a magnetic binding protein comprising a magnetic binding domain, as defined above.

Nucleic acids encoding the specific magnetic binding domains described above are also provided.

A further aspect of the invention provides a plasmid or other expression vector comprising a nucleic acid sequence according to the invention. Cells, such as bacterial, yeast, mammalian, or plant cells comprising a nucleic acid according to the invention are also provided. Such cells, or cell-free systems comprising such nucleic acid sequences, may be used to produce the heterologous protein or polypeptide covalently attached to the magnetic binding protein, as defined above.

The invention will now be described by way of example only with reference to the following figures:

Figure 1

Phage ELISA of magnetite selected Adhirons. (A) Phage ELISA results where a colourchange indicates positive binding. The right hand panel shows ELISA controls of M13 phage, non-magnetite selected phage, and the selected phage pool after the final panning round. The left panel shows the range of binding of 72 clones from the final phage pool. (B) Frequency distribution showing the percentage of clones displaying a certain absorbance in the AP-Phage ELISA.

Figure 2

Sequence analysis of selected binding loops. (A) Seq2logo Kullback-Leibler plots of the amino acid sequences from the two binding loops. Residues in the positive area of the graph are enriched in the loops, and those in the negative area are depleted. (B) Frequency distribution of residue type in both binding loops

Figure 3

A schematic model of the coiled coil scaffold and the different peptide sequences displayed at the variable loop region. Figure 4

The E8CC construct.

Figure 5

UV-Vis spectra showing non-DNA binding of MIA E8 Adhiron peptide and E8CC coiled coil peptide.

Figure 6

Protein analysis of MIAAvi. SDS-PAGE band and m/z peak at predicted position for the correct molecular weight.

Figure 7

A schematic showing the design for the attachment of an antibody to a MIA protein using a sulfo SMCC linker group which forms a maleimide thioether bond with the sulfhydryl group of the terminal cysteine. The reaction mechanism diagrams were taken from the reagent manuals available on the Thermo Fisher Scientific webpages.

Figure 8

TEM image and grainsize analysis of the magnetite NPs used to attach the antibody. Average NP diameter was 5.58 nm ± 3.13 nm.

Figure 9

ELISA showing the binding of the MIA-RAGE complex to Strep-plate (cyan) and to magnetite NPs (black). Complex 1 and 2 are the same complexes but synthesised using different reaction conditions.

Figure 10

Gene string design ready to insert the scFv DNA seguence. The DNA seguence is shaded to aid identification of the function of each region. Figure 11

The manipulation of magnetite nanoparticles using a magnet.

Figure 12

Purification of GFP-magnetite binding protein-magnetite fusion protein from total cell lysate. Tot is the total crude lysate, Sol is the soluble lysate, Unb is the unbound material after nanoparticle addition, Cys and DTT refers to the concentration of either cysteine or reducing agent added to bring about release of the GFP.

Figure 13

GFP levels in different fractions of a) a GFP-intein-affimer fusion protein and b) a GFP- intein-peptide fusion protein. In both proteins the magnetite binding domain comprises peptide A3.

Figure 14

Binding of peptide E8 to magnetite. MB=monobody scaffold, CC=coiled coil scaffold.

Experimental

Isolation of peptides capable of binding magnetite particles with Adhiron scaffolds

The Adhiron scaffold is a well expressing protein. The library panned was an Adhiron library of over 1.3 x 1O ¹⁰ different sequences on two beta helix loops fused to a truncated pill coat protein of M13 bacteriophage (Tiede C et al Protein Eng. Des. Sei. 2014, 1 , 145-155).

Preparation of nanoparticles for phage display.

A total of 5 ml 0.2M Fe(acac)3 solution in benzyl ether was injected at 10 ml/h into a 15 ml solution of benzyl ether containing oleic acid, oleylamine and tetradecanediol at 290 °C. The solution was refluxed at 290 °C under argon for 4 h with constant stirring and then allowed to cool to room temperature. The nanoparticles were precipitated by adding anhydrous acetone and stirred for 2 h, magnetically extracted, and re-suspended in hexane containing oleic acid and oleylamine. This process was repeated five times to ensure that that unreacted iron compounds and benzyl ether were removed from the solution. The final product was stored in hexane under nitrogen in a sealed vial. A phase transfer process was used to exchange the nanoparticles into an aqueous phase ¹ . 5 ml of the nanoparticle solution was mixed with 5 ml of a 1.7% solution of TMAOH in degassed water and stirred for 12 h. Anhydrous acetone was added and the mixture centrifuged and the supernatant discarded. A small amount of TMAOH (25%) was added to the nanoparticles before re-suspension in water. When the particles were used for subsequent experiments, the nanoparticles were washed with degassed water three times.

Phage Display Panning

Magnetite nanoparticles were incubated with the phage library in 2x blocking buffer (BB) (Sigma) prepared in nitrogen sparged phosphate buffered saline with Tween (PBS-T) for one hour on a rotating bloodwheel. Nitrogen sparging minimises formation of alternative iron oxides on the particle surface. The particles were washed three times with fresh nitrogen sparged PBS-T to remove unbound phage. Bound phage were eluted with 0.2 M glycine pH 2.2, followed by triethylamine. The eluted phage were used to infect E. coli cells to amplify the phage pool and in addition the MNP particles were also mixed directly with E. coli cells to facilitate any further infection of bound phage. Cells were cultured and the phage isolated and used in a subsequent panning round.

Phage ELISA

Magnetite nanoparticles were mixed with 300 pl of 2 x blocking buffer (BB) (Sigma) prepared in phosphate buffered saline with tween (PBST) and supplemented with 2 pl of phage. The phage were allowed to bind to the particles over one hour with mixing, before transferring them magnetically using a KingFisher robot (Thermo) into 1 ml wash solution comprising 2 x BB PBST for 5 minutes with mixing. The particles were magnetically removed from the wash and placed into 2 x BB PBST containing rabbit Anti-fd bacteriophage (Sigma) at 1/1000 ratio for 1 hour in a final volume of 150 pl. This was followed by a washing step as before and then transferred into 150 pl of 2 x BB PBST containing Anti-Rabbit IgG alkaline phosphatase (Sigma) conjugated antibody at a 1/30,000 ratio for 1 h. Finally the particles were washed again before being introduced into 150 pl of freshly prepared BluePhos microwell reagent (KPL) for 15 minutes. The particles were then magnetically removed and the absorbance at 600 nm of each well was measured using a microplate reader. DNA sequencing

Based on the phage ELISA results, the 48 clones which showed the highest intensity were selected and used to infect E. coli cells and the phagemid vectors were subsequently extracted by mini-preparation. The Adhiron coding regions were sequenced using an M 13 promoter primer (Beckman Coulter Genomics) and the sequences aligned.

Protein production

The MIA- 1 and 2 coding sequences, as well as the control Adhiron were amplified by PCR and introduced into pPR-IBAl expression vectors (IBA) via conventional restriction cloning using Bsal restriction sites. The resulting plasmids encode a C- terminal StrepII tag ² to facilitate purification. The primers used were:

• Adhiron-IBAlf 5'-

ATGGTAGGTCTCAAAATGAAAAAGATTTGGTTGGCTCTGGCTGGTC-3'

• Adhiron-IBAl r 5'-ATGGTAGGTCTCAGCGCTCGCGGCCGCAGCGTCAC-3'.

A codon for cysteine was introduced on the end of the C-terminal tag via site- directed mutagenesis using primers:

• IBAlCys-f 5'-CGCAGTTCGAAAAATGCTAATAAGCTTGATCC-3'

• IBAlCys-r 5'-GGATCAAGCTTATTAGCATTTTTCGAACTGCG-3'.

The presence of the cysteine codon was confirmed by DNA sequencing. The target proteins were produced in E. coli BL21 (DE3) RP cells (Stratagene) using autoinduction medium (Formedium) at 37°C with vigorous shaking. The cells were harvested by centrifugation (20 minutes at 3,000 x g), resuspended in phosphate buffered saline (PBS) and lysed via sonication, and insoluble debris removed by centrifugation (45 minutes at 16,000 x g). StrepII tagged proteins were isolated from the soluble fraction by application to Streptactin resin (GE Healthcare) and elution with 2.5 mM d-Desthiobiotin dissolved in PBS. The eluate was loaded directly onto a HiTrap Heparin column (GE Healthcare) which was washed with PBS containing 500 mM NaCI. Bound Adhirons were eluted with PBS containing 750 mM NaCI and the eluate was dialysed against ultrapure water using 3.5 kDa MWCO snakeskin tubing (Thermo Scientific).

The results of the biopanning are discussed in detail in Rawlings et al (Supra). DNA sequencing of the Adhiron coding sequences of 48 clones ranked in terms of signal intensity in the phage ELISA. The binding loop sequences showed a strong preference for basic residues and in particular in a predominance of loops on lysine (16.4% of residues) and histidine (9.3% of residues) in loop 1. This is in contrast to the starting library which had approximately equal residues. This is summarised in Figure 1 and 2.

The top binding sequences found under DNA free conditions were:

Loopl : QRAQSVSKK G4

Loopl : QKFVPKSTN Loop2: PKKSKIELK H4

Loopl : PAGKKHKPA B8

Loopl : ISKKKTFYV B9

Loopl : QMKKSKMWK Loop2: QLRRNPEAH C3

Loopl : IYKKKTLRL G3

Production of alternative scaffolds to the Adhiron scaffold:

Coiled coil peptide displaying scaffold

Coiled coils contain two or more a helices wrapped around each other, creating a protein complex. An a-helix is one type of secondary structure a protein can adopt. It has 3.6 amino acid residues per turn allowing hydrogen bonding between the carbonyl (C=O) and amine (-NH) groups of the amino acids. A heptad repeat unit of amino acids is characteristic of coiled coil proteins. The residues of the heptad are labelled from a to g (and a' to g' on the opposite strand) based on their position in the coil hydrophobicity, charge and position of the residues within the helices dictates how they coil. Residues occupying the a and d positions are on the same surface and are usually hydrophobic residues such as alanine, valine, leucine or isoleucine, whereas charged residues such as glutamic acid, aspartic acid and lysine tend to occupy positions e and g. The electronic properties of opposing strands dictate the form that the complex takes. A coiled coil, containing two strands, can run parallel to each other, or alternatively can run in an antiparallel orientation. This orientation depends on the interactions between a/a' and d/d' (which often produce a hydrophobic core for the stabilisation of the complex). The coiled coil structure designed by Gurnon, D. G.; Whitaker, J.; Oakley, M. G., Design and characterization of a homodimeric antiparallel coiled coil. J. Am. Chem. Soc. 2003, 125 (25), 7518-7519, as modified in order to display variable peptides as a hairpin loop, linking the two helices. The formation of a bridge was predicted to increase the stability of the structure by reducing an unfavourable entropic effect upon interaction between two 'free' helical peptides. The aim was to synthesise an alternative protein scaffold that would closely mimic the characteristics of transmembrane biomineralisation proteins, but increase the solubility and expression yields compared to the wild-type proteins. The protein scaffolds displaying the active peptide regions of MmsF and Mmsl3 (and a control protein, AcrB), were designed and ordered as Gene Strings from Thermo Fischer Scientific. Ultimately, the MmsCC mimics would then be used in bio-inspired RTCP reactions for controlled magnetite synthesis. The peptide sequences displayed I the variable loop region of the scaffold are shown in figure 3. The figure also gives the peptide sequence of the homodimeric coiled coil domain.

The DNA for the MmsCC and AcrBCC were supplied by Dr Andrea Rawlings of the Staniland group at the University of Sheffield. The DNA sequences were incorporated into the pPR- IBA1 vector (IBA GmbH, Gottingen, Germany) inserted between the Bsal restriction sites. This plasmid was used to ensure efficient transformation in chemically competent E. coli cells, utilising T7 promoter-based expression. Also, the vector contained the gene sequence for carbenicillin/ampicillin resistance, aiding specific growth of bacterial colonies that had enveloped the recombinant DNA - preventing contamination. Prior to the construction of the recombinant DNA plasmid for MmsFCC, Mmsl3CC and AcrBCC, the DNA sequences for the inserts were designed and ordered as Gene Strings from Thermo Fisher Scientific. Complimentary forward and reverse primers were also synthesised for the amplification of the DNA sequence using PCR. The primers contained Bsal restriction sites for digestion and ligation into the pPR-IBAl vector.

Optimisation of the CC protein expression and purification steps was essential to ensure an optimum yield of the correct protein product, using the simplest and least timeconsuming methods. Each of the proteins, MmsFCC, Mmsl3CC and AcrBCC were produced following transformation of the recombinant DNA into chemically competent BL21 (DE3) E. coli cells (which is a widely used strain for T7 expression). Firstly, the process of protein expression was analysed and optimised, following by the purification process. Initially, two routes of protein expression were compared: expression using autoinduction; and expression using IPTG. Autoinduction appeared to produce higher yields of protein, when compared to expression using IPTG, therefore this route was analysed more extensively to further increase protein yields.

To optimise the purification protocol for the CC proteins, four mediums were tested with varying nutrient content. Each media was a mixture of salt, yeast extract and tryptone, in varying proportions. The four autoinduction media were: LB, 2 x YT, TB an SB. Two temperatures were also compared: 25 °C and 37 °C with all conditions measured between 0 and 60 hours. Cultures of E. coli containing known sequences of the CC DNA were grown and added to the autoinduction mediums containing the desired antibiotics before incubation at the temperatures tested. At each time point a sample of the autoinduction reaction was taken and spun to pellet the cells. For the dot blot technique to work the E. coli cells producing the CC protein needed to be lysed using sonication, resulting in the release of DNA. Sonication as a form of lysis appeared to work well. The lysed cell pellets were resuspended in dot blot buffer which contained salts and denaturant and a sample was added to a nitrocellulose membrane - this provided quantification of the total amount of protein produced. Following this, the remaining sample was subjected to centrifugation to remove the insoluble debris and leave the soluble protein fraction in the supernatant.

Further analysis of the intensities observed for the total protein fractions when subjected to the different growth conditions. This indicated a significant difference between the use of an SB autoinduction medium compared to the other medias, with TB media appearing to give the next best yield. This was expected as these mediums have a higher nutrient content compared to LB and 2 x YT mediums. The plot indicates that optimum protein production in SB media occurs at 37 °C around 48 hours. After 48 hours, the expression levels begin to drop as the E. coli levels increase to levels that are no longer sustainable at those conditions; oxygen is being used up resulting in levels lower than necessary to keep the cells replicating, causing some cell death. Higher temperatures often result in faster bacterial replication times therefore it makes sense that the expression levels would be higher at 37 °C compared to 25 °C at any given time point.

With protein expression optimised, the protein purification protocol was considered. The CC proteins were designed to incorporate a Hise purification tag to the N-terminus which was available for purification and detection purposes. A TAA stop codon was added to the end of the DNA sequence to prevent expression of the Strep-II tag encoded in the pPR- IBA1 plasmid. The Hise sequence is a small tag that was unlikely to result in any change of protein folding or conformation and so was the chosen purification tag for the CC proteins. As described above, it had been shown that the CC peptide had promising thermal stability which enabled purification to be carried out at room temperature or in a cold room (4 °C). The presence of the tag and the thermal stability of the complex ensured an uncomplicated start to planning the purification, however the relatively low solubility of the protein needed to be addressed. Therefore, a denaturation and refolding step was required as part of the protein purification process.

Following protein expression using autoinduction, the cells containing the CC protein were pelleted and resuspended in PBS buffer before the first round of lysis. The dot blot showed that the majority of the protein was in the total protein fraction with much lower amounts in the soluble fraction therefore the cells were then pelleted again using centrifugation, before resuspension in a denaturing buffer containing GuHCl. Refolding of the proteins was then achieved on a Ni-NTA column using washes with decreasing concentration of denaturation buffer. An additional wash step using a solution with a low imidazole concentration was used to remove any contaminants that were not tightly associated to the column; before the His6-tagged, CC proteins were eluted using a higher concentration of imidazole to displace the proteins. The Ni-NTA affinity chromatography purification was achieved using reusable Ni-NTA resin in a disposable column on the bench-top or alternatively using a HisTrap™ HP column supplied by GE Healthcare on the AKTA. The purification process was carried out on the bench-top with addition of the sample and buffers through a syringe and on the AKTA system. The purity of each of the CC proteins was assessed using both of the methods. Following purification, the eluted protein fractions were analysed using SDS-PAGE.

Each of the SDS-PAGE gels analysed appeared to contain protein of the desired molecular weight (15 kDa for MmsFCC; 16 kDa for AcrBCC and 17 kDa for Mmsl3CC), with an intense band observed for each of the samples. Suggesting the production of the CC proteins had been a success. However, some of the gels also, show a range of other bands, indicating that the protein samples contain contaminants, or have not refolded correctly, possibly leading to the formation of dimers or higher order oligomers. These samples were purified on the bench-top using addition via a syringe. Other gels were obtained from CC samples purified on the AKTA, based in the cold room. The eluted purified proteins in each of these cases appear to be much purer, with no extra bands present at unpredicted molecular weights. Each of the samples appeared to be clean, and correctly folded. Therefore, AKTA purification appeared to be superior to bench-top purification. This is likely due to the accuracy of the addition of the refolding buffers, with the correct volumes or the correct concentration added at the required time. Preventing combination and dilution of the buffers that may have occurred during manual use of the HisTrap™ column. It may also have been due to the activity of the resin in the different columns. The SDS-PAGE analysis of the purified CC proteins appeared to show the samples are clean and correctly folded but the proteins were also analysed using western blotting to ensure the observed protein bands were the His6-tagged proteins predicted. The protein was denatured/linearised and run on an SDS-PAGE gel to separate species present in the mixture by molecular weight. The contents of the gel were then transferred onto the nitrocellulose membrane, before the His6-tag fused to the N-terminus of the CCs was detected using a HRP-conjugated antibody system for the recognition of the tag. A photometric assay was used to amplify the HRP signal, catalysing the conversion of the chemiluminescent substrate for protein detection. The luminescent substrate was oxidised by the HRP, resulting in the emission of light. It was this photo-emission that was visualised. A signal would only be visible is the tag was present for the primary antibody to bind to and the only tagged proteins present were the CCs. Therefore, if a band was at the correct molecular weight, it could be accurately assumed that the protein was the protein of interest.

Magnetic Nanoparticle Binding Peptide

To demonstrate that the coiled coil could be used to present magnetite binding domains, the E8 magnetite binding peptide was displayed on the scaffold, see Figure 4.

This was synthesised in the same way as above and was shown to produce the peptide expressed on the scaffold and behaved in the same manner as other peptides tested on the scaffold.

The peptide E8 when expressed on Adhiron proteins was found to bind DNA molecules. The interaction of the coiled coil E8 construct (E8CC) was tested using the 260/280nm reading on a Nanodrop 2000 and compared to that of the Adhiron construct (MIA E8) and was not found to significantly bind DNA (see Figure 5). This appeared to show an advantage of using the coiled-coil scaffold over Adhiron-based scaffolds. This also had improved gel filtration properties over the Adhiron scaffold.

Attachment of Adhiron Magnetite binding proteins with magnetic binding MIA biotinylation for attachment

Biotinylation is the process of covalently attaching biotin to a protein. It can be used for attachment of other molecules to a protein through the biotin-avidin or biotin-streptavidin interaction, but also for labelling, detection and purification methods. This interaction is highly specific and one of the strongest known non-covalent interactions to form with a Ka of 10 ¹⁵ M- ¹ . Biotinylation is a useful method of labelling and attachments for proteins and other macromolecules due to the rapid bond formation between biotin and avidin. In addition to this once the biotin moiety has been attached it is insensitive to experimental conditions such as pH, temperature or solvent and due to biotins small size (244 g mol' ¹ ) there is minimal interference with the activity or folding of the protein, which is often much larger.

Introduction of a single C-terminal cysteine residue

Peptide displaying scaffolds are often designed to not include any cysteine residues, this enables a single cysteine amino acid to be added to the protein sequence at a later stage for specific targeted attachment - preventing heterogeneous labelling of the protein. Heterogeneous labelling attachment often leads to decreased binding between the protein and target ligand. The cysteine residue can be added to any point along the polypeptide chain, however it is most commonly fused to the end of the chain, therefore making it more likely to be accessible for binding and not disrupt the protein secondary or tertiary structure. The side chain of cysteine contains a thiol (-SH) group and so can be attached to drugs, dyes or antibodies through a disulphide bridge or maleimide linkage, both of these strategies were investigated.

QuikChange™ site-directed mutagenesis

A single cysteine residue was added to the C-terminus of the MIA proteins to enable homogeneous attachment of a variety of molecules. The extra amino acid was attached using the QuikChange™ site-directed mutagenesis method which enables site-specific mutations using double-stranded plasmid DNA as a template for the reaction. Forward and reverse primers that were complimentary to the opposite strands of the plasmid were designed; these primers also contained the sequence for the cysteine insertion mutation (the sequence of primers used is IBA1 Cys F: CGCAGTTCGAAAAATGCTAATAAGCTTGATCC IBA1 Cys R: GGATCAAGCTTATTAGCATTTTTCGAACTGCG. The bold letters indicate the inserted cysteine.

PCR was used to amplify the mutated MIA-containing pPR-IBAl vector, the reaction product yielded the new mutated plasmid, however this plasmid contained nicks. The DNA template (that did not contain the cysteine mutation) was digested using a Dpnl enzymatic digest, which cut all the remaining methylated parental DNA. The desired, mutated DNA was then transformed into chemically competent E. coli cells (XL10 Gold). The bacteria contained ligase to 'fix' the nicks in the plasmid and therefore the new DNA, containing the cysteine residue was able to infect the cells and colonies containing the desired mutation grew on LB agar plates (containing the required antibiotic). Individual colonies that had grown on the plates overnight were picked and grown in and LB medium containing antibiotic. The cells were then pelleted using centrifugation and prepared for sequencing using a MiniPrep kit. Sequencing for mutated MIAs A3 (A3C), E8 (E8C) and the control protein (non-binding Adhiron; controlC) came back positive - indicating the presence of a single C-terminal cysteine residue for each of the proteins. With the mutation reaction appearing to be a success, the protein was then produced. Protein expression and purification was conducted in the same way as the non-mutated MIA protein production (see section above. The cysteine residue didn't appear to affect the protein yields using these expression and purification methods, when compared to the yields achieved for the non-mutated MIAs.

Once the mutated proteins had been synthesised and dialysed into water, they were characterised using mass spectrometry. The resulting mass spectra showed the presence of dimeric species formed via disulphide bonds between the inserted cysteine residues.

The reversibility of the disulphide bridge formation was investigated, to ensure the cysteine residues could be prepared so that they were accessible for further labelling and attachment applications. The reduction of the bond was tracked using SDS-PAGE. The gels suggested that the intermolecular disulphide bridge that forms between the cysteine tagged MIA proteins can be easily reduced using reducing agents such as DTT or TCEP, ensuring that the cysteine tag is available for protein labelling and attachment. The bands observed around 12 kDa show the reduced cysteine tagged proteins, adopting a monomeric structure, whereas the bands present around 25 kDa are the dimerised MIA samples. The proteins were also analysed using western blotting to ensure that the proteins present on the SDS-PAGE were the desired proteins, which containing a Strep-II tag. This tag was recognised and bound using a HRP-conjugated antibody. The detection of bands using a chemiluminescent response indicated that the proteins were Strep-II and they were the proteins of interest.

The initial formation of dimers for A3C and E8C is promising as it suggests that the cysteine tag is present and accessible. Even more promising is that this disulphide bond can be reduced for binding of A3C or E8C to other thiol containing species.

Attachment via a maleimide linker

Although formation of intermolecular disulphide interactions has been shown through dimerisation of the cysteine tagged MIAs, the formation of this redox active link proved difficult for other molecules. Another common form of attachment via the thiol group of the cysteine sidechain uses a maleimide interaction. The maleimide functional group interacts specifically with sulfhydryl groups, such as the primary amine of cysteine, this reaction occurs at around neutral pH (pH= 6.5-7.5) therefore making it an ideal reaction to use as the MIAs are stable when stored in water or PBS. The reaction can be quenched at a higher pH, promoting the hydrolysis of the maleimide group. Another favourable feature of the reaction is that a TCEP reducing agent (shown to be successful in reducing the link between the MIA dimers) does not have to be excluded as the maleimide is not sensitive to it. A further advantage of using the maleimide linkage is that unlike disulphide bridges, the reaction is irreducible.

A fluorescent dye was attached via the C-terminal cysteine. The dye chosen for this function was the DyLight 650 Maleimide (Thermo Fisher Scientific), which had a useful Ex/Em of 652/672 nm. The dye gives impressive far-red fluorescence and has high water solubility enabling a good dye-protein ratio in protein samples stored in water. The labelling reaction occurs through the interaction between the reduced sulfhydryl group of the cysteine and the maleimide-activated dye, producing a stable thioether bond between the species.

The extent of fluorescent labelling was measured using a number of different methods: UV/Vis spectroscopy; SDS-PAGE; mass spectrometry; fluorescent spectroscopy/fluorescent microscopy. Firstly, UV/Vis spectroscopy was used to give an indication of the success of the labelling; giving the percentage of protein labelled using the maximum absorbance of the sample at 655 nm. The calculation used is shown below:

The dilution factor used for each labelling reaction was 2 (diluted in 1 x PBS); the protein concentration used was 50 μ M; the molar extinction coefficient for the DyLight 650 maleimide dye was 2.5 x 10 ⁵ M' ¹ cm' ¹ . An average measurement of A _max at 655 nm for A3fluoro and E8fluoro respectively was 2.7 and 5.5 resulting in the number of moles of dye per mole of protein calculated as: A3ffluoro = 0.43 and E8fluoro = 0.88. There is only one available cysteine residue (reactive sulfhydryl group) per protein and therefore values of 0.43 and 0.88 correspond to 43 % and 88 % labelled sample. SDS-PAGE and mass spectrometry analysis were used to confirm this calculation. Two labelling reactions were run in parallel to determine the method for optimum labelling. The first reaction was performed in the presence of TCEP and the second without (all reducing agent was removed prior to the reaction by dialysis). SDS-PAGE was used to compare the products obtained for each of the reactions Two intense bands for each of the wells containing fluorescently tagged MIA with the addition of the TCEP reducing agent were observed on the gel, the lower band around 12 kDa, the expected molecular weight for A3C and E8C; the upper band around 13 kDa, the expected molecular weight for the A3fluoro and E8ffluoro complexes. This suggested the labelling reaction produced the desired fluorescently tagged MIA complex to some extent, however the intensities of the bands observed on the gel suggest a lower percentage of labelled sampled compared to the UV/Vis measurements and calculation. Also, the upper band around 13 kDa for A3fluoro and E8fluoro was much less intense when TCEP was not used as a reducing reagent as the labelling reaction was run and there was a much more intense band around 25 kDa due to the likely formation of MIA dimers. This suggests that in the absence of a reducing agent, the formation of disulphide bonds between the cysteine residues of the MIAs is much more favourable than the formation of a thioester bond between the cysteine and maleimide activated dye. It may be possible to move the equilibrium towards thioester formation if a higher excess of dye was added to the reaction, however, the dye is relatively expensive and increasing the concentration required for labelling the MIA is not desirable.

ESI mass spectroscopy was used to confirm the extent of successful labelling of the MIA samples. Using the intensities of the peaks, the calculated percent of sample labelled closely matched the calculated percentage using the UV/Vis calculation above. For A3fluoro, using the data collected using mass spectrometry, the proportion on labelled sample was calculated to be 45 %, for E8fluoro the value was calculated as 95 %. It is possible these results varied when using SDS-PAGE as the dye may have been degraded by the harsh conditions used in the gel, however this has not been confirmed. Also, interestingly, the observed peaks for the fluorescently labelled MIAs vary slightly compared to the predicted m/z of the complexes. It is clear that the sample has been labelled as the values are around that expected for the addition of the dye and the errors calculated for both A3C and E8C samples were the same, each peak was 65 Da smaller than predicted from the labelling. This may be due to the dye having a slightly different molecular weight than stated by the supplier, or if may be due to degradation of the dye during mass spectrometry. However, no additional peaks at m/z = 65 appeared on the spectrum.

Finally, to test that a fluorescent signal was measureable for the DyLight 650-tagged MIAs, NPs coated in E8fluoro were analysed using fluorescent microscopy. A fluorescent signal on a black background should be observed, indicating the attachment of a fluorophore to the MNP (linked via the MIA). Due to extensive wash steps, no unbound fluorophore was present in the sample (this was confirmed in a control reaction). The fluorescently-tagged MNPs were analysed by fluorescent microscopy using a 633/647 nm laser line and Cy5 common filter set. The DyLight 650 dye can be used as an alternative to the Cy5 fluorescent dye, which is also a far-red-fluorescent dye and exhibits comparable performance. This showed the detection of the fluorophore on the NPs. This characterisation technique was used as an initial proof of principle and successfully showed that the fluorescently tagged MIAs had bound to the MNPs and the fluorescence was not quenched by the magnetite NPs.

Antibody attachment (RAGE and AP)

At Swansea University, a-RAGE, a novel biological target for ovarian cancer had been identified and was known to be over expressed in diseased tissue. Furthermore, ligands that bind this receptor are subjected to rapid internalization and membrane recycling. Antibodies that bind this cell surface receptor have been extensively studied and have been attached to magnetic nanoparticles for the development of targeted delivery and hyperthermia-based treatments of ovarian cancer. High expression of this receptor had been noted in a variety of human tumors including brain, breast, colon, colorectal, lung, prostate, oral squamous cell, and ovarian cancer, as well as lymphoma and melanoma. Within ovarian cancer specifically, high expression has been detected by immunohistochemistry. The strong association between the RAGE receptor and tumor has also been identified as therapeutically relevant, and mice deficient in the target receptor were resistant to chemically-induced skin carcinogenesis.

Following verification of the association between increased target expression and ovarian cancer, and the differential expression between malignant and non-malignant tissues, Professor Steve Conlan and his group at Swansea University had undertaken extensive analysis of antibodies that target this specific receptor. Rapid internalization and membrane recycling of bound ligands has been reported. Internalisation of binding antibodies within ovarian cancer cell lines was explored. Internalisation in as little as fifteen minutes had been observed (which is significantly faster than reported for the anti-human epidermal growth factor receptor 2 (HER2) antibody, Trastuzumab).

With the addition of a maleimide linked fluorescent dye showing that attachment via the MIA cysteine is possible through the formation of a thioether bond, new, the attachment of new, novel bodies. Antibody attachment using maleimide linkage to the MIA was investigated. AP and RAGE antibodies were used to test the attachment mechanism and optimise protocol. The design of the complex was independent of the antibody used and could be applied to the attachment of any antibody or protein to the cysteine labelled MIAs. Two complexes were designed (shown in figure 6): the first containing a single sulfo- SMCC linker; the second containing a cleavable dipeptide flanked by two sulfo SMCC groups. In the initial complex the sulfo-SMCC linker was attached to the MIA through the maleimide group, forming a thioether bond with the sulfhydryl group of the cysteine. The antibody would then be attached to the linker through the NHS ester group by formation of an amide bond with available lysine groups on the surface of the antibody. Attachment in this orientation prevented blocking of the MIA binding site which was previously witnessed for NHS ester binding to the lysine groups. Also, it would minimise the interference with the intramolecular disulphide bonds within the antibody, preventing a change in structure which could inhibit specific binding to the complimentary receptors and antigens. A schematic of this design can be found in figure 6.

The dipeptide complex was considered as it would enable selective cleavage of the MIA and therefore MNP from the antibody is required, giving a degree of flexibility to the applications the complex could be used for. The Val-Cit dipeptide site sits at the centre of the complex and is cleavable using Cathepsin B enzymatic cleavage. The peptide was designed and ordered from Thermo Scientific. The serine groups of the peptide act as a small spacer between the lysine and Vit-Cit, minimising steric bulk at the dipeptide region, aiding accessibility of the protease. The lysine residues flanking the peptide were used to attach the sulfo-SMCC linkers through the NHS ester group, forming an amide bond between the primary amine of the lysine sidechain and the NHS ester of the linker. The maleimide group was then used to form a thioether bond between the linker and the cysteine sidechain of the MIA and available sidechains on the antibody.

For the attachment of the sulfo-SMCC maleimide groups to the sulfhydryl groups on the antibody, 2-Mercaptoethylamine.HCI (2-MEA) was used to reduce any disulphide bonds. 2-MEA is a mild reducing agent, selectively reducing disulphide bridges present in hinge regions of IgG heavy chains. This method of IgG antibody reduction leaves intact essential intramolecular disulphide bonds and generates two functional half-antibodies with surface sulfhydryl groups accessible for labelling.

Binding of the dipeptide complex (MIA-x-pep-x-AB) was analysed using mass spectroscopy. The desired complex consisted of the peptide (KSVCitSK) flanked by two Sulfo-SMCC groups which were then linked to the MIA and antibody. The spectrum showed that the most intense peak was for the unbound Sulfo-SMCC with m/z = 431, this was predicted as the cross-linker was added in excess to the peptide. There were also two other major peaks observed with m/z equal to 1142 and 11556; these peaks correspond to the predicted peak position of the dipeptide fused to a single Sulfo-SMCC group and the dipeptide flanked by two Sulfo-SMCC groups respectively This indicated that a longer reaction time was required, however any unreacted product could be removed using sizeexclusion chromatography before use.

With mass spectrometry confirming the successful formation of the linker group, this linker could then be incubated with MIA and antibody. As both groups were attached to the linker through the formation of a thioether bond resulting in the reaction between reduced cysteine residues and a maleimide group, it was difficult to control the reaction. An excess of linker and cysteine-tagged MIA were used as it was likely the antibody contained multiple available sulfhydryl groups that would react. The formation of the final complex (MIA-x-pep-x-AB) was analysed using an ELISA. No detection antibody was required when the antibody used during the MIA conjugation was AP. The complex was allowed to bind through the MIA to MNPs; the presence of the AP antibody was then detected using a BluePhos detection reagent, resulting in a colorimetric response when the antibody was present. Negative controls applied included the dipeptide, the AP antibody and the Sulfo- SMCC cross-linker, each incubated with MNPs before the washing and detection steps. There was no colour changed observed for these controls, suggesting that if a colour change to blue/purple was detected, the complete complex would be present. There was a colour change from colourless to blue observed for the MIA-x-pep-AP complex. There was also a colour change observed for the MIA-x-pep-RAGE complex; this ELISA used a secondary antibody to specifically target the RAGE antibody. The intensity of the colour change was similar for both antibody complexes, suggesting that this could be a universal approach for attaching antibodies to MNPs.

Another complex designed and synthesised connected the cysteine-tagged MIA to the antibody directing through a Sulfo-SMCC linker. This complex was non-cleavable, suiting some applications and was advantageous over the production of the complex containing the peptide linker as the synthesis could be controlled. The MIA and antibody attach to the linker through different bonds and mechanisms, therefore one bond could be formed before the other. Using the manufacturer's instructions, the amide bond between the linker and antibody was generated first, followed by the thioether between the MIA and linker. It did not require the reduction of intramolecular disulphide bonds in the antibody, which could affect the antibody specify and function.

A number of different synthesis conditions were tested, including the use of varying ratios of Sulfo-SMCC to MIA. A 5-fold excess of the cross-linker to MIA was found to produce optimum results. This was analysed using an ELISA to test the binding intensity of the MIA-x-AP complex to MNPs. The greatest binding intensity was observed using a 5-fold excess of Sulfo-SMCC during synthesis. The complex was also analysed using western blot with a Strep-II tag specific antibody. A3C, Sulfo-SMCC and MIA-x-RAGE were run on a SDS-containing gel. The bands of the gel were then transferred to a nitrocellulose membrane, which was treated with Strep-II specific, HRP-conjugated antibody. There was no band present in line with the observed A3C band, indicating that there is no MIA present that has not be conjugated to the antibody.

This complex, consisting of A3C, fused to an a-RAGE antibody via a Sulfo-SMCC crosslinker was taken further to assess the capability of the complex to be used in vivo for biomedically useful applications.

Analysis of the ability of the antibody complex to bind the MNPs

The antibody is much larger than the MIA so analysis was required to ensure that binding ability of the MIAs to the NPs was not compromised. MNPs with an average diameter of 5 nm were used to study the biomedical implications of the ADC as smaller NPs should be taken up more easily by the cell. Visualisation of the NPs using TEM and analysis of their diameter is shown in figure 8. The image does show a small amount of impurity present in the sample, however this small amount was not expected to cause any problems as it is unlikely to be magnetic and will therefore be removed by magnetic extraction.

A binding ELISA was used to test the binding strength of the MIAs to the MNPs for the generation of the new ADC. Two independent ELISAs were performed. The first was an ELISA on the MNPs and the second was an ELISA on streptavidin coated plates. The HRP- conjugated antibody targeted the RAGE antibody that had been bound to the MIA, oxidation of the detection reagent by the HRP then produced a colorimetric response when the RAGE antibody was present. Positive binding resulted in a colour change from yellow to an intense blue/purple colour. This colour change was measured using UV-Vis spectroscopy, measuring the absorbance of the sample at 600 nm. A higher absorbance reading indicated a more intense colour change and therefore better binding.

The ELISA on the streptavidin coated plates was essential to show that the RAGE antibody was bound exclusively via the MIA. The Strep-II tag on the MIA was confirmed to bind directly to the plate, whereas the RAGE antibody alone did not bind to the plate. Therefore, for a positive binding result, the MIA must be present to bind to the plate and the RAGE antibody must be of been present in order to bind the HRP-conjugated antibody for detection. The data suggested that a minimal binding response was observed for the NPs, MIA or RAGE antibody alone, however when the ADC was analysed substantial binding was detected (figure 9, ADC labelled 'RAGE-MIA).

Once it had been confirmed that the RAGE had bound directly to the MIA and ELISA was performed on the MNPs. The same HRP-conjugated antibody was used for detection of binding. No binding colour change was observed for the NPs or RAGE antibody alone however binding was apparent for the RAGE-MIA ADC.

Analysis of the ELISA assays confirmed that the RAGE-antibody was bound to the MIA via the sulfo-SMCC linker. It also showed that the RAGE antibody bound MIA was still able to form strong interactions with the magnetite NPs and that a steric clash arising from the bulky antibody did not inhibit binding.

ScFv antibody attachment for biomedical applications

ScFv (single-chain variable fragments) are active, low molecular weight regions of antibodies which are used in preference to the full-size monoclonal antibody. These fragments are generated by creating a protein fusion between variable regions of the heavy (V _H ) and light (V _L ) chains. The chain regions are often connected by a short peptide region and the complete complex is then able to act as a functional antigen-binding fragment. There are numerous advantages of using scFv antibodies in place of monoclonal antibodies, including the possible production of these protein fusions using bacterial systems. Another advantage the small size is the rapid clearance from the blood stream and shorter wait times between administration and surgery when used in vivo for clinical diagnoses and therapies. The scFv used in this project was shMFE-23 which was designed and acquired from Prof Kerry Chester and her group at UCL. ShMFE is a known tumour associated antigen. ⁷ A carcinoembryonic antigen (CEA) specific scFv. The scFv was isolated from phage display libraries for the use in radioimmunodetection and radioimmunotherapy and since it's discover has been shown to successfully target colon cancer and has been in surgery to treat colorectal cancers. The scFv is interesting due to its tested diagnostic and therapeutic ability. It offers promising applications for the field of biomedicine - reducing the unfavourable entropic effect associated with the binding of the free peptide to MNPs and increasing thermal stability above 37 °C, therefore reducing thermal degradation in vivo compared to full molecular weight antibodies.

The aim was to attach the scFv to the MNPs by via protein fusion. Protein fusion design

A Gene String was designed and ordered from ThermoScientific that would be used to clone in the scFv generated at UCL. The initial DNA sequence of green fluorescent protein (GFP) was encoded into the gene string to aid the visualisation of the uptake of the complex by E. coli and as colonies with the gene present would appear green. Nhel and Notl restriction sites flanked the GFP gene to enable the exchange between GFP and the scFv DNA sequence using restriction digest and ligation.

A PelB leader sequence was added to the N-terminus to enable expression in the periplasm as inclusion body formation could be a problem when cloning the scFv. The PelB sequence self-cleaves and therefore would not be present in the purified protein. A Hise purification tag was incorporated into the DNA sequence of the gene string and was used for detection in blots and ELISAs as well as purification. It was followed by a TEV cleavage site which would allow cleavage of the His6 tag which is required for purification but also binds to some other transition metal ions, therefore the binding strength of the Hise tag to magnetite was investigated. This interaction could compete for MNP binding.

The DNA sequence for an AviTag was added to the designed gene string following the purification tag at the N-terminus (see figure 10) to offer the possibility of using specific biotinylation. If it became a problem that the AviTag was not at a terminus the TEV cleavage site would allow it to become an N-terminal tag. Therefore, fluorescent markers etc. may be incorporated via biotinylation. A short linker was included, acting as a spacer group to ensure that when the scFv insert replaced GFP the presence of bulky scFv near the Adhiron peptide would not alter the structure or sterically hinder the peptide binding. Also, it would allow the protein to protrude through any dextran coating on the NPs (currently used by UCL). Finally, as suggested, an Adhiron peptide (A3) is present at the C-terminus which will bind strongly and specifically to magnetite [100] facets allowing the protein to act as a coating for the magnetic nanoparticles and therefore increasing the biocompatibility of the MNPs. The A3 sequence is used as this has one of the highest binding strengths to magnetite and appears to be able to displace any bound DNA.

PURIFICATION OF PROTEINS USING MAGNETITE

Figure 11 shows the ability to manipulate and attract magnetite nanoparticles in a liquid such as water or a lysate, using a magnet. A number of different so called MagTag fusions of a magnetic binding protein linked to a heterologous protein or polypeptide have been generated to act as proof of concept studies for the isolation of target proteins/enzymes. MagTag - Adhiron (also known as an affimer) scaffold or peptide scaffold) is expressed as a fusion to the target protein, allowing it to be magnetically captured from cell lysate (either soluble fraction or crude lysate). Between the target protein and MagTag we have engineered different release mechanisms. This is either a self-cleaving protein (intein) which cleaves under certain chemical stimuli (e.g. addition of reducing agents such as DTT), or a specific protease recognition site allowing the target protein to be released by the addition of certain proteases (e.g. TEV protease).

In the example below, we bind a green fluorescent protein (GFP)-MagTag fusion to a suspension of magnetite nanoparticles and then release GFP using a DTT activated intein. This can be performed in both crude lysate (total lysate) as well as in the soluble lysate. There is limited washing of the nanoparticles required and the protein is released within a few hours (i.e. just a change from lysate to phosphate buffered saline and DTT). This is shown in Figure 12.

The GFP allows us to track the fluorescence throughout the various stages of the purification and compare different MagTag sequences. We compared a lead Adhiron scaffold-magnetite binding domain peptide (A3) with a combination peptide formed from the lead peptides acting as the so called MagTag. Using a serial dilution of purified GFP we could quantify the level of GFP in the various fractions. This analysis reveals that although the peptide MagTag construct does result in purified GFP it binds less well than the Adhiron/affimer version.

This is shown in Figure 13.

Influence of the presence of DNA on the binding of magnetite binding proteins, magnetite binding domains, and their scaffold proteins on magnetite nanoparticle binding.

When different peptides were screened using the Adhiron (also called the Affimer) peptide scaffold, peptides were found to have different affinities for the magnetite nanoparticles in the presence of DNA. This has the potential to affect the ability to use such constructs in the presence of DNA, such as cell lysates. This was also affected by the scaffold protein itself. The results below show the results using Affimer scaffolds and also the coiled coil peptide scaffold described above

This was tested as follows: ELISA Methodology for DNA free conditions (This is taken and adapted from the supplementary methods of the paper Rawlings et al. Chem Sci 2015.)

Magnetite nanoparticles (50 pL, 10 mg mL' ¹ ) are mixed with 300 pl of 2 x blocking buffer (BB) (Sigma) prepared in phosphate buffered saline with 0.1% tween (PBST). This is supplemented with 2 pl of amplified phage from a phage display experiment. The phage bind to the particles for hour with mixing.

Washing step. The particles are magnetically collected and placed into 1 ml wash solution comprising 2 x BB PBST for 5 minutes with mixing. Repeat 2 times. The particles are magnetically removed from the wash and placed into 2 x BB PBST containing rabbit Anti- fd bacteriophage (Sigma) at 1/1000 ratio for 1 hour in a final volume of 150 pl.

This is followed by a washing step as before, and then transferred into 150 pl of 2 x BB PBST containing Anti-Rabbit IgG alkaline phosphatase (Sigma) conjugated antibody at a 1/30,000 ratio for 1 hour.

Finally the particles are washed again before being introduced into 150 pl of freshly prepared BluePhos microwell reagent (KPL) for 15-30 minutes to produce a yellow to blue colour change. The particles are then magnetically removed and the absorbance at 600 nm of each well is measured using a microplate reader.

The absorbance of each well is compared to the absorbance arising BluePhos reagent alone. Samples with an absorbance greater than four times the control sample are considered to have a significant binding interaction.

ELISA Methodology for DNA present conditions

1 ml culture of E. coli cells expressing His-6 tagged binding proteins from an IPTG inducible promoter are collected via centrifugation and the supernatant discarded.

The cell pellet is resuspended in 300 pL IxPBS and the cells lysed on ice using sonication (20 seconds sonication in 5 second bursts using 40% duty). Magnetite nanoparticles (50 pL, 10 mg mL' ¹ ) are mixed with the lysate for 1 hour with mixing.

Washing step. The particles are magnetically collected and placed into 1 ml wash solution comprising 2 x BB PBST for 5 minutes with mixing and repeated 2 times. The particles are magnetically removed from the wash and placed into 2 x BB PBST containing Anti-6xHis tag alkaline phosphatase conjugated antibody at 1/5000 ratio for 1 hour in a final volume of 150 pl.

This is followed by a washing step as before.

The particles are magnetically recovered and introduced into 150 pl of freshly prepared BluePhos microwell reagent (KPL) for 15-30 minutes to produce a yellow to blue colour change. The particles are then magnetically removed and the absorbance at 600 nm of each well is measured using a microplate reader.

The absorbance of each well is compared to the absorbance arising from blank BluePhos reagent. Samples with an absorbance greater than four times that of blank BluePhos reagent are considered to have a significant binding interaction.

This best peptides found for magnetite are shown in the following table:

Top Affimer binder .sequences in the presence of DNA. Loop 1 sequences

Top Affimer binding sequences in the absence of DMA. Loop 1 sequences.

Top peptide binding sequences the absence of DNA

Binding to magnetite was also assayed using the following protocol:

Gel binding protocol

This was a protocol developed to assess the binding of the E8 peptide and to show that it can work. Typically an adapted ELISA (enzyme-linked immunosorbent assay) would be used to characterise the extent of binding to magnetite nanoparticles. The adapted method was to incubate a blocked nanoparticle solution with the protein of interest, wash and detect the presence and amount of the protein by incubating the particles with a protein specific antibody. In most cases it was a terminal poly-histidine tag that was being detected with a polyclonal antibody. The need for an alternative protocol arose from the fact that the terminally poly-histidine tagged monobody (also known as FN type III) (Koide et al. J Mol Biol'. 2012;415(2):393-405. doi: 10.1016/j.jmb.2011.12.019 ) protein was not detectable by the anti-poly-histidine antibody (as seen by running a Western blot and a dot-blot). This may have been caused by the poly-histidine tag being buried within the protein globule, thus making it inaccessible for antibody binding.

Amino acid sequences

Three proteins were used to produce the binding results. His-CC-E8 (a coiled coil scaffold previously shown to bind magnetite nanoparticles), His-MB-E8 (the monobody scaffold containing the E8 binding loop), His-MB-Ctrl (the monobody scaffold containing a control sequence previously shown not to bind magnetite nanoparticles)

His-CC-E8:

MGSHHHHHHHHGSTENLYFQGPSMKQLEKELKQLEKELQAIEKQLAQLQWKAQARKK KLAQLKKK LQAAHMYTKAQTGMKQLEKELKQLEKELQAIEKQLAQLQWKAQARKKKLAQLKKKLQA

His-MB-E8

MGSSHHHHHHGGVPRDLEVVAATPTSLLISWDAPAVTVRYYRITYGETGGNSPVQEF TVPGSKSTA TISGLKPGVDYTITVYAVTGAHMYTKAQTPISINYRTEID

His-MB-Ctrl

MGSSHHHHHHGGVPRDLEVVAATPTSLLISWDAPAVTVRYYRITYGETGGNSPVQEF TVPGSKSTA TISGLKPGVDYTITVYAVTGDWWEAGVFMPISINYRTEID

*the loop regions are highlighted in bold

Protein production

The protein coding DNA sequences were inserted into either pPR-IBAl vector for the coiled-coil construct or pET28a vector for the monobody constructs. In both cases the vectors contained an inducible T7 promoter upstream of the protein coding region. E.coli BL21 DE3 Rosetta cells were transformed with either of the plasmid and grown in autoinduction medium (Formedium) for 40 hours at 37 degrees with 225 revolutions per minute shaking. The monobody producing cells were harvested, lysed and the soluble fractions were subjected to pre-packed 1ml AminTrap His-Tag Protein Purification Columns (Expedeon, Abeam, UK) using an AKTA pure (GE Healthcare) purification system. The coiled-coil scaffold producing cells were harvested and lysed in an 8M GuHCI buffer, the soluble fraction was then incubated with free Amintra Ni-NTA resin (Expedeon, Abeam, UK) and washed with buffers containing decreasing amounts of GuHCI to refold the coiled- coil scaffold. In both cases the eluted protein was further purified using gel filtration chromatography (Superdex 200 Increase 10/300 GL, GE Healthcare) with final elution in PBS at pH 7.4. The presence of the correct protein was assessed by comparing the obtained gel filtration chromatograms to standard calibration curves for globular proteins and by determining exact molecular weight by electrospray ionization mass spectrometry. The correct folding of the purified proteins was determined by taking circular dichroism (CD) spectroscopy readings on Jasco J-810 spectropolarimeter and comparing the spectra to known literature values. The protein concentration was obtained by taking absorbance readings at 280 nm and adjusting the values for the absorbance coefficient of each protein.

Magnetite binding assay

Magnetite nanoparticle were prepared using a Room Temperature Co-Precipitation method (Macolo et al. Materials (Basel). 2013;6(12):5549-5567. Published 2013 Nov 28. doi: 10.3390/ma6125549 ) that produces magnetic nanoparticles averaging at size of 30 nanometers. A 30 mg/ml magnetite mixture was sonicated on ice for 1 minute at 40 % amplitude to separate aggregate particles. Proteins being studied were prepared in equal molar concentrations and were serially diluted in PBS pH 7.4. 35 pl of the magnetite solution was then incubated with 250 pl of each protein dilution at room temperature for 5 hours and 30 minutes on a rotator in room temperature. The protein bound particles were then magnetically and washed with PBS-T once. The washed particles were incubated with a 100 mM glycine solution at pH 2 for 5 minutes to elute the bound protein. The solution was then neutralized with 1 M tris pH 7.4. The particles were then again collected magnetically and the solution containing the protein was removed. The protein solution was mixed with and SDS-loading buffer(0.2 M Tris-HCI, 0.2 M DTT, 0.4 M SDS, 277 mM, 8.0% (w/v) Bromophenol blue, 6 mM Glycerol: 4.3 M) and incubated at 95 degrees for 10 minutes. 9 pl of each sample was then applied to individual wells on a 12 % polyacrylamide stacked Tris-Glycine gels and run until full separation. The gels were then stained for an hour each with InstantBlue® Coomassie Protein Stain (Expedeon, Abeam, UK). The extent of particle binding for each protein was assessed by quantifying the intensities (with respect to the background) of the bands corresponding to the correct protein size. The quantification was done using the volume tools option found in the Image Lab gel analysis software package (BioRad, USA). Results are shown in Figure 14

FePt binding peptides

A phage display library of peptides was screened in a similar manner to those used to look for peptides that bound magnetite to identify a high affinity binding peptide to be used as a biotemplating molecule in Lio FePt synthesis. Initially a three panning round phage display protocol was performed using a 7 amino acid peptide phage library, which identified 4 sequences of high affinity binding peptides: VYNHMRQ, HARMPWT, YQGASEN, EMSHFIA. The first two of the four sequences made up over 80 % of the identified peptides and were recognised as the most likely to bind with the highest affinity to Lio phase FePt due to their commonality.

CoPt binding peptides

The production of CoPt binding peptides is shown in Jarraid R. M. et al. (Bioconjugate Chem 2020, 31, 1981-1994), incorporated herein in its entirety. The paper shows the use of CoPt binding peptides to produce CoPt nanoparticles and shows that the production of binding peptides can be produced. The five lead peptides identified in that paper are listed below:

KTHEIHSPLLHK (LSI)

HNKHLPSTQPLA (LS2)

KSLSRHDHIHHH (LS3)

SVSVGMKPSPRP (LS4)

VISNHRESSRPL (LS5)