key: cord-0319794-physdyhx
authors: Macpherson, Alex; Birtley, James R.; Broadbridge, Robert J.; Brady, Kevin; Tang, Yalan; Joyce, Callum; Saunders, Kenneth; Bogle, Gregory; Horton, John; Kelm, Sebastian; Taylor, Richard D.; Franklin, Richard J.; Selby, Matthew D.; Laabei, Maisem; Wonfor, Toska; Hold, Adam; Vadysirisack, Douangsone; Shi, Jiye; van den Elsen, Jean; Lawson, Alastair D.G.
title: The chemical synthesis of knob domain antibody fragments
date: 2021-06-17
journal: bioRxiv
DOI: 10.1101/2021.06.16.448769
sha: 1e2d56517c7cd22bef9cf8030de1f88cc4233929
doc_id: 319794
cord_uid: physdyhx

Cysteine-rich knob domains found in the ultralong complementarity determining regions of a subset of bovine antibodies, are capable of functioning autonomously as 3-6 kDa peptides. While they can be expressed recombinantly in cellular systems, in this paper we show that knob domains are also readily amenable to chemical synthesis, with a co-crystal structure of a chemically synthesised knob domain in complex with antigen showing structural equivalence to the biological product. For drug discovery, following immunisation of cattle, knob domain peptides can be synthesised directly from antibody sequence data, combining the power and diversity of the bovine immune repertoire with the ability to rapidly incorporate non-biological modifications. We demonstrate that, through rational design with non-natural amino acids, paratope diversity can be massively expanded, in this case improving the efficacy of an allosteric peptide. As a potential route to further improve stability, we also performed head-to-tail cyclisation, exploiting the unusual proximity of the N- and C-termini to synthesise functional, fully cyclic antibody fragments. Lastly, we highlight the stability of knob domains in plasma and, through pharmacokinetic studies, use palmitoylation as a route to extend the plasma half-life of knob domains in vivo. This study presents an antibody-derived medicinal chemistry platform, with protocols for solid-phase synthesis of knob domains; together with characterisation of their molecular structures, in vitro pharmacology and pharmacokinetics.

Around 10 % of bovine IgM and IgG antibodies are endowed with ultralong CDRH3 1 , where a knob domain, a series of mini loops stabilised by 2-5 disulphide bonds, is held atop a β-ribbon stalk, over 40 Å from the neighbouring CDRs [2] [3] [4] . This striking motif arises as a consequence of limited immune gene segment diversity and is aided by priming mechanisms to introduce cysteine residues during somatic hypermutation 5 .

We have previously shown that knob domains can function autonomously to bind antigen independently of the bovine antibody scaffold 6, 7 . This creates antibody fragments of just 3-6 kDa, so small as to be considered peptides. One notable advantage is that, through immunisation, the vast diversity of the bovine immune system can be exploited to produce high affinity and structurally complex peptides.

Small antibody fragments have previously been isolated from sharks and camelids, the VNAR 8 and VHH 9 respectively, which are attractive due to their small paratopes and, in the case of certain VHHs, long CDR loops 10 . Due to their size and structural complexity, recombinant antibody production is typically reliant on DNA engineering and mammalian or microbial cellular machinery. Conversely, A functional, full-length anti-HER2 VHH has been produced by solid-phase synthesis 11 . However, due to the inefficiency of performing solid-phase peptide synthesis on long polypeptide chains, this approach is exceptional.

Chemical synthesis of antibody fragments would be advantageous in certain situations. For example, for antibody fragments where the Fc domain is absent and binding to serum albumin has not been engineered, extension of the plasma half-life (t1/2) by the neonatal Fc receptor is impossible 12 . For fragments of less than 60 kDa, renal clearance will markedly reduce exposure 13 . While protease instability can further reduce active concentrations of peptidic or antibody fragment-based drugs.

Medicinal chemists employ a range of peptide design strategies to attenuate renal clearance and extend t1/2, such as increasing molecular weight by PEGylation 14 or conferring reversible binding to serum albumin through conjugation of fatty acids, such as palmitic acid 15 . Head-to-tail cyclisation of peptides is an established approach to limit proteolysis 16 , as is stabilisation of tertiary structure by disulphide bonds 17 ; a feature which is already inherent to cysteine-rich knob domains. For drug manufacture, working independently of cell systems may lower manufacturing costs 18 , while the stability conferred by an abundance of disulphide bonds may remove the requirement for a temperature-controlled supply chain.

The first knob domain peptides were isolated through bovine immunisation with human complement C5 protein 6 . C5 is a 188 kDa soluble glycoprotein of the complement cascade and the target of the approved monoclonal antibodies eculizumab 19 and ravulizumab 20 . It remains the focus of intensive therapeutic research, with further antibodies 21 , peptides 22 , aptamers 23 and small molecules 24 in clinical and pre-clinical development 25 , with a view to treating complement induced inflammation and autolysis.

Of the four knob domains which we reported to tightly bind C5, three were functionally modifying and two were demonstrably allosteric (as defined by partial antagonism at asymptotic concentrations) 7 . Notably, one knob domain, K92, demonstrated that selective allosteric inhibition of the alternative pathway can be achieved through C5 7 . Co-crystal structures of the C5-knob domain complexes highlighted the molecular interactions underpinning binding and showed that the knob domain peptides adopted 3-strand β-sheet topologies and were constrained by varying numbers of disulphide bonds 7 . Broadly similar folds are endorsed by the cysteine-rich defensin 26 and various venom peptides 27 which are ubiquitous as tools of innate immunity and defence, across all clades, and which are also the focus of drug discovery efforts 28, 29 .

Here, we produce previously reported C5 modifying knob domain peptides 6,7 by chemical synthesis and explore their biological activity, structure and in vivo pharmacokinetics. This study presents antibody-derived medicinal chemistry as a simple approach to modify knob domains to aid the discovery of drug candidates.

Four knob domains have previously been reported as tightly binding C5 when expressed recombinantly in a HEK293 cell line and purified from a cleavable protein scaffold: K8, K57, K92 and K149, with each peptide stabilised by two or three disulphide bonds ( Figure 1B) 7 . By surface plasmon resonance (SPR), we measured equilibrium dissociation constants (KD) for binding of these recombinantly expressed peptides to human C5 (17.8 nM for K8, 1.4 nM for K57, <0.6 nM for K92, and 15.5 nM for K149) 6 .

In this study we performed solid-phase peptide synthesis by two methods: firstly, a site-directed method (henceforth suffixed chemSD), whereby cysteines in the K149 peptide were specifically protected and deprotected to form disulphide bonds in a preordained manner; and secondly, using a free energy method (henceforth suffixed chemFE), where, for all four peptides, thermodynamiccontrolled air oxidation was used to obtain the minimum energy form of the disulphide bonds.

For peptides produced by both methods, liquid chromatography/mass spectrometry (LC/MS) confirmed that purities were in excess of 95 % and that masses consistent with the predicted amino acid sequences were unanimously present. A complete list of the knob domain peptide sequences is shown in S1, with accompanying LC/MS characterisation shown in S2.

Binding of knob domain peptides to human C5 was measured by SPR, using a multi-cycle kinetics method. The chemical knob domains produced by the free energy method bound C5 with high affinity ( Figure 1C and Table 1 ), equivalent to values previously reported for the purified recombinant forms of the peptides 6 . For K149, which has two disulphide bonds, the adjacent cysteines C15 and C16 are unable to pair, giving rise to only two potential disulphide bonding arrangements: K149chemSD-A (C2-C15, C16-C22) and K149chemSD-B (C2-C16, C15-C22), shown in Figure 1A . Both forms were synthesised in a site directed manner, and, while both bound C5, K149chemSD-B displayed approximately 35-fold lower affinity, with a markedly slower on rate. When produced by the free energy method, K149chemFE bound C5 with equal affinity to the higher affinity K149chemSD-A form.

As mispairing of K149 disulphide bonds was tolerated to a certain extent, we next tested the effect of removing disulphide bonds entirely, through reduction and capping of cysteines with iodoacetamide (IAM). Following high resolution LC/MS analysis, to confirm uniform capping had occurred (shown in S3), binding to C5 was again measured by SPR (shown in S4). For K149chemFE, reduction and capping of cysteines entirely abrogated binding, while for K92chemFE, it resulted in a substantial drop in affinity from 411 pM to 2 µM, predominantly mediated by a decrease in on rate, potentially due to a loss of tertiary structure. Removal of disulphide bonds in K57chemFE also affected binding affinity but a reasonable KD of 76 nM was retained.

Having demonstrated binding to C5, biological function was evaluated in a range of complement ELISAs and erythrocyte haemolysis assays that were specific for either alternative (AP) or classical pathway (CP) activation. In complement activation ELISAs, which tracked C5b-6, a complex formed from the an activated product of C5, K57chemFE was a potent and fully efficacious inhibitor of both the CP and AP ( Figure 2E and 2F); K8chemFE was a partial inhibitor of the CP and AP ( Figure 2A ); while K92chemFE partially inhibited the AP and showed slight, dose dependent enhancement of the CP ( Figure 2C ). Consistent with earlier studies with biologically derived K149 7 , K149chemFE was a nonfunctional, silent binder of C5 (data not shown).

The peptides were counter screened using ELISAs which measured deposition of C3d, an activation product from complement C3, in response to AP and CP activation (shown in S5). A lack of activity in these assays confirmed that the chemical knob domain peptides did not affect activation of either of the human C5 homologs, C3 or C4, which are both upstream of C5 in the complement cascade.

Behaviour in erythrocyte haemolysis assays was consistent with the ELISAs. K57chemFE was a potent and fully efficacious inhibitor of complement mediated cell lysis, K92chemFE was active solely in the AP-driven haemolysis assay ( Figure 2D ) and K8 was a partial inhibitor for the CP and weakly active in the AP assay ( Figure 2B ). Importantly, these observations in the ELISA and haemolysis assays closely mirror those previously reported with the biological forms of the peptides 7 . In haemolysis assays, K57chemFE was broadly equivalent to two previously characterised C5 inhibitors ( Figure 2G and 2H): RA101295 30 , a close analogue of the UCB-Ra Pharma macrocyclic peptide Zilucoplan 31 , which is currently in phase III trials, and SOBI002 32, 33 , an affibody from Swedish Orphan Biovitrum, which was discontinued after showing transient adverse effects in a phase 1 trial 34 . While the three inhibitors display equivalent efficacy, RA101295 exhibited approximately ten-fold higher potency than K57chemFE and SOBI002.

To permit comparison of the molecular structure of a chemically synthesised knob domain, a crystal structure of K92chemFE in complex with C5 was determined at a resolution of 2.57 Å ( Figure 3A , data collection and refinement statistics are shown in Table 2 ). The C5-K92chemFE complex was crystallised under the same conditions previously reported for the C5-biological K92 (K92bio) complex (PDB accession: 7AD6) 7 . A stringent mFo-DFc simulated annealing omit map of the C5-K92chemFE complex, with the peptide deleted from the model, shows clear electron density for the peptide at 1.0 sigma ( Figure 3C) , while the final structure shows the fold and disulphide bond arrangement of K92chemFE to be contiguous to K92bio ( Figure 3B ). Analysis with the macromolecular structure analysis tool PDBPiSA 35 , confirmed that the molecular interactions which sustain binding to C5 are consistent between K92chemFE and K92bio (Shown in S6).

Examination of the K92chemFE binding interface with C5 revealed that electrostatic interactions were comparatively sparse, with the structure suggesting just ten hydrogen bonds were present, which were predominantly mediated via the polypeptide backbone. We noted a bifurcated pocket on the β-chain of C5, adjacent to the residues A12 and I13 of K92chemFE, and attempted to rationally design mutations to enable new interactions to be made with C5 within this region, with the goal of improving the biology efficacy. Residue A12 offered scope to contact the α-chain on C5 via N805C5 (numbering based on the mature C5 sequence), which the crystal structure of the complex suggested did not have strong non-covalent molecular interactions with K92, as well as access a cavity on the C5 β-chain, which is flanked by residues L152C5, Q532C5 and V534C5 ( Figure 4A ). We therefore designed mutants for A12, using Molecular Operating Environment's (MOE, version 2019 36 ) residue scan function (shown in S7 and S8). We predominantly focused on non-natural and D-forms of amino acids, which have been shown to be beneficial in improving the stability, potency, permeability, and bioavailability of peptide-based therapies 37 .

The final eight designs, shown in S9, contained mono and bicyclic side chains, such as D-tryptophan, benzothiazole and 2-oxo-histidine; acetylated and methylated lysine; L-and D-forms of ornithine, as well as the natural amino acid asparagine. The mutants were tested in SPR multi-cycle kinetics experiments and in AP and CP complement ELISAs. While all the mutants were readily synthesised, the assays identified three mutations of note: asparagine (A12N [shown in Figure 4C and 4E]), Dtryptophan (A12d-W [shown in Figure 4B and 4E]) and acetylated lysine (A12K[Ac]), shown in Figure 4D and 4E.

Our SPR data (Table 1 and Figure 5E -H), show that we did not improve binding affinity relative to K92chemFE in any of these mutants. While the A12N mutation did not result in a loss of affinity, both A12d-W and A12K(Ac) suffered 130 and 166-fold reductions in affinity for C5, with KD values of 53.5 nM and 68.1 nM, respectively. However, while all three mutations resulted in an acceleration in koff, for A12N, there was a compensatory improvement in kon ( Table 1) .

As shown earlier, K92chemFE has no effect on the CP but is demonstrably allosteric in the AP ELISA, with an IC50 of 84 nM and an Emax of just 27 %. Consistent with the K92chemFE parent compound, all mutations had no effect in the CP ELISA ( Figure 6F ). In the AP ELSA, the A12N mutation gained approximately a log of potency (IC50 of 4 nM) but did not affect Emax to any great extent ( Figure 6G ).

Similarly, the A12d-W mutant appeared not to have lost potency or efficacy, despite having a markedly lower affinity than the parent ( Figure 6I ). For drug binding studies in vitro, potent compounds frequently have faster kon, which speeds their onset of action 38 , subsequently an increase in kon can compensate for a reduction in binding affinity in a short assay, which may explain the gain in potency for the A12N mutant. Conversely, to prolong drug action in vivo, the endurance of a drugtarget complex is usually more contingent on koff 38, 39 .

Interestingly, the A12K(Ac) mutant also displayed a slight increase in potency but exhibited a significant increase in Emax, with a two-fold increase in efficacy observed ( Figure 6H ). For allosteric compounds, biological efficacy is due to an effect on the protein target's conformation that is uncorrelated with binding affinity 40 , but a conformationally constrained, or rigidified, compound may be beneficial, particularly for small molecules 41 . While we were unsuccessful in increasing the affinity of the complex, we show that the rational incorporation of non-proteogenic amino acids into knob domains can be a route to modulate biological efficacy and our data suggest that D-amino acids may also be tolerated, which may be beneficial when exploring routes to attenuate proteolysis.

For development of peptide drug candidates, head-to-tail cyclisation can extend exposure in vivo by preventing exopeptidase cleavage 42 . To create small, cyclic antibody fragments, we synthesised headto-tail cyclised forms of our knob domains, which are subsequently termed: K8chemFE cyclic , K57chemFE cyclic , K92chemFE cyclic and K149chemFE cyclic .

All cyclic knob domains bound C5 (Table 1 and Figure 5A -D). For K57chemFE cyclic and K149chemFE cyclic affinity was improved, but binding was attenuated in K8chemFE cyclic and K92chemFE cyclic ( Figure 5H ). Biological activity was again tested in AP and CP ELISAs. The activity of K57chemFE cyclic and K92chemFE cyclic was not obviously affected by cyclisation, while K8chemFE cyclic displayed a modest loss of potency relative to K8chemFE ( Figure 6A -E). Consistent with the parent compound, K149chemFE cyclic had no discernible effect on complement activation, despite binding C5 (data not shown). Overall, these data suggest that knob domains are amenable to cyclisation and, while some optimisation may be required in certain cases, this approach can be explored to improve stability.

Finally, we sought to explore the pharmacokinetic profile of chemically synthesised knob domain peptides in rodents. As target binding may influence pharmacokinetics, we first tested for binding to C5 protein from Rattus norvegicus (rat). By SPR, K8chemFE was cross reactive with rat C5 protein (as measured by single-cycle kinetics, shown in S10), as well as C5 from other species (data not shown), while K57chemFE was specific for human C5. To explore the effect of fatty acid conjugation, we synthesised a palmitoylated form of K57chemFE, with attachment via a Gly-Ser-Ser-Gly linker at the N-terminus, and determined an apparent four-fold reduction in binding for affinity for human C5 by SPR (Table 1 and S11).

To test if the knob domains were resistant to proteolysis by virtue of their abundant disulphide bonds, we employed HPLC mass spectrometry to track the stability of K8chemFE, K57chemFE and K57chemFEpalmitoyl in human, rat and Mus musculus (mouse) plasma at 37 o C, over a period of 24 hours. The unmodified knob domains were predominantly stable in human plasma, with K8chemFE not degraded in plasma from humans, mice or rats ( Figure 7C ). For K57chemFE, in excess of 75 % remained intact after 24 hours in human plasma, although markedly more proteolysis was observed in rodent plasma ( Figure 7A ). The addition of a fatty acid to K57chemFE-palmitoyl conferred protection from proteolysis in rodent plasma ( Figure 7B ), by virtue of binding to serum albumin.

The pharmacokinetics of K8chemFE, K57chemFE and K57chemFE-palmitoyl were measured following dosing via an intravenous bolus to Sprague Dawley rats ( Figure 7D ). Following administration at 10 mg/kg, K57chemFE was eliminated in a rapid manner typical of low molecular weight peptides and proteins (t1/2 = 17 mins/ plasma clearance [Clp] = 10.6 ml/min/kg), which is typically due to glomerular filtration in the kidney. In contrast, K8chemFE tightly bound rat C5 protein and adopted target-like pharmacokinetics, resulting in markedly improved exposure (t1/2 = ~9 hours/ Clp = 3.3 ml/min/kg). Due to poor solubility, K57chemFE-palmitoyl was tested at a lower dose of 1 mg/kg; as expected, the conjugation of a palmitic fatty acid extended exposure, relative to unmodified K57chemFE (t1/2 = 1.6 hours/ Clp = 0.8 ml/min/kg). This suggests that conjugation of fatty acids, potentially in combination with cyclisation or other chemical modifications, is a viable route to extend the biological exposure of chemical knob domains in vivo.

Our data suggest that, despite their biological origin, knob domain peptides are viable chemical entities and that this greatly aids their potential developability as drug candidates. In chemical form they present with high potency and are inherently stable in plasma. Their pharmacokinetics are akin to conventional peptides -moreover, unusually for antibody derived molecules, chemical modification offers a tractable route to improve their pharmaceutical properties.

We have shown, using various examples, formats and modifications, that knob domains are amenable to chemical synthesis. Crucially, the chemically derived peptides bind antigen with high affinity and are active in biological assays, with K57chemFE offering a comparative in vitro profile to SOBI002 and a peptide with a similar core sequence to zilucoplan, both of which are clinical stage C5 inhibitors.

We present chemical knob domains as the first examples of antibody fragments which are readily amenable to medicinal chemistry approaches. For drug discovery, medicinal chemistry typically focuses on the optimisation of drug potency, pharmacokinetics and biodistribution. Here we exemplified chemical strategies that may be applied to improve the stability and biology efficacy of knob domain peptides.

To modulate biological efficacy, this study used structure-based drug design with non-natural amino acids to access latent interactions which were not employed in the natural paratope. Non-natural amino acids are commonly used to increase the binding affinity of peptides. Compstatin, a macrocyclic inhibitor of complement C3 which, in a modified form, was recently approved for paroxysmal nocturnal haemoglobinuria 43 , is one such example; where incorporation of a hydrophobic 1-methyltryptophan results in a 264-fold increase in activity 44 . Similarly, non-proteinogenic residues have been used in the generation of high affinity human leukocyte antigen blockers, which competitively displace the antigenic peptide on major histocompatibility complex receptors to prevent T-cell recognition of the complex 45 . To improve stability, adjacent D-amino acids are more resistant to degradation by natural proteases than their L-enantiomeric counterparts, and this represents a strategy to employ against proteolysis 37 , through stabilization of backbone conformation and elimination of the cleavage site 46 . For knob domain peptides, this approach is complementary to the optimisation of the natural amino acid sequence, which occurs in vivo within the germinal centres of the secondary lymphoid organs when a cow is exposed to an antigen. The rational application of nonnatural amino acids, in this post-immune setting, can bring chemical space and physiochemical properties, that are not at the disposal of the bovine immune system, into play.

There are several methods for incorporation of non-natural amino acids into antibody fragments or other small proteins, as part of attempts to either improve affinity, conjugate toxins for antibody drug conjugates 47 or to introduce chemical handles for photocoupling 48 Our data suggest that the knob domains are highly resistant to plasma proteolysis in vitro, potentially due to their network of disulphide bonds. We have shown that the proximity of the N-and C-termini can give rise to fully cyclic antibody fragments, which may provide a route to reduce proteolysis even further, especially when used in conjunction with other chemical approaches.

When dosed in vivo, knob domains exhibit typically peptide-like pharmacokinetics and in unmodified form are likely subject to renal clearance, unless target-mediated drug disposition is achieved through tight binding to a target. We have shown that by using chemical synthesis, simple modifications such as palmitoylation can be readily incorporated to attenuate renal clearance and produce lead-like molecules, which are consistent with once daily dosing.

This chemical method may also provide a route to decouple small antibody fragments from the cost constraints associated with manufacturing in cell systems. The cost of antibody therapies has come under scrutiny in the face of global COVID-19 pandemic and alternatives are being sought to address the so called 'access gap', whereby 80 % of the global licenced antibody sales are within the United States, Europe and Canada 52 . We hope that the abundance of disulphide bonds and apparent stability of the knob domains may remove the requirement for a cold supply chain, in a similar manner to other peptide drugs.

We present a chemical biology approach for bovine antibody-derived knob domain peptides, demonstrating structural and functional equivalence from biologically expressed and chemically Purification and cyclisation of knob domains. It was found that the most expedient and high yielding way to obtain these cyclic peptides with greater than two disulphide bonds was to reversed phase high performance liquid chromatography (RP-HPLC) purify the linear sequence and immediately initiate cyclisation without a freeze-drying step, otherwise much insoluble polymeric material resulted.

Cyclisation was achieved by using thermodynamic-controlled air oxidation to obtain the minimum energy form of the disulphides in the sequence, employing a mixture of reduced and oxidised glutathione.

Crude peptides were dissolved in dimethyl sulfoxide (DMSO)/ water and treated with tris (2carboxyethyl) phosphine (TCEP) to ensure the disulphide bonds were fully reduced. The peptides were purified by RP-HPLC, using a Varian Prostar system equipped with two 210 pumps and a 355 UV spectrophotometer. Running buffers were for pump A, solvent A, 0.1% (v/v ammonium acetate in water, pH 7.5 -7.8), and for pump B, solvent B, 100% acetonitrile. The peptide was introduced to a prep RP-HPLC column (C18 Axia, 22 mm x 250 mm, 5 micron particle, size 110 angstrom pore size, Phenomenex). The linear sequence was eluted from the column by running a gradient between solvents A and B, 5% B to 65% B over 60 minutes. Linear peptide was identified by electrospray ionisation mass spectrometry.

The solution of the linear peptide (approximately 50 mL) was added to 500 mL of a cyclisation buffer, (0.2 M phosphate buffer, pH 7.5, containing 1mM EDTA, 5mM reduced glutathione, and 0.5 mM oxidised glutathione). The solution was stirred at room temperature for 48 hours. After which, a small sample was analysed by analytical HPLC to assess the level of cyclisation.

When the cyclisation was deemed sufficiently complete, the whole buffer containing the peptide was pumped onto a preparative RP HPLC column (C-18 Axia as above). The cyclic peptide was eluted using a gradient between solvent A (0.1 % TFA in water) and solvent B (0.1% TFA in acetonitrile) of 5% B to 65% B in 60 minutes. Fractions identified as the correct compound were freeze dried before analysis.

For head-to-tail cyclisation, we used Gly-Cys as the point of cyclisation and used a thioester strategy employing native chemical ligation. Generation of a thioester upon the c-terminal Gly residue does not result in racemisation; while cyclisation using the purified side chain deprotected thioester improves solubility, giving cleaner products and better yields.

N-terminal tert-butyloxycarbonyl (boc) protected peptides were made by Fmoc peptide synthesis using glycine loaded 2-chlorotrityl resin, with cleaving from the resin via 1% TFA treatment. The peptide benzyl-thioesters were generated through reaction with benzyl-mercaptan and 1benzotriazole-tris-dimethyl. The resulting protected peptidyl-thioester species underwent a complete side chain cleavage, using the normal TFA scavenger cleavage regime. Finally, the peptides were HPLC purified, and then cyclised head-to-tail via native chemical ligation, (by dissolving the peptides in phosphate buffer, pH 7.8, with a 3-fold molar excess of TCEP present, overnight). The head-to-tail cyclised peptides were freeze dried, and intra-chain disulphide bond formation was again achieved using thermodynamic-controlled air oxidation with a mixture of reduced and oxidised glutathione, to obtain the minimum energy form of the disulphides in the sequence, as described previously. to the bound complex structure, were computed. We noted that MOE frequently inserted rotamers that did not in fact fill the desired pocket but pointed out into solvent. To avoid this, we decided to place virtual atoms (a simple carbohydrate chain) on the outside of the pocket, such that they occupied the same space as the undesired rotamers, thereby producing a steric clash and forcing MOE to choose only side chain orientations that would fill the desired pocket.

This resulted in a ranked list of the 70 candidate amino acids. We selected only those amino acids that were predicted by MOE to improve the complex's overall stability and the knob domain's affinity for its target (dAffinity < 0; dStability < 0). This left 26 candidate amino acids, including 8 natural and 18 non-natural amino acids (shown in S8). Rather than further relying on MOE's energy-based ranking, we took all 18 non-natural amino acids into consideration.

We then manually eliminated a number of candidate side chains due to potentially unstable chemistry as well as difficulties in obtaining the required building blocks to easily synthesize them. Due to their availability, we also chose to synthesise L-and D-ornithine mutants, which were not selected through our MOE analysis. The side chains chosen for synthesis are marked in S9.

Plasma stability. Rat, mouse and human plasma were collected using lithium heparin as an anticoagulant. Stability was assessed at a concentration level of 1.25 µg/mL for K57chemFE-Palmitoyl, 6.25 µg/mL for K57chemFE and 3.75 µg/mL for K8 chemFE, over a 24 hour period at room temperature. The UPLC column was equilibrated with the initial mobile phase conditions at a flow rate of 0.4 mL/min and the analytes were separated using a water, acetonitrile and formic acid gradient described below. Solvent A was 100 % water 0.1 % formic acid and solvent B 100 % acetonitrile. 

Use of a single VH family and long CDR3s in the variable region of cattle Ig heavy chains

Structural Diversity of Ultralong CDRH3s in Seven Bovine Antibody Heavy Chains

Conservation and diversity in the ultralong third heavy-chain complementarity-determining region of bovine antibodies

Reshaping antibody diversity

The Unusual Genetics and Biochemistry of Bovine Immunoglobulins

Isolation of antigen-specific, disulphide-rich knob domain peptides from bovine antibodies

The allosteric modulation of complement C5 by knob domain peptides

A new antigen receptor gene family that undergoes rearrangement and extensive somatic diversification in sharks

Naturally occurring antibodies devoid of light chains

Sequence and structure of VH domain from naturally occurring camel heavy chain immunoglobulins lacking light chains

VHH characterization. Comparison of recombinant with chemically synthesized anti-HER2 VHH

Crystal structure at 2.8 A of an FcRn/heterodimeric Fc complex: mechanism of pH-dependent binding

Mechanisms Influencing the Pharmacokinetics and Disposition of Monoclonal Antibodies and Peptides

PEG-peptide conjugates

Preparation, purification, and characterization of a reversibly lipidized desmopressin with potentiated anti-diuretic activity

Seamless proteins tie up their loose ends

Multifaceted roles of disulfide bonds. Peptides as therapeutics

Peptides: New processes, lower costs

Eculizumab, a terminal complement inhibitor, improves anaemia in patients with paroxysmal nocturnal haemoglobinuria

Ravulizumab (ALXN1210) in patients with paroxysmal nocturnal hemoglobinuria: results of 2 phase 1b/2 studies

Antibody engineering to generate SKY59, a long-acting anti-C5 recycling antibody

Development of RA101348, a Potent Cyclic Peptide Inhibitor of C5 for Complement-Mediated Diseases

Derivation of RNA aptamer inhibitors of human complement C5

A small-molecule inhibitor of C5 complement protein

Clinical promise of next-generation complement therapeutics

Crystal structure of defensin HNP-3, an amphiphilic dimer: mechanisms of membrane permeabilization

Structural venomics reveals evolution of a complex venom by duplication and diversification of an ancient peptide-encoding gene

Building upon Nature's Framework: Overview of Key Strategies Toward Increasing Drug-Like Properties of Natural Product Cyclopeptides and Macrocycles

Editorial: From Peptide and Protein Toxins to Ion Channel Structure/Function and Drug Design

Inhibition of complement C5 protects against organ failure and reduces mortality in a baboon model of Escherichia coli sepsis

Preclinical Evaluation of RA101495, a Potent Cyclic Peptide Inhibitor of C5 for the Treatment of Paroxysmal Nocturnal Hemoglobinuria

Combined Inhibition of C5 and CD14 Attenuates Systemic Inflammation in a Piglet Model of Meconium Aspiration Syndrome

Development of Affibody® C5 inhibitors for versatile and efficient therapeutic targeting of the terminal complement pathway: 114

Safety and Tolerability of SOBI002 in Healthy Volunteers Following Single and Repeated Administration

Inference of macromolecular assemblies from crystalline state

Molecular Operating Environment (MOE)

Impact of non-proteinogenic amino acids in the discovery and development of peptide therapeutics

Binding kinetics and mechanism of action: toward the discovery and development of better and best in class drugs

The drug-target residence time model: a 10-year retrospective

Modulating Target Protein Biology Through the Re-mapping of Conformational Distributions Using Small Molecules

Importance of Rigidity in Designing Small Molecule Drugs To Tackle Protein-Protein Interactions (PPIs) through Stabilization of Desired Conformers

Reversible lipidization for the oral delivery of salmon calcitonin

Pegcetacoplan versus Eculizumab in Paroxysmal Nocturnal Hemoglobinuria

Hydrophobic effect and hydrogen bonds account for the improved activity of a complement inhibitor, compstatin

Design of azidoproline containing gluten peptides to suppress CD4+ T-cell responses associated with celiac disease

Chemical modifications designed to improve peptide stability: incorporation of non-natural amino acids, pseudo-peptide bonds, and cyclization

Antibody conjugates with unnatural amino acids

Site-specific photocoupling of pBpa mutated scFv antibodies for use in affinity proteomics

Expanding the genetic code of Escherichia coli

Systematic Activity Maturation of a Single-Domain Antibody with Non-canonical Amino Acids through Chemical Mutagenesis

Synthesis of proteins by automated flow chemistry

Expanding access to monoclonal antibody-based products: A global call to action

The rational design of affinity-attenuated OmCI for the purification of complement C5

Solid Phase Peptide Synthesis: A Practical Approach

The PyMOL Molecular Graphics System, Version

Selby 1 , Maisem Laabei 2 , Toska Wonfor 2 , Adam Hold 1 , Douangsone Vadysirisack 4

United Kingdom *email alex.macpherson@ucb.com S2. Knob domain QC: HPLC chromatograms and MS spectra

Ac) (Molecular mass

Full list of symbols for non-natural amino acids scanned in MOE