PEDS Advance Access published online on April 23, 2007
Protein Engineering Design and Selection, doi:10.1093/protein/gzm015
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Alternative antibody Fab' fragment PEGylation strategies: combination of strong reducing agents, disruption of the interchain disulphide bond and disulphide engineering
1 UCB-Celltech, 216 Bath Road, Slough, SL1 4EN, UK
2 To whom correspondence should be addressed. E-mail: david.humphreys{at}celltech.ucb-group.com
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
Antigen-binding fragments (Fab') of antibodies can be site specifically PEGylated at thiols using cysteine reactive PEG–maleimide conjugates. For therapeutic Fab'–PEG, conjugation with 40 kDa of PEG at a single hinge cysteine has been found to confer appropriate pharmacokinetic properties to enable infrequent dosing. Previous methods have activated the hinge cysteine using mildly reducing conditions in order to retain an intact interchain disulphide. We demonstrate that the final Fab–PEG product does not need to retain the interchain disulphide and also therefore that strongly reducing conditions can be used. This alternative approach results in PEGylation efficiencies of 88 and 94% for human and murine Fab, respectively. It also enables accurate and efficient site-specific multi-PEGylation. The use of the non-thiol reductant tris(2-carboxyethyl) phosphine combined with protein engineering enables us to demonstrate the mono-, di- and tri-PEGylation of Fab fragments with a range of PEG size. We present evidence that PEGylated and unPEGylated Fab' molecules that lack an interchain disulphide bond retain very high levels of chemical and thermal stability and normal performance in PK and efficacy models.
Keywords: disulphide engineering/Fab'/PEGylation/periplasm
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
PEGylation has found favour due to the ability of PEG to beneficially alter the solubility, immunogenicity, pharmacokinetics, aggregation and proteolytic susceptibility of therapeutic proteins (see Suzuki et al., 1984
; Cunningham-Rundles, 1992; Caliceti and Veronese, 2003; Harris and Chess, 2003). PEGylation of proteins can be achieved with a number of different conjugation methods including random chemical attachment at lysines, site-specific chemical attachment at sugars (Rodwell et al., 1986) and cysteines (Chapman et al., 1999) and site-specific enzymatic attachment at glutamines (Sato, 2002). Random attachment can result both in loss of functionality and in product heterogeneity. This loss of functionality can be especially acute with antibodies due to interference with, or steric occlusion of the antigen-binding site (Kitamura et al., 1991). Product heterogeneity can complicate downstream purification or specification issues.
Antigen-binding fragments (Fab') of antibodies are attractive for use as therapeutics due to their monovalent antigen binding, lack of innate effector functions and ability to be produced in a range of large-scale expression hosts such as Escherichia coli. Fab' fragments have a very short circulating half-life in mammals since they lack the receptor-mediated recycling encoded on the Fc fragment of antibodies and are subject to renal filtration (Covell, et al., 1986; Buchegger et al., 1992). Conjugation of various sizes of PEG to Fab' has been shown to increase the circulating half-life of PEG in a controllable manner (see Chapman 2002). Furthermore, addition of 40 kDa of PEG is required to confer a serum half-life approaching that of an IgG to facilitate infrequent therapeutic dosing regimes (Chapman et al., 1999; Leong et al., 2001; Knight et al., 2004; Wei et al., 2005).
Site-specific PEGylation at solvent accessible cysteine residues is attractive since maleimide-based reaction chemistries can be used. Such chemistries are simple and efficient, and are performed under mild conditions that do not disrupt the protein. Typically, a single cysteine is activated through mild reduction with a thiol-based compound such as ß-mercaptoethylamine (ß-MA), ß-mercaptoethanol (ß-ME), glutathione or cysteine before modification with a single 40-kDa PEG–maleimide. Reductive activation is required prior to PEGylation because the target cysteine can be capped with small thiol reactive molecules during fermentation/purification (Carter et al., 1992). The use of thiol-based reductants suffers to some extent from formation of thiol adducts between the reductant and the thiol being activated (Begg and Speicher, 1999). Formation of these adducts can reduce the efficiency of the PEGylation reaction. Non-thiol reducing agents such as tris(butyl) phosphine (TBP) and tris(2-carboxyethyl) phosphine (TCEP) cannot form thiol adducts. TCEP is water soluble, odourless, is active over a wide pH range and has been used for the activation of scFv thiols (Han and Han, 1994; Albrecht et al., 2004; Cline et al., 2004). However, since TBP and TCEP are strong reducing agents, they are not suitable for the activation of a single hinge thiol, whilst retaining an intact interchain disulphide bond.
Protein engineering can be used to remove the interchain disulphide bond by mutation of the cKappa and CH1 cysteines to serine. The short (
2 h) serum permanence of thioether linked F(ab')2 engineered in this way was not affected by the loss of the interchain disulphide (Rodrigues et al., 1993). Fab' with non-covalently associated LC and HC has also been shown to have unchanged in vitro stability during purification processes (Humphreys et al., 1997). These observations suggested that Fab' lacking interchain disulphide bonds might also have sufficient stability in vivo to survive the long periods conferred by PEGylation with 40 kDa PEG. Hence, engineered removal of the interchain cysteines enables new approaches for PEGylation to be considered.
In this work, we made mono-PEGylated Fab' that was engineered to lack its interchain disulphide bond. We show that serum half-life was essentially unchanged relative to that of a Fab–PEG with an interchain disulphide bond. Thereafter, we investigated the ability to produce mono-, di- and triPEGylated Fab' with efficiencies up to 88–94% after reduction with TCEP. Multi-PEGylated Fab' with non-covalently associated LC and HC is shown to have unchanged affinity in vitro and efficacy in vivo. We also present data that compare the chemical stability of Fab' that retains or lacks interchain disulphide bonds. The combined use of non-thiol reducing agents and protein engineering enables the use of alternative, highly efficient PEGylation strategies for antibody Fab' fragments.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
Fab' engineering, expression and purification
The codons for the interchain disulphide bond cysteines in human cKappa (Kabat residue 214) or human CH1 (Kabat residue 233) constant regions were changed to serine using oligonucleotide-encoded polymerase chain reaction (PCR) mutagenesis. Fab' was expressed in the periplasm of E. coli by fusion of Fab' coding regions to OmpA signal peptides and induced by the addition of IPTG to cultures. Fermentation, extraction of periplasmic material and purification were performed essentially as previously reported with minor modifications (Humphreys et al., 1997, 1998). Fab–PEG was purified using SP Sepharose HP (GE-Healthcare). Material was loaded at pH 4.5 in a low conductivity sodium acetate buffer, such that PEG does not bind. Fab–PEG and unreacted Fab' were eluted separately using a shallow NaCl gradient. Purified Fab–PEG used in the following experimentation is quantified to account for both the Fab and the PEG components and not just the Fab' component.
The notation used throughout uses single letter amino acid code to illustrate whether a constant domain or hinge residue is a cysteine (C) or a serine (S). For example, ‘Fab-X LC-C, HC-S, hinge-CAA’ is a Fab with v-regions of antibody; ‘X’ that retains the cKappa and hinge cysteines, but has a CH1 cysteine to serine mutation. To make a Fab construct (i.e. one with no upper hinge/hinge region), a stop codon was introduced immediately after the CH1 interchain cysteine. Figure 1 shows examples of the various Fab and Fab' formats used in this study.
|
Three humanised model Fab's were used in these studies. Fab-A binds to the extracellular domain of a human cytoplasmic membrane protein and acts here as an example of a ‘typical’ (high affinity, stable, well expressed, standard purification) humanised Fab. Fab-B neutralises a soluble human cytokine and was used to enable simple in vivo efficacy studies to be performed. Fab-C neutralises a second human cytokine and was used as an additional exemplification of the di-PEGylated ‘no-hinge’ Fab construct.
Reduction and PEGylation of Fab'
Reductions and PEGylations were performed in 0.1 M phosphate, pH 6.0, with 2 mM EDTA. For comparisons of various reductants (Fig. 2), 5 mM of each reductant and Fab at 10 mg/ml was used; we thereby tested a fixed molar ratio of reductant to Fab' throughout. ‘No-hinge’ Fab was di-PEGylated with 20 kDa linear-PEG using TCEP as reductant under the same conditions. In all cases, reduction was done at room temperature (24°C) for 30 min, before desalting on a PD-10 column (GE-Healthcare) followed by mixing with a 5-fold molar excess of PEG–maleimide over Fab'. The 40-kDa PEG (branched 2 x 20 kDa) was from Nektar, while the linear 20- and 30-kDa PEGs were from NOF.
|
Analysis of extent of Fab' PEGylation
Approximately 50 µg samples were analysed by size-exclusion high-performance liquid chromatography (HPLC) on analytical Zorbax GF-450 and GF-250 columns in series (Agilent). These were developed with a 30-min isocratic gradient of 0.2 M phosphate, pH 7.0, with 10% ethanol at 1 ml/min. Detection is by absorbance at 214 and 280 nm. PEGylation efficiencies were calculated by the comparison of the relative area under the peak of each molecular species.
Pharmacokinetics of Fab–PEG in rats
An amount of 300 µg of Fab'–PEG per animal group was 125I-labelled using Bolton and Hunter reagent (GE-Healthcare) to a specific activity of 0.22–0.33 µCi/µg.
Male Sprague–Dawley rats of 220–250 g (Harlan) were injected intravenously (i.v.) or subcutaneously (s.c.) with 13 µg of 125I-labelled Fab'–PEG variants while under halothane anaesthesia (n = 6 per group). Serial arterial bleeds from the tail were taken at 0.5, 2, 4, 6, 24, 48, 72 and 144 h post-administration for i.v. doses or 3, 6, 24, 30, 48, 72 and 144 h post-administration for s.c. doses. Samples were counted using a COBRA Autogamma counter (Canberra Packard). Data were plotted and area under the curve (AUC) were calculated using GraphPad Prism (GraphPad Software Inc.) and is expressed as percent injected dose hour (% i.d h). The t1/2
was defined by time points 0.5 h, 2 h, 4 h, and 6 h, while the t1/2ß was defined by time points 24 h, 48 h, 72 h and 144 h for i.v. doses. The means and standard errors of means (SEM) of data for all six animals are shown in Table I.
|
Mouse antigen-binding efficacy model: i.v. dosed anti-cytokine mAb and intraperitoneally (i.p.) dosed human cytokine
Male Balb/c mice (21 g) were injected i.v. with a single dose [3 mg/kg in 100 µl phosphate-buffered saline (PBS)] of Fab'-B LC-C HC-C hinge-CAA-40 kDa PEG, Fab'-B LC-S HC-S hinge-CAA-40 kDa PEG, or Fab'-X LC-C HC-C hinge-CAA-40 kDa PEG (irrelevant control), 1, 3 or 7 days prior to an i.p. injection of human cytokine (3 ng/kg in 100 µl PBS vehicle). After 120 min, mice were killed by cervical dislocation, and peritoneal lavage was performed [3 ml PBS with 0.25% bovine serum albumin (BSA), 12 mM HEPES]. A total leucocyte count was performed using a Coulter Counter. For identification of neutrophils, 50 µl peritoneal lavage fluid was stained with 1 : 300 dilution of anti-CD45–CyChrome mAb and 1 : 300 dilution of anti-GR-1–PE mAb (anti-Ly6G/Ly6C) for 20 min (4°C, in the dark). Leucocytes were washed once in wash buffer (PBS, 0.25% BSA, 12 mM HEPES), resuspended in 300 µl wash buffer and analysed by flow cytometry. Neutrophils were identified as CD45+GR-1HIGH. The 3-mg/kg dose was the minimum dose found to have the maximum efficacy in an in vivo titration experiment in a related animal model with a 24-h read-out (data not shown).
Thermal stability. Periplasmic Fab' is extracted from cell paste and partially clarified away from host proteins by overnight agitation in 100 mM Tris, pH 7.4/10 mM EDTA at 60°C. The ability of various constructs to survive this step, relative to a control overnight 30°C extraction in the same buffer was used to demonstrate the presence of a stabilising interchain disulphide bond.
Guanidine denaturation. Proteins were used at a concentration of 65 µg/ml in 20 mM Tris–HCl (pH 7.4) with 100 mM NaCl. Different GuHCl concentrations were added in 0.2 M increments ranging from 0 to 6 M. Samples were incubated overnight at 25°C to achieve equilibrium. The fluorescence excitation wavelength was set to 295 nm. The emission spectra were recorded from 310 to 375 nm, with the slit width set to 5 nm from a Cary Eclipse fluorimeter, (Varian).
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
Construction of new Fab and Fab' disulphide structures
The three cysteine residues that compose the interchain disulphide bond and the single hinge cysteine were individually changed to serine using PCR mutagenesis. These single mutations were also combined to create all possible double and triple mutations. Furthermore, by the placement of a stop codon immediately after CH1 Cys233, a Fab coding region was created. The notation used to describe these disulphide engineered Fab is described in the Methods section, while Figure 1 illustrates the various Fab' and Fab formats used in this study. No significant differences of expression level were observed for any of these constructs for either Fab-A or -B, with the normal range from E. coli fermentation being 900–1200 mg/l.
Comparison of reductants used for activation of ‘classic’ Fab' and ‘
-inter’ Fab'
Five reducing agents were compared for their ability to perform a mild activation of the hinge cysteine found in LC-C HC-C, hinge-CAA Fab': TCEP, dithiothreitol (DTT), ß-MA, ß-ME and glutathione (reduced) (GSH). There are a total of three solvent accessible cysteines in this Fab' format that can theoretically be activated by reducing agents. Therefore, the strength of the reducing agent defines whether one, two or three PEG molecules are added to the Fab'. HPLC analysis cannot accurately discern between the addition of two or three PEG molecules when they are found in approximately equal abundance in a reaction mixture, hence these are shown as combined ‘multi-PEGylation’ in the results. Figure 2 shows that TCEP is a stronger reductant than all of the thiol reductants, because it is incapable of activating the hinge thiol without breaking a large proportion of the interchain disulphide bonds. Approximately 95% of the ‘classic’ Fab' is PEGylated (combination of mono- and multi-PEGylated), but approximately two-third of the PEGylated material is actually multi-PEGylated. ß-MA was shown to be the best reductant for achieving mono-PEGylation of Fab'. Under these un-optimised conditions, 55% of the ‘classic’ Fab' is mono-PEGylated while <5% of the protein is multi-PEGylated. DTT lies between these two extremes, because it achieves slightly more mono-PEGylation than ß-MA, but at the cost of an increase in the percentage of protein that is also multi-PEGylated. ß-ME and GSH are both shown to be rather weak reducing agents since only 25% of the material is mono-PEGylated and there is an almost undetectable level of breakage of the interchain disulphide as evidenced by the lack of multi-PEGylated product. These data suggested that the use of stronger reducing agents was not an effective method for increasing mono-PEGylation efficiency, but rather suggested that reagents such as TCEP could be useful in alternative PEGylation strategies where the final PEGylated product lacks an interchain disulphide bond.
Fab' function is unaffected by the loss of interchain disulphide bond
To investigate the effect of the loss of the interchain disulphide bond on Fab' function, ‘classic’ and ‘-inter’ versions of Fab'-B were modified with a single, 40-kDa, branched PEG on the hinge and then tested for circulating half-life, affinity and in vivo function.
Figure 3 shows that the pharmacokinetic profile of the ‘classic’ and ‘-inter’ Fab–PEG proteins is identical within experimental error. When the data are analysed for t1/2, t1/2ß phases and AUC (Table I), we find that the details of the pharmacokinetics are also unchanged. The ‘-inter’ Fab–PEG form of Fab-B was found to be 0.13 nM, whereas the ‘classic’ Fab-PEG of Fab-B was identical within experimental error at 0.10 nM. Both of these compare well with unPEGylated ‘classic’ Fab'-B which has an affinity of 0.12 nM. Therefore, neither absence of the interchain disulphide bond of Fab-B or its hinge mono-PEGylation affects antigen binding in vitro. Similar results were observed with Fab-A. The affinity of unPEGylated ‘classic’ Fab-A was 1.3 nM whereas the unPEGylated ‘-inter’ version was 1.5 nM, confirming that the absence of the disulphide bond was not critical for retaining antigen binding. Mono-PEGylated ‘classic’ Fab-A (branched 40-kDa PEG) had an affinity of 1.8 nM while di-PEGylation with 2 x 20 kDa or 2 x 30 kDa linear PEG (LC-C HC-S, hinge-CAA) resulted in affinities of 1.4 and 1.6 nM, respectively. These affinities are identical within experimental error showing that neither absence of the interchain disulphide or attachment of PEG alters the ability to bind antigen.
|
) and ‘
) forms of Fab-B monoPEGylated at the hinge with 1 x 40 kDa, branched PEG administered i.v. (filled symbols) or s.c. (open symbols).The cytokine antigen for Fab-B attracts neutrophils to the (peritoneal) site of its administration in an in vivo mouse model. Hence, neutrophils can be counted from peritoneal lavage fluid as a read-out of the biologically active concentration of the antigen. This cellular migration can be blocked by the presence of a neutralising Fab'. Figure 4 shows the results of i.v. mono-PEGylated ‘classic’ and ‘
-inter’ Fab'-B administered 1, 3 or 7 days before the i.p. administration of the antigen. For effective antigen neutralisation to take place, the Fab–PEG needed to have sufficient dose, circulating half-life, antigen affinity and access to the peritoneum, all of which depended upon the structural stability of the Fab' arm. The data in Fig. 4 show that even with pre-administration of as little as 3 mg/kg Fab-PEG, 1 week before administration of antigen both the ‘classic’ and ‘-inter’ forms are similarly effective in blocking neutrophil migration. Collectively, these results show that Fab–PEG lacking an interchain disulphide bond has normal in vivo stability and functionality.
|
The use of Fab–PEG formats lacking an interchain disulphide bond facilitates multi-PEGylation
Total PEG mass approaching 40 kDa has been observed as representing a nominal threshold for achieving the kind of long PK that enables infrequent therapeutic dosing regimes in humans. Avoidance of steric clashes with the CDR's also leads to a preference for the attachment of a single, branched, 40 kDa PEG at the hinge. Multi-PEGylation of Fab' is attractive since it offers a greater range of flexibility for choice of PEG size and point of attachment than mono-PEGylation. However, multi-PEGylation of Fab retaining the interchain disulphide is technically challenging. Weaker reductants that enable retention of the interchain disulphide are less efficient at activating the hinge cysteines and so result in low overall efficiencies (see ß-ME and GSH, Fig. 2). Conversely, the use of stronger reductants increases the risk of breaking the interchain disulphide bond and inappropriate PEGylation of its cysteines. These lead to a product heterogeneity issue that can be difficult to resolve with standard chromatography (Knight, et al., 2004).
Protein engineering can be used to define the number and location of target cysteines, eliminating any theoretical risk of post-purification product heterogeneity. Protein engineering combined with the use of strong reductants such as TCEP can result in very high levels of reduction and, therefore, also high levels of PEGylation. Disulphide engineering of Fab' constant regions and hinges therefore enables discrete and accurate PEGylation at 1, 2 or 3 cysteines, but only where the final product lacks an interchain disulphide bond.
We have performed multi-PEGylation of Fab'-A with both wild-type and disulphide engineered constant regions. TCEP reduction of a ‘classic’ Fab' exposes three solvent accessible cysteines that can be reacted with PEG–maleimide. Tri-PEGylation with linear, 20-kDa PEG–maleimide was performed using un-optimised conditions with an estimated efficiency of >50%. The tri-PEGylated form was purified, and its serum permanence was tested in a rat model. Reduction of the LC-C HC-S, hinge-CAA construct with TCEP exposes two cysteines for modification with PEG. PEGylation efficiency for this construct was generally 85–90%, illustrating that both of these cysteines were available for reaction with PEG–maleimide. PEGylation efficiency of this or other constructs was not affected by the range of PEG sizes tested in this work (e.g. linear 20 and 30 kDa, branched 40 kDa). Table I shows the serum permanence for Fab-A LC-C HC-S, hinge-CAA after PEGylation with 2 x 20 kDa linear PEG and 2 x 30 kDa linear PEG. This confirms that the serum half-life of PEGylated Fab' is not affected by the absence of the interchain disulphide, and that the addition of 60-kDa PEG confers a longer serum half-life than branched 40-kDa PEG as measured by AUC analysis.
Physical stability of Fab' lacking interchain disulphide bonds
Our Fab' fragments are synthesised in and extracted/purified from the periplasm of E. coli. To effect a simple reduction in the number of host proteins and proteolytic fragments of Fab' prior to chromatography, we perform periplasmic extraction at elevated temperatures. It was therefore highly desirable that our engineered Fab' fragments should be stable enough to survive such incubations. Figure 5 shows the percent recovery of various Fab' formats after an overnight incubation at 60°C. It is clear that Fab' that are engineered to be incapable of forming an interchain disulphide bond (see bars 2, 3 and 6) are insufficiently stable under such conditions. Those that retain a native interchain disulphide (bars 1 and 5) survive the overnight 60°C incubation. Of note is the LC-C HC-S hinge-CAA construct (bar 4). This has a disruption to its natural disulphide pair, but is clearly still capable of efficiently forming an interchain disulphide. Inclusion of the cysteine capping agent N-ethyl maleimide at 10 mM in extraction buffer strongly suggested that this interchain disulphide was formed within the periplasm of live cells rather than during the extraction process itself (D.P.H, data not shown).
|
Fab' lacking an interchain disulphide bond is stable in the mammalian circulation, but are unstable in an E. coli extract incubated overnight at 60°C. These observations lead us to investigate the stability of various disulphide forms of Fab' fragments in more detail. Differential scanning calorimetry of purified Fab'-A showed the following rankings: LC-S HC-S, hinge-CAA, 64°C; LC-C HC-S, hinge-CAA, 69°C and LC-C HC-C ‘no-hinge’ and LC-C HC-C, hinge-CAA both at 71°C. These data confirmed that purified ‘
-inter’ Fab' is indeed very stable even though there is a small thermal stability penalty for the loss of the interchain disulphide bond.
The chemical stabilities of these purified disulphide variants were studied using guanidine chloride denaturation. The data in Fig. 6 show that the mid-point for the unfolding transition of Fab-A LC-C HC-C, hinge-CAA is 3.7 M GuHCl while the ‘no-hinge’ Fab version appears to be slightly more stable at 3.9 M GuHCl. The LC-C HC-S, hinge-CAA variant which also contains a stabilising interchain disulphide bond appears to be slightly less stable at 3.6 M GuHCl, suggesting that although disulphide bonded, the alignment of the constant domains may have been slightly strained in this molecule. The LC-S HC-S, hinge-CAA variant that lacked the interchain disulphide was 1 M less stable with a mid-point of transition of 2.6 M GuHCl. Collectively, these data demonstrated that Fab variants lacking an interchain disulphide bond retained significant thermal and chemical stability, but also confirmed that the interchain disulphide bond played a significant role in maximising the innate stability of these two polypeptides.
|
max.Optimal Fab format for multi-PEGylation
An ideal therapeutic Fab–PEG should be derived from a Fab' construct that is high yielding and easy to purify; the PEGylation efficiency should be high and the final PEGylated product should be easy to separate from non-PEGylated material. For multi-PEGylation, a number of these areas overlap. For example, a desire to perform periplasmic extractions at elevated temperatures dictates that the Fab' has an intact disulphide bond. High efficiencies of PEGylation are always desirable on process efficiency grounds, but with multi-PEGylation a very high percent formation of the di- or tri-PEGylated final product is especially desirable in order to simplify its chromatographic separation from un-reacted or partially PEGylated Fab' material. This concern becomes incrementally more important as more PEG molecules are added per Fab'. We found that di-PEGylation offered the best balance of these concerns since it could be performed on a Fab that contained an interchain disulphide bond, thereby allowing the use of elevated extraction temperatures and resulted in high PEGylation efficiencies.
We made a Fab fragment by the introduction of a stop codon after the Cys233 of human 
1 CH1. Hence, both LC and HC polypeptides terminated in disulphide bonded cysteines. Fab-C was made in ‘no-hinge’ Fab format and modified with 2 x 20 kDa, linear PEG. After optimisation of conditions, this reaction reached 88% di-PEGylation. Fab-C ‘no-hinge’ 2 x 20kDa PEG has been shown to have normal serum permanence in cynomolgous monkeys in a 56-day study. The PK of this di-PEGylated Fab was identical within experimental variation to that of a previously tested Fab–PEG with 1 x 40 kDa, branched PEG (data not shown). Di-PEGylated ‘no-hinge’ Fab-C has also been shown to be stable upon storage. After 28 days at 4°C, pH 5.0, no measurable loss of protein was observed, as studied by size-exclusion chromatography (S.P.H., unpublished data).
We have also prepared and tested a di-PEGylated murine equivalent of the ‘no-hinge’ construct. The serum permanence of this molecule in mice was very similar to other mouse Fab' that had been mono-PEGylated with 1 x 40 kDa, branched PEG (data not shown). Murine ‘no-hinge’ Fab have also shown very high levels of di-PEGylation efficiency, routinely >90% with the highest efficiency observed being 94%, as measured by HPLC analysis.
Disulphide engineering of novel interchain disulphide bonds
The efficiency with which interchain disulphide bonds were formed between the cKappa and hinge cysteines surprised us. The hinge cysteine is five amino acids distant in linear sequence from the CH1 cysteine that normally disulphide bonds to cKappa. Hence, the formation of an interchain disulphide in LC-C HC-S, hinge –CAA suggested that there was considerable flexibility/mobility in these regions. To test the structural limits of cKappa–hinge disulphide formation, we made variants of the C-terminus of cKappa and tested their abilities to form an interchain disulphide with the hinge cysteine by the combination with a ‘HC-S, hinge-CAA’ CH1 region. Conversely, we also made upper hinge variants in order to increase or decrease its potential flexibility and combined these with a wild-type cKappa. Fab'-A containing these cKappa–CH1 combinations were expressed in small scale cultures, and periplasmic samples were tested for the presence of disulphide bonded Fab'-A (or free LC and HC) by sodium dodecyl sulphate–polyacrylamide gel electrophoresis/immunoblotting with an anti-CH1 reagent.
The cKappa and hinge variants constructed along with their ability to form an interchain disulphide bond are shown in Table II. All of the upper hinge variants were capable of forming an interchain disulphide bond with wild-type cKappa. Most of the cKappa variants were also found to be capable of forming a disulphide with the hinge cysteine of a HC-S, hinge-CAA Fab'. Only Phe209Cys and Arg211Cys (Kabat numbering) were unable to disulphide bond to the hinge. Collectively, these data suggest that in the absence of an interchain disulphide bond, there may be considerable structural flexibility in the C-terminus of human cKappa and in the human 1 upper hinge. This is consistent with previous observations using structural overlay of Fab fragments where the cKappa–CH1 pair was seen to be a ‘tight unit’, but the disulphide bond was seen as a ‘somewhat mobile tether’, (Röthlisberger et al., 2005).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
Conjugation of PEG can confer increased half-life and solubility and decreased immunogenicity, enzymatic degradation and drug toxicity to proteins (Mahmood and Green, 2005
). PEGylated therapeutic proteins including insulin, growth hormones, cytokines, enzymes, antibodies and antibody fragments have been used successfully in the clinic in the last decade.
The PEGylation strategies used on the above proteins differ considerably, from multiple random lysine conjugations of small (2–5 kDa) PEG molecules to cysteine-specific mono-PEGylation with 40 kDa, branched or 100 kDa, linear PEG molecules. For antibody Fab' fragments, cysteine-specific PEGylation at the hinge has found favour since it reduces the possibility for steric interference of PEG with the antigen-binding site. Chapman et al. (1999) found that random attachment of 2 x 25 kDa, linear PEG resulted in loss of 29% of the Fab's binding affinity, attachment of 3 x 5 kDa, linear PEG caused a loss of 38% of binding affinity, while attachment of 6 x 5 kDa PEG resulted in loss of 83% of binding affinity. Koumenis et al. (2000) also observed a loss of 
2-fold of an IC50 after the addition of three or more 20 kDa, linear PEG to a F(ab')2. Clearly, a monoPEGylated fraction purified from a random PEGylation reaction is highly, likely to contain fractions of molecules that have lost some or all of their binding affinity. Previous work has shown that an addition of 40 kDa of PEG is required to confer a half-life approaching that of an IgG onto a Fab' fragment. To date, this has been achieved using a mild reductive activation of a single hinge thiol followed by reaction with a 40-kDa, branched PEG–maleimide.
In this work we have demonstrated that Fab' fragments that lack interchain disulphide bonds retain normal in vitro antigen affinity and in vivo pharmacokinetics and efficacy in rodents. Because of this fact, stronger non-thiol reductants such as TCEP can be used and hence open up the possibility of an increased PEGylation efficiency. We have also shown that purified Fab' lacking interchain disulphide bonds retains high thermal and chemical stability, unfolding at 64°C and 2.6 M guanidine, respectively. McDonagh et al. (2006) have used a similar rationale to selectively alter the cysteine composition of the four interchain disulphide bonds of a human IgG1 antibody in order to facilitate a more efficient and homogeneous attachment of drug conjugates. Although no data on physical stability was published, they also observed no significant effect on expression yield, purification properties, antigen affinity, in vitro and in vivo efficacy or PK.
Combined use of strong non-thiol reductants and disulphide engineering enables one to site specifically PEGylate at as many cysteines as is desired. However, chromatography practicalities mean that addition of 1, 2 or 3 PEG molecules is preferred over higher order PEGylations. Furthermore, when Fab' fragments are synthesised in E. coli, it is preferable that the Fab' retains its interchain disulphide bond in order to facilitate its primary recovery from the periplasm using a heated extraction process. We found that after balanced consideration of these factors, di-PEGylation of a LC-C HC-C, ‘no-hinge’ (Fab) is preferred. This has been achieved with an efficiency of 88% with a humanised Fab and 94% with a murine Fab'.
In addition to high efficiencies of PEGylation, multi-PEGylation offers the flexibility of using different sizes of PEG molecules to achieve a number of different PK outcomes. For example, the PK conferred by a single, 40 kDa, branched PEG can now be achieved by the addition of 2 x 20 kDa, linear PEG molecules, whilst PK longer than that conferred by a 40 kDa, branched PEG can be achieved simply by the conjugation of 2 x 30 kDa, linear PEG or 3 x 20 kDa, linear PEG. This offers considerable flexibility when it comes to the fine-tuning of the pharmacokinetic profile of a therapeutic candidate. We have demonstrated di- and tri-PEGylation with 20 and 30 kDa PEG resulting in the addition of 40 and 60 kDa of total PEG to various formats of Fab and Fab'. It is clearly possible to reduce or extend this even further by the conjugation of smaller or larger PEG molecules.
One surprising aspect of the work presented here is the efficiency with which the hinge cysteine239 disulphide bonded with the C-terminal cKappa cysteine214. There are five amino acids of the upper hinge N-DKTHT-C between the CH1 Cys233 and the hinge Cys239. This stretch of protein is known to be highly flexible and disordered in solution (Endo and Arata, 1985; Ito and Arata, 1985) and, for this reason, is generally absent from crystal structures of Fab' fragments. However, this unusual disulphide bond formed very efficiently in the E. coli periplasm.
To achieve high PEGylation efficiencies, it is necessary to ensure that the cysteines do not re-oxidise after reduction before the reaction with the maleimide group can be completed. We had considered that spatial separation of the LC and HC cysteines involved in formation of the disulphide bond might have been important. However, the PEGylation efficiency of the LC-C HC-C, ‘no-hinge’ Fab was not measurably different (at 88%) from that of the LC-C HC-S, hinge-CAA Fab', suggesting that linear separation of the cysteines is not important. We made cKappa and upper hinge variants to test the limits of this flexibility in disulphide bond formation. We found that the hinge cysteine has an even greater apparent reach than expected, because it was also capable of disulphide bonding with Asn210Cys of cKappa. Röthlisberger et al. (2005) have also shown by overlay of crystal structures that the C-termini of Fab containing disulphide bonds are highly mobile. These observations appear to be in agreement and confirm that there is a potential for a considerable degree of flexibility in both the C-terminus of cKappa and the upper hinge region.
In other classes and isotypes of antibody, the interchain disulphide bond utilises a CH1 cysteine at a different position (e.g. Cys127 of 4 IgG) in the linear sequence, but both this and the Cys233 of 1 CH1 must be capable of forming an interchain disulphide with both cKappa and cLambda cysteines. These cysteines do not appear to occupy exactly the same 3-dimensional location in Fab crystal structures. A need to successfully accommodate these subtle structural variations perhaps provides some insight into the flexibility of these protein regions.
That the interchain disulphide bond has been retained throughout antibody evolution suggests that it has an important role to play. This conclusion may appear to be at odds with our observations of its dispensability in terms of Fab' structural stability. Perhaps, it has a more important role to play during the antibody selection, affinity maturation and secretion processes that occur in B-cells than the final stability of the Fab arm. The Fab's studied in this work have all been carefully selected for their beneficial humanisation, affinity and expression properties—and so in essence represent on balance ‘good’ v-region pairs. However, the situation may be quite different within the ER lumen or cell surface of maturing B-cells. Here, immature v-regions that form slightly imperfect pairs must still function tolerably well in order to support their own maturation process. Röthlisberger et al. (2005) have shown that both the Fab constant regions and the interchain disulphide bond confer incremental stability benefits upon v-regions that have low intrinsic stability, but have a much smaller or no positive influence upon stable v-region pairs. Hence, it would seem possible that the role of Fab constant regions is to mate together slightly imperfect v-region pairs while the role of the interchain disulphide bond is to irreversibly cement such pairings so that they can survive passage through the ER, golgi and out of the cell.
In summary, we have shown that Fab' that lack interchain disulphide bonds either through mutagenesis or reduction/modification of cysteines remain sufficiently stable to perform normally as therapeutic Fab' or Fab–PEG. We have taken advantage of this information to create efficient alternative PEGylation strategies.
|
Footnotes |
|---|
Abbreviations: BSA, bovine serum albumin; ß-ME, ß-mercaptoethanol; ß-MA, ß-mercaptoethylamine; DTT, dithiothreitol; GuHCL, guanidine hydrochloride; GSH, glutathione (reduced); TBP, tris(butyl) phosphine; TCEP, tris(2-carboxyethyl) phosphine; Fab', Fragment antigen binding; PCR, polymerase chain reaction; HPLC, high-performance liquid chromatography; PBS, phosphate-buffered saline; AUC, area under the curve; PEG, polyethylene glycol; PK, pharmacokinetics; LC, light chain; HC, heavy chain; IPTG, isopropyl ß-D-thiogalactopyranoside; EDTA, ethylenediaminetetraacetic.
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
We would like to thank Andrew Popplewell for careful reading of this manuscript and Frances Stringer for discussion of cyno PK data.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
Albrecht H., Burke P.A., Natarajan A., Xiong C.Y., Kalicinsky M., Denardo G.L., Denardo S.J. (2004) Bioconj. Chem. 15:16–26.[CrossRef][Web of Science][Medline]
Begg G.E. and Speicher D.W. (1999) J. Biomol. Tech. 10:17–20.
Buchegger F., Pelegrin A., Hardman N., Heusser C., Lukas J., Dolci W., Mach J.P. (1992) Int. J. Cancer 50:416–422.[CrossRef][Web of Science][Medline]
Caliceti P. and Veronese F.M. (2003) Adv. Drug Deliv. Rev. 55:1261–1277.[CrossRef][Web of Science][Medline]
Carter P., et al. (1992) Bio/Technology 10:163–167.[CrossRef][Medline]
Chapman A.P., Antoniw P., Spitali M., West S., Stephens S., King D.J. (1999) Nat. Biotechnol. 17:780–783.[CrossRef][Web of Science][Medline]
Chapman A.P. (2002) Adv. Drug Deliv. Rev. 54:531–545.[CrossRef][Web of Science][Medline]
Cline D.J., Redding S.E., Brohawn S.G., Psathas J.N., Schneider J.P., Thorpe C. (2004) Biochemistry 43:15195–15203.[CrossRef][Medline]
Covell D.G., Barbet J., Holton O.D., Black C.D.V., Parker R.J., Weinstein J.N. (1986) Cancer Res. 46:3969–3978.
Cunningham-Rundles C., Zhuo Z.H.O.U., Griffith B., Keenan J. (1992) J. Immunol. Methods 152:177–190.[CrossRef][Web of Science][Medline]
Endo S. and Arata Y. (1985) Biochemistry 24:1561–1568.[CrossRef]
Han J.C. and Han G.Y. (1994) Anal. Biochem. 220:5–10.[CrossRef][Web of Science][Medline]
Harris J.M. and Chess R.B. (2003) Nat. Drug Disc. 2:214–221.[CrossRef]
Humphreys D.P., Chapman A.P., Reeks D.G., Lang V., Stephens P.E. (1997) J. Immunol. Methods 209:193–202.[CrossRef][Web of Science][Medline]
Humphreys D.P., et al. (1998) J. Immunol. Methods 217:1–10.[CrossRef][Web of Science][Medline]
Ito W. and Arata Y. (1985) Biochemistry 24:6467–6474.[CrossRef][Medline]
Kitamura K., et al. (1991) Cancer Res. 51:4310–4315.
Knight D.M., Jordan R.E., Kruszynski M., Tam S.H., Gile-Komar J., Treacy G., Heavner G.A. (2004) Platelets 15:409–418.[CrossRef][Web of Science][Medline]
Leong S.R., et al. (2001) Cytokine 16:106–119.[CrossRef][Web of Science][Medline]
McDonagh C.F., et al. (2006) Protein Eng. Des. Sel. 19:299–307.
Mahmood I. and Green M.D. (2005) Clin. Pharmacokinet. 44:331–347.[CrossRef][Web of Science][Medline]
Rodrigues M.L., Snedecor B., Chen C., Wong W.L.T., Garg S., Blank G.S., Maneval D., Carter P. (1993) J. Immunol. 151:6954–6961.[Abstract]
Rodwell J.D., Alvarez V.L., Lee C., Lopes A.D., Goers J.W.F., King H.D., Powsner H.J., McKearn T.J. (1986) Proc. Natl Acad. Sci. USA 83:2632–2636.
Röthlishberger D., Honegger A., Plückthun A. (2005) J. Mol. Biol. 347:773–789.[CrossRef][Web of Science][Medline]
Sato H. (2002) Adv. Drug Del. Rev. 54:487–504.[CrossRef][Web of Science][Medline]
Suzuki T., Kanbara N., Tomono T., Hayashi N., Shinohara I. (1984) Biochim. Biophys. Acta. 788:248–255.[CrossRef][Medline]
Wei S., et al. (2005) Immunobiology 210:109–119.[CrossRef][Web of Science][Medline]
Received October 16, 2006; revised January 9, 2007; accepted February 14, 2007.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||