PRESENT INVESTIGATION
`
`The research has been devoted to different protein engineering strategies for
`
`design of robust and predictable protein purification systems, addressing different
`
`problems faced in both large-scale and small-scale protein production, and can be
`
`divided into the two main objectives:
`
`(i) Destabilization of an affinity ligand to allow elution at milder conditions(I).
`
`(ii) Stabilization of affinity ligands towardsalkaline conditions(II, I, IV, V).
`
`(i) One of the problems in the recovery of antibodies by protein A-based affinity
`
`chromatography is the low pH, which is normally essential to elute the bound
`
`material
`
`from the column. Some antibodies are not able to withstand these
`
`conditions and suffer from irreversible inactivation. Therefore, it is interesting to
`
`outline a strategy to circumvent this problem by using milder elution conditions.
`
`Turn/loop engineering is considered a possible strategy to achieve destabilization
`
`of a protein. Thus, destabilized mutants of domain Z derived from SPA were
`
`designed.
`
`(ii) Another problem with a proteinaceous affinity ligand is its sensitivity to
`
`alkaline conditions. Many applications in the pharmaceutical and biotechnological
`
`field, such as large-scale production of antibodies for therapeutic use, require
`
`extreme attention to minimize contamination. In order to remove contaminants, a
`
`cleaning-in-place (CIP) step is often integrated in the purification protocol, often
`
`implying high pH. Unfortunately, most protein-based affinity media show high
`
`fragility towards this extremely harsh environment, making them less attractive as
`
`binding ligands. Asparagine has been shown to be the major contributor to the
`
`alkaline
`
`fragility. A protein engineering strategy consisting in replacing
`
`asparagine residues with other amino acids was applied. Three different affinity
`
`ligands,
`
`important
`
`in large-scale production of different
`
`target molecules,
`
`i.e.
`
`antibodies and albumin, have been remarkably stabilized. These proteins are the
`
`albumin-binding domain (ABD)(II, III) and the IgG-binding domain C2 of SPG
`
`(IV), as well as the IgG-binding domain Z of SPA (V).
`
`4]
`
`Page 41
`
`KASHIV EXHIBIT 1064
`IPR2019-00797
`
`Page 41
`
`KASHIV EXHIBIT 1064
`IPR2019-00797
`
`

`

`In this thesis small bacterial surface domains descending from Staphylococcus
`
`aureus and Streptococcus strains have been used. These are very suitable
`
`frameworks for different protein engineering and protein design methods. The
`
`framework of the C2 domain derived from SPG is composed of four B-strands
`
`crossed diagonally by an a-helix. The protein is ideal for protein engineering
`
`purposes since it is stable and well characterized, lacks cysteines, and is easily
`
`produced in large quantities. Both the albumin-binding domain (ABD) derived
`
`from SPG, and the Z domain derived from SPA, are three-helical proteins. The
`
`three-dimensional structure is well known and neither contains cysteines. Both
`
`proteins are very thermally and chemically stable and easily produced in large
`
`quantities. All three domains have potential
`
`to be used as affinity ligands for
`
`isolation of antibodies and their fragments, or albumin.
`
`6. Destabilization of Z to allow milder elution conditions
`
`(1)
`
`Affinity chromatography based upon SPA is
`
`frequently used for antibody
`
`recovery. However, due to the strong interactions of the SPA-IgG complex, low
`
`pH buffers are often required in order to elute the bound material
`
`from the
`
`column. Unfortunately, some antibodies are not able to withstand these conditions
`
`and suffer from irreversible inactivation. Therefore,
`
`it
`
`is most
`
`interesting to
`
`outline a strategy to circumvent this problem by using milder elution conditions.
`
`A possible strategy is the one outlined in paper I consisting in constructing
`
`destabilized mutants of protein Z derived from the B domain of SPA. For
`
`destabilization of proteins, turn/loop engineering can be one possible strategy
`
`according to several studies (Predki et al., 1996; Nagi and Regan, 1997; Nagi et
`
`al., 1999). Loop/turn sequences have been proposed to have some degree of
`
`tolerance towards protein engineering when they are not intimately involved in
`
`the function of the protein (Brunetetal., 1993; Castagnoli et al., 1994; Predki and
`
`Regan, 1995). These connecting segments can vary widely in both sequence and
`
`length without affecting the structure or the function (Leszczynski and Rose,
`
`1986; 1997; Viguera and Serrano, 1997). However, by engineering these
`
`segments the thermodynamic stabilities and folding pathways can be significantly
`
`42
`
`Page 42
`
`Page 42
`
`

`

`affected (Predki et al., 1996; Nagi and Regan, 1997; Nagi et al., 1999). By
`
`lengthening the loop, protein destabilization can be obtained due to higher energy
`
`requirements associated with the closure of longer loops comparedto shorter ones
`
`(Nagi and Regan, 1997; Nagi et al., 1999).
`
`In order to destabilize protein Z, two variants were constructed using site-directed
`
`mutagenesis of the second turn ofthis three-helix bundle domain composedofthe
`
`Asp-Pro-Ser-Gln sequence (Fig. 4). The rationale of engineering the second turn
`
`relies on earlier reports, in which it has been established that the third helix does
`
`not contribute with a binding surface to the Fc-fragment of the IgG-antibody
`
`(Jendeberg et al., 1996; Tashiro et al., 1997). However, when helix 3 is partly
`
`deleted the conformational stability and the affinity to IgG are substantially
`
`decreased, suggesting that helix 3 is essential to the formation of the global chain
`
`fold in solution (Bottomley et al., 1994). Hence,it is believed that the third helix
`
`is stabilizing the molecule and also exhibits some flexibility (Jendeberg et al.,
`
`1996). Thus, engineering of the second turn between helix 2 and helix 3 probably
`
`would affect the stability of the protein, allowing disruption of the Z-IgG complex
`
`at milder conditions.
`
`A suitable residue to introduceis glycine since the size of this amino acid is small
`
`and the side chain is a single hydrogen thus, minimizing interactions. The glycine
`
`residue also introduces exceptional flexibility of the peptide bond since a large
`
`range of permissible backbone dihedral angles are allowed (Predki et al., 1996;
`
`Nagi and Regan, 1997; Nagiet al., 1999). Additionally, glycine has been shown to
`
`be a poor a-helix former but commonly occurs in loop/turn sequences. In one
`
`mutant denoted Z6G, the second turn was extended with six glycine residues in
`
`order to evaluate the significance of the loop length. Thus, the total number of
`
`loop residues for this engineered variant was ten comparedto four in native Z. In
`
`the other mutant denoted ZL4G,the original turn sequence was exchanged for
`
`glycinesin order to evaluate the importanceofthe turn forming residues.
`
`43
`
`Page 43
`
`Page 43
`
`

`

`Helix 3
`Helix 2
`Helix 1
`> +> +>
`VDNKFN KEQONAFYEIL HLPNLN EEQRNAFIOQSLKD DPSQ
`SANLLAEAKKLNDA QAPK
`
`
`
`GGGGGG
`
`GGGG
`
`z
`Z6G
`ZLAG
`
`Fig. 4. The primary sequence of the Z-domain and the engineered variants. In
`
`Z6G the turn is extended with six glycines, whereas in ZL4G the original turn is
`
`exchanged for glycines.
`
`6.1. Characterization and proof-of-concept
`
`The two loop-engineered variants, Z6G and ZL4G, were characterized and
`
`compared to the native Z-molecule according to structure, stability, and function,
`
`using different technologies such as CD, and Biacore™. Also, the behavior as
`
`resin-bound ligands in affinity chromatography was investigated. Using the CD
`
`technology, it can be concluded that the a-helical structure of protein Z is retained
`
`for both mutants. It should be noted howeverthat changesin tertiary structure is
`
`difficult to detect with this technology. A small decrease in the a-helical content
`
`is detected compared to native Z, probably due to higher three-dimensional
`
`flexibility of the mutated proteins, since the method measures the mean residue
`ellipticity (MRE) (deg cm’ dmol’) andnot the real structure. Z6G exhibits the
`
`lowest ct-helical content. This is partly explained by the increase of loop forming
`
`amino acids, but not the numberof a-helical forming amino acids, resulting in
`
`decreased MRE.
`
`According to the stability analyses using CD, significant destabilization has been
`
`achieved for both engineered variants. A two-state folding mechanism without
`
`formation of any stable intermediates is assumed for protein Z, since this has been
`
`reported for the B-domain (Bai et al., 1997). Z is both chemically and thermally
`
`stable and does not reach a well-defined unfolded state in our experiments, which
`
`is necessary in order to be able to baseline-correct the data.
`
`In addition,
`
`the
`
`mutants do not show well-defined folded states even at low concentrations of
`
`denaturant at 20°C. Thus,all signals must be transformed to MRE to compare the
`
`different constructs. Both loop-engineered proteins are more susceptible to
`
`44
`
`Page 44
`
`Page 44
`
`

`

`chemical as well as thermal denaturation, and unfold at lower concentrations of
`
`GdnHCl and lower temperatures, compared to Z. A feasible way of comparing
`
`different mutants is to compare the Cm- and Tm-values, defined as midpoints of
`
`the denaturant- or temperature-induced transition. These midpoint values can be
`
`estimated from a graph with MRE versus the concentration of GdnHCl and
`
`temperature respectively. As can be seen in table 3, Z6G is significantly
`
`destabilized compared to Z. Similar Tm-values are also observed at pH 4
`
`compared to pH 7.3 in table 3, suggesting that further destabilization is not
`
`achieved when lowering the pH to 4. The reversibility of the denaturation is
`
`retained for the engineered variants. Thus, neither of the loop-engineered variants
`
`is prone to aggregation during refolding. The lower stability of Z6G is in
`
`accordance with a study of a four-helix bundle protein, ROP,
`
`in which loop
`
`substitution and lengthening were analyzed (Nagi and Regan, 1997; Nagi etal.,
`
`1999),
`
`Table 3. Midpoints of the denaturation transition at pH 7.3.
`
`Cm (M)
`
`Tm (°C)
`
`Z
`
`ZLAG
`
`Z6G
`
`4A
`
`2.4
`
`2.0
`
`>75
`
`55
`
`45
`
`The binding behavior of the mutated proteins looks different compared to Z, when
`
`analyzed by Biacore™based on the surface plasmon resonance technology (Table
`
`4). The observed decreased binding affinities are almost entirely due to the
`
`increased dissociation rate constants. The difference in binding free energy
`
`suggests that the Z6G-IgG complex is the weakest. Thus, this may indicate that
`
`the interaction to IgG can be interrupted at milder conditions. Interestingly, ZL4G
`
`also differs slightly in the k,-value at pH 7.4, which might be explainedby either a
`
`structural rearrangementthat results in rate-limiting in the recognition process, or
`
`decreased charge repulsion since an aspartate has been removed from ZL4G,
`
`which has a pI of about 5. However, the pI for IgG is only marginally acidic or
`
`neutral.
`
`45
`
`Page 45
`
`Page 45
`
`

`

`Table 4. Binding characteristics to humanpolyclonal IgG.
`
`k,
`(M"'s"')
`
`10.1x10*
`
`8.3x10*
`
`ZLAG
`
`Z6G
`
`ka
`(st)
`
`Ka
`(M")
`
`AAG
`(kcal mol")
`
`18.3x10°
`
`17.8x10°
`
`0.6x10"
`
`0.5x10’
`
`0.4
`
`0.5
`
`The proof-of-concept is the behavior of the NHS-immobilized loop-engineered
`
`constructs in a standard affinity chromatography protocol. As predicted, a higher
`
`proportion of bound IgG can be eluted at milder pH from the mutant columns
`
`compared to the Z-column (Table 5). At a pH of 4.5 almostall IgG can be eluted
`
`from the mutant columns compared to 70% for Z. Notably,
`
`the interaction is
`
`overall weaker for the mutants, which is also verified by the Biacore™. A minor
`
`part of the IgG does not bind to the mutant columns and is washed out probably
`
`because different subclasses of polyclonal IgG may have different affinity for the
`
`Z-domain (Langone, 1982). The elution is probably accomplished by means of
`
`protonation of the binding site and not destruction of the Z-domain. This is
`
`supported by the CD analyses, in which the structure and the stability are not
`
`significantly changed when lowering the pH to 4. Therefore, the changed elution
`
`behavior
`
`is probably due to larger
`
`flexibility of the structure, and as
`
`a
`
`consequencean increased accessibility of the entire binding site composed of both
`
`the IgG-surface, as well as the surface of the Z-molecule. However, structural
`
`changes of the IgG-molecule at low pH cannotbe ruled out.
`
`Table 5. % eluted IgG at different pH.
`
`pH 4.5 (%)
`
`pH 4.75 (%)
`
`pH 5.0 (%)
`
`Z
`
`ZLAG
`
`Z6G
`
`70
`
`93
`
`97
`
`49
`
`78
`
`82
`
`31
`
`64
`
`67
`
`46
`
`Page 46
`
`Page 46
`
`

`

`The difference between the two turn/loop-engineered variants is interesting. All
`
`analyses indicate that lengthening the loop sequence affects this protein’s special
`
`properties more than replacing the native loop sequence. The closure of the longer
`
`loop in Z6G requires more energy. Also, by occupying a larger volumein space,
`
`the loop might affect the flexibility of the third helix. In the case with ZL4G, the
`
`native sequenceis replaced resulting in a loss of a proline residue. This proline is
`
`conserved in all domains of SPA, indicating that it may be important (Tashiro et
`
`al., 1997). Also, prolines exhibit low conformational freedom and are commonly
`
`occurring in turns/loops. Thus, exchanging this proline for a glycine probably
`
`gives the turn and therebythethird helix a larger flexibility.
`
`In conclusion, the results presented in this paper show that turn/loop engineering
`
`of a three-helix bundle protein can have large effect on the overall stability and
`
`function of the protein. Accordingly, the interaction with a target molecule can be
`
`weakened. The described ligands circumvent
`
`the harsh elution conditions
`
`associated with SPA, by allowing a milder affinity chromatography procedure.
`
`This should be applicable in the production of certain monoclonal antibodies,
`
`which are susceptible to extremes of pH-values. In addition, the reverse might be
`
`an interesting application,
`
`in which IgG is immobilized to the matrix and the
`
`mutants described here function as purification handles of sensitive target
`
`proteins.
`
`7, Stabilization of ABD towards alkaline conditions (II,
`
`It)
`
`HSA has been extensively studied and has found widespread use both in
`
`therapeutic and biotechnological applications. HSA is traditionally obtained by
`
`fractionation from human plasma. However, the potential hazard of contamination
`
`with human viruses and the low supply have resulted in development of
`
`alternative recombinant production systems in yeast (Quirk et al., 1989). A cost-
`
`effective purification strategy may be to use the albumin-binding domain (ABD)
`
`derived from SPG (Olsson et al., 1987), as a matrix-bound affinity ligand.
`
`In
`
`industrial large-scale production protocols CIP-treatment, often implying high pH,
`
`47
`
`Page 47
`
`Page 47
`
`

`

`is routinely included in order to remove contaminants from the purification
`
`devices thus,
`
`requiring high alkaline stability of the matrix-bound ligand.
`
`However, many proteinaceousligands including ABD showhighfragility towards
`
`this treatment.
`
`In paper II,
`
`it
`
`is shown that a rationale including a protein
`
`engineering strategy can be successful
`
`in enabling a protein-based affinity
`
`medium to withstand the harsh conditions associated with a NaOH-based CIP-
`
`step.
`
`ABDisa three-helix bundle consisting of 46 amino acids (Kraulis et al., 1996).
`
`Thus,
`
`the structural resemblance to the three-helix structure of the protein A
`
`domainsis strikingly. Despite that no homology exists between these sequences.
`
`Four asparagines (Asn9, Asn23, Asn26, and Asn27)
`
`(Kraulis et al., 1996;
`
`Johansson et al., 2002) reside within the domain (Fig. 5). Since the asparagines
`
`have been reported to be particular sensitive towards alkaline conditions a single
`
`variant denoted ABD*,
`
`in which all four residues were exchanged for residues
`
`less fragile, was made using site-directed mutagenesis.
`
`In order to choose
`
`appropriate residues, comparison with homologous sequences of albumin-binding
`
`proteins from other strains was performed. Asn9 in helix 1 was therefore
`
`substituted for
`
`leucine. Asn23 and Asn26 in helix 2 were substituted for
`
`aspartates. Finally, Asn27 in the turn connecting helix 2 and 3 was exchanged for
`
`lysine.
`
`
`
`>—————__ <+_—_>+———_
`
`LAEAKVLANRELDK YGVS DYYKNLIN NAKT VEGVKALIDEILA ALP
`
`Helix 1
`
`Helix 2
`
`Helix 3
`
`'
`
`L
`
`14
`
`D DK
`
`Fig. 5. The primary sequence of ABD.Thesubstituted asparaginesare indicated.
`
`In paperIII, the possibility of improving the performance of ABD* asan affinity
`
`ligand in HSA-purification, by genetically fusing two monomer units,
`
`is
`
`48
`
`Page 48
`
`Page 48
`
`

`

`evaluated. Also,
`
`linker engineering is performed to achieve a robust affinity
`
`ligand for use in a predictable purification system. Additionally, directed coupling
`
`to the chromatography medium is compared to a non-directed strategy.
`
`Multimerization of affinity ligands for coupling to a chromatography matrix might
`
`result in an increased binding capacity of the resin, since more functional domains
`
`should be available for capture of the target. However, careful considerations
`
`must be taken to the design of the connecting linker regions (Alfthan et al., 1995;
`
`Arai et al., 2001). In order to optimize stability and function of the multimer,
`
`properties such as the linker composition and length are essential. Also, the linker
`
`should not be protease sensitive during expression in the host organism of choice.
`
`Development of a linker, which also can stand the harsh CIP-procedures, would
`
`render multimerization possible and thereby be of great
`
`interest. Since linker
`
`regions often are flexible and have undefined three-dimensional structures, the
`
`amino acids in a linker would be more sensitive when exposed for high pH than
`
`well-defined and stable structures within the domains. Therefore,
`
`it seems
`
`reasonable that shorter linkers also would decrease the sensitivity of the molecule.
`
`Three different linkers (A: VDANS, B: VDADS and C: GGGSG)wereevaluated.
`
`Thefirst linker existed in the cloning vector and also includes asparagine, giving
`
`the possibility of evaluating the impact of asparagines on the fragility of a linker
`
`region towards alkaline conditions.
`
`In the second linker,
`
`the asparagine was
`
`exchanged for aspartate, since the Asn-Ser sequence has been reported to be
`
`particular fragile (Geiger and Clarke, 1987). Finally, the third linker includes only
`
`glycine and serine since linkers composed of these amino acids have earlier been
`
`reported to be stable, highly flexible, and also successful in separating domains
`
`(Argos, 1990; Alfthan et al., 1995; Takeda et al., 2001). Also, linkers occurring in
`
`nature are often rich in small polar amino acids such as glycine and serine (Argos,
`
`1990). Our designed linkers are only 5 residues in length. Notably, the existing
`
`linkers connecting the albumin-binding domains in SPG are approximately 30
`
`amino acids, and the linkers separating the IgG-binding domains are about 15
`
`residues (Olsson et al., 1987). To date, these long linkers have no known obvious
`
`biological function. Notably, the IgG-binding domains in protein A are separated
`
`by shorter linker regions of 8 to 13 amino acids (Uhlénet al., 1984).
`
`49
`
`Page 49
`
`Page 49
`
`

`

`7.1. Characterization and proof-of-concept
`
`When substituting the asparagine residues it
`
`is
`
`important
`
`that properties
`
`concerningstructure, stability, and function are not negatively affected. Therefore,
`
`the characteristics of the mutant were analyzed using CD and Biacore™.
`
`In
`
`addition, the behavior towards alkaline conditions was analyzed using an affinity
`
`chromatography protocol with an integrated CIP-step. The four mutations in
`
`ABD*do not impose any major secondary structure changes according to the CD-
`
`analysis. The a-helical structure is retained for ABD* with only a small decrease
`
`in a-helical content comparedto parental ABD.
`
`As observed for the Z molecule (paper I), ABD is chemically and thermally stable
`
`and does not reach a defined unfolded state in our experiments. As in the case
`
`with protein Z, the CD-signals can then be transformed to MRE to compare the
`
`two variants.
`
`Table 6. Midpoints of the denaturationtransition at pH 7.
`
`ABD
`
`ABD*
`
`Cm (M)
`
`42
`
`>5
`
`Tm (°C)
`
`70
`
`>80
`
`The ABD molecule exhibits similar Cm- and Tm-values as protein Z (Table 3, 6).
`
`Additionally,
`
`in accordance to Z both ABD and ABD* show fully reversible
`
`denaturation transitions. Interestingly, the engineered variant ABD* exhibits both
`
`increased chemical and thermal stability. A possible explanation for this behavior
`
`might be the leucine that has replaced the asparagine in the first helix. This
`
`particular asparagine is pointing somewhatinwards to the interior region of the
`
`protein, whereasthe other three residues are located on the surface exposed to the
`
`solvent (Kraulis et al., 1996; Johanssonet al., 2002), and might have less effect on
`
`the thermostability. The substitution of Asn9 results in introduction of a residue
`
`with higher helix propensity. Additionally,
`
`leucine is more hydrophobic in
`
`character compared to asparagine. Therefore,
`
`it can be argued that a leucine
`
`instead of Asn9 mightstabilize the hydrophobic core. Thus, stabilizing one region
`
`mayresult in stabilization of the entire protein. This is in accordance with a study
`
`50
`
`Page 50
`
`Page 50
`
`

`

`of iso-l-cytochromec, in which the substitution of an asparagine centered in the
`
`middle of an a-helix for an isoleucine resulted in a significant increase in the
`
`transition temperature, Tm (Das et al., 1989; Hickey et al., 1991). Increased
`
`thermostability due to substitutions of boundary residues has also been described
`
`for the B1 domain of SPG (Malakauskas and Mayo, 1998). However, increased
`
`thermostability of ABD* due to favorable charge-charge interactions,
`
`i.e. salt
`
`bridges, of the surface mutations cannot be ruled out.
`
`The binding behavior of ABD*is similar to the binding behavior of native ABD
`
`as analyzed by Biacore™ (Table 7). The dissociation rate constants are almost
`
`identical,
`
`indicating that the stability of the ABD-HSA complex has not been
`
`affected. The k,-values differ slightly and the calculated affinity constant
`
`is
`
`therefore decreased for the mutant. This might be due to a small structural
`
`rearrangement in ABD*, or increased charge repulsion since ABD* and albumin
`
`have pl of about 5 and the analysis is run at about7.4.
`
`Table 7. Binding characteristics to HSA.
`
`AAG
`Ka
`ka
`ka
`(M")
`(s"')
`(M™s"')
`(kcal mol")
`
`
`‘ABD~—sOdLIxl0*——~S™—«*&L.:.22109.0x10" (0)
`ABD*
`2.5x10*
`1.3x10°
`1.9x10/
`0.9
`
`Finally, ABD* has obtained a significantly increased stability towards alkaline
`
`conditions. When incubated in 0.5 M NaOH for 24 hours at room temperature,
`
`only a slight decrease in binding capacity is detected using Biacore™. On the
`
`other hand, a total loss of HSA-binding capacity is revealed for ABD. The final
`
`proof-of-concept, consisting in immobilization to a standard affinity matrix,
`
`further support this observation. ABD*is fully active after a total of 34 minutes of
`
`exposure to 0.5 M NaOH.Also, the selectivity towards HSAis retained. ABD in
`
`contrast is significantly less stable and shows a 50% decrease of activity after the
`
`same time of exposure.
`
`Apparently, one or several of the asparagines in ABD are modified to such an
`
`extent that the protein’s binding capacity is severely decreased. -Asn-Gly- and -
`
`31
`
`Page 51
`
`Page 51
`
`

`

`Asn-Ser- sequences are knownto be particular sensitive towards high pH (Geiger
`
`and Clarke, 1987; Robinson and Robinson, 1991). Additionally, the -Asn-Ala-
`
`sequenceis considered sensitive due to the small size of alanine thus,facilitating
`
`the deamidation mechanism. Interestingly, none of the asparagine residues in
`
`ABDis succeeded by a glycine or a serine. Instead,
`
`they are linked to large,
`
`branched, and hydrophobic residues
`
`that are supposed to protect
`
`towards
`
`deamidation, except Asn27 that is succeeded by an alanine (Geiger and Clarke,
`
`1987; Kossiakoff, 1988; Robinson and Robinson, 1991). Specifically, Asn9 is
`
`followed by arginine, Asn23 by leucine, and Asn26 by asparagine. Interestingly,
`
`the first three asparagines are situated in a-helices, whereas Asn27 is the first
`
`residue of the second turn. a-helices may havestabilizing effects on deamidation
`
`(Kosky et al., 1999; Xie and Schowen, 1999). Therefore,
`
`it
`
`is tempting to
`
`speculate if the most probable site for degradation is Asn27.
`
`The three different dimers, ABD*dimerA, ABD*dimerB, and ABD*dimerC,with
`
`different linkers were compared to their monomeric counterpart ABD* according
`
`to capacity and stability. From these data it can be concluded that dimerization
`
`results in higher capacity compared to the monomeras investigated by Biacore™.
`
`The increased capacity of the dimer
`
`is
`
`further confirmed in an affinity
`
`chromatography purification of HSA using standard protocols. Therefore,
`
`dimerization of an affinity ligand may be a convenient way of increasing the
`
`binding capacity of the ligand. Multimers of higher order might
`
`increase the
`
`capacity further. However, earlier studies made on staphylococcal protein A
`
`indicate that there was no advantage in using more than two domains (Ljungquist
`
`et al., 1989). This is of course dependenton the specific protein-proteinpair.
`
`Additionally, a remarkable increase in binding capacity is observed for the
`
`directed immobilization method using the C-terminal cysteine, compared to the
`
`non-directed amine
`
`coupling utilizing NHS-chemistry. Applying thioether
`
`chemistry when immobilizing to affinity chromatography adsorbents presents
`
`several advantages. Since the introduced C-terminal thiol is the only one in the
`
`sequence of ABD, no amino acids involved in the capture of HSA are used for
`
`coupling to the solid support. Therefore, the binding site of ABD* should be
`
`accessible. These results stress the importance of how an affinity ligand is
`
`presented to its target molecule. Another advantage with the thioether coupling is
`
`52
`
`Page 52
`
`Page 52
`
`

`

`the higher resistance towards high pH compared to the NHS-chemistry normally
`
`used. According to the supplier, the thioether linkage can withstand pH in the
`
`interval 2-14, whereas the NHS-couplingis stable in the interval of 3-12. Thus,if
`
`the resin is going to be exposed for alkaline conditions the thioether chemistry
`
`should be considered for optimization of the purification system.
`
`Interestingly, the function of the affinity ligand is retained when decreasing the
`
`numberof linker residues from 30 in native SPG to 5 amino acids. Also, recent
`
`results of the three-dimensional structure of ABD indicate that one amino acid of
`
`the N-terminallinker region can bepart of the first a-helix of ABD (Johansson et
`
`al., 2002). However,
`
`three of the C-terminal amino acids in the domain are
`
`suggested to be flexible, resulting in a flexible part between the two domains of
`
`about seven amino acids. Hence, the nature of the linker must be considered on a
`
`case-by-case basis since it depends on the characteristics of the individual
`
`domains such assize and function.
`
`ABD* and the three different ABD*dimers exhibit remarkably high stability
`
`towards alkaline conditions in relation ta other proteinaceous molecules. After 7h
`
`of exposure towards 0.1 M NaOH, ABD*, ABD*dimerA, and ABD*dimerC still
`
`retain about 85% of the original capacity. The observed deactivation is a
`
`combination of several parameters such as the domain stability, linker stability,
`
`and also the stability of the coupling chemistry. ABD*dimerB (VDADS)is the
`
`mostresistant construct towards alkaline cleaning-in-place conditions and retains
`
`as much as 95%ofthe activity after 7 hours exposure to 0.1 M NaOH. Thus,this
`
`variant would therefore be a very interesting candidate for large-scale purification
`
`of HSA. The increased stability of the VDADSlinker is most probably due to the
`
`lack of the sensitive asparagine residue, since this residue is the only difference
`
`between linker A and B. The -Asn-Ser- sequenceis knownto besensitive towards
`
`cleavage in alkaline conditions (Geiger and Clarke, 1987). Surprisingly,
`
`the
`
`GGGSGlinker shows the samestability as the VDANS linker. None of these
`
`aminoacids have been reported to exhibit sensitivity towards alkaline conditions.
`
`However, the GGGSG sequence increasesthe flexibility of the linker due to the
`
`rotational freedom of the glycine residues (Argos, 1990) and thereby, might the
`
`stability decrease (Robinson and Sauer, 1998).
`
`53
`
`Page 53
`
`Page 53
`
`

`

`In conclusion, paper II shows that by exchanging the asparaginesin a protein for
`
`other residues, a stabilization of the domain towards alkaline conditions can be
`
`achieved. Since noneof the asparagines in ABDare involvedin the interaction to
`
`albumin or are important for other properties, all can be exchanged in a single
`
`variant without negatively affecting the characteristics of the domain. It must be
`
`stressed that this might notbe feasible for all proteins. There might be asparagines
`
`that are not possible to replace. Additionally, in paper III the importance of how
`
`an affinity ligand is presented to its target molecule is pointed out, and by using
`
`directed immobilization the binding capacity can be significantly increased. Also,
`
`dimerization of ABD* results in an affinity ligand with increased HSA-binding
`
`capacity compared to the monomeric counterpart. Noteworthy, the composition of
`
`the linker connecting monomer units in multimers is
`
`important
`
`in order to
`
`withstand harsh CIP-treatment thus asparagines should be avoided. Therefore,
`
`ABD*dimerBrepresents an interesting ligand for large-scale purification of HSA.
`
`8. Stabilization of C2 towards alkaline conditions (IV)
`
`The IgG-binding region of SPG has been thoroughly characterized, and exploited
`
`in commercially available protein-based affinity chromatography media. The IgG-
`
`binding region consists of three domains, Cl, C2, and C3, exhibiting high
`
`sequence homology (Olsson et al., 1987). The sensitivity of these domains to
`
`alkaline pH hasearlier been reported (Gowardet al., 1991).
`
`In this paper, the emphasis wasto stabilize the C2 domain, whichhasthe potential
`
`to be used in large-scale purification of monoclonal antibodies. This ligand is a
`
`small protein constituted of only 55 residues. The structure is composed of a B-
`
`sheet crossed diagonally by a single o-helix (Fig. 6). Thus,
`
`the structure is
`
`different to the earlier stabilized ABD domain (paperII). The fragility of the C2
`
`domain was addressed by designing single mutants to resolve the contribution of
`
`each asparagine residue to the deactivation of the domain.
`
`In addition,
`
`the
`
`possibility of further stabilization by aspartate and glutamine replacements was
`
`investigated, since also these residues can be chemically modified (Stephenson
`
`and Clarke, 1989; Wright, 1991a, b; Tomizawaet al., 1995). A first generation of
`
`mutated variants was assigned to asparagine replacements since this residue is
`
`known to be the most sensitive amino acid towards alkaline conditions. The C2
`
`54
`
`Page 54
`
`Page 54
`
`

`

`domain contains three asparagine residues (Fig. 6), which were substituted for
`
`alanines in a single point mutation fashion, since no alternative amino acids were
`
`found in the homologous sequences. Since alanine is a small amino acid and not
`
`known to introduce any unusual conformational properties in the structure, it is
`
`suitable for this purpose. Also, a double mutant, C2n736a, as well as a triple
`
`mutant, C2n7,34,36,, were constructed. Thereafter, a second generation of mutants
`
`was generated using the most stable mutated variant, C2n7.36,, as scaffold. Two
`
`single mutants and one triple mutant were designed,
`
`in which the aspartate
`
`residues were substituted for glutamate residues, C2n7,36ap3sz, C2n736ap39r, and
`
`C2n7,36ap21,45.46E,
`
`thereby retaining the charge of the domain. Furthermore,
`
`in
`
`contrast to aspartate residues no studies have been reported on isomerization of
`
`glutamate residues, probably due to the extra -CH> group of glutamates. C2736,
`
`was also used as scaffold for the single glutamine replacement. The glutamine
`
`was substituted for alanine resulting in C2Nn7,36AQ31A-
`
`BI
`
`B2
`
`a
`
`B3
`
`B4
`
`<+_-—W' +> <cM\!_—_—_
`
`+>
`
`<>
`
`T YKLVING KT LKGETTT EAVD AATAEKVFKQYAND NGVDG EWTY DDATKTFTVT E
`
`’
`
`4
`
`A
`
`Fig. 6. The primary sequence of C2. The asparagines were exchanged for
`
`alanines.
`
`8.1. Characterization
`
`The different mutants were characterized and comparedto the native C2-molecule
`
`using CD and Biacore™, Additionally,
`
`IEF and an affinity chromatography
`
`protocol were used to evaluate the sensitivity towards alkaline conditions. The
`
`two single mutations, N7A and N36A,result in constructs with retained affinity to
`
`55
`
`Page 55
`
`Page 55
`
`

`

`the Fc-fragm