The Effect of Buffers on Protein
`Conformational Stability
`
`Sydney O. Ugwu and Shireesh P. Apte*
`
`B
`
`uffers used to formulate proteins should not serve as
`substrates or inhibitors. They should exhibit little or no
`change in pH with temperature, show insignificant pen-
`etration through biological membranes, and have max-
`imum buffer capacity at a pH where the protein exhibits opti-
`mal stability. In conformity with the proposition that “Nature
`designs the optimum molecules,” buffers should mimic the an-
`tidenaturant properties of nature exhibited by osmolytes (1–5)
`that are independent of the evolutionary history of the proteins
`(6, 7). Such properties may include preferential exclusion from
`the protein domain (8–11) and stabilization without changing
`the denaturation Gibbs energy (⌬Gd) (12).
`Conformational instability refers not only to unfolding, ag-
`gregation, or denaturation but also to subtle changes in local-
`ized protein domains and the alteration of enzyme catalytic
`properties (13) that may result from buffer-component bind-
`ing, proton transfer, and metal or substrate binding effects di-
`rectly or indirectly mediated by buffers or by buffers themselves
`acting as pseudosubstrates.
`Salts can affect protein conformation to the extent that the
`anions or cations of the salt could be potential buffer compo-
`nents. When the salt concentration is much larger than that of
`the buffer, the salt becomes the effective buffer in the reaction.
`The mechanisms or combinations thereof by which buffers
`may cause protein stabilization (or destabilization) are com-
`plex and not well understood. The problem is compounded by
`the inability to definitively differentiate between various pro-
`tein stabilization mechanisms, the small free energies of stabi-
`lization of globular proteins (14–16), and a paucity of review
`manuscripts on this subject in the literature. The authors ad-
`dress some of these issues as they relate to buffers used in the
`formulation of proteins. The effect of buffers that may be used
`in the extraction, purification, dialysis, refolding, or assay of
`proteins on protein conformation is not discussed.
`
`Buffer effects on freeze drying
`Change in pH as a result of buffer salt crystallization. When inor-
`ganic salts are used as buffers, the freezing point of the mono-
`ionized species (salt) can be different from that of the non-
`ionized (i.e., free acid or base) species and from its higher ionized
`species. This difference leads to the freezing of one form before
`
`www.pharmtech.com
`
`PHOTODISC, INC.
`
`The extent to which a particular protein may
`be stabilized or destabilized by a buffer
`depends on many factors, thereby making
`the selection of a buffer for formulating a
`specific protein a formidable challenge.The
`authors describe qualitative and
`semiqualitative correlations to help in the
`selection of a buffer for a particular protein
`and formulation.
`
`Sydney O. Ugwu, PhD, of Baxter IV
`Systems in Murray Hill, New Jersey, is
`currently at NeoPharm Inc. (Waukegan, IL).
`Shireesh P. Apte, PhD, is a lead
`scientist at Baxter IV Systems (Murray Hill,
`NJ) and may be contacted at Alcon Research
`Inc., 6201 S. Freeway, Ft. Worth, TX 76134,
`tel. 817.551.4901, fax 817.551.8626,
`shireesh.apte@alconlabs.com.
`
`*To whom all correspondence should be addressed.
`
`86 Pharmaceutical Technology MARCH 2004
`
`KASHIV EXHIBIT 1072
`IPR2019-00791
`
`Page 1
`
`

`

`List of abbreviations
`ACES:N-2-acetamido-2-aminoethane sulfonic acid
`2,3-BPG: 2,3-bis phosphoglycerate
`CHES:1-[N-cyclohecyamino]-ethane sulfonic acid
`CLARP: caspase-like-apoptosis-regulatory-protein
`DIPSO:3-[N,N-bis(hydroxyethyl)amino]-2-hydroxypropane sulfonic acid
`G-CSF: granulocyte colony stimulating factor
`Good’s buffers: zwitterionic buffers containing aminoalkyl sulfonate (e.g.,DIPSO,
`MES,HEPPSO,HEPES)
`HEPES:4-(2-hydroxyethyl)piperazine-N’-2-ethane sulfonic acid
`HEPPSO:[N-(2-hydroxyethyl)piperazine-N’-2-hydroxypropane sulfonic acid
`KCl: potassium chloride
`NaCl: sodium chloride
`NADPH: nicotinamide adenine dinucleotide phosphate
`Na2HPO4: disodium phosphate
`(NH4)2SO4: ammonium sulfate
`NTB: nitrothiobenzoate
`MES: 2-(N-morpholino)ethane sulfonic acid
`MOPS: 3-(N-morpholino-2-hydroxypropane sulfonic acid
`PIPES: [piperazine-N,N’-bis(ethane sulfonic acid)]
`POPSO: piperazine-N,N’-bis(2-hydeoxypropane sulfonic acid)
`PBS: phosphate buffered salineTAPS: N-tris[hydroxymethyl]methyl-3-
`aminopropane sulfonic acid
`TEA: triethylamine
`TES: 1% sodium dodecyl sulfate ⫹ 5mM EDTA ⫹ 10 mM TRIS-HCL
`TRIS-HCL: Tris-(hydroxymethyl)aminomethane hydrochloride
`
`the other during the freezing phase of lyophilization (17). Such
`a phenomenon has been linked to drastic changes in pH of the
`liquid medium during freezing, which can lead to the denatu-
`ration of the protein being lyophilized (18, 19). If an ampho-
`teric molecule were to function as a buffer containing both
`acidic and basic groups on one molecule, one would expect neg-
`ligible pH shifts to occur during the crystallization of this zwit-
`terionic molecule (20). Such is indeed the case for various or-
`ganic buffers broadly categorized as aminoalkylsulfonate
`zwitterions (21). Good et. al. prepared and disclosed such buffers
`in their classic publication (22).
`Researchers have shown that replacing the Na⫹ cation with
`the K⫹ cation in a phosphate buffer could significantly decrease
`the pH shift during the freezing stage of lyophilization (23). A
`potassium phosphate buffer at pH 7.2 exhibited a eutectic point
`at a temperature greater than ⫺10 ⬚C. However, the sodium
`cation counterpart showed a eutectic point at a temperature
`below ⫺20 ⬚C. Monoclonal antibodies against HBV and L-se-
`lectin, humanized IgG, as well as monomeric and tetrameric ␤-
`galactosidase exhibited less aggregation when subjected to
`freeze–thaw cycles with a potassium phosphate buffer than with
`a sodium phosphate buffer (24). Similarly, the propensity of re-
`combinant hemoglobin to denature as a result of phase sepa-
`ration from a polyethylene glycol–dextran matrix was reduced
`when NaCl was replaced with KCl in the formulation buffer. In
`this case, the sodium phosphate buffer did not exhibit a pH shift
`during freezing owing to inhibition of crystallization of dis-
`odium phosphate by the polymer (25). However, replacing NaCl
`with KCl did decrease the phase separation caused by anneal-
`88 Pharmaceutical Technology MARCH 2004
`
`ing at ⫺7 ⬚C because of the propensity for KCl, but not NaCl,
`to form a stable glass at this temperature (26). Also, the specific
`surface area of freeze-dried bovine IgG from solutions con-
`taining NaCl was found to be significantly higher than those
`containing KCl (27). Annealing also increases the surface ac-
`cumulation of proteins at the ice–liquid interface so that the
`formation of a stable glass at the annealing temperature is es-
`pecially important to minimize denaturation caused by such a
`mechanism (28).
`The rate of aggregation of recombinant human interleukin-1
`receptor antagonist (rhIl-1ra) was greater in mannitol–phosphate
`formulations than in glycine–phosphate formulations, possibly
`owing to the inhibition of the selective crystallization of the
`dibasic salt by glycine during freezing, thereby preventing large
`localized pH changes in the frozen matrix (29).
`The effect of various buffer solutions on freezing damage to
`rabbit-muscle-derived lactate dehydrogenase, type II (LDH,
`isoionic point [pI] ⫽ 4.6) was examined with sodium phos-
`phate, TRIS-HCl, HEPES, and citrate buffers (50 mM, pH 7.0)
`and pH 7.4 (30). The activity recovery was directly proportional
`to enzyme concentration and was the lowest in the sodium
`phosphate buffer (31). The activity increased in the following
`order: citrate ⬍ Tris ⬇ potassium phosphate ⬍ HEPES. The
`low activity recovery in the sodium phosphate buffer was at-
`tributed to its significant pH shift on freezing (32). The study
`revealed no clear pattern relating recovery of activity after freez-
`ing to the freezing rate because an intermediate freezing rate
`gave the highest recovery of activity (31). The researchers hy-
`pothesized that the slowest freezing method actually resulted
`in a greater degree of supercooling and better thermal equili-
`bration throughout the volume of liquid such that, when ice
`crystals nucleated, the freezing rate actually was faster than de-
`signed. Therefore, the study illustrates the need to control the
`extent of supercooling by seeding the cooling liquid when com-
`paring the effects of buffers or lyoprotectants on the stability
`of freeze-dried proteins.
`In another study, the decreased solid-state stability of
`lyophillized recombinant human interferon ␥ (pI ⫽ 10.3) in
`sodium succinate buffer as compared with sodium glycolate
`buffer was attributed to a pH shift occuring in the former (19).
`In this case, however, the decrease in pH on freezing was at-
`tributed to crystallization of the monosodium form of succinic
`acid. In addition, it was not immediately clear why a similar
`crystallization effect would not be observed with the use of
`sodium salt of the glycolic acid. In any event, the Na⫹ cation
`can be replaced with the K⫹ cation in inorganic buffers as a first
`approximation to potentially increase the stability of proteins
`during the freezing stage of lyophillization.
`Influence on specific surface areas of lyophillized cakes. An ag-
`gregation mechanism involving partial denaturation at the
`ice–freeze concentrate interface has also been linked to an in-
`crease in protein degradation (27). Denaturation induced by
`this mechanism can be reduced by incorporating surface active
`agents in the formulation (33). Studies have shown the copper
`complexing ability of several zwitterionic N-substituted
`aminosulfonic acid buffers to correlate with surface activity at
`pH 8.0 as measured by alternating current polarography (34).
`
`www.pharmtech.com
`
`Page 2
`
`

`

`Increasing surface activity was correlated to the number of hy-
`droxyl groups; thus, the following order: DIPSO (three hydroxyl
`groups) ⬎ HEPPSO, POPSO (two) ⬎ PIPES (no hydroxy
`groups), showing no surface activity.
`In another study, the increase of Cu⫹2 toxicity observed in
`the marine dinoflagellate A. carterae in the presence of more
`than 10 mM HEPES at pH 8.0 was attributed to the surfactant
`activity of this buffer, as measured using an array of various
`electrochemical techniques (35). Such buffers exhibiting some
`degree of surface activity could be used to potentially inhibit
`freeze-induced damage to proteins that involve the partial un-
`folding of proteins after adsorption to the ice surface.
`Influence on thiol–disulfide interchange. The aggregation of
`lyophilized natriuretic peptide (ANP, pI ⬎ 10) was significantly
`reduced when the concentration of acetic acid buffer at pH 4.0
`was increased from 5 to 15 mM before lyophilization (36). The
`mechanism of aggregation was attributed to alkali induced ␤-
`elimination from the disulfide linkage to form a free thiolate ion.
`The thiolate anion subsequently underwent thiol–disulfide in-
`terchange with ANP to form the disulfide-linked multimers.
`However, it was not apparent why a phase transition of osten-
`sibly incompletely crystallized mannitol after lyophilization
`from a glass to a crystal upon storage would trigger an increase
`of local pH in the lyophilized product (that was attributed to
`the generation of thiolate ions).
`Influence on excipient properties of crystallinity and glass transition.
`␤-galactosidase was lyophilized in a range of sodium phosphate
`buffer concentrations (10–200 mM, pH 7.4) containing varying
`amounts of mannitol (0–500 mM). A larger mannitol concen-
`tration without buffer caused aggregation presumably as a result
`of the complete crystallization of mannitol. The residual activity
`was preserved at buffer–mannitol concentrations at which the
`buffer presumably prevented the crystallization of mannitol (37).
`The glass-transition temperature of lyophilized rhIl-1ra
`containing 1% sucrose, 4% mannitol, and 2% glycine decreased
`from 46 to 26 ⬚C when the buffering agent sodium citrate was
`replaced with sodium phosphate (38). This result was consis-
`tent with the observation that the lyophilized product was
`more stable in citrate than in phosphate buffer containing
`these excipients.
`
`Chaotrope–Kosmotrope effects
`Chaotropic anions are water-structure breakers and destabilize
`proteins. Kosmotropic anions are polar water-structure mak-
`ers and generally stabilize proteins (39–41). A study involving
`aqueous column chromatography on a size-exclusion cross-
`linked dextran gel showed that a chaotrope interacts with the
`first layer of immediately adjacent molecules somewhat less
`strongly than would bulk water in its place and that a polar kos-
`motrope interacts more strongly (42). The ability of anions to
`make or break water structure closely parallels the Hofmeister
`series (43). A continual decay in activity of an immobilized fu-
`sion protein (organophosphorus hydrolase of pI ⫽ 8.3 and
`green fluorescent protein) was observed in reaction mixtures
`containing 1-[N-cyclohexylamino]-ethane sulfonic acid (CHES)
`at pH 6.9. This decay in activity was fully restored, along with
`fluorescence, upon washing with PBS buffer. The researchers
`90 Pharmaceutical Technology MARCH 2004
`
`concluded that the sulfonate was more chaotropic than the
`phosphate anion (44).
`The solution and thermal stability of the tetrameric enzyme,
`phosphoenolpyruvate carboxylase (PEPC, pI ⫽ 6.0) was de-
`termined at pH 6.2 in MES buffer in the presence of various
`salts by temperature-accelerated enzyme activation and by size
`exclusion chromatography (45). Results showed that kos-
`⫺2, F⫺, OAc⫺) and glu-motropic anions (HPO4⫺2, citrate⫺3, SO4
`
`
`tamate stabilized the enzyme most effectively and that Cl⫺ and
`Br⫺ were destabilizing. The effect of cations ranged from rela-
`tively inert (e.g., Na⫹ and K⫹) to destabilizing (e.g., NH4⫹, Li⫹,
`
`(CH3)4N⫹ ). The observed stabilization by specific salts was in-
`terpreted in terms of the positive surface-tension increment
`and the water-structuring effects conferred on the solution by
`these agents. The destabilization by some salts was associated
`with the dissociation of the tetrameric enzyme into its dimeric
`and monomeric forms.
`
`Effect of buffer heat of ionization
`When protein conformation is protonation dependent (i.e., the
`enthalpy of denaturation or the association constant—for bind-
`ing between substrate and ligand—varies with pH), the ob-
`served denaturation or binding enthalpy often varies with the
`kind of buffer used in the study. This variation exists because
`the experimentally measured enthalpy (⌬Hobs) at a given pH is
`determined by two values: the ionization enthalpy of the par-
`ticular buffer used (⌬Hion) and the enthalpy of the denatura-
`tion or binding process corrected for buffer effects (⌬Hb). These
`enthalpies are related to ⌬Hobs by
`
`⌬Hobs
`
`⫽ ⌬Hb
`
`⫹ (n)⌬Hion
`
`in which n is the number of protons released (positive sign) or
`taken up (negative sign) by the buffer during denaturation or
`binding.
`A graph of ⌬Hion against ⌬Hobs at various pH values (46) (see
`Figure 1) can be used to obtain the “true” enthalpy change on
`denaturation or binding, to “deconvolute” the enthalpic and
`entropic components of reactions involving a change in pro-
`tein conformation, and to estimate the pKa values of the ion-
`izable group(s) in the protein involved in the reaction.
`A similar study was undertaken to investigate the possibility
`that the uptake or release of protons was responsible for the
`anomalous heat-capacity change obtained during complexa-
`tion of dihydrolipoyl acetyltransferase to dihydrolipoyl dehy-
`drogenase in the multienzyme complex of Bacillus stearother-
`mophilus (47). The effect of the buffer heat of ionization was
`similarly studied for the acylation of ␣-chymotrypsin (48), the
`binding of NADPH to ␣-ketoglutarate (49), and the hexokinase-
`catalyzed phosphorylation of sugars by ATP (50).
`Studies on the helix-forming thermodynamic propensity scales
`of various amino acid residues indicate that an amino acid residue
`located at a solvent-exposed position of an ␣ helix differently
`affects the stability of the protein (51–53). Such stabilization in
`proteins is very similar to that found in short helical peptides of
`the same amino acid sequence, thereby indicating the funda-
`mental character of the observed thermodynamic propensity
`
`www.pharmtech.com
`
`Page 3
`
`

`

`Figure 1: Dependence of protein unfolding on the heat of ionization of
`buffer. Figure shows enthalpies of binding for the Src SH2 domain
`binding to the hmT peptide at various pH values as a function of ⌬Hion
`of the buffer (adapted from Reference 46).
`
`(54, 55). Where the unfolding of such proteins appears to be
`linked to the protonation of a solvent-exposed amino acid
`residue—such as that of the major cold-shock protein of Es-
`cherichia coli CspA—studies have shown that the magnitude of
`the denaturation temperature is inversely correlated with the
`buffer’s heat of ionization (56) (see Figure 2). The figure shows
`the following buffers and their heats of ionization (kJ/mol): ca-
`codylate (⫺4), phosphate (⫺1), HEPES (⫹3), citrate (⫺11),
`PIPES (⫹12), MOPS (⫹23), and imidazole (⫹36).
`
`The antioxidant effect of buffers
`Some Good’s buffers are efficient scavengers of hydroxyl radicals
`with rate constants of ~109 /M s (57). Tris, tricine, and HEPES
`(in that order) were shown to inhibit the loss of a competitive
`solute, thymine, in radiolyzed water. HEPES and Tris but not phos-
`phate inhibited the rate of auto-oxidation of hemoglobins A (pI
`⫽ 6.9) and S (58). The mechanism was not specifically attributed
`to free radical scavenging but rather by the binding of the phos-
`phate anion to the 2,3-BPG binding site at pH 7.0. This binding
`favored a shift to the deoxy state that was linked to more rapid
`methemoglobin formation. In contrast, HEPES and Tris, being
`positively charged at pH 7.0, were not bound as readily as phos-
`phate to the 2,3-BPG electropositive region in the hemoglobin
`molecule. HEPES and MOPS also accelerated the decomposition
`rate of the oxidant, peroxynitrite (ONOO⫺) (59). The ability of
`ONOO⫺ to stimulate current good manufacturing practice
`(CGMP) formation in cultured endothelial cells in the presence
`of HEPES and MOPS but not phosphate was attributed to the ox-
`idant’s reaction with the buffers to release NO in a Cu(I) catalyzed
`reaction (60). In contrast, the binding of phosphate or phospho-
`rylated compounds to acidic fibroblast growth factor (aFGF) sig-
`nificantly reduced the copper catalyzed oxidation of its free thiol
`groups, thereby reducing aggregation (61).
`The rate of Fe(II) auto-oxidation was substantially larger in
`phosphate and bicarbonate than in HEPES, MOPS, Tris, or MES
`buffers (50 mM, pH 6.5–7.0). Furthermore, the rate of Fe(II)
`92 Pharmaceutical Technology MARCH 2004
`
`Figure 2: CsPA unfolding in various buffers (adapted from Reference 56).
`
`auto-oxidation of Fe(II) chelates with oxygen ligands was higher
`than the auto-oxidation rate of Fe(II) chelates with nitrogen
`ligands (62). Results indicated that phosphate buffer could
`chelate with Fe(II), thereby promoting its oxidation even in the
`absence of free hydroxyl radicals (63). In addition, another study
`was conducted to compare the hydroxyl radical quenching abil-
`ity of phosphate, carbonate, and bicarbonate buffers. Results
`showed that phosphate buffer quenched hydroxyl radicals less
`efficiently than did carbonate or bicarbonate buffers (64).
`The ability of buffers to scavange free radicals assumes in-
`creased importance with the emergence of depot protein for-
`mulations administered by intramuscular or subcutaneous in-
`jection in conjunction with the absence of glycosylation in
`recombinantly produced proteins. Nonglycosylated proteins
`are more prone to denaturation by free radical attack than O-
`linked glycosylated proteins (65–67). However, studies have not
`recognized that other factors such as protection against free-
`radical–mediated denaturation and/or a decrease in the amide
`proton exchange rate may have been partly responsible for the
`sustained activity of recombinant bovine granulocyte colony
`stimulating factor (rbG-CSF, pI ⫽ 6.6), when formulated in or-
`ganic buffers rather than in acetate buffer (68, 69).
`
`Buffer effect on thiol–disulfide interchange reactions
`Proteins administered through a controlled-release system such
`as polymeric matrices containing powdered proteins are ex-
`posed to high water activity (70). The moisture-induced ag-
`gregation of several proteins is caused by intermolecular s–s
`bond formation via the thiol–disulfide interchange reaction
`(31). The aggregation of bovine serum albumin caused by such
`reaction is substantially reduced when the initiating buffer (5
`mM phosphate, 150 mM NaCl, pH ⫽ 7.3) contained 1 M
`sodium phosphate (71). The addition of 4 M NaCl did not cause
`the same level of inhibition as the sodium phosphate. Because
`the thiolate anion (rather than the thiol) is the reactive species
`in the thiol–disulfide interchange, it is possible that the phos-
`phate anion prevents the nucleophilic attack of the thiolate
`anion on the disulfide linkage (72, 73). This effect may be partly
`a result of charge repulsion because in at least another case, in-
`
`www.pharmtech.com
`
`Page 4
`
`

`

`Figure 3: Optimization of charge–charge interactions by Monte Carlo
`analysis (adapted from Reference 74).
`
`Figure 4: Electrostatic interactions of native protein (adapted from
`Reference 74).
`
`creasing the concentration of buffer (acetate) also seemed to
`decrease the extent of thiol–disulfide interchange (36).
`
`Effect of buffer–salt concentration
`An excellent review showed that charge–charge interactions
`were better optimized in the enzymes (E) than in proteins with-
`out enzymatic functions (N), relative to randomly distributed
`charge constellations obtained by the Monte Carlo technique
`(74) (see Figure 3). Proteins belonging to the mixed ␣␤ type
`were electrostatically better optimized than pure ␣-helical or
`␤-strand structures. Proteins stabilized by disulfide bonds
`showed a lower degree of electrostatic optimization. Finally, the
`electrostatic interactions in a native protein were effectively op-
`timized by rejection of the conformers that lead to repulsive
`charge–charge interactions (see Figure 4). The implication of
`this computational analysis is that salt or buffer-mediated elec-
`trostatic or binding effects are likely to be more pronounced in
`enzymes rather than in proteins; in higher evolutionary fold-
`ing classes that use the ␣Ⲑ␤ or the ␣ ⫹ ␤folds rather than in
`pure ␣ or pure ␤ folds; and in proteins that have relatively fewer
`disulfide bonds in their primary structure, all other factors being
`equal (75).
`The larger the difference between the pI and the pH of in-
`terest, the greater the net charge on the protein. This implies
`that the ability of ionic compounds to cause either stabilization
`or destabilization of the protein by binding to specific residues
`(not kosmotropic or chaotropic effects) should increase as the
`difference between pI and pH becomes greater (76). This effect
`is even greater for proteins in which the ionic contributions
`substantially affect protein stability (see Figure 3). Classical pro-
`tein electrostatics dictates that the electrostatic contributions
`to stability should be maximal at the pI and hence the salt de-
`pendence on stability at a given pH also should be determined
`by the distance to the isoionic point, pI. The difference in esti-
`mates of the stability of a protein obtained using either guani-
`dine hydrochloride (a charged denaturant) or urea (an un-
`charged denaturant) also is dependent on the contribution of
`electrostatic interactions to protein stability (77).
`94 Pharmaceutical Technology MARCH 2004
`
`Increasing the buffer (acetate or phosphate) concentration
`from 50 mM to 1 M caused a 3- and 10-fold increase in the ther-
`mal stability of P. amagasakiense glucose oxidase (pI ~4.4) at
`pH 6.0 and 8.0, respectively, and a 100-fold stabilization at pH
`7.0. The thermal stability also was enhanced by 1 M (NH4)2SO4,
`which stabilized the enzyme 100–300 fold at 50 ⬚C and pH 7–8,
`and 2 M potassium formate (KF), which stabilized the enzyme
`as much as 36 fold at 60 ⬚C at pH 6.0. In all instances, the deg-
`lycosylated enzyme was stabilized to a lesser extent than the na-
`tive enzyme (78).
`Another study showed that a similar increase in phosphate
`buffer concentration (from 25 to 70 mM) increased the rate of
`reactivation of Cyanidium caldarium latent nitrate reductase
`when incubated at pH 7.5 and 0 ⬚C (79). This result was pos-
`tulated to occur because of the dissociation of the nitrate
`reductase–inhibitor complex by an increase in the ionic strength
`of the buffer.
`The aggregation rate of an acidic fibroblast growth factor (pI
`⫽ 5–6) decreased as the concentration of phosphate buffer was
`increased at pH 7.4 (80, 81).The extent of stabilization by var-
`ious phosphorylated anionic polymers was a result of the in-
`teraction between the electropositive heparin binding site on
`the protein and the anion and was proportional to the chain
`length of the phosphorylated anionic polymer.
`The concentration of urea needed to denature a photointer-
`mediate of the photoactive yellow protein was greater in citrate
`than in acetate buffer at pH 5.0 (82). The slope m of the plot
`between the free energy of unfolding ⌬Gu and denaturant con-
`centration was lesser in citrate buffer, which suggested that fewer
`denaturant molecules binded to the protein on denaturation
`in citrate than in acetate buffer (83).
`A recombinant Aspergillus fumigatus phytase (pI ⫽ 4.7–5.2)
`demonstrated better thermostability at 65 and 90 ⬚C in acetate
`than in citrate buffer at pH 5.5. In addition, the enzyme had
`a greater heat tolerance in the presence of low concentration
`(10 mM) than in high concentration (200 mM) of either buffer
`at 65 ⬚C. Because the heat stability of the enzyme originates
`from its ability to refold completely into the native-like, fully
`
`www.pharmtech.com
`
`Page 5
`
`

`

`stabilizing in the native protein, and these favorable coulombic
`interactions are reduced at low ionic strength. Above the 200-
`mM salt concentration, the results were consistent with the
`Hofmeister series. Researchers also demonstrated that the ther-
`mostability of Csp increases as the destabilizing effect of salt
`decreases, probably due to a greater favorable optimization of
`salt bridges and hydrogen bonds in the thermophilic as com-
`pared to the mesophilic species (74, 92, 93).
`Similarly, the thermal stability of calf skin collagen type I (pI
`~9) in 50-mM acetic acid (pH ⫽ 3.0) depended on salt con-
`centration (94). At salt concentrations ⬍20 mM, the salts re-
`duced the denaturation temperature. However, between 20 and
`500 mM, they either increased or decreased the denaturation
`temperature in a salt-specific manner that correlated with their
`anion position in the Hofmeister series.
`⫺2 significantly improved
`The orthophosphate anion HPO4
`not only the thermal stability but also the activity of the en-
`doxylanase (pI ⫽ 10.6) at pH 7.0 in 40-mM MOPS buffer. When
`K2HPO4 concentration was increased from 50 mM to 1.2 M,
`the Tm value of xylanase increased from 60.0 ⬚C to 74.5 ⬚C. The
`xylanase activity at 0.6-M K2HPO4 was 2.3-fold higher than that
`at 50-mM K2HPO4 and 120.2-fold higher than that in 40-mM
`MOPS buffer. The K+ cation contributed to the thermal stabi-
`lization until 0.6 M, after which the stabilizing effect of the phos-
`phate anion became dominant at K2HPO4 concentrations
`⬎ 0.6 M (95).
`Dnase (pI ⫽ 3.9–4.3) is a phosphodiesterase capable of hy-
`drolyzing polydeoxyribonucleic acid. Ca⫹2 ion at concentrations
`>10 mM stabilized the enzyme against aggregation at 37 ⬚C when
`formulated at pH 6.3, at which the enzyme is stable to deamida-
`tion (96, 97). Other divalent cations such as Mn⫹2, Mg⫹2, and
`Zn⫹2 did not stabilize the enzyme. The effect of Ca⫹2 was attrib-
`uted to specific binding to the active site and preventing aggra-
`gation by causing a conformational change in the protein (98).
`Phosphate buffer was better than sulfate or imidazole at in-
`hibiting the rate of thermal aggregation and denaturation in ␤-
`lactoglobulin (pI ⫽ 5.13) at pH 6.7 (99). The researchers spec-
`ulated that a lysine and arginine-rich region on the edge of the
`␤ strands A, E, and F could act as a nucleation center for fur-
`ther unfolding of the protein molecule because of a high
`surface-charge density. Because arginine and lysine residues can
`act as sites for phosphate binding, the net charge density is re-
`duced along with the propensity for further unfolding and ag-
`gregation (100). Moreover, the net charge on the protein is neg-
`ative at pH 6.7, and the magnitude of the net coulombic
`repulsion between the anionic buffer and the protein also can
`decrease the propensity for denaturation. Anecdotal evidence
`suggests that the conformational stability of proteins toward
`denaturation increases if anionic buffers are used above the pI
`(and conversely, if cationic buffers are used below the pI). This
`effect is similar to the specific example cited previously and may
`be viewed as being analogous to the “salting out” effect pro-
`duced by kosmotropes.
`Mobility increments of a 20-mer phosphodiester oligonu-
`cleotide were compared for a Tris buffer and various Group I
`monovalent cations. Organic amines such as TRIS and several
`Good’s buffers bind to the DNA not only by means of electro-
`
`www.pharmtech.com
`
`Figure 5: The effect of buffers on the deoxyguanine triphosphate
`(dGTP) incorporation (adapted from References 85 and 86). Increasing
`concentrations of KCl were used with a fixed concentration of each of
`the buffers. For each buffer, dashed lines and solid-colored symbols
`represent those results.
`
`active conformation after heat denaturation, the results sug-
`gested that the refolding was affected by buffer specificity (84).
`The relative effectiveness of various buffers at pH 7.2 for the
`deoxynucleotidyl transferase catalyzed polymerization of the
`deoxynucleoside triphosphates (dATP, dCTP, and dGTP) onto
`an oligonucleotide initiator decreased in the following order:
`cacodylate ⬎ MES ⬎ HEPES ⬎ TRIS ⬎ phosphate (85, 86) (see
`Figure 5). The differences in the effectiveness of the buffers
`could be attributed neither to differences in ionic strength nor
`to differences in the amounts of de-protonated buffer ions. The
`poor effectiveness of the enzyme in a potassium phosphate
`buffer is most likely a result of the phosphate ion functioning
`as a competitive inhibitor for the triphosphates.
`The half life of L-amino acid oxidase (pI ⫽ 4.8) from the
`Gram-positive bacterium Rhodococcus opacus at 37 ⬚C increased
`more than 20 fold by incubating the enzyme in a glycine-NaOH
`⫽ 938 min) compared with the half life when TEA-
`buffer (t1/2
`⫽ 35 min) and a potassium phosphate buffer (t1/2
`⫽
`HCl (t1/2
`44 min) were used (87). The buffer pH was 8.6 for all three
`buffers. The half life of hydroxynitrile lyase (pI ⫽ 4.1) activity
`decreased in the presence of citrate and acetate buffers at pH
`3.75 compared with the half life when phosphate buffer was
`used (88).
`Formulations of LysB28ProB29 human insulin analog (Huma-
`log, pI ⫽ 5.5) comprising TRIS or L-arginine buffer at pH 7.4 re-
`mained stable against aggregation for markedly longer periods
`of time than formulations containing a phosphate buffer (89).
`The stability of the ␣-helical Greek key caspase recruitment
`domain from the CLARP kinase protein at pH 8.0 (pI ⫽ 5.3)
`decreased in the presence of moderate salt concentrations ⬍200
`mM and then exhibited an increase at higher salt concentra-
`tions (90). Similar results were obtained for the cold shock pro-
`tein (Csp) from the thermophilic organism Bacillus caldolyti-
`cus (91). Results suggested that electrostatic interactions are
`96 Pharmaceutical Technology MARCH 2004
`
`Page 6
`
`

`

`static interactions but also by hydrogen bonds primarily to the
`purine or pyrimidine rings (101).
`Remarkable increases in protein stability can be achieved by
`improving the coulombic interactions among charged groups
`on the protein surface. When the hyperexposed Asp49 residue
`of Ribonuclease T1, an acidic protein with a pI value of ⬃3.5,
`was substituted with a histidine, the resulting mutant was 1.1
`kcal/mol more stable at pH 6.0 than the wild-type enzyme (102).
`A buffer molecule that would screen this hyperexposed residue
`could potentially improve enzyme stability. Indeed, results
`showed that the conformational stability of the protein was al-
`most doubled with the addition of 0.2 M Na2HPO4 at pH 7.0
`(103). Tetraprotonated spermine and Mg⫹2 also stabilize Rnase
`T1 by preferential binding to the folded protein (104). As an-
`other example, when two residues of the hexameric glutamate
`dehydrogenase enzyme from th