The effect of net charge on the solubility,
`activity, and stability of ribonuclease Sa
`
`KEVIN L. SHAW,1,4 GERALD R. GRIMSLEY,1 GENNADY I. YAKOVLEV,2
`ALEXANDER A. MAKAROV,2 AND C. NICK PACE1,3
`1Department of Medical Biochemistry and Genetics, Texas A&M University, College Station, Texas 77843, USA
`2Engelhardt Institute of Molecular Biology, Moscow 119991, Russia
`3Department of Biochemistry and Biophysics and The Center for Advanced Biomolecular Research, Texas A&M
`University, College Station, Texas 77843, USA
`(RECEIVED January 28, 2001; FINAL REVISION March 14, 2001; ACCEPTED March 20, 2001)
`
`Abstract
`
`The net charge and isoelectric pH (pI) of a protein depend on the content of ionizable groups and their pK
`values. Ribonuclease Sa (RNase Sa) is an acidic protein with a pI ⳱ 3.5 that contains no Lys residues. By
`replacing Asp and Glu residues on the surface of RNase Sa with Lys residues, we have created a 3K variant
`(D1K, D17K, E41K) with a pI ⳱ 6.4 and a 5K variant (3K + D25K, E74K) with a pI ⳱ 10.2. We show that
`pI values estimated using pK values based on model compound data can be in error by >1 pH unit, and
`suggest how the estimation can be improved. For RNase Sa and the 3K and 5K variants, the solubility,
`activity, and stability have been measured as a function of pH. We find that the pH of minimum solubility
`varies with the pI of the protein, but that the pH of maximum activity and the pH of maximum stability do
`not.
`Keywords: Ribonuclease Sa; isoelectric pH; net charge; electrostatic interactions; protein solubility; en-
`zyme activity; protein stability
`
`The net charge on a protein at any given pH is determined
`by the pK values (pKs) of the ionizable groups (Tanford
`1962). The net charge on a protein is zero at the isoelectric
`point (pI), positive at pHs below the pI, and negative at pHs
`above the pI. Ribonuclease Sa (RNase Sa) has 7 Asp, 5 Glu,
`2 His, 0 Lys, and 5 Arg residues. Consequently, the wild-
`type protein has an excess of acidic residues giving it a
`pI ⳱ 3.5 and a net charge ∼−7 at pH 7 (Hebert et al. 1997).
`We have previously reported studies of the thermodynamics
`of folding and conformational stability of RNase Sa (Pace et
`al. 1998) and several mutants (Hebert et al. 1998; Grimsley
`et al. 1999; Pace et al. 2000). In this study, we have replaced
`Asp and Glu residues with Lys residues singly and in com-
`
`Reprint requests to: C. Nick Pace, Department of Medical Biochemistry
`and Genetics, Texas A&M University, College Station, Texas 77843–
`1114, USA; e-mail: nickpace@tamu.edu; fax: 979-847-9481.
`4Present address: Department of Biology, Grove City College, Grove
`City, Pennsylvania 16127, USA.
`Article and publication are at www.proteinscience.org/cgi/doi/10.1110/
`ps.440101.
`
`bination to produce variants with different pIs. The residues
`selected for replacement, Asp 1, Asp 17, Asp 25, Glu 41,
`and Glu 74, are shown in Figure 1. All of these residues are
`well exposed to solvent and do not form ion pairs or hy-
`drogen bonds. We report here the experimentally deter-
`mined isoelectric points for RNase Sa (0K) and the 2K
`(D17K + E41K), 3K (2K + D1K), 4K (3K + D25K), and 5K
`(4K + E74K) variants, and we compare these values to the
`calculated isoelectric points. The addition of five lysines is
`expected to raise the pI to above 10 and reverse the net
`charge from −7 to +3 at pH 7. The ability to predict the pI
`of a protein is useful for developing methods to purify pro-
`teins by ion exchange chromatography (Janson andRyden
`1989). We also report the dependence of solubility on pH
`for RNase Sa and the 3K and 5K variants. Theory predicts
`that the solubility of a protein will be minimal near the pI
`(Tanford 1961), and we test this in a system where the
`minimum possible changes in structure are used to vary the
`pI. Protein solubility is important in diseases such as Al-
`zheimer’s disease (Kaytor andWarren 1999), in the devel-
`
`1206
`
`Protein Science (2001), 10:1206–1215. Published by Cold Spring Harbor Laboratory Press. Copyright © 2001 The Protein Society
`
`MYLAN INST. EXHIBIT 1099 PAGE 1
`
`MYLAN INST. EXHIBIT 1099 PAGE 1
`
`

`

`Effect of net charge on the properties of RNase Sa
`
`Fig. 1. Ribbon diagram of RNase Sa. The acidic residues changed to lysines are indicated in ball and stick representation. Also given
`is the percent solvent exposure of the side chain and the oxygens in the carboxyl groups estimated by the method of Lee and Richards
`( 1971). The figure was generated with the program MOLSCRIPT (Kraulis 1991). The PDB identifier for RNase Sa is 1RGG (Sevcik
`et al. 1996).
`
`opment of recombinant proteins for treating diseases such as
`fast-acting Lys-Pro insulin (Bakaysa et al. 1996), and to
`many biochemists, especially X-ray crystallographers
`(McPherson 1998) and NMR spectroscopists (Harris 1986).
`We also report the enzyme activity as a function of pH for
`RNase Sa and the 3K and 5K variants. This is of interest
`because the natural substrate for the ribonucleases is nega-
`tively charged and the net charge on the enzyme might
`influence the steady-state enzyme kinetics. Finally, we re-
`port the pH dependence of the thermodynamics of folding
`and the conformational stability of RNase Sa and the 3K
`and 5K variants. The pH dependence of protein stability
`depends both on the net charge of the protein and on the
`difference in pKs of the ionizable groups between the folded
`and unfolded states. Our system may allow us to assess the
`relative importance of these two effects. There is wide-
`spread interest in the pH dependence of protein stability
`(Tanford 1970; Matthew andRichards 1982; Anderson et al.
`1990; Pace et al. 1990; Dao-Pin et al. 1991; Yang andHonig
`1993; Antosiewicz et al. 1994; Barrick et al. 1994; Tan et al.
`
`1995; Schaefer et al. 1998; Warwicker 1999; Pace et al.
`2000; Whitten andGarcia-Moreno 2000).
`
`Results
`
`The measured and calculated pI values for RNase Sa and the
`2K, 3K, 4K, and 5K variants are given in Table 1. The
`difference between the measured and calculated pI values
`will be discussed below.
`The solubilities of RNase Sa and the 3K and 5K variants
`are given in Table 2, and the pH dependences are shown in
`Figure 2. It is clear that the solubility minimum shifts with
`the pI of the protein from near pH 3.5 for RNase Sa to >pH 9
`for the 5K variant. Despite this, the minimum solubility of
`RNase Sa and the 5K variant is about the same, ∼2 mg
`mL−1. In contrast, the 3K variant is less soluble, having a
`minimum solubility of 0.5 mg mL−1 near pH 7.
`The steady-state kinetic parameters KM and kcat for the
`hydrolysis of poly inosinic acid (poly(I)) were determined
`as a function of pH for RNase Sa and the 3K and 5K
`
`www.proteinscience.org
`
`1207
`
`MYLAN INST. EXHIBIT 1099 PAGE 2
`
`MYLAN INST. EXHIBIT 1099 PAGE 2
`
`

`

`Shaw et al.
`
`Table 1. Isoelectric points of RNase Sa and the 2K, 3K, 4K,
`and 5K variants
`
`Sa variant
`
`Experimental pIa
`
`Calculated pIb
`
`Wild type
`2K
`3K
`4K
`5K
`
`3.5
`4.6
`6.4
`8.4
`10.2
`
`4.4 (3.9)
`5.3 (4.8)
`6.6 (6.5)
`8.2 (8.2)
`9.1 (9.1)
`
`Net charge
`at pH 7c
`
`−7
`−3
`−1
`+1
`+3
`
`a pIs of the wild-type and the 2K, 3K, and 4K variants were measured using
`isoelectric focusing gels and are accurate to ≈0.1 pH units. The pI of the 5K
`variant was estimated by determining the pH of immobility using continu-
`ous polyacrylamide gel electrophoresis and buffer systems recommended
`by McLellan (1982). This method is accurate to ≈0.3 pH units.
`b Obtained by calculating the titration curve of the protein using pK values
`for the amino acid side chains and the ␣-COOH and ␣-NH3
`+ termini
`determined from peptide/model compound studies: Asp ⳱ 4.1, Glu ⳱ 4.5
`(Nozaki and Tanford, 1967); His ⳱ 6.6 (McNutt et al. 1990); Tyr ⳱ 9.6,
`Lys ⳱ 10.4, Arg ⳱ 12.5, ␣-COOH ⳱ 3.6, ␣-NH3
`+ ⳱ 7.8 (Tanford 1968).
`The values in parentheses were obtained using the average pKs for the side
`chains of Asp, Glu, and His measured in folded proteins: Asp ⳱ 3.4 ± 1.1,
`Glu ⳱ 4.1 ± 0.8 and His ⳱ 6.5 ± 0.9. The Asp and Glu values are based
`on >200 measured pK values in 24 proteins and they were compiled and
`kindly provided to us by William Forsyth and Andy Robertson at the
`University of Iowa. The His values are based on 58 measured pK values in
`21 proteins and they were compiled and kindly proved by Steve Edgcomb
`and Kip Murphy at the University of Iowa.
`c Calculated using the pK values from the model compound data in the
`footnote above. The net change is rounded to the nearest integer.
`
`variants. The results are given in Table 3. The pH depen-
`dence of log (kcat/KM) is shown in Figure 3. The pH of
`maximum activity is ∼6.5 for RNase Sa and the 3K variant,
`and ∼7.0 for the 5K variant. The 3K variant has somewhat
`less activity than wild-type RNase Sa, whereas the 5K vari-
`ant is significantly less active than both. At the pH of op-
`timum activity, (kcat/KM) is reduced from 1187 (mM sec)−1
`for wild-type RNase Sa to 105 (mM sec)−1 for the 5K vari-
`ant. The decrease in activity for both variants results entirely
`from a decrease in kcat.
`
`Table 2. Solubility as a function of pH for RNase Sa and the
`3K and 5K variantsa
`
`Wild type
`
`3K
`
`pH
`
`2.3
`2.9
`3.6
`4.0
`4.5
`4.8
`5.0
`5.2
`5.4
`
`S (mg mL−1)
`
`11.0
`3.7
`2.2
`2.7
`4.3
`7.6
`9.8
`17.7
`23.3
`
`pH
`
`5.4
`5.7
`5.9
`6.5
`7.3
`7.7
`8.3
`9.3
`
`S (mg mL−1)
`
`4.3
`2.7
`1.7
`0.6
`0.5
`0.6
`0.7
`1.7
`
`5K
`
`S (mg mL−1)
`
`7.8
`8.5
`5.9
`4.6
`4.1
`3.2
`2.3
`5.3
`
`pH
`
`6.7
`6.7
`7.4
`7.9
`8.5
`8.8
`9.3
`10.2
`
`a Measured at 25°C in the presence of 10 mM buffer as described in
`Materials and Methods.
`
`1208
`
`Protein Science, vol. 10
`
`Fig. 2. Solubility of RNase Sa (␱) and the 3K (䊉) and 5K (⌬) variants as
`a function of pH. The lines have no theoretical significance.
`
`Thermal denaturation curves were determined by moni-
`toring the circular dichroism at 234 nm. The curves for the
`variants are similar to those previously shown for wild-type
`RNase Sa (Pace et al. 1998). All curves were analyzed by
`assuming a two-state mechanism. The analysis yields the
`melting temperature, Tm, and the van’t Hoff enthalpy
`change at Tm, ⌬Hm. These parameters and the differences in
`conformational stability, ⌬(⌬G), at pH 7 in 30 mM MOPS
`are given in Table 4. We also show the calculated ⌬⌬G
`values obtained by summing the appropriate ⌬⌬G values of
`the single lysine variants. The good agreement between the
`measured and calculated ⌬⌬G values shows that the muta-
`tional effects are additive.
`The pH dependence of the conformational stability of
`RNase Sa and the 3K and 5K variants was also studied by
`thermal denaturation. Thermal denaturation curves were de-
`termined as a function of pH from pH 2 to pH 10 and
`analyzed by assuming a two-state mechanism as previously
`described (Pace et al. 1998). The thermodynamic param-
`eters are summarized in Table 5. The free energy change for
`folding at any temperature, ⌬G(T), can be calculated using
`the modified Gibbs-Helmholtz equation,
`
`⌬G(T) ⳱ ⌬Hm(1 − T/Tm) +⌬C p[(T − Tm) − Tln(T/Tm)].
`(1)
`
`We use a value of 1.52 kcal mole−1 K−1 for ⌬Cp in
`Equation 1 to calculate ⌬G(25°C), based on our earlier stud-
`ies of the conformational stability of RNase Sa (Pace et al.
`1998). Using the results in Table 5 between pH 2 and pH 5,
`where it is not necessary to correct for the ⌬H for buffer
`ionization, gives ⌬Cp ⳱ 1.31 ± 0.18 kcal mole−1 K−1. This
`lower value of ⌬Cp would increase the ⌬G values by, at
`most, 4%. Changes in amino acid side chains on the exterior
`
`MYLAN INST. EXHIBIT 1099 PAGE 3
`
`MYLAN INST. EXHIBIT 1099 PAGE 3
`
`

`

`Table 3. Kinetic parameters characterizing the cleavage of
`poly(I) as a function of pH for RNase Sa and the 3K and
`5K variantsa
`
`Wild type
`
`3K
`
`5K
`
`pH
`
`4.02
`4.47
`5.05
`5.60
`6.05
`6.50
`7.06
`7.50
`8.01
`8.50
`4.00
`4.50
`5.00
`5.50
`6.00
`6.50
`7.06
`7.48
`8.00
`8.48
`4.47
`5.04
`5.60
`6.05
`6.50
`7.06
`7.50
`8.10
`8.50
`
`kcat(s−1)
`
`KM (mM)
`
`kcat/KM (mM s)−
`
`2.2
`14.2
`59
`131
`175
`178
`146
`95
`42.1
`5.6
`1.5
`7.5
`24
`70
`98
`112
`88
`51
`26
`8
`0.3
`1.2
`4.7
`7.7
`11.1
`10.5
`6.7
`3.5
`1.3
`
`0.25
`0.24
`0.21
`0.18
`0.16
`0.15
`0.19
`0.26
`0.29
`0.34
`0.75
`0.57
`0.33
`0.26
`0.18
`0.14
`0.14
`0.19
`0.20
`0.24
`1.10
`0.64
`0.47
`0.23
`0.13
`0.10
`0.10
`0.11
`0.14
`
`8.8
`59
`281
`728
`1094
`1187
`768
`365
`145
`16
`3.3
`13
`73
`269
`544
`800
`629
`268
`130
`33
`0.3
`1.9
`10
`34
`85
`105
`67
`32
`9.3
`
`a Measured at 25°C in the presence of 0.05 M Tris, 0.1 M potassium
`chloride and 0.05 M sodium acetate as described in Materials and Methods.
`
`of a protein are expected to have only a small effect on ⌬Cp
`because the side chains will be largely exposed to solvent in
`both the folded and unfolded conformations. Therefore we
`decided to use the more accurately determined value of 1.52
`kcal mole−1 K−1. Note that all three proteins are maximally
`stable near pH 5, with stabilities (kcal mole−1) of 7.6 for
`RNase Sa, 5.4 for 3K RNase Sa, and 5.6 for 5K RNase Sa.
`
`Discussion
`
`Isoelectric points of RNase Sa
`and the charge reversal variants
`
`Knowing the pI of a protein is helpful to protein chemists
`developing purification schemes or performing experiments
`in which the protein’s solubility is a crucial factor. (As
`discussed below, proteins tend to be the least soluble near
`their pI.) The pI is usually determined experimentally by
`isoelectric focusing (Righetti et al. 1981); however, one can
`estimate a protein’s pI by calculation if the amino acid
`
`Effect of net charge on the properties of RNase Sa
`
`Fig. 3. pH dependence of log (kcat/KM) for the cleavage of poly(I) by
`RNase Sa (␱) and the 3K (䊉) and 5K (⌬) variants.
`
`composition and the pKs of the ionizable groups are known.
`Usually the pKs of the groups are not known, so values
`obtained from model compound studies are used for the
`calculation. Because these pKs may differ significantly
`from the pKs of groups in the folded protein, calculated pIs
`often disagree with experimentally measured pIs. A number
`of methods have been proposed for the theoretical determi-
`nation of the pIs of proteins (Sillero andRibeiro 1989; Pa-
`trickios andYamasaki 1995). Typically, these methods give
`results that are within ±1 pH unit of the experimental pI.
`In Table 1 we give the experimentally determined pIs for
`RNase Sa and the 2K, 3K, 4K, and 5K variants. The value
`of 3.5 obtained for RNase Sa is in exact agreement with our
`previously reported value (Hebert et al. 1997). Thus, RNase
`Sa is an acidic ribonuclease like RNase T1, which has a pI
`of 3.8 (Hebert et al. 1997). In contrast, barnase, another
`well-studied microbial ribonuclease, is a basic protein with
`a pI of 9.2 (Bastyns et al. 1996). Replacing three acidic
`residues with lysines (3K) raises the pI close to neutrality,
`and replacing five acidic residues with lysines (5K) raises
`the pI above 10. By reversing five charges on RNase Sa, we
`have changed it from one of the most acidic proteins to one
`of the most basic proteins. This is shown in Table 6, where
`we give the experimentally determined pIs and calculated
`pIs for a number of well-studied proteins.
`When the pIs are calculated for RNase Sa and the 3K and
`5K variants using pK values taken from model compound
`studies (Table 1), the pI is overestimated for RNase Sa and
`the 2K and 3K variants and underestimated for the 4K and
`5K variants. When the calculation is performed using the
`average pKs for Asp, Glu, and His measured in folded pro-
`teins, the agreement between measured and calculated im-
`proves for RNase Sa and the 2K and 3K variants. The pKs
`of the ionizable groups of RNase Sa have been measured by
`
`www.proteinscience.org
`
`1209
`
`MYLAN INST. EXHIBIT 1099 PAGE 4
`
`MYLAN INST. EXHIBIT 1099 PAGE 4
`
`

`

`Shaw et al.
`
`Table 4. Parameters characterizing the thermal unfolding of RNase Sa and the
`charge reversal variants at pH 7a
`
`Sa variant
`
`Wild type
`D1K
`D17K
`D25K
`E41K
`E74K
`2K (E41K,D17K)
`3K (E41K,D17K,D1K)
`4K (E41K,D17K,D1K,D25K)
`5K (E41K,D17K,D1K,D25K,E74K)
`
`b
`Tm
`(°C)
`
`47.2
`48.7
`43.3
`50.2
`42.9
`51.1
`39.5
`40.4
`44.4
`46.8
`
`⌬Hm
`c
`(kcal mole−1)
`
`⌬⌬Gd
`(kcal mole−1)
`
`⌬⌬Ge
`(kcal mole−1)
`
`91
`89
`90
`93
`92
`94
`83
`78
`84
`82
`
`—
`0.4
`−1.1
`0.9
`−1.2
`1.1
`−2.2
`−1.9
`−0.8
`−0.1
`
`—
`—
`—
`—
`—
`—
`−2.3
`−1.9
`−1.0
`0.1
`
`a pH 7, 30 mM MOPS.
`b Midpoint of the thermal unfolding curve. The standard deviation is ±0.3°C.
`c Enthalpy change at Tm. The standard deviation is ±5%.
`⳱ ⌬Tm × ⌬Sm (wild type). See Becktel and Schellman (1987) or Pace and Scholtz
`d ⌬⌬Gmeas
`(1997) for a discussion of this method. A positive sign indicates an increase in stability.
`e Calculated by summing the appropriate ⌬⌬G values of the single mutant variants.
`
`NMR (D. Laurents, pers. comm. Similar measurements on
`the 5K variant are in progress.) When these values are used
`in the calculation, the pI ⳱ 3.7, in good agreement with the
`measured value. Similarly, if the pK values measured for
`RNase Sa are used for the 5K variant,
`the calculated
`pI ⳱ 10.1, again in excellent agreement with the measured
`pI. The calculated pI for the 5K variant in Table 1 is low
`because RNase Sa has eight Tyr that are largely buried and
`they all have pKs >11. This is much higher than the
`pK ⳱ 9.6 that is assumed for tyrosine -OH groups on the
`basis of model compound studies. This may be true with
`other proteins because Tyr side chains are 76% buried on
`average in a large sample of folded proteins (Lesser and
`Rose 1990).
`The fact that the pKs of the ionizable groups in model
`compounds may differ significantly from the pKs of the
`same groups in folded proteins is reflected by the calculated
`and measured pI values shown in Table 6. In general, acidic
`proteins have their pIs overestimated and basic proteins
`have their pIs underestimated when calculated using pKs
`from model compound studies. This is expected. For Asp
`side chains, a pK ⳱ 4.1 is observed with uncharged model
`compounds, but a pK ⳱ 3.4 is the average value observed
`in proteins (See the footnotes to Table 1). This results
`mainly because proteins have a net positive charge in the
`region where Asp and Glu carboxyls titrate, and this will
`tend to lower their pKs relative to those measured in un-
`charged model peptides (Antosiewicz et al. 1996). The con-
`verse will be true for basic proteins, where the proteins will
`have a net negative charge above the pI that will tend to
`raise the pKs of the ionizable groups. In addition to net
`charge, other environmental effects can change pKs. For
`example, we have shown that the completely buried and
`hydrogen bonded carboxyl of Asp 76 in RNase T1 has a pK
`
`1210
`
`Protein Science, vol. 10
`
`Table 5. Parameters characterizing the thermal unfolding of
`RNase Sa and the 3K and 5K variants between pH 2 and 10
`
`Sa variant
`
`Wild type
`
`3K
`
`5K
`
`pH
`
`2
`3.3
`5
`7
`8.3
`9
`10
`2
`3.3
`4
`5
`6
`7
`8.3
`9
`10
`2
`3
`4
`5
`6
`7
`8
`9
`10
`
`a
`Tm
`(°C)
`
`27.3
`42
`54.2
`47.2
`39.5
`35.4
`30.1
`27.9
`44.4
`50.8
`49.4
`50.0
`40.7
`31.9
`29.1
`22.3
`25.6
`41.4
`52.4
`53.3
`51.3
`47.3
`44.2
`42.7
`36.4
`
`⌬Hm
`b
`(kcal
`mole−1)
`
`⌬Sm
`c
`(cal K−1
`mole−1)
`
`⌬G (25°C)d
`(kcal
`mole−1)
`
`56
`85
`108
`91
`92
`78
`72
`60
`83
`89
`91
`84
`77
`70
`62
`50
`62
`77
`85
`87
`85
`82
`76
`72
`59
`
`186
`270
`330
`284
`294
`253
`237
`199
`261
`275
`282
`260
`245
`230
`205
`169
`208
`245
`261
`267
`262
`256
`240
`228
`191
`
`0.42
`3.88
`7.59
`5.11
`3.75
`2.36
`1.15
`0.56
`4.15
`5.48
`5.44
`4.99
`3.24
`1.46
`0.80
`−0.48
`0.12
`3.31
`5.35
`5.62
`5.22
`4.50
`3.70
`3.50
`1.85
`
`a Midpoint of the thermal unfolding curve. The standard deviation is
`±0.3°C.
`b Enthalpy change at Tm. The standard deviation is ±5%.
`⳱ ⌬Hm/Tm.
`c ⌬Sm
`d Calculated with Equation 1 using the Tm and ⌬Hm values in this table and
`a value of 1.52 kcal K−1 mole−1 for ⌬Cp (Pace et al. 1998).
`
`MYLAN INST. EXHIBIT 1099 PAGE 5
`
`MYLAN INST. EXHIBIT 1099 PAGE 5
`
`

`

`Table 6. pH of maximum stability, pHmax and isoelectric points
`of some well-studied proteins
`
`Protein
`
`Pepsin
`RNase Sa
`RNase T1
`␤-Lactoglobulin
`RNase Sa2
`RNase Sa (3K)
`RNase Sa3
`Whale myoglobin
`RNase Ba
`Horse cytochrome C
`RNase A
`Chymotrypsinogen A
`Staph nuclease
`RNase Sa (5K)
`Human lysozyme
`T4 lysozyme
`BPTI
`Hen lysozyme
`
`pHmax
`
`—
`∼5.0c,d
`4–5
`—
`∼4.5d
`4–5c
`∼5.5d
`∼6.5g
`5–6h
`∼7.5j
`7–9f
`—
`∼9.0k
`∼5.0c
`∼4.5l
`∼5.5m
`∼5.0o
`5–7p
`
`Experimental pI
`
`Calculated pIa
`
`2.9b
`3.5c,e
`3.8e
`5.1b
`5.3e
`6.4c
`7.2e
`8.0b
`9.2i
`9.4b
`9.6b
`9.6b
`>10.0k
`10.2c
`10.5b
`>10.5n
`10.6b
`11.2b
`
`3.2
`4.4
`4.3
`4.8
`7.4
`6.6
`6.2
`8.4
`9.0
`9.7
`9.3
`9.4
`9.7
`9.1
`10.3
`10.0
`10.2
`10.5
`
`a Calculated using pK values determined from model compound studies as
`described in footnote b of Table 1.
`b From Table 3 of Patrickios and Yamasaki (1995).
`c From this paper.
`d From Pace et al. (1998).
`e From Hebert et al. (1997).
`f From Pace et al. (1990).
`g From Puett (1973).
`h From Pace et al. (1992).
`i From Bastyns et al. (1996).
`j From Walter Englander, personal communication.
`k From Whitten and Garcia-Moreno (2000).
`l From Takano et al. (2000).
`m From Dao-pin et al. (1991).
`n From Becktel and Base (1987).
`o From Makhatadze et al. (1993).
`p From Pfeil and Privalov (1976) and Schaefer et al. (1998).
`
`∼0.5 (Giletto andPace 1999), but that the carboxyl group
`introduced in N44D in RNase T1 has a pK ∼5.7 (Hebert et
`al. 1998). Thus, it is always best to measure rather than
`estimate the pI of a protein unless the measured pKs of the
`ionizable groups are available.
`
`Solubility of the RNase Sa and the 3K and 5K variants
`
`Protein solubility has become increasingly important as it
`has become evident that many diseases result from the in-
`solubility of proteins. For example, a genetic variant of
`␣1-antitrypsin forms inclusion bodies in the liver and this
`can lead to early onset emphysema (Carrell andGooptu
`1998). In addition, the plaques found in patients suffering
`form Alzheimer’s disease are formed by the precipitation of
`A␤
`1–42, a peptide generated by the inappropriate cleavage of
`amyloid precursor protein (Kaytor andWarren 1999). Solu-
`bility can also be important in treating diseases. The devel-
`opment of a fast-acting form of insulin (Lys-Pro insulin)
`
`Effect of net charge on the properties of RNase Sa
`
`was prompted by the fact that formulations of human insulin
`were not solubilized rapidly enough after they were injected
`(Bakaysa et al. 1996). Solubility is also important in both of
`the primary techniques used to determine the structure of
`proteins: X-ray crystallography (McPherson 1998) and
`NMR (Harris, 1986).
`The relationship between protein solubility and pH has a
`long history in protein chemistry (Cohn andEdsall 1943;
`Tanford 1961; Arakawa andTimasheff 1985; Ries-Kautt
`and Ducruix 1997). To a first approximation, the solubility
`of a protein is proportional to the square of the net charge on
`the protein (Tanford 1961). Consequently, proteins are ex-
`pected to be least soluble near their isoelectric points, and
`the solubility should increase as the pH is raised or lowered
`as the magnitude of the net charge increases. Our results
`support this. Figure 3 shows that the minimum solubility of
`RNase Sa and the 3K and 5K variants occurs within a pH
`unit of the pIs and the solubility increases markedly at both
`higher and lower pH.
`
`Enzyme activity of RNase Sa
`and the 3K and 5K variants
`
`Our hope was that the charge reversal mutants would retain
`most of their enzyme activity. Of the five Glu and Asp
`residues that were considered (Fig. 1), only Glu 41 appeared
`to be important to substrate binding or catalysis. Glu 41 in
`RNase Sa is conserved in all of the microbial ribonucleases
`where it forms hydrogen bonds to the guanine base of the
`nucleotide (Sevcik et al. 1990; Hebert et al. 1997). Thus, it
`is important to both specificity and binding. Previous stud-
`ies have shown that the equivalent residues in barnase (Glu
`60) and RNase T1 (Glu 46) were important to the activity.
`The value of kcat/Km for RNA hydrolysis is reduced by 30%
`in the E60Q mutant of barnase (Bastyns et al. 1994) and by
`95% for the E46Q mutant of RNase T1 (Steyaert et al.
`1991). It seemed likely that replacing Glu with Lys would
`be much more detrimental to the enzyme activity than re-
`placing Glu with Gln. Thus, it was surprising that kcat/Km
`for the 3K variant is only reduced by ∼33%, and kcat/Km for
`the 5K variant is only reduced by 90% (Table 3). In both of
`these mutants, Glu 41 has been replaced by Lys. It is even
`more surprising that the decrease in activity is entirely
`caused by a decrease in kcat. Based on the Km value, these
`mutants appear to bind substrate better than wild-type
`RNase Sa. Perhaps, the loss of the hydrogen bond to the
`inosine base in E41K RNase Sa is compensated for by more
`favorable electrostatic interactions, because the net charge is
`∼−6 for RNase Sa, ∼1 for the 3K variant, and ∼+4 for the 5K
`variant near pH 7 (using the measured pKs of all charged
`groups). The activity of the 5K variant is lower that that of
`RNase Sa and the 3K variant over the pH range pH 4– pH
`8.5, but the pH of maximum activity is similar: ∼6.5 for
`RNase Sa and 3K variant, and ∼7 for the 5K variant (Fig. 3).
`
`www.proteinscience.org
`
`1211
`
`MYLAN INST. EXHIBIT 1099 PAGE 6
`
`MYLAN INST. EXHIBIT 1099 PAGE 6
`
`

`

`Shaw et al.
`
`Thus, unlike solubility, the pH dependence of the enzyme
`activity does not vary significantly with the pI of the en-
`zyme. This is not surprising because the pKs of the ioniz-
`able groups on the enzyme actually involved in binding and
`catalysis will influence the pH dependence to a greater ex-
`tent than the net charge on the protein (Fersht 1999).
`
`problem to analyze. Note that even though RNase Sa is a
`very acidic protein (net charge ∼−6 at pH 7) and the 5K
`variant a very basic protein (net charge ∼+4 at pH 7) they
`have the same stability at pH 7. In contrast, the 3K variant
`is about 2 kcal mole−1 less stable than both RNase Sa and
`the 5K variant.
`
`Conformational stability of the single
`charge reversal mutants at pH 7
`
`pH dependence of the conformational stability
`of RNase Sa and the 3K and 5K variants
`
`As shown in Figure 1, the Glu and Asp residues selected for
`charge reversal are highly accessible to solvent and not
`hydrogen bonded. In addition, when Lys residues are sub-
`stituted for the Asp and Glu residues, calculations show that
`+ group of the Lys residues should still be too far
`the -NH3
`from any negative charges to form an ion pair. Thus, the
`main effect of these charge reversals on stability should be
`through a change in the long-range electrostatic interactions
`with the rest of the charged groups in RNase Sa. The actual
`net charge on RNase Sa at pH 7 will be ∼−6 based on the
`measured pKs of all charged groups (Laurents et al. un-
`publ.). Consequently, replacing a negative charge with a
`positive charge is expected to lead to an increase in stability.
`This is supported by calculations using Coulomb’s law and
`a dielectric constant of 80 to sum up the electrostatic inter-
`actions. If we assume that electrostatic interactions are neg-
`ligible in the unfolded states, then the predicted increases in
`stability are (kcal/mole): 2.9 for D1K, 3.2 for D17K, 2.5 for
`D25K, 1.2 for E41K, and 2.3 for E74K. The results in Table
`4 show that, in fact, much smaller increases in stability are
`observed for three of the single variants and that a decrease
`in stability is observed for the other two. We have discussed
`some of these data and related results in two previous papers
`and reached the following conclusions. First, protein stabil-
`ity can sometimes be increased by improving the electro-
`static interactions among charged groups on the surface of a
`protein (Grimsley et al. 1999). Second, the discrepancy be-
`tween the observed and calculated changes in stability leads
`to the surprising suggestion that the free energy of the de-
`natured state may be decreased more than that of the native
`state by these single charge reversal mutations (Pace et al.
`2000). This makes it difficult to predict which charge re-
`versals will be most likely to increase the stability of a
`protein.
`
`Conformational stability of
`2K, 3K, 4K, and 5K at pH 7
`The observed ⌬⌬G values of the multiple mutants can be
`approximated remarkably well by summing the appropriate
`⌬(⌬G) values of the single mutants (Table 4). At first
`glance, this seems surprising. However, at pH 7 each charge
`is interacting with ∼19 other charges in both the folded and
`unfolded states of the protein so this is clearly a difficult
`
`1212
`
`Protein Science, vol. 10
`
`In 1924, a model for the electrostatic interactions of a pro-
`tein was proposed that smeared the net charge over the
`surface of a sphere with a dielectric constant, D, that was
`impenetrable to solvent (Linderstrom-Lang 1924). This
`model predicts that proteins should be most stable near their
`pIs, where their net charge is zero, because unfavorable
`electrostatic interactions resulting from either an excess of
`positive or negative charges will be minimized. Fig. 4
`shows the conformational stabilities estimated from thermal
`denaturation curves, ⌬G(25°C), as a function of pH for
`RNase Sa and the 3K and 5K variants. The maximum sta-
`bility for all three proteins is near pH 5, and clearly does not
`shift with the pI of the protein. At pH 5, RNase Sa and the
`3K and 5K variant have approximate net charges of −4, +2,
`and +5, respectively. Thus, the 3K variant is the least stable
`of the three proteins even though it has the net charge clos-
`est to 0 at this pH. Table 6 gives the pH of maximum
`stability, pHmax, for many of the proteins listed. It is inter-
`esting that the pHmax for eight of the 15 proteins is >2 pH
`units from their experimental pIs. These results show clearly
`that the Linderstrom-Lang model cannot account for the pH
`dependence of the stability of these proteins.
`The more important factor that determines the pH depen-
`dence of protein stability is the difference in the pK values
`of the ionizable groups in the folded and unfolded confor-
`mations of a protein. This subject has been discussed by
`several authors (Tanford 1962, 1970; Matthew and Gurd
`1986; Yang and Honig 1993; Antosiewicz et al. 1994; El-
`cock 1998; Schaefer et al. 1998; Warwicker, 1999). Experi-
`mental studies of the pH dependence of stability have been
`reported for RNase A (Pace et al. 1990), RNase T1 (McNutt
`et al. 1990; Pace et al. 1990; Hu et al. 1992; Yu et al. 1994),
`T4 Lysozyme (Anderson et al. 1990; Dao-Pin et al. 1991),
`barnase (Pace et al. 1992; Makarov et al. 1993; Oliveberg et
`al. 1995), ovomucoid third domain (Swint-Kruse andRob-
`ertson, 1995), chymotrypsin inhibitor 2 (Tan et al. 1995),
`N-terminal domain of L9 (Kuhlman et al. 1999), and staph-
`ylococcal nuclease (Whitten andGarcia-Moreno 2000).
`As shown in Figure 4, the maximum stability for RNase
`Sa and the 3K and 5K variants occurs near pH 5. We will
`first discuss the marked decrease in stability between pH 5
`and 2. An analysis of this region shows that the three pro-
`teins take up ∼2 protons when they unfold. This shows that
`some of the groups in the unfolded protein have higher pKs
`
`MYLAN INST. EXHIBIT 1099 PAGE 7
`
`MYLAN INST. EXHIBIT 1099 PAGE 7
`
`

`

`Effect of net charge on the properties of RNase Sa
`
`than the same groups in the folded protein. The only groups
`that titrate in this pH range are the carboxyl groups. The pKs
`of carboxyl groups in folded proteins are generally consid-
`erably lower than observed in uncharged peptide models.
`The average pKs in folded proteins are less than in peptide
`models by 0.7 for Asp, 0.4 for Glu, and 0.1 for His (See
`footnote b in Table 1). The fact that the effect is larger for
`Asp than Glu reflects the fact that this is caused, in part, by
`the large positive charge that develops on proteins at low pH
`where the carboxyls titrate (Antosiewicz et al. 1996). In the
`unfolded protein, the pKs will be higher because the mol-
`ecule expands and the effective dielectric constant
`in-
`creases. However, it is clear that the carboxyl pKs will still
`be less than in peptide models (Elcock 1998). Only in 6M
`GdnHCl or 8M urea in the presence of 1M salt do the pKs
`in unfolded proteins approach the pKs observed in model
`compounds (Whitten and Garcia-Moreno 2000).
`RNase Sa has 13 carboxyls, and all but three have lower
`pKs than in model compounds (D. Laurents, unpubl.). The
`3K and 5K variants have fewer carboxyls than does RNase
`Sa, but the carboxyls with the lowest pKs have not been
`removed. In addition, the 3K and 5K variants have greater
`positive charges at low pH than RNase Sa, so the pKs of the
`carboxyl groups are expected to be even lower. Conse-
`quently, the pH dependence is quite similar for RNase Sa
`and the 3K and 5K variants at low pH and can be explained
`by the difference in pKs between the groups in the folded
`and unfolded states of the protein.
`Above pH 5, the dependence on pH is less and ∼0.8
`protons are released when RNase Sa or the 3K or 5K vari-
`ants unfold. Again, this is expected based on the pKs that
`we have measured in RNase Sa. Both the ␣-amino group
`and the side chain of His 53 have considerably higher pKs
`in folded RNase Sa than in model compounds and can easily
`account for the pH dependence observed above pH 5. We
`will attempt to account completely for the pH dependence
`of RNase Sa and the 5K variant when we have completed
`measurements of the pKs in the folded and unfolded states
`of these proteins.
`
`Materials and methods
`
`Glycine, diglycine, sodium acetate, MES (2–[N-morpholino]eth-
`anesulf