Stability and function: two constraints in the evolution of
`barstar and other proteins
`Gideon Schreiber, Ashley M Buckle and Alan R Fersht*
`
`Cambridge Centre for Protein Engineering, Medical Research Council Centre, Hills Road, Cambridge CB2 2QH, UK
`
`Background: Barstar is the intracellular inhibitor of
`barnase, an extracellular RNAse of Bacillus amylolique-
`faciens. The dissociation constant of the barnase-barstar
`complex is 10 - 14 M with an association rate constant
`between barnase and barstar of 3.7x10 8 s- 1 M- 1. The
`rapid association arises in part from the clustering of four
`acidic residues (Asp35, Asp39, Glu76 and Glu80) on the
`barnase-binding surface of barstar. The negatively
`charged barnase-binding surface of barstar effectively
`'steers' the inhibitor towards the positively charged active
`site of bamase.
`Results: Mutating any one of the four acidic side chains
`of barstar to an alanine results in an approximately two-
`fold decrease in the association rate constant, while the
`dissociation rate constant increases from five orders of
`
`magnitude for Asp39-*Ala, to no significant change for
`Glu80-*Ala. The stability of barstar is increased by all
`four mutations, the increase ranging from 0.3 kcal mol- 1
`for Asp35-*Ala or Asp39-Ala, to 2.1 kcal mol-1 for
`Glu80-*Ala.
`Conclusions: The evolutionary pressure on barstar for
`rapid binding of barnase is so strong that glutamate is
`preferred over alanine at position 80, even though it does
`not directly interact with bamase in the complex and sig-
`nificantly destabilizes the inhibitor structure. This, and
`other examples from the literature, suggest that proteins
`evolve primarily to optimize their function in vivo, with
`relatively little evolutionary pressure to increase stability
`above a certain threshold, thus allowing greater latitude
`in the evolution of enzyme activity.
`
`Structure 15 October 1994, 2:945-951
`Key words: electrostatic interactions, molecular evolution, molecular recognition, protein engineering, protein stability
`Introduction
`There is a continuous evolutionary pressure for changes
`in protein sequences, and conserved regions are those
`that resist this pressure. The most conserved regions of
`sequence in a family of homologous proteins are usually
`found in the active site, especially the residues directly
`involved in activity. Conversely, residues outside active
`sites are much less conserved, especially those on the
`surface of a protein [1-3]. There are thus many ways of
`achieving the same tertiary structure without affecting
`the stability of a protein, whereas a restricted set of
`residues is required for the optimization of a specific
`activity. This poses a question: what is the compromise
`between the optimization of structural stability and the
`optimization of activity in the evolution of proteins?
`
`positively charged residues in the active site, leading to
`reduced activity [11]. Increased protein activity achieved
`by site-directed mutagenesis was shown, for example,
`for insulin [12] and for human growth hormone binding
`its receptor [13].
`
`Barstar is the intracellular inhibitor of barnase, an extra-
`cellular RNAse of Bacillus amyloliquefaciens. Barstar is
`necessary for survival of barnase-producing cells, since
`intracellular barnase activity is lethal to the organism.
`Barstar consists of a single chain of 89 amino acids, of
`M, 10 211 [14]. The solution structure of free barstar has
`been solved by NMR [15], while the crystal structure of
`a barnase-barstar complex has been solved to high
`resolution [16,17]. Barstar is composed of three parallel
`oL-helices packed against a three-stranded 13-sheet, with a
`short fourth helix serving as a cap. The ionic residues in
`the barnase-binding site of barstar are exclusively nega-
`tively charged, with four acidic residues on the barstar
`surface; Asp35 and Asp39 are located on
`-helix 2 (which
`partly blocks the barnase active site cleft in the complex).
`Asp39 is especially important in the interaction with
`barnase since it makes hydrogen bonds with Arg83,
`Arg87, and the catalytic histidine (residue 102) of
`barnase. Asp39 effectively mimics the reactive phosphate
`group of an RNA substrate [16,17]. Glu76 and Glu80
`are both located on ot-helix 4 of barstar. Glu76 forms a
`salt bridge with Arg59 of barnase. Although Glu80 does
`not directly interact with barnase in the complex, it is
`7 A from the barnase surface and makes indirect, water-
`mediated, hydrogen bonds with barnase (Fig. 1). There is
`
`Both stability and activity of proteins in vitro can be
`increased substantially [4-13]. Comparisons of homolo-
`gous proteins from thermophiles and mesophiles have
`shown that related proteins can perform the same
`function, yet have very different stabilities [4-6]. Protein
`stability can also be increased by site-directed mutagen-
`esis, without affecting the activity of the protein.
`Examples of this include the results of deleting the salt
`bridge between Asp12 and ArgllO in barnase [7], of
`introducing mutations to alanine in a helix of barnase
`and T4 lysozyme [8,9], and the results of the multiple
`mutations which convert barnase to binase, its very
`close homologue from Bacillus intermedius, and which
`increase stability by up to 3.3 kcal mo-1l [10]. Increased
`stability of barnase was also achieved by mutation of
`
`*Corresponding author.
`
`© Current Biology Ltd ISSN 0969-2126
`
`945
`
`MYLAN INST. EXHIBIT 1098 PAGE 1
`
`MYLAN INST. EXHIBIT 1098 PAGE 1
`
`

`

`946 Structure 1994, Vol 2 No 10
`
`Fig. 1. Cross-section through the
`barnase-barstar interface, showing
`some important protein-protein interac-
`tions and the residues mutated in this
`study. Hydrogen bonds are drawn as
`broken lines. This figure was drawn
`with the MOLSCRIPT program [321.
`
`a predominance of positively charged over negatively
`charged residues (Lys27, Arg59, Arg83 and Arg87 versus
`Asp54, Glu60 and Glu73) in the active site of barnase
`[11,17]. The highly electrostatic nature of the
`barnase-barstar interaction results in a very high associ-
`ation rate constant of 3.7x108 S- 1 M-1 , which is about
`100 times faster than usually observed for the association
`of two protein molecules [18,19]. The association rate
`constant (subsequently referred to as the 'on rate') can be
`increased by removing any one of the three acidic
`residues in the barstar-binding site of barnase, whereas
`removal of any basic active-site residues of barnase
`reduces the on rate by up to 10-fold. Since the associ-
`ation rate constant is strongly dependent on salt
`concentration, it is clear that electrostatic forces are very
`important for the association of these two molecules [18].
`
`The aim of this work is to investigate, using site-directed
`mutagenesis, the apparent compromise between barstar
`stability and activity.
`
`Results
`The negatively charged binding surface of barstar has an
`important role in barnase-barstar association
`The barnase-binding surface of barstar comprises four
`negatively charged residues (Asp35, Asp39, Glu76 and
`Glu80), with no positively charged residues on this face
`of the protein structure (Fig. 2). Mutating any one of
`these acidic residues to an alanine decreases the associ-
`ation rate constant by about two-fold, with no significant
`variation among the four mutations (Table 1). In
`contrast, removal of positively charged side chains from
`the active site of barnase results in a decrease in the
`association rate constant by a factor of 2-10 [18]. These
`four acidic residues on the surface of barstar are all
`fully exposed, which might explain the lack of variation
`
`in on rates. However, Arg83 and Arg87 are partially
`buried in a small cleft in the surface of the active site of
`barnase, while Lys27 and Arg59 are fully exposed,
`resulting in the latter having an increased influence on
`the on rate reaction.
`
`Dissociation rate constants for barstar mutations vary by
`up to five orders of magnitude compared with wild-type:
`the Asp39-Ala mutation results in a dissociation rate
`constant of 0.9 s- 1, compared with 3.7x10- 6 s-1 for wild-
`type barstar, corresponding to a change in the free
`energy of binding (AAG) of 7.7 kcal mol-1 (Table 1).
`This large decrease in binding energy can be explained
`by the many salt bridges and hydrogen bonds between
`Asp39 and barnase active-site residues Arg83, Arg87 and
`HislO02 (Fig. 1); the Asp35-*Ala mutation increases the
`dissociation rate constant (off rate) by three orders of
`magnitude to 0.0038 s-1 (AAG=4.5 kcal mol-1). The
`side chain of Asp35 makes hydrogen bonds with the
`barnase Arg59 backbone NH, and also makes van der
`Waals interactions with the arginine side chain; the
`Glu76-*Ala mutation increases the off rate by a factor of
`six to 2.1x10 -5 (AAG=1.4 kcal mol-'). This loss in
`binding energy probably results from the removal of a
`salt bridge between Glu76 in barstar and Arg59 in
`barnase. The mutation Glu80-.Ala does not significantly
`effect the dissociation rate constant, since Glu80 does
`not directly interact with barnase.
`
`The stability of barstar is increased by mutating any one of
`the four acidic residues on the barnase-binding site
`The free energy of unfolding in water (AGHU2 was
`measured as a function of the change in barstar fluor-
`escence upon titration with urea. The entire data set was
`fitted to a two-state unfolding transition curve as de-
`scribed previously [20]. Mutating any one of the four
`acidic residues in the barnase-binding site of barstar
`
`MYLAN INST. EXHIBIT 1098 PAGE 2
`
`MYLAN INST. EXHIBIT 1098 PAGE 2
`
`

`

`Stability and function in the evolution of barstar Schreiber et al. 947
`
`Fig. 2. Representations of the structure
`of barstar, showing the barnase-binding
`surface and locations of the residues
`mutated in this study. Left; molecular
`surface of barstar colour coded
`according to electrostatic potential (cal-
`culated by GRASP [33]). Positively
`charged regions are coloured blue, neg-
`atively charged regions red. Right;
`backbone of barstar, drawn in the same
`orientation.
`
`Table 1. Association (k1) and dissociation (k_ l) rate constants of barnase
`with wild-type and mutant barstar at pH 8. a
`
`Table 2. Changes in the free energies of unfolding of wild-type barstar
`and mutant proteins.a
`
`Barstar
`
`k x 10-8 k_ 1 x 106 Ki(pM)
`(s- 1 M- 1 )
`(5-1)
`
`AGb
`(kcal mol-1)
`
`AAG c
`(kcal mol- 1)
`
`Barstar
`
`Lurea]50 /
`(M)
`
`AG H2 0
`(kcal mol-1)b
`
`A AG H2 0
`(kcal mol-1)c
`
`Wild-type
`Asp35-*Ala
`Asp39-Ala
`Clu76-Ala
`Clu80-.Ala
`
`3.7
`1.9
`1.9
`2.0
`2.0
`
`3.7
`3800
`900000
`21
`5.2
`
`0.01
`20
`4100
`0.1
`0.025
`
`-19.0
`-14.5
`-11.3
`-17.6
`18.5
`
`-4.5
`- 7.7
`-1.4
`-0.5
`
`Wild type
`Asp35-Ala
`Asp39-*Ala
`Glu76-Ala
`Glu8O-.Ala
`
`4.19
`4.45
`4.42
`4.84
`5.90
`
`- 5.28
`- 5.60
`- 5.57
`- 6.09
`- 7.42
`
`0
`0.3
`0.3
`0.8
`2.1
`
`aAll rate constants were measured in 50mM Tris-HCI buffer at 25 C. Ki was
`calculated from the equation Ki = kl/k_ 1. bThe free energy of binding was
`calculated from: AG = -RTInK i. CThe difference in free energy of binding
`wild-type and mutant proteins AAG = AGwt - AGutant Standard errors for
`association and dissociation rate constants are 15 %, which results in an
`error in AG of + 0.11 kcal mol - 1.
`
`results in an increase in the stability of the folded protein
`(Table 2). The largest increases in stability were found for
`mutations Glu76-Ala (0.8 kcal mol-1) and Glu80-Ala
`(2.1 kcal mol-1). Deletion of the Glu80 side chain
`increases the total stability of the protein by about 40%. In
`contrast, mutation of either Asp35 or Asp39 increases the
`overall stability by only 0.3 kcal mol-1. None of the
`mutations affect the overall structure of the protein, as
`monitored by far UV circular dichroism (data not shown).
`
`The increased stability of barstar mutants has an
`electrostatic component
`The addition of 300 mM NaCl increases the stability of
`wild-type barstar by 0.6 kcal mol- 1 (Table 3). We
`interpret this as resulting from increased electrostatic
`screening between the four clustered acidic side chains.
`Under these conditions, the increase in the free energy
`of unfolding for the mutant proteins, relative to wild-
`type, is 0.05 kcal mol- 1 for Asp35-Ala, 0.1 kcal mol-1
`for Asp39-Ala, 0.6 kcal mol- l for Glu76-Ala and
`1.4 kcal mol- 1 for Glu80-,Ala. A comparison between
`the change in free energy of unfolding for the different
`mutant proteins relative to wild-type, with and without
`the addition of NaCl, shows that the greater the contri-
`bution the mutation makes to the stability of the protein
`in the absence of NaCl, the greater is the dependence of
`
`aThe free energy of unfolding was determined by fluorescence changes on de-
`naturation with urea at 25°C, 50 mM Tris-HCI pH 8, 10 mM DTT [20]. bThe free
`energy of unfolding in water was determined by multiplying the average value
`of m (1.26 kcal mol 2) by [ureal 50o. CThe difference in free energy of unfold-
`ing wild-type and mutant proteins AGUF = Gwt-AGmutant. Standard
`errors for [urea]50% are of a magnitude of i 0.06 M urea, which corresponds
`to a free energy (AG) of f 0.075 kcal mol- 1.
`
`stability on the ionic strength of the solution. This is
`further evidence to support the argument that electro-
`static repulsion makes a major contribution to the
`inherent instability of the wild-type structure. The
`Glu80-Ala mutation is an extreme case, where the
`addition of 300 mM salt actually destabilizes the
`structure by 0.1 kcal mol-1.
`
`Increasing the ionic strength of the solution from
`300 mM to 700 mM NaCl increases the stability of
`wild-type barstar by an additional 1.15 kcal mol-1. The
`increased salt concentration does not significantly alter
`the relative free energy of mutant versus wild-type
`unfolding (AAGH2
`for the mutants Asp35-Ala,
`Asp39-Ala, and Glu76-Ala. However, the mutation
`Glu80-Ala causes a further decrease of 0.3 kcal mol- in
`the free energy of unfolding, relative to wild-type.
`
`Discussion
`We have investigated the compromise between structural
`stability and activity in the evolution of barstar. In any
`such study, the requirements of the protein in vivo must
`be borne in mind. An increased activity in vitro by itself
`does not necessarily mean that the activity in vivo will be
`
`MYLAN INST. EXHIBIT 1098 PAGE 3
`
`MYLAN INST. EXHIBIT 1098 PAGE 3
`
`

`

`948 Structure 1994, Vol 2 No 10
`
`Table 3. Changes in the free energies of unfolding of wild-type barstar and mutant proteins in the presence of 300 mM and 700 mM NaCl.a
`
`300 mM NaCI
`
`700 mM NaCI
`
`Barstar
`
`AAGo300
`AAG H2 0
`[urea] 50 /o AG H20
`U-F
`-F
`(kcal mol - 1) (kcal mol- 1) mM NaClb
`(M)
`
`[urea]50%
`(M)
`
`AAG H20
`AG H2 0
`F
`U-F
`U-
`(kcal mol- 1) (kcal mol- 1)
`
`AAGo700
`mM NaClb
`
`Wild-type
`Asp35-Ala
`Asp39-Ala
`Glu76--Ala
`Glu80-*Ala
`
`4.67
`4.71
`4.77
`5.11
`5.8
`
`- 5.88
`- 5.93
`- 6.01
`- 6.44
`- 7.30
`
`0
`0.1
`0.1
`0.6
`1.4
`
`0.6
`0.3
`0.4
`0.4
`- 0.1
`
`5.58
`5.63
`5.76
`5.99
`6.45
`
`-7.03
`-7.09
`-7.26
`-7.55
`-8.13
`
`0
`0.1
`0.2
`0.5
`1.1
`
`1.8
`1.5
`1.7
`1.5
`0.7
`
`aThe free energy of unfolding was determined by fluorescence changes on denaturation with urea at 25°C, 50 mM Tris-HCI pH 8, 10 mM DTT + 300 mM
`or 700 mM NaCI as for Table 2. bThe difference in free energy of unfolding in the presence of NaCI.
`
`increased as well, and, if it does, such an increase might
`not be favourable for the system as a whole. The
`problem is simplified when studying a protein with a
`specific function that can be increased without affecting
`other cellular functions. The only known role of barstar
`is to be the intracellular inhibitor of barnase. Further,
`when analyzing the effects of single mutations in a
`protein, we are just looking at the fine-tuning of the
`protein to its environment and role.
`
`Barstar is necessary for survival of barnase-producing
`cells, since intracellular barnase activity would be lethal
`to the organism. As a consequence, barstar is required to
`bind rapidly and tightly to any intracellular, active
`barnase molecules. The strongly negatively charged
`binding surface of barstar, effectively 'guides' the protein
`during its association with the positively charged barnase
`active site [17,18]. A mutation of any one of these acidic
`side chains of barstar decreases the association rate
`constant by a factor of about two. We have shown that
`barstar achieves this very fast binding at the expense of
`its stability. The free energy of folding of reduced wild-
`type barstar is around 5 kcal mol-1, which is at the lower
`end of the range of stabilities measured for other small
`globular proteins [21]. Clusters of like-charged residues
`on the surface of proteins have been shown previously
`to have a destabilizing effect on protein stability [22].
`For example, the repulsion energy between two aspartic
`acid side chains in a-helix1 of barnase (Asp8 and Aspl2)
`was estimated to be 0.3 kcal mol-1 [7]. Mutations of
`Lys27 and Arg59 in the positively charged barnase active
`site increases the stability of the protein by 0.36 kcal
`mo1' and 0.64 kcal mol-1, respectively [11]. A similar
`repulsion energy was measured for a charged His-Lys
`pair in subtilisin [23]. Barstar residues Asp35 and Asp39
`are located on the same helix, and their side chain
`carboxyl groups, which are well defined, are separated
`by about 6 A in the NMR solution structure of the
`reduced, free wild-type protein [15]. Mutating Asp35 or
`Asp39 to alanine increases the free energy of unfolding
`by about 0.3 kcal mol- 1. The addition of salt diminishes
`most of this increase in stability, suggesting some elec-
`trostatic repulsion between the acidic side chains of
`these residues.
`
`Glu76 and Glu80 are located on a-helix 4 of barstar. The
`increase in free energy of unfolding upon mutating these
`two residues to alanine was found to be 0.8 kcal mol-1
`and 2.1 kcal mol-1, respectively. The dependence of
`AAG H2? on salt concentration shows that there is also
`some electrostatic repulsion between these two residues.
`Since the presence of alanine mutations at residues 76
`and 80 also causes increases in stability of 0.5 kcal mol-'
`and 1.1 kcal mol 1, respectively, even in the presence of
`700 mM NaCl, we cannot exclude the possibility that
`some factor other than electrostatic repulsion might also
`be contributing to the inherent instability in this region
`of the wild-type structure, for example the difference in
`helix-forming tendencies between alanine, glutamic acid
`and aspartic acid [24]. The intracellular ionic strength in
`bacteria depends on the osmolarity'of the growth
`medium, and was found to be at around 200 mM for
`osmolarities up to 0.2 osM in Escherichia coli cells [25].
`The instability due to glutamic acid at positions 76 and
`80 is still very significant at this ionic strength.
`
`The presence of a glutamic acid residue at position 80,
`rather than an alanine, shows to what extent in evolution
`stability will be sacrificed for activity. A glutamic acid at
`this position has no substantial effect on the dissociation
`rate constant of the barnase-barstar complex, and
`increases the association rate constant by a factor of only
`two. Yet, the protein is destabilized by 2.1 kcal mol-1 in
`the presence of this negatively charged residue, relative to
`an alanine in the mutant. To our knowledge, the increase
`in stability due to this mutation is larger than any other
`in the literature. Smaller values have been found for
`single mutants in barnase, T4 lysozyme and insulin
`[9,10,12]. Our results suggest that rapid, and not just
`tight, binding are the predominant factors in the
`evolution of barstar.
`
`Is there a structural basis for the measured differences in
`AA G uH2of unfolding for the four barstar mutations? We
`find that mutations at positions 35 and 39 stabilize the
`protein by only 0.3 kcal mol-, yet mutations at positions
`76 and 80 increase stability by 0.8 kcal mol-1 and
`2.1 kcal mol-1, respectively. There are two plausible
`explanations for these differences; . Firstly, since Glu80 is
`
`MYLAN INST. EXHIBIT 1098 PAGE 4
`
`MYLAN INST. EXHIBIT 1098 PAGE 4
`
`

`

`Stability and function in the evolution of barstar Schreiber et al. 949
`
`near to the carboxyl termini of helix 2 and helix 4, there
`could be an unfavourable electrostatic interaction
`between the side chain and the helix macro-dipoles.
`Fig. 2 shows that Glu80 could be sterically restricted by
`the repulsion between this side chain and the negative
`charges at positions 35, 39, 76, and the carboxyl termini
`of o-helices 2 and 4. This could explain why the Glu80
`side chain is well defined in the NMR structure (with a
`similar orientation as in the crystal structure of the
`complex). It is interesting to note that the side chains of
`Asp35 and Glu76 are also well defined in the NMR
`solution structure, and show similar orientations in the
`free and complexed structures [15,17] (the Asp39 side
`chain is not well defined in the NMR structure). This
`pre-ordering of side chain conformations will be entrop-
`ically unfavourable, and will result in destabilization of
`the protein. On the other hand, it will contribute to the
`stability of the complex, since the entropic loss upon
`formation of the complex will be lowered. Secondly, the
`Glu80 mutation could remove a favourable interaction in
`the unfolded state, thereby stabilizing the folded
`structure. This is unlikely, however, since there is suf-
`ficient evidence from NMR studies that barstar has little
`residual unfolded structure (MJ Lubienski and
`M Bycroft, personal communication).
`
`Analysis of activity versus stability in the barnase active
`site, using site-directed mutagenesis, gave similar results
`to those of barstar, although stability is compromised to a
`lesser extent and activity losses due to mutations are
`much greater than found in barstar [11]. Also, the
`example of barnase is more complicated since barnase
`has evolved to bind both barstar and RNA, and so any
`compromise between structure and activity will reflect
`this. We can also ask the question: if the active site of
`barnase has small areas of negative charge (Asp54, Glu73,
`and Glu60 -
`the last two are important in the
`barnase-RNA substrate interaction [26]), why does
`barstar not have any complementary positively charged
`regions on its binding surface? One explanation is that
`the many electrostatic interactions between barnase and
`barstar result in a complex of sufficiently high stability,
`and that in barstar evolution, there existed further
`pressure for rapid binding (rather than only tight
`binding). This was achieved by simply increasing the
`negative charge on the binding site.
`
`Are our findings general? When increased stability is of
`major importance, as in thermophilic bacteria, we see
`that homologous enzymes can have very different stabil-
`ities. In the mapping of stability versus activity of
`bacterial tyrosyl-tRNA synthetase from E. coli versus
`Bacillus stearothermophilus it was possible to improve sig-
`nificantly the activity of the thermophilic enzyme by
`constructing chimeric proteins with the E. coli enzyme.
`But this was achieved at the expense of stability, which,
`in this case, is crucial in the evolution of the enzyme [6].
`However, when increased stability over a certain
`threshold of about 5 kcal mol-1 to 15 kcal mol- l
`is not
`of major importance, it is relatively easy to improve the
`
`stability by mutagenesis without affecting activity [9,10].
`The best example so far to illustrate this point is in the
`systematic stepwise evolution of barnase to its very
`closely related homologue, binase. A chimeric protein
`containing all the stabilizing changes has a similar activity
`to both parents, but has a significantly higher stability
`than either [10].
`
`Biological implications
`During evolution, every protein is subjected to a
`constant pressure for change, resulting from the
`insertion of random mutations during replication.
`Despite this, we find a strong tendency for con-
`serving residues related to activity, but not to
`stability. An implication of this is that activity is
`highly specified, while there are many possible
`ways to achieve sufficient stability. In other words,
`a protein structure has a much higher tolerance
`for changing structural amino acids than for
`changing functional ones. In recent years, at least
`four cases where stability can be greatly increased
`with no compromise in activity have been ident-
`ified using site-directed mutagenesis. These
`suggest that there are many ways to achieve suf-
`ficient stability for a protein with a given
`structure. The problems frequently encountered
`in engineering an increased stability for a
`protein probably stem, therefore, from our lack of
`understanding of the factors that govern protein
`stability.
`
`Here we report the results of mutating four
`residues of the nuclease inhibitor barstar that are
`important in the formation of a complex between
`barstar and its target nuclease, barnase. These
`mutations increase the stability of barstar but
`decrease the reaction rates of complex formation
`and dissociation. We conclude that barstar has
`compromised its stability to a large extent to
`optimize its specific activity in vivo. In evol-
`utionary terms, the latitude that a protein has
`with respect to stability will result in an ability to
`change its sequence readily, allowing the opti-
`mization of enzyme activity. This will also give a
`protein the freedom needed for adjusting itself to
`the constant changes in the evolutionary process.
`
`Materials and methods
`Protein expression and purification
`Site-directed mutagenesis of barstar was performed by the
`method of Sayers et al. [27]. The oligonucleotides used to
`introduce an alanine codon were: for Asp35; 5'-CCA TAA
`AGC GGC CAG GTT TTC-3', for Asp39; 5'-GGT CAG
`ACA AGC CCA TAA AGC-3', for Glu76; 5'-GCT TTC
`GCT GCA CGG AAA-3', and for Glu80; 5'-GTC GCA
`GCC TGC CGC TTT CGC-3'. Mutant plasmids were ident-
`ified by direct sequencing. Barstar and barnase expression and
`purification have been described [22,28]. Barstar has two
`
`MYLAN INST. EXHIBIT 1098 PAGE 5
`
`MYLAN INST. EXHIBIT 1098 PAGE 5
`
`

`

`both measurements. Barnase-barstar association was measured
`under second order conditions, the concentrations of both
`proteins being equal. A concentration of 0.25 pM was used for
`most experiments (where virtually all of the protein is bound
`within the first 0.5 s of the reaction). The association is effec-
`tively irreversible at these concentrations (except for the barstar
`mutant Asp39-Ala, where a range of concentrations was
`checked to confirm the result): i.e. [E]+[I] - [E.I], where [E] is
`the barnase concentration,
`[I] the barstar concentration and
`[E.I] the complex concentration. The analytical solution [30]
`for second order conditions, where [E]o=[I]o, is:
`
`l/([E]o-[I.E])-l/[E]o=klt
`
`(1)
`
`Only the first 80% of the reaction amplitude was used for
`fitting the data to a second order equation because small errors
`in measuring the initial concentrations of barnase or barstar
`will result in transforming the second order reaction to a
`pseudo-first order one, once most of the protein is complexed.
`Dissociation rate constants ('off rates') were measured by first
`allowing the barnase-barstar complex to form, then 'chasing'
`off the mutant barstar with 5-10 times the concentration of
`wild-type barstar. A [3H]bamase-barstar complex was used for
`complexes with dissociation rate constants of 5x10-3 s-
`or
`more. In this case, [3H]barnase was 'chased' off with unlab-
`elled barnase. Complexed and non-complexed [3H]barnase
`were separated using an anion exchange column (DE52). The
`[3H]bamase-barstar complex remains bound to the column at
`NaCl concentrations below 50 mM, allowing the elution of
`[3H]barnase. The procedure used is very similar to that
`described previously [18]. Dissociation rate constants greater
`then 5x10-3 s- 1 were measured with a stopped-flow fluori-
`meter, using wild-type barstar to chase off mutant barstar from
`the barnase-barstar complex. This method takes advantage of
`the differences in behaviour of fluorescence between wild-
`type and mutant complexes, relative to the free form, as
`described previously [18], and the different fluorescence of
`reduced and oxidized wild-type barstar on excitation at 230
`nm and 280 nm. Both 'chasing' procedures were used for
`complexes having 'off rates' of around 510-3 s- 1, and gave
`results that agree within experimental error. Dissociation rate
`constants were calculated by fitting the results to a single
`exponential equation.
`
`Equilibrium denaturation of barstar in urea was performed in
`50 mM Tris-HC,
`10 mM DTT, from 0-8.0 M urea as
`described [28]. For fitting the data to a two-state unfolding
`transition curve, an average m value derived from 20 individual
`barstar unfolding curves (wild-type as well as mutant curves),
`of m=1.26 kcal mo- 2 was used. The free energy of unfolding
`was calculated from (AG=[urea] 50%xm)
`[31].
`
`References
`1. Mark6ta, J.J.M. & Sternberg, M.J.E. (1988). Analysis and prediction
`of the location of catalytic residues in enzymes. Protein Eng. 2,
`127-138.
`2. Chothia, C. & Lesk, A.M. (1986). The relation between the diver-
`gence of sequence and structure in proteins. EMBOJ. 5, 823-826.
`3. Schulz, G.E. & Schirmer, R.H. (1979). Principles of Protein
`Structure. Springer-Verlag.
`4. Argos, P., Rossmann, M.G., Grau, U.M., Zuber, H., Frank, G. &
`Tratschin, J.G. (1979). Thermal stability and protein structure.
`Biochemistry 18, 5698-5703.
`5. Imanaka, T., Shibazaki, M. & Takagi, M. (1986). A new way of
`enhancing the thermostability of proteases. Nature 324, 695-697.
`6. Guez-lvanier, V., Hermann, M., Baldwin, D. & Bedouelle, H.
`(1993). Mapping the stability determinants of bacterial tyrosyl
`
`950 Structure 1994, Vol 2 No 10
`
`cysteine residues (positions 40 and 82) which can form a disul-
`phide bridge [14]. The double Cys-Ala mutant protein has a
`reduced affinity for barnase, with a free energy of binding
`1.5 kcal mol-1 less than wild-type barstar ([29], and
`G Schreiber and AR Fersht, unpublished data). We have
`chosen therefore to study the barnase-barstar interaction using
`the reduced form of wild-type barstar. Barstar was unfolded in
`urea in the presence of 10 mM dithiothreitol for 30 min,
`refolded and extensively dialyzed before use. Interestingly,
`despite both the reduced and the oxidized forms being stable
`and active, the two cysteine residues cannot be reduced or
`oxidized readily without prior unfolding of the protein, even
`in the presence of high concentrations of reducing or oxidizing
`agents. We attribute this to the significant conformational
`movements that would be required before a disulphide bridge
`could form [17].
`
`Biochemical measurements
`There are three tryptophan residues in the vicinity of the
`barnase-barstar interface: Trp38 and Trp44 of barstar, and
`Trp35 of barnase. Since tryptophan emission is quenched in a
`polar environment relative to a non-polar one, it is possible to
`measure a change in the overall emission of the two proteins
`upon association. The association and dissociation rate
`constants for the barnase-barstar complexes (wild-type and
`mutants) were measured in 50 mM Tris-HCI buffer pH 8 at
`25°C, as described previously [18]. The association of oxidized
`barstar with barnase results in a decrease in the fluorescence
`intensity when excited at 280 nm. When reduced wild-type
`barstar was used, no change in fluorescence intensity was
`recorded under these conditions. Interestingly, an increase in
`the fluorescence intensity was recorded upon excitation at 230
`nm (Fig. 3). The biophysical basis of this phenomena will be
`discussed elsewhere (G Schreiber, R Golbik and AR Fersht,
`unpublished data). For barstar mutants where a change in fluo-
`rescence intensity could be measured on excitation at 230 nm
`as well as at 280 nm, similar rate constants were calculated for
`
`A AO
`
`().()8
`U.Ue
`
`0.06
`
`a)o 0.04
`)
`
`CG
`
`-) 0.02
`0.00
`
`0)
`so
`
`
`02
`
`-0.02
`
`-v SA
`
`0
`
`0.02
`
`0.04
`
`0.08
`
`0.1
`
`0.1
`
`0.06
`Time (s)
`
`Fig. 3. Time course for formation of the barnase-barstar complex
`using reduced or oxidized barstar, on excitation at 230 nm and
`280 nm. Solutions of the two proteins (0.25 1±M each) in 50 mM
`Tris buffer pH 8 were mixed in a stopped-flow fluorimeter, and
`the change with time of the emission fluorescence was recorded
`using a 315 nm cut-off filter. The data were fitted to a second-
`order equation (see equation 1 in Materials and methods).
`x=oxidised barstar excited at 280 nm; filled circles=oxidized
`barstar excited at 230 nm; open circles=reduced barstar excited
`at 280 nm; triangles=reduced barstar excited at 230 nm.
`
`MYLAN INST. EXHIBIT 1098 PAGE 6
`
`MYLAN INST. EXHIBIT 1098 PAGE 6
`
`

`

`Stability and function in the evolution of barstar Schreiber et al. 951
`
`transfer RNA synthetases by an experimental evolutionary approach.
`J. Mol. Biol. 234, 209-221.
`7. Horovitz, A., Serrano, L., Avron, B., Bycroft, M. & Fersht, A.R.
`(1990). Strength and cooperativity of contributions of surface salt
`bridges to protein stability. J. Mol. Biol. 216, 1031-1044.
`8. Serrano, L., Neira, J.L., Sancho, J. & Fersht, A.R. (1992). Effect