Báo cáo khoa học: Creation of a new eye lens crystallin (Gambeta) through structure-guided mutagenic grafting of the surface of bB2 crystallin onto the hydrophobic core of cB crystallin pot

This new protein, Gambeta, consists of 61 residues pos-sessing the same identity at structurally equivalent positions in bB2- and cB crystallin, 91 surface residues unique to bB2 crystal

structure-guided mutagenic grafting of the surface of bB2 crystallin onto the hydrophobic core of cB crystallin

Divya Kapoor1, Balvinder Singh2, Karthikeyan Subramanian1and Purnananda Guptasarma1

1 Division of Protein Science & Engineering, Institute of Microbial Technology, Chandigarh 160 036, Council of Scientific & Industrial Research, New Delhi, India

2 Division of Bioinformatics, Institute of Microbial Technology, Chandigarh 160 036, Council of Scientific & Industrial Research, New Delhi, India

We have developed a novel protein engineering

tech-nique that we hope will facilitate the rational dissecting

out and independent re-assembly of the various

struc-tural features and residue-packing schemes used in

nat-ure to build the interiors and surfaces of various

structurally homologous b sheet-based proteins

Recently, we provided a ‘proof-of-principle’

demon-stration of this technique [1], which we call ‘protein

surface grafting’, by using it to notionally segregate

and re-assort the structural stability features of one

b sheet-based thermophile enzyme (a Cel12A cellulase)

with the functional features of a structurally related mesophile enzyme (another Cel12A cellulase), to produce a variant enzyme bearing a still-functioning, transplanted active surface derived from the mesophile enzyme, but resembling the thermophile enzyme in most other respects [1] The successful creation of such

a meso-active, thermo-stable enzyme encouraged us to explore the workability of our surface grafting approach further, to extend it from grafting of ‘active surfaces’ to grafting of ‘whole-protein surfaces’ or

‘whole-protein interiors’

Keywords

beta sheet remodeling; lens structural

proteins; protein engineering; protein folding

and stability; protein surface grafting

Correspondence

P Guptasarma, Division of Protein Science

& Engineering, Institute of Microbial

Technology, Chandigarh 160 036, Council of

Scientific & Industrial Research, New Delhi,

India

Fax: +91 172 2690585

Tel: +91 172 2636680, ext 3301

E-mail: pg@imtech.res.in

(Received 14 February 2009, revised 2 April

2009, accepted 14 April 2009)

doi:10.1111/j.1742-4658.2009.07059.x

The degree of conservation of three-dimensional folds in protein superfami-lies is greater than that of amino acid sequences Therefore, very different groups of residues (and schemes of residue packing) can be found displayed upon similar structural scaffolds We have previously demonstrated the workability of a protein engineering-based method for rational mixing of the interior features of an all-beta enzyme with the substrate-binding and catalytic (surface) features of another enzyme whose sequence is not similar but which is structurally homologous to the first enzyme Here, we extend this method to whole-protein surfaces and interiors We show how two all-beta Greek key proteins, bB2 crystallin and cB crystallin, can be recombined

to produce a new protein through rational transplantation of the entire sur-face of bB2 crystallin upon the structure of cB crystallin, without altering the latter’s interior This new protein, Gambeta, consists of 61 residues pos-sessing the same identity at structurally equivalent positions in bB2- and cB crystallin, 91 surface residues unique to bB2 crystallin, and 27 interior residues unique to cB crystallin Gambeta displays a mixture of the struc-tural⁄ biochemical characteristics, surface features and colligative properties

of its progenitor crystallins It also displays optical properties common to both progenitor crystallins (i.e retention of transparency at high concentra-tions, as well as high refractivity) The folding of a protein with such a

‘patchwork’ residue ancestry suggests that interior⁄ surface transplants involving all-beta proteins are a feasible engineering strategy

Abbreviation

Gdm.HCl, guanidinium hydrochloride

Briefly, in our approach, candidate proteins that can

be subjected to surface grafting are required to fulfil

two structural criteria: firstly, the donor and recipient

proteins must have polypeptide backbones that can be

structurally superimposed (to within an RMSD

£ 2.0 A˚); secondly, the surface areas subjected to

graft-ing must predominantly be b sheet-based structures,

and associated loop structures, with very little or no

helical content The reason for the latter criterion is

that, within any strand participating in a

multi-stranded sheet on a protein’s surface, alternating

resi-dues face away from the sheet in opposite directions

The sheet itself (described by the strand backbones

and hydrogen bonds) thus acts like a separator that

physically separates two distinct groups of mutually

interacting residues that have already evolved to pack

independently of each other – one facing the solvent,

and the other facing the protein’s interior It is our

contention that when the backbone atoms of two

structurally homologous all-beta proteins are

superim-posable (despite poor sequence homology), the two

proteins appear to have somehow evolved very

differ-ent residue–residue packing schemes for ‘surface’ and

‘interior’ residues within the b sheet(s) under

consider-ation, compatible with the same set of backbone atom

coordinates This compatibility automatically generates

scope for the success of mutation-based replacement of

one entire set of residues with a completely different

set of analogous residues from a structurally

homolo-gous protein This is what we call grafting

It is necessary to note that, although the

compatibil-ity of the ‘original’ and ‘replacement’ sets of residues

with the same set of backbone atom coordinates

defi-nitely presages, or even predicts, success in grafting, it

does not automatically guarantee success because of

uncertainties involved with respect to the mechanisms

of chain folding A b sheet can bring together strands

that are widely separated in the primary sequence

Therefore, residues constituting the solvent-exposed

surface of a b sheet are generally non-contiguous and

are sourced from all over the protein’s sequence

Large-scale mutagenic replacements of surface residues

can conceivably affect the mechanisms by which chains

achieve their folded three-dimensional (native)

struc-tures; indeed, as is widely appreciated, sometimes even

a single mutation can drastically affect folding, unless

compensating mutations occur elsewhere in the

pro-tein, and there is no gainsaying that sufficient numbers

of mutually compensating mutations would be made in

a surface grafting experiment involving tens or

hun-dreds of residues undergoing replacement Therefore,

theoretical verification of packing compatibility does

not prove that folding will lead to the desired

structure(s) It is necessary to perform the experiment, and see whether this indeed occurs

Our grafting approach – successfully demonstrated here using the whole surfaces of two structurally homologous proteins – involves the performance of five systematic steps that combine structural (bioinformat-ics) analyses with genetic engineering and protein bio-chemistry: (a) superimposition of the polypeptide backbones of any two significantly structurally homolo-gous all-beta proteins; (b) identification of all pairs of residues located at structurally analogous positions in the two proteins; (c) segregation of such pairs of resi-dues into separate sets, i.e those contributing atoms to the surface, and those contributing to the hydrophobic interior; (d) site-directed replacement of residues consti-tuting the surface of one protein by analogous residues occurring in the other protein; and (e) expression, purification and characterization of the mutant

For our ‘whole-surface’ grafting experiment, we selected two b sheet-rich proteins sharing extensive structural homology: the vertebrate bovine lens struc-tural proteins bB2 crystallin [2] and cB crystallin [3] The proteins are of different lengths: cB crystallin is 174 residues long, while bB2 crystallin is 201 residues long

in its full-length form, but only approximately 175 or

177 residues long in its truncated form, depending on exactly how its N- and C-terminal extensions have been removed, or truncated In comparing the amino acid sequence of cB crystallin with that of the equivalent (truncated) form of bB2 crystallin consisting of only the core domain structures without the terminal extensions,

61 residues with the same identity (approximately 35%) are used at structurally equivalent positions in the two proteins, while another 22 residues of similar nature (approximately 12%) are used at other structurally equivalent positions, bringing the total homology to 47% Thus, over half the residues used by the two pro-teins at structurally equivalent positions are different with respect to both their identity and their nature Both bB2 (Protein Data Bank accession 2BB2) [2] and cB (Protein Data Bank accession 1AMM) [3] con-sist of two double Greek key domains Each of these domains, approximately 80–85 residues long, consists

of two interacting Greek key motifs, each approxi-mately 40 amino acids long As already mentioned, bB2 has N- and C-terminal sequences that extend beyond the core two-domain motif The inter-domain linkers joining the N- and C-terminal domains in the two proteins are very different from each other in structure, as well as sequence, with the linker in cB being bent into a V-shape that allows the two domains

to interact intramolecularly (such that the protein is a monomer), while the linker in bB2 is extended (causing

the protein to form a homodimer in which the N- and

C-terminal domains of two different chains interact

like the two domains of cB)

Both proteins belong to a superfamily of proteins

displaying limited sequence homology but high

struc-tural homology [4] Here, we show that a ‘recombined’

amino acid sequence created through residue

altera-tions in cB crystallin, involving (a) substitution of

some residues with analogous residues from bB2

crys-tallin, (b) insertion of certain bB2 residues with no

analogs in cB, and (c) deletions of other cB residues,

leads to formation of a soluble protein that displays

many of the structural stability characteristics of cB

crystallin and most of the surface characteristics of

bB2 crystallin In addition, this new protein, which we

call Gambeta, displays certain characteristics that are

not seen in either of its progenitors Our results also

shed some light on the evolution of monomeric versus

multimeric structural arrangements in the bc crystallin

superfamily

Results and discussion

Using the cB crystallin sequence as a template, many substitution mutations were first made in silico to replace the surface residues of cB crystallin with struc-turally analogous bB2 residues Certain bB2 surface res-idues, including those constituting the solvent-exposed inter-domain linker, have no counterparts in cB; conse-quently, these were inserted into the cB sequence Fur-ther, certain cB surface residues were deleted, as these have no structurally analogous residues in bB2 Details

of the above changes are given in Table S1 A new syn-thetic gene incorporating all the above changes was expressed in Escherichia coli The amino acid sequence

of the protein product of this gene, named ‘Gambeta’,

is defined in column 9 of Table S1, and also shown in the top part of Fig 1, which displays the amino acid sequence of Gambeta in a structure-based sequence alignment with the amino acid sequences of its progeni-tor crystallins, cB and bB2 Figure 1 also provides

Fig 1 The surface ⁄ interior transplant (Top) The green font represents residues that are not present in the progenitors, Structure-based sequence alignment showing N- and C-terminally truncated bB2 crystallin (Protein Data Bank accession 2BB2) in blue, cB crystallin (Protein Data Bank accession 1AMM) in red⁄ orange, and Gambeta crystallin in a combination of blue (for bB2-derived residues) and red ⁄ orange (for cB-derived residues) The inter-domain linker region separating the two double Greek key domains is marked Residues presenting side chains

to the aqueous solvent are highlighted by green shading of structurally equivalent positions in all three sequences Of the surface regions sub-jected to transplantation, those involving contiguous residues are mainly from loops separating b strands, while single surface residues flanked

by core ⁄ interior residues are from strands (Bottom) Schematic diagram showing the relationship of Gambeta to its two progenitors The surface

is shown in green for all three proteins, and the notional boundary separating the surface from other regions is shown in white Residues of bB2 are shown in blue, while those of cB are shown in red ⁄ orange In Gambeta, colors denote the origins of the residues from the two progenitors.

details about which residues occur upon the surfaces of

bB2 and cB Furthermore, the bottom part of Fig 1

shows a schematic representation of the

transplanta-tion This part of the figure emphasizes the conceptual

point that our construction of a new protein, using a

synthetic gene encoding surface residues carefully

selected from bB2, and interior residues sourced from

cB, may be viewed as a ‘surface transplantation’

experi-ment or an ‘interior transplantation’ experiexperi-ment,

depending on one’s perspective

Expression of Gambeta and

confirmation of its identity

Gambeta was overexpressed from a synthetic gene

cloned between the NdeI and XhoI restriction sites of

the expression vector pET-23a, with a C-terminal 6xHis

affinity tag, in E coli strain BL21-DE3pLysS The

DNA sequence and sequencing chromatograms of the

synthetic gene are shown in Fig S1A,B Figure S2A

shows the overexpression levels of several

Gambeta-expressing clones, showing similar yields of 100–

110 mgÆL)1 of culture After confirmation of the

sequence of the encoding gene by DNA sequencing, we

selected the clone shown in lane 2 of Fig S2A for

pro-tein production Gambeta was expressed and purified

by Ni-nitrilotriacetic acid immobilised metal affinity

chromatography (IMAC) chromatography under

non-denaturing conditions Figure S2B shows a

representa-tive purification profile of Gambeta from a 1 L culture,

involving elution of bound protein by 250 mm

imidaz-ole, which was later removed by dialysis against 20 mm

Tris pH 8.0 Figure S2C shows that the purified protein

has an intact mass of 21 746 Da, which is only 148 Da

less than the mass of 21 894 Da expected for the 188

amino acid residue Gambeta chain; this error is well

within the permitted range of errors for mass

measure-ments using MALDI-TOF mass spectrometry in the

linear mode The identity of the protein was further

confirmed by MALDI-TOF-based peptide mass

finger-printing, with 1–2 Da accuracy, involving detection of

the masses of trypsinolytic peptides in the mass range of

500–5000 Da Peptides detected by peptide mass

finger-printing provided a very high coverage (83%) of the

sequence of the C-terminally 6xHis-tagged form of

Gambeta, as shown in Fig S2D, confirming that the

protein produced and purified was indeed Gambeta

Gambeta is a dimer like bB2

It is known that cB crystallin is a monomer whereas

bB2 crystallin is a homodimer [5] To determine the

quaternary structural characteristics of Gambeta, the

protein was subjected to gel filtration chromatography

on an analytical Superdex-200 SMART column as shown in Fig 2A (the column’s calibration profile is shown in Fig S3) For comparison, control samples of bB2 and cB crystallin were also chromatographed under identical experimental conditions Gambeta was found to elute predominantly at approximately 1.66 mL from this column of 2.4 mL bed volume, with

a minor fraction of the population also seen to elute

as a soluble aggregate at the void volume (0.9 mL) The elution of Gambeta at 1.66 mL indicates that it is

a dimer, with a hydrodynamic volume similar to that

of the bB2 control, which elutes at approximately 1.70 mL The bB2 control and Gambeta were pro-duced in the truncated form, without the N- and C-terminal extensions that normally exist in bB2 Oth-ers who have similarly produced bB2 without exten-sions have also observed that bB2 exists in a predominantly dimeric state; however, reports suggest that there is usually an accompanying minority popu-lation of tetrameric bB2 present with the dimeric pop-ulation [4,5] We did not find any evidence of a minority tetrameric population, either with the bB2 control or with Gambeta This could be due to the fact that the C-terminal affinity tag used to produce Gamb-eta and its progenitor controls acts like bB2’s natural C-terminal extension, sterically inhibiting further asso-ciations once a dimeric state has formed In this context, it is important to note that previous studies with truncated bB2 used no affinity tags, but instead used naturally occurring histidines in bB2’s sequence for metal affinity-based purification In any case, the important point to note is that the control cB crystal-lin protein elutes at 1.85 mL, as a monomer, despite being identical to bB2 and Gambeta with respect to its C-terminal affinity tag, indicating that Gambeta’s dimerization is due to its possessing the inter-domain linker peptide and surface features of bB2 crystallin

Gambeta’s structural ⁄ biochemical characteristics are derived partly from bB2 and partly from cB

Figure 2B shows that the CD spectrum of Gambeta has a typical negative band maximum at 216–218 nm, with a mean residue ellipticity of about )9000 degressÆ

cm2Ædmol)1 The shape of Gambeta’s spectrum resem-bles that of the spectra of bB2 and cB, which are also shown in Fig 2B While this indicates that Gambeta is

a folded all-beta protein like both of its progenitors, Gambeta also appears to have a higher b sheet content than both bB2 and cB, suggesting that it is somewhat more folded than its progenitors Figure 2C,D shows

the guanidinium hydrochloride (Gdm.HCl)-induced

and urea-induced denaturation transition curves for

Gambeta, bB2 and cB, as plots of changes in the mean

residue ellipticity at 216 nm with increasing

concentra-tions of denaturant For Gambeta, data are omitted

for concentrations between 0.3 and 1.75 m Gdm.HCl,

because the protein showed precipitation at these

dena-turant concentrations (this behavior is discussed

below) Neither of the control progenitor proteins

showed this behavior Although different initial mean

residue ellipticities are involved in all three cases, it is

clear from these transitions that Gambeta’s unfolding

closely parallels that of bB2, rather than that of cB

crystallin Although cB shows great resistance to

urea-mediated unfolding (as reported previously [7]), at least

over the time scale of an overnight incubation

(approx-imately 12 h), both Gambeta and bB2 showed some

unfolding in urea, with similar profiles of partial

unfolding

The wavelengths of maximal fluorescence emission (emkmax) and the emission intensities of tryptophan residues tend to be acutely sensitive to the polarity of their environment within a protein The emkmax of a solvent-exposed tryptophan is usually approximately 352–353 nm, while that of a buried tryptophan tends

to be blue-shifted to a lower wavelength, to a degree that is dependent on the extent of burial [6] There are four tryptophan residues in cB, all of which lie buried within its structural core [3] In contrast, only three of these tryptophans are conserved at structurally equiva-lent positions in bB2, and there are two additional tryptophans on its surface [2] As the core of Gambeta

is derived from cB and its surface is derived from bB2,

it inherits all four of the cB tryptophans together with both of the bB2 surface tryptophans, making six tryptophans in all Figure 3A shows the fluorescence emission spectra of all three proteins, obtained at matched concentrations Although Gambeta, bB2 and

Fig 2 Quaternary and secondary structure⁄ stability of purified Gambeta and its progenitors (A) Gel filtration elution profiles on an analytical Superdex-200 SMART column (B) Far-UV CD spectra (C) Changes in mean residue ellipticity at 216 nm following overnight incubation in various molarities of Gdm.HCl (D) Changes in mean residue ellipticity at 216 nm following overnight incubation in various molarities of urea.

cB have six, five and four tryptophans, respectively,

cB shows the most intense emission, followed by

Gambeta and bB2, in that order This indicates that

the tryptophans of Gambeta and bB2 are quenched by

their local structural environments compared with

the cB tryptophans Theemkmaxof bB2 is between the

emkmaxvalues of Gambeta and cB, and the shape of its

fluorescence emission envelope is quite distinct from

those of Gambeta and cB All three proteins display

emkmax values below 335 nm (Gambeta at

appro-ximately 335 nm, bB2 at approappro-ximately 331 nm, and

cB at approximately 325 nm), indicating that Gambeta

has an extremely well-folded structure in which the six

aromatic tryptophan residues are protected from

the aqueous solvent as effectively as those in bB2 and

cB

That the aromatic residues of all three proteins exist

in largely immobile environments, in association with

chiral structural elements, is also evident from the fact

that they display near-UV CD spectra that show marked spectral features (Fig 3B) Although near-UV

CD spectra are merely spectral signatures that cannot

be interpreted further, it is noteworthy that the spec-trum of Gambeta resembles that of bB2 much more than it resembles that of cB, at least in terms of inten-sity The same spectral features are seen in all three proteins, indicating that Gambeta has folded into a structure that is similar to that of its progenitors The reason that Gambeta’s spectrum is more like that of bB2 could be that Gambeta derives five of its six tryptophans from bB2 (including three that are buried

at equivalent structural positions in bB2 and cB, and two that exist only in bB2), with only one tryptophan sourced solely from cB

We further examined the denaturant-induced unfolding transitions of Gambeta and its progenitor controls by monitoring changes in fluorescence emis-sion and plotting variations in both emisemis-sion intensity and emkmax values with denaturant concentration We noted above that precipitation of Gambeta between 0.3 and 1.75 m Gdm.HCl interferes with CD-based monitoring of unfolding Such precipitation does not interfere with fluorescence-based monitoring of emkmax values during unfolding (which are independent of protein concentration), but can affect emission intensi-ties This is because precipitation can affect protein concentrations and therefore intensities, whereas fluo-rescence emission wavelength maxima are not sensitive

to scattering of light if the emkmax is sufficiently dis-placed from the wavelength of excitation (in nm), as this prevents the Rayleigh and Raman scatter from contaminating the emission spectrum The data for the Gdm.HCl- and urea-induced transitions are shown

in Fig 4A,B, which show actual protein emission intensities at 370 nm as a function of denaturant concentrations At 370 nm, shifting of the protein’s emission spectrum towards longer wavelengths (accompanying exposure of buried tryptophan resi-dues) causes emission intensities to rise progressively with unfolding All three proteins show changes in emission intensity at 370 nm with increasing denatur-ant concentration, although, as already mentioned, urea does not cause unfolding of cB The unfolding

of Gambeta by Gdm.HCl appears to be considerably less cooperative than that of its progenitors, whereas, with urea, unfolding of both Gambeta and bB2 appears not to be very cooperative In Fig 4A, no intensity data for Gambeta at 370 nm are presented for Gdm.HCl concentrations between 0.3 and 1.75 m; this is because of the precipitation of Gambeta seen

at these concentrations of Gdm.HCl, which is reversed

at higher concentrations of the denaturant

A

B

Fig 3 Tertiary structural features of purified Gambeta and its two

progenitors (A) Fluorescence emission spectra (B) Near-UV CD

spectra.

The data showing the changes in emkmax with the

two denaturants are shown in Fig S4A,B As for the

changes in emission intensities, the unfolding transition

of Gambeta is far less cooperative than the unfolding

transitions of bB2 and cB, both of which show more

or less cooperative unfolding transitions The

concen-tration at which half the population is unfolded (Cm)

of Gdm.HCl is far lower for Gambeta than for cB,

but considerably higher than the Cm of unfolding of

bB2 With urea, Gambeta unfolded entirely

non-cooperatively over the entire range of urea

concentra-tions used Thus, although Gambeta unfolds to the

same extent that bB2 unfolds (i.e fully), bB2’s

unfold-ing was completed in a sharp cooperative transition

between 1.0 and 2.5 m urea, whereas Gambeta’s

unfolding occurs slowly and in a monotonic fashion

between 0 and 7.0 m urea Gambeta appears to have

derived its relative resistance to unfolding by urea from cB, which shows hardly any unfolding even upon overnight incubation in 7.0 m urea

We also monitored the unfolding transition by examining changes in the ratio of emission intensities

at 320 and 370 nm with denaturant concentration, to see whether there is any inhomogeneity of behaviour with respect to the relative exposure of tryptophans in various parts of Gambeta and its control progenitors during denaturant-mediated unfolding The data are presented in Fig S4C,D Gambeta’s unfolding is clearly seen to be biphasic in the presence of Gdm.HCl

or urea Although the same is not seen clearly in the Gdm.HCl-induced denaturation profiles of the two progenitors, the intensity ratio data in Fig S4C indicate that cB shows very subtle unfolding Such unfolding involves two phases spanning the same range

of concentrations over which Gambeta shows this behavior in the urea-induced denaturation data in Fig S4D It is possible that the two phases of Gambeta unfolding seen in Fig S4C,D relate to independent unfolding of the two domains, which could be related to the fact that the protein shows non-cooperativity of unfolding

To explore the reversibility of the unfolding transi-tions, far-UV CD spectra of all three proteins were obtained after removal of 6 m Gdm.HCl or 7 m urea

by dialysis The data are presented in Figs S5 and S6, together with spectra of protein unexposed to denatur-ant The data show that both Gambeta and bB2 dis-play poor refolding, through dialysis, from the completely unfolded state achieved through overnight incubation in 7 m urea, while cB remains unaffected

by the treatment In contrast, all three proteins refold poorly from the completely unfolded states achieved through overnight incubation in 6 m Gdm.HCl

In summary, the far-UV CD transitions in Fig 2C,D as well as the emkmax value and intensity ratio transitions indicate that Gambeta is unfolded non-cooperatively by urea It is interesting that all six tryptophans of Gambeta become exposed to the aque-ous solvent even though the secondary structure of Gambeta does not fully unravel in the presence of these denaturants

Binding of Gambeta to the calcium-mimic dye Stains-all is intermediate to that of bB2 and cB

Figure 5A shows CD spectra induced in the otherwise achiral (and therefore non-dichroic) dye Stains-all in the presence of Gambeta, bB2 and cB crystallin The dye Stains-all binding simulates calcium-binding in a

A

B

Fig 4 Chemical denaturation of Gambeta and its two progenitors.

(A) Fluorescence emission intensities at 370 nm as a function of

Gdm.HCl concentration (B) Fluorescence emission intensities at

370 nm as a function of urea concentration.

protein, and both bB2 and cB crystallin are reported

to bind to this dye [8,9] Figure 5A shows that, for

equivalent concentrations of protein and Stains-all, the

strength of the approximately 650 nm negative band

(known as the J band) induced in Stains-all by its

binding to Gambeta is intermediate to that of bB2 and

cB crystallins As with the other crystallins, Gambeta

also shows a positive band in the region of 670–

700 nm at higher concentrations of dye to protein

(Fig S7); the control spectrum with the dye alone is

close enough to the zero line to be indistinguishable

Although, in our experiment, all three proteins show

the J band at approximately 650 nm, the reported

wavelength for Stains-all bound to full-length bB2 is

closer to 660 nm [8,9] As our spectra in Fig 5A are

the first spectra ever reported for Stains-all binding to

truncated bB2, the three spectra shown must be

com-pared only with each other and not with other reports

in the literature

Gambeta precipitates upon heating like

cB does

There have been no reports that bovine bB2 crystallin precipitates upon heating In contrast, cB is reported to precipitate upon heating, and the nature of its thermal aggregates has also been studied [10] It is known that there is a partial melting of structure that leads to aggregation in the temperature range 65–80C, with actual structural unfolding (as discerned by differential scanning calorimetry) occurring largely above 85 C, when aggregation is prevented from occurring [7] Figure 5B shows the responses of Gambeta and its con-trol progenitor proteins to heating The behavior of Gambeta is entirely like that of cB, in that there is a dramatic change in mean residue ellipticity at 216 nm upon heating, which leads to the mean residue elliptic-ity quickly being reduced to zero because the protein precipitates and disappears from the light path

Fig 5 Further characterization of Gambeta and its progenitors (A) CD spectra of the 650 nm J band induced in the calcium-mimic dye, Stains-all upon protein binding (B) Temperature-induced changes in mean residue ellipticity at 216 nm (owing to thermal unfolding, aggrega-tion or precipitaaggrega-tion) (C) Changes in refractive index (nD) with increasing protein concentration (D) Scattering (turbidity) shown by a 0.1 mgÆmL)1Gambeta solution at 600 nm as a function of increasing Gdm.HCl concentration.

However, although the thermal precipitation is like that

of cB, the temperature at which the thermal

precipita-tion occurs is actually much closer to the temperature

at which bB2 (which is quite thermostable) displays a

very minor partial unfolding transition, without either

undergoing complete thermal unfolding or thermal

pre-cipitation When the respective cuvettes were removed

from the Peltier-controlled chamber of the CD

spec-trometer after heating and cooling, all the protein was

found to have formed visible aggregates in the case of

Gambeta and the cB control, but no precipitates

what-soever were visible in the case of the bB2 control

Gambeta shows cold precipitation like

its progenitors

Cold precipitation is defined as a tendency to

precipi-tate out of solution at low temperatures in a

concen-tration-dependent fashion Cold precipitation is one of

the most defining characteristics of cB crystallin In a

concentration-dependent manner, with greater

precipi-tation seen at higher protein concentrations, cB

precip-itates visibly out of aqueous solution upon cooling to

temperatures below 10C [11,12] What is especially

interesting about this cold precipitation is that it is

fully reversible, i.e the solution clears when the

tem-perature is returned room temtem-perature, with no

appar-ent change in the protein’s characteristics, suggesting

that no profound structural change is involved in the

phenomenon In contrast to this behavior of cB (and

all other c isoforms, which show reversible cold

precipitation), mouse cN crystallin has been reported

to show irreversible cold precipitation [13] No other

crystallin, including full-length bB2 crystallin, has been

reported to show such precipitation

Gambeta was found to readily precipitate out of

solution in the refrigerator, in a

concentration-depen-dent manner, over a period of hours, for all

concentra-tions exceeding 10–12 mgÆmL)1 However, unlike the

reversible cold precipitation shown by cB crystallin,

the cold precipitation of Gambeta was found to be

irreversible, in that no re-dissolution of the protein

could be detected upon return to room temperature

Interestingly, whereas full-length bB2 has never been

reported to show cold precipitation, our truncated bB2

control showed cold precipitation at these

concentra-tions just like the cB control and Gambeta samples

Interestingly, the cold precipitation of N- and

C-termi-nally truncated bB2 was also found to be irreversible,

like that of Gambeta, explaining where this

irrevers-ibility comes from It is possible that others who have

worked with truncated bB2 have not made this

obser-vation previously because they have not used protein

concentrations high enough for the phenomenon to manifest itself; even cB and the other c crystallins are known to show cold precipitation only at concentra-tions exceeding 10–15 mgÆmL)1 [11,12] It is possible that the N- and C-terminal extensions of full-length bB2 (which protect it from associating beyond the dimeric state) somehow also protect it from cold pre-cipitation, and that we have removed this protection

by truncating Gambeta and bB2 It may be recalled that we truncated these two proteins in order to allow proper comparison with cB, as cB lacks these terminal extensions

Gambeta is soluble at ultra-high concentrations like other lens crystallins

The crystallins are special proteins, in that they exist in the fiber cells of the vertebrate ocular lens at concentra-tions in excess of 100 mgÆmL)1; in the center of a lens, crystallin concentrations can even reach 500–

600 mgÆmL)1 [14] The crystallins in the lens remain soluble at such high concentrations, and form clear solutions of high refractive index that help the lens to focus light onto the retina [15] There is a natural gradi-ent of crystallin concgradi-entrations in the lens, increasing from the periphery to the center This is associated with

a corresponding gradient of refractive index that helps the lens to correct for spherical aberration by bending light to lesser and lesser extents as the periphery is approached, to compensate for shape changes

As Gambeta is derived from two progenitor crystal-lins, we examined whether it is soluble at high concen-trations, and also whether it generates highly refractive solutions We were able to concentrate Gambeta to approximately 280 mgÆmL)1, with no evidence of pre-cipitation or aggregation Figure 5C shows the mea-sured change in the refractive index for yellow light (nD) against Gambeta concentration, together with similar plots for the control bB2 and cB crystallins, over the concentration range of 0–50 mgÆmL)1 The plot shows that Gambeta possesses the most important properties of any crystallin, i.e high solubility and the ability to form clear and transparent solutions of high refractive index Furthermore, the slope of the increase

in refractive index with protein concentration was highest for cB crystallin, followed by Gambeta and bB2 crystallin

Gambeta precipitates at intermediate concentrations of Gdm.HCl

As noted above, far-UV CD data for Gambeta could not be collected between 0.3 and 1.75 m Gdm.HCl,

because Gambeta shows a tendency to precipitate This

precipitation frustrates any attempts to examine

whether or not the homodimeric form of Gambeta can

be dissociated into folded monomers by low

concentra-tions of the denaturant The precipitation is visible

within a few tens of minutes after addition of the

Gdm.HCl We wished to examine whether this

precipitation is an equilibrium phenomenon, showing a

specific concentration-dependent trend, or a

non-equilibrium phenomenon like most forms of protein

aggregation We wish to point out what is a prima facie

reason to believe that this may be an equilibrium

phenomenon In the CD data presented in Fig 2C, the

data points prior to 0.3 m Gdm.HCl and after 1.75 m

Gdm.HCl show continuity; it may be argued that this

would not have been the case if the precipitation were a

non-equilibrium phenomenon However, we wished to

examine this further by monitoring the extent of

precipitation in various concentrations of Gdm.HCl

Overnight incubation would allow sufficient time for a

non-equilibrium phenomenon to precipitate most or all

of the protein over the entire relevant range of

Gdm.HCl concentrations, creating discontinuity in the

data

To explore this further, we measured and plotted

the turbidity (A600) of solutions of 0.1 mgÆmL)1

Gambeta incubated overnight in the presence of

various concentrations of Gdm.HCl, as shown in

Fig 5D The data reveal a Gaussian distribution of

turbidity with increasing Gdm.HCl concentration, with

a peak at approximately 1 m Gdm.HCl and no

discon-tinuity, indicating that phenomenon is an equilibrium

one A search of the biochemical literature showed

that another protein, rusticyanin, also shows similar

precipitation behavior in the presence of a certain

range of concentrations of Gdm.HCl [16] We have

not yet characterized these precipitates further, but we

describe this precipitation behavior to emphasize that

this is one aspect of Gambeta that is not directly

attributable to either of its progenitors

Experimental procedures

Design of Gambeta

As the geometries of the arrangement of domains are

different in the two proteins, the N-terminal domains of bB2

and cB crystallin were superimposed separately from the

C-terminal domains of the two proteins using LSQMAN

software [17] This showed that the backbone atoms of the

N-terminal domains can be superimposed with an RMSD of

0.9 A˚, while those of the C-terminal domains can be

superimposed with an RMSD of 1.05 A˚ The residues at

structurally analogous positions in the two proteins identi-fied by LSQMAN are listed in columns 2 and 3 of Table S1 For the two domains, backbone atoms of a total of 150 residues can be seen to be superimposed with individual RMSD values £ 2.00 A˚, while eight more residues superim-pose with individual RMSD values of 2.00–3.00 A˚ Details

of structurally analogous residues are given in column 4 of Table S1, with RMSD values in column 5 We used a combi-nation of visual and software analysis by AreaImol, as implemented in CCP4 [18], to assess the solvent accessibility

of each residue, to identify residues contributing atoms to the formation of the surface in each protein Details of specific residue pairs contributing to surface formation are given in column 6 of Table S1 Of these surface residues, some are conserved and so there was no need to alter these during surface grafting Details of conserved residues are given in column 7 of Table S1 Column 8 in Table S1 list the action taken, i.e whether the residue in cB was (a) mutated and replaced with the structurally analogous residue occur-ring in bB2, (b) left unaltered, or (c) deleted, or (d) whether

a specific residue from bB2 had to be inserted to create a bB2-derived surface in Gambeta, while conserving the hydrophobic core of cB Mutations were inserted in silico into a gene encoding the sequence of c crystallin, and the sequence of this gene was optimized for expression in E coli using gene designer dna 2.0 software This approach differs considerably from another very interesting approach that also uses cB crystallin’s structural scaffold to generate libraries of crystallins with novel binding properties [19], as our approach is rational while the library approach is combinatorial, and our approach also results in a folded structure despite making a much larger number of changes

Gene synthesis and cloning

A gene with a sequence designed as described above (shown

in Fig S1) was produced through contract synthesis by Ocimum Biosolutions (Hyderabad, India) in a pUC-19 (https://www.dna20.com/tools/genedesigner.php) plasmid The gene was amplified from this plasmid by PCR using for-ward primer 5¢-ACTTATACTATCCATATGGGTAAAAT CATCTTCTTTGAACAGG-3¢ and reverse primer 5¢-ACT-TATACTATCCTCGAGCCACTGCATATCACGGATAC GACGC-3¢ The forward primer incorporated an NdeI site and the reverse primer incorporated an XhoI site (both underlined) to allow digestion and cloning of this amplicon between the NdeI and XhoI sites of the expression vector pET-23a, enabling expression with an N-terminal methionine and a C-terminal extension incorporating 6xHis residues and two other residues (Leu and Glu) from the XhoI site placed immediately upstream of the stop codon The pET-23a plasmid incorporating the clone was transformed into the XL1-Blue strain for making of plasmid stocks and sequenc-ing, and into the BL21(DE3)pLysS strain for expression