proteinsSTRUCTURE O FUNCTION O BIOINFORMATICS

Structural insights into the aggregationbehavior of Murraya koenigii miraculin-like proteinbelow pH 7.5Purushotham Selvakumar, Nidhi Sharma, Prabhat Pratap Singh Tomar, Pravindra Kumar, and

Ashwani Kumar Sharma*

Department of Biotechnology, Indian Institute of Technology Roorkee, Roorkee 247 667, India

ABSTRACT

Murraya koenigii miraculin-like protein (MKMLP) gradually precipitates below pH 7.5. Here, we explore the basis for this

aggregation by identifying the aggregation-prone regions via comparative analysis of crystal structures acquired at several

pH values. The prediction of aggregation-prone regions showed the presence of four short peptides either in beta sheets or

loops on surface of the protein. These peptides were distributed in two patches far apart on the surface. Comparison of

crystal structures of MKMLP, determined at 2.2 A resolution in pH 7.0 and 4.6 in the present study and determined at

2.9 A in pH 8.0 in an earlier reported study, reveal subtle conformational differences resulting in gradual exposure of

aggregation-prone regions. As the pH is lowered, there are alterations in ionic interactions within the protein interactions

of the chain with water molecules and exposure of hydrophobic residues. The analysis of symmetry-related molecular inter-

faces involving one patch revealed shortening of nonpolar intermolecular contacts as the pH decreased. In particular, a

decrease in the intermolecular distance between Trp103 of the aggregation-prone peptide WFITTG (103–108) unique to

MLPs was observed. These results demonstrated that aggregation occurs due to the cumulative effect of the changes in inter-

actions in two aggregation-prone defined regions.

Proteins 2014; 82:830–840.VC 2013 Wiley Periodicals, Inc.

Key words: acidic pH; aggregation-prone peptides; crystal structure; conformational changes; hydrophobic interactions.

INTRODUCTION

Understanding the underlying reasons for protein

aggregation, both in vivo and in vitro, is a major chal-

lenge for protein biochemists. Protein aggregation results

in a number of human pathologies including Alzhei-

mer’s, Parkinson’s, and Creutzfeldt-Jakob diseases; and

the systemic amyloidoses associated with immunoglobu-

lin light chain, transthyretin, lysozyme, and Beta-2

microglobulin.1 There are specific regions of amino acid

sequence, termed “aggregation prone,” which plays a

major role in determining the tendency of proteins to

aggregate.2 These “aggregation-prone” regions are not

exposed in a native protein. However, proteins can be

destabilized by heat, pH, denaturants, etc. exposing these

regions and leading to aggregation.3–5

Miraculin-like proteins (MLPs) exhibit significant

sequence identity (�39%–55%) to miraculin protein, a

24.6 kDa plant protein purified from red berries of

Richadella dulcifica.6,7 Both proteins belong to Kunitz

superfamily and have sequence similarity (�30%) to soy-

bean Kunitz family trypsin inhibitors.8 Murraya koenigii

miraculin-like protein (MKMLP), a 21.4 kDa protein

with trypsin inhibitory activity, was purified and charac-

terized from seeds of Murraya koenigii belonging to Ruta-

ceae family.9 Despite being a member of Kunitz

superfamily, MKMLP demonstrated some distinct features.

It formed a distinct cluster with MLPs in phylogenetic

Additional Supporting Information may be found in the online version of this

article.

Abbreviations: MKMLP, Murraya koenigii miraculin-like protein; MLPs,

Miraculin-like proteins; STI, soybean Kunitz inhibitor; PQS, Protein Quater-

nary Structure server; ANS, 8-anilino-1-naphthalene sulfonate.

Grant sponsor: Department of Science and Technology, Government of India;

Grant number: SR/SO/BB-002/2007.

*Correspondence to: Ashwani Kumar Sharma, Department of Biotechnology,

Indian Institute of Technology Roorkee, Roorkee-247 667, India.

E-mail: aksbsfbs@yahoo.co.in

Received 30 May 2013; Revised 3 October 2013; Accepted 21 October 2013

Published online 22 November 2013 in Wiley Online Library (wileyonlinelibrary.

com).

DOI: 10.1002/prot.24461

830 PROTEINS VVC 2013 WILEY PERIODICALS, INC.

analyses and showed major differences at primary and sec-

ondary specificity sites in reactive loop when compared

with classical Kunitz inhibitors like soybean Kunitz inhibi-

tor (STI). The conventional Arg/Lys at P1 position in

MKMLP has been replaced by an Asn residue suggesting

that the protein may not act as a typical substrate-like

inhibitor because of the absence of a residue essential for

trypsin specificity.10 The crystal structure of MKMLP

determined at 2.9 A exhibited a classical b-trefoil fold

similar to Kunitz family inhibitors with major conforma-

tional differences limited to loop regions. The unique fea-

tures of the MKMLP structure was the presence of three

disulfide bridges and two short 310 helices.10 The

MKMLP, native and heat treated, was found to be highly

stable against proteolysis due to the disulfide bridges.

Unlike classical Kunitz inhibitors, MKMLP was function-

ally unstable at higher temperature.11 The reason for this

loss in inhibitory activity has been attributed to the

absence of stabilization of reactive loop conformation

where Asn13, which plays an important role in stabilizing

the reactive loop conformation in STI, is replaced by

Ala.12 MKMLP shows bioinsecticidal activities and possess

an N-terminal signal sequence with possible plasma

membrane-spanning motif for plasma membrane indicat-

ing the translocation of protein from the site of synthe-

sis.10,13 Another important unique feature of MKMLP, as

compared with most Kunitz members which are stable

over a broad range of pH,14,15 is its solubility properties

at acidic pH. The protein gradually precipitates below pH

7.5 with an increasing rate of precipitation as pH is low-

ered.9 The present study explores the basis for this aggre-

gation by identifying the aggregation-prone regions and

analyzing the subtle conformation changes by comparative

crystal structure analysis in different pH conditions. Here

we report crystal structures of MKMLP at 2.2 A resolution

determined at pH 7.0 and 4.6, and also compare the three

structures determined at pH 8 (previously reported), 4.6,

and 7 to unravel the molecular basis of aggregation

behavior.

MATERIALS AND METHODS

Purification, crystallization and datacollection

Purification of MKMLP was carried out as described

earlier9,16 Briefly, MKMLP was purified by crushing of

M. koenigii seeds and soaking overnight at 4�C in 30 mL

of 50 mM Tris–HCl buffer, pH 7.5. The homogenate was

cleared by centrifugation at 12,000g for 1 h and the

supernatant was used for purification. The protein was

purified by combination of anion exchange and size

exclusion chromatography. Also, affinity chromatography

using Cibacron blue 3GA was done for single step purifi-

cation. The protein was crystallized by the sitting-drop

vapor diffusion method under two pH conditions. The

precipitant solutions were 4M ammonium acetate, 0.1M

BIS-TRIS propane, pH 7.0 and 4M ammonium acetate,

0.1M sodium acetate trihydrate, pH 4.6. Drops were pre-

pared by mixing 1 mL protein solution with 1 mL precipi-

tant solution and were equilibrated against 50 mL

reservoir solution. For cryoprotection, crystals briefly

exposed to well solution containing 20% glycerol were

mounted in cryoloops prior to collection of X-ray dif-

fraction data. Data were collected on a MAR 345dtb

image-plate system using Cu Ka radiation generated by a

Bruker Microstar-H rotating-anode generator operated at

45 kV and 60 mA, and equipped with Helios optics. The

crystals belonged to the monoclinic space group C121,

with unit-cell parameters a 5 101.51 A, b 5 45.69 A,

c 5 38.78 A for crystal grown at pH 7 and a 5101.62 A,

b 5 45.42 A, c 5 38.79 A for crystal grown at pH 4.6.

The crystals contained one molecule in asymmetric unit.

Diffraction was observed to 2.2 A resolution (Fig. S1,

Supporting Information). The diffraction data were proc-

essed and scaled with iMOSFLM and SCALA program in

CCP4i suite.17

Structure solution and refinement

The structure was solved by the molecular replacement

method using MOLREP17 with the structure of the

MKMLP [PDB ID: 3IIR] as the search model. The initial

model obtained with molecular replacement was refined

using REFMAC 5.217 first as a rigid body, and subse-

quently, it was refined using restrained refinement. Model

building was conducted in manual mode in Coot,18 fol-

lowed by refinement in REFMAC 5.2. The alternate cycles

of refinement and model building were performed for all

the data in the resolution range 50.5–2.2 A. The water

molecules were added according to the criteria that each

water molecule must make at least one stereochemically

reasonable hydrogen bond, that it should be well defined

in (2mjfobsj–Djfcalcj) and (mjfobsj–Djfcalcj) electron density

maps. The water molecules were added and removed in

subsequent refinement and model building cycles as per

the above criteria. The stereochemistry of final model was

analyzed by PROCHECK19 and MOLPROBITY.20 The

protein structures were examined using molecular visual-

ization software Coot18 and PyMOL.21 The atomic coor-

dinates have been deposited in the Protein Data Bank.

Prediction of aggregation-prone regions

Aggregation-prone regions were predicted using differ-

ent programs employing empirical and structure-based

algorithms such as Tango,22 FoldAmyloid,23 Aggres-

can,24 Waltz,25 Zyggregator,26 and Pasta.27

Accession numbers

The atomic coordinates and structure factors have

been deposited into the Protein Data Bank under the

Insights into Aggregation Behavior of MKMLP

PROTEINS 831

following accession codes: 3ZC8 (pH 7 structure) and

3ZC9 (pH 4.6 structure).

RESULTS

Three-dimensional structure of MKMLP atpH 7.0 and pH 4.6

Quality of the model

The refinement data in Table I shows that both models

are well refined with excellent stereochemistry and crys-

tallographic R-factor values. The deviations in bond

lengths and angles are within reasonable limits from ideal

values. The electron density is well defined in both struc-

tures except at the C-terminal end residues 183–190 so

they were not used for model building.

In pH 7 structure, the temperature factors of residues

in five loops namely L2 (24–30), L3 (35–43), L4 (48–57),

L6 (80–89), and L12 (168–174) were found to be higher

than the average value. For the residues involving loops

L1 (1–16), L5 (63–74, reactive loop), L7 (94–104), L8

(110–114), L9 (128–134), L10 (140–151), and L11 (162–

163) the temperature factors were less than the average

value. For pH 4.6 structure, the temperature factors of

residues in six loops namely L2 (22–30), L3 (35–43), L4

(48–57), L6 (80–89), L8 (110–114), and L12 (168–175)

were higher and the residues included in six loops L1

(1–16), L5 (63–74, reactive loop), L7 (94–104), L9 (128–

134), L10 (140–151), and L11 (162–163) were found to

be less compared with average value.

Overall structure

The three-dimensional structures of MKMLP, deter-

mined at 2.2 A resolution, at both pH values are similar

with a few exceptions. The pH 7 structure was predicted

to be monomeric but the pH 4.6 structure was predicted

to be dimeric by the Protein Quaternary Structure server

(PQS).28 The superposition of core region gave RMSD

of 1.971 A for 110 Ca atoms between two MKMLP

structures. The overall crystal structure consists of 12

antiparallel b-strands, loops connecting the b-strands,

one a-helix, and a short 310 helix (Fig. 1). In both struc-

tures, except for first and last b-strands, the correspond-

ing residues forming b-strands are same. b-strand 1 is

shortened by two residues (residues 17–21) and b-strand

12 by one residue (176–180) in pH 4.6 structure as com-

pared with pH 7.0 structure (residues 17–23 and 175–

180). The corresponding loop regions are also changed

in two structures where residues 24–30 (L1), 168–174

(L12) form loop region in pH 7.0 structure and residues

22–30 (L2), 168–175 (L12) form loop region in pH 4.6

structure. Both MKMLP structures, like Kunitz family

inhibitor structure, exhibits a typical b-trefoil fold with

six of the strands arranged in a barrel structure and

other six forms a triangular lid on the barrel. A pseudo-

threefold internal symmetry with symmetry axis roughly

parallel to barrel axis divides the structure into three

repeating units. Each unit consists of approximately 60

amino acids arranged in four b-strands. In addition to

one a-helix and a short 310 helix, the presence of short

stretches of distorted helices within loops was observed.

In pH 7 structure, these are present in loops L1 (residues

6–8 and 13–15), L2 (26–28), L4 (residues 51–53), L5

(residues 63–65), L6 (residues 84–87), and L10 (residues

144–147). In pH 4.6 structure, six short stretches of dis-

torted helices were observed similar to pH 7 structure

except for the L2 (residues 26–28). The presence of heli-

ces which constitutes almost 6% of the structure is a

unique feature of MKMLP and substantiates our earlier

CD results which demonstrated a, b pattern for the pro-

tein. The residues forming sheets, helices, and loops are

indicated in Table II.

The crystal structure of MKMLP at higher resolution

of 2.2 A showed remarkable improvement. The diffrac-

tion data was collected at room temperature in case of

2.9 A. Although the overall three-dimensional structures

of MKMLP determined at 2.2 A in two pH conditions

were similar to the earlier reported structure of MKMLP

at 2.9 A in pH 8.0 condition, a few noticeable differences

were observed. The superposition of core regions of 2.2

and 2.9 A structures gave an RMSD of 0.322 A (Fig. 2).

Table ICrystal Parameters, Data Collection, and Structure Refinement

Crystal data and intensity statistics pH 7 pH 4.6

Space group C 1 2 1 C 1 2 1Unit-cell parameters (�)a 101.51 101.61b 45.69 45.42c 38.78 38.79Resolution range (�) 50.5–2.2 50.5–2.2Completeness (%) 96.3 (74.9) 90.0 (69.0)Rmerge

a(%) 0.09 (0.27) 0.07 (0.21)Multiplicity 3.3 (3.1) 3.5 (3.3)Mean I/sigma (I) 10.3 (4.4) 13.7 (5.7)Refinement and model statisticsTotal no. of reflections 28,501 28,271No. of reflections (used) 8605 8647Percentage observed 99.6 92.1Wilson B-factor (�2) 20.9 24.6Crystallographic R-factor (%) 19.5 19.4Free R-factor (%) 24.2 23.8Average B factor (�2) 16.6 19.1RMSD bonds (�) 0.01 0.01RMSD angles (�) 1.3 1.2Validation by MOLPROBITYRamachandran plotFavored (%) 95.5 93.5Allowed (%) 4.5 5.5Outliers (%) 0 1PDB code 3zc8 3zc9

The values in parentheses refer to statistics in the highest bin.aRmerge 5 RhklRijIi(hkl) – <I(hkl)>j/RhklRiIi(hkl), where Ii(hkl) is the intensity

of an observation and <I(hkl)> is the mean value for its unique reflection; sum-

mations are overall reflections.

P. Selvakumar et al.

832 PROTEINS

The major changes include a longer extended helix from

five to eight residues (residues 115–122 as compared

with 118–122) and conformational changes at b-strand 1

and loop 2 involving residues 17–30. The orientation

and length of b-strand 1 and loop 2 is quite different in

two structures solved at different resolutions. The elec-

tron density was well defined in both cases. One reason

could be that data were collected in different conditions.

Interestingly, the conformational differences in two high

resolution structures determined in pH 7.0 and 4.6 con-

ditions involve the same region. The conformational

changes at b-strand 1 and loop 2 involving residues 17–

30, therefore, could be attributed to different pH condi-

tions. In 2.9 A resolution structure at pH 8, residues 18–

25 form b-strand 1 and residues 26–27 form the loop 2

while in 2.2 A resolution structures, residues 17–23 form

b-strand 1 and residues 24–30 form loop 2 at pH 7.0

and residues 17–21 form b-strand 1 and residues 22–30

form loop 2 at pH 4.6. It is to be noted that b-strand 2

(residues 31–34) is shortened in 2.2 A resolution com-

pared with 2.9 A resolution structures (residues 28–34)

because of increased length of loop 2. A significant

change in conformation in this region is seen. Superposi-

tion of residues 17–30 between structures at pH 4.6 and

7, pH 7 and 8, and pH 4.6 and 8 gave RMSD of 0.142,

1.246, and 1.236 A respectively [Fig. 3(a–c)]. Also, the

differences in the presence of distorted helices were

observed. The structures at 2.2 A showed the presence of

distorted helices L2 (residues 26–28) and L6 (residues

84–87), and absence of distorted helices L3 (residues 37–

39) and L7 (residues 98–101) which are found in 2.9 A

structure.

Reactive loop

The exposed reactive site loop of MKMLP (P4–P40)adopts a characteristic canonical conformation found in

classical Kunitz inhibitors like STI.10 The lower B-factors

and well defined electron density for residues of reactive

site loop were observed in 2.2 A structures indicating

conformational stability of the region. Like in 2.9 A

structure at pH 8.0, the reactive loop in both high reso-

lution structures exhibits a well defined electron density

with a typical canonical conformation. A well defined

unoccupied electron density at Asn64 confirms the pres-

ence of glycan moieties. Only MLPs possess this

Figure 1Overall structure of MKMLP. Part figures (a) and (c) represent cartoon models at pH 7 and 4.6 showing b-sheets, a-helices, loops, and putative

reactive loop indicated in yellow, red, blue, and magenta, respectively. The disulfide bridges are shown in stick in blue (C41–C85, C144–C147, andC140–151). The structures were submitted in PDB database (PDB ID code: 3ZC8 for pH 7 and 3ZC9 for pH 4.6 structures). Part figures (b) and

(d) represent the final electron density map around the region (Tyr63–His 70) in the MKMLP pH 7 and 4.6 structure, respectively. The map wascalculated using 50.56–2.24 A data and contoured at 1.0r. [Color figure can be viewed in the online issue, which is available at

wileyonlinelibrary.com.]

Insights into Aggregation Behavior of MKMLP

PROTEINS 833

glycosylation motif at active site loop and even miraculin

lacks the same. The superposition of Ca atoms of reac-

tive loop when compared with pH 7 and 8, pH 4.6 and

8, and pH 7 and 4.6 gave an RMSD of 0.329, 0.328, and

0.091, respectively [Fig. 3(d)]. The orientation of P2 resi-

due Asn64 is changed in both pH 7 and 4.6 structure

compared with pH 8 structure. Also there is slight

change in orientation of P30 residue Ile68 in pH 4.6

structure.

Crystallographic symmetry analysis

The pH 7 and 4.6 structures had single molecule in

the asymmetric unit whereas pH 8 structure had two.

The symmetry-related molecule interface area shows that

in all structures it involved region 145–152-KSCVFLCN

(Fig. 4). In pH 7 and 4.6 structures, the symmetry-

related molecules generated are within 4 A contain this

region. The loop region L7 (94–104) and a1 helix region

(115–122) are also in close contact with KSCVFLCN

(145–152) peptide region. The distance between two

Table IISecondary structural element details of MKMLP structures at pH 7 and 4.6 conditions

pH 7 MKMLP structure 4.6 MKMLP structure

Loops 1-16 L1 Loops 1-16 L124-30 L2 22-30 L235-43 L3 35-43 L348-57 L4 48-57 L463-74 L5 63-74 L580-89 L6 80-89 L6

94-104 L7 94-104 L7110-114 L8 110-114 L8128-134 L9 128-134 L9140-151 L10 140-151 L10162-163 L11 162-163 L11168-174 L12 168-175 L12

Sheets 17223

31234

44247

58262

9>>>>>=>>>>>;

Subdomain A

b1 Sheets 17221

31234

44247

58262

9>>>>>=>>>>>;

Subdomain A

b1b2 b2b3 b3b4 b4

75279

90293

1052109

1242127

9>>>>>=>>>>>;

Subdomain B

b5 75279

90293

1052109

1242127

9>>>>>=>>>>>;

Subdomain B

b5b6 b6b7 b7b8 b8

1352139

1522156

1642167

1752180

9>>>>>=>>>>>;

Subdomain C

b9 1352139

1522156

1642167

1762180

9>>>>>=>>>>>;

Subdomain C

b9b10 b10b11 b11b12 b12

Helices 115-122 a1 Helices 115-122 a1158-161 a2 158-161 a2

Disulfide bridges Cys41-Cys89, Disulfide bridges Cys41-Cys89,Cys140-Cys151 Cys140-Cys151Cys144-Cys147 Cys144-Cys147

Figure 2Structural superimposition of Ca atoms of three MKMLP structures.

The structures at pH 8, 7, and 4.6 showing similar overall fold are rep-resented in green, blue, and red ribbons, respectively. The glycine rich

loop undergoing conformational change is shown in box. [Color figure

can be viewed in the online issue, which is available atwileyonlinelibrary.com.]

P. Selvakumar et al.

834 PROTEINS

Trp103 residues in loop region (94–104) in adjacent sym-

metry molecules decreases with pH. In pH 8 structures,

the distance between two Trp103 residues in adjacent

symmetry molecules is 4.0 A, in pH 7 structure it is 3.7

A, and in pH 4.6 structure it is 3.5 A (Fig. 5).

DISCUSSION

Analysis of MKMLP aggregation

The pH-induced reversible aggregation of MKMLP is

unique among Kunitz family members. Most of the clas-

sical members, like STI, are stable at a broad range of

pH and temperature, and there are no reports about

aggregation at low pH. The reason for this unique

behavior may lie in the differences in the primary struc-

ture compared with other classical Kunitz members.

Many studies have revealed that certain peptides in pro-

tein sequences initiate or mediate aggregation. These

regions have also been successfully predicted. There must

be differences in the number of aggregation-prone

regions in MKMLP compared with other members. The

analysis of these regions in three-dimensional structure

determined in different conditions would seem to be a

reasonable approach for understanding the aggregation

behavior of MKMLP. For aggregation to happen there

should be more than one region for intermolecular con-

tact on a protein molecule. In the crystal structure deter-

mined at pH 4.6, the purified MKMLP used for

crystallization was in Tris buffer at pH 7.5 and precipi-

tant sodium acetate was at pH 4.6. The protein solution

and precipitant were mixed in equal proportions. There-

fore, structure represented here may not be a true struc-

ture at pH 4.6 but it is definitely a structure below pH 7.

Peptide sequences of MKMLP predicted tobe involved in aggregation

Six programs were used for prediction of aggregation-

prone regions of MKMLP. Since these programs employ

various parameters for prediction so there were subtle

differences in the results (Fig. S2, Supporting Informa-

tion). We have considered consensus from all the results

and have found four regions involved in aggregation dis-

tributed across the MKMLP sequence. These regions

involve peptides YYLVSVI (17–23), WFITTG (103–108),

SCVFLCN (146–152), and VFGVVIV (173–179). Tango

algorithm predicts the b-aggregation propensity in

sequences considering physico-chemical parameters such

as pH, temperature, and ionic strength. MKMLP was

assessed from pH 4 to 8. Among predicted regions by

Tango there are no charge residues. At pH 8, Tango pre-

dicts that three peptides, YYLVSVIG (17–24), WFITTGGV

(103–110), and VFGVVIVP (173–180) are involved in

aggregation. Prediction at pH 7, 6, 5, and 4 reveals

another aggregation region, CVFLC (147–151) with scores

highest at pH 6 and pH 5 (Fig. 6; and Table S1, Support-

ing Information). These results indicate increased aggrega-

tion tendency of MKMLP below pH 7.5.

Sequence comparisons reveal that other related mem-

bers lack predicted consensus aggregation-prone peptides

found in MKMLP sequence. Compared with MKMLP,

results for STI showed only two aggregation prone pep-

tide regions, TYYILS (16–20) and LKFDSFAVIMLCVG

(74–88). STI lacks the WFITTG (103–108) peptide and a

peptide containing hydrophilic amino acids (DDKCG) is

present in STI as compared with the corresponding pep-

tide SCVFLCN (146–152) in MKMLP (Fig. 7).

Structural insights into the aggregationbehavior of MKMLP

We hypothesize that the aggregation of MKMLP below

pH 7.5 is driven by the exposure of aggregation-prone

regions. Aggregation can be seen as an anomalous type

of protein–protein interaction. Tertiary protein structure

is governed mainly by electrostatic and hydrophobic

interactions and proteins are most likely to aggregate at

Figure 3Superimposition involving residues 17–30 and putative reactive loop inthree MKMLP structures at different pH exhibiting variation in second-

ary structure elements are shown. The structures at (a) pH 8, (b) pH 7,and (c) pH 4.6 are represented in green, blue, and red cartoons, respec-

tively. (d) Superimposition of putative reactive loop residues Ala62 (P4)to Ile69 (P40) of three MKMLP structures at different pH conditions

showing variation in conformation of Asn64 (P2) at pH 8 structure

(green) compared with pH 7 (blue) and pH 4.6 (red sticks). [Color fig-ure can be viewed in the online issue, which is available at

wileyonlinelibrary.com.]

Insights into Aggregation Behavior of MKMLP

PROTEINS 835

their isoelectric points, where they bear no net charges.

Hydrophobic interactions are the main mode through

which nonpolar patches of the protein surface are

shielded by water molecules arranged in an ordered

structure. When two nonpolar patches come together,

the water molecules are expelled, increasing their

entropy. This increase is the main driving force for pro-

tein association.29 Crystal structures of MKMLP deter-

mined at pH 8.0, 7.0, and 4.6 were analyzed for subtle

conformational differences that can provide clues about

aggregation below pH 7.5. Most of the predicted four

aggregation-prone peptides are involved in native b-sheet

formation, except for SCVFLCN (146–152) and residues

103–108 of WFITTG, which are in surface loops. In

MKMLP the predicted aggregation-prone peptides form

two patches/regions (Fig. 8). Patch 1 involves the pre-

dicted aggregation prone peptides YYLVSVI (17–23),

VFGVVIV (173–179), and the glycine rich loop region

GGAGGGG (24–30). Patch 2 involves peptides WFITTG

(103–108) and SCVFLCN (146–152). These two patches

are far from each other and present in slightly opposite

orientations with respect to each other. The data demon-

strated the potential of these peptides to form intermo-

lecular b-sheets that are not seen in native structure.

Most importantly SCVFLCN (146–152) showed b-

aggregation propensity only below pH 8, consistent with

the biochemical data showing decrease in solubility

below pH 7.5.11 The conformational changes in these

Figure 4Crystallographic symmetry-related molecular interface at different pH conditions involving KSCVFLCN (145–152). (a) At pH 8, (b) pH 7, and (c)pH 4.6. Black sticks represent KSCVFLCN (145–152) loop region which forms interface residues in all structures. [Color figure can be viewed in the

online issue, which is available at wileyonlinelibrary.com.]

Figure 5Crystallographic symmetry-related molecule interface showing distances between two Trp103 at different pH involving WFITTG (103–108). Loca-tion of Trp103, which is part of WFITTG (103–108), is shown in blue sticks. The distance between two symmetrical Trp103 of molecules gets

decreased from (a) pH 8, (b) pH 7, and (c) pH 4.6. The red loop region indicates SCVFLCN (146–152). [Color figure can be viewed in the online

issue, which is available at wileyonlinelibrary.com.]

P. Selvakumar et al.

836 PROTEINS

two regions and the resulting alteration in electrostatic

or hydrophobic interaction will be mainly responsible for

promoting aggregation below pH 7.5. The analysis of

crystal structures determined at three pH conditions

showed a pattern in conformational changes as the pH

was lowered. Conformational changes were observed at

aggregation prone peptides YYLVSVI (17–23), SCVFLCN

(146–152), and WFITTG (103–108). In Patch 1, the two

main structural alterations were observed around the

glycine-rich region involving the aggregation-prone pep-

tide YYLVSVI (17–23). First, the change in orientation of

glycine rich loop when pH 8 and 7, and pH 4.6 struc-

tures are compared (Fig. 3). Second, a shortening of b1

and b2 strands and lengthening of loop L2. The compar-

ison of the conformation in this region in three struc-

tures showed that b1 strand shortened from 18–25 at

pH 8.0 to 17–23 at pH 7.0 to 17–21 at pH 4.6 structure.

This observation implies that Val22 in aggregation-prone

peptide and Ile23 are not involved in native b-sheet for-

mation in the pH 4.6 structure and exists in loop region,

suggesting partial unfolding in region which, in turn,

exposes aggregation-prone peptide YYLVSVI (17–23).

Likewise, b2 strand is shortened in structures determined

at pH 7.0 and 4.6. Thus, the conformational changes at

low pH lead to partial unfolding, exposing aggregation-

prone regions and thereby promoting hydrophobic inter-

protein interactions. These interactions would promote

intermolecular associations leading to aggregation. In

Patch 2, there are changes in the interaction within the

molecule and at the interface of the crystallographic

symmetry-related molecule. The SCVFLCN (146–152)

peptide formed the interface in all structures (Fig. 4). In

pH 7 structure, Val148 forms one hydrogen bond with

water and Phe149 forms two hydrogen bonds with water.

Leu150 forms hydrogen bond with main chain of

His139. None of these interactions are seen in pH 4.6

structure [Fig. 9(a)]. Also, WFITTG (103–108) is in close

contact with this interface. Interesting results were seen

on analyzing the distances between two adjacent sym-

metrical Trp103 residues. The distance decreases when

compared with pH 8 to pH 4.6 structures (Fig. 5). Taken

together these results demonstrate the alterations that are

seen in pH 4.6 structure: ionic interactions are lost and

hydrophobic interactions become more prominent. These

changes are a prerequisite for intermolecular association.

The differences among the interactions involved in

two patches at pH 4.6 and 7 provided important

Figure 6Plots of b-aggregation propensity for MKMLP. At all pH values three

peptides are predicted to be aggregation prone. An additional peptide

CVFLC (147–151) is found only below pH 8. Peaks in brown (pH 8),green (pH 7.5), purple (pH 7), yellow (pH 6), and blue (pH 5) signifies

increase in b-aggregation score. [Color figure can be viewed in theonline issue, which is available at wileyonlinelibrary.com.]

Figure 7Multiple sequence alignment of MKMLP, miraculin, and STI. The predicted aggregation-prone regions in MKMLP are shown and underlined inred. Boxes indicate comparative critical amino acid sequence changes and deletions. MKMLP shows presence of more hydrophobic residues com-

pared with miraculin and STI. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Insights into Aggregation Behavior of MKMLP

PROTEINS 837

observations. For Patch 1, the pH 4.6 structure lost criti-

cal hydrogen bonding interactions compared with the

pH 7 structure. These include interactions between

Ileu23–Gly25, Gly175, Gly24–158Tyr, Gly25–Gln47, and

Gly28. Also certain hydrogen bonding interactions are

gained at pH 4.6 compared with pH 7. These include

interactions between Gly24–Gly175, Ala26–Arg162, and

Val179–Ala181. Superimposition of Ile23 at both pH

conditions reveals a flip of the carbonyl in the pH 4.6

structure. Due to this flip, Ile23 cannot hydrogen bond

to Gly175 because distance between them increases from

2.8 to 5.1 A [Fig. 9(b–d)]. For Patch 2, interactions with

water molecules involving the hydrophobic residues

Val148, Phe149, and Cys151 were lost in the pH 4.6

structure. The details of the above interactions are

described in the Table S2 of the Supporting Information.

The results suggest that partial unfolding and changes in

interactions are responsible for aggregation behavior.

Clearly, these alterations in the two patches results in

intermolecular contacts that lead to visible precipitation.

Peptide WFITTG (103–108) is unique to MLPs and may

be responsible for initiation, particularly W103, which

plays a crucial role in promoting hydrophobic interaction

between two molecules at low pH. When symmetry-

related molecules are analyzed the distance between adja-

cent W103 sidechains are decreased from 3.9 A at pH 8

to 3.5 A at pH 4.6. These signify dominance of hydropho-

bic interactions, and this slight shift in intramolecular

associations might trigger two similar molecules to associ-

ate leading to aggregation (Fig. 10). Also, the biological

assembly of pH 7 structure was predicted to be mono-

meric and pH 4.6 structure as dimeric by PQS, which

uses crystal symmetry matrices to generate symmetry-

related copies of the chains and by considering the buried

surface area between pairs of chains.

8-Anilino-1-naphthalene sulfonate (ANS) fluorescence

studies demonstrated a linear increase in fluorescence

intensity with increasing temperatures with no substan-

tial blue shift indicating that conformational changes

does not significantly expose hydrophobic pockets. How-

ever, there is sharp increase in fluorescence intensity

below pH 5 indicating relaxation.11,30 Strong evidences

exist that translocation of proteins across a variety of

membranes and membrane insertion involves non-native

or denatured states and sometimes molten globule struc-

tures.31 In some cases conformational changes are seen

at low pH conditions.32,33 The reversible nature of

aggregation can be explained by the presence of three

disulfide bonds in MKMLP one more than classical Kun-

tiz inhibitors, a unique conserved feature seen only in

MLPs. The two cysteines involved in the extra disulfide

are found in SCVFLCN (146–152) which borders the

exposed hydrophobic patch VFL residues. These two cys-

teines provide structural constraint and prevent further

aggregation by not allowing adjacent residues to form

intermolecular b-sheets. These features suggest that apart

from protease inhibition there is a possibility of addi-

tional functions linked to MKMLP associated with the

aggregation process because clearly there is a route to

prevent amyloid formation.

CONCLUSIONS

The overall three-dimensional structures of MKMLP

determined at 2.2 A resolution were similar to the earlier

Figure 9Ionic and hydrogen bonding interactions in MKMLP structures at pH 7

and 4.6. (a) Water molecule and ionic interactions of hydrophobic resi-

dues Val148, Phe149, and Leu150 residues of SCVFLCN (146–152) ofPatch 2 region of MKMLP at pH 7 are shown. Red spheres indicate

water molecules shown with electron density contoured at 1.0 r. (b)Superimposition of Ile23 residue of Patch 1 region for pH 7 and 4.6

show displacement of carboxyl group. (c) In pH 7 structure Ile23 formshydrogen bond to Gly175. (d) In pH 4.6 structure Ile23 cannot interact

with Gly175 as distance between them increased due to the flip. [Color

figure can be viewed in the online issue, which is available atwileyonlinelibrary.com.]

Figure 8Surface view of MKMLP structure showing location of two aggregationprone patches. Patch 1 region shown in blue includes residues 17–30

and 173–179. Patch 2 region shown in red consists of residues 103–108

and 146–152.

P. Selvakumar et al.

838 PROTEINS

reported structure at 2.9 A. However, a remarkable

improvement in the model was observed and certain key

structural details were revealed in the high resolution

structures. The presence of increased helix content (resi-

dues 115–122) is seen, which is consistent with earlier

circular dichroism studies. The presence of helices is a

unique feature found in MLPs as compared with classical

Kunitz inhibitors. Also, flexibility in the glycine rich

region (GGAGGGG—24–30) is evident where drastic

alteration in conformation is observed. The comparison of

crystal structures grown in three different pH conditions

(pH 8.0, 7.0, and 4.6) provided the structural basis for the

aggregation of MKMLP below pH 7.5. The analysis of

aggregation-prone regions revealed four aggregation-prone

peptides distributed in two patches present far apart on

the surface of the protein. The subtle pH-dependent con-

formational changes resulted in alterations to electrostatic

and hydrophobic interactions. A gradual exposure of

aggregation-prone peptides in two patches was observed.

In Patch 1, a partial unfolding due to the shortening of

b-strand of aggregation-prone peptide and change in ori-

entation of glycine-rich loop is observed in low pH struc-

ture. Comparison of the three crystal structures at the

symmetry-related molecular interface involving Patch 2,

revealed increased hydrophobic interactions due to the

juxtaposition closing in of two symmetry-related molecule

as the pH decreased. The distance between Trp103 in

aggregation-prone peptide WFITTG (103–108) in Patch 2

decreased when the symmetry-related molecular interface

was compared in three structures. The peptide WFITTG

(103–108) is unique to MLPs and, therefore, W103 may be

responsible for initiating aggregation. The results indicate

that the aggregation in MKMLP below pH 7.5 is a result

of subtle conformational change accompanied by alteration

in electrostatic and hydrophobic interaction, and thereby

exposing the aggregation-prone peptides in two patches as

pH decreases. The exposure of aggregation-prone regions

results into increased nonpolar intermolecular contacts

where Patch 1 and Patch 2 of one molecule interact with

Patch 1 and Patch 2 of adjacent molecules, respectively.

The aggregation results from the cumulative effect of these

hydrophobic interactions between the two defined respec-

tive patches among adjacent MKMLP molecules as pH is

lowered.

ACKNOWLEDGMENTS

The crystal structure analysis was performed at Macro-

molecular Crystallographic Unit, IIC at IIT Roorkee. We

acknowledge Ms Sonali Dhindwal for the technical help.

P. Selvakumar, N. Sharma, and P.P.S. Tomar thank, DBT

and CSIR, Government of India for financial support,

respectively.

REFERENCES

1. Stefani M, Dobson CM. Protein aggregation and aggregate toxicity:

new insights into protein folding, misfolding diseases and biological

evolution. J Mol Med 2003;81(11):678–699.

2. Tartaglia GG, Pawar AP, Campioni S, Dobson CM, Chiti F,

Vendruscolo M. Prediction of aggregation-prone regions in struc-

tured proteins. J Mol Biol 2008;380(2):425–436.

3. Colon W, Kelly JW. Partial denaturation of transthyretin is sufficient

for amyloid fibril formation in vitro. Biochemistry 1992;31(36):

8654–8660.

4. Calamai M, Chiti F, Dobson CM. Amyloid fibril formation can pro-

ceed from different conformations of a partially unfolded protein.

Biophys J 2005;89(6):4201–4210.

5. Kelly JW. The alternative conformations of amyloidogenic proteins

and their multi-step assembly pathways. Curr Opin Struct Biol

1998;8(1):101–106.

6. Theerasilp S, Kurihara Y. Complete purification and characterization

of the taste-modifying protein, miraculin, from miracle fruit. J Biol

Chem 1988;263(23):11536–11539.

7. Hirai T, Sato M, Toyooka K, Sun H-J, Yano M, Ezura H. Miraculin,

a taste-modifying protein is secreted into intercellular spaces in

plant cells. J Plant Physiol 2010;167(3):209–215.

Figure 10The ribbon diagram showing proposed MKMLP aggregation model involving Patch 1 (red) and Patch 2 (blue) regions involved in intermolecular

associations. A total of six MKMLP molecules are represented and aggregation-prone patches form the interface between two molecules. [Color fig-ure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Insights into Aggregation Behavior of MKMLP

PROTEINS 839

8. Selvakumar P, Gahloth D, Tomar PPS, Sharma N, Sharma AK.

Molecular evolution of miraculin-like proteins in soybean kunitz

super-family. J Mol Evol 2011;73(5-6):369–379.

9. Shee C, Sharma AK. Purification and characterization of a trypsin

inhibitor from seeds of Murraya koenigii. J Enzyme Inhib Med

Chem 2007;22(1):115–120.

10. Gahloth D, Selvakumar P, Shee C, Kumar P, Sharma AK. Cloning,

sequence analysis and crystal structure determination of a

miraculin-like protein from Murraya koenigii. Arch Biochem Bio-

phys 2010;494(1):15–22.

11. Shee C, Islam A, Ahmad F, Sharma AK. Structure-function studies

of Murraya koenigii trypsin inhibitor revealed a stable core beta

sheet structure surrounded by a-helices with a possible role for a-

helix in inhibitory function. Int J Biol Macromol 2007;41(4):410–

414.

12. Iwanaga S, Yamasaki N, Kimura M, Kouzuma Y. Contribution of

conserved Asn residues to the inhibitory activities of Kunitz-type

protease inhibitors from plants. Biosci Biotechnol Biochem 2005;

69(1):220–223.

13. Gahloth D, Shukla U, Birah A, Gupta GP, Ananda Kumar P,

Dhaliwal HS, Sharma AK. Bioinsecticidal activity of Murraya

koenigii miraculin-like protein against Helicoverpa armigera and

Spodoptera litura. Arch Insect Biochem Physiol 2011;78(3):132–

144.

14. Roychaudhuri R, Sarath G, Zeece M, Markwell J. Stability of the

allergenic soybean Kunitz trypsin inhibitor. Biochim Biophys Acta

2004;1699(1):207–212.

15. Azarkan M, Dibiani R, Goormaghtigh E, Raussens V, Baeyens-

Volant D. The papaya Kunitz-type trypsin inhibitor is a highly sta-

ble b-sheet glycoprotein. Biochim Biophys Acta 2006;1764(6):1063–

1072.

16. Shee C, Sharma AK. Storage and affinity properties of Murraya koe-

nigii trypsin inhibitor. Food Chem 2008;107(1):312–319.

17. Bailey S. The CCP4 suite: programs for protein crystallography.

Acta Crystallogr D Biol Crystallogr 1994;D50:760–763.

18. Emsley P, Cowtan K. Coot: model-building tools for molecular

graphics. Acta Crystallogr D Biol Crystallogr 2004;60(12):2126–

2132.

19. Laskowski RA, MacArthur MW, Moss DS, Thornton JM. PRO-

CHECK: a program to check the stereochemical quality of protein

structures. J Appl Crystallogr 1993;26(2):283–291.

20. Chen VB, Arendall WB, Headd JJ, Keedy DA, Immormino RM,

Kapral GJ, Murray LW, Richardson JS, Richardson DC. MolProbity:

all-atom structure validation for macromolecular crystallography.

Acta Crystallogr D Biol Crystallogr 2009;66(1):12–21.

21. DeLano WL. The PyMOL molecular graphics system. San Carlos,

CA: DeLano Scientific LLC; 2002.

22. Fernandez-Escamilla A-M, Rousseau F, Schymkowitz J, Serrano L.

Prediction of sequence-dependent and mutational effects on the

aggregation of peptides and proteins. Nat Biotechnol 2004;22(10):

1302–1306.

23. Garbuzynskiy SO, Lobanov MY, Galzitskaya OV. FoldAmyloid: a

method of prediction of amyloidogenic regions from protein

sequence. Bioinformatics 2010;26(3):326–332.

24. Oscar C-SO, de Groot NS, Aviles FX, Josep V, Xavier D, Salvador V.

AGGRESCAN: a server for the prediction and evaluation of “hot

spots” of aggregation in polypeptides. BMC Bioinf 2007;8(65).

25. Maurer-Stroh S, Debulpaep M, Kuemmerer N, de la Paz ML,

Martins IC, Reumers J, Morris KL, Copland A, Serpell L, Serrano L.

Exploring the sequence determinants of amyloid structure using

position-specific scoring matrices. Nat Methods 2010;7(3):237–242.

26. Tartaglia GG, Vendruscolo M. The Zyggregator method for predict-

ing protein aggregation propensities. Chem Soc Rev 2008;37(7):

1395–1401.

27. Trovato A, Seno F, Tosatto SCE. The PASTA server for protein

aggregation prediction. Protein Eng Des Sel 2007;20(10):521–523.

28. Henrick K, Thornton JM. PQS: a protein quaternary structure file

server. Trends Biochem Sci 1998;23(9):358.

29. Levy Y, Onuchic JN. Water mediation in protein folding and molec-

ular recognition. Annu Rev Biophys Biomol Struct 2006;35:389–415.

30. Patel GK, Shee C, Gahloth D, Selvakumar P, Sharma AK. Stability

of Murraya koenigii miraculin-like protein in different physicochem-

ical conditions. Med Chem Res 2011;20(9):1542–1549.

31. Bychkova VE, Pain RH, Ptitsyn OB. The ‘molten globule’ state is

involved in the translocation of proteins across membranes? FEBS

Lett 1988;238(2):231–234.

32. Van der Goot FG, Gonzalez-Manas JM, Lakey JH, Pattus F. A

‘molten-globule’ membrane-insertion intermediate of the pore-

forming domain of colicin A. Nature 1991;354:408–410.

33. Owers Narhi L, Philo JS, Li T, Zhang M, Samal B, Arakawa T.

Induction of a-helix in the b-sheet protein tumor necrosis factor-a:

acid-induced denaturation. Biochemistry 1996;35(35):11454–11460.

P. Selvakumar et al.

840 PROTEINS