Critical Regions for the Sweetness of Brazzein1 Provided by Elsevier - Publisher Connector

FEBS 27205 FEBS Letters 544 (2003) 33^37

View metadata, citation and similar papers at core.ac.uk brought to you by CORE Critical regions for the sweetness of brazzein1 provided by Elsevier - Publisher Connector

Zheyuan Jina, Vicktoria Danilovaa, Fariba M. Assadi-Porterb, DavidJ. Aceti b, John L. Markleyb,Go«ran Hellekanta;Ã aDepartment of Animal Health and Biomedical Sciences, University of Wisconsin^Madison, Madison, WI 53706, USA bDepartment of Biochemistry, University of Wisconsin^Madison, Madison, WI 53706, USA

Received25 February 2003; revised4 April 2003; accepted4 April 2003

First publishedonline 2 May 2003

Edited by Gianni Cesareni

Brazzein is very water-soluble. Its isoelectric point (pI = 5.4) Abstract Brazzein is a small, heat-stable, intensely sweet protein consisting of 54 amino acid residues. Based on the wild-type is lower than those of other sweet proteins, which all have brazzein, 25 brazzein mutants have been produced to identify pI s 7.0 [5]. Brazzein is remarkably heat-stable, andits sweet critical regions important for sweetness. To assess their sweet- taste remains after incubation at 80‡C for 4 h. Chemical mod- ness, psychophysical experiments were carried out with 14 hu- i¢cation studies suggested that the surface charge of the mol- man subjects. First, the results suggest that residues 29^33 and ecule is important andledto the conclusion that Arg, Lys, 39^43, plus residue 36 between these stretches, as well as the Tyr, His, Asp, andGlu are important for brazzein’s sweetness C-terminus are involved in the sweetness of brazzein. Second, andshouldbe studiedfurther [5,6]. In addition, studies with charge plays an important role in the interaction between braz- other sweet proteins, thaumatin andmonellin, have suggested zein and the sweet taste receptor. a role for lysine and/or carboxyl groups in the sweet protein^ ß 2003 Federation of European Biochemical Societies. Pub- receptor(s) interaction [7^9]. lished by Elsevier Science B.V. All rights reserved. The structure of brazzein was determined by 1H nuclear Key words: Brazzein; Sweet protein; Mutagenesis; magnetic resonance (NMR) spectroscopy in solution at pH Sweetness determinant; Sweetness evaluation; Taste 5.2 and22‡C [10]. The study revealed that brazzein contains one short K-helix (residues 21^29) and three strands of anti- parallel L-sheet (strand I, residues 5^7; strand II, residues 44^ 50; strand III, residues 34^39) held together by four disul¢de 1. Introduction bonds. The authors proposed that the small connecting loop containing His31 andthe randomcoil loop aroundArg43 Six sweet proteins have been discovered over the last 30 were the possible determinants of the molecule’s sweetness. years. The latest discovered is brazzein, isolated from the fruit Site-directed mutagenesis was used to change surface resi- of Pentadiplandra brazzeana Baillon [2]. Brazzein is a single- dues of WT brazzein at di¡erent locations along its sequence chain polypeptide of 54 amino acid residues with four intra- [11]; andtaste studiesof 14 brazzein variants showedthat molecular disul¢de bonds, no free sulfhydryl group, and no most mutations decreased sweetness but that two increased carbohydrate [3]. It is rich in lysine but contains no methio- sweetness. On the basis of these studies, the authors suggested nine, threonine or tryptophan. Brazzein exists in two forms in that the N- andC-termini and L-turn aroundArg43 are in- the ripe fruit. The major form contains pyroglutamate (pGlu) volvedin the sweetness of brazzein [1]. at its N-terminus; the minor form is without the N-terminal Here we present results from a larger set of brazzein var- pGlu (des-pGlu1). Taste comparisons of chemically synthe- iants, which includes multiple mutations at several speci¢c sizedbrazzein anddes-pGlu1brazzein revealedthat the latter positions, aimedat delineatingfurther how changes of charges protein has about twice the sweetness of the former [4].We andsidechains a¡ect the sweetness of brazzein. The ultimate use ‘WT (wild type) brazzein’ to denote recombinant des- goal, unrealizedas yet, is to predictsweetness from the chem- pGlu1 brazzein. ical structure of a compound, or from the amino acid sequence of a sweet protein. Because New Worldprimates andother mammals donot perceive brazzein as sweet, inves- *Corresponding author. Fax: (1)-608-262 7420. tigations of its sweetness must be carriedout with humans or E-mail address: [email protected] (G. Hellekant). OldWorldprimates. This investigation describesandanalyzes 1 For brazzein mutants, the same nomenclature system was followed the e¡ects of site-directed mutations of brazzein on its sweet- as in the previous publication [Assadi-Porter et al., Arch. Biochem. ness as perceivedby human subjects. Biophys. 376 (2000) 259^265]. For example, Asp29Ala indicates the Asp of residue 29 was replaced by Ala; Arg19Ile20ins is a double insertion with Arg insertedat position 19 andIle insertedat 2. Materials and methods position 20; mutant Ala2Ala31 indicates the additive mutations of Ala2ins andHis31Ala. 2.1. Preparation of brazzein variants A protein expression system for WT brazzein, developed in Esche- Abbreviations: CNBr, cyanogen bromide; ddH2O, doubly deionized richia coli, allowedbrazzein variants to be producede⁄ciently [1]. water; MALDI, matrix-assistedlaser desorption/ionization; NMR, Brazzein mutants were preparedby site-directedmutagenesis(Quick nuclear magnetic resonance; pGlu, pyroglutamic acid; SCM, single- Change1 PCR kit, Stratagene) for substitutions of speci¢c amino chain monellin; SNase, staphylococcal nuclease; WT, wildtype; WT acids based on the template gene encoding WT and expressed as a brazzein, recombinant des-pGlu1 brazzein fusion protein (staphylococcal nuclease (SNase)^Met^brazzein). The

FEBS 27205 21-5-03 Cyaan Magenta Geel Zwart 34 Z. Jin et al./FEBS Letters 544 (2003) 33^37 brazzein molecule was releasedfrom the fusion protein at the unique verbal labels: ‘barely detectable’, ‘weak’, ‘moderate’, ‘strong’, ‘very Met linkage by CNBr (cyanogen bromide) cleavage, then puri¢ed by strong’, and‘the strongest imaginable’. The intensity of the sweetness cation exchange chromatography, followedby reverse-phase high-per- was later convertedto a numerical value. The datawere averaged formance liquidchromatography puri¢cation to remove all unfolded between subjects, and standard deviations were calculated. Sweetness or mis-folded brazzein proteins [11]. scores were ¢rst evaluatedwith repeatedmeasurements ANOVA Because brazzein contains no tryptophan, its extinction coe⁄cient (analysis of variance) followedby pairwise comparisons of the scores was determined at 205 nm instead of 280 nm. The extinction coe⁄- for di¡erent variants using Fisher’s least signi¢cant di¡erences. Prob- cient O205 of each brazzein variant was calculatedfrom measurements ability P 9 0.05 was considered to be signi¢cant. of the absorbance of solutions at 280 and205 nm accordingto the O1mg=ml formula [12]: 205 = 27.0+120(A280/A205). All protein solutions were scannedat 195^290 nm at mediumspeed 3. Results with a Cary Win UV Scan spectrophotometer (Varian Analytical Instruments, Walnut Creek, CA, USA). Doubly deionized water 3.1. Characterization of brazzein variants (ddH2O) was usedas the blank, andthe baseline was subtracted 1 automatically. Samples were diluted by factors that provided absor- Fig. 1 shows the 1D H NMR spectra of two of the most bance values of about 0.7 at 205 nm; higher absorbances were sweet brazzein variants (Asp29Ala, Asp29Lys) andthree avoided to minimize e¡ects of stray light. The concentrations of braz- of the least sweet brazzein variants (Glu36Ala, Glu36Lys, zein solutions were determined from the absorbance at 205 nm and Glu36Gln). Each spectrum exhibits peaks at low frequency the extinction coe⁄cient for the particular variant. The molecular weights of the brazzein variants were con¢rmedwith (around0.4 ppm from methyl groups in the protein core) matrix-assistedlaser desorption/ionization (MALDI) mass spectrom- andat high frequency (around10.1 ppm from the hydroxyl etry. Samples were run on a Bruker Bi£ex III MALDI time-of-£ight proton of Tyr11) that are characteristic for folded protein. mass spectrometer in the linear mode using K-cyano-4-hydroxycin- Similar results obtainedfor the other mutants indicatedthat namic acidas the matrix. The external calibration was performed they too were correctly folded. with bovine insulin and ubiquitin as standards. One-dimensional (1D) 1H NMR spectroscopy was usedto check that the mutants were folded correctly. Each brazzein protein (0.5^1.0 3.2. Human evaluations 2 1 mM) was dissolved in 10% H2O at pH 7.0, and1D H NMR spectra Fig. 2 shows the average sweetness scores of all stimuli were acquiredwith a Bruker DMX 500 MHz spectrometer in a 5 mm used. In Fig. 2 we use black columns to indicate signi¢cantly 1H probe at 298 K. increasedsweetness, gray columns for no change, striped 2.2. Stimuli and sensory analysis columns for signi¢cantly less sweet than WT brazzein, and WT brazzein, brazzein variants, monellin, andsingle-chain monellin white columns for no di¡erence from water. The bars denote (SCM) were dissolved in ddH2O. The solutions were adjusted to pH S.E.M. 7.0 with 0.1 M NaOH or HCl. The concentrations of monellin and Four brazzein mutants (Asp29Ala, Asp29Lys, Asp29Asn, SCM were determined spectrophotometrically using a calculated mo- 31 31 andGlu41Lys) were scoredsigni¢cantly sweeter than WT lar extinction coe⁄cient O280 of 15 930 M cm [13] anda mass of 11 197 Da for monellin and11 050 Da for SCM, respectively (SWISS- brazzein. Three brazzein mutants (Ala2ins, Asp2Asn, and PROT). Gln17Ala) were scoredas sweet as WT brazzein. In eight The taste panel consistedof six females andeight males (ages 17^70 mutants the sweetness decreased signi¢cantly but the score years) of reportedgoodhealth andnormal sense of taste. The exper- imental protocol was approvedby the University of Wisconsin^Mad- was di¡erent from that of water, and in 10 mutants the sweet- ison Human Subjects Committee. The subjects tasted26 brazzeins, ness score did not di¡er signi¢cantly from that of water. two monellins as positive controls, anddeionizedwater as the nega- To visualize the structural correlations for these results, we tive control in double blind experiments. The sample volume was 150 have indicated the position of each mutation on the three- W W 31 l. All proteins were usedat a concentration of 100 gml . The dimensional backbone of brazzein (Fig. 3). One dramatic solutions were delivered with a micropipet to the anterior part of the subject’s tongue. The subject tastedthe compoundwithout any time change in sweetness occurredwith mutants at or near constraints, then expectorated, followed by ad lib. rinsing with tap Asp29: mutation of Asp29 to Ala, Lys or Asn gave signi¢- water within a 1 min interval. Each subject testedeach stimulus three cantly increasedsweetness, while mutation of His31 to Ala or times, and the presentations were randomized. Arg33 to Asp reduced sweetness; mutating Lys30 to Asp or Between each presentation the subjects were askedto score the sweetness of the stimulus on a LabeledMagnitudeScale (LMS) Arg33 to Ala yielded compounds tasting not signi¢cantly dif- [14]. This scale is a semantically labeledscale, which we usedfor ferent from water. Glu41Lys gave the highest score of sweet- rating the intensity of a taste sensation. The scale is composedof ness; however, Arg43Ala hadno sweetness. Notably, the mu-

Asp29Lys

Asp29Ala

Glu36Gln

Glu36Lys

Glu36Ala WT 0.6 0.5 0.4 0.3 0.2 10.2 10.0 9.8 9.6 9.4 9.2 9.0 Chemical Shift (ppm) Chemical Shift (ppm)

1 2 Fig. 1. 1D H NMR spectra (500 MHz) of WT brazzein and¢ve representative mutants in 10% H2O at pH 7.0 and25‡C. All spectra exhibit features indicative of proper protein folding, including the expected chemical shift dispersion with peaks present at high and low frequency.

FEBS 27205 21-5-03 Cyaan Magenta Geel Zwart Z. Jin et al./FEBS Letters 544 (2003) 33^37 35

35 Strong 30

Moderate 15

Sweetness Score 10

Weak 5

Barely Detectable 0 SCM Water Ala2ins Tyr8Ala Lys5Ala Lys6Ala Monellin Lys6Asp Arg19ins Asp2Asn Tyr54del Glu36Ala Lys15Ala Gln17Ala His31Ala Arg43Ala Arg33Ala Glu36Lys Glu36Gln Glu41Lys Asp29Ala Asp50Ala Arg33Asp Asp29Lys Lys30Asp Ala2Ala31 Asp29Asn Brazzein WT Arg19Ile20ins Fig. 2. Results of psychophysical experiments with brazzein mutants andmonellins. Data were averagedfor 14 subjects. Error bars are S.E.M. Column patterns indicate di¡erent levels of sweetness in comparison with WT brazzein: black, signi¢cantly sweeter than WT brazzein; gray, not di¡erent from WT brazzein; both striped and white, signi¢cantly less sweet than WT brazzein. White columns also indicate that the scores were not di¡erent from that of water. tations at the N-terminus of brazzein (Ala2ins, Asp2Asn) 4. Discussion brought no changes in sweetness. Furthermore, substitutions of Glu36 to Ala, Lys, or Gln all abolishedsweetness. In ad- The present study included a considerable number of mu- dition, mutation of Lys6 to Asp, insertion of Arg at position tants from our earlier study [1] but useda di¡erentmethodof 19, doubleinsertions of Arg andIle at position 19, anddele- sensory evaluation. In the following we discuss the possible tion of Tyr at the C-terminus all yielded compounds with structure^sweetness relationship andthe sensory methodsap- taste indistinguishable from water. pliedin this andthe earlier study.The current set of mutants

Fig. 3. Diagram showing the three-dimensional backbone of brazzein [10] with the position of mutations studied. The residues are color-coded to indicate the taste properties of mutants at these positions relative to those of WT brazzein: red, increased sweetness; black, the same sweetness; light blue, decreased sweetness; dark blue, taste equivalent to water. Intramolecular disul¢de bonds are indicated as yellow lines.

FEBS 27205 21-5-03 Cyaan Magenta Geel Zwart 36 Z. Jin et al./FEBS Letters 544 (2003) 33^37 is discussed in the context of the three-dimensional structure in psychophysical experiments. Methods for such evaluations of brazzein [10]. Because brazzein’s foldis constrainedby four in humans range from simple tasting at the ‘lab bench’ to disul¢de bridges and is so thermostable, it is likely that the quantitative measurements of gustatory sensation described structure determined at pH 5.2 and 22‡C provides a valid in a number of textbooks andscienti¢c articles [14,17,18]. structural model for brazzein at pH 7.0 and 37‡C (the con- Two di¡erent methods to evaluate sweetness on brazzein mu- ditions of the sensory evaluations). tants were usedhere andin our previous study.In our earlier The two brazzein mutants with the largest increase in sweet- study we used a stepwise scale to measure sweetness and drew ness were Glu41Lys andAsp29Lys, while mutations in the conclusions from the observedthresholdconcentrations. Here immediate vicinity of these regions essentially rendered braz- we useda well-establishedmethodof sensory analysis [14] zein tasteless. Thus, the mutations that dramatically changed that allowedus to convert semantic expression into a contin- the sweetness of brazzein are localizedwithin two regions uous function andworkedwith above-thresholdconcentra- (AspLysHisAlaArg29À33 andTyrAspGluLysArg 39À43). This tions. Furthermore, as has been demonstrated with the sweet suggests that these are the critical regions of the molecule proteins thaumatin andmonellin, sweetness of high-potency for eliciting sweetness. These two regions have been previously compounds does not increase linearly as with sucrose, but suggestedto be important for its sweetness as determined rather non-linearly, asymptotically approaching maximal re- from a comparison of the NMR solution structure of brazzein sponse [19]. This complicates comparison of the current re- with taste analyses of 14 brazzein variants [1,10]. In the three- sults with those from our earlier study, even though 11 of the dimensional structure of brazzein, the region containing variants studied were the same. Arg33 is close to residues Tyr54 and Tyr51 (particularly the In future studies of brazzein, it will be interesting to inves- aromatic ring) and Arg33 itself is hydrogen-bonded to Asp50 tigate the e¡ects of changing Lys30, His31, andLys42 to [15]. This shows that the region containing Arg33 is in close neutral, negative, or positively chargedresidues.In fact, braz- contact with the C-terminus. Additionally, deletion of Tyr54 zein is a su⁄ciently small protein so that all of its amino abolishedthe sweetness of brazzein almost completely [1]. acids, except for the cysteines, which are involved in disul¢de Overall, these data indicate that the C-terminus is a necessary bridges, have some surface exposure. It may be worthwhile to component for sweetness in brazzein. systematically mutate all residues with surface exposure to Mutations of the negatively chargedAsp29 residue,to either thoroughly study the structure^activity relationships. Interest- a neutral or positively chargedresidue(Asp29Ala, Asp29Asn, ing mutants couldbe labeledwith 15N and/or 13C for detailed Asp29Lys), all markedly increased sweetness. Similar types of NMR analysis of possible structural changes. As a by-product mutations performedat the Glu36 site (Glu36Ala, Glu36Gln, of these studies, brazzein variants identi¢ed to have enhanced Glu36Lys) all decreased the sweetness to the level of no taste. sweet qualities could become candidates for a new generation This suggests that at these sites, charge is important for elicit- of low-caloric natural sweeteners. ing sweetness, whereas the length or orientation of the side In summary, our results suggest a multi-point interaction chain plays a lesser role. between brazzein andits receptor in which charge plays a Research on thaumatin suggestedthat the positive charge signi¢cant role. Our ¢ndings also suggest that residues 29^ of lysine residues is important in the determination of its 33 and residues 39^43, plus residue 36 in the peptide connect- sweetness [7]. Furthermore, recent computer modeling inves- ing these stretches, as well as the C-terminus are involvedin tigations of interactions between the sweet taste receptor and determining the sweetness of brazzein. sweet proteins point to the importance of electrostatic com- plementarity between the protein andreceptor [16]. The au- Acknowledgements: We thank Dr. Qin Zhao for help in collecting 1D thor proposedthat sweet proteins activate the T1R2^T1R3 NMR data. We express our sincere gratitude to Monsanto (St. Louis, MO, USA) for their support. receptor by interacting with the free form II (active state) of the receptor andstabilizing it. Most of the preferredbinding solutions for the docking of SCM with the receptor were References centeredon a large cavity of the T1R3 extracellular domain, which has an average negative charge. Complementary to it, [1] Assadi-Porter, F.M., Aceti, D.J. and Markley, J.L. (2000) Arch. SCM has a positive surface. Similar docking studies were Biochem. Biophys. 376, 259^265. [2] Ming, D. andHellekant, G. (1994) FEBS Lett. 355, 106^108. performedon computer modelsof thaumatin andbrazzein, [3] Ming, D., Markley, J.L. andHellekant, G. (1995) Pept. Res. 8, andagain it was foundthat the surface of the sweet protein 113^114. interacting with the receptor is predominantly positive [16]. [4] Izawa, H., Ota, M., Kohmura, M. andAriyoshi, Y. (1996) Bio- This prediction is consistent with the present results, which polymers 39, 95^101. [5] Ming, D., Hellekant, G. andZhong, H. (1996) Acta Bot. Yun- show that changing the negative Asp29 to a neutral (Ala, Asn) nancia 18, 123^133. or positive (Lys) residue increased sweetness, with Asp29Lys [6] Ming, D. (1994) Ph.D. thesis, 150 pp., University of Wisconsin at exhibiting the largest e¡ect. Introduction of a positive charge Madison. at another site in brazzein (Glu41Lys) also greatly increased [7] Kaneko, R. andKitabatake, N. (2001) Chem. Senses 26, 167^ the sweetness. These results suggest that charge plays an im- 177. [8] Slootstra, J.W., De Geus, P., Haas, H., Verrips, C.T. andMe- portant role in controlling whether brazzein is perceivedas loen, R.H. (1995) Chem. Senses 20, 535^543. sweet or tasteless. Because its pI is so low (5.4), brazzein [9] Kohmura, M., Nio, N. andArioshi, Y. (1992) Biosci. Biotech. may have a higher potential than the other sweet proteins Biochem. 56, 1937^1942. for engineering enhancedsweetness through the introduction [10] Caldwell, J.E., Abildgaard, F., Dzakula, Z., Ming, D., Hellekant, G. andMarkley, J.L. (1998) Nat. Struct. Biol. 5, 427^431. of positive charges at critical sites. [11] Assadi-Porter, F.M., Aceti, D.J., Cheng, H. and Markley, J.L. To elucidate the e¡ects of mutations on brazzein sweetness, (2000) Arch. Biochem. Biophys. 376, 252^258. it was important that accurate sensory analysis be carriedout [12] Scopes, R.K. (1974) Anal. Biochem. 59, 277^282.

FEBS 27205 21-5-03 Cyaan Magenta Geel Zwart Z. Jin et al./FEBS Letters 544 (2003) 33^37 37

[13] Pace, C.N., Vajdos, F., Fee, L., Grimsley, G. and Gray, T. (1995) taste Chemoreception (Mathlouthi, M., Kanters, J.A. andBirch, Protein Sci. 4, 2411^2423. G.G., Eds.), pp. 237^267, Elsevier Applied Science, London. [14] Green, B.G., Dalton, P., Cowart, B., Sha¡er, G., Rankin, K. and [18] Bartoshuk, L.M. (2000) Chem. Senses 25, 447^460. Higgins, J. (1996) Chem. Senses 21, 323^334. [19] DuBois, G.E. et al. (1991) in: Sweeteners: Discovery, Molecular [15] DeRider, M.L. (2001) Ph.D. thesis, 135 pp., University of Wis- Design, andChemoreception (Walters, D.E., Orthoefer, F.T. and consin at Madison. DuBois, G.E., Eds.), pp. 261^276, American Chemical Society, [16] Temussi, P. (2002) FEBS Lett. 526, 1^4. Washington, DC. [17] DuBois, G.E., Walters, D.E. andKellogg, M.S. (1991) in: Sweet-

FEBS 27205 21-5-03 Cyaan Magenta Geel Zwart