Do guanidinium and tetrapropylammonium ionsspecifically interact with aromatic amino acidside chains?Bei Dinga,b, Debopreeti Mukherjeea, Jianxin Chena,b, and Feng Gaia,b,1

aDepartment of Chemistry, University of Pennsylvania, Philadelphia, PA 19104; and bUltrafast Optical Processes Laboratory, University of Pennsylvania,Philadelphia, PA 19104

Edited by Michael L. Klein, Temple University, Philadelphia, PA, and approved December 16, 2016 (received for review November 1, 2016)

Many ions are known to affect the activity, stability, and structuralintegrity of proteins. Although this effect can be generally attributedto ion-induced changes in forces that govern protein folding,delineating the underlying mechanism of action still remains chal-lenging because it requires assessment of all relevant interactions,such as ion–protein, ion–water, and ion–ion interactions. Herein, weuse two unnatural aromatic amino acids and several spectroscopictechniques to examine whether guanidinium chloride, one of themost commonly used protein denaturants, and tetrapropylammo-nium chloride can specifically interact with aromatic side chains.Our results show that tetrapropylammonium, but not guanidinium,can preferentially accumulate around aromatic residues and that tet-rapropylammonium undergoes a transition at ∼1.3 M to form aggre-gates. We find that similar to ionic micelles, on one hand, suchaggregates can disrupt native hydrophobic interactions, and on theother hand, they can promote α-helix formation in certain peptides.

guanidinium | tetrapropylammonium | Hofmeister ions |unnatural amino acid | 2D IR

The stability of a protein, or more precisely, the free energydifference between its folded and unfolded states, can bemodulated by various solution properties. For example, addition ofanother solute to a protein solution can result in either a decreaseor an increase in this protein’s stability (1–3). The most noticeableexample in this regard is the Hofmeister series (4–6), a group ofions that are ranked based on their protein-denaturing abilities.Among this series, the guanidinium ion (Gdm+) is the most widelyused chemical agent in protein denaturation due to its strongdestabilizing effect and high solubility in water. Consequently, itsmechanism of action has been subjected to extensive studies (7–14).Although different interpretations have been put forth (15–17), thegenerally accepted notion is that Gdm+ denatures a protein bypreferentially interacting with its peptide groups (7, 11, 18), in-cluding certain side chains (11, 19–22). In particular, it has beenhypothesized that Gdm+, which exists in aqueous solution as a rigid,flat object (12, 18, 19), can engage in stacking interactions withamino acids consisting of planar side chains, such as arginine (Arg)(19), asparagine (Asn) (11, 20), glutamine (Gln) (11, 20), and ar-omatic residues (20–22). However, to the best of our knowledge,the only experimental evidence that supports this hypothesis comesfrom crystallographic data (20–22), which show that the guanidi-nium group of Arg is more frequently found to be stacked againstthe side chain of tyrosine (Tyr) or tryptophan (Trp) in proteins.Therefore, additional experimental studies that can directly probesuch stack interactions in solution are needed.Another ion in the Hofmeister series that has a similar protein-

denaturing capability to Gdm+ is tetrapropylammonium (TPA+)(23). However, a recent study by Dempsey et al. (24) found thatalthough TPA+, like Gdm+, can denature a tryptophan (Trp) zip-per β-hairpin (i.e., trpzip1) at a sufficiently high concentration(>1.0 M), it stabilizes this β-hairpin at lower concentrations andalso the α-helical conformation of an alanine-based peptide. Thisobservation led them to conclude that a TPA+ ion, which has four

roughly flat faces, can interact favorably with planar aromatic sidechains but is excluded from interacting with peptide backbone dueto its large size (24). Although the finding that TPA+ can denaturetrpzip1 is consistent with an earlier molecular dynamics (MD)simulation (23), which revealed that TPA+ has a longer residencetime on the indole ring of Trp than on other nonaromatic sidechains in the peptide melittin, the stabilizing actions of TPA+ re-main to be explained (24). In addition, results obtained with di-electric spectroscopy (25) and ultrafast infrared (IR) spectroscopy(26) on a series of tetra-n-alkylammonium salts indicated that ionsconsisting of long alkyl chains, such as TPA+, can cause the motionof nearby water molecules to slow down, owing to their largerhydrophobic surfaces and hence the so-called hydrophobic hydra-tion. However, it is not clear how such hydration property of TPA+

would contribute to its interaction with different protein moieties.Although previous studies (23, 24) have provided many insights

on the protein denaturation mechanisms of Gdm+ and TPA+ ions,their molecular actions have not been directly probed from theperspective of the protein. In this regard, it would be quite advan-tageous to devise an experimental approach that is capable of re-vealing information on specific interactions of interest, such asinteractions between Gdm+/TPA+ ions and a specific amino acidside chain. To achieve this goal, herein we use multiple spectroscopicmethods and two unnatural amino acids, p-cyano-phenylalanine(PheCN) and 5-cyano-tryptophan (TrpCN), to directly assesswhether protein aromatic side chains can preferentially interactwith Gdm+ and TPA+ ions. Both the C≡N stretching vibration(27, 28) and fluorescence quantum yield (29, 30) of PheCN andTrpCN have been shown to be sensitive to their local environment.Thus, the premise of our study is that any specific ion–side chain

Significance

Ever since the discovery by Franz Hofmeister in 1888 that a seriesof salts (now commonly referred to as the Hofmeister series) canhave different but consistent effects on the solubility and hencestability of proteins, many studies have been devoted to un-derstanding how such effects are achieved. However, examiningany specific mode of ion–protein interactions has proven to bedifficult. Herein, we demonstrate, using guanidinium and tetra-propylammonium as examples, that it is possible to use multiplespectroscopic methods and unnatural amino acids that havedistinct, environment-sensitive fluorescence and infrared prop-erties to explicitly assess how these ions interact with a specifictype of amino acid side chains, thus yielding microscopic insightsinto their protein-denaturing mechanisms.

Author contributions: B.D. and F.G. designed research; B.D., D.M., and J.C. performedresearch; B.D. analyzed data; and B.D., D.M., and F.G. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. Email: gai@sas.upenn.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1618071114 PNAS | January 31, 2017 | vol. 114 | no. 5 | 1003–1008

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

June

30,

202

1

http://crossmark.crossref.org/dialog/?doi=10.1073/pnas.1618071114&domain=pdfmailto:gai@sas.upenn.eduhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplementalwww.pnas.org/cgi/doi/10.1073/pnas.1618071114

interaction will manifest itself as a change in the local hydrationand electrostatic environment of the spectroscopic probe used,which can be assessed by either infrared (IR) or fluorescencespectroscopy. Several previous studies (27, 31–33) have demon-strated that 2D IR spectroscopy is ideally suited to study suchchanges because it is capable of reporting the frequency–fre-quency correlation function of the IR vibration in question andhence the dynamics of motions (arising from both hydration andprotein conformational heterogeneities) that contribute to theoverall bandwidth of the IR spectrum. Our results show thatcontrary to the conventional expectation, Gdm+ does not specif-ically stack on aromatic side chains. On the other hand, consistentwith molecular dynamics simulations (23), TPA+ exhibits a strongaffinity toward aromatic side chains, and at ∼1.3 M, TPA+ ag-gregates to form clusters that provide an interfacial region, which,similar to that found in lipid membranes and micelles (34), canhelp anchor Trp side chains and also promote α-helix formation inalanine-based peptides, due to dehydration and hence a lowerdielectric environment.

Results and DiscussionTo access whether Gdm+ and TPA+ can specifically interact withthe two aforementioned aromatic unnatural amino acid sidechains (i.e., PheCN and TrpCN) in a polypeptide environment, wefirst examined the dynamics of the C≡N stretching vibration oftwo peptides, Gly-PheCN-Gly and Gly-TrpCN-Gly (hereafter re-ferred to as GFCNG and GWCNG, respectively), in the presenceand absence of Gdm+ and TPA+ ions using Fourier transforminfrared (FTIR) and 2D IR spectroscopies. We chose thesepeptide sequences to ensure the full exposure of the respectivearomatic side chains to solvent and to avoid any side chain–sidechain interactions. The latter could complicate interpretation ofthe spectroscopic results. We then further verified our finding onTrpCN–TPA

+ interactions using a longer peptide with the fol-lowing sequence: SWCNTAENGKATWCNK (hereafter referredto as 2WCNP). In addition, we also used an alanine-based pep-tide (sequence: GKAAAAKWCNAAAWCNKAAAAKG) and anantimicrobial peptide (sequence: INWKGIAAMAKKLL), bothof which can be induced to form α-helix by TFE, to test the helix-inducing ability of TPA+ ions. Wherever applicable, we also car-ried out fluorescence and circular dichroism (CD) measurementsto substantiate the IR results.

Linear and 2D IR Studies. As shown (Fig. 1 and SI Appendix, TableS1 and Fig. S1), the FTIR spectra of GFCNG and GWCNGobtained under different solvent conditions indicate that bothGdm+ and TPA+ (at high concentrations) affect the C≡Nstretching vibrations of PheCN and TrpCN in these peptides. Spe-cifically, in 6.0 M GdmCl solution the C≡N stretching bandsof both peptides show a small but measurable red-shift (i.e.,∼1.6 cm−1 for GFCNG and ∼1.3 cm−1 for GWCNG) in comparisonwith their respective peak frequencies in pure water. In addition,the width of these bands shows a slight increase (

For an oscillator that is capable of forming hydrogen bonds withwater, such as the CN group in the current case, the exponentialcomponent in Eq. 1 is typically attributed to the dynamics of hy-drogen bonds formed between the oscillator and water molecules(27, 31–33). As shown (SI Appendix, Table S1), for both GFCNGand GWCNG, the major effect of 6.0 M GdmCl is manifested by amodest increase in the offset B value, which, according to previousstudies (31), could be attributed to the increase in viscosity. Thus,this finding is consistent with the linear IR results and also theabovementioned notion that GdmCl does not increase in any sig-nificant manner the inhomogeneity of the environment sampled byPheCN and TrpCN in these peptides. On the contrary, the effect of2.0 M TPACl on the spectral diffusion dynamics of the C≡Nstretching vibration of both peptides is much more pronounced (SIAppendix, Table S1), with a substantial increase in the B value inboth cases. This result indicates that at a concentration of 2.0 M,TPA+ ions can effectively slow down certain dynamic motions inboth peptides that contribute to the spectral diffusion dynamics ofthe C≡N stretching vibration and hence its IR bandwidth. Thus,this finding corroborates the conclusion reached above that TPA+

ions specifically accumulate around aromatic side chains, likelythrough the stacking interactions suggested by Mason et al. (23).A closer comparison of the spectral diffusion dynamics of the

C≡N probes in GFCNG and GWCNG obtained in 2.0 M solution ofTPACl further indicates that TPA+ ions affect the hydration statusof PheCN and TrpCN differently (Fig. 3 and SI Appendix, Fig. S3). Itis apparent that for GFCNG, interacting with TPA

+ ions does notresult in any significant change in the hydrogen-bonding compo-nent of the spectral diffusion dynamics, whereas for GWCNG, in-teractions between TrpCN and TPA

+ ions essentially eliminate thispicosecond dynamic component. In other words, these results showthat interaction with TPA+ ions causes the TrpCN side chain to bemore dehydrated than that of PheCN, a picture that is consistentwith the MD simulation of Mason et al. (23), which suggested thatTPA+ ions interact more strongly with the indole ring of Trp thanwith other hydrophobic groups in the antimicrobial peptidemelittin. Despite this stronger interaction, however, based on the

2D IR results alone we cannot rule out the possibility that thereare water molecules surrounding the TrpCN side chain in thepresence of 2.0 M TPA+. Nonetheless, these water molecules, ifany, should be less mobile than bulk water and hence do notcause the C≡N stretching frequency to fluctuate on the timescaleof our experiment, as observed in other exmples (32, 39).

Fluorescence Studies. To help better understand the results obtainedwith linear and nonlinear IR spectroscopic methods, we also per-formed fluorescence measurements. It has been shown that thefluorescence quantum yields of PheCN and TrpCN are sensitive tointeractions with water molecules (29, 30), with a relation thathydration increases/decreases the fluorescence intensity of PheCN/TrpCN. Thus, both unnatural amino acids can in principle be usedas fluorescence probes to assess how Gdm+ and TPA+ affect thelocal hydration status of PheCN and TrpCN in GFCNG and GWCNG.However, we only carried out fluorescence experiment on GWCNGbecause Cl− is known to significantly quench the fluorescence ofPheCN (29). As shown (Fig. 4, Inset), the TrpCN fluorescence in-tensity of GWCNG in 6.0 M GdmCl solution is slightly lower thanthat in pure H2O. However, addition of 2.0 M TPACl leads to asignificant increase in the fluorescence intensity of TrpCN (∼9.7times) in comparison with that obtained in pure H2O. Taken to-gether, these results indicate that TPA+ ions (at 2.0 M), but notGdm+ ions (at 6.0 M), show preferential interactions with TrpCN(and hence Trp), resulting in either partial or complete dehydrationof its side chain, as concluded from the IR experiments.

Concentration-Dependent Studies. To gain further insight into theinteraction between TPA+ and aromatic side chains, we carried outsimilar linear and nonlinear IR studies on GFCNG and GWCNG insolutions of different molar concentrations of TPACl ([TPACl]).As expected, for both peptides, increasing [TPACl] leads to a de-crease/increase in the frequency/bandwidth of the C≡N stretchingvibration (SI Appendix, Table S3 and Figs. S4–S6), which is ac-companied by an increase in the static component in the respectiveC≡N spectral diffusion dynamics (SI Appendix, Table S3). In ad-dition, although the hydrogen-bonding dynamics component ofGFCNG does not show a significant dependence on [TPACl], thesame component of GWCNG becomes practically undetectablewithin our experimental uncertainty when [TPACl] is increased toabout 1.5 M (SI Appendix, Figs. S7–S10). As indicated (SI Appendix,

Fig. 2. Representative 2D IR spectra of GWCNG in the C≡N stretching fre-quency region obtained under different solvent conditions and at differentwaiting times, as indicated.

0.0

0.2

0.4

0.6

0.0 1.0 2.0 3.0 4.0 5.0

Waiting Time, T (ps)

CLS

-1

H2O 6.0 M GdmCl 2.0 M TPAClH2

Fig. 3. CLS−1 versus T plots of the C≡N stretching vibration of GWCNG obtainedunder different solvent conditions, as indicated.

Ding et al. PNAS | January 31, 2017 | vol. 114 | no. 5 | 1005

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

June

30,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdf

Fig. S11), a further fluorescence titration experiment reveals thatthe fluorescence intensity of TrpCN in GWCNG exhibits an unusualdependence on [TPACl]—it increases linearly as a function of[TPACl] with one slope in the range of 0–1.3 M but a steeper slopebetween 1.3 and 2.0 M (Fig. 4). This indicates that there is a distincttransitional event occurring at about 1.3 M. A dielectric spectros-copy study by Buchner et al. (25) on tetrapropylammonium bro-mide (TPABr) aqueous solutions showed that the dispersionamplitude of the water trapped in the hydration shell of TPA+ ionsincreases with [TPABr] and reaches a maximum at 1.3 M, which,based on neutron scattering experiments (40), was attributed toTPA+ aggregation taking place at concentrations above this criticalvalue. Similarly, the study of Bakker and coworkers (26) on thedynamics of water surrounding tetra-n-alkylammonium ions withlong alkyl chains also suggested ion aggregation. Thus, taking thoseprevious findings into consideration, our IR and fluorescence re-sults provide direct evidence that aggregated and nonaggregatedTPA+ ions have different effects on the hydration dynamics of theside chain of TrpCN and, perhaps more importantly, that aggre-gated TPA+ ions are more effective in dehydrating TrpCN (andhence Trp). The latter is reminiscent of the phenomenon that Trpprefers the interfacial region of lipid membranes (41) and associ-ation with this region leads to dehydration of its indole moiety (42).Thus, this result suggests that above the critical concentration of1.3 M, TPA+ ions aggregate to form clusters that exhibit a distinctwater-hydrophobic interface. Furthermore, considering the findingof Dempsey et al. (24) that TPA+ shows a denaturing effect towardtrpzip1 only when its concentration is higher than 1 M, we believethat the aggregated and nonaggregated TPA+ ions have differenteffects on protein hydrophobic interactions and that the protein-denaturing ability of TPA+ arises from its aggregated clusters.

Effect on a Longer Peptide. To further validate the findings obtainedwith GWCNG, we used the same approach to study a longer pep-tide, 2WCNP. As shown (SI Appendix, Table S4 and Fig. S12), thelinear IR spectroscopic properties of the C≡N stretching vibrationof 2WCNP obtained in 6.0 M GdmCl and 2.0 M TPACl solutionsare almost identical to those of GWCNG obtained under the samesolvent conditions. However, the C≡N stretching vibrational band

of 2WCNP in pure water is broader than (by ∼1.1 cm−1) and red-shifted from (by ∼0.8 cm−1) that of GWCNG. This suggests that incomparison with GWCNG, the longer 2WCNP peptide can sample alarger conformational space, as expected. In fact, the 2D IRspectra of 2WCNP in pure water (SI Appendix, Fig. S13) show thatthere are two resolvable spectral features within the profile of thelinear C≡N stretching band, at ∼2,225 (strong) and ∼2,217 cm−1(weak), indicative of two distinct population ensembles. As shown(SI Appendix, Fig. S14), the CD spectrum of 2WCNP in pure waterconsists of not only a broad, negative-going band at 198 nmexpected for a disordered peptide but also a weak CD band at∼245 nm. Previously, we have shown that in a well-folded peptidestructure, a pair of nearby TrpCN residues can produce a strong CDcouplet at this wavelength due to excitonic coupling (43). Thus, theCD spectrum of 2WCNP is not only consistent with the 2D IRresults but also suggests that the minor population ensemble cor-responds to a more compact conformation wherein the two aro-matic side chains are in close proximity. In 6.0 M GdmCl solutionthis excitonic CD band only becomes weaker, which supports thenotion that Gdm+ ions do not specifically and strongly interact withthe side chain of TrpCN. One the other hand, this excitonic CDband is not detectable in 2.0 M TPA solution, suggesting, onceagain, preferential binding of TPA+ to TrpCN side chains. Whatis more, the CD spectrum of 2WCNP in 2.0 M TPACl solution iscomposed of two negative-going peaks, centered at ∼205 and∼225 nm, suggesting that TPA+ ions promote formation of(partial) α-helical structures. This finding is consistent with thestudy of Dempsey et al. (24), which showed that 2 M TPAClpromotes α-helix formation in an alanine-based peptide.As shown (Fig. 5), the spectral diffusion dynamics of the C≡N

stretching vibration of 2WCNP in pure water exhibits a similar hy-drogen-bonding kinetic phase comparing to that of GWCNG;however, the corresponding static component is much larger. Thisis well expected for a longer peptide consisting of different sidechains because a greater portion of the conformational dynamicscontributing to the linear IR bandwidth will occur on a timescalethat is too slow to be explicitly time resolved. As similarly observedfor GWCNG, addition of 2.0 M TPACl to 2WCNP solution essen-tially abolishes any observable hydrogen-bonding dynamics andalso substantially increases the static component in the spectraldiffusion dynamics (Fig. 5). Thus, these results provide confirma-tion that TPA+ ions have a strong affinity toward Trp and theunderlying Trp–TPA+ interactions lead to dehydration of the in-dole ring. This notion is further substantiated by the fact that theTrpCN fluorescence intensity of 2WCNP in 2.0 M TPACl solution is∼10 times larger than that in pure water (SI Appendix, Fig. S15).Addition of 6.0 M GdmCl results in a decrease in the static com-ponent of the spectral diffusion dynamics of the C≡N stretchingvibration of 2WCNP (Fig. 5). This result has two implications: (i) itsupports the notion that Gdm+ ions do not strongly interact withprotein aromatic side chains and (ii) it shows that as observed in aprevious study (31), Gdm+ reduces the distribution of peptideconformational states that interconvert on a timescale too slow tobe captured by 2D IR measurements.

Implication on the Protein Denaturizing Mechanisms of TPA+ andGdm+. Because of its enormous importance in chemistry and bi-ology, the effect of ions on the activity and structural integrity ofproteins has been the subject of extensive studies (5). For a specificion, its mechanism of action is often assessed by examining itsability to alter the hydrogen-bonding structure and dynamics ofwater, its affinity toward specific amino acid side chains, and itsinteraction with protein backbone units. For Gdm+ and TPA+, ithas been suggested that they denature proteins by directly inter-acting with various protein components (7, 11, 18–24). By takingadvantage of two amino acid-based spectroscopic probes andmultiple spectroscopic techniques, herein we assess whether theseions can alter the hydration and conformational dynamics of

0.0

0.4

0.8

1.2

0.0 0.7 1.4 2.1

0.0

0.4

0.8

1.2

325 425 525

[TPACl] (M)

Nor

mal

ized

Inte

nsity

Wavelength (nm)

Nor

mal

ized

Inte

nsity H2O GdmCl TPAClH2

Fig. 4. Normalized TrpCN fluorescence intensity (integrated area) of GWCNGas a function of TPACl concentration. The lines represent linear regressionsof the data obtained below [TPACl] = 1.2 M and above [TPACl] = 1.4 M.These two lines cross at a TPACl concentration of ∼1.3 M. Shown in the insetare the TrpCN fluorescence spectra of GWCNG obtained under different sol-vent conditions, as indicated.

1006 | www.pnas.org/cgi/doi/10.1073/pnas.1618071114 Ding et al.

Dow

nloa

ded

by g

uest

on

June

30,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfwww.pnas.org/cgi/doi/10.1073/pnas.1618071114

aromatic side chains in a peptide environment. Taken together, ourresults show that Gdm+ does not strongly bind to aromatic sidechains and hence is more likely to denature proteins through in-teractions with the backbone. Of course, as a salt, GdmCl can alsoscreen electrostatic interactions that stabilize a protein’s nativestate. On the other hand, TPA+ does interact favorably with aro-matic side chains, especially the indole ring of Trp. Furthermore,TPA+ exhibits a behavior common to detergents or surfactants (44)as it undergoes a distinct transition when the concentration reachesa critical value (∼1.3 M) to form aggregates or clusters. Althoughthe current study cannot reveal the structural characteristics ofsuch TPA+ clusters, we believe that their mode of interaction withproteins is, at least to some extent, similar to that of ionic surfac-tants. This is because similar to the surfactant SDS, which denaturesglobular proteins by forming micelles and produces a denatured-state rich in α-helices (44), TPA+ clusters are able to disrupt thenative hydrophobic interactions among aromatic side chains in aβ-hairpin, causing it to unfold (24), and at the same time, areable to promote α-helical structure formation in an alanine-based peptide (24). Despite these similarities, clusters formedby TPA+ ions are expected to be different from SDS micelles inmany aspects. This is because the polar group of TPA+, unlikethat of SDS or other surfactant molecules, is situated at thecenter of the molecular ion and hence renders it impossible toyield a membrane-like environment as SDS does. Indeed, ourCD results (SI Appendix, Figs. S16 and S17) show that althoughTPA+ clusters can induce α-helix formation in an alanine-basedpeptide, as observed by Dempsey et al. (24), they do not increasethe α-helicity of a Trp3/TrpCN mutant of the antimicrobial peptidemastoparan-X that is known to form an amphipathic α-helix uponbinding to the interfacial region of lipid micelles or membranes (30).These results suggest that unlike membrane-mimic micelles formedby surfactants, a single TPA+ cluster is unable to provide a con-tinuous water-hydrophobic interface large enough to host an entireα-helix. Instead, the helix-promoting ability of TPA+ clusters islikely achieved by interacting with aromatic and hydrophobic sidechains in the peptide, which brings the corresponding backboneunit(s) to a relatively low dielectric environment and hence increasesthe possibility of local, backbone hydrogen bond formation (45). Thispicture is consistent with the fact that the α-helix–promoting abilityof TPA+ clusters depends on the peptide sequence. For an alanine-rich α-helix, most side chains are not only identical but also quiteuniformly and symmetrically distributed along the helical axis. This

uniform and symmetrical distribution of TPA+-interacting hydro-phobic side chains allows multiple TPA+ clusters to be accumulatedaround the peptide, which, collectively, yields a channel-like envi-ronment favorable for α-helix folding. However, for a peptide withmany different side chains, although simultaneous interactions withmultiple TPA+ clusters are still feasible, there is no guarantee thatthese clusters are adjacent to each other, thus reducing the possibilityof forming a continuous membrane-like environment across theentire peptide and hence α-helix formation. Finally, it is worth notingthat unlike clusters formed by protein-protecting cosolvents (46) thatare excluded from the protein surface, the excluded-volume orcrowding effect of TPA+ clusters is more localized and, as a result,will not increase the protein stability through this entropic cause.

ConclusionsThere are numerous examples in the literature documenting andinvestigating the effect of ions on the physical and/or chemicalproperties of proteins and, among which, studies on protein-denaturing ions, such as Gdm+, are especially prevalent due to theirimportant utility in protein science. Despite previous efforts, how-ever, we still lack a molecular-level understanding of the protein-denaturing mechanism of many ions, especially organic ions. This isbecause such ions can, in principle, engage in specific interactionswith protein side chains, an aspect that is difficult to assess experi-mentally. Herein, we use two unnatural amino acids, PheCN andTrpCN, both of which have been shown to be useful as site-specificIR and fluorescence probes of protein local environment, andmultiple spectroscopic techniques to examine whether Gdm+ andTPA+ specifically interact with aromatic side chains in a peptideenvironment. Our 2D IR results show no evidence of this interactionfor Gdm+ because it does not change the spectral diffusion dynamicsof the site-specific IR probes in any significant manner at the con-centration of 6.0 M. Thus, as suggested by other studies (7, 9, 11,47), Gdm+ most likely denatures proteins by binding to theirbackbone units or charged side chains. On the other hand, TPA+

does exert a strong effect on the spectral and dynamic properties ofthose probes and, at 2.0 M, causes TrpCN to be dehydrated.Moreover, our fluorescence data show that TPA+ undergoes atransition in ionic aggregation state at ∼1.3 M, akin to the micelli-zation behavior of surfactants. Taken together, these findings sug-gest a protein-denaturing action, similar to that found for ionicmicelles, wherein TPA+ clusters disrupt native hydrophobic inter-actions by accumulating around hydrophobic side chains. In addi-tion, as observed for micelles and membranes, the TPA+ clustersappear to have an interfacial region that promotes α-helix formationin certain peptides and also provides an anchoring position for theside chain of TrpCN (and hence Trp). Furthermore, considering theimportance and complexity of the ion–protein interaction problem,we believe that the current study demonstrates an approach of usingsite-specific vibrational and fluorescent probes in different proteinmoieties to help tease out relevant molecular information aboutspecific interactions between the ion and side chain of interest.Currently, we are using different unnatural amino acid-based IRprobes and 2D IR spectroscopy to investigate whether Gdm+ ionsspecifically interact with other types of amino acid side chain.

Materials and MethodsAll peptides used in the current study were synthesized using standard9-fluorenylmethoxy-carbonyl (Fmoc) methods. FTIR spectra were collected ona Thermo Nicolet 6700 FTIR spectrometer. Fluorescence spectra were col-lected on a Jobin Yvon Horiba Fluorolog 3.10 spectrofluorometer. CD spectrawere measured on an Aviv 62A DS spectropolarimeter (Aviv Associates). The2D IR spectra were collected on a home-built 2D IR apparatus in a boxcargeometry with heterodyne detection. The details of the sample preparationand all spectroscopic measurements are given in SI Appendix.

ACKNOWLEDGMENTS. We gratefully acknowledge financial support fromthe National Institutes of Health (Grant P41-GM104605).

0.0

0.3

0.6

0.9

0.0 1.0 2.0 3.0 4.0 5.0

Waiting Time, T (ps)

CLS

-1

H2O 6.0 M GdmCl 2.0 M TPACl

T (ps)

Fig. 5. CLS−1 versus T plots of the C≡N stretching vibration of 2WCNPobtained under different solvent conditions, as indicated.

Ding et al. PNAS | January 31, 2017 | vol. 114 | no. 5 | 1007

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

June

30,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdfhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1618071114/-/DCSupplemental/pnas.1618071114.sapp.pdf

1. Hua L, Zhou R, Thirumalai D, Berne BJ (2008) Urea denaturation by stronger disper-sion interactions with proteins than water implies a 2-stage unfolding. Proc Natl AcadSci USA 105(44):16928–16933.

2. Canchi DR, García AE (2013) Cosolvent effects on protein stability. Annu Rev PhysChem 64:273–293.

3. Ganguly P, Hajari T, Shea JE, Van Der Vegt NF (2015) A mutual exclusion of urea andtrimethylamine N-oxide from amino acids in mixed solvent environment. J Phys ChemLett 6(4):581–585.

4. Zhang Y, Cremer PS (2010) Chemistry of Hofmeister anions and osmolytes. Annu RevPhys Chem 61:63–83.

5. Lo Nostro P, Ninham BW (2012) Hofmeister phenomena: An update on ion specificityin biology. Chem Rev 112(4):2286–2322.

6. Xie WJ, Gao YQ (2013) A simple theory for the Hofmeister series. J Phys Chem Lett4(24):4247–4252.

7. Makhatadze GI, Privalov PL (1992) Protein interactions with urea and guanidiniumchloride: A calorimetric study. J Mol Biol 226(2):491–505.

8. Mayo SL, Baldwin RL (1993) Guanidinium chloride induction of partial unfolding inamide proton exchange in RNase A. Science 262(5135):873–876.

9. Monera OD, Kay CM, Hodges RS (1994) Protein denaturation with guanidine hydro-chloride or urea provides a different estimate of stability depending on the contri-butions of electrostatic interactions. Protein Sci 3(11):1984–1991.

10. Schellman JA (2003) Protein stability in mixed solvents: A balance of contact in-teraction and excluded volume. Biophys J 85(1):108–125.

11. O’Brien EP, Dima RI, Brooks B, Thirumalai D (2007) Interactions between hydrophobicand ionic solutes in aqueous guanidinium chloride and urea solutions: Lessons forprotein denaturation mechanism. J Am Chem Soc 129(23):7346–7353.

12. Lim WK, Rӧsgen J, Englander SW (2009) Urea, but not guanidinium, destabilizesproteins by forming hydrogen bonds to the peptide group. Proc Natl Acad Sci USA106(8):2595–2600.

13. Sharp KA, Vanderkooi JM (2010) Water in the half shell: Structure of water, focusingon angular structure and solvation. Acc Chem Res 43(2):231–239.

14. England JL, Haran G (2011) Role of solvation effects in protein denaturation: Fromthermodynamics to single molecules and back. Annu Rev Phys Chem 62:257–277.

15. Vanzi F, Madan B, Sharp K (1998) Effect of the protein denaturants urea and gua-nidinium on water structure: A structural and thermodynamic study. J Am Chem Soc120(41):10748–10753.

16. Batchelor JD, Olteanu A, Tripathy A, Pielak GJ (2004) Impact of protein denaturantsand stabilizers on water structure. J Am Chem Soc 126(7):1958–1961.

17. Scott JN, Nucci NV, Vanderkooi JM (2008) Changes in water stucture induced by theguanidinium cation and implications for protein denaturation. J Phys Chem A 112(43):10939–10948.

18. Mason PE, Neilson GW, Dempsey CE, Barnes AC, Cruickshank JM (2003) The hydrationstructure of guanidinium and thiocyanate ions: Implications for protein stability inaqueous solution. Proc Natl Acad Sci USA 100(8):4557–4561.

19. Mason PE, et al. (2004) The structure of aqueous guanidinium chloride solutions. J AmChem Soc 126(37):11462–11470.

20. Chakrabarti P, Bhattacharyya R (2007) Geometry of nonbonded interactions involvingplanar groups in proteins. Prog Biophys Mol Biol 95(1-3):83–137.

21. Burley SK, Petsko GA (1986) Amino-aromatic interactions in protein. FEBS Lett 203(2):139–143.

22. Flocco MM, Mowbray SL (1994) Planar stacking interactions of arginine and aromaticside-chains in proteins. J Mol Biol 235(2):709–717.

23. Mason PE, et al. (2009) Specificity of ion-protein interactions: Complementary andcompetitive effects of tetrapropylammonium, guanidinium, sulfate, and chlorideions. J Phys Chem B 113(10):3227–3234.

24. Dempsey CE, Mason PE, Jungwirth P (2011) Complex ion effects on polypeptideconformational stability: Chloride and sulfate salts of guanidinium and tetrapropy-lammonium. J Am Chem Soc 133(19):7300–7303.

25. Buchner R, Hölzl C, Stauber J, Barthel J (2002) Dielectric spectroscopy of ion-pairingand hydration in aqueous tetra-n-alkylammonium halide solutions. Phys Chem ChemPhys 4(11):2169–2179.

26. van der Post ST, Scheidelaar S, Bakker HJ (2013) Water dynamics in aqueous solutionsof tetra-n-alkylammonium salts: Hydrophobic and coulomb interactions disen-tangled. J Phys Chem B 117(48):15101–15110.

27. Ma J, Pazos IM, Gai F (2014) Microscopic insights into the protein-stabilizing effect oftrimethylamine N-oxide (TMAO). Proc Natl Acad Sci USA 111(23):8476–8481.

28. Zhang W, Markiewicz BN, Doerksen RS, Smith AB, III, Gai F (2016) C≡N stretchingvibration of 5-cyanotryptophan as an infrared probe of protein local environment:What determines its frequency? Phys Chem Chem Phys 18(10):7027–7034.

29. Pazos IM, Roesch RM, Gai F (2013) Quenching of p-cyanophenylalanine fluorescenceby various anions. Chem Phys Lett 563:93–96.

30. Markiewicz BN, Mukherjee D, Troxler T, Gai F (2016) Utility of 5-cyanotryptophanfluorescence as a sensitive probe of protein hydration. J Phys Chem B 120(5):936–944.

31. Chung JK, Thielges MC, Fayer MD (2011) Dynamics of the folded and unfolded villinheadpiece (HP35) measured with ultrafast 2D IR vibrational echo spectroscopy. ProcNatl Acad Sci USA 108(9):3578–3583.

32. Ghosh A, Qiu J, DeGrado WF, Hochstrasser RM (2011) Tidal surge in the M2 protonchannel, sensed by 2D IR spectroscopy. Proc Natl Acad Sci USA 108(15):6115–6120.

33. King JT, Kubarych KJ (2012) Site-specific coupling of hydration water and proteinflexibility studied in solution with ultrafast 2D-IR spectroscopy. J Am Chem Soc134(45):18705–18712.

34. Mukherjee S, Chowdhury P, Gai F (2006) Tuning the cooperativity of the helix-coiltransition by aqueous reverse micelles. J Phys Chem B 110(24):11615–11619.

35. Pazos IM, Gai F (2012) Solute’s perspective on how trimethylamine oxide, urea, andguanidine hydrochloride affect water’s hydrogen bonding ability. J Phys Chem B116(41):12473–12478.

36. Getahun Z, et al. (2003) Using nitrile-derivatized amino acids as infrared probes oflocal environment. J Am Chem Soc 125(2):405–411.

37. Woys AM, et al. (2010) 2D IR line shapes probe ovispirin peptide conformation anddepth in lipid bilayers. J Am Chem Soc 132(8):2832–2838.

38. Kwak K, Park S, Finkelstein IJ, Fayer MD (2007) Frequency-frequency correlationfunctions and apodization in two-dimensional infrared vibrational echo spectroscopy:A new approach. J Chem Phys 127(12):124503.

39. Zhang L, Yang Y, Kao YT, Wang L, Zhong D (2009) Protein hydration dynamics andmolecular mechanism of coupled water-protein fluctuations. J Am Chem Soc 131(30):10677–10691.

40. Polydorou NG, Wicks JD, Turner JZ (1997) Hydrophobic interaction of tetrapropy-lammonium ions in water: A neutron diffraction and reverse Monte Carlo study.J Chem Phys 107(1):197–204.

41. de Planque MRR, et al. (1999) Different membrane anchoring positions of tryptophanand lysine in synthetic transmembrane α-helical proteins. J Biol Chem 274(30):20839–20846.

42. de Jesus AJ, Allen TW (2013) The role of tryptophan side chains in membrane proteinanchoring and hydrophobic mismatch. Biochim Biophys Acta 1828(2):864–876.

43. Mukherjee D, Gai F (2016) Exciton circular dichroism couplet arising from nitrile-derivatized aromatic residues as a structural probe of proteins. Anal Biochem 507:74–78.

44. Otzen D (2011) Protein-surfactant interactions: A tale of many states. Biochim BiophysActa 1814(5):562–591.

45. Honig BH, Hubbell WL, Flewelling RF (1986) Electrostatic interactions in membranesand proteins. Annu Rev Biophys Chem 15:163–193.

46. Hilaire MR, Abaskharon RM, Gai F (2015) Biomolecular crowding arising from smallmolecules, molecular constraints, surface packing, and nano-confinement. J PhysChem Lett 6(13):2546–2553.

47. Meuzelaar H, Panman MR, Woutersen S (2015) Guanidinium-induced denaturation bybreaking of salt bridges. Angew Chem Int Ed 54(50):1–6.

1008 | www.pnas.org/cgi/doi/10.1073/pnas.1618071114 Ding et al.

Dow

nloa

ded

by g

uest

on

June

30,

202

1

www.pnas.org/cgi/doi/10.1073/pnas.1618071114