Protein Stability and Structure in HIC: Hydrogen Exchange

Nov 30, 2010 - Unfolding of a model protein on ion exchange and mixed mode chromatography surfaces. Adrian M. Gospodarek , Diana E. Hiser , John P...
1 downloads 0 Views 2MB Size
pubs.acs.org/Langmuir © 2010 American Chemical Society

Protein Stability and Structure in HIC: Hydrogen Exchange Experiments and COREX Calculations Adrian M. Gospodarek, Marissa E. Smatlak, John P. O’Connell, and Erik J. Fernandez* Department of Chemical Engineering, University of Virginia, Charlottesville, Virginia 22904-4741, United States Received September 21, 2010. Revised Manuscript Received November 15, 2010 Hydrogen exchange mass spectrometry (HXMS) coupled to proteolytic digestion has been used to probe the conformation of bovine β-lactoglobulin (BLG), bovine R-lactalbumin (BLA), and human serum albumin (HSA) in solution and while adsorbed to the hydrophobic interaction chromatography media Phenyl Sepharose 6FF. All three proteins show evidence of EX1 exchange kinetics, indicating a loss of stability on the surface. HX protection patterns for all three proteins also indicate that the unfolded form is only partially solvent exposed. The hydrogen-deuterium exchange patterns of BLG and BLA on the surface suggest a structure that resembles each protein’s respective solution phase molten globule state. The low stability of Domain II of HSA observed on Phenyl Sepharose 6FF also suggests a link to solution stability because Domain II is frequently cited as the least stable domain in solution unfolding pathways. COREX, an algorithm used to compute protein folding stabilities, correctly predicts solution hydrogen-deuterium exchange patterns for BLG and offers insight into its adsorbed phase stabilities but is unreliable for BLA predictions. The results of this work demonstrate a link between solution-phase local stability patterns and the nature of partially unfolded states that proteins can adopt on HIC surfaces.

1. Introduction Hydrophobic interaction chromatography (HIC) is a valuable purification tool that provides a means of separating proteins based on their apparent hydrophobicities.1-3 HIC stationary phases have a variety of backbone and hydrophobic ligand chemistries leading to differing degrees of hydrophobicity and protein binding capabilities. Industrial applications are driven toward the more hydrophobic HIC resins because of their ability to bind effectively proteins with lower salt concentration requirements. However, these same resins can lead to greater protein unfolding and compromised recoveries.4-6 HIC resins target the hydrophobic amino acids of proteins during the adsorption process, many of which tend to be found within the interior of the protein, or its hydrophobic core. This interaction can cause a change in protein conformation, and if significant and irreversible, protein function and therapeutic efficacy can be lost. Furthermore, yield losses can also occur, even when unfolding is reversible, because unfolded molecules will elute differently from the folded molecules.7,8 Designing HIC processes that optimize selectivity while preventing protein unfolding remains a challenge for many proteins, especially considering the many variables that influence adsorption, including resin, salt type and concentration, pH, and temperature.9

*To whom correspondence should be addressed. E-mail: erik@virginia. edu. (1) Queiroz, J. J. Biotechnol. 2001, 87, 143–159. (2) Hjerten, S. J. Chromatogr., A 1973, 87, 325–331. (3) Mccue, J. T. Methods Enzymol. 2009, 463, 405–414. (4) Ingraham, R.; Lau, S.; Taneja, A.; Hodges, R. J. Chromatogr., A 1985, 327, 77–92. (5) Kato, Y.; Kitamura, T.; Hashimoto, T. J. Chromatogr., A 1984, 298, 407– 418. (6) Fausnaugh, J.; Kennedy, L.; Regnier, F. J. Chromatogr., B 1984, 311, 141– 155. (7) Wu, S. W.; Figueroa, A.; Karger, B. L. J. Chromatogr., A 1986, 371, 3–27. (8) Benedek, K. J. Chromatogr., A 1988, 458, 93–104. (9) Gagnon, P.; Grund, E.; Lindb€ack, T. BioPharm 1995, 8, 1–9.

286 DOI: 10.1021/la103793r

Although processing windows are often found, it may be that none exists. Therefore, reliable predictive approaches will be valuable in both, finding optimal process conditions as well as identifying cases where proteins are too unstable for HIC. Furthermore, if protein unfolding occurs, then it might be another process variable for manipulating retention and selectivity.10 One option for anticipating when and how proteins will unfold on surfaces is to consider their solution-phase folding stability. Many proteins are known to unfold in solution via a two-step process through a folding intermediate whether the unfolding is thermal or with denaturant, such as urea or guanidine-HCl.11 Such intermediates, characterized by regions of preserved native structure, represent local minima in Gibbs energy in the folding pathway of proteins.12,13 The structures in solution of many of these intermediates have been determined by techniques such as NMR and CD.14,15 There is preliminary evidence that single domain proteins may unfold only partially on hydrophobic chromatography surfaces.16,17 This phenomenon may be even more prevalent for multidomain proteins. In this work, we hypothesize that regions of preserved structure in intermediates found in solution offer insight into the regions that are less prone to unfold on HIC surfaces. Then, structural information could be gained without adsorption or chromatography experiments. Another approach to predicting protein stability on HIC surfaces involves conformational modeling. COREX,18 a statistical representation of the (10) Lindahl, L.; Vogel, H. Anal. Biochem. 1984, 140, 394–402. (11) Foss, J. G.; Schellman, J. A. J. Phys. Chem. 1959, 63, 2007–2012. (12) Kim, P. S.; Baldwin, R. L. Annu. Rev. Biochem. 2003, 51, 459–489. (13) Bryngelson, J. D.; Wolynes, P. G. J. Phys. Chem. 1989, 93, 6902–6915. (14) van Mierlo, C. P.; Steensma, E. J. Biotechnol. 2000, 79, 281–298. (15) Fanali, G.; De Sanctis, G.; Gioia, M.; Coletta, M.; Ascenzi, P.; Fasano, M. JBIC, J. Biol. Inorg. Chem. 2009, 14, 209–217. (16) Mcnay, J.; Fernandez, E. J. Chromatogr., A 1999, 849, 135–148. (17) Engel, M. F.; Visser, A. J.; van Mierlo, C. P. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 11316–11321. (18) Hilser, V. J.; Freire, E. J. Mol. Biol. 1996, 262, 756–772.

Published on Web 11/30/2010

Langmuir 2011, 27(1), 286–295

Gospodarek et al.

Article

protein conformation ensemble, has been developed to predict the stability of protein in solution on the basis of the protein crystal structure and empirical parametrizations of Gibbs energy. COREX provides residue-level stability predictions that can be correlated with experimental hydrogen-deuterium exchange rates. COREX has successfully predicted residue-level stability patterns in solution for several proteins including staphylococcal nuclease, HEWL, and equine lysozyme.18 In this work, we also hypothesize that, regardless of pathway, solution-phase local stability patterns of proteins are similar to patterns of partial unfolding on hydrophobic chromatography surfaces. Then, computational models, such as COREX, may provide insight into regions of proteins that are unstable and more prone to unfolding on HIC surfaces. These two hypotheses are investigated by identifying regions on proteins that interact with HIC surfaces and experience conformational changes and relating such changes to known solution stability trends. Hydrogen-exchange mass spectrometry (HXMS) coupled to proteolytic digestion is used to probe conformations of three model proteins in solution and also while adsorbed on an HIC surface, Phenyl Sepharose 6FF. Also, exchange-rate patterns for proteins in solution and adsorbed to HIC surfaces are compared. Finally, hydrogen-deuterium exchange rates for the proteins in solution and regions of the proteins that show protection when adsorbed are compared with solution stability predictions by COREX.

2. Theory 2.1. Isotopic (Deuterium) Labeling. The exchange of a buried amide hydrogen, H, for a deuterium, D, can be modeled by the following19 kint

ku

ku

NðHÞ r s s f UðHÞ s f UðDÞ r f NðDÞ s s kf

ð1Þ

kf

ð2Þ

and the overall first-order rate constant for exchange, kobsd, is kobsd ¼ ku

ð3Þ

This limiting regime of exchange kinetics is commonly referred to as the EX2 limit. For very unstable regions of tertiary structure kf , kint

ð4Þ

and the overall rate constant for exchange is kobsd ¼ ku 3 kint

kint 1 ¼ ku kobsd

Langmuir 2011, 27(1), 286–295

expð - ΔGi =RTÞ Q

ð8Þ

where ΔGi is the relative free energy of a given conformational state a protein can theoretically exist in and Q is the conformational partition function Q ¼

N X

expð - ΔGi =RTÞ

ð9Þ

i¼0

The smaller the computational folding unit, the more states that must be generated during the calculation, at a cost of increased computation time. However, it also gives a more exhaustive landscape of the possible states. The stability constant of an individual residue, κf,i, is proportional to the summed probabilities of all states in which that residue is in a region that has retained its native conformation and thus is considered folded, Pf,i, over the summed probabilities of all states in which that residue is unfolded, Pnf,i Kf , i ¼

Pf , i Pnf , i

ð10Þ

Stability constants are closely related to hydrogen exchange protection factors

ð6Þ

(19) Hvidt, A.; Nielsen, S. O. Hydrogen Exchange in Proteins; Academic Press: New York, 1966; p 369.

ð7Þ

where R is the universal gas constant and T is the temperature. The stability constants are calculated by generating an ensemble of states that vary from completely folded to completely unstructured. The protein is broken up into folding units with a folding unit capable of comprising a varying number of residues. The COREX algorithm examines all possible states of the protein in terms of what units are folded or unfolded and assigns a Boltzmann-weighted probability for each state based on the free energy of that conformation

ð5Þ

This limiting regime of exchange kinetics is commonly referred to as the EX1 limit. Under EX2, the protection factor, PF, is defined as PF ¼

ΔGf , i ¼ - RT ln Kf , i

Pi ¼

where ku, kf, and kint, are the unfolding, folding, and intrinsic exchange rates and N and U represent the amide in its native and unfolded state, respectively. For very stable regions of tertiary structure kf . kint

Mass spectrometry (MS) can be used to detect the extent of deuterium labeling because protein mass will increase by 1 Da for each hydrogen that is exchanged for a deuterium. As a protein changes conformation, the number of backbone amides that are solvent exposed can change. If a protein exchanges via the EX2 limit, then a single peak appears in the mass spectrum that increases in mass with an increase in labeling time. If a protein exchanges via the EX1 limit, then the labeled protein will exhibit a bimodal distribution in the mass spectrum. The lighter mass peak represents protein molecules that have not unfolded during the labeling period, and the heavier peak arises from protein molecules that have unfolded. As labeling time increases, the amplitude of the heavier peak will increase with time as that of the lighter peak decreases. Globular proteins can demonstrate aspects of both EX1 and EX2 behavior because different regions of the protein molecule can differ in stability. 2.2. COREX. The COREX algorithm calculates solution stability constants for each residue of a protein using a statistical mechanical representation of the protein conformational ensemble.18 In brief, the local stability constants of residue i, κf,i, are related to their free energy differences, ΔGf,i, through

PFi ¼

Pf , i - Pf , xc, i Pnf , i þ Pf , xc, i

ð11Þ

where Pf,xc,i is the summed probability of all states in which residue i is folded yet solvent exposed, that is, exchange-competent.18 DOI: 10.1021/la103793r

287

Article

Gospodarek et al.

2.3. Measuring and Calculating Fractions of Labeled Peptide. The extent to which a peptide has been labeled with deuterium is determined by20 D mt - m0 ¼ N m100 - m0

ð12Þ

where D is the number of deuterated amides and N is the number of exchange-competent residues in a peptide. Also, mt is the mass of a peptide after a given labeling time, m0 is the nondeuterated mass of that peptide, and m100 is the fully deuterated mass of that peptide. In the present cases, exchange-competent refers to all peptide residues except proline, which does not have an amide hydrogen, and the N-terminal residue of the peptide, which does not have a backbone amide. Back-exchange of deuterium for hydrogen occurs when the deuterated protein molecules are introduced back into H2O solvents used in high-performance liquid chromatography (HPLC) to resolve peptides. Therefore, the residue immediately after the N-terminal residue of a peptide is also not counted; the back-exchange of this residue is unusually high.21 Equation 12 also accounts for back-exchange experienced during the time in between sample quenching and introduction into the MS.20 Both mt and m100 experience back-exchange, so the error is systematic and does not affect the true value of D/N. 2.4. Fraction-Labeled Peptide Calculated Using COREX. Under EX2 exchange kinetics, the COREX predicted fraction label of a peptide is determined by N P

D ¼ N

1 - expð - kobsd, i tÞ

i

N

ð13Þ

where t is time and the observed deuterium labeling rate for residue i, kobsd,i, is kobsd, i ¼ kint, i  PFi- 1

ð14Þ

COREX provides calculated kint,i for each residue, allowing for the prediction of the first-order rate constant for exchange.

3. Materials and Methods 3.1. Materials. Proteins used in this study were purchased from Sigma-Aldrich (St. Louis, MO). The proteins investigated were bovine R-lactalbumin (BLA), bovine β-lactoglobulin (BLG), and human serum albumin (HSA). Potassium phosphate, ammonium sulfate, calcium chloride, ethylenediaminetetraacetic acid (EDTA), citric acid, and guanidine hydrochloride (GuHCl) were purchased from Fisher Scientific (Houston, TX) and were of HPLC-grade quality or better. Tris(2-carboxyethyl)phosphine hydrochloride (TCEP) was purchased from Thermo Scientific (Rockford, IL). The GE Healthcare HIC resin used in this study, Phenyl Sepahrose 6 Fast Flow (high substitution), was purchased from Fisher Scientific. An Ultrafree-MC centrifugal filter device was purchased from Fisher Scientific for the separation of supernatant liquid from resin particles. 3.2. Solution-Phase Hydrogen-Deuterium Exchange.

We mixed 5 μL of 1 mg/mL protein solution with 45 μL of labeling buffer at room temperature. The protein solutions and labeling buffers were prepared at pH 7.0, 25 mM potassium phosphate, and 1.5 M ammonium sulfate. Experiments with BLA included 12 mM calcium chloride in the solutions for added

(20) Zhang, Z.; Smith, D. L. Protein Sci. 1993, 2, 522–531. (21) Bai, Y.; Milne, J. S.; Mayne, L.; Englander, S. W. Proteins 1993, 17, 75–86.

288 DOI: 10.1021/la103793r

stability. Three labeling times of 2.5, 10, and 40 min were examined. After labeling, 6 μL of quench buffer (100 mM potassium phosphate, pH 1.5) kept in an ice bath was added, bringing the final solution pH to 2.6, near the pH minimum of the hydrogen-deuterium exchange reaction. The sample was kept at room temperature for 1 min before 47 μL of desorption buffer was added to the solution. The desorption buffer for BLA and BLG had pH 2.6, 100 mM citric acid, 2.2 M GuHCl, 600 mM TCEP, and 100 mM EDTA in D2O. HSA required a higher concentration of GuHCl to unfold fully the protein for proteolytic digestion, so 8 M GuHCl, 100 mM TCEP, and 23 mM EDTA were used for the HSA desorption buffer. After the addition of desorption buffer, the sample was allowed to sit in ice for 2 min before being placed at room temperature for 1 min. Solutionphase samples were placed at room temperature for 1 min to replicate the time between sample quenching and introduction into the MS for the adsorbed phase experiments, where there are two additional centrifugation steps (1 min each).

3.3. Adsorbed Phase Hydrogen-Deuterium Exchange.

Protein solution (100 μL), 4 mg/mL for BLG, 10 mg/mL for BLA, and 1 mg/mL for HSA were equilibrated overnight at room temperature with 200 μL of resin slurry in a glass vial with gentle agitation. These concentrations were chosen to ensure adsorption in the linear region of each protein’s respective adsorption isotherm,22 where monolayer coverage is observed with minimal protein-protein interactions. To initiate labeling, we added 1800 μL of labeling buffer to the glass vial. Three labeling times of 2.5, 10, and 40 min were examined. To bring the final solution pH to 2.6, 240 μL of quench buffer kept in an ice bath was added. A 500 μL sample was immediately pipetted into a Ultrafree-MC centrifugal filter device and centrifuged at 7600 rcf for 1 min. The filter unit was removed and transferred to a new Ultrafree-MC centrifugal filter device where 200 μL of desorption buffer was added. The sample was allowed to sit in ice for 2 min to provide sufficient time for desorption of protein from the resin. The sample was centrifuged at 7600 rcf for 1 min. UV absorbance at 280 nm of the eluted samples along with mass balances revealed that less than 6 and 4% of the originally adsorbed BLG and BLA remained on the resin. All of the HSA was effectively eluted. 3.4. HPLC-MS. Samples were injected into a 500 μL stainless steel sample loop using a 500 μL glass syringe. HSA samples were diluted with sample pump solution prior to injection to decrease the concentration of GuHCl from 8 to 2.9 M to facilitate digestion. A sample pump (LabAlliance Series I) pumped 95% distilled, deionized water (ddH2O), 5% acetonitrile, 0.06% trifluoroacetic acid (TFA) solution, and the injected sample at 150 uL/min for BLG and HSA and 100 uL/min for BLA through the sample loop and into an immobilized pepsin column (2.1 mm inner diameter by 300 mm length) where proteolytic digestion took place. Pepsin preferentially cleaves at the C-terminal side of phenylalanine, leucine, tryptophan, tyrosine, alanine, glutamic acid, and glutamine, allowing consistent peptide fragments to be generated for different runs. Peptides exiting the column were trapped, desalted, and concentrated on a C8-desalting column (TR1/25109/02, 1 mm inner diameter by 8 mm length, Michrom Bioresources). After this desalting step (4 min for BLG, 9 min for HSA, and 6 min for BLA), flow was switched from the sample pump to the Surveyor MS HPLC pump to elute the peptides off the C8 column. A longer residence time in the pepsin column was necessary to improve BLA digestion. A XBridge C18 column (186003563, 2.1 inner diameter by 50 mm length, 3.5 μm pore size, Waters) downstream of the C8 column was used for HSA for improved resolution with its large number of peptides. For the solution and adsorbed phase studies, a short gradient run was employed to minimize back-exchange but still effectively resolve peptides. For BLG and BLA, the treatment for peptide desorption was a 10 min gradient of 95% solvent A (ddH2O, 0.1% formic acid, 0.01% TFA) and 5% solvent B (acetonitrile, 0.8% (22) Chen, J.; Cramer, S. M. J. Chromatogr., A 2007, 1165, 67–77.

Langmuir 2011, 27(1), 286–295

Gospodarek et al.

Article Table 1. BLG Reporter Peptides Identified in Short Gradient Runs

peptide

sequence

position

exchangeable amides

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

LIVTQTMKGLD LIVTQTMKGLDIQKVAGTW LIVTQTMKGLDIQKVAGTWYSLAM IQKVAGTWY LIVTQTMKGLDIQKVAGTWYSLAMAASD ISLLDAQSAPLRVYVEELKPTPEGDL YVEELKPTPEGDL AEKTKIPAV KIDALNENKVL KIDALNENKVLVL NENKVLVL DTDYKKYLL DTDYKKYLLF LLFCMENSAEPE VRTPEVDDEAL VRTPEVDDEALEKFDKALKALPMHIRL EALEKFDKALKALPMHIRL ALEKFDKALKALPMHIRL EKFDKALKALPMHIRL EKFDKALKALPMHIRLSFNPTQL

1-11 1-19 1-24 12-19 1-28 29-54 42-54 73-81 83-93 83-95 88-95 96-104 96-105 103-114 123-133 123-149 131-149 132-149 134-149 134-156

9 17 22 6 26 21 9 6 9 11 6 7 8 9 8 23 16 15 13 19

Table 2. BLA Reporter Peptides Identified in Short Gradient Runs peptide

sequence

position

exchangeable amides

1 2 3 4 5 6 7 8 9 10

EQLTKCEV FRELKDLKGYGGVSLPEWVC TTFHTSGYDTQA IVQNNDSTEYGL IVQNNDSTEYGLF QINNKIWCKDDQNPHSSNICN ISCDKFLDDDLTDDIM CVKKILDKVGINY CVKKILDKVGINYWLAHKAL WLCEKL

1-8 9-28 29-40 41-52 41-53 54-74 75-90 91-103 91-110 118-123

6 18 10 10 11 18 14 11 18 4

formic acid) to 65% solvent A, followed by a 2 min gradient from 65% solvent A to 10% solvent A, followed by 4 min at 10% solvent A. For HSA, the procedure was a 17 min gradient of 70% solvent A (ddH2O, 0.1% formic acid, 0.01% TFA) and 30% solvent B (acetonitrile, 0.8% formic acid) to 40% solvent A, followed by a 2 min gradient from 40% solvent A to 10% solvent A, followed by 4 min at 10% solvent A. Peptides were eluted directly to an LTQ linear electrospray ionization quadrupole ion trap mass spectrometer (Thermo Finnigan, San Jose, CA). Data were collected in a positive ion, selective ion monitoring, zoom scan, profile mode with an ESI voltage of 4.3 kV, a capillary temperature of 250 °C, and sheath gas flow rate of 40 units. Tables 1-3 show the reporter peptides identified for this study for BLG, BLA, and HSA, respectively. Sequence coverage of 67, 83, and 70% was obtained for BLG, BLA, and HSA, respectively. 3.5. COREX Calculations. COREX calculations were performed with the Fyrestar software package generously provided by the Hilser lab at the University of Texas Medical Branch. X-ray structures of the proteins specifying the spatial positions of the individual amino acid residues were required for the COREX calculations. PDB files 1BSQ and 1F6S were used for BLG and BLA, respectively. Experimental Gibbs energies of folding for the whole protein were used to calculate an entropy-scaling factor needed for the calculation of local stability data. The values were 38.8 and 42.9 kJ/mol for BLG and BLA23,24 to give entropyscaling factors of 1.003 and 0.929, respectively. These free energies take into account stabilizing effects of (NH4)2SO4 in solution.25 HSA was not included in the COREX calculations because its (23) Sakurai, K.; Goto, Y. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 15346– 15351. (24) Ikeguchi, M.; Kuwajima, K.; Sugai, S. J. Biochem. 1986, 99, 1191–1201. (25) Gloss, L. M.; Placek, B. J. Biochemistry 2002, 41, 14951–14959.

Langmuir 2011, 27(1), 286–295

size, 585 residues, would require excessive computational time to produce the ensemble needed for reliable stability constants and protection factors.

4. Results 4.1. BLG and BLA Solution Phase HDX Studies. MS spectra were collected for solution phase labeling of BLG and BLA and compared with COREX solution stability predictions. The MS spectra for the 20 reporter peptides of BLG and 10 reporter peptides of BLA at 2.5, 10, and 40 min deuterium labeling were converted to fraction labeled values, as described by eq 12 and compared with COREX fraction labeled values, as described by eqs 13 and 14. The fractions labeled for each BLG peptide at 2.5, 10, and 40 min are plotted and compared with COREX predictions in Figure 1. The COREX predictions are shown in black bars, and experimental values are shown in white bars. COREX does well in predicting the general trend of hydrogen-deuterium exchange for most of the BLG reporters at all exchange times (Figure 1a-c). Residues 3-11, 3-19, and 3-24, represented by peptides 1-3, are predicted by COREX to become subsequently more protected with increasing peptide length, which is accurately reflected at all experimental labeling times. Furthermore, COREX accurately predicts solvent protection to increase from residues 85-95 to 97-105 and then decrease from 97 to 105 to 105-114 as is found experimentally with peptides 9 through 14. A similar trend is predicted for peptides 15-20 and confirmed experimentally. The fraction labeled for the reporter peptides of BLA is compared with COREX predictions for all labeling times in Figure 2. COREX, in general, does not do a good job of predicting the rate of labeling at short time for the different regions of BLA. DOI: 10.1021/la103793r

289

Article

Gospodarek et al. Table 3. HSA Reporter Peptides Identified in Short Gradient Runs

peptide

sequence

position

exchangeable amides

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

DAHKSEVAHRFKDLGEENFKA FAQYLQQCPFEDHVKLVNE FAKTCVADESAENCDKSLHTLFGDKLCT LRETYGEM VRPEVDVM CTAFHDNEETFLKKYLYE FLKKYLYEIARRHPYFYAPELL LLFFAKRYKAAFTECCQAADKA DELRDEGKASSAKQRLKC ASLQKFGERAFKAWAVARL WAVARLSQRFPKAEF FVESKDVCKNYAEAKDVF VFLGMF LRLAKTYETTL KVFDEFKPLVEEPQ FEQLGEYKFQNAL LVRYTKKVPQVSTPTL VEVSRNLGKVGSKCCKHPEAKRMPCAEDYL VVLNQLCVLHEKTPVSDRVTKCCTESLVNRRPCF EVDETYVPKEFNAET FTFHADICTLSEKERQIKKQTAL LSEKERQIKKQTAL VELVKHKPKATKEQLKAVMDDF VEKCCKADDKETCF FAEEGKKLVAASQ

1-21 27-45 49-76 80-87 116-123 124-141 134-155 154-175 183-200 201-219 214-228 309-326 325-330 347-357 373-385 395-407 408-423 424-453 455-488 492-506 507-529 516-529 530-551 555-568 568-580

19 17 26 6 6 16 18 20 16 17 12 16 4 9 9 11 12 26 30 12 21 12 19 12 11

Figure 1. Fraction labeled for BLG reporter peptides at (a) 2.5, (b) 10, and (c) 40 min of solution phase deuterium labeling as predicted by COREX using eqs 13 and 14 (black, top) and from experimental studies (white, bottom). The negatives of the experimental values are shown for easier comparisons of COREX and experiment.

Figure 2. Fraction labeled for BLA reporter peptides at (a) 2.5, (b) 10, and (c) 40 min of solution phase deuterium labeling as predicted by COREX using eqs 13 and 14 (black, top) and from experimental studies (white, bottom). The negative of the experimental values are shown to ease comparison between COREX and experimental.

At 2.5 min of labeling, the experimental data show that most of the protein has not picked up a deuterium label with the exception of peptides 1 and 10, representing the N and C terminus, which are more than 30% labeled. COREX, however, predicts that 7 of the 10 reporter peptides should be labeled >40%. Furthermore, COREX does not accurately predict the variation of labeling from N- to C-terminus. Experimentally, the greatest labeling is observed at the N- and C-termini with minimal

variation along the sequence, with the exceptions of shallow minima at peptide 3 (residues 31-40), and peptide 8 (residues 93-103). COREX correctly predicts a high degree of labeling at the C-terminus but fails to capture the exchange behavior of the remainder of the molecule. It correctly identifies the shallow minima at peptide 3 but incorrectly predicts a substantial amount of variation in labeling along the remainder of the sequence.

290 DOI: 10.1021/la103793r

Langmuir 2011, 27(1), 286–295

Gospodarek et al.

Figure 3. Fraction labeled for BLG reporter peptides in solution, 1.5 M (NH4)2SO4, pH 7.0 at 1 mg/mL (9) and while adsorbed on Phenyl Sepharose, 1.5 M (NH4)2SO4, pH 7.0 at 10 mg/mL (0) at (a) 2.5, (b) 10, and (c) 40 min. Lines added to help guide the eye. Dashed line gives cutoff for peptides that retained >30% solvent protection when adsorbed. Error bars represent 95% confidence intervals by Student t test with the Bonferroni correction. Closed circles identify peptides that are statistically different in solution and adsorbed.

Control experiments with BLG, BLA, and HSA were performed at 10 min of solution labeling with and without the addition of ammonium sulfate and calcium chloride to observe their effect on the conformation of the proteins. No significant effects were observed because fraction-labeled values overall varied by only 2, 1, and 1% for BLG, BLA, and HSA between the two different conditions (data not shown). 4.2. BLG, BLA, and HSA Adsorbed Phase Studies. MS spectra were collected for adsorbed phase labeling of BLG, BLA, and HSA and compared with respective solution phase labeling of each protein. The MS spectra for the 20 reporter peptides of BLG, 10 reporter peptides of BLA, and 25 reporter peptides of HSA at 2.5, 10, and 40 min deuterium labeling were converted to fractionlabeled values, as described by eq 12, and compared with values of fraction labeled in solution. Figure 3 compares the labeling of BLG reporter peptides in solution and when adsorbed. A Student t test (95% confidence) with the Bonferroni correction identified 13 reporter peptides that are labeled significantly more in the presence of surface: 3, 4, 5, 6, 7, 8, 11, 12, 14, 16, 17, 18, and 19. Peptides 1, 2, 9, 10, 13, 15, and 20 show no statistically significant change in labeling when adsorbed. Furthermore, peptides 6, 7, 9, 10, 12, and 13 are the only peptides that show >30% protection after 40 min of labeling when adsorbed. With the exception of peptide 2, no other peptide can statistically be considered fully labeled; instead, the fraction labeled is between 0.70 and 0.80 for the other peptides, suggesting that these regions are not fully unfolded but rather only destabilized when adsorbed on the surface. Langmuir 2011, 27(1), 286–295

Article

Figure 4. Fraction labeled for BLA reporter peptides in solution, 1.5 M (NH4)2SO4, pH 7.0 at 1 mg/mL (9) and while adsorbed on Phenyl Sepharose, 1.5 M (NH4)2SO4, pH 7.0 at 10 mg/mL (0) at (a) 2.5, (b) 10, and (c) 40 min. Lines added to help guide the eye. Error bars represent 95% confidence intervals by Student t test with the Bonferroni correction. Closed circles identify peptides that are statistically different in solution and adsorbed.

The common aspect of the regions that retain >30% protection, peptides 6, 7, 9, 10, 12, and 13 (residues 31-54, 85-95, and 98-105), is a high degree of β-sheet content. Of the 57 residues of BLG that are in β-sheet conformations, sequence coverage for 30 of them was measured. Nineteen of these are among the 48 residues that are in protected regions (40%). The unprotected regions consist primarily of turns, R-helices, and coiled structure. Of the 23 R-helical residues, sequence coverage for 21 of them was measured. Only 2 of these 21 (10%) retain any protection when adsorbed on Phenyl Sepharose. Figure 4 compares the labeling of BLA reporter peptides in solution and while adsorbed. A Student t test (95% confidence) with the Bonferroni correction identified peptides 1-8 as labeled more when adsorbed. At 10 min, 7 of the 10 reporter peptides are fully labeled. Peptides 3, 8, and 9, residues 31-40, 93-103, and 93-110 show protection at 10 min: 20, 39, and 35%, respectively. These results are in good agreement with previous studies of BLA under the same experimental conditions.26 The characteristic feature of these protected regions is a high degree of R-helix content. Peptides 3, 8, and 9 consist of 40, 64, and 67% R-helix content. Sixteen of the 36 R-helical residues of BLA are in these protected regions. Although BLA is stabilized by CaCl2 by 21.4 kJ/mol,24 the calcium-binding region, spanning residues 78-89, is not protected. Therefore, peptide 7 (residues 77-90) is fully labeled at 10 and 40 min when adsorbed on Phenyl Sepharose. Figure 5 compares the labeling of HSA reporter peptides in solution and while adsorbed. A Student t test (95% confidence) (26) Fogle, J. L.; O’Connell, J. P.; Fernandez, E. J. J. Chromatogr., A 2006, 1121, 209–218.

DOI: 10.1021/la103793r

291

Article

Gospodarek et al.

Figure 6. COREX protection factors for BLG. Horizontal lines above the bars represent how well COREX protection factors match whether a region is folded or unfolded on Phenyl Sepharose. Thick lines represent regions that were correctly predicted by COREX. These are regions that retained more than the average solvent protection, 36%, when adsorbed and were predicted by COREX to have average PFs higher than the median PF, 7.8, and those retaining 30% protection after 40 min of labeling when adsorbed. Similar to BLG, HSA shows few peptides that can be considered to be fully labeled at 40 min. Other than peptides 3, 8, 19, and 24, the other 20 peptides have fractionlabeled values between 0.70 and 0.85, suggesting that like BLG, HSA is not fully unfolded on the surface but instead retains partial protection. Both EX1 and EX2 behavior was observed over the different regions of HSA when adsorbed to Phenyl Sepharose. Of the 25 reporter peptides, 14 exhibit EX1 behavior, a characteristic of a very unstable region of tertiary structure that has unfolded. All nine of the reporter peptides that span Domain II exhibited EX1 behavior, with the exception of Peptide 15. The observation that (27) Englander, S. W.; Mayne, L.; Bai, Y.; Sosnick, T. R. Protein Sci. 1997, 6, 1101–1109.

292 DOI: 10.1021/la103793r

almost all of Domain II exhibits EX1 behavior implies that this domain is the most unstable when adsorbed onto Phenyl Sepharose. Four of the 8 peptides in Domain I and 4 of the 10 peptides in Domain III featured EX1 behavior. The remainder of the peptides showed EX2 behavior, suggesting more stable regions of tertiary structure where some native-like state persists. All three of the proteins investigated show a measure of unfolding on Phenyl Sepharose. BLA unfolds completely on the surface with the exception of a few regions, whereas BLG and HSA unfold but retain some degree of native-like structure in most. Furthermore, different regions of the molecules are affected to different extents by the surface, as evidenced by variability in the degree of labeling. Many of the more protected regions are characterized by welldefined secondary structure, suggesting what regions of a protein may retain protection when adsorbed on a HIC surface. 4.3. COREX Protection Factors versus Adsorbed State Protection of BLG and BLA. We tested the ability of COREX to offer insight into what regions of a protein will or will not unfold on a surface by comparing solution-phase protection factor patterns predicted by COREX with adsorbed phase exchange patterns observed by HX-MS. Figure 6 shows residue-by-residue protection factors for BLG calculated using COREX. The horizontal lines represent how well COREX protection factors match whether a region is folded or unfolded on Phenyl Sepharose. Regions that COREX identified protection correctly or incorrectly are highlighted in Figure 6 with thick or thin overbars, respectively. Specifically, regions that retained >36% solvent protection when adsorbed and were predicted by COREX to have average PFs higher than the median PF, 7.8, are represented by thick lines. Regions that retained