Effect of Benzyl Alcohol on Recombinant Human Interleukin-1 Receptor Antagonist Structure and Hydrogen–Deuterium Exchange

Abstract

Benzyl alcohol, a preservative commonly added to multidose therapeutic protein formulations, can accelerate aggregation of recombinant human interleukin-1 receptor antagonist (rhIL-1ra). To investigate the interactions between benzyl alcohol and rhIL-1ra, we used nuclear magnetic resonance to observe the effect of benzyl alcohol on the chemical shifts of amide resonances of rhIL-1ra and to measure hydrogen–deuterium exchange rates of individual rhIL-1ra residues. Addition of 0.9% benzyl alcohol caused significant chemical shifts of amide resonances for residues 90–97, suggesting that these solvent-exposed residues participate in the binding of benzyl alcohol. In contrast, little perturbation of exchange rates was observed in the presence of either sucrose or benzyl alcohol.

INTRODUCTION

Antimicrobial preservatives are required to maintain sterility of therapeutic products in multidose formulations and drug delivery devices.^1–5 Development of stable formulations of therapeutic proteins that contain antimicrobial agents is greatly complicated, however, by the tendency of proteins to aggregate in the presence of these additives.^6–11 Such aggregation can reduce the biological activity of proteins^12,13 and, more importantly, impact the safety of the formulation by eliciting immunogenicity in patients.^14–16

Benzyl alcohol is one of the most commonly used preservatives in multidose therapeutic protein formulations.¹⁷ However, it has been shown to cause aggregation for several proteins, including recombinant human growth hormone,⁶ recombinant human interferon-γ,^7,18 and recombinant human interleukin-1 receptor antagonist (rhIL-1ra).^19–21 In earlier studies, no changes in rhIL-1ra secondary structure were observed by infrared (IR) spectroscopy upon addition of 0.9% (w/v) benzyl alcohol to rhIL-1ra solutions.¹⁹ However, subtle changes in tertiary structure determined from a decreased near-ultraviolet (UV) Circular Dichroism (CD) signal and increased 8-anilinonaphthalene-1-sulfonic acid (ANS) fluorescence revealed an increase in partially unfolded species within the protein population.¹⁹ A small increase (3%) of the a/b ratio,²² calculated from second-derivative UV spectra for rhIL-1ra in the presence of 0.9% (w/v) benzyl alcohol, also indicated greater solvent exposure of aromatic residues.¹⁹ Previous IR spectroscopy experiments¹⁹ also showed that the hydrogen–deuterium exchange (HX) rate averaged over all exchangeable residues of rhIL-1ra is accelerated at short time spans (<1.5 min) in the presence of 0.9% (w/v) benzyl alcohol at 37°C and pH 6.5. To explain these results, it was suggested that benzyl alcohol binds to rhIL-1ra and shifts the population toward partially unfolded species with increased mobility and greater time-averaged solvent exposure, which in turn lead to greater propensity to aggregate.¹⁹

Because the IR technique for measuring HX yields only a “global” rate of exchange averaged over all exchanging residues, it was not possible to use IR-based HX measurements to locate the site where benzyl alcohol might bind to rhIL1-ra. Indeed, an alternate explanation for the increased rates of HX for rhIL1-ra in the presence of benzyl alcohol might be that benzyl alcohol directly accelerates HX by nonspecifically diffusing into the protein, causing an increase in global HX without binding to a specific site.

In the current study, we used nuclear magnetic resonance (NMR) spectroscopy to map chemical shifts in the presence of benzyl alcohol and to measure HX for individual residues of rhIL-1ra in order to identify any specific benzyl alcohol binding site. Measurement of individual residue HX rates by NMR is a well established technique that has been used for several proteins (see Refs. 23–29) and has recently been used to examine the interactions of ANS with rhIL-1ra.³⁰ This allowed us to differentiate between the possibilities of specific binding versus nonspecific partitioning of benzyl alcohol into the structure of IL-1ra. Furthermore, we mapped chemical shifts at two temperatures in order to probe whether the interactions of benzyl alcohol with IL-1ra might be hydrophobic in nature, and to examine whether any potential conformational changes in rhIL-1ra as a function of temperature might affect benzyl alcohol binding.

MATERIALS AND METHODS

Materials

Both ¹⁵N-labeled rhIL-1ra and unlabeled rhIL-1ra were produced by and purified (>99%) at Amgen Inc., Thousand Oaks, California. The ¹⁵N-labeled rhIL-1ra stock formulation contained 6.6 mg/mL protein in 10 mM sodium acetate buffer (pH 5.2) and 5 mM ethylenediaminetetraacetic acid (EDTA). rhIL-1ra stock solution was maintained at −20°C until needed. The unlabeled rhIL-1ra stock formulation contained 220 mg/mL protein in citrate buffered saline:10 mM sodium citrate buffer (pH 6.5), 140 mM sodium chloride, and 0.5 mM EDTA. All other chemicals were purchased from Sigma (St. Louis, MO) or Fisher Scientific (Waltham, MA) and were of reagent grade or higher quality.

Dialysis and Sample Preparation

Dialysis for buffer exchange from the stock formulation into 10 mM sodium citrate buffer (pH 6.5), 140 mM sodium chloride (“buffer”) was performed against excess solvent (2 × 1L volumes) overnight using Pierce 10000 Da MWCO Slide-A-Lyzer dialysis cassettes (Rockford, IL). Following dialysis, the labeled protein for each experiment was concentrated to a volume of 120 μL (250 mg/mL, 14.5 mM) using two Millipore Centricon® cellulose YM-3 (3000 Da nominal molecular weight limit) centrifugal concentrators (Billerica, MA) for 6–8 h at 7000 rpm (6500 g). Fresh deuterated buffer was prepared for each experiment with the appropriate excipient (sucrose or benzyl alcohol) at a pH meter reading of 6.1, which corresponds to pH 6.5 for nondeuterated buffer (i.e., pD = pH + 0.4).

NMR Spectroscopy

Resonance assignments were confirmed using standard ¹⁵N-total correlation spectroscopy (TOCSY) and nuclear Overhauser enhancement spectroscopy (NOESY) experiments performed on 25 mg/mL (1.45 mM) ¹⁵N-labeled rhIL-1ra samples in 10 mM sodium citrate buffer, 140 mM sodium chloride at pH 6.5 containing 10% D₂O. Spectra were acquired at 25°C and 37°C on a Varian INOVA 500 MHz spectrometer (Santa Clara, CA) equipped with a room temperature HCN pulse (¹ H, ¹³ C, ¹⁵ N). The Varian Biopack suite (Santa Clara, CA) of pulse sequences (with minor modifications) was used for data acquisition. All spectra were processed using NMRPipe software (Washington, DC),³¹ and visualization and analysis were performed using Sparky software (Sanf Francisco, CA).³² Backbone amide resonances were assigned by comparison with previously reported ¹H and ¹⁵N chemical shifts.³³

A two-dimensional ¹⁵N heteronuclear single quantum correlation (HSQC) experiment was used to determine ¹H and ¹⁵N chemical shifts. A three-dimensional (3D) TOCSY–¹⁵N-HSQC experiment with a mixing time of 80 ms and a spinlock field of 5800 Hz provided confirmation of H^α, H^β, and H^γ shifts. A 3D NOESY–¹⁵N-HSQC spectrum with a 100 ms mixing time was also collected for further confirmation of proton chemical shifts and sequential connectivities. Spectra were apodized in all dimensions using a cosine-squared window function and then zero filled. Additionally, all indirect dimensions were linear predicted to twice the number of acquired points prior to apodization, and indirect dimensions in 3D experiments were subject to baseline correction. Proton chemical shifts are referenced to water at 4.773 ppm at 25°C and 4.649 ppm at 37°C,^{34 15}N dimensions were indirectly referenced to water as well using NMRPipe (see Ref. 35 for full protocol). Additionally, all experiments used a recycle time of 1.5 s; NOESY spectra used eight transients/FID (Free Induction Decay), whereas TOCSY and HSQC spectra used 16 transients/FID.

The protein primary sequence was obtained from the protein data bank (keyword: 1ILR). Minimum chemical shifts (Δδ_min) were calculated using ccpNMR Analysis software (Cambridge, UK) as follows³⁶:

$$\Delta {\delta }_{\min }(\mathrm{ppm})={[\Delta \delta {({}^1\mathrm{H}\mathrm{ppm})}^2+0.17\Delta \delta {({}^{15}\mathrm{N}\mathrm{ppm})}^2]}^{\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$2$}\right.}$$ (1)

Measurement of HX

Prior to recording data for exchange, the instrument parameters were optimized on a 10% D₂O sample prepared at a labeled protein concentration identical to that of the exchange experiment to follow (i.e., 60 μL at 250 mg/mL diluted 10-fold to a final volume of 600 μL with 480 μL nondeuterated buffer and 60 μL deuterated buffer). To initiate exchange, 540 μL of deuterated buffer (with sucrose or benzyl alcohol if required) was added to 60 μL of concentrated labeled protein in H₂O for a final labeled protein concentration of 25 mg/mL and final D₂O concentration of 90%. The first spectrum of each series was recorded 10 min after the exchange was initiated (data collection required 8.5 min per HSQC spectrum and was started to be centered around each time point). At 25°C, subsequent spectra were recorded at the following times after the exchange was initiated: 10, 20, 30, and 45 min; 1, 1.25, 1.5, 2, 2.5, 3, 4, 5, 6, 8, 10, 12,16, 20, and 24 h; and continuing in 6 h increments up to 90 h. For the sample in buffer alone (no benzyl alcohol or excipients), additional time points were taken as far out as 1 week, but it was determined that the data collected in the first 90 h was sufficient for fitting rate constants as the remainder of the residues exchanged very slowly. At 37°C, subsequent spectra were recorded at 10 min intervals until 2 h after the exchange was initiated. All spectra were processed using a Lorentz-to-Gauss apodization function in both dimensions, and the volumes of the resulting crosspeaks were fit as Gaussians and normalized to the volume of the first time point (10 min) using NMR-Pipe. The volumes were then fit to an exponential decay according to Eq. 2 ³⁷ using the CurveFit software (La Jolla, CA)³⁸:

$$H=H_0\exp (-k_{\mathrm{obs}}t)+C$$ (2)

with the initial intensity H₀, the baseline C, and the observed exchange rate k_obs as the free variables in the fit.

In addition to examining the effect of 0.9% benzyl alcohol on rhIL-1ra, we also tested separate rhIL-1ra samples in the presence of 0.5 M sucrose for comparison with previous results that used IR spectroscopy at 25°C.³⁹ For experiments in the current study, we chose to examine solutions conditions similar to the commercialized formulation buffer and the previous HX data.^19,39 Protein solutions in buffer alone (i.e., 10 mM sodium citrate, 140 mM NaCl, pH 6.5) are the reference states for comparing the effects of 0.9% benzyl alcohol and 0.5 M sucrose.

RESULTS

Cross-Peak Assignments

Recombinant human interleukin-1 receptor antagonist peak assignments were obtained from³⁰ and verified independently using ¹⁵N-TOCSY and NOESY spectra, and are listed in Table 1. Our observations show a systematic offset of 1.3 ppm in the N^α chemical shift as compared with the previously assigned values,³³ most likely due to a difference in referencing protocols, but are otherwise consistent.

Table 1.

Chemical Shift Assignments (ppm) and HX Rates for rhIL-1ra at 25°C in 10 mM Sodium Citrate Buffer, 140 mM Sodium Chloride at pH 6.5 Containing 10% D₂O

Residue	Observed 25°C		Buffer 25°C kobs	0.9% Benzyl Alcohol 25°C kobs (min−1)	0.5 M Sucrose 25°C kobs (min−1)	Buffer 37°C kobs	0.9% Benzyl Alcohol 37°C kobs (min−1)
	Na	HN
11 Met	119.3	8.02	fast	fast	fast	fast	fast
12 Gin	124.7	9.21	fast	fast	fast	fast	fast
13 Ala	126.5	8.66	fast	fast	fast	fast	fast
14 Phe	123.5	9.73	2.6E-04 ± 3.E-05	3.7E-04 ± 5.E-05	medium	2.4E-03 ± 3.E-04	2.5E-03 ± 4.E-04
15 Arg	119.1	9.05	1.1E-03 ± 5.E-05	1.5E-03 ± 7.E-05	1.4E-03 ± 1.E-04	2.3E-03 ± 1.E-03	1.4E-03 ± 1.E-03
16 Ile	120.5	9.06	3.2E-04 ± 1.E-05	2.6E-04 ± 2.E-05	3.0E-04 ± 7.E-05	slow	slow
17 Trp	124.0	8.64	8.4E-04 ± 5.E-05	slow	4.8E-03 ± 2.E-03	fast	fast
18 Asp	120.2	9.41	slow	3.2E-04 ± 1.E-04	slow	6.0E-04 ± 2.E-04	6.9E-04 ± 1.E-04
19 Val	109.4	6.76	1.8E-03 ± 6.E-05	2.2E-03 ± 8.E-05	2.5E-03 ± 1.E-03	1.3E-02 ± 2.E-03	1.1E-02 ± 1.E-03
20 Asn	119.4	8.36	7.9E-04 ± 3.E-05	9.0E-04 ± 4.E-05	7.4E-04 ± 6.E-05	slow	slow
22 Lys	119.6	8.67	2.2E-03 ± 4.E-04	slow	1.7E-03 ± 3.E-04	slow	slow
23 Thr	113.7	8.91	4.0E-04 ± 4.E-05	5.7E-04 ± 6.E-05	7.5E-04 ± 1.E-04	3.4E-03 ± 5.E-04	4.3E-03 ± 5.E-04
24 Phe	122.1	9.58	3.1E-04 ± 2.E-05	7.4E-04 ± 3.E-04	slow	5.4E-04 ± 4.E-04	1.1E-03 ± 3.E-04
25 Tyr	116.5	9.12	slow	slow	slow	1.9E-04 ± 9.E-05	4.9E-04 ± 1.E-04
26 Leu	120.8	9.23	1.6E-03 ± 7.E-05	1.9E-03 ± 6.E-05	2.1E-03 ± 2.E-04	2.9E-04 ± 3.E-04
27 Arg	124.2	8.90	2.6E-03 ± 9.E-05	3.4E-03 ± 1.E-04	3.2E-03 ± 2.E-04	5.2E-02 ± 3.E-03	medium
30 Gin	116.9	7.66	2.1E-02 ± 2.E-04	2.4E-02 ± 5.E-04	2.5E-02 ± 1.E-03	medium	medium
31 Leu	124.7	8.83	8.6E-02 ± 2.E-02	9.5E-02 ± 2.E-02	1.3E-01 ± 6.E-02	fast	fast
32 Val	125.6	9.19	slow	slow	slow	slow	slow
33 Ala	118.9	7.29	slow	slow	slow	slow	1.7E-04 ± 1.E-04
34 Gly	113.4	9.83	3.2E-04 ± 2.E-05	3.4E-04 ± 5.E-05	slow	6.9E-04 ± 2.E-04	5.8E-04 ± 2.E-04
35 Tyr	121.3	9.26	fast	fast	fast	slow	6.8E-04 ± 2.E-04
36 Leu	125.2	8.65	slow	slow	slow
37 Gin	119.9	8.45	fast	fast	fast	8.3E-03 ± 2.E-03	6.8E-03 ± 1.E-03
40 Asn	114.5	7.69	fast	fast	fast	fast	fast
41 Val	116.0	7.43	fast	fast	fast	fast	fast
42 Asn	117.6	7.79	slow	3.6E-03 ± 1.E-04	4.0E-04 ± 6.E-05	1.0E-03 ± 3.E-04	6.8E-04 ± 3.E-04
44 Glu	119.5	7.14	5.4E-02 ± 6.E-03	6.9E-02 ± 3.E-03	6.2E-02 ± 4.E-03	medium	medium
45 Glu	127.5	8.86	4.3E-02 ± 5.E-03	2.1E-02 ± 1.E-03	6.7E-02 ± 7.E-03	fast	fast
46 Lys	123.5	8.06	2.7E-02 ± 3.E-03	3.0E-02 ± 1.E-03	2.4E-02 ± 6.E-03	medium	1.1E-02 ± 1.E-02
47 Lie	121.9	9.13	2.5E-04 ± 2.E-05	3.3E-04 ± 6.E-05	slow
48 Asp	128.7	9.31	slow	4.1E-04 ± 1.E-04	slow	8.5E-04 ± 3.E-04	7.5E-04 ± 2.E-04
49 Val	120.2	8.84	3.9E-02 ± 1.E-02	fast	fast	medium	medium
50 Val	125.5	8.04	slow	slow	slow	slow	slow
52 Lie	122.4	8.38	fast	fast	fast	fast	fast
56 Ala	122.0	8.28	1.8E-01 ± 3.E-02	medium	1.7E-01 ± 5.E-02
57 Leu	118.3	9.30	3.1E-04 ± 2.E-05	slow	slow	3.8E-04 ± 2.E-04	4.3E-04 ± 2.E-04
59 Leu	122.2	10.46	slow	9.2E-04 ± 3.E-04	slow	slow	slow
60 Gly	109.9	9.33	slow	9.2E-04 ± 4.E-04	slow	slow	slow
62 His	115.7	9.83	fast	fast	fast	fast	fast
63 Gly	111.4	9.15	fast	fast	fast	fast	fast
64 Gly	108.4	8.30	fast	fast	fast
65 Lys	116.2	7.25	fast	fast	fast	1.3E-03 ± 6.E-04	slow
66 Met	116.4	7.67	1.9E-04 ± 2.E-05	2.6E-04 ± 3.E-05	4.1E-04 ± 8.E-05	2.4E-03 ± 2.E-04	1.4E-03 ± 3.E-04
67 Cys	117.6	9.10	slow	slow	slow	slow
68 Leu	124.9	8.08	slow	slow	slow	slow
69 Ser	114.0	9.15	slow	slow	slow	5.8E-04 ± 2.E-04	3.5E-04 ± 1.E-04
70 Cys	117.5	7.58	3.3E-04 ± 9.E-05	3.5E-04 ± 3.E-05	slow	2.8E-03 ± 5.E-04	2.1E-03 ± 3.E-04
72 Lys	125.2	8.33	fast	fast	fast	fast	fast
73 Ser	118.3	8.38	fast	fast	fast	fast	fast
76 Glu	120.7	8.09	fast	fast	fast	fast	fast
77 Thr	120.7	7.91	slow	2.9E-02 ± 7.E-04	slow
78 Arg	123.2	8.51	fast	fast	fast	slow	slow
79 Leu	123.5	8.64	fast	fast	fast	medium	medium
80 Gin	125.8	9.35	slow	slow	slow
81 Leu	121.2	8.26	fast	fast	fast	fast	fast
82 Glu	122.4	9.17	slow	slow	slow	slow	slow
83 Ala	128.5	8.95	fast	fast	fast	fast	fast
84 Val	125.0	7.70	slow	3.6E-04 ± 7.E-05	3.5E-04 ± 1.E-04	1.0E-03 ± 2.E-04	1.47E-03 ± 1.29E-04
85 Asn	120.2	8.24	fast	fast	fast	fast	fast
86 Ile	128.1	9.00	1.1E-01 ± 1.E-02	medium	medium	fast	fast
87 Thr	108.6	7.45	medium	medium	medium	fast	fast
89 Leu	120.3	6.48	4.5E-02 ± 3.E-03	5.5E-02 ± 1.E-02	2.5E-02 ± 1.E-03	medium	9.82E-03 ± 4.83E-03
90 Ser	117.1	10.02	fast	fast	fast	fast	fast
91 Glu	128.3	10.03	fast	fast	fast	medium	medium
92 Asn	116.3	8.22	fast	fast	fast	fast	fast
93 Arg	121.1	7.40	fast	fast	fast	fast	fast
96 Asp	119.9	7.65	fast	fast	fast
97 Lys	122.0	7.67	fast	fast	fast	fast	fast
98 Arg	112.5	7.49	fast	fast	fast	fast	fast
99 Phe	120.4	7.81	3.0E-03 ± 8.E-05	3.2E-03 ± 5.E-04	2.7E-03 ± 3.E-04	2.5E-02 ± 5.E-03	1.28E-02 ± 2.80E-03
101 Phe	122.8	9.54	slow	slow	slow	slow	5.12E-04 ± 3.52E-04
102 Ile	121.5	10.25	slow	slow	9.1E-04 ± 1.E-04	5.9E-04 ± 5.E-04
103 Arg	130.3	9.06	3.3E-02 ± 2.E-03	2.6E-02 ± 4.E-03	4.2E-02 ± 9.E-03	fast	fast
104 Ser	121.8	8.56	5.0E-02 ± 6.E-03	6.2E-02 ± 1.E-02	8.1E-02 ± 3.E-02	fast	fast
105 Asp	124.5	8.64	4.1E-02 ± 2.E-03	4.2E-02 ± 1.E-02	4.6E-02 ± 4.E-03	1.3E-03 ± 1.E-03
106 Ser	117.1	8.51	fast	fast	fast	medium	medium
109 Thr	109.2	7.58	1.9E-01 ± 1.E-02	medium	1.5E-01 ± 2.E-02	fast	fast
110 Thr	119.5	9.72	7.6E-04 ± 3.E-05	1.2E-03 ± 2.E-04	2.4E-03 ± 2.E-04	5.3E-03 ± 8.E-04	1.56E-02 ± 1.21 E-03
111 Ser	120.3	8.77	1.7E-02 ± 4.E-04	2.4E-02 ± 1.E-03	2.5E-02 ± 9.E-04	slow	9.91E-04 ± 7.38E-04
112 Phe	126.4	9.26	medium	medium	medium	slow	4.05E-04 ± 2.06E-04
113 Glu	126.3	9.09	8.8E-04 ± 8.E-05	slow	slow
114 Ser	120.4	8.90	3.1E-04 ± 2.E-05	slow	3.4E-04 ± 4.E-05	1.9E-03 ± 7.E-04	slow
115 Ala	128.0	8.19	3.3E-03 ± 1.E-03	7.0E-04 ± 3.E-04	4.0E-03 ± 1.E-03	1.2E-03 ± 1.E-03	slow
119 Gly	112.9	10.41	3.3E-02 ± 7.E-03	4.1E-02 ± 9.E-03	2.8E-02 ± 5.E-03	medium	medium
121 Phe	121.4	9.47	slow	slow	slow	1.5E-02 ± 2.E-03	1.35E-02 ± 1.20E-03
122 Leu	124.9	8.92	2.8E-04 ± 7.E-05	slow	slow	2.6E-03 ± 3.E-04	3.43E-03 ± 3.95E-04
123 Cys	120.3	9.14	8.9E-04 ± 5.E-05	3.8E-04 ± 2.E-04	slow	4.8E-04 ± 2.E-04
124 Thr	109.4	9.17	3.7E-04 ± 2.E-05	6.5E-04 ± 9.E-05	5.9E-04 ± 3.E-05	2.1E-03 ± 3.E-04	5.44E-03 ± 6.91E-04
125 Ala	122.3	9.35	5.6E-04 ± 3.E-05	8.0E-04 ± 7.E-05	7.2E-04 ± 3.E-05	5.5E-03 ± 1 .E-03	6.09E-03 ± 9.60E-04
126 Met	117.5	8.70	fast	fast	fast	fast	fast
127 Glu	119.2	7.96	fast	fast	fast	fast	fast
128 Ala	123.6	8.90	fast	fast	fast	fast	fast
129 Asp	114.8	8.66	3.9E-02 ± 2.E-03	3.7E-02 ± 1.E-02	4.4E-02 ± 5.E-03	5.3E-03 ± 9.E-04	6.37E-03 ± 8.66E-04
130 Gln	116.9	8.63	9.2E-04 ± 3.E-05	1.1E-03 ± 1.E-04	1.2E-03 ± 6.E-05	fast	fast
132 Val	127.5	8.11	slow	slow	slow	slow	slow
133 Ser	124.2	9.07	9.3E-04 ± 7.E-05	slow	3.1E-04 ± 1.E-04	slow	slow
134 Leu	117.4	8.49	4.9E-02 ± 3.E-03	medium	6.2E-02 ± 7.E-03
135 Thr	114.6	9.40	slow	8.4E-04 ± 2.E-04	8.4E-04 2.E-04	slow	6.92E-04 ± 1.76E-04
136 Asn	127.0	7.53	fast	fast	fast	fast
137 Met	119.7	8.35	1.9E-01 ± 1.E-02	medium	medium	5.6E-03 ± 3.E-03	6.77E-03 ± 1.14E-03
139 Asp	117.7	8.48	fast	fast	fast	fast	fast
140 Glu	118.8	7.56	fast	fast	fast	fast	fast
141 Gly	106.7	8.15	fast	fast	fast	fast	fast
142 Val	116.8	7.79	fast	fast	fast	fast	fast
143 Met	117.0	7.98	fast	fast	fast	fast	fast
144 Val	125.2	9.62	7.2E-04 ± 2.E-05	8.1E-04 ± 8.E-05	1.1E-03 ± 4.E-05	4.9E-03 ± 7.E-04	5.32E-03 ± 7.04E-04
145 Thr	111.9	8.41	fast	fast	fast	fast	fast
146 Lys	123.1	6.71	4.3E-03 ± 1.E-04	5.3E-03 ± 2.E-04	5.5E-03 ± 2.E-04	2.7E-02 ± 2.E-03	3.42E-02 ± 3.13E-03
147 Phe	117.9	8.87	1.8E-03 ± 3.E-04	slow	2.2E-02 ± 2.E-03	slow	slow
148 Tyr	121.6	9.49	3.2E-04 ± 1.E-04	slow	7.6E-04 ± 3.E-04	3.2E-04 ± 1.E-04	1.20E-03 ± 2.12E-04
149 Phe	127.3	8.79	1.6E-02 ± 6.E-04	2.1E-02 ± 2.E-03	6.3E-02 ± 9.E-03	medium	medium
150 Gln	125.2	8.75	3.9E-04 ± 2.E-05	5.2E-04 ± 7.E-05	medium	slow	slow
151 Glu	128.9	9.16	fast	fast	fast	fast	fast
152 Asp	126.2	8.58	fast	fast	fast	fast	fast
153 Glu	127.1	7.80	slow	slow	slow	fast	fast

Values for k_obs were fit using Eq. 2 (± standard error).

HX rate constants (k_obs) were also measured for rhIL-1ra in buffer alone (“buffer”) and in the presence of 0.9% benzyl alcohol, at both 25°C and 37°C or 0.5M sucrose (at 25°C only).

HX rate constants for residues labeled “fast” indicate cross-peaks that appeared in spectra collected in 10% D₂O, but had fully exchanged before the first time point at 10 min following initiation of HX by dilution to 90% D₂O.

HX rate constants for residues labeled “medium” exchanged within the duration of the HX experiment, but the level of noise in the data precluded a good fit (i.e., standard error > k_obs).

HX rate constants for residues labeled “slow” did not show significant exchange (i.e., less than 10% exchanged) during the 90 h duration of the HX experiment, and noise in the data often did not allow for a good fit (i.e., standard error > k_obs).

Chemical Shift Mapping

Figures 1 and 2 display ¹H–¹⁵N-HSQC spectra obtained at 25°C for rhIL-1ra in the presence of 0.9% benzyl alcohol and 0.5 M sucrose, respectively. The spectra (in red) are overlaid with the spectrum measured in buffer alone (in black) for comparison. Minimum chemical shifts between spectra with and without either 0.9% benzyl alcohol or 0.5 M sucrose were calculated according to Eq. 1 (Materials and Methods) and are displayed graphically in Figure 3. Residues showing the most significant shifts greater than 0.08 ppm (90 Ser, 91 Glu, 93 Arg, 96 Asp, and 97 Lys for shifts resulting from the addition of 0.9% benzyl alcohol; 63 Gly and 64 Gly for shifts resulting from the addition of 0.5 M sucrose) are mapped onto the crystal structure displayed in Figure 4.

Figure 1.

Recombinant human interleukin-1 receptor antagonist ¹H–¹⁵N-HSQC spectra collected with (red) and without (black) 0.9% benzyl alcohol at 25°C. Buffer in each case contained 10 mM sodium citrate, 140 mM sodium chloride, pH 6.5.

Figure 2.

Recombinant human interleukin-1 receptor antagonist ¹H–¹⁵N-HSQC spectra collected with (red) and without (black) 0.5 M sucrose at 25°C. Buffer in each case contained 10 mM sodium citrate, 140 mM sodium chloride, pH 6.5.

Figure 3.

Minimum rhIL-1ra chemical shifts for the addition of 0.9% benzyl alcohol (open squares) or 0.5 M sucrose (closed diamonds) to buffer at 25°C, 0.9% benzyl alcohol to buffer at 37° C (open circles). 63 Gly and 64 Gly show significant shifts (i.e., >0.08 ppm) upon addition of 0.5 M sucrose to the buffer. The prominent peak caused upon addition of 0.9% benzyl alcohol is caused by residues 90 Ser, 91 Glu, 93 Arg, 96 Asp, and 97 Lys.

Figure 4.

Mapping of chemical shift changes on the structure of rhIL-1ra. Numbers on the crystal structure label residues that shifted significantly (i.e., >0.08 ppm). Residues colored red (90 Ser, 91 Glu, 93 Arg, 96 Asp, and 97 Lys) shifted significantly upon addition of 0.9% benzyl alcohol to buffer at 25°C. Residues colored blue (63 Gly and 64 Gly) shifted significantly upon addition of 0.5 M sucrose to buffer at 25°C. The rhIL-1ra crystal structure was obtained from the protein data bank (keyword: 1ILR) and displayed using Pymol software (Portland, OR).

Because of the dependence of rhIL-1ra structure on temperature,⁴⁰ spectra were also collected for 2 h at 37°C. Previous studies showed that after 168 h of incubation at 25°C, the extent of aggregation is limited to a few percent even in the presence of benzyl alcohol.⁴⁰ The rate of rhIL-1ra aggregation is increased at 37°C with respect to the rate at 25°C, especially in the presence of benzyl alcohol.⁴⁰ However, within the 2-h time frame for these experiments, the extent of rhIL-1ra aggregation was negligible. In the absence of benzyl alcohol, nearly all the cross-peaks shifted when the temperature was increased from 25°C to 37°C (Fig. 5), with residues 62 His, 63 Gly, and 64 Gly exhibiting the greatest shifts (Fig. 6).

Figure 5.

Recombinant human interleukin-1 receptor antagonist ¹H–¹⁵N-HSQC spectra collected at 25°C (black) and 37°C (red). Buffer in each case contained 10 mM sodium citrate, 140 mM sodium chloride, pH 6.5.

Figure 6.

Minimum rhIL-1ra chemical shifts caused by increasing temperature from 25°C to 37°C. Closed squares represent rhIL-1ra chemical shifts observed in buffer alone. Open circles represent the effect of the temperature change on chemical shifts for rhIL-1ra residues in buffer plus 0.9% benzyl alcohol. In both cases, 62 His, 63 Gly, and 64 Gly exhibit the greatest temperature-induced shifts.

At both 25°C and 37°C, the largest chemical shifts that are observed upon addition of 0.9% benzyl alcohol occur at residues 90 Ser, 91 Glu, 93 Arg, and 97 Lys. These shifts are nearly identical in magnitude at the two temperatures. The minimum chemical shifts for all rhIL-1ra residues at 37°C upon addition of 0.9% benzyl alcohol to buffer are mapped onto the crystal structure in Figure 7. The residues showing the largest shifts cluster on a single area on the protein’s surface, whereas the residues located in the protein interior exhibit smaller shifts. The overall lack of chemical shift change suggests that benzyl alcohol does not introduce a large conformation change in rhIL-1ra. Thus, we used HX to determine whether the addition of benzyl alcohol impacted the dynamic properties of the protein.

Figure 7.

Minimum rhIL-1ra chemical shifts at 37°C upon addition of 0.9% benzyl alcohol to buffer, calculated using Eq. 4. Red residues, minimum chemical shifts greater than 0.1 ppm; orange residues, minimum chemical shifts between 0.06 and 0.1 ppm; yellow residues, minimum chemical shifts between 0.03 and 0.06 ppm,; white residues, minimum chemical shifts less than 0.03 ppm; and black residues, unassigned.

HX Measurements

To allow for comparison with previous IR-based measurements of HX,¹⁹ which showed that overall HX at short times (<1.5 min) is accelerated by benzyl alcohol and inhibited by the addition of sucrose. Global and individual HX rates were measured by NMR, by dilution of a protonated sample into D₂O-containing buffer and rapid collection of ¹H–¹⁵N-HSQC spectra. Global exchange rates for rhIL-1ra were calculated from the sum of all quantifiable ¹H–¹⁵N-HSQC crosspeak volumes and are presented in Figure 8. The data were collected at 25°C over 90 h, and the peak intensities were normalized to the integrated intensity of the first spectrum, which was recorded at 10 min after the exchange was initiated. At 25°C, global HX of rhIL-1ra was inhibited by the addition of 0.5 M sucrose in comparison with HX in buffer alone. Addition of 0.9% benzyl alcohol to rhIL-1ra resulted in only minor differences in the overall HX rate at 25°C as compared with the exchange in buffer alone. These NMR data, recorded at times starting at 10 min after the HX was initiated, are consistent with IR spectroscopic measurements for global HX at 25°C.¹⁹

Figure 8.

Global HX for rhIL-1ra at 25°C from the sum of all observed ¹H–¹⁵N-HSQC cross-peaks as a function of the remaining peak volume (intensity). The peak intensities were normalized to the first recorded spectra at 10 min after the exchange was initiated. Open circles, rhIL-1ra in buffer only (10 mM sodium citrate, 140 mM sodium chloride, pH 6.5); closed squares, rhIL-1ra in buffer with 0.5 M sucrose; and triangles, rhIL-1ra in buffer with 0.9% benzyl alcohol. Each series displays the HX from one experiment at each solution condition.

Hydrogen–deuterium exchange rate constants for individual residues were analyzed. The residues were grouped into three classes: those fully exchanged before the first time point (fast), those exchanging during the 90-h course of the experiment (medium), and those that were exchanged less than 10% over the course of the experiment (slow). Rate constants were obtained for medium exchangers by fitting the decay in cross-peak volume over time with Eq. 2. Predictably, the fastest exchanging residues (Fig. 9) are all located on the protein surface [confirmed by NAC-CESS (http://wolf.bms.umist.ac.uk/naccess) with an H₂O_volume of 1.4ų], whereas residues in the protein core exchange more slowly. Only minor perturbations in HX exchange properties were noted in the presence of the excipients tested. The classification (slow, medium, or fast) for each residue did not change in the presence of 0.9% benzyl alcohol or 0.5 M sucrose as compared with the sample containing buffer alone, although some of the residues with medium rate constants did show statistically significant changes between these three solution conditions (see Table 1). These data suggest that neither benzyl alcohol nor sucrose greatly change the slow and medium timescale dynamics of rhIL-1ra. All residues that exhibited particularly large chemical shifts in the presence of benzyl alcohol (residues 90, 91, 93, 96, and 97) were “fast” exchangers, as is typical for solvent-exposed sites.

Figure 9.

Hydrogen–deuterium exchange rate constants (k_obs) for individual rhIL-1ra residues at 37°C. Residues colored red were fully exchanged before the first data point at 10 min, residues colored orange were exchanged during the 10–120 min observation time, and yellow residues did not appreciably exchange (i.e., <10%) during the 10–120 min observation time. Unassigned residues are colored black.

The normalized rhIL-1ra global exchange rate is not affected by the addition of 0.9% benzyl alcohol (Fig. 8). Furthermore, none of the observed rate constants for individual residues measured at 37°C changed classification (e.g., from medium to fast) upon addition of 0.9% benzyl alcohol (see Table 1). Also, there were no changes in classification when rate constants measured at 25°C and 37°C in the absence of benzyl alcohol were compared (data not shown). Rate constants for individual residues at 37°C were fit using Eq. 2 and classified according to the same criteria used for the data at 25°C, and are listed in Table 1.

In the NMR measurement of HX at both 25° C and 37° C, many cross-peaks had already disappeared (i.e., dropped below the minimum contour level) by the first time point at 10 min after the exchange was initiated. The presence of benzyl alcohol has little effect on the intensities of the remaining cross-peaks (data not shown). Likewise, there were no significant differences between the remaining cross-peak intensities observed in the presence or absence of benzyl alcohol after 120 min of HX exchange (data not shown), indicating similar exchange rates for these slower exchanging residues in both solution conditions.

On the basis of these observations from NMR experiments, any changes to HX caused by addition of benzyl alcohol must affect only rapidly exchanging residues within the first 10 min after initiation of the exchange. This is consistent with the previously reported global HX rates measured by IR, where the most significant changes to rhIL-1ra HX at 37°C due to addition of benzyl occurred even before the first recorded IR measurement at 2 min.¹⁹

DISCUSSION

The significant chemical shift changes observed for rhIL-1ra residues 90–97 in the presence of benzyl alcohol (Fig. 7) are due to the changes in the local chemical environment of these residues, and suggest that this region forms the benzyl alcohol binding site for rhIL-1ra. Previous isothermal titration calorimetry experiments suggested weak hydrophobic binding of benzyl alcohol to rhIL-1ra.¹⁹ To examine the relative hydrophobicity of 90–97 region, a hydrophobicity plot for rhIL-1ra (Fig. 10) was generated on the Kyte–Doolittle scale.⁴¹ On this scale, values greater than zero represent hydrophobic regions, whereas values less than zero represent hydrophilic regions. Notably, residues 90 Ser, 91 Glu, 93 Arg, 96 Asp, and 97 Lys have the most negative values, indicating the hydrophilic nature of this region of the protein. The other residues of interest based on minimum shift analysis (62 His, 63 Gly, and 64 Gly) are clustered close to zero on the plot in Figure 10, indicating that they are not significantly hydrophobic or hydrophilic. Cation–π interactions involving 94 Lys and 97 Lys have been proposed as a potential contact site for reversible rhIL-1ra dimerization,^42,43 and further irreversible aggregation.⁴⁴ Notably, site-directed mutagenesis replacing 94 Lys with alanine significantly reduced aggregation.⁴² It is possible that benzyl alcohol is binding by the same cation–π interactions instead of by hydrophobic interactions as was previously proposed.¹⁹ Additional support for this point of view is provided by recent NMR studies of the interactions of ANS with rhIL-1ra,³⁰ wherein ANS was shown to bind in the region of 90 Ser, 93 Arg, and 95 Gln.

Figure 10.

A hydrophobicity plot for rhIL-1ra generated on the Kyte–Doolittle scale (http://expasy.org/cgi-bin/protscale.pl) with a window size of seven, using a linear weight variation model. On this scale, values greater than zero represent hydrophobic regions and values less than zero represent hydrophilic regions.

At 37°C, benzyl alcohol-induced aggregation is much more rapid than at 25°C.⁴⁰ Increasing the temperature from 25°C to 37°C in the absence of benzyl alcohol perturbed the observed chemical shifts, especially at residues 62 His, 63 Gly, and 64 Gly. But chemical shift perturbations observed upon addition of benzyl alcohol occurred predominately at the same residues and were essentially identical at both temperatures. This suggests that the temperature change did not create or remove benzyl alcohol binding sites. A similar observation was made for ANS binding to rhIL-1ra at 25°C and 37°C.³⁰

Addition of sucrose partially inhibits the increases in rhIL-1ra aggregation observed when the temperature is raised from 25°C to 37°C.⁴⁵ Interestingly, addition of sucrose results in substantial chemical shift changes at the same location (at residues 62 His, 63 Gly, and 64 Gly) that is most affected by temperature.

At the pH of interest, HX of solvent-exposed residues such as 90–97 occur too quickly (i.e., within 10 min of initiating the exchange) to be observed by the type of NMR measurements used in this study. However, the consistency in exchange rates at medium and slow timescales for all residues in the presence and absence of benzyl alcohol (e.g., slowly exchanging residues are slow in both conditions) indicates that the increased aggregation rates seen in the presence of benzyl alcohol cannot be explained by benzyl-induced increases in medium and slow dynamic processes occurring within the protein interior. This conclusion, coupled with the strongly localized chemical shift effects caused by addition of benzyl alcohol, supports the hypothesis that benzyl alcohol is binding to a specific surface region rather than diffusing nonspecifically into the protein.

The effect of addition of 1 M sucrose on the HX rate of rhIL-1ra in 10 mM sodium citrate, 140 mM sodium chloride at 25°C was measured previously using IR spectroscopy.³⁹ The authors observed that the addition of sucrose inhibited the HX rate at short times (<1.5 min), which was attributed to preferential exclusion of the sugar from the protein’s surface, but little difference in the rate of HX was observed for longer times. Our global HX results (which are obtained for times longer than 10 min) show only slight decreases in HX rates upon addition of sucrose (Fig. 8), consistent with the earlier work.

CONCLUSIONS

Recent HX studies with interferon-γ¹⁸ and recombinant human granulocyte colony-stimulating factor⁴⁵ revealed that benzyl alcohol causes only minor perturbations to the protein’s structure, whereas the protein core is largely unaffected. These limited structural perturbations are sufficient to populate partially unfolded species prone to irreversible aggregation. Our rhIL-1ra results show that benzyl alcohol binds to the protein’s surface, but does not induce global conformational destabilization or result in increased tendency to populate completely unfolded states, which would be manifest in greatly increased rates of HX. Thus, the increased rates of aggregation of rhIL-1ra seen in the presence of benzyl alcohol are likely the result of increased populations of only partially and conformationally perturbed species that are prone to irreversible aggregation, a conclusion that is consistent with the previous study of benzyl alcohol effects on rhIL-1ra aggregation.¹⁹