Skip to main content
NCBI home page
ACS Omega logoLink to ACS Omega
. 2020 Dec 4;5(49):31845–31857. doi: 10.1021/acsomega.0c04692

Effects of Excipients on the Structure and Dynamics of Filgrastim Monitored by Thermal Unfolding Studies by CD and NMR Spectroscopy

Houman Ghasriani 1, Grant E Frahm 1, Michael J W Johnston 1, Yves Aubin 1,*
PMCID: PMC7745408  PMID: 33344838

Abstract

graphic file with name ao0c04692_0009.jpg

Product excipients are used to confer a number of desirable properties on the drug substance to maintain or improve stability and facilitate drug delivery. This is especially important for products where the active pharmaceutical ingredient (API) is a recombinant protein. In this study, we aimed to determine if excipients and formulation conditions affect the structure and/or modulate the dynamics of the protein API of filgrastim products. Samples of uniformly labeled 15N-Met-granulocyte-colony stimulating factor (GCSF) were prepared at 100 μM (near formulation concentration) with various concentrations of individual components (polysorbate-20 and -80, sorbitol) and three pH values. Nuclear magnetic resonance (NMR) spectroscopy techniques were applied to measure chemical shift perturbation (CSP) to detect structural changes, and relaxation parameters (T1, T2, and heteronuclear Overhauser effect) were measured to probe the effects on protein backbone motions. In parallel, the same solution conditions were subjected to protein thermal unfolding studies monitored by circular dichroism spectropolarimetry (CD). Detergents (polysorbate-20 and 80) do not induce any observable changes on the protein structure and do not modify its dynamics at formulation concentration. Lowering pH to 4.0, a condition known to stabilize the conformation of filgrastim, as well as the addition of sorbitol produced changes of the fast motion dynamics in the nanosecond and picosecond timescale. NMR-derived order parameters, which measure the local conformational entropy of the protein backbone, show that lowering pH leads to a compaction of the four-helix bundle while the addition of sorbitol relaxes helices B and C, thereby reducing the mobility of loop CD. CSPs and measurements of protein dynamics via NMR-derived order parameters provide a description in structural and motional terms at an atomic resolution on how formulation components contribute to the stabilization of filgrastim products.

Introduction

Therapeutic recombinant proteins must be formulated with excipients to stabilize the protein active pharmaceutical ingredient (API) during manufacture, storage, and delivery of the product to patients. Various categories of excipients are available to the manufacturer to develop a formulation appropriate for their product. These include buffering agents to control pH, amino acids to control pH and other functions, osmolytes, sugar, and carbohydrates to help stabilize protein during environmental stress and lyophilization, salts to adjust tonicity, proteins, polymers, and surfactants as competitive inhibitors of surface adsorption and denaturation, chelators, antioxidants, and preservatives. It can be said that excipients have two modes of action: indirect and direct. An indirect mode acts on the environment of the protein API and a direct mode acts on the API itself, such as binding or modifying the hydrogen bonding character (essentially H-bond length). A given excipient could have both modes of actions. Formulations are developed and tested using normal and stress conditions to better stabilize the API under all conditions; it may be subjected to during manufacturing, storage, transportation, and delivery to patients. Kamerzell and co-workers have reported a comprehensive review on protein–excipient interactions and the various biophysical methods used in formulation development.1 Here, we apply nuclear magnetic resonance (NMR) spectroscopy to identify and characterize, if present, direct interactions between the various excipients and the protein API in filgrastim products under normal conditions of storage and temperature (no stress imparted to protein API).

Filgrastim products, such as the innovator product Neupogen from Amgen, are used to treat neutropenia that is characterized by a low absolute neutrophil count as a result of chemotherapy bone marrow transplant or other conditions. Filgrastim is a recombinant methionyl granulocyte-colony stimulating factor (Met-GCSF). The mature human GCSF protein contains 174 amino acid residues that fold into a four-helix bundle.2,3 Part of the loop between helix A and helix B is folded into a small helix that is positioned perpendicular to the bundle axis. The primary sequence includes five cysteine residues where the last four form two disulfide bridges. Filgrastim products are formulated at pH 4.0 to stabilize cysteine-17 that is partially solvent-exposed.4 This low pH is also known to provide conformational stabilization5,6 via a proposed cation−π interaction between histidine-79 and tryptophan-118 and within the triad made by histidine-52, tryptophan-58, and histidine-156.7 The excipients include sorbitol and polysorbate-80 (PS-80) and -20 (PS-20). Protein–excipient interactions of sorbitol and PS-80 excipients were investigated by NMR spectroscopy using chemical shift perturbation studies. No detectable perturbations were observed at formulation concentration and up to 20 times the concentration. However, a loss of signal intensity without line broadening was observed. This suggested that potential protein–excipient interactions might be too weak to induce detectable changes in the magnetic environment of backbone amide pairs but could affect protein dynamics. NMR spectroscopy can measure protein dynamics over a very wide range of motions, from very fast motions in the pico-nanoseconds timescale, down to slow motions in the micro-millisecond timescale.8 In the present study, we aimed to measure NMR relaxation parameters of backbone amide nitrogen, namely T1, T2, and 1H–15N-heteronuclear nuclear Overhauser effect to probe for the presence of protein–excipient interactions and shed light on the cation−π interaction that stabilizes the conformation at formulation pH. Relaxation parameters were then used to calculate order parameters for each residue of the polypeptide backbone. These provide a description of the amplitude of backbone internal motions that are correlated with the entropy, thereby the free energy of protein conformation.9−12 Atomic-level structural information can only be obtained by NMR spectroscopy under native conditions where protein molecules remain monodispersed in solution. Upon unfolding, resonance dispersion is lost leading to severe overlap. However, unfolding is usually associated with aggregation, which produces tremendous loss of NMR spectral quality due to signal broadening. Therefore, NMR measurement of binding, structural perturbations, and nanosecond–picosecond backbone motions were carried out in the absence of any stress to the protein to probe effects of excipients in the product vial. In addition, thermal unfolding studies by circular dichroism spectropolarimetry (CD) were carried out in parallel to obtain a readout of the overall conformational stability of the protein under the various solution conditions tested in the NMR relaxation measurements.

Results and Discussion

Sample Preparation

A single dose of 500 μL of Neupogen contains 300 μg of Met-GCSF in 10 mM sodium acetate buffer at pH 4.0 (30 μM) with the following excipients: 274 mM sorbitol and 0.03 mM polysorbate-80. One biosimilar version of this product has substituted the detergent with polysorbate-20. To study interactions between each component and the protein active pharmaceutical ingredient, CD and NMR measurements were carried out on samples containing 15N-labeled Met-GCSF with only a single component (parameter) of the formulation (pH variation, concentration of a single excipient at the formulation pH). Therefore, the strategy chosen for sample preparation consisted of adding an appropriate amount of a single excipient from a stock solution to a given concentration of labeled protein sample to ensure that the desired concentration of detergents was obtained, and that they were above their respective critical micelle concentrations (CMCPS-80 = 0.012 mM and CMCPS-20 = 0.06 mM). When possible, analyses were carried out at formulation concentration of 30 μM of Met-GCSF, but for sensitivity reasons of the NMR relaxation measurements, a protein concentration of 100 μM was utilized to obtain adequate signal-to-noise required to measure peak intensities (60:1) in a reasonable amount of time (60 h total acquisition time for all three experiments).

CD Spectropolarimetry

Thermal unfolding curves recorded by CD did not reveal any detectable effect of individual excipients on the stability of the conformation of the API when compared to the API in buffer at pH 4.0 (Figure 1A). These observations were somewhat expected and suggest that polysorbate and sorbitol are merely providing a stabilizing environment in situations of stress that can be encountered during all phases of the product lifecycle. In addition, API–excipient interactions may be too subtle to affect the melting temperature of the API as measured by CD. Far-UV CD was used here to monitor the protein secondary structure and the effects on secondary structure elements induced by conformational changes. In contrast, pH has a significant effect on conformational stability, as previously observed.6,13 Melting temperatures decreased as pH increased. In our hands, these were 68.2 ± 1.0, 58.9 ± 0.6, and 57.2 ± 0.5 °C, at pH 4.0, 5.0, and 6.0, respectively. A 10 °C shift at pH 4.0 compared to pH 5.0 (Figure 1B) is indicative of a dramatic increase in conformational stability.

Figure 1.

Figure 1

Open in a new tab

Fractional change in ellipticity at 222 nm with temperature for Filgrastim as monitored by far-UV circular dichroism. Each spectrum represents the mean of at least three separate experiments for panel (A) and at least two separate experiments for panel (B). (A) Effects of excipients: Met-GSCF (red); Met-GCSF + PS80 (black); Met-GCSF + PS20 (yellow); and Met-GCSF + sorbitol (cyan), all samples at pH 4. (B) Effects of pH: Met-GSCF-pH 4.0 (red); Met-GSCF-pH 5.0 (green); and Met-GSCF-pH 6.0 (blue).

Chemical Shift Perturbation

The magnetic environment surrounding nuclei directly influences their chemical shifts.8 Factors that influence the electronic density surrounding a given NMR active nucleus, such as torsion angles, proximity of electron-rich groups such as neighboring aromatic side chains, binding of molecules, dielectric strength of the solvent (ionic strength), pH, temperature, to name only these, will induce changes in the chemical shift of that nucleus. Any changes in one or several of these factors will produce chemical shift perturbations (CSP), making them potentially good probes of API–excipient interactions and conformational perturbations. It is important to stress here that the observation of a CSP does not automatically mean that the conformation has changed since some of the above factors, such as ionic strength and pH, do influence electronic density. Finally, CSPs can be very useful probes to investigate, for example, weak binding interactions (Kd values in the high millimolar range) by simply shifting the equilibrium to the bound state with an excess of the ligand and measuring CSPs that are indicative of the interaction. If the assignment of the resonances of the protein is known, the binding site on the API can be derived from the CSPs and the dissociation constant can be calculated.

Example of the Use of CSPs to Derive Binding Site and Dissociation Constant of ANS to Filgrastim

The molecular probe 8-anilinonaphthalene-1-sulfonic acid (ANS) experiences extrinsic fluorescence upon binding protein surfaces. Initially, the measurement of CSPs was used to monitor the binding of ANS to filgrastim samples (Figure 2). Considering that the probe is aromatic, it was expected that the magnitude of the CSP measured for ANS binding to filgrastim (Figure 2A) would be larger, such as tenths of ppm per titration point, as observed for ANS binding to calmodulin14 or interleukin-1 receptor agonist.15 Mapping of CSPs on the surface of the structure of filgrastim (Figure 2B) shows that the largest number of CSPs are observed in the vicinity of the small helix formed by residues 47–55 with the second group of CSPs affecting residues at the beginning of loop 56–70 and residues at the end of the C-terminal helix. This may be indicative of two binding sites or a single binding site located near or at the small helix that induced a small change of the loop AB (in the area of residues 47–55) conformation that also induce the CSP observed at the end of the C-terminal helix (helix-D) (Tyr-165, Arg-169, and His-170). To distinguish between these two possibilities, CSPs alone are not sufficient. Measurement of internuclear distances between the probe and the protein would be required to allow the proper docking of the probe to its binding site(s), or carrying out structure determination using X-ray crystallography.16 ANS may bind to proteins in some cases via ionic interactions between its sulfate group and positively charged side chains.15 This mode of binding was suggested to explain sample precipitation at higher protein concentrations. In our study, ANS titration experiments using a filgrastim concentration of 100 μM protein led to precipitation of the sample at high ANS concentrations. In fact, two side-chain resonances assigned to Arg-146 and Gln-145 show CSPs, which may support the thesis that the probe may be involved in such a positive charge–sulfate interaction. However, considering that the sample is at pH 4.0, the ANS amine group is protonated, which would dampen an electrostatic interaction with these charged residues on the protein surface. Moreover, many surrounding residues in both binding sites provide neutral and hydrophobic side chains to interact with the aromatic rings of the probe.

Figure 2.

Figure 2

Open in a new tab

Binding of 8-anilinonaphtalene-1-sulfonic acid (ANS) to samples of 30 μM 1H–15N-Met-GCSF in 10 mM sodium deuteron-acetate buffer at pH 4.0 measured by NMR and fluorescence spectroscopy. (A) Expansion of the overlay of 2D-1H–15N-HSQC (tryptophan HN-indol resonances are not shown) (0 [red], 30 [green], and 87 [blue] μM ANS). Labeled resonances with their corresponding peaks are expanded to show their chemical shift perturbations (CSP) upon ANS addition. (B) Mapping CSPs on structures. The structure of ANS is in the upper left corner of this panel and residues experiencing CSP (see text) are represented by their side chains as sticks where a subset is labeled. Regions of residues 47–55 are circled. (C) Dissociation constants (Kd) of residues showing CSPs calculated using eq 1. (D) Top panel: fluorescence spectrum at 1:1 ANS/GCSF molar ratio for two attempts (blue and red) and ANS only in gray and black. Bottom panel: binding of ANS measured by fluorescence with Kds of 143 and 231 μM for attempt #1 in blue and #2 in red calculated using eq 1. The dotted line indicates the 1:1 ANS/GCSF molar ratio.

The dissociation constant was derived by measuring CSP of well-resolved signals and assuming a 1:1 ratio of ANS with the residue corresponding to that signal (Figure 2C). Values derived from the NMR data compared well with those measured using fluorescence spectroscopy (Figure 2D).

Effects of Polysorbate Detergents

Analysis of potential interactions or effects of detergents on 15N-Met-GCSF were carried out on samples that were prepared by appropriately diluting the labeled protein and detergents from their respective concentrated stock solutions of protein and detergents to ensure that the protein-to-detergent ratio was known. In all samples, the minimum detergent concentrations (polysorbate-20 or -80) studied were at the formulation concentration, which are above their respective CMCs (see the Experimental Section for details).

Polysorbate-80

The addition of polysorbate-80 to the NMR sample did not induce any observable CSPs. Measurement of relaxation parameters (T1, T2, and HetNOE) did not reveal any informative elements (see Figure S1) that could inform on our previous studies whereby a slight decrease of line intensities, i.e., the entire amide signal envelope, without line broadening was observed at high polysorbate concentration.7 This behavior is similar to samples that have high ionic strength, such as 50+ mM salt concentrations. These samples absorb the electric field component of the excitation pulse, thus requiring longer excitation pulses, and they attenuate the detected signal, which result in a decrease of signal intensity. However, polysorbate is not ionic, it does not contribute to the ionic strength of the sample and relatively low concentrations were used (0.3 mM). This loss of signal intensity at high polysorbate concentrations remains unexplained. In contrast, a study by Singh et al.17 reported differential intensity changes for some cross peaks in the 1H–13C two-dimensional (2D) spectrum of the Fab fragment of an unknown IG-g1 monoclonal antibody upon titration of PS-80. However, no changes of the contour map of the Fc were observed. Unfortunately, they do not report any effects on the envelope of the one-dimensional (1D) trace of their spectra, which would have indicated a similar behavior. As reported earlier, NMR analysis of products with high detergent concentrations requires partial or total detergent removal.18

Polysorbate-20

A study by Chang et al.19 has detected weak interactions between PS-20 and filgrastim and has measured a binding constant via isothermal calorimetry with no changes on the protein secondary structure. Titration of polysorbate-20 in samples of filgrastim showed small but detectable CSP that were mapped on the structure (Figure 3). Like polysorbate-80, no effects on the dynamics have been observed (see Figure S2). The presence of CSPs in this case was initially puzzling, considering that the concentration of detergent is above its CMC. The addition of more PS-20 should simply drive the formation of more micelle assemblies because the concentration of detergent monomer remains constant at the CMC and should not induce any CSP, as observed with PS-80 micelles. The only difference between PS-20 and PS-80 is the relatively shorter length of the fatty acid moiety resulting in a smaller size of the micelle particles. Notably, about 20% of the total alkyl chains of PS-20 are short 10 carbons or less (∼10% C10; ∼10% C8; and ∼1% C6) (USP National Formularies). These shorter chain detergents, similar to alkyl glucoside detergents, probably have higher CMC, thus allowing higher concentrations of short-chain detergent monomers that can interact with filgrastim. Filgrastim does not present any hydrophobic surfaces that would allow preferential binding of long alkyl chain detergent. Molecules like ANS even do not bind primarily via hydrophobic interactions because of the absence of such hydrophobic regions. This explains the binding of PS-20 to filgrastim observed by Chang et al.19 Using CSPs, we derived dissociation constants (Figure 3C), which are in the 200–600 μM range between residues located at the top of the fold.

Figure 3.

Figure 3

Open in a new tab

Titration of polysorbate-20 to samples of 30 mM 1H–15N-Met-GCSF in 10 mM sodium deutero-acetate buffer at pH 4.0 measured by NMR. (A) Expansion of the overlay of 2D-1H–15N-HSQC (red 0 μM, green 60 μM about PS-20 1× CMC, cyan 587 μM 10× CMC, and blue 1487 mM 30× CMC PS-20) showing a subset of labeled resonances experiencing CSP upon PS-20 addition. (B) Mapping CSPs on Met-GCSF structure. Note that most residues involved in PS-20 interactions are hydrophobic. (C) Dissociation constants Kd (reported in μM in parenthesis) for residues showing CSP calculated using eq 2.

Effects of pH

Melting curves monitored by Far-UV circular dichroism showed (Figure 1B) a shift of 10° of the melting temperature upon a pH decrease from 6.0 to 4.0. This indicated a gain of conformational stability, also reported in other studies,4,13,20,21 and could be explained, as we proposed earlier,7 by the formation of the nonbonding cation−π interaction at pH 4.0. Calculations of order parameters for backbone amides were carried out from measurements of six relaxation parameters (T1, T2, and hetNOE, measured at two fields: 600 and 700 MHz) using the model-free approach with the implementation of the protocol developed by d’Auvergne and Gooley.22−26 The model-free formalism means that no a priori model is selected; it is optimized along the calculation pathway to account for all experimental measurements. Also, this protocol allows for no a priori diffusion tensor to be used; it can therefore be optimized during the calculation of O2. An oblong diffusion tensor was selected to keep calculation times within weeks. The x, y, and z axes could be optimized during the calculations. NMR-derived order parameters provide a description of the amplitude of backbone internal motions that are correlated to the entropy, thereby the free energy of protein conformation.11 The order parameters for each residue along the primary sequence at three pHs are shown in Figure S3. In this format, there is one feature that strikes the eye: the order parameters for the CD-loop (residues 126–145) show an increase of mobility for these residues. This may a priori seem counterintuitive considering that a gain in thermal stability resulting from the formation of a nonbonding cation−π interaction should be associated with a reduction of the polypeptide backbone mobility. Further examination of the data shown in Figure S4 did not shed light on the issue. However, a similar behavior was observed with human growth hormone (hGH).27 Led and co-workers have observed that upon raising the temperature, hGH gained conformational stability. Order parameters derived from NMR data showed that a long flexible loop gained mobility while helices became more rigid. In that study, order parameters for all residues within a structural element, whether helix or loop, were summed and averaged. Using this method, order parameters were summed and averaged for each secondary structure elements (helices A, B, C, D, and AB-, BC-, and CD-loops) at the three pH values. Figure 4A shows that upon lowering pH to 4.0, the amplitude of backbone amide motions of all helices and loops decreased as opposed to the CD-loop that increased, and to a lesser extend loop AB. This gain of overall rigidity of the conformation is consistent with a gain in stability. This could be explained by a tighter helical bundle resulting from the formation of cation−π interactions at pH 4.0. This nonbonding interaction between His-79 on helix-B and Trp-118 on helix-C leads to a reduction of mobility between these helices. The second cation−π interaction while weaker, based on the smaller CSPs, involves His-52 located on the short helix of the AB-loop, Trp-58 on the AB-loop, and His-156 on helix-D and can “attach” the AB-loop to helix-D, thereby explaining why this loop gain rigidity at lower pH. In addition, the protonation of the carboxyl side chains of Glu-93, Glu-98, and Asp-104 could contribute to reduce repulsive electrostatic interactions. These interactions are illustrated in Figure 4B. These three residues are located at the end of helix-B, BC-loop, and beginning of helix-C, respectively. Therefore, tightening up of the helix bundle from the above interactions thus provides more degrees of freedom for the CD-loop, which results in smaller order parameters (larger amplitudes of motion) for this loop.

Figure 4.

Figure 4

Open in a new tab

Effects of pH on the dynamics of 100 μM 1H–15N-Met-GCSF in 10 mM sodium deuteron-acetate buffer measured by NMR. (A) Order parameters for structural elements (see text) at pH 4.0, 5.0, and 6.0. (B) Identification of loops AB and CD structural elements and potential interactions on the structure of Met-GCSF. (C) Cartoon representations to illustrate the observed change of order parameters upon lowering pH. At left, the cartoon illustrates a reduction of motion amplitude of helices (shorter arrows on helices) leading to a significant increase of motion of loop CD (longer arrows on loops). At pH 6.0, the acidic side chains of Asp-104, Glu-93, and Glu-98 are deprotonated, thus inducing repulsive negatively charged interactions. In addition, histidine side chains are not protonated and cannot participate in stabilizing cation−π interactions. Lowering pH to 4.0 causes protonation of histidine residues and acidic side chains. The newly formed nonbonding cation−π interactions and absence of electrostatic repulsive interactions lead to a slight compaction of the four-helix bundle (increase of order parameter, gain in rigidity, shorter arrows on helices) with a resulting increase in degrees of freedoms of loops AB and CD (longer arrows on loops) associated with a decrease of order parameters for these elements at pH 4.0.

Effect of Sorbitol

The effects of the bulking agent sorbitol on the structure and dynamics of filgrastim were examined at one time (1×: 274 mM) and three times the formulation concentration (3×: 749 mM). Overlay of 2D 1H–15N-HSQC shows a large number of small CSPs (CSP = 0.001–0.040 ppm) that are less than the width at half height and increase with sorbitol concentration. Mapping these on the structure (Figure 5) reveals that sorbitol affects the overall protein fold through a change of solvent activity arising from the exclusion of water at the protein surface (first hydration shell) and replacement with the sorbitol–water mixture. This observation may be an example of ordering and compaction of the protein fold in the presence of high concentrations of sugars, such as been previously proposed with sucrose.28 Analysis of spectral overlay of well-resolved signals indicates that 63 proton resonances move upfield upon increasing sorbitol concentration, while 42 move downfield with Trp-58 showing the largest perturbation (Figure 5). The chemical shift of amide protons is very sensitive to bond length in both intramolecular and intermolecular hydrogen bonding. Variation of hydrogen bond strength has the following effect on proton chemical shifts. Strengthening of H-bonds decreases the donor (HN)–acceptor (C=O) distance producing a concomitant lengthening of the N–H bond length, resulting in a downfield chemical shift. The weakening of hydrogen bonding produces the opposite, an upfield chemical shift.29 If these amide proton upfield shifts are the result of a weakening of hydrogen bonds, especially inside helices, then the protein fold may be somewhat more “relaxed” and looser on conformational stability. The explanation for these upfield shifts may lie elsewhere. Previously, we have observed that increasing salt concentrations and varying anion type (buffer → Cl– → HPO42–) produced upfield changes of proton chemical shifts.7 It is conceivable that sorbitol may produce an analogous effect to the Hofmeister ion effect on protein stability. The presence of strongly hydrated anions such as SO4 or HPO42– reduces the amount of free water, resulting in a less hydrated protein and leading to a more stable conformation.30,31

Figure 5.

Figure 5

Open in a new tab

Titration of sorbitol to samples of 30 μM 1H–15N-Met-GCSF in 10 mM sodium deutero-acetate buffer at pH 4.0 measured by NMR. (A) Expansion of the overlay of 2D-1H–15N-HSQC (red 0 mM, green 274 mM 1× formulation concentration of sorbitol, and 749 mM 3× formulation concentration of sorbitol). A subset of resonances experiencing CSPs upon sorbitol addition are labeled. (B) All detected CSPs are mapped in red on the structure.

A selection of 42 residues with well-resolved resonances that exhibit downfield amide proton CSPs may experience a slight but detectable strengthening of an existing hydrogen bond, i.e., the shortening of the donor–acceptor distance, or the formation of a transient hydrogen bond resulting from a modification of the local conformational exchange. Examination of the ensemble of 10 structures derived from NMR spectroscopy3 (PDBID: 1GNC) provides some avenues to shed light on the above observed CSPs. Protein structures derived from X-ray crystallography data are obtained by fitting electron density maps. They have the advantage of producing precise local geometries (bond angles, etc.) but represent the protein conformation in a crystal that sometimes has local differences from its solution structure that experience motions. In contrast, NMR structures are models that fit all NMR measurements (experimental constraints such as NOE-derived internuclear distances, torsion angles, etc.) without any violations. The quality of NMR structures is reported in terms of root-mean-square (RMS) deviation between all models, which is directly dependent on the number of experimental constraints used in the calculation. Often times, loops and N- and C-termini are not well defined due to the lack of geometrical constraints resulting from a local conformational exchange or simply because they cannot be measured. The root-mean-square deviation (RMSD) for the backbone atoms is only 2.8 Å (3.4 Å for all atoms). Therefore, this low-resolution ensemble offers a set of 10 models that can be examined to identify possible transient interactions that could appear or be strengthened upon sorbitol addition. Analysis of these 10 models for the above 42 residues suggests that two types of hydrogen bonding may explain the observed CSP: intraresidue (C5 hydrogen bonds) and inter-residue hydrogen bonds.

Raines and co-workers have suggested the formation of C5 hydrogen bonds in β-sheet secondary structure elements that could afford a non-negligible contribution (0.25 kcal/mol) to the stabilization of the protein conformation.29 In this case, the donor–acceptor distance must be less than 2.5 Å with backbone Φ and Ψ torsion angles greater than 140° to allow an overlap of the carbonyl oxygen lone pair (ns) and the amide N–H σ* orbitals forming a “five-membered ring,” thereby the notation “C5 hydrogen bond”. The resonance of Trp-58 shows the largest downfield amide proton shift of 0.04 ppm. This residue in model #9 has the appropriate geometry to form a C5 hydrogen bond (Figure 6). This geometry is also present for Ser-63 in model #4 and Gly-100 in model #6 although with a smaller CSP of 0.012 ppm. Ile-56 can interact with Ser-53 in several models as depicted in Figure 6. All other residues exhibiting negative CSPs (downfield shifts) are mainly populating the extremities of helices and are involved in hydrogen bonding within α-helices. These perturbations may be seen as sensors indicating a tightening up of α-helices extremities (small reduction of the propensity of helices to fray) upon sorbitol addition.32,33 The presence of small CSP in the middle of helices may be consistent with a slight compaction of the fold whereby helices may be adopting a slight curvature. This would explain the presence of both upfield (lengthening of some hydrogen bond) and downfield (shortening of hydrogen bond) chemical shift changes that are observed. These observations may reveal that transient interactions (hydrogen bonds here) may have longer lifetime in the presence of some excipients, namely sorbitol in this case.

Figure 6.

Figure 6

Open in a new tab

Mapping of residues showing downfield amide proton shifts on the structure of Met-GCSF. For the ribbons, cyan represents small changes (0.020 ppm ≤ CCSD < 0.025 ppm), yellow represents moderate changes (0.025 ≤ CCSD < 0.030 ppm), and red represents largest changes (0.030 ppm ≤ CCSD). Residues showing largest shifts are identified (boxes) with the potential hydrogen bonds formed upon sorbitol addition (see text for details).

The small CSPs of a few ppb may be indicative of small but detectable stabilizing interactions by NMR but too small to shift thermal unfolding curves measured by CD. This may stem from the fact that a rise in temperature during the unfolding experiment may be sufficient to counteract the stabilizing effect afforded by the bulking agent upon increasing thermal agitation of the sorbitol–water mixture and filgrastim. Therefore, we set to measure dynamic-dependent parameters to see if sorbitol induces any changes in backbone dynamics in the fast-motional regime (picosecond to nanosecond). Here, measurements were only carried out at 3 times the formulation concentration (793 mM) at two fields. It is expected that at such a high concentration, viscosity effects will slow molecular tumbling (increase of the correlation time) of protein molecules, which in turn will increase T1 and decrease T2 while having little to no effects on the heteronuclear NOE (see Figure S4). As above, we calculated the order parameter (O2) with six relaxation parameters (T1, T2, and hetNOE measured at two fields: 600 and 700 MHz) using the model-free approach protocol in an attempt to determine the presence of any effects of sorbitol on the dynamics. Considering that we were trying to compare two samples with different viscosities (0 and 793 mM sorbitol), no diffusion tensor model was a priori selected to allow the search of the total tensor space. This resulted in a rotational diffusion tensor at high sorbitol content that was strikingly different than the tensor of the protein in pH 4.0 buffer (see Figure S6). The calculated order parameter per residue (see Figure S5) is again represented as average per structural elements as above in the pH study. At high sorbitol concentration, helices B and C along with loop BC are experiencing an increase of backbone amplitude of motion, while loop AB and loop CD show a reduction of motion amplitude (Figure 7A). This observation may be explained by the following. Replacement of the hydration shell with sorbitol produces a reduction of the polarity of the surrounding water, resulting in a weakening of the core hydrophobic interactions that keep the four helices packed together. Examination of the structure (Figure 7B,C) suggests that Helix B, Loop BC, and Helix C can move as an ensemble, which would induce stretching of loops AB and CD, thereby reducing their degrees of freedom. These structural elements do experience a slight reduction of mobility according to the order parameters.

Figure 7.

Figure 7

Open in a new tab

Effects of sorbitol on the dynamics of 1H–15N-Met-GCSF in 10 mM sodium deuteron-acetate buffer at pH 4.0 measured by NMR. (A) Order parameters calculated (see text) for Met-GCSF at 0 and 793 mM sorbitol. (B) Structure of Met-GCSF where helices B and C are colored cyan. (C) Cartoon representations of the observed change of order parameters upon addition of sorbitol. The addition of sorbitol induces a decrease of order parameters for helices B and C and loop BC. In this case, the trio helix B–loop BC–helix C experienced an increase of the amplitude of motion of their backbone amides (longer arrows on helices), which in turn reduces the flexibility of loop CD and to a lesser extent loop AB by reducing the number of degrees of freedom of these elements (shorter arrows). The dynamics of helices A and D are not influenced by sorbitol.

Conclusions

Excipients do not only act in an indirect fashion, such as providing a stable aqueous milieu for the protein to tumble into. If this were the case, one would expect that a single solution would fit all products. It is clearly not the case; therefore, manufacturers deploy significant resources to find the best mix of excipients for their particular product. Chemical shift perturbation and NMR-derived order parameters have both contributed to shed light on how excipients contribute to enhance the stability of the drug substance. Probing changes of backbone dynamics in the presence of product excipients could provide new insights into chemical modifications such as asparagine deamidation as these are influenced by backbone motions. While some findings proposed here could be applied to other protein therapeutics, a wider set of drug substances, including monoclonal antibody fragments, should be studied by NMR in parallel with other biophysical techniques to develop a more generalized understanding of excipient interactions at the atomic level.

Experimental Section

Sample Preparation

Expression and purification of 15N-Met-GSCF were carried out exactly as described previously.7 All samples of 15N-Met-GCSF for NMR spectroscopy were prepared from a stock protein solution at ∼100 μM protein concentration in 10 mM sodium acetate at pH 4 final volumes of 550 μL, while samples for CD used a formulation concentration of 30 μM of the protein. Samples at pH 5.0 and 6.0 were buffer-exchanged with 10 mM deuterated sodium acetate buffer at pH 5.0 and 6.0, respectively, using a 15 mL Amicon Ultrafiltration device with a 10 kDa molecular weight cutoff. The final concentration of 15N-Met-GCSF and the actual pH values after the buffer exchange were measured at ∼100 μM at pH 4, 5, and 6. Samples for relaxation measurements of excipients (sorbitol, polysorbate-80 and -20) were prepared by adding the appropriate amounts of individual excipients from concentrated stock solutions into a 100 μM 15N-Met-GCSF sample such as to minimize protein dilution. The concentrations of various excipients were sorbitol (0, 274, 749 mM for chemical shift perturbation (CSP) experiments and 793 mM for relaxation experiments), Polysorbate-80 (0, 30, 100, 300 μM), and Polysorbate-20 (0, 600, 1000, and 1400 μM). The initial 15N-Met-GCSF concentration was 100 μM; therefore, the molar ratio of [polysorbate 80]/[15N-Met-GCSF] of these four samples computed to 0:1 (0 × CMC), 0.33:1 (2.5 × CMC), 1:1 (5 × CMC), and 3:1 (25 × CMC). Stock solutions of 0.1, 1.0, and 9.0% (w/w) polysorbate 80 were used to prepare the desired detergent concentrations in the final NMR samples. Data collection of relaxation rates was acquired at 600 and 700 MHz at 25 °C. Samples containing polysorbate 20 (Sigma-Aldrich, St. Louis, MO) were prepared at concentrations of 0, 600, 1000, and 1400 μM. Here, the initial 15N-Met-GCSF concentration was 30 μM; thus, the molar ratio [polysorbate 20]/[15N-Met-GCSF] of these four samples computed to 0:1 (0 × CMC), 20:1 (11 × CMC), 35:1 (18 × CMC), and 50:1 (25 × CMC). A stock solution of 2.5% (w/w) polysorbate 20 was used to obtain the desired detergent concentrations in the final NMR samples.

Circular Dichroism Spectropolarimetry

Samples were prepared to the appropriate concentrations of protein and excipients and at pH 4 with a final volume for each condition of 600 μL. Sample analyses were run in a 1 mm path length quartz Suprasil cuvette (Hellma, Mullheim, Germany) on a Jasco 815 spectropolarimeter (Jasco International Co. Ltd., Tokyo, Japan) equipped with a Peltier thermal control unit. Jasco’s Spectra Manger Software controlled both the instrument and thermal control unit. Sample analyses included thermal denaturation between 20 and 90 °C at a rate of 1 °C/min, while monitoring Far-UV CD signal (in millidegrees) at 222 nm, with a data pitch of 1 nm and response time of 1 s. The data was recorded every 1 °C with an absolute error of ±0.5 °C on the measurement. Measurements were repeated three times for obtaining an estimate of the measurement error in the form of standard deviation. All spectra were corrected for buffer and/or excipient signal where appropriate. Fractional (normalized) change in signal was calculated according to previously published studies using the CDpal software.34−37

ANS Binding by Fluorescence Spectroscopy

8-Anilinonaphtalene-1-sulfonic acid (ANS, Sigma-Aldrich, St. Louis, MO) was dissolved in deuterated dimethyl sulfoxide (DMSO)-d6 (Cambridge Isotope Laboratories Incorporated, Andover, MA), at various concentrations (ranging between 1 and 30 μg/μL) and was added incrementally to ca. 200 μL of solution of 15N-Met-GCSF in 96-well plates (Costar 3603, Tissue Culture Treated polystyrene, Corning Incorporated, Corning, NY). Samples containing 10 and 30 μM of 15N-Met-GCSF in 10 mM sodium acetate-d3 at pH 4.0 were prepared. At each protein concentration, a total of 12 data points (titrations) with increasing molar ratios of ANS were collected such that [ANS]/[15N-Met-GCSF] were ranging between 0 and 27. Fluorescence measurements were carried out in duplicate for each protein concentration using a Biotek spectrometer model Epoch/SynergyMX. Excitation was carried out at 360 nm and emission at 400–600 nm. Data were analyzed using the Gen5 software.

ANS Binding by NMR Spectroscopy

This binding study was carried out with an initial 15N-Met-GCSF protein concentration of 30 μM with an ANS stock solution of 1 mg/mL (dissolved in 10 mM deuterated sodium acetate, pH 4.0), or an initial labeled protein concentration of 100 μM with a 29 mg/mL ANS (dissolved in DMSO-d6). Two-dimensional 15N-HSQC spectra acquired for 15N-Met-GCSF at both concentrations supplemented with ANS in molar ratios of [ANS]/[15N-Met-GCSF] = 0, 1, 3, and 6, respectively. Measurements were done in duplicate. A dissociation constant was calculated using eq 1(16)

graphic file with name ao0c04692_m001.jpg 1

Chemical shift changes, measured as combined chemical shift difference (CCSD),38 were monitored with increasing amounts of ANS, by recording 15N-HSQC experiments on 15N-Met-GCSF at 25 °C. In total, data for seven titrations with increasing molar ratios of ANS were collected, such that [ANS]/[15N-MET-GCSF] were 0, 0.1, 0.3, 1.0, 3.0, 6.0, and 9.0, respectively. 15N-Met-GCSF showed a strong tendency to aggregate at ANS molar ratio larger than 9.0.

A dissociation constant was calculated for each residue that showed chemical shift perturbation using eq 2(38)

graphic file with name ao0c04692_m002.jpg 2

where Δδobs is the change in the observed chemical shift from the free state, Δδmax is the maximum shift change that occurs at saturation, n is the number of equivalent sites, [P]t is the total protein concentration, and [L]t is the total ligand concentration. While Δδobs, [P]t, and [L]t were known for each titration step, Δδmax, n, and Kd were optimized in a nonlinear least-square fit of the model to NMR data using Microsoft Excel.

Monitoring 15N-Met-GCSF Chemical Shift Changes versus Polysorbate 20

Polysorbate 20 was ordered from four different sources: TCI America (Portland, OR, manufactured in Japan), Sigma-Aldrich (St. Louis, MO, manufactured in Switzerland), Bio-Rad Laboratories Incorporations (manufactured in USA), and Acros Organics (Fair Lawn, NJ, manufactured in Belgium).

Chemical shift changes were monitored by 15N-HSQC and 13C-HSQC as a function of added polysorbate 20. Three series of titrations were conducted. First, polysorbate 20 was titrated into NMR buffer (10 mM deuterated sodium acetate-d3, pH 4.0) at concentrations of 206, 403, and 1411 μM and 15N-HSQC and 13C-HSQC spectra were recorded for each concentration. Following the buffer-only titration, a sample of 30 μM 15N-Met-GCSF was prepared and titrated with polysorbate 20, such that the final polysorbate concentrations were 0, 203, and 406 μM, corresponding to molar ratios [polysorbate 20]/[15N-Met-GCSF] of 0:1 (0 × CMC), 6.8:1 (3.7 × CMC), 13.5:1 (7.4 × CMC), and 15N-HSQC and 13C-HSQC spectra were recorded for each concentration. Finally, a sample of 100 μM 15N-Met-GCSF was prepared and titrated with polysorbate 20, such that the final polysorbate concentrations were 0, 32, 64, 100, 203, and 406 μM, corresponding to molar ratios [polysorbate 20]/[15N-Met-GCSF] of 0:1 (0 × CMC), 0.3:1 (0.6 × CMC), 0.6:1 (1.2 × CMC), 1:1 (1.8 × CMC), 2:1 (3.7 × CMC), and 4:1 (7.4 × CMC) and 15N-HSQC and 13C-HSQC spectra were recorded for each concentration. The stock solutions of polysorbate 20 that were used for titrations were either at 2.5% (w/w) or at 1% (w/w) and were prepared from polysorbate 20 purchased from Sigma-Aldrich (St. Louis, MO, manufactured in Switzerland).

Effect of Sorbitol

d-Sorbitol was purchased from Sigma-Aldrich, St. Louis, MO. d-Sorbitol stock solution (∼4.0 M) was prepared in 10 mM deuterated sodium acetate at pH 4.0.

15N-Met-GCSF samples with final concentrations of 100 μM and final volumes of 550 μL were prepared in 10 mM deuterated sodium acetate at pH 4.0. For chemical shift changes, 15N-HSQC experiments were acquired at three different sorbitol concentrations, namely 0, 274, and 749 mM. Initially, experiments were carried out at 25 °C. However, as the backbone amide nitrogen for Trp-58 is well resolved at 45 °C, the 15N-HSQC experiments were repeated at 45 °C. For spin relaxation studies, T1, T2, and heteronuclear NOE experiments were acquired at two different sorbitol concentrations, namely 0 and 749 mM. Relaxation experiments were acquired at both 600 and 700 MHz fields at 25 °C. The collected spin relaxation data for the sorbitol study was analyzed in nmr-relax25,26 as described in the NMR Spectroscopy section below.

NMR Spectroscopy

Sensitivity-enhanced, temperature-compensated, interleaved experiments from the Bruker library (Topspin, version 3.6) were utilized for samples at pH 4, 5, and 6. Data were acquired on a Bruker 600 MHz Avance III and a Bruker 700 MHz Avance III-HD (Milton, ON) equipped with TCI with cryo-probes. The pulse program hsqct1etf3gpsitc3d from the manufacturer library was used to determine relaxation times T1 using the following delays of 20, 40, 120, 200 (×2), 300, 450, 600, 800, 1000, 1200 (×2), 1400, and 1800 ms. The T2 relaxation time were determined using the pulse program hsqct2etf3gpsitc3d.2 with delays of 0, 8.48, 16.96, 25.44 (×2), 33.92, 59.36, 84.8, 110.24, 135.68 (×2), 161.12, and 195.04 ms. Total data acquisition time are on the order of 60:24 h each for T1 and T2 experiments and 12 h for the heteronuclear NOE experiments.

Heteronuclear NOE experiments were collected using the pulse program hsqcnoef3gpsi3d. The spectral widths for direct and indirect dimensions were 16.0 and 29.0 ppm, respectively, with center frequencies for the two carriers being set at water resonance and 117.5 ppm, respectively. A total of 256 points were acquired in the indirect dimension corresponding to an acquisition time of 100 ms. The interscan delay was set to 1.5 s for T1 and T2 experiments and to 3.0 s for the heteronuclear NOE experiment. For all subsequent 15N-based NMR experiments, these settings were used, with the exception of interscan delay, which was set to 1 s for 15N-HSQC experiments. For error estimations in R1 and R2 rates, duplicate measurements were performed for two different relaxation delays. For heteronuclear NOE experiments, an error of 5% was assigned based on the work of Montelione and co-workers.39,40

Spectra were processed and relaxation rates and heteronuclear NOEs were extracted using nmrPipe.41 Chemical shifts were extracted using Sparky version 3.1 and NMRFam-Sparky version 1.2.42,43

Calculation of order parameters through model-free formalism,22−24,44−46 the program nmr-relax, version 4.0.325,26,47 installed on a Mac Pro with 3 GHz eight-core Intel Xeon E5 processors was employed. The first model of the ensemble of solution NMR structures of Met-GCSF (pdb accession code: 1GNC) was chosen for setting up dipolar interactions and chemical shift anisotropy mechanisms. Six data sets (i.e., R1, R2, and heteronuclear NOE collected at both 600 and 700 MHz) were used as input data for each pH experiment. Adopting the d’Auvergne protocol, initially local correlation times for each residue were determined, and models with internal correlation times (tm0, tm1, etc.) were selected based on Akaike’s Information Criteria (AIC) and chi-square in nmr-relax. The total computational time for this step was ∼48 h. The diffusion tensor was then chosen (DIFF_MODEL = “ellipsoid”) and optimized through iterations with Newton’s minimization algorithm (MIN_ALGOR = “newton”) until a solution was found. The total number of iterations for ellipsoid as diffusion tensor was 6 (7 cycles including the initial “init” cycle), with each iteration taking anything from 90 min to several days to reach completion. Visual inspection of diffusion tensors showed that after five cycles of calculations the optimal tensor was found, and the calculations had converged. The number of Monte-Carlo steps was set to 50, and the size of the grid search was 11 × 11 (GRID_INC = 11) for all calculations. This analysis procedure was repeated for every experimental pH (4.0, 5.0, and 6.0).

Acknowledgments

The help of Sara Ahmadi and Derek Hodgson is greatly acknowledged for the preparation of isotopically labeled Met-GSCF samples for this study. Dr. Roger Tam and Dr. Simon Sauvé are acknowledged for the critical reading of the manuscript.

Glossary

Abbreviations

ANS

8-anilinonaphtalene-1-sulfonic acid

API

active pharmaceutical ingredient

CMC

critical micelle concentration

CSP

chemical shift perturbation

HetNOE

heteronuclear nuclear Overhauser effect

HSQC

heteronuclear single quantum correlation

Met-GCSF

recombinant methionyl granulocyte-colony stimulating factor

NMR

nuclear magnetic resonance spectroscopy

PS-20 and -80

polysorbate-20 and -80

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.0c04692.

  • Effects of polysorbate-80 on relaxation parameter of Met-GCSF at 600 MHz (Figure S1); effects of polysorbate-20 on relaxation parameter of Met-GCSF at 600 MHz (Figure S2); effects of pH 4.0 and 6.0 on relaxation parameter of Met-GCSF at 600 MHz (Figure S3); effects of sorbitol at 0 and 793 mM on relaxation parameter of Met-GCSF at 600 MHz (Figure S4); and order parameters calculated for Met-GCSF at 0 and 793 mM sorbitol and optimized diffusion tensor for Met-GCSF in the absence and presence of 793 mM sorbitol (Figure S5) (PDF)

The authors declare no competing financial interest.

Supplementary Material

ao0c04692_si_001.pdf (1.3MB, pdf)

References

  1. Kamerzell T. J.; Esfandiary R.; Joshi S. B.; Middaugh C. R.; Volkin D. B. Protein–excipient interactions: Mechanisms and biophysical characterization applied to protein formulation development. Adv. Drug Delivery Rev. 2011, 63, 1118–1159. 10.1016/j.addr.2011.07.006. [DOI] [PubMed] [Google Scholar]
  2. Hill C. P.; Osslund T. D.; Eisenberg D. The structure of granulocyte-colony-stimulating factor and its relationship to other growth factors. Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 5167. 10.1073/pnas.90.11.5167. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Zink T.; Ross A.; Luers K.; Cieslar C.; Rudolph R.; Holak T. A. Structure and dynamics of the human granulocyte colony-stimulating factor determined by NMR spectroscopy. Loop mobility in a four-helix-bundle protein. Biochemistry 1994, 33, 8453–8463. 10.1021/bi00194a009. [DOI] [PubMed] [Google Scholar]
  4. Arakawa T.; Prestrelski S. J.; Narhi L. O.; Boone T. C.; Kenney W. C. Cysteine 17 of recombinant human granulocyte-colony stimulating factor is partially solvent-exposed. J. Protein Chem. 1993, 12, 525–531. 10.1007/BF01025117. [DOI] [PubMed] [Google Scholar]
  5. Ricci M. S.; Sarkar C. A.; Fallon E. M.; Lauffenburger D. A.; Brems D. N. pH Dependence of structural stability of interleukin-2 and granulocyte colony-stimulating factor. Protein Sci. 2003, 12, 1030–1038. 10.1110/ps.0230103. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Narhi L. O.; Kenney W. C.; Arakawa T. Conformational changes of recombinant human granulocyte-colony stimulating factor induced by pH and guanidine hydrochloride. J. Protein Chem. 1991, 10, 359–367. 10.1007/BF01025250. [DOI] [PubMed] [Google Scholar]
  7. Aubin Y.; Hodgson D. J.; Thach W. B.; Gingras G.; Sauve S. Monitoring Effects of Excipients, Formulation Parameters and Mutations on the High Order Structure of Filgrastim by NMR. Pharm. Res. 2015, 32, 3365–3375. 10.1007/s11095-015-1713-3. [DOI] [PubMed] [Google Scholar]
  8. Cavanagh J.; Fairbrother W. J.; Plamer A. G.; Rance M.; Skelton N. J.. Protein NMR Spectroscopy, Principles and Practice, 2nd ed.; Elsevier Academic Press: San Diego, 2007. [Google Scholar]
  9. Akke M.; Brueschweiler R.; Palmer A. G. NMR order parameters and free energy: an analytical approach and its application to cooperative calcium(2+) binding by calbindin D9k. J. Am. Chem. Soc. 1993, 115, 9832–9833. 10.1021/ja00074a073. [DOI] [Google Scholar]
  10. Sharp K. A.; O’Brien E.; Kasinath V.; Wand A. J. On the relationship between NMR-derived amide order parameters and protein backbone entropy changes. Proteins 2015, 83, 922–930. 10.1002/prot.24789. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Yang D.; Kay L. E. Contributions to Conformational Entropy Arising from Bond Vector Fluctuations Measured from NMR-Derived Order Parameters: Application to Protein Folding. J. Mol. Biol. 1996, 263, 369–382. 10.1006/jmbi.1996.0581. [DOI] [PubMed] [Google Scholar]
  12. Yang D.; Mok Y.-K.; Forman-Kay J. D.; Farrow N. A.; Kay L. E. Contributions to protein entropy and heat capacity from bond vector motions measured by NMR spin relaxation. J. Mol. Biol. 1997, 272, 790–804. 10.1006/jmbi.1997.1285. [DOI] [PubMed] [Google Scholar]
  13. Herman A. C.; Boone T. C.; Lu H. C.. Characterization, Formulation and Stability of Neupogen (Filgrastim), a Recombinant Human Granulocyte-Colony Stimulating Factor. In Formulation, Characterization and Stability of Protein Drugs; Pearlman R.; Wang J., Eds.; Plenum Press: New York, 1996; pp 303–328. [DOI] [PubMed] [Google Scholar]
  14. Ishida H.; Huang H.; Yamniuk A. P.; Takaya Y.; Vogel H. J. The Solution Structures of Two Soybean Calmodulin Isoforms Provide a Structural Basis for Their Selective Target Activation Properties. J. Biol. Chem. 2008, 283, 14619–14628. 10.1074/jbc.M801398200. [DOI] [PubMed] [Google Scholar]
  15. Latypov R. F.; Liu D.; Gunasekaran K.; Harvey T. S.; Razinkov V. I.; Raibekas A. A. Structural and thermodynamic effects of ANS binding to human interleukin-1 receptor antagonist. Protein Sci. 2008, 17, 652–663. 10.1110/ps.073332408. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Schonbrunn E.; Eschenburg S.; Luger K.; Kabsch W.; Amrhein N. Structural basis for the interaction of the fluorescence probe 8-anilino-1-naphthalene sulfonate (ANS) with the antibiotic target MurA. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 6345–6349. 10.1073/pnas.120120397. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Singh S. M.; Bandi S.; Jones D. N. M.; Mallela K. M. G. Effect of Polysorbate 20 and Polysorbate 80 on the Higher-Order Structure of a Monoclonal Antibody and Its Fab and Fc Fragments Probed Using 2D Nuclear Magnetic Resonance Spectroscopy. J. Pharm. Sci. 2017, 106, 3486–3498. 10.1016/j.xphs.2017.08.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Panjwani N.; Hodgson D. J.; Sauve S.; Aubin Y. Assessment of the effects of pH, formulation and deformulation on the conformation of interferon alpha-2 by NMR. J. Pharm. Sci. 2010, 99, 3334–3342. 10.1002/jps.22105. [DOI] [PubMed] [Google Scholar]
  19. Chang P. K.; Prestidge C. A.; Barnes T. J.; Bremmell K. E. Impact of PEGylation and non-ionic surfactants on the physical stability of the therapeutic protein filgrastim (G-CSF). RSC Adv. 2016, 6, 78970–78978. 10.1039/C6RA16254A. [DOI] [Google Scholar]
  20. Ricci M. S.; Sarkar C. A.; Fallon E. M.; Lauffenburger D. A.; Brems D. N. pH Dependence of structural stability of interleukin-2 and granulocyte colony-stimulating factor. Protein Sci. 2003, 12, 1030–1038. 10.1110/ps.0230103. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Kolvenbach C. G.; Narhi L. O.; Philo J. S.; Li T.; Zhang M.; Arakawa T. Granulocyte-colony stimulating factor maintains a thermally stable, compact, partially folded structure at pH 2. J. Peptide Res. 1997, 50, 310–318. 10.1111/j.1399-3011.1997.tb01472.x. [DOI] [PubMed] [Google Scholar]
  22. d’Auvergne E. J.; Gooley P. R. The use of model selection in the model-free analysis of protein dynamics. J. Biomol. NMR 2003, 25, 25–39. 10.1023/A:1021902006114. [DOI] [PubMed] [Google Scholar]
  23. d’Auvergne E. J.; Gooley P. R. Model-free model elimination: a new step in the model-free dynamic analysis of NMR relaxation data. J. Biomol. NMR 2006, 35, 117–135. 10.1007/s10858-006-9007-z. [DOI] [PubMed] [Google Scholar]
  24. d’Auvergne E. J.; Gooley P. R. Set theory formulation of the model-free problem and the diffusion seeded model-free paradigm. Mol. BioSyst. 2007, 3, 483–494. 10.1039/b702202f. [DOI] [PubMed] [Google Scholar]
  25. d’Auvergne E. J.; Gooley P. R. Optimisation of NMR dynamic models II. A new methodology for the dual optimisation of the model-free parameters and the Brownian rotational diffusion tensor. J. Biomol. NMR 2008, 40, 121–133. 10.1007/s10858-007-9213-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. d’Auvergne E. J.; Gooley P. R. Optimisation of NMR dynamic models I. Minimisation algorithms and their performance within the model-free and Brownian rotational diffusion spaces. J. Biomol. NMR 2008, 40, 107–119. 10.1007/s10858-007-9214-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Vinther J. M.; Kristensen S. M.; Led J. J. Enhanced Stability of a Protein with Increasing Temperature. J. Am. Chem. Soc. 2011, 133, 271–278. 10.1021/ja105388k. [DOI] [PubMed] [Google Scholar]
  28. Kendrick B. S.; Chang B. S.; Arakawa T.; Peterson B.; Randolph T. W.; Manning M. C.; Carpenter J. F. Preferential exclusion of sucrose from recombinant interleukin-1 receptor antagonist: Role in restricted conformational mobility and compaction of native state. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 11917. 10.1073/pnas.94.22.11917. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Newberry R. W.; Raines R. T. A prevalent intraresidue hydrogen bond stabilizes proteins. Nat. Chem. Biol. 2016, 12, 1084–1088. 10.1038/nchembio.2206. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Ahmed M.; Namboodiri V.; Singh A. K.; Mondal J. A. On the intermolecular vibrational coupling, hydrogen bonding, and librational freedom of water in the hydration shell of mono- and bivalent anions. J. Chem. Phys. 2014, 141, 164708 10.1063/1.4899070. [DOI] [PubMed] [Google Scholar]
  31. Metrick M. A.; MacDonald G. Hofmeister ion effects on the solvation and thermal stability of model proteins lysozyme and myoglobin. Colloids Surf., A 2015, 469, 242–251. 10.1016/j.colsurfa.2015.01.038. [DOI] [Google Scholar]
  32. Fesinmeyer R. M.; Peterson E. S.; Dyer R. B.; Andersen N. H. Studies of helix fraying and solvation using 13C′ isotopomers. Protein Sci. 2005, 14, 2324–2332. 10.1110/ps.051510705. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Fierz B.; Reiner A.; Kiefhaber T. Local conformational dynamics in alpha-helices measured by fast triplet transfer. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 1057–1062. 10.1073/pnas.0808581106. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Niklasson M.; Andresen C.; Helander S.; Roth M. G.; Zimdahl Kahlin A.; Lindqvist Appell M.; Martensson L. G.; Lundstrom P. Robust and convenient analysis of protein thermal and chemical stability. Protein Sci. 2015, 24, 2055–2062. 10.1002/pro.2809. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Sakakura M.; Tamura M.; Fujii N.; Takeuchi T.; Hatanaka T.; Kishimoto S.; Arata Y.; Takahashi H. Structural mechanisms for the S-nitrosylation-derived protection of mouse galectin-2 from oxidation-induced inactivation revealed by NMR. FEBS J. 2018, 285, 1129–1145. 10.1111/febs.14397. [DOI] [PubMed] [Google Scholar]
  36. Wennerstrand P.; Blissing A.; Martensson L. G. In Vitro Protein Stability of Two Naturally Occurring Thiopurine S-Methyltransferase Variants: Biophysical Characterization of TPMT*6 and TPMT*8. ACS Omega 2017, 2, 4991–4999. 10.1021/acsomega.7b00801. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Smith H. I.; Guthertz N.; Cawood E. E.; Maya-Martinez R.; Breeze A. L.; Radford S. E. The role of the IT-state in D76N β2-microglobulin amyloid assembly: A crucial intermediate or an innocuous bystander?. J. Biol. Chem. 2020, 295, 12474–12484. 10.1074/jbc.RA120.014901. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Williamson M. P. Using chemical shift perturbation to characterise ligand binding. Prog. Nucl. Magn. Reson. Spectrosc. 2013, 73, 1–16. 10.1016/j.pnmrs.2013.02.001. [DOI] [PubMed] [Google Scholar]
  39. Andrec M.; Montelione G. T.; Levy R. M. Estimation of Dynamic Parameters from NMR Relaxation Data using the Lipari–Szabo Model-Free Approach and Bayesian Statistical Methods. J. Magn. Res. 1999, 139, 408–421. 10.1006/jmre.1999.1839. [DOI] [PubMed] [Google Scholar]
  40. Li Y. C.; Montelione G. T. Human type-alpha transforming growth factor undergoes slow conformational exchange between multiple backbone conformations as characterized by nitrogen-15 relaxation measurements. Biochemistry 1995, 34, 2408–2423. 10.1021/bi00008a003. [DOI] [PubMed] [Google Scholar]
  41. Delaglio F.; Grzesiek S.; Vuister G. W.; Zhu G.; Pfeifer J.; Bax A. NMRPipe: a multidimensional spectral processing system based on UNIX pipes. J. Biomol. NMR 1995, 6, 277–293. 10.1007/BF00197809. [DOI] [PubMed] [Google Scholar]
  42. Goddard T. D.; Kneller D. G.. SPARKY 3; University of California: San Francisco, 2008.
  43. Lee W.; Tonelli M.; Markley J. L. NMRFAM-SPARKY: enhanced software for biomolecular NMR spectroscopy. Bioinformatics 2015, 31, 1325–1327. 10.1093/bioinformatics/btu830. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Lipari G.; Szabo A. Model-free approach to the interpretation of nuclear magnetic resonance relaxation in macromolecules. 1. Theory and range of validity. J. Am. Chem. Soc. 1982, 104, 4546–4559. 10.1021/ja00381a009. [DOI] [Google Scholar]
  45. Lipari G.; Szabo A. Model-free approach to the interpretation of nuclear magnetic resonance relaxation in macromolecules. 2. Analysis of experimental results. J. Am. Chem. Soc. 1982, 104, 4559–4570. 10.1021/ja00381a010. [DOI] [Google Scholar]
  46. Clore G. M.; Szabo A.; Bax A.; Kay L. E.; Driscoll P. C.; Gronenborn A. M. Deviations from the simple two-parameter model-free approach to the interpretation of nitrogen-15 nuclear magnetic relaxation of proteins. J. Am. Chem. Soc. 1990, 112, 4989–4991. 10.1021/ja00168a070. [DOI] [Google Scholar]
  47. Bieri M.; d’Auvergne E. J.; Gooley P. R. relaxGUI: a new software for fast and simple NMR relaxation data analysis and calculation of ps-ns and mus motion of proteins. J. Biomol. NMR 2011, 50, 147–155. 10.1007/s10858-011-9509-1. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

ao0c04692_si_001.pdf (1.3MB, pdf)

Articles from ACS Omega are provided here courtesy of American Chemical Society

RESOURCES