Anion Effect on Phase Separation of Polyethylene Glycol-8000–Sodium Salt Two-Phase Systems

Abstract

Aqueous two-phase systems (ATPSs) are formed when two nonionic polymers, or a single polymer and salt, are mixed in water above a specific concentration, resulting in the emergence of phase separation and the formation of two immiscible aqueous phases. The solvent properties of the aqueous media within the phases of ATPSs rely on the specific composition of the co-solutes and the arrangement of the hydrogen bond network within each phase. Here, we investigate the anion effect of various sodium salts on the enhancement or destabilization of polyethylene glycol (PEG)–salt ATPS formation. Relatively small changes in ATPS ionic composition were shown to result in significant changes in solute partitioning. Additionally, we previously established that the arrangement of hydrogen bonds within the coexisting phases of ATPSs is different, as evidenced by Attenuated Total Reflection—Fourier Transform Infrared (ATR-FTIR) spectroscopic analysis of OH-stretch bands. The hydrogen bond arrangement was shown to abruptly change at concentrations below the threshold of macroscopic phase separation in the ATPSs. Using dynamic light scattering (DLS), we observed a correlation between these abrupt changes in H-bond arrangement and the detection of agglomerate formation in both polymer–polymer and polymer–salt systems.

1. Introduction

Liquid water, commonly referred to as the universal solvent, is frequently misunderstood as a passive matrix for diffusive macromolecules in biological systems. Water, in its own right, is a complex, highly structured fluid due to its ability to form an extensive network of hydrogen bonds (H-bonds) [1,2]. The structure and dynamics of water as a solvent are driven by both water–water interactions and the introduction of a solute molecule to this highly complex network structure. The physicochemical properties of aqueous solutions rely heavily on the nature and concentration of a given solute. The many types of solute–solvent interactions include electrostatic, dipole–dipole, dipole-induced dipole, hydrogen bonding, and electron pair donor–acceptor interactions. Experimentally observed differences in the solvent properties of water in aqueous solutions have been shown to originate from (or correlate with) the rearrangement of water’s H-bond network [3,4,5,6,7,8].

Vibrational spectroscopy (e.g., infrared (IR) and Raman) can provide microscopic information on both water–water and water–solute molecular interactions. The spectral distributions of high and low wavenumber parts within the broad band in the IR spectra (4000–2500 cm⁻¹) have been assigned to reflect the inhomogeneity of the H-bond network of a given aqueous solution [9,10,11,12,13,14,15,16,17,18]. Various groups have assigned specific wavenumber positions to water molecules existing in different H-bonded environments [19,20,21,22,23,24,25,26,27,28,29]. Components at lower optical frequencies are generally assigned to water molecules forming strong, ice-like, H-bonds, while those at higher frequencies are assigned to water molecules in an environment with weaker and/or distorted H-bonds [5,19,21,26,30].

We established a model that decomposes the OH-stretch band into four Gaussian components, which were assigned to four different subpopulations of water [4,5,6,7,8,26]. Through fitting the OH-stretch band in pure water and various aqueous solutions with a model using the sum of three, four, or five Gaussian distributions with floated peak frequencies, amplitudes, and widths, we found that the best and most reliable fits were obtained with four Gaussians with peak locations, which were in good agreement with the well-accepted literature values [3,21,26]. The position of each Gaussian component has since been confirmed and fixed for all analyses of pure water and aqueous solutions at 3080 cm⁻¹, 3230 cm⁻¹, 3400 cm⁻¹, and 3550 cm⁻¹ [3,4,6,7,8]. It should be noted however, that there is some disagreement in defining the specific wavenumber, position, and assignment of these components; this is most likely due to differences in the experimental conditions and techniques [19,20,22,23,24,25,26,27,28,29]. Although the estimated relative contributions of these components depend on the solute type and concentration, the resulting Gaussian components from our FTIR spectrum decompositions do not have well-defined physical meanings; therefore, we do not assign specific vibrational transitions to individual molecules but rather provide an approximation of four simultaneously existing water subpopulations. From the analysis of the different assignments of these and other differently positioned components used in the literature [5,11,12,19,21,25,26,30], we assigned these four subpopulations of water as follows: water with four tetrahedrally arranged H-bonds (3080 cm⁻¹); water with four distorted H-bonds (3230 cm⁻¹); water with three or four loosely arranged H-bonds (3400 cm⁻¹); and water with three, two, or one H-bond(s) (3550 cm⁻¹).

Light scattering spectroscopic techniques (e.g., dynamic and static light scattering) are commonly used to characterize polymer solution structure and dynamics [6,7,31,32]. Dynamic light scattering (DLS) can be used to estimate the size of macromolecules in solution by characterizing the time evolution of Brownian motion-driven fluctuations in the intensity of light scattered by a specific sample [33]. We have previously reported the emergence of polymer agglomerates in both PEG–polymer and PEG–Na₂SO₄ ATPSs prior to visual phase separation using DLS [7]. The emergence of these agglomerates, detected via DLS, strongly correlates with the abrupt re-arrangement of H-bonds in aqueous mixtures of polymer–polymer and polymer–salt ATPSs [6,7].

The OH-stretch band of water is highly sensitive to interactions between ions and water molecules, with different sensitivity to cations compared to anions. Anions directly interact with H atoms and therefore more strongly affect the OH stretch [21]. The theoretical effect of salts on the H-bond network of water has been shown to correlate with the empirical Hofmeister series of ion effects [34]. One study [35] on the influence of the Hofmeister series of sodium salts on the solvent properties of water showed that the relative effects of examined salts (Na₂SO₄, NaF, CH₃COONa, NaCl, NaBr, NaI, NaClO₄, and NaSCN) were strongly correlated with the linear combination of the ionic water structural entropy and anion static polarizability. The change in the anion of an electrolyte has also been observed to produce a greater effect on aqueous two-phase systems (ATPSs) formed with polyethylene glycol and various salts than an equivalent change in the cation [36]. Here, we explore the effect of anions in various sodium salts on H-bond strength and the effect on phase separation of polymer/salt ATPSs via bionodal determination, ATR-FTIR spectroscopic analysis, and DLS.

2. Materials and Methods

2.1. Materials

Polyethylene glycol (PEG-8000, cat. P2139 lot. SLCP8509) with an average molecular weight of 8000 g/mol, Na₃HPO₄.12H₂O (cat. 222003, lot MKCS0002), NaBr (cat. 310506, lot MKCT2413), and NaNO₃ (cat. S5506, lot MKC2019) were obtained from EMD Millipore Sigma (Burlington, MA, USA). Na₂HPO₄.7H₂O (cat. S373, lot 190833), NaH₂PO₄.H₂O (cat. S369, lot 194198), NaClO₄ (cat. S490, lot 225908), and Na₂SO₄ (cat. S421, lot 195740) were obtained from Fisher Scientific (Hampton, NH, USA). A sodium phosphate buffer (NaPB) solution at pH 7.4 was prepared by mixing 21.7 g of Na₂HPO₄.7H₂O with 2.6 g of NaH₂PO₄.H₂O in up to 200 mL of water. All the solutions were prepared with UltraPure water (ELGA PURELAB^® flex2 water purification system, 18.2 MΩ, TOC ≤ 5 ppb) and 0.01 wt.% NaN₃ (Sigma, cat. S2002, lot MKCQ1781) as a preservative.

2.2. Methods

Aqueous two-phase systems (ATPSs) for FTIR and DLS measurements were prepared by mixing appropriate amounts of the aqueous stock solutions for PEG-8000 and specific salt to final compositions (measured by weight), as listed in

Table 1. Compositions of individual PEG–salt ATPSs used in this study. All systems contained 0.01 M sodium phosphate buffer (NaPB), pH 7.4.

Component 1	Weight Fraction	Component 2	Weight Fraction
PEG-8000	0.106	Na2SO4	0.0698
PEG-8000	0.179	Na2HPO4	0.040
PEG-8000	0.156	NaH2PO4	0.110
PEG-8000	0.087	NaNO3	0.311
PEG-8000	0.061	NaClO4	0.347
PEG-8000	0.060	Na2SO4	0.052
PEG-8000	0.098	NaBr	0.410

. For the ATR-FTIR and DLS measurements, each ATPS was vigorously mixed and quickly aliquoted, and each aliquot was diluted with 0.01 M sodium phosphate buffer (NaPB), pH 7.4, to 50–100 wt.% of the initial mixture.

Binodal curves were obtained experimentally using the turbidometric titration method [37]. Different biphasic systems, with known PEG-8000 and salt concentrations, were prepared by weight in assay tubes. The tubes were vigorously mixed and examined for turbidity (i.e., evidence of a two-phase system). Diluent (in this case, 0.01M NaPB, pH 7.4) was added dropwise, and the assay tubes were reweighed, vigorously mixed, and again examined for turbidity. This process was repeated until a single homogenous phase was obtained. A minimum of seven points were determined for each set of ATPSs.

Absorbance spectra for each sample were measured in two separately prepared solutions using an ALPHA II Fourier Transform Infrared (FT-IR) spectrometer equipped with a Platinum ATR single-reflection diamond module (Bruker Scientific, LLC, Billerica, MA, USA). All the measurements were performed at ambient temperature (approximately 23 °C) using 24 scans for each sample and 24 scans for background in the spectral range of 4000–1000 cm⁻¹ with a resolution of 4 cm⁻¹. The spectra were reproducible to within ±1 cm⁻¹. The ATR-FTIR spectra were analyzed using custom software written in Wolfram Mathematica and run under version 12. Details on the code and protocol used can be found in [8].

Hydrodynamic size and size distributions were estimated by dynamic light scattering (DLS, Malvern ZS Zetasizer Nano ZSP) with a He-Ne laser (633 nm, 4 mW) as a light source, as described in [7]. Analysis was performed using the light scattering software DTS application. The scattering light was collected at a 173° backscattering angle and a wavelength of 633 nm at ambient temperature (~25 °C). Ten scans of 10 s each were performed under the studied conditions. The reported values were an average of at least six measurements.

3. Results

Polyethylene glycol (PEG) with an average molecular weight of 8000 g/mol was combined with different sodium salts (sulfate, chlorate, bromide, nitrate, and various phosphates) to study the anion effect on phase separation. By examining various PEG–salt phase diagrams established under the same conditions (e.g., buffer, pH, and temperature), we can establish the concentration of phase-forming components necessary for two phases to emerge. For each set of systems, the experimental data obtained for the binodal curves were adjusted to the empirical equation suggested by Merchuk and co-workers [38]:

$$W_{PEG}=a\exp [b{\left(W_{salt}\right)}^{0.5}-c{\left(W_{salt}\right)}^3]$$

(1)

where $W_{PEG}$ and $W_{salt}$ are the polymer and salt compositions in mass fraction, and $a,b,$ and $c$ are adjustable parameters obtained by nonlinear regression. Fit parameters can be found in Table S1 in Supplementary Materials. The resulting phase diagrams for the seven PEG–sodium salt ATPSs are shown in Figure 1. The area below each binodal line corresponds to the compositions that result in a single homogenous solution, whereas the area above each binodal line corresponds to the compositions that result in biphasic systems.

For visual comparison, each binodal curve from the seven sets of PEG–sodium salt ATPSs are shown together in Figure 2. Here, we observe that by changing the anion in the sodium salt used to form PEG–salt ATPSs the binodal line shifts either to the right or to the left. These shifts indicate that either increased salt concentration is required to form two phases (right) or less salt is required for ATPS formation (left).

From these data, we can conclude that, empirically, the anion effect on the efficiency of phase separation in PEG–sodium salt systems is as follows: PO₄³⁻ > HPO₄²⁻ > SO₄²⁻ > H₂PO₄⁻ > ClO₄⁻ > NO₃⁻ > Br⁻. The electrostatic nature of the anion appears to play a role in the enhancement or destabilization of phase separation in PEG–sodium salt ATPSs. The above series of anions can be compared to the Hofmeister series, which describes the order of ions in terms of chaotropic (structure-breaking) to kosmotropic (structure-making) capacity in aqueous solutions. The order of the co-solvents is determined empirically based on their capacity to induce protein salting-out. For the anions used in this study, the order of kosmotropicity to chaotropicity is as follows [39,40]: SO₄²⁻ > H₂PO₄⁻ > Br⁻ > NO₃⁻ > ClO₄⁻. Our data appear to follow a similar trend for the kosmotropes when compared to the classical Hofmeister series; however, we observe a switch in positions between ClO₄⁻ and Br⁻ in terms of which salt requires a higher concentration for the emergence of biphasic systems. A comparison of the structural parameters of the first hydration shell of Ca²⁺ in aqueous solutions reported that CaBr₂ had a hydration number of 6, versus that of 7–8 for Ca(ClO₄)₂ [41]. The hydration number for Br- determined via neutron diffraction and X-ray absorption can range from 6 to 7.4, whereas that of ClO₄⁻ was determined to be 8 or greater [42,43]. Multiple groups have suggested an inverse relationship between the hydration strength of an ion and the concentration required to form a two-phase system [44,45,46,47,48,49,50]. The data presented here support this hypothesis.

In a similar manner to what we observe here, previous works have also reported a switch between phosphates and sulfates in terms of which anion is more effective at phase formation in PEG–salt systems [36,50,51]. Our data suggest that phosphates may be more efficient than sulfates at the same, or higher, valency. In the classical Hofmeister series, the only phosphate studied was H₂PO₄⁻, and our data for this monovalent anion in comparison to SO₄²⁻ align in terms of kosmotropicity.

Kosmotropicity is related to high charge density (e.g., small or multiply charged ions), in which ion–water molecular interactions are stronger than those of water–water. Chaotropicity is related to low charge density (e.g., singly charged ions), in which ion–water molecular interactions are weaker than those of water–water. Marcus [52] proposed the use of ΔG_HB as a unitless value representing the effect of a given solute on the H-bond structure of water, where values ≤ −0.1 represent structure-breaking, chaotropic ions; values ≥ 0.1 represent structure-making, kosmotropic ions; and values between −0.1 and 0.1 are borderline and can be defined either way depending on the system being measured. For the sodium salts measured in this study, the anion order is as follows: ClO₄⁻ < Br⁻ < NO₃⁻ < SO₄²⁻ < H₂PO₄⁻ < HPO₄²⁻ < PO₄³⁻. We therefore decided to examine the H-bond strength of mixtures of ATPSs formed with PEG-8000 and the same sodium salts.

Two examples of the FTIR OH-stretch spectral band decomposition into four Gaussian components (i.e., subpopulations of the water H-band structure) can be seen in Figure 3. The analysis of the OH-stretch spectral band for dilutions of various PEG–salt ATPSs shows that all fractions of the water subpopulations I–IV change with the dilution of the mixture but that these changes appear to follow similar trends in relation to the Hofmeister and Marcus anion arrangements and the ion effect on H-bond strength. Figure 4 shows that the H-bond strength in PEG–sodium salt ATPSs decreases from the greatest to the least extent in the following order: NaClO₄ > NaNO₃ > Na₂SO₄ > Na₂HPO₄ > NaH₂PO₄. Conversely, we see an increase in H-bond strength for the system containing Na₃PO₄. The shift of the OH stretch to higher frequencies (i.e., an increase in water subpopulation IV, 3550 cm⁻¹) suggests that these salts are less efficient in their phase-forming behavior due to a weakening of the H-bond network. These results show empirical proof of the theoretical concepts regarding the salt effect on H-bond arrangement in PEG–salt ATPSs [36,51].

From the above results, we can assume that chaotropic sodium salts diminish phase separation in PEG–salt ATPSs, whereas kosmotropic sodium salts enhance phase separation. The kosmotropic sodium salts appear to increase the strength of the H-bond network, while the converse is observed with chaotropic sodium salts. To further study the effect of H-bond strength on phase separation in PEG–sodium salt systems, we pursued the anion effect on the formation of PEG agglomerates prior to macroscopic phase separation. We have previously shown [7] that the size and concentration of these agglomerates depend on the chemical nature of the phase-forming components in two-polymer systems. Here, we used dynamic light scattering (DLS) to measure the hydrodynamic diameter of the agglomerates present in various dilutions of PEG–sodium salt systems.

By using DLS, we can present the size of the PEG-8000 agglomerates as a function of the relative contribution to each of the four water subpopulations (3080 cm⁻¹, 3230 cm⁻¹, 3400 cm⁻¹, and 3550 cm⁻¹), as shown in Figure 5. Displayed like this, we can clearly see a trend of decreasing H-bond strength as agglomerate size increases for the NaClO₄- and NaNO₃-containing systems across all four populations, supporting the chaotropicity of the anions in these sodium salts. The changes in water fractions I and IV as a function of agglomerate size follow a similar trend that is in line with the Hofmeister series, where we observe an increase in water subpopulation I as follows: PO₄³⁻ > H₂PO₄⁻ > HPO₄²⁻ > SO₄²⁻ > NO₃⁻ > ClO₄⁻. The reverse trend is observed for water subpopulation IV. Interestingly, PEG agglomerate size does not appear to be correlated with the anion valency of the sodium salts used in this study.

4. Discussion

Liquid water is highly complex, in part due to an extensive network of hydrogen bonds (H-bonds) which range from forming strong, rigid “ice-like” structures to those which are weaker and less rigid. This network of H-bonds is highly dynamic and extremely sensitive to the presence of (co)solutes, with different solutes causing different changes in the H-bond network of aqueous solutions. We use a model to estimate the H-bond matrix via empirically measured IR spectra of aqueous solutions. H-bonding between OH groups in aqueous solutions results in a broad absorption spectrum between 4000 and 2500 cm⁻¹. The decomposition of this OH-stretch band into Gaussian peaks has been conducted by multiple groups to define assignments of specific H-bond environments in a given aqueous solution [3,4,5,6,7,8,11,12,19,21,25,26,28,30]. Our model consists of four subpopulations of water, whose positions were determined by the best and most reliable fits using the sum of two, three, four, or five Gaussians with floated central frequencies for pure water and single-component aqueous solutions. Fitting the OH-stretch band using our model consistently showed that the best fit was always obtained with exactly four components. These four components are fixed at positions, each assigned to water molecules existing in different H-bonded environments: water with four tetrahedrally arranged H-bonds (3080 cm⁻¹); water with four distorted H-bonds (3230 cm⁻¹); water with three or four loosely arranged H-bonds (3400 cm⁻¹); and water with one, two, or three H-bond(s) (3550 cm⁻¹) [3,4,6,7,8]. Our assignment of these components is only an approximation of the complex H-bond network existing in water; however, these assignments provide an insight into the structure of water molecules in the hydration shells of dissolved ions.

In the case of aqueous two-phase systems (ATPSs), each phase-forming component produces a different H-bond network domain with dissimilar solvent properties that co-exist in a single solution until the concentration of one or both polymer(s)/salt exceeds a threshold in which the domains become immiscible [3,4,6,7,8,53]. The network of H-bonds in two-polymer and single-polymer–salt ATPSs are altered by the presence and quantity of salt/osmolyte additives which can either favor or disfavor phase separation [6,7,16,54,55,56,57]. We also previously provided strong evidence of the abrupt rearrangement of this H-bond matrix in aqueous mixtures of biphasic systems prior to macroscopic phase separation [6,7].

The data presented here align with the previously observed inverse correlations between anion valency and salt concentration that are required to form an ATPS in which the stronger the hydration of the ion, the lower the concentration required for phase separation [23,26]. It has been suggested that this phenomenon occurs due to an increased salting-out capacity of higher valence anions, leaving less “free” water to hydrate PEG in comparison to lower valence anions [36]. Although we do not use a high-frequency Gaussian peak in our model to specifically represent “free” or “quasi-free” water populations (~3600–3700 cm⁻¹) [20,22,23,24,27], we do observe a shift to higher wavenumber frequencies for lower valence chaotropic salts, suggesting a lower efficiency in phase-forming behavior. Furthermore, our previously published work on single-solute aqueous solutions provided results in agreement with Pavelec et al., who used a decomposition model to fit Raman band profiles of water and aqueous mixtures with six Gaussian components, which did include a high-frequency component [5]. The results from their spectral decomposition of PEG/water mixtures showed that the relative intensity of water subpopulations at peaks positioned at 3497 cm⁻¹ and 3606 cm⁻¹ decreased with increasing PEG concentration, which is in agreement with our findings [3,5].

Here, we provide further empirical evidence supporting the phenomenon observed in various PEG–salt ATPSs in which anions with a higher valence, and increased salting-out capacity, require lower concentrations of salt/polymer for the observance of phase separation. The shift of the OH stretch to higher frequencies suggests that less efficient phase-forming behavior may be due to a weakening of the H-bond network.