Solvent Interaction Analysis: A New Lens for Protein Structure and Diagnostics

1. Introduction

Types of Biphasic and Multiphasic Systems in Water

The formation of multiple aqueous phases when different water-soluble compounds are mixed represents a widespread phenomenon. The number and characteristics of phases that are formed depend on the physicochemical properties of the constituent compounds. The complexity of these systems can be remarkable—Albertsson documented an aqueous system containing six polymers (including Dextran sulfate, Dextran, and four hydroxypropyl Dextrans with varying substitution levels) that separated into 18 distinct phases [1]. A comprehensive catalog by Mace et al. identified over 300 phase-separated systems, ranging from two to six aqueous phases, generated from mixtures of various polymers and surfactants [2]. While the relationship between the number of phase-forming components and the resulting number of phases remains unclear, all these systems share an important characteristic: they are reversible with respect to component concentrations. Diluting such a system below critical thresholds causes the originally separated phases to merge into a single phase, while restoring the component concentrations regenerates the multi-phase system.

These systems have proven valuable for creating stable density gradients in aqueous media, with phase density differences as small as 0.001 g/cm³. Such gradients have enabled the separation of nanoparticles [3] and the diagnosis of medical conditions through erythrocyte fractionation, including sickle cell disease [4] and iron deficiency anemia [4,5]. They have also facilitated the isolation of reticulocyte-enriched fractions from blood samples [6].

While multi-phase systems demonstrate the versatility of aqueous phase separation, most research has focused on simpler aqueous two-phase systems (ATPSs) formed by just two phase-forming compounds. In these binary systems, two immiscible aqueous phases emerge once the concentrations of both components exceed specific threshold values. Under controlled environmental conditions (constant temperature and composition), ATPS formation occurs spontaneously and reversibly, producing thermodynamically stable, equilibrium systems. A clear interface separates the two phases, each enriched in one of the phase-forming components while both retain substantial water content (typically exceeding 80% on a molal basis). The kinetics of phase separation vary considerably with system composition: polymer-salt systems generally separate rapidly (within minutes), whereas polymer-polymer systems (such as PEG-Ficoll) typically require longer equilibration times (hours).

ATPSs can also be formed in solutions of single compounds under appropriate conditions. Examples include the termo-responsive polymers [7,8], stimulus-responsive polypeptides [9,10], and surfactants [11,12,13]. In these cases, once a critical temperature threshold is reached, the system separates into two phases: one enriched in the compound and another essentially free of it.

The diversity of ATPS compositions is substantial. Common configurations include two different polymers [1,14,15,16,17], a polymer with a salt [1,14,15,16,17] or surfactant [18,19], two different surfactants [20], and ionic liquids [21]. Additionally, ATPSs can be formed from proteins and polysaccharides [22,23,24,25], a category that has attracted increased attention due to discoveries that living cells naturally generate multiple membrane-less organelles (MLOs, also known as biomolecular condensates, BMCs) through highly regulated biological liquid-liquid phase separation processes [26,27,28,29,30,31,32,33,34,35,36,37,38,39,40]. Importantly, cells represent highly sophisticated multi-phase systems, as they typically contain very large number of different MLOs/BMCs [41,42,43] .

2. Phase Diagrams as a Means to Describe ATPS

As depicted in Figure 1, phase diagram displays a curved boundary (binodal curve) that demarcates two distinct regions. Below this curve lies the single-phase region, where mixtures appear visually homogeneous. Above the binodal curve exists the two-phase region, where phase separation occurs. When a system’s composition falls within the two-phase region, it spontaneously separates into two coexisting phases, whose compositions are represented by points on the binodal curve. The tie line connects the two phases compositions and passes through the system overall composition point. As polymer (or polymer-salt) concentrations decrease, tie lines become progressively shorter, eventually converging at the critical point (labeled C in Figure 1), where the two phases become identical in composition [1,14,15,44,45].

A key feature of ATPS is their compositional flexibility. Multiple systems can share identical phase compositions while differing in volume ratios. For instance, points A_o, A₁ and A₂ in Figure 1 represent different overall compositions that all yield the same top phase (T) and bottom phase (B) compositions, but with varying phase volume ratios. Conversely, compositions located off a given tie line (such as point D) will produce phases with entirely different compositions.

ATPS preparation typically involves combining stock solutions of polymers and salts by mass. Since each component dilutes the others, careful calculation of stock solution concentrations is necessary to achieve the desired final composition. Stock solutions should be prepared gravimetrically. For hygroscopic polymers, like dextran and Ficoll, it is essential to prepare aqueous solutions first and then determine actual polymer content through freeze-drying or another suitable analytical method.

The binodal curve can be established through turbidimetric titration [46], while tie lines require analytical determination of polymer and/or salt concentrations in both phases. Unlike conventional organic solvent-water systems, a given polymer pair (or polymer-salt combination) can form numerous distinct two-phase systems with varying properties simply by adjusting the overall composition.

Polymer molecular weight significantly affects phase diagrams in both polymer-polymer and polymer-salt systems [14,47,48,49]. Generally, the binodal curve shifts downward as molecular weight increases, indicating that lower polymer concentrations suffice for phase separation at higher molecular weights. This trend supports the hypothesis that volume-exclusion effects drive phase separation [50].

However, polymer chemical nature proves more influential than size alone. Comparing phase diagrams for dextran-70/PVP and dextran-70/Ucon systems in identical media (0.15 M NaCl in 0.01 M sodium phosphate buffer, pH 7.4) reveals that dextran-PVP requires higher polymer concentrations for phase separation despite PVP's smaller size (~12 kDa) compared to Ucon (~4 kDa). Similarly, in polymer-salt ATPS [51,52,53], polypropylene glycol (PPG-725, 725 Da) forms two phases with sodium citrate at much lower concentrations than the larger PEG-4000 or Ucon-4000 (both ~4 kDa).

The binodal curves for PEG-4000 and Ucon-4000 (polymers of equivalent molecular weight) in sodium citrate mixtures are distinctly positioned relative to each other on the phase diagram. These observations demonstrate that chemical characteristics outweigh molecular size in determining phase separation behavior. These observations also hint to the important idea that phase separation might occur via the effect of solutes on the solvent properties of water (see below). The specific chemical or physicochemical features of polymers that govern phase separation in aqueous mixtures remain to be fully elucidated.

Salt identity dramatically influences polymer-salt ATPS phase diagrams as well. For example, binodal curves for potassium versus sodium carbonate in PPG-400 solutions [53,54,55] differ markedly, as do those for sodium chloride versus sodium nitrate with PPG-425 [53,56].

Salt additives also affect phase diagrams in polymer-polymer systems, with effects studied more extensively in these systems (pages 103-122 in [14]) than in PEG-salt systems, though some examples exist [57,58,59]. The magnitude and direction of salt additive effects vary depending on the specific polymer pair, salt type, and concentration. The influence of salts on dextran-PVP and dextran-polyvinyl alcohol (PVA) phase diagrams is considerably more pronounced than in dextran-PEG or dextran-Ficoll systems (pages 103-116 in [14]). Salt additions can alter phase volume ratios, signaling compositional changes even when binodal curves appear unchanged.

Interestingly, dextran and polyvinyl alcohol can form ATPS in water even when completely miscible in the dry state (pages 141-152 in [14]). This observation supports the conclusion [60] that phase separation in aqueous polymer mixtures originates from differential effects on water structure, resulting in the formation of immiscible hydrogen bond networks that share the same aqueous character.

Phase diagram analysis serves two important purposes. First, it provides fundamental insights into phase separation mechanisms across different ATPS types. Second, it guides rational optimization of conditions for protein and solute distribution. However, the accumulated phase diagram data has not yet yielded sufficient mechanistic understanding of phase separation processes.

Among theoretical frameworks developed to explain phase separation in aqueous polymer mixtures, the binodal model introduced by Guan and colleagues [61,62] remains the most effective for accurately representing phase diagrams. This semi-empirical approach conceptualizes each point along the binodal curve as representing a saturated solution of one phase-forming component within a solution of the other component. While the original model relied on excluded volume principles, Ferreira et al. [63] proposed that the dissolution behavior of one compound in solutions of another may be governed by how that second compound alters water's solvent characteristics.

Applying this modified binodal framework yielded excellent results when describing phase diagrams for various polymer-polymer aqueous two-phase systems, including combinations of dextran, Ficoll, poly(ethylene glycol)-8000, and Ucon. The model also successfully characterized both a newly investigated system containing trimethylamine N-oxide (TMAO) and polypropylene glycol-400, as well as a previously documented TMAO and poly(ethylene glycol)-600 system. Additionally, it was demonstrated that this modified approach applies effectively to single polymer-salt and polymer-ionic liquid aqueous two-phase systems as well.

The key finding from our investigation is that phase separation in mixtures of various compounds (including polymers and salts) results from how these substances modify water's solvent properties, thereby restructuring the water environment in their aqueous solutions. Regarding solute distribution manipulation, this relates directly to the properties of the coexisting phases themselves. These properties must be characterized for all ATPS, whether or not detailed phase diagram information is available. Importantly, comprehensive phase diagram knowledge is not essential for understanding and controlling solute partitioning in ATPS.

3. Physicochemical Fundamentals of Aqueous Two-Phase Partitioning

Research has demonstrated that solute partitioning in ATPSs—ranging from small organic molecules to proteins and nucleic acids—generally occurs independently of direct interactions between solutes and phase-forming polymers. This conclusion has been proposed and subsequently validated through multiple studies [64,65,66,67,68,69,70].

The driving force behind unequal solute distribution in ATPS is attributed to differential solute-solvent interactions between the coexisting phases. The primary interaction types contributing to this phenomenon include: dipole-dipole and dipole-induced dipole interactions; hydrogen bonding interactions; and electrostatic interactions (ion-ion, ion-dipole, ion-induced dipole). When these interaction contributions are characterized for a given solute in a particular ATPS, the partition coefficient in a newly designed ATPS with known solvent interaction properties can be predicted with 90-95% accuracy [64,65].

The nature and magnitude of solute-solvent interactions for any compound depend on solvent properties, though the relative contributions of different properties vary based on the solute's structural and physicochemical characteristics. Polarity represents the most commonly used classification term for solvents and refers to a solvent's ability to interact with and dissolve polar or ionic substances. Contemporary understanding defines polarity as encompassing all possible specific and non-specific interactions between solvent and potential solute, excluding interactions leading to chemical transformation of the solute. These interactions include electrostatic interactions, dipole-dipole and dipole-induced dipole interactions, hydrogen bonding, and electron pair donor-acceptor interactions.

Importantly, polarity describes the solvent's potential behavior in relationship with the solute rather than being an absolute property of the pure solvent. Numerous polarity scales exist based on different probes and spectroscopic techniques (NMR, IR, UV/Visible absorption, emission spectroscopy). As noted by Ab Rani et al. [71], no single universal measure of polarity exists; all scales provide estimates, and different scales yield different values for the same solvent. Rather than being "right" or "wrong," the utility of an empirical polarity scale lies in its ability to explain and predict other solvent-dependent phenomena [71].

Solvent polarity of aqueous media in solutions of several ATPS-forming polymers (dextran, Ficoll, and PEG of various molecular weights) has been measured using water-soluble carboxylate-substituted anionic betaine Reichardt's solvatochromic dye, thymol blue, sulfonephthalein dyes, fluorescein, and eosin [72,73]. Studies have demonstrated that the solvent polarity of coexisting phases in dextran-Ficoll and dextran-PEG ATPS differs measurably [64,65]. These differences in solvent polarity, combined with variations in hydrophobic and electrostatic properties between phases [14], provide the basis for considering polymer-polymer ATPS as analogous to organic solvent-water biphasic systems, with the critical distinction that both phases maintain aqueous character.

Since single-parameter polarity scales cannot adequately represent the multitude of possible solute-solvent interactions, Kamlet and Taft developed multi-parameter polarity scales based on the Linear Solvation Energy Relationship (LSER). This approach incorporates three scales: hydrogen bond d onor (HBD) acidity (α; which is a measure of a molecule's ability to provide a proton (H⁺) to a hydrogen-bond acceptor), hydrogen bond acceptor (HBA) basicity (β; which refers to the thermodynamic tendency of a molecule or atom to act as a hydrogen-bond acceptor), and dipolarity/polarizability (π*; which describe nonspecific intermolecular interactions that define a solvent's ability to stabilize a solute through electrostatic forces. Here, solvent dipolarity refers to the permanent dipole moment of solvent molecules that describes the solvent's ability to orient its molecules around a solute to stabilize charge or dipoles through electrostatic (Coulombic) interactions. It is considered an anisotropic effect because it depends on the direction and orientation of the molecules. On the other hand, solvent polarizability measures how easily the electron cloud of a solvent molecule can be distorted by an external electric field (such as that from a solute). This distortion induces a temporary dipole, facilitating dispersion and induction forces. Being an isotropic effect, it acts uniformly in all directions regardless of molecular orientation) [74,75,76]. The combination of these three scales provides a superior description of a solvent's ability to participate in solute-solvent interactions compared to any single-parameter scale. The LSER model [77] can be expressed as:

XYZ = (XYZ)₀ + sπ* + aα + bβ (1)

where XYZ is the solute property in a given solvent, (XYZ)₀ = same property in a reference state, s, a, b = solute-dependent coefficients, and π*, α, β = respective solvent parameters.

The π* scale was generated by averaging values from seven solvatochromic dyes with strong, symmetric solvatochromic absorption spectra. Subsequently, 45 different dyes have been employed to generate π*-values for over 200 solvents [78]. The use of multiple dyes aims to avoid dye-specific values.

For aqueous media in polymer solutions and ATPS phases, the selection of water-soluble solvatochromic dyes is limited. For studying π*-values, 4-nitroanisole was selected [64] based on commercial availability, chemical stability, and water solubility.

As emphasized for ionic liquid studies [71], the precise π*-value lacks fundamental physical meaning; rather, it serves as a guide to solvent effects on solute species sensitive to solvent dipole interactions. The magnitude of these effects must be considered in context: the π* difference between 40% Ucon solution and polymer-free media is 0.022, while the difference between methanol and ethanol is 0.06 (three times larger). In dextran-75-Ficoll-70 ATPS, the π* difference between coexisting phases is only 0.003, yet this small difference significantly affects the distribution of small compounds and proteins.

4-Nitrophenol is commonly used as the probe for β measurements, with β-values calculated from wavelength of maximum absorbance and π*-value. The polymer effects on β are relatively modest—the difference between β values for aqueous media in 40% dextran-75 solution versus polymer-free media is 0.033, compared to 0.15 for the methanol-ethanol difference. The β difference between coexisting phases in dextran-75-PEG-600 ATPS is merely 0.005.

Water-soluble solvatochromic Reichardt's carboxylated betaine dye sodium {2,6-diphenyl-4-[4-(4-carboxylato-phenyl)-2,6-diphenylpyridinium-1-yl])phenolate} is used as a single probe for α measurements. The α-value is calculated from wavelength of maximum absorbance and π*-value. Data indicate that solvent HBD acidity (α) decreases with increasing polymer concentration for all examined polymers. The polymer effects on α are quite substantial—the α-value difference between 40% dextran-40 solution and polymer-free media is 0.223, considerably larger than the 0.10 difference between methanol and ethanol. In polymer-polymer ATPS, α-value differences between coexisting phases range from 0.0 (Ficoll-70-PEG-6000) to 0.181 (dextran-75-Ucon), depending on polymer and salt composition. These findings align with previously reported data on PEG's effect on water's HBD acidity.

Electrostatic interactions between a partitioned solute and the aqueous media in the phases constitute important components of the forces governing solute partitioning. Analysis of partitioning behavior of a homologous series of dinitrophenylated (DNP) amino acid sodium salts with aliphatic alkyl side-chains of increasing length (glycine, alanine, norvaline, norleucine, α-amino-n-octanoic acid) enables characterization of phase electrostatic property differences. Partition coefficient logarithms for these compounds can be described as:

log K^DNP-AA_i = Cⁱ + Eⁱ N_c (2)

where K^DNP-AA_i = partition coefficient of DNP-amino acid sodium salt in i-th ATPS; N_c is equivalent number of CH₂ groups in side-chain; E and C = constants characterizing the i-th ATPS. Parameter E represents average logK per CH₂ group, readily convertible to free energy of transfer of this group between phases. This parameter characterizes the difference in hydrophobic properties between phases. In octanol-water systems, E ≈ 0.50, while in polymer-polymer ATPS it varies from ~0.005 to ~0.12. Parameter C represents the contribution of the polar moiety (DNP-NH-CH-COONa moiety in this case). The advantages and limitations of using this moiety as a probe for electrostatic properties have been extensively discussed. Solvent properties in typical polymer-polymer and polymer-salt ATPS vary with different additives, and these properties affect partition behavior of proteins such as lysozyme and α-chymotrypsinogen A.

Solvent dipolarity/polarizability (Δπ*) differences vary within similar ranges in both dextran-PEG and PEG-Na₂SO₄ ATPSs, as do electrostatic property differences (C). Hydrogen bond basicity (Δβ) differences are typically larger in PEG-Na₂SO₄ ATPS compared to dextran-PEG ATPS, a pattern also observed for hydrogen bond acidity (Δα). The only universal rule applicable to proteins and other soluble compounds is that for K > 1, partition coefficient increases with increasing phase polymer (or polymer and salt) concentrations, and for K < 1 partition coefficient decreases under similar conditions. Partition behavior of any solute (including proteins) in ATPS is independent of molecular weight and size. In a given ATPS, only the nature and spatial arrangement of chemical moieties exposed to the solvent determine partition behavior.

4. Characterization of Individual Proteins with Solvent Interaction Analysis

Experimental evidence shows [79] that two peptides sharing identical amino acid compositions but differing in sequence can exhibit partition coefficients that vary more than three-fold. The magnitude of these differences, as well as the actual partition coefficient values for small organic molecules, can fluctuate significantly depending on the specific polymer and ionic composition of the employed ATPS.

As mentioned earlier, the partition coefficient characterizes a solute's partition behavior and is calculated as the solute concentration ratio between the top and bottom phases. It has been demonstrated that a compound's partition behavior in an ATPS is extremely responsive to structural modifications of a solute [14,49,64,66,67,80,81,82,83,84]. This is further illustrated by Table 1 representing partition coefficients for different mutants of staphylococcal nuclease A and bacteriophage T4 lysozyme.

In fact, Table 1 presents an example of the impressive sensitivity of this approach. Protein partitioning has additionally been demonstrated to vary based on the nature and spatial organization of chemical groups accessible to the solvent [88,89].

For determining an individual protein's partition coefficient, the recommended procedure [90]. involves preparing multiple identical polymer and salt additive mixtures at concentrations above those desired in the selected ATPS. A "free volume" of water or sample solution must then be incorporated to reach the target polymer and salt composition. This free volume typically ranges from 50–250 μl for total ATPS volumes of approximately 400–1500 μl, depending on the specific ATPS composition. To achieve varied solute concentrations within the ATPS, different volumes of solute solution combined with corresponding water volumes (totaling the free volume) can be added.

Following solute addition, the systems undergo mixing and centrifugation to accelerate phase separation. Aliquots are then collected from the separated phases, diluted as necessary, and analyzed using appropriate concentration assays. This workflow can be performed manually in 2 mL vials (see video demonstration at www.youtube.com/watch?v=0h9bTSO7Zv4), using microtubes or deep-well plates with liquid-handling workstations [91,92,93,94], or in microchip formats [95,96,97,98,99,100,101,102,103,104,105,106]. Figure 2 provides a graphical representation of the complete process.

Various analytical techniques can measure protein concentrations in the diluted phase aliquots [92]. Available methods include: HPLC (sensitive yet relatively time-intensive); mass spectrometry/MS (rapid and sensitive, though requiring internal standards for concentration determination); o-phthaldialdehyde/OPA assay (fast, dependable, and highly sensitive for proteins containing adequate lysine residues); Bradford assay (straightforward and quick, albeit less sensitive than OPA); UV absorbance measurements (generally requiring larger protein quantities due to limited sensitivity); and ELISA or similar affinity-based assays (highly sensitive and suitable for analyzing specific proteins in biological fluids including serum, plasma, urine, cerebrospinal fluid/CSF, tissues, or cell extracts). Calibrating the analytical method in diluted phases from both top and bottom is advisable, though not invariably essential.

Once phase aliquots are withdrawn and diluted, they can be refrigerated or frozen prior to protein concentration measurement. The protein partition coefficient is calculated from the protein concentrations (or analytical signals proportional to concentration). During partition coefficient determination, protein concentrations (or analytical signals) measured in top phases are plotted against corresponding bottom phase measurements. Figure 3 shows a representative plot. The linear curve's slope represents the protein partition coefficient K, while the intercept relates to phase component interference with the analytical signal and may approach zero. The plot's linearity indicates the protein partition coefficient K remains independent of protein concentration, implying identical species distribute between the coexisting phases across the experimental protein concentration range. Deviation from linearity between upper and lower phase solute concentrations suggests that elevated protein concentrations can generate species exhibiting partition behavior distinct from species present at lower concentrations, typically indicating protein aggregate formation. Nonlinear curve in Figure 3 illustrates typical protein aggregation behavior. Partition coefficient K-value determination accuracy typically achieves values superior to 3%, though this depends on the analytical assay employed and may be considerably lower when immunoassays are utilized. The target protein concentration range to be examined depends on the particular application and analytical assay sensitivity. For diagnostic applications, economic and throughput considerations typically necessitate using single sample volumes.

The complete protein partition analysis process in ATPS encompasses two procedures: first, partitioning an individual protein or biological fluid in a specific ATPS, and second, measuring the given protein's concentration in each phase. Once phases are separated and diluted, they can be stored at low temperature and prepared for subsequent analytical assay.

For proteins, the partition coefficient varies with ATPS polymer composition. If the goal is protein isolation through aqueous two-phase extraction, achieving extreme protein distribution is desirable. For protein analysis/characterization purposes, the protein partition coefficient (K) should fall within an analytically reliable range, typically defined as 0.1 to 10. More extreme protein distributions prove difficult to measure accurately, as concentrations in the two phases differ by over an order of magnitude. Generally, ATPSs formed by dextran-70 or dextran-40 (molecular weights of 70 kDa or 40 kDa, respectively) combined with Ficoll-70 (molecular weight of 70 kDa), or Ficoll-70 combined with PEG (molecular weight ranging from 600 to 8000 Da) prove suitable for protein analysis/characterization, though PEG–salt ATPSs may also be employed.

Once an ATPS with suitable polymer composition that produces protein partitioning with K-values in the aforementioned range (0.1 < K < 10) is identified, the ATPS composition requires fine-tuning to deliver appropriate conditions for the selected objective. For instance, the objective might be detecting alterations in protein sequence (mutations, splicing), post-translational modifications (PTMs) including glycosylation or phosphorylation, conformational changes, partner binding, or chemical modifications, such as oxidation or deamidation. In these scenarios, the goal involves designing conditions, where the K-value differences between intact and modified proteins are sufficiently large to be analytically reliable. The optimal approach for achieving this objective is manipulation of the ionic compositions of the ATPS.

Table 2 presents examples of different K-values for human serum albumin when exposed to various concentrations of different drugs [107].

It merits noting that the data from [107] shown in Table 2 were generated with the objective of exploring conformational changes of the protein induced by binding of drugs.

The data listed in Table 2 showed that the ratio of partition coefficient of albumin at various drug concentrations on a logarithmic scale (lnK_ij) to the partition coefficient of HSA in the absence of the drug (lnK₀) may be described as:

Ln(K_ij/K_o) = A + B exp(C[drug]_j), (3)

where A, B, and C are constants. These constants reflect the drug’s effect on the conformational change of HSA.

The clearance times for the drugs investigated in [107] are known [108], and the analysis showed that there is a linear relationship described as:

exp(-CT) = -0.42 _{± 0.06} + 0.20 _{± 0.03}1/C + 0.33 _{± 0.04}exp(-B) (4)

N = 10; r² = 0.920; SD = 0.097; F = 40

where CT is drug clearance time.

These results showed that HSA's partitioning behavior in the ATPS is influenced by its interactions with various drugs in a manner that depends on drug concentration. While all examined drugs increased the HSA partition coefficient, each affected protein partitioning differently, with the concentration-dependent relationship for each drug following an exponential pattern. These findings suggest that all tested drugs alter the albumin's conformation. The relationship between the exponential equation parameters that describe HSA partitioning in the presence of a specific drug and that drug's clearance rate from the human body implies that drug-induced conformational changes in HSA, which are reflected in how the HSA-drug complex partitions in ATPS, are connected to the drug's blood clearance efficiency. This findings also demonstrate the practical value of using protein partitioning in ATPS as a useful analytical tool for diverse applications.

Finally, the reason for describing this technique as the solvent interaction analysis (SIA) will be addressed. It has been demonstrated by Madeira et al. [64,65,66,81,83,109], Ferreira et al. [68,110,111], and da Silva et al. [69,112] that Eq. (4) with an additional parameter representing electrostatic interactions in the phases of ATPSs describes partition coefficients of small organic compounds and proteins in ATPSs formed by different pairs of various polymers, but of the same ionic composition, or in polymer–salt ATPSs with fixed ionic composition. It has been established [64,65,66,68,69,81,83,109,110,111,112] that the contributions of different solute–solvent interactions (dipole–dipole, hydrogen bonding, and electrostatic interactions) are specific for a solute and are completely independent of the nature, molecular weights, and concentrations of phase-forming polymers or nonionic additives capable of affecting the solvent properties of the phases. This information validates the previously suggested [14] idea that solutes do not directly interact with phase-forming polymers, but interact only with the aqueous media in the two phases. The solvent properties of the aqueous media may be manipulated by different additives, such as osmolytes [68,69,110,111,112], and these manipulations affect the partition coefficients of proteins. Therefore, the partition coefficient of a solute (such as a protein) may serve as a measure of the difference between protein interactions with the aqueous media in the two coexisting aqueous phases, thereby defining this technique the solvent interaction analysis (SIA) technology.

Ferreira et al. [113] demonstrated that TMAO, a nonionic additive, can be used to control how solutes distribute themselves in aqueous two-phase systems. The researchers investigated the partitioning behavior of various proteins and small organic molecules in two different aqueous two-phase systems: one containing Dextran-75, Polyethylene glycol (PEG-600), 0.15 M sodium sulfate, and 0.01 M sodium/potassium phosphate buffer at pH 7.4, with TMAO concentrations ranging from 0 to 1.95 M; and another consisting of Ficoll-70, Polyethylene glycol (PEG-8000), and 0.01 M sodium/potassium phosphate buffer at pH 7.4, with TMAO concentrations between 0 and 1.5 M. The study revealed that protein partitioning responds to increasing TMAO concentrations in a protein-dependent manner. To characterize how the solvent properties of water differ between phases—including hydrophobicity, electrostatic characteristics, solvent dipolarity/polarizability, hydrogen bond donor acidity, and hydrogen bond acceptor basicity—the researchers employed two approaches: examining the partitioning patterns of homologous sodium salts of dinitrophenylated amino acids containing aliphatic alkyl side-chains, and applying the solvatochromic comparison technique. The findings indicate that TMAO influences solute partitioning by modifying the solvent characteristics of each phase, supporting the hypothesis that the distribution of proteins and small organic molecules is determined by solute-water interactions within the coexisting phases.

Figure 4 summarizes some of the major advantages of the ATPS-based SIA in application to proteins and shows that this technique represents a unique tool for characterizing various features ranging from 3D conformation, to structural dynamics, interactions, and posttranslational modifications. By measuring peculiarities of partition of proteins between two immiscible aqueous phases, it also represents a powerful label-free analytical platform to separate, identify, and characterize the diverse molecular forms (proteoforms) generated by genetic variations or modifications (like phosphorylation or glycosylation), which are notoriously difficult to separate using conventional methods.

5. Structure-Based Proteomic Approaches for Biomarker Discovery

The field of proteomics as the large-scale study of proteins, particularly, their structure and functions, may be compared to space exploration, because in the addition to the expected advancement of knowledge within the biological universe, there are developments and advancements of new technologies. The practical outcome of proteomic research is discovery of new drug targets and biomarkers that could be used in clinical and medical practice. In regards to biomarkers, it should be noted that the number of discovered biomarkers [114,115,116,117,118,119,120,121,122] at this time greatly exceeds those already validated and implemented in clinical laboratories [123,124,125,126,127,128,129,130,131,132,133,134]. It has been reported [135] that less than 1.5 new tests for protein biomarkers per year were implemented in clinical practice over the last 16 years. The term “biomarker” refers to “any substance, structure, or process that can be measured in the body or its products and influence or predict the incidence of outcome or disease” [136]. A protein biomarker is present in the body or excreted by the body and/or changed as a biological response to a particular disease.

Different types of natural compounds may be considered for potential disease biomarkers. Advantages of proteins in this regard are their immense diversity, dynamic turnover, and secretion in biological fluids. The number of different proteoforms (taking into account posttranslational modifications) exceeds number of genes, metabolites, and mRNA transcripts by more than an order of magnitude, which increases the chances of discovering a marker or a panel of markers for each disease state [126]. Because the ultimate goal of biomarker discovery is usually development of a blood test, blood is a logical fluid to use for protein biomarker discovery. Human plasma is viewed as the most comprehensive human proteome, representing all body tissues and both physiological and pathological processes. This comprehensiveness, together with the accessibility of plasma and the vast medical laboratory infrastructure already in place for its analysis, ensures that it will remain the preferred diagnostic material in the foreseeable future.

Proteins in a biological fluid have multiple features that may change under pathological conditions. Features for which changes are currently measurable with existing technologies include [137]: (i) concentration; (ii) structural features such as presence and ratio of different proteoforms, particular post-translational modifications (PTMs); presence of single or multiple point mutations, truncations, etc.; (iii) functional features, such as protein–protein, protein–nucleic acid, and protein–small molecule interactions.

Currently, the search for protein biomarkers is following a long established trend, in which essentially all existing diagnostic clinical tests measure the biomarker concentration in serum, plasma, urine, or CSF. There are two main reasons for this strategy. First, it is assumed that human pathologies or any physiological state are reflected as changes in the presence/absence or the levels of proteins in a given biological fluid [138,139]. Second, different technologies of affinity based measurements of the biomarker concentrations are available and being continuously improved and advanced [140,141,142,143,144,145,146,147,148,149,150]. The current clinical chemistry laboratory tests are based almost exclusively on affinity-based assays for measuring concentrations of proteins or other biomarkers in the patients' biological fluids.

The choice of the biomarker concentration as a diagnostic indicator is understandable based on its historic use; however, it may not be the optimal choice [138] of species (proteoforms) grouped under a single denominator [151,152,153]. According to Bults et al. [151] measuring a “protein” is meaningful only if the method it is measured with and the molecular property this method responds to are defined. Different proteoforms of plasma proteins may be quantified by mass spectrometric immunoassay (MSIA) [154,155,156,157,158,159,160,161,162,163,164,165,166,167,168]. Variability of a given protein biomarker level in plasma/serum among different individuals is, diagnostically, an even more important issue. An example of a systematic study of variability in proteins levels is presented by Liu et al. [169]. Levels of 342 different proteins in 232 plasma samples were collected longitudinally from pairs of healthy monozygotic and dizygotic twins at intervals of 2–7 years and examined [169] with SWATH-MS, a high-throughput targeting mass spectrometry method [170,171,172]. Substantial variability in plasma proteins levels were observed between individuals for 174 out of 342 proteins examined [169], especially, but not exclusively, for low abundance proteins generally considered to be promising biomarkers of diseases.

Vernardis et al. studied [173] changes in the plasma proteome using mass-spectrometry of blood plasma from several cohorts including from 10 to 200 participants before and after a glucose tolerance test at 4 time points (0, 30, 60, 120 minutes); and before and after caloric restriction. Proteins responded across individuals or in an individual-specific manner. Among the proteins, which changed in response to both caloric restriction (CR) and refeeding (RF), there were multiple proteins (with 60 biomarkers detected in total) that are known as disease biomarkers from previous studies [174]. For instance, acute CR decreased cystatin C (CST3) levels. CST3 is known as a biomarker of kidney function, since levels correlate with glomerular filtration rate [175]. Its decrease in plasma during refeeding may suggest, among other factors [176], an increase in kidney function. Other examples include α1-acid glycoprotein (AGP1, also known as orosomucoid-1, ORM1) and α1-acid glycoprotein ORM2, acute-phase proteins that might regulate energy homeostasis [177]. These proteins responded to CR and were restored under RF. Moreover, CR affected the levels of insulin-like growth factor 2 (IGF2), a growth factor important for fetal development as well as adipocyte proliferation [178]. IGF2 decreased under CR and increased under RF. Insulin-like growth factor binding protein 3 (IGFBP3) has high homology to IGF2 and one of its roles is to facilitate IGF’s transfer to receptors [179]. Coagulation factors like F9 (factor IX) and F13B (factor XIIIB) were also among the proteins significantly affected by caloric restriction.

It was found that short-term nutritional interventions result in a substantial change in the proteome. For instance, the CR experiment affected more than a tenth of the high-abundant functional plasma proteins in all study participants. The analysis also revealed a substantial individual-specific proteomic response that was not detected to the same degree at the metabolome level, but also that both complementary nutritional challenges affected a similar set of core proteins [173].

Chen et al. examined [180] effect of short-term (24 h) sleep deprivation on proteome of peripheral blood in 14 males and 18 females. The results of plasma proteomics analysis showed that, compared with Day1, expression levels of 316 proteins were changed on day 2 with most of the proteins (n = 292) expression reduced, and only for 24 proteins increased. Compared with Day1, expression levels of 372 proteins were changed on day 3, of which for78 proteins they were increased and for 294 proteins decreased. In addition, compared with day 2, the expression of 179 proteins was changed (100 increased and 79 decreased) on day 3.

Bjorkum et al examined [181] changed proteins in human blood serum after loss of 6 hour sleep at night in eight females recruited by volunteer request. Peripheral venous whole blood was sampled at 04:00 am, after 6 hours of sleep and after 6 hours of sleep deprivation. Blood serum from each subject was depleted before protein digestion by trypsin and iTRAQ (Isobaric Tags for Relative and Absolute Quantitation) labeling. Labeled peptides were analyzed by mass spectrometry connected to a LC system. Out of 725 proteins identified in human blood serum. 34 proteins were significantly differentially expressed after 6 h of sleep deprivation at night. Out of 34 proteins, 14 proteins were up-regulated, and 20 proteins were down-regulated. Histone H4 (H4) and protein S100-A6/Calcyclin (S10A6) were upregulated more than 1.5-fold.

The study reported in [181] suggested that acute sleep deprivation, at least in females, affects expression levels of proteins that also are changed under pathological conditions like impaired coagulation, oxidative stress, immune suppression, neurodegenerative related disorder, and cancer.

Robbins et al [182] studied changes in plasma proteomic in 654 healthy human participants caused by 20-week endurance exercise training. It was found that the expression levels of 453 proteins changed after exercise training. Of these, the levels of 306 proteins (68%) increased approximately 40% of which constituted secreted proteins.

Corlin et al [183] investigated effects of modifiable lifestyle risk factors for cardiovascular disease on expression levels of 1305 circulating plasma proteins in 897 participants with mean age 46±8 years and in 1121 participants with mean age 52 years. Participants were free of hypertension, diabetes mellitus, and clinical cardiovascular disease. It was found that expression levels of 53 proteins were significantly associated with current smoking, those of 49 proteins were correlated with number of packs of cigarettes smoked. Expression levels of 30 proteins were found to be associated with alcohol consumption, and 5 proteins correlated with at least 1 of the physical activity risk factors.

The common practice in clinical chemistry is to apply a single threshold rule. According to this rule, the presence of disease is detected when the biomarker concentration exceeds or falls off a common population wide threshold. Analysis [184] of longitudinal changes in the plasma level of cancer antigen 125 (CA-125) in women with ovarian cancer implied that using the population wide threshold may be not the optimal approach, likely due to CA-125 level variability [185]. CA-125 is used in clinical practice to monitor the effectiveness of treatment and to detect the recurrence of disease after treatment. It is not recommended for screening, as it is known to miss about 50% of ovarian cancer at early stages. According to Drescher et al. [184], CA-125 may be used as an ovarian cancer screening biomarker under longitudinal analysis. Annual measurements of CA-125 levels were conducted [184] in a serial fashion so that each patient had their own individual baseline. The above data imply that the concentration of a plasma protein biomarker may not be the optimal diagnostic characteristic.

Another important feature for protein biomarkers is the protein structure. It is widely recognized that essentially all proteins exist in plasma as a family of different proteoforms. The ratio of structurally different proteoforms as a diagnostic biomarker has been used in clinical practice for a long time. For example, transferrin occurs in serum as several proteoforms, with different sialic acid residue contents. Heavy alcohol consumption leads to up to 15-fold increases in the fraction of the proteoforms with reduced sialic acid residue contents. This fraction is called carbohydrate-deficient transferrin (CDT) [186]. The ratio of CDT to total transferrin concentrations in serum is an accepted clinical biomarker of chronic alcohol consumption and congenital disorder of glycosylation (CDG) [187]. The ratio of glycated albumin level to that of glycated hemoglobin (HbA1c) in serum is highly correlated with insulin secretory function [188] and is currently considered a better marker in type 1 diabetes for glycemic variability than HbA1c level alone [189]. The ratio of levels of structurally different isoforms of platelet amyloid precursor protein is reported [190] as a promising biomarker for presymptomatic stage Alzheimer disease. Creutzfeldt–Jakob disease may be differentiated from other dementias with the ratio of phosphorylated tau protein isoforms' levels to that of total tau protein in CSF [191]. The ratio of isoforms for a given protein is independent of the protein level in biological fluid, benefitting the performance of the ratio as a diagnostic biomarker.

Different proteoforms of a given protein in biological fluid may be quantified by the SIA method [154,155,156,157,158,159,160,161,162,163,164,165,166,167,168]. It is necessary, however, to establish what particular quality of proteoform or other features is related to a given disease state. Protein activity and function are mainly defined by structure, which can be regulated by PTMs, allowing rapid response to external/internal stimuli within (milli-) seconds [192]. The most stable PTMs observed for over 50% of all human proteins are different patterns of glycosylation [192]. Mass spectrometric analysis of carbohydrates in glycoproteins [193,194,195,196,197,198,199] demonstrated an association between particular glycosylation profiles and diseases for numerous proteins [199,200,201,202,203]. For example, glycomic profiling of serum samples from 52 women with low malignant potential tumors, 147 patients with epithelial ovarian cancer, and 100 healthy women showed [204] that the glycan compositions were significantly different and diagnostically promising. Differences between glycosylation profiles of following proteins in patients with ovarian cancer and healthy women were reported: CA-125 [205], haptoglobin [206,207], α1-acidglycoprotein and α1-anticymotrypsin [206], clusterin, complement factor H, and hemoplexin [208], leucine-rich alpha-2-glycoprotein [209], and other glycoproteins [210,211]. Furthermore, significant differences between the glycosylation profiles of proteins in plasma from patients with different types of breast cancer and controls were reported [212,213,214,215,216,217,218,219]. Glycosylation of several plasma proteins, such as α1-acidglycoprotein, α1-antitrypsin, fetuin, haptoglobin, transferrin [220,221], ceruloplasmin [222], thrombospondin-1 [223], alpha-2-macroglobulin [224], and certain other proteins [225,226,227,228] has changed in patients with pancreatic cancer relative to those in patients with chronic pancreatitis and healthy controls. Glycosylation of different plasma proteins changes in colorectal [229,230,231,232], gastric [233], liver [234,235,236,237,238], lung [239,240,241], brain cancer [212], and prostate cancers [242,243,244,245], melanoma [246], schizophrenia [247], and other diseases. Implementation of glycomic analysis in a clinical setting is currently a challenge. It might be easier to use the SIA method to display changes in the glycoprotein partition coefficient as an indication of altered glycosylation profile.

The method of solvent interaction analysis is suitable for the discovery and clinical monitoring of changes in all the above features of protein biomarkers as described below.

6. IsoPSA as an Illustration of the Power of SIA

Prostate-Specific Antigen (PSA)-based screening has substantially improved the early detection of prostate cancer (PCa); however, it is limited by poor specificity for clinically significant disease, particularly within the 4–10 ng/mL “gray zone,” where elevated PSA levels are frequently attributable to benign prostatic conditions rather than malignancy [248,249,250,251,252,253]. Consequently, contemporary clinical guidelines recommend a shared decision-making approach to PSA screening in average-risk men aged 55–69 years, emphasizing the need to balance the potential reduction in prostate cancer–specific mortality against the risks of overdiagnosis and overtreatment associated with indiscriminate biopsy and treatment [254,255,256]. Collectively, these limitations highlight a critical unmet need for a blood-based biomarker capable of more accurately identifying patients harboring actionable, high-grade prostate cancer.

At the molecular level, malignant transformation of prostate epithelium is associated with alterations in PSA isoforms, including post-translational changes in glycosylation patterns, protein conformation, and intermolecular interactions, relative to PSA derived from benign tissue [242,244,245,257,258] (see Figure 5). IsoPSA assay instead of only measuring the total amount of PSA in the blood analyzes the structural changes (isoforms) of the protein. It is designed to detect these variants using an ATPS, generating a ratiometric index that reflects the composite physicochemical properties of circulating PSA [259]. This index has been shown to discriminate high-grade prostate cancer (HG-PCa; Gleason score ≥7/Grade Group ≥2) from low-grade disease (LG-PCa) and benign histopathology on biopsy, using the proprietary IsoClear Platform (Cleveland Diagnostics, Inc.) [260,261,262].

The clinical value of IsoPSA in informing biopsy decision-making is grounded in the well-established limitations of total PSA, particularly within the “gray zone”, where confounding factors such as benign prostatic hyperplasia, inflammation, and pharmacologic influences can significantly impair specificity [254,255,256].

These factors limit the ability of total PSA to reliably distinguish HG-PCa from LG-PCa or non-malignant conditions. In parallel, multiparametric Magnetic Resonance Imaging (mpMRI), while an important adjunctive tool, demonstrates reduced diagnostic certainty in patients with PI-RADS 1–3 findings, where false-negative, indeterminate, and false-positive results remain clinically relevant limitations [263].

IsoPSA addresses these diagnostic challenges through a distinct, mechanism-based approach. Rather than quantifying total PSA concentration or targeting predefined molecular isoforms, the assay evaluates cancer-associated structural alterations in PSA in an agnostic manner. Malignant prostate cells produce PSA with altered glycosylation profiles, conformational states, and protein-binding characteristics, which collectively influence its partitioning behavior and are captured by the IsoPSA index [242,244,245,257,258,259].

In a prospective, multicenter clinical validation study of men aged ≥50 years with total PSA ≥4 ng/mL undergoing diagnostic biopsy, IsoPSA achieved an area under the receiver operating characteristic curve (AUC) of approximately 0.78 for the detection of high-grade prostate cancer [260]. At prespecified IsoPSA index cutoffs, the assay maintained high sensitivity for HG-PCa while enabling an estimated reduction of up to about 45% in unnecessary biopsies among men at low risk, reflecting favorable tradeoffs between sensitivity, specificity, and the proportion of avoided biopsies.

Longitudinal data further supports the prognostic relevance of baseline IsoPSA values. In a retrospective study of men with elevated PSA followed for up to 30 months, patients with low IsoPSA indices (IsoPSA index ≤ 6) demonstrated a very low subsequent risk of clinically significant prostate cancer (csPCa), with estimated risks of approximately 0.4%, 2.5%, and 6.3% at 12, 24, and 30 months, respectively, compared with roughly 5.9%, 31.7%, and 49.5% among those with high IsoPSA indices [261]. In this context, IsoPSA showed high sensitivity and specificity for csPCa and robust negative and positive predictive value at clinically relevant index thresholds, supporting its usefulness to inform biopsy decisions in important clinical contexts (e.g., early detection, monitoring, and active surveillance).

In addition, IsoPSA has been evaluated in combination with mpMRI and PI-RADS (Prostate Imaging Reporting and Data System) scoring to refine biopsy decision-making. In patients with elevated PSA who underwent MRI, an elevated IsoPSA Index (for example, ≥6.0) and PI-RADS 4–5 lesions were both significant predictors of csPCa. Among men with negative or equivocal MRI findings (PI-RADS 1–3), a low IsoPSA ratio (≤6) was associated with a predicted csPCa probability of <5%, suggesting that integration of IsoPSA with mpMRI can improve risk stratification in equivocal imaging scenarios and may permit safe omission of biopsy in carefully selected patients [262]. Taken together, these data from validation, longitudinal, and mpMRI-integrated studies demonstrate statistically informative, reliable, and clinically meaningful IsoPSA performance for pre-biopsy risk stratification, with strong negative and positive predictive characteristics, that directly supports its role in guiding biopsy decisions [260,261,262].

7. Conclusions

In conclusion, this review highlights several critical paradigm shifts in molecular diagnostics and biophysical analysis.

First, Solvent Interaction Analysis (SIA) stands out as a uniquely powerful methodology, offering unparalleled quantitative insights into protein-aqueous media interactions. Unlike conventional biophysical tools, SIA achieves an exceptional level of sensitivity, capturing subtle thermodynamic and hydration changes that were previously unmeasurable.

Second, the traditional reliance on monitoring the expression levels of protein biomarkers is proving inadequate for the molecular diagnostics of complex diseases. Quantitative abundance alone fails to account for the heterogeneous nature of these disorders, often leading to inconsistent diagnostic outcomes.

Finally, shifting the analytical focus to structural alterations in proteins represents a far more robust frontier. Detecting early changes in protein conformation, misfolding, or post-translational modifications can substantially improve the early detection of oncological and neurological disorders, ultimately paving the way for more timely and precise clinical interventions.