Quantitative Assessment of the in-vitro Binding Kinetics of Antisickling Aromatic Aldehydes with Hemoglobin A: A Universal HPLC-UV/Vis Method to Quantitate Schiff-Base Adduct formation

Abstract

Aromatic aldehydes act as allosteric effectors of hemoglobin (AEH), forming Schiff-base adducts with the protein to increase its oxygen (O₂) affinity; a desirable property in sickle cell disease (SCD) treatment, as the high-O₂ affinity hemoglobin (Hb) does not polymerize and subsequently prevents erythrocytes sickling. This study reports the development, validation, and application of a weak cation-exchange HPLC assay – quantifying the appearance of Hb-AEH adduct – as a “universal” method, allowing for the prioritization of AEH candidates through an understanding of their Hb binding affinity and kinetics. Concentration- and time-dependent Hb binding profiles of ten AEHs were determined with HPLC, followed by the appropriate nonlinear modeling to characterize their steady-state binding affinity (K_D^ss), and binding kinetics second-order association (k_on) and first-order dissociation (k_off) rate constants. Vanillin-derived AEHs exhibited enhanced binding affinity to Hb, primarily due to their faster k_on. Across AEH, k_on and k_off values are strongly correlated (r=0.993, n=7), suggesting that modifications of the AEH scaffold enhanced their interactions with Hb as intended, but inadvertently increased their Hb-AEH adduct dissociation. To our knowledge, the present study is the first to provide valuable insight into Hb binding kinetics of antisickling aromatic aldehydes, and the assay will be a useful platform in screening/prioritizing drug candidates for SCD treatment.

1. Introduction

Sickle cell disease (SCD) is a genetic disorder caused by substitution of glutamic acid (Glu6) in the β-chain of normal adult hemoglobin (HbA) with a valine (Val6) to form sickle hemoglobin (HbS). Under hypoxic conditions, the non-polar βVal6 of one deoxygenated HbS (deoxyHbS) molecule interacts with a hydrophobic pocket of an adjacent deoxyHbS molecule to form insoluble polymers inside erythrocytes, or red blood cells (RBCs), deforming the RBCs into sickle shape.[1–5] The rigid sickled cells occlude small blood vessels, triggering several secondary pathologies, such as painful crises, organ damage, oxidative stress, hemolysis, impaired microvascular blood flow, morbidity, and mortality.[4,6–10] Several synthetic allosteric effectors of hemoglobin (AEH), particularly those containing aromatic aldehydes, have been identified to ameliorate in-vitro and/or in-vivo RBC sickling.[4,8,9,11–18] The mechanism of action of aromatic aldehyde AEHs entails forming a “transiently covalent” interaction via Schiff-base reaction between the aldehyde group and the nitrogen atom of the N-terminal valine residue of α-globin chains (α1Val1 and α2Val1) of liganded or relaxed state (R-state) Hb.[4,8,9,11–18] The Schiff-base formation leads to an increase in Hb oxygen (O₂) affinity, preventing the hypoxia-induced HbS polymerization and/or leads to direct polymer destabilization, resulting in inhibition of erythrocyte sickling.[4,8,9,11–18] Well-known examples of antisickling aromatic aldehydes are Tucaresol, 5-Hydroxymethyl-2-furfural (5-HMF), vanillin (4-Hydroxy-5-methoxybenzaldehyde), and several of their derivatives, as well as more recent compounds, e.g. PP-14, VZHE-039, and GBT-440 (Voxelotor).[4,8,9,11–17,19] Tucaresol and 5-HMF have previously been studied in early clinical testing,[4,8,9,18] while GBT-440 was approved for the treatment of SCD in 2019,[4,9,20,21], and PP-14 and VZHE-039 undergoing preclinical studies for the treatment of SCD.[9,11,12]

The degree of changes in O₂-affinity of Hb and the ensuing inhibition of sickling of RBCs by AEH have been shown to positively correlate with modification of Hb as a result of the Schiff-base interaction.[11,13,14] The Schiff-base reaction is a concentration- and time-dependent binding.[13,14] Glucose, an endogenous moiety, is also known to bind to Hb via Schiff-base reaction in a concentration- and time-dependent manner.[22,23] The Schiff-base further undergoes Amadori rearrangement to form HbA1_C, which is used widely in clinic as a surrogate endpoint and diagnose for diabetes.[24] Unlike 5-HMF, where dissociation from Hb is slower than association to Hb, glucose is found to be 1,000 times slower in associating to Hb than dissociating, followed by the Amadori rearrangement.[22]

Several studies have reported the use of reverse phase HPLC-UV/Vis for the quantification of the Schiff-base formed by 5-HMF or other aromatic aldehydes with Hb in both non-biological and biological matrices.[13,25–27] One such validated method, refer to as the “reference method” usess ultracentrifugation to separate 5-HMF-bound and unbound Hb.[28] The method involves first reacting 5-HMF with Hb and passing through a filter with a cutoff 10 kD and using reverse phase HPLC-UV/Vis to determine the concentrations of Hb bound and unbound 5-HMF, followed by using nonlinear regression to estimate the forward rate constant (k₁) and backward rate constant (k₋₁), alongside with steady-state binding affinity (K_D). Unfortunately, the reference method is only suitable for hydrophilic compounds, e.g. 5-HMF but not lipophilic compounds, due to non-specific binding to the ultracentrifugation membrane.[28] Moreover, the method requires validation for each individual AEHs.[28] LC-MS/MS has also been reported for quantification of 5-HMF in biological matrices with the advantages of improved detection limit, and not having to deal with absorption interference compared to UV detection, and identification of multiple interest of analytes in complex system.[29,30] Besides the high cost involved in this LC-MS system, it also requires more tedious sample preparation processes. As an alternative to these methods, we have developed and validated a weak cation exchange HPLC-UV/Vis assay for Hb adduct (hereafter referred to as “universal method”), supporting the study of the binding kinetics of AEHs with Hb. This assay was used to characterize the steady-state binding affinity (K_D^ss), second-order association (k_on) and first-order dissociation (k_off) rate constants of several previously studied and/or reported antisickling compounds, including 5-HMF, 2,5-DMP, Vanillin, TD-7, TD-8, TD-9, PP-14, INN-310 (or SAJ-310), VZHE-039, and GBT-440 (Figure 1) with HbA, providing insight into the binding kinetics of aromatic aldehydes interactions with Hb. The previously reported reference method[28] was used to indirectly confirm the validation of the universal method described here.

Figure 1:

Previously studied or reported antisickling aromatic aldehydes

2. Materials and Methods Section:

2.1. Study approvals

Hemoglobin for the binding studies were obtained from normal whole blood, which was collected from adult donors (> 18 years) after informed consent, in accordance with regulations of the IRB for Protection of Human Subjects (IRB# HM1) by the VCU Human Research Protection Program/Institutional Review Board.

2.2. Materials and reagents

NERL^® high purity water (Fisher Scientific, Nazareth, PA), Bis-tris (Sigma-Aldrich, St.Louis, MO) (209.24 g/mol), Potassium cyanide (Sigma-Aldrich, St.Louis, MO) (KCN, 65.12 g/mol), Sodium chloride (Sigma-Aldrich, St.Louis, MO) (NaCl, 58.44 g/mol), AFSC Hemo Control (Cat. No. 5331) (Helena laboratories, Beaumont, TX), Methanol, HPLC grade (Burdick and Jackson, Morristown, NJ), Hydrochloric acid, ACS 37% (Sigma-Aldrich, Milwaukee, WI) (HCl,36.46 g/mol), Sodium cyanoborohydride (sigma-aldrich, St.Louis, MO) (NaBH3CN, 62.84 g/mol), Sodium borohydride (sigma-aldrich, St.Louis, MO) (NaBH4, 37.84 g/mol), EDTA.2Na, (0.5M, pH 8.0) (Quality Biological, Gaithersburg, MD), Dimethyl sulfoxide (DMSO), ≥ 99.9%, A.C.S. Reagent (Sigma-Aldrich, Milwaukee, WI) (78.13 g/mol), 5-Hydroxymethyl-2-furfural (5-HMF) (Sigma, St. Louis, MO) (126.11 g/mol), 2,5-Furandicarboxaldehyde (2,5-DMF) (Sigma-Aldrich, Milwaukee, WI) (124.09 g/mol), Vanillin ReagentPlus^® 99% (Sigma, St. Louis, MO) (152.15 g/mol), GBT-440 (Voxelotor, 337.37 g/mol) (MedChemExpress, NJ), Chemicals provided by Dr. Martin Safo and Yan Zhang’s labs including: TD-7 (273 g/mol), TD-8 (273 g/mol), TD-9 (273 g/mol), INN-310 (243 g/mol), VZHE-039 (259 g/mol), PP-14 (287 g/mol) (Department of Medicinal Chemistry, VCU).

2.3. Methods

2.3.1. Overall HbA binding study of AEH using cation-exchange HPLC-UV/Vis assay – Universal Method

For each AEH, varying initial concentrations ([AEH]₀ with less than 2%DMSO) were added to natural HbA solution ([HbA]₀ of 0.1 or 0.2 mM) in a 96-well plate and incubated on a rotary shaker at 37 °C. After AEH addition, aliquots were taken at different time intervals until steady state (SS) was achieved. Sodium cyanoborohydride/sodium borohydride mixture (NaBH₃CN/NaBH₄ = 1:1 v/v, 50 mM) was added to terminate the Schiff-base reaction, fixate the Hb-AEH adduct, and reduce the free reactive aldehydes from the mixture.[12,14] HbA and Hb-AEH adduct concentrations were quantitated by HPLC-UV/Vis with a weak cation-exchange column (PolyCAT^A, 35 mm × 4.6 mm) at ambient column temperature and a detection wavelength of 410 nm. A linear gradient wash was applied with buffer A (20 mM Bis-Tris, 2 mM KCN, pH 6.5) and buffer B (20 mM Bis-Tris, 2 mM KCN, 200 mM NaCl, pH 6.6) over 20 minutes (Mobile Phase A 100% – 60% from 0 min to 10 mins, and 60% to 0% from 10 mins to 15 mins, along with a column equilibrium period with 100% Phase A from 16 mins to 20 mins) with flow rate of 1.5 mL/min.

2.3.2. Assay validation

Assay selectivity was assessed with water/stopping solution (V:V = 1:1) in the presence of 2% DMSO as a control. A mixture of Hb variants, including HbF, HbA, HbS, HbC (Helena Laboratories, Beaumont, TX) was qualitatively assessed in order to confirm the separation of Hb according to their different isoelectric point (pI).[13] The separation of HbA-AEH adduct and HbA mixture were monitored for all AEH. Three HbA calibration curves (0.003–0.4 mM) were prepared and linearity, sensitivity, accuracy, and precision were evaluated. Linearity was assessed using linear regression with 1/y weighting factor applied in order to cover large [AEH]0 range. Sensitivity (slope), accuracy (DNF = (Predicted concentration – Nominal concentration) / Nominal concentration) and precision (RSD = SD / Mean) were evaluated. eLOQ was estimated as the lowest observed nominal concentration which has acceptable of %DFN (< 20%) and %RSD (<20%), corrected by the dilution factor. Concentrations below eLOQ were not included in the data set. A 3D capture from (200 nm to 600 nm) was set up on UV/Vis to compare the absorbance of HbA and HbA-AEH adduct.

2.3.3. Data analysis

A sigmoidal B_max model (Equation 1) was fit to the steady-state HbA-AEH adduct concentration – AEH₀ profiles by nonlinear regression (Scientist^®) characterized with maximal HbA-AEH adduct formed (B_max), equilibrium dissociation constant (K_D^ss), and Hill coefficient (n). A bimolecular interaction (AEH:Hb = 1:1) model was fit to the HbA-AEH adduct concentration-time profiles across [AEH]₀ using Scientist^® (version 3.0, Micromath) to estimate primary and secondary binding parameters, i.e., forward rate constant (k_on) and backward rate constant (k_off), as well as dissociation constant (K_D) and equilibrium half-life (t_1/2^eq), using the equations below (Equation 2– 3).

$$\text{Hb-AEH adduct }=\frac{B_{max}*{AE{H}_0}^n}{{K_D}^n+{AE{H}_0}^n},$$ - Equation 1

$${\text{K}}_{\text{D}}=\frac{k_{off}}{k_{on}},$$ - Equation 2

$${\text{t}}_{1/2}^{\text{ eq}}=\frac{0.693}{k_{on}*H{b}_0+k_{off}},$$ - Equation 3

3. Results

3.1. Compounds and HbA selection for study

We chose several previously-studied antisickling aromatic aldehydes for our study based on their diverse psychochemical, structural and functional/biological properties: 5-HMF, 2,5-DMP, Vanillin, TD-7, TD-8, TD-9, PP-14, INN-310, VZHE-039, and GBT-440 (Figure 1). In particular, VZHE-039, PP-14, and INN-310 were chosen for their potently modifying Hb (100% HbS-AEH adduct).[11,12] The isomers TD-7, TD-8, and TD-9 were chosen because of their different HbS binding profiles with varying Hb modification potencies,[14] which may be due to different primary target (Hb) binding affinity and/or binding kinetics, as well as ALDH-mediated metabolism that reduced AEH concentrations in RBC. Vanillin, the parent compound of the eight chosen benzaldehydes with the least potent antisickling activity,[14] was included as a negative control. GBT-440, currently approved for the treatment of SCD, was chosen as the positive clinical reference.[16,17] Two furaldehydes, 5-HMF and 2,5-DMF, were included to enlarge structural variability for structure-activity relationship (SAR) study.[15]

Aromatic aldehyde AEHs bind to the N-terminal αVal1 amines at the α-cleft of Hb, which is identical in HbA and HbS.[3] Also, HbS differs from normal HbA by only a single base substitution, resulting in the substitution of the polar β6-glutamic acid by the neutral amino acid, valine – the overall fold of both structures are identical with only minor localized structural differences at the βVal6 mutation site.[3] Crystal structures with either HbA or HbS show identical AEH binding. It is therefore expected that the binding kinetics of AEH to HbA and HbS will be similar, necessitating the use of HbA, which is easier to obtain in large quantity. Consistently, our earlier comparative studies using reference and universal method to measure binding affinity of 5-HMF with HbA and HbS resulted in similar K_D value.[28]

3.2. HbA and AEH concentration range and sampling schedule

Based on several pilot studies, preliminary data, literature, and rational design, the final [AEH]₀ range and sampling schedule were designed to cover the full steady state (SS) concentration- and time-dependent HbA binding profiles of AEH. For example, 5-HMF in the presence of HbA has been found to decline and approached equilibrium within 24–30 hrs, with a binding affinity of 0.25 mM.[28] Therefore, 5-HMF was designed to cover concentration range from 0.06 to 4 mM with aliquots collection up to 24 hrs. 2,5-DMF, another furaldehyde, showed enhanced in-vitro apparent binding affinity than 5-HMF in whole blood; therefore, lower concentrations were included in the scheduling (0.025, 0.05, 0.1, 0.25, 0.5, 1, 2, 4, 8 mM). Since 100% HbA modification was not achieved with 5.8 mM of vanillin from an initial pilot study, higher [vanillin]₀ up to 20 mM (about 10-fold higher than the estimated K_D = 1.7 mM from its pilot study) was used in the definitive study.

TD-7, TD-8, and TD-9 are isomers; however, major differences in their abilities to induce Hb-AEH adduct formation in SCD whole blood have been reported.[14] At 2 mM, TD-7 showed 66% HbS modification, whereas TD-8 and TD-9 only showed 18% and 43%, respectively. Therefore, more condense sampling was scheduled up to 5 hrs for TD-7 and covered lower [TD-7]₀. In contrast, aliquots were collected up to 24 and 27 hrs for TD-8 and TD-9, respectively, and higher [TD-8]₀ and [TD-9]₀ (2 and 4 mM) were studied. In-vitro HbS binding kinetics of INN-310 in whole blood was more like TD-7 than other AEH;[14] therefore, similar concentration and time-dependency study were designed for both compounds.

In-vitro studies showed that 1 mM and 2 mM of PP-14 resulted in 44.4% and 90.8% of inhibition of cell sickling, respectively.[11] PP-14 may behave more like fast and potent binding compound, such as TD-7. Therefore, concentration range of 0.02 mM to 2 mM were chosen, and aliquots were collected up to 4 hrs, according to the experimental design of TD-7.

Unlike the above AEH with kinetic profiles that can be characterized with 0.2 mM HbA, VZHE-039 achieved SS fast (< 6 mins). Therefore, 0.1 mM of HbA was used for VZHE-039 to slow down the overall kinetics. A condense sampling schedule (every 3 mins) was performed in order to capture the kinetics at early reaction time and optimize parameter estimates.

According to the pilot study of GBT-440, SS was achieved almost instantaneously. High GBT-440 concentrations show possible secondary reactions with evidence of new peaks observed on chromatogram after long incubation time, as well as a decrease of the primary HbA-adduct concentration after 2 mins. Therefore, the primary HbA-AEH adduct was assumed to achieve steady state by 2 mins. Since the kinetic study was not achievable under these experimental conditions, only the SS concentration-dependency study was performed. According to its pilot study, lower [GBT-440]₀ (below K_D: 0.02, 0.04, 0.06, 0.08 mM) were used in the definitive study.

3.3. Partial validation of cation-exchange HPLC-UV/Vis assay – Universal Method

To quantify the Schiff-base interaction between AEH and Hb, a cation-exchange HPLC method was used to study HbA binding profiles of 5-HMF, 2,5-DMP, Vanillin, TD-7, TD-8, TD-9, PP-14, IND-310, VZHE-039, and GBT-440. The adduct formation between these AEHs and HbA is captured by HPLC, which monitors the elution of compound-modified and unmodified HbA.[13] The principle of separation is based on differences in isoelectric point (pI) of Hb variants and the pH of the mobile phases. The HbA-AEH adduct decreases the pI of protein, which allows its separation from unmodified HbA. A chromatogram of blank water shows no peak, while 0.003 mM HbA in water shows a single peak. We also note that the HbA peak separates from Hb variants such as HbF, HbS, and HbC. The HbA-AEH adduct with lower pI elutes prior to HbA with a resolution of 2.4. Representative chromatograms are reported as supplementary material (Figures S1–S7)

The peak area is used for quantification. HbA calibration curves (1/y weighted) are linear (R² of 0.992–0.999) with an eLOQ of 0.003 mM and have acceptable precision (%RSD of 4%–15%) and accuracy (%DFN of −20% to 4.5%). The time-dependent reduction of HbA reflects the formation of HbA-AEH adduct (Figure 2). The entire initial HbA ([HbA]₀) was modified for most of the studied AEH, except for GBT-440, where only about 50% HbA were modified to the HbA-AEH adduct and the remaining 50% HbA unmodified (Figure 2). Consistently, unlike the other AEHs that bind two molecules with a Hb tetramer,[11,12,14,15] GBT-440 binds Hb in 1:1 ratio.[16] HbA and HbA-AEH adduct show similar UV/Vis spectra with estimated extinction coefficients of 3.16 and 3.10 AU*mM⁻¹cm⁻¹ at λ_max of 416 nm, respectively (Figure 2). Since no reference chemical compound for HbA-AEH adduct is available, using the HbA calibration curve for HbA-AEH adduct quantification is reasonable, as HbA-AEH adduct and HbA have the same absorptivity.

Figure 2.

Representative mass balance assessment. Experimental data (symbol) were fitted (line) with sigmoidal B_max model and the reverse sigmoidal B_max model for HbA-AEH adduct and HbA, respectively. Mass balance (HbA-AEH adduct + HbA) was fitted by linear regression for visual inspection. Mass balance assessment for VZHE-039 incubated with 0.1 mM HbA solution for 0.4 hrs. (B) Mass balance assessment for GBT-440 incubated with 0.2 mM HbA solution for 0.03 hrs.

3.4. Concentration-dependent steady-state (SS) binding study of AEH with HbA

Steady-state concentration-dependency studies were designed to estimate the primary binding affinity and capacity of each of the compounds. All AEH showed the expected saturable concentration-dependent HbA binding, where HbA-AEH adduct concentrations ultimately achieved SS under the experimental conditions (Figures 3). The entire [HbA]₀ was modified by AEH, resulting in saturation, except for GBT-440, where only about 50% HbA molecules were modified.

Figure 3.

Concentration-dependency profile of AEH binding to HbA using reverse-phase HPLC assay. A sigmoidal B_max model (dashed lines), characterized by maximal AEH-HbA adduct formed/AEH bound (B_max), dissociation equilibrium constant (KD^ss) and Hill coefficient/sigmoidicity factor (n), was fit to the experimentally measured AEH-adduct concentration-dependent SS profiles (marker symbols). (A) HbA binding curves of TD-7, TD-8, TD-9, Vanillin, and 5-HMF. (B) HbA binding curves of GBT-440, VZHE-039, PP-14, INN-310, Vanillin, and 5-HMF.

The sigmoidal B_max model fit all AEH adequately, with good R² values (0.976 to 0.999) and precision of parameter estimates (%COV of 3% to 34%) (Table 1). B_max/HbA₀ values were close to 1, except for vanillin, INN-310, and GBT-440 (Table 1). Consistently positive values for n in SS binding indicate allosteric effects, supported by X-ray crystallography studies, where two AEH molecules are observed binding to the N-terminal Val1 of each of the α-globin chain [11,12,14,15] (with the exception of GBT-440 where only one molecule binds[16]).

Table 1.

Final parameter estimates of SS concentration-dependency of HbA-AEH adduct formation in HbA solution using sigmoidal B_max model. Data are presented as point estimate (PE) with coefficient of variance (COV).

Compound	R2	Bmax /HbA0	KD [mM]	n
5-HMF	0.998	1.0 (6%)	0.37 (16%)	1.1 (10%)
2,5-DMF	0.997	1.0 (15%)	0.19 (25%)	1.7(21%)
Vanillin	0.990	0.70 (7%)	1.7 (16%)	1.9 (24%)
TD-8	0.997	0.90 (8%)	0.58 (19%)	1.4(17%)
TD-9	0.999	0.90 (3%)	0.25 (7%)	2.0 (12%)
TD-7	0.997	0.85 (5%)	0.19 (10%)	1.8 (15%)
INN-310	0.976	0.70 (8%)	0.19 (21%)	2.4 (20%)
PP-14	0.988	0.80 (6%)	0.16 (13%)	2.8 (30%)
VZHE-039	0.986	0.80 (14%)	0.10 (24%)	1.9 (34%)
GBT-440	0.991	0.65 (3%)	0.11 (9%)	1.6 (16%)

Compared to 5-HMF, 2,5-DMF showed a 2-fold enhanced binding affinity (smaller K_D) (Table 1), while compared to the parent compound vanillin, all benzaldehyde or vanillin derivatives exhibited enhanced HbA binding affinity (Table 1). Other tested AEHs also showed enhanced binding affinity over 5-HMF or vanillin; namely, INN-310, PP-14, VZHE-039, and GBT-440 (Table 1).

3.5. Time-Dependent binding study of AEH with HbA

Time-dependency studies were performed for characterization of kinetic reactions via the estimation of the corresponding rate constants and equilibrium half-lives. All AEH showed time-dependent HbA-AEH adduct formation under the designed experimental conditions and selected binding profiles are shown for 5-HMF and VZHE-039 (Figure 4). For all AEH, the 1:1 binding model adequately fits data with good R² values (0.897 to 0.992) and precision of parameter estimates (%COV of 5% to 26%) (Table 2). As expected, increasing initial AEH concentrations ([AEH]₀) led to a more rapid achievement of equilibrium (steady-state), as well as higher HbA-AEH adduct formation. The kinetic binding model adequately described the observed data across a wide range of AEH₀ concentrations and time points and resulted in similar K_D estimates compared to the K_D obtained from SS concentration-dependency studies using Sigmoidal B_max model (Table 3).

Figure 4.

Time-dependency profile of AEH binding to HbA (0.1 or 0.2 mM) using reverse-phase HPLC assay. A bimolecular kinetic binding model (dashed lines), characterized by association rate constant (k_on) and dissociation rate constant (k_off), was fit to the experimentally measured final HbA-AEH adduct concentration-time profiles (marker symbols). (A) HbA binding curves of 5-HMF. (B) HbA binding curves of VZHE-039.

Table 2.

Final parameter estimates of time-dependency of HbA-AEH adduct formation in HbA solution, using the simple kinetic binding model. Data are presented as point estimate (PE) with coefficient of variance (COV).

Compound	R2	kon [mM−1*hr−1]	koff [hr−1]	KD [mM]	t1/2eq [hr]
5-HMF	0.934	0.46 (11%)	0.13 (25%)	0.28	3.1
2,5-DMF	0.980	0.54 (6%)	0.065 (14%)	0.12	4.5
Vanillin	0.956	0.16 (13%)	0.69 (16%)	4.3	0.96
TD-8	0.935	0.18 (10%)	0.10 (26%)	0.55	5.0
TD-9	0.979	0.76 (8%)	0.18 (15%)	0.23	2.1
TD-7	0.992	9.8 (5%)	1.5 (9%)	0.15	0.20
PP-14	0.958	16 (12%)	2.2 (19%)	0.14	0.13
INN310	0.897	5.0 (11%)	0.70 (25%)	0.14	0.41
VZHE-039	0.983	61 (9%)	5.8 (13%)	0.10	0.058

Table 3.

Comparison of the final parameter estimates of SS concentration-dependency of HbA-AEH adduct formation for 5-HMF by the sigmoidal B_max model, using two different HPLC-UV/Vis assay methods. Data were presented as point estimate (PE) with coefficient of variance (COV).

	Reference Method[28]	Universal Method
R 2	0.994	0.998
B max / HbA 0	0.74 (10%)	1.0 (6%)
B max	0.092 (10%)	0.20 (6%)
KD (mM)	0.28 (17%)	0.37 (16%)
n	1.7 (20%)	1.1 (10%)

The tested AEH achieved equilibrium on different time scales (Figure 4, Table 2). At low [AEH]₀ – assuming first-order conditions – 5-HMF, TD-8, and TD-9 equilibrated slowly with HbA, in 12 – 20 hrs, with an equilibrations half-life of 3.1, 5.0, 2.1 hrs, respectively (Table 2). The slow equilibration rates were caused by their small k_on and k_off rate constants. TD-7, being an isomer with TD-8 and TD-9, exhibited significantly more rapid equilibration in less than 1 hr due to its large k_on and k_off values (Table 2). The three TD isomers differ by either the position of the methoxy moiety or the methoxypyridine moiety on the benzene ring. TD-7 has a meta-located methoxy group relative to the aldehyde, while TD-9 has the methoxy group located at the para position on the benzene ring. TD-8 differs from TD-9 by the position of the methoxypyridine moiety, which is located para and ortho to the aldehyde moiety, respectively. These structural differences appear to affect their binding kinetics with HbA. VZHE-039, which is structurally similar to TD-7 but with an ortho-positioned hydroxyl group instead of meta-positioned methoxy group relative to the aldehyde moiety, showed the fastest binding kinetics, achieving equilibrium in less than half an hour (Table 2).

PP-14 and INN-310, which showed slightly lower binding affinity (large K_D) than VZHE-039, exhibited slower overall binding kinetics than VZHE-039 (Table 2). GBT-440, on the other hand, showed the same binding affinity as VZHE-039 in the SS concentration-dependency study (Table 1), but very fast reaction kinetics which could not be characterized under the current experimental conditions. Vanillin – without a bulky side-chain available for supplemental HbA interactions besides the Schiff base binding – exhibits a small k_on and relatively large k_off, resulting in poor HbA affinity (K_D of 4.3 mM) (Table 2).

4. Discussion

Synthetic aromatic aldehydes or AEH have potential use in the treatment of SCD with one such compound, Voxelotor (GBT-440) recently approved for such indication.[4,9,20,21] Their primary antisickling mechanism is the transiently covalent modification of Hb by Schiff-base interaction between the AEH aldehyde moiety and the N-terminal Val1 amines of the α-subunits, leading to the prevention of hypoxia-induced polymerization with concomitant inhibition of RBC sickling.[4,18] Due to the transient nature of the Schiff-base interaction, the onset and duration of the aromatic aldehyde antisickling effect is mostly dependent on the binding kinetics and affinities of these candidate drugs. However, there are currently no reliable analytical methods to fully and efficiently characterize their in-vitro Hb binding kinetics. We report the validation and application of a weak cation-exchange HPLC-UV/Vis assay (Universal Method) to characterize the steady-state binding affinity (K_D), second-order association (k_on) and first-order dissociation (k_off) rate constants of a total of ten previously-studied antisickling aromatic aldehydes. These compounds are characterized by varying physicochemical, structural, and functional/biological properties.

The universal assay method was designed to measure the Hb adduct rather than each individual AEH, thus allowing its use across a series of AEH. It has the advantage of not requiring assay validation for each AEH individually, and overcomes the lack of chemical standard/compound for each Hb-AEH adduct. The universal method of quantifying Hb-AEH adduct is also valid for Hb binding kinetic studies, as the adduct formed is the underlying mechanism for the AEHs pharmacologic activity. The binding estimates for 5-HMF obtained from the previously reported reference assay[28] and our universal assay using the same Sigmoidal B_max model are statistically the same (Table 3), validating the prospective use of the universal assay. A simple kinetic binding model (1:1 binding) was used because of its good model selection criteria, indicating an acceptable fit of the experimental data and resulting in reliable final point estimates for k_on and k_off (Table 2). The different assumptions on AEH to Hb binding ratio between SS concentration- and time-dependent models resulted in slightly different final estimates for K_D (Table 4). However, after accounting for the imprecision of the final estimates as represented by the coefficient of variance (COV%), the differences between KD^ss and K_D^kinetic estimates can be considered negligible, indirectly confirming the proper use of both models for Hb binding affinity and binding kinetic characterization.

Table 4.

Comparison of the final K_D^ss and K_D^kinetic estimates from SS concentration-dependency and time-dependency of HbA-AEH adduct formation in HbA solution, using sigmoidal B_max model and simple kinetic binding model, respectively. Data are presented as point estimate (PE) with coefficient of variance (COV).

Compound	KDss [mM]	KDKinetic [mM]
5-HMF	0.37 (16%)	0.28 (36%)
2,5-DMF	0.19 (25%)	0.12 (20%)
Vanillin	1.7 (16%)	4.3 (29%)
TD-8	0.58 (19%)	0.55 (36%)
TD-9	0.25 (7%)	0.23 (23%)
TD-7	0.19 (10%)	0.15 (14%)
INN-310	0.19 (21%)	0.14 (31%)
PP-14	0.16 (13%)	0.10 (36%)
VZHE-39	0.10 (24%)	0.10 (22%)
GBT-440	0.11 (9%)	NA

The concentration-dependent SS binding study shows the di-aldehyde furaldehyde, 2,5-DMF to exhibit 2-fold enhanced binding affinity (smaller K_D) than the mono-aldehyde furaldehyde 5-HMF (Table 1), possibly due to its additional aldehyde group on the furan ring. The enhanced HbA binding affinity for all substituted benzaldehydes (TD-8, TD-9, TD-7, INN-310, PP-14, VZHE-039, and GBT-440) over the parent compound vanillin (Table 1), is consistent with X-ray crystallography findings (vide infra), which show these compounds to make additional pyridinyl-methoxy side chain interactions with the protein.[11,12,14,16] It is also clear that the positions of the methoxy group and/or the pyridinyl-methoxy group on the benzene ring determine K_D (TD-7<TD-9<TD-8) (Table 1). As noted in the result section, TD-7 and TD-9 have substituted methoxy group relative to the aldehyde, at the meta and para positions, respectively, while TD-8 differs from TD-9 by the position of the methoxypyridine moiety, which is located para and ortho to the aldehyde moiety, respectively (Figure 1). INN-310 shares the same meta-positioned methoxy group on the benzene ring as TD-7, where the only difference is the methylhydroxy group on the pyridine ring, which resulted in identical binding affinity estimates (K_D = 0.19 mM). PP-14, VZHE-039, and GBT-440 share the same ortho-positioned-hydroxy group on the benzene ring and exhibited higher HbA binding affinity, with K_D of 0.16 mM, 0.10 mM, and 0.11 mM, respectively.

The structural differences also appear to affect the compounds binding kinetics with HbA. As observed above for the concentration-dependent SS binding study, the time-dependent binding study also showed that the addition of a methoxy-pyridine side chain to the benzene ring of vanillin to give TD-8, TD-9, TD-7, INN-310, PP-14, VZHE-039, in general, leads to an increase of k_on, which is a desirable affect for drug design (Table 2). However, TD-7, PP-14, and VZHE-039 also show an undesirable increase of k_off (Table 2). These benzaldehyde derivatives unlike vanillin make several additional interaction with the protein beside the common Schiff-base interaction. Of the three TD isomers, TD-7 exhibited significantly more rapid equilibration in less than 1 hr due to its large k_on and k_off values when compared to TD-8 and TD-9 (Table 2). It clear that the ortho-positioned hydroxyl group in VZHE-039 instead of the meta-positioned methoxy group in TD-7, explains why the former exhibits the faster binding kinetics, achieving equilibrium in less than half an hour (Table 2). PP-14 and INN-310 exhibited slower overall binding kinetics than VZHE-039 (Table 2), while GBT-440, although showed the same binding affinity as VZHE-039 (Table 1), exhibited very fast reaction kinetics that could not be measure under the current experimental conditions.

All AEHs across furaldehydes and benzaldehydes, except vanillin and 2,5-DMF, showed a strong correlation (r = 0.993, n=7) between their k_on and k_off values (Figure 5). This observation indicates that the current molecular design strategy did enhance interaction with Hb (as intended), but also inadvertently enhanced dissociation from Hb. This strong correlation also indicates proportional changes of k_on and k_off in the same direction, resulting in a less than 5-fold improvement of K_D from the top candidate VZHE-039, relative to the previous lead compound 5-HMF. However, the overall Hb binding kinetics of VZHE-039 is more rapid than 5-HMF, as reflected by a more than 50-fold shorter t_1/2^eq under the current experimental conditions, due to the increase of both k_on and k_off.

Figure 5.

Correlation of k_on and k_off for AEH. Final point estimates (blue circle marker symbols) were fitted with liner regression (r = 0.993), except two outliners (orange triangle marker symbols), i.e., 2,5-DMF and vanillin.

The crystal structures of some AEH molecules in complex with liganded Hb have been elucidated, which can provide insight into the differences in AEH binding affinities to Hb. Consistent with TD-7 showing the lowest K_D (highest binding affinity) among the three isomers, TD-7 (0.19 mM), TD-8 (0.58 mM), and TD-9 (0.25 mM), the electron density of bound TD-8 and TD-9 were found to be very weak and sparse compared to TD-7,[14] suggesting weaker interactions of TD-8 and TD-9 with Hb than TD-7.

TD-7 and VZHE-039 share the same methylhydroxyl group on the pyridine ring. However, TD-7 differs from VZHE-039, with the former having a meta-positioned methoxy group while the latter has an ortho-hydroxyl group. The methylhydroxyl group on the pyridine ring was incorporated in both compounds to make interactions with the protein, more specifically with a surface-located αF-helix to directly weaken polymer interactions.[9,12,14] The αF-helix is implicated in polymer stabilization through hydrogen-bond interactions with other polymer strands.[31–34] Consistent with this idea, natural mutations, e.g. αAsn78→Lys (Hb Stanleyville), that abrogate these interactions, lead to increased solubility/less sickling of deoxy HbS hetero tetramers, mitigating the severity of SCD.[32] Interestingly, while the methylhydroxyl of VZHE-039 makes the predicted interactions with the αF-helix, that of TD-7 does not, instead rotating away from the αF-helix.[12] Consistently, VZHE-039 showed the lowest K_D (0.1 mM), i.e., VZHE-039 < INN-310 ⁓ TD-7 < TD-9 < TD-8 < vanillin. We also note that VZHE-039 interactions with the αF-helix has resulted in a secondary O₂-independent antisickling activity in addition to the primary O₂-dependent antisickling activity, which may provide clinical benefits over other aromatic aldehydes.[12] As expected from GBT-440 high affinity to Hb, the crystal structure of this compound with Hb shows the isopropyl-pyrazole making significant protein interactions.[16] Interestingly, this moiety failed to make close interactions with the αF-helix, which, in addition to binding only one molecule, explains GBT-440 lack of O₂-independent antisickling activity.[9,16]

As noted above, the current molecular design strategy led to enhanced association, by design, but also enhanced dissociation, albeit unintentionally. Several reasons may explain the undesirable increasing of k_off: 1) The stability of AEH side chains binding with HbA, 2) The electric density of imine carbon and nitrogen that affect the Schiff-base hydrolysis process, or 3) The kinetic features of the Schiff-base reaction. For example, to explain the kinetic difference between 5-HMF and TD-7 (k_on = 0.46 and 9.8 hr⁻¹mM⁻¹, k_off = 0.13 and 1.5 hr⁻¹, respectively), crystal structures were compared. A pair of 5-HMF or TD-7 molecules form Schiff-base interactions with the N-terminal αVal1 nitrogen of the two α-subunits in a symmetrical fashion.[14,15] The additional interactions for 5-HMF involves a strong and intricate network of water-mediated and direct hydrogen bonds.[15] On the other hand, in TD-7, the additional interactions to the proteins are predominantly hydrophobic and weak hydrogen bonds,[14] which may be unstable in an aqueous solution and unable to facilitate the dissociation of AEH from Hb. The Schiff-base hydrolysis reaction of the HbA-AEH adduct may be facilitated with the increasing electropositive and electronegative nature of the imine carbon of AEH and nitrogen of Hb, respectively (Scheme 1). A computational approach using Mulliken charge analysis (Software: Gaussian 09) was applied to calculate the electric density of Schiff-base formed by each AEH. Compared to INN-310 and TD-7, the intramolecular hydrogen bond of PP-14 and VZHE-039 increased the electronegativity of the nitrogen atom (−0.37 for INN-310 and TD-7, −0.54 for PP-14 and VZHE-039) and the electropositivity of the carbon atom (0.071 for INN-310 and TD-7, 0.198 for PP-14 and VZHE-039) of the Schiff-base, which is favorable for the decomposition of the Schiff base (k_off: VZHE-039 > PP-14 > TD-7 > INN-310).

Scheme 1:

Proposed formation and hydrolysis of Schiff-base between AEH and Hb

The reversible “transient covalent” Schiff-base reaction involves intermediates.[35] The kinetic features between the formation and hydrolysis of Schiff-base may be determined by any of the steps shown in Scheme 1. The dissociation (k_off) used in our kinetic model is an overall rate constant that includes all the steps of Schiff-base hydrolysis and is ultimately determined by the rate-limiting part in the Schiff-base hydrolysis reaction steps. To explain the overall increase of k_off for tested AEH and to design a novel AEH with a smaller k_off, further research on the kinetic profile of Schiff-base hydrolysis will be necessary.

5. Conclusions

Current pharmacologic treatment options for SCD are limited to Hydroxyurea (HU), and more recently, L-glutamine, Voxelotor, and Crizanlizumab.[4,9,20,21] A promising alternative approach, currently being investigated by groups in academia and in industry, is to prevent hypoxia-induced RBC sickling by increasing the affinity of Hb for O₂. Aromatic aldehydes are known to bind to Hb and exhibit this pharmacologic activity.[4,8,9,11–18] We report for the first time the validation and application of a weak cation-exchange HPLC-UV/Vis assay (Universal Method) for ten diverse antisickling aromatic aldehydes, which provided insight into the binding kinetics of aromatic aldehydes interactions with HbA. The Universal Method described in this report avoids the use of ultrafiltration, which is cumbersome and impossible for compounds with a high degree of binding to the filtration membrane, and does not require the availability of a chemical standard/compound for each Hb-AEH adduct individually. Using this analytical method, coupled with a modeling approach, key molecular features to enhance Hb binding have been identified, allowing for the rational design of better drug candidates. While Hb binding kinetics may be a major determinant of in-vivo potency, onset, and duration of effect, other drug properties, such as metabolic stability and plasma protein binding, will have to be considered to extrapolate in-vivo ADME and PK/PD properties for further druggability evaluation. The Universal Method, however, can serve as platform technology to allow for the rapid screening and prioritization of AEH drug candidates as part of further development.

Highlights.

• Aromatic aldehydes form Schiff-base adduct with hemoglobin to prevent sickling

• A universal HPLC method to characterize and quantitate Schiff-base formation

• A platform for screening/prioritizing candidates for sickle cell disease treatment