Efficient Sub-1 Minute Analysis of Selected Biomarker Catecholamines by Core-Shell Hydrophilic Interaction Liquid Chromatography (HILIC) with Nanomolar Detection at a Boron-Doped Diamond (BDD) Electrode

A rapid, sensitive method for the separation of catecholamine biomarkers (CAs), of importance in traumatic brain injury (TBI) and in Parkinson’s disease (PD), has been successfully developed using hydrophilic interaction liquid chromatography (HILIC). Dopamine (DA), epinephrine (EPI), and norepinephrine (NE) are known to be three to fivefold elevated above normal in traumatic brain injury (TBI) patients. HILIC facilitates the rapid and efficient separation of these polar biomarkers, which can be poorly retained by reversed-phase liquid chromatography (RPLC), while electrochemical detection (ECD) at the boron-doped diamond (BDD) electrode provides enhanced nanomolar detection. Three HILIC columns were compared, namely the superficially porous (core-shell) Z-HILIC column and the Z-cHILIC and Z-HILIC fully porous columns. The core-shell Z-HILIC showed the highest efficiency with a rapid separation within 60 s. The HILIC method utilizing the core-shell Z-HILIC column was initially optimized for the simultaneous analysis of DA, EPI, and NE using UV detection. The advantages of using the BDD electrode over UV detection were explored, and the improved limits of detection (LODs, S/N = 3) measured were 40, 50, and 50 nM for DA, EPI, and NE, respectively. Method validation is reported in terms of the linearity, repeatability, reproducibility, and LODs. Furthermore, the proposed method was successfully applied to the real sample analysis of urinary CAs following phenylboronic acid (PBA) solid phase extraction (SPE) pretreatment.


Introduction
Catecholamines (CAs) are biogenic amines consisting of a catechol group with an amine side chain, which are synthesized from the amino acid tyrosine [1]. Among the most critical CAs in biological systems are dopamine (DA), epinephrine (EPI), and norepinephrine (NE) [2]. They are released from the adrenal glands and function as neurotransmitters that facilitate intercellular communication within the nervous system [3]. Therefore, they have an essential biological role for many physiological functions, and their levels can be regarded as key biomarkers for several diseases and neurological disorders such as some forms of cancer, Parkinson's disease (PD), and Alzheimer's disease (AD) [2,4,5].
DA is the most abundant CA, and it has a major influence on behavior and physical function. These include learning, motivation, memory [5], movement [6], and heart and kidney function [7]. EPI and NE also play important roles as they control the regulation of the sympathetic nervous system. Based upon their concentration levels, they affect the muscular and cardiovascular function [8]. The overproduction of specific CAs in urine and plasma is utilized for biomarking for types of tumors, such as phaeochromocytomas, neuroblastomas, and ganglioneuromas [9,10]. cortex and striatal in mouse samples [28], six antidepressant drugs in blood [50], and phenolic compounds in whiskey samples [51].
To the best of our knowledge, this is the first application of HILIC core-shell stationary phases coupled with a BDD electrode for the rapid and sensitive detection of CAs. Amperometric ECD is especially suitable for the sensitive determination of electroactive substances due to its high selectivity, relative simplicity, and low cost. The optimization of the separation and detection conditions, including HILIC column selection, was carried out prior to employing the method for the simultaneous analysis of the selected biomarkers in urine samples.
An aqueous filter membrane (0.2 µm) was purchased from Millipore Ltd., Ireland (Cork, Ireland). Phenylboronic acid cartridges (PBA, 100 mg) were obtained from Agilent Technologies (Cork, Ireland). Stock solutions were prepared daily at a concentration of 100 mM in 0.1% formic acid and stored in the dark at 4 • C. The stock solutions were stable for one week and used for further dilution with mobile phase (80:20) ACN: H 2 O to obtain the desired concentrations of working standards.
For urine sample analysis, urine samples were obtained from two healthy individuals within the research group. A 24-h urine collection was carried out by collecting the urine in a special container over a full 24-h period. The urine samples were kept away from light and stored in the fridge throughout the 24-h period. Lastly, the urine samples were pre-treated using PBA solid phase extraction (SPE) as described elsewhere [32].
Briefly, the PBA cartridge was initially equilibrated with 1 mL of 80:20 ACN: H 2 O (v/v) containing 1% formic acid, and then with 1 mL of 50 mM phosphate buffer (pH 10). A 3 mL of buffered urine sample (prepared by mixing 1 mL urine and 2 mL of phosphate buffer, pH 8.5) was then loaded to the cartridge. The cartridge was washed with 1 mL of (50:50 v/v) ACN: 10 mM phosphate buffer pH 8.5. Eventually, the analytes were eluted from the cartridge with 1 mL of 80:20 ACN: H 2 O (v/v) containing 1% formic acid and analyzed by HILIC-BDD.

Cyclic Voltammetry (CV)
An electrochemical workstation (CH Instrument, Bee Cave, TX, USA) was used to determine the electrochemical behavior of CAs. An electrochemical cell consisting of silver/silver chloride (Ag/AgCl/3M KCl) as a reference electrode (BASi Analytical Instruments, West Layette, IN, USA), BDD as the working electrode (Windsor Scientific, Slough Berkshire, UK), and a Pt wire as a counter electrode (Sigma-Aldrich, Dublin, Ireland) was utilized. Ammonium formate buffer of pH 3 in 85% ACN was used as the supporting electrolyte for all voltammetric measurements. A stock solution of 100 mM for each CAs was prepared in 0.1% formic acid.

HPLC Instrumentation
An Agilent 1290 Infinity II UHPLC system (Agilent Technologies, Waghaeusel, Germany) consisting of 1290 Infinity II High Speed Pump (model G7120A), degasser (model G7116B), multi-sampler (model G7167B), and diode array detection (DAD) (G7117B) system was used. Agilent Open Lab CDS workstation software was used for the data analysis. A  Germany). The ECD setting is described in [51]. An Antec Flexcell thin layer flow cell with a cell volume of 0.7 µL (Apex Scientific, Co., Kildare, Ireland) was used for ECD. The flow cell consists of a three electrode configuration with a working BDD electrode (8 mm diameter), a HyREF (Pd/H2) reference electrode and carbon-loaded polytetrafluoroethylene (PTFE) counter electrode. A CHI660E electrochemical workstation was used for data analysis (CH Instrument, Bee Cave, TX, USA).

Chromatographic Conditions
Isocratic separation mode was examined for the separation of the selected biomarker CAs by varying the percentage of ACN (75-90%), the ammonium formate pH (3)(4)(5) and concentration (10-20 mM), and flow rate (0.5-1.5 mL/min). The optimum mobile phase composition consisted of 85% ACN (v/v) and 10 mM ammonium formate at pH 3. The selective mobile phase was prepared by mixing 85% ACN with 66 mM ammonium formate (pH 3) to give 10 mM buffer concentration. The mobile phase was filtered and sonicated for 15 min. Each column was equilibrated with the mobile phase for 20 min prior to injection. The separation time using the core-shell Z-HILIC column was less than 60 s with a flow rate of 1.5 mL/min and injection volume of 5 µL. The column temperature was set at 23 • C. LC-UV detection was carried out at 280 nm, while +1.3 V in oxidative mode was used in LC-ECD.

Validation Procedures
The statistical analysis was applied in the method validation for the determination of the limit of detection (LOD), linearity, repeatability, accuracy, and precision. LODs were determined utilizing a signal to noise (S/N) ratio of 3. The linearity was based on linear regression analysis of the calibration curve between various standard concentrations versus peak areas of each standard (50 to 500 µM for LC-UV, and 0.1 to 25 µM for LC-BDD)). The intra-and inter-day results were evaluated via triplicate runs of each standard mixture with concentration 500 µM for LC-UV and 25 µM for LC-BDD. The relative standard deviation (RSD%) was obtained to express the precision of the method.

Electrochemical Behavior of the CAs Using Cyclic Voltammetry (CV)
The direct electrochemical responses of DA, EPI, and NE at the bare BDD electrode were recorded in 85% ACN with 10 mM ammonium formate buffer at pH 3. In brief, CAs exhibit similar electrochemical behavior as they oxidize to their corresponding o-quinones as illustrated in Scheme 1. CV analysis revealed well-defined oxidation peaks for DA, EPI, and NE on the forward scan at potentials of +0.82, +0.9, and +0.85 V, respectively. The quasi-reversible reactions of DA, EPI, and NE on the BDD electrode were indicated by the corresponding cathodic peaks at +0.07, −0.27, and +0.11 V, respectively, as illustrated in Figure 1.
The electrochemical behavior of the CAs has been extensively reported in the literature [52][53][54][55]. Due to the similar redox potentials of the CAs, it is challenging to detect them simultaneously. Several reports have investigated new methods for CA detection based on electrode modification [43,56,57]. However, often due to the complexity, difficulty, and irreproducibility of electrode modification, hyphenation with rapid and robust separation methods is highly effective. The electrochemical behavior of the CAs has been extensively reported in the literature [52][53][54][55]. Due to the similar redox potentials of the CAs, it is challenging to detect them simultaneously. Several reports have investigated new methods for CA detection based on electrode modification [43,56,57]. However, often due to the complexity, difficulty, and irreproducibility of electrode modification, hyphenation with rapid and robust separation methods is highly effective.  The electrochemical behavior of the CAs has been extensively reported in the literature [52][53][54][55]. Due to the similar redox potentials of the CAs, it is challenging to detect them simultaneously. Several reports have investigated new methods for CA detection based on electrode modification [43,56,57]. However, often due to the complexity, difficulty, and irreproducibility of electrode modification, hyphenation with rapid and robust separation methods is highly effective.

Comparison of Zwitterionic HILIC Columns
Three different commercially available zwitterionic columns were tested for the separation of the CAs. Separation using two fully porous Z-HILIC columns was compared to the core-shell Z-HILIC column under the same separation conditions ( Figure 2). The core-shell Z-HILIC phase uses a novel zwitterionic stationary phase bonded to a robust hybrid silica particle, while the Z-cHILIC and Z-HILIC fully porous columns contain a zwitterionic phosphorylcholine functional group and sulfobetaine functional group, respectively.

Comparison of Zwitterionic HILIC Columns
Three different commercially available zwitterionic columns were tested for the separation of the CAs. Separation using two fully porous Z-HILIC columns was compared to the core-shell Z-HILIC column under the same separation conditions (Figure 2). The coreshell Z-HILIC phase uses a novel zwitterionic stationary phase bonded to a robust hybrid silica particle, while the Z-cHILIC and Z-HILIC fully porous columns contain a zwitterionic phosphorylcholine functional group and sulfobetaine functional group, respectively. The present study confirmed that the core-shell column provided superior performance over the fully porous columns in terms of efficiency and speed. The column efficiency is indicated as the number of theoretical plates (N) per column, and the RSD (%) values are reported in Table 1. Table 1 indicates that the core-shell Z-HILIC column provided an almost twofold increase in efficiency in comparison to the two fully porous columns. The separation time was reduced by approximately 60% using the core-shell column ( Figure 3). The present study confirmed that the core-shell column provided superior performance over the fully porous columns in terms of efficiency and speed. The column efficiency is indicated as the number of theoretical plates (N) per column, and the RSD (%) values are reported in Table 1. Table 1 indicates that the core-shell Z-HILIC column provided an almost twofold increase in efficiency in comparison to the two fully porous columns. The separation time was reduced by approximately 60% using the core-shell column ( Figure 3). retention time of the analytes can shed light on the mechanism involved in the separation. A linear plot would indicate ion exchange retention between the analytes and the stationary phase [32,61]. However, a slightly curved line plot indicates the presence of other mechanisms, such as HILIC partitioning and electrostatic interactions [32]. To obtain a rapid HILIC separation in less than 60 s, faster flow rates were examined, and 1.5 mL/min with a corresponding backpressure of 360 bar was chosen, as illustrated in Figure S2.  Selectivity (α) is another important factor that indicates the ability of the stationary phase to discriminate between analytes. α values were calculated between DA and EPI, and EPI and NE for each column (Table 2, Figure 2). The core-shell Z-HILIC column provided the best α value between DA and EPI at 1.3, while the Z-cHILIC provided the highest α value between EPI and NE at 1.84. Selectivity (α) values greater than 1 indicate that the three columns provided sufficient selectivity for the separation of CAs [58]. However, as the core-shell Z-HILIC column displayed higher efficiency and rapid separation, it was selected for all subsequent experiments.

Optimization of HILIC Separation Conditions
The chromatographic separation of the selected biomarker CAs using the core-shell Z-HILIC column was initially performed at a flow rate of 0.5 mL/min and UV detection at 280 nm and was optimized in terms of the mobile phase composition, buffer pH, and concentration. Different mobile phase compositions were tested with varying percentages of ACN in 10 mM ammonium formate buffer, as illustrated in Figure 3. A higher concentration of ACN in the mobile phase increase the retention times (Figure 3), as the polar analytes have a higher degree of partitioning in the thin water-enriched layer on the stationary phase [32].
Furthermore, it is important to highlight that, as the concentration of the organic solvent is increased, the pH of the mobile phase also increases. The measured pH in the presence of high organic solvent content ( s w pH) differed considerably from the actual pH measured on water ( w w pH) using the pH meter calibrated with aqueous buffers [32,59,60]. This can potentially aid the deprotonation of the silanol group and increase the secondary interaction between the analyte and the silanol group, which would cause peak distortion and tailing [32,59].
Also, Figure S1 demonstrates the major effect of buffer pH on the peak shapes and peak intensities for the CAs. Increasing the pH from 3 to 5 caused peak broadening and tailing, and the peak heights were reduced by 18% for DA, 33% for EPI, and 31% for NE. Similar findings were reported by Abhinav et al. for the separation of CAs using HILIC [32]. It is worth noting that CAs are very reactive molecules, and they are kept in vesicles with acidic media in the brain as they can be easily degraded to their corresponding o-quinones forms in an alkaline environment [4]. Therefore, developing a separation method with a suitable pH value requires careful attention with CAs.
In addition, as the concentration of the ammonium formate buffer at pH 3 increased from 10 to 20 mM, the retention times of the analytes decreased slightly, and the peaks became sharper [33]. The relationship between the inverse of buffer concentration and the retention time of the analytes can shed light on the mechanism involved in the separation. A linear plot would indicate ion exchange retention between the analytes and the stationary phase [32,61]. However, a slightly curved line plot indicates the presence of other mechanisms, such as HILIC partitioning and electrostatic interactions [32]. To obtain a rapid HILIC separation in less than 60 s, faster flow rates were examined, and 1.5 mL/min with a corresponding backpressure of 360 bar was chosen, as illustrated in Figure S2.

Optimization of Detection Potential for LC-ECD
The influence of the detection potential on the oxidation of the CAs was investigated in the range of +1.0 to +1.7 V using amperometric detection. Figure 4 shows the hydrodynamic voltammograms of the CAs under the optimum separation conditions with the maximum oxidation currents occurring at +1.3 V.

Optimization of Detection Potential for LC-ECD
The influence of the detection potential on the oxidation of the CAs was investigated in the range of +1.0 to +1.7 V using amperometric detection. Figure 4 shows the hydrodynamic voltammograms of the CAs under the optimum separation conditions with the maximum oxidation currents occurring at +1.3 V.

Method Validation
Linear calibration curves were obtained by the injection of a mixture of CAs with six different concentrations under the optimal separation conditions, initially with LC-UV (Table S1) and then with LC-BDD. The linear calibration range (0.1-25 µM), regression equation, correlation coefficients, and LODs are presented in Table 3 for LC-BDD. Linear regression analysis was performed by plotting the peak areas versus concentrations with a correlation coefficient (R 2 ) of more than 0.99. The linearity of the three CAs was evaluated by triplicate injection of different concentrations of the standard mixtures.
The LODs (calculated at a S/N of 3) were 40 nM for DA and 50 nM for both EPI and NE with ECD. The LODs of CAs detected by the BDD electrode were lower than those obtained by UV, indicating the higher sensitivity achievable with the BDD electrode. The LODs reported in the present work were lower than those previously reported methods using the capillary electrophoresis (CE) chemiluminescence (CL) method [62], and using a magnetic molecularly imprinted polymer (MMIP) for the sorbent extraction and CE analysis [63].
However, the LODs of CAs using LC-MS [64] and LC-fluorescence detection [29] were lower than the present LODs, shown in Table 4. The nanomolar detection achievable with LC-BDD indicates its suitability in providing early diagnosis of TBI and cancer types previously stated. Repeatability and intermediate precision (intra-and inter-assay) were calculated as RSD%. The intra-day and inter-day precision levels based on retention time were in the range of 0.35-0.42% and 0.4-0.65%, respectively. This indicates good method repeatability, further confirming the applicability of this method for analysis of CAs (Table 5).

Method Validation
Linear calibration curves were obtained by the injection of a mixture of CAs with six different concentrations under the optimal separation conditions, initially with LC-UV (Table S1) and then with LC-BDD. The linear calibration range (0.1-25 µM), regression equation, correlation coefficients, and LODs are presented in Table 3 for LC-BDD. Linear regression analysis was performed by plotting the peak areas versus concentrations with a correlation coefficient (R 2 ) of more than 0.99. The linearity of the three CAs was evaluated by triplicate injection of different concentrations of the standard mixtures.
The LODs (calculated at a S/N of 3) were 40 nM for DA and 50 nM for both EPI and NE with ECD. The LODs of CAs detected by the BDD electrode were lower than those obtained by UV, indicating the higher sensitivity achievable with the BDD electrode. The LODs reported in the present work were lower than those previously reported methods using the capillary electrophoresis (CE) chemiluminescence (CL) method [62], and using a magnetic molecularly imprinted polymer (MMIP) for the sorbent extraction and CE analysis [63].
However, the LODs of CAs using LC-MS [64] and LC-fluorescence detection [29] were lower than the present LODs, shown in Table 4. The nanomolar detection achievable with LC-BDD indicates its suitability in providing early diagnosis of TBI and cancer types previously stated. Repeatability and intermediate precision (intra-and inter-assay) were calculated as RSD%. The intra-day and inter-day precision levels based on retention time were in the range of 0.35-0.42% and 0.4-0.65%, respectively. This indicates good method repeatability, further confirming the applicability of this method for analysis of CAs (Table 5).

Urine Sample Analysis
The applicability of the method was examined by evaluating its use in determining the concentrations of the three selected CAs in urine samples using the optimal LC-BDD method following PBA SPE [66][67][68][69]. In this study, PBA SPE, as described by Abhinav et al. [32], was used for urine sample pretreatment using the elution solvent under acidic conditions with 80% ACN making it a suitable and efficient step prior to HILIC separation. The recovery results obtained for each analyte after spiking urine samples at three different CAs concentration levels are shown in Table 6. RSD expresses the variation of peak areas for each analyte (n = 6 each). The utilized SPE method provided good recovery values ranging from 90% to 94%. These finding are in agreement with previously reported values [32]. DA and NE were successfully quantified in the unspiked urine samples, and the calculated concentrations are presented in Table 7. EPI was not quantifiable in the urine samples. This might be due to the low concentration of EPI, below the detection limit, which was also reported by Xu et al. [62]. The expected ranges of urinary CAs over a 24 h period were 65-400 µg/L DA, 0-20 µg/L EPI, and 15-80 µg/L NE [70]. The calculated values of DA and NE in urine samples were in the normal range of their concentration in urine. Figure 5 shows a representative chromatogram of a urine sample spiked with a mixture of CAs (3 µM). DA and NE were successfully quantified in the unspiked urine samples, and the calculated concentrations are presented in Table 7. EPI was not quantifiable in the urine samples. This might be due to the low concentration of EPI, below the detection limit, which was also reported by Xu et al. [62]. The expected ranges of urinary CAs over a 24 h period were 65-400 µg/L DA, 0-20 µg/L EPI, and 15-80 µg/L NE [70]. The calculated values of DA and NE in urine samples were in the normal range of their concentration in urine. Figure 5 shows a representative chromatogram of a urine sample spiked with a mixture of CAs (3 µM).

Conclusions
An efficient, rapid, and sensitive separation method for dopamine (DA), epinephrine (EPI), and norepinephrine (NE) has been optimized, combining the superior performance of a core-shell Z-HILIC column coupled to nanomolar ECD at a BDD electrode. The core-shell Z-HILIC column provided higher efficiency with separation in less than 60 s in comparison to the fully porous Z-HILIC stationary phases. The proposed method represents sufficient precision and linearity to enable the rapid separation of DA, EPI, and NE with sensitive detection at a downstream BDD electrode.
The applicability of the developed method was evaluated by determining the CA levels in urine samples following SPE based on PBA. The concentrations of DA and NE in unspiked urine samples were determined, and their values were in the expected range for these CAs in urine. This rapid LC-BDD method could be effectively applied for the screening of CAs overproduction in urine samples for patients with TBI, pheochromocytoma, and paraganglioma [12,71,72].
Informed Consent Statement: Not applicable.