Hydrophilic interaction liquid chromatography with mass spectrometry for the separation and identification of antisense oligonucleotides impurities and nusinersen metabolites.

With the development of therapeutic oligonucleotides for antisense and gene therapies, the demand for analytical methods also increases. For the analysis of complex samples, for example plasma samples, where the use of mass detection is essential, hydrophilic interaction liquid chromatography is a suitable choice. The aim of the present work was to develop a method for separation and identification of the oligonucleotide impurities and metabolites by hydrophilic interaction liquid chromatography. First of all, the effects of different chromatographic conditions (e.g. pH of the aqueous part of the mobile phase, buffer concentration, column temperature) on the retention and separation of phosphorothioate oligonucleotides standards on the amide stationary phase were investigated. A set of model oligonucleotides containing a fully modified 21mer and its typical impurities (shortmers and oligonucleotides with different number of thiophosphate modifications) was used. The results showed that the concentration of the salt in the mobile phase as well as its pH, are the most influential parameters with regard to peak shape and separation. The knowledge gained was applied to the analysis of an unpurified 18mer oligonucleotides, analogues of the drug nusinersen used for the treatment of spinal muscular atrophy. The successful separation and identification of twenty-six and twenty-eight impurities was performed with the developed HILIC method. The method was applied to analysis of nusinersen metabolites of serum samples of patients treated with Spinraza.

Mixed-mode column allows simple direct coupling with immobilized enzymatic reactor for on-line protein digestion.

Liquid chromatography coupled with mass spectrometry is widely used in the field of proteomic analysis after off-line protein digestion. On-line digestion with chromatographic column connected in a series with immobilized enzymatic reactor is not often used approach. In this work we investigated the impact of chromatographic conditions on the protein digestion efficiency. The investigation of trypsin reactor activity was performed by on-line digestion of N-α-benzoyl-L-arginine 4-nitroanilide hydrochloride (BAPNA), followed by separation of the digests on the mixed-mode column. Two trypsin column reactors with the different trypsin coverage on the bridged ethylene hybrid particles were evaluated. To ensure optimal trypsin activity, the separation temperature was set at 37.0 °C and the pH of the mobile phase buffer was maintained at 8.5. The on-line digestion itself ongoing during the initial state of gradient was carried out at a low flow rate using a mobile phase that was free of organic modifiers. Proteins such as cytochrome C, enolase, and myoglobin were successfully digested on-line without prior reduction or alkylation, and the resulting peptides were separated using a mixed-mode column. Additionally, proteins that contain multiple cysteines, such as α-lactalbumin, albumin, ß-lactoglobulin A, and conalbumin, were also successfully digested on-line (after reduction and alkylation). Moreover, trypsin immobilized enzymatic reactors were utilized for over 300 injections without any noticeable loss of digestion activity.

Liquid Chromatography Methods for Analysis of mRNA Poly(A) Tail Length and Heterogeneity.

Messenger RNA (mRNA) is a new class of therapeutic compounds. The current advances in mRNA technology require the development of efficient analytical methods. In this work, we describe the development of several methods for measurement of mRNA poly(A) tail length and heterogeneity. Poly(A) tail was first cleaved from mRNA with the RNase T1 enzyme. The average length of a liberated poly(A) tail was analyzed with the size exclusion chromatography method. Size heterogeneity of the poly(A) tail was estimated with high-resolution ion-pair reversed phase liquid chromatography (IP RP LC). The IP RP LC method provides resolution of poly(A) tail oligonucleotide variants up to 150 nucleotide long. Both methods use a robust ultraviolet detection suitable for mRNA analysis in quality control laboratories. The results were confirmed by the LC-mass spectrometry (LC MS) analysis of the same mRNA sample. The poly(A) tail length and heterogeneity results were in good agreement.

Phosphorothioate oligonucleotides separation in ion-pairing reversed-phase liquid chromatography: effect of temperature.

Analysis of diastereomers of phosphorothioate oligonucleotides in ion-pairing reversed-phase liquid chromatography is affected not only by the character and concentration of ion-pairing system, but also by the separation temperature. In this work, eight ion-pairing systems at two concentrations buffered with acetic acid were used with octadecyl column to investigate the effects of temperature (in the range from 20 °C to 90 °C) on retention, diastereomeric separation, resolution of mers of different length and resolution of oligonucleotides with different number of phosphorothioate linkages. It was observed that elevated temperature suppresses the diastereomeric separation and oligonucleotide peaks become narrower. This improves the resolution of n and n-1 mers at elevated temperature. Plots of ln k (k = retention factor) versus reciprocal absolute temperature show that for 100 mM ion-pairing systems the increase in temperature does not lead to simple decrease in oligonucleotides retention as generally observed in reversed-phase liquid chromatography. The aim of this work is to improve chromatographic method for analysis of phosphorothioate oligonucleotides.

Phosphorothioate oligonucleotides separation in ion-pairing reversed-phase liquid chromatography: Effect of ion-pairing system.

Ion-pairing reversed-phase liquid chromatography was utilized for the analysis of native and phosphorothioate oligonucleotides of the identical sequence but different amount and position of phosphorothioate modifications. The effects of ion-pairing amines nature (alkyl chain length, hydrophobicity) and concentration on retention, n/n-1 resolution, and diastereomeric separation of phosphorothioate oligonucleotides were investigated using octadecyl column. Eight different ion-pairing agents at two concentrations (10 mM and 100 mM) buffered with acetic acid were investigated. The resolution of n and n-1 mers oligonucleotides improved with hydrophobicity and concentration of ion-pairing amines. Ion-pairing amines with moderate hydrophobicity were most successful in suppressing diastereomeric resolution. However, a partial separation of phosphorothioate oligonucleotide diastereomers was observed with all ion-pairing systems, which resulted in wider peaks compared to phosphorodiester oligonucleotides of the same sequence. This phenomenon complicates the n and n-1 mers separation of oligonucleotides with high degree of phosphorothioate modifications. Separation of oligonucleotides with different number of phosphorothioate modifications was observed, which may be useful for therapeutic oligonucleotide analysis. The aim of the work was to identify the ion-pairing systems useful for chromatographic characterization of phosphorothioate oligonucleotides.

Contribution of ionic interactions to stationary phase selectivity in hydrophilic interaction chromatography.

We compared the separation selectivities of 19 different hydrophilic interaction chromatography columns. The stationary phases included underivatized silica and hybrid particles, cyano-bonded silica, materials with neutral ligands such as amide, diol, pentahydroxy, and urea, zwitterionic sorbents, and mixed-mode materials with amine functionalities. A set of 77 small molecules was used to evaluate the columns. We visualized the retention behavior of the different columns using retention time correlation plots. The analytes were classified as cations, anions, or neutral based on their estimated charge under the separation conditions. This involved adjusting the dissociation constants of the analytes for the acetonitrile content of the mobile phase and experimentally determining the pH of the mobile phase containing 70% acetonitrile. The retention correlation plots show that the selectivity differences strongly depended on ionic interactions. Comparisons of the neutral stationary phases (e.g., diol vs. amide) showed more similar selectivity than did comparisons of neutral columns versus columns with cation or anion exchange activity (bare silica or amine columns, respectively). The zwitterionic columns did not behave as perfectly neutral. The correlation plots indicated that they exhibited either cation or anion exchange activity, although to a lesser degree than the silica and amine-containing stationary phases.

Effect of ion-pairing reagent hydrophobicity on liquid chromatography and mass spectrometry analysis of oligonucleotides.

We performed a systematic study of thirteen alkylamines used as ion-pairing reagents for ion-pair reversed-phase liquid chromatography (IP RP LC) separations of oligonucleotides on a C18 column. We proposed a method to classify the hydrophobicity of alkylamines by their retention in RP LC. The IP reagent hydrophobicity correlated with the retention and resolution of oligonucleotides in the corresponding IP mobile phases. The baseline resolution was achieved up to 30 mer for hydrophilic, or up to 50 mer for hydrophobic IP reagents. Hydrophobic alkylamines permitted useful oligonucleotide separations at relatively low buffer concentrations, such as 5-10 mM alkylamine-acetate IP systems. These buffers were compatible with mass spectrometry detection, however, replacement of acetic acid with hexafluoroisopropanol in the mobile phase improved the MS signal by 2-3 orders of magnitude. Experiments with native and chemically modified oligonucleotides highlighted the mixed-mode nature of IP RP LC. When using hydrophobic IP reagents, the ionic retention mechanism of oligonucleotides is enhanced while hydrophobic retention is diminished.

Impact of Nonspecific Adsorption to Metal Surfaces in Ion Pair-RP LC-MS Impurity Analysis of Oligonucleotides.

Nonspecific adsorption has been a consistent challenge in the analysis of oligonucleotides. Nonspecific adsorption is a result of interactions between charged acidic analytes and adsorption sites present in metallic surfaces located in the fluidic path of chromatography systems. Due to their high surface area, adsorption to column frits is especially concerning. Poor peak shape, low recovery and compromised LOQ have been associated with this phenomenon. Alternative methods including substitution of stainless steel for different hardware materials and mobile phase additives have been explored in an attempt to minimize this issue. Chemical modification of metal surfaces using hybrid surface technology (HST) by-passes the limitation of stainless steel construction material by forming a hybrid organic/inorganic layer that acts as a barrier and limits nonspecific interactions. In this study we explore the implications of this new technology in sensitive analysis and determination of relative impurity levels of oligonucleotides. Higher relative impurity levels and better reproducibility were obtained with columns using HST.

Assessing the impact of nonspecific binding on oligonucleotide bioanalysis.

Aim: Accurate and reliable quantification of oligonucleotides can be difficult, which has led to an increased focus on bioanalytical methods for more robust analyses. Recent advances toward mitigating sample losses on liquid chromatography (LC) systems have produced recovery advantages for oligonucleotide separations. Results & methodology: LC instruments and columns constructed from MP35N metal alloy and stainless steel columns were compared against LC hardware modified with hybrid inorganic-organic silica surfaces. Designed to minimize metal-analyte adsorption, these surfaces demonstrated a 73% increase in 25-mer phosphorothioate oligonucleotide recovery using ion-pairing reversed-phase LC versus standard LC surfaces, most particularly upon initial use. Conclusion: Hybrid silica chromatographic surfaces improve the performance, detection limits and reproducibility of oligonucleotide bioanalytical assays.

Mitigation of analyte loss on metal surfaces in liquid chromatography.

The adsorptive loss of acidic analytes in liquid chromatography was investigated using metal frits. Repetitive injections of acidic small molecules or an oligonucleotide were made on individual 2.1 or 4.6 mm i.d. column frits. Losses were observed for adenosine 5'-(α,ß-methylene) diphosphate, 2-pyridinol 1-oxide and the 25-mer phosphorothioate oligonucleotide Trecovirsen (GEM91) on stainless steel and titanium frits. Analyte adsorption was greatest at acidic pH due to the positive charge on the metal oxide surface. Analyte recovery increased when a series of injections was performed; this effect is known as sample conditioning. Nearly complete recovery was achieved when the metal adsorptive sites were saturated with the analyte. A similar effect was achieved by conditioning the frits with phosphoric, citric or etidronic acids, or their buffered solutions. These procedures can be utilized to mitigate analyte loss. However, the effect is temporary, as the conditioning agent is gradually removed by the running mobile phase. Metal frits modified with hybrid organic/inorganic surface technology were shown to mitigate analyte-to-metal surface interactions and improve recovery of acidic analytes. Quantitative recovery of a 15-35 mer oligodeoxythymidine mixture was achieved using column hardware modified with hybrid surface technology, without a need for column conditioning prior to analysis.

Characterization and comparison of mixed-mode and reversed-phase columns; interaction abilities and applicability for peptide separation.

In this work, two mixed-mode columns from a different manufacturers and one marketed as a reversed-phase column were characterized and compared in the terms of their interaction abilities, retentivity, peak symmetry, and applicability for peptide separation. All the tested columns contain octadecyl ligand and positively charged modifier, i.e. pyridyl group for the reversed-phase column XSelect CSH C18, quaternary alkylamine for mixed-mode column Atlantis PREMIER BEH C18 AX, and permanently charged moiety (details not available from the manufacturer) for mixed-mode column Luna Omega PS C18. For detailed characterization and comparison of their interaction potential, several approaches were used. First, a simple Walters test was performed to estimate hydrophobic and silanophilic interactions of the tested columns. The highest values of both parameters were observed for column Atlantis PREMIER BEH C18 AX. To investigate the effect of pH and buffer concentration on retention, mobile phases composed of acetonitrile and buffer (ammonium formate, pH 3.0; ammonium acetate pH 4.7 and pH 6.9) in various concentrations (5mM; 10mM; 15mM and 20mM) were used. The analysis of permanently charged compounds was used to describe the electrostatic interaction abilities of the stationary phases. The most significant contribution of electrostatic interactions to the retention was observed for Atlantis PREMIER BEH C18 AX column in the mobile phase with buffer of pH 3.0. A set of ten dipeptides, three pentapeptides and one octapeptide was used to investigate the effects of pH and buffer concentration on retention and peak symmetry. Each of the tested columns provides the optimal peak shape under different buffer pH and concentration. The gradient separation of the 14 tested peptides was used to verify the application potential of the tested columns for peptide separation. The best separation was achieved within 4 minutes on column Atlantis PREMIER BEH C18 AX.

Using Hybrid Organic-Inorganic Surface Technology to Mitigate Analyte Interactions with Metal Surfaces in UHPLC.

Interactions of analytes with metal surfaces in high-performance liquid chromatography (HPLC) instruments and columns have been reported to cause deleterious effects ranging from peak tailing to a complete loss of the analyte signal. These effects are due to the adsorption of certain analytes on the metal oxide layer on the surface of the metal components. We have developed a novel surface modification technology and applied it to the metal components in ultra-HPLC (UHPLC) instruments and columns to mitigate these interactions. A hybrid organic-inorganic surface, based on an ethylene-bridged siloxane chemistry, was developed for use with reversed-phase and hydrophilic interaction chromatography. We have characterized the performance of UHPLC instruments and columns that incorporate this surface technology and compared the results with those obtained using their conventional counterparts. We demonstrate improved performance when using the hybrid surface technology for separations of nucleotides, a phosphopeptide, and an oligonucleotide. The hybrid surface technology was found to result in higher and more consistent analyte peak areas and improved peak shape, particularly when using low analyte mass loads and acidic mobile phases. Reduced abundances of iron adducts in the mass spectrum of a peptide were also observed when using UHPLC systems and columns that incorporate hybrid surface technology. These results suggest that this technology will be particularly beneficial in UHPLC/mass spectrometry investigations of metal-sensitive analytes.

Method for evaluation of ionic interactions in liquid chromatography.

In this work we utilized basic and acidic analytes to investigate the ionic interaction participation in retention behavior of selected reversed-phase and polar columns. The test analytes included nitrate, benzenesulfonate and trimethylphenylammonium ions. The fully aqueous mobile phase comprising 10 mM dichloroacetic acid buffered with ammonia solution to desirable pH was used for retention experiments. Developed method was utilized to study the ionic interactions of stationary phases in pH range between 2.5 and 9.0. We demonstrate that selected sorbents used for reversed-phase and hydrophilic interaction chromatography separations exhibit cation- or anion-exchange interactions. We compare the results to novel Atlantis PREMIER BEH C18 AX mixed-mode column that combines reversed-phase and anion-exchange interaction modes. We evaluated the relative retention strength of selected columns for anionic and cationic analytes.

Evaluating MISER chromatography as a tool for characterizing HILIC column equilibration.

Hydrophilic Interaction Liquid Chromatography (HILIC) is a technique for retaining polar analytes that uses polar stationary phases and acetonitrile-rich mobile phases. While this technique has several advantages over reversed-phase liquid chromatography (RPLC), one main drawback is the reported need for longer column equilibration. The reason for this is not fully understood and is a topic of current investigation. In order to better understand and reduce the equilibration needs, accurate characterization of column equilibration under varying conditions is required. The current method of characterizing HILIC column equilibration produces limited data points per test, or low time resolution, and is highly dependent on the column and probe compounds being used. There is a need for an improved method for characterizing HILIC column equilibration, especially if trends across stationary phases are to be observed. In this work, MISER, or Multiple Injections in a Single Experimental Run, is evaluated as a possible tool for characterizing HILIC column equilibration. MISER improves time resolution by allowing for replicate injections without interruption of data collection, enabling a more thorough evaluation of column equilibration compared to traditional techniques. Experimental results gathered using MISER show that equilibration of a BEH Amide column is notably shorter when equilibrating from acetonitrile to mobile phases containing higher percentages of water. Column equilibration to a 10% aqueous mobile phase was found to be approximately 5-fold faster than equilibration to a 3% aqueous mobile phase.

Utility of linear and nonlinear models for retention prediction in liquid chromatography.

Linear solvation strength model in reversed-phase liquid chromatography assumes linear relationship between ln k and Φ. In this work we show that this assumption is true only in narrow range of mobile phase strength. The ln k versus Φ relationship could be more accurately described by three-parametric non-linear model in a wide range of eluent strength. We investigated the consequences of non-linearity on retention prediction accuracy and analyte retention behavior in reversed-phase chromatography. When the ln k versus Φ is measured in narrow range of mobile phase strength (ΔΦ ~ 0.1-0.2) both linear and nonlinear models provide comparable retention prediction results. We propose that the linear trend of ln k versus Φ relationship is obtained in the range flanking the elution factor ke (value of retention factor at the column end). We calculated and plotted changes of retention factor of analytes along the column. The visualization illustrates the ranges of retention factor values participating in separation during gradient. For typical gradient slopes employed in liquid chromatography practice and small molecules the elution factor ke value is between 2 and 8. As a simplified generalization for typical gradient slopes we propose using linear ln k versus Φ trend in the k range between 1 and 30. The spreadsheet was utilized to compare the retention prediction accuracy of linear and non-linear retention models. When fitting ln k versus Φ trend in k range 1-30 the simple linear model is in good agreement with nonlinear model with retention time prediction error 0.3-4.7% (for gradient slope 0.013-0.260).

Retention loss of reversed-phase chromatographic columns using 100% aqueous mobile phases from fundamental insights to best practice.

This work deals with experimental investigations pertaining to the impact of chemical (electrolyte concentration from 0 to 100 mM, dissolved nitrogen gas from 0 to 6.7  ×  10-4 M in water; surface chemistry including hexylphenyl, polyphenyl, C30, C18, and C8; surface coverage in C18-bonded chains from 1.5 to 3.5 µmol/m2; presence of surface dopant), physical (hydrostatic pressure of water from 50 to 500 bar; temperature from 27 ∘C to 75 ∘C), and structural parameters (average pore size from 50 Å to 400 Å; pore connectivity) on the dewetting kinetics of water from the hydrophobic mesopores of particles packed in RPLC columns. The results are explained from physico-chemical viewpoints involving intrusion and extrusion Laplace pressures, advancing and receding contact angles, surface tension of water, vapor pressure of water, 3D reconstruction of the actual mesoporous structure, pore connectivity, and the hysteresis in nitrogen adsorption and desorption isotherm onto reversed-phase chromatographic materials. A model of water dewetting consistent with the observations and the physical interpretations is then proposed. Finally, the most relevant practical solutions (pressurizing the column in absence of flow, pore size enlargement, using phenyl-bonded phase, polar embedded or surface doped C18-bonded phases, reducing the C18 surface coverage, doping the silica surface, lengthening of the alkyl-bonded chains, applying low temperatures, purging and degassing the mobile phase with helium gas) are suggested in order to eliminate or at least minimize the retention loss of RPLC columns when using fully aqueous mobile phases.

Mismatch between sample diluent and eluent: Maintaining integrity of gradient peaks using in silico approaches.

The mismatch of elution strength between the sample diluent and the eluent causes undesirable peak deformations for large sample volumes in gradient liquid chromatography. The solution to that problem consists in diluting the sample solution in a weak solvent. But the minimum dilution factor has to be determined by the user given some performance objectives. In silico approaches are applied in this work to find such adequate dilution factors. Two calculations methods are proposed for the prediction of peak distortions. The first comprehensive method is based on solving numerically the mass balance equations for all the analytes and the strong solvent. An excellent agreement between the experimental and the calculated gradient chromatograms is observed (sample diluent: acetonitrile/water, 50/50, v/v; injection volume: 15 µL; linear gradient: 5%-95% acetonitrile during 3 min) for five compounds (acetanilide, coumarin, benzoin, bi-naphthol, and dibutylphthalate) injected into a 2.1 × 50 mm column packed with 1.7 µm XBridge-C18 particles. This first method happens to be highly time-consuming and impractical for common users. Experimental work and calculation times are then minimized by applying a second method based on the basics of retention and dispersion of injected pulses. Despite being less accurate than the first method, the agreement between the experimental and calculated peak width remains physically meaningful allowing the experimenter to rapidly guess the required sample dilution factor for any combination of injected volume and strong solvent concentration in the sample solution.

Kinetic mechanism of water dewetting from hydrophobic stationary phases utilized in liquid chromatography.

An experimental protocol was designed to accurately measure the dewetting kinetics of aqueous mobile phases from reversed-phase liquid chromatography (RPLC) columns. The protocol enables the determination of the losses in the wetted surface area and internal pore volume (leading to undesirable retention losses) of RPLC columns as a function of the dewetting time. It is used to evaluate the impact of the buffer/salt concentration in water (0-100 mM), nitrogen concentration dissolved in water (0-6.7 × 10-4 M), column temperature (300-358 K), and of the internal structure (pore connectivity) of the stationary phase on the dewetting kinetics of various RPLC packing materials. From a fundamental viewpoint, the experimental facts demonstrate that dewetting kinetics are not solely driven by the pore size of the stationary phase and the contact angle with water. Temperature has a major influence on dewetting kinetics as it controls the nucleation rate of isolated water vapor bubbles over the entire mesoporous network. Additionally, the internal microstructure of the stationary phase (characterized by its internal porosity, pore size distribution, and pore connectivity) influences the rate at which the water vapor bubbles grow and coalesce in the entire particle volume. From a more practical viewpoint, the retention loss of RPLC columns due to water dewetting can be eliminated or at least minimized by (1) adjusting the surface and bonding chemistries to reduce the receding contact angle, (2) elevating the column outlet pressure, (3) operating at the lowest possible temperature, (4) minimizing the pore connectivity of the stationary phase (e.g., by increasing the degree of surface functionalization from C8 to C18-bonded phases), and (5) by degassing the aqueous mobile phase from any dissolved gases.

Impact of frit dispersion on gradient performance in high-throughput liquid chromatography.

The impact of porous frits (2.1 mm i.d. × 1 mm thick, 0.2 µm porosity, 20% void, 0.69 µL) in short 0.5-5 cm long × 2.1 mm i.d. columns packed with sub-2 µm superficially porous particles on gradient performance was investigated using a low dispersive i-class ACQUITY UPLC system (25 cm × 75 µm outlet tube + 250 nL optical cell). In order to maximize data accuracy, the sample dispersion through a single frit was measured from four independent methods: (1) plate height subtraction, (2) peak capacity versus efficiency plot, (3) flow reversal, and (4) direct measurement. Frit dispersion increases non-linearly with increasing flow rate. The corresponding volume variances of small molecules were measured at 0.16 ±â€¯0.04, 0.24 ±â€¯0.05, and 0.31 ±â€¯0.06 µL2 at flow rates of 0.1, 0.2, and 0.3 mL/min, respectively. These observed variances are lower than and consistent with the maximum volume variance of 0 . 692 ∼ 0.5 µL2 expected for a mixer-like behavior. The peak capacity of short columns were then calculated for mixtures of peptides using a general model of gradient elution by considering (or not) the actual analyte dispersion taking place in the outlet frit, in the post-column connecting tube, and in the detection cell. The results show that the sole presence of the outlet frit is responsible for about 50% loss in peak capacity (relative to the expected gradient performance in absence of frit and post-column tube) for a 1 cm long column operated under standard gradient steepness. The actual structure and volume of column frits needs adjustment if one wants to take full advantage of the true performance of short high-throughput columns packed with sub-2 µm particles.

Semi-preparative high-resolution recycling liquid chromatography.

A semi-preparative high-resolution system based on twin column recycling liquid chromatography was built. The integrated system includes a binary pump mixer, a sample manager, a two-column oven compartment, two low-dispersion detection cells, and a fraction manager (analytical). It addresses challenges in drug/impurity purification, which involve several constraints simultaneously: (1) small selectivity factors (α < 1.2, poor resolution), (2) mismatch of elution strength between the sample diluent and the eluent causing severe band fronting or tailing, (3) diluent-to-eluent mismatch of viscosity causing viscous fingering and unpredictable band deformation, (4) low abundance of the impurity relative to the active pharmaceutical ingredient (API) (<1/100), and (5) yield and purity levels to be larger than 99% and 90%, respectively. The prototype system was tested for the preparation of a trace impurity present in a concentrated solution of an API, estradiol. The ultimate goal was to collect ∼1 mg of impurity (>90% purity) for unambiguous structure elucidation by liquid state nuclear magnetic resonance (NMR 600 MHz and above). First, the particle size (3.5 µm) used to pack the 4.6 mm × 150 mm long twin columns is selected so that the speed-resolution of the recycling process is maximized at 4000 psi pressure drop. Next, the production rate of the process is also maximized by determining the optimum number (7) of cycles and the corresponding largest sample volume (160 µL) to be injected. Finally, the process is fully automated by programming the time events related to (1) sample cleaning, (2) transfer of the targeted impurity from one to the second twin column, and (3) impurity collection. The process was tested without interruption during one week for the collection of a trace impurity (α = 1.166, strong acetonitrile-methanol sample diluent, concentration ∼2 mg/L) from a concentrated (10 g/L) stock solution (60 mL total) of estradiol. The process enriches the impurity content relative to the API by about a factor ∼5000. For the lack of a sufficient collected amount (∼120 µg only) of the pure impurity (purity 50% only), NMR experiments could not provide reliable results. Instead, the combination of LC-MS (single ion monitoring) and UV absorption spectra (λmax shift) revealed that the targeted impurity was likely the low-abundant enol tautomeric form of the ketone estrone, a possible intermediate or by-product of the synthesis reaction of estradiol.