Development of a unified gradient theory for ion-pair chromatography using oligonucleotide separations as a model case.

Ion-pair chromatography is the de facto standard for separating oligonucleotides and related impurities, particularly for analysis but also often for small-scale purification. Currently, there is limited understanding of the quantitative modeling of both analytical and overloaded elution profiles obtained during gradient elution in ion-pair chromatography. Here we will investigate a recently introduced gradient mode, the so-called ion-pairing reagent gradient mode, for both analytical and overloaded separations of oligonucleotides. The first part of the study demonstrates how the electrostatic theory of ion-pair chromatography can be applied for modeling gradient elution of oligonucleotides. When the ion-pair gradient mode is used in a region where the electrostatic surface potential can be linearized, a closed-form expression of retention time can be derived. A unified retention model was then derived, applicable for both ion-pair reagent gradient mode as well as co-solvent gradient mode. The model was verified for two different experimental systems and homo- and heteromeric oligonucleotides of different lengths. Quantitative modeling of overloaded chromatography using the ion-pairing reagent gradient mode was also investigated. Firstly, a unified adsorption isotherm model was developed for both gradient modes. Then, adsorption isotherms parameter of a model oligonucleotide and two major synthetic impurities were estimated using the inverse method. Secondly, the parameters of the adsorption isotherm were then used to investigate how the productivity of oligonucleotide varies with injection volume, gradient slope, and initial retention factor. Here, the productivity increased when using a shallow gradient slope combined with a low initial retention factor. Finally, experiments were conducted to confirming some of the model predictions. Comparison with the conventional co-solvent gradient mode showed that the ion-pairing reagent gradient leads to both higher yield and productivity while consuming less co-solvent.

Building machine-learning-based models for retention time and resolution predictions in ion pair chromatography of oligonucleotides.

Support vector regression models are created and used to predict the retention times of oligonucleotides separated using gradient ion-pair chromatography with high accuracy. The experimental dataset consisted of fully phosphorothioated oligonucleotides. Two models were trained and validated using two pseudo-orthogonal gradient modes and three gradient slopes. The results show that the spread in retention time differs between the two gradient modes, which indicated varying degree of sequence dependent separation. Peak widths from the experimental dataset were calculated and correlated with the guanine-cytosine content and retention time of the sequence for each gradient slope. This data was used to predict the resolution of the n - 1 impurity among 250 000 random 12- and 16-mer sequences; showing one of the investigated gradient modes has a much higher probability of exceeding a resolution of 1.5, particularly for the 16-mer sequences. Sequences having a high guanine-cytosine content and a terminal C are more likely to not reach critical resolution. The trained SVR models can both be used to identify characteristics of different separation methods and to assist in the choice of method conditions, i.e. to optimize resolution for arbitrary sequences. The methodology presented in this study can be expected to be applicable to predict retention times of other oligonucleotide synthesis and degradation impurities if provided enough training data.

Selectivity limits of and opportunities for ion pair chromatographic separation of oligonucleotides.

Here it was investigated how oligonucleotide retention and selectivity factors are affected by electrostatic and non-electrostatic interactions in ion pair chromatography. A framework was derived describing how selectivity depends on the electrostatic potential generated by the ion-pair reagent concentration, co-solvent volume fraction, charge difference between the analytes, and temperature. Isocratic experiments verified that, in separation problems concerning oligonucleotides of different charges, selectivity increases with increasing surface potential and analyte charge difference and with decreasing co-solvent volume fraction and temperature. For analytes of the same charge, for example, diastereomers of phosphorothioated oligonucleotides, selectivity can be increased by decreasing the co-solvent volume fraction or the temperature and has only a minor dependency on the ion-pairing reagent concentration. An important observation is that oligonucleotide retention is driven predominantly by electrostatic interaction generated by the adsorption of the ion-pairing reagent. We therefore compared classical gradient elution in which the co-solvent volume fraction increases over time versus gradient elution with a constant co-solvent volume fraction but with decreasing ion-pair reagent concentration over time. Both modes decrease the electrostatic potential. Oligonucleotide selectivity was found to increase with decreasing ion-pairing reagent concentration. The two elution modes were finally applied to two different model antisense oligonucleotide separation problems, and it was shown that the ion-pair reagent gradient increases the selectivity of non-charge-based separation problems while maintaining charge-difference-based selectivity.

A Retention-Matching Strategy for Method Transfer in Supercritical Fluid Chromatography: Introducing the Isomolar Plot Approach.

A strategy to match any retention shifts due to increased or decreased pressure drop during supercritical fluid chromatography (SFC) method transfer is presented. The strategy relies on adjusting the co-solvent molarity without the need to adjust the back-pressure regulator. Exact matching can be obtained with minimal changes in separation selectivity. To accomplish this, we introduce the isomolar plot approach, which shows the variation in molar co-solvent concentration depending on the mass fraction of co-solvent, pressure, and temperature, here exemplified by CO2-methanol. This plot allowed us to unify the effects of the co-solvent mass fraction and density on retention in SFC. The approach, which was verified on 12 known empirical retention models for each enantiomer of six basic pharmaceuticals, allowed us to numerically calculate the apparent retention factor for any column pressure drop. The strategy can be implemented either using a mechanistic approach if retention models are known or empirically by iteratively adjusting the co-solvent mass fraction. As a rule of thumb for the empirical approach, we found that the relative mass fraction adjustment needed is proportional to the relative change in the retention factor caused by a change in the pressure drop. Different proportionality constants were required to match retention in the case of increasing or decreasing pressure drops.

Impact of stationary-phase pore size on chromatographic performance using oligonucleotide separation as a model.

A combined experimental and theoretical study was performed to understand how the pore size of packing materials with pores 60-300 Å in size affects the separation of 5-50-mer oligonucleotides. For this purpose, we developed a model in which the solutes were described as thin rods to estimate the accessible surface area of the solute as a function of the pore size and solute size. First, an analytical investigation was conducted in which we found that the selectivity increased by a factor of 2.5 when separating 5- and 15-mer oligonucleotides using packing with 300 Å rather than 100 Å pores. We complemented the analytical investigation by theoretically demonstrating how the selectivity is dependent on the column's accessible surface area as a function of solute size. In the preparative investigation, we determined adsorption isotherms for oligonucleotides using the inverse method for separations of a 9- and a 10-mer. We found that preparative columns with a 60 Å-pore-size packing material provided a 10% increase in productivity as compared with a 300 Å packing material, although the surface area of the 60 Å packing is as much as five time larger.

Analytical and preparative separation of phosphorothioated oligonucleotides: columns and ion-pair reagents.

Oligonucleotide drugs represent an emerging area in the pharmaceutical industry. Solid-phase synthesis generates many structurally closely related impurities, making efficient separation systems for purification and analysis a key challenge during pharmaceutical drug development. To increase the fundamental understanding of the important preparative separation step, mass-overloaded injections of a fully phosphorothioated 16mer, i.e., deoxythymidine oligonucleotide, were performed on a C18 and a phenyl column. The narrowest elution profiles were obtained using the phenyl column, and the 16mer could be collected with high purity and yield on both columns. The most likely contribution to the successful purification was the quantifiable displacement of the early-eluting shortmers on both columns. In addition, the phenyl column displayed better separation of later-eluting impurities, such as the 17mer impurity. The mass-overloaded injections resulted in classical Langmuirian elution profiles on all columns, provided the concentration of the ion-pairing reagent in the eluent was sufficiently high. Two additional column chemistries, C4 and C8, were also investigated in terms of their selectivity and elution profile characteristics for the separation of 5-20mers fully phosphorothioated deoxythymidine oligonucleotides. When using triethylamine as ion-pairing reagent to separate phosphorothioated oligonucleotides, we observed peak broadening caused by the partial separation of diastereomers, predominantly seen on the C4 and C18 columns. When using the ion-pair reagent tributylamine, to suppress diastereomer separation, the greatest selectivity was found using the phenyl column followed by C18. The present results will be useful when designing and optimizing efficient preparative separations of synthetic oligonucleotides.

Investigation of factors influencing the separation of diastereomers of phosphorothioated oligonucleotides.

This study presents a systematic investigation of factors influencing the chromatographic separation of diastereomers of phosphorothioated pentameric oligonucleotides as model solutes. Separation was carried out under ion-pairing conditions using an XBridge C18 column. For oligonucleotides with a single sulfur substitution, the diastereomer selectivity was found to increase with decreasing carbon chain length of the tertiary alkylamine used as an ion-pair reagent. Using an ion-pair reagent with high selectivity for diastereomers, triethylammonium, it was found the selectivity increased with decreased ion-pair concentration and shallower gradient slope. Selectivity was also demonstrated to be dependent on the position of the modified linkage. Substitutions at the center of the pentamer resulted in higher diastereomer selectivity compared to substitutions at either end. For mono-substituted oligonucleotides, the retention order and stereo configuration were consistently found to be correlated, with Rp followed by Sp, regardless of which linkage was modified. The type of nucleobase greatly affects the observed selectivity. A pentamer of cytosine has about twice the diastereomer selectivity of that of thymine. When investigating the retention of various oligonucleotides eluted using tributylammonium as the ion-pairing reagent, no diastereomer selectivity could be observed. However, retention was found to be dependent on both the degree and position of sulfur substitution as well as on the nucleobase. When analyzing fractions collected in the front and tail of overloaded injections, a significant difference was found in the ratio between Rp and Sp diastereomers, indicating that the peak broadening observed when using tributylammonium could be explained by partial diastereomer separation.

Investigation of robustness for supercritical fluid chromatography separation of peptides: Isocratic vs gradient mode.

We investigated and compared the robustness of supercritical fluid chromatography (SFC) separations of the peptide gramicidin, using either isocratic or gradient elution. This was done using design of experiments in a design space of co-solvent fraction, water mass fraction in co-solvent, pressure, and temperature. The density of the eluent (CO2-MeOH-H2O) was experimentally determined using a Coriolis mass flow meter to calculate the volumetric flow rate required by the design. For both retention models, the most important factor was the total co-solvent fraction and water mass fraction in co-solvent. Comparing the elution modes, we found that gradient elution was more than three times more robust than isocratic elution. We also observed a relationship between the sensitivity to changes and the gradient steepness and used this to draw general conclusions beyond the studied experimental system. To test the robustness in a practical context, both the isocratic and gradient separations were transferred to another laboratory. The gradient elution was highly reproducible between laboratories, whereas the isocratic system was not. Using measurements of the actual operational conditions (not the set system conditions), the isocratic deviation was quantitatively explained using the retention model. The findings indicate the benefits of using gradient elution in SFC as well as the importance of measuring the actual operational conditions to be able to explain observed differences between laboratories when conducting method transfer.

Optimizing Column Length and Particle Size in Preparative Batch Chromatography Using Enantiomeric Separations of Omeprazole and Etiracetam as Models: Feasibility of Taguchi Empirical Optimization.

The overreaching purpose of this study is to evaluate new approaches for determining the optimal operational and column conditions in chromatography laboratories, i.e., how best to select a packing material of proper particle size and how to determine the proper length of the column bed after selecting particle size. As model compounds, we chose two chiral drugs for preparative separation: omeprazole and etiracetam. In each case, two maximum allowed pressure drops were assumed: 80 and 200 bar. The processes were numerically optimized (mechanistic modeling) with a general rate model using a global optimization method. The numerical predictions were experimentally verified at both analytical and pilot scales. The lower allowed pressure drop represents the use of standard equipment, while the higher allowed drop represents more modern equipment. For both compounds, maximum productivity was achieved using short columns packed with small-particle size packing materials. Increasing the allowed backpressure in the separation leads to an increased productivity and reduced solvent consumption. As advanced numerical calculations might not be available in the laboratory, we also investigated a statistically based approach, i.e., the Taguchi method (empirical modeling), for finding the optimal decision variables and compared it with advanced mechanistic modeling. The Taguchi method predicted that shorter columns packed with smaller particles would be preferred over longer columns packed with larger particles. We conclude that the simpler optimization tool, i.e., the Taguchi method, can be used to obtain "good enough" preparative separations, though for accurate processes, optimization, and to determine optimal operational conditions, classical numerical optimization is still necessary.

Chemometric evaluation of the combined effect of temperature, pressure, and co-solvent fractions on the chiral separation of basic pharmaceuticals using actual vs set operational conditions.

The need to determine the actual operational conditions, instead of merely using the set operational conditions, was investigated for in packed supercritical fluid chromatography (SFC) by design of experiments (DoE) using a most important type of compounds, pharmaceutical basics, as models. The actual values of temperature, pressure, and methanol levels were recorded and calculated from external sensors, while the responses in the DoE were the retention factors and selectivity. A Kromasil CelluCoat column was used as the stationary phase, carbon dioxide containing varying methanol contents as the mobile phase, and the six racemates of alprenolol, atenolol, metoprolol, propranolol, clenbuterol, and mianserin were selected as model solutes. For the retention modeling, the most important term was the methanol fraction followed by the temperature and pressure. Significant differences (p<0.05) between most of the coefficients in the retention models were observed when comparing models from set and actual conditions. The selectivity was much less affected by operational changes, and therefore was not severely affected by difference between set and actual conditions. The temperature differences were usually small, maximum ±1.4°C, whereas the pressure differences were larger, typically approximately +10.5bar. The set and actual fractions of methanol also differed, usually by ±0.4 percentage points. A cautious conclusion is that the primary reason for the discrepancy between the models is a mismatch between the set and actual methanol fractions. This mismatch is more serious in retention models at low methanol fractions. The study demonstrates that the actual conditions should almost always be preferred.

Evaluation of scale-up from analytical to preparative supercritical fluid chromatography.

An approach for reliable transfer from analytical to preparative scale supercritical fluid chromatography was evaluated. Here, we accounted for the conditions inside the columns as well as to the fact that most analytical instruments are volume-controlled while most preparative scale units are mass-controlled. The latter is a particular problem when performing pilot scale experiments and optimizations prior to scaling up to production scale. This was solved by measuring the mass flow, the pressure and the temperature on the analytical unit using external sensors. Thereafter, it was revealed with a design of experiments approach that the methanol fraction and the pressure are the two most important parameters to control for preserved retention throughout the scale-up; for preserved selectivity the temperature was most important in this particular system. Using this approach, the resulting chromatograms from the preparative unit agreed well with those from the analytical unit while keeping the same column length and particles size. A brief investigation on how the solute elution volume varies with the volumetric flow rate revealed a complex dependency on pressure, density and apparent methanol content. Since the methanol content is a parameter of great importance to control during the scale up, we must be careful when changing operational and column design conditions which generates deviations in pressure, density and methanol content between different columns.

Choice of Model for Estimation of Adsorption Isotherm Parameters in Gradient Elution Preparative Liquid Chromatography.

The inverse method is a numerical method for fast estimation of adsorption isotherm parameters directly from a few overloaded elution profiles and it was recently extended to adsorption isotherm acquisition in gradient elution conditions. However, the inverse method in gradient elution is cumbersome due to the complex adsorption isotherm models found in gradient elution. In this case, physicochemically correct adsorption models have very long calculation times. The aim of this study is to investigate the possibility of using a less complex adsorption isotherm model, with fewer adjustable parameters, but with preserved/acceptable predictive abilities. We found that equal or better agreement between experimental and predicted elution profiles could be achieved with less complex models. By being able to select a model with fewer adjustable parameters, the calculation times can be reduced by at least a factor of 10.

A closer study of peak distortions in supercritical fluid chromatography as generated by the injection.

In SFC the sample cannot be dissolved in the mobile phase, so it is often dissolved in pure modifier, or another liquid, sometimes resulting in serious distortions of the eluted peak profiles already at moderately high injection volumes. It is suspected the reasons for these effects are solvent strength mismatch and/or viscosity mismatch. This study presents a systematic and fundamental investigation of the origin of these peak deformations due to the injection solvent effects in SFC, using both systematic experiments and numerical modeling. The first set of experiments proved that the injection volume and the elution strength of the sample solution had a major impact of the shapes of the eluted peaks. Secondly, the sample band elution profile was numerically modeled on a theoretical basis assuming both un-retained and retained co-solvent injection plugs, respectively. These calculations quantitatively confirmed our first set of experiments but also pointed out that there is also an additional significant effect. Third, viscous fingering experiments were performed using viscosity contrast conditions imitating those encountered in SFC. These experiments clearly proved that viscous fingering effects play a significant role. A new method for determination of adsorption isotherms of solvents was also developed, called the "Retention Time Peak Method" (RTPM). The RTPM was used for fast estimation of the adsorption isotherms of the modifier and required using only two experiments.

Evaluation of co-solvent fraction, pressure and temperature effects in analytical and preparative supercritical fluid chromatography.

A chemometric approach is used for studying the combined effect of temperature, pressure and co-solvent fraction in analytical and preparative supercritical fluid chromatography (SFC). More specifically, by utilizing design of experiments coupled with careful measurements of the experimental conditions the interaction between pressure, temperature and co-solvent fraction was studied with respect to productivity, selectivity and retention in chiral SFC. A tris-(3,5-dimethylphenyl) carbamoyl cellulose stationary phase with carbon dioxide/methanol as mobile phase and the two racemic analytes trans-stilbene oxide (TSO) and 1,1'-bi-2-naphthol (BINOL) were investigated. It was found for the investigated model system that the co-solvent fraction and pressure were the parameters that most affected the retention factors and that the co-solvent fraction and column temperature were most important for controlling the selectivity. The productivity in the preparative mode of SFC was most influenced by the co-solvent fraction and temperature. Both high co-solvent fraction and temperature gave maximum productivity in the studied design space.

Investigation of plateau methods for adsorption isotherm determination in supercritical fluid chromatography.

The Perturbation Peak (PP) method and Frontal analysis (FA) are considered as the most accurate methods for adsorption isotherms determination in liquid chromatography. In this study we investigate and explain why this is not the case in Supercritical Fluid Chromatography (SFC), where the PP method does not work at all, using a modern analytical system. The main reason was found to be that the solute to be studied must be dissolved in the MeOH reservoir before it is mixed with CO2. Since the solute occupies a certain partial volume in the reservoir, the larger the solute content the larger this fractional volume will be, and the final MeOH fraction in the mobile phase will then be smaller compared to the bulk mobile phase without solute in the modifier. If the retention of small injections on the concentration plateaus, i.e., "analytical-size" perturbation peaks, is sensitive to small variations of MeOH in the eluent, this will seriously decrease the accuracy of the PP method. This effect was verified and compensated for and we also demonstrated that the same problem will occur in frontal analysis, another concentration plateau method.

Fast estimation of adsorption isotherm parameters in gradient elution preparative liquid chromatography II: the competitive case.

Experimental competitive adsorption isotherms were successfully determined directly from overloaded elution profiles in gradient elution mode using an extended inverse method. This approach differs from the existing methods in one important aspect - no isocratic experiments are necessary which makes it possible to study adsorption of substances whose retention factors vary strongly with the mobile-phase composition. The approach was verified with simulated binary data and with experimental data from gradient separations of a cyclohexanone/cycloheptanone mixture. For the synthetic data, the original adsorption isotherm parameters were found using a two-step estimation procedure. In the first step analytical peaks were used to estimate the "analytical" part of the Langmuir equation and in the second step the association equilibrium parameters were estimated from two simulated overloaded elution profiles. For the experimental data, a three-step approach was used. The two first steps were used to reduce the calculation time so that parameter estimation could be performed on an ordinary computer. In the first step, analytical peaks were used to estimate the "analytical" part of the bi-Langmuir equation. In the second step, initial guesses for all other parameters were determined separately for each solute using the faster Rouchon algorithm. In the final and third step, the more accurate orthogonal collocation on finite elements algorithm, was used to fine-tune the isotherm parameters. The model could accurately predict the shape of overloaded elution profiles. The shape of the adsorption isotherms agreed well with those determined with the standard isocratic method, although the numerical values were not the same. The extended inverse method is well suited for process optimization where few experiments and accurate predictions are important.

Determination of adsorption isotherms in supercritical fluid chromatography.

In this study we will demonstrate the potential of modern integrated commercial analytical SFC-systems for rapid and reliable acquisition of thermodynamic data. This will be done by transferring the following adsorption isotherm determination methods from liquid chromatography (LC) to supercritical fluid chromatography (SFC): Elution by Characteristic Points (ECP), the Retention Time Method (RTM), the Inverse Method (IM) and the Perturbation Peak (PP) method. In order to transfer these methods to SFC in a reliable, reproducible way we will demonstrate that careful system verification using external sensors of mass flow, temperature and pressure are needed first. The adsorption isotherm data generated by the different methods were analyzed and compared and the adsorption isotherms ability to predict new experimental elution profiles was verified by comparing experiments with simulations. It was found that adsorption isotherm data determined based on elution profiles, i.e., ECP, IM and RTM, were able to accurately predict overloaded experimental elution profiles while the more tedious and time-consuming PP method, based on small injections on concentration plateaus, failed in doing so.

Fast estimation of adsorption isotherm parameters in gradient elution preparative liquid chromatography. I: the single component case.

The inverse method is a numeric method for fast estimation of adsorption isotherm parameters directly from overloaded elution profiles. However, it has previously only been used for isocratic experiments. Here we will extend the inverse method so it can be used for gradient elution too. This extended inverse method will make it possible to study the adsorption of substances whose retention factor vary strongly with the mobile-phase composition, like peptides and proteins, where the classic methods will fail. Our extended inverse method was verified using both simulations and real experiments. For simulated overloaded elution profiles we were able to determine almost exact Langmuir adsorption isotherm parameters with the new approach. From real experimental data, bi-Langmuir adsorption parameters were estimated using both the perturbation peak method and the extended inverse method. The shape of the acquired adsorption isotherms did match over the considered concentration range; however, the adsorption isotherm parameters found with the two methods were not the same. This is probably due to the fact that adsorption isotherm estimated with the inverse method is only a good approximation up to the highest eluted concentration in the used chromatograms. But this is not a serious drawback from a process point of view where the main objective is to make accurate predictions of elution profiles. The bi-Langmuir adsorption isotherm obtained with both methods could accurately predict the shape of overloaded elution profiles.

Enantioseparation of omeprazole--effect of different packing particle size on productivity.

Enantiomeric separation of omeprazole has been extensively studied regarding both product analysis and preparation using several different chiral stationary phases. In this study, the preparative chiral separation of omeprazole is optimized for productivity using three different columns packed with amylose tris (3,5-dimethyl phenyl carbamate) coated macroporous silica (5, 10 and 25 µm) with a maximum allowed pressure drop ranging from 50 to 400 bar. This pressure range both covers low pressure process systems (50-100 bar) and investigates the potential for allowing higher pressure limits in preparative applications in a future. The process optimization clearly show that the larger 25 µm packing material show higher productivity at low pressure drops whereas with increasing pressure drops the smaller packing materials have substantially higher productivity. Interestingly, at all pressure drops, the smaller packing material result in lower solvent consumption (L solvent/kg product); the higher the accepted pressure drop, the larger the gain in reduced solvent consumption. The experimental adsorption isotherms were not identical for the different packing material sizes; therefore all calculations were recalculated and reevaluated assuming identical adsorption isotherms (with the 10 µm isotherm as reference) which confirmed the trends regarding productivity and solvent consumption.

Characterization of an unusual adsorption behavior of racemic methyl-mandelate on a tris-(3,5-dimethylphenyl) carbamoyl cellulose chiral stationary phase.

An interesting adsorption behavior of racemic methyl mandelate on a tris-(3,5-dimethylphenyl)carbamoyl cellulose chiral stationary phase was theoretically and experimentally investigated. The overloaded band of the more retained enantiomer had a peculiar shape indicating a type V adsorption isotherm whereas the overloaded band of the less retained enantiomer had a normal shape indicating a type I adsorption behavior. For a closer characterization of this separation, adsorption isotherms were determined and analyzed using an approach were Scatchard plots and adsorption energy distribution (AED) calculations are combined for a deeper analysis. It was found that the less retained enantiomer was best described by a Tóth adsorption isotherm while the second one was best described with a bi-Moreau adsorption isotherm. The latter model comprises non-ideal adsorbate-adsorbate interactions, providing an explanation to the non-ideal adsorption of the more retained enantiomer. Furthermore, the possibility of using the Moreau model as a local model for adsorption in AED calculations was evaluated using synthetically generated raw adsorption slope data. It was found that the AED accurately could predict the number of adsorption sites for the generated data. The adsorption behavior of both enantiomers was also studied at several different temperatures and found to be exothermic; i.e. the adsorbate-adsorbate interaction strength decreases with increasing temperature. Stochastic analysis of the adsorption process revealed that the average amount of adsorption/desorption events increases and the sojourn time decreases with increasing temperature.