Computer-assisted chiral HPLC method development for ozanimod on polysaccharide- based chiral stationary phases: possibilities and limitations

doi:10.21203/rs.3.rs-4523941/v1

Download PDF

Article

Computer-assisted chiral HPLC method development for ozanimod on polysaccharide- based chiral stationary phases: possibilities and limitations

https://doi.org/10.21203/rs.3.rs-4523941/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

In this study, a direct high performance liquid chromatography method was developed to determine the enantiomeric purity of the immunomodulatory drug, ozanimod. A systematic method development process was followed, incorporating risk assessment, identification of critical analytical procedure parameters (APP), initial screening of stationary phases, and software-assisted optimization of method parameters. Eight different polysaccharide-based chiral columns (Lux i-Amylose-1, Lux Amylose-2, Chiralpak AD, Chiralcel OD, Lux Cellulose-1, Lux Cellulose-2, Lux Cellulose-3, and Lux Cellulose-4) were selected to assess chiral separation of enantiomers under polar organic elution mode. The most promising results were obtained using a methanol (MeOH)-2-propanol (IPA) mixture on the Chiralpak AD column.

Following this, systematic modeling was conducted using DryLab software to optimize method conditions, including isocratic eluent composition (%IPA in MeOH), temperature, and flow rate. Baseline separation was achieved within fifteen minutes using the optimized parameters: Chiralpak AD column thermostated at 10°C, and a mobile phase of MeOH:IPA:diethylamine (DEA), 70:30:0.1 (v/v/v), delivered at a flow rate of 0.8 mL/min. The developed method was validated according to the International Council on Harmonization guideline Q2(R2) for chiral impurity determination in ozanimod samples. Additionally, in silico robustness testing was conducted to determine tolerance limits for critical separation parameters and their impact on enantioresolution. Our findings demonstrate the utility of DryLab, typically employed for reversed-phase achiral separations, in optimizing chiral methods even in polar organic mode. The software's limitations for this purpose are also discussed.

Physical sciences/Chemistry/Physical chemistry/Chemical physics

Physical sciences/Chemistry/Analytical chemistry

Physical sciences/Chemistry/Supramolecular chemistry

ozanimod

Zeposia

chiral separation

polysaccharide-type chiral column

QbD

DryLab

computer-assisted method development

Ozanimod emerges in the clinical landscape as one of the promising candidates for the treatment of multiple sclerosis [1, 2]. Offering an oral administration route, it represents a more patient-friendly alternative when juxtaposed with conventional injectable therapies. The unique capability to modulate the immune response positions makes this substance extremely valuable for therapeutic uses, in particular the management of relapsing forms of multiple sclerosis [3]. From the chemical perspective, ozanimod can also be identified as a sphingosine analogue, meticulously crafted to selectively target S1P receptors. This engineered structure allows effective interactions with the receptor system, eliciting the desired pharmacological responses, while minimizing off-target effects [2, 4].

Ozanimod contains a singular chiral carbon atom, implying the existence of two enantiomeric forms, namely S-enantiomer (ozanimod) and R-ozanimod (Fig. 1).

In therapeutic applications, the more effective S-enantiomer is employed. Consequently, regulatory authorities mandate the determination of the R-antipode in ozanimod sample as part of the necessary assessments. Despite this, current literature reveals a missing context concerning the chiral separation studies of ozanimod. However, there are few articles and some patents available regarding the enantioselective synthesis of the drug [5–8]. This discrepancy highlights a research gap in the field, emphasizing the need for investigations into chiral separation techniques for ozanimod, as it could provide valuable insights into the stereochemical aspects of the drug, facilitating its characterization and ensuring accurate determination of enantiomeric purity, which is crucial for therapeutic efficacy and regulatory compliance.

Today, a variety of methods are utilized for chiral analysis, including capillary electrophoresis (CE), gas chromatography (GC), high performance liquid chromatography (HPLC), and supercritical fluid chromatography (SFC). Among them, direct HPLC enantioseparation in combination with a chiral stationary phase (CSP) is the gold standard in the field [9, 10]. The most popular and effective CSPs are commonly based on polysaccharides. The polysaccharide backbone, whether amylose or cellulose, introduces conformational chirality established with specific glycosidic linkages. Structurally diverse pendant groups enable radial expansion of the polymer and the formation of chiral grooves. These grooves comprise a polar layer with carbamate groups for polar interactions, and a hydrophobic layer housing aromatic ring for π-π and hydrophobic interactions, enhancing the chiral recognition [11–13]. Additionally, these columns also offer versatility, as they can be employed in normal-phase (NP), reversed-phase (RP), and polar organic (PO) modes [14]. Despite being toxic and environmentally unfriendly, many official compendial enantioselective methods still up to this day use NP mode. However, numerous studies have shown that polar organic mode, using simple mobile phase compositions, such as neat alcohols or acetonitrile (ACN), can be equally effective or even surpass NP, offering several advantages, such as environmental friendliness, shorter run times and improved peak shapes. Not only neat eluents but mixtures of methanol (MeOH) and 2-propanol (IPA) or ACN and alcohols are commonly used in polar organic mode to alter peak shape, selectivity, and enantiomer elution order (EEO) [15–17]. One of the interesting effects of PO mode is the hysteresis of retention times and sometimes even EEO changes. This was first observed and described by Németh and Horváth in 2018 on the amylose tris(3,5-dimethylphenylcarbamate) stationary phase and highlights the importance of careful selection of eluent composition and all-around interpretation of chromatographic results [18–20].

It is crucial to emphasize that the development of chiral separations often relies on conventional methods. These methods typically involve identification of suitable chiral selectors, followed by a single optimization step of method parameters through trial and error or the one-factor-at-a-time (OFAT) approach. It is evident that solely depending on such traditional development strategies can result in prolonged processes and suboptimal outcomes. In contrast, multivariate methodologies have demonstrated their effectiveness in overcoming these limitations. By employing experimental designs, they allow for the simultaneous evaluation of critical method parameters while monitoring their impact on selected performance indicators. These comprehensive approaches promote a deeper understanding with fewer experimental iterations, facilitating the identification of relationships between variables [21, 22].

Hence, it is not surprising that current regulatory guidelines such as International Council of Harmonization (ICH) Q14[23] and the United States Pharmacopeia (USP) General Chapter < 1220>[24] also foster the integration of Analytical Quality by Design (AQbD) principles within the life cycle framework. This integration aims to enhance method robustness throughout the method’s intended use. One significant advantage of AQbD lies in its capacity to transition from a rigid workpoint to more flexible methodologies through the incorporation of defined tolerances within a multidimensional Design Space. This integration enhances decision-making processes rooted in risk and knowledge, thereby helping to prevent exhaustive investigations into out-of-specification occurrences.

Currently, several chromatography modeling software have been made commercially available. Three main groups of chromatography modeling software are available, based on different operational principles. The software of the first group uses a pure empirical “black-box” approach to uncover patterns and relationships of a set of experimental — Design of Experiments — data. To this group belong tools like Fusion QbD, JMP, MODDE, Statistica and others. The second group uses molecule structures along with physicochemical properties and specific molecule descriptors sometimes combined with experimental data to predict expected chromatographic behavior of compounds. Examples are ChemSpider, ChromSword, EluEx, PALLAS and others. The third class uses a mechanistic design to connect experimental data points, i.e., one or more chromatographic variables with modeling responses, typically critical resolution (software packages like DryLab, ACD Labs, and others).

While molecule-structure based modeling solutions may seem appealing, predicted separations often exhibit significant errors compared to experimentally observed separations. Additionally, these solutions may prove inadequate, particularly when dealing with structural isomers, where identical molecular formulas result in identical retention times [25]. Presently, statistical-based development predominates in chiral chromatography. Although it is recognized for its efficacy, it is often deemed less efficient in terms of resource utilization and inattentive use of any “black-box” tool may result in oversight of some chromatographically relevant parameters. As stated in USP < 1220> “When available, mechanistic models can be used to understand the effect of procedure parameters on performance. The use of mechanistic models can reduce experimental work and provide a reliable estimate of the behavior of the analytes of interest. “ [24]

Bearing this in mind, we selected one of the most established, chromatography-based modeling tools (DryLab) to investigate modeling benefits regarding time-, and performance-improvements of chiral separations in PO mode. It is worth mentioning that this tool was originally designed for modeling RP-systems but has demonstrated versatility over its 38-year history [26, 27], spanning various chromatographic applications including separation of small, and large molecules in various modes of chromatography such as ion-pairing reversed phase (IP-RP) [28], ion-exchange (IEX) [29, 30], hydrophobic interaction (HIC)[31] and hydrophilic interaction chromatography (HILIC) [32], and most recently SFC [33]. DryLab has demonstrated success in optimizing the enantioseparation of racemic 4-dinitrophenyl amino acids using quinine carbamate-type chiral stationary phase with water-based mobile phase[34] and atropoisomers[35] on a derivatized β-cyclodextrin bonded stationary phase using reversed-phase mode. In a previous study, we discussed how a general modeling approach can be applied to separate achiral analytes with closely similar structures in reversed-phase mode using chiral stationary phases [36]. However, it is important to acknowledge that in such cases, the prediction accuracy of virtual separation models, particularly for polysaccharide CSPs, may be somewhat lower. For instance, a few years ago, Wagdy successfully tested DryLab's applicability on several CSPs containing immobilized antibiotic groups like teicoplanin and vancomycin. However, modeling on a polysaccharide phase (amylose carbamate) was proven unfeasible [37]. Nevertheless, it can remain an effective modeling tool in refining method selectivity even in those scenarios.

Having all this said, our study aimed to revisit modeling possibilities for chiral separation by developing an enantioselective method for determination of the enantiomeric purity of ozanimod, following the principles outlined in the current ICH Q14 guideline and by leveraging the most recent modeling options provided by DryLab.

2.1. Materials

Ozanimod and R-ozanimid were from Egis Pharmaceuticals PLC (Budapest, Hungary). Gradient grade HPLC eluents (MeOH, IPA, ACN) were obtained from Merck KGaA (Darmstadt, Germany). Lux i-Amylose-1 [Amylose tris(3,5-dimethylphenylcarbamate)], Lux Amylose-2 [Amylose tris(5-chloro-2-methylphenylcarbamate)], Lux Cellulose-1 [Cellulose tris(3,5-dimethylphenylcarbamate)], Lux Cellulose-2 [Cellulose tris(3-chloro-4-methylphenylcarbamate)], Lux Cellulose-3 [Cellulose tris(4-methylbenzoate)], Lux Cellulose-4 [Cellulose tris(4-chloro-3-methylphenylcarbamate)] (150 × 4.6 mm, particle size 5 µm) were obtained from Phenomenex (Torrance, CA, USA). Chiralpak AD [Amylose tris(3,5-dimethylphenylcarbamate)] and Chiralcel OD [Cellulose tris(3,5-dimethylphenylcarbamate)] ((250 × 4.6 mm; particle size 10 µm) were the products of Daicel Corporation (Tokyo, Japan).

2.2. LC-UV analysis and software

LC-UV analyses were carried out on a JASCO HPLC system equipped with a JASCO PU-2089 Plus quaternary gradient pump, AS-4050 autosampler, MD-2010 Plus diode array detector and CO-2065 Plus column oven (Jasco Corporation, Tokyo, Japan). For the instrument control and data processing ChromNAV (Ver. 2.0, Jasco Corporation, Tokyo, Japan) was used. Acquired experimental results were processed and exported into standardized AIA/ANDI-formats (*.cdf). Retention modeling was accomplished with DryLab 4 (Ver. 4.5) modeling software (Molnár-Institute, Berlin, Germany). Modeling steps included experimental design (DoE), peak-tracking, extensive design space studies and in silico robustness testing.

All samples were dissolved in pure MeOH, stock solutions were prepared at 1 mg/mL concentrations, further dilutions were made using pure MeOH as sample solvent. The ratio between distomer-eutomer was 1:3 during the preliminary experiments (100 µg/mL vs. 300 µg/mL). Injection volume was 2.0 µL, each experiment was replicated 3 times. To ensure proper run-to-run re-equilibration of the chromatographic system, we flushed the column with 10 effective column volumes (CV) prior to injection. For the analyses, detection wavelength was uniformly set to 270 nm.

3.1. Primary data collection and defining the Analytical Target Profile

In accordance with the ICH Q14 guideline, collecting prior data from either the company’s proprietary development or from external knowledge is crucial for choosing suitable analytical technology and development strategy. At this stage, clear method objectives (performance criteria) — independent of the analytical technology in use — should also be set up and captured within the Analytical Target Profile (ATP).

Although there are no enantioseparation methods explicitly described in the literature for ozanimod, Németh and Horváth previously investigated this compound as a model substance in PO mode. Utilizing an amylose tris(3,5-dimethylphenylcarbamate) chiral stationary phase, they effectively achieved enantioseparation [18].

Furthermore, the molecular formula of ozanimod suggests lipophilic properties (logP = 3.73), with a pH-sensitive basic secondary amine embedded within its structure (pK_a=8.96).

From a chromatographic perspective, employing a high pH with a basic additive is necessary to minimize the ionization of the secondary amine. Considering the logP value and available literature, it is advisable to dissolve ozanimod in methanol.

With all of this in view, we defined our ATP goals: The enantiomeric purity testing method should accurately quantify R-ozanimod in the ozanimod sample, meeting accuracy and precision requirements of 100.0% ± 3.0% and ≤ 2.0%, respectively. It should also be able to allow quantification of the enantiomeric impurity of at least 0.05% in ozanimod samples, while also ensuring adequate peak symmetry (T_f in the usual range of 0.8-2.0) Additional expectations included short run time, preferably less than 20 minutes, fixed EEO (distomer eluting before the eutomer) and resolution values (R_s) > 2.0 even if the method is in routine use (established robustness). Building upon these considerations, chiral HPLC employing polysaccharide columns under PO mode has been selected as the preferred technique. Furthermore, environmental concerns were taken into consideration, with the polar organic mode chosen for its eco-friendliness, avoiding toxic eluents such as hexane.

3.2. Risk Assessment (RA) and identification of potential critical analytical procedure parameters (APPs)

As a subsequent step, the ICH Q14 guideline strongly advocates for the implementation of quality risk management to support the development of a robust analytical procedure, thereby mitigating the risk of poor performance and the potential reporting of inaccurate results.

In the context of quality risk assessment, straightforward tools such as the Ishikawa diagram, risk matrix, and Failure Mode and Effects Analysis (FMEA) analysis are frequently employed. When the risk is clearly defined, selecting an appropriate risk management tool, and identifying the necessary information to address the risk become more straightforward [38].

Ishikawa diagrams are typically well-suited for categorizing APPs to select critical parameters [38]. In our scenario, parameters concerning the separation system—specifically, the type of stationary phase and applied conditions — have been identified as critical APPs. The resulting heat map from the risk matrix outlined in Table 1 further highlights the risks associated with column chemistry and eluent composition. In this representation, red denotes situations of higher risk, necessitating further investigation, analysis, or attention. Conversely, green, and yellow boxes indicate a low to medium impact on performance criteria, accompanied by a reduced likelihood of occurrence. In our case, both the identified APPs and the critical APPs with potential impact on analytical procedure attributes (APA) are shown in Fig. 2.

Table 1. Risk assessment regarding chiral HPLC (red denotes situations of higher risk, yellow medium, while green no or low risk)

3.3. Initial stationary phase selection

Previous RAs have underscored the critical significance of selecting the appropriate stationary phase. However, due to the complex separation processes in chiral chromatography, it is not possible to reliably predict the accurate resolving capability of a chiral selector. Consequently, preliminary screening studies are typically necessary. To evaluate the enantioseparation capabilities of various stationary phases, we selected eight chiral columns (Chiralpak AD, Chiralcel OD, Lux i-Amylose-1, Lux Amylose-2, Lux Cellulose-1, Lux Cellulose-2, Lux Cellulose-3, and Lux Cellulose-4) and conducted some initial experiments on them using methanol and acetonitrile as mobile phases, supplemented with 0.1% diethylamine (DEA) as a basic additive to control the ionization of ozanimod.

Moreover, drawing from our prior expertise and scientific literature on amylose-based columns, we explored the feasibility of DEA-modified IPA and various binary mixtures of IPA - MeOH. As known, in case of an amylose-based column, the tridimensional structure of the chiral selector can be influenced by different alcohol ratios, potentially leading to distinct chiral environments and, consequently, varied enantiorecognition properties. Throughout the preliminary study, we maintained other parameters fixed, with the column temperature set to 25°C. For Lux columns the flow rate was set to 0.5 mL/min, while for Daicel columns, it was programmed to 0.7 mL/min, irrespective of the chosen eluent. The results of the preliminary experiments are summarized in Table 2, and representative chromatograms depicted in Fig. 3.

Table 2

The results of preliminary screening with the retention factor of the ozanimod enantiomers (k₁ and k₂), resolution (R_s) and enantiomeric elution order (EEO).
Column	Mobile phase*	k₁	k₂	R_s	EEO
Lux Cellulose-1	MeOH	1.16		-	-
Lux Cellulose-1	ACN	0.88	0.98	0.32	S < R
Lux Cellulose-2	MeOH	1.88	2.23	2.52	S < R
Lux Cellulose-2	ACN	2.59	3.20	3.23	S < R
Lux Cellulose-3	MeOH	1.08	1.22	1.21	S < R
Lux Cellulose-3	ACN	0.47		-	-
Lux Cellulose-4	MeOH	1.25	1.62	3.60	S < R
Lux Cellulose-4	ACN	1.94	2.14	1.23	S < R
Chiralcel OD	MeOH	1.97	2.09	0.55	R < S
Chiralcel OD	ACN	1.05		-	-
Lux i-Amylose-1	MeOH	0.85		-	-
	MeOH:IPA 70:30	0.58		-	-
	MeOH:IPA 50:50	0.54		-	-
	MeOH:IPA 30:70	0.56		-	-
	IPA	0.42		-	-
	ACN	1.33		-	-
Chiralpak AD	MeOH	2.30		-	-
	MeOH:IPA 70:30	1.65	1.96	2.10	R < S
	MeOH:IPA 50:50	1.52	1.77	1.51	R < S
	MeOH:IPA 30:70	1.50	1.65	1.10	R < S
	IPA	0.57		-	-
	ACN	1.97	2.48	0.51	S < R
Lux Amylose-2	MeOH	0.01		-	-
	MeOH:IPA 70:30	0.03		-	-
	MeOH:IPA 50:50	0.05		-	-
	MeOH:IPA 30:70	0.14		-	-
	IPA	0.52		-	-
	ACN	1.34	2.20	2.03	R < S
* with 0.1% DEA

These results indicated that except for the Lux i-Amylose-1 column, all chiral selectors successfully were able to deliver the desired chiral selectivity for the enantiomers. In some cases, initial resolution values far exceeded the ATP specified minimum of 2.0. Notably, on the Lux Cellulose-2 column with ACN and MeOH (Fig. 3A and B), and the Lux Cellulose-4 column with methanol, resolution values reached as high as R_s=3.6 (Fig. 3C). However, with cellulose-based columns — except for Chiralcel OD in combination with MeOH — the elution order enantiomer (EEO) was found to be not ideal, as the distomer tended to elute first. It is important to acknowledge that despite this undesired EEO, the high-resolution values could potentially facilitate the quantification of enantiomeric impurities without significant interference with the eutomer.

In contrast, both the EEO and the targeted minimum resolution threshold were conveniently achieved with Lux Amylose-2 when operated with ACN (Fig. 3E) and Chiralpak AD with MeOH-IPA mixtures (Fig. 3F).

While further optimization on Lux Amylose-2 with ACN might seem logical, a closer examination of the chromatogram revealed concerning results: despite achieving a high-resolution value and the appropriate EEO, peak symmetries were unacceptable. Additionally, observed plate numbers were significantly lower than anticipated, even after reasonable adjustments to the flow rate, column temperature and DEA amount, indicating that further improvement in peak shapes was hardly possible. As a result, we decided to proceed with the selection of the Chiralpak AD column with IPA - MeOH mixtures containing 0.1% DEA for further method development instead.

3.4. Design Space Modeling

After the initial selection of the stationary phase, the subsequent step involved establishing a systematic connection between the identified critical APPs and critical APAs. To unveil these correlations, we utilized the DryLab modeling tool. Like the modeling methodologies employed in achiral developments, we began by selecting a meaningful DoE. Building upon the findings of the previous RA step, we chose to investigate the simultaneous effect of the critical APPs within the model: eluent composition (%B = %IPA in MeOH), temperature (T), and flow rate (F).

Given this scenario, the software necessitated a comprehensive 2x2 full factorial experimental plan with two parameters — %B and T — changed experimentally. Variation in flow rate on an experimental basis was deemed unnecessary, as the applied theories of the tool account for this parameter's variability. Following the modeling software’s recommendations, when selecting appropriate ranges for the four model input experiments, it was preferable for the two temperature values to differ by at least 20–30°C. Considering that the manufacturer's recommendation for a polysaccharide column typically sets the maximum operational temperature at 40°C, choosing 10°C and 40°C as input values seemed reasonable. However, selecting the two inputs for %B presented a more compelling challenge. While the proprietary modeling functions have been proven effective in reliably modeling more complex separation cases, in this instance, it was imperative to identify an input range that would yield reproducible results and an optimal modeling fit, ensuring that there was no critical change in chiral recognition. Similar best practices and modeling principles have also been described by Fekete for large molecule applications [39].

As previously mentioned for amylose-based CSPs, such as Chiralpak AD, enantiorecognition is influenced by both the choice of eluent and the ratio of IPA to MeOH. Therefore, it was necessary to conduct a preliminary screening to assess how the percentage of IPA in MeOH affects enantiorecognition, selectivity, and retention. In this screening, the percentage of IPA was incrementally increased by 10% in MeOH, ranging from 0% IPA to 100% IPA. All mobile phases contained 0.1% DEA. The results are illustrated in Fig. 4.

It is evident that enantioseparation can only be achieved in specific eluent mixtures, and notably, the elution order of enantiomers also changes after reaching 30 v/v% of methanol in 2-propanol. However, it is observed that between 40 and 70 v/v% MeOH in IPA, enantiorecognition remains relatively consistent, indicating that this range is well-suited for subsequent modeling. At the same time, it is important to emphasize that selecting inputs outside this range may result in significant changes in separation mechanisms. In such cases, modeling algorithms may only provide results with limited accuracy. Based on these considerations, we selected the following conditions as inputs:

Chiralpak AD column, 40 °C, IPA:MeOH:DEA 30:70:0.1 (v/v/v), 0.7 mL/min
Chiralpak AD column, 10 °C, IPA:MeOH:DEA 60:40:0.1 (v/v/v), 0.7 mL/min
Chiralpak AD column, 40 °C, IPA:MeOH:DEA 60:40:0.1 (v/v/v), 0.7 mL/min
Chiralpak AD column, 10 °C, IPA:MeOH:DEA 30:70:0.1 (v/v/v), 0.7 mL/min

The four acquired chromatograms are depicted in Fig. 5.

Based on the input experiments, DryLab computed a 2-dimensional (%B-T) heat map, showcasing by default the changes in separation (R_s) across the calculated modeling range. This contextualized information can be utilized to delineate the region(s) where the ATP is fulfilled (Fig. 6A). It is evident that around 30%B and 10–15°C, the required enantioseparation can be attained. As the initial objective of the method was also rapid analysis (< 20 min), this heat map was expanded to a third dimension, incorporating flow rate within the range of 0.50-1.00 mL/min. The resultant 3D model is illustrated in Fig. 6B. Here, it's important to emphasize that according to chromatographic fundamentals, modifying the flow rate or column dimensions generally doesn't influence selectivity in isocratic elution. However, such adjustments may affect kinetic performance of the column, as well as changes in backpressure.

Flexible modeling options also allowed us to compute, visualize and study all combination of method parameters fulfilling the selected single or multiple method attributes—within the red displayed area the required analytical targets will be met. Obviously, when considering all critical APAs captured within the ATP, a more restricted space, termed the Method Operational Design Region (MODR), could be derived, illustrating all parameter combinations where both R_s > 2.0 and the analysis time is under 15 min (Fig. 6C).

In our study, the final modeling step involved validating the MODR. We meticulously chose nine work points (Table 3) within the model and conducted measurements to compare the real retention time (measured data) with the virtual retention time (modeling results), as depicted in Fig. 7. While the modeling results showed a general good agreement with reality, there was slightly more deviation compared to typical accuracy results reported by other authors [40].

Table 3

The selected working points for the model’s MODR-verification. These points were deliberately selected at various percentages of %B (IPA%), temperatures, and flow rates to comprehensively assess the MODR within the model.
Working point	%B	Temperature (°C)	Flow (ml/min.)
1	25	10	0.7
2	30	10	0.9
3	30	10	1.0
4	30	15	0.8
5	30	15	0.9
6	30	15	1.0
7	32	15	0.8
8	32	15	0.9
9	32	15	1.0

The high R² coefficient value (0.988) obtained from the total MODR-verification indicated excellent and reproducible modeling capability across the range of selected points. Run times were tightly clustered around the fitted linear curve, with a root mean square error (RMSE) of approximately 0.5 minutes. The calculated reduced R² coefficient (RR²) value was close to 1 (0.86), suggesting that a satisfactory number of verification points were selected to validate the correlative capabilities of the fitted model.

These verification values, especially in consideration of the added flow rate experimental points, underscore the modeling software’s suitability to reliably predict the enantioseparation within the previously selected modeling range in polar organic mode using various MeOH - IPA mixtures (as shown earlier in Fig. 4.).

3.5 Method validation

The final stage of our development process involved evaluating the performance of a chosen working point for the Chiralpak AD column. We aimed to ensure that the developed analytical procedure meets the related performance criteria objectives outlined in ICH Q2(R2).

Building on the validated MODR, we first selected our final working condition: IPA:MeOH:DEA 30:70:0.1 (v/v/v) with a flow rate of 0.8 ml/min and a column temperature of 10°C. Figure 8 displays the chromatogram obtained for ozanimod, containing 0.1% chiral impurity, using these optimized parameters.

Furthermore, through the integration of AQbD principles, we enhanced the rationalization of development processes by leveraging information from the previous modeling dataset and generated model. With the utilization of flexible modeling options, we assessed the prospective method performance under the selected working conditions. The experiment-free model robustness option considered all likely deviations for each chromatographic parameter, considering their specified tolerance levels. It then conducted a full-factorial analysis around the chosen point to assess the impact of these deviations on the overall performance of the method.

Specific values for in silico testing of the selected working condition and the defined tolerances are outlined in Table 4.

Table 4

Optimal working conditions selected based on the model and parameters intervals selected for *in silico* robustness testing.
Parameter	Optimal value	Level/Limit
Flow rate	0.8 mL/min	+/- 0.2
Temperature	10°C	+/- 5
IPA %	30%	+/- 5

The software calculated results indicated that 100% success rate, meaning that all the virtual combinations (3³ = 27 cases) passed R_s > 2, and t_R < 15 min.

Once the modeling tool affirmed robust separation was achieved, subsequent validation steps aimed to determine precision, accuracy linearity, limits of detection (LOD), and limit of quantification (LOQ) for the determination of R-ozanimod in ozanimod samples was investigated.

The limit of detection (LOD) and limit of quantification (LOQ) of R-ozanimod in ozanimod samples were established at concentration levels where signal-to-noise ratios of 3:1 and 10:1 were achieved, respectively. The LOD was 0.27 µg/mL, while the LOQ was 0.90 µg/mL for R-ozanimod, respectively. The linearity range was investigated in the range of 0.1% − 0.3% regarding a target concentration of 1 mg/mL ozanimod.

The equation y = 1.1542x + 0.1224 with R² = 0.9991 using 7 different points was derived to represent the relationship between concentration in µg/ml and peak area. The 95% confidence intervals of the y-intercepts included zero, and the residuals exhibited random distributions. Accuracy and precision were evaluated through intraday and interday (intermediate precision) assessments, which involved performing five replicate injections at three different concentration levels. Injections were performed on the same day and repeated over two consecutive days.

The accuracy for R-ozanimod, expressed as average recovery percentages, ranged from 99.08–101.04%. Intraday precision, represented by RSD percentages, fell within the range of 0.09–1.56%, while the RSD for intermediate precision was below 1.79%. These validation results confirm that the method is sensitive, linear, accurate, and precise for determining the chiral impurity in an ozanimod sample.

In line with the recommendation of ICH Q14, as part of method lifecycle management the classification of the most relevant APPs was summarized in Table 5. According to ICH Q14 critical APPs with high impact on method performance are marked as established conditions (ECs). Moreover, the proven acceptable ranges (PARs) for other (non-critical or non-EC) APPs calculated with the aid of in silico robustness testing are also summarized in Table 5. The role of PARs is to facilitate routine applicability of the method by providing higher flexibility compared to a traditional rigid workpoint report.

Table 5

Classification of most relevant APPs in accordance with ICH Q14 and established PARs based on *in silico* robustness testing.
Analytical procedure parameter (APP)	Optimal value	Proven acceptable range (PAR)	Classification
Stationary phase type	Chiralpak AD, 150x4.6mm, 10 µm	-	EC
Flow rate (ml/min.)	0.8	± 0.2	non-EC
Column temperature (°C)	10	± 5	non-EC
Mobile phase composition (% IPA)	30	± 5	non-EC

In conclusion, the development of a direct HPLC method for determining the enantiomeric purity of ozanimod showcases the efficacy of employing polysaccharide-based chiral columns under the polar organic mode. Through rigorous scouting and optimization phases, it was found that the Chiralpak AD column combined with a MeOH:IPA mobile phase gave the most ideal solution to perform baseline separation within 15 min.

In our work, we effectively implemented systematic elements of the Analytical Quality by Design methodology. Following the ICH Q14 and ICH Q2(R2) guideline, we conducted an initial risk assessment and identified critical APPs, linking them to a systematic model design space. Optimization was facilitated using a mechanistic modeling tool, DryLab software. The resulting validated method, aligned with ICH Q2(R2) guidelines, provided a robust means of assessing chiral impurities in ozanimod samples. Additionally, in silico robustness testing provided insights into critical separation parameters and their impact on enantioresolution. This study also highlights the utility of DryLab software, traditionally designed for reversed-phase achiral separations, in assisting chiral method optimization even in polar organic mode. However, it also revealed its bottlenecks in modeling chiral separations. One of the most critical stages involves selecting the correct eluent compositions for a given chiral stationary phase before extensive modeling. Once this selection is made, the modeling tool could reliably predict the optimal method conditions based on only a few experimental runs.

Acknowledgement

This work was funded by the National Research, Development and Innovation Office, Hungary (grant: NKFIH FK 146930). This work was supported by the ÚNKP-23-3-II and ÚNKP-23-2-III New National Excellence Program of the Ministry for Culture and Innovation funded by the National Research, Development and Innovation fund. The authors are highly appreciating the comments of Gábor Németh (Egis Plc, Budapest, Hungary).

Credit authorship contribution statement

Elek Ferencz: Methodology, Writing – original draft. Zoltán-István Szabó: Methodology, Writing – original draft, Writing – review & editing, Conceptualization. Arnold Zöldhegyi: Formal analysis, Methodology, Writing – original draft, Software. Gergely Dombi: Investigation, Methodology, Visualization. Gergely Molnár: Writing – review & editing, Project administration, Investigation. Máté Dobó: Methodology, Writing- original draft; Erzsébet Varga: Investigation, Methodology. Imre Molnár: Funding acquisition, Visualization, Resources, Conceptualization, Supervision. Gergő Tóth: Conceptualization, Methodology, Writing – original draft, Software.

Data Availability

The datasets used in the present study are available from the corresponding author on reasonable request.

Declaration of Competing Interest

Imre Molnár is the founder, while Arnold Zöldhegyi is employed by Molnár-Institute for Applied Chromatography, the developer of the DryLab software. All other authors declare no competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

L. Rasche, F. Paul, Ozanimod for the treatment of relapsing remitting multiple sclerosis, Expert Opinion on Pharmacotherapy 19 (2018) 2073–2086. https://doi.org/10.1080/14656566.2018.1540592.
G. Lassiter, C. Melancon, T. Rooney, A.M. Murat, J.S. Kaye, A.M. Kaye, R.J. Kaye, E.M. Cornett, A.D. Kaye, R.J. Shah, O. Viswanath, I. Urits, Ozanimod to Treat Relapsing Forms of Multiple Sclerosis: A Comprehensive Review of Disease, Drug Efficacy and Side Effects, Neurology International 2020, Vol. 12, Pages 89–108 12 (2020) 89–108. https://doi.org/10.3390/NEUROLINT12030016.
M. Fronza, L. Lorefice, J. Frau, E. Cocco, An overview of the efficacy and safety of ozanimod for the treatment of relapsing multiple sclerosis, Drug Design, Development and Therapy 15 (2021) 1993–2004. https://doi.org/10.2147/DDDT.S240861.
Y.N. Lamb, Ozanimod: First Approval, Drugs 80 (2020) 841–848. https://doi.org/10.1007/S40265-020-01319-7/METRICS.
F. Uthoff, J. Löwe, C. Harms, K. Donsbach, H. Gröger, Chemoenzymatic Synthesis of a Chiral Ozanimod Key Intermediate Starting from Naphthalene as Cheap Petrochemical Feedstock, Journal of Organic Chemistry 84 (2019). https://doi.org/10.1021/acs.joc.8b03290.
C. Cianferotti, G. Barreca, V. Bollabathini, L. Carcone, D. Grainger, S. Staniland, M. Taddei, Enantioselective Synthesis of Ozanimod, the Active Pharmaceutical Ingredient of a New Drug for Multiple Sclerosis, European Journal of Organic Chemistry 2021 (2021). https://doi.org/10.1002/ejoc.202100058.
J.H. M. Porcs-Makkay, G. Simig, E. Molnár, J. Pethő, B. Volk, L. Szlávik, E.Szabó, Process for the production of ozanimod, WO2018/215807A1, 2018. https://doi.org/WO2018/215807A1.
M. Buchlovic, L. Chalupova, L. Vyklicky, Process for preparing ozanimod, 2022.
L.A. Papp, Z.I. Szabó, G. Hancu, L. Farczádi, E. Mircia, Comprehensive Review on Chiral Stationary Phases in Single-Column Simultaneous Chiral–Achiral HPLC Separation Methods, Molecules 2024, Vol. 29, Page 1346 29 (2024) 1346. https://doi.org/10.3390/MOLECULES29061346.
H. Liu, Z. Wu, J. Chen, J. Wang, H. Qiu, Recent advances in chiral liquid chromatography stationary phases for pharmaceutical analysis, Journal of Chromatography A 1708 (2023) 464367. https://doi.org/10.1016/J.CHROMA.2023.464367.
R. Dallocchio, A. Dessì, B. Sechi, P. Peluso, Molecular Dynamics Simulations of Amylose- and Cellulose-Based Selectors and Related Enantioseparations in Liquid Phase Chromatography, Molecules 2023, Vol. 28, Page 7419 28 (2023) 7419. https://doi.org/10.3390/MOLECULES28217419.
B. Sechi, A. Dessì, R. Dallocchio, N. Tsetskhladze, B. Chankvetadze, M. Pérez-Baeza, S. Cossu, G. Jibuti, V. Mamane, P. Peluso, Unravelling dispersion forces in liquid-phase enantioseparation. Part I: Impact of ferrocenyl versus phenyl groups, Analytica Chimica Acta 1278 (2023) 341725. https://doi.org/10.1016/J.ACA.2023.341725.
P. Peluso, B. Sechi, G. Lai, A. Dessì, R. Dallocchio, S. Cossu, E. Aubert, R. Weiss, P. Pale, V. Mamane, B. Chankvetadze, Comparative enantioseparation of chiral 4,4’-bipyridine derivatives on coated and immobilized amylose-based chiral stationary phases, Journal of Chromatography A 1625 (2020) 461303. https://doi.org/10.1016/J.CHROMA.2020.461303.
E.P. Sousa, M.E. Tiritan, R. V. Oliveira, C.M.M. Afonso, Q.B. Cass, M.M.M. Pinto, Enantiomeric resolution of kielcorin derivatives by HPLC on polysaccharide stationary phases using multimodal elution, Chirality 16 (2004) 279–285. https://doi.org/10.1002/CHIR.20031.
B. Chankvetadze, Recent trends in preparation, investigation and application of polysaccharide-based chiral stationary phases for separation of enantiomers in high-performance liquid chromatography, TrAC Trends in Analytical Chemistry 122 (2020) 115709. https://doi.org/10.1016/J.TRAC.2019.115709.
I. Matarashvili, I. Shvangiradze, L. Chankvetadze, S. Sidamonidze, N. Takaishvili, T. Farkas, B. Chankvetadze, High-performance liquid chromatographic separations of stereoisomers of chiral basic agrochemicals with polysaccharide-based chiral columns and polar organic mobile phases, Journal of Separation Science 38 (2015) 4173–4179. https://doi.org/10.1002/JSSC.201500919.
Z.-I. Szabó, M. Foroughbakhshfasaei, R. Gál, P. Horváth, B. Komjáti, B. Noszál, G. Tóth, Chiral separation of lenalidomide by liquid chromatography on polysaccharide-type stationary phases and by capillary electrophoresis using cyclodextrin selectors, Journal of Separation Science 41 (2018). https://doi.org/10.1002/jssc.201701211.
S. Horváth, G. Németh, Hysteresis of retention and enantioselectivity on amylose tris(3,5-dimethylphenylcarbamate) chiral stationary phases in mixtures of 2-propanol and methanol, Journal of Chromatography A 1568 (2018) 149–159. https://doi.org/10.1016/J.CHROMA.2018.07.033.
G. Németi, R. Berkecz, T.M. Le, Z. Szakonyi, A. Péter, I. Ilisz, High-performance liquid chromatographic enantioseparation of azole analogs of monoterpene lactones and amides focusing on the separation characteristics of polysaccharide-based chiral stationary phases, Journal of Chromatography A 1717 (2024) 464660. https://doi.org/10.1016/J.CHROMA.2024.464660.
M. Dobó, M. Foroughbakhshfasaei, P. Horváth, Z.I. Szabó, G. Tóth, Chiral separation of oxazolidinone analogues by liquid chromatography on polysaccharide stationary phases using polar organic mode, Journal of Chromatography A 1662 (2022). https://doi.org/10.1016/j.chroma.2021.462741.
M. Wadie, E.M. Abdel-Moety, M.R. Rezk, M.A. Tantawy, Eco-friendly chiral HPLC method for determination of alfuzosin enantiomers and solifenacin in their newly pharmaceutical combination: Method optimization via central composite design, Microchemical Journal 165 (2021) 106095. https://doi.org/10.1016/J.MICROC.2021.106095.
M. Dobó, G. Dombi, I. Köteles, B. Fiser, C. Kis, Z.I. Szabó, G. Tóth, Simultaneous Determination of Enantiomeric Purity and Organic Impurities of Dexketoprofen Using Reversed-Phase Liquid Chromatography—Enhancing Enantioselectivity through Hysteretic Behavior and Temperature-Dependent Enantiomer Elution Order Reversal on Polysaccharide Chiral Stationary Phases, International Journal of Molecular Sciences 25 (2024) 2697. https://doi.org/10.3390/IJMS25052697/S1.
Internationa Council for Harmonisation, ICH Q14 - Analytical Procedure Development - draft, 2024.
Food an Drug Administration, Analytical procedures lifecycle - draft, n.d.
L.R. Snyder, J.W. Dolan, D.C. Lommen, Drylab® computer simulation for high-performance liquid chromatographic method development: I. Isocratic elution, Journal of Chromatography A 485 (1989) 65–89. https://doi.org/10.1016/S0021-9673(01)89133-0.
J.W. Dolan, D.C. Lommen, L.R. Snyder, Drylab® computer simulation for high-performance liquid chromatographic method development: II. Gradient Elution, Journal of Chromatography A 485 (1989) 91–112. https://doi.org/10.1016/S0021-9673(01)89134-2.
L.A. Larew, B.A. Olsen, J.D. Stafford, M. V. Wilhelm, Comparison of theory-based and empirical modeling for the prediction of chromatographic behavior in the ion-pairing separation of benzodiazepine-derived pharmaceutical compounds, Journal of Chromatography A 692 (1995) 183–193. https://doi.org/10.1016/0021-9673(94)00877-C.
S. Fekete, A. Beck, J. Fekete, D. Guillarme, Method development for the separation of monoclonal antibody charge variants in cation exchange chromatography, Part II: pH gradient approach, Journal of Pharmaceutical and Biomedical Analysis 102 (2015) 282–289. https://doi.org/10.1016/J.JPBA.2014.09.032.
S. Fekete, A. Beck, J. Fekete, D. Guillarme, Method development for the separation of monoclonal antibody charge variants in cation exchange chromatography, Part I: Salt gradient approach, Journal of Pharmaceutical and Biomedical Analysis 102 (2015) 33–44. https://doi.org/10.1016/J.JPBA.2014.08.035.
V. D’Atri, R. Pell, A. Clarke, D. Guillarme, S. Fekete, Is hydrophobic interaction chromatography the most suitable technique to characterize site-specific antibody-drug conjugates?, Journal of Chromatography A 1586 (2019) 149–153. https://doi.org/10.1016/J.CHROMA.2018.12.020.
B. Bobály, V. D’Atri, A. Beck, D. Guillarme, S. Fekete, Analysis of recombinant monoclonal antibodies in hydrophilic interaction chromatography: A generic method development approach, Journal of Pharmaceutical and Biomedical Analysis 145 (2017) 24–32. https://doi.org/10.1016/J.JPBA.2017.06.016.
J. Neumann, S. Schmidtsdorff, A.H. Schmidt, M.K. Parr, Retention modeling of therapeutic peptides in sub-/supercritical fluid chromatography, Separation Science Plus (2024) 2300239. https://doi.org/10.1002/SSCP.202300239.
M. Lämmerhofer, P. Di Eugenio, I. Molnar, W. Lindner, Computerized optimization of the high-performance liquid chromatographic enantioseparation of a mixture of 4-dinitrophenyl amino acids on a quinine carbamate-type chiral stationary phase using drylab, Journal of Chromatography B: Biomedical Sciences and Applications 689 (1997) 123–135. https://doi.org/10.1016/S0378-4347(96)00366-0.
L. Zhang, Y. Hu, E. Galella, F.P. Tomasella, W.P. Fish, Separation of atropisomers by chiral liquid chromatography and thermodynamic analysis of separation mechanism, J Pharm Anal 7 (2017) 156–162. https://doi.org/10.1016/J.JPHA.2017.03.003.
E. Ferencz, É.-K. Kelemen, M. Obreja, G. Tóth, M. Urkon, A. Zöldhegyi, E. Sipos, Z.-I. Szabó, The Applicability of Chromatographic Retention Modeling on Chiral Stationary Phases in Reverse-Phase Mode: A Case Study for Ezetimibe and Its Impurities, International Journal of Molecular Sciences 2023, Vol. 24, Page 16097 24 (2023) 16097. https://doi.org/10.3390/IJMS242216097.
H.A. Wagdy, R.S. Hanafi, R.M. El-Nashar, H.Y. Aboul-Enein, Predictability of Enantiomeric Chromatographic Behavior on Various Chiral Stationary Phases Using Typical Reversed Phase Modeling Software, Chirality 25 (2013) 506–513. https://doi.org/10.1002/CHIR.22154.
International Council for Harmonisation, ICH Q9(R1) - Quality risk management, 2005.
S. Fekete, I. Molnár, Software-Assisted Method Development in High Performance Liquid Chromatography - Chaps. 9–12, World Scientific (Europe), 2019.
R. Kormány, I. Molnár, J. Fekete, D. Guillarme, S. Fekete, Robust UHPLC separation method development for multi-API product containing amlodipine and bisoprolol: The impact of column selection, Chromatographia 77 (2014) 1119–1127. https://doi.org/10.1007/s10337-014-2633-9.

Competing interest reported. Imre Molnár is the founder, while Arnold Zöldhegyi is employed by Molnár-Institute for Applied Chromatography, the developer of the DryLab software. All other authors declare no competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Download PDF

Editorial decision: Revision requested
29 Jul, 2024
Reviews received at journal
26 Jul, 2024
Reviews received at journal
24 Jul, 2024
Reviews received at journal
22 Jul, 2024
Reviewers agreed at journal
18 Jul, 2024
Reviewers agreed at journal
18 Jul, 2024
Reviewers agreed at journal
18 Jul, 2024
Reviewers agreed at journal
17 Jul, 2024
Reviewers agreed at journal
17 Jul, 2024
Reviewers invited by journal
16 Jul, 2024
Editor assigned by journal
16 Jul, 2024
Editor invited by journal
09 Jun, 2024
Submission checks completed at journal
05 Jun, 2024
First submitted to journal
03 Jun, 2024

You are reading this latest preprint version

Computer-assisted chiral HPLC method development for ozanimod on polysaccharide- based chiral stationary phases: possibilities and limitations

Status:

Version 1

Abstract

Figures

1. Introduction

2. Materials and methods

2.1. Materials

2.2. LC-UV analysis and software

3. Results and Discussion

3.1. Primary data collection and defining the Analytical Target Profile

3.2. Risk Assessment (RA) and identification of potential critical analytical procedure parameters (APPs)

3.3. Initial stationary phase selection

3.4. Design Space Modeling

4. Conclusion

Declarations

Acknowledgement

References

Additional Declarations

Status:

Version 1

Working point	%B	Temperature (°C)	Flow (ml/min.)
1	25	10	0.7
2	30	10	0.9
3	30	10	1.0
4	30	15	0.8
5	30	15	0.9
6	30	15	1.0
7	32	15	0.8
8	32	15	0.9
9	32	15	1.0

Working point	%B	Temperature (°C)	Flow (ml/min.)
1	25	10	0.7
2	30	10	0.9
3	30	10	1.0
4	30	15	0.8
5	30	15	0.9
6	30	15	1.0
7	32	15	0.8
8	32	15	0.9
9	32	15	1.0

Working point	%B	Temperature (°C)	Flow (ml/min.)
1	25	10	0.7
2	30	10	0.9
3	30	10	1.0
4	30	15	0.8
5	30	15	0.9
6	30	15	1.0
7	32	15	0.8
8	32	15	0.9
9	32	15	1.0