Enabling Large‐Volume Injections in Hydrophilic Interaction Chromatography of Oligonucleotides Through In‐Line Mixing

TL;DR

This paper shows how in-line mixing improves HILIC separations for oligonucleotides by preventing peak distortion and breakthrough during large-volume injections.

Contribution

The study demonstrates that in-line mixing enables large-volume injections in HILIC without compromising separation quality for oligonucleotides.

Findings

In-line mixing prevents analyte breakthrough and peak splitting during large-volume injections of oligonucleotides in HILIC.

Gradient HILIC with in-line mixing allows up to 40 µL injections of oligonucleotides dissolved in 100% water without separation loss.

Large-volume injection with in-line mixing enables detection of main compounds and minor impurities in pharmaceutical antisense oligonucleotides.

Abstract

Hydrophilic interaction chromatography (HILIC) is an attractive separation mode for the analysis of therapeutic oligonucleotides (ONs). ONs are very polar compounds that are commonly dissolved in highly aqueous media, whereas HILIC eluents often comprise a high percentage of organic solvent. This solvent mismatch can cause breakthrough and peak splitting. In this study, we investigated the effects of the sample solvent composition and injection volume on the HILIC separation of nucleobases and ONs, and to what extent an in‐line mixer between injector and column can mitigate breakthrough and peak splitting. Using isocratic HILIC with nucleobases as medium‐polar, less‐retentive test compounds, we illustrated that an injection solvent of 90% water results in peak broadening, which deteriorates with increasing injection volume, leading to serious peak deformations and asymmetries. Here, in‐line mixing did not improve peak shapes, as the mixer volume adds to band broadening and the homogenization of the injection solvent with the HILIC eluent does not lead to on‐column focusing conditions. When analyzing an ON mixture (15/16/17‐mer) dissolved in a weak solvent (30% water) by gradient HILIC, injection volumes could be increased up to 20 µL (5.7% of the column volume) without losing separation performance. However, when the ON mixture was dissolved in 100% water, injection volumes above 2 µL caused severe peak distortion and extensive analyte breakthrough in gradient HILIC. When injecting the ONs in 100% water with a 100‐µL in‐line mixer, no analyte breakthrough and peak splitting were observed for injection volumes up to 40 µL, while maintaining baseline separation of the ONs. The usefulness of in‐line mixing in HILIC was demonstrated by the analysis of a pharmaceutical antisense ON (ASO). Large‐volume injection of this ON in 100% water allowed for the separation and detection of the main compound and adjacent minor impurities, without breakthrough or severe peak distortion.

Figures

Conventional LC setup (top) and in‐line mixing setup (bottom).

Normalized HILIC chromatograms of the nucleobase mixture dissolved in (A + C) water–ACN (20:80, v/v) or (B +D) water–ACN (90:10, v/v) using the indicated injection volume; (A + B) no mixer installed; (C + D) with mixer A (35 µL) or mixer B (100 µL) installed between injector and column. Eluent, 100 mM AF (pH 3)–ACN (10:90, v/v); flow rate 0.5 mL min−1; nucleobases: A, adenosine; U, uracil; C, cytosine.

Normalized chromatograms obtained during gradient HILIC of the ON mixture dissolved in (A) water–ACN (30:70, v/v) and (B) 100% water using the indicated injection volumes. Gradient: solvent A, 25 mM AA in ACN–water (40:60, v/v); solvent B, 25 mM AA in ACN–water (80:20, v/v); gradient program, 60% B for 1 min, 60%–30% B linearly in 14 min, 30% B for 1 min using an overall flow rate of 0.5 mL min−1. ON elution order: N − 1, FLP, and N + 1.

Chromatograms obtained during gradient HILIC of the ON mixture dissolved in (A) water–ACN (30:70, v/v) and (B) water using the indicated injection volume with no mixer, mixer A or, mixer B installed. (C) Percentage breakthrough (area %) calculated from the chromatograms provided in (B). For gradient conditions and ON elution order, see Figure 3.

Gradient HILIC of ASO X dissolved in water (50 µg mL−1) employing in‐line mixing: (A) full chromatogram and (B) zoom of the 11–14.5 min retention interval. Injection volume, 20 µL; in‐line mixer, mixer B; solvent A, 25 mM AA in ACN–water (40:60, v/v); solvent B, 25 mM AA in ACN–water (80:20, v/v). Gradient: 65% B for 1 min, linear decrease to 35% B in 14 min, 1 min at 35% B. Flow rate: 0.3 mL min−1.

Full text

Introduction

1

Therapeutic oligonucleotides (ONs) are a class of pharmaceuticals that have shown rapid development over the last two decades [1]. ONs cover a broad range of active compounds, such as antisense ONs (ASOs), small interfering RNA (siRNA), messenger RNA (mRNA), and aptamers. Their ability to modulate protein expression by selective targeting of mRNA, preventing protein translation by binding to proteins to block their activity, or by acting as templates for therapeutic proteins, makes them valuable modalities for the treatment of, for example, genetic, neurodegenerative, and viral diseases [2, 3]. In 1998, the first ASO was approved by the Food and Drug Administration (FDA), and since then, a large number of ONs have entered clinical trials, leading to more than 20 registered ONs [4].

To establish safe and effective ON formulations, analytical workflows have to be developed that can characterize critical quality attributes (CQA) of the complex ON molecules. In 2024, the European Medicines Agency (EMA) published guidelines on the characterization of ONs [5], of which physical, structural, and impurity analysis are important parts. Chromatography has proven to be essential for resolving ON variants and impurities from the main active pharmaceutical ingredient, with ion‐pairing reversed‐phase chromatography (IP‐RPLC) and anion exchange chromatography (AEX) being predominantly used in industry. More recently, there has been an increased attention for the use of hydrophilic interaction chromatography (HILIC) for the analysis of ONs, which are highly polar compounds due to their negatively charged backbones. HILIC exhibits a unique selectivity and has also been shown to be useful for the separation of ON diastereomers [6]. Opposed to AEX and IP‐RPLC, HILIC does not require eluents with high concentrations of nonvolatile salts (AEX) or ion‐pairing reagents (IP‐RPLC), providing high compatibility with mass spectrometry (MS) [7]. Moreover, HILIC is valuable for multi‐attribute analysis of ONs as part of two‐dimensional LC (2D‐LC) setups, where HILIC has been used both as a first‐dimension (^1^D) and second‐dimension (^2^D) separation mode [8, 9, 10, 11].

While being a powerful separation mode, HILIC is prone to peak deformations when relatively large volumes (typically 1%–10% of column volume [8]) and/or highly aqueous samples are injected. Under these conditions, analyte breakthrough and/or splitting of analyte peaks can be observed. Large‐volume injections may be desirable when low analyte concentrations need to be determined, for instance, in the case of impurity analysis, in 2D‐LC applications with HILIC in ^2^D, or when (semi‐)preparative HILIC is employed for compound purification. Various injection strategies have been proposed to minimize the adverse effects of injecting relatively large volumes of aqueous samples on HILIC columns (Table S1). For instance, Jaffuel et al. extensively described how at‐column dilution (ACD) can be used to improve chromatographic performance for highly aqueous samples in preparative HILIC [12]. In ACD, the solvent strength of the injected sample is lowered with a weak solvent in an in‐line mixer. This weak solvent is introduced to the sample plug before the mixer using a T‐piece and an auxiliary pump. Although ACD requires an additional pump in the setup, performance optimizing injection sequence (POISe) offers a simpler solution and setup by injecting a plug of weak solvent alongside the sample, reducing the effective solvent strength of the injection plug. This plug can be introduced before and/or after the sample. However, large‐volume injections of aqueous samples remain challenging with this technique [13]. In‐line mixing offers another option that has not yet been used in one‐dimensional (1D) HILIC. In this approach, a mixer is placed between the injector and the HILIC column to improve the mixing of the weaker eluent with the stronger sample solvent (Figure 1). Here, in‐line mixing has shown to moderate solvent mismatch of the sample and eluent in RPLC, while no auxiliary pump is needed [14], which simplifies the instrumental setup.

Conventional LC setup (top) and in‐line mixing setup (bottom).

ONs can be characterized more comprehensively when the HILIC separation is coupled to a second separation dimension with orthogonal selectivity. When HILIC is utilized in 2D‐LC, it often requires solvent modulation between the LC dimensions to ensure good HILIC performance, especially when HILIC is utilized in ^2^D. Offline and passive modulation is used in most of the employed modulation strategies that have been reported since 2006 (Figure S1 and Table S2). Offline modulation is labor intensive, involving fractionation of the ^1^D effluent, usually followed by evaporation of the eluent and reconstitution of the analytes in a weak HILIC solvent. Offline modulation strategies are frequently used when the ^1^D is (semi‐)preparative LC and the volumetric flow rate is relatively large. When the ^1^D effluent is highly organic, passive modulation (i.e., direct transfer of the unmodified ^1^D effluent to the ^2^D) can be employed, as the ^1^D effluent often is a weak solvent in HILIC (Table S2) and does not cause peak deformations. Stationary‐phase–assisted modulation (SPAM) and ACD have been successfully used to achieve 2D‐LC with HILIC as ^2^D, but the number of studies reported thus far is quite limited (Table S2). More recently, in‐line mixing modulation (ILMM) has been introduced in the IP‐RPLC–HILIC analysis of ONs and polymers, allowing for injection of relatively large fractions of the aqueous ^1^D effluent on the HILIC column in ^2^D. Comparisons with passive modulation and active‐solvent modulation (ASM) were made [9].

Recently, Guillarme et al. published an extensive protocol for the characterization of ONs by HILIC [15]. In their work, the authors highlighted the importance of injection conditions when analyzing ONs. To prevent or overcome incompatibility issues, the authors recommended (i) injection of ONs in a weak solvent, (ii) use of small injection volumes, or (iii) to use POISe. In our present work, we considered in‐line mixing to facilitate large‐volume injections of aqueous samples in HILIC with minimal instrumental adjustments. For this, we studied the effect of injection solvent composition (highly aqueous or organic) and injection volume (up to 40 µL) with and without an in‐line mixer (volume of 35 or 100 µL) on the performance of isocratic HILIC of medium polar nucleobases and of gradient HILIC of highly polar ONs. The applicability of in‐line mixing was evaluated using HILIC analysis of a therapeutic ASO sample dissolved in water.

Materials and Methods

2

Chemicals and Reagents

2.1

Adenine (A; ≥ 99%), ammonium acetate (AA; BioXtra, ≥ 98%), ammonium formate (AF), cytosine (C; ≥ 99%), and uracil (U; ≥ 99.0%) were obtained from Sigma‐Aldrich (Darmstadt, Germany). Acetonitrile (ACN; HPLC‐R and LC–MS grade) and formic acid (FA; 99% ULC/MS—CC/SFC grade) were obtained from Biosolve (Valkenswaard, Netherlands). All water used was deionized by a Milli‐Q Purification System (Merck Millipore, Burlington, Massachusetts, USA). A phosphorothioated 16‐mer GalNAc‐conjugated ON (FLP, full‐length product) and a shortmer (N − 1) and longmer (N + 1) thereof were provided by AstraZeneca, as well as a 16‐mer ASO (ASO X).

Sample Preparations

2.2

The nucleobase samples were prepared by making 1 mg mL^−1^ stock solutions of A, U, and C in ACN–water (50:50, v/v). From these stock solutions, a stock mixture was made at 0.2 mg mL^−1^ each in 30% ACN in water. From this, standard solutions were made at concentrations of 50, 25, 10, 5, 2.5, and 1.25 µg mL^−1^. Each standard was made in both 10% ACN and 80% ACN in water.

A 1 mg mL^−1^ stock solution of the FLP, N − 1, and N + 1 was made in water. From these stock solutions, mixtures were made in 100% water and 40% ACN in water, respectively, containing 0.2 mg mL^−1^ of each ON. Dilutions were made from the mixtures at concentrations of 100, 50, 20, 10, 5, and 2.5 µg mL^−1^. For each concentration, a standard was made in 100% water and 70% ACN in water. A 1 mg mL^−1^ stock solution of ASO X was prepared in 100% water and diluted to 100 or 50 µg mL^−1^ prior to injection for a 10 or 20 µL injection, respectively.

LC Instrumentation

2.3

The LC measurements were performed on an Agilent 1290 Infinity II UHPLC (Agilent, Waldbronn, Germany). The system was equipped with a 1290 high‐speed pump (G7120A) containing a Jet Weaver V100 mixer, a multisampler (G7167B), a multicolumn thermostat (G7116B) with an 8‐position column‐switching valve, and a DAD FS detector (G7117A) containing a max‐light cartridge cell (optical path length, 10 mm; volume, 1 µL). The autosampler operated at a draw and injection speed of 100 µL min^−1^. The DAD collected data at 260 nm with a 4.0 nm bandwidth at 10 Hz. The column oven was set to 60°C unless specified differently.

To minimize sample adsorption, a bioinert HILIC column and bioinert in‐line mixers were used [16]. A Waters Corporation ACQUITY Premier BEH Amide (1.7 µm, 2.1 × 150 mm, 130 Å) column was used for all HILIC separations. For in‐line mixing, an Agilent bioinert Jet Weaver V35/100 (BI JW, G7132‐68135) was used. Two valve positions of the column‐switching valve were equipped with a 35‐ and 100‐µL mixer (mixer A and mixer B, respectively), whereas a third position was equipped with a piece of tubing of negligible volume relative to the mixers’ volumes (denoted as “no mixer”).

HILIC Analysis

2.4

For the analysis of nucleobases, solvent A consisted of 100 mM AF in water adjusted to pH 3 with formic acid, and solvent B was ACN. The nucleobases were analyzed with isocratic HILIC using 10% A at a flow rate of 0.5 mL min^−1^ and a column temperature of 30°C.

The ON mixture of FLP, N + 1, and N − 1 was analyzed by HILIC using a gradient of 25 mM AA in ACN–water (40:60, v/v) (eluent A) and 25 mM AA in ACN–water (80:20, v/v) (eluent B) at an overall flow rate of 0.5 mL min^−1^. The gradient program started with a 1 min hold at 60% B, followed by a linear decrease to 30% B in 14 min, and a hold of 1 min at 30% B. ASO X was measured using a gradient program starting with a 1 min hold at 65% B, followed by a linear decrease to 35% B in 14 min, and a hold of 1 min at 35% B using an overall flow rate of 0.3 mL min^−1^.

Data Processing

2.5

MATLAB R2023b was used to visualize the chromatograms. Peak parameters (i.e., peak position, area, height, asymmetry, tailing, width at 10%, and width at 50% peak height) were obtained using the OpenLAB CDS software (Agilent Technologies, Waldbronn, Germany). Subsequent processing and visualization of the peak parameters were performed with Microsoft Excel (version 16.16.27).

Results and Discussion

3

Effect of Injection Solvent in Isocratic HILIC of Nucleobases

3.1

HILIC methods typically use eluents with high percentages of ACN (e.g., 95%) and a low percentage of water (e.g., 5%) in combination with polar stationary phases exhibiting amide, diol, or sulfobetaine functional groups at their surface. HILIC methods allow for the retention of polar compounds that are usually not (well) retained in reversed‐phase LC. However, highly polar analytes often have low solubility in ACN, and therefore are dissolved in solvents rich in water. When the sample solvent is highly aqueous, sample injection onto the HILIC column introduces a mismatch between the composition of the injection solvent and the eluent, potentially resulting in analyte breakthrough and/or peak deformation. This especially holds true when relatively large volumes (1%–10% of the column volume [8]) of aqueous solvent are injected onto the HILIC column. To study and highlight the issue of injecting such volumes of aqueous solvent in HILIC, we first analyzed a test mixture of the nucleobases A, U, and C by isocratic HILIC using a mobile phase of 100 mM AF (pH 3)–ACN (10:90, v/v). Nucleobases are medium polar and, hence, not strongly retained and more prone to breakthrough, which allows straightforward study of the effect of the sample solvent composition and injection volume.

Figure 2A shows the results obtained for the nucleobase test mixture that was dissolved in water–ACN (20:80, v/v), which is only slightly stronger than the eluent. The injection volume (V inj) was 1, 2, 5, 10, 20, or 40 µL, corresponding to 0.3%, 0.6%, 1.4%, 2.9%, 5.7%, or 11.4% of the column dead volume (V 0), respectively. The overall injected mass of A, U, and C was kept constant by proportionally decreasing the injected concentrations when the injection volume was increased. When injecting 1–2 µL, three narrow and well‐separated peaks are observed, corresponding to the nucleobases in order of increasing hydrophilicity, with base A eluting first, followed by U and C. From an injection volume of 5 µL and larger, the nucleobase peaks clearly started to broaden, whereas from 20 µL onwards, peak splitting and further deformation occurred. The adverse influence of raising the injection volume on peak shapes can be mainly ascribed to volume overloading effects. When the nucleobase mixture was dissolved in highly aqueous solvent (water–ACN (90:10, v/v)) and analyzed by isocratic HILIC, the chromatogram already showed broadened peaks when using the smallest injection volume of 1 µL (Figure 2B). Further increasing the injection volume quickly resulted in very severe peak deformation and peak splitting. Clearly, injection of analytes in a strong solvent in isocratic HILIC employing a relatively weak eluent negatively affects peak shape and the attainable resolution.

Normalized HILIC chromatograms of the nucleobase mixture dissolved in (A + C) water–ACN (20:80, v/v) or (B +D) water–ACN (90:10, v/v) using the indicated injection volume; (A + B) no mixer installed; (C + D) with mixer A (35 µL) or mixer B (100 µL) installed between injector and column. Eluent, 100 mM AF (pH 3)–ACN (10:90, v/v); flow rate 0.5 mL min−1; nucleobases: A, adenosine; U, uracil; C, cytosine.

Next, we studied the effect of incorporating an in‐line solvent mixer between the injector and HILIC column, aiming to homogenize the sample plug with the eluent, thereby lowering the overall elution strength of the injection plug. When injecting 5 µL of the nucleobase mixture dissolved in water–ACN (90:10, v/v) with the mixer A (35 µL volume) or mixer B (100 µL volume) in place, the peak shapes of the nucleobases further deteriorated (Figure 2C). We suggest that under these conditions the homogenization of the injection solvent with the eluent in the mixers does not sufficiently lower the solvent strength of the injection plug and allow for the analytes to focus on the column head. In contrast, the relatively large mixer volume further adds to the injection band broadening. The latter was confirmed by injection of 5 µL of the nucleobase mixture dissolved in water–ACN (20:80, v/v) with the mixers installed (Figure 2D). The widths of the peaks observed for the nucleobases are considerably larger than obtained without the in‐line mixers (cf. Figure 1A).

Effect of Injection Solvent in Gradient HILIC of ONs

3.2

For HILIC of ONs, we investigated the effect of injection solvent and volume by analyzing the mixture of FLP, N + 1, and N − 1. These ONs are relatively large and highly polar and, therefore, usually are well retained on HILIC stationary phases. ONs exhibit an “on–off”‐like retention behavior [17], and typically require gradient elution to be separated in a reasonable time, even when the ONs are rather similar. In this study, we used a gradient composed ACN–water (40:60, v/v) as solvent A and ACN–water (80:20, v/v) as solvent B, both comprising 25 mM AA.

Gradient HILIC of the ON mixture was carried out using different injection solvents and volumes, while keeping the overall injected ON mass constant. As a bench mark, Figure 3A shows the obtained chromatograms when using water–ACN (30:70, v/v) as ON‐mixture solvent and increasing the injection volume from 1 to 40 µL. For all injection volumes in the 1–20 µL range, good and similar resolution was obtained for the three ONs, which eluted in order of increasing size and showed sharp and symmetric peaks of virtually constant width. Notably, the elution strength of the sample solvent was slightly lower than that of the eluent at the start of the gradient, providing on‐column focusing conditions and thus allowing injection of relatively large volumes without causing band broadening. At an injection volume of 40 µL, the ON peaks remained baseline‐separated, but started to deform and split due to volume overloading and/or sub‐optimal mixing of the injection solvent and eluent.

Normalized chromatograms obtained during gradient HILIC of the ON mixture dissolved in (A) water–ACN (30:70, v/v) and (B) 100% water using the indicated injection volumes. Gradient: solvent A, 25 mM AA in ACN–water (40:60, v/v); solvent B, 25 mM AA in ACN–water (80:20, v/v); gradient program, 60% B for 1 min, 60%–30% B linearly in 14 min, 30% B for 1 min using an overall flow rate of 0.5 mL min−1. ON elution order: N − 1, FLP, and N + 1.

Using 100% water as sample solvent in gradient HILIC analysis of the test ONs, only the 1‐ and 2‐µL injections yielded well‐shaped ON peaks (Figure 3B), with resolutions of 3.3 and 3.6 for the N − 1/FLP and FLP/N + 1 peak pairs when injecting 1 µL volume (Figure S2). When 5 µL volume was injected, part of the ON molecules did not show proper retention as they were smeared over a 5‐min wide elution range (7–12 min). For 100% aqueous mixture injections of 10 µL and larger, major breakthrough occurred (51% or more of total peak area), where the ONs largely eluted at t_0_. We concluded that under these conditions, the sample plug does not sufficiently mix with the weak starting eluent. As a result, (large) part of the injected ONs were not focused on the HILIC column head, but either spread over the HILIC stationary phase (injection volume, 5 µL or 1.4% of column volume) or eluted with the unmixed water‐rich injection plug at t_0_ (injection volume, ≥ 10 µL or 2.9% of column volume).

The effect of in‐line mixing on the HILIC performance when injecting the test ONs dissolved in 30% or 100% water was studied after installing mixer A or B between the injector and the column. Figure 4A,B shows the 5‐ to 15‐min retention intervals of the chromatograms obtained for the 30% and 100% water injections, respectively, without and with the mixers in place (see Figure S3 for full chromatograms).

Chromatograms obtained during gradient HILIC of the ON mixture dissolved in (A) water–ACN (30:70, v/v) and (B) water using the indicated injection volume with no mixer, mixer A or, mixer B installed. (C) Percentage breakthrough (area %) calculated from the chromatograms provided in (B). For gradient conditions and ON elution order, see Figure 3.

For the 20‐µL injection of the ON mixture in water–ACN (30:70, v/v), the incorporation of mixers A or B had no adverse effects: The ON peak widths and resolution remained unchanged (Figure 4A, bottom), with only a 2% increase in peak width at half height when using mixer B compared to when no mixer is used. Even better, inclusion of the mixers restored the peak shapes of the ONs when injecting 40 µL (Figure 4A, top), indicating mixing of the injection plug with the weaker eluent at the start of the gradient was sufficient. When injecting 20 µL of the ON mixture dissolved in 100% water, installation of the in‐line mixers mitigated complete breakthrough of the ONs (Figure 4B(bottom),C). Yet, mixer A did not prevent peak splitting, with part of the injected ONs eluting before their anticipated retention times. When the injection volume was increased to 40 µL, mixer A also failed to prevent breakthrough (Figure 4B(top),C). Apparently, a 35‐µL mixer volume is not sufficiently large to fully homogenize the 100% aqueous sample solvent with the eluent when injection volumes are 20 µL or more. However, when using an in‐line mixer with a volume of 100 µL (mixer B), breakthrough and peak splitting could be completely prevented for the injection of both 20 and 40 µL of the ON mixture in water. While the resolution between the peaks decreased in comparison with the 1‐µL injection, the separation was still sufficient. For the 40‐µL injection, resolutions for the N − 1/FLP and FLP/N + 1 peak pairs were 2.3 and 2.6, respectively (Figure S2), as for the 1‐µL injection, the corresponding resolutions were 3.3 and 3.6, respectively. The ONs exhibited somewhat wider peaks (0.22 vs. 0.15 min at 50% peak height) as compared to results obtained upon injection of much smaller volumes of the ON mixture in weak solvent, but baseline resolution of the N − 1, FLP, and N + 1 was still achieved. The observed decrease in resolution most likely is caused by peak broadening due to the larger injection volumes and the related increase in the dwell volume when using the in‐line mixer. This is confirmed when looking at the peak widths at half height obtained for the 10 µL injections with water–ACN (30:70, v/v) as sample solvent when using no mixer, mixer A, or mixer B. Under these measurement conditions, where on‐column focusing takes place, the peak widths at half height increased with only 0.9% and 2.8% for mixer A and B, respectively.

These last results convincingly show that in‐line mixing can facilitate the injection of relatively large volumes of highly aqueous ON samples in HILIC, thereby circumventing the notorious solvent mismatch between sample and eluent that can cause serious peak deformations and loss of resolution. Notably, for achieving satisfactory ON peak shapes when injecting aqueous samples, it is essential that the mixing of the sample solvent and eluent lead to conditions that induce on‐column focusing. This way, the band broadening caused by both the strong injection solvent and the volume of the in‐line mixer can be largely mitigated, albeit not completely prevented as discussed before. In practice, this means that the gradient should start at a sufficiently high percentage of weak solvent. This was the case in the HILIC experiments reported in Figure 3, where the gradient started at 64% ACN, while the estimated percentage ACN at the elution time is 55.7% for the first eluting ON (N − 1). Figure S4 illustrates what happens when the gradient starts at 56% and 1–20 µL of the ON mixture in water is injected without and with mixer B installed. For the 1‐µL injection, the three ONs were still separated, but the resolution is clearly affected when compared with the earlier gradient conditions (Figure 3B). With mixer B in place, the resolution even further deteriorated (no baseline separation). Injection of 5 µL resulted in almost complete breakthrough, which could be largely mitigated by installing mixer B, but the achieved ON‐resolution was very poor. When 20 µL was injected with mixer B in the setup, the ON resolution was totally lost due to major peak deformation and smearing. Clearly, the starting percentage ACN of the gradient was too close to the elution ACN percentages of the ONs, and in‐line mixing of the sample solvent and eluent does not lead to on‐column focusing circumstances.

HILIC Analysis of Therapeutic ASO Using In‐Line Mixing

3.3

In order to study the feasibility of the developed in‐line mixing approach for the impurity profiling of ONs, a sample of a therapeutic ASO (denoted as ASO X) was analyzed by HILIC. For quality assurance, it is important to be able to separate and detect (very) minor impurities next to relatively large amounts of the main product. Therefore, it can be attractive to have the option to inject relatively large volumes of diluted sample to maintain favorable detection limits. To evaluate the achievable HILIC performance, a solution of ASO X in water with a concentration of 50 µg mL^−1^ was injected employing volumes of 1, 10, and 20 µL with no mixer or mixer B installed (Figure S5 and Figure 5). As expected, direct injection (no mixer) of the aqueous sample led to breakthrough of the ASO, which significantly increased in severity with rising injection volume. With mixer B in place, breakthrough and peak splitting were prevented, as demonstrated for the 20‐µL injection in Figure 5. When zooming in on the baseline near the main peak, the peaks of several (partly resolved) minor impurities can be discerned (Figure 5B). Assuming a similar detector response, the three ON impurities eluting before the main ASO represent about 0.2%, 0.6%, and 3.5%, respectively. The estimated injected concentration of the least abundant impurity is about 100 ng mL^−1^. Noticeably, these minor impurities will be difficult to detect reliably when the injection volume would have to be limited to 1 µL or smaller to prevent adverse effects of the mismatch of injection solvent and eluent. This underlines the importance of the possibility to inject relatively large volumes of aqueous samples in HILIC.

Gradient HILIC of ASO X dissolved in water (50 µg mL−1) employing in‐line mixing: (A) full chromatogram and (B) zoom of the 11–14.5 min retention interval. Injection volume, 20 µL; in‐line mixer, mixer B; solvent A, 25 mM AA in ACN–water (40:60, v/v); solvent B, 25 mM AA in ACN–water (80:20, v/v). Gradient: 65% B for 1 min, linear decrease to 35% B in 14 min, 1 min at 35% B. Flow rate: 0.3 mL min−1.

Conclusions

4

We investigated how incompatibility problems arising in HILIC during the injection of relatively large volumes of aqueous ON samples could be solved by implementing an in‐line mixer between the injector and column. Isocratic HILIC of medium‐polar nucleobases highlighted the issue of injecting aqueous samples in HILIC, which often results in strongly deformed peak shapes. Installing an in‐line mixer does not help prevent the issue, as it does not lead to conditions for on‐column focusing and only adds to volume overloading. However, for gradient HILIC of ONs, in‐line mixing was demonstrated to be a viable means for avoiding the detrimental effects of the mismatch of sample solvent and eluent. The results showed that the larger the mixer volume, the larger the volume of aqueous solvent that can be injected without causing adverse effects. A 100‐µL in‐line mixer allowed up to 40 µL of aqueous ON sample to be injected without causing breakthrough or peak splitting of the analytes. The potential of the in‐line mixing approach was illustrated by the analysis of a therapeutic ASO dissolved in water, allowing detection of minor impurities below the 1% level. In‐line mixing can be regarded a practical strategy to mitigate solvent mismatches in HILIC, providing an easy‐to‐implement alternative for techniques such as POISe and ACD.

Author Contributions

Joshka Verduin: conceptualization, methodology, formal analysis, investigation, data curation, visualization, writing – original draft. Luca Tutiš: conceptualization, methodology, formal analysis, investigation, data curation, visualization, writing – original draft. Antonia Kritsima: investigation, writing – review and editing. Andrea F. G. Gargano: conceptualization, supervision, funding acquisition, project administration, writing – review and editing. Govert W. Somsen: conceptualization, supervision, funding acquisition, project administration, writing – review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: jssc70372‐sup‐0001‐SuppMat.docx.