Ion-Pairing Hydrophilic Interaction Chromatography for Impurity Profiling of Therapeutic Phosphorothioated Oligonucleotides

Therapeutic oligonucleotides often contain closely related impurities that are difficult to separate using conventional LC methods. This study evaluates ion-pairing hydrophilic interaction chromatography (IP-HILIC) as an MS-compatible alternative for impurity profiling. A phosphorothioated, N-acetylgalactosamine-conjugated 16-mer antisense ON was used as a model compound, along with potential impurities such as shortmers, longmers, PS–PO conversions, deaminated, and nonconjugated forms.

The results show that ion-pairing reagents enhance selectivity by reducing the influence of phosphate groups on retention, highlighting the role of nucleobases and conjugates in separation. Using triethylamine acetate at pH 6.3 and elevated column temperature, IP-HILIC successfully resolved deaminated impurities from both full-length and nonconjugated ONs—something conventional MS-compatible LC methods and MS resolution alone cannot achieve. This demonstrates the strong potential of IP-HILIC for advanced oligonucleotide impurity profiling.

The original article

Ion-Pairing Hydrophilic Interaction Chromatography for Impurity Profiling of Therapeutic Phosphorothioated Oligonucleotides

Luca Tutiš*, Paul D. Ferguson, David Benstead, Adrian Clarke, Carl Heatherington, Chris Gripton, Christina J. Vanhinsbergh, Govert W. Somsen, and Andrea F. Gargano*

Anal. Chem. 2025, 97, 29, 15717–15726

https://doi.org/10.1021/acs.analchem.5c01407

licensed under CC-BY 4.0

Selected sections from the article follow. Formats and hyperlinks were adapted from the original.

Therapeutic oligonucleotides (ONs) are short strands of DNA or RNA used for the treatment of diseases that cannot be efficiently targeted by other drug modalities. (1) Antisense oligonucleotides (ASOs) represent an important class with the largest number of approved drugs. (1,2) ASOs function as gene-expression inhibitors or splicing modulators, aiming to modulate protein expression. (2) ASOs are typically synthesized by solid-phase supported synthesis (SPSS) and the sequential additions of single nucleotides to a growing chain, employing four reactions per nucleotide: deblocking (detritylation), coupling, oxidation/sulfurization, and capping. During synthesis, side reactions, reaction failures, and degradation reactions can occur. (3) This commonly results in a full-length product (FLP) comprising a considerable number of closely related impurities, including shorter and longer sequences (shortmers and longmers, respectively). Moreover, to improve the pharmacokinetic properties of ONs, chemical modifications are often introduced, (4) such as 2′-substituted riboses, methylated bases, or sulfurization of the phosphate group. The latter results in phosphorothioated (PS) groups, which significantly reduce enzymatic (nuclease) degradation in the body compared to phosphodiester ONs, (5) but also increases sample complexity as each PS introduces a chiral center, leading to 2ⁿ diastereomers for an ON with n PS groups. (6) Additionally, to aid targeted drug delivery, the ON can be covalently conjugated with a ligand to increase cellular uptake, e.g., a carbohydrate moiety, such as N-acetylgalactosamine (GalNAc), which facilitates receptor-mediated uptake of ONs specifically to hepatocytes in the liver. (7) As of 2025, there are at least nine candidate ONs in clinical trials that exploit GalNAc conjugation. (8) Clearly, the molecular complexity of the current ONs poses a significant analytical challenge.

In order to ensure quality, efficacy, and safety of therapeutic ON products, impurity profiling is imperative. Ion-pairing reversed-phase liquid chromatography (IP-RPLC) is the predominant analytical technique used for this purpose in the ON industry. (9) In IP-RPLC, positively charged ion-pair reagents (IPRs) are used to increase the retention of the negatively charged ONs on the hydrophobic stationary phase, leading to a length-based separation, as every additional nucleotide increases the apparent hydrophobicity. Impurities that comprise the same number of nucleotides as FLP are, therefore, typically not or poorly resolved. Separations of sequence variants have been reported, (10) but usually the sequence differs significantly. In this respect, the deaminated product, a sequence variant impurity that only differs on one position and could cause off-target effects, (11) is one of the most challenging impurities to separate and quantify. During deamination, a cytosine or 5-methylcytosine (MeC) of the ON is converted to uracil or thymine (T), respectively, corresponding to a loss of ammonia and a gain of water. (3) The molecular mass of this impurity differs by less than 1 Da from that of the FLP, resulting in overlapping isotope patterns in mass spectrometry (MS), hindering their reliable distinction. So far, three approaches have been described to quantify the deamination impurities in ONs. First, weak anion exchange chromatography (WAX) employing a high-pH eluent was used to resolve the deaminated products from the FLP. (12) The second, MS-based approach focused on statistical modeling of the subtle changes in the isotope pattern when the impurity coelutes with the FLP, (13) and recently, supercritical fluid chromatography has been reported to separate DA impurities from the FLP. (14) However, these approaches have limitations, such as MS incompatibility of the WAX method due to the high concentrations of involatile salts used in the eluent, the need for high-resolution and sensitive MS, which is not prevalent in a quality control setting, or it has not been applied to PS ONs. Besides the aforementioned deamination impurity, deamination of adenine and guanine (G) to hypoxanthine and xanthine, respectively, can also occur, but these processes are much slower.

Alongside IP-RPLC, strong anion exchange chromatography (AEX) and hydrophilic interaction chromatography (HILIC) have been used for impurity profiling of ON products. (15,16) AEX and HILIC methods have a different separation selectivity compared to IP-RPLC, enabling, besides the separation of shortmer and longmer impurities, the separation of oxidized phosphorothioate (PO) impurities from phosphorothioated FLPs. This difference is related to the difference in acidity (PS has a lower pK_a) and hydrophilicity (PS is more hydrophobic) of the PO impurity, which are factors that contribute to the separation of the methods, respectively. (17−20) However, separation of deaminated products also remains challenging with these LC modes. In the present work, we studied the usefulness of IP-HILIC as an alternative mode for the impurity profiling of phosphorothioated ONs. IP-HILIC has been shown to be highly useful for glycoprotein analysis, where negatively charged IPRs (usually trifluoroacetic acid) shield positive charges of the protein backbone, thereby enhancing the contribution of glycosylation to retention and providing unique resolution of protein glycoforms. (21−26) The application of IP-HILIC to ON analysis has thus far been quite limited and has not been extensively studied. (27−29) Gong has demonstrated the potential of IP-HILIC to resolve longmer impurities as well as some chemically modified ONs. (30) In this research, we investigated the use of IP-HILIC for the impurity profiling of therapeutic ONs using a representative GalNAc-conjugated, fully phosphorothioated 16-mer ASO (FLP) as a model compound. We hypothesized that the addition of positively charged IPRs to the HILIC eluent reduces the contribution of highly negatively charged phosphate backbone of ONs on retention, thereby increasing the HILIC selectivity toward the composition of nucleobases and presence of GalNAc conjugation. We especially aimed at resolving more closely related impurities comprising the same number of nucleotides as the FLP. For that, we systematically explored the effect of IPR physicochemical properties, hydrophobicity and concentration, column temperature, and eluent pH on ON retention and chromatographic performance. The potential of IP-HILIC for impurity profiling was evaluated by analyzing the shortmer, longmer, PS–PO converted, deaminated, and nonconjugated (NC) products of the FLP, and applying IP-HILIC-MS to FLP containing low levels of added impurities.

Experimental Section

LC-UV and LC-MS Instrumentation

For IP-HILIC and IP-RPLC, an Agilent 1290 Infinity II UHPLC system was used, consisting of a 1290 binary pump (G7120A) containing a V35 jet weaver mixer and a 1290 Multisampler (G7167B) with draw and eject speeds set to 100 and 400 μL/min, respectively. It contained a 1290 MCT (G7116B) with InfinityLab Quick-Connect heat exchangers (P/N: G7116-60015) and a 1290 DAD FS detector (G7117A) containing a Max-Light Cartridge Cell (10 mm; V = 1 μL). UV absorbance was measured at 260.0 nm with a bandwidth of 4.0 nm at 10 Hz. For the AEX measurements, the same modules were used, except for the pump, which was a 1260 Bioinert quaternary pump (G5654A).

LC-UV-MS measurements were performed on a 1290 Agilent infinity II UHPLC system, consisting of a binary pump (G4220A) and a HiP Sampler (G4226A). Moreover, it contained a column compartment (G1316C) and a DAD detector (G4212A). The LC system was hyphenated to a Thermo Q-Exactive Orbitrap MS equipped with a heated electrospray ionization (ESI) source set to the negative mode. The following parameters were used: ISCID, 20 eV; capillary temperature, 275 °C; spray voltage, 3.30 kV; S-lens RF, 50; sheath gas, auxiliary gas, and sweep gas flow, 60, 10, and 0 units, respectively. Full scan experiments were performed with a resolution of 140.000, 4 microscans, AGC target of 1 × 10⁶, maximum IT of 200 ms, and scan range of 400–2500 m/z.

IP-RPLC and AEX

The IP-RPLC conditions utilized were essentially those according to Rentel et al. (9) Briefly, the eluent was generated by mixing 5 mM TBuAA in ACN–water (10:90, v/v) (A) and 5 mM TBuAA in ACN–water (80:20, v/v) (B) at pH 7, both containing 1 μM EDTA. An Acquity UPLC BEH C18 column (2.1 mm × 150 mm, 1.7 μm d_p, 130 Å) (P/N: 186002353) from Waters was used. The gradient program started with a 1 min hold at 40% B, followed by a linear increase of 40–60% B in 20 min. The eluent was held at 60% B for 2 min, followed by a flush to 100% B for 2 min. The flow rate was 0.2 mL/min, and the column oven temperature was 60 °C.

For AEX, a TSKgel DNA-STAT column (4.6 mm × 100 mm, 5 μm d_p) (P/N: 0021962) from TOSOH Bioscience was used. The eluent was generated by mixing 20 mM Tris–HCl (pH 8) in ACN–water (10:90, v/v) (A) and 20 mM Tris–HCl (pH 8) with 2 M NaCl in ACN–water (10:90, v/v). The gradient program started with a 1 min hold at 35% B, followed by a linear increase of 35–55% B in 20 min and a hold of 2 min at 55% B. The flow rate was 0.4 mL/min, and the column oven temperature was 25 °C.

HILIC and IP-HILIC

All (IP-)HILIC experiments were performed on an Acquity UPLC BEH amide column (2.1 mm × 150 mm, 1.7 μm d_p, 130 Å) (P/N: 186004802) except for the evaluation of triethylamine acetate (TEAA) concentration (Figure 4), which was obtained using a Premier BEH Amide with the same column dimensions. The IPRs DEA, TEA, TPA, and TBuA, as well as AA were tested at a concentration of 15 mM in eluent A and B, where the pH of the aqueous part of the eluents was adjusted with acetic acid or ammonia to pH 7 for the IPRs and AA, respectively, resulting in the alkylamine salts (e.g., diethylamine acetate (DEAA)). Eluent A consisted of ACN–water (20:80, v/v), and eluent B consisted of ACN–water (90:10, v/v). The gradient program started with a 1 min hold at 100% B, followed by a linear decrease of B to 30% from 1 to 50 min, and ending with a 3 min hold at 30% B. The flow rate was 0.2 mL/min. For all ON samples, the injection volume was 1 μL unless specified differently. The poly(dT) ladder solution was injected at a volume of 5 or 10 μL. The column oven temperature was 45 °C.

For the evaluation of the effect of TEA concentration (5–100 mM), a stock solution of 1 M TEAA at pH 7 in water was used. Eluent A was obtained by diluting the TEAA solution to the desired concentration using Milli-Q water. To prepare eluent B, we added different volumes of TEAA stock solution (0.5–10% v/v) to ACN to achieve the desired IP concentration. This resulted in a change of water content in B; to adjust for this the gradient started with a 1 min hold at different % B depending on the ion concentration: 81, 82, 83, 85 (5, 10, 25, 50 mM, respectively, to normalize to the 100 mM TEAA gradient of 90–60% MPB); followed by a linear decrease to 54, 55, 56, 57% B from 1 to 21 min, and a hold at the highest % B for 2 min. The flow rate was 0.2 mL/min. In the analysis of the effect of pH (4.8–9) of the eluent and column temperature (5–80 °C), eluent A consisted of Milli-Q water with TEAA and eluent B consisted of ACN with TEAA. The gradient started with a 1 min hold at 80% B, followed by a linear decrease to 66.7% B from 1 to 21 min, and a hold at 66.7% B for 2 min. The flow rate was 0.2 mL/min. Final conditions for the analysis of mixtures of FLP with impurities were 25 mM TEA in both eluents A and B, an eluent pH of 6.3, and column temperature of 80 °C, applying an 80–66.7% B gradient as mentioned above.

Data Processing

LC-UV data was processed using MATLAB 2024b. Peak widths were extracted from Agilent Openlab CDS software (version 2.7). MS data was processed with Thermo Scientific Freestyle version 1.8.65.0. Mass deconvolution was performed with UniDec. (31) The MS data presented are available at the following Zenodo Repository link: https://zenodo.org/records/14927266. Log P and pK_a values were obtained from PubChem. (32)

Results and Discussion

In HILIC, the retention of analytes is primarily based on polar interactions with the stationary phase, largely originating from the hydrophilic moieties of the retained molecule. HILIC elution is typically achieved by using a gradient starting with high amounts of ACN and gradually increasing the water percentage, where the (buffer) salt concentration is kept constant throughout the run and prepared at a specific pH. HILIC is MS-compatible and increasingly used for ON characterization as its selectivity differs slightly from IP-RPLC, and in principle, no IPRs are needed in the eluent. (18−20,33−35) Given the highly hydrophilic nature of ONs, the percentage of water used for elution ranges between 30 and 70%, suggesting a limited contribution of hydrophilic partitioning to the HILIC retention mechanism. The retention of ONs is most probably driven by hydrogen bonding and ionic interactions. The highly polar, negatively charged phosphate groups of ONs add strongly to retention, allowing separation of shortmer, longmer, and PO impurities from a phosphorothioated FLP on amide-based stationary phases. The resulting separation is relatively similar to IP-RPLC, where the number of phosphate moieties on the ON (that interact with IPRs in IP-RPLC) drives the retention.

In our work, we aimed to develop an HILIC method that can resolve ONs based on their nucleobase composition and chemical modifications by reducing the influence of the phosphate backbone on retention. To accomplish this, we introduced positively charged IPRs into the mobile phase (IP-HILIC). Cationic IPRs are known to interact with the negatively charged ONs, neutralizing their charge and altering their interaction with the HILIC stationary phase. Under these conditions, the HILIC retention will be less dependent on the phosphate backbone (and thus the number of nucleotides) and, resultingly, will be largely driven by the nucleobase composition and conjugated groups of the ON, unlike IP-RPLC, where the IPRs significantly increase retention due to its hydrophobicity. This process is schematically visualized in Figure 1 and more extensively in Figure S1.

Anal. Chem. 2025, 97, 29, 15717–15726: Figure 1. Schematic representation of the effect of IPRs on the eluent for ONs separation in RPLC and HILIC.

IP-HILIC: Effect of Ion-Pairing Reagent Type on Retention and Selectivity, and Comparison with IP-RPLC and AEX

To assess the effect of ion-pairing on ON retention in HILIC, IPRs of different physicochemical properties and hydrophobicity (DEAA, TEAA, TPAA, and TBuAA) were tested by analyzing a poly(dT) ladder applying a linear gradient from 90% ACN to 41% ACN (i.e., 100 to 30% B) on a BEH amide column. This sample does not contain any PS modifications and therefore no diastereomers, allowing us to evaluate the effect of IPRs in the eluent on ON peak widths and separation resolution. Using a concentration of 15 mM of IPR in the eluent, (partial) separation of the five ONs was obtained (Figure 2), where the HILIC retention of the poly(dT) ONs decreased with increasing IPR hydrophobicity. Compared with an eluent containing no IPR but 15 mM AA instead, a retention-time reduction of approximately 30 min was observed for the 15-mer poly(dT) when using TBuAA in the eluent. The interaction of the positively charged IPRs with the negatively charged phosphate groups decreases the apparent hydrophilicity of the ONs and, thus, their retention. Along these lines, the retention increases with ON size as every additional nucleotide adds a phosphate group, increasing the overall hydrophilicity of the ON.

Anal. Chem. 2025, 97, 29, 15717–15726: Figure 2. HILIC-UV of a poly(dT) ladder comprising 15-, 20-, 25-, 30-, and 35-mer using an eluent containing 15 mM of AA, DEA, TEA, TPA, or TBuA (pH 7) using a linear gradient from 100 to 30% B in 50 min on a BEH amide column at 45 °C, with eluents A and B consisting of 90:10 and 20:80 ACN–water (v/v), respectively. UV absorbance detection was performed at 260 nm, and baselines were subtracted using blank measurements (see Figure S2). For other conditions, see the Experimental Section.

IP-HILIC-MS for ON Impurity Profiling

For testing the potential of the optimized IP-HILIC method for impurity profiling, a mixture of FLP and N – 1, N + 1, PO, DA, and NC was prepared with the impurities at 2% in weight relative to FLP. The mixture was analyzed using IP-HILIC-UV-MS using an eluent comprising 25 mM TEAA (pH 6.3) and a column temperature of 80 °C. Figure 5 (top left) shows the obtained total-ion (TIC) and extracted-ion chromatograms (EICs) for each ON species. The TIC is dominated by the FLP signal, but in the EICs, the low-level impurities can be detected at suitable signal-to-noise ratios, confirming the good MS compatibility of the IP-HILIC method, despite the presence of TEAA as IPR. The detected impurities can now be selectively assigned according to their masses recorded with MS. The obtained impurity profile corresponds to what could be expected from the exploring IP-HILIC-UV experiments (Figure S14): the NC, N – 1, and DA impurities elute before the FLP, and the PO and N + 1 are eluting after FLP, indicating satisfactory LC-MS interfacing. The DA and N – 1 impurities are well separated from the FLP, coelute together. Nevertheless, as their masses differ significantly, they can still be adequately differentiated by MS.

Anal. Chem. 2025, 97, 29, 15717–15726: Figure 5. Left: IP-HILIC-UV-MS of (top) mixture of FLP and impurities at 2% and (bottom) mixture of NC with DA-NC at 2% total-ion chromatograms (TIC) and extracted-ion chromatograms (EICs; ±5 ppm m/z) for 1696.32 (FLP), 1611.07 (N – 1), 1783.58 (N + 1), 1694.33 (PO), 1697.32 (FLP and DA), 1719.52 (NC), and 1718.52 (NC and DA-NC) for the mixtures are plotted. Right: mass spectra showing the isotope patterns of the (A, B) [M – 4H]4– ions and (C, D) [M – 3H]3– ions observed in the peaks highlighted in the corresponding TICs. See the Experimental Section for the experimental conditions.

Conclusions

This study focused on the use of IP-HILIC-MS for the separation and assignment of impurities of phosphorothioated ONs comprising the same number of nucleotides as the FLP. We systematically studied the influence of IP-HILIC separation parameters, such as type and concentration of IPR, eluent pH, and column temperature, using a poly(dT) ladder and a model FLP (GalNAc-conjugated PS 16-mer ASO) with several of its related impurities (N – 1, N + 1, PO, DA, and NC). We showed that IPRs in the HILIC eluent (i) reduce the relative contribution of the highly polar phosphate moieties to ON retention, (ii) alter the HILIC separation selectivity inducing resolution of DA, PO, and NC impurities from the FLP while maintaining separation of N – 1 and N + 1 from the FLP, and (iii) reduce peak widths of phosphorothioated ONs, most probably due to reduction of diastereomer separation. Optimal resolution for the model FLP and its related impurities was obtained using an IP-HILIC eluent of pH 6.3 with 25 mM TEA, and a column temperature of 80 °C. The feasibility of the optimized method for impurity profiling was indicated by IP-HILIC-MS analysis of the model FLP containing several impurities at the 2% level. Comparison with current IP-RPLC, AEX, and HILIC methods showed that only IP-HILIC was able to resolve the DA product, an essential but difficult to detect impurity, from the FLP. Moreover, the selectivity and sensitivity IP-HILIC-MS showed the suitability of revealing a large number of minor impurities in an unpurified standard ON. Therefore, we believe that IP-HILIC-MS shows interesting potential for impurity profiling of therapeutic ONs.