Skip to main content
Bioanalysis logoLink to Bioanalysis
. 2024 Jun 6;16(13):651–667. doi: 10.1080/17576180.2024.2350266

Comparison of multiple bioanalytical assay platforms for the quantitation of siRNA therapeutics

Karan Agrawal a,*, Laurelle K Calliste a, Shaofei Ji b, Shengsheng Xu a, Stephen A Ayers a, Wenying Jian a
PMCID: PMC11389733  PMID: 39254503

Abstract

Aim: Oligonucleotide therapeutics can be quantified using various bioanalytical methods, and these methods have been compared extensively. However, few comparisons exist where the same analyte is evaluated by multiple assay platforms.

Materials & methods: Hybrid LC-MS, SPE-LC-MS, HELISA and SL-RT-qPCR methods were developed for an siRNA analyte, and samples from a pharmacokinetic study were analyzed by all four methods.

Results: All assay platforms provided comparable data, though higher concentrations were observed using the non-LC-MS assays. Hybrid LC-MS and SL-RT-qPCR were the most sensitive methodologies, and SL-RT-qPCR and HELISA demonstrated the highest throughput.

Conclusion: Each assay platform is suitable for oligonucleotide bioanalysis, and the ultimate choice of methodology will depend on the prioritization of needs such as sensitivity, specificity and throughput.

Keywords: : capture probe, hybridization, LC-MS, ligand-binding assay, locked nucleic acids, oligonucleotides, short interfering RNA (siRNA), stem loop-reverse transcription-quantitative PCR

Plain language summary

Article highlights.

Introduction

  • Oligonucleotide therapeutics can be quantified using various bioanalytical methods, with LC-MS and hybridization ELISA (HELISA) approaches commonly used to support regulatory bioanalytical submissions.

  • Several comparisons between oligonucleotide bioanalytical assays have been published, but few evaluate the same analyte by multiple assay platforms. Therefore, it is difficult to identify the extent to which assays differ due to factors inherent to the assay platform versus factors inherent to the compound being analyzed.

Materials & methods

  • To better understand the strengths and limitations of common oligonucleotide assay platforms, hybrid LC-MS, solid phase extraction-LC-MS (SPE-LC-MS), HELISA and stem loop-reverse transcription-quantitative PCR (SL-RT-qPCR) workflows were developed for an siRNA analyte, and samples from a pharmacokinetic study were analyzed by all four methods.

Results & discussion

  • All assay platforms evaluated provided comparable data for the study samples, though HELISA and SL-RT-qPCR tended to generate higher observed concentrations relative to the LC-MS assays, possibly due to quantification of both the parent analyte and its metabolites, which indicates a lack of specificity.

  • Hybrid LC-MS and SL-RT-qPCR demonstrated the highest sensitivity, and SL-RT-qPCR and HELISA demonstrated the highest throughput.

  • Evaluation of assay performances indicate that all assay formats could generally be validated to standards necessary to support regulatory bioanalytical submissions.

Conclusions

  • Each assay platform evaluated is suitable for oligonucleotide bioanalysis, and the ultimate choice of methodology will depend on which factors among sensitivity, specificity, and throughput need to be prioritized.

1. Background

Interest in developing oligonucleotides as therapeutics has increased in the last decade, with nine antisense oligonucleotides (ASOs) and five siRNA molecules having gained regulatory approval in this period [1]. These therapeutics act by binding to complementary endogenous RNA sequences, and either modulating gene expression by a variety of mechanisms or triggering RNA degradation through endonucleolytic cleavage [1]. To understand the efficacy and safety of these novel therapeutics and enable these compounds to progress through the drug discovery and development process, accurate and reliable bioanalytical methods are needed to quantify their concentrations during nonclinical and clinical studies.

LC-MS assays tend to be the preferred bioanalytical approach for the quantification of siRNA therapeutics, with four out of the five currently approved siRNA therapeutics being supported by LC-MS assays for regulatory submissions [2]. This trend appears to be reflected in the broader oligonucleotide bioanalysis community, with a 2019 survey finding that ∼75% of respondents preferred to use LC-MS for oligonucleotide analysis [3]. From a sample preparation standpoint, most of these LC-MS assays tend to use either solid-phase extraction (SPE) or liquid-liquid extraction (LLE). Other common bioanalytical platforms for oligonucleotide bioanalysis include hybridization ELISA (HELISA), stem-loop reverse transcription quantitative PCR (SL-RT-qPCR), and LC-Fluorescence, all of which use analyte-specific probes or primers to isolate or enrich the target oligonucleotide strand [4–10]. Of these alternative workflows, HELISA and LC-Fluorescence have been used to support regulatory submissions for oligonucleotide therapeutics [2,11,12]. More recently, the hybridization-based sample preparation strategy has been paired with LC-MS analysis for oligonucleotide bioanalysis [13–21], but to the best of our knowledge, this workflow has not been employed as part of a regulatory submission.

The relative advantages and disadvantages of these bioanalytical workflows have been extensively reviewed elsewhere [4–10]. Briefly, SPE- or LLE-LC-MS assays tend to use generic reagents, require shorter method development time, and have high specificity with the potential for metabolite identification, but have comparatively poorer sensitivity and throughput [4–10]. On the other hand, HELISA and qPCR assays tend to have comparatively greater sensitivity and throughput, but require more extensive method development and analyte-specific reagents, and are generally not capable of discriminating between the parent oligonucleotide and its truncated metabolites [4–10]. Hybrid LC-MS assays, tend to have lower limits of quantitation (LLOQs) ≤1 ng/ml [13–21], which approaches those of HELISA or qPCR assays. However, this improvement in sensitivity comes at the cost of a need for analyte-specific reagents and potentially increased method development time, though these assays retain the throughput, specificity and metabolite identification potential of other LC-MS/MS assays [13–21]. It should be noted that most of these comparisons do not evaluate multiple assay platforms developed for the same analyte. As a result, it is difficult to identify the extent to which assays differ due to factors inherent to the assay platform versus factors inherent to the compound being analyzed, leading to a potentially non-representative comparison.

To the best of our knowledge, only two studies have compared the same oligonucleotide analytes by multiple bioanalytical methods. Dillen et al. analyzed a 13-mer ASO using hybrid LC-MS, SPE-LC-MS and HELISA [13], and Sips et al. analyzed two 16-mer ASOs using hybrid LC-MS and SPE-LC-MS [15]. Both studies demonstrated comparability between the various techniques for sensitivity, robustness and analyte recovery using spiked samples [13,15], and Dillen et al. also showed comparability in pharmacokinetic profiles when evaluating in vivo samples by all three techniques [13]. However, no such comparisons exist for siRNA analytes. Additionally, no comparisons have been made between LC-MS and/or HELISA workflows and qPCR workflows.

In this manuscript, we present the development of hybrid LC-MS, SL-RT-qPCR, HELISA and SPE-LC-MS assays for a 21-mer lipid-conjugated siRNA therapeutic (SIR-2), and evaluate samples collected from a pre-clinical pharmacokinetic study by all four techniques. Assay performances are evaluated to understand differences in sensitivity, robustness and the potential for each assay to be validated in a regulatory setting. Additionally, pharmacokinetic profiles are evaluated to understand the quantitative performance of each assay for in vivo samples. The overall study provides greater insight into the comparability of the four most common bioanalytical techniques for oligonucleotide therapeutics.

2. Materials & methods

2.1. Chemicals & reagents

SIR-2 (10 mg/ml) and ISTD-3 (1 mg/ml) reference materials were generated in-house at purities ≥90% and provided as pre-dissolved solutions in phosphate-buffered saline (PBS), pH 7.4. ISTD-3 is a lipid-conjugated siRNA molecule unrelated to SIR-2, and is used as an analog internal standard for the LC-MS-based assays. Locked nucleic acid (LNA) probes against SIR-2 used in the Hybrid LC-MS and HELISA workflows were custom synthesized by Qiagen, LLC (MD, USA), and provided at purity ≥85%. Sheep anti-digoxigenin antibody was purchased from Roche Diagnostics (IN, USA) and was ruthenium-labeled in-house. Control dipotassium ethylenediaminetetraacetic acid (K2EDTA) plasma was purchased from BioIVT (NY, USA). Type 1 water used in this study was deionized and filtered through a Milli-Q® Academic system (MilliporeSigma, MA, USA) prior to use. All reagents used for the SL-RT-qPCR assay were custom synthesized by or part of commercially available kits from ThermoFisher Scientific (MA, USA).

Other reagents purchased included: ammonium bicarbonate, bovine serum albumin (BSA), cysteine, EDTA, sodium chloride, dibasic sodium phosphate, Triton X-100 and Tween20 from Sigma Aldrich (MO, USA); Acetonitrile and methanol from MilliporeSigma; 2-chloroacetamide, Dynabeads™ MyOne™ Streptavidin C1 magnetic beads, nuclease-free water, proteinase K solution (20 mg/ml) and tris(2-carboxyethyl)phosphine (TCEP) from ThermoFisher Scientific; Tris(hydroxymethyl)aminomethane (Tris) from Alfa Aesar (MA, USA); N-Dimethylbutylamine (DMBA) from TCI Chemicals (OR, USA); and 10× Dulbecco's PBS, hexafluoro-2-propanol (HFIP) and tris-buffered saline from J.T. Baker (PA, USA).

Mobile phases for the LC-MS-based assays consisted of 0.1% (v/v) DMBA, 10 μM EDTA and 0.5% (v/v) HFIP in water (mobile phase A) and 0.1% (v/v) DMBA, 10 μM EDTA and 0.5% (v/v) HFIP in methanol (mobile phase B). Mobile phases were not subjected to pH adjustment, and were stored at room temperature and used within 48 hr of preparation to avoid loss of sensitivity associated with mobile phase aging [13,16].

2.2. Stock & working solutions

Pre-dissolved SIR-2 reference material was directly used as a stock solution for all four workflows. A 10 μg/ml sub-stock was prepared in PBS from the stock solution, and calibrators and quality control (QC) samples were prepared in plasma from independent dilutions of this sub-stock solution. Calibration standards and QC samples were prepared fresh in Eppendorf DNA LoBind microcentrifuge tubes (NY, USA), and disposed after use. The SIR-2 reference material and 10 μg/ml sub-stock solutions were stored at -20°C.

Pre-dissolved ISTD-3 reference material was directly used as a stock solution for the LC-MS-based workflows. A 10 μg/ml sub-stock was prepared in PBS from the stock solution, and the working ISTD-3 solutions were prepared in LC-MS mobile phase B from dilutions of this sub-stock. All internal standard solutions were prepared in Eppendorf DNA LoBind microcentrifuge tubes, and the stock and sub-stock solutions were stored at -20°C while the working solutions were disposed after use.

2.3. Sample preparation & analysis

2.3.1. Hybrid LC-MS

SIR-2 was isolated from plasma according to our previously published generic internal standard hybrid LC-MS protocol [21], except using a 50 μl sample aliquot in place of the 10 μl sample aliquot. All other aspects of the protocol remained the same. Following elution of the SIR-2 antisense strand, 200 ng/ml ISTD-3 working solution (10 μl/well) was added to samples receiving internal standard, and the samples were centrifuged at 3000 × g for 5 min at ambient conditions. The supernatant was transferred to a clean Eppendorf DNA LoBind deep-well 96-well plate for injection.

LC-MS/MS experiments were performed on a Shimadzu Nexera 40 series binary pump LC system (Columbia, MD, USA) coupled to an AB Sciex TripleQuad 7500 mass spectrometer (MA, USA). Chromatographic separation was achieved using a ThermoFisher Scientific DNAPac™ RP 2.1 × 50 mm, 4 μm column with gradient elution at a flow rate of 0.250 ml/min for 9.00 min. The gradient (% mobile phase B) used was: initial – 15%, 0.75 min – 15%, 4.00 min – 95%, 6.00 min – 95% and 6.20 min – 15%. A switching valve was employed such that the post-column flow was diverted to waste before 1.00 min and after 5.00 min. The sample manager and column oven were maintained at 8 and 75°C, respectively. MS/MS was performed by electrospray ionization operated in negative ion mode using multiple reaction monitoring experiments (dwell time 50 ms) with mass transitions of 780.4 >845.0 (SIR-2) and 697.4 >593.3 (ISTD-3). The precursor ions for SIR-2 and ISTD-3 represent the -9 and -10 charge states, respectively. Mass spectrometric parameters included an ion spray voltage of -3500 V; source temperature of 500°C; collision, curtain, nebulizer and auxiliary gases set at 9 arbitrary units, 40 psi, 35 psi and 50 psi, respectively; entrance potential of -10 V; collision cell exit potential of -25 V; and collision energies of -25 V (SIR-2) and -32 V (ISTD-3).

Data were processed using SciexOS version 3.1.0 (Sciex), and a quadratic regression model with 1/x2 weighting was applied to the concentration-response plot, where x was the ratio of analyte to internal standard concentration and y was the ratio of analyte to internal standard mass spectrometric response.

2.3.2. SL-RT-qPCR

For sample analysis by SL-RT-qPCR, plasma samples initially underwent a 100× dilution using 0.25% (v/v) Triton X-100 in PBS. A 10 μl aliquot of the diluted samples was transferred to a 96-well PCR plate (VWR, Radnor, PA, USA), and 6 μl of 5× stem-loop RT primer was added. The PCR plate was transferred to a Bio-Rad T100 thermal cycler (Hercules, CA, USA) and incubated at 95°C for 5 min, 80°C for 2 min, 70°C for 2 min, 60°C for 2 min, 45°C for 2 min, and then held at 4°C for at least 10 min to anneal the primer to the antisense strand of SIR-2. An 8 μl aliquot of the annealed product was transferred to a clean 96-well PCR plate containing 7 μl/well of RT enzyme mix that had been prepared according to manufacturer's instructions [22]. The PCR plate was then transferred to the thermal cycler and incubated at 16°C for 30 min, 42°C for 1 h, 85°C for 5 min, and then held at 4°C for at least 15 min, after which 75 μl of nuclease-free water was added to all samples. An 8.4 μl aliquot of this diluted reverse transcribed product was transferred to a clean 96-well PCR plate containing 16.6 μl/well of qPCR reaction mix that had been prepared according to manufacturer's instructions [22]. After mixing, two 10 μl aliquots of this sample were transferred to adjacent wells of a 384-well plate (ThermoFisher Scientific), and real-time qPCR analysis was performed on a QuantStudio 12K Flex instrument (ThermoFisher Scientific). For the real-time qPCR analysis, samples were initially held at 95°C for 20 s, after which 40 temperature cycles were performed where the temperature was held at 95°C for 1 s and then reduced to 60°C and held for 20 s.

Data were processed using Excel for Microsoft 365 (WA, USA), and an unweighted linear regression model was applied to the concentration-response plot, where x was log10-transformed analyte concentration and y was the Ct value (i.e., the number of cycles needed for the signal to exceed a threshold during the qPCR analysis).

2.3.3. HELISA

The HELISA workflow used in this study was adapted from a previously published method [23]. Plasma samples underwent a 10× dilution using Low-EDTA Tris-EDTA, pH 8.0 buffer (VWR) and a 50 μl aliquot of the diluted samples was added to a 96-well PCR plate (VWR). Biotinylated capture probes and digoxigenin-labeled detection probes were prepared at 10 nM each in a buffer consisting of 10 mM Tris, 500 mM sodium chloride, 1 mM EDTA and 0.05% (v/v) Tween20 in water and 50 μl of this probe mixture was added to the PCR plate containing the diluted samples. Samples were incubated in an Eppendorf ThermoMixer C at 95°C for 15 min, 40°C for 30 min and 12°C for at least 40 min to hybridize the probes to the antisense strand of SIR-2. A 45 μl aliquot of the hybridized sample was transferred to an MSD Gold 96-well Streptavidin SECTOR plate (Meso Scale Diagnostics, MD, USA) that had been pre-washed with 0.1% (v/v) 2-chloroacetamide and 0.05% (v/v) Tween-20 in 1× Dulbecco's calcium and magnesium free PBS, pH 6.9. The MSD plate was incubated on a shaking platform for 30 min at ambient conditions, washed and then incubated for 1 h following the addition of 50 μl of 0.5 μg/ml ruthenium labeled anti-digoxigenin antibody in tris-buffered saline containing 1% (v/v) BSA, 150 mM Tris, 128 mM sodium chloride, 0.1% (v/v) 2-chloroacetamide and 0.5% (v/v) Tween-20. After a final wash, 150 μl of 1× MSD Read Buffer T (Meso Scale Diagnostics) was added and the plate was read on an MSD Sector S600 instrument (Meso Scale Diagnostics) within 5 min of adding the read buffer.

Data analysis was performed using Watson LIMS 7.5 (ThermoFisher Scientific), and a four-parameter logistic (4PL Marquardt) regression model with 1/y2 weighting was applied to the concentration–response plot, where x was the analyte concentration and y was the electrochemiluminescent signal intensity.

2.3.4. SPE LC-MS

The SPE workflow used in this study was adapted from a previously published method [15]. Briefly, a Clarity® OTX 100 mg/well 96-well SPE plate (Phenomenex, CA, USA) was conditioned with 1 ml of methanol and then equilibrated with 1 ml of 50 mM dibasic sodium phosphate in water, pH 5.5. A 470 μl sample aliquot consisting of 50 μl plasma, 20 μl of quenching solution (80 mM Tris, 40 mM EDTA and 4 μg/ml ISTD-3 in water), 140 μl of Clarity® OTX Lysis-Loading Buffer v2.0 (Phenomenex) and 260 μl of the equilibration buffer was then added to the equilibrated SPE plate. To minimize non-analyte matrix components from the sample, the SPE plate was washed thrice with 1 ml of 50 mM dibasic sodium phosphate in water, pH 5.5:acetonitrile (1:1, v/v), once with 0.5 ml of water and once with 0.5 ml of 100 mM ammonium bicarbonate in water, pH 10. SIR-2 was eluted by three 0.4 ml additions of 100 mM ammonium bicarbonate in water, pH 10:acetonitrile (1:1, v/v), after which 20 μl/well of 10 mg/ml cysteine in the elution buffer was added to each sample. The eluate was evaporated to dryness under nitrogen, reconstituted in 200 μl of mobile phase A:mobile phase B:plasma (95:5:0.05, v/v/v), and transferred to a clean Eppendorf DNA LoBind deep-well 96-well plate for injection.

LC-MS/MS instrumentation and conditions used for the analysis of the SPE LC-MS samples were the same as those used for the Hybrid LC-MS samples. The only changes were the use of a Waters XBridge C18 2.1 × 50 mm, 2.7 μm column (MA, USA) to achieve chromatographic separation, and a 12.0 min LC gradient. The gradient (% mobile phase B) used was: initial – 15%, 0.75 min – 15%, 4.00 min – 95%, 5.00 min – 95%, 5.20 min – 15%, 5.40 min – 15%, 5.60 min – 95%, 5.80 min – 95%, 6.00 min – 15%, 6.20 min – 15%, 6.40 min – 95%, 6.60 min – 95%, 6.80 min – 15%, 7.00 min – 15%, 7.20 min – 95%, 7.40 min – 95%, 7.60 min – 15%. The LC flow rate was 0.25 ml/min between 0.00 and 4.50 min and between 7.80 and 12.00 min, and 0.6 ml/min between 4.60 and 7.60 min. All other instrument and data processing conditions remain the same as previously described for the Hybrid LC-MS method.

2.4. Assay performance evaluation

The performances of each of the four assays for the quantitation of SIR-2 were evaluated using assessments of precision and accuracy, selectivity, and dilution integrity. Additionally, in the case of the LC-MS-based assays, recovery and matrix effects were assessed.

For precision and accuracy evaluation, eight non-zero calibration standards and at least four levels of QC samples were freshly prepared in control plasma. The calibration curve ranges and QC sample concentrations for each assay are listed in Table 1. For the LC-MS-based assays, two replicates of the calibration curve (one replicate each at the start and end of the analytical run) and six replicates of each QC sample were analyzed, and for the SL-RT-qPCR and HELISA assays, a single replicate of the calibration curve and six replicates of each QC sample were analyzed. Dilution integrity was assessed by spiking control plasma with SIR-2 at a concentration greater than the upper limit of quantitation for each assay and diluting 100× with control plasma prior to analysis (n = 3). Selectivity was established by analyzing double-blank matrix samples from six different lots of control plasma (n = 1 per lot), and evaluation of the same lots when spiked at the respective assay low QC concentrations with SIR-2 (n = 3 per lot).

Table 1.

Calibration curve ranges and quality control sample concentrations for the bioanalytical assays evaluated for the quantification of SIR-2.

Technique Hybrid LC-MS SL-RT-qPCR HELISA SPE-LC-MS
Calibration curve range (ng/ml) 0.200–200 0.128–2000 1.00–250 5.00–500
LLOQ QC concentration (ng/ml) 0.200 0.128 1.00 5.00
Low QC concentration (ng/ml) 0.600 0.500 3.00 15.0
Low-medium QC concentration (ng/ml) 8.00 2.50 N/A N/A
Medium QC concentration (ng/ml) 80.0 50.0 30.0 200
High QC concentration (ng/ml) 160 1000 200 400
Dilution QC concentration (ng/ml) 10,000 5000 10,000 10,000

HELISA: Hybridization ELISA; LLOQ: Lower limit of quantitation; N/A: Not applicable; QC: Quality control; SL-RT-qPCR: Stem-loop reverse transcription quantitative PCR; SPE: Solid phase extraction.

Samples to evaluate recovery and matrix effects were prepared by post-spiking double-blank matrix or water samples, respectively that had been subjected to the entire sample preparation process, with solutions containing double-stranded SIR-2 that represented the concentration of an extracted low, medium, or high QC sample at 100% recovery (n = 3 per concentration). Recovery was determined by comparing the average analyte responses for extracted low, medium or high QC samples in the analytical run (n = 6 per concentration) to their respective post-spiked matrix samples. Matrix effects were evaluated as matrix factors by comparing the average analyte responses for the post-extraction spiked matrix samples to the respective post-extraction spiked water samples.

2.5. Analysis of pre-clinical plasma samples

Plasma samples obtained from cynomolgus monkeys dosed parenterally at one of four doses of SIR-2 (n = 1 selected for analysis from each dose) were analyzed by all four bioanalytical assays. Plasma was collected pre-dose (0 h) and at 14 timepoints post-dose and stored at -80°C until analysis. Samples were thawed once at room temperature, and aliquots for all four bioanalytical assays were taken within 2 h of thawing. Both LC-MS-based assays used 50 μl aliquots for analysis, whereas the SL-RT-qPCR and HELISA assays each used 10 μl aliquots that were further diluted prior to analysis. Plasma concentration profiles for SIR-2 were generated for each bioanalytical assay. Using the Hybrid LC-MS data as a reference, concentrations at each time for each dose group obtained by the other assays were compared using Spearman's correlation and Bland-Altman ratio analysis. All data analyses were performed using GraphPad Prism 9 (CA, USA). Ethical approval for this study was obtained from the Institutional Animal Care and Use Committee of Charles River-Nevada (Protocol #20314229), and reviewed and approved by Johnson & Johnson.

3. Results & discussion

Bioanalytical assays for oligonucleotide therapeutics are constantly evolving, and multiple assay platforms currently exist that can quantify these analytes with varying degrees of sensitivity and specificity. The chromatography- and/or mass spectrometry-based assays such as SPE- and LLE-LC-MS tend to be more specific, capable of distinguishing between the parent molecule and its metabolites, but typically lack sensitivity [4–10]. On the other hand, luminescence-based assays such as HELISA or SL-RT-qPCR tend to be more sensitive, but frequently provide a ‘total’ concentration that is the sum of the parent molecule and its truncated metabolites [4–10]. Pairing hybridization-based sample preparation with chromatographic separation, as has been done for hybrid LC-MS and LC-Fluorescence assays, can lead to assays with both high sensitivity and high specificity [4–10], but these assays typically have low throughput and require extensive method development. For the purposes of supporting late-stage drug development, SPE- and LLE-LC-MS, LC-Fluorescence and HELISA assays have all been used to support regulatory submissions for ASO and siRNA therapeutics [2,11,12], whereas SL-RT-qPCR and hybrid LC-MS assays have not yet been applied in this context. In fact, no current regulatory guidance is available for the qualification or validation of qPCR assays or hybrid LC-MS assays, though in the latter case the regulatory acceptance criteria can be extrapolated from guidance for both chromatographic and ligand-binding assays [24].

Despite the extensive comparison of each assay platform mentioned above to one or more of the other workflows, it is rare to find a comparison that evaluates the same analyte by more than one technique. To the best of our knowledge, only two publications exist that compare ASO analytes by more than one analytical technique, and only one of these publications evaluates in vivo samples [13,15]. Therefore, it is difficult to truly compare the various assay platforms and understand how data generated by one workflow can be related to data generated by an orthogonal technique. This is particularly the case with siRNA analytes, where no comparisons of the same analyte by multiple assay platforms have been performed. We therefore attempted to assess the performances of hybrid LC-MS, SL-RT-qPCR, HELISA and SPE-LC-MS assays by quantifying a lipid-conjugated siRNA molecule in plasma, and subsequently analyzed pharmacokinetic samples obtained from a pre-clinical species by all four techniques. A high-level overview of each of the four workflows is presented in Figure 1, and comparisons of metrics for each assay platform based on published methods are summarized in Table 2.

Figure 1.

Figure 1.

Workflows for the four bioanalytical methods used to analyze SIR-2.

Table 2.

Comparison of major features of common bioanalytical methods for the analysis of siRNA analytes in plasma.

  Hybrid LC-MS SL-RT-qPCR HELISA SPE-LC-MS
Sample preparation format Hybridization Hybridization and amplification Hybridization Solid phase extraction
Detection format MS/MS Fluorescence qPCR Electrochemiluminescence MS/MS or HRMS
Sample aliquot volume range 10–50 μl Unspecified Unspecified 20–50 μl
Typical sensitivity 0.5–25 ng/ml 0.007 ng/ml 0.3–4 ng/ml 10–20 ng/ml
Typical dynamic range 400–1000× 9000000× 200–600× 100–1000×
Throughput Low High High Low
Metabolite differentiation from parent siRNA Yes No No Yes
Custom reagent needs Analyte-specific probe Analyte-specific primers and probe Analyte-specific probes (capture and detection) None
Assay development time Days – weeks Days – weeks Weeks – months Days
Challenges Probe design, chromatography, throughput Primer design, specificity Probe design, specificity Chromatography, sensitivity, ruggedness, throughput
Regulatory guidance available? No§ No Yes (ligand-binding assay criteria) Yes (chromatographic assay criteria)
Representative published workflows [19,21] [25] [23,26] [15,29–31]

The actual sample aliquot volume used is unspecified by the representative publication. However, the sample was diluted at least 100× prior to analysis, and 20 μl of the diluted sample was used.

The actual sample aliquot volume used is unspecified by the representative publications. However, in the case of both methods, the sample was diluted 10–20× prior to analysis, and 50 μl of the diluted sample was used.

§

No regulatory guidance is available specifically for hybrid LC-MS assays. However, guidance for this type of assay is usually extrapolated from available guidance for both chromatographic and ligand-binding assays.

HELISA: Hybridization ELISA; SL-RT-qPCR: Stem-loop reverse transcription quantitative PCR; SPE: Solid-phase extraction.

3.1. Bioanalytical assay development

3.1.1. Hybrid LC-MS

The method development for the Hybrid LC-MS workflow, including capture probe design, optimization of the sample preparation conditions and evaluation of generic versus analyte-specific internal standards is described in our previous publication on hybrid LC-MS assays for oligonucleotides [21]. In the present assay, the same capture probe was utilized, and apart from the sample aliquot volume, the sample preparation process was the same. Based on our findings that the addition of a generic internal standard after the sample preparation process was sufficient to ensure good assay performance, we opted to use ISTD-3, which is a lipid-conjugated siRNA with an unrelated sequence as an analog internal standard rather than a specific internal standard as in our previous study [21]. Representative chromatograms for this assay are available in Supplementary Figure S1.

The change in sample aliquot volume from 10 to 50 μl was governed by a desire to gain additional sensitivity, given that we were now working with a larger pre-clinical species (i.e., cynomolgus monkeys). As we have previously demonstrated, the aliquot volume-response plot for SIR-2 was linear between 10 μl and 50 μl of plasma at both 100 ng/ml and 10,000 ng/ml of SIR-2 when using 90 pmol of LNA capture probe [21]. Given that we had previously achieved an LLOQ of 0.500 ng/ml for SIR-2 using a 10 μl sample aliquot on a Sciex TripleQuad 7500 mass spectrometer [21], extrapolating from the aliquot volume-response plot in our previous study, we hypothesized that we could achieve an LLOQ of 0.100 ng/ml using a 50 μl sample aliquot. This LLOQ concentration was tested during method development, and while the average accuracy of multiple LLOQ QC replicates was between 70% and 130% (which we considered acceptable for a discovery-grade assay), the coefficient of variance (%CV) was typically ∼35–40%, which we did not consider an acceptable level of precision. The signal-to-noise ratio at 0.100 ng/ml using a 50 μl sample aliquot was ∼3, which may explain the high variability observed for this QC level. By increasing the target LLOQ concentration to 0.200 ng/ml, we were able to obtain acceptable accuracy and precision during method development for the LLOQ QC, and therefore we proceeded with this LLOQ and a 1000-fold calibration curve range for assay performance assessment. The observed LLOQ of 0.200 ng/ml is very similar to the LLOQs we were able to achieve with the SL-RT-qPCR assay.

3.1.2. SL-RT-qPCR

The workflow for the SL-RT-qPCR assay largely followed the protocol recommended by the assay kit manufacturer [22]. The stem-loop RT primer, and qPCR primers and probe were custom designed by ThermoFisher Scientific based on the raw sequence of the antisense strand and their proprietary stem-loop sequence. Initial attempts at analyzing undiluted plasma samples were unsuccessful due to observations of potential non-specific amplification and overall poor accuracy and precision. Similar observations in at least one other publication were attributed to matrix components in plasma contributing to poor assay performance, and in that publication, the matrix effects were minimized by performing an initial 100× dilution of all samples using 0.25% (v/v) Triton X-100 in PBS [25]. We followed the same dilution scheme and also observed a significant improvement in assay performance.

Despite minimizing the matrix effects, overall assay robustness, particularly the precision and accuracy were still outside the thresholds of bias within ±30.0% (±45.0% at the LLOQ), %CV ≤30.0% (≤45.0% at the LLOQ) and intra-replicate Ct %CV ≤2.00% that we considered appropriate for a discovery-grade assay. Systematic evaluation of each step of the sample preparation process indicated that increasing the hold time from the manufacturer-recommended 30 min to 60 min for the 42°C incubation during reverse transcription helped ensure acceptable assay performance. The final LLOQ established for the qPCR assay was 0.128 ng/ml.

3.1.3. HELISA

The starting point for method development of the HELISA workflow was the sandwich ELISA protocol published by Thayer et al. [23]. The LNA probes used for capture and detection were adapted from the LNA capture probe used by the hybrid LC-MS assay, with the first 11 nucleotides from the 5′ end used as the detection probe, and the next 10 nucleotides being used as the capture probe. The position and number of LNAs remained the same as the Hybrid LC-MS probe, with three LNAs incorporated into the capture probe, and four LNAs incorporated into the detection probe. The capture probe was labeled with biotin which was separated from the 3′ terminus by a triethyleneglycol spacer, and the detection probe was labeled with digoxigenin at the 5′ terminus.

Each step of the workflow was evaluated to determine if further optimization was possible, and if using buffers commonly used in our laboratory in place of those used by Thayer et al. [23] would significantly affect assay performance. Factors that helped improve the assay sensitivity included replacing the hybridization buffer with the capture buffer used in the Hybrid LC-MS workflow, and increasing the initial 95°C hold used to denature the siRNA from 5 to 15 min. Replacing the published washing and blocking buffers as well as the MSD read buffer with buffers commonly used in our laboratory did not significantly affect assay performance, and therefore these changes were made to ensure convenient access to reagents. The incorporation of a proteinase K digest prior to probe hybridization was also evaluated, but this step worsened overall assay performance, and was therefore abandoned. Given that a proteinase K digest step is not always included in hybridization-based oligonucleotide bioanalytical assays [13–15], and was not used by Thayer et al. in their workflow [23], we did not attempt to further optimize this step at this time.

The final LLOQ established for SIR-2 at the end of our method development was 1.00 ng/ml, which is an order of magnitude greater than that of the hybrid LC-MS and SL-RT-qPCR assays. We did observe considerably higher signal in the blank samples than we typically see for MSD assays, which could be mitigated by using an alternative detection reagent or decreasing the challenge ratio used during the ruthenium-labeling of the detection antibody. However, given that our LLOQ is on the same order of magnitude as that of Thayer et al. for their siRNA analyte [23], it is possible that the poor sensitivity is a limitation of the assay format. Thayer et al. have previously demonstrated an LLOQ of ∼0.4 ng/ml in plasma for a different siRNA using a two-step HELISA format [26]. Our evaluation of a similar workflow was unsuccessful due to extremely high background signal that most likely resulted from the S1 nuclease used in this assay format failing to cleave single-stranded oligonucleotides. It is possible that we incorporated LNAs too close to the nucleation site [26], and redesigning the probe with alternative LNA positions would help in this case. Additionally, we could explore incorporation of polyethylene glycol or other additives into the detection step of the two-step HELISA assay, which has been shown to improve ligation efficiency and therefore sensitivity [27]. Therefore, it appears there is still work to be done in order to truly achieve a sensitive HELISA methodology for SIR-2, but the methodology as developed is sufficient for now to assess the assay performance against other bioanalytical techniques.

3.1.4. SPE-LC-MS

The SPE-LC-MS workflow was adapted from the protocol published by Sips et al. [15], with minor changes made to the published workflow. Rather than using both a carrier oligonucleotide (added at the start of sample preparation) and an internal standard (added at the end of sample preparation), an analog lipid-conjugated siRNA molecule, ISTD-3, was added at the start of the sample preparation process and used as the internal standard during data processing. TCEP was not included as part of the wash buffer since we determined that adding antioxidant at the end of the sample preparation process was more beneficial than adding it during the workflow, and the reconstitution buffer was changed to match the mobile phases used during sample analysis.

A further change made to Sips et al.'s workflow was the addition of 20 μl of 10 mg/ml cysteine as an antioxidant to all samples following elution from the SPE plate. During method development, a steady decrease in the precursor ion peak intensity and a corresponding increase in peak intensity of a mass that was 16 Da lower were observed in the MS full scan from high-resolution mass spectrometry, and almost 100% conversion was achieved after 72 h of post-sample preparation storage. Accurate mass data were well aligned with the loss of sulfur (-32 Da) and addition of oxygen (+16 Da), indicating the desulfurization of phosphorothioate bonds due to oxidants present in the SPE samples, and this phenomenon has been previously reported in ASO drug formulations [28]. It was experimentally determined that addition of cysteine at the end of the sample preparation process was effective to arrest this conversion that primarily took place during sample dry-down and autosampler storage.

With respect to the LC-MS conditions, we have previously discussed the need to use the polymer-based ThermoFisher Scientific DNAPac column to achieve chromatographic separation for the Hybrid LC-MS assay [21]. However, with respect to the SPE-LC-MS method, we could not use the DNAPac column, and instead had to use the particle-based Waters XBridge BEH C18 column to achieve chromatographic separation. The change in analytical column is because the lipid conjugate on the sense strand of SIR-2 binds almost irreversibly to the polymer stationary phase of the DNAPac column. The SPE workflow extracts both the sense and antisense strands of an siRNA, unlike the Hybrid LC-MS workflow that selectively extracts the antisense strand. We attempted multiple aggressive wash cycles to clean the DNAPac column once extracts from the SPE-LC-MS workflow had been injected on it with little success. However, the same initial gradient on the C18 column followed by several wash cycles and a longer re-equilibration period was sufficient to minimize both lipid and analyte carryover when injecting extracts from the SPE-LC-MS assay. Representative chromatograms for this assay are available in Supplementary Figure S1.

The LLOQ established for the SPE-LC-MS assay at the end of method development was 5.00 ng/ml, which is considerably poorer than the LLOQs established for any of the other bioanalytical assays assessed in this manuscript. However, this LLOQ is on par with most published SPE-LC-MS methods for siRNA analytes [15,29–31].

3.2. Assay performance evaluation

Each bioanalytical assay was assessed independently for its quantitative performance of SIR-2. In the accuracy and precision evaluation, the calibration standards and QC samples for all assays demonstrated bias within ±25.0% of the nominal concentration (±30.0% at the LLOQ) and %CV ≤25.0% (≤30.0% at the LLOQ) (Table 3), which we considered acceptable in a discovery setting. Dilution QC samples that were subjected to a 100× dilution with control plasma prior to analysis demonstrated bias within ±25.0% of the nominal concentration and %CV ≤25.0% for all assays (Table 3), which indicates that samples with SIR-2 concentrations above the calibration range of each assay can still be quantified following appropriate dilutions with control plasma. None of the plasma lots assessed for selectivity demonstrated significant interferences for SIR-2 when assessed as blank matrices. When spiked at the respective low QC concentrations, at least five out of the six selectivity lots demonstrated bias within ±25.0% of the nominal concentration and %CV ≤25.0% for all assays (Table 4).

Table 3.

Precision and accuracy, recovery and matrix effects data for SIR-2 in plasma.

    LLOQ QC Low QC Low-Medium QC Medium QC High QC Dilution QC
Hybrid LC-MS Nominal concentration (ng/ml) 0.200 0.600 8.00 80.0 160 10000
Precision and accuracy 7.67 (28.8) 3.19 (13.7) 9.60 (6.20) 14.3 (8.81) -2.62 (6.22) -22.6 (7.40)
Recovery (%) N/A 88.5 ± 20.4 N/A 103 ± 6.48 103 ± 10.0 N/A
Matrix factor N/A 1.09 ± 0.32 N/A 1.06 ± 0.07 0.92 ± 0.09 N/A
SL-RT-qPCR Nominal concentration (ng/ml) 0.128 0.500 2.50 50.0 1000 5000
Precision and accuracy 5.25 (7.95) 18.6 (7.50) 14.9 (7.94) 20.1 (7.69) -14.5 (8.70) -5.96 (10.9)
HELISA Nominal concentration (ng/ml) 1.00 3.00 N/A 30.0 200 10000
Precision and accuracy -4.93 (5.80) -0.648 (15.3) N/A 11.5 (15.7) 11.6 (6.09) 7.62 (12.7)
SPE-LC-MS Nominal concentration (ng/ml) 5.00 15.0 N/A 200 400 10000
Precision and accuracy 6.60 (8.53) 23.0 (17.1) N/A -6.62 (15.4) 13.7 (24.0) -18.7 (14.1)
Recovery (%) N/A 43.4 ± 11.4 N/A 31.7 ± 4.39 36.4 ± 13.6 N/A
Matrix factor N/A 1.02 ± 0.12 N/A 0.94 ± 0.09 1.13 ± 0.33 N/A

Samples were diluted 100× with control plasma prior to analysis.

Determined by comparing to signals in post-extraction spiked plasma samples to post-extraction spiked water samples, which is a relative assessment of matrix effects. Absolute matrix effects which also consider the impact of reagents used during sample preparation may be different from this value.

Data are presented as %Bias (%CV) or as Arithmetic Mean ± Standard Deviation (n = 6 for each QC level, except for the dilution QC where n = 3).

HELISA: Hybridization ELISA; LLOQ: Lower limit of quantitation; N/A: Not applicable; QC: Quality control; SL-RT-qPCR: Stem-loop reverse transcription quantitative PCR; SPE: Solid-phase extraction.

Table 4.

Selectivity data for SIR-2 in plasma (n = 3 for each lot).

Technique Hybrid LC-MS SL-RT-qPCR HELISA SPE-LC-MS
Nominal concentration (ng/ml) 0.600 0.500 3.00 15.0
Matrix Lot 1 -15.6 (3.63) 18.7 (6.05) -18.1 (9.42) 6.81 (11.1)
Matrix Lot 2 6.81 (3.30) 9.42 (9.36) -4.02 (11.1) 4.15 (12.5)
Matrix Lot 3 1.33 (4.10) 19.2 (4.43) -15.0 (14.0) -11.1 (8.40)
Matrix Lot 4 1.71 (19.1) 12.9 (1.36) 489 (4.33) -3.39 (6.51)
Matrix Lot 5 5.79 (9.30) 10.6 (2.80) -9.19 (16.9) -9.96 (10.6)
Matrix Lot 6 -3.25 (23.6) 23.0 (3.54) -18.2 (13.6) -11.4 (3.03)

Lots analyzed are the same for each technique. Data are presented as %Bias (%CV).

HELISA: Hybridization ELISA; SL-RT-qPCR: Stem-loop reverse transcription quantitative PCR; SPE: Solid-phase extraction.

Recovery across the calibration range was between 88.5 and 103% for the Hybrid LC-MS assay, and between 31.7 and 43.4% for the SPE-LC-MS assay (Table 3). The recovery for the Hybrid LC-MS assay is consistent with our previous evaluation of SIR-2 using a 10 μl sample aliquot [21], indicating that the increased sample aliquot volume did not saturate the binding capacity of the capture probe. The recovery for the SPE-LC-MS assay is ∼3× lower than that observed for the Hybrid LC-MS assay, demonstrating that despite both sample preparation techniques using the same LC-MS hardware for analysis, the differences in sensitivity between the two workflows is at least in part due to the lower recovery. Recovery was not assessed for the SL-RT-qPCR and HELISA methods since the sample preparation protocols in both cases did not allow for a convenient assessment of recovery.

Both LC-MS workflows had matrix factors between 0.92 and 1.13 (Table 3), which illustrates no significant matrix effects. However, in our experiment, matrix factors are a relative assessment of matrix effects in which post-extraction spiked blank plasma samples are compared with post-extraction spiked blank water samples rather than unextracted neat solution, and therefore do not consider the impact buffers or other components of the sample preparation process may have on analyte signal. We did attempt to evaluate absolute matrix effects during method development by comparing a post-extraction spiked blank plasma samples to neat solutions of SIR-2, but noted severe and irreversible non-specific binding of SIR-2 in neat solution at the relevant concentrations, which rendered the results of this test meaningless.

The chromatograms of samples extracted by the SPE-LC-MS assay demonstrate much higher background and lower signal intensities when compared with the chromatograms of samples extracted by the Hybrid LC-MS assay (Supplementary Figure S1), which is representative of ion suppression due to buffers or other components of the SPE sample preparation process. This in turn illustrates that the relative matrix factors calculated for the SPE-LC-MS assay are not truly representative of the overall matrix effects. In part, this ion suppression can be attributed to the fact that the C18 column used during LC separation cannot adequately resolve SIR-2 from some reagent components associated with the SPE sample preparation process. It is unclear if using a different analytical column would be helpful in this respect, since we have already discussed the reason the DNAPac column used for the Hybrid LC-MS assay was not suitable for use with the SPE LC-MS assay, and no other stationary phases are routinely used for oligonucleotide bioanalysis. On the other hand, hybrid LC-MS assays tend to produce relatively clean sample extracts that are generally free of matrix components [17], which would indicate that the relative matrix factors observed for the Hybrid LC-MS assay are representative of absolute matrix effects. Thus, the absolute matrix effects for the SPE-LC-MS assay paired with the observed lower recovery would explain the ∼25× difference in sensitivity compared with Hybrid LC-MS.

If we were to consider the accuracy and precision results with a view to eventually validate any of the assay platforms for the analysis of GLP or clinical samples, three out of the four workflows show great promise in this regard. Except for the precision of the LLOQ QC, the Hybrid LC-MS assay generally had bias within ±15% of the nominal concentration and %CV ≤15% (Table 3), which meets the current regulatory standard for chromatographic assays [24]. There is some debate as to whether the use of a hybrid sample preparation process would permit applying an expanded acceptance criteria (such as that used to validate ligand-binding assays) [32], but given that other hybrid LC-MS methods for oligonucleotides have also been demonstrated to meet the current regulatory standard for chromatographic assays [16,18–20], this may be a moot point. The SL-RT-qPCR assay was very precise, with %CV ≤10% across the calibration range, and biases were generally within ±20% of the nominal concentrations (Table 3). While there are currently no formal regulatory acceptance criteria for qPCR-based bioanalytical assays, the performance of the SL-RT-qPCR assay for SIR-2 is acceptable according to the criteria of bias within ±20–25% and %CV ≤20–25% as shown by a few white papers and reviews based on current practices in the bioanalytical community [33–36]. The HELISA workflow generally had bias within ±20% and %CV ≤15% across the calibration range (Table 3), which meets current regulatory guidance acceptance criteria for ligand-binding assays [24].

The SPE-LC-MS assay did not meet regulatory acceptance criteria for accuracy and precision, with multiple QC levels having bias or %CV >15% (Table 3). It is unclear why this specific application of SPE-LC-MS did not meet regulatory acceptance criteria when this technique is one of the more mature workflows for oligonucleotide bioanalysis, and others have successfully used SPE workflows to support regulatory submissions for siRNA therapeutics [2]. Poor analyte recovery (Table 3) and high chromatographic background (Supplementary Figure S1) could contribute to the lack of acceptable accuracy and precision.

3.3. Analysis of pre-clinical plasma samples

The four bioanalytical assays were able to quantify SIR-2 in the same sets of plasma samples that had been collected from cynomolgus monkeys dosed parenterally at one of four doses (n = 1 selected for analysis from each dose) with varying degrees of success. Pharmacokinetic profiles for SIR-2 as estimated by each bioanalytical assay are available in Figure 2.

Figure 2.

Figure 2.

Pharmacokinetic profiles of SIR-2 in cynomolgus monkey plasma when analyzed by four different bioanalytical methods. SIR-2 was dosed parenterally at one of four doses (n = 1 selected for analysis from each dose), and plasma was collected pre-dose and at 14 timepoints post-dose. The same samples were analyzed by all four methods at the same time, and sample analyses were started within a 2-h period. The x-axis is unitless and is presented as an ordinal scale that represents each timepoint sampled rather than the actual times post-dosing.

For the two highest dose groups (10× Dose and 30× Dose), all four assays were able to quantify SIR-2 at almost all timepoints. Typical timepoints at which SIR-2 could not be quantified by one or more assays were the first timepoint post-dose or the final two timepoints, where analyte concentrations are typically low or undetectable. For the second-lowest dose group (3× Dose), Hybrid LC-MS and SL-RT-qPCR were able to quantify SIR-2 at almost all timepoints, while HELISA was able to quantify SIR-2 at six timepoints, and SPE-LC-MS was able to quantify SIR-2 at only three timepoints. For the lowest dose group (1× Dose), Hybrid LC-MS and SL-RT-qPCR were able to quantify SIR-2 at almost all timepoints, while HELISA was able to quantify SIR-2 at four timepoints, and SPE-LC-MS was unable to quantify SIR-2 at any timepoint. No samples had to be diluted to ensure they fit within the quantifiable range of any of the bioanalytical assays, indicating that the ULOQs for all assays were appropriate. However, the variation in successful quantitation of SIR-2 for the different dose groups indicates the necessity for highly sensitive assays for oligonucleotide quantitation.

Evaluating the comparability of observed SIR-2 concentrations in the various samples across the four bioanalytical assays, concentrations obtained by SL-RT-qPCR, HELISA and SPE-LC-MS are all strongly correlated to concentrations obtained by Hybrid LC-MS when assessing all collected data independent of dose group and time (Figure 3A–C). However, correlation is not truly indicative of comparability since data can be highly correlated but still be offset from one another either on a constant or proportionate basis [37]. Comparing the various assays to Hybrid LC-MS using Bland-Altman ratio analysis, SPE-LC-MS is most similar with an average ratio of 0.992, followed by SL-RT-qPCR with an average ratio of 1.26, and then HELISA with an average ratio of 1.60 (Figure 3D–F). Hybrid LC-MS was chosen as the reference assay for this comparison since it utilizes both a hybridization-based sample preparation and LC-MS-based separation and detection, and therefore we hypothesize this assay platform will provide the highest specificity. It should also be noted that all comparisons between assays only considered datapoints where both assays were able to quantify SIR-2.

Figure 3.

Figure 3.

Comparison of SIR-2 concentrations in cynomolgus monkey plasma obtained by different bioanalytical methods to those obtained by Hybrid LC-MS. Assays were compared by Spearman's correlation (A–C) and Bland-Altman ratio analysis (D–F). Tested bioanalytical methods included stem-loop reverse transcription quantitative PCR (A & D), hybridization ELISA (B & E) and solid phase extraction LC-MS (C & F).

A more precise interpretation of the Bland-Altman analysis results is that the average Bland-Altman ratio reflects the similarity in detection methodology to the Hybrid LC-MS assay (which in turn is a reflection of the specificity of the assay platform), whereas the width of the limits of agreement reflects the similarity in sample preparation technique. Starting with the similarity of sample preparation technique, HELISA and Hybrid LC-MS have almost identical sample preparation workflows consisting of SIR-2 denaturation, hybridization of solid support-conjugated capture probes to the antisense strand of SIR-2, and removal of unbound material followed by detection (Figure 1), which is reflected in the much narrower limits of agreement (1.27–1.94) around an average ratio of 1.60. SL-RT-qPCR and SPE-LC-MS both utilize very different sample preparation techniques compared with Hybrid LC-MS (Figure 1), where in one case matrix components are selectively removed prior to elution of the analyte (SPE-LC-MS), and in the other case, the target sequence is amplified prior to detection (SL-RT-qPCR). Thus, the limits of agreement for these assay platforms are considerably wider than that of HELISA.

Turning to the average Bland-Altman ratio, SPE-LC-MS and Hybrid LC-MS utilize mass spectrometry for detection and are therefore highly selective and able to distinguish between SIR-2 and its potential metabolites. Thus, observed concentrations by each assay reflect concentrations of SIR-2 only, which should be consistent between assays, and this is demonstrated by the average Bland-Altman ratio of ∼1. The HELISA and SL-RT-qPCR workflows use non-specific detection methodologies that cannot distinguish between SIR-2 and its potential metabolites that may also be captured during the sample preparation. Thus, the observed concentrations by these assay platforms represent the sum of the concentrations of SIR-2 and any truncated metabolites captured during sample preparation. The average Bland-Altman ratios of 1.60 (HELISA) and 1.26 (SL-RT-qPCR) indicate that metabolite concentrations were reported as part of the observed SIR-2 concentrations for these assays.

The hypothesis that the SIR-2 concentrations observed by the HELISA and SL-RT-qPCR assays includes metabolite concentrations is further supported when the pharmacokinetic profiles obtained by each assay for the 10× Dose and 30× Dose samples are overlaid. These two highest dose groups were chosen since they have nearly complete data from all four assays. Concentrations by SL-RT-qPCR and HELISA are overall elevated relative to the LC-MS assays (Supplementary Figure S2). Additionally, when performing the Hybrid LC-MS sample analysis, we did include theoretical MRMs for the most common truncated metabolites (5′ n-1, 5′ n-2, 3′ n-1 and 3′ n-2) and saw MS signals that appeared to correspond to these analytes in the 30× Dose samples only. However, since we did not have reference standards for these metabolites, we cannot confirm their identity or estimate their concentrations in the samples.

3.4. Overall comparison of bioanalytical assays

Considering the four bioanalytical assays evaluated in this manuscript, it is clear that each workflow has its own strengths and weaknesses when analyzing a lipid-conjugated siRNA molecule in plasma. Specific parameters for the assays when evaluating SIR-2 in plasma are presented in Supplementary Table S1.

From a sensitivity standpoint, Hybrid LC-MS and SL-RT-qPCR are almost an order of magnitude more sensitive than HELISA or SPE-LC-MS. On the other hand, SL-RT-qPCR and HELISA cannot typically distinguish between a parent oligonucleotide and its metabolites, and therefore could provide an overestimate of the concentration. From just these two pieces of information, it would appear that Hybrid LC-MS is the best assay platform for future analyses of SIR-2 given that this workflow is both highly sensitive and highly specific. However, the Hybrid LC-MS assay required a much higher sample volume than the SL-RT-qPCR assay to achieve comparable sensitivity, which could be a problem if working with smaller pre-clinical species. The SL-RT-qPCR assay also has a wider dynamic range, which is useful when analyzing samples from multiple dose groups such as in a toxicology or toleration study, since fewer samples need dilution prior to analysis. Furthermore, if we accept the view that “total” concentrations of the antisense strand and its metabolites better represent the bioactive concentration of the siRNA molecule [30,31], then the lack of specificity associated with HELISA and SL-RT-qPCR could be considered an asset.

From a throughput standpoint, considering the analysis of the cynomolgus monkey plasma samples described above, despite all four assays taking approximately the same amount of time for the actual sample preparation, only the HELISA and SL-RT-qPCR assays were able to provide sample concentration data within the same day. It took almost 20 h to acquire all data for the Hybrid LC-MS samples, and over 24 h for the SPE-LC-MS samples. Given that this study analyzed only 60 subject samples, and most pre-clinical and clinical studies have hundreds or thousands of samples, it could take a much longer time to generate all the data by LC-MS assays than HELISA or SL-RT-qPCR.

Finally, considering the method development time for each new analyte once a workflow has been established in a laboratory, as well as the need for critical reagents, SPE-LC-MS is the most generic of the assays since the sample preparation workflow has no analyte-specific components and therefore can be directly applied to each new compound. This significantly reduces method development time, and for the most part, method development will consist of verifications of assay sensitivity and perhaps minor modifications to the SPE or LC-MS conditions. All the other assays require analyte-specific probes or primers, which must be custom ordered for each new project. Each probe or primer must be designed and then evaluated for effectiveness, though once a suitable probe or primer is found, most of the method development is complete for the hybrid LC-MS and SL-RT-qPCR assays since the remaining assay parameters are typically generalizable across analytes. For SL-RT-qPCR, the same set of reagents can be applied to a set of analytes with same raw sequences but different chemical modifications, which adds efficiency in developing assays for analog molecules. HELISA workflows require evaluation of buffers, incubation times and minimum required dilutions for each new analyte, so this workflow will typically have the longest method development time.

For SIR-2 in particular, as we move through the later stages of drug development, we have determined that SL-RT-qPCR would be the best assay since it offers the required high sensitivity and throughput, and has acceptable specificity. Hybrid LC-MS would be an ideal orthogonal assay that can be used to analyze selected samples for metabolite detection or quantitation with high sensitivity and specificity. Since only selected samples will be analyzed for this purpose, throughput will be less of a concern.

4. Conclusion

This manuscript presents the development of four common oligonucleotide bioanalytical assays and applies them toward quantifying a 21-mer lipid-conjugated siRNA therapeutic in plasma obtained from a pre-clinical pharmacokinetic study. All four assays were able to quantify the therapeutic with varying degrees of success in the study samples, with the Hybrid LC-MS and SL-RT-qPCR assays quantifying the therapeutic at almost all timepoints, followed by HELISA and then the SPE-LC-MS assay. Data obtained from all four assays was comparable and highly correlated, though systemic bias was observed between assays that could be attributed to potential existence of metabolites and assay variability. Overall, any of the evaluated assays could be used in oligonucleotide therapeutic programs to quantify the analyte at various stages of the drug discovery and development process.

There is no “right” answer when choosing a bioanalytical assay for oligonucleotide analysis. In addition to considering sensitivity, specificity and throughput, some consideration must be given to the stage of the drug program in which samples are being analyzed, as well as the instrumentation and expertise currently available to the laboratory. As research continues in the area of oligonucleotide bioanalysis, it is hoped that more comparisons will be made between different assay platforms for the same analytes, so that a more comprehensive picture can emerge regarding the impact of oligonucleotide therapeutic-associated modifications, conjugations or nucleotide length on inter-assay comparability, as well as the appropriate use of each methodology during the drug discovery and development process.

Supplementary Material

Supplementary Figures S1-S2 and Table S1
IBIO_A_2350266_SM0001.pdf (287.6KB, pdf)

Acknowledgments

The authors gratefully acknowledge the technical support provided by V Annavarapu during assay qualification for the SPE-LC-MS assay and the subsequent sample analysis; and the advice and assistance provided by Y Bao during the method development process for the SL-RT-qPCR assay, and by K Johnson during data analysis and interpretation. Additionally, L Dillen, B Geist and T-Y Yuan are acknowledged for providing feedback and insights during the preparation of this manuscript.

Supplemental material

Supplemental data for this article can be accessed at https://doi.org/10.1080/17576180.2024.2350266

Author contributions

K Agrawal, SA Ayers and W Jian designed the study; K Agrawal, LK Calliste, S Ji and S Xu developed and qualified the bioanalytical assays for SIR-2 and performed the sample analyses; and K Agrawal and W Jian performed data analysis and interpretation and wrote the manuscript. All authors reviewed and approved the submitted manuscript.

Financial disclosure

The authors have no financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Competing interests disclosure

The authors have no competing interests or relevant affiliations with any organization or entity with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Writing disclosure

No writing assistance was utilized in the production of this manuscript.

Ethical conduct of research

All animal studies were conducted in accordance with the Johnson & Johnson Animal Welfare Policy. Ethical approval for this study was obtained from the Institutional Animal Care and Use Committee of Charles River-Nevada (Protocol #20314229), and reviewed and approved by Johnson & Johnson.

References

Papers of special note have been highlighted as: • of interest; •• of considerable interest

  • 1.Egli M, Manoharan M. Chemistry, structure and function of approved oligonucleotide therapeutics. Nucleic Acids Res. 2023;51(6):2529–2573. doi: 10.1093/nar/gkad067 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Xing J, Arya V, Reynolds KS, et al. Clinical pharmacology of RNAi-based therapeutics: a summary based on FDA-approved small-interfering RNAs. Drug Metab Dispos. 2023;51(2):193–198. doi: 10.1124/dmd.122.001107 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Infographic: the exciting world of oligonucleotides [Internet]. Abingdon (United Kingdom): Bioanalysis Zone; 2019 Sept 16. [cited 2023 Sept 1]. Available from: http://www.bioanalysis-zone.com/infographic-exciting-world-oligonucleotides_yuduoligo/ [Google Scholar]
  • 4.Wang L. Oligonucleotide bioanalysis: sensitivity versus specificity. Bioanalysis. 2011;3(12):1299–1303. doi: 10.4155/bio.11.111 [DOI] [PubMed] [Google Scholar]
  • 5.Wang L, Meng M, Reuschel S. Regulated bioanalysis of oligonucleotide therapeutics and biomarkers: qPCR versus chromatographic assays. Bioanalysis. 2013;5(22):2747–2751. doi: 10.4155/bio.13.234 [DOI] [PubMed] [Google Scholar]
  • 6.Wang L, Ji C. Advances in quantitative bioanalysis of oligonucleotide biomarkers and therapeutics. Bioanalysis. 2016;8(2):143–155. doi: 10.4155/bio.15.234 [DOI] [PubMed] [Google Scholar]
  • 7.Wang L, Meng M, Bennett P. Recent advances in bioanalysis and metabolite profiling of oligonucleotide therapeutics. Bioanalysis Zone. 2019. [cited 2023 Sept 1]; [7 screens]. http://www.bioanalysis-zone.com/recent-advances-bioanalysis-metabolite-profiling-oligonucleotide-therapeutics_yuduoligo/ [Google Scholar]
  • 8.Kotapati S, Deshpande M, Jashnani A, et al. The role of ligand-binding assay and LC-MS in the bioanalysis of complex protein and oligonucleotide therapeutics. Bioanalysis. 2011;13(11):931–954. doi: 10.4155/bio-2021-0009 [DOI] [PubMed] [Google Scholar]; •• Comparison of ligand-binding and LC-MS assays for oligonucleotide bioanalysis, including discussion on how to choose an appropriate bioanalytical workflow for each stage of the drug discovery and development process.
  • 9.Boyanapalli R, Singh I, Faria M. Peptides and oligonucleotide-based therapy: bioanalytical challenges and practical solutions. In: Kumar S, editor. An Introduction to Bioanalysis of Biopharmaceuticals. Cham (Switzerland): Springer; 2022. p. 131–155. doi: 10.1007/978-3-030-97193-9_6 [DOI] [Google Scholar]; •• Comparison of ligand-binding, LC-MS and RT-qPCR assays for oligonucleotide bioanalysis.
  • 10.White A. Ligand-binding assays vs chromatographic platforms for oligonucleotide quantification. Bioanalysis. 2023;15(2):53–55. doi: 10.4155/bio-2023-0032 [DOI] [PubMed] [Google Scholar]
  • 11.Clinical Pharmacology and Biotherapeutics Review(s): Nusinersen. Silver Spring (MD): US Food and Drug Administration; 2016. http://www.accessdata.fda.gov/drugsatfda_docs/nda/2016/209531Orig1s000ClinPharmR.pdf [Google Scholar]
  • 12.Clinical Pharmacology and Biotherapeutics Review(s): Inotersen. Silver Spring (MD): US Food and Drug Administration; 2018. http://www.accessdata.fda.gov/drugsatfda_docs/nda/2018/211172Orig1s000ClinPharmR.pdf [Google Scholar]
  • 13.Dillen L, Sips L, Greway T, et al. Quantitative analysis of imetelstat in plasma with LC-MS/MS using solid-phase or hybridization extraction. Bioanalysis. 2017;9(23):1859–1872. doi: 10.4155/bio-2017-0145 [DOI] [PubMed] [Google Scholar]; •• Comparison of hybrid LC-MS, SPE-LC-MS and HELISA assays for the same ASO analyte, including assessment of in vivo samples.
  • 14.Basiri B, Sutton JM, Hooshfar S, et al. Direct identification of microribonucleic acid miR-451 from plasma using liquid chromatography mass spectrometry. J Chromatogr A. 2019;1584:97–105. doi: 10.1016/j.chroma.2018.11.029 [DOI] [PubMed] [Google Scholar]
  • 15.Sips L, Ediage EN, Ingelse B, et al. LC-MS quantification of oligonucleotides in biological matrices with SPE or hybridization extraction. Bioanalysis. 2019;11(21):1941–1954. doi: 10.4155/bio-2019-0117 [DOI] [PubMed] [Google Scholar]; •• Comparison of hybrid LC-MS and SPE-LC-MS assays for the same ASO and siRNA analytes.
  • 16.Li P, Gong Y, Kim J, et al. Hybridization liquid chromatography-tandem mass spectrometry: an alternative bioanalytical method for antisense oligonucleotide quantitation in plasma and tissue samples. Anal Chem. 2020;92(15):10548–10559. doi: 10.1021/acs.analchem.0c01382 [DOI] [PubMed] [Google Scholar]
  • 17.Jiang D, Yuan L. Microflow LC-MS/MS to improve sensitivity for antisense oligonucleotides bioanalysis: critical role of sample cleanness. Bioanalysis. 2022;14(21):1365–1376. doi: 10.4155/bio-2022-0201 [DOI] [PubMed] [Google Scholar]
  • 18.Li P, Dupuis J-F, Vrionis V, et al. Validation and application of hybridization liquid chromatography-tandem mass spectrometry methods for quantitative bioanalysis of antisense oligonucleotides. Bioanalysis. 2022;14(9):589–601. doi: 10.4155/bio-2022-0015 [DOI] [PubMed] [Google Scholar]
  • 19.Yuan L, Dupuis J-F, Mekhssian K. A novel hybridization LC-MS/MS methodology for quantification of siRNA in plasma, CSF and tissue samples. Molecules. 2023;28(4):1618. doi: 10.3390/molecules28041618 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Yuan L, Dupuis J-F, Vrionis V, et al. A validated capillary microsampling liquid chromatography-tandem mass spectrometry method for quantification of antisense oligonucleotides in mouse serum. Bioanalysis. 2023;15(12):683–694. doi: 10.4155/bio-2023-0096 [DOI] [PubMed] [Google Scholar]
  • 21.Agrawal K, Kang L, Ji S, et al. Evaluating the use of locked nucleic acid capture probes in hybrid LC-MS/MS analysis of siRNA analytes. Bioanalysis. 2023;15(18):1129–1146. doi: 10.4155/bio-2023-0079 [DOI] [PubMed] [Google Scholar]; • Method development workflow for LNA capture probe-based hybrid LC-MS assays to analyze siRNA analytes.
  • 22.TaqMan™ small RNA assays user guide. Waltham (MA): Applied Biosystems; 2019. (Publication Number: 4364031, Revision: H). https://assets.thermofisher.com/TFS-Assets/LSG/manuals/4364031_TaqSmallRNA_UG.pdf [Google Scholar]
  • 23.Thayer MB, Humphreys SC, Chung KS, et al. POE immunoassay: plate-based oligonucleotide electro-chemiluminescent immunoassay for the quantification of nucleic acids in biological matrices. Sci Rep. 2020;10:10425. doi: 10.1038/s41598-020-66829-6 [DOI] [PMC free article] [PubMed] [Google Scholar]; • Method development workflow for LNA capture probe-based sandwich HELISA assays to analyze siRNA analytes.
  • 24.Bioanalytical method validation and study sample analysis . Geneva (Switzerland): International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use; 2022. (ICH harmonised guideline; no. M10). https://database.ich.org/sites/default/files/M10_Guideline_Step4_2022_0524.pdf [Google Scholar]
  • 25.Castellanos-Rizaldos E, Brown CR, Dennin S, et al. RT-qPCR methods to support pharmacokinetics and drug mechanism of action to advance development of RNAi therapeutics. Nucleic Acid Ther. 2020;30(3):133–142. doi: 10.1089/nat.2019.0840 [DOI] [PubMed] [Google Scholar]; • Method development workflow for SL-RT-qPCR assays to analyze siRNA analytes.
  • 26.Thayer MB, Lade JM, Doherty D, et al. Application of locked nucleic acid oligonucleotides for siRNA preclinical bioanalytics. Sci Rep. 2019;9:3566. doi: 10.1038/s41598-019-40187-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Haegele JA, Boyanapalli R, Goyal J. Improvements to hybridization-ligation ELISA methods to overcome bioanalytical challenges posed by novel oligonucleotide therapeutics. Nucleic Acid Ther. 2022;32(4):350–359. doi: 10.1089/nat.2021.0100 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Krotz AH, Mehta RC, Hardee GE. Peroxide-mediated desulfurization of phosphorothioate oligonucleotides and its prevention. J Pharm Sci. 2005;94(2):341–352. doi: 10.1002/jps.20235 [DOI] [PubMed] [Google Scholar]
  • 29.Wheller R, Summerfield S, Barfield M. Comparison of accurate mass LC-MS and MRM LC-MS/MS for the quantification of a therapeutic small interfering RNA. Int J Mass Spectrom. 2013;345–347:45–53. doi: 10.1016/j.ijms.2012.09.012 [DOI] [Google Scholar]
  • 30.Agarwal S, Simon AR, Goel V, et al. Pharmacokinetics and pharmacodynamics of the small interfering ribonucleic acid, Givosiran, in patients with acute hepatic porphyria. Clin Pharmacol Ther. 2020;108(1):63–72. doi: 10.1002/cpt.1802 [DOI] [PubMed] [Google Scholar]
  • 31.McDougall R, Ramsden D, Agarwal S, et al. The nonclinical disposition and pharmacokinetic/pharmacodynamic properties of N-acetylgalactosamine-conjugated small interfering RNA are highly predictable and build confidence in translation to human. Drug Metab Dispos. 2022;50(6):781–797. doi: 10.1124/dmd.121.000428 [DOI] [PubMed] [Google Scholar]
  • 32.Jenkins R, Duggan JX, Aubry A-F, et al. Recommendations for validation of LC-MS/MS bioanalytical methods for protein biotherapeutics. AAPS J. 2015;17(1):1–16. doi: 10.1208/s12248-014-9685-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Yang T-Y, Doddareddy R. Considerations in the development and validation of real-time quantitative polymerase chain reaction and its application in regulated bioanalysis to characterize the cellular kinetics of CAR-T products in clinical studies. Bioanalysis. 2021;13(2):115–128. doi: 10.4155/bio-2020-0221 [DOI] [PubMed] [Google Scholar]
  • 34.Ma H, Bell KN, Loker RN. qPCR and qRT-PCR analysis: regulatory points to consider when conducting biodistribution and vector shedding studies. Mol Ther Methods Clin Dev. 2021;20:152–168. doi: 10.1016/j.omtm.2020.11.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Hays A, Islam R, Matys K, et al. Best practices in qPCR and dPCR validation in regulated bioanalytical laboratories. AAPS J. 2022;24(2):36. doi: 10.1208/s12248-022-00686-1 [DOI] [PubMed] [Google Scholar]
  • 36.Wissel M, Poirier M, Satterwhite C, et al. Recommendations on qPCR/ddPCR assay validation by GCC. Bioanalysis. 2022;14(12):853–863. doi: 10.4155/bio-2022-0109 [DOI] [PubMed] [Google Scholar]
  • 37.Giavarina D. Understanding Bland Altman analysis. Biochem Med (Zagreb). 2015;25(2):141–151. doi: 10.11613/BM.2015.015 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Figures S1-S2 and Table S1
IBIO_A_2350266_SM0001.pdf (287.6KB, pdf)

Articles from Bioanalysis are provided here courtesy of Taylor & Francis

RESOURCES