Oligonucleotide drugs:analysis and purification

SummaryTen core conclusions

1Backbone chemistry: phosphodiester and phosphorothioate

Natural nucleic acids are linked by phosphodiester (PO) bonds. Therapeutic oligonucleotides almost all replace one of the non-bridging oxygens with sulfur, forming a phosphorothioate (PS) linkage. This single-atom substitution has three decisive consequences: markedly greater resistance to nucleases, stronger plasma protein binding and hence longer residence in the body, and a chiral phosphorus centre at every linkage.

Phosphodiester (PO) linkage · dinucleotide fragment

The natural backbone. The non-bridging oxygen carries a negative charge and is the source of charge for AEX separation. Readily degraded by nucleases.

Phosphorothioate (PS) linkage · dinucleotide fragment

P=O replaced by P=S. Incomplete sulfurization leaves a P=O impurity differing from the main product by only 16 Da — resolvable by AEX, often co-eluting in IP-RP.

PS chiral phosphorus centre · Rp configuration

Four different groups on the phosphorus → a stereocentre.

PS chiral phosphorus centre · Sp configuration

A diastereomer of Rp, with slightly different chromatographic behaviour.

Sulfurizing reagent

A 3H-1,2-benzodithiol-3-one 1,1-dioxide type reagent, which carries out the P(III)→P(V) sulfurization in the synthesis cycle.

2The 2′ sugar modification landscape and conformational locking

The 2′ position is the main battleground of oligonucleotide chemistry. Modification has three aims: greater nuclease stability, higher hybridization affinity for the target RNA (T_m), and lower immunogenicity and off-target activity. From a purification standpoint, 2′ modification directly changes the molecule's hydrophobicity and hydrophilicity, and therefore directly changes retention in IP-RP and HILIC — which is why a method has to be redeveloped when the modification pattern changes on the same sequence.

2′-H · deoxyribose (DNA)

Required in the central region of a gapmer — only a DNA:RNA hybrid duplex is recognized by RNase H1.

2′-OH · natural ribose (RNA)

The most hydrophilic; the 2′-OH takes part in alkaline hydrolysis, so pure RNA cannot use denaturing AEX at pH 12.

2′-F · 2′-fluoro

Locks the C3′-endo conformation and markedly raises T_m; used alternately with 2′-OMe in siRNA.

2′-OMe · 2′-O-methyl

Occurs naturally in human RNA with low immunogenicity; the main modification in siRNA.

2′-MOE · 2′-O-methoxyethyl

The signature modification of marketed ASOs. It increases hydrophobicity appreciably, markedly strengthening IP-RP retention.

LNA · 2′-O,4′-C-methylene bridge

The bridge locks C3′-endo, and each LNA unit raises T_m by 2–8 °C.

cEt · constrained ethyl (methyl on the bridge)

A refinement of LNA that keeps the high affinity while improving the hepatotoxicity profile.

UNA · unlocked nucleic acid

The C2′–C3′ bond is opened, giving very high flexibility; used in the siRNA seed region to suppress off-target activity.

3Terminal chemistry, protecting groups and phosphoramidite monomers

5′-vinylphosphonate (5′-VP)

Replaces the hydrolysable P–O bond with a carbon–carbon double bond, resisting phosphatase degradation and stabilizing loading of the antisense strand into RISC. Adds one stable negative charge.

5′-DMT protected nucleoside

A strongly hydrophobic group of MW 303 — precisely the physical basis of the DMT-on purification strategy; its orange cation can be used to monitor coupling efficiency on line.

Phosphoramidite monomer

DMT protects the 5′-OH, cyanoethyl (CE) protects the phosphorus, and diisopropylamino is the leaving group. Under ICH Q11, commercially sourced phosphoramidites with a defined impurity profile can generally serve as starting materials.

Abasic site (AP site)

The product of purine glycosidic bond cleavage during acidic DMT removal (−117 / −133 Da), which can undergo β-elimination and break the chain. Explicitly listed by regulators as a new structural unit impurity of safety concern.

2′-deoxyguanosine — the site most susceptible to depurination

The guanine N9 glycosidic bond is the most readily cleaved during the acidic DMT removal step. The number and position of G in the sequence directly affect the level of abasic impurity formed.

4The solid-phase synthesis cycle and the chemical origin of impurities

The synthesis column is itself a chromatography column: CPG or polystyrene support is packed in a 100–800 mm bore synthesis column, and reagents flow through in a four-step cycle. Every additional nucleotide is another opportunity to introduce an impurity.

This figure explains why the difficulty of purifying long products (sgRNA, long ASOs) rises not linearly but as a cliff edge, and why the industry has developed alternative synthetic strategies such as fragment (blockmer) coupling.

The text below works through this classification class by class, with a consistent template for each:structural features → key impurities → purification route → analytical scheme。

01ASO — antisense oligonucleotide

1.1 Structural features

A single strand of 15–25 nt, usually fully thioated (all PS). There are two main mechanistic designs: Gapmer(a central DNA gap recruits RNase H1 to cleave the target mRNA, with high-affinity modifications in both wings) and steric-block(fully modified, non-cleaving, used to modulate splicing). Common modifications include 2′-MOE, 2′-OMe, LNA, cEt and MsPA.

The 2′ modifications differ between the wings and the gap, so hydrophobicity is distributed unevenly along the chain. This means that N−1 species arising from deletions at different positions shift retention differently in IP-RP, appearing as a cluster of adjacent peaks rather than a single impurity peak — the most common difficulty in developing an ASO purity method.

1.2 Key impurities

1.3 Purification routes

1.4 Analytical scheme

• Main purity:IP-RP-UV at 260 nm. Column: hybrid silica C18, 130 Å / 1.7 μm, 60 °C; mobile phase TEA-HFIP or HAA; raise to 80–90 °C for GC-rich sequences.
• Orthogonal purity:Denaturing AEX, 20 mM NaOH pH 12.4, NaCl 330 → 900 mM over 31.5 min; specifically for P=O and N−1.
• Third dimension:CGE, Bis-Tris/borate buffer with 27% PEG 35,000, −25 kV, 30 °C.
• Structural confirmation:LC-HRMS for the intact molecular weight (within 15–20 ppm) plus MS/MS sequencing (a-B / w ions; ECD/EThcD improves coverage for a fully PS, 2′-MOE backbone).
• Degree of sulfurization:³¹P NMR to quantify the PS/PO ratio.
• Dedicated abasic and deamination methods:WAX (DEAE weak anion exchange column) with cysteine-analogue derivatization, column temperature ≤20 °C.

02siRNA and miRNA mimics — small interfering RNA / microRNA mimic

2.1 Structural features

A duplex of 21+21 nt (the antisense strand may reach 23 nt where there is a 3′ overhang), with the sense and antisense strands differing in chemistry and possibly in length. The typical modification pattern alternates 2′-OMe and 2′-F, with two to four terminal PS linkages against exonucleases and an optional 5′-VP at the 5′ end of the antisense strand. miRNA mimics and agomirs are structurally isomorphous with siRNA, so the purification and analytical strategies transfer directly.

Note: the alternating modification pattern shown is a general industry illustration; the modification pattern differs for each product.

2.2 Key impurities (additional to the ASO impurity profile)

2.3 Purification route

2.4 Analytical scheme: paired denaturing and non-denaturing methods

SEC × IP-LC-MS two-dimensional chromatography can also give both non-denaturing and denaturing information in one experiment; the denaturing/non-denaturing boundary for HILIC is that the duplex is stable below 55 °C and fully melted above 65 °C, so60 °C is a broad transition zone to be avoided。

03GalNAc-conjugated oligonucleotides

3.1 Structural features

A trivalent N-acetylgalactosamine (GalNAc) cluster is conjugated through a linker to the 3′ or 5′ end of the oligonucleotide, binding the asialoglycoprotein receptor (ASGPR) highly expressed on hepatocytes and being endocytosed. The triantennary design gives affinity several orders of magnitude higher than monovalent. GalNAc can be used with siRNA, ASO and miRNA.

GalNAc ligand unit (N-acetylgalactosamine with linker, schematic)

The actual product has three such units converging on one attachment point through a trivalent scaffold. The acetyl groups on the sugar must be removed at the end of synthesis — incomplete deacetylation is a characteristic impurity of conjugated products.

List of GalNAc-related impurities

• Completely unconjugated oligonucleotide — the same net charge as the product
• Partly conjugated species (1× or 2× GalNAc instead of 3×)
• Linker degradation products (hydrolysis of the amide or ester bond)
• Incompletely removed acetyl protecting groups on the sugar
• Residual conjugation reagents and catalytic metals (Pd, for example)

3.2 Purification: three workable routes

Additional: the monolith routePatent example— a quaternary ammonium monolith (8 mL, pore size 1300–2000 nm), 20 mM Na₂HPO₄ with 10% ACN at pH 8.5, NaBr 0 → 1 M, 0 → 55% B over 15 min, 30 mg loaded per run (tested to 50 mg), kept below 35 °C to protect guanine-rich quadruplex structures, giving over 200 mg of pure material with single-strand fraction purity of 88–93%.

3.3 Analytical scheme

• IP-RP-UV / MS as the main method:the hydrophobicity of the conjugated, partly conjugated and unconjugated species differs considerably, so they separate well; HRMS confirms the mass corresponding to each arm.
• The GalNAc cluster is a hydrophilic structure, so HILIC is also an effective orthogonal method.
• Where conjugation is carried out after purification (purify the oligonucleotide, conjugate with the ligand precursor, then lyophilize), a dedicated purity and conjugation-rate method is needed after the conjugation step, and the pre-conjugation method cannot simply be carried over.

04Other conjugates: peptide / lipid / PEG / antibody / radiolabel / fluorophore

Cholesterol — a lipophilic conjugation ligand

In the same family of lipophilic ligands as C16–C24 fatty acids, and can be attached at the 5′, 3′ or 2′ position. Hydrophobicity rises sharply and IP-RP retention lengthens considerably, usually requiring a higher organic proportion and precautions against precipitation on the column.

6-FAM — the most common fluorescent label

The standard reporter group for probe products. The fluorophore itself absorbs in the UV and interferes with purity quantification at 260 nm, so dual wavelengths or a correction factor are needed.

Biotin — an affinity tag

Used in capture and separation applications. Moderately hydrophobic, giving a resolvable shift in IP-RP that makes the labelling rate easy to monitor.

05sgRNA and crRNA·tracrRNA — CRISPR guide RNA

5.1 Structural features and purification difficulties

A single guide RNA (sgRNA) is a long RNA of about 100 nt, made up of an approximately 20 nt targeting sequence, the crRNA repeat region, a linker loop and the tracrRNA scaffold, with extensive hairpin secondary structure. The termini usually carry 2′-OMe with PS modification against nucleases. It can also be split into crRNA and tracrRNA, synthesized separately and annealed.

5.2 Purification route

• Two or more steps are unavoidable:DMT-on IP-RP (300 Å wide-pore medium) for coarse separation → detritylation with 3% DCA → AEX or RP polishing.
• Switch the ion-pairing reagent to HAA:published comparisons show 100 mM hexylammonium acetate at pH 7.0 resolving a 60mer better than TEAA. Reference analytical conditions: hybrid silica C18, 300 Å / 2.5 μm / 4.6×100 mm, 60 °C, 30 → 63.5% ACN over 47 min; A = 5% ACN in 100 mM HAA pH 7.0, B = 80% ACN in 100 mM HAA pH 7.0.
• The pore size must be increased:100–130 Å up to 30 nt; 300 Å for 40–200 nt; 1000–4000 Å for mRNA scale. On the support side, above 100 nt a 2000 Å CPG (loading only 10–20 μmol/g) or a polystyrene support is recommended.
• Alternative strategies:fragment synthesis (blockmer, coupling 20–40 nt fragments) avoids the yield cliff of synthesizing a long chain in one pass; the in vitro transcription (IVT) route is entirely different, with an impurity profile dominated by dsRNA, truncated transcripts, residual DNA template and enzymes, and purification based mainly on AEX, reversed phase and affinity.

5.3 Analytical scheme

• HILIC-MS is particularly suitable for sgRNA:A = 100 mM aqueous ammonium acetate with 1% ACN, column temperature 75 °C, with no ion-pairing reagent to contaminate the system.
• SEC needs a large pore:a 30 nm pore diol SEC column suits 60–120mers; molecules above about 50 bases cannot enter smaller pores and elute together at the exclusion limit.
• CE / CGE Resolves long-chain RNA better than HPLC and is an important orthogonal method for releasing long-chain products.

06miRNA inhibitors / antagomirs — microRNA inhibitor

Mature miRNAs are 21–25 nt long. An inhibitor is a fully modified single strand fully complementary to the target miRNA, and an antagomir usually carries a cholesterol conjugate to improve cellular uptake. Purification and analysis follow the ASO framework directly, with two differences:

• Full modification (usually all 2′-OMe or a mix with LNA) shifts hydrophobicity upward overall, so the IP-RP gradient must start higher; the strong conformational locking of LNA units makes the secondary structure more stable, requiring a higher column temperature.
• A cholesterol-conjugated antagomir is handled as a lipid conjugate under class 04: 300 Å pore size with a C8/C4 phase, at a raised organic proportion.

07Aptamers and ApDCs

7.1 Structural features

A single-stranded DNA or RNA of 25–80 nt, obtained by SELEX, binding a target protein with high affinity through a particular three-dimensional fold (hairpins, internal loops and, above all, G-quadruplexes formed by guanine-rich sequences). They commonly carry 2′-F / 2′-OMe modifications, an inverted dT cap at the 3′ end against exonucleases, and PEGylation to extend the half-life.

7.2 Additional considerations for ApDCs

An aptamer–drug conjugate can be built in three ways: (1) covalent conjugation (amine / thiol / DBCO cyclooctyne → MMAE, paclitaxel and others); (2) direct incorporation of a small-molecule nucleoside analogue during synthesis (5-FU, gemcitabine); (3) physical non-covalent intercalation (an anthracycline intercalating into a GC region, for example). Type (1) requires separation of the unconjugated aptamer from conjugates of different DAR, and the strong hydrophobicity of the small-molecule payload makes IP-RP inevitable; type (3), being a non-covalent complex, is destroyed by any chromatography, so only free drug content and encapsulation efficiency can be controlled.

08CpG immunoadjuvant ODNs

Release points common to all three classes:endotoxin control is especially critical(these are immunostimulants themselves, and endotoxin contamination confounds efficacy readouts), as is strict sequence confirmation — the position and number of CpG motifs determine activity directly.

09Primers, probes and diagnostic-grade oligonucleotides

Particular points for labelled primers and probes:there is a very wide range of terminal modifications — functional groups (phosphate, NH₂-C6/C12, SH-C6, carboxyl), click chemistry (alkyne, azide, DBCO, maleimide, Acrydite), affinity tags (biotin, digoxigenin), lipophilic groups (cholesterol, lipids, PEG), and fluorophore/quencher pairs (FAM/TET/HEX/TAMRA/Cy series with BHQ1/2/3, Dabcyl, Eclipse, MGB). Every class of label changes the molecule's hydrophobicity, so purification of labelled products almost always falls to IP-RP, and because the label itself absorbs in the UV, purity quantification must use dual wavelengths or a correction factor.

10Product–solution overview matrix

11A framework for orthogonality between chromatographic modes

The horizontal axis is the difference in net charge (the dimension AEX resolves) and the vertical axis the difference in hydrophobicity (the dimension IP-RP and HIC resolve). Where an impurity sits on this plot determines directly which mode should be used to remove it.

Placing the impurities on this two-dimensional plot makes the purification and analytical strategy derivable:

• Lower right (N−1/N−2, N+1)— a clear charge difference and almost no difference in hydrophobicity → AEX is the workhorse and IP-RP can only assist.
• Upper left (DMT-on full-length product, residual protecting groups, lipid/GalNAc conjugates)— a large difference in hydrophobicity → IP-RP or HIC deals with it in one step.
• Centre-left (P=O desulfurization)— a slight change in charge/pKa and almost none in hydrophobicity → AEX can resolve it, IP-RP often co-elutes.
• Lower left (deamination +1 Da, abasic, diastereomers)— near zero in both main dimensions, and thereforea blind spot for conventional chromatography, requiring chemical derivatization (WAX with a cysteine analogue), ³¹P NMR or digestion-MS instead.
• Upper right (excess single strand in siRNA)— requires the size and conformation dimensions under non-denaturing conditions (SEC, non-denaturing IP-RP).

12AEX anion exchange preparative purification

The mechanism separates bynet negative chargeand resolves N−1/N−2 best. The usual range is 12–50 nt (some papers extend this to 80 nt). High-pH (11–12) denaturation has become the industry default starting point: at around pH 12 the Watson-Crick hydrogen bonds are broken and G-quadruplexes disassemble, so the separation reflects net charge and chain length purely; at the same time guanine N1-H (pKa ≈ 9.4) and thymine N3-H (pKa ≈ 9.9) are deprotonated, adding a further base-composition selectivity.

12.1 Choice of elution salt: an underrated lever

• NaCl— the most common and cheapest, with a gradient of 0 → 1.0–2.0 M; eluting power may be insufficient for fully PS oligonucleotides, which are more hydrophobic.
• NaBr— mechanistic modelling gives a selectivity coefficient K_Br/OH = 1.1 against K_Cl/OH ≈ 5 (with selectivity reversal), and the conclusion is that switching from NaCl to NaBrbrings elution appreciably earlier with narrower peaks. NaBr (0 → 1 M) is chosen as the optimal elution salt in the published patent.
• NaClO₄— the most chaotropic with the highest eluting power, suiting long chains and strong secondary structure; corrosivity and safety concerns keep industrial uptake low.Title-level evidence only
• Organic modifier— 7–15% acetonitrile in the mobile phase suppresses secondary hydrophobic interaction (10–15% in the published patent; 7% in a domestic case).
• Urea— mentioned in the literature but not adopted by any process-grade source, being avoided for the desalting burden and the risk of cyanate (carbamylation).

12.2 Published practice on denaturing conditions

13IP-RP ion-pair reversed-phase preparative purification

The oligonucleotide phosphate backbone is entirely negatively charged and is unretained on a bare reversed-phase column. An alkylamine (cation) forms an ion pair with the backbone, neutralizing the charge and providing hydrophobicity so the molecule is retained by chain length and hydrophobicity.

13.1 Two sets of measured data usable directly as starting points

13.2 DMT-on / DMT-off: the central lever in IP-RP

The 5′-DMT (MW 303) is retained as a strongly hydrophobic "handle". Because only chains that completed the final coupling carry the DMT, while every failed sequence was capped during synthesis and carries none, the selectivity changes from a small difference in chain length to the presence or absence of a large hydrophobic group, giving a very large difference in k′. Published sources note that DMT-on reversed phase is particularly effective for 40–150 nt long chains.

14HIC hydrophobic interaction chromatography: the underrated "green reversed phase"

Industry reviews place HIC alongside RP-HPLC and IEX as one of the three main purification methods at commercial scale, describing it as "similar in principle to RP-HPLC but without organic solvent"; sustainability studies also list it as a near-term green alternative to RP-HPLC.

The cost of HIC:what elutes is a concentrated ammonium sulfate solution, so the desalting burden is even greater than with AEX, and the corrosivity of ammonium sulfate to stainless steel must be assessed.

15SEC, TFF and mixed mode

15.1 The division of labour between SEC and TFF

SEC desalting (dextran gels with an exclusion limit around 5 kDa) has low volumetric productivity (loading up to about 30% of column volume) and dilutes heavily, and at industrial scale has largely been replaced by UF/DF (TFF) — the regulatory dossiers for marketed products all describe post-chromatography desalting as ultrafiltration rather than SEC. SEC's main role in the oligonucleotide field has shifted to the analytical side: duplex integrity, aggregates, and absolute molecular weight by SEC-MALS.

• Membrane cut-off:oligonucleotide molecular weights run from about 6–7 kDa (20mer) to about 14 kDa (siRNA duplex), and by the usual TFF rule of one third to one fifth of the target molecular weight this gives a 1–3 kDa MWCO (5 kDa is usable for an siRNA duplex). Inference The specific figures are not given directly in authoritative sources and must be confirmed experimentally.
• A neat piece of design:a published patent uses 100 mM sodium acetate pH 3 with 40% ethanol as the TFF buffer, removing acid-labile protecting groups while desalting — the TFF unit doubling as a reaction step and saving an operation.

15.2 Mixed mode (MMC)

See the three routes for class 03 GalNAc conjugates. The value of mixed mode is that where two species show no difference in any single physical property (the charge of conjugated versus unconjugated species, or the charge of a peptide conjugate after neutralization), only a stationary phase that can exploit several interactions at once will separate them. Besides triple-ligand phases, the literature also reports orthogonal AEX with mixed mode, and RP / weak anion exchange mixed mode, for purifying modified oligonucleotides.

16Orthogonal combination strategies and evidence from marketed drugs

16.1 Published process information for marketed and clinical-stage drugs

Information gap:the specific purification conditions for most marketed siRNA and ASO productshave no published source— the drug substance manufacturing section of a review report usually describes the unit operations without giving parameters. Targeted searching of a company's "process for the preparation" patent family is the only workable way to supplement this.

17The scale ladder, DAC and operating parameters

Published industry reference: synthesis column configurations cover the full 100–800 mm range; the purification side has chromatography systems, industrial prep-HPLC, fully automated TFF and two-column chromatography systems (MCSGP); batch sizes of the order of 10 kg with annual capacity reaching ton scale Manufacturer's published information, not independently verified

17.1 Dynamic axial compression (DAC) columns

DAC keeps the bed compressed with a hydraulic piston, eliminating voids from solvent shock and settling, and is the only practical way to pack rigid small-particle (10–30 μm) media at industrial scale. Published measurement: 50 mm ID, 200 mm bed height, ODS 15 μm / 120 Å, 35% slurry concentration, 6.4 MPa packing pressure, with efficiency and column pressure essentially unchanged after seven cycles of repacking.

18Chromatography media selection: from parameters to grades

The preceding chapters give parameter windows (pore size, carbon load, capacity, pH tolerance, linear velocity). This chapter maps those parameters onto specific purchasable grades. One point has to be made first:no single brand covers the whole chain of oligonucleotide purification— silica-based media dominate the reversed-phase stage (IP-RP), while AEX, HIC and process-scale desalting must use soft gel or polymer media. Media selection is therefore inherently a multi-brand exercise.

18.1 Daisogel: the workhorse for silica-based preparative reversed phase

The Daisogel product structure follows two clear rules, and once these are understood selection can largely be done in one's head: pore size and surface area correspond in fixed pairs（60/100 Å→450，120→300，200→200，300→100，1000→25，2000→15 m²/g）； carbon load falls as pore size rises(the ODS-RPS series: 60 Å 19% → 120 Å 17% → 200 Å 12% → 300 Å 7%). Particle sizes cover 3–5 µm (analytical), 7–20 µm (preparative) and 40–60 µm (coarse separation) — the only one of the four brands with a complete particle size range for industrial scale-up.

18.2 COSMOSIL: a selectivity toolbox for the analytical and semi-preparative end

The strength of COSMOSIL is the diversity of its phase chemistry and its analytical-grade performance, including an RNA series developed specifically for nucleic acids. Naming rule: the number prefixing the grade is the particle size in µm — 5C18-MS-II is a 5 µm C18-MS-II.

18.3 HPLCONE / SILICAONE: analytical columns and method development

HPLCONE is a column brand built on SILICAONE ultra-pure spherical fully porous silica. The published substrate specifications are silica purity ≥99.99% with total metal impurities below 10 ppm, with 3 µm and 5 µm analytical particle sizes, and a usable pH range of 1.5–12 for the ODS stationary phase.

18.4 JNC Cellufine: filling in AEX, HIC and process-scale desalting

The first three brands are all silica-based reversed-phase systems. The three parts genuinely missing from the oligonucleotide purification chain — denaturing AEX, HIC and process-scale desalting — need soft gel media. Cellufine is a crosslinked cellulose substrate, and the trade-offs against agarose, polymethacrylate and PS-DVB need to be set out clearly.

Trade-offs of a cellulose substrate against the alternatives

18.5 Example media configurations for the whole chain, by product type

Note: this table is a selection suggestion based on each manufacturer's published parameters, intended to narrow the range for bench trials, and is not a performance commitment. All capacities, purities and yields must be calibrated at bench scale for the specific molecule and scale; the Cellufine DBC data are protein-based and must be re-measured for oligonucleotides.

19A four-tier orthogonal framework

20IP-RP analysis: the ion-pair reagent system is the first decision

20.1 Column temperature, pore size and hardware

• Three reasons for a column temperature of 50–80 °C:(1) it breaks secondary structure (hairpins, G-quadruplexes), avoiding split and tailing peaks; (2) it raises mass transfer and lowers viscosity, improving efficiency for large molecules; (3) it makes N−1 and other homologues elute by chain length in an orderly way, reducing sequence dependence. A conventional single-stranded ASO defaults to 60 °C; GC-rich and G-rich sequences use 80–90 °C; siRNA denaturing methods use 75 °C and non-denaturing methods 40 °C or below (typically 20 °C). Note that high temperature with high pH accelerates stationary phase degradation: a silica-based column at pH 8.5–9.5 and 80 °C is at the tolerance boundary, so hybrid particles or a polymer substrate should be chosen.
• Manufacturers disagree on pore sizeTo be measured: the general rule is "300 Å for biomolecules above 5 kDa", but oligonucleotide-specific guidance says "130 Å suits single strands of 2–100mer". In practice: use 130 Å for 16–25mer ASOs (the larger surface area usually gives better resolution); use 300 Å for non-denaturing siRNA analysis, sgRNA (about 100 nt) and lipid or GalNAc conjugates. The disagreement suggests 5 kDa is a conservative general boundary, and that the chain flexibility of an oligonucleotide lets it enter smaller pores than a globular protein of the same molecular weight — comparison by measurement on the specific molecule is advisable.
• Metal adsorption is a major problem:oligonucleotides have a high affinity for Fe³⁺ on stainless steel surfaces, causing peak tailing, low recovery, carryover between injections and a first-injection effect. Inert hardware (high-performance surface treatment / PEEK lining / bio-inert) brings carryover below the noise level — and over the last five years this has become standard understanding.

21AEX analysis, HILIC, SEC, LC-MS and CE

21.1 AEX analysis: the pellicular strong anion exchanger is the industry's most common platform

• Structure: an 8 μm non-porous solvent-compatible core with a 130 nm quaternary ammonium functionalized latex layer (5% crosslinked), column capacity about 40 μeq
• No restriction over pH 4–10; pH 2.5–4 and 10–12.5 require an equimolar co-ion; the recommended maximum for denaturing work is pH 12.4; temperature below 85 °C; pressure limit 4000 psi
• Non-denaturing gradient (pH 8): 20–25 mM Tris, NaClO₄ 70 → 202 mM over 22 min (about 5 mM/mL)
• Denaturing gradient (pH 12.4): 20 mM NaOH, NaCl 330 → 900 mM over 31.5 min (about 15 mM/mL), 1.2 mL/min; up to 100% ACN/MeOH can be added as an organic modifier

21.2 HILIC and SEC

21.3 LC-MS and CE

De-adduction strategy:Na⁺/K⁺ adducts are the main interference. In IP-RP the protonated alkylamine competes with alkali metals for backbone binding sites and suppresses adduct formation; more hydrophobic or more hindered amines (HA, DBA, DIEA) give fewer adducts; desalt during sample preparation (ethanol precipitation, desalting column, ultrafiltration); use low-sodium glassware or polypropylene throughout. An ageing mobile phase increases sodium and potassium adducts markedly — another reason for the one-week shelf life.

Sequence characterisationMcLuckey nomenclature is used (the a/b/c/d and w/x/y/z series), with a-B and w ions dominating for DNA; fragmentation depends strongly on precursor charge state, chain length, chemical modification and collision energy, and must be optimised charge state by charge state; a modified backbone (full PS + 2′-MOE, for example) can be covered more completely with ECD / EThcD. Supplementary approaches include nuclease P1 digestion with LC-MS (to assess consistency of thiophosphate stereochemistry) and cyclic ion mobility (cIMS) (to separate isomeric impurities).

The place of CE:CGE separates by migration through a sieving medium according to charge-to-size ratio, which is different from both IP-RP (hydrophobicity) and AEX (charge) — a third orthogonal axis. Its advantages are high resolution of N−1/N+1, no organic solvent and a very small sample requirement; its drawbacks are quantitative precision slightly below HPLC, difficulty coupling directly to MS, and the need to normalise peak areas by migration time. cIEF has limited value for pure nucleic acids (no isoelectric point differences) and is used mainly for protein–oligonucleotide conjugates No authoritative literature was found for routine use on drug substance; we suggest not including it in routine release。

22A systematic review of the impurity profile

The classification framework in general industry use comes from a safety working group's recommendation: it divides impurities intoA process-related impurity(residual solvents, metals, reagents, amidite-derived impurities, resin leachables, bioburden / endotoxin) andproduct-related impurities(sequence variants, backbone variants, modification variants, degradation products).

23The release testing panel

24The regulatory view: which ICH guidelines apply and which do not

Key points of the control strategy:(1) define diagnostic specified impurities (the N−1 level as an indicator of complete detritylation, for example) and list them separately in the specification; (2) set intermediate specifications and hold times; (3) justify the amidite as a starting material under ICH Q11 and control its reactive impurities; (4) base impurity qualification on stability data, safety data, the proposed clinical dose and impurity levels in the toxicology batches.

Materials are about 51% water and about 49% organic (of which wash solvents account for half);detritylation alone accounts for nearly half the material in the synthesis stage。

25How to read the yield figures correctly

26Green chemistry levers (in order of effect)

27MCSGP multi-column countercurrent gradient purification

The principle is a two-column cycle in which the mixed front- and back-shoulder fractions are recycled continuously between the columns rather than collected and reprocessed, breaking the purity–yield trade-off. Its greatest engineering advantage is that itrequires no change of resin or buffer, so an existing batch method can be carried over directly.

In an optimised single-column comparison MCSGP reached 94% yield against 60% for batch. The manufacturer's overall claims (to be treated with caution): an absolute yield gain of 10–50% at the target purity, with typically above 90% recovery at 92% purity for oligonucleotides; solvent consumption −75%, in-process testing −96%, production time −40 to −70%. Peer-reviewed evidence: a dedicated paper on two-column AEX MCSGP continuous purification of a conjugated siRNA, and a methodology study using dynamic UV control to improve robustness.

28TFF, drying and single use

• TFF is the industry standard desalting unit:in the processes of marketed products there is one UF after synthesis and another after chromatography (specifically to remove mobile phase components). Membrane cut-off is set at 1/3–1/5 of the target molecular weight, that is 1–3 kDa MWCO (5 kDa is usable for an siRNA duplex)Inference。
• TFF doubles as a reaction unit:a published patent uses 100 mM sodium acetate pH 3 with 40% ethanol as the TFF buffer, removing acid-labile protecting groups while desalting; the example scale is a 0.1 m² membrane cassette processing 4.2 L, with an acetate concentration range of 50–250 mmol/L.
• Lyophilisation is the industry default final state:the drug substance of every marketed siRNA and ASO finishes with lyophilisation. Spray drying is discussed as a continuous alternative (with clear advantages in energy use and cycle time), but no published evidence was found of a marketed oligonucleotide drug substance using spray drying。
• Single-use chromatography has limited applicability in this fieldInference: (1) commercial batches are 10–40 kg with column volumes of tens of litres, so single-use cost is uneconomic; (2) the RP route uses a high proportion of acetonitrile, limiting the compatibility of single-use plastics; (3) oligonucleotides carry no biological contamination risk, so single use's main selling points (eliminating cross-contamination and cleaning validation) are worth less. No industrial case of single-use preparative chromatography for oligonucleotides was found.

29A decision table for purification process development

30A quick reference for choosing analytical methods

31Principal information gaps (marked as such)

In writing this document we systematically recorded 10 placeswhere published information is genuinely insufficient —5 on the process side、5 on the analytical side, covering industrial-scale loading and resin lifetime, the specific purification conditions of marketed products, the breakdown of solvent usage, conflicting statements on pore size boundaries, calibration of extinction coefficients for modified species, release limits and recommended membrane MWCO values. At these pointsno inferred values have been used as filler; each is marked instead as "published information insufficient or from a single source". The item-by-item description of each gap and its scope of impact is an internal record and requires a passcode to view.

AppendixPrincipal source categories

The technical parameters and process information throughout this document are drawn from five categories of published source, totalling 42 items：Regulatory and pharmacopoeial documents（4）、 Peer-reviewed literature（13）、Patent literature（5）、Chromatography and media manufacturers' application notes / technical white papers（13）、 Industry analyses and reference works(7). All content comes from published regulatory review documents, patents and the published literature, and involves no non-public information from any particular company. The item-by-item source list (journal titles, patent families and document origins) is an internal record and requires a passcode to view.