Swapping a non-bridging phosphodiester oxygen for sulfur along the oligonucleotide backbone — nuclease resistance, plasma protein binding, and the chemistry behind every marketed antisense drug
Every DNA and RNA strand is held together by a phosphodiester backbone: a phosphate group bridging the 3′ carbon of one deoxyribose/ribose to the 5′ carbon of the next, with two non-bridging oxygens completing the tetrahedral phosphorus. This backbone is chemically optimized by billions of years of evolution for one purpose — being cut. Nucleases recognize and hydrolyze this exact linkage as part of normal nucleic acid turnover, and an unmodified antisense oligonucleotide injected into plasma is destroyed before it can reach a meaningful fraction of target cells.
Circulating and intracellular nucleases fall into two broad mechanistic classes, and both act on the unmodified phosphodiester bond:
Exonucleases (dominant in serum): • 3′→5′ exonucleases (e.g., snake venom phosphodiesterase, SVPD, used as the standard in vitro surrogate) processively remove nucleotides from the 3′ terminus — the single largest degradation route for a linear oligo in plasma • 5′→3′ exonucleases contribute less in serum but matter intracellularly • Because exonucleases attack from the termini inward, protecting just the last 2–5 linkages at each end (a "PS-capped" design) already blocks most exonucleolytic loss even without full-length thioation
Endonucleases (dominant intracellularly): • DNase I and related endonucleases cleave internal phosphodiester bonds, generating fragments from anywhere along the strand • RNase A/RNase-family enzymes rapidly degrade unmodified RNA (t½ often <1 minute in serum) — one reason RNA-based oligos almost always carry 2′-modifications in addition to backbone chemistry • Internal protection requires modifying most or all backbone linkages, not just the termini
Why chemistry, not delivery alone, is required: • Even with a lipid nanoparticle or GalNAc conjugate to solve cellular uptake, an oligo with a native backbone released into the cytoplasm or endosome still meets nucleases at every step • Degradation products (mononucleotides, short fragments) are pharmacologically inert and can trigger innate immune sensors (TLR9, cGAS) if CpG-rich, adding toxicity without benefit • Backbone modification is therefore the first and most fundamental chemistry decision in oligonucleotide drug design — it is applied before any consideration of sugar modification, conjugation, or delivery vehicle
Historical context: • Fritz Eckstein first synthesized phosphorothioate-modified nucleotides in 1966–1969 while studying nuclease mechanism, not drug stability • The therapeutic potential was recognized in the late 1980s (Paul Zamecnik, Stanley Crooke/Isis Pharmaceuticals, now Ionis) as antisense technology matured • Every antisense drug approved to date (fomivirsen 1998 through inotersen, nusinersen, and later gapmers) uses a full or partial phosphorothioate backbone as its foundational chemistry
Phosphorothioate oligonucleotides are built on standard DNA/RNA synthesizers using the same phosphoramidite cycle used for native oligos, with one substitution: after each coupling step, the transient P(III) phosphite triester is oxidized not by iodine/water (which yields P=O) but by a sulfur-transfer reagent that yields P=S. This single reagent swap, repeated at every cycle, is what converts an entire oligonucleotide into a fully thioated phosphorothioate (PS) backbone — or, combined selectively, into mixed-backbone or gapmer designs.
Phosphoramidite cycle with sulfurization:
1. Detritylation: 4,4′-dimethoxytrityl (DMT) protecting group removed from 5′-OH with dichloroacetic acid, freeing the site for the next coupling 2. Coupling: incoming phosphoramidite monomer, activated by tetrazole or a tetrazole substitute (ETT, DCI), reacts with the free 5′-OH to form a P(III) phosphite triester linkage — this is the step that determines sequence 3. Capping: unreacted 5′-OH groups (coupling failures, ~0.5–1% per cycle) are acetylated with acetic anhydride/N-methylimidazole to prevent them from extending into deletion-sequence impurities 4. Oxidation/sulfurization: the P(III) phosphite triester is converted to a stable P(V) center • Standard route (native PO): I2/H2O/pyridine oxidizes to phosphate triester (P=O) • Phosphorothioate route: a sulfur-transfer reagent is used instead – PADS (phenylacetyl disulfide) — widely used, clean byproducts – TETD (tetraethylthiuram disulfide) — fast kinetics, common on large-scale GMP synthesizers – DDTT (xanthane hydride derivative) — high sulfurization efficiency, low epimerization – Beaucage reagent (3H-1,2-benzodithiol-3-one 1,1-dioxide) — original reagent, still used in specialty applications • Reaction converts the P(III) phosphite to a P(V)=S phosphorothioate triester in typically 99.5–99.9% yield per coupling under optimized conditions
Mixed-backbone and gapmer synthesis: • Because the oxidation/sulfurization step is chosen independently at every cycle, a single synthesis run can place PO at some linkages and PS at others • Clinical gapmers typically use full PS backbone (20/20 linkages in an 18–20mer) combined with 2′-O-methoxyethyl (2′-MOE) or locked nucleic acid (LNA) wings — the backbone chemistry and the sugar chemistry are independent design axes
Stereochemistry — the Rp/Sp problem: • Standard sulfurization creates a new stereocenter at phosphorus with each PS linkage: two non-equivalent substituents (=S and –O–) around a trivalent phosphorus core produce Rp and Sp diastereomers • Uncontrolled synthesis produces a statistical ~50:50 mixture at each linkage; an 18-mer with 17 PS linkages therefore yields 2^17 (~131,000) diastereomeric species in the crude product • Diastereomers differ meaningfully in nuclease resistance (Rp linkages are generally more nuclease-resistant than Sp at the immediately adjacent bond), RNase H1 recruitment efficiency, protein binding, and hybridization thermodynamics • Stereopure (stereodefined) synthesis platforms — most notably Wave Life Sciences' PS chemistry and related chiral auxiliary or chiral catalyst methods — control each new stereocenter during synthesis, producing single-diastereomer oligonucleotides with improved and more predictable pharmacology; these remain more synthetically demanding and costly than stereorandom production
The functional payoff of sulfurization is quantified experimentally, not assumed. Standard assays expose PS oligonucleotides to purified exonucleases, endonucleases, or full biological matrices (serum, cerebrospinal fluid, cell lysate) and track intact full-length oligo over time by denaturing PAGE, anion-exchange HPLC, or LC-MS. A fully phosphorothioated 20mer typically shows serum half-life extended from ~2 minutes (PO) to 40–60+ hours (PS) — a 100- to 200-fold improvement that is the single largest stability gain available from a backbone modification alone.
Why sulfur blocks hydrolysis:
• Nuclease catalysis proceeds through a pentacoordinate phosphorane transition state stabilized by active-site divalent metal ions (typically Mg2+ or Mn2+) that coordinate the non-bridging oxygens and activate a water nucleophile • Sulfur is a larger, softer, more polarizable atom than oxygen and coordinates divalent metal cations far more weakly — this alone significantly destabilizes the transition state geometry most nucleases require • The larger van der Waals radius of sulfur (~1.8 Å vs. ~1.4 Å for oxygen) also introduces steric clash within active sites evolved for a smaller oxygen substituent • Because the effect is centered on the scissile phosphate's own geometry, resistance is roughly linkage-local: an Rp-configured sulfur at a given position more strongly protects the immediately adjacent 3′ bond than a distant one, which is why exonucleases (which must process every linkage sequentially from the terminus) are blocked more completely than endonucleases (which can cleave at any remaining unprotected internal site)
Standard assay protocols:
1. Serum/plasma stability assay: • Oligo incubated in 90–100% fetal bovine serum or human plasma at 37°C • Aliquots removed at 0, 0.5, 1, 2, 4, 8, 24, 48, 72 h • Quenched, extracted, resolved by 20% denaturing PAGE or ion-pair reverse-phase HPLC • Percent full-length band quantified by densitometry or UV254 peak area vs. t=0 • First-order decay fit yields apparent t½
2. SVPD (snake venom phosphodiesterase) exonuclease assay: • Purified 3′-exonuclease from Crotalus adamanteus venom, a standard surrogate for serum exonuclease activity • Defined enzyme units per mL, fixed oligo concentration, timepoints over minutes-to-hours • Because SVPD is a pure single-enzyme system, this assay isolates exonuclease resistance specifically, complementing the mixed-nuclease information from whole serum
3. DNase I endonuclease assay: • Tests internal-cleavage resistance separately from exonuclease resistance • Ca2+/Mg2+-dependent; typically run at low nM DNase I concentrations against µM oligo • PS backbones show smaller fold-improvements here than in exonuclease assays because the terminal-protection mechanism does not directly apply to internal cleavage sites — a rationale for combining PS backbone with 2′-sugar modifications (2′-MOE, 2′-F, LNA) that further block endonuclease and RNase-family attack
4. LC-MS linkage mapping: • High-resolution LC-MS/MS of partial digests localizes exactly which linkages survive longest, informing stereochemistry (Rp/Sp) and gapmer design decisions
The same physicochemical property that blunts nuclease attack — a bulky, polarizable, weakly metal-coordinating sulfur substituent replacing oxygen — also makes phosphorothioate oligonucleotides unusually "sticky" toward plasma and cell-surface proteins. Fully thioated oligos bind serum albumin and α2-macroglobulin with affinities in the low-to-mid micromolar range, are >90% protein-bound in circulation, and interact with dozens of intracellular and surface proteins (including a well-characterized role for the receptor-mediated uptake protein pathway used by GalNAc-conjugated ASOs). This behavior transforms the pharmacokinetics from that of a typical biologic to something closer to a small-molecule drug — with corresponding benefits and liabilities.
Protein binding mechanism: • The phosphorothioate backbone presents a highly anionic, sulfur-rich surface that engages basic and amphipathic protein-binding pockets through electrostatic and hydrophobic contacts distinct from the shallow interactions seen with native phosphodiester DNA • Human serum albumin is the dominant binding partner by abundance (~40 g/L in plasma); α2-macroglobulin and other acute-phase proteins contribute additional binding capacity • Protein binding is saturable at high local concentrations, which is one reason PS-ASO pharmacokinetics are markedly non-linear across dose ranges used in clinical development
Pharmacokinetic consequences: • Reduced renal clearance: protein-bound oligo is not freely filtered at the glomerulus, extending circulating exposure compared to a small unbound anion of similar size • Two-compartment PK: rapid initial plasma clearance (α-phase, minutes to a few hours, driven by distribution into liver and kidney) followed by a much slower terminal elimination phase (β-phase, days to weeks) reflecting release from tissue depots • Tissue accumulation: PS-ASOs concentrate heavily in liver (hepatocytes and Kupffer cells) and kidney (proximal tubule, largely via megalin-mediated reabsorption after glomerular filtration of a smaller free fraction) — this underlies both the utility of PS chemistry for liver-targeted GalNAc-conjugate drugs and the dose-limiting renal/hepatic toxicity seen in some programs • CNS/CSF administration (e.g., intrathecal nusinersen) bypasses systemic protein-binding kinetics largely, giving CSF half-life on the order of weeks to months and enabling infrequent dosing (roughly every 4 months at steady state)
Toxicity trade-offs linked to the same chemistry: • Complement activation: highly anionic PS backbones can activate the alternative complement pathway at high plasma concentrations, producing acute infusion-related reactions (flushing, hypotension) — a class effect managed clinically by slow infusion rates and dose fractionation • Pro-inflammatory cytokine induction: PS-ASOs can engage innate immune receptors and scavenger receptors on immune cells, contributing to injection-site reactions and, in some programs, thrombocytopenia • Hepatotoxicity and nephrotoxicity: tissue accumulation described above correlates with the dose-limiting toxicities observed across the PS-ASO class in chronic dosing regimens, motivating potency improvements (better sugar chemistry, conjugation) that allow lower absolute doses
Mipomersen (Kynamro), a 20mer full-PS gapmer targeting ApoB-100 mRNA approved in 2013 for homozygous familial hypercholesterolemia, illustrates the trade-off directly: weekly subcutaneous 200 mg dosing achieved durable ~25% LDL-C reduction, but a boxed warning for hepatotoxicity (dose-dependent hepatic fat accumulation, documented in >60% of patients by imaging) and common injection-site reactions (>80% of patients) — both mechanistically linked to full-length PS backbone protein interactions — contributed to its 2018 market withdrawal in the US, underscoring why later-generation ASOs pair PS backbone with more potent sugar chemistries to permit lower, less frequent dosing.
Nuclease resistance and protein binding are necessary but not sufficient for a functional antisense drug — the oligonucleotide also needs a mechanism to silence its target. The dominant clinical architecture is the gapmer: a central block of ~8–10 unmodified or minimally modified DNA nucleotides (the "gap") flanked by two short wings (typically 5 nucleotides each) bearing a 2′-sugar modification such as 2′-MOE or locked nucleic acid (LNA). The entire construct — gap and wings alike — carries a full phosphorothioate backbone. Binding to the complementary mRNA forms a short RNA:DNA heteroduplex within the gap, which is specifically recognized and cleaved by endogenous RNase H1, permanently destroying the target transcript in a catalytic, multi-turnover fashion.
Why the gap must stay unmodified DNA: • RNase H1 is a Mg2+-dependent endonuclease that specifically recognizes the minor-groove geometry of an RNA:DNA heteroduplex — it does not efficiently cleave RNA:RNA or RNA:2′-modified-DNA duplexes • 2′-MOE and LNA wings dramatically increase binding affinity (each LNA substitution can raise duplex Tm by 2–8°C) and further boost nuclease resistance, but a 2′-modified strand paired with target RNA does not present the correct minor-groove width for RNase H1 catalysis • The unmodified central DNA gap (minimum ~5–6 contiguous deoxynucleotides) preserves the correct heteroduplex geometry, so RNase H1 cleaves the target mRNA within the gap region while the PS backbone throughout the whole oligo (gap and wings) protects the drug itself from degradation before and after that catalytic event
Catalytic, multi-turnover pharmacology: • Because RNase H1 cleavage destroys the target RNA but leaves the ASO largely intact, a single oligonucleotide molecule can direct cleavage of multiple target transcripts sequentially — this catalytic mechanism (as opposed to stoichiometric steric-blocking mechanisms used by some non-gapmer, fully modified ASOs) contributes to the potency and long duration of action seen with PS gapmers in tissue • Knockdown of target mRNA of 70–90% at steady state is typical for well-optimized clinical gapmers in accessible tissues (liver, CNS)
Clinical translation timeline: • Fomivirsen (Vitravene, 1998) — first antisense drug approved by FDA, intravitreal, targets CMV IE2 mRNA for CMV retinitis; full PS backbone, first-generation (no 2′-sugar modification), withdrawn commercially as HAART reduced CMV retinitis incidence • Mipomersen (Kynamro, 2013) — ApoB-100 gapmer, subcutaneous, illustrates PS-driven hepatotoxicity trade-offs (see Stage 4) • Nusinersen (Spinraza, 2016) — not a gapmer (a splice-modulating steric-block ASO), but shares the full-PS-backbone/2′-MOE chemistry platform; delivered intrathecally for spinal muscular atrophy, dosed roughly every 4 months at maintenance due to the long CSF residence time conferred by nuclease resistance and protein binding described in Stages 3–4 • Inotersen (Tegsedi, 2018) — TTR gapmer for hereditary transthyretin amyloidosis, subcutaneous weekly dosing • The chemistry validated across this drug class — full phosphorothioate backbone plus 2′-modified wings — remains the default starting point for essentially every gapmer ASO program in clinical development today
A single RNase H1 cleavage event inside the DNA gap is sufficient to commit a target mRNA to degradation by cellular 3′→5′ and 5′→3′ exonucleases, while the PS-protected gapmer itself survives to engage additional transcripts — pharmacokinetic modeling of hepatic ASO programs estimates an effective in-tissue functional half-life for the intact drug measured in weeks, several orders of magnitude longer than the sub-2-minute plasma half-life of the unmodified phosphodiester backbone this entire chemistry platform was built to overcome.