Extended Nucleic Acid (exNA): A Novel, Biologically Compatible Backbone that Significantly Enhances Oligonucleotide Efficacy in vivo

Metabolic stabilization of therapeutic oligonucleotides requires both sugar and backbone modifications, where phosphorothioate (PS) is the only backbone chemistry used in the clinic. Here, we describe the discovery, synthesis, and characterization of a novel biologically compatible backbone, extended nucleic acid (exNA). Upon exNA precursor scale up, exNA incorporation is fully compatible with common nucleic acid synthetic protocols. The novel backbone is orthogonal to PS and shows profound stabilization against 3’- and 5’-exonucleases. Using small interfering RNAs (siRNAs) as an example, we show exNA is tolerated at most nucleotide positions and profoundly improves in vivo efficacy. A combined exNA-PS backbone enhances siRNA resistance to serum 3’-exonuclease by ~ 32-fold over PS backbone and > 1000-fold over the natural phosphodiester backbone, thereby enhancing tissue exposure (~ 6-fold), tissues accumulation (4- to 20-fold), and potency both systemically and in brain. The improved potency and durability imparted by exNA opens more tissues and indications to oligonucleotide-driven therapeutic interventions.


Introduction
Oligonucleotide therapeutics, including small interfering RNAs (siRNAs), are a new class of medicine that enable the modulation of disease-causing genes [1][2][3][4][5] . Five siRNAs have been FDA-approved, with many more candidates in late-stage clinical trials [5][6] . The basis of successful oligonucleotide therapeutic platforms is an optimized chemical architecture that imparts metabolic stabilization to support robust, safe, and sustained modulation of gene expression in a tissue of interest 1,5,[7][8][9][10][11][12][13][14] . Phosphorothioate (PS) modi cation of the oligonucleotide backbone (i.e., internucleotide phosphodiester linkages) is currently the default strategy for stabilization of the 5'-and 3'-ends of oligonucleotides, despite several backbone modi cations being discovered. Indeed, PS is the only backbone modi cation used in clinic 1,5,15 , in part because the PS backbone is structurally similar to the canonical phosphodiester (PO) backbone, enabling high compatibility with RNA-binding protein machinery, such as Argonaute 2 (Ago2), needed for gene silencing 12,15,[16][17][18][19] . Incorporating a few terminal PS-backbone modi cations signi cantly enhances 5' and 3'-exonuclease resistance in vivo, extending duration of effect from weeks to months 12,19 . However, 3'truncated metabolites are detectable in vivo following months-long duration 18, 19 . This nding suggests 3'-end degradation continues to impede long-term e cacy and durability of oligonucleotide therapeutics, perhaps because PS does not alter backbone structure enough to prevent nuclease recognition.
In nature, single carbon (methyl group) attachment to nucleobases and amino acids is used to slightly alter the local structure of DNA, RNA, and protein to modulate nucleic acid-protein interactions and regulate gene expression 20 (Fig. 1). This biological process inspired us to apply an analogous approach to the backbone of siRNA with the goal of reducing interaction with nucleases (to enhance stability) while maintaining interaction with Ago2 (to maintain e cacy).
Here, we describe the discovery, synthesis, and evaluation of a novel, structurally extended backbone chemistry named "extended-Nucleic Acid (exNA)", where an extra carbon is incorporated 3' of phosphate ( Fig. 1) [21][22][23] . We demonstrate that exNA imparts profound improvements in exonuclease stability, while minimally impacting overall structure, charge, and thermodynamic stability, of oligonucleotides. exNA is well tolerated at most backbone positions in the context of fully chemically modi ed siRNAs and is orthogonal to PS. Incorporating both exNA and PS at the 3'-terminal profoundly impacts siRNA clearance and tissue accumulation to improve potency and duration of effect. The exNA-PS backbone opens a new chapter in enhanced stabilization of oligonucleotides, widening the clinical utility and application of the oligonucleotide therapeutics class.
Impact of exNA modi cation on RNA duplex thermostability. Stability of the siRNA duplex impacts the e cacy of siRNA-mediated gene silencing. Therefore, we sought to understand the general impact of exNA on RNA duplex thermostability. We paired an RNA strand incorporating a single exNA-ribouridine (rxU) with a fully complementary strand or a strand with a single mismatch at the position opposite the exNA insert, and measured thermostability of each duplex. For the fully complementary duplex, rxU had little impact on stability (ΔT m = -3 o C), con rming that extra carbon incorporation is well accommodated in the RNA duplex structure. Notably, the destabilizing impact of rxU in the context of a G-U mismatch (wobble) was more pronounced (ΔT m = -6 o C). exNA incorporation had no effect or reduced the energetic penalty in the other mismatch contexts (ΔT m = + 5 o C and ΔT m = 0 o C for rxU-C and rxU-U, respectively) ( Fig. 2). exNA modi cation profoundly enhances 3'-and 5'-exonuclease resistance in vitro. To evaluate the impact of exNA structure on the exonuclease resistance of oligonucleotides in vitro, we incorporated one or two 2'-OMe modi ed exNA at the 3'-or 5'-terminus of a model oligonucleotide strand with PO ( ex PO) or PS ( ex PS) backbones. Terminal PO and PS modi cations without exNA were used in control compounds (see methods). We rst tested the stability of each compound against the 3'-exonuclease, snake venom phosphodiesterase I (SVPD). At the conditions tested, the PO control was fully degraded, whereas the PS control showed better stability (Half-life 0.03 vs 1.1 for 3'-PO and 3'-PS, respectively, Fig. 3A), consistent with previous reports 25 . Surprisingly, ex PO supported higher 3'-exonuclease resistance than the PS control (~ 9 versus 1.1 half-life). Furthermore, ex PS exhibited a half-life that was ~ 36-fold higher than the PS control, and ~ 1000-fold higher than the PO control. We next tested the stability of each compound against the 5'-exonuclease, bovine spleen phosphodiesterase II (BSP). PO control degraded rapidly, whereas ex PO signi cantly enhanced 5'-exonuclease resistance but was not as much as PS backbone.
( Fig. 3B) when exNA is incorporated at 1st nucleotide. When exNA incorporated at 2nd nucleotide, the stabilization effect was more prominent than introduction to 1st position. We observed identical stabilization with PS under the condition tested, which is expected as introduction of exNA at 2nd position directly modulate the structure of the rst 5'-exonuclease cleavage site.
Our results suggest that exNA modi cation could be used as an alternative to PS for 3' and 5'stabilization of oligonucleotides or used in combination with PS to impart potentially the highest stability of any biologically compatible backbone reported to date. We elected to further investigate the latter strategy.
Position-dependent impact of exNA on siRNA e cacy in vitro. During the process of siRNA loading into Ago2 and subsequent target RNA binding and cleavage, each nucleotide of the siRNA has its own speci c interaction with the proximal amino acids of Ago2 7, 16, 26-29 . To examine the impact of the extended backbone structure of exNA on siRNA e cacy in cells, we systematically replaced each nucleotide position of a fully chemically modi ed antisense strand ( ex AS1-20) or sense strand ( ex SS1-15) with either 2'-OMe-exNA-Ur or 2'-F-exNA-Ur to create a panel of exNA-containing siRNAs. Moreover, to evaluate the impact of multiple exNA insertions at the 3'-end of the antisense strand, ex AS21-24 were designed to contain one, two, three, or four 3'-end exNA inserts, respectively (Table S4). All siRNAs contained 2'-OMe or 2'-F ribose modi cations and terminal PS linkages (Fig. 4) and targeted a previously validated human Huntingtin (Htt) mRNA sequence (for all sequences, see Table S4) [30][31][32] . Interestingly, the exNA was fully tolerated proximal to the mRNA cleavage site of Ago2 (antisense position 9-10). This nding is consistent with our previous report evaluating another backbone modi cation, structurally locked internucleotide (E)-vinylphosphonate ( i E-VP) 36 . This, combined with the lack of clear crystal structure resolution in this region, suggests that Ago2 can exibly accommodate both constraining and lengthening backbones at position 9-10 19 . The most exciting observation was the complete tolerance for incorporation of up to four exNA at the 3' end ( Fig. 4and Fig. S3). We even observed a slight but reproducible (with different sequences, data not shown) increase in potency (~ 2fold) with a single or double exNA incorporation. Terminal nucleotides do not contribute to target recognition [40][41][42] and the lack of full complementarity at the 3' terminal caused by exNA may protect the antisense strand from target RNA-induced Ago2 unloading 46 . It is also possible that 3'-end exNA modi cation may induce better structural tting into the Paz region of the Ago2 or lower a nity enhance the rate of catalytic cleavage. exNA modi cation block enhances stabilization of fully chemically modi ed antisense strand against 3' exonuclease.
Typically, stabilization imparted by a chemical modi cation can be further enhanced when the modi cation is incorporated in a block (i.e., multiple positions in a row) 43 . Because up to four exNA modi cations were tolerated at the 3'-end of the antisense strand in the context of chemically modi ed siRNA, we sought to better understand the differential impact of 3'-end single, double, triple, and quadruple exNA blocks on antisense strand stability. We analyzed the 3' stability of ex AS21-24 from our panel of fully modi ed siRNAs (Table S4) using the SVPD assay. We found that exNA incorporation effects were cumulative, where incorporation of two exNA provides further stabilization compared to a single exNA. Further improvement in stability for triple and quadruple exNA incorporation, however, was minimal ( Fig. S4).
Collective results from in-vitro siRNA e cacy and 3'-exonuclease resistance studies demonstrate that 3'terminal exNA block modi cation signi cantly improves 3'-exonuclease resistance without compromising Ago2 function, suggesting it might enhance overall e cacy and durability of oligonucleotides in vivo. Because enhanced stability seemed to plateau with double exNA incorporation, this block was selected for in vivo evaluation.
3'-exNA modi cation robustly improves the plasma clearance pro le of DCA-conjugated siRNA in mice.
Upon delivery into the bloodstream, fully modi ed siRNAs are predominantly degraded by serum 3'exonuclease 44,45 . Clearance kinetics from the bloodstream profoundly impact the rate of tissue distribution, where higher tissue exposure (Area Under the Curve, AUC) correlates with higher tissue accumulation. We therefore sought to evaluate the effect of 3'-exNA modi cation on oligonucleotide clearance kinetics in the bloodstream in vivo.
The previously characterized Htt-targeting fully chemically modi ed siRNAs 30 were synthesized with and without a double 3'-exNA-PS block (Table S8, Fig. 5A) and conjugated to docosanoic acid (DCA), which enables widespread tissue distribution to, and e cacy in, extrahepatic tissues 46 . A 10 mg/kg dose of DCA conjugated fully modi ed siRNAs without double exNA (D49, or PS) or with double exNA (D50, or exNA-PS) was subcutaneously administered to FVB mice (see methods), and plasma samples were collected at 0, 5, 15, and 30 minutes, and 1, 2, 3, 6, 9, and 24 hours (Fig. 5B). Incorporation of exNA had a profound impact on siRNA plasma clearance pro le; D50 exhibited a ~ 3-fold enhancement in C max (504 vs 1660 pmol/mL), 6.3-fold increase in AUC 0 − inf (164 vs 1034 pmol/mL min), and 2-fold shift (187 vs 305 min) in mean residence time compared to D49 (Fig. 5C). Our ndings con rm the signi cant, negative impact of 3' degradation during siRNA absorption and distribution, and demonstrate the effectiveness of exNA in inhibiting this degradation.
To determine whether the profound effect of exNA modi cation on plasma clearance kinetics translated into changes in tissue accumulation, mice used for measuring plasma pharmacokinetics were sacri ced at 2 weeks post-injection and siRNA accumulation in tissues was evaluated by the PNA hybridization assay 47 . D50 (exNA-PS) exhibited 9.7-, 4.7-, 3.5-, 14.9-, and 8.9-fold higher accumulation in liver, kidney, muscle, heart, and fat, respectively, compared to D49 (PS) (Fig. 5D). To con rm whether enhanced tissue accumulation of D50 is maintained over a longer time course, D49 and D50 were administered subcutaneously (10 mg/kg) and tissues were collected at 4 weeks post-injection. D50 exhibited 16-, 2-, 5.8-, 7.9-, and 6.4-fold higher accumulation than D49 in liver, kidney, heart, muscle, and fat, respectively (Fig. 5E). Our results indicate that simple addition of two carbons (double exNA modi cation) to the 3' end of siRNA chemical architecture can profoundly enhance liver and extrahepatic tissue accumulation and retention.
To determine whether the observed enhancement in tissue accumulation by 3'-exNA modi cation translated into functional modulation of target gene expression, we also quanti ed Htt mRNA expression in collected tissues at 4 weeks post injection (10 mg/kg of D49 or D50). We intentionally selected a relatively low dose and longer timeline for gene expression analysis to increase the likelihood of detecting potential differences in D49 and D50 silencing e cacy. At the selected dose and timeline, D49 did not induce statistically signi cant silencing (compared to non-targeting control) in any tissue, except for fat. By contrast, D50 supported productive silencing in all tissues tested (45%, 48%, 34% and 64% in liver, kidney, muscle, heart and fat respectively), with statistically signi cant difference observed between D49 and D50 in all tissues but heart (Fig. 5F).
To con rm that the effects are transferable to other targets, we synthesized previously-validated fully chemically modi ed (including PS) siRNA targeting myostatin without and with 3'-exNA (D52 and D53, respectively, Table S8) and found that exNA modi cation signi cantly enhanced myostatin silencing in quadriceps (Fig. S5). Thus, the effects of this new backbone stabilization on extrahepatic gene silencing e cacy are reproducible in a different target and sequence context.
3'-exNA modi cation enhances siRNA potency in mouse central nervous system (CNS). To evaluate the generalizability of the observed e cacy improvement in the context of a different siRNA scaffold, we synthesized Apolipoprotein E (ApoE) and Htt targeting siRNA in a di-valent, CNS-active con guration discovered by our group 48 . In this di-valent scaffold, two siRNAs are covalently connected through a linker at the 3' end of the sense strand. The increase in size and cooperativity of cellular interactions enables broad distribution in the CNS and potent target silencing. Fully chemically modi ed di-valent siRNA targeting ApoE and Htt were synthesized with and without 3'-end double exNA modi cation (exNA-PS or PS, respectively, Fig. 6A, Table S8).
Fully chemically modi ed di-valent siRNA targeting ApoE and Htt were previously shown to induce robust silencing at a 5-nmol dose. To increase the likelihood of detecting differences in silencing e cacy between PS and exNA-PS divalent siRNAs, we delivered compounds at lower concentrations (2.5 nmol and 1.25 nmol for ApoE-targeting and Htt-targeting compounds, respectively) to the CNS via intracerebroventricular injection (Fig. 6B). Htt-RNAs exist both in nucleus and cytoplasm, both of which can be detected by Quantigene 2.0 assay. Thus, we measured Htt protein expression level, which directly associates with cytoplasmic Htt-mRNA silencing level. At one-month post injection, gene silencing was observed in all regions (except thalamus), and silencing e cacy was enhanced by the addition of 3' terminal exNA. The effects and trends were similar for both ApoEand Htt-targeting compounds (Fig. 6C-D), con rming that the effects of exNA modi cation are sequence and target independent. To con rm whether enhancement of e cacy by exNA modi cation is maintained over a longer time frame, we injected 20 nmols of exNA-PS or PS siRNAs (Htt-targeting) into mice, and measured Htt expression at sixmonths post injection. Consistent with one-month results, silencing e cacy was signi cantly enhanced by the addition of 3' terminal exNA (Fig. 6E).
Collectively, our results suggest that 3' terminal exNA modi cation enhances siRNA e cacy, potency, and durability independent of target sequence and scaffold.

Discussion
Metabolic stability is critical to the clinical success of oligonucleotide drugs. Although many stabilityenhancing backbone modi cations have been identi ed, including phosphoroamidate morpholino oligonucleotides (PMOs), Peptide Nucleic Acids (PNAs), Phosphoramidate (PN), and Mesylphosphoramidates (PN-MS), PS is largely the only siRNA backbone modi cation used in the clinic. This is primarily because PS provides signi cant stabilization and modulates PK properties 50 without impacting PO-backbone charge and architecture (e.g., spacing between phosphates), enabling siRNA to maintain compatibility with Ago2. However, 5' and 3' exonucleases recognize nucleic acid backbone by sensing the relative distance between phosphate charges [51][52][53] , and PS modi ed oligonucleotides appear to be susceptible to 3' nucleases after months-long duration in vivo.
To investigate the strategy to further enhance metabolic stability without compromising siRNA e cacy, we describe here a novel non-natural nucleic acid backbone, exNA, where an extra carbon is inserted between the 5'-OH and 5'-carbon of the nucleoside. This extra carbon insertion has no impact on charge and does not introduce signi cant stereo-constraint, preventing disruption to Watson-crick base pairing. As a result, exNA has minimal impact on duplex thermostability, enhances tolerance for mismatches, and is compatible with Ago2. Indeed, exNA shows compatibility with Ago2 in a wider range of backbone positions in the siRNA compared to a previously published inter-nucleotide modi cation, (E)vinylphosphonate ( i E-VP), suggesting an extended exible backbone structure is better tolerated than a structurally constrained backbone. On the other hand, the extra carbon in exNA slightly extends the relative distance between backbone phosphate charges, causing a more profound stabilizing effect against 3'-exonuclease than that of PS and equal stabilization effect against 5'-exonuclease. Thus, exNA represents a new class of backbone chemistry with potential application in engineering therapeutic oligonucleotides.
Whereas exNA may inhibit initial nuclease binding to the oligonucleotide, PS induces metabolic stabilization by destabilizing the formation of reaction intermediates when the nuclease cleaves the backbone. Our results, therefore, suggest that blockage of initial nuclease binding has a more profound impact on the nuclease resistance properties of oligonucleotides than that of the stabilizing mechanism of PS. However, exNA is an orthogonal modi cation to PS, presenting an opportunity to combine exNA and PS in one linkage to inhibit nucleases by two different mechanisms. Indeed, we nd that backbones with combined exNA-PS modi cation synergistically enhances 3'-exonuclease resistance. The enhanced stability effect seems to plateau with two exNA-PS modi cations per strand, which could suggest that the rate-limiting step of 3'-exonuclease is binding to the oligonucleotide. Indeed, the co-crystal structure of 3'exonuclease with substrate shows that only the rst few nucleotides are recognized 51 . Our work suggests than an exNA-PS combinatorial backbone may represent the best option for extensive oligonucleotide stability.
Although our current evaluation focused on exNA and exNA-PS backbone modi cations in the context of siRNAs, the susceptibility to degradation by 5' and 3' nucleases is shared across the oligonucleotide therapeutics class. Thus, the enhanced stability and pharmacokinetics/dynamics (PK/PD) conferred by terminal exNA-PS modi cation should be transferable to CRISPR guides, ASOs, mRNAs, tRNAs, and other RNA therapies [54][55][56] . Interactions between oligonucleotides and their effector proteins (Cas9, RNaseH, etc.) are less transferable across the therapeutic class. Therefore, future work must systematically investigate the tolerability of exNA incorporation at internal positions in other RNA therapies.
There are numerous chemistries being explored for oligonucleotide therapies. However, the clinical utility of these chemistries depends on manufacturability. exNA modi cation is orthogonal to not only PS, 2'-OMe, and 2'-F, but also 2'-methoxyethyl (MOE), bridged (or locked) nucleic acids (BNAs and LNAs), PN, Ms-PN backbones, and other reported chemical modi cations 5,57 . Here, we focus on most commonly used 2'-OMe, 2'-F, and PS in clinic, but further characterization of other combinations is expected to produce a similar level of impact on plasma PK pro les, tissue accumulation, and potencies. Another important factor dictating utility is ease/scalability of synthesis. exNA nucleosides can be synthesized from commercially available nucleosides in a straightforward manner. Indeed, we demonstrate that exNA conversion can be done in just 2 steps (Wittig reaction and Hydroboration). exNA phosphoramidites require no speci c condition to incorporate into oligonucleotides on solid support, which can be handled the same way as commercially available nucleoside phosphoramidites.
One of our most striking observations was the impact of 3' exNA modi cation on plasma clearance kinetics, which resulted in signi cantly enhanced tissue accumulation at both 2-and 4-weeks post injection. This result is likely due to enhanced stability against 3' exonuclease and indicates that 3' degradation during absorption and distribution signi cantly impacts tissue accumulation of oligonucleotides. However, it is also possible that exNA-PS modi cation alters plasma protein binding properties to impact PK properties and tissue accumulation, which should be investigated in future work.
Oligonucleotide delivery and functionality are well established for liver, with the tri-N-acetyl-galactosamine (tri-GalNAc) conjugate serving as the basis for several FDA approved and clinically advanced siRNA drugs 5 . Therefore, demands in the RNA therapeutics eld have shifted towards delivery to extrahepatic tissues. Here, we show that, even with non-optimal conjugate mediated delivery, the enhanced tissue accumulation of exNA-PS modi ed siRNAs translates into improved functional e cacy in extrahepatic tissues. This observation was not dependent on delivery con guration, as we see it for systemic delivery with a hydrophobic conjugate and direct administration into cerebrospinal uid with a di-valent siRNA scaffold; and thus, is likely applicable to other delivery modalities (e.g., antibodies, peptides, etc.). The enhanced e cacy and durability of paired exNA and PS modi cation may therefore offer an opportunity to advance the clinical utility of RNA drugs in extrahepatic tissues.
In summary, exNA is a synthetically accessible, scalable, and biologically compatible backbone that provides profound stability against nucleases. This modi cation has the potential to expand the clinical utility of siRNA and potentially other oligonucleotide therapies to more tissues and indications. Declarations maintaining infrastructure for NMR, and Dr. Yanglan Tan and Dr. SonNgoc Nguyen for measurement of high-resolution mass analysis. For antisense strands used in the in-vitro screen, 5'-terminus was coupled with phosphorylating reagent, bis(2-cyanoethyl)-N,N-diisopropylphosphoramidite (ChemGenes) to have 5'-phosphate. Metabolic stability of 5'-terminal phosphate against endogenous phosphatase is an essential parameter that de nes potency and duration of effect of siRNA in vivo. 5'-(E)-vinylphosphonate (VP) is a well-known phosphatase-resistant 5'-phosphate analogue allowing for prolonged duration of effect of siRNA in vivo.
Thus, for all guide strands used in in-vivo experiments, 5'-vinyl tetra phosphonate (pivaloyloxymethyl) and 2'-O-methyl uridine 3'-CE phosphoramidite (Hongene Biotechnology Co., Ltd.) were coupled as a 5'terminal nucleotide to generate the 5'-(E)-VP moiety. The coupling of all monomers building blocks were conducted under standard conditions on solid support using 5-(benzylthio)-1-H-tetrazole (BTT) as an activator. Cholesterol-conjugated oligonucleotides were prepared on a 500-Å LCAA-CPG support, where the cholesterol moiety is bound to tetra-ethylenglycol through a succinate linker (ChemGenes, co.). For DCA conjugated oligonucleotide synthesis, docosanoic acid was directly attached via an amide bond to a controlled pore glass (CPG) functionalized by a C7 linker, as described 58 . For divalent sense strand synthesis, in-house synthesized solid support was used as described 48 . For all 2'-OH RNA synthesis, Thermostability assay. 1 μM guide strand and 1 μM sense strand were annealed in a 10 mM sodium phosphate buffer (pH 7.2) containing 100 mM NaCl and 0.1 mM EDTA by heating at 95°C for 1 min and cooled down gradually to room temperature. T m measurement was performed with a temperature controller. Both heating and cooling curves were measured over a temperature range from 20 to 95°C at 1.0 °C/min for three times.
Peptide Nucleic Acid (PNA) Hybridization Assay for Tissue siRNA Quanti cation. Quanti cation of antisense strands in tissues was performed using a PNA hybridization assay as described 61 . Brie y, tissues were lysed in MasterPure tissue lysis solution (EpiCentre) containing 0.2 mg/mL proteinase K (Invitrogen). Sodium dodecyl sulfate (SDS) was precipitated from lysates by adding 3 M potassium chloride and pelleted centrifugation at 4,000 × g for 15 min. siRNA in the supernatant were hybridized to a Cy3-labeled PNA probe fully complementary to the antisense strand (PNABio). Samples were analyzed by HPLC (Agilent) over a DNAPac PA100 anion-exchange column (ThermoFisher). Cy3 uorescence was monitored, and peaks were integrated. The nal concentrations were ascertained using calibration curves. as the zero position for guidance, then measured +/-0.8 mm medial-lateral and -0.2 mm anterior-posterior (AP), and nally lowered -2.5mm dorsal-ventral (DV) into the ventricle. The injection started one minute after the needle was inserted, and the needle was removed one minute after the injection was completed.
Once injection was complete, the mice were placed on a heating pad until fully awake and returned to their primary housing facility. Each animal received 0.1cc/10g KETOFEN (ketoprofen) subcutaneously once fully awake. All mice were housed individually post-op to allow healing and avoid losing stitches.
Rodent tissue collection. For e cacy studies, mice were sacri ced one month after ICV siRNA injections. The mice were deeply anesthetized with tribromoethanol and perfused intracardially with 20mL cold 1X PBS buffer. The brains were immediately sliced into 1.0mm sections using a brain matrix (Kent Scienti c Corporation, RBMA-200C). Brain sections oating in cold 1X PBS were punched using 1.5mm sterile biopsy punches. The brain punches were submerged in RNAlater or ash frozen on dry ice and stored at -80°C for mRNA and protein analyses.
Statistical methods for CNS mRNA and protein silencing data. To assess the average effect of exNA modi cation, the data was t to a Gamma family generalized mixed effects linear model with a log link, using the R package glmmTMB 63 . In this model brain region and exNA modi cation status were used as xed effects and mouse as a random effect Scheme Scheme 1 is available in the Supplementary Files section. Figure 1 Chemical structure of extended Nucleic Acid (exNA) with methyl inserts (permanent structural modulation) and natural epigenetic modi cation with methyl adduct (removable by endogenous demethylase).    Tables S4 and S5 for sequences, and Table   S6 for all IC 50 values used to calculate e cacy changes in the gure. a Potency change calculated based on IC 50 by lipid-mediated uptake.