Novel Cluster and Monomer-Based GalNAc Structures Induce Effective Uptake of siRNAs in Vitro and in Vivo

Abstract

GalNAc conjugation is emerging as a dominant strategy for delivery of therapeutic oligonucleotides to hepatocytes. The structure and valency of the GalNAc ligand contributes to the potency of the conjugates. Here we present a panel of multivalent GalNAc variants using two different synthetic strategies. Specifically, we present a novel conjugate based on a support-bound trivalent GalNAc cluster, and three others using a GalNAc phosphoramidite monomer that was readily assembled into tri- or tetravalent designs during solid phase oligonucleotide synthesis. We compared these compounds to a clinically used trivalent GalNAc cluster both in vitro and in vivo. In vitro, cluster-based and phosphoramidite-based scaffolds show a similar rate of internalization in primary hepatocytes, with membrane binding observed as early as 5 minutes. All tested compounds provided potent, dose-dependent silencing, with 2-4% of injected dose recoverable from liver after 1 week. The two preassembled trivalent GalNAc clusters showed higher tissue accumulation and gene silencing relative to di-, tri- or tetravalent GalNAc conjugates assembled via phosphoramidite chemistry.

Graphical Abstract

Introduction

Conjugation of N-acetylgalactosamine has become a major clinical strategy for delivery of oligonucleotides to hepatocytes.^1-6 Such conjugates are efficiently internalized by binding to the asialoglycoprotein receptor (ASGPR), which naturally recognizes and internalizes glycoproteins containing terminal galactose (Gal) or N-acetyl galactosamine (GalNAc) moieties.⁷ Tetravalent and trivalent GalNAc clusters have much higher affinity for ASGPR than smaller sugar clusters.^8-9 ASGPR is constitutively endocytosed and recycled about every 15 minutes,¹⁰ and the rate of internalization is increased by ligand binding.¹¹ The ligand is released inside endosomes due to reductions in pH¹² and Ca²⁺ concentration.¹³

Pioneering approaches to the synthesis of galactose and GalNAc–oligonucleotide conjugates were developed in the late 1990s by groups at Johns Hopkins University,^14-17 Leiden University,^18-20 ISIS Pharmaceuticals (later Ionis)^21-22 and Ribozyme Pharmaceuticals (later Sirna, then Merck).²³ Most of this work was focused on synthetic approaches to GalNAc assembly with only minimal data on the in vivo uptake and efficacy of the oligonucleotide conjugates. This lack of in vivo efficacy was partly due to the relatively early stage of development of oligonucleotide backbone chemistry at the time. Nevertheless, as backbone chemistry continued to evolve,^24-28 it became clear that improved backbones in combination with GalNAc held significant clinical promise.^29-30

Thus, a patent application on the structure of a GalNAc conjugate suitable for delivery of siRNAs to hepatocytes was published in 2009³¹ and two landmark 2014 papers demonstrated the dramatic impact of GalNAc conjugation on the in vivo efficacy of ASOs³² and siRNAs.³³ Further chemical tuning of the siRNA scaffold enhanced both potency and duration of effect in vivo.^{29-30, 34} At least nine GalNAc conjugates are currently in clinical trials.^{28, 35}

This clinical promise has spurred a great deal of exploration into structure-function relationships of GalNAc conjugates,³⁶ as well as work exploring synthetically more accessible GalNAc structures that can be readily applied. Synthetic approaches to multivalent GalNAc conjugates can largely be divided into two families: those based on assembly of multiple monomeric GalNAc-containing phosphoramidites,^{23, 36-40} and those based on a pre-assembled, support-bound or post-synthetically conjugated GalNAc cluster.^{22, 32-33, 36, 41-46} In this work we describe the synthesis of a novel trivalent GalNAc cluster and a novel monomeric GalNAc phosphoramidite that we have used to assemble multimeric GalNAc clusters on solid phase (Figure 1). We compare their uptake and efficacy properties with a clinically used trivalent GalNAc construct.

Figure 1.

Overview of GalNAc conjugate structures used in this work. Left: Cluster-based approaches: Gal₃C1 is a clinically used proline-based GalNAc conjugate.³³ Gal₃C2 is a novel aminopropanediol-based trivalent GalNAc cluster. Right: Monomer based structures: Gal₃M1 and Gal_2-4M2 are clusters assembled using a GalNAc monomer phosphoramidite without (M1 series) and with (M2 series) TEG spacer.

Results and Discussion

We synthesized and tested six multivalent GalNAc conjugate structures using two strategies: pre-assembled trivalent clusters or GalNAc-phosphoramidite-based solid phase cluster assembly. The pre-assembled cluster strategy requires more solution-phase synthetic work, but allows diverse types of chemical linkages to be used within the cluster. The monomer (GalNAc phosphoramidite) strategy allows a simpler solution-phase synthesis and the rapid solid-phase synthesis of conjugates of different valency and spacing, but requires the inclusion of phosphate linkages within the cluster and involves wastage of a typically large excess of GalNAc phosphoramidite required during solid phase synthesis.

Synthesis of trivalent PEG-spaced GalNAc cluster

The geometry and hydrophobicity of multivalent GalNAc structures has been shown to affect binding affinity toward ASGPR.⁵ We designed a trivalent GalNAc cluster Gal₃C2 with two variations from the clinically used trivalent structure Gal₃C1: the cyclic hydroxyproline moiety was replaced with an acyclic aminopropanediol group, and the amidoalkane linker proximal to each GalNAc sugar was replaced with a triethylene glycol moiety (Figure 1). This had the effect of lengthening the structure by two atoms relative to the clinically used structure, potentially adding flexibility and reducing hydrophobicity.

To synthesize this support-bound GalNAc cluster, we first coupled monobenzylated dodecanedioic acid with protected Tris-based synthon 1, then deprotected the tert-butyl esters in one pot to yield the tri-acid 3 in 89% yield (Scheme 1). To this compound we coupled peracetylated GalNAc functionalized at the anomeric position with a triethylene glycol aminolinker, to yield the tri-GalNAc cluster 5 in 78% yield. We removed the benzyl group by catalytic hydrogenation in near-quantitative yield, then coupled to the amino group of 3-O-DMT-protected 1-amino-2,3-propanediol. The secondary hydroxyl group was then used as an attachment point to controlled pore glass (CPG) support to yield target structure 8 at a loading of 42.9 μmol/g. This solid support was effective for the synthesis of oligonucleotides in high yield and purity.

Scheme 1.

Synthesis of a novel trivalent GalNAc cluster (Gal₃C2) on CPG support.

Synthesis of GalNAc monomer phosphoramidite and associated designs

To explore the impact of GalNAc valency and the importance of monosaccharide spacing, we also developed a GalNAc monomer phosphoramidite. We synthesized a GalNAc phosphoramidite containing a short amidoalkane linkage, with the idea that we could increase the spacing within the final clusters by interspersing a commercially available tetraethyleneglycol phosphoramidite as desired (Figure 1, compare Gal₃M1 with Gal_2-4M2).

The monomeric GalNAc phosphoramidite was based on threoninol (Scheme 2). Fmoc-L-threoninol was quantitatively converted to the benzylidene acetal 10 by treating with benzaldehyde dimethyl acetal and catalytic camphor sulfonic acid. The Fmoc protecting group was then removed using a solution of 1% DBU and 1% piperidine. Attempts to begin from unprotected threoninol, avoiding Fmoc protection altogether, was less successful due to the poor solubility of threoninol as well as the challenge of selective acetal formation in the presence of a free amine.

Scheme 2.

Synthesis of phosphoramidite 16 and solid support 17 for the monomer-based route to GalNAc conjugates.

We then treated the amine 11 with ε-caprolactone in the presence of DIPEA. Reaction at 75°C for 3 days produced the extended hydroxyl–amide compound 12. We then treated 12 with the oxazoline of triacetyl-GalNAc⁴⁷ in the presence of trimethylsilyl triflate to yield the β-glycoside 13 in 76% yield. The benzylidene acetal was then removed and the primary alcohol of the threoninol moiety protected as the dimethoxytrityl (DMT) derivative 15. The secondary alcohol was either phosphitylated (to produce 16) or succinylated and coupled to solid support (compound 17). The phosphoramidite was synthesized in seven steps and 40% overall yield.

Using a combination of the support-bound monomer and the monomer phosphoramidite, we synthesized several GalNAc-modified oligonucleotides (numbered Gal₃M1 and Gal_2-4M2, Figure 1 and Table S1). For the M2 series, we used a commercially available tetraethylene glycol phosphoramidite to provide additional space between GalNAc units. The phosphoramidite 16 could also be used with branching phosphoramidites³⁷ or in loop structures.

We annealed the GalNAc-modified sense strands with complementary antisense strands (Table S1) and used these siRNAs for all subsequent experiments. We used fully chemically modified siRNAs based on the alternating pattern of 2'-O-methyl RNA (2'-O-Me-RNA) and 2'-fluoro RNA (2'-F-RNA) first identified by Allerson et al.⁴⁸ We also included phosphorothioate linkages at all termini since terminal stabilization contributes to potency and duration of effect in conjugate-mediated delivery.²⁹ The duplexes consisted of 20 base pairs and a 2-nucleotide overhang on the 3'-end of the guide strand only. For fluorescence-based biodistribution, a Cy3 dye was included at the 5'-end of the sense strand. For gene silencing experiments and evaluation of biodistribution using the PNA hybridization assay, the siRNAs were not dye-labeled.

In vitro kinetics of GalNAc-siRNA uptake

To determine whether our novel cluster- and monomer-based GalNAc scaffolds were capable of siRNA delivery and comparable to a clinically used trivalent cluster, we carried out a study on the kinetics of conjugate uptake in primary rat hepatocytes in vitro.

Primary rat hepatocytes were obtained by treating fresh rat liver with collagenase then purifying the hepatocytes using Percoll sedimentation. After plating, we treated the cells with Cy3-labeled, cyclophilin B (PPIB)-targeted, GalNAc-modified siRNAs Gal₃C1, Gal₃C2 and Gal₄M2 and monitored their uptake by microscopy at 5, 10, 15, 30 and 60 minutes. At the appropriate timepoints, the cells were washed with phosphate-buffered saline, fixed with paraformaldehyde, and imaged by fluorescence microscopy.

The three GalNAc conjugates showed very similar kinetics of uptake (Figure 2 and Figure S1). Specifically, within 5 min the cells show significant association predominantly with cell surfaces (see inset figures) while the amount of internalization increases over time, consistent with a recycling time of ~15 min.¹⁰ The total amount of fluorescence associated with the cells continued to increase throughout the 60 min experiment. The kinetics and extent of uptake were similar for the three compounds. Therefore, we chose to investigate the uptake and efficacy of both the cluster- and monomer-based scaffolds in vivo.

Figure 2.

GalNAc conjugates Gal₃C1, Gal₃C2 and Gal₄M2 associate with primary rat hepatocytes in vitro, as monitored by fluorescence microscopy. All three clusters show similar kinetics of uptake and extent of association with hepatocytes under these conditions. Quantitation of the fluorescence signal is shown in Supporting Figure S1.

In vivo biodistribution by fluorescence microscopy

To explore the liver accumulation of the GalNAc conjugates in vivo, we prepared five Cy3-labeled GalNAc conjugates and injected them subcutaneously into FVBN/J mice at a dose of 10 mg/kg. After 48 h, tissues were formalin-fixed, paraffin-embedded, sectioned and imaged (Figure 3).

Figure 3.

Conjugates Gal₃C2 and Gal₄M2 appear to show the highest level of association with hepatocytes by fluorescence microscopy of tissue sections 48 h after subcutaneous injection of all five GalNAc conjugates. Fluorescence was quantitated using Fiji software and is shown below. Significance (one way ANOVA) is calculated relative to PBS (**** P<0.0001).

We observed significant differences in the apparent liver accumulation of dye-labeled siRNA conjugates depending on the structure of the GalNAc cluster. We observed that cluster Gal₃C2 (containing longer, more hydrophilic arms) showed better liver accumulation (i.e. higher fluorescence) than Gal₃C1. For monomer-based assemblies, the inclusion of TEG spacers did not significantly impact liver accumulation in the context of trivalent conjugates. But the tetravalent conjugate (Gal₄M2) showed significantly higher liver fluorescence compared to all of the trivalent conjugates (both cluster based and monomer-based). With these fluorescence-based uptake data in hand, we synthesized non-fluorescently labeled compounds and conducted studies of efficacy and of liver accumulation using a PNA hybridization assay to recover the antisense strand, a method which is not dependent on fluorescently-labeled siRNAs.⁴⁹

In vivo efficacy and liver accumulation of unlabeled siRNA conjugates

Non-dye-labeled conjugates Gal₃C1, Gal₃C2 and Gal₄M2 were injected subcutaneously into FVBN/J mice at 2.5 or 10 mg/kg. After one week in the liver was measured using the QuantiGene bDNA assay.⁵⁰

All three GalNAc conjugates showed dose-dependent gene silencing relative to PBS-treated animals or non-targeting control GalNAc conjugates of identical configuration (Figure 4a). The trivalent clusters Gal₃C1 and Gal₃C2 showed a slightly higher degree of silencing than the tetravalent compound Gal₄M2; this difference did not reach statistical significance when all samples were analyzed by one-way ANOVA with Bonferroni’s post-hoc analysis, though it did reach significance (p < 0.0001) after post-hoc reanalysis of a smaller subset of the data (Gal₃C1 vs. Gal₄M2, 10 mg/kg doses only). Since compound Gal₄M2 showed the highest degree of distribution to hepatocytes by fluorescence imaging (Figure 3) after 48 hours, this trend toward lower silencing surprised us. Therefore, we measured the liver accumulation using a PNA hybridization assay,⁴⁹ a method which is not dependent on the presence of a fluorescent dye (Figure 4b). Using this method, the tissue concentrations of siRNA were parallel to the silencing efficacy when both measured after one week.

Figure 4.

Trivalent GalNAc clusters Gal₃C1 and Gal₃C2 show higher efficacy and tissue accumulation than linear tetravalent conjugate Gal₄M2. (a) Dose-dependent silencing of cyclophilin B after a single subcutaneous dose in mice, compared to PBS or a non-targeting control (**** P<0.0001). (b) Liver accumulation based on PNA hybridization assay.

For GalNAc conjugates of antisense oligonucleotides, it was recently shown that lower valent clusters had higher-than-expected efficacy: Seth et al. showed that ASOs with divalent GalNAc conjugates showed comparable efficacy to the trivalent conjugates.^51-52 This might be unique to applications in antisense oligonucleotides, since the single-stranded, phosphorothioate-modified backbone can also directly interact with cell surface receptors (including ASGPR).^52-53 To explore whether this also applies to siRNAs, we took advantage of the modular nature of our M series of conjugates to compare the efficacy and liver accumulation of di-, tri- and tetravalent siRNA conjugates (Figure 5).

Figure 5.

Linear di-, tri- and tetravalent siRNA conjugates show dose-dependent silencing (a) which correlates with liver accumulation as measured by a PNA hybridization assay (b). (**** P<0.0001).

At the higher dose of 10 mg/kg, all three linear conjugates showed a high and comparable extent of gene silencing. However, at the lower dose of 2.5 mg/kg, the divalent conjugate Gal₂M2 showed significantly reduced silencing (Figure 5a). This correlated with a trend toward lower tissue accumulation by this conjugate (Figure 5b).

Calculations using the liver tissue concentration, the total injected dose, liver weight and mouse weight showed that approximately 4% of the 10 mg/kg injected dose of the trivalent conjugates could be recovered from the liver (Table 1). The linear conjugates (M-series) showed about 2% of injected dose in liver at both doses, and showed only a slight increase with increasing GalNAc valency. This suggests that the geometry of the GalNAc conjugate may be more important than its valency (for valencies ≥3), consistent with the observations of Westerlind et al.⁵⁴

Table 1.

Percentage of injected dose of antisense strand of non-dye-labeled siRNA recovered from liver.^a

Conjugate	Injected Dose
	2.5 mg/kg	10 mg/kg
Gal2M2	1.7 %	2.0 %
Gal3M2	2.4 %	2.0 %
Gal4M2	2.1 %	2.2 %
Gal3C1	2.8 %	3.6 %
Gal3C2	2.0 %	4.6 %

^aCalculated from the injected dose and liver tissue concentrations shown in Figures 4-5, based on 1.5 g liver weight and 20 g total mouse weight.

Conclusions

We have synthesized a variant cluster-based GalNAc on solid support and a GalNAc monomer phosphoramidite which can be assembled, with or without spacers, to produce multivalent GalNAc conjugates. Both types of conjugates bound cultured primary hepatocytes and led to effective silencing of an endogenous gene in vivo. Importantly, fluorescence-based estimates of liver accumulation did not correlate with accumulation of unlabeled compounds as measured by a PNA hybridization assay or with the efficacy of those unlabeled compounds. However, there was a good correlation between liver accumulation of the unlabeled compounds and the conjugate efficacy.

All conjugates, including di-, tri- and tetravalent linear assemblies of GalNAc sugars based on assembling GalNAc phosphoramidites, showed potent, dose-dependent silencing in the liver. The divalent GalNAc conjugate showed reduced potency relative to the tri- and tetravalent conjugates, which contrasts with observations using single-stranded antisense oligonucleotides.^51-52 Preassembled trivalent clusters showed higher liver accumulation and efficacy than linear, phosphoramidite-based conjugates. Nevertheless, the activity of the phosphoramidite-based conjugates was still very high, and given the synthetic ease and versatility of this phosphoramidite approach, these compounds may enable a wide variety of groups to study drug delivery using GalNAc conjugates.

Materials and Methods

Animal studies.

FVBN/J mice were used in all mouse studies, in compliance with UMMS IACUC protocol #A-2411. All in vivo studies were blinded. For biodistribution studies, equal numbers of male and female animals (n = 2 per compound) were injected subcutaneously with 10 mg/kg Cy3-labeled siRNA or PBS. After 48 hours, animals were euthanized and perfused with PBS. Liver tissue was paraffin-embedded, sliced into 4 μm sections, and stained with DAPI to visualize nuclei. Fluorescence images were acquired with a Leica DMi8 inverted microscope (40x). PBS liver sections were used to set background fluorescence settings. As there were no apparent sex-dependent differences in compound biodistribution, female animals were used in subsequent experiments. For efficacy studies, female animals (n = 6 per compound) were injected subcutaneously with 2.5 or 10 mg/kg of unlabeled siRNA (cyclophilin B-targeting or non-targeting sequences), or PBS. After one week, animals were euthanized and livers were collected for mRNA quantification as described previously, using the QuantiGene branched DNA assay.⁵⁰ Briefly, 1.5 mm punch biopsies were lysed in QuantiGene homogenizing buffer with added Proteinase K. Liver lysate was analyzed for target (cyclophilin B, PPIB) mRNA levels, which were normalized to housekeeping (hypoxanthine guanine phosphoribosyltransferase, HPRT) mRNA levels and presented as percent of PBS-treated control.

Statistical Methods.

Data were analyzed using GraphPad Prism 6 software. In vivo data were analyzed using one-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test. Statistical comparisons reported against the non-targeting controls.

Cell culture & kinetic study.

Primary rat hepatocytes were obtained as previously described.⁵⁵ Briefly, rat livers were perfused with collagenase prior to dissection. The resultant cell suspension was filtered and hepatocytes were purified using Percoll gradient sedimentation. Hepatocytes were plated in serum-free culture medium (Lonza) at 25,000 cells per well in a 96-well cell imaging plate (Eppendorf) and maintained at 37°C and 5% CO₂. After four hours, plated hepatocytes were stained with NucBlue Live ReadyProbes reagent for nuclear staining for 15 minutes. Medium was then replaced with serum-free culture medium containing 5 mM CaCl₂ and 0.5 μM of siRNA with an untreated control. At each appropriate timepoint, cells were washed three times with phosphate-buffered saline (PBS), fixed with 4% paraformaldehyde, and rinsed three additional times with PBS. Fluorescence images were acquired with a Leica DMi8 inverted microscope (40x magnification). Untreated cells were used to set background fluorescence settings.

Image fluorescence quantitation.

After image acquisition, images were exported in 16-bit TIFF format from the Leica LAS X software and processed in Fiji v1.51n⁵⁶ to quantify the kinetics of oligo uptake into cells. Briefly, the freehand tool was used to manually select individual cells and the mean fluorescence intensity was obtained by using the “Measure” function. Background fluorescence was corrected for by measuring the mean fluorescence of an area with no cells and subtracting this value from the cellular fluorescence intensity.

Synthetic organic methods.

CPG support allowing synthesis of oligonucleotides containing the clinically used trivalent GalNAc cluster C1 was synthesized according to the literature method.³³

3,3'-((2-(12-(benzyloxy)-12-oxododecanamido)-2-((2-carboxyethoxy)methyl)propane-1,3-diyl)bis(oxy))dipropionic acid (3).

12-(benzyloxy)-12-oxododecanoic acid⁵⁷ (2) (1.50 g, 4.89 mmol) was added under magnetic stirring to the solution of amine 1⁵⁸ (2.15 g, 4.25 mmol) in CH₂Cl₂ (20 mL). After full dissolution of solids, EDC·HCl (0.98 g, 5.1 mmol), HOBt (0.69 g, 5.1 mmol) and Et₃N (0.77 mL, 5.53 mmol) were added and the reaction mixture was stirred overnight. Then it was diluted with CH₂Cl₂ (100 mL), and washed with water (3×50 mL), 10% citric acid in water (3×50 mL), saturated sodium bicarbonate in water (3×50 mL), brine (50 mL). After drying with sodium sulfate, solution was evaporated to obtain amide (2) as colorless oil that was further used without purification. The oil was dissolved in 25% TFA in CH₂Cl₂ (20 mL) and the reaction mixture was stirred overnight. After evaporation, the viscous oil was co-evaporated with toluene (2×20 mL), then dissolved in ethyl acetate (3 mL) and precipitated with petroleum ether (16 mL). After cooling (+4°C for 2 h), the liquid was decanted, the solid was re-dissolved in ethyl acetate (3 mL) and precipitation was repeated to yield the product (3) as a white solid (2.36 g, 89%). R_f 0.2 (ethyl acetate); ¹H NMR (500 MHz, DMSO-d₆) δ 12.16 (s, 3H), 7.39 – 7.28 (m, 5H), 6.90 (s, 1H), 5.07 (s, 2H), 3.59 – 3.50 (m, 12H), 2.41 (t, J = 6.3 Hz, 6H), 2.33 (t, J = 7.4 Hz, 2H), 2.03 (t, J = 7.3 Hz, 2H), 1.57 – 1.47 (m, 2H), 1.46 – 1.38 (m, 2H), 1.26 – 1.19 (m, 12H). ¹³C NMR (126 MHz, DMSO-d₆) δ 172.80, 172.63, 172.51, 136.33, 128.43, 127.98, 127.94, 68.17, 66.69, 65.29, 59.52, 35.89, 34.60, 33.49, 28.92, 28.91, 28.85, 28.68, 28.53, 28.45, 25.30, 24.50; ESI-HRMS m/z calc’d for [C₃₂H₄₉NO₁₂-H]⁻ 638.3177, found 638.3156. calc’d for [C₃₂H₄₉NO₁₂+H]⁺ 640.3333, found 640.3321.

N-(12-(benzyloxy)-12-oxododecanamido)-tris[2,5,8,11,18-pentaoxa-14-aza-15-oxo-19-(3,4,6-tri-O-acetyl-2-acetamido-2-deoxy-β-D-galactopyranosyloxy)nonadecyl]methane (5).

2-(2-(2-(2-(((3,4,6-tri-O-acetyl-2-acetamido-2-deoxy-β-D-galactopyranosyloxy)ethoxy)ethoxy)ethoxy)ethylammonium trifluoroacetate⁵⁹ (4) (8.63 g, 13.56 mmol) was added to the solution of tri-acid (3) (2.36 g, 3.77 mmol) in CH₂Cl₂ (50 mL) under magnetic stirring. After full dissolution of solids, Et₃N (4.19 mL, 30.16 mmol), EDC·HCl (2.89 g, 15.08 mmol) and HOBt (2.04 g, 15.08 mmol) were added and the reaction mixture was stirred overnight. Then it was diluted with CH₂Cl₂ (100 mL) and washed with water (3×50 mL), 10% citric acid in water (3×50 mL), saturated sodium bicarbonate in water (3×50 mL) and brine (50 mL). After drying with sodium sulfate, the solution was evaporated and the residue was purified by column chromatography on silica gel with a linear gradient of CH₂Cl₂–MeOH from 100:0 to 90:10 v/v to obtain 5 as a white solid (3.96 g, 78%); R_f 0.65 (dichloromethane–methanol, 87:13 v/v); ¹H NMR (500 MHz, DMSO-d₆) δ 7.92 (t, J = 5.7 Hz, 3H), 7.81 (d, J = 9.3 Hz, 3H), 7.41 – 7.29 (m, 5H), 6.99 (s, 1H), 5.21 (d, J = 3.4 Hz, 3H), 5.07 (s, 2H), 4.96 (dd, J = 11.2, 3.4 Hz, 3H), 4.55 (d, J = 8.5 Hz, 3H), 4.05 – 3.98 (m, 9H), 3.88 (dt, J = 11.1, 8.8 Hz, 3H), 3.80 – 3.74 (m, 3H), 3.61 – 3.47 (m, 45H), 3.40 – 3.37 (m, 6H), 3.23 – 3.15 (m, 6H), 2.35 – 2.26 (m, 8H), 2.10 (s, 9H), 2.04 (t, J = 7.4 Hz, 2H), 1.99 (s, 9H), 1.88 (s, 9H), 1.77 (s, 9H), 1.55 – 1.48 (m, 2H), 1.45 – 1.39 (m, 2H), 1.26 – 1.16 (m, 12H). ¹³C NMR (126 MHz, DMSO-d₆) δ 173.84 , 172.63, 170.40, 170.11, 170.04, 169.74, 169.48, 136.36, 128.50, 128.06, 127.99, 101.02, 78.55, 70.59, 69.96, 69.88, 69.77, 69.64, 69.52, 69.16, 68.41, 68.36, 68.34, 67.36, 66.78, 65.35, 61.53, 59.57, 49.39, 38.57, 35.92, 33.53, 28.97, 28.96, 28.89, 28.74, 28.69, 28.50, 25.38, 24.55, 22.84, 20.59, 20.53, 20.50. ESI-HRMS m/z calc’d for [C₉₈H₁₅₇N₇O₄₅+H]⁺ 2153.0290, found 2153.0279. Calc’d for [C₉₈H₁₅₇N₇O₄₅+2H]²⁺ 1077.1085, found 1077.0176.

N-(11-carboxy-undecanamido)-tris[2,5,8,11,18-pentaoxa-14-aza-15-oxo-19-(3,4,6-tri-O-acetyl-2-acetamido-2-deoxy-β-D-galactopyranosyloxy)nonadecyl]methane (6).

Compound 5 (3.96 g, 1.84 mmol) was co-evaporated with absolute methanol (2 x 25 mL) and dissolved in absolute methanol (25 mL) followed by addition of 10% (w/w) palladium on activated carbon (196 mg, 0.18 mmol). The mixture was stirred under hydrogen pressure (1 atm) for 30 min, then the catalyst was filtered off and the solution was evaporated to yield 5 as a white solid (3.68 g, 97%). R_f 0.6 (CH₂Cl₂–methanol, 87:13 v/v); ¹H NMR (500 MHz, DMSO-d₆) δ 7.92 (t, J = 5.7 Hz, 3H), 7.81 (d, J = 9.2 Hz, 3H), 6.99 (s, 1H), 5.21 (d, J = 3.4 Hz, 3H), 4.96 (dd, J = 11.2, 3.4 Hz, 3H), 4.55 (d, J = 8.5 Hz, 3H), 4.06 – 4.00 (m, 9H), 3.88 (dt, J = 11.2, 8.9 Hz, 3H), 3.80 – 3.75 (m, 3H), 3.61 – 3.46 (m, 45H), 3.39 – 3.36 (m, 6H), 3.22 – 3.16 (m, 6H), 2.29 (t, J = 6.4 Hz, 6H), 2.17 (t, J = 7.4 Hz, 2H), 2.10 (s, 9H), 2.04 (t, J = 7.4 Hz, 2H), 1.99 (s, 9H), 1.88 (s, 9H), 1.77 (s, 9H), 1.50 – 1.40 (m, 4H), 1.25 – 1.19 (m, 12H). ¹³C NMR (126 MHz, DMSO-d₆) δ 174.60, 172.62, 170.38, 170.10, 170.03, 169.73, 169.46, 142.47, 101.01, 70.57, 69.96, 69.88, 69.76, 69.64, 69.52, 69.16, 68.41, 68.34, 68.33, 67.35, 66.78, 61.53, 59.56, 49.38, 38.57, 35.91, 33.73, 29.01, 28.99, 28.89, 28.82, 28.69, 28.64, 25.38, 24.58, 22.83, 20.58, 20.52, 20.50. ESI-HRMS m/z calc’d for [C₉₁H₁₅₁N₇O₄₅+H]⁺ 2062.9821, observed 2062.9832. Calc’d for [C₉₁H₁₅₁N₇O₄₅+2H]²⁺ 1031.9950, observed 1031.9962.

N-(12-((3-O-(4,4’-dimethoxytrityl)-2-hydroxypropyl)amino)-12-oxododecanamido)-tris[2,5,8,11,18-pentaoxa-14-aza-15-oxo-19-(3,4,6-tri-O-acetyl-2-acetamido-2-deoxy-β-D-galactopyranosyloxy)nonadecyl]methane (7).

1-Amino-3-O-(4,4’-dimethoxytrityl)propan-2-ol⁶⁰ (862 mg, 2.14 mmol) was added to the solution of acid (6) (3.68 g, 1.78 mmol) in CH₂Cl₂ (50 mL) under magnetic stirring. After full dissolution of solids, Et₃N (743 μL, 5.34 mmol), EDC·HCl (682 mg, 3.56 mmol) and HOBt (480 mg, 3.56 mmol) were added and the reaction mixture was stirred overnight. Then it was diluted with CH₂Cl₂ (400 mL) and washed with water (3×150 mL), 5% citric acid in water (3×150 mL), saturated sodium bicarbonate in water (3×150 mL), brine (150 mL). After drying on sodium sulfate, the solution was evaporated and the residue was purified by column chromatography on silica gel with linear gradient of CH₂Cl₂–MeOH from 100:0 to 95:5 v/v to obtain compound (7) as a white solid (3.30 g, 76%); R_f 0.6 (CH₂Cl₂–methanol, 87:13 v/v); ¹H NMR (500 MHz, DMSO-d₆) δ 7.94 (t, J = 5.7 Hz, 3H), 7.84 (d, J = 9.2 Hz, 3H), 7.67 (t, J = 5.7 Hz, 1H), 7.41 – 7.36 (m, 2H), 7.33 – 7.16 (m, 7H), 7.00 (s, 1H), 6.91 – 6.81 (m, 4H), 5.21 (d, J = 3.3 Hz, 3H), 5.01 – 4.93 (m, 4H), 4.56 (d, J = 8.5 Hz, 3H), 4.05 – 3.99 (m, 9H), 3.88 (dt, J = 11.1, 8.9 Hz, 3H), 3.81 – 3.75 (m, 3H), 3.73 – 3.71 (m, 6H), 3.71 – 3.65 (m, 2H), 3.61 – 3.45 (m, 48H), 3.40 – 3.37 (m, 6H), 3.22 – 3.16 (m, 6H), 3.00 – 2.94 (m, 1H), 2.91 – 2.81 (m, 1H), 2.29 (t, J = 6.5 Hz, 6H), 2.10 (s, 9H), 2.05 (t, J = 7.7 Hz, 2H), 1.99 (s, 9H), 1.88 (s, 9H), 1.77 (s, 9H), 1.46 – 1.37 (m, 4H), 1.23 – 1.18 (m, 12H). ¹³C NMR (126 MHz, DMSO) δ 172.61, 172.52, 170.37, 170.08, 170.01, 169.71, 169.45, 158.04, 145.15, 135.88, 129.78, 128.97, 127.80, 127.78, 127.69, 126.61, 113.12, 112.82, 101.00, 85.21, 70.58, 69.95, 69.87, 69.76, 69.63, 69.51, 69.14, 68.84, 68.40, 68.34, 67.35, 66.77, 65.78, 61.52, 59.55, 49.37, 42.64, 38.56, 35.97, 35.91, 35.42, 29.06, 29.03, 28.92, 28.88, 28.81, 28.72, 25.38, 22.83, 20.57, 20.50, 20.49. ESI-HRMS m/z calc’d for [C₁₀₉H₁₇₂N₈O₄₈+H]⁺ 2362.1342, observed 2362.1351. Calc’d for [C₁₀₉H₁₇₂N₈O₄₈+2H]²⁺ 1181.5710, observed 1181.5699.

TRIS-(GalNAc-PEG)-CPG (8).

To the solution of compound 7 (3.3 g, 1.6 mmol) in pyridine-DMF (8 mL, 1:1 v/v), DMAP (109 mg, 0.9 mmol), succinylated LCAA-CPG (1.0 g), and diisopropylcarbodiimide (2.16 mL, 13.8 mmol) were added and the mixture was left at ambient temperature for 48 h. Then, as done previously,⁶¹ unreacted carboxylates were capped by adding a solution of pentafluorophenol (760 mg, 4.13 mmol) in pyridine (2 mL) and allowing the mixture to react for 20 h. The solid was filtered and washed with pyridine (15 mL), and then the solution of piperidine (0.5 mL) in pyridine (5 mL) was added and left for 5 minutes. After filtration the modified CPG was washed with pyridine (20 mL), acetonitrile (20 mL), and dried in vacuo. CPG was suspended in solution of acetic anhydride (0.5 mL), 2,6-lutidine (0.9 mL) and 1-methylimidazole (0.5 mL) in THF (8.1 mL) and left for 2 h. Then CPG was washed with CH₂Cl₂ (20 mL), MeOH (20 mL), MeCN (20 mL), diethyl ether (20 mL), and dried in vacuo to yield modified CPG 8 (1.08 g, loading 42.9 μmol/g).

(9H-fluoren-9-yl)methyl ((4R,5R)-4-methyl-2-phenyl-1,3-dioxan-5-yl)carbamate (10):

A solution of Fmoc-L-threoninol (7.0 g, 21.4 mmol), camphor-10-sulfonic acid (496 mg, 2.14 mmol) and benzaldehyde dimethyl acetal (4.18 mL, 27.8 mmol) in acetonitrile (30 mL) was refluxed for 4 h. The reaction mixture was then allowed to cool at RT and the excess solvent was evaporated under reduced pressure. The residue was purified by flash column chromatography using ethyl acetate in pentane to afford the benzylidene-protected compound 10 in quantitative yield as a sticky solid; ¹H NMR (500 MHz, CDCl₃): δ 1.31 (d, J=6.26 Hz, 3 H) 3.74 (dd, J=9.92, 1.53 Hz, 1 H) 4.12 - 4.17 (m, 1 H) 4.21 (br dd, J=19.00, 1.75 Hz, 1 H) 4.19 - 4.30 (m, 3 H) 4.41 - 4.53 (m, 2 H) 5.63 (s, 2 H) 7.34 - 7.48 (m, 7 H) 7.56 (d, J=7.02 Hz, 2 H) 7.66 (d, J=7.48 Hz, 2 H) 7.81 (d, J=7.48 Hz, 2 H); ¹³C NMR (126 MHz, CDCl₃): δ 17.63, 47.29, 48.79, 66.99, 71.80, 75.31, 101.68, 120.04, 125.13, 125.16, 125.94, 127.11, 127.74, 128.41, 129.13, 137.86, 141.35, 143.90, 143.92, 156.53; ESI-HRMS: m/z calc’d for [C₂₆H₂₅NO₄+Na]⁺ 438.1676, observed 438.1677.

(4R,5R)-4-methyl-2-phenyl-1,3-dioxan-5-amine (11):

Compound 10 (5.0 g, 12.0 mmol) was dissolved in CH₂Cl₂ (25 mL) containing 1% DBU and 1% piperidine. The reaction mixture was stirred overnight at RT. Thereafter, the excess of solvent was evaporated under reduced pressure. The residue was purified by flash column chromatography using methanol in CH₂Cl₂ to afford compound 11⁶² as a white solid (2.0 g, 87%); ¹H NMR (400 MHz, CDCl₃) δ 1.30 (d, J=6.25 Hz, 3 H) 1.82 (br s, 2 H) 2.60 (m, 1 H) 4.02 - 4.23 (m, 3 H) 5.54 (s, 1H) 7.35 - 7.38 (m, 3 H) 7.49 - 7.51 (m, 2 H); ¹³C NMR (101 MHz, CDCl₃) δ 17.82, 49.01, 73.95, 75.86, 101.82, 126.03, 128.33, 128.97; ESI-HRMS: m/z calc’d for [C₁₁H₁₅NO₂+H] ⁺ 194.1176, observed 194.1174.

6-hydroxy-N-((4R,5R)-4-methyl-2-phenyl-1,3-dioxan-5-yl)hexanamide (12):

A solution of amine 11 (1.4 g, 7.25 mmol) and ε-caprolactone (4.0 mL, 36.25 mmol) in 1:1 mixture of DIPEA and THF (12.6 mL) was heated at 75 °C for 3 days. The reaction mixture was then allowed to cool at RT and the excess of solvent was evaporated under reduced pressure. The residue was purified by flash column chromatography using methanol in CH₂Cl₂ to afford the alcohol 12 as a colorless oil (2.0 g, 90%); ¹H NMR (500 MHz, CDCl₃) δ 1.23 (d, J=6.41 Hz, 3 H) 1.35 - 1.45 (m, 2 H) 1.51 - 1.61 (m, 2H) 1.68 (quin, J=7.63 Hz, 2 H) 2.16 (br s, 1 H) 2.27 (t, J=7.48 Hz, 2 H) 3.58 (t, J=6.48 Hz, 2 H) 4.00 (dd, J=9.46, 1.53 Hz, 1 H) 4.10 (s, 2 H) 4.16 (dd, J=6.33, 1.60 Hz, 1 H) 5.58 (s, 1 H) 6.34 (br d, J=9.31 Hz, 1 H) 7.33 - 7.42 (m, 3 H) 7.49 (d, J=6.71 Hz, 2 H); ¹³C NMR (126 MHz, CDCl₃) δ 17.63, 25.38, 32.26, 36.59, 46.57, 62.38, 71.75, 75.23, 101.50, 125.83, 128.35, 129.07, 137.82, 173.14; ESI-HRMS m/z calc’d for [C₁₇H₂₅NO₄+H] ⁺ 308.1856, observed 308.1874.

6-(3,4,6-tri-O-acetyl-2-acetamido-2-deoxy-β-D-galactopyranosyloxy)-N-((4R,5R)-4-methyl-2-phenyl-1,3-dioxan-5-yl)hexanamide (13):

The previously described GalNAc oxazoline⁴⁷ (4.22 g, 12.8 mmol) and the alcohol 12 (4.72 g, 15.3 mmol) were dissolved in dichloroethane (42 mL) and stirred for 30 min. Thereafter, TMSOTf (1.16 mL, 6.4 mmol) was added to the reaction and the mixture was stirred at RT for 24 h. The reaction mixture was poured into an ice cold solution of sodium bicarbonate and extracted with dichloromethane, washed with water and dried over sodium sulfate and filtered. Excess solvents were removed under reduced pressure and the residue was purified by flash column chromatography using methanol in CH₂Cl₂ to afford the glycoside 13 as a sticky white solid (6.2 g, 76%); ¹H NMR (500 MHz, CDCl₃) δ 1.24 (d, J=6.41 Hz, 3 H) 1.42 (td, J=14.19, 7.02 Hz, 2 H) 1.55 - 1.76 (m, 4 H) 1.86 (br s, 1 H) 1.90 - 2.08 (m, 8 H) 2.14 (s, 3 H) 2.23 - 2.33 (m, 2 H) 3.48 (ddd, J=9.46, 8.01, 5.57 Hz, 1 H) 3.85 - 3.97 (m, 3 H) 4.01 (dd, J=9.54, 1.60 Hz, 1 H) 4.06 −4.22 (m, 5 H) 4.74 (d, J=8.39 Hz, 1 H) 5.32 - 5.40 (m, 2 H) 5.60 (s, 1 H) 6.06 (d, J=8.54 Hz, 1 H) 6.27 (br d, J=9.46 Hz, 1 H) 7.39 (br d, J=7.32 Hz, 2 H) 7.34 - 7.42 (m, 1 H) 7.50 (d, J=6.81 Hz, 2 H); ¹³C NMR (126 MHz, CDCl₃): δ 17.66, 20.71, 20.74, 23.38, 25.16, 25.36, 25.73, 28.63, 36.44, 46.59, 51.84, 61.51, 66.86, 69.67, 69.84, 70.55, 71.74, 75.22, 100.87, 101.56, 125.82, 128.36, 129.09, 137.77, 170.33, 170.41, 170.46, 170.50, 173.20; ESI-HRMS: m/z calc’d for [C₃₁H₄₄N₂O₁₂+H] ⁺ 637.2967, observed 637.3098.

N-[6-(3,4,6-tri-O-acetyl-2-acetamido-2-deoxy-β-D-galactopyranosyloxy)-hexanoyl]-L-threoninol (14):

To a suspension of compound 13 (1.0 g, 1.57 mmol) and 10% Pd(OH)₂/C (100 mg) in ethanol (10 mL) was added Et₃SiH (2.26 mL, 14.13 mmol) dropwise in three portions (one aliquot every 2 h), and then the reaction mixture was stirred at RT for another 2h. The reaction mixture was filtered through Celite, and washed with ethanol (3 x 10 mL). The combined filtrate was concentrated under reduced pressure and the crude product thus obtained was purified by flash column chromatography using methanol in CH₂Cl₂ to afford the diol 14 as a white solid (0.72 g, 83%); ¹H NMR (500 MHz, DMSO-d₆) δ 0.98 (d, J=6.41 Hz, 3 H) 1.17 - 1.33 (m, 2 H) 1.41 - 1.54 (m, 4 H) 1.78 (s, 3 H) 1.90 (s, 3 H) 2.00 (m, 3 H) 1.98 - 2.12 (m, 5 H) 3.28 - 3.34 (m, 1 H) 3.37 - 3.46 (m, 2 H) 3.58 - 3.73 (m, 2 H) 3.81 −3.91 (m, 2 H) 4.01-4.07 (m, 3 H) 4.47 - 4.56 (m, 3 H) 4.97 (dd, J=11.29, 3.36 Hz, 1 H) 5.21 (d, J=3.36 Hz, 1 H) 7.26 (d, J=8.85 Hz, 1H) 7.81 (d, J=9.31 Hz, 1 H); ¹³C NMR (126 MHz, DMSO-d₆): δ 20.59, 20.92, 20.94, 20.99, 23.27, 25.47, 25.64, 29.22, 35.83, 49.81, 55.90, 61.05, 61.92, 64.74, 67.18, 69.20, 70.29, 70.96, 101.40, 169.72, 170.10, 170.38, 170.48, 172.77; ESI-HRMS m/z calc’d for [C₂₄H₄₀N₂O₁₂+H]⁺ 549.2654, observed 549.2684.

6-(3,4,6-tri-O-acetyl-2-acetamido-2-deoxy-β-D-galactopyranosyloxy)-N-((2R,3R)-1-O-(4,4'-dimethoxytrityl)-1,3-dihydroxybutan-2-yl)hexanamide (15):

To a solution of compound 14 (5.0 g, 9.1 mmol) in pyridine (50 mL) was added dimethoxytrityl chloride (3.40 g, 10.0 mmol) under argon atmosphere. The reaction mixture was then stirred at RT overnight, quenched with saturated aq. sodium bicarbonate solution (100 mL) and extracted with EtOAc (3 x 100 mL). The organic phase was washed with brine (2 x 200 mL), dried over anhydrous sodium sulphate and concentrated under reduced pressure. The residue thus obtained was purified by flash column chromatography using methanol in dichloromethane containing 1% Et₃N to afford the DMT-protected glycoside 15 as a white solid (4.8 g, 91%); ¹H NMR (500 MHz, CDCl₃): δ 1.13 (d, J=6.41 Hz, 3 H), 1.30 - 1.52 (m, 2 H), 1.60 (br d, J=6.10 Hz, 1 H), 1.68 (dt, J=13.73, 6.71 Hz, 3 H), 1.88 (s, 3 H), 2.00 (s, 3 H) 2.04 (s, 3 H) 2.13 (s, 3 H), 2.20 - 2.30 (m, 2 H), 3.28 (br dd, J=9.61, 3.51 Hz, 1 H) 3.40 (dd, J=9.54, 4.35 Hz, 1 H) 3.45 - 3.54 (m, 1 H), 3.80 (s, 6 H), 3.87 - 3.99 (m, 4H), 4.04 - 4.21 (m, 3 H), 4.73 (d, J=8.39 Hz, 1 H), 5.31 - 5.40 (m, 2 H), 6.12 (dd, J=8.62, 2.06 Hz, 2 H), 6.84 (dd, J=8.85, 1.68 Hz, 4 H) 7.22 - 7.32 (m, 8 H) 7.38 (d, J=7.48 Hz, 2 H); ¹³C NMR (126 MHz, CDCl₃): δ 19.99, 20.71, 23.36, 25.21, 25.83, 28.61, 36.48, 51.80, 53.38, 55.26, 61.50, 65.26, 66.85, 68.70, 69.69, 69.87, 70.58, 86.79, 100.90, 113.33, 127.09, 127.92, 128.04, 129.91, 129.92, 135.38, 135.48, 144.34, 158.63, 158.66, 170.45, 173.55; ESI-HRMS: m/z calc’d for [C₄₅H₅₈N₂O₁₄+Na⁺] 873.3780, observed 873.3793.

6-(3,4,6-tri-O-acetyl-2-acetamido-2-deoxy-β-D-galactopyranosyloxy)-N-((2R,3R)-1-O-(4,4'-dimethoxytrityl)-3-O-(N,N-diisopropylamino-(2-cyanoethyloxy)phosphonamidic)-1,3-dihydroxybutan-2-yl)hexanamide (16):

To a stirred solution of DMT-protected glycoside 15 (2.6 g 3.06 mmol) in dry CH₂Cl₂ (25 mL), N,N-diisopropylethylamine (2.66 mL, 15.3 mmol) and 2-cyanoethyl N,N-diisopropylamino chlorophosphoramidite (0.88 mL, 3.97 mmol) were added under argon atmosphere. The reaction mixture was stirred at RT for 6 h. After completion of reaction on analytical TLC examination, the reaction mixture was diluted with EtOAc (100 mL) and washed with saturated KCl solution (2 x 50 mL). The organic phase was dried over anhydrous sodium sulphate and concentrated under reduced pressure. Purification was done by flash column chromatography using methanol in dichloromethane containing 1% Et₃N to afford the phosphoramidite 16 as a white solid (2.8 g, 88% yield); ¹H NMR (500 MHz, CDCl₃): δ 0.89 (t, J=7.10 Hz, 1 H) 1.00 (d, J=6.71 Hz, 2 H) 1.11 - 1.34 (m, 18 H) 1.36 - 1.51 (m, 2 H) 1.54 −1.74 (m, 4 H) 1.89 (d, J=3.81 Hz, 3 H) 1.99 - 2.07 (m, 6 H) 2.09 - 2.15 (m, 3 H) 2.16 - 2.29 (m, 2 H) 2.31 - 2.49 (m, 1 H) 2.55 - 2.69 (m, 1 H) 2.77 (td, J=6.18, 2.75 Hz, 1 H) 3.11 (br dd, J=9.16, 6.10 Hz, 1 H) 3.15 - 3.26 (m, 1 H) 3.40 - 3.58 (m, 5 H) 3.70 (br s, 1 H) 3.76 - 3.83 (m, 1 H) 3.80 (d, J=2.75 Hz, 6 H) 3.86 - 4.03 (m, 4 H) 4.07 - 4.25 (m, 4 H) 4.34 (br s, 1 H) 4.72 (d, J=8.39 Hz, 1 H) 5.34 - 5.40 (m, 2 H) 5.64 - 5.93 (m, 1 H) 5.90 (d, J=9.00 Hz, 1 H) 6.22 (dd, J=8.39, 5.34 Hz, 1 H) 6.79 - 6.86 (m, 4 H) 7.20 - 7.34 (m, 8 H) 7.38 - 7.45 (m, 2 H); ³¹P NMR (202.5 MHz, CDCl₃): δ 148.11, 147.81; ESI-MS: m/z calc’d for [C₅₄H₇₅N₄O₁₅P+Na]⁺ 1073.4, observed 1071.9.

GalNAc-hexyl-threoninol-CPG (17).

A solution of GalNac monomer 15 (170 mg, 0.2 mmol), succinic anhydride (50 mg, 0.50 mmol), DMAP (10 mg), Et₃N (0.14 mL, 0.5 mmol) in CH₂Cl₂ (3 mL) was stirred at RT for 1 h. The reaction was diluted with CH₂Cl₂ (20 mL) and washed with brine (50 mL). The organic layer was dried over anhydrous sodium sulphate and concentrated under reduced pressure. The crude residue was coevaporated with dry pyridine (3 x 2 mL). The triethylammonium salt thus obtained was stirred with HATU (152 mg, 0.4 mmol) and DIPEA (0.70 mL, 4.0 mmol) in DMF (7.0 mL) at RT for 10 minutes. This reaction mixture was then transferred to a centrifuge tube containing 1.0 g of LCAA-CPG 1000Å. The reaction mixture was gently shaken at RT overnight. The solid support was filtered and successively washed with CH₂Cl₂, MeOH, CH₂Cl₂ and diethyl ether (25 mL each).

For capping, the derivatized CPG was suspended with acetic anhydride (3 mL), pyridine (3 mL) and DMAP (50 mg) and was gently shaken at RT for 2 h. The support was filtered, successively washed with CH₂Cl₂, MeOH, CH₂Cl₂, diethyl ether (25 mL each) and then air dried. The loading of CPG was 25-28 μmol/g as determined by absorbance of the DMT cation at 495 nm.

Oligonucleotide synthesis and purification.

Oligonucleotides were synthesized using an Applied Biosystems 394 DNA/RNA synthesizer at 1 μmol scale or an AKTA OligoPilot 10 at 5 μmol scale. Oligonucleotides were synthesized on 1000Å LCAA CPG functionalized with Unilinker (ChemGenes, 40 μmol/g) or GalNAc monomer/trimer-functionalized CPG. Activation was achieved with 5-benzylthio-1H-tetrazole (BTT, 0.25 M in acetonitrile). Oxidation was achieved using 0.02 M iodine in a mixture of THF, water, and pyridine. Sulfurization was accomplished with DDTT (0.1 M, ChemGenes).

Oligonucleotides were deprotected with dimethyl sulfoxide (100 μl) and 40% methylamine (250 μl) at 65 °C for 15 minutes. Then the solvents were evaporated to dryness in a centrifugal evaporator and oligonucleotides were resuspended in 1 mL RNase-free water. Oligonucleotides were purified by reverse phase chromatography using an Agilent 1200 HPLC using C-18 columns and triethylammonium acetate (TEAA, pH 6.0) buffers with acetonitrile gradient. Pure fractions were desalted using NAP-25 columns (GE Healthcare). Oligonucleotide masses were verified by LC-MS analysis on an Agilent 6530 Accurate-Mass Q-TOF using linear gradients of methanol in water, both containing 9 mM Et₃N and 100 mM 1,1,1,3,3,3-hexafluoroisopropanol.