Introduction of sugar-modified nucleotides into CpG-containing antisense oligonucleotides inhibits TLR9 activation

Antisense oligonucleotides (ASOs) are synthetic single-stranded oligonucleotides that bind to RNAs through Watson–Crick base pairings. They are actively being developed as therapeutics for various human diseases. ASOs containing unmethylated deoxycytidylyl-deoxyguanosine dinucleotide (CpG) motifs are known to trigger innate immune responses via interaction with toll-like receptor 9 (TLR9). However, the TLR9-stimulatory properties of ASOs, specifically those with lengths equal to or less than 20 nucleotides, phosphorothioate linkages, and the presence and arrangement of sugar-modified nucleotides—crucial elements for ASO therapeutics under development—have not been thoroughly investigated. In this study, we first established SY-ODN18, an 18-nucleotide phosphorothioate oligodeoxynucleotide with sufficient TLR9-stimulatory activity. We demonstrated that an unmethylated CpG motif near its 5′-end was indispensable for TLR9 activation. Moreover, by utilizing various sugar-modified nucleotides, we systematically generated model ASOs, including gapmer, mixmer, and fully modified designs, in accordance with the structures of ASO therapeutics. Our results illustrated that introducing sugar-modified nucleotides in such designs significantly reduces TLR9-stimulatory activity, even without methylation of CpG motifs. These findings would be useful for drug designs on several types of ASOs.

degradation.Representative examples of steric-blocking ASOs are splice-switching oligonucleotides (SSOs), which modulate pre-mRNA splicing and repair defective transcripts to restore the production of functional proteins.Another example is anti-miRNA ASOs that hybridize to the target miRNA and inhibit its function.In stericblocking ASOs, 2′-sugar-modified nucleotides are often used throughout the length of the oligonucleotides.For instance, nusinersen (Spinraza), an approved 18-mer SSO used for the treatment of spinal muscular atrophy, consists of 2′-MOE nucleotides throughout its length.On the contrary, BNAs are generally placed at intervals in steric-blocking ASOs; this is called a mixmer design.CDR132L is a 16-mer anti-miRNA ASO, which contains 7 LNAs with a mixmer design (5′-aTgGcTgTaGactgTT-3′: upper and lower cases indicate LNA and DNA, respectively) 12,13 .In addition to the sugar modifications, most ASOs that are currently in clinical use or under clinical trials are phosphorothioated at internucleotide linkages to enhance membrane permeability and stability, except for SSOs with a phosphorodiamidate morpholino backbone 10 .The application of sugar-modified nucleotides with high binding affinity to complementary RNA has enabled the development of shorter ASOs.Indeed, the sugar-modified ASOs currently in clinical development are generally equal to or less than 20-mer in length 14,15 .
Understanding the biological characteristics of oligonucleotides, such as their potential to activate innate immunity via toll-like receptors (TLRs), is crucial for ensuring the safety of ASOs.Single-stranded, synthetic oligodeoxynucleotides (ODNs) with unmethylated deoxycytidylyl-deoxyguanosine dinucleotide (CpG) motifs can activate an immune response through interaction with TLR9 16,17 .The potency of immune stimulation via TLR9 depends on several factors, including the ODN sequence and base modifications.For instance, the absence or cytosine methylation of CpG motifs reduces the immunostimulatory properties of CpG ODNs 18 .However, these results were mostly obtained using relatively long CpG ODNs, such as a 24-mer CpG ODN called ODN2006 and its variants, which contrasts with the shorter ASOs currently in development [19][20][21] .The backbone of ASOs is a single-stranded ODN, and it is assumed that ASOs containing CpG motifs have the potential to induce immune responses through interaction with TLR9.However, few reports describe the TLR9-stimulatory properties of ASO therapeutics, with consideration of their characteristic features-equal to or less than 20-mer in length, phosphorothioate linkages, existence of sugar-modified nucleotides, and their arrangement in oligonucleotides: gapmer or mixmer design 22 .
Therefore, in the present study, we aimed to examine the effect of sugar-modifications of ASO therapeutics on TLR9 activation.To achieve this, we first established an 18-mer CpG-containing phosphorothioate oligonucleotide that can efficiently activate TLR9.We then systematically introduced sugar-modified nucleotides into the 18-mer CpG oligonucleotide and comprehensively examined their effects on TLR9-stimulatory activity.
Next, to examine the effects of cytosine methylation in the CpG motifs, SY-ODN18 variants with methylated CpG motifs were designed and evaluated for TLR9 activity (Fig. 2b).Cytosine methylation in both CpG motifs or only in the 5′-CpG motif markedly decreased the TLR9-stimulatory activity of SY-ODN18 (Fig. 2b, 2B-4 and 2B-5), whereas methylation only in the 3′-CpG motif did not significantly affect the activity (Fig. 2b,  2B-6).These results indicate that the CpG motif near the 5′-end of SY-ODN18 contributes more substantially to its TLR9-stimulatory activity than the other CpG motif.Notably, substitution of the 5′-CpG motif with GpC reduced TLR9 activation capacity close to the basal levels (~ 10%) (Fig. 2a, 2A-4), whereas cytosine methylation reduced it to ~ 20% (Fig. 2b, 2B-5).These observations indicate that cytosine methylation does not completely inhibit TLR9 activation by the CpG motif.
Initially, we introduced these sugar-modified nucleotides in a gapmer design.In recently developed gapmer ASOs, sugar-modified cytidines are usually methylated by default.Therefore, we replaced two cytidines in the wing regions of SY-ODN18 with methylated sugar-modified cytidines (2′-MOE, LNA, ENA, or BNA NC (N-Me)) (Fig. 4, 4-5 to 4-8).Thus, the 5′-CpG motifs in these variants were both methylated and sugar-modified.As a result, we observed that the TLR9-stimulatory activities of these variants were suppressed to less than 10% (Fig. 4, 4-5 to 4-8).This reduction was comparable to that obtained by substituting the 5′-CpG motif with GpC (Fig. 4, 4-3), but clearly more significant than that obtained by methylating cytosine of the 5′-CpG motif (Fig. 4, 4-4).These results indicate that introducing sugar-modified nucleotides with cytosine methylation into the wing regions of SY-ODN18 leads to a robust mitigation in TLR9-stimulatory activity.To examine whether the sugar modification alone could suppress the TLR9 activation, we prepared LNA-cytosine phosphoramidite (without methylation, see Fig. 3b) 7,25,26 , and then synthesized an LNA gapmer variant of SY-ODN18 with the unmethylated 5′-CpG motif (Fig. 4, 4-9).The TLR9-stimulatory activity of this variant was reduced to similar levels as that of the LNA gapmer with a methylated 5′-CpG motif (Fig. 4, 4-6 and 4-9), suggesting that cytosine methylation of the 5′-CpG motif is not required for the reduction in TLR9-stimulatory activity when LNAs are introduced in the wing regions.
We conducted further analysis to investigate whether the introduction of sugar-modified nucleotides could attenuate TLR9-stimulatory activity without cytosine methylation.The incorporation of 2′-OMe or 2′-MCE modification in the gapmer design led to a reduction in TLR9-stimulatory activity to approximately 10% (Fig. 4, 4-10 and 4-11), supporting the idea that sugar modification alone can suppress TLR9 activation.As the remaining activity of the 2′-OMe variant was relatively higher compared to that of other gapmer variants, we hypothesized whether cytosine methylation within the 5′-CpG motif could further diminish this activity.To investigate this hypothesis, we utilized 2′-O-methyl-5-methylcytidine phosphoramidite to synthesize a SY-ODN18 variant with cytosine-methylated 2′-OMe in the 5′-CpG motif (Figs.3b, 4, 4-12).As depicted in Fig. 4 (4-10 vs. 4-12), methylation of 2′-O-methylcytidine in the 5′-CpG motif led to a slight yet significant decrease in the TLR9-stimulatory activity of the 2′-OMe variant.These findings suggest that 2′-OMe modification and cytosine methylation can jointly attenuate TLR9-stimulatory activity.
Next, we introduced sugar-modified nucleotides into designs utilized for steric-blocking ASOs.Stericblocking ASOs employing 2′-modification analogs, such as nusinersen, had all nucleotides substituted with such analogs.Therefore, oligonucleotides were prepared in which all SY-ODN18 nucleotides were replaced with 2′-OMe or 2′-MCE (fully modified variants).In contrast, BNA analogs are often introduced in mixmer designs due to their higher affinity to RNAs.Thus, we employed a design in which every other nucleotide was substituted with LNA, ENA, or BNA NC (N-Me), which had been shown to induce splice-switching effectively in our previous study 27 .For comparison, we introduced 2′-OMe, 2′-MCE, or 2′-MOE in the same mixmer design.Note that cytosines were not methylated in these variants.In addition, sugar modifications were not incorporated in the cytidines of the two CpG motifs in the mixmer variants.
The fully modified variants displayed minimal TLR9-stimulatory activity, essentially the same as the background levels without oligonucleotides (~ 5% of that observed for SY-ODN18; Fig. 5, 5-3 and 5-4).Despite only having half the number of modified nucleotides, the mixmer variants also showed a robust reduction in TLR9stimulatory activity, similar to the fully modified variants, regardless of the type of sugar modification (Fig. 5, 5-5 to 5-10).These results clearly indicated that the introduction of 2′-modifications or BNA analogs in a mixmer design could abolish TLR9-stimulatory activity, even in the absence of methylations and sugar modifications in the cytidines of the CpG motifs.

Discussion
In this study, we selected an 18-mer phosphorothioate ODN exhibiting sufficient TLR9-stimulatory activity from the deletion variants of ODN2006.Utilizing this variant (SY-ODN18), we analyzed the effects of substituting CpG motifs with GpC and cytosine methylation on TLR9-stimulatory activity.Additionally, employing various sugar-modified nucleotides, we systematically generated model ASOs mirroring the chemical structures of ASO therapeutics under development, including gapmers, mixmers, and fully modified designs.Our findings demonstrate that the 5′-CpG motif is indispensable for TLR9 activation and that the introduction of sugar-modified nucleotides alone significantly diminishes TLR9-stimulatory activity.SY-ODN18 exhibited the most potent activity among 18-mer variants of ODN2006 (Fig. 1, 1-8 to 11).It contains two CpG motifs, and the base length of the spacer between the two motifs is six.The 5′-CpG motif is located at the 2 nd and 3 rd positions from the 5′ end (Fig. 1, 1-10, highlighted in orange).Another 18-mer variant with two CpG motifs, in which the 5′-CpG motif was at the 7 th and 8 th positions from the 5′ end (Fig. 1, 1-9, highlighted in light green), showed approximately 40% weaker TLR9-stimulatory activity compared to SY-ODN18.These results suggest that a CpG motif close to the 5′ end is required for efficient activation of TLR9.A previous report has shown that two CpG motifs are necessary for TLR9 activation, with the CpG motif near the 5′ end being particularly important, and that high TLR9 activity is observed when the base length between the two CpG motifs is between 6 and 10 20 .The sequence of SY-ODN18 meets these conditions, reinforcing that these features are essential for TLR9-stimulatory activity.
According to its position, the 5′-CpG motif of SY-ODN18 could function as a 5′-xCx motif.By solving the crystal structure of TLR9 in complex with a 6-mer ODN containing a 5′-xCx motif and a 10-mer ODN with a central CpG motif, Ohto et al. demonstrated that the 5′-xCx and CpG motifs bind to distinct sites on TLR9 and cooperatively drive dimerization and subsequent activation of TLR9 23,28 .They also reported that methylating the cytosine in the 5′-xCx motif or moving the "C" to the third position (i.e., 5′-xxC) reduced the affinity for TLR9.In the present study, we showed that the cytosine methylation or GpC replacement of the 5′-CpG motif greatly reduced TLR9-stimulatory activity (Fig. 2a and b).Together with the previous findings, these data suggest that the 5′-CpG motif of SY-ODN18 binds to the 5′-xCx motif-binding site, thereby promoting dimerization of TLR9.
Oligonucleotides need to bind to both the 5′-xCx motif and CpG motif-binding sites to fully activate TLR9 23,29 .A single oligonucleotide molecule containing both 5′-xCx and CpG motifs may bind to these sites simultaneously.However, based on the distance between the 5′-xCx and CpG motif-binding sites in the crystal structure of TLR9, it is proposed that 5′-xCx and CpG motifs in a single oligonucleotide can bind to these binding sites at the

Figure 3 .
Figure 3. Sugar-modified nucleotides used in this study.(a) Chemical structures of DNA and the sugarmodified nucleotides; sugar modifications are shown in blue.(b) Chemical structures of cytosines and 5-methylcytosines used in this study.The sugar modifications and 5-methyl groups are shown in blue and red, respectively.