Investigation of Sequence-Penetration Relationships of Antisense Oligonucleotides

Abstract

A major limitation for the development of more effective oligonucleotide therapeutics has been a lack of understanding of their penetration into the cytosol. While prior work has shown how backbone modifications affect cytosolic penetration, it is unclear how cytosolic penetration is affected by other features including base composition, base sequence, length, and degree of secondary structure. We have applied the chloroalkane penetration assay, which exclusively reports on material that reaches the cytosol, to investigate the effects of these characteristics on the cytosolic uptake of druglike oligonucleotides. We found that base composition and base sequence had moderate effects, while length did not correlate directly with degree of cytosolic penetration. Investigating further, we found that degree of secondary structure had the largest and most predictable correlations with cytosolic penetration. These methods and observations add an additional layer of design for maximizing the efficacy of new oligonucleotide therapeutics.

Graphical Abstract

Developing better oligonucleotide therapeutics requires better understanding of the features that promote efficient cytosolic delivery. This study uses the chloroalkane penetration assay to explore features of base sequence, length, and secondary structure that affect cytosolic delivery of druglike oligonucleotides.

Introduction

Approximately 80% of the disease-related proteome pose challenges for drug development using small molecule therapeutics. Most of these proteins lack an obvious binding pocket for a small organic compound, or are otherwise deemed ‘undruggable’.^[1] RNA therapeutics are a class of drugs that can bypass these challenges by targeting the mRNA of a disease-related protein and modulating its expression.^[2,3] Antisense oligonucleotides (ASOs), a prominent class of RNA therapeutics, are short oligonucleotide sequences that hybridize to the target mRNA through Watson-Crick-Franklin base pairing.^[4,5] They reduce the amount of expressed protein through RNase H1-mediated degradation^[6,7] or through alteration of mRNA splicing.^[8,9]

There are 10 ASO drugs currently approved for use in humans,^[10,11] and many more are in clinical trials. ASOs are chemically modified on their backbone, sugar, and nucleobases in order to prevent degradation by nucleases and improve tissue and cell penetration.^[12] Common chemical modifications include substitution of phosphodiesters for phosphorothioates (PS) in the backbone and 2′-O-methoxyethyl (MOE) or 2′-O-methyl groups on the ribose. These modifications improve pharmacokinetic and pharmacodynamic properties by increasing target affinity, nuclease resistance, and cell penetration.^[13–19]

It is often assumed that the druglike properties of ASOs, including extent of cell penetration, heavily depend on backbone chemical modifications and not the base sequence.^[20] However, there is strong evidence that base sequence and length play a critical role in the cell penetration and activity of ASOs. For example, Brand et al. showed that base sequence affects the iontophoretic delivery of ASOs across mouse skin by comparing ASOs with analogs that have the same length and base content but reverse sequence.^[21] Peyman et al. compared the cell penetration of homonucleotide 16mer ASOs and observed that the 16mer poly-G ASO penetrated cells most efficiently. Moreover, adding 3-non-base pairing guanosines on either end of an ASO with moderate activity enhanced its activity by almost 10-fold.^[22] Wang et al. also demonstrated effects of base sequence on penetration, concluding that guanosine-rich ASOs were most cell-penetrant.^[23] In similar studies using coverslip compression to drive delivery, Chen et al. found that poly-T ASOs were the most efficiently delivered ASOs to the nucleus.^[24] Similar to these studies on ASO base composition, there are some publications that have studied effects of ASO length on delivery, but without reaching an obvious consensus. Loke et al. saw an inverse linear relationship between length of poly-T ASOs and their intracellular delivery.^[25] Such a relationship was also observed by Khatsenko et al. in the rat intestinal adsorption of ASOs with PS and 2′-MOE modifications.^[26] Brand et al. showed that ASOs with up to 20 bases had less efficient transport as the length increased, but also that some ASOs did not follow this trend. Also, the longer ASOs they tested (20–40 nucleotides long) did not have a clear length-transport relationship and the differences were attributed to base sequence.^[21] Chen et al. saw an opposite pattern when they measured the intranuclear delivery of different lengths of poly-T ASOs upon coverslip compression. In that study, shorter oligonucleotides (12– 30 nucleotides) penetrated cells more efficiently as the length increased, whereas the delivery efficiency of longer oligonucleotides (60–120 nucleotides) with length.^[24] Finally, in two independent papers, Seth et al. and Staarup et al. showed that decreasing the length of an 18mer ASO by 6–8 bases in combination with introduction of locked nucleic acid modifications resulted in compounds that were more potent and less toxic than the parent 18mer oligonucleotides.^[27,28]

These works to date on the effects of sequence and length on ASO delivery are sometimes contradictory and difficult to compare for several reasons. The data span over two decades of work in the field, and most studies used ASOs that did not have the most common modifications used in current ASO drugs and clinical candidates (PS-modified, with 2ʹ-O-methyl or 2ʹ-MOE modifications). Most studies also involved physical delivery methods such as iontophoresis or coverslip compression.^[21,24] However, the most common delivery method for current ASO drugs is gymnosis, which is delivery unassisted by physical or chemical methods.^[29] In addition, most of these experiments examined activity, which depends on many factors other than cell penetration, and/or total cellular uptake of the ASO. Measuring total cellular uptake includes material that reached the cytosol or the nucleus (productive uptake), but also material trapped in endosolysosomal compartments (non-productive uptake). In order to design more therapeutically efficient ASOs, it will be important to understand the effects of sequence and length on productive uptake of ASOs with chemical modifications commonly used in current ASO drugs.^[30]

In this work we use the Chloroalkane Penetration Assay (CAPA) to measure directly the cytosolic penetration of druglike ASOs.^[31–33] Because CAPA measures cytosolic penetration independent from ASO function, this approach allows a more direct examination of the effects of base sequence and ASO length on productive uptake. CAPA (Figure 1) uses a HeLa cell line that stably expresses HaloTag, which is a modified haloalkane dehalogenase enzyme that covalently labels itself with chloroalkane-tagged compounds (ct-compounds).^[34] First, cells are pulsed with serial dilutions of ct-oligonucleotides; during this step, if the ct-oligonucleotide is taken up by the cell and manages to escape the endosome it will react irreversibly with HaloTag, which is only present in the cytosol. After this pulse step, cells are washed and chased with ct-TAMRA, a red fluorescent dye that readily enters cells and rapidly reacts with any unreacted HaloTag. After free dye is washed away, the red fluorescence associated with the cells is quantified using a benchtop flow cytometer and normalized based on control cells that have been treated only with ct-TAMRA (100% fluorescence) and cells that have been treated with no ct-oligonucleotide and no ct-TAMRA (0% fluorescence). The normalized data are plotted against concentration of ct-oligonucleotide, producing a sigmoidal curve from which one can derive a CP₅₀ value, which is the concentration of the ct-oligonucleotide at which 50% of HaloTag was blocked.^[32] The smaller the CP₅₀, the more cell-penetrant the ct-compound. CAPA has been used extensively for peptide and protein drugs,^[35–39] and we recently reported its expansion into measuring cell penetration efficiency of RNA therapeutics including ASOs.^[33] Importantly, CAPA is low-cost, high-throughput, and specific to productive uptake since ct-oligonucleotides must escape endosomes before they can react with HaloTag in the cytosol.

Figure 1. Schematic of the Chloroalkane Penetration Assay (CAPA).

HeLa cells that stably express HaloTag in the cytosol are (1) pulsed with serial dilutions of the ct-oligonucleotide and incubated for 24 h at 37°C in 5% CO₂. (2) Next, they are washed and chased with ct-dye for 15 min at room temperature. (3) They are then washed again, after which dye fluorescence is measured by benchtop flow cytometry. (4) The data are normalized based on controls (no ct-oligonucleotide and no ct-dye for 0% fluorescence and no ct-oligonucleotide for 100% fluorescence), plotted against ct-oligonucleotide concentration, and fitted to a sigmoidal curve to yield the CP₅₀, the concentration of the ct-oligonucleotide at which 50% of HaloTag has reacted with the molecule.

Results

Effects of base composition and sequence on cytosolic penetration

For this study, we chose nusinersen as a starting point for comparing ASOs with different sequences, compositions, and lengths.^[40] Nusinersen is an FDA-approved drug that treats spinal muscular atrophy by correcting the splicing of SMN2 mutations that exclude exon 7.^[41] It is an 18mer ASO uniformly modified with a PS backbone and 2ʹ-MOE sugars.

Initially, we focused on the effects of base composition on cytosolic penetration. We prepared nusinersen and its reverse complement sequence (Figure 2A) with PS backbones and 2ʹ-MOE sugars. The chloroalkane tag was added via a strain-promoted cycloaddition reaction using a 5ʹ azide group and a dibenzylcyclooctyne-chloroalkane (SI Figures 1–3). We performed CAPA using stably transfected HeLa cells with 24 hours of incubation with serial dilutions of each chloroalkane-tagged oligonucleotide (ct-oligonucleotide). Chloroalkane-tagged nusinersen (ct-nusinersen) had a CP₅₀ of 2.87 ± 0.06 μM and its reverse complement had a CP₅₀ of 10.62 ± 0.05 μM, indicating almost 4-fold poorer cytosolic penetration than ct-nusinersen (Figure 2B, 2C). Nusinersen and its reverse complement have the same length and chemical modifications, so based on the significant difference in their CP₅₀ values (SI Table 2) we can conclude that base composition is important for the cytosolic penetration of oligonucleotides.

Figure 2. Effects of base composition and base sequence on cytosolic penetration of nusinersen, its reverse complement, and scrambled analogs.

(A) Base sequences and CP₅₀ values for ct-nusinersen, its reverse complement, and five scrambled analogs. (B) CAPA dose-dependence curves showing the degree of cytosolic penetration at concentrations between 15 and 0.02 μM for the ct-reverse complement and between 5 and 0.007 μM for ct-nusinersen following 24 h incubation in HeLa cells. (C) CP₅₀ values for ct-nusinersen and ct-reverse complement, derived from the curve fits shown in Figure 2B. **** denotes p < 0.0001. (D) CAPA dose-dependence curves showing the degree of cytosolic penetration for ct-nusinersen and scrambled analogs at concentrations between 5 and 0.007 μM following 24 h incubation in HeLa cells. (E) CP₅₀ values for ct-nusinersen and scrambled analogs, derived from the curve fits shown in Figure 2D. (F) Percent occupied HaloTag after 24 h of incubation with 1.67 μM ct-oligonucleotide. All data show results from four independent trials or three for ct-reverse complement, and error bars show standard error of the mean. CP₅₀ values are reported as average and standard error of the mean from four independent curve fits to four independent trials (or three trials for ct-reverse complement). Nearly all pairwise comparisons in panels E and F show statistically significant differences, ranging from p = 0.05 to p = 0.0001, calculated using an ordinary one-way ANOVA test (SI Tables 4, 6).

To examine more subtle effects of base sequence, we next generated sequences for scrambled analogs of nusinersen using a random scrambling algorithm (see SI). We selected five scrambled analogs for which the predicted secondary structure did not differ dramatically from that of nusinersen, as predicted using the RNA Fold Web Server.^[42] These five oligonucleotides (SCR1-SCR5, Figure 2A) were also prepared as uniformly modified PS, 2ʹ-MOE oligonucleotides with a 5ʹ chloroalkane tag. We carried out CAPA with these ct-oligonucleotides (Figure 2D) and their CP₅₀ values ranged between 2.04 ± 0.05 μM and 4.63 ± 0.35 μM (Figure 2A, 2E). Similar differences were also seen comparing the percent occupied HaloTag for each of these ct-oligonucleotides at different concentrations (Figure 2F, SI Figure 4). For example, at 1.67 μM, SCR1 occupied 44.6 ± 0.6% of cellular HaloTag, while SCR3 occupied only 23.6 ± 1.9% of HaloTag (SI Table 3). These data show statistically significant differences for most of the oligonucleotides for percent occupancy at 1.67 μM and 5 μM concentrations (SI Tables 4, 5) and also for overall dose-dependence of cytosolic penetration (SI Table 6). Considering that nusinersen and the five scrambled analogs have the same length, molecular weight, nucleotide composition, and chemical modifications, and that they have similar predicted extent of secondary structure, we can conclude that specific nucleotide sequences have an effect on the cytosolic penetration of nusinersen-like ASOs.

Effects of length on cytosolic penetration

To test the effects of length, we designed seven oligonucleotides by either removing or adding flanking nucleotides on the 5ʹ and 3ʹ ends of nusinersen (Figure 3A). We added nucleotides in roughly the same purine:pyrimindine ratio as nusinersen (SI Table 7). When tested in CAPA, the CP₅₀ values for these ct-oligonucleotides spanned an order of magnitude in range (Figure 3A, 3C). Removal of two nucleotides from nusinersen led to a slight increase in cytosolic penetration, with CP₅₀ values of 1.99 ± 0.11 μM for the ct-16mer compared to 2.87 ± 0.06 μM for ct-nusinersen which is an 18mer. By contrast, adding two nucleotides to nusinersen resulted in a 20mer oligonucleotide with a CP₅₀ greater than 5 μM. At 5 μM, the highest concentration tested, the ct-20mer only blocked 33.5 ± 1.4% of cytosolic HaloTag (SI Table 8, SI Figure 5). Surprisingly, adding additional nucleotides resulted in oligonucleotides with a wide range of CP₅₀ values. The 22-mer, 24-mer, 26-mer, 28-mer, and 30-mer oligonucleotides had CP₅₀ values of 0.72 ± 0.10 μM, 0.60 ± 0.08 μM, 2.38 ± 0.24 μM, 0.97 ± 0.12 μM, and greater than 5 μM, respectively (Figure 3A–C). The wide differences in the ability of these oligonucleotides to penetrate the cytosol can also be observed at a single concentration point by comparing the percentage of occupied HaloTag at 1.67 μM ct-oligonucleotide. At this concentration, the 20-mer and 30-mer occupied approximately 18% and 2% of the cytosolic HaloTag, respectively, while the 22mer and 24mer occupied 70–74% of HaloTag (Figure 3D, SI Table 8).

Figure 3: Effects of oligonucleotide length on cytosolic penetration.

(A) Base sequences, T_m values associated with hairpin formation as calculated by IDT OligoAnalyzer^™, and CP₅₀ values for ct-nusinersen and analogs with different lengths. (B) CAPA dose-dependence curves showing the degree of cytosolic penetration at concentrations between 5 and 0.007 μM following 24 h incubation in HeLa cells. (C) CP₅₀ values for ct-nusinersen and analogs with different lengths, derived from the curve fits shown in Figure 3B. (D) Percent occupied HaloTag after 24 hours of incubation with 1.67 μM ct-oligonucleotide. All data show results from three independent trials or four trials for ct-nusinersen, and error bars show standard error of the mean. CP₅₀ values are reported as average and standard error from three independent curve fits to three independent trials, or four for ct-nusinersen. Nearly all pairwise comparisons in Figure 3C and 3D show statistically significant differences, ranging from p = 0.05 to p = 0.0001, calculated using an ordinary one-way ANOVA test (SI Tables 9, 11).

Clearly, degree of cytosolic penetration after 24 hours for phosphorothioate, 2ʹ-O-methoxyethyl oligonucleotides does not vary consistently with oligonucleotide length or molecular weight. One factor that differed among these oligonucleotides was the expected degree of secondary structure (SI Figure 6) as predicted by the IDT OligoAnalyzer^™ (https://www.idtdna.com/oligoanalyzer). SI Figure 6 shows the lowest-energy secondary structure(s) that each oligonucleotide is predicted to adopt at 37°C. Notably, the 20-mer, 26-mer, and 30-mer oligonucleotides are the only oligonucleotides in this series with predicted T_m values above 37 °C (SI Table 18) and these are the oligonucleotides with the poorest cytosolic penetration. This led us to hypothesize that extent of secondary structure might be an explanation for the large differences in cytosolic penetration observed for the ct-oligonucleotides of different lengths.

Effects of secondary structure on cytosolic penetration

To evaluate the predictions made by the RNA Fold algorithm on degree of secondary structure among the oligonucleotides of different lengths, we ran the chloroalkane-tagged oligonucleotides on an agarose gel (Figure 4A). Staining with ethidium bromide revealed patterns in both intensity and mobility of the bands. Specifically, the 20mer, 26mer, and 30mer had the darkest overall intensities when stained with ethidium bromide (Figure 4A–B). Also, the 20mer, 26mer, and 30mer had greater mobility through the gel relative to their overall length, indicating that they formed a more compact structure than the other five oligonucleotides (Figure 4A,C). These data support the prediction that these oligonucleotides possess a greater degree of secondary structure, and that their more compact structure correlates with poorer overall cytosolic penetration.

Figure 4. Degree of secondary structure for chloroalkane-tagged oligonucleotides of different lengths, visualized by agarose gel electrophoresis.

(A) Image of a 4% agarose gel run with 3 micrograms of each oligonucleotide, stained with ethidium bromide. A ssDNA ladder is provided for reference, though these oligonucleotides have slower overall mobility than ssDNA oligonucleotides of the same length due to the chloroalkane tag and due to PS, and 2ʹ modifications. Image is representative of three independent trials (SI Figure 7). (B) Band intensity of each ct-oligonucleotide, quantified using ImageJ.^[47] (C) Band mobility of each ct-oligonucleotide. The dotted line shows a linear fit to the averaged data points for the 16-mer, 18-mer, 22-mer, 24-mer, and 28-mer oligonucleotides. The dotted line illustrates that, compared to the other oligonucleotides tested, the 20-mer, 26-mer, and 30-mer oligonucleotides had greater mobility relative to their length. All data shown are the average values from three independent trials, and error bars show standard error of the mean.

To further examine the effects of secondary structure on cytosolic penetration, we tried using heating and cooling cycles to alter the degree of secondary structure for individual oligonucleotides. Oligonucleotides were heated to 95°C, then either slow-cooled over the course of 120 minutes or fast-cooled on ice. Prior to repeating CAPA measurements with these altered samples, we measured absorbance at 260 nm. Any differences in absorption before and after the heating and cooling cycle were ascribed to a hyperchromic shift due to a change in the overall degree of secondary structure. Generally, the UV absorbance increased for the slow-cooled and fast-cooled samples compared to the unheated controls (SI Figure 8, SI Tables 12, 13), which is consistent with the oligonucleotides having less overall secondary structure after the heat-cool cycle.^[48] We proceeded to carry out CAPA with these samples using the same experimental conditions that we used for the unheated samples (SI Figure 9). However, there were only small differences in the CP₅₀ values (only 0.4-fold change at most) and these slight changes did not correlate to changes in absorbance at 260 nm (SI Figures 10, SI Tables 14, 15).

Finally, we further examined the effects of secondary structure on cell penetration by testing a scrambled version of nusinersen that we intentionally designed to have a high degree of hairpin formation (SI Figure 11). On agarose gels, this scrambled analog had higher mobility consistent with hairpin formation (SI Figure 12). The CP₅₀ value of the scrambled analog was 5.71 ± 0.43 μM, indicating approximately 2-fold poorer cytosolic penetration than nusinersen (Figure 5). While it does not control for all possible differences between these oligonucleotides, this experiment provides further evidence that a higher degree of secondary structure formation results in poorer cytosolic penetration for PS, 2ʹ-MOE-modified oligonucleotides.

Figure 5: Comparing cytosolic penetration for nusinersen and a scrambled analog with high hairpin content.

(A) Base sequences, T_m values associated with hairpin formation as calculated by IDT OligoAnalyzer^™, and CP₅₀ values for nusinersen and the scrambled hairpin. (B) CAPA dose-dependence curves showing the degree of cytosolic penetration at concentrations between 15 and 0.02 μM for the ct-hairpin and between 5 and 0.007 μM for ct-nusinersen following 24 h incubation in HeLa cells. (C) CP₅₀ values for ct-hairpin and ct-nusinersen, derived from the curve fits shown in Figure 5B. Data show results from three independent trials for ct-hairpin or four trials for ct-nusinersen, and error bars show standard error of the mean. CP₅₀ values are reported as average and standard error from three independent curve fits to three independent trials (or four for ct-nusinersen). *** denotes p < 0.001 calculated using an ordinary one-way ANOVA test (SI Table 16).

Discussion

ASOs are a promising drug modality for treating diseases for which small molecule therapeutics are difficult or impossible to develop. They are also promising therapeutics for treating some rare genetic diseases in an allele-specific, N-of-1 manner.^[49] However, there is still limited knowledge and lack of consensus on how sequence, length, and other characteristics of ASOs affect their cytosolic penetration. In this study, we used the drug nusinersen as a parent compound to explore the effects of base composition, base sequence, length, and secondary structure on cytosolic penetration. Overall, we observed that differences in base sequence alone had a mild effect on cytosolic penetration (up to a 2-fold difference when comparing ct-nusinersen, ct-SCR1–5, and ct-hairpin). However, changes in base composition had a larger effect (a roughly 4-fold difference between ct-nusinersen and its reverse complement) and changes in overall extent of secondary structure had the largest effect (up to 10-fold differences among ct-oligonucleotides of different lengths and different degrees of secondary structure).

Previous studies have also shown that base sequence and base composition affect delivery efficiency. Many of these prior studies used PS-modified ASOs without 2ʹ modifications,^[21–23] so the data in this study extends these investigations to more modern, druglike oligonucelotides. Most of these prior studies also studied some manner of facilitated delivery, but in this work we examined gymnotic delivery. Also, prior studies examined activity and/or total cellular uptake, but in this work we examined cytosolic delivery without interference from endosomally trapped material. As we described previously for a variety of chemically modified ASO and siRNA therapeutics,^[33] these data provide a valuable measurement of productive uptake for oligonucleotide therapeutics.

Prior studies of oligonucleotides of different lengths showed inconsistent or contradictory results. For example, early work in the field observed a linear inverse relationship between length and overall oligonucleotide uptake.^[25,26] More recently, activity of PS-modified ASOs was found to be variable with ASO length between 15 and 25 nucleotides, with a 19mer ASO most active in cell culture and in mice when administered by subcutaneous injection.^[50] By contrast, a study that used flow cytometry to measure the total cellular uptake of dye-labeled, PS and 2ʹ-modified ASOs ranging from 7 to 30 nucleotides observed that longer ASOs had more efficient total cellular uptake.^[51] Finally, a study using chemically modified gapmers that were delivered in cells using transfection reagents showed that longer ASO was more active against the target while also minimizing off-target effects.^[52] In this work, our data addresses only extent of delivery to the cytosol, rather than total uptake or cellular activity. Our data do not address broader effects on total cellular uptake or subtle effects on activity once ASOs are delivered to the interior of the cell, which likely contribute to some of the discrepancies with prior work in this area. However, our data suggest that formation of secondary structure is a major factor for cytosolic penetration for oligonucleotides, and we recommend that continuing work in this area pay special attention to predicted and measured degree of secondary structure for PS, 2ʹ-modified ASOs.

There is some prior work that suggested that secondary structure formation can affect ASO activity. In a recent study that used PS-modified gapmers flanked with LNA-modified residues, the addition of a small (3bp) or medium (5bp) double-stranded stem to the 3ʹ end of the gapmer resulted in a more active oligonucleotide, while the addition of a 7bp stem did not improve activity.^[53] When the same length stems were added on the 5ʹ end the ASO with the 5bp stem was the least active, which implies that the local context of the secondary structure also matters. The authors hypothesized that the increasing efficiency of the oligos with the hairpin structure was a result of higher stability against exonucleases, which especially relevant for gapmers which have unmodified residues.^[54] A recent study by Østergaard et al. found that hairpin-forming ASOs are active, but the more stable the hairpin the less active the ASO.^[55] Our data agree with these studies, and provide an explanation beyond ASO folding, annealing to the mRNA, and recruiting cellular factors for activity. Instead, our data suggest that hairpin formation decreases cytosolic penetration and that this accounts for changes in activity previously observed in cellular models.

The mechanism by which increased secondary structure interferes with cytosolic delivery is unclear. Importantly, CAPA does not deconvolute effects on different parts of the internalization pathway. Thus, the effects of hairpin structure on cytosolic penetration may be due to decreased overall association with the membrane, decreased efficiency of uptake into endosomes, decreased efficiency of escape from endosomes, or some combination of these factors. Matile and colleagues have recently suggested that PS-modified oligonucleotides enter the cells through a “thiol-mediated uptake” mechanism mediated by transient covalent interactions with thiols on cell surface proteins.^[56,57] This model was inspired by prior work which showed that antisense oligonucleotides with phosphorothioate backbones interact with many cell-surface proteins, most of them cysteine-rich.^[3,15,16,58] The thiol-mediated uptake model was supported by several studies that selectively modulated the interactions of PS-oligonucleotides with cell surface thiols.^[59] In this model of PS-oligonucleotide uptake, hairpin structure would be expected to decrease cytosolic uptake because phosphorothioate groups within hairpins are closer together and less accessible than phosphorothioates in unstructured oligonucleotides (Figure 6). Thus, well-structured oligonucleotides would be less able to form multiple, cooperative transient interactions with thiols on the cell surface. It remains unclear whether these transient interactions alter initial association with the membrane, efficiency of uptake into endosomes, or efficiency of escape from endosomes or lysosomes. While mechanistic details remain to be investigated, our data support the thiol-mediated uptake model, or any model in which multiple backbone groups interact with multiple cell surface groups in a transient manner at any stage of internalization.

Figure 6. Effect of secondary structure formation on cell penetration via a thiol-mediated uptake model.

Phosphorothioate oligonucleotides with a low degree of secondary structure (left) can more easily interact with multiple surface thiols. Phosphorothioate oligonucleotides with a higher degree of secondary structure (right) cannot interact with multiple surface thiols as easily. In this model, fewer thiol-mediated interactions may lead to decreased efficiency at one or more steps of internalization, including initial association with the membrane, endosomal uptake, and/or endosomal escape.

Because we noted up to 10-fold differences in cytosolic penetration among oligonucleotides with different extents of secondary structure, we hypothesized that the same oligonucleotide might have different cytosolic penetration efficiencies after a heating/cooling cycle. If this were true, this could be a major source of variance for all ASO research. However, we confirmed that heat/cool cycles do not appear to alter cytosolic penetration for ASO-length, PS,2ʹ-MOE-modified oligonucleotides. We attribute this observation to the likelihood that hairpin (or duplex) formation for small, highly modified oligonucleotides has rapid kinetics such that folded states are in equilibrium with unfolded states at room temperature. Based on these results, we conclude that one does not need to worry about cell penetration efficiency being affected by a short, druglike ASO being kinetically trapped in a specific secondary structure.

While many different factors undoubtedly have a synergistic effect on the cytosolic penetration of oligonucleotides, our data suggest that a low degree of secondary structure correlates with improved penetration. At room temperature, these secondary structures may be in equilibrium with unfolded states, but even transient formation of hairpin structures may decrease cytosolic penetration efficiency in a thiol-mediated uptake model (Figure 6). It remains to be tested whether purposeful design to avoid structure-forming sequences might improve ASO activity in applications where there is the flexibility to alter ASO sequence. Future work can also examine the effects of secondary structure with greater detail, including the effects of overall thermodynamic stability of the secondary structure, effects of specific secondary structure features such as stem length and loop length, and strategic placement of chemically modified backbones to promote or decrease structure to optimize activity.