Aptamers have great potential as therapeutic or diagnostic ligands, and numerous aptamers have been developed to bind to a variety of clinically relevant cell targets. However, assessment of aptamer function has varied quite dramatically between different groups, with methods often not directly assessing the first step in any target-specific downstream events, target binding. Indeed, a previous evaluation revealed potential challenges in using aptamers for cell typing37. Here we sought to establish and apply standardized protocols to identify aptamers with the ability to engage their targets on cells both in vitro and in vivo as a means to evaluate their future potential for the targeted delivery of therapeutics.
Due to their innate negative charge, we have observed that aptamers can demonstrate non-specific interactions with proteins or other cellular components, likely those predisposed to bind other polyanions (e.g., heparin sulfate) or scavenger receptors, which are known to bind nucleic acids and are expressed on many cell types38,39. Additionally, it is well known that dead or sickly cells can purge or non-specifically uptake macromolecules as their membranes lose integrity40. Indeed, when incubated for prolonged times (>6 h) with live cells, we observed significant levels (>5-fold) of non-specific cell staining. This coupled with the fact that different cell types display vastly different levels of non-specific background binding means care needs to be taken to account for these variations when confirming target specificity. For example, the oft-used PSMA (+) LNCaP prostate cancer cells display very high levels of non-specific binding (>10-fold) when compared to the PSMA (−) cell line PC3 cells.
In vitro, we have found that the inclusion of a non-specific competitor, ssDNA, can help to minimize non-specific cell staining and that this effect varies with cell line. Other anionic competitors (e.g., tRNA, dextran sulfate) likely function similarly. However, the presence of competitors is unable to completely block non-specific uptake or compensate for the variable levels of non-specific staining observed across different cell lines. As sites with positive charge potential have previously been reported to be sites of aptamer binding41,42,43, one might expect the presence of poly-anionic competitor to also adversely affect specific binding. However, we have previously observed that the overall electrostatic contributions to the formation of the binding complex can actually be quite small (~18%44). Thus, charge–charge interactions may not be as important as one might expect. For the aptamers assessed here that displayed specific binding for their targets (C2.min, Waz, A9.min, SGC8c, and E07), the presence of a non-specific competitor, even at 1 mg/mL, had only minor effects on the aptamer’s cell staining ability (see Fig. 3 and Supplementary Figs. 11–21, 22–32, 33–43, 99–109, and 121–131). In contrast, the presence of the non-specific competitor had a more significant effect on the level of background, non-specific cell labeling, and suppressed cell staining (over 10-fold) on some cell lines (Figs. 1 and 2).
Perhaps more importantly, when we omitted non-specific competitor from our binding correlation studies across multiple cell types (Fig. 3 and Table 3), two of the four aptamers we confirmed as specific for their targets, C2.min and Waz, were no longer statistically significant (P > 0.05), a consequence of increased non-specific binding by the non-targeting control sequence, C36. As we know that both C2.min and Waz are, specific for the human transferrin receptor, this serves to further highlight the need for including blocking agents in such analyses, as their absence may not only lead to the identification of false positive, but also false negatives.
Of the compounds we tested, 13 of the 15 molecules have reported Kd values or apparent binding affinities of <200 nM (see Supplementary Table 5). Thus, with the exception of XE0-mini, 2-2(t), and Waz, all experiments were performed under conditions in which the aptamer was in at least a 5 to 10-fold and sometimes >100-fold excess of the reported affinity constant for each aptamer. For the molecules outside this range, XE02-mini has a reported affinity of 1.5 μM by flow cytometry using conditions similar to those we employed here and would be expected to show some signal under the conditions we utilized. 2-2(t) was reported to demonstrate cell binding activity by flow cytometry at 1 μM19. The discrepancy observed between our work and both of these reports may rest on the choice of control aptamer used (XE02-mini binding was compared to the naïve aptamer library) and/or the lack of blocking agents (neither study employed a non-specific competitor).
It is possible that the binding affinity for all the compounds that failed to function in our assays are significantly worse on cellular targets than the values reported in their parent publications. However, the fact that we do not see any target-specific binding or cell uptake even at concentrations as high as 1 µM limits the utility of these molecules for the targeted delivery of therapeutics. For comparison, clinically approved antibody drug conjugates (e.g., Trastuzumab Emtansine, Brentuximab Vedotin, Enfortumab Vedotin) and those in clinical trials (e.g., Mirvetuximab soravtansine, Trastuzumab duocarmazine) or their parent antibody, which are being utilized for targeted delivery, typically demonstrate apparent binding affinities on cells under assay conditions similar to those used here in the low to sub nanomolar range (see for example refs. 45,46,47,48).
In our comparison of aptamer and antibody binding, for a given aptamer/antibody pair, there was some variability across cell types. For example, the signal from Waz on HeLa-PSMA cells using the ‘internalizing and binding with blocking’ protocol was lower than that from the ‘binding’ protocol (Fig. 3B). Similarly, the signal for E07 assayed with ‘internalizing and binding with blocking’ was lower than that observed for the ‘binding’ protocol (Fig. 3F). In some cases, these discrepancies may have arisen from differences in the background uptake levels, as our non-targeting control sequence, C36, was used to normalize the data. However, in other cases, differences in binding may have been due to target/epitope variations (e.g., variation in glycosylation, the presence of splice variants, etc.) on different cell types. Thus, as an additional check to validate aptamer specificity, we used siRNA to knock down the target receptor. Importantly, the results from our analysis using siRNA identified the same ‘hits’ as our correlative study across cell types.
There were a number of discrepancies between our in vitro work and those previously reported in the aptamer literature. In an attempt to reconcile some of these differences between our work and the literature, we also performed a number of additional experiments, which can be found in Supplementary Figs. 193–201. For example, in our studies the often used anti-PSMA aptamer A10.3 and its minimized variant A10.3.2, failed to demonstrate any significant correlation to anti-PSMA antibody binding to PSMA positive cell lines. Both aptamers showed a slight increases in binding on the PSMA positive LNCaP cells, with A10.3 performing better than A10.3.2 (see also Supplementary Figs. 50 and 61); however, using siRNA to knock down PSMA expression, we were unable to correlate aptamer binding with protein knockdown (Fig. 4 and Table 4). These aptamers also both failed to show any appreciable binding to 22Rv1 cells, another prostate cell line that naturally expresses PSMA, or either of the two engineered PSMA-expressing cell lines we tested (HeLa-PSMA and PC3-PSMA, Fig. 3D, E; also see Supplementary Figs. 44, 46, 52, 55, 57, and 63). As these molecules are two of the most widely used aptamers and reported to bind the human protein when expressed natively on human cells (22Rv1 and LnCAP10,11,12), ectopically on mouse cell lines (B16-PSMA, CT26-PSMA49,50), and even bind to canine cells expressing the canine PSMA (canine hemangiosarcoma cell lines51), we performed experiments on recombinant PSMA protein to reconcile these results. Interestingly, while A10.3 and A10.3.2 both fail to bind PSMA produced from mammalian cells, the molecules showed modest binding for protein produced from insect cells (Supplementary Fig. 193). As protein produced in insect cells was the target for the original selection11, the lack of activity on human protein may be due to differences in glycosylation. Similar selectivity in binding based on glycosylation has been shown in a selection against the EGFR receptor52. For this reason, it is important to characterize aptamer-target binding in a cellular context to determine specificity.
The aptamers GL21.T18, CL415, and EpDT321 demonstrated the lowest levels of correlation with target receptor expression (Table 3). They also failed to yield any target-specific cell labeling in our siRNA knockdown experiments, even when performed under conditions that favor staining (no blocking agents). In additional supplementary experiments using flow cytometry, the anti-AXL aptamer, GL21.T, also failed to stain HEK293T cells overexpressing AXL using all three of our assay formats (Supplementary Fig. 194). Similarly, in contrast to the anti-EGFR aptamer E07.min, the reported anti-EGFR aptamer CL4 failed to stain A431 cells, an EGFRhigh cell line often employed for studies involving this receptor (Supplementary Fig. 195). In the case of the reported anti-EpCAM aptamer EpDT3, we performed additional experiments to assess prolonged, 24 h, incubation time on cell staining using HT29 cells, as a recent report suggested EpCAM endocytosis and subsequent cell staining by this molecule to be slow53. However, here again, the aptamer failed to display any signal above that observed for a non-specific control (Supplementary Fig. 196).
Similarly, both of the reported anti-HER2 aptamers we tested, SE-15-8-mini20 and 2-2(t)19, failed to show any significant binding to SkBR3 cells, the only HER2high cell line we tested. Interestingly, both of these aptamers showed increased staining when assessed for internalization and binding on Jurkat cells, the only outright HER2Negative cell line we tested. Together, these results suggest that neither HER2 aptamer is specific for the reported target, a conclusion supported by our siRNA data (Fig. 4E and Supplementary Fig. 186).
Again, it is important to note the standardized in vitro binding conditions we have utilized to test some of these molecules are somewhat different than conditions originally reported for each aptamer selection. However, molecules best suited for use as detection agents or for the development of therapeutics, arguably, need to be robust and maintain function in physiologically relevant biological buffers (e.g., DPBS, media, HBSS, etc.). Since folding conditions and buffers do vary between studies, there is the potential for aptamers to require specific conditions (e.g., temperature, time, divalent cations) to adopt a functional form. Therefore, to further examine these effects, we tested the aptamers that did not demonstrate specific signal in our assays for target binding when folded under previously reported conditions. However, when compared to the binding of aptamers folded under literature conditions with those folded under our standardized conditions, no changes in fluorescence signal were observed (Supplementary Fig. 199).
For our in vivo studies, we used IVIS imaging to look for aptamer localization to tumors following intravenous (tail vein) injection. Tumors were grown to ~0.5–1.0 cm in diameter and size matched for comparison to controls to minimize the potential for artifacts that may result from variation in vascularization54,55. Our choices of tumor models for these experiments, PSMA-expressing 22Rv1 cells or PC3-PSMA cells, express high levels of hTfR and PSMA (>10-fold staining as assessed by aptamer staining using flow cytometry), but low or no levels of EGFR and PTK7. As such, we limited our in vivo analyses to the in vitro validated anti-hTfR aptamers C2.min and Waz and the anti-PSMA aptamer A9. We also extended this analysis to the other two anti-PSMA aptamers, A10-3 and A10-3.2, given their prevalence in the aptamer literature. We note that while we didn’t assess the in vivo function of E07 or SGC8c, their ability to localize and stain tumors in vivo has previously been reported56,57.
In vivo systems are significantly more complex than assays in culture. It is thus, perhaps, not surprising that a strong specific signal in cell culture does not guarantee successful translation to in vivo activity. It is also important to note that while composed of 2′F RNA, none of the molecules we tested are fully stabilized molecules and degrade in serum; all of the aptamers displayed half-life in serum of between ~5 to 11+ hours (Supplementary Table 3). Based on their small size (~15 kDa), aptamer clearance is fast. When fit using a two-compartment model, the half-life for the first phase of clearance for all aptamers tested was ~2 min, followed by a longer, ~2 h, phase. Thus, for all of these molecules tested, plasma clearance, not stability, appears to be the limiting factor in vivo. This fact allowed us to directly compare in vivo activities of the aptamers based on monitoring the fluorescence signal, eschewing the more traditional dual-probe hybridization approach often used to confirm the presence of full length functional aptamers4.
In vivo, Waz displayed a specific signal above a non-specific control (C36) in both tumor models, a fact we confirmed by performing additional analyses with point mutants of this aptamer, which demonstrated diminished (Waz.X) or abolished function (Waz.GGG). The in vitro activity of these molecules correlated well with their in vivo function. Surprisingly, C2.min and A9.min did not demonstrate any activity over that observed for our non-targeting control, C36. Similarly, but consistent with our in vitro results, neither A10.3 nor A10.3-2 demonstrated significant tumor staining. Thus, unlike the other molecules tested, Waz is capable of engaging its target in vivo. However, while tumor localization is an encouraging proof of concept, since Waz does not cross react with the murine transferrin receptor, additional studies are required to better understand its targeting capability in more realistic systems.
However, it is important to note that even for Waz, the signal in the tumor above background in our images at 12 h was only ~2–3-fold as assessed by NIR imaging. These data, in fact, are similar in magnitude to the signals we observed using another aptamer on which we have recently reported, E358. Together, these data demonstrate that 2′F modified RNA aptamers can target tumors even without further stabilization or PEGylation. Alterations to slow clearance rates (e.g., PEGylation) are likely to improve on these results but will require additional aptamer stabilization. Such alterations might also allow molecules such as C2.min and A9.min, which are specific for their targets in vitro, a greater opportunity to bind when utilized in vivo.
Interestingly, despite the fact that both C2.min and Waz bind the same target with similar affinities (C2 actually binds ~4-fold more tightly) and that the molecules share similar serum stabilities and clearance rates (Supplementary Table 3), only Waz functioned in vivo. There is likely more at play when moving from in vitro to in vivo. Indeed, while we have previously reported that C2.min showed no activity on mouse cell lines, in more recent analyses using DY650 labeled aptamers, which provides for increased signal over the AF488 labeled molecules used in our earlier studies, we have observed some cross-reactivity with murine transferrin receptor (Supplementary Fig. 197). Additionally, C2.min competes with both human transferrin13 and, to a lesser extent, murine transferrin (Supplementary Fig. 198), which could serve to suppress tumor targeting in vivo.
In the case of the anti-PSMA aptamer, A9.min, Dassie et al. reported on the ability of an anti-PSMA aptamer, A9g, to localize to PC3 PSMA + cells following systemic injection59. A9g is, in fact, the same aptamer sequence as A9.min, but linked to the fluorescent dye CW800 via an amide linkage. In light of this difference, we performed additional studies using an amide linked AF750 in an attempt to reconcile the difference in our results. These studies too, failed to demonstrate tumor staining (Supplementary Fig. 200). Thus, the observed discrepancy between that work and our own may reflect differences in the cell lines used, with the line employed by Dassie et al. potentially displaying more favorable target expression levels, or that the resulting tumors displayed more favorable targeting characteristics (e.g., large tumor size and/or increased leaky vasculature). Alternately, the observed differences could be a result of the choice of fluorescent dye (AF750 versus CW800).
To the extent that these in vivo studies may be affected by our choice of fluorescent dye, it is also important to note that in all of our studies, we could not rule out that the identity of the dye molecule (DY650, AF750) or its placement at the 5′ end of the aptamer via a 5′ thioester linkage did not adversely affect aptamer function. Therefore, for many of the molecules that failed to function, experiments were also performed with an alternate dye, AF488, to rule out this possibility. For example, anti-PSMA aptamers that failed to function bearing DY650 also failed to function with AF488 (Supplementary Fig. 201).
In summary, we have described standardized conditions to assess aptamer suitability for cell surface targeting, since an aptamer’s reported affinity for a target or its reported activity does not necessarily indicate how robustly it can function on cells or in vivo. We applied our assay conditions to the analysis and characterization of 15 cell surface targeting aptamers reported from the literature and identified the anti-transferrin receptor aptamer, Waz, as a robust candidate for future in vivo studies. The majority of other aptamers we tested failed to function effectively as cell targeting agents.
Through our investigation, we have found that a number of different factors can skew apparent aptamer function, such as variable levels of non-specific uptake between different cells. However, the use of appropriate non-targeting controls, inclusion of blocking agents, and validation of target engagement through knock down or knock in experiments would eliminate those non-specific and somewhat misleading interactions. Going forward, researchers can apply these assay conditions to determine an aptamer’s effectiveness before wasting time on more costly functional experiments. We do caution, however, that aptamer function in vitro may not necessarily translate to aptamer function in vivo.
Taken as a whole, our work should provide a framework for future studies in the development and validation of aptamers to cell surface targets. Aptamers that demonstrate high target specificity and efficient cell staining are likely to provide greater success in delivering a diagnostic or therapeutic payload while avoiding off target effects. As such, we expect that more careful and detailed aptamer characterization and target validation will lead to the identification of more robust aptamers in the future. This, in turn, should translate to the development of more effective aptamer-based diagnostics and therapeutics.
