Discovery of indole-modified aptamers for highly specific recognition of protein glycoforms

Glycosylation is one of the most abundant forms of post-translational modification, and can have a profound impact on a wide range of biological processes and diseases. Unfortunately, efforts to characterize the biological function of such modifications have been greatly hampered by the lack of affinity reagents that can differentiate protein glycoforms with robust affinity and specificity. In this work, we use a fluorescence-activated cell sorting (FACS)-based approach to generate and screen aptamers with indole-modified bases, which are capable of recognizing and differentiating between specific protein glycoforms. Using this approach, we were able to select base-modified aptamers that exhibit strong selectivity for specific glycoforms of two different proteins. These aptamers can discriminate between molecules that differ only in their glycan modifications, and can also be used to label glycoproteins on the surface of cultured cells. We believe our strategy should offer a generally-applicable approach for developing useful reagents for glycobiology research.

1. In the SI, the authors state that several rounds of traditional SELEX were carried out prior to MPPD due to the limited throughout of FACS. This pre-selection should be discussed in the main text. There is also some confusion regarding round numbering. For example, is round 0 in the main text for the RB aptamer actually the 6th round of total selection? Some clarification is needed. 2. Statements like "Based on the clear enrichment after two rounds of selection,…" and "Although the round 2 pool…" are misleading because 7 rounds of selection were actually carried out in this example. In all cases, the pre-selection rounds should be included when discussing the number of rounds used. 3. When the authors discuss the "naïve" library (e.g. Figure 3B), is this the initial library or the library following pre-enrichment? 4. The sequence of i-6 is nearly 50% indole-modified dUTP. Is it possible that having a high level of indole moieties results in a modest affinity for N-glycans regardless of aptamer sequence? It will be important to demonstrate that scrambled control of i-6, containing the same number of indole modifications, is unable to bind RB. 5. Aptamers isolated against fetuin have very few indole modifications (Table S2). F-1 has none at all, yet still binds fetuin with comparable affinity to f-4. It is important that the authors show that the fetuin aptamers require the indole moiety to bind fetuin. If not, these results argue against their primary assertion that the indole moiety is essential for selective recognition of protein glycoforms. 6. As far as I can tell, all characterization experiments described in this manuscript were carried out with aptamer particles (i.e. immobilized aptamers). Do the free, non-immobilized aptamers bind their targets? This is an important point because a requirement that the aptamers be bound to beads potentially limits the utility of this approach. 7. Some discussion of aptamer affinity is warranted. For example, are the Kd values reported herein adequate for potential downstream applications, such as detection of Fetuin in the blood? How does the affinity of these indole-modified aptamers compare to previously reported aptamers targeting protein glycans? 8. Chemical structures of the competitor glycans in Figure 5D would be helpful.
Reviewer #2: Remarks to the Author: In the manuscript "Discovery of indole-modified aptamers for highly specific recognition of protein glycoforms", Yoshikawa et al. present a particle-display-based workflow to generate and screen for aptamers containing indole-modified bases for recognizing and binding with glycosylated proteins. By using the workflow, the authors produced several aptamer candidates, which can distinguish glycosylated RNase B (RB) from non-glycosylated RNase A (RA). They showed that the interaction between RB and aptamer candidates was not significantly affected by free mannose, Man5, and Lewis A glycans. The authors further showed the workflow's potential to distinguish differently glycosylated fetuins. Finally, the authors suggest the RB-recognizing aptamer can cross-react with other glycoproteins by showing the aggregation of aptamer-coupled beads on paraformaldehydefixed Dictyostelium discoideum cell surface. While the topic is of great interest to protein glycosylation studies, the developed workflow lacks a comprehensive characterization of the binding epitopes of the screened aptamers. It remains unclear which part(s) of the target proteins were recognized and bound by the aptamers, which limits its applications. The successful generation of aptamers discriminating sialylated and asialylated fetuins is encouraging, but the characteristics of the binding epitopes are again missing. The bead aggregation experiment, unfortunately, in my opinion, leaves more questions than it answers. A good demonstration of the developed aptamer's potential applications in glycobiology will significantly strengthen the manuscript. Specific comments/questions: 1. The authors should comprehensively characterize the binding epitopes of the generated aptamers. For instance, free Man5 reduced the interaction of RB and the RB-specific aptamer by ~20% ( Figure 4D). Whether the aptamer can bind to other high-mannose-type glycans? Does the aptamer recognize a specific glycan type attached on RB? If not, can the authors specifically define which glycans the aptamer can bind and with which affinities? 2. The Kd of the protein-glycosylation-recognizing aptamer is at the low-µM level, which is not better than existing lectins. The authors may want to provide evidence to demonstrate the advantages of aptamers. 3. The bead aggregation experiment ( Figure 6) does not disclose what exactly the aptamer beads bound on the cell surface. The authors should provide more convincing evidence to show the aptamer can recognize protein glycoforms at a global level. 4. In the aptamer screening workflow, would it be possible to use all glycoproteins enriched from cell lysate (or even live cells) with and without a specific glycosidase treatment as the binding targets? A highly specific glycosidase may help screen aptamers recognizing a specific glycoform.
1. "In the SI, the authors state that several rounds of traditional SELEX were carried out prior to MPPD due to the limited throughout of FACS. This pre-selection should be discussed in the main text. There is also some confusion regarding round numbering. For example, is round 0 in the main text for the RB aptamer actually the 6th round of total selection? Some clarification is needed." We completely agree, and have added a description of the pre-selection to the main text and now make an explicit distinction between the rounds of pre-selection and the rounds of particle display. Figure 3 has been updated accordingly.

"Statements like "Based on the clear enrichment after two rounds of selection,…"
and "Although the round 2 pool…" are misleading because 7 rounds of selection were actually carried out in this example. In all cases, the pre-selection rounds should be included when discussing the number of rounds used." We agree that the references to selection rounds were previously unclear, and believe that the changes made in response to comment #1 should address any ambiguities.
3. "When the authors discuss the "naïve" library (e.g. Figure 3B), is this the initial library or the library following pre-enrichment?" This refers to the initial library, prior to pre-enrichment. We have added a sentence clarifying this in the manuscript.

"
The sequence of i-6 is nearly 50% indole-modified dUTP. Is it possible that having a high level of indole moieties results in a modest affinity for N-glycans regardless of aptamer sequence? It will be important to demonstrate that scrambled control of i-6, containing the same number of indole modifications, is unable to bind RB." We agree that this is an important consideration, and have run a scrambled-sequence control that has a randomized variable region containing the same distribution of each nucleotide. The scrambled sequence exhibited low levels of binding relative to the forward-primer negative control, indicating that the modification indeed has some baseline ability to interact with N-glycans. Nevertheless, the scrambled control showed significantly weaker RB binding than i-6, confirming that the aptamer sequence and structure is critically involved in glycan recognition. These results have been added to the SI ( Figure S3, shown below), with a sentence describing the experiment in the main text. (Table S2). F-1 has none at all, yet still binds fetuin with comparable affinity to f-4. It is important that the authors show that the fetuin aptamers require the indole moiety to bind fetuin. If not, these results argue against their primary assertion that the indole moiety is essential for selective recognition of protein glycoforms."

"Aptamers isolated against fetuin have very few indole modifications
To address the reviewer's suggestion, we ran a control experiment with a natural DNA version of the f-4 aptamer that does not contain indole modifications. This natural DNA sequence showed similar performance to the forward-primer negative control, confirming that the indole is necessary for fetuin recognition. We have added these data to the SI ( Figure S8, shown below) and added a sentence regarding this control experiment to the main text.
It should be noted that the affinity of the fetuin aptamers does seem to correlate with the presence of the modification. We believe that the reviewer is referring to f-2 rather than f-1, which contains two indole-modified groups; f-2 lacks an indole moiety and had the lowest binding of all aptamers tested in that assay, while f-4 and f-7, which contained the most modifications, had the strongest binding. We would also like to point out that the random region used for the fetuin selection was designed to include fewer indole moieties (VNVV) 10 , such that the presence of the four modification present in aptamer f-4 is greater than would be expected due to chance. We realize that this was not discussed clearly in the main text, and so we have added a description of the library used for the fetuin selection to the revised text.

"As far as I can tell, all characterization experiments described in this manuscript were carried out with aptamer particles (i.e. immobilized aptamers). Do the free, non-immobilized aptamers bind their targets? This is an important point because a requirement that the aptamers be bound to beads potentially limits the utility of this approach."
We thank the reviewer for this thoughtful comment. We have utilized microscale thermophoresis (MST) to interrogate the binding between Cy5 labeled aptamer and unlabeled glycoprotein target in solution. First, we examined the RB aptamer i-6, and found consistent K D measurements for the bead-based (~29 μM) and MST (~25 μM) assays. We also conducted the assay for RA, and observed weak binding for i-6 (K D ~150 μM). We were surprised to see evidence of binding to RA, as we observed no signal in bead-based assays at RA concentrations of up to 150 μM. However, we believe that the discrepancy is understandable considering the many differences between the two binding assays and the fact that MST assay is sensitive to fluorophore's environment.
We also conducted MST with aptamer f-4 for both fetuin and asialofetuin (n = 3). The MST results confirmed binding of the aptamer to fetuin, with a similar K D of ~10 μM (versus ~6 μM for the bead-based assay) and no measurable binding to asialofetuin.
We believe these MST experiments demonstrate that the aptamers can function in solution, and provide a valuable confirmation of binding of both the RB and fetuin aptamers. These experiments have been added to the SI section ( Figures S4 and S9).
7. "Some discussion of aptamer affinity is warranted. For example, are the Kd values reported herein adequate for potential downstream applications, such as detection of Fetuin in the blood? How does the affinity of these indole-modified aptamers compare to previously reported aptamers targeting protein glycans?" We have added discussion regarding previous aptamers towards protein glycans and the implication of their K D s to the manuscript. Briefly, we measured K D s similar to those observed for lectins (which are multimers) as well as aptamers and antibodies selected for glycans. These affinities are several orders of magnitude weaker than aptamers selected for glycoproteins in two previously-published reports cited in the manuscript (Díaz-Fernández et al. 2019, Li et al. 2008. However, the specificities of those aptamers are much lower, and we believe that the higher affinities are likely due to the aptamer forming more interactions with the protein epitopes. It is also possible that there are simply inherent differences between the glycoprotein targets themselves. We would like to note that if higher affinity is desired and less specificity is necessary, this method could be tuned by adjusting the concentrations and gating strategy for the counter-target. We selected the glycoproteins targeted in this study because they are well characterized, which was advantageous for the creation and validation of the aptamer generation pipeline, but believe this workflow can be applied to create aptamers towards a wide variety of other glycoprotein targets. Figure 5D would be helpful."

"Chemical structures of the competitor glycans in
The structures of the competitor molecules are now shown in Figure 4E. Please note that this figure has been modified to include five additional competitor molecules.

Reviewer 2:
The reviewer was generally enthusiastic about the topic and our aptamer discovery methodology. However, s/he felt that the initial manuscript needed more comprehensive characterization of the binding epitopes of the screened aptamers. We thank the reviewer for the support and have addressed these concerns in the revision.
1. "The authors should comprehensively characterize the binding epitopes of the generated aptamers. For instance, free Man5 reduced the interaction of RB and the RB-specific aptamer by ~20% ( Figure 4D). Whether the aptamer can bind to other high-mannose-type glycans? Does the aptamer recognize a specific glycan type attached on RB? If not, can the authors specifically define which glycans the aptamer can bind and with which affinities?" We agree that it is unclear exactly what the aptamers are binding to, and that additional epitope characterization is necessary. To answer this question, we extensively expanded the competition assays conducted in the manuscript. First, we obtained four additional high-mannose N-glycan standards that are present on RB and repeated the binding assay to determine if the aptamer was sensitive to these terminal branching mannose residues. We found that each of these N-glycan standards demonstrated similar inhibition of binding, indicating that the aptamer is not sensitive to the terminal mannose residues and perhaps recognizes the GlcNac core of the N-glycan. Second, since competition from the Man5 N-glycan was minimal at 20 μM, we increased the concentration to 50 μM to provide a more compelling case that aptamer binding is inhibited due to binding to the free form of the Man5 N-glycan. We observed a substantial decrease in signal, confirming that the aptamer is also binding to the glycan, albeit with a slightly lower affinity. Finally, although we had determined that the aptamer could not be inhibited by very high concentrations (up to 10 mM) of the free mannose monosaccharide, we never investigated whether the N-acetyl glucosamine (GlNac) monosaccharide, which is present in the core of the N-glycan, could inhibit binding. We therefore conducted a competition assay with 10 mM GlcNac, and observed slight inhibition of aptamer binding to RB, indicating modest affinity for the GlcNac monosaccharide. As a whole, these experiments suggest that the aptamer is recognizing part of the conserved N-glycan core, including GlcNac residues.
Finally, we would like to note that we wanted to create a truly comprehensive "map" of specificity for each glycan structure using the strategy described above. Unfortunately, we quickly learned that this is not feasible because well-defined glycans are simply not available for many of the structures of our interest, due to the challenges associated with their synthesis. Given this, we believe the above additional experiments provide sufficient insights into the RB aptamer's binding epitopes, and have added these competition assay results to the main text ( Figure 4C-E, shown below) as well as a discussion of the results.

"
The Kd of the protein-glycosylation-recognizing aptamer is at the low-µM level, which is not better than existing lectins. The authors may want to provide evidence to demonstrate the advantages of aptamers." We believe that the primary advantage our aptamer selection system provides over lectins is that they can be generated through an in vitro selection process to generate entirely novel reagents, whereas lectins must be identified from molecules that already exist in nature. We would also like to note that lectins are typically expressed as multimers, which greatly enhances their K D through avidity, and typically have monovalent affinities in the very high micromolar to millimolar range. We have added some discussion regarding the benefits of aptamers over lectins to the Conclusion section of the manuscript.
3. "The bead aggregation experiment ( Figure 6) does not disclose what exactly the aptamer beads bound on the cell surface. The authors should provide more convincing evidence to show the aptamer can recognize protein glycoforms at a global level." We believe that the additional competition binding assay experiments provide solid evidence that the aptamer can recognize high-mannose N-glycans, which we believe provides sufficient evidence that the aptamer can recognize N-glycans in a global context.

"
In the aptamer screening workflow, would it be possible to use all glycoproteins enriched from cell lysate (or even live cells) with and without a specific glycosidase treatment as the binding targets? A highly specific glycosidase may help screen aptamers recognizing a specific glycoform." We agree that the reviewer's suggestion would be an excellent and worthwhile idea for a future experiment, and although this would fall outside the scope of the current manuscript, we strongly believe that the workflow developed in this project could serve as the foundation for such an effort.