Antibody-controlled actuation of DNA-based molecular circuits

DNA-based molecular circuits allow autonomous signal processing, but their actuation has relied mostly on RNA/DNA-based inputs, limiting their application in synthetic biology, biomedicine and molecular diagnostics. Here we introduce a generic method to translate the presence of an antibody into a unique DNA strand, enabling the use of antibodies as specific inputs for DNA-based molecular computing. Our approach, antibody-templated strand exchange (ATSE), uses the characteristic bivalent architecture of antibodies to promote DNA-strand exchange reactions both thermodynamically and kinetically. Detailed characterization of the ATSE reaction allowed the establishment of a comprehensive model that describes the kinetics and thermodynamics of ATSE as a function of toehold length, antibody–epitope affinity and concentration. ATSE enables the introduction of complex signal processing in antibody-based diagnostics, as demonstrated here by constructing molecular circuits for multiplex antibody detection, integration of multiple antibody inputs using logic gates and actuation of enzymes and DNAzymes for signal amplification.

The explanation about why the background reaction if fig 2 is drawn as irreversible (whereas it is described in the text as thermodynamically infavorable) is given only in SI S19 point 5. This a a bit late.
Fig2b, the number written in the plots are not explained. They probably represent toehold length, but this should be stated somewhere. Also, is this data the same as presented in Fig1 (but converted to nM)? It is not clear.
In Fig2e, it is a bit weird to plot the background intensity as a function of antibody, since this reaction is by definition the one that does NOT depend on [antibody]. Also, the use of a linear scale in these heat maps flattens the plots and as a consequence, the shape of the S/B plot cannot be guessed from the background and signal plots. A log scale would probably fix that.
The author do not really adress the limit of detection of their method. ( even if it is possibly not the most relevant point here, as the purpose is more to show the possibility to link antibody chemistry to DNA Strand displacement circuits). However, when they introduce downstream reporting strategy (with additional aplification modules introduced in fig4), the authors could compare the LOD with the direct circuit presented in the beginning of the paper. Since theranostics are used as a motivation, it would be interesting to discuss the relevance of this LOD to clinical applications.
The presentation of DNA encoded "molecular circuitry" (and associated references) could be better balanced and not overly focused on strand displacement cascades. A couple of reference to enzymatic approaches could help the uninformed reader to obtain an informative picture of the field, especially since enzymatic circuit typically provide the amplification effect that could be required here. This is also important since the paper later uses an elegant way to control enzymatic activity with DNA strands.
While the rate acceleration of 20~100 described in tge paper is fine, one can only wonder if it could be pushed further, and how. The only design element that has been optimized is the length of the toehold, but how the total length of BO was selected is not explained. Is the current choice better than a slightly longer, or slightly shorter duplex? This may be relevant since the antibody imposes geometric constraints on the intramolecular reaction.
Typos: -separate -circuitery Reviewer #2 (Remarks to the Author): In the current work, the authors introduced Antibody Templated Strand Exchange (ATSE) in constructing molecular circuits. The use of the characteristic bivalent Y-shaped architecture of antibodies is certainly powerful, by promoting DNA strand exchange both thermodynamically and kinetically. However, ATSE is not as generic as the authors claimed in this manuscript. The intracellular application (e.g. smart drug delivery and even the rewiring of natural signalling pathways in vivo based on DNA circuits) of this antibody-based approach is very likely impossible. The conjugation and purification of DNA-peptide complex is quite laborious. Peptides without cysteine residue can not be conjugated to DNA-NH2 according to the protocol in Supplementary Information. Furthermore, after conjugation with DNA, binding affinity of peptides to antibodies may be dramatically disturbed. Therefore, this manuscript is an interesting extension/continuation of work that has been previously done by authors. However, it does not have sufficient novelty to warrant publication in Nature Communications.
The experiments are well-designed and the presentation is clear. The following specific points should be properly addressed: [1] Authors determined dissociation constant (Kd) for fluorescein labeled HA binding to the antibody ( Figure S8, P4, line80) and the value of Kd was calculated to be 0.2 nM. In ASTE process, it is DNA-peptide binding to antibody not fluorescein-peptide. How is the Kd for fluorescein-DNA labeled HA to the antibody？Does DNA attachment significantly influence the binding affinity of HA to antibody?
[2] This question refers to Question 1. Authors extended ASTE approach to anti-HIV1-p17 antibody. The S/B ratio for anti-HIV1-p17 antibody was ~20, compared to ~100 for anti-HA antibody ( Figure 1d). The reason for the lower ATSE rate was proposed to be the lower affinity of HIV1-p17 to antiboby, with a Kd of 42 nM based on literature. It is better to calculate the Kd of DNA-HIV1-p17 to antibody and then compare the two Kd values of DNA-peptide binding to antibody. Moreover, it will be great to include extra data of a DNA-peptide/antibody system with a Kd between HA and HIV1-p17 systems.
[3] Since the background reaction only starts to contribute significantly to the observed fluorescence after 60min, when the ATSE reaction has reached completion (P7, line 145), why the analysis of Figure 2e and 2f was based on data at 180 min not 60 min? Is there a time dependency of the estimation of optimal initial conditions ( Figure 2g and 2h)?
[4] This question refers to Question 3. In Figure 3, the fluorescence levels of those DNA circuits was measured after 3 h. How about the results after 1 h?
[5] Authors claimed that "activity levels as estimated from Fig 4a are at least 20-fold higher after 1 h in cubation in the presebnce of antibody compared to the background reaction". How did they get this conclusion?
Minor points: [6] To make this manuscript easier to read, it is better to point out which system (HA or HIV1-p17) is employed in text, Figure 2 and relevant supporting figures.
[7] In Figure 4a, the pentagram of P1 and the spherosome of P2 should be switched.

Reviewer #3 (Remarks to the Author):
The manuscript describes the "antibody templated strand exchange" or ATSE, which specifically translates the presence of antibodies to unique DNA strand outputs. Subsequently, ATSE makes monoclonal antibodies (and potentially their specific targets) compatible with the well established molecular computing systems powered by nucleic acids. The design is elegant and very smart. The reaction mechanisms is carefully studied by experiments and modeling. Finally, the ATSE reactions are coupled with a few simple logic gates and enzyme catalyzed signal-amplification reactions to fully prove the concept.
The study is quite thorough and most claims are well supported by substantial amount of data and careful analyses. Given the interest of molecular computing in nanotechnology and synthetic biology communities, this work should be interesting to a broad range of readers.
Nevertheless I have the following concerns/questions that I want the authors to address: (1) According to main text, Figure 1b shows the ATSE with toehold length increasing from 1nt to 6nt. However, the DNA sequences shown in Table S1 says otherwise. BxOx0 has 3nt overhang (TAG) for Ix to bind. So while Figure 1b shows T=1,2,3...6 , the toehold length should actually be 4,5,6...9. Otherwise the data do not make sense: for example it is thermodynamically highly unfavorable for the background reaction to take place when toehold is 4 nt because it would result in an higher enthalpy state (i.e. less base pair formed). This should be clarified in the main text.
(2) It is mentioned multiple times in the manuscript that ATSE works because of a. the formation of a stable cyclic structure that involves bivalent interactions and b. higher local concentration of reactants or faster intramolecular interactions. I think this explanation is superficial and misleading to say the least. To me, it works because of one reason: entropy gain, i.e. the release of the output strand from the antibody-DNA complex. Consider the following thought experiment: an "antibody" that has one additional binding pocket for "O" so that it would still be part of the complex and stay nearby after being displaced from "B". The strand displacement will never happen at appreciable level if BO forms more base pairs than BI because "O" can easily come back to kick "I" off "B". Am I wrong?
(3) I strongly suggest replacing Figure 2a with Figure S7, which shows the formation of Ab(I2) and Ab(BO)2. Otherwise readers can be left wondering why those complexes are not included in the model (but in fact they were).
(4) It would be nice to show a zoom-in view of the first ~30 minutes of Figure 2d. As presented it is hard to see the level of "O" (green curve).
(5) Can authors provide explanations as to why the signal-to-background ratio in Figure 3b (OR gate, roughly 3.5:1) and Figure 4b (enzyme activity control, roughly 3:1) are much worse than what's shown in Figure 1 (at least 20:1, as high as 100:1)?
(6) I find the sentence that starts at line 347 and ends at line 349 very hard to understand. Exactly what is considered 0 and 1 fluorescent unit? The explanation provided in SI Section 2a seems much clearer to me. My personal preference of course does not matter but I think breaking the sentence in question into two, one explaining negative control (0) and another explaining positive control (1), may help reader like me.
(7) Typo: line 134, "lenght" should be "length". Dear Dr Pastore, Thank you for sending us the reviewers' assessment of our manuscript. We are pleased that all reviewers appreciated our work and have used their feedback to further improve our manuscript. Please find a point-by-point response below.
In figure 1 the obtained fluorescence values were normalized to controls that either lack BO or are spiked with 5.5 nM of free output. N.u. stands for normalized unit, which corresponds to the fluorescence that is generated when 1 nM output displaces the reporter duplex. We have added the following sentence to the legend of Figure 1 for clarification "One normalized unit (n.u.) represents the fluorescence generated by the displacement of Rep by 1 nM O." [2] The explanation about why the background reaction if fig 2 is drawn as irreversible (whereas it is described in the text as thermodynamically infavorable) is given only in SI S19 point 5. This is a bit late.
In principle the reaction is reversible, but since any product O that is formed is quickly trapped by reaction with the reporter duplex it is effectively irreversible and modelled accordingly. As suggested, we have moved the explanation from the SI to the main text.
("The background reaction was assumed to be effectively irreversible as DNA strand O is sequestered by an excess of reporter complex F·Q.") E.g. in the experiment using a toehold of 3 nucleotides, both the ATSE reaction (which dominates in the first 60 min) and the background reaction (after the ATSE has reached completion) can be modeled accurately. To avoid any confusion, we now also mention in the main text that an antibody concentration of 2 nM was used. ("optimization of the ODE model to experimental data obtained in the presence of 2 nM anti-HA antibody ( Fig. 2c and Table S6).")

of the S/B plot cannot be guessed from the background and signal plots. A log scale would probably fix that.
We agree that using log scales makes it easier to interpret these plots and have adjusted Figures 2e-2h accordingly. The background reaction is indeed not dependent on antibody concentration. However, to allow easy comparison of 2e -2h we plotted all of them in a similar manner as heat maps. In this way it is easier to see how plots 2g and 2h are derived from plots 2e and 2f.
[5] The author do not really adress the limit of detection of their method. ( even if it is possibly not the most relevant point here, as the purpose is more to show the possibility to link antibody chemistry to DNA Strand displacement circuits). However, when they introduce downstream reporting strategy (with additional aplification modules introduced in fig4), the authors could compare the LOD with the direct circuit presented in the beginning of the paper. Since theranostics are used as a motivation, it would be interesting to discuss the relevance of this LOD to clinical applications. Reaching a low LOD was indeed not the prime motivation for this study. Nonetheless, we agree that it may be useful to provide an LOD of the direct circuit in order to compare our method with other analytical methods. We therefore determined the LOD for the ATSE reaction with direct detection using the reporter duplex as described in Figure 1, yielding an LOD of 23 pM. A figure showing the LOD determination has been added to the SI ( Figure S5), and we mention the LOD in the main text. ("The limit of detection (LOD) was determined to be 23 pM under these conditions (Fig. S5."). Since the circuits presented in figure 4 were not optimized for LOD, a comparison of the LODs of the downstream systems to that of the direct circuit is not very useful. E.g. the β-lactamase-BLIP system has a relatively large S/N because the enzyme is not fully inhibited in the absence of any trigger strand.
[6] The presentation of DNA encoded "molecular circuitry" (and associated references) could be better balanced and not overly focused on strand displacement cascades. A couple of reference to enzymatic approaches could help the uninformed reader to obtain an informative picture of the field, especially since enzymatic circuit typically provide the amplification effect that could be required here. This is also important since the paper later uses an elegant way to control enzymatic activity with DNA strands We agree that enzymatically controlled DNA circuits are another important class of molecular circuits. We now explicitly mention this in the introduction, including references to representative papers. ("DNA-based molecular circuits capable of adaptation, oscillations and bistability have been created using enzymes involved in DNA synthesis and degradation such as DNA polymerases, nickases and exonucleases. [7] While the rate acceleration of 20~100 described in tge paper is fine, one can only wonder if it could be pushed further, and how. The only design element that has been optimized is the length of the toehold, but how the total length of BO was selected is not explained. Is the current choice better than a slightly longer, or slightly shorter duplex? This may be relevant since the antibody imposes geometric constraints on the intramolecular reaction. We realize that we did not explicitly discuss our rationale for choosing a duplex length of 24 nucleotides. Antibodies can accommodate different lengths of the duplexes due to a certain flexibility in the hinge-region of the antibody. In our previous work we found that bivalent peptide-dsDNA constructs with duplex lengths of 20 and 35 nucleotides bind equally well and both form tight 1:1 complexes, whereas linkers of 50 nucleotides also form 2:2 complexes. We have added a short discussion of this important design consideration to the beginning of the Results section where we introduce our experimental system, including references to our previous work. ("The length of the strands was chosen such that the bivalent peptide-dsDNA product BI contained 24 base pairs, a length that we have previously shown to be optimal for bivalent binding to the target antibody. 33,34 ").

Reviewer 2
General: ATSE is not as generic as the authors claimed in this manuscript. The intracellular application (e.g. smart drug delivery and even the rewiring of natural signalling pathways in vivo based on DNA circuits) of this antibody-based approach is very likely impossible. The conjugation and purification of DNA-peptide complex is quite laborious. Peptides without cysteine residue can not be conjugated to DNA-NH2 according to the protocol in Supplementary Information. Furthermore, after conjugation with DNA, binding affinity of peptides to antibodies may be dramatically disturbed.
We agree that intracellular applications of ATSE are at present nearly impossible, but we also did not make any claims in our manuscript in this direction. While the preparation of peptide-DNA conjugates indeed requires careful HPLC purification, conjugation efficiencies in our hands are 80-100%. In addition, commercial suppliers have started to offer peptide-DNA conjugates (BioSynthesis, LifeTein). The remark that peptides without a cysteine cannot be conjugated is peculiar, since in those cases one can add a cysteine at the N-or C-terminus of the peptide, as we did for both the HA and HIV-p17 epitopes. In fact it is more challenging when the original peptide epitope contains one or more native cysteines, e.g. in disulfide-containing cyclic peptides. In those cases, alternative bioconjugation methods are available, however, such as strain-promoted click chemistry. While conjugation of the peptide to DNA can affect the binding affinity, we have verified that this is not the case for the two examples shown in this work (see below).
[1] Authors determined dissociation constant (Kd) for fluorescein labeled HA binding to the antibody ( Figure S8, P4, line80)  The fluorescently labeled HA peptide was used for determination of the K d , because we also used this probe to determine kinetic parameters using stopped-flow fluorescence anisotropy. The latter experiments could not be done using fluorescently labeled peptide-DNA conjugates, because the increase in fluorescence anisotropy was too small to reliably detect using our stopped-flow machine. However, we agree with the referee that conjugation to DNA could affect the binding affinity of the peptide to the antibody. We therefore carefully repeated the antibody binding assays with a fluorescently labeled peptide-DNA conjugate. Similar K d values of 0.27 and 0.78 nM were obtained for the fluorescently labeled peptide and fluorescently labeled peptide-DNA conjugates, respectively. These additional titration experiments have been added to the supporting information ( Figure S9).
[2] This question refers to Question 1. Authors extended ASTE approach to anti-HIV1-p17 antibody. The S/B ratio for anti-HIV1-p17 antibody was ~20, compared to ~100 for anti-HA antibody (Figure 1d) DNA-peptide binding to antibody. Moreover, it will be great to include extra data of a DNA-peptide/antibody system with a Kd between HA and HIV1-p17 systems.
As suggested by the referee, we also repeated these binding experiments both for the peptide epitope and the peptide-DNA conjugates. Again, similar K d values of 21 and 16 nM were obtained for the peptide and peptide-DNA conjugates, respectively. These additional titration experiments have been added to the supporting information ( Figure  S10). We have adjusted the K d values mentioned in the text to the values obtained for the peptide-DNA conjugates.
[3] Since the background reaction only starts to contribute significantly to the observed fluorescence after 60min, when the ATSE reaction has reached completion (P7, line 145), why the analysis of Figure 2e and 2f was based on data at 180 min not 60 min? Is there a time dependency of the estimation of optimal initial conditions (Figure 2g and 2h)?
The ATSE reaction modelled in Fig2d indeed reaches completion after 60 minutes, but this was for a relatively high antibody concentration and an optimal toehold length. We chose to show the data after 180 min because the ATSE will be slower at lower concentrations of antibody and/or BO and I (modelled in Fig2e and f). The time point of 180 min was chosen to best accommodate these diverse reaction conditions. In fact, time is one of the parameters that determine the optimal signal-to-background and similar plots could be generated showing the various parameters as a function of e.g. time and BO/I concentration for a fixed antibody concentration, or time and antibody concentration for a fixed concentration of BO/I.
[4] This question refers to Question 3. In Figure 3, the fluorescence levels of those DNA circuits was measured after 3 h. How about the results after 1 h? Since the AND and NAND gates are based on two sequential toehold-mediated strand displacement reactions (three if you include the ATSE reaction itself) the rates of these reactions are slower than the rates of the single toehold-mediated strand exchange reactions (the OR and NOR gates) and require 3h to reach completion. To keep the analysis of the logic circuits consistent, we report the fluorescence intensities of all gates after 180 minutes.
[5] Authors claimed that "activity levels as estimated from Fig 4a are at least 20-fold higher after 1 h in cubation in the presebnce of antibody compared to the background reaction". How did they get this conclusion?
The activity levels in absence and presence of antibody were determined by deriving the slope of the emission ratio at 1 hour incubation (i.e. the change in emission ratio as a function of time). This slope is a measure for the activity of the DNAzyme. We have included a figure in the supporting information ( Figure S6) showing our data analysis and have included a reference to this figure in the main text.
[6] To make this manuscript easier to read, it is better to point out which system  is employed in text, Figure 2 and relevant supporting figures. We agree. We have adjusted the legends of figures 2 and 4 and all relevant figures in the SI accordingly. We have also added this information at relevant places in the main text.
[7] In Figure 4a, the pentagram of P1 and the spherosome of P2 should be switched. We thank the reviewer for catching this mistake, we have switched the pentagram and spherosome as suggested.