Development of A4 antibody for detection of neuraminidase I223R/H275Y-associated antiviral multidrug-resistant influenza virus

The emergence and spread of antiviral drug-resistant viruses have been a worldwide challenge and a great concern for patient care. We report A4 antibody specifically recognizing and binding to the mutant I223R/H275Y neuraminidase and prove the applicability of A4 antibody for direct detection of antiviral multidrug-resistant viruses in various sensing platforms, including naked-eye detection, surface-enhanced Raman scattering-based immunoassay, and lateral flow system. The development of the A4 antibody enables fast, simple, and reliable point-of-care assays of antiviral multidrug-resistant influenza viruses. In addition to current influenza virus infection testing methods that do not provide information on the antiviral drug-resistance of the virus, diagnostic tests for antiviral multidrug-resistant viruses will improve clinical judgment in the treatment of influenza virus infections, avoid the unnecessary prescription of ineffective drugs, and improve current therapies.

• Affinities measured by ELISA are used for a rough estimation. Since the authors measured the affinity of the antibody by surface plasmon resonance, which is the state-of-art technique for such, the ELISA measurements should be kept as supplemental information or omitted since it does not add anything to the paper. Moreover, the methodology for measuring the affinity by ELISA was not described in the material and methods section.
• There is no methodology for docking described in the paper and also there is no methodology for free energy calculation described in the paper. Please clarify this.
• Different from protein structure modelling and despite its evolution in the past years, proteinprotein docking methods are still only valid after experimental validation. Given two structures, a docking algorithm will always find many docking conformations even if the two proteins don't actually bind. Without experimental validation of the docking structure, it's impossible to assess its quality and validity. Without further evidence there is no way to substantiate the claims made. Moreover, all the features pointed out for increased binding affinity consider hypothetical enthalpic changes, when in fact it could be entropic contributions.
HA, which commonly renders the virus resistant to antiviral drugs, including oseltamivir. Furthermore, the authors describe the incorporation of this antibody within a lateral flow immunoassay. Even though the flow of experiments is logical, I do not believe that the manuscript at its current stage justifies the conclusions. Most importantly, all tests are performed with recombinant proteins or virus. In order to lay claim that this platform, in particular in combination with the LFI, would have any impact on clinical management of influenza virus infection, it would be absolutely essential that the authors evaluate performance of their assay directly in patient samples. Many diagnostic platforms have ultimately failed and a rigourous assessment using nasal swabs or broncheoalveolar lavage samples would be needed to assess the performance of this. Furthermore, the authors do not discuss the relevance of other drug resistance-associated variants in Influenza virus subtypes. Even though I223R/H275Y is an important resistence-associated variant, there are others playing an important role in other influenza virus subtypes. As far as I understand from the manuscript, the here-described antibody is specific to the H1N1 strain, even though it is not clear whether the authors tested other influenza virus subtypes. This would need to be included to ascertain the diagnostic validity of the assay.

Reviewer #3: None
We appreciate the reviewers for valuable comments to improve our manuscript. The changes in the manuscript and the answers to the reviewers' comments are as follows:

Reviewer comments
In this work, the authors develop a new monoclonal antibody capable of specifically recognizing an influenza neuraminidase double-mutant which confers drug resistance. The goal stated by them is to create a point-of-care fast diagnosis method for drug-resistant strains. This is a real problem identified by the authors and they go in the direction of solving it. This is a very respectable work but, despite the paper's merits, it cannot be framed as a protein engineering paper neither as diagnostic development paper since it lacks in both areas.
I find that there are major issued that have to be addressed to substantiate their claims before it can be accepted in a publication such as Nature Communications.

Question 1
They show the antibody recognizing the double-mutant protein and not recognizing the wildtype. However, they never use single-mutants in their study which also confers drug resistance and is more prevalent them the double mutants. It's paramount to test the single mutants to show the diagnostic value of the antibody. Answer) Following the reviewer's suggestion, we tested the detection of single-mutant influenza virus using A4 antibody. Because H275Y mutation is the most frequently observed drug-resistant mutation, [1] we examined the diagnostic ability of A4 antibody for pH1N1/H275Y mutant virus. The pH1N1/H275Y mutant virus (H275Y mutation A/Korea2785/2009 pdm: NCCP 42017) was obtained from the National Culture Collection for Pathogens (NCCP) operated by the Korea National Institute of Health (KNH). Figure R1a shows binding activity of purified A4 antibody to H275Y NA by competition ELISA. The A4 antibody bound to H275Y NA in a concentration-dependent manner with K d of 0.12 µM. Figure R1b displays the interaction between A4 antibody and pH1N1/H275Y mutant virus (10 7 PFU/mL) by dot-blot analysis. A4 antibody was applied to pH1N1/H275Y mutant virus and HRP-conjugated anti-human IgG Fc was applied for detection. For the comparison, I223Y/H275Y pH1N1 (10 7 PFU/mL) and wt NA were also examined. As shown 2 in Figure R1b, the dot was observable only from the double-mutant virus. This suggests the low affinity of A4 antibody to the single-mutant influenza virus. Figure R1. (A) Binding activity of purified A4 antibody to H275Y NA by competition ELISA. (B) Interaction of A4 antibody to I223Y/H275Y pH1N1 (10 7 PFU/mL), H275Y pH1N1 (10 7 PFU/mL), and wt NA (0.5 mg/mL) by dot-blot analysis.
We also applied A4 antibody for the detection of pH1N1/H275Y mutant virus by using colorimetry, SERS, and LFA. Figure R2a is absorption spectra of A4-Au NPs in the presence of pH1N1/H275Y mutant virus. The presence of single-mutant virus in the A4-Au NP solutions caused little change in absorption spectra shift. The SERS-based immunoassay result for H275Y pH1N1 (10 6 PFU) also shows negative signals ( Figure R2b). Lastly, the micrograph of the LFA after detection of single-mutant virus (10 7 PFU) exhibits no test line.
Taken together, we concluded that the A4 antibody can recognize the I223R/H275Y pH1N1 virus specifically. Although we reported the detection of multidrug-resistant virus in this manuscript, the identification of single-mutant influenza virus is also important as mentioned by the reviewer.
To achieve this goal, we had been developed the methods for pH1N1/H275Y mutant virus. [2][3][4] Moreover, we will report the other state-of-art results for the accurate identification of drugresistant virus soon.  "Additionally, we tested the recognition of single-mutant influenza NA protein (H275Y NA) using A4 antibody. H275Y mutation is the most frequently observed drug-resistant mutation. 10 The A4 antibody bound to H275Y NA in a concentration-dependent manner with K d of 0.12 µM ( Figure S4A). Figure S4B displays the interaction between A4 antibody and pH1N1/H275Y mutant virus (10 7 PFU/mL) by dot-blot analysis. For the comparison, I223R/H275Y pH1N1 (10 7 PFU/mL) and wt NA were also examined. As shown in Figure   S4B, the dot was observable only from the double-mutant virus. This suggests the low affinity of A4 antibody to the single-mutant influenza virus." (Line 20, Page 17).

4
"The presence of wt pH1N1 virus and H275Y pH1N1 virus in the A4-Au NP reaction solutions caused little change in color or absorption spectral shift ( Figure 4A, S6C, S7A).".
"For the detection of influenza viruses, the immune substrates were reacted with I223R/H275Y pH1N1, wt pH1N1, or H275Y pH1N1, and then the immunoprobes were reacted ( Figure S8). Figure 5a is the SERS-based immunoassay results for I223R/H275Y pH1N1 (blue spectrum) and wt pH1N1 (black spectrum). The number of both viruses is 1,500 PFU. The SERS-based immunoassay result for H275Y pH1N1 was shown in Figure S7B.
When the sample includes I223R/H275Y mutant virus, Au NPs on a nanoplate (NPs-on-plate) structures can be constructed through the immunoreaction of A4-I223R/H275Y pH1N1-HA.
This NPs-on-plate architectures can provide significantly enhanced SERS signals. In contrast, very weak SERS signals were obtained when the sample has wt pH1N1 or H275Y pH1N1 because A4 does not bind to the wt influenza virus or single-mutant virus." (Line 1, Page 22).
"On the other hand, wt pH1N1 virus and H275Y pH1N1 virus are not able to interact with A4-Au NPs; thus, a signal from the test line is not observed. Figure 6A, S7C is the micrographs of the LFAs after detection of wt, double-mutant, and signle-mutant viruses.
When the I223R/H275Y pH1N1 virus samples were applied, the red test lines were observed clearly. Importantly, even in the cases of high concentrations of wt pH1N1 virus (10 6 PFU) and H275Y virus (10 6 PFU), only the control line was observed in the absence of the I223R/H275Y pH1N1 virus." (Line 9, Page 23).

Question 2
They frame the paper as a diagnostic development work, but all the tests are done against recombinant protein or laboratory produced viruses. It's paramount that their very interesting tools are tested on real human samples where it should be compared to traditional diagnostic methods such as DNA sequencing and/or qPCR to prove its ability to accurately differentiate the virus in real samples. Sensitivity and specificity numbers should be calculated after these tests to show the validity of the methods.   This verified that the A4 antibody can recognize the I223R/H275Y pH1N1 virus specifically in real sample. The sensitivity and specificity of A4-based LFA developed for rapid antiviral multidrug-resistant influenza virus diagnostic tests are 100% (14/14) and 100% (14/14), respectively." (Line 11, Page 24).

Question 3-1
They show computational studies for docking of the antibody-antigen complex without providing any methodological detail. Free energy number is discussed but it's never explained how does the numbers were obtained. Different from protein structure modeling and despite its evolution in the past years, protein-protein docking methods are still only valid after experimental validation. Given two structures, a docking algorithm will always find many docking conformations even if the two proteins don't actually bind. Without experimental validation of the docking structure, it's impossible to assess its quality and 7 validity. Without further evidence, there is no way to substantiate the claims made. Moreover, all the features pointed out for increased binding affinity consider hypothetical enthalpic changes, when in fact it could be entropic contributions.

Answer)
With respect to the docking simulations of the epitope in the CDR of A4, a total of 20 conformations of the epitope were generated with the genetic algorithm. Among these putative binding conformations, clustered together had similar binding modes differing by less than 1.5 Å in positional root-mean-square deviation. The lowest-energy configuration in the top-ranked cluster was selected as the final structural models for the antibody-epitope complexes. To explain these, we have added a paragraph in the revised manuscript as follows.
"Docking simulations of wt and I223R/H275Y NA in the CDR of A4 3D structure of A4 obtained in the precedent homology modeling served as the receptor model in docking simulations with wt and I223R/H275Y mutant NA. The epitope structures were extracted from the X-ray crystal structure of (PDB entry: 4B7R). 37 Docking simulations were carried out using the AutoDock program to estimate the binding free energy (ΔG bind ) of the epitope in the complementarity-determining region (CDR) of A4, which can be expressed mathematically as follows. 38 The weighting parameters for van der Waals contacts (W vdW ), hydrogen bonds (W hbond ), electrostatic interactions (W elec ), entropic penalty (W tor ), and ligand dehydration free energy (W sol ) were set to 0.1485, 0.0656, 0.1146, 0.3113, and 0.1711, respectively, as in the original AutoDock program. r ij stands for the interatomic distance, and A ij , B ij , C ij , and D ij are associated with the well depth and the equilibrium distance in the potential energy function.
The hydrogen bond term has the additional weighting factor (E(t)) to describe the angledependent directionality. To compute the electrostatic interaction energy between A4 antibody and the epitopes, we used the sigmoidal function with respect to r ij proposed by Mehler et al. as the distance-dependent dielectric constant. 39 In the entropic penalty term, N tor indicates the number of rotatable bonds in the epitope. In the hydration free energy term, S i and V i denote the atomic solvation energy per unit volume and the fragmental atomic volume, respectively, while Occ i max represents the maximum occupancy of each atom in the epitope. 40 8 All the energy parameters in Eq. (1) were extracted from the original AutoDock program to derive the binding modes of wt and I223R/H275Y mutant NA in the CDR of A4.
Among 20 conformations generated with the genetic algorithm, those clustered together had similar binding modes differing by less than 1.5 Å in positional root-meansquare deviation. The lowest-energy configuration in the top-ranked cluster was selected as the final structural models for antigen-antibody complexes." (Line 22, Page 11).

Question 3-2
They show computational studies for docking of the antibody-antigen complex without providing any methodological detail. Free energy number is discussed but it's never explained how does the numbers were obtained. Different from protein structure modeling and despite its evolution in the past years, protein-protein docking methods are still only valid after experimental validation. Given two structures, a docking algorithm will always find many docking conformations even if the two proteins don't actually bind.
Without experimental validation of the docking structure, it's impossible to assess its quality and validity. Without further evidence, there is no way to substantiate the claims made. Moreover, all the features pointed out for increased binding affinity consider hypothetical enthalpic changes, when in fact it could be entropic contributions.

Answer)
We agreed that the binding modes derived from docking simulations had to be validated with experimental approaches. Therefore, we carried out the mutational analysis at positions His94 in the light chain and Trp33 in the heavy chain in order to assess the importance of the hydrophobic interactions to stabilize the epitopes in the CDR of A4. These mutant A4 antibodies were purified with the same method as A4 antibody. Figures   with the aromatic residues in CDR seems to be a determinant for selective binding to A4 antibody." (Line 7, Page 19).

Question 3-3
They show computational studies for docking of the antibody-antigen complex without providing any methodological detail. Free energy number is discussed but it's never explained how does the numbers were obtained. Different from protein structure modeling and despite its evolution in the past years, protein-protein docking methods are still only valid after experimental validation. Given two structures, a docking algorithm will always find many docking conformations even if the two proteins don't actually bind. Without experimental validation of the docking structure, it's impossible to assess its quality and validity. Without further evidence, there is no way to substantiate the claims made. Moreover, all the features pointed out for increased binding affinity consider hypothetical enthalpic changes, when in fact it could be entropic contributions.
Answer) The binding free energy function used in this work included not only the enthalpic term but also the entropic term that is proportional to the number of rotatable bonds in the epitope. To place an emphasis on this point, we added a sentence in the revised manuscript as "In the entropic penalty term, N tor indicates the number of rotatable bonds in the epitope. In the hydration free energy term, S i and V i denote the atomic solvation energy per unit volume and the fragmental atomic volume, respectively, while Occ i max represents the maximum occupancy of each atom in the epitope. 40 " (Line 13, Page 12).

Question 4
There are no considerations or references on phage display library specifications. This is very important for an antibody display paper.

Answer)
Previously constructed large naïve human Fab phage display library (3 × 10 10 ) in Korea Research Institute of Bioscience and Biotechnology was used for antibody screening.

11
We added the description in the revised manuscript as "For antibody screening, previously constructed large naïve human antigen-binding fragment (Fab) phage display library (3 × 10 10 ) in Korea Research Institute of Bioscience and Biotechnology was used. 31 " (Line 21, Page 8).

Additional reviewer comments
Those are the major points that I firmly believe should be addressed to make this paper, that has a lot of potentials, a very important piece of work that will push the current state-of-theart on influenza diagnosis. Other minor points are described below.

Question 5
Affinities measured by ELISA are used for a rough estimation. Since the authors measured the affinity of the antibody by surface plasmon resonance, which is the state-of-art technique for such, the ELISA measurements should be kept as supplemental information or omitted since it does not add anything to the paper. Moreover, the methodology for measuring the affinity by ELISA was not described in the material and methods section. were pre-incubated at 37 °C for 1 -2 h. The mixture was then added to each well previously coated with 100 ng of NA. Anti-human Fc-HRP (Thermo, 1:10,000 v/v) was added to the wells. All incubations were carried out at 37 for 1 h. Color was developed with OptEIA TMB Substrate (BD), and the absorbance was measured at 450 nm in a microtiter plate reader.
Affinity was determined as the antigen concentration required to inhibit 50% of binding activity and K d value was calculated from a Klotz plot.
We modified the manuscript as "Microtiter wells were coated with the purified NA (100 ng) in 50 mM sodium carbonate buffer (pH 9.6) at 4 overnight, blocked with BSA (2%) in PBS, and washed with PBST. A reaction mixture containing purified antibody (10 nM) and various concentrations (10 −11 -10 −5 M) of NA as a competing antigen were pre-incubated at 37 °C for 1 -2 h. The mixture was then added to each well previously coated with 100 ng of NA. HRPconjugated goat anti-human IgG (Pierce) was used for the detection of bound IgG. Color was developed with the 3,3′,5,5′-Tetramethylbenzidine substrate reagent set (BD Biosciences), and the absorbance at 450 nm was measured using a microtiter plate reader (Emax; Molecular 12 Devices). Affinity was determined as the antigen concentration required to inhibit 50% of binding activity and binding affinity (K d ) value was calculated from a Klotz plot." (Line 14, Page 10).

Question 6
There is no methodology for docking described in the paper and also there is no methodology for free energy calculation described in the paper. Please clarify this.
Answer) Please, refer the answer of Question 3-1.

Question 7
Different from protein structure modelling and despite its evolution in the past years, proteinprotein docking methods are still only valid after experimental validation. Given two structures, a docking algorithm will always find many docking conformations even if the two proteins don't actually bind. Without experimental validation of the docking structure, it's impossible to assess its quality and validity. Without further evidence there is no way to substantiate the claims made. Moreover, all the features pointed out for increased binding affinity consider hypothetical enthalpic changes, when in fact it could be entropic contributions.

Question 8
Line 34: It abbreviates neuraminidase to "NA" which has never been used in the text and so far.

Answer)
We changed the NA to neuraminidase in the revised manuscript.

Answer)
We changed the Delbecco to Dulbecco in the revised manuscript.

Question 10
Line 130: why did you used insect expression system. Instead of mammalian expression system?
Answer) Previously, soluble NA protein from the 1918 H1N1 (A/Brevig Mission/1/1918) strain was successfully expressed using a baculovirus expression system and crystalized for structural analysis.
Line 138: the centrifugation for 1h at 16,000g was to precipitate the protein or cell debris? I assume the protein was in the supernatant and that was used for affinity chromatography purification, but the text suggests that the "protein was protein was obtained by centrifugation" which implies that the protein was pelleted.

Answer)
We agree with the reviewer's comment. The corresponding sentence in the manuscript has been changed as "After the cell lysate was sonicated to reduce its viscosity, the cell debris was removed by centrifugation for 1 h at 16,000 g. The soluble protein from the cell supernatant was applied to Ni-Nitrilotriacetic acid agarose resin (Qiagen), washed, and eluted with buffer (50 mM Tris-HCl, 0.5 M NaCl, 0.5 M imidazole, pH 8.0)." (Line 13, Page 8).

Question 12
Line 146: where the antibody library comes from? Is there a reference for this library? Is it human? Is it scFv os Fab format? What is the promoter? What is the helper phage used?

Answer) Previously constructed large naïve human Fab library (3 × 10 10 ) in Korea Research
Institute of Bioscience and Biotechnology was used for antibody screening.

Question 13
Line 147: How many washes were performed in each round? It's very important for the authors to give details of the method used to obtain the antibody so others can successfully follow. round.
"Four rounds of panning were conducted, and the stringency of selection was increased with each round by gradually increasing the number of washes from 10 to 40." (Line 6, Page 9).

Question 14
Line 156: How was the soluble Fab expressed? Was it fused to g3p? Is there an amber stop codon between the Fab gene and g3p so one can express soluble Fab when using a nonsuppressor strain?

Answer) Soluble Fab expression was induced in E. coli TG1 cells at 30°C overnight by
adding isopropyl β-D-1-thiogalactopyranoside to a final concentration of 1 mM.
We added the description in the revised manuscript as "To screen individual clones for specific binding to I223R/H275Y NA, 500 colonies were randomly selected from the output plate after the third or fourth round of panning, cultured in Superbroth medium containing 100 μg/mL ampicillin until optical density of 0.5, and induced for Fab expression in Escherichia coli TG1 cells at 30 °C overnight by adding isopropyl β-D-1thiogalactopyranoside to a final concentration of 1 mM." (Line 9, Page 9).

Question 15
Line 157: how was the Fab detected? Is there a tag so one can use a labeled anti-tag antibody?
Please give more details.

Answer)
A micortiter plate was coated with 100 ng of I223R/H275Y NA in coating butter (0.5 M carbonate buffer, pH 9.6) and incubated at 4 °C overnight. After blocking, Goat F(ab')2 Anti-Human IgG (Fab')2-HRP (Abcam) antibody was used for the colorimetric detection of bound clones using the tetramethylbenzimidine substrate.
We added the description in the revised manuscript as "In detail, a microtiter plate was coated with 100 ng of I223R/H275Y NA in coating buffer (0.05 M carbonate buffer, pH 9.6) and

Question 2
Furthermore, the authors do not discuss the relevance of other drug resistance-associated variants in Influenza virus subtypes. Even though I223R/H275Y is an important resistanceassociated variant, there are others playing an important role in other influenza virus subtypes.
As far as I understand from the manuscript, the here-described antibody is specific to the H1N1 strain, even though it is not clear whether the authors tested other influenza virus subtypes. This would need to be included to ascertain the diagnostic validity of the assay.

Answer)
We tested the present mutant virus sensing methods by using four kinds of influenza virus subtypes. The reverse genetics system was used to generate different subtypes of the I223R/H275Y influenza virus. The expression plasmids for the eight-plasmid reverse genetic  1 9 3 4 (H 1 N 1 )

CL TL
A / B ri s b a n e / 1 0 / 2 0 0 7 (H 3 N 2 ) A / c a n in e / I welcome the possibility of reviewing the paper for the 2nd time. In the first version the authors developed a new antibody aiming to improve Influenza diagnosis, however, many things remained to be clarified.
In this 2nd revision, the authors have addressed the major concerns I had previously: -They have tested the antibody against the single mutant; -They tested the antibody against donor samples; -They provided some experimental validation on the antibody-antigen binding mechanism; -The methodology section has been significantly improved, providing details regarding the phage display methodology and the docking methodology.
I believe the paper has scientific merit and recommend it to be published at Nature Communications.

Best wishes, André
Reviewer #4: Remarks to the Author: Review of Author's response to reviewer #2 questions. Reviewer #2 Question #1. Reviewer#2 raised a very valid concern on the application of the LFA in a clinical setting without vigorous validation of the assay performance using real clinical material. The authors responded by including a small study (n=14) using NP/Op swabs spiked with 103 pfu pH1N1 viruses with I223R/H275Y markers. A few issues remain: 1). Sensitivity is often an issue for LFA assays. the authors did not demonstrate a dose effect with the detection of the I223R/H275Y in NP/OP swabs. 2). Matrix effect from different NP/OP swabs/broncheoalveolar lavage samples may also impact the sensitivity of the LFA detection, a larger number of the actual clinical samples collected from different patients are needed in order to evaluate the performance of this assay. 3). A more vigorous study using actual clinical specimen that are conducted side-by-side with sequencing are indeed needed to truly demonstrate the sensitivity and specificity of the assay and its true value in the actual clinical setting. Question 2 "Furthermore, the authors do not discuss the relevance of other drug resistance-associated variants in Influenza virus subtypes. Even though I223R/H275Y is an important resistanceassociated variant, there are others playing an important role in other influenza virus subtypes. As far as I understand from the manuscript, the here-described antibody is specific to the H1N1 strain, even though it is not clear whether the authors tested other influenza virus subtypes. This would need to be included to ascertain the diagnostic validity of the assay. " Reviewer #2 raised a good point. The authors did not sufficiently addressed this question. I223R/H275Y is a recognized marker for A(H1N1) oseltamivir/zanamivir resistance, but not necessarily for other subtypes. Furthermore, there are several other genetic markers that were identified to be associated for drug resistance, and not all drug resistance are based on neuraminidase. The authors should consider modifying the language in manuscript to address this limitation, including in the title to avoid overstatement of the study findings, for example the title can be modified to: "Development of novel A4 antibody for detection of neuraminidase I223R/H275Y associated antiviral multidrug-resistant influenza viruse" 1 We appreciate the reviewers for valuable comments to improve our manuscript. The changes in the manuscript and the answers to the reviewers' comments are as follows:

Reviewer comments
I welcome the possibility of reviewing the paper for the 2nd time. In the first version the authors developed a new antibody aiming to improve Influenza diagnosis, however, many things remained to be clarified.
In this 2nd revision, the authors have addressed the major concerns I had previously: -They have tested the antibody against the single mutant; -They tested the antibody against donor samples; -They provided some experimental validation on the antibody-antigen binding mechanism; -The methodology section has been significantly improved, providing details regarding the phage display methodology and the docking methodology.
I believe the paper has scientific merit and recommend it to be published at Nature Communications.
Best wishes, André Answer) Thank you for the recommendation of our manuscript in Nature Communications.
We appreciate for your valuable comments to improve our manuscript.

Reviewer comments
Reviewer#2 raised a very valid concern on the application of the LFA in a clinical setting without vigorous validation of the assay performance using real clinical material. The authors responded by including a small study (n=14) using NP/Op swabs spiked with 10 3 pfu pH1N1 viruses with I223R/H275Y markers. A few issues remain:

Question 1
Sensitivity is often an issue for LFA assays. the authors did not demonstrate a dose effect with the detection of the I223R/H275Y in NP/OP swabs.

Answer)
Following the reviewer's suggestion, the dose effect of I223R/H275Y detection in nasopharyngeal swab samples was demonstrated. Figure R1 shows the results of LFA-based   In addition to the previous data, an additional 26 nasopharyngeal swab samples were examined using the developed LFA, as the reviewer suggested. Figure R2 shows the results of a newly tested LFA after detecting the mutant virus in human nasopharyngeal swab samples. The test line was only observed in the presence of the I223/H275Y pH1N1 viruses.
In this study, a total of 40 nasopharyngeal swab samples were tested. According to the guideline of National Institute of Food and Drug Safety Evaluation of Korea, a minimum of 20 sample results are required for approval of an in vitro diagnostic system. Therefore, the current results are suitable for assessing the performance of A4 antibody-based LFA assay.
The figure was included in Supplementary Information and the manuscript was changed as "Totally, we tested 40 human nasopharyngeal swab samples ( Figure S12). The sensitivity and specificity of A4-based LFA developed for rapid antiviral multidrug-resistant influenza virus diagnostic tests are 100% (40/40) and 100% (40/40), respectively. According to the guideline of National Institute of Food and Drug Safety Evaluation of Korea, a minimum of 20 sample results are required for approval of an in vitro diagnostic system." (Line 21, Page 24).

Question 3
A more vigorous study using actual clinical specimen that are conducted side-by-side with sequencing are indeed needed to truly demonstrate the sensitivity and specificity of the assay and its true value in the actual clinical setting.

Answer)
Following the reviewer's suggestion, sequencing tests were performed and compared to the LFA assay results. The viral sequence was identical in all samples tested as shown in Figure R3. The sensitivity and specificity of A4-based LFA assay for antiviral multidrug-resistant influenza virus diagnosis is determined as 100% (26/26) and 100% (26/26), respectively ( Figure R2). (Previous Question) Furthermore, the authors do not discuss the relevance of other drug resistance-associated variants in Influenza virus subtypes. Even though I223R/H275Y is an important resistance associated variant, there are others playing an important role in other influenza virus subtypes. As far as I understand from the manuscript, the here-described antibody is specific to the H1N1 strain, even though it is not clear whether the authors tested other influenza virus subtypes. This would need to be included to ascertain the diagnostic validity of the assay.
Reviewer #2 raised a good point. The authors did not sufficiently addressed this question.
I223R/H275Y is a recognized marker for A(H1N1) oseltamivir/zanamivir resistance, but not necessarily for other subtypes. Furthermore, there are several other genetic markers that were identified to be associated for drug resistance, and not all drug resistance are based on neuraminidase. The authors should consider modifying the language in manuscript to address this limitation, including in the title to avoid overstatement of the study findings, for example the title can be modified to: "Development of novel A4 antibody for detection of neuraminidase I223R/H275Y associated antiviral multidrug-resistant influenza virus" Answer) The goal of this study is to find an antibody to the I223R/H275Y mutant virus, an antiviral multidrug-resistant virus for both zanamivir and oseltamivir. After careful evaluation of A4 antibody, we were successful in detecting the mutant virus in several ways. However, as the reviewer pointed out, there are certainly other genetic markers of drug resistance and not all drug resistance is based on NA. In agreement with the reviewer's suggestion, the title of the manuscript has been modified to limit the current approach to the detection of neuraminidase I223R/H275Y associated antiviral multidrug-resistant influenza virus only.