We would like to further reply to the response provided by Luo and Mignot on our initial comments regarding their publication in PLOS ONE (1). The “additional” data provided by Luo and Mignot in their response published on May 4, 2018 has provided further insights regarding their attempts to replicate the findings in our publication (2).

KEY INSIGHT: Authors’ chose NOT to disclose data in PLOS ONE (1) that confirmed our data (2):
If the “additional” data provided by Luo et al. in their comments from May 4, 2018 (please see https://figshare.com/s/3e...) had been included in the original Luo et al. publication (1), it would have immediately alerted the reader that Luo et al. actually confirmed using the same ProImmune assay our findings related to the difference in binding of the NP epitopes and the lack of binding of the HA epitope (2).
Luo et al. claim, “These data were not included in our [Luo et al.’s PLOS ONE] manuscript because we felt this was a detail and it was just too much to discuss.” However, by Luo et al. not including these confirmatory ProImmune data in their original manuscript (1), neither the PLOS ONE Reviewers nor the readers of the PLOS ONE publication by Luo et al. (1) were given the chance to assess the discrepancy between the results of their ProImmune experiments (that actually confirmed Ahmed et al.’s ProImmune experiments) and the different results that Luo et al. obtained using a DQB0602-EBV displacement assay that they performed in-house. This was the same DQB0602-EBV displacement assay that appeared in their retracted publication (3).

Luo et al. introduced a bias by concluding that the results of the ProImmune assay (repeated by them) had to be wrong because it did not demonstrate binding for what they consider “established” DQ0602 binders (see https://figshare.com/s/a8... in the Response Letter by Luo et al.). The advantage of using a commercial assay performed by a third party such as ProImmune (4) is that it avoids operator bias. The ProImmune REVEAL® assay has been cited in multiple publications (5). It is our opinion that Luo et al.’s critique of the ProImmune REVEAL®assay as “…the flawed REVEAL® assay” is unjustified. Their decision not to display their own results with the ProImmune REVEAL® assay in their original publication (1) is ironic. The data from their own experiments with the ProImmune REVEAL assay actually confirm our own ProImmune data in our original publication (2). We wonder why the authors disclosed these data only in their subsequent response provided on May 4, 2018 and not in the original publication (1).

The reader needs to be aware that this group’s previous publication in Science Translational Medicine was retracted (3) due to what some of the authors claimed was the inability to reproduce their in-house ELISPOT data. A third-party organization called “Retraction Watch” quickly identified that the ELISPOT images had been graphically manipulated (please see the web link provided in reference 6). In that web link (6), observers identified that the “spots” seen in one well of the ELISPOT plate were “copied”, “rotated”, and “pasted” in another well of the plate and this “process” was repeated again in multiple figures in that retracted publication (3).

Indeed, after our Comment sent to PLOS ONE on March 25, 2018 and published on that date in the Comments section of the Luo et al. article (1) PLOS ONE webpage, they have at last provided a more detailed response. In their published response of May 4, 2018, Luo and Mignot have finally disclosed “additional” data whose significance they have previously chosen to suppress. These “additional” data contradict the conclusions made by Luo et al. (1) and confirm the reproducibility of our findings (2). It seems that the authors chose to withhold critical information on their own experiments with the ProImmune REVEAL® assay that actually confirmed our results (2).
We now comment on the nine points in the Response Letter posted by Luo et al. on May 4, 2018 to help the reader and the Editors draw his/her own conclusions:

1. In our previous Comment Letter, we indicated that Luo et al. had misrepresented a key finding in our publication (2) by stating that, “… NP…and HCRTr2…did not appear to bind to DQ0602 [in the studies of Luo et al.] unlike what was reported by the Proimmune array by Ahmed et al...”.
a. This underlined statement from Luo et al. is what we are objecting to because it misleads the reader of the Luo et al. paper (1) to conclude that the Ahmed et al. paper (2) suggested that both NP and HCRTr2 were binding to HLA DQB1*06:02. The Ahmed et al. paper (2) demonstrated that only NP bound to HLA DQB1*06:02.
b. Luo et al. apologized in their Response Letter for this confusion and indicated that it stemmed from the difficulties they had with replicating the findings of our publication (2). In their response on May 4, 2018, they demonstrate for the very first time that they indeed performed the ProImmune REVEAL® assay (please see their ProImmune Report indicated in the link provided by Luo et al. at https://figshare.com/s/3e...). The data in their ProImmune Report was generated on July 20, 2015 but was not shown in the current paper (1) that was published 2 years later. On pages 13-14 of that report, one sees that the 15-mer influenza nucleoprotein (NP) peptides containing “isoleucine” in row #27 (RELILYDKEEIRRI) and row #30 (ILYDKEEIRRIWRQA) have ProImmune REVEAL® scores at 24 hours of 9.6 and 12.5, respectively. In contrast, the 15-mer NP peptides containing “methionine” in row #28 (MRELILYDKEEMRRI) and row #31 (ILYDKEEMRRIWRQA) have no binding to HLA DQB1*06:02 as indicated by ProImmune REVEAL® scores at 24 hours of 0.0 and 0.0, respectively. These findings of Luo et al. confirm the ProImmune results in the Ahmed et al. publication (2) in which the 21-mer NP peptides containing “isoleucine” (RELILYDKEEIRRIWRQANNG) demonstrated a ProImmune REVEAL® score at 24 hours of 16.9 compared to the 21-mer NP peptide containing “methionine” (RELILYDKEEMRRIWRQANNG) that demonstrated a REVEAL score at 24 hours of 0.5. Therefore, the effect of the “I” to “M” substitution on influenza NP binding to HLA DQB1*06:02 reported in our publication (2) has been replicated (and confirmed) by Luo et al.
c. Furthermore, on page 17 of the ProImmune report from Luo et al. (https://figshare.com/s/3e...), the 15-mer peptides representing the native HCRT receptor 2 (row #129 = FLNPTDYDDEEFLRY, row #130 = TDYDDEEFLRYLWRE, and row #131 = DEEFLRYLWREYLHP) demonstrate no binding to HLA DQB1*06:02 as indicated by the ProImmune REVEAL® scores at 24 hours of 0.0, 0.0, and 0.0, respectively. These results by Luo et al. are exactly the same as what we reported with the 21-mer native HCRT receptor 2 peptide LNPTDYDDEEFLRYLWREYLH that had a ProImmune REVEAL® score at 24 hours of 0.0 (2).
d. Finally, on page 17 of the ProImmune report from Luo et al. (https://figshare.com/s/3e...), the 15-mer peptide representing the influenza HA273-287 peptide from A(H1N1)pdm09 strain, A/California/07/2009 (row #136, AMERNAGSGIIISDT) from their current paper (1) and earlier retracted publication (3) demonstrates a ProImmune REVEAL® score at 24 hours of 0.0. This finding replicates our results for the 15-mer HA273-287 (AMERNAGSGIIISDT) and 21-mer HA271-291 (AFAMERNAGSGIIISDTPVHD) peptides that had ProImmune REVEAL® scores at 24 hours of 0.0 and 0.0, respectively (2). Therefore, Luo et al. has confirmed no binding of the HA fragment. In contrast to the conclusion of Luo et al. in their PLOS ONE publication (1), all of the data reported in Table S3 of our publication (2) could be replicated by Luo et al. using the ProImmune REVEAL® assay with the epitope likely to be binding HLA DQB1*06:02 reflecting the relevant sequence.


2. In the Response Letter by Luo et al., they claim that the retraction of their previous paper (3) was only due to flawed ELISPOT result. As we have clarified previously, this retracted paper was actually found to have ELISPOT images that have been graphically manipulated. The Retraction note (3), did not exonerate or comment on any other data in that paper (3), and did not mention, distinguish or exculpate the DQB0602-EBV displacement assay.

3. a. We brought to the attention of Luo et al. that they did not actually test in their PLOS ONE publication (1) the 21-mer peptide indicated in our publication (2) but instead a 15-mer of the nucleoprotein peptide derived from within the 21-mer peptide. Again, they have given the same problematic response, “We had in fact ordered these exact peptides but these were tested later…We did not include this data at the time the paper was accepted as it seemed like a detail.” We are concerned with their lack of attention to detail. Length matters in binding assays.

b. Regarding the robustness of the ProImmune REVEAL® assay used in our publication (2), we tested both positive and negative control peptides reported in the literature by other investigators independently testing peptide binding to HLA DQB1*06:02. We observed consistent results between the binding reported in the ProImmune REVEAL® assay and those publications (7 and 8). Some of these data were shown at the IABS meetings on narcolepsy on October 23, 2015 in Geneva, Switzerland and again on March 26, 2018 in Brussels, Belgium. Immediately after each of these meetings, we submitted our slide presentations to the IABS organizers as part of their general request to meeting presenters so have these data on record. In those meetings, we showed the ProImmune REVEAL® assay data that demonstrate that the positive control (very strong binding influenza A hemagglutinin peptide) RGYFKMRTGKSSIMRS showed a REVEAL® score at 24 hours of 194.9 which has been independently shown to have an IC50 = 369 nM (7) thus confirming the validity of the ProImmune Assay. The negative control (non-binding insulin A-2 peptide) showed a REVEAL® score at 24 hours of 0.0 which has also been independently shown to have an IC50 of greater than 100,000 nM (8). For the reader not familiar with IC50 measurements, small numbers (369 nM) translate into stronger binding while large numbers (100,000 nM) translates into weaker binding.

We have included in the Appendix of this Letter the ProImmune REVEAL® assay data (Figure 1) reported in our original publication (2) and also the ProImmune REVEAL® assay results (Figure 2) for the panel of “positive” and “negative” binders of HLA DQB1*06:02 reported in the literature (7 and 8). Luo et al. claim that they “screened the literature for known DQ0602 binders…found one CMV…and one EBV sequence that were clearly both binders.” Luo et al. also found an HSV2 peptide that “is also a good binder”. Luo et al. noted that in Table S3 of our publication (2), the ProImmune REVEAL® assay reported no binding of HA273-287 (AMERNAGSGIIISDT) nor homologous-peptide sequences from other strains. This was surprising to them as they had tested the HA69 (AMERNAGSGIIISDT) related peptides many times with their DQ0602-EBV displacement assay and found these to be strong binders.

However, regarding the influenza HA and HSV2 peptides that Luo et al. refer to above, we would like to again draw the reader’s attention to Figure 2 (provided by us in this Comment) that reports the ProImmune REVEAL® assay testing of the panel of “positive” and “negative” binders of HLA DQB1*06:02 reported in the literature (7 and 8). In this table, one sees that the 16-mer influenza HA peptide (RGYFKMRTGKSSIMRS) that would be expected to be a very strong binder to HLA DQB1*06:02 based on the published IC50 = 369 nM (7) does indeed correlate to a ProImmune REVEAL® assay score at 24 hours of 194 (very strong binder). Therefore, the ProImmune REVEAL® assay does detect influenza HA peptides as being strong binders (e.g., RGYFKMRTGKSSIMRS) but only when they actually bind to HLA DQB1*06:02. In contrast, the different HA peptide claimed to be a binder by Luo et al. (e.g., AMERNAGSGIIISDT) most likely is NOT a binder to HLA DQB1*06:02 based on the ProImmune REVEAL® assays demonstrating NO binding in our studies (2) nor those conducted by Luo et al. (https://figshare.com/s/3e...) that they disclosed only AFTER their publication (1).

In Figure 2, one will note that we also tested the overlapping peptides from HSV-2 (UL49105-190) that have published IC50 values ranging from 25,000 nM to 100,000 nM (8). What one does, indeed, see with the ProImmune REVEAL® assay are decreasing REVEAL® scores of 1.7, 1.4, 0.8, 0.6, 0.3 that would be expected with increasing IC50 values. We additionally tested the insulin B5-15 analogs containing single arginine substitutions and with measured IC50 values reported in the literature (8). Testing of these insulin B5-15 analog peptides using the ProImmune REVEAL® assay demonstrated that peptides with increasing IC50 from 3,000 nM, to 52,000 nM, to 100,000 nM correlated with decreasing REVEAL® scores at 24 hours of 1.7, 0.3, and 0.2, respectively (as would be expected). Finally, we tested the IA-2 (islet antigen-2 peptide361-371, LTLLQLLPKGA, with IC50 > 100,000 nM) and, as expected, the ProImmune REVEAL® assay yielded a REVEAL® score at 24 hours of 0.0. Therefore, these data show that the ProImmune REVEAL® assay is valid and highly reproducible.

c. With the “additional” data that Luo et al. have decided to only show after our Response Letter, the reader can easily see that the ProImmune REVEAL® assay conducted by them demonstrated no binding of the HA peptide identified in their retracted publication (3). This replicates our ProImmune studies (2) that similarly did not find any binding of HA peptides with DQB1*06:02 despite testing both 15-mer and 21-mer versions of the HA peptide.

d. Multiple points have been raised in this section that need to be clarified by us:
i) Luo et al. in response to our Response Letter now claim that the “CHO-HCRTR2 cell line was in fact a commercial cell line that stably expressed HCRTR2 and that we [Luo et al.] bought more than 10 years ago for potential radioreceptor binding assays (TM0508, GenScript, CHO-K1/OX2)”. However, this new statement is clearly contradicting what is written in their original paper (1) in the “Materials and methods” section on page 8 where they clearly indicate, “CHO-HCRTR2 cell line was made in a CHO-K1 host cell line (purchased from ATCC, https://www.atcc.org/prod...) [57, 58]”. A reader examining the “Materials and methods” section would have clearly understood that the source of the cell line was ATCC and that the CHO-HCRTR2 cell line was made in-house. This new claim of having used a commercial cell line purchased more than 10 years ago is difficult to believe because the source of those cells is GenScript, not ATCC as is clearly written in the PLOS ONE paper of Luo et al (1).

ii) In their Response letter, Luo et al. go one step further to provide an email where they asked us “for the exact cell line used in the Ahmed and Steinman article [2]…” and allege that they “…were told this was proprietary and not available”. The reader is kindly asked to review the link to the email contained in the response letter of Luo et al. (see https://figshare.com/s/95...) and needs to be made aware of the following background (well known to Luo et al.) that provides context for the email exchange between Guo Luo and Sohail Ahmed. In their “retracted” Science Translational Medicine paper from July 30, 2014 (3), the reader should note on Page 13 in the “Acknowledgments” section three key pieces of information: 1) one of the two industry funding sources to Emmanuel Mignot was GlaxoSmithKline (GSK) SPO #104642 (the other being Jazz Pharmaceuticals, SPO #108095), 2) “Competing interests” section clearly indicates that Emmanuel Mignot is a paid consultant for both Jazz Pharmaceuticals and GSK (the maker of the Pandemrix vaccine) and 3) Emmanuel Mignot is a co-inventor on a patent that has been filed to use the HCRT epitopes for diagnosis and on modification of the pHA1 2009 epitope in influenza vaccines, that intellectual property rights for narcolepsy diagnosis are owned by Stanford, and rights for making vaccine improvements are owned by GSK. This disclosure indicates that the group of Luo et al. and their industry funding partner, GSK (and maker of the Pandemrix vaccine), have a financial and commercial interest related to the patent claims for mimicry of the HA antigen with the hypocretin ligand. These patent claims should have been clearly noted in their PLOS ONE publication (1) but were NOT. The reader is asked to see the first page of the PLOS ONE publication (1) where under “Competing Interests”, Luo et al. have stated, “The authors have declared that no competing interests exists.”

The reader should also note on page 21 of the Luo et al. publication (1) that under the “Acknowledgments”, Luo et al. have stated, “We also thank …GSK and clinicians through Europe for providing Pandemrix® samples… . We appreciate their contributions for study design and data collection.” This “Acknowledgment” by Luo et al. in the PLOS ONE publication (1) would suggest a role for GSK in this publication, and we will explain why this is relevant:
The reader should be aware that Novartis Vaccines’ non-influenza franchise was sold to GSK (public announcement on April 2014) while the Novartis Vaccines’ influenza franchise was sold to another company, CSL Ltd (public announcement October 2014) and is currently handled by Seqirus. This separation of the Novartis Vaccines franchise into a “non-influenza” and “influenza” sale to different companies was clearly indicated by Ahmed et al. at the IABS narcolepsy meeting on October 23, 2015 in Geneva, Switzerland and again on March 26, 2018 in Brussels, Belgium so is well-known to those attending the meetings. The sale of these two parts of Novartis Vaccines to different buyers legally resulted in all data, supplies, products, and information generated by Novartis Vaccines in the “influenza franchise” and sold to CSL being kept strictly separate from the “non-influenza franchise” that was sold to GSK.

iii) Guo Luo (who we all now know works in the Mignot laboratory at Stanford) contacted Sohail Ahmed on June 8, 2016 by email without identifying himself and using an email address of disanpe@gmail.com instead of his Stanford email (please see these emails contained in the link provided in the Response letter of Luo et al at https://figshare.com/s/95...). Guo Luo was requesting the host line Chem-1 and Chem-1/HCRTR2/1, and Sohail Ahmed (who had left Novartis Vaccines and was working at Roche Pharmaceutical) logically requested that Guo Luo identify himself to better understand how he could help him in his request for material that were now the legal property of CSL Ltd (Seqirus). When Guo Luo indicated that he was working in the Mignot lab at Stanford (that had been collaborating closely with GSK), Sohail Ahmed correctly informed Guo Luo that the influenza franchise (and all related data) was acquired by another company, CSL. Due to the confidentiality agreements in place (and also respective intellectual property) between Novartis Vaccines, CSL, and GSK, Sohail Ahmed could not further help Guo Luo as this would be a breach of the separation agreements between CSL, GSK, and Novartis Vaccines related to the separate sales of the “influenza franchise” to CSL and the “non-influenza franchise” to GSK (Sohail Ahmed also cc’d co-author Lawrence Steinman in that email). Guo Luo replied on June 16 that he understood and thanked Sohail Ahmed. Guo Luo forwarded this email to his supervisor, Emmanuel Mignot who writes, “This is not good… Then they should not publish”. This description of the facts of the email exchange are in contrast to what Luo et al. allege in their Response Letter as “We asked Ahmed et al. for the exact cell line used in the Ahmed and Steinman article, but were told this was proprietary and not available…”. These allegations are refuted in the email stream.

iv) Luo et al. go on to indicate that “Having recently lost all NIH and industry grant funding following the retraction, we were not in a position to buy an extra cell line costing several thousand dollars, so we used this older cell line instead.” The reader is asked to make note of the information we have provided in point “d” (subsection “i”) above. Luo et al did not use an older cell line from GenScript. They are claiming this only now, after having read our initial Response Letter to PLOS ONE. Luo et al. clearly used an (in-house) CHO-HCRTR2 cell using a CHO-K1 host cell line purchased from a different company, ATCC as indicated in their PLOS ONE publication (1). In our original Response Letter, we indicated that the cell-line used in our publication (2) could still be purchased from Eurofins (as clearly indicated in our original publication (2) on page 11 under the section “Assays for antibody binding to human HCRT receptor” where it states, “…cloned cell lines made in the Chem-1 host expressing high levels of recombinant human HCRT receptor 1 or HCRT receptor 2 on the cell surface were purchased from Eurofins…”. The cell line that we purchased (HTS045C) cost 9,495 USD but if this was too expensive for Luo et al., they could have easily purchased the frozen cells containing HCRT receptor 2 (HTS045RTA) that cost 627.15 USD. The reader should also keep in mind that the ProImmune REVEAL® assay studies performed by Luo et al. (that they have only recently disclosed to us and have been discussed in the beginning of this Letter) typically cost more than 20,000 USD! Again, the ProImmune REVEAL® assay carries a high price tag in part because of the quality and fidelity of this platform for determining peptide binding to various HLA and is why it has been used in multiple prestigious publications in the literature (5)! Therefore, Luo et al. had two options for testing the same cell line expressing the HCRT receptor 2 that we used but chose neither. They instead chose to attempt transfection of different cell lines (in-house) and have generated data in their PLOS ONE publication (1) that severely challenge the fidelity of their in-house transfection procedure, their protein isolation procedure, the antibodies for detection, and the procedures for microscopic visualization of HCRT receptor 2 (HCRTr2).
e. Regarding the clarification that we requested about Luo et al. claiming to have tested the same sera as in the Ahmed et al. publication (2), Luo et al. replied that, “…we were unfortunately unable to test the exact same blood samples. It is my understanding that the critical post Focetria® blood samples are proprietary to Novartis.” While this would have been a sufficient response, Luo et al. go on to say, “Of note, it would appear that the difference in reactivity is mostly related to the Focetria® post vaccination samples, which we hypothesize may not have been handled the same way as other samples.” While the intention behind such a statement by Luo et al. in their Response Letter appears elusive, we can reassure the reader that the Ahmed et al. publication (2) involved twenty scientific and clinical expert co-authors spanning Novartis Vaccines (Italy), Novartis Institutes of Biomedical Research (USA), Novartis Pharma AG (Switzerland), National Institute for Health and Welfare (Finland), academic and university hospital of Siena (Italy), VisMederi Srl (Italy), and Stanford University. Experiments were performed by the respective scientific centers that were blinded to the identity of the samples to limit any bias and ensure equal handling of all samples.
4. In light of the ProImmune experiments conducted by Luo et al. almost two years ago but only recently disclosed by Luo et al. in their Response Letter, we believe that the reader now has the full picture regarding our concerns about the validity of the Luo et al. paper (1) in general. We agree with Luo et al.’s statement that “the NetMHCII 2.2 [used by Luo et al.] does not adequately predict in vitro DQ binding…”.
5. Luo et al.’s response to our concerns regarding the multiple bands above and below the expected molecular weight for HCRT receptor 2 in their Western blotting experiments is very worrisome, again. Luo et al. indicated, “Our goal was not to show that any particular commercially available antibody was 100% specific for HCRTR2, but rather to have a control sera known to recognize HCRTR2 as positive control for our assays. It really did not matter that much if it was 100% specific.” It is NOT clear why Luo et al. are referring to control sera in the quote above? The Western blot in the Luo et al. publication (1) was supposed to be detecting HCRTR2 using a commercially available antibody (primary anti-HCRTR2 antibody, Cat# WH0003062M1-100UG, Sigma-Aldrich as per the Luo et al. publication [1]) and has nothing to do with control sera.
The reader is requested to look at Fig. 4 in our original publication (2) to understand how we performed a Western blot for detecting another protein called nucleoprotein (NP). The reader will see an SDS-PAGE in Panel A that shows the multiple bands of protein that separate out when undergoing electrophoresis on a gel (2). An arrow on the right-hand side indicates “NP” and is indicating where this NP protein band should be located based on its molecular weight being close to near 51 kD (the left hand column titled “M” is indicating the loading of a molecular weight standard which contains proteins of known molecular weights that will generate a banding pattern from top to bottom to enable an investigator to compare with their experimental sample. In Panel B and C, we have used two different commercial antibodies specific for NP (InA245 and InA108, HyTest Ltd.) that clearly detected NP (2). We intentionally allowed the exposure to be slightly longer than necessary because lane # 16 and lane #17 appeared to be negative and we wanted to be sure were not possibly missing very low amounts of NP below the level of detection in lane #16 and #17. The faint bands that we saw in lane #16 and #17 were confirmed by mass spectrometry (2) to reflect trace amounts of NP protein that were at the lower limit of detection by Western blotting. This longer exposure is what can cause additional “background” bands to show up in experiments but would always be less intense than the band that is specific for the protein being targeted by the antibody. One can already see that a very strong specific band was detected by the anti-NP antibody in the same row all the way across in the Western blot. In our Supplementary Data (Fig. S9), we show the entire Western blot so the reader can see the specificity of the protein detection by the antibody and for maintaining transparency. However, Luo et al. (who claimed to be trying to replicate our results) did not use either of these two antibodies but a different one (primary anti-HCRTR2 antibody, Cat# WH0003062M1-100UG, Sigma-Aldrich [1]). The S4 Fig. in the Luo et al. publication (1) shows their Western blot with many bands detected above and below the arrow near 43 kd that is labeled HCRTR2. What is very concerning is that the exact opposite effect is being seen with their antibody. The protein of interest has a very faint band while all the background bands are very intense. Furthermore, the band indicated by the arrow is also seen in the control column (CHO-K1) that has not been transfected by HCRTR2. Our impression is that all of these bands are reflecting background and neither the lane with CHO-K1 nor CHO-HCRTR2 are showing antibody detection specific for HCRTR2. This is consistent with the microscopy performed by Luo et al. using yet another in-house transfected cell line (HEK293) that showed a non-specific “linear” reaction (1) instead of the punctate staining typical for a G-protein coupled receptor such as HCRTR2 (2).
Luo et al. point out in their Response Letter that they have used an anti-HCRTR2 antibody (Cat#WHOOO3062M1-100UG, Sigma Aldrich) that is monoclonal (1), and this monoclonal antibody was not used by Ahmed et al (2). This statement of theirs makes the results of their Western blotting (1) even more worrisome because a monoclonal antibody for HCRTR2 detection would have even higher specificity for HCRTR2 and even fewer “background bands”. However, Luo et al. dismiss this discrepancy in their Western blotting (1) by stating, “We do not blindly believe that commercially available products are always what they advertised…”. In their Response Letter, they provide a repeat experiment showing an entire gel (Panel A) to demonstrate equal amount of whole cell lysate protein in both lanes but then do NOT show the entire image of the Western blot with the same monoclonal antibody (that should correlate with the entire gel shown in Panel A). They again mention “other cross-reacting bands are still present in other areas of the gel” but “…(data not shown)” hoping that this is acceptable scientific methodology.
6. Regarding the arbitrary cutoffs assigned to the ELISA experiments of Luo et al. (1), Luo et al. indicate in their Response Letter that the cutoff they used in the PLOS ONE publication is based on Tanaka et al. publication (9) and, regarding the samples in which Tanaka et al. detected anti-HCRT receptor 2 antibodies, “…we concluded that these few positives are most likely false positive.” Luo et al. go on to state, “We indeed do not believe that any of the very few positives we observed (1) are real, since they did not replicate using the 3 techniques we used.” What is problematic with this statement is that it echoes the same biased approach that they used to dismiss the positive ProImmune REVEAL® assay results that they generated (and which confirmed our ProImmune REVEAL® assay results). Also, for Luo et al. to claim replication of antibodies to HCRT receptor 2 in Pandemrix-vaccinated patients, they required that the same sera demonstrate positivity in all THREE of their “in-house” assays. We have already described in detail, in our previous response letter, our concerns with some of these assays (e.g., the LACK of fidelity of their in-house transfection procedure, protein isolation procedures, antibody detection, and microscopy visualization of HCRT receptor 2).
7. Luo et al. dismiss the specificity of the work performed by Yamanaka et al. (10) because “it was impossible to exclude the possibility that the Hcrt gene promoter used…was somewhat leaky…”. As mentioned previously, Luo et al. dismiss the work of Tanaka et al. (9) stating, “…we concluded that these few positives are most likely false positive…”. Luo et al. dismiss their own data generated using the ProImmune REVEAL® assay because this widely used commercial assay cited in multiple prestigious scientific publications (5) was unable to replicate data generated by Luo et al. in-house with their EBV-bio binding assay (from their “retracted” publication [3]). But then Luo et al. state, “We believe a scientific finding must be consistent across multiple techniques and laboratories, otherwise it should be viewed with suspicion until replication…”. Therefore, why does not Luo et al. accept their ProImmune REVEAL® assay replication of our ProImmune REVEAL® assay results (2); the similar findings among their studies (1), Tanaka’s studies (9), and ours (2) [regarding detection of HCRT receptor 2 antibodies]; and the physiological experiments of Yamanaka (confirming the presence of an HCRT receptor 2 autoreceptor)? These investigations report a scientific finding that is consistent across multiple techniques and laboratories. Instead, Luo et al. disregard (and even avoid disclosing) commercially tested experiments (e.g, ProImmune REVEAL® assay results) and choose to use scientific methodologies that are incorrectly applied and generate data “in-house” to support what they believe has to be the truth (despite some of their own experiments suggesting the contrary).
8. Regarding the projected population coverage for HLA DQB1*06:02 in Canada and the European countries such as Finland, Luo et al. believe the suggestions indicated in our Response Letter are inaccurate and are not sure where we found these statistics. We would like to direct the reader to the original publication of Ahmed et al. (2) and to the Supplementary information (Table S1). This table indicates that the Immune Epitope Database (IEDB) Analysis Resource (http://tools.immuneepitop...) was utilized to obtain allele genotypic frequency and the projected population coverage for the narcolepsy risk allele HLA-DQB1*06:02 in various countries and the world. This web-based tool indicates an HLA-DQB1*0602 Allele (genotypic frequency) of 17.1 % for Finland and Projected Coverage of HLA-DQB1*0602 allele of 31.3% of that population.
These data were presented also at the IABS meetings on narcolepsy on October 23, 2015 in Geneva, Switzerland that probably Luo et al. were not able to attend due to the investigation surrounding their “retracted paper” (3) around that time period. At that meeting, additional information was shown including an HLA-DQB1*0602 Allele (genotypic frequency) of 2.84 % for Finland and Projected Coverage of HLA-DQB1*0602 allele of 5.6% of that population (with the clearly indicated caveat that since no information is available for the frequency of HLA-DQB1*0602 allele in Canada, the frequency of DR2 was used for estimation as it is associated with narcolepsy and linkage disequilibrium exists for many DR-DQ types). This information was also included in a paper (11) cited in our original Letter to PLOS ONE.
Luo et al. believe our estimates are incorrect and claim to have typed control Canadian samples of European ancestry and obtained a DQ0602 carrier frequency of around 23%, but were surprised that is was difficult for them to find a published reference on this topic except a study reporting the frequency of DRB1*15:01 (allele with 95% linkage with DQ0602). Using this as a reference point, they hypothesize a 21% carrier/phenotype frequency for Canada. For Finland, they estimate a carrier/phenotype frequency of 28% based on the closely linked DRB1*15:01~DQB1*06:02 haplotype. While both Luo et al. and we are trying to approximate the HLA-DQB1*0602 allele genotypic frequency for Canada based on linkage disequilibrium, Luo et al. are quick to conclude, “There is no doubt that carrier frequency in the Quebec population is above 20%.” We are not so sure.
9. Luo et al. state, “Our initial response to this long comment [from Ahmed and Steinman] was short and to the point, as we really did not want to argue. It was just stating: ‘We thank the reviewers for their thoughtful comments. Sometimes experiments cannot be replicated…’.” Luo et al. then claim, “Plos [ONE] nonetheless asked to respond point by point. We [Luo et al.] were waiting for the response of the editors when the comment we were responding to [that of Ahmed and Steinman’s] was posted (just prior to the IABS meeting mentioned above).” Similar to the email provided by Luo et al. regarding the exchange of information between Luo Guo and Sohail Ahmed discussed above in point # 3d(iii), Luo et al.’s version of the facts need to be clarified by us! We submitted a Letter of PLOS ONE that was received and acknowledge by the Editors of PLOS ONE on December 27, 2017. We were informed by the Editors on January 12, 2018 that discussion with the authors [Luo et al.] was in process. On March 5, 2018, we informed the Editors about the upcoming IABS meeting on March 26-27, 2018. On March 19, 2018, due to their ongoing discussions with the authors, we were advised by the Editors that it would be possible to post a public comment on the PLOS ONE article’s webpage to raise to the public’s attention in advance of the IABS conference. On March 25, 2018, we were able to prepare a web-based letter and submit to the PLOS ONE article’s webpage. Luo et al. were aware of our concerns already in January 2018 and responded finally on May 4, 2018. Hopefully, this clarification provides the reader with the appropriate temporal context.

Signed,

S. Sohail Ahmed and Lawrence Steinman

Disclosure: Steinman and Ahmed have a patent application, AVOIDING NARCOLEPSY RISK IN INFLUENZA VACCINES.

APPENDIX

FIGURE 1. https://figshare.com/s/04...
FIGURE 2. https://figshare.com/s/1b...

REFERENCES
1. Luo G et al. (2017) Absence of anti-hypocretin receptor 2 autoantibodies in post pandemrix narcolepsy cases. Plos ONE 12(12): e0187305.
2. Ahmed et al. (2015) Antibodies to influenza nucleoprotein cross-react with human hypocretin receptor 2. Science Translational Medicine 7(294): 294ra105.
3. De la Herran-Arita et al. (2013) CD4+ T cell autoimmunity to hypocretin/orexin and cross-reactivity to a 2009 H1N1 influenza A epitope in narcolepsy. Science Translational Medicine 5(216): 216ra176. RETRACTED 30 JULY 2014
4. https://www.proimmune.com...
5. https://www.proimmune.com...
6. https://retractionwatch.c...
7. Sidney J et al. (2010) Divergent motifs but overlapping binding repertoires of six HLA-DQ molecules frequently expressed in the worldwide human population. Journal of Immunology 185: 4189-4198.
8. Ettinger R and Kwok W. (1998) A peptide binding motif for HLA-DQA1*0102/DQB1*0602, the class II MHC molecule associated with dominant protection in insulin-dependent diabetes mellitus. Journal of Immunology 160: 2365-2373.
9. Tanaka S et al. (2006) Detection of autoantibodies against hypocretin, hcrtrl, and hcrtr2 in narcolepsy: anti-Hcrt system antibody in narcolepsy. Sleep 29(5): 633-8.
10. Yamanaka A et al. (2010) Orexin directly excites orexin neurons through orexin 2 receptor. Journal of Neuroscience 30(38): 12642-52.
11. Ahmed SS and Steinman L (2017) Narcolepsy and influenza vaccination-induced autoimmunity. Annals Translational Medicine 5(1): 25.