Mitochondrial DNA Instability Is Common in HIV-Exposed Uninfected Newborns

Round 1

Reviewer 1 Report

The further response adequately addresses my concerns.

Reviewer 2 Report

I thank the authors for adapting the manuscript based on my review. I feel that the manuscript has been improved and I'm satisfied with the point-to-point reply provided and with the changes. As already said in my first review the work is very valuable and I recomend it for publication.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

The authors present a novel approach to study mitochondrial genotoxicity in HEU children performing deep mtDNA sequencing and looking for first time in an exhaustive manner at mtDNA deletions and heteroplasmy rates. The manuscript is well written and the methods outstanding making the work very valuable and certainly suitable for publication.

However I would like to raise the following considerations, which should be clarified or better explained in the discussion section:

• Although the results of the comparison between HEU children and controls are striking, the total number of patients (n=32 vs n=15) remain very small. This point should be raised in the discussion when highlighting possible limitations of the study.
• Additionally a possible influence of the time point of blood collection (d3 vs d7) on the results should also be discussed.
• The authors explain how mtDNA deletions preferentially accumulate with age in post-mitotic tissues such as the brain or skeletal muscles rather than in the blood. Even in patients with inherited mitochondrial disorders such levels of mtDNA deletions are encountered in the blood, as blood cells have a high division rate allowing for rapid elimination of altered mtDNA deletions (lines 210-216). These facts inevitably raise the question about the significance of the study findings and call for a longitudinal assessment of mtDNA deletions over time in HEU children. This point should be included in the discussion.
• The studied population of HEU children is very homogeneous and all the children were directly and indirectly exposed to the same ARV drugs. It would be interesting to study different type and amount of exposure in a larger cohort of HEU children. Especially, it would be valuable to study HEU not exposed to neonatal PEP in order to assess the influence of direct exposure on mtDNA deletions.
• The findings of this study are clear and well presented. However the link between these findings and the previously described health impairments of the HEU children remain purely hypothetical. This point should be stressed out more in the discussion.

Author Response

We thanks the Reviewer #1 for the comments which help us to improve the manuscript

The authors present a novel approach to study mitochondrial genotoxicity in HEU children performing deep mtDNA sequencing and looking for first time in an exhaustive manner at mtDNA deletions and heteroplasmy rates. The manuscript is well written and the methods outstanding making the work very valuable and certainly suitable for publication.

However I would like to raise the following considerations, which should be clarified or better explained in the discussion section:

• Although the results of the comparison between HEU children and controls are striking, the total number of patients (n=32 vs n=15) remain very small. This point should be raised in the discussion when highlighting possible limitations of the study.

We agree with the reviewer that the number of samples is small. We added at the end of the discussion a “study limitations” paragraph which was missing in the original version and we included this remark (Line 326-327). Given the cost of such experiments and the design bias inherent in this kind of post-hoc analysis, we have decided to develop a study dedicated to this aim with more recent samples.

• Additionally a possible influence of the time point of blood collection (d3 vs d7) on the results should also be discussed.

This point is now addressed in the “study limitations” paragraph as follows (Line 327-330): “control samples were collected on average 4 days before those of the PROMISE trial. To the best of our knowledge, it is unlikely that the time frame for the acquisition of mtDNA deletion will occur specifically during these 4 days.”

• The authors explain how mtDNA deletions preferentially accumulate with age in post-mitotic tissues such as the brain or skeletal muscles rather than in the blood. Even in patients with inherited mitochondrial disorders such levels of mtDNA deletions are encountered in the blood, as blood cells have a high division rate allowing for rapid elimination of altered mtDNA deletions (lines 210-216). These facts inevitably raise the question about the significance of the study findings and call for a longitudinal assessment of mtDNA deletions over time in HEU children. This point should be included in the discussion.

We modified the discussion to further strengthen this point as follows (Lines 315-320: “Future steps will require further investigation of deleted mtDNA molecules later in HEU children’s lives, in both qualitative and quantitative terms. Different mechanisms of elimination of damaged mitochondria have been described but are only efficient up to a certain threshold of heteroplasmy [44]. However, should they become inefficient, mtDNA deletions observed herein are likely to contribute, at least partially, to the health impairments observed in HEU children, especially those with increased deletion contents.”   

•   The studied population of HEU children is very homogeneous and all the children were directly and indirectly exposed to the same ARV drugs. It would be interesting to study different type and amount of exposure in a larger cohort of HEU children. Especially, it would be valuable to study HEU not exposed to neonatal PEP in order to assess the influence of direct exposure on mtDNA deletions.

We definitely agree with the reviewer and as noted in the answer to point 1, we are fully committed to launch a new study with such objectives among HEU under the current PMTCT program. Nonetheless, we find that the message of the manuscript was important enough to be shared with the community. To some extent this suggestion was included in the text (Lines 279-286): “No reported studies investigated the effect of various ARVs on the occurrence of mtDNA deletions in humans. Given the new guidelines and practices, HIV-infected pregnant women are now receiving triple ARV therapy and it is now recommended that HEU children receive an extended prophylaxis up to the end of breastfeeding in order to tackle MTCT of HIV [39,40]. Such intensification of ARV exposure, up to triple-drug prophylaxis for the child, as well as the prolonged exposure to more than one year strongly advocate for a full assessment of mtDNA integrity and long term clinical follow up.”

• The findings of this study are clear and well presented. However the link between these findings and the previously described health impairments of the HEU children remain purely hypothetical. This point should be stressed out more in the discussion.

We agree with the reviewer and we have modified the text accordingly. Lines 316-322: “Mechanisms of elimination exist and are efficient up to a certain threshold of heteroplasmy [44]. However, should they become inefficient, mtDNA deletions observed herein are highly likely to contribute, at least partially, to the health impairments observed in HEU children, especially those with increased deletion contents. It is striking that cardiac conduction dysfunctions, muscle weakness or elevated lactate concentrations reported in HEU children [45–47] are also hallmarks of most mitochondrial disorders.” And Line 332-333: “Fifthly, the study is purely descriptive and cannot predict any causality for future health outcomes.”

Author Response File: Author Response.docx

Reviewer 2 Report

The authors study mitochondrial genotoxicity in HEU (HIV-exposed uninfected) infants, compared with unexposed controls. This is an important topic given the large numbers of HEU infants born every year. The reasons for some adverse health outcomes in HEU infants are not well-understood and mitochondrial genotoxicity (due to ART or HIV itself) is a plausible hypothesis, with a small amount of inconclusive literature to date. 

Unfortunately however, this paper has a number of methodological problems, some rather serious, which mean that the conclusions cannot be justified. 

Specific comments.

Material and Methods.

Line 82 - suggests that some mothers received lamivudine as well as zidovudine in pregnancy, however the results table appear to report only ZDV - please clarify.

Line 93 - was DNA extracted from DBS immediately after DBS collection? If so, was this true both for HEU and controls? If not, were there any differences in DBS storage time before DNA extraction?

Line 106/107 - Please add the nt positions as well as the primer sequences, both in the text here, and also to the figures. I have BLASTed the stated primers. The first pair appear to be for an amplicon spanning nt 7017-15044, which corresponds with the stated size. However I BLAST the other two primers to nt 4830 and 4243. Is there an error in the stated sequences? 

Line 134 - states that the same results (deletions) were observed from cell pellets as well as DBS. This statement may be taken to mean that the findings were validated by detecting the same deletions in both DBS and cell pellets from the same subjects. In fact it appears (supplementary table) that this is not the case - in fact it is simply that some subjects had DNA extracted from DBS and some from cell pellets. This should be clarified.

Specific concerns about the sequencing methodology:

Minimal depth of 100 reads per base needs clarifying. I assume this means that 100% of bases in a sample required coverage of at least 100x. However 100x is far too low for low-level heteroplasmy analyses. It would be better to set a criterion like >1000x coverage at >95% of base positions.

There does not appear to be any true validation of the deletion findings. The authors state that they repeated each sample, but this was only from the library stage. This does not therefore control for PCR artefacts during the amplicon generation step, which are probably a likely cause of false positive findings. The authors should perform a spike-in control with a known single deletion to confirm the specificity and sensitivity of the assay. They should also consider confirming some of the deletions found by specific PCR.

I cannot be justified to report heteroplasmic deletions at very low heteroplasmy levels (e.g. <1%). Even with good coverage (and therefore a good number of supporting reads), the inherent noise in NGS means it is not possible to reliably report heteroplasmy to <1-2%. 

How heteroplasmic (and presumably homoplasmic) point mutations were dealt with needs to be described in the methods. What were criteria for pathogenicity?

Results.

Line 154 - 'within the limits' - clarify what this means (presumably normal range for bloods and a certain centile for weight)?

Table 1 - when is maternal viral load taken (at birth)? Plus clarification as above re ARV regimens.

Figure 1A - resolution was too low for me to review this. Please also add primer / amplicon positions to the plots.

Figure 1B - heteroplasmic point mutations are not reported as they were not 'pathogenic', however some were detected (presumably therefore non-pathogenic). I would suggest it is better to report all point mutations, and then stratify by type.

Discussion.

Line 249 - the authors rightly suggest that longer follow-up of mtDNA integrity is required in HEU infants. However I would assume that such samples might have been available from the PROMISE trial? Is this the case, and if so, could they be examined - e.g. a 6 month or 12 month follow-up sample.

Other.

Line 294 - Given that this paper is a post-hoc analysis of samples from the PROMISE trial, did the original consent process cover genetic analyses? If not, was additional IRB permission requested for this?

Supplementary table.

It is very clear from the 'deletion' column on this table (and indeed from Figure 1C) that a high proportion of the deletions report sit at or just inside the primer positions for one of the long-PCR amplicons. This strongly suggests that many / all of the deletions detected may be accounted for by long-PCR artefacts. 

'Sequencing depth' -  is this the lowest coverage seen at any base position? It would be more usual to report a mean / median across the whole mtDNA genome and perhaps a 5th centile of coverage.

'Heteroplasmy' - comment as above re. reporting of very low-level heteroplasmies (e.g. <1-2%).

'Point mutation' - comment as above re. reporting of point mutations.

Author Response

We thank the reviewer #2 for the comments which help us to improve the manuscript

The authors study mitochondrial genotoxicity in HEU (HIV-exposed uninfected) infants, compared with unexposed controls. This is an important topic given the large numbers of HEU infants born every year. The reasons for some adverse health outcomes in HEU infants are not well-understood and mitochondrial genotoxicity (due to ART or HIV itself) is a plausible hypothesis, with a small amount of inconclusive literature to date. 

Unfortunately however, this paper has a number of methodological problems, some rather serious, which mean that the conclusions cannot be justified. 

Specific comments.

Material and Methods.

Line 82 - suggests that some mothers received lamivudine as well as zidovudine in pregnancy, however the results table appear to report only ZDV - please clarify.

We agree with the reviewer that this point is confusing. In fact, Line 83 describes the various ARV regimens taken during pregnancy by all the mothers in the PROMISE-PEP trial. However, in the analysis restricted to 32 HEU children, mothers were among those who received ZDV only. We clarified this in lines 85-86: “It is noteworthy that in the present study, all 32 mothers of HEU children were among those who received ZDV only.

Line 93 - was DNA extracted from DBS immediately after DBS collection? If so, was this true both for HEU and controls? If not, were there any differences in DBS storage time before DNA extraction?

DBS for HEU children were collected between 2009 and 2012. They were stored at -80°C until the end of the trial and then transferred to Montpellier, France for definitive storage (-80°C). DBS for control children were collected in France and for some of them during 2012. The latter were used to verify that long term storage had no impact on the results of the experiment. This point is mentioned in the text Line 89-92: “We also analyzed anonymously 15 HIV-unexposed uninfected neonates collected in France for the systematic screening of inherited diseases at birth in 2019 (n = 9) and in 2012 (n = 6) to check the absence of bias due to DNA alteration with time on dried blood spots (DBS).”. and Line 143-145: “We also checked for the absence of mtDNA deletions in control neonates with the same DBS duration of conservation (8 years) to avoid biased results due to DNA alteration.” DNA extractions were performed in 2020.

Line 106/107 –Please add the nt positions as well as the primer sequences, both in the text here, and also to the figures. I have BLASTed the stated primers. The first pair appear to be for an amplicon spanning nt 7017-15044, which corresponds with the stated size. However I BLAST the other two primers to nt 4830 and 4243. Is there an error in the stated sequences? 

We thank the reviewer for pointing out this mistake. It is now corrected as follows (Line 109-114) : “Briefly, mtDNA (NC_012920.1) from DBS DNA extracts was amplified in two overlapping fragments of 8009 bp and 8994 bp using the following forward and reverse primer 5’-TACGTTGTAGCCCACTTCCACT-3’ (position 7018-7039) and 5’-GCCCGATGTGTAGGAAGAG-3’ (position 15036-15018), and 5’- AACTTCGGCTCACTCCTTGG-3’ (position 140832-14851) and 5’- AGTAACGTCGGGGCATTCCG-3’ (position 7225-7206) and the long range LA Taq DNA polymerase (Takara Bio Europe SAS, Saint-Germain-en-Laye, France).”

Line 134 - states that the same results (deletions) were observed from cell pellets as well as DBS. This statement may be taken to mean that the findings were validated by detecting the same deletions in both DBS and cell pellets from the same subjects. In fact it appears (supplementary table) that this is not the case - in fact it is simply that some subjects had DNA extracted from DBS and some from cell pellets. This should be clarified.

We agree with the reviewer, the text was not precise enough. We have modified it accordingly (Line 141-143). “The mtDNA deletions were evidenced in total DNAs extracted from dried blood spots but the same results were observed with total DNAs extracted from dried cell pellets.” was replaced by “The mtDNA deletions were evidenced in total DNAs extracted from dried blood spots but the same rate of deletion was observed with total DNAs extracted from dried cell pellets.” 

Specific concerns about the sequencing methodology:

Minimal depth of 100 reads per base needs clarifying. I assume this means that 100% of bases in a sample required coverage of at least 100x. However 100x is far too low for low-level heteroplasmy analyses. It would be better to set a criterion like >1000x coverage at >95% of base positions.

We confirm that “minimal depth of 100 reads per base” stands for “100% of bases in a sample required coverage of at least 100x”. Then, we decided to set a minimal number of reads with an identical sequence at 10 in order to be analysed which allows us to exclude putative artefactual deletions. This method is described in M&M section Line 131-134 by “For this study, we determine the deletion detection threshold for each mtDNA position by the detection of at least ten reads out of the total reads, therefore, the cutoff varied for each position depending on the sequencing depth for a given position.” This method was published (Goudenège D, Bris C, Hoffmann V, Desquiret-Dumas V, Jardel C, Rucheton B, Bannwarth S, Paquis-Flucklinger V, Lebre AS, Colin E, Amati-Bonneau P, Bonneau D, Reynier P, Lenaers G, Procaccio V. eKLIPse: a sensitive tool for the detection and quantification of mitochondrial DNA deletions from next-generation sequencing data. Genet Med. 2019;21:1407-1416). Together, with the lowest sequencing depth (n=523) the minimal heteroplasmy rate would be 1.9% and with the highest sequencing depth (n=28671), this rate decreases to 0.003%.

There does not appear to be any true validation of the deletion findings. The authors state that they repeated each sample, but this was only from the library stage.

Thank you for pointing this out, which allows us to clarify the text. Indeed, out of the 32 samples included in the study, 22 have been repeated from the PCR amplification step, library preparation included the PCR amplification step in our language. This is now included in the text as follows (lines 137-139) : “To ensure the reproducibility of the mtDNA deletion detection method, we applied twice the entire workflow (library preparation, PCR emulsion, sequencing and bioinformatics analysis) on three samples.” was replaced by “To ensure the reproducibility of the mtDNA deletion detection method, we applied the entire workflow (library preparation, PCR emulsion, sequencing and bioinformatics analysis) two times on 22 out of the 32 HEU samples.”

This does not therefore control for PCR artefacts during the amplicon generation step, which are probably a likely cause of false positive findings. The authors should perform a spike-in control with a known single deletion to confirm the specificity and sensitivity of the assay. They should also consider confirming some of the deletions found by specific PCR.

It is noteworthy that the vast majority of our patients routinely explored with this technique show no evidence of deletions, as do our controls in this paper. This absence of deletions in a very large number of samples tested in our laboratory indicates that the technique does not artifactually amplify deletions. Regarding reproducibility, hereafter is an example of a deletion rate explored twice (2 different libraries) in the urine of a patient. The identification and quantification are perfectly reproducible.

Patient with a single mtDNA deletion on urine sample :

1st library :

9907-15448 3,48 %

2nd library :

9907-15448 à 3,4%

Finally, this technique has been the subject of two publications in which these different quality controls were explored and validated with a sensitivity of 0.5% in a conservative manner for diagnostic purposes. Herein, we have a research purpose and we believe that description of events with subclinical impacts is equally significant because they may reflect subclinical instability that may alert us to possible adverse effects. This finality has justified the use of a minimal of 10 reads carrying a deletion as a threshold of analysis and consequently a variable sensitivity according to the sequencing depth. We strengthen this choice in the M&M section (line 134-137). “This level of significance differed from the threshold presented in the published method which was set in a more conservative manner for diagnostic purposes. However, in the context of a research study, events of lower frequency may be indicative of a subclinical genetic instability.” and include this point in the study limitations paragraph (line 333-337). “Finally, we used a minimal number of read of 10 to include the deletion in the analysis. The use of this threshold allows for the description of low frequency events and renders possible the reporting of low-level genetic instability. Control samples and routine practice clearly demonstrate the success of this threshold.”  

Goudenège D, Bris C, Hoffmann V, Desquiret-Dumas V, Jardel C, Rucheton B, Bannwarth S, Paquis-Flucklinger V, Lebre AS, Colin E, Amati-Bonneau P, Bonneau D, Reynier P, Lenaers G, Procaccio V. eKLIPse: a sensitive tool for the detection and quantification of mitochondrial DNA deletions from next-generation sequencing data. Genet Med. 2019;21:1407-1416.

Bris C, Goudenège D, Desquiret-Dumas V, Gueguen N, Bannwarth S, Gaignard P, Rucheton B, Trimouilla A, Allouche S, Rouzier C, Jardel C, Slama A, Barth M, Verny C, Spinazzi M, Cassereau C, Colin E, Armelle M, Pereon Y, Martin-Negrier ML, Paquis-Flucklinger V, Letournel F, Lenaers G, Bonneau D, Reynier P, Amati-Bonneau P, Procaccio V. Improved detection of mitochondrial DNA instability in mitochondrial genome maintenance disorders. Genet Med. 2021, In Press.

I cannot be justified to report heteroplasmic deletions at very low heteroplasmy levels (e.g. <1%). Even with good coverage (and therefore a good number of supporting reads), the inherent noise in NGS means it is not possible to reliably report heteroplasmy to <1-2%. 

We do not believe that these low deletion rates are artefacts as there is no background in the vast majority of our patients routinely explored with this technique, as well as in the control samples explored in this article. The rate of deletion was identical if samples were from DBS or dried cell pellets and DBS from healthy children of 8 years of age were also negative. On the other hand, these low deletion rates have been fully validated in two published articles (Bris et al 2018 and 2021). The fact that we ensure in this paper is that a deletion with the same breaking point that is present on at least ten copies guarantees the absence of accidental detection.

How heteroplasmic (and presumably homoplasmic) point mutations were dealt with needs to be described in the methods. What were criteria for pathogenicity?

We modified the text accordingly.

Line 146-154: “Mitochondrial DNA variants were detected, annotated and prioritized as previously described [30]. In short, Next Generation Sequencing reads were analyzed with a dedicated in-house pipeline for coverage analysis, variant calling, annotation and prioritization. The identified variants were next analyzed individually to determine biological relevance using a funnel strategy with the above criteria: (i) status in Mitomap database, (ii) frequency in mtDNA variants databases (Genbank, GnomAD and Hellix), (iii) conservation percentage of the nucleotide and (iv) classification of the variant by in-silico pathogenicity predictor tools (MitImpact, APOGEE, SIFT and POLYPHEN)”

Results.

Line 154 - 'within the limits' - clarify what this means (presumably normal range for bloods and a certain centile for weight)?

We agree with the reviewer and clarify the text as follows (lines 172-176) : “Nineteen (59.4%) of the HEU children in our study originate from Burkina Faso. The sex ratio was close to 1:1 but of 2:1 (Boy:Girl) in the control group. Their growth indicators were in the normal range according to WHO standards and did not differ from the values of the other children enrolled in the trial when matched by country (data not shown).”

Table 1 - when is maternal viral load taken (at birth)? Plus clarification as above re ARV regimens.

We apologize if this information was not clear. Maternal viral load was taken at D7. This is specified in the title of the Table 1 but we added this information in the text (line 177)

Figure 1A - resolution was too low for me to review this. Please also add primer / amplicon positions to the plots.

We join to this revised manuscript the original pictures. We modified the figure according to the comment.

Figure 1B - heteroplasmic point mutations are not reported as they were not 'pathogenic', however some were detected (presumably therefore non-pathogenic). I would suggest it is better to report all point mutations, and then stratify by type.

As our article is oriented towards the clinical applications of this exploration, we have only focused on the exploration of pathogenic nucleotide variants according to the diagnostic criteria outlined above. We do not consider it is useful to include non-pathogenic variants without clinical consequences in this list.

Discussion.

Line 249 - the authors rightly suggest that longer follow-up of mtDNA integrity is required in HEU infants. However I would assume that such samples might have been available from the PROMISE trial? Is this the case, and if so, could they be examined - e.g. a 6 month or 12 month follow-up sample.

Indeed, we have the samples to perform such investigations and they are currently in the process to be analysed. We decided not to wait for these results to submit this manuscript for two main reasons: (i) children enrolled in the PROMISE-PEP trial had received one-year prophylactic regimen. The trial arms consisted in a 3TC or lopinavir/ritonavir regimens, which added an extra-exposure to ARV that may have an impact on the rate of mitochondrial deletion. Furthermore, the trial did not follow children receiving no treatment. Therefore, the only study design to analyse such results is a comparison between the two arms, when the study herein is a cross sectional observation. (ii) We believe that the results reported here are sufficiently original and deserve to be disseminated to the scientific community.

Other.

Line 294 - Given that this paper is a post-hoc analysis of samples from the PROMISE trial, did the original consent process cover genetic analyses? If not, was additional IRB permission requested for this?

Samples were collected from 2009 to 2012 and at that time consent for genetic studies was not the norm. We did not inform the different country IRBs because we were addressing somatic genetic and not constitutional genetic. This point was discussed with the country PIs during scientific advisory meetings.

Supplementary table.

It is very clear from the 'deletion' column on this table (and indeed from Figure 1C) that a high proportion of the deletions report sit at or just inside the primer positions for one of the long-PCR amplicons. This strongly suggests that many / all of the deletions detected may be accounted for by long-PCR artefacts. 

Indeed with the wrong set of primers, it was impossible for the reviewer to visualize the localisation of the deletion with regards to the position of the primers. This is now corrected. As the long-range PCR fragments overlap, deletion encompassing either primer is experimentally possible. In addition, this soft-clipping technique used by eKLIPse has been used in other articles:

Piro-Mégy C, Sarzi E, Tarrés-Solé A, Péquignot M, Hensen F, Quilès M, Manes G, Chakraborty A, Sénéchal A, Bocquet B, Cazevieille C, Roubertie A, Müller A, Charif M, Goudenège D, Lenaers G, Wilhelm H, Kellner U, Weisschuh N, Wissinger B, Zanlonghi X, Hamel C, Spelbrink JN, Sola M, Delettre C. Dominant mutations in mtDNA maintenance gene SSBP1 cause optic atrophy and foveopathy. J Clin Invest. 2020;130:143-156.

Piotrowska-Nowak A, Krawczyński MR, Kosior-Jarecka E, Ambroziak AM, Korwin M, Ołdak M, Tońska K, Bartnik E. Mitochondrial genome variation in male LHON patients with them.11778G> A mutation. Metab Brain Dis. 2020;35:1317-1327.

Also it has to be kept in mind that such techniques do not give information regarding the full length mtDNA molecule. Duplication was reported by others in various pathological situations and ageing, we therefore suggested that this event may occur in these samples. These aberrant molecules resulting from abortive replication may generate uncommon sequences. This point was mentioned in the text line 264-271.

Sequencing depth' -  is this the lowest coverage seen at any base position? It would be more usual to report a mean / median across the whole mtDNA genome and perhaps a 5th centile of coverage.

We apologize that this information was not clear. The sequencing depth is the mean coverage across the whole mtDNA genome. It is now corrected.

'Heteroplasmy' - comment as above re. reporting of very low-level heteroplasmies (e.g. <1-2%).

'Point mutation' - comment as above re. reporting of point mutations.

Round 2

Reviewer 2 Report

The authors have very largely addressed my previous concerns and the manuscript has been significantly improved.

However I have a few points that still require complete clarification:

• Fig 1C - the response and legend indicate that long-PCR amplicon / primer positions have been added to the figure, but it is not clear where these are.
• I am still unclear about the verification of deletions and their heteroplasmy levels by repeat sequencing. I understand that the whole workflow has been repeated from the 'library prep' stage, but does this include going right back to the initial DNA extract and repeating the long-range PCR step?
• I remain unhappy about the choice of heteroplasmy thresholds for reporting deletions:
  • Firstly the authors vary this according to the read depth per base position. This makes it very difficult to compare between samples or to make a meaningful statement about what proportion of subjects showed some deletions. I suggest the following approach instead:
    • 1) decide what is an acceptable minimum coverage that all samples need to achieve for all (or perhaps 95%) of base positions. Exclude (or repeat) those samples which fail to meet this minimum coverage requirement.
    • 2) then determine what is the minimum deletion heteroplasmy threshold that can be reported for ALL samples (e.g. at 10 reads minimum for the variant) based on this minimum accepted coverage (i.e. do not vary reporting of deletion heteroplasmy on a base pair by base pair basis). For example if you require minimum coverage of 1000x for all samples then minimum reported heteroplasmy will be 1%.
    • 3) Report only those deletions exceeding this  minimum heteroplasmy threshold as the main results. If the authors wish to show lower level heteroplasmic deletions they could present supplementary data where they analyse a subset of samples which achieved higher coverage.
  • Secondly the authors refer to previous publications validating the deletion detection method, but they acknowledge that those papers only validated down to 0.5% heteroplasmy. If the authors with to present any deletions at <0.5% heteroplasmy then they certainly need to perform their own validation (it does not matter whether they are reporting these for 'research' or for 'clinical' interest - the data still need to be valid!).

Author Response

Thank you for all your inputs which improve the manuscript.

1. Fig 1C - the response and legend indicate that long-PCR amplicon / primer positions have been added to the figure, but it is not clear where these are.

We apologize, as we did not submit the revised figure. It has now been corrected and we also modified the legend as follows (line 246): “Products of long-range PCR amplification are depicted on top of the diagram.”

2. I am still unclear about the verification of deletions and their heteroplasmy levels by repeat sequencing. I understand that the whole workflow has been repeated from the 'library prep' stage, but does this include going right back to the initial DNA extract and repeating the long-range PCR step?

Our verification did not include a new DNA extraction, but it included new long-range PCRs. It is now explained in the text as follows (line 138-139): “To ensure the reproducibility of the mtDNA deletion detection method, we applied the entire workflow (long-range PCRs, library preparation, PCR emulsion, sequencing and bio-informatics analysis)…”

3. I remain unhappy about the choice of heteroplasmy thresholds for reporting deletions:

1. Secondly the authors refer to previous publications validating the deletion detection method, but they acknowledge that those papers only validated down to 0.5% heteroplasmy. If the authors with to present any deletions at <0.5% heteroplasmy then they certainly need to perform their own validation (it does not matter whether they are reporting these for 'research' or for 'clinical' interest - the data still need to be valid!).--

As for any experimental assay, some results are detectable but not accurately quantifiable. We may have been unclear in our choice of words, but we never said that heteroplasmy rate was invalidated below 0.5%. We only said that the technique has a sensitivity of 0.5% which means that all samples with a heteroplasmy of 0.5% or above are correctly quantified. It doesn’t mean that heteroplasmy below this threshold are invalid, they are simply outside the range that permits precise quantification. Nevertheless, we are convinced that all the deletions with heteroplasmy below 0.5% are indeed valid because neither the control sample of this study nor the thousand samples tested in routine laboratory showed deletion. All these low frequency deletions are true events. The confusion was between the lower limit of quantification and the lower limit of detection.

We clarified the text accordingly as follows (line 130-138): “For this study, we determine the lower limit of detection for each mtDNA position by the detection of at least ten reads out of the total reads, given that none of the control samples presented deleted mtDNAs in our routine practice. As a consequence, the heteroplasmy rate may be lower than the one previously published [29]. This threshold of quantification was set for diagnostic purposes and in a rather conservative manner. However, in the context of a research study, events of lower frequency may be indicative of a subclinical genetic instability. Therefore, a heteroplasmy rate from 0.01% to 0.5% will be not accurately quantified.” And line 218-219: “Out of them, 16 (25%) were in the linear range of quantification.”

1. Firstly the authors vary this according to the read depth per base position. This makes it very difficult to compare between samples or to make a meaningful statement about what proportion of subjects showed some deletions. I suggest the following approach instead:

1. 1) decide what is an acceptable minimum coverage that all samples need to achieve for all (or perhaps 95%) of base positions. Exclude (or repeat) those samples which fail to meet this minimum coverage requirement.
2. 2) then determine what is the minimum deletion heteroplasmy threshold that can be reported for ALL samples (e.g. at 10 reads minimum for the variant) based on this minimum accepted coverage (i.e. do not vary reporting of deletion heteroplasmy on a base pair by base pair basis). For example if you require minimum coverage of 1000x for all samples then minimum reported heteroplasmy will be 1%.
3. 3) Report only those deletions exceeding this  minimum heteroplasmy threshold as the main results. If the authors wish to show lower level heteroplasmic deletions they could present supplementary data where they analyse a subset of samples which achieved higher coverage.

We understand the reviewer’s concern for the threshold variations with samples. The reviewers’ suggestion can be applied but it does fix the lower limit of quantification as the threshold of positivity which is, as mentioned above and in our point of view, not correct. To some extent, we are in the same situation concerning the description of viral quasi-species in an infected patient which adopt the same analytical plan. With a variable threshold, we may underreport these events in samples with the lowest sequencing depth. This is now added in the limitation paragraph (lines 326-327): “As this is an inference, we may underreport these deletion events for samples with low sequencing depth.”