Referee 2's review:

Summary
This paper looks at the allele-specific DNA methylation of an upstream CpG-rich region of the IGF2 gene (called DMR0) and studies its methylation status in normal and pathological samples. More specifically, the authors analyze the DNA methylation status of 6 CpG dinucleotides within DMR0. Previous work had suggested that DMR0 is methylated on the silent maternal allele. The authors present results suggesting the opposite, i.e. methylation of DMR0 on the paternal chromosome, which is also methylated at the linked H19DMR (IC1). Their analysis of normal, BWS, SRS and Wilms tumor samples is consistent with a direct relationship between the DNA methylation at IC1 and DMR0 during normal development and in the pathogenesis of BWS and SRS, but not during tumorigenesis. This is the first study of DMR0 methylation in relation to BWS and SRS.

Comments
The main claims of the paper are that the IGF2 DMR0 is in fact methylated on the expressed paternal allele and that its methylation status responds to that of the linked H19 DMR (IC1).

They show a direct link between gain of methylation (in BWS) or loss of methylation (in SRS) at IC1 and epigenetic changes at DMR0. The mechanism regulating this long-range epigenetic interplay do not appear to operate in Wilms tumors, since hypermethylation at IC1 is in fact correlated with hypomethylation at DMR0. This lead to the conclusion that epigenetic defects at IC1 have different consequences during development (normal or pathological) and during tumorigenesis.

The studies are well performed and the data is solid. Much of the study relies on the availability of the appropriate patient and tumor samples, coupled with good techniques to analyze CpG methylation at DMR0.

The study is based only on the analysis of only 6 CpG dinucleotides in this region, but previous work were also based on 6 or even only 3 CpGs. In the M&M, the authors point out that their CpGs include the 3 sites previously studied by Cui et al. in colorectal cancer (ref. 21), but they should also indicate the overlap, if any, with the sites analyzed in the studies involving LOH or UPD material (refs 19 & 20). This is particularly important, since these are the only two studies referred to where the methylation at DMR0 was suggested to be of maternal origin.

For the allele-specific methylation data presented in Fig. 1C, the authors should specify the category to which belong the BWS and SRS samples (UPD, IC defects ?).
This is the first analysis of the parent-of-origin of the DMR0 methylation in normal (non-UPD or LOH) samples and together with the analyses of BWS and SRS patients provide strong evidence for a higher methylation level on the paternal allele. This finding, which is opposite to what is observed in the mouse placenta (DMR0 is biallelically methylated in the embryo), shows that in human, all IGF2 DMRs are methylated on the same allele as IC1. Since their results are different from those previous published, the authors should address this discrepancy head on in the discussion. The most extensive set of data supporting maternal-specific methylation (ref. 20 Sullivan et al.) is all on tumor samples. Even if these cells are genetically mUPD at IGF2, hypomethylation at DMR0 might still be selected for in these cells if, as suggested here in Wilms tumors, cancer cells
Their results also show a clear correlation between methylation at DMR0 and at IC1 in BWS and SRS patients. Their results also suggest that the long-range interactions mediating such epigenotype switch operate differently during tumorigenesis. In itself, this is an interesting observation which could have further implication for our understanding of epigenetic abnormalities associated with cancer.

Presentation
The text is generally difficult to read because of several typos, inconsistencies, and the use of compound adjective in profusion. Adjectives such as “parent of origin” should be written with or without hyphens, but using the same rule throughout. Hyphens are often missing, such as in Beckwith-Wiedemann Syndrome (BWS), insulin-like growth factor 2.
Comments and recommended changes (underlined) about specific sections are presented below.

1. Abstract
• DMRs are not “germline” or “somatic” ; this is an inappropriate use of the term. DMRs are simply DNA sequences. It is their epigenetic marking, for example by DNA methylation, which can be acquired at different stages. The same mistake in made in the Introduction, in discussing the mouse Igf2 DMRs and in the Discussion.
• “ ...disorder such as Beckwith-Wiedemann (BWS) and Silver-Russell (SRS) syndromes...”.
• “The effect of DMR0 methylation changes in the aetiology...”

2. Author summary
• This section is not well edited and needs to be re-written. There are several typos, missing hyphens and punctuation mistakes, such as:
• dependant > dependent
• “ only active on parental copy” > “ only active on one parental copy”
• “sulin like” > “insulin-like”
• “Silver Russell Syndrome” > “Silver-Russell syndrome”
• “birth disorder. Beckwith-...” > “birth disorder Beckwith-...”
• The term “gene” is used inappropriately in reference to an “allele”. IGF2 is active on the paternally inherited allele, not gene. There’s only one IGF2 gene in our cells; we have two alleles of this gene.
• The sentence starting with “It is believed...” and the following one need to be re-written. The authors must clearly state and distinguish what are the actual results previous obtained by others, and what are the hypotheses proposed on the basis of these results. The fact that differential DNA methylation has previously been documented at DMR0 is not an hypothesis, nor are the (few) results suggesting that it is the maternal allele which is methylated.

3. Intro
• Several mistakes involving lack of hyphens, particularly at “insulin-like”, “Beckwith-
Wiedemann”.
• “ the closely linked Igf2 and H19 genes...”
• add a specific reference for the data on the deletion of the maternal H19 DMR.
epigenetic marks at those DMRs which can be qualified as being “gametic (primary)” or
“somatic (secondary)”, NOT the DMRs themselves.
• “...influenced by CTCF binding and DNA methylation.”

4. Results
• The final sentence of the first paragraph does not seem to make sense; I agree with the first alternative, not to the second. The results obtained with the non-UPD BWS and SRS patients (Fig.1B) would suggest a trans de novo methylation (of the maternal IC1 and paternal DMR0, assuming that DMR0 is normally methylated on the maternal allele), or that DMR0 is in fact methylated on the paternal allele, like H19 DMR (requiring cis methylation of both maternal DMRs). It is this latter possibility which the authors confirm in their subsequent experiments.
• In the analysis of the Wilms tumor samples, the authors discuss two samples from BWS cases; why are these results not included in any of the figures, even if they represent single data point?

5. Discussion
• I believe that the results in BWS are similar to what was observed in the choroid plexus in the mouse (Feil et al. Dev. 1994): biallelic methylation at IC1 and (at least) some IGF2 DMRs. In both cases, silencing of H19 and biallelic expression of IGF2/Igf2 is observed. Could this provide a non-pathological example where a similar epigenotype switch occurs? Perhaps the authors could point out this parallel in their discussion.
• For the sentence starting with “Thus far loss of methylation...”, it should be clear that this applies only to tumors, since they show gain of methylation as an epimutation at the DMR0 in BWS.

M&M
• EZ DNA : from which company/vendor?

Fig.1 and 2
What is meant by “expected normal methylation”? Instead of a general statements about
methylation levels at DMRs in general, why not refer to the results presented in Fig. 1C ?

**********
N.B. These are the comments made by the referee when reviewing an earlier version of this paper. Prior to publication the manuscript has been revised in light of these comments and to address other editorial requirements.