Article activity feed

  1. Author Response

    Reviewer #2 (Public Review):

    Fukuda et al. use whole genome bisulfite sequencing (WGBS) and RNA sequencing (RNA-seq) data obtained from sperm and human primordial germ cells (hPGCs), as well as KRAB-ZFP protein ChIP-seq data obtained from HEK293T cells, to study the relationship between DNA methylation, KRAB-ZFP binding and genome-wide transcription of LINE-1 (L1), SVA and LTR12 retrotransposons.

    This work aims overall to elucidate pathways silencing retrotransposons in the male germline, in particular making new (and known) links between ZFPs and DNA methylation. The focus here ends up being on immobile retrotransposons, as L1s (bound by ZNF93 and ZNF649) and SVAs (bound by ZNF28 and ZNF257) capable of mobilization either do not have binding sites for the identified ZFPs, or have far fewer than their older relatives. The relationships between L1, ZNF93 and ZNF649 has been reported previously (Jacobs et al., 2014, Nature; Fernandes et al., 2018, bioRxiv). That older retrotransposons have more binding to these ZFPs, and are more methylated in hPGCs, is based mainly on correlation. Overall, I thought the subject matter was interesting but I have substantial reservations around the analyses, particularly the more novel results related to SVA. As the work stands it is not clear whether ZNF257 or ZNF28 are reinforcing DNA methylation on SVA. The claims around this repression being transcriptionally-directed or varying significantly amongst individuals, for biological as opposed to technical reasons, appear preliminary at this stage.

    Specific comments:

    1. The use of WGBS to analyse very young retrotransposons, like L1HS and SVA_F, has potential caveats. One of the most important of these is that the CpG islands most likely to be differentially methylated for these elements, in somatic cells at least, are internal to their sequences (e.g. PMID: 33186547). This includes the SVA VNTR sequence, which is where the vast majority of proposed ZNF28 and ZNF257 binding motifs reside (Fig. 2F). Does WGBS, using only uniquely mapped reads as done here, resolve these regions sufficiently to identify differential methylation?

    First of all, thank you for your valuable comments on our manuscript. We confirmed that VNTR in SVA_A derived sequences can be uniquely mapped and DNA methylation in VNTR in SVA_A could be analyzed (please see New Supplementary Fig. S1I and J).

    1. Why does Fig. 1D have only 36 full-length L1HS copies? The definition of a full-length L1 here (>90% consensus length) should yield (from memory) >300 reference L1HS copies.

    According to our threshold for full-length, we obtained 319 full-length L1HS copies. However, the most of them could not measure DNA methylation levels due to low mappability (please see New Supplementary Fig. S1B).

    1. It would be useful to explain the inclusion of LTR12 as a representive ERV, as opposed to, say HERVK, which has been studied in hPGC like cells recently (PMID: 35075135).

    Thank you for your comment. We added the following sentence. “A subset of LTR transposons, including LTR12, function as enhancers (Deniz et al., 2020). It was recently reported that LTR5s, which are Hominidae-specific LTR-type transposons and hypomethylated in hPGCs (DNA methylation levels < 10%), can function as enhancers to promote hPGC differentiation (Xiang et al., 2022). Therefore, in the case of LTR12C, maintaining DNA methylation might be beneficial for hPGC development because it suppresses inappropriate activation of transposon-embedded enhancer function.” p18, 311-316

    1. The exo ChIP-seq for a variety of ZFPs was obtained from published data generated using HEK293T cells, whereas the WGBS is from sperm and hPGCs. What evidence can the authors point to be reasonably sure that the ZFP binding patterns from HEK293Ts carry over to the male germline in vivo?

    It is not known whether ZNF binding pattern in HEK293Ts is same, similar or totally different in male germline in vivo, and we do recognize this is the strong shortage of this paper. Good antibody for KRAB-ZNFs, availability of human PGCs and establishment of low-cell number input ChIP-seq (or Cut and TAG) for KRAB-ZNFs are required to address this issue. From our analysis, all we can say is KRAB-ZNFs we identified are candidate factors for retroelement silencing during human male germ cell development.

    1. In Fig. 3C it appears quite a few L1PA3s have ZNF649 peaks and yet the motif for ZNF649 has two mismatches to the L1PA3 consensus (Fig. 3F). Yes, L1HS has one more mismatch than L1PA3. It would be useful to explain further why two mismatches are acceptable whereas 3 completely abolishes binding.

    We added following sentences. “Although highly methylated L1 copies had two mismatches within the ZNF649 binding motif, one at the third position (T→G) and one at the sixth position (A→T) (Figure 3G), a minor fraction of the ZNF649 binding motif had the same base composition at these sites (Figure 3D). Thus, these two mismatches may not abrogate ZNF649 binding.” p11, 181-185

    1. Line 244: "More than 90% of full-length SVA_B-F copies could be analyzed by SVA amplicon-seq". What is the basis for this calculation? Presumably the amplicon-seq doesn't give information as to where on the genome the SVA resides.

    The criterion of analyzed copy is more than 10 CpGs within each copy are covered by at least five reads. I mapped computationally generated 100-bp reads from full-length SVA copies to investigate how much reads from each copy are uniquely aligned to genome. Even in the youngest SVA type, SVA_F, more than 10% of reads can be uniquely mapped in the most of copies (please see New Supplementary Fig .S1A).

    Was this evaluation helpful?
  2. Evaluation Summary:

    The majority, but not all retrotransposons undergo massive reprogramming of their methylation states during germ cell development. This manuscript tests the contribution of binding motifs for KRAB-Zinc Finger Proteins (KZFPs) and the position of retrotransposons relative to genes to explain the variable methylation dynamics of different retrotransposon families, namely L1, SVA and LTR12, as well as potential inter-individual variation during male germ cell development in humans, using an integrative analyses of available sequencing datasets. By bringing insights into the complex regulation of retrotransposons, this study could be of particular interest to the epigenetics community. Some additional analyses would strengthen the inferences made.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 and Reviewer #2 agreed to share their name with the authors.)

    Was this evaluation helpful?
  3. Reviewer #1 (Public Review):

    Fukuda et al investigated patterns of DNA methylation in human male germ cells, aiming to explain why some elements in the human genome are resistant to the global wave of DNA demethylation occurring during primordial germ cell differentiation. Many of these are transposable elements of various kinds, notable targets of repression mechanisms in germ cells to prevent their spread between generations of individuals. The authors have found links between subfamilies of transposable elements and the presence of binding sites for specific KRAB zinc finger proteins, a large family of epigenetic regulators. The authors then further dissect these findings, with correlation to differentially methylated loci between individuals in sperm.

    The study is well designed and although much of the initial analysis was done with a limited sample size (sperm from 2 donors), the important conclusions have been replicated in a second dataset with a few more donors, and further confirmed them with independent experiments by amplicon sequencing. The result are compatible with current knowledge of the KZFP family, as they are known to affect DNA methylation levels in stem cells. The findings related to interindividual DNA methylation level differences are interesting, especially in regard to their potential to trigger biological problems such as infertility. This is also somewhat related to recent publications in the field on metastable epialleles associated with transposable elements in both mice and human, a few of which have been linked to KZFPs so far.

    There is a few weaknesses, but they are appropriately discussed by the authors themselves. For example, they acknowledge that they are working with an incomplete set of KZFP binding factors binding sites - the family contains 350 protein-coding genes in humans, with data for binding for approximately 230 of them. Also, the sample size is low and some of the wording related to conclusions could be toned down to. Finally, most conclusions of the paper are purely descriptive / correlative - again, the authors acknowledge this in a late section of the manuscript, but some sentences in other parts are worded in a way that could lead some readers that these are facts and not speculation.

    Nonetheless it is reporting an important and novel finding that expands our understanding of the ways KZFPs can subtly affect human biology and paves the way for further studies.

    Was this evaluation helpful?
  4. Reviewer #2 (Public Review):

    Fukuda et al. use whole genome bisulfite sequencing (WGBS) and RNA sequencing (RNA-seq) data obtained from sperm and human primordial germ cells (hPGCs), as well as KRAB-ZFP protein ChIP-seq data obtained from HEK293T cells, to study the relationship between DNA methylation, KRAB-ZFP binding and genome-wide transcription of LINE-1 (L1), SVA and LTR12 retrotransposons.

    This work aims overall to elucidate pathways silencing retrotransposons in the male germline, in particular making new (and known) links between ZFPs and DNA methylation. The focus here ends up being on immobile retrotransposons, as L1s (bound by ZNF93 and ZNF649) and SVAs (bound by ZNF28 and ZNF257) capable of mobilisation either do not have binding sites for the identified ZFPs, or have far fewer than their older relatives. The relationships between L1, ZNF93 and ZNF649 has been reported previously (Jacobs et al., 2014, Nature; Fernandes et al., 2018, bioRxiv). That older retrotransposons have more binding to these ZFPs, and are more methylated in hPGCs, is based mainly on correlation. Overall, I thought the subject matter was interesting but I have substantial reservations around the analyses, particularly the more novel results related to SVA. As the work stands it is not clear whether ZNF257 or ZNF28 are reinforcing DNA methylation on SVA. The claims around this repression being transcriptionally-directed or varying significantly amongst individuals, for biological as opposed to technical reasons, appear preliminary at this stage.

    Specific comments:

    1. The use of WGBS to analyse very young retrotransposons, like L1HS and SVA_F, has potential caveats. One of the most important of these is that the CpG islands most likely to be differentially methylated for these elements, in somatic cells at least, are internal to their sequences (e.g. PMID: 33186547). This includes the SVA VNTR sequence, which is where the vast majority of proposed ZNF28 and ZNF257 binding motifs reside (Fig. 2F). Does WGBS, using only uniquely mapped reads as done here, resolve these regions sufficiently to identify differential methylation?

    2. Why does Fig. 1D have only 36 full-length L1HS copies? The definition of a full-length L1 here (>90% consensus length) should yield (from memory) >300 reference L1HS copies.

    3. It would be useful to explain the inclusion of LTR12 as a representive ERV, as opposed to, say HERVK, which has been studied in hPGC like cells recently (PMID: 35075135).

    4. The exo ChIP-seq for a variety of ZFPs was obtained from published data generated using HEK293T cells, whereas the WGBS is from sperm and hPGCs. What evidence can the authors point to to be reasonably sure that the ZFP binding patterns from HEK293Ts carry over to the male germline in vivo?

    5. In Fig. 3C it appears quite a few L1PA3s have ZNF649 peaks and yet the motif for ZNF649 has two mismatches to the L1PA3 consensus (Fig. 3F). Yes, L1HS has one more mismatch than L1PA3. It would be useful to explain further why two mismatches are acceptable whereas 3 completely abolishes binding.

    6. Line 244: "More than 90% of full-length SVA_B-F copies could be analyzed by SVA amplicon-seq". What is the basis for this calculation? Presumably the amplicon-seq doesn't give information as to where on the genome the SVA resides.

    Was this evaluation helpful?