An Orphan CpG Island Drives Expression of a let-7 miRNA Precursor with an Important Role in Mouse Development

Round 1

Reviewer 1 Report

This study by Adrian Bird and his colleagues considered CpG islands that were not known to be associated with annotated gene promoters, but could be associated with the expression of non-coding RNAs. In their screen performed in embryonic stem (ES) cells, they discovered several candidates, including a CpG island that was linked to a long transcript (called KY467470) that expresses three microRNAs of the let-7 family (let7a-1, let-7f-1 and let-7d). They studied the expression and evolutionary conservation of this lncRNA transcript and that of the three let-7 microRNAs, that have the same seed sequence. They then set out to explore its function by targeted the CpG island in ES cells, followed by the derivation of chimaeric mice and a knockout mouse line.  Interestingly, homozygous knockout mice showed reduced viability with growth and metabolic defects, most likely due to the loss of the lncRNA produced let-7 microRNAs.  The precise origin of this phenotype is not further explored in this study, but is quite interesting given that all the let-7 microRNAS have the same seed sequence..

The presented data are clear and of high quality. The new findings on KY467470, its let7 microRNAs and their roles in mouse development will be of interest to the field, and this reviewer has a few minor suggestions only:

1)      Where in the transcript is positioned the RT-PCR primer pair that was used to show the loss of KY467470 expression in the homozygous animals? Does the deletion of the CpG island ablate all transcription across the ~15-kb region along which RNAseq signal was detected in ES and differentiated cells (as shown in figure 1A. Can this be ascertained by using additional primer pairs?

 

2)      In both the female and the male homozygous animals, an increased gain in weight is observed between weeks 4 and 16 (Figure 3J, K), compared to WT controls. The authors show that this correlates with metabolic marker changes, but it would be important to explore to what extent this phenotype could be linked to altered food intake as well. Could the authors present data, or comment on this possibility?

 

3)      Several lncRNAs that are functionally important have been shown to have a particular nuclear localisation. The authors show by RT-PCR amplification that KY467470 RNA is enriched in the nucleus.  Besides presence at the transcription site, does the RNA show a particular sub-nuclear localisation? This could be explored by RNA-FISH.

Author Response

Reviewer 1

Comments and Suggestions for Authors

The presented data are clear and of high quality. The new findings on KY467470, its let7 microRNAs and their roles in mouse development will be of interest to the field, and this reviewer has a few minor suggestions only:

1)      Where in the transcript is positioned the RT-PCR primer pair that was used to show the loss of KY467470 expression in the homozygous animals? Does the deletion of the CpG island ablate all transcription across the ~15-kb region along which RNAseq signal was detected in ES and differentiated cells (as shown in figure 1A. Can this be ascertained by using additional primer pairs?

We have added information about the location of these qPCR primers to the manuscript and have also added additional qPCR data using primer pairs further downstream (see Fig. 1A, 3B). These data confirm that CpG island deletion causes the transcript to be severely depleted along the length of the region showing an RNAseq signal.

2)      In both the female and the male homozygous animals, an increased gain in weight is observed between weeks 4 and 16 (Figure 3J, K), compared to WT controls. The authors show that this correlates with metabolic marker changes, but it would be important to explore to what extent this phenotype could be linked to altered food intake as well. Could the authors present data, or comment on this possibility?

We have not analysed these KO mice using metabolic cages and therefore do not know if there are any alterations to their food intake or activity. We note however that Let-7 has been previously implicated in regulation of metabolism and have now added a brief discussion of this point (p7, 1st paragraph). It is likely therefore that the defect is due to metabolic defects rather than appetite, but we agree that future work is needed to specifically test this.

3)      Several lncRNAs that are functionally important have been shown to have a particular nuclear localisation. The authors show by RT-PCR amplification that KY467470 RNA is enriched in the nucleus. Besides presence at the transcription site, does the RNA show a particular sub-nuclear localisation? This could be explored by RNA-FISH.

We did not test whether the RNA forms a “cloud” in the nucleus, similar to e.g. Xist, or is localised in other ways. As the RNA levels of the flanking two genes were unchanged in its absence, we don’t expect that KY467470 forms a silencing compartment, but primarily serves as a precursor for the 3 miRNAs. While of interest, we do not consider that knowledge of intra-nuclear localisation would affect the conclusions of our study.

Reviewer 2 Report

1.    (Fig 1) Although you have previously identified this region as a CpG island, is there evidence that it is differentially methylated? It would help to see bisulfite sequencing next to your strand specific RNA-sequencing to show.  

2.    (Fig 2C/D) If the cap helps to facilitate nuclear export and translation by the ribosomes, why would you expect that they remain untranslated and localized to the nucleus? Please provide further data to support a reason for a miRNA being capped but retained in the nucleus.     

3.    (Fig 3C/F/I) This is very interesting to see that the effect of deleting one locus of let7a/f has varied effects on expression levels. However, in considering the number and overall physiology of the homozygote knock out mice, it would be interesting to perform some invitro assays to determine how deleting the other loci of let7a/f on cellular physiology, especially for metabolism. Since there appears to be a metabolic phenotype associated with the knockout. 

4.    (Fig 3H) This should be an entire figure to itself or as a supplement. Instead of displaying this as a checklist, it would be very compelling to show the actually phenotypical data for each category. 

5.    (Fig 3 H) It would also help to show the survival curve comparing the knock out to the WT.

6.    (Fig3 I) As there are a very few homozygous knock-out mice, you should assess the embryonic phenotype of these offspring. Are the KO embryos larger than the WT during development to a certain point? 

7.    (Fig 3 and Fig 4) because of the obvious differences in body weight, size, and lipids, you should show expression data for metabolic genes. Also indicate any publications that have previously identified a role for let7 in metabolism. I am sure that this would have implications for cancer cell homeostasis. 

8.    (Discussion line 235) this should be the explanation for Figure 2D     

9.    To further validate the in vivo data that you have shown here, it would be helpful to go back to the ESCs and other cells used in your previous publication and figure 1

Author Response

Reviewer 2

Comments and Suggestions for Authors
1.    (Fig 1) Although you have previously identified this region as a CpG island, is there evidence that it is differentially methylated? It would help to see bisulfite sequencing next to your strand specific RNA-sequencing to show.  

Publicly available DNA methylation data indicate that this CGI is unmethylated in all tissues tested, which also agrees with the widespread expression of the transcript reported here. We have added this information and source reference to the manuscript (p4, 1st paragraph).

2.    (Fig 2C/D) If the cap helps to facilitate nuclear export and translation by the ribosomes, why would you expect that they remain untranslated and localized to the nucleus? Please provide further data to support a reason for a miRNA being capped but retained in the nucleus.

We would not expect the miRNAs themselves to be capped. We have only tested KY467470, the precursor RNA for its capping status. It has been shown for other ncRNAs like Xist or Airn that those transcripts are capped and polyadenylated, but retained within the nucleus. Therefore there are established precedents for capped RNA to be retained in the nucleus.

3.    (Fig 3C/F/I) This is very interesting to see that the effect of deleting one locus of let7a/f has varied effects on expression levels. However, in considering the number and overall physiology of the homozygote knock out mice, it would be interesting to perform some invitro assays to determine how deleting the other loci of let7a/f on cellular physiology, especially for metabolism. Since there appears to be a metabolic phenotype associated with the knockout.

This indeed would be very interesting, but is beyond the scope of this manuscript. We anticipate that future work will look into this issue. We now cite previous evidence implicating Let-7 in metabolic regulation (p7, 1st paragraph).

4.    (Fig 3H) This should be an entire figure to itself or as a supplement. Instead of displaying this as a checklist, it would be very compelling to show the actually phenotypical data for each category.

Showing all data in the manuscript would mean a very long list. To make this information accessible, we have now added a link to this mouse line on the IMPC webpage (see last paragraph of Results; p5/6) which gives a comprehensive list of all parameters analysed. We also discuss the different assays for significance adopted by IMPC and in our study (described in Materials and Methods; p12, 3rd paragraph), which leads to some extra phenotypes being called on the IMPC website. Briefly, our calling relies on comparison to a reference phenotype and is somewhat more conservative than that used by IMPC.

5.    (Fig 3 H) It would also help to show the survival curve comparing the knock out to the WT.

We only know that homozygous KO animals are found at reduced numbers at P14, but are present at expected Mendelian ratios at E14.5. Therefore, the lethality affecting a proportion of KO mice must occur between E14.5 and P14. To explain this more fully, we have added a new panel to Fig. 3 (Fig. 3J) and a sentence into the manuscript (p5, 2nd paragraph). Future work will shed more light on the precise timeframe during which the lethality occurs.

6.    (Fig3 I) As there are a very few homozygous knock-out mice, you should assess the embryonic phenotype of these offspring. Are the KO embryos larger than the WT during development to a certain point?

At E14.5, homozygous KO embryos appear phenotypically normal, which we now also have mentioned in the manuscript (p5, 2nd paragraph).

7.    (Fig 3 and Fig 4) because of the obvious differences in body weight, size, and lipids, you should show expression data for metabolic genes. Also indicate any publications that have previously identified a role for let7 in metabolism. I am sure that this would have implications for cancer cell homeostasis.

In-depth characterisation of metabolic genes affected by this mutation is beyond the scope of this study. We have, however, added a brief discussion of previous work that implicated Let-7 in metabolic regulation (Discussion, p7, 1st paragraph). In particular, we thank the reviewer for pointing out that over-expression of a negative regulator of Let-7 processing also leads to increased body size, most probably via effects on metabolism.

8.    (Discussion line 235) this should be the explanation for Figure 2D    

The reviewer is correct. To make this clearer, we have now added a sentence to the Discussion.

9.    To further validate the in vivo data that you have shown here, it would be helpful to go back to the ESCs and other cells used in your previous publication and figure 1

Analysing the effect of the KO in different mouse tissues is beyond the scope of this study, but may be addressed by future work.

Reviewer 3 Report

In this short manuscript, Koerner and colleagues study how an orphan CpG island (CGI) acts as a promoter and drives expression of let-7 microRNAs. Using transgenesis in the mouse, they deleted this element and confirmed an effect on the expression of at least 2 different let-7 microRNAs. Although most microRNA are thought to be redundant, the deletion resulted in many phenotypes (e.g. viability, blood lipids). I have the following comments:

 

1. The authors mention that the link between the orphan CGI and KY467470 was identified by visual inspection of the data. Were they particularly lucky to find this locus, or are they many similar CGI linked to “novel” transcripts in the mouse genome? The RNA-seq data is at hand so it should not be very difficult to develop a simple algorithm to quantify their observation across the mouse genome.

 

2. For the results presented in part of Fig3 and in Fig4, there is a discussion of significant differences, but I can’t find statistical evidences (test statistics, p-values, etc)?

 

3. What was the rationale for looking at the levels of gamma-delta T-cell over other cell types?

 

4. I understand that a large number of phenotypes were compared between WT and mutant mice. Was that number of phenotypes considered when reporting significant differences (i.e. accounting for the multiple testing problem)?

 

5. The notion that let-7 microRNA affects body size is not novel as it has been shown in mice that disruption of Lin28, which is required for let-7 biogenesis, impacts growth (PMID: 20512147). This should be discussed in the manuscript.

 

6. MINOR. Line #96. I think the authors referred to Fig1C instead of 3A. Line #111, Fig3F-G instead of Fig3E-F. Fig3I is discussed before Fig3H.

Author Response

Reviewer 3

Comments and Suggestions for Authors
In this short manuscript, Koerner and colleagues study how an orphan CpG island (CGI) acts as a promoter and drives expression of let-7 microRNAs. Using transgenesis in the mouse, they deleted this element and confirmed an effect on the expression of at least 2 different let-7 microRNAs. Although most microRNA are thought to be redundant, the deletion resulted in many phenotypes (e.g. viability, blood lipids). I have the following comments:

1. The authors mention that the link between the orphan CGI and KY467470 was identified by visual inspection of the data. Were they particularly lucky to find this locus, or are they many similar CGI linked to “novel” transcripts in the mouse genome? The RNA-seq data is at hand so it should not be very difficult to develop a simple algorithm to quantify their observation across the mouse genome.

Expression of transcripts from orphan CGIs is quite common in the RNA-Seq dataset. We chose to focus this manuscript on a specific example rather than including the results of our genome-wide analysis. In our opinion a discussion of these global genomic issues would distract from the specific problem addressed by the present study.


2. For the results presented in part of Fig3 and in Fig4, there is a discussion of significant differences, but I can’t find statistical evidences (test statistics, p-values, etc)?

We now have added information about the statistical analysis in the Methods section (p12).

3. What was the rationale for looking at the levels of gamma-delta T-cell over other cell types?

We analysed an entire panel of immune cells for a phenotype and gamma-delta T-cells where the only ones which had a phenotypic difference. To make this clearer, we have now inserted a sentence into the manuscript (p5, last paragraph before Discussion). The IMPC database lists additional phenotypes, as now reported in the last Results paragraph.

4. I understand that a large number of phenotypes were compared between WT and mutant mice. Was that number of phenotypes considered when reporting significant differences (i.e. accounting for the multiple testing problem)?

Information about the statistical analysis now has been added to the Methods section (p12).


5. The notion that let-7 microRNA affects body size is not novel as it has been shown in mice that disruption of Lin28, which is required for let-7 biogenesis, impacts growth (PMID: 20512147). This should be discussed in the manuscript.

We now have added consideration of this issue to the Discussion (p7, 1st paragraph).

6. MINOR. Line #96. I think the authors referred to Fig1C instead of 3A. Line #111, Fig3F-G instead of Fig3E-F. Fig3I is discussed before Fig3H.

Thanks – these errors have now been corrected.

Round 2

Reviewer 3 Report

No additional comments.