Recent @CellCellPress study claiming HERVK transcription drives senescence has a fundamental RNA-seq analysis probl… https://t.co/gfVQCsIC7x
Note: most HERVK copies on the genome exist as so-called solo LTR5_Hs elements, where recombination has eliminated… https://t.co/nWfOC4LnYE
At the outset, Fig. 1C shows a critical RNA-seq heatmap of late (LP) vs early passage (EP) human mesenchymal progen… https://t.co/VCQfBH1JKM
Immediate impression: 1) 20% is a minor change and incongruent with the immunoblots shown later. 2) LTR5_Hs, which… https://t.co/XRo9fCsX2h
Curious about this concerning heatmap, I downloaded the WT LP vs EP RNA-seq datasets and analysed these with STAR/T… https://t.co/oCvJdZSySb
Genome-wide HERVK is 5% higher in the LP cells, and for HERVK-int I see an absolute difference of 9%. LTR5_Hs is vi… https://t.co/denNL7aAi5
Next, I used the IGV to look at the most highly expressed HERVK copies. These predominately involve "read-through"… https://t.co/kj3oPUPqgP
The high intronic RNA-seq signal in PRAG1 got my attention, so I looked next at the CDKN2A (P16) locus and again we… https://t.co/sJ2UZAQ6MZ
The RNA-seq seems to have been via rRNA depletion, and featureCounts QC suggests ~35% of the reads align to introns… https://t.co/senpPDMzQM
Locus-level analyses of HERVK transcription are not shown anywhere in the paper. These are feasible with short-read… https://t.co/CY2FeuXUGQ
To explain HERVK "upregulation", WGBS is then done (Fig. 1K). A highly significant (***) difference in methylation… https://t.co/4hcfKXho7F
In the supplement (Fig. S1K), a similar WGBS analysis is performed for HERVK "proviral loci" but the differences ar… https://t.co/CXYpyYnX1x
Surprisingly, an analysis of the methylation-sensitive LTR5_Hs promoter is not shown. Why? That is the key sequence… https://t.co/l6ww9qf8g6
Fig. 1 provides the rationale for the downstream experiments (e.g. perturbing HERVK activity, the caveats of which… https://t.co/x4fgQph5fB
A couple of papers that would be worth referring to here. Firstly, this excellent review from @sophie_lanciano and… https://t.co/WQ159aVBut
Secondly, @tashjansz and I wrote this review on ERVs in cancer but the concepts are also applicable to studies of a… https://t.co/8LMcij1yzD
It is imperative before making big claims about ERVs in biology to use the right tools to measure their activity. L… https://t.co/NaOghcwhPb
Without these and other foundational experiments, particularly when the data seem problematic on closer inspection,… https://t.co/a4PydfP5sN
We thank Dr. Faulkner for his comments.
1) As for the comment about differences in expression and methylation levels of HERVK being not significant between “EP” and “LP”, we would like to clarify here that the LPs (P9, pre-senescent) in WT hMPCs subjected to RNA-seq and WGBS were not fully senescent. Normally, WT hMPCs were replicatively senescent in their late passages (LP, P > 12), whereas prematurely aged HGPS and WS hMPC models exhibited growth arrest at passages 8 or 9 (termed LP in prematurely aged models).
Since HGPS and WS-hMPCs senesce much earlier than their WT counterparts, when comparing all 6 cell lines, we frequently used P3 as the EP for WT, HGPS and WS hMPCs and P9 as the LP for WT, HGPS WT hMPCs, but P9 WT hMPCs are still pre-senescent. We thank Dr. Faulkner for noting that in Figure 1 C in the online version, compared to EP (P3), the LP (P9, pre-senescent) WT hMPCs did not show dramatic transcript changes in HERVK, which is indeed in line with the notion that the LP (P9) WT hMPCs are not that “aged”. Therefore, for the sequencing data, the LP in WT hMPCs were at passage 8 or 9, at which point it should be the middle passage (pre-senescent) for WT hMPCs. This is why the expression and methylation differences of HERVK were not as marked between the EP (P3) and LP (P9, pre-senescent) of WT hMPCs. If we compare young cells (EP (P3) WT) and “real” senescent cells (LP (P9) WS) in RNA-seq, we can see a striking difference in HERVK (Figure 1). However, when we performed immunoblotting in “replicatively senescent (RS)” hMPCs alone, LP WT hMPCs were referred to as P > 12, at which point WT hMPCs displayed a more pronounced senescent phenotype. Because of the word limit of the main text, we moved to this definition in the Methods section.
2) Dr. Faulkner mentioned that based on the RNA-seq data, there was a limited increase in genome-wide HERVK. The fold-change values may vary due to mapping, counting, and normalization methods by different analysis tools and versions. In our study, we used STAR and TEtranscripts to map and quantify the expression level of repetitive elements. Then, the differential analysis between LP (P9, pre-senescent) and EP (P3) was performed by DESeq2 (in TEtranscripts pipeline). The difference [loga(fold change)] in HERVK-int between LP (P9, pre-senescent) and EP (P3) WT hMPCs was 0.273528592 (Figure 1). The P value was 2.68E-08, and the adjusted P value was 2.07E-07. Again, we would like to emphasize that for WT cells, LP (P9) is still pre-senescent yet a much higher increase in HERVK is observed in “real” senescent cells, i.e., LP (P9) WS hMPCs and HGPS hMPCs (Figure 1).
3) In addition, we also implemented multiple software programs to calculate the expression level changes of repetitive elements, including RepEnrich2 (PMID: 25012247) and Software for Quantifying Interspersed Repeat Expression (SQuIRE) (PMID: 30624635) (Figure 2). Using these RE analysis approaches, we found that HERVK-int exhibited uniformly increase expression in various senescence models, albeit the expression level changes varied from different software (Figure 2).
4) We also appreciate the comments about LTR5_Hs. We separated RepeatMakser-annotated HERVK elements into two classes, 1) HERVK elements with LTR5_Hs promoters and 2) HERVK elements without LTR5_Hs promoters, and calculated the expression levels of HERVK elements for each group. Interestingly, we found that the expression levels of HERVK elements without LTR5_Hs promoters remained relatively stable during senescence, whereas the expression levels of HERVK elements with LTR5_Hs promoters were dramatically upregulated in senescent hMPCs (Figure 3). These data indicated that the HERVK elements from solo LTR elements may not contribute to the elevated expression levels identified during hMPC aging.
5) As a matter of fact, with particular care, we also analyzed the expression levels of REs in a locus-specific manner using 5 different software programs (i.e., Telescope (PMID:31568525), SQuIRE (PMID: 30624635), TEtranscripts (PMID: 26206304), TElocal (https://github.com/mhammell-laboratory/TElocal), and ERVmap (PMID: 30455304)) based on multiple RE annotations (i.e., RepeatMasker annotation, Telescope HERV annotation, ERVmap annotation, and custom annotation which incorporated the Telescope annotation into the RepeatMasker annotation), and they all point to the notion that HERVK HML2_1q22 is consistently upregulated in senescent hMPCs (Figures 4-8). We also verified the increased level of HML2_1q22 in senescent hMPCs using a polyclonal antibody against HML2_1q22 (HML2_1q22 provirus ancestral Env polyprotein, ERVK-7 (Endogenous Retrovirus group K Member 7 Env polyprotein), #302427, United States Biological) (Figure 9). These results imply that HERVK HML2_1q22 is a potential senescence-sensitive locus, which awaits to be further validated by third-generation long-read sequencing and more mechanistic studies.
6) For the comments on the analysis of the methylation-sensitive LTR5_Hs promoter, in fact, we have analyzed the methylation levels of LTR5_Hs, HERVk-int, and HERVK-int flanked with LTR5_Hs. We found that both the methylation levels of HERVK-int and those flanked by LTR5_Hs were decreased in replicatively senescent hMPCs and prematurely senescent hMPCs (Figure 10). Although the methylation levels of LTR5_Hs were slightly higher in HGPS than in WT, they also decreased with passage (EP (P3) vs. LP (P9) HGPS) (Figure 10). This could be caused by the HGPS-specific biological events (e.g., progerin (a mutant form of Lamin A)-associated epigenetic changes). Since the HGPS-specific effect is beyond the focus of this study, we mainly focused on analyzing the methylation levels of HERVk-int.
In addition, the statistical test used for this analysis is paired student’s t-test. Our previous legend is “Violin plot showing the CpG DNA methylation levels (mCG/CG) for HERVK-int in RS (LP (P9, pre-senescent) vs. EP (P3) [WT]) and prematurely senescent hMPCs. The white dots represent the median values, and the white lines represent the values within the interquartile range (IQR) from smallest to largest. **p < 0.01, ***p < 0.001 (paired t test)”. In the final version, because our word count exceeded the limit, as the editor requested, many cuts were made to the legend. The strategy to analyze the DNA methylation level changes in cellular aging was included in the Methods section, and the statistical analysis was summarized in the “QUANTIFICATION AND STATISTICAL ANALYSIS” section. HERVK provirus was defined in a previous study (PMID: 22067224). We also show one example of HERVK HML2_1q22. We observed a reduced methylation level of HERVK HML2_1q22 in senescent cells (Figure 11).
7) We really apologize for missing these important references and the right tools in the final versions of this paper (they were present in the earlier versions but were accidentally removed during multiple rounds of revisions when we prepared the manuscript). Additionally, we completely agree about long-read nanopore sequencing, which is being performed by our team (being suggested in the “Research Limitations” section of the paper). In addition, we also thank Dr. Faulkner for his insightful comments on the "read-through" non-coding transcription or high intronic signal. We did observe this when analyzing the data. It may be associated with aging-associated “read-through”. Although it is a very interesting topic, it also goes beyond the scope of this paper.
Although we have already realized the limitations of second-generation RNA-seq in the analysis of genomic repetitive elements because of majorly enriching/capturing short read lengths of RNA fragments (we discuss it in the “Research Limitations” section of the paper), we still thank this approach for leading us to discover aging-associated upregulation of HERVK as starting immature evidence. Therefore, we turned to multiple layers of gene expression analyses, including RNA FISH, DNA FISH, immunostaining, immunoblotting, etc. All these gene expression results strongly support our conclusion that HERVK is substantially upregulated during cellular and tissue aging.
Finally, we sincerely thank Dr. Faulkner for those useful comments and discussions. Since most of the authors of the paper are technically inconvenient to communicate on Twitter, for any colleagues who are interested in the bioinformatics part of the paper, please feel free to contact me via email. Thanks a lot! My email: zunpengliu@163.com.
We sincerely thank Dr. Faulkner for all those useful comments and discussions. Please see our full responses in the attachment.
Figures can be found here:
https://www.dropbox.com/s/558f8dxjn256c10/To Professor Geoff Faulkner‘s comments- Figures.pdf?dl=0
Thanks a lot!
Attach files by dragging & dropping,
selecting them, or pasting
from the clipboard.
Uploading your files…
We don’t support that file type.
with
a PNG, GIF, or JPG.
Yowza, that’s a big file.
with
a file smaller than 1MB.
This file is empty.
with
a file that’s not empty.
Something went really wrong, and we can’t process that file.