Telomeres reforged with non-telomeric sequences in mouse embryonic stem cells

Telomeres are part of a highly refined system for maintaining the stability of linear chromosomes. Most telomeres rely on simple repetitive sequences and telomerase enzymes to protect chromosomal ends; however, in some species or telomerase-defective situations, an alternative lengthening of telomeres (ALT) mechanism is used. ALT mainly utilises recombination-based replication mechanisms and the constituents of ALT-based telomeres vary depending on models. Here we show that mouse telomeres can exploit non-telomeric, unique sequences in addition to telomeric repeats. We establish that a specific subtelomeric element, the mouse template for ALT (mTALT), is used for repairing telomeric DNA damage as well as for composing portions of telomeres in ALT-dependent mouse embryonic stem cells. Epigenomic and proteomic analyses before and after ALT activation reveal a high level of non-coding mTALT transcripts despite the heterochromatic nature of mTALT-based telomeres. After ALT activation, the increased HMGN1, a non-histone chromosomal protein, contributes to the maintenance of telomere stability by regulating telomeric transcription. These findings provide a molecular basis to study the evolution of new structures in telomeres.

the shift of copy number alteration profiles from a subclonal mix in the pre-ALT samples to a monoclonal architecture in the post-ALT samples to characterize the evolutionary bottleneck the initiation of ALT represents. Overall the manuscript summarizes an impressive body of work that will drive the field forward, but several aspects of the text require further consideration.
Major comments: -Have you provided a complete sequence of the mTALT elements in the supplementary data? If possible, do so to augment the following result: Line 141: "A terminal restriction fragment (TRF) assay revealed that only post-ALT cells showed discrete bands resulting from telomere non-cutter enzymes" Annotate the restriction sites in the sequence of mTALT and list the expected fragment sizes. Does these predictions match the some of the observed bands? -You nicely used the SNP allele frequencies to identify the chr11 locus as the active locus. Beyond that, is the SNP data sufficient to decide if both alleles of the chr11 mTALT locus are used for ALT or only one allele? Be more explicit if you can clarify this point or if the alleles are too similar to make a conclusion.
-Line 227-228: "The amount of various repetitive sequences including rRNA did not significantly (Extended Data Fig. 3c)." Sentences is incomprehensible without the verb, and will change its meaning depending on which verb you will add. Completion necessary. -Line 235: You indicate that "certain cells" were selected upon ALT activation. Do you have any reason to believe that this was not a monoclonal event happening in a single cell that then expanded? It would be helpful if you would discuss the possible alternatives in the light of the data you have gathered.
-In human cancer inactivation of the proteins ATRX and DAXX are frequently observed in association with ALT. You should briefly mention in one or two sentences if their mouse homologs are affected or not in your proteomics data. Atrx for instance is featured quite prominently in figure 5b, and appears to be as affected as Hmgn1, and is after FDR correction close to significance in the Volcano plot in Figure 5a. -In line 152 and line 340 you mention the "right part" of the chromosomes 13 and 11, respectively. You may be more precise here using terms such as long arm / short arm and telomeric side / centromeric side etc.. In the current form it is easy to misunderstand these sentences.
-Line 362-364: "We showed that the actual ALT template sequence form heterochromatic telomere and promoted the transcription from insulator motif toward centromere." This sentences is incomplete and does not make sense in the current form. Reformulate! -Does the fact that some mouse strains have the chr11 copy of mTALT while other have not impact the likelihood that the former evolve ALT compared to the later? Addressing this question in the Discussion may inspire follow-up analysis.
-For many of the computational tools used in the analysis you mention the versions, but do not cite the related scientific publications. This includes BWA-mem, Picard-Tools, Samtools, MACS2, DESeq2, Computel, TelomereHunter and Telomerecat. I would suggest that you carefully check for which of the third party software you have applied scientific publications exist which you then may cite in the manuscript.
- Figure 1 c in the Extended Data Appendix indicates that PD800/PD100 log2 ratio shows an enrichment of TGAGGG and TTCGGG singleton repeats above the expected log2 ratio. This is in contrast to the statement you make in line 146-149, where no differences are reported. Reformulate this part to match the observations you do report.
Minor comments: -The first sentences of the abstract appears unprecise. How about: "Telomeres are part of a highly refined system for maintaining the stability of linear chromosomes." -Line 76-78: "but ~15% of human cancer cells are known to maintain telomeres by a telomeraseindependent telomere maintenance mechanism, which is called the alternative lengthening of telomeres (ALT)" A more precise summery would be: "but ~15% of human cancer cells are known to maintain telomeres by telomerase-independent telomere maintenance mechanisms, which are summarized under the term alternative lengthening of telomeres (ALT)" -Line 346-347: "telomere shortening dependent-chromosomal fusions" appears to be an odd use of the hyphen. Do you mean "telomere-shortening-dependent chromosomal fusions"? -Sometimes upper and sometimes lower case is used to refer to figures. Decide for one form.
-In figure 5 the first panel is named aa instead of a.
Reviewer #3 (Remarks to the Author): The authors provide evidence showing that, in mammals, telomeres can be reconstituted with a nontelomeric unique sequence (mTALT) used for repairing DNA damage at telomeres and for adding new telomeric sequences in ALT mESs. Using MS and epigenomics data before or after ALT activation, the authors show the expression of a non-coding mTALT transcript. After ALT activation, they find that HMGN1 protein levels increase and contribute to telomere stability through telomeric transcription. They propose that the non-canonical telomeric sequence has a functional role in cancer and evolution.
Overall, the data are interesting and for the most part the conclusions are supported from the observations shown. What is missing is further evidence on the functional role of HMGN1 in mTALT transcription and on R-loops formation a well as on the specificity of the findings to HMGN1 itself. The work would also benefit from the addition of few important controls as well as from a more detailed explanation on the proposed working model (in view of recent findings).
Specifically: -the authors provide no evidence on the specificity of HMGN1 in telomere maintenance. This is important because their MS list (Fig 5) also includes other members of the HMG group i.e. HMGA, HMGN3. Would knocking down these protein targets lead to similar findings (e.g. R-loops, mTALT transcription, etc.)?
-is the effect of HMGN1 on mTALT transcription specific? Other targets (within and outside this region) should be included.
-On Figure 5L, the authors need to include an RNAaseH1 treatment to control for the presence of Rloops.
-The connection of HMGN1 to R loops (or in the resolution of R-loops) remains unclear in the present work. Is HMGN1 recruited to R-loops? An HMGN1 ChIP followed by DRIP could further highlight this point.
-Is depletion of HMGN1 leading to an increase in γH2AX-associated DNA damage? If so, is the increased DNA damage levels due to R-loops? An RNaseH1 treatment in shHMGN1 cells could help answer this question. -R-loops are known to be generated during abrupt changes in transcription demands or when transcription is blocked due to DNA damage. This should be discussed with respect to mTALT transcripts in shHMGN1 cells.
Other comments: Figure 5b/Suppl. Figure 4a-c: The GO numbers of GO terms are not shown. Please provide the GO number for all GO terms shown. Figure 5b: Is the heatmap shown referring to the number of peptides or is this a log2 fold change as shown in figure 5a? The term "count" should be explained in the figure legend. Figure 5d: the authors should validate the protein levels of a few more mass spec targets (beyond HMGN1). It is not entirely clear why the authors have chosen to focus on HMGN1 alone or why they have neglected other factors e.g. HMGA or HMGN3 in the MS list that seem equally relevant. Figure 5f: do the three independent replicates mentioned represent biological or technical replicates? This information should be provided here and elsewhere in the manuscript. Figure 6: please provide a legend that summarizes what is shown in the figure.

Reviewer #1
Q: Overall, the figures are well presented, and it is an interesting study. I found the manuscript quite hard to follow, and the text requires substantial editing. The phenomenon of non-telomeric sequences being propagated in ALT telomeric regions is not totally novel, and I feel the interpretation that this may represent a new ALT cancer type is overstated.
A: We agree with the reviewer's comment. We have extensively revised the manuscript and employed a professional editing company (ENAGO) to edit our manuscript. We also toned down our discussion on the possibility of a new ALT cancer type. A more detailed description of our revision is provided below.
Major points: Q1. In the abstract, the authors state that "we show that mammalian telomeres can also be completely reconstituted using a non-telomeric unique sequence", whilst later they say that "canonical telomeric repeats were duplicated along with mTALTs". I don't think that it is correct that the mTALT sequence "constructs" new telomeres, I think it is more that the sequence becomes propagated through the telomeres. This should be clarified/discussed. I feel that statements eluding to a new ALT cancer type are not sufficiently supported by the data/phenotype. The presence of telomeric sequences within the mTALT sequences is also indicative of the cells activating ALT pathways, but mTALT sequences infiltrating the terminal regions and then becoming amplified. A1. Thank you for your constructive comments. As you pointed out, telomeres of ALT mESCs do contain telomeric repeats as well as mTALT sequences. We did not intend to present mTALT as the sole component of ALT telomeres, but some expressions we used would be confusing to readers. As you adequately pointed out, mTALTs were replicated with telomeric repeats, so we agree that the expression of 'construct' may be misleading. Therefore, we replaced several expressions with alternatives to explain mTALT as a constituent of telomere contents.
We revised the manuscript as follows: (Page 2, line 47) Here, we show that mammalian telomeres can also be completely reconstituted using a non-telomeric unique sequence.
 (line 43) Here we show that mammalian telomeres can also exploit non-telomeric, unique sequences in addition to telomeric repeats.
(Page 5, line 121) Our findings suggest that the robust telomere system based on simple repeats can be replaced with the ALT mechanism reconstructing telomere with new sequences even in mammalian cells. (Page 13, line 328) This is the first description of a specific mammalian ALT template that has the ability to protect telomere even when telomerase activity exists and eventually reconstruct telomeres.
 (line 362) This is the first description of a specific mammalian ALT template that can protect telomeres even when telomerase activity exists and that can eventually be incorporated into telomeres.
An interesting question that was raised during this research was whether mTALT was randomly (or accidentally) mobilized to be included in telomeres in ALT cells. We would like to think that mTALT is not just a random sequence, but a sequence that bears some characteristics to be easily recruited to telomeres. Underlying our thought is the fact that mTALT has been selected as a telomere-protecting sequence at least twice.
One is the process of repairing telomeric damage in the 129/Ola strain, which is manifested by the duplication of mTALT from chromosome 13 to chromosome 11, and the other is the ALT activation process in the ALT survivors that emerged in the telomerase-deficient ES cells of the 129/Ola origin. Thus, we think that mTALT has some characteristics related to the propensity to be recruited to telomeres. The authors hope that the reviewer agrees with our thoughts.

Q2. A similar phenomenon has been reported previously in two papers published back to back in Cancer Research by Roger Reddel's and Brad Johnson's groups, in which SV40
sequences become propagated through human ALT telomeres (Fasching et al and Marciniak et al, 2005). Both studies report a reduction in ALT phenotypes, concomitant with the presence of aberrant sequences in the telomere regions. These studies detract from the novelty somewhat, and must be cited and discussed in the context of this work.
A2. Thanks for the great suggestion. We added the following sentences in the discussion section.
(line 370) "In humans, tags flanked by telomeric repeats are reported to have the potential to be duplicated to other chromosomes 28 . For example, the SV40 sequence can be integrated into telomeres during immortalisation of the Werner syndrome cell line 29,30 . It is notable that SV40-based ALT cells also did not show the common ALT marker, APB, which suggests a similar mechanism of ALT to that of ALT mESCs we report here. One difference between the previously reported cases and our findings is that the mTALT sequence we described naturally exists in the mouse genome as a source of templates for ALT telomeres, which underscores its cell-intrinsic potential of reshaping telomeres." Q3. I didn't find the method for the C-circle assay. If the telomeres contain abundant mTALT sequences, they may still be forming extrachromosomal circular DNA, but the circles would not be detectable by the C-circle detection method. Sequence content should be taken into account with the ALT assays, and it would be interesting to see data using an mTALT probe.
A3. Thank you for the constructive suggestion. We performed C-circle assays without any restriction enzymes. In the revision, to be more precise, we repeated the C-circle assay with mTALT-specific probes as the reviewer suggested (New Extended Data Fig.   1e). We also added a positive control, U2OS, which was reported to bear a large number of C-circles. Consistently, we did not find any difference of C-circle between PD100 cells and PD800 cells.
We also described the details of the method of the C-circle assay as follows. In TRF assay, we used a telomeric repeat-specific probe so we could detect the length of fragments of telomeric repeats present between two mTALTs. We can predict the length of mTALT fragments remaining at the 5' and 3' terminal regions of mTALT after restrictions. In the case of AluI/MboI the length of the remaining mTALT is 150 bps. For HinfI, the length is 880 bps. So, we can predict the difference of each TRF signal will be 730 bps. In practice, the major TRF size of AluI/MboI case is ~1.2 kbps and that of HinfI is ~1.9 kbps. From the result, we can also find that the length of telomeric repeats between two mTALTs is ~1 kbps. The result was also confirmed by in silico digestion. The figure below is attached to Extended Data Fig. 2. The matched explanation appears in line 187 of the revised manuscript.

Q2. You nicely used the SNP allele frequencies to identify the chr11 locus as the active locus.
Beyond that, is the SNP data sufficient to decide if both alleles of the chr11 mTALT locus are used for ALT or only one allele? Be more explicit if you can clarify this point or if the alleles are too similar to make a conclusion. Chr. 11 mTALT PCR analysis.

A2. Thank you very much for your suggestion. That is a very interesting question
Allele frequency of Chr.11 amplicon Q3. Line 227-228: "The amount of various repetitive sequences including rRNA did not significantly (Extended Data Fig. 3c)." Sentences is incomprehensible without the verb, and will change its meaning depending on which verb you will add. Completion necessary.
A3. Sorry for the mistake. We added the verb 'change'.

Q4. Line 235: You indicate that "certain cells" were selected upon ALT activation. Do you
have any reason to believe that this was not a monoclonal event happening in a single cell that then expanded? It would be helpful if you would discuss the possible alternatives in the light of the data you have gathered.
A4. We thank the reviewer for the point. As the reviewer pointed out, we cannot rule out the possibility that ALT was activated in a single cell, and then the cell was

Q2. is the effect of HMGN1 on mTALT transcription specific? Other targets (within and outside this region) should be included.
A2. Thank you for the comment. HMGN1 does not have enzymatic activity on its own, so the resultant effect on chromatin may differ concerning interacting partners. As HMGN1 is a global chromatin protein, we do not think HMGN1 has only telomerespecific effects. However, as shown in our data, HMGN1 localizes to telomeres and has a protective role for telomeres. Another study showed HMGN1 interacted with a major shelterin protein, TRF2 (Ok-Hee et al., 2011). Thus, it will be a fair statement that HMGN1 may affect transcriptional networks in a genome-wide way, but it also regulates telomere physiology.
To examine these points in the most stringent way, we performed RNA-seq with control PD800 cells and HMGN1-depleted PD800 cells during revision. When the differentially expressed genes were sorted according to their q-value (adjusted p-value), Hmgn1 was the most significantly decreased gene. Notably, the next significantly decreased transcript was mTALT, which suggests HMGN1 regulates mTALT transcription critically. Another interesting observation was that the expression of Tcstv3, the gene located inside mTALT region, was increased. HMGN1 seems to affect the directionality of transcription because Tcstv3 is transcribed to the opposite direction of mTALT. We put this result in main figure 5i.
Notably, no evidence was found that HMGN1 regulates regions adjacent mTALT specifically or regulates genes related to ALT maintenance. First, when Hmgn1 was knocked down, genes close to mTALT (such as Ptchd3, Metrnl, B3gntl1, and Zfp750) did not change specifically (refer to the volcano plot below). Many other genes including those on chromosome11 (refer to the dot plot below) were affected by HMGN1 depletion. Second, when examining the Gene Ontology (GO) of the genes that were downregulated under hmgn1 knockdown, GO terms included multicellular organism development, positive regulation of cell proliferation, cell adhesion, and a few unrelated to terms. These terms are not likely to be directly related to ALT mechanism. Therefore, we think that HMGN1 is involved in the ALT mechanism by regulating mTALT transcription in a strand-specific manner. Although we cannot fully exclude the possibility that HMGN1 regulates ALT-related genes, we could not find ALTspecific terms with RNA-seq of HMGN1-depleted cells. Figure 5L, the authors need to include an RNAaseH1 treatment to control for the presence of R-loops.

A3. Sorry for the confusion that we may have caused by using abbreviations without
definition. 'RH' in the figure was RNAseH1 treatment control. To reduce the potential confusion, we changed 'RH' to 'RNaseH1'.

Q4. The connection of HMGN1 to R loops (or in the resolution of R-loops) remains unclear
in the present work. Is HMGN1 recruited to R-loops? An HMGN1 ChIP followed by DRIP could further highlight this point.

HMGN1-interacting factors have enzymatic roles in the chromatin opening process and R-loop formation. It can be interesting to study the role of HMGN1-interacting proteins in telomere and ALT.
As suggested by the reviewer, it will be interesting to examine whether HMGN1 is directly recruited to R loops. Unfortunately, it is too hard for us to perform ChIP and DRIP in the same sample, so we chose an alternative way. To observe the interaction of HMGN1 and R-loop, we performed proximity ligation assay (PLA), which can capture protein-protein interaction, in this case, the interaction of anti-HMGN1 antibody and S9.6 R-loop specific antibody. We compared three cells, pre-ALT PD100, post-ALT PD800 and HMGN1-depleted PD800 (which have lowered HMGN1 and R-loop in telomeres). PD800 cells showed more PLA signals than PD100 cells as expected.
HMGN1 depletion lowered PLA signals in PD800 cells. These data indicate that HMGN1 and R-loop were located close enough to be detected by PLA, and PLA signals were sufficiently specific.

Thus, HMGN1 and R-loop may have a complicated inter-connectivity that HMGN1
leads to R-loop formation as well as R-loop recruits HMGN1. It will be a very interesting topic for further study. However, we want to streamline the logical flow of this paper for readers' understanding, especially focusing on the function of HMGN1 onto R-loop formation. So please understand the situation in which we deliver the data above solely to reviewers. As the reviewer pointed out, HMGN1 depletion increased telomeric DNA damage.

Q5. Is depletion of HMGN1 leading to an increase in
However, we found that HMGN1 depletion also lowered TERRA expression and R-loop formation. Thus, it is unlikely that increased DNA damage was due to R-loops as suggested by the reviewer. Instead, we interpreted these results as TERRA/R-loop executed a protective role in mTALT-based telomeres.
During revision, we additionally experimented to test the effect of RNaseH enzyme on telomere protection (Extended data Fig. 8g). We used a lentivirus-based shRNA construct to deplete RNaseH1, and found that the lowered RNaseH induced the increased telomeric DNA damage. In this case, the telomeric R-loop formation was increased. Based on these findings, we can say that an elaborate balance of R-loop is important for stable telomere maintenance. Excess or too little amounts of R-loop may provoke problems in telomeres. We added a paragraph in discussion to elaborate these notions as follows (line 449):

'R-loops can induce genome instability by interfering with transcription and
replication. In particular, DSB formation and DNA loss may occur if a fork collapse occurs following replication fork stalling. However, several reports allude to potential R-loop contribution to the maintenance of the genome, particularly the telomere. This is possible by chromatin regulation 34 , priming DNA replication 35 and promotion of intertelomere HR 36 . Of note, R-loops can prevent telomeric replication fork collapse through HR which prevents telomeres from becoming dysfunctional 37 . In a study of ALT cancer cells, when the amount of RNaseH1 was depleted or overexpressed, abrupt telomere shortening occurred 38 . In other words, the precisely controlled telomeric Rloop makes an important contribution to telomere stability without seriously harming telomeric integrity.'

Q6. R-loops are known to be generated during abrupt changes in transcription demands or
when transcription is blocked due to DNA damage. This should be discussed with respect to mTALT transcripts in shHMGN1 cells.
A6. Thank you for the suggestion. We did not provide an adequate explanation for the formation process of R-loops. Following your comments, we added a paragraph in the discussion as follows (line 435): 'R-loops can be generated during abrupt changes in transcriptional demands or when transcription is blocked 20 . The formation of telomeric R-loops of ALT mESC can also be explained by the chromatin decompaction effect produced by HMGN1 which increases telomere transcription, and the loosened chromatin structure which may assist the interaction of transcribed RNA and DNA. In addition, the increased telomere damage level after ALT activation may interfere with transcription progress and promote the generation of co-transcriptional R-loops. Reasons for the increased telomere damage after ALT activation could not be accurately determined but it may be due to the lowered shelterin density which promotes stable telomere replication and protection. Another interesting aspect is that the formation of R-loops can be increased when there is a conflict between DNA replication and transcription machinery (socalled transcription-replication collisions, TRCs) 33 . While telomere replication proceeds from the subtelomere to the end of the chromosome, the direction of mTALT transcription is the reverse. Head-on TRCs specifically promote R-loop formation.

Thus, the association between the mTALT sequence characteristics and HMGN1
function produced telomeric R-loops and regulated telomere physiology.

Other comments
Q7. Figure 5b/Suppl. Figure 4a- The authors have written a very nice rebuttal and have done a good job in addressing the reviewers' comments. The manuscript is substantially improved.
I still have some problems with the manuscript, mostly relating to the idea that the ALT mechanism can be "converted" to using non-telomeric sequences, and that this represents a "backup telomere maintenance mechanism capable of producing new telomeres" (taken from the Introduction). I don't find this explanation compelling. It seems more likely that sequences that are amenable to heterochromatinization can infiltrate telomeres and contribute to telomere capping function. I think this could be more clearly articulated.
I also have a problem with the novelty. First, this study is in mice, in which the significance, utility, mechanistic characterization and relevance of ALT is not clear. Second, a similar process has been observed in human ALT cells (as mentioned in my previous review). Although, these back to back studies have now been mentioned in the Discussion, they are not discussed fully or described in the Introduction, which I feel is important to set the scene for this work.
In the abstract, please state ALT as alternative lengthening of telomeres (not alternative telomere lengthening), telomere loss does not directly induce telomere recombination.
In the Introduction, please correct the first sentence, change the phrase "telomeres can be "broken" by cellular stresses".
I'm also not convinced by the telomere analysis in Fig 1, for instance in Fig 1j are we not supposed to be seeing TTAGGG repeats being replaced by mTALT from PD100 to PD800? I cannot see this in the figure. Also, have these cell lines been STR profiled for authenticity? I can't find this in the methods. The chromosomes look very different at PD100 compared to PD800.
Reviewer #2 (Remarks to the Author): My comments have been adequatly addressed.
Reviewer #3 (Remarks to the Author): The authors have provided a wealth of information in the revised version, which I think merit publication.
Q: I still have some problems with the manuscript, mostly relating to the idea that the ALT mechanism can be "converted" to using non-telomeric sequences, and that this represents a "backup telomere maintenance mechanism capable of producing new telomeres" (taken from the Introduction). I don't find this explanation compelling. It seems more likely that sequences that are amenable to heterochromatinization can infiltrate telomeres and contribute to telomere capping function. I think this could be more clearly articulated.
A: We agree with the reviewer because our expression that the ALT mESC will use a new ALT mechanism that is different from previously known ALT mechanisms is not accurate, or supported by experimental evidence. It will be an important task for us to define what mechanism ALT mESC uses in our future studies. We discussed about the heterochomatic feature of mTALT in the discussion section, but, as the reviewer properly pointed out, we did not clarify how this feature actually affected the telomere protection function. Therefore, in our second revision, we simply explain the situation in which nontelomeric sequences co-exist with telomeric repeats and propose that this ALT phenomenon may be well conserved.
While we totally agree with the reviewer, we still would like to emphasize what we think is novel and interesting in our findings. Although we have not yet presented an exact mechanism for mTALT amplification, we still think it is a very interesting phenomenon that the mTALT sequence has an inherent ability to protect telomeres. There is a similarity to the situation in which the SV40 sequence is inserted into telomere and replicated, but there is a difference in that the sequence existing in the genome, rather than an external sequence, is contributing to the protection of the telomere. In any case, by showing that both human and mouse cells can stably maintain telomeres while harboring non-telomeric sequences, we successfully and meaningfully expanded our knowledge on the mammalian telomere physiology.
Following the reviewer's suggestions, we have changed the text in the second revision as follows.

"Our findings suggest that the robust telomere system based on simple repeats can convert to the ALT mechanism making use of non-telomeric sequences even in mammalian cells."
 Our findings suggest that non-telomeric sequences from an internal genomic region can be parts of telomeres even in mammalian cells.

"The evolutionary conservation of this ALT mechanism implies that cells have a backup telomere maintenance mechanism capable of producing new telomeres."
 The evolutionary conservation of this ALT phenomenon implies that eukaryotes have a robust system to cope with the loss or inactivation of telomerase.

"We found that a specific subtelomeric element, the mouse template for ALT (mTALT), is used for repairing telomeric DNA damage as well as for the development of new telomeric sequences in ALT-dependent mouse embryonic stem cells."
 We found that a specific subtelomeric element, the mouse template for ALT (mTALT), is used for repairing telomeric DNA damage as well as for composing portions of telomeres in ALT-dependent mouse embryonic stem cells.
Q: I also have a problem with the novelty. First, this study is in mice, in which the significance, utility, mechanistic characterization and relevance of ALT is not clear.
A: Thank you for your thoughtful discussion. We agree with the reviewer that there have not been many studies on the ALT mechanism with mouse model. In addition, the fact that the average telomere length of the widely used reference mice model is ten times longer than that of humans suggests that it is difficult to apply the mouse model directly to human biology. However, despite these limitations, the mouse studies significantly contributed to the telomere biology as a whole. First of all, most of the telomere-binding proteins have been discovered through mouse cell studies, and these proteins are well preserved in humans. Considering the versatility of the telomere protection mechanism, the mouse model may be considered to be an important axis of mammalian telomere research including ALT. The problem has been rather that there has not been an adequate mouse model to study ALT physiology. Our telomerasedeficient mESC model can be proven as a model for longitudinal studies on ALT initiation and maintenance in mammalian cells. Second, the mouse was the first animal model that showed that the recombination-based telomere maintenance mechanism could have a physiological meaning even in the presence of telomerase. Considering that most ALT mechanisms are based on recombination, it is reasonable that a mechanism similar to ALT may be important in the mouse development. In addition, when telomerase-based cancer was induced in mice and a telomerase inhibition strategy was introduced, cancers with the ALT mechanism occurred. Although it is limited to certain conditions, it is true that ALT can function in mice.
ALT mechanisms have been studied in terms of the mechanism of telomere maintenance in cancers, but studies on the functions of these mechanisms in real individuals have been limited. Although we have not yet been able to present a specific mechanism for telomere maintenance in the mESC study, ALT mESC can be an important model to investigate ALT mechanisms regarding the gold standard definition of ALT that the ability to maintain telomere length without telomerase. We can infer that the loss of the telomerase gene is not a necessary prerequisite for the mTALT sequence to be replicated because we observed in some wild mice strains that the mTALT sequence was replicated in natural conditions with functional telomerase present. If we use our model to clarify the activation process of this mechanism, it could contribute to a new understanding of the stress and physiological situations in which ALT can function in mammals in vivo.
Reflecting these considerations on the pros and cons of the mouse embryonic stem cell model for telomere biology including ALT, we added the following paragraph to the discussion section of the revised manuscript.
"Although there have been reports that the ALT-like mechanism is involved in telomere maintenance during mouse development, the mouse organismal model that maintains telomeres only by ALT has not been established. Considering that telomeres of mice are considerably longer than that of humans, it is still unclear whether the mechanism of telomere maintenance found in mice can be applied universally to various mammals including humans. In addition, further studies are needed to determine whether the ALT phenomenon identified in this study works in the same way as in human ALT cancer cells in which mechanisms based on homologous recombination or break-induced replication operate." "Our report is the first in mammalian models to longitudinally track and describe the telomeric changes integrating the non-telomeric sequence existing in the genome. Considering that this phenomenon is widely conserved in various model organisms such as S. cerevisiae, S. pombe, C. elegans and specific human immortalized cell line (AG11395), the characteristics of ALT revealed in this study will contribute to the expansion of our knowledge on telomere biology" In addition, we changed the title and the sentence in abstract to clearly show our study focuses on the mouse embryonic stem cell model. The title was changed from "Mammalian telomeres reforged with nontelomeric sequences" to "Telomeres reforged with non-telomeric sequences in mouse embryonic stem cells.
In abstract, "Here we show that mammalian telomeres can also exploit non-telomeric, unique sequences in addition to telomeric repeats" was changed to:  "Here we show that mouse telomeres can exploit non-telomeric, unique sequences in addition to telomeric repeats."

Q: Second, a similar process has been observed in human ALT cells (as mentioned in my previous review).
Although, these back to back studies have now been mentioned in the Discussion, they are not discussed fully or described in the Introduction, which I feel is important to set the scene for this work.
A : In Introduction, we added the description of AG11395 and explained that APB was not found, thus introducing the possibility that ALT may not be all the same mechanisms.
"While telomeric repeats and unique sequences are used to maintain telomeres in these various species, telomeres of human ALT cancers seem to consist only of telomeric repeats and their variants."  "Telomeres of human ALT cancers seem to consist only of telomeric repeats and their variants.
Interestingly, there is a distinct human cell line, AG11395, which is a SV40-transformed Werner mutant fibroblast. Telomeres of AG11395 cells contain extensive amounts of SV40 sequences and telomeric repeats. The cells showed some of typical ALT characteristics, but lacked ALT-associated promyelocytic leukaemia bodies (APBs), a molecular marker of ALT. This case implies the possibility that ALT may be a multi-faceted mechanism." Q: In the abstract, please state ALT as alternative lengthening of telomeres (not alternative telomere lengthening), telomere loss does not directly induce telomere recombination.
A: We changed "alternative telomere lengthening" to "alternative lengthening of telomeres". We also revised the following sentence.
"In ALT, telomere loss can induce telomere recombination by which specific sequences can be recruited into telomeres; however, to date, only canonical telomeric repeat-based telomeres have been found in mammals."  "ALT mainly utilizes recombination-based replication mechanisms and the constituents of ALT-based telomeres vary depending on models." Q: In the Introduction, please correct the first sentence, change the phrase "telomeres can be "broken" by cellular stresses".
A: We have revised the sentence as follows.
"All eukaryotic cells have linear chromosomes which inevitably cause the end replication and the end protection problem."  "Ends of linear chromosomes should handle two problems: 'the end replication problem' in which DNA replication machinery cannot completely replicate ends of lagging strands and 'the end protection problem' in which chromosomal ends should be discriminated from internal double-strand breaks (DSBs).
We also changed "broken" to "damaged".
Q: I'm also not convinced by the telomere analysis in Fig 1, for instance in Fig 1j are we not supposed to be seeing TTAGGG repeats being replaced by mTALT from PD100 to PD800? I cannot see this in the figure. Also, have these cell lines been STR profiled for authenticity? I can't find this in the methods. The chromosomes look very different at PD100 compared to PD800.
A: The FISH results clearly showed that the pattern shown in the results of the WGS analysis (Fig. 1c, 1h) and the results of mmqPCR (Fig. 1i) are the phenomena occurring at the actual ends, not at interstitial repeats. Precisely speaking, the telomere length decrease occurs only when proceeding from PD100 to PD350. Although there was a change in copy number in WGS, mTALT amplification was not detected in FISH due to too small amount of mTALTs (PD100 : ~2 copy, PD350 : ~4 copy). When proceeding from PD350 to PD450, the recovery of TTAGGG and mTALT recruitment occur simultaneously. In PD450, mTALT is not located at the entire ends and shows partial localization. And from PD450 to PD800, mTALT spreads to the entire ends and coexist with telomere repeats (yellow signal). To show this a little more clearly, additional images were taken and replaced the previous figures.
When it comes to morphology of chromosome, as shown in the previous paper ( The identity score using about 2.3 million SNPs for PD100 and PD800 cell lines was 94.28%, indicating that the WGS data ensure the authenticity. The reviewer might think that some chromosomes looked like human chromosomes. When the data were mapped onto the human reference genome, the rates were 23.44% and 24.28% for PD100 and PD800, respectively. Therefore, the cell lines were not contaminated by other cell lines, and PD100 and PD800 originated from the same cell line.