Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Formal reply to “Alternative lengthening of telomeres is not synonymous with mutations in ATRX/DAXX”

The Original Article was published on 10 March 2021

Replying to R. R. Reddel and A. de Nonneville. Nature Communications https://doi.org/10.1038/s41467-021-21794-0 (2021)

The study of Nonneville and Reddel partially overcomes a limitation of our original work1. Namely, it integrates the in silico analysis of the genomic footprints of telomere maintenance mechanisms (TMMs) of the Pan-Cancer Analysis of Whole Genomes (PCAWG) dataset1 with molecular assays of ALT status based on C-circle assays.

As PCAWG was in its essence a computational reanalysis of existing data from the ICGC and TCGA projects, generation of such gold standard molecular data was outside of the project’s scope. In consequence, we compared truncating mutations of the ATRX and DAXX genes against activating mutations of TERT to pick out strong representatives of the two classes of TMMs that were identifiable by sequencing data alone and removed all other samples from the training set of the machine learning procedure.

The data that Nonneville and Reddel now used was published after we submitted our study early in 2017 to the internal revision process of the PCAWG consortium, and thus could not be included in our work. Therefore, the current manuscript is a great opportunity to learn more about ALT-positive cancers that lack genetic mutations in ATRX or DAXX.

The Nonneville and Reddel study use C-circle assays for 167 pancreatic neuroendocrine tumors (PaNET) and melanomas, roughly 6% of the full PCAWG dataset, to establish a molecular gold standard for the ALT status, assuming a one-to-one correspondence of C-circle positive signals and ALT-positive status. They then integrate this data with the genomic footprints that were derived by us using the TelomereHunter software2.

Strikingly, they find that five of the features identified in the original study are sufficient to train classifiers that perfectly separates C-circle positive from C-circle negative cases if trained cohort wise (Table 1). Besides telomere content, all these features are singleton telomere variant repeat counts, including the, before the Sieverling et al. study uncharacterized, depletion of TTTGGG singletons and the enrichment of TTCGGG singletons.

Table 1 Comparison of features for ALT, respectively, ATRX/DAXXtrunc prediction.

Not surprisingly, they report that combining melanoma and PaNET cases for classifier training reduces the accuracy to 93–94%. This is likely due to the combination of increasing biological and technical heterogeneity, as not only the baseline frequencies of genomic alterations differ substantially between melanoma and PaNET, but different tumor cellularity and batch effects during the generation of the sequencing data also impact the underlying data. Interestingly, using Lee et al.3 C-circle assay data as target variable in a novel ROC-curve analysis detects also for the ALT probability score proposed in Sieverling et al.1 a robust performance (Fig. 1). Choosing a comparable sensitivity-specificity trade-off for both models, the ALT probability model reaches at a specificity level of 93% a sensitivity of 87% which is close to the 93.98% sensitivity reported for Nonneville et al. classifier. This corresponds to a score threshold of 0.5, in contrast to the arbitrarily chosen threshold of 0.75 in Nonneville et al. Separating the melanoma and PaNET cases supports the necessity of cancer-type specific classifiers, as the C-circle status of the former is predict far less accurately by the ALT probability score than the latter (Fig. 1).

Fig. 1: Receiver operating characteristic analysis.
figure1

The curve shows the True positive rate (sensitivity) and the Fales positive rate (1 – specificity) for all possible thresholds on the ALT probability score (Sieverling et al.1). The predicted class label is the C-circle status reported from Lee et al., while the red cross depicts the performance of the classifier proposed in the Matters Arising article for which classification results are reported for only one threshold.

In their study, Nonneville and Reddel suggest that TVR counts correlate with ATRX/DAXXtrunc regardless of CCA result. Remarkably, their dataset only contains three ATRX/DAXXtrunc samples that are C-circle-negative. As depicted in Fig. 1, indeed two of these strongly derivate in all TVR counts, but drawing such a general conclusion from two out of three observations may be premature. Therefore, it is necessary to either conduct additional ALT assays, such as APB-staining or telomere FISH, to generate an undebatable gold standard for these three samples, and to collect more data from the “C-circle-”/“ATRX/DAXXtrunc + ” subgroup to further substantiate this hypothesis. The presented data on the TTCGGG singleton count raises another interesting question. Figure 1 of Nonneville and Reddel shows that not only all “C-circle + ”/“ATRX/DAXXtrunc + ” but also about 40% of the “C-circle + ”/“ATRX/DAXXtrunc – ” cases show a clear enrichment of TTCGGG singleton counts. If indeed the TVR patterns in general and the TTCGGG pattern, in particular, are linked to the disruption of the ATRX/DAXX complex rather than the ALT status, this would imply for these cancer samples the presence of alterations beyond the genetic mutations that constitute the ATRX/DAXXtrunc subgroup, for instance, a downregulation of the complex on the protein level.

Another highly relevant observation is the exclusion of breakpoint and telomere insertion count as features by the Akaike information criterion, which were among the most predictive features in the Sieverling et al. study (Table 1). Does this imply that these features are a mere side effect of ATRX/DAXX disruption and that a more genomically stable ALT subtype with intact ATRX/DAXX complex but increased C-circle levels exist?

The most relevant function of the ATRX/DAXX complex for the ALT phenotype is the maintenance of a condensed chromatin state at the telomeres and elsewhere in the genome. Loss of this function probably increases the likelihood of genomic breaks in the chaperoned loci, and enables the induction of break-induced replication (BIR) that is implied in the recombination of telomeric sequences in ALT4.

Future studies integrating multiple omics layers will clarify in how far this function is simply disrupted by additional mechanisms in ALT positive tumors without ATRX/DAXXtrunc mutations, and if alternatives with less deleterious impact on genomic stability exist. In any case, the new data emphasizes that ALT positivity does not fully overlap with the ATRX/DAXXtrunc annotation. In consequence, the derived score in our original study should be renamed from the ALT probability score into a genomic ATRX/DAXXtrunc score to reflect this discovery.

Methods

The ROC-curve analysis was conducted with the ROCR package5. The ATRX/DAXXtrunc score, respectively, ALT probability score from Sieverling et al.1 was used to evaluate the performance for the individual cohorts and for both cohorts together. As the model in the Matters Arising study does only report a binary classification but no continuous score, the reported performance values of the combined model were marked by a red cross in the diagram. The underlying data is provided as source data in XLSX format.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

All used data is available as a supplement to Lee et al.3 (Supplementary Table 2), Sieverling et al.1 (Supplementary Table 1), and the Matters arising article. For the purpose of patient id mapping between the different datasets the icgc data portal was used: https://dcc.icgc.org/pcawgSource data are provided with this paper.

References

  1. 1.

    Sieverling, L. et al. Genomic footprints of activated telomere maintenance mechanisms in cancer. Nat. Commun. 11, 1–13 (2020).

    Article  Google Scholar 

  2. 2.

    Feuerbach, L. et al. TelomereHunter – in silico estimation of telomere content and composition from cancer genomes. BMC Bioinformatics 20, 272 (2019).

    Article  Google Scholar 

  3. 3.

    Lee, M. et al. Telomere sequence content can be used to determine ALT activity in tumours. Nucleic Acids Res. 46, 4903–4918 (2018).

    CAS  Article  Google Scholar 

  4. 4.

    Zhang, J.-M., Yadav, T., Ouyang, J., Lan, L. & Zou, L. Alternative lengthening of telomeres through two distinct break-induced replication pathways. Cell Rep. 26, 955–968.e3 (2019).

    CAS  Article  Google Scholar 

  5. 5.

    Sing, T., Sander, O., Beerenwinkel, N. & Lengauer, T. ROCR: visualizing classifier performance in R. Bioinformatics 21, 3940–3941 (2005).

    CAS  Article  Google Scholar 

Download references

Acknowledgements

Lina Sieverling, Sabine Hartlieb, and Frank Westermann provided valuable comments to this manuscript. The author received funding within the DFG Research Consortium FOR 2674.

Author information

Affiliations

Authors

Contributions

LF has performed the reanalysis of the data and wrote the manuscript.

Corresponding author

Correspondence to Lars Feuerbach.

Ethics declarations

Competing interests

The author declares no competing interests.

Additional information

Peer review information Nature Communications thanks Trey Ideker and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Source data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Feuerbach, L. Formal reply to “Alternative lengthening of telomeres is not synonymous with mutations in ATRX/DAXX”. Nat Commun 12, 1551 (2021). https://doi.org/10.1038/s41467-021-21796-y

Download citation

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing