On 2026-01-19 13:00:58, user Gene C Koh wrote:
Gene Ching Chiek Koh, Serena Nik-Zainal
Department of Genomic Medicine, University of Cambridge, CB2 0QQ, UK.
We commend Kanwal et al. for their timely evaluation of the in vivo mutagenic potential of CX-5461. This follows our report that CX-5461 induces substantial mutagenesis in cultured mammalian cells1. The authors analysed samples from four patients treated with CX-5461, including marrow aspirates, trephine biopsies, PBMCs, and skin lesions collected at early treatment timepoints (baseline; days 1, 2, or 9; and end-of-treatment of a 21-/28-day cycle), and used error-corrected duplex sequencing to detect low-frequency mutations. They concluded that CX-5461 exposure did not increase single-/ double-base substitution or indel burdens, nor reproduced the mutational signatures reported in our in vitro study. While we welcome their contribution, several methodological and interpretive shortcomings limit the conclusions that can be drawn.
1. Data presentation<br />
Figures 1–3 present absolute mutation counts instead of frequencies normalized to total informative duplex bases per sample. In duplex sequencing, normalization is a basic requirement to account for variability in sequencing depth and library complexity; without it, true mutation accrual or fold-change differences versus controls (if any) cannot be assessed reliably.
2. Experimental controls, assay sensitivity, and performance<br />
The study lacks essential positive and negative controls making it impossible to evaluate whether the sequencing and analytical processes used by the authors have worked. Clinical samples with known mutational signatures detectable through this approach should have been included to confirm assay sensitivity and substantiate a true negative finding. This is fundamental. Samples from patients unexposed to CX-5461 were also required as negative controls to establish background variability, affording confidence intervals and statistical robustness.<br />
Moreover, the authors have not shown awareness of the assay’s limit of detection (LOD). What is the smallest measurable fold-change at the reported sequencing depth? Without this, one cannot determine the smallest mutational differences that could have been missed. The authors have not disclosed quality-control metrics required to understand whether sufficient data quality was achieved for detecting differential mutagenesis. P/S: TwinStrand kit has an error rate ~0.5e-7 to 1e-7 depending on the protocols, and this can be considerably higher if DNA quality is low or from fixed biopsies.
3. Lack of curation, comparisons to literature<br />
The reported mutation counts did not make sense (baseline values exceeding treated samples, patient samples sometimes lower than kit control). The authors should perform some ‘sanity check’ comparisons with published mutation frequencies of respective normal adult tissues from other duplex-sequencing studies2,3. Analytical rigour would include, for example, examining whether detected variants represent driver mutations from clonal haematopoiesis or occurred in genes under post-treatment selection. Such analyses would have demonstrated critical evaluation of data quality and biological relevance.
4. Cell-type considerations, sampling window<br />
Most analysed compartments—PBMCs, MACS-sorted marrow fractions—are dominated by mature, non-dividing cells that rarely fix new mutations. A more relevant population for assessing mutagenicity is the haematopoietic stem and progenitor cells (HSPCs), typically <0.5% of marrow cells. A null result in the analysed compartments could just mean no widespread mutation fixation in mature immune cells; it does not exclude the possibility of mutagenesis in progenitors below the detection threshold of the current assay.<br />
In addition, samples were taken at very early timepoints (days 1, 2, 9, or EOT) of the first treatment cycle. At such intervals, mutagenic events are unlikely to have become fixed, as mutagen-induced DNA damage will need time to become embedded through DNA repair and replication. Exposure in terminally-differentiated cells might yield no detectable mutations. If exposure occurs on dividing cells, mutational footprints may only become detectable months or years after exposure. The current dataset lacks the temporal window necessary to assess cumulative in vivo mutagenicity.
5. Expected evidence of prior treatments <br />
All four patients reportedly had “measurable, relapsed, or refractory advanced haematologic malignancies without any standard therapeutic options available”4. Although treatment histories were not provided, these patients likely received multiple prior therapies (e.g., doxorubicin, cyclophosphamide, etc) that could induce characteristic mutational signatures in normal haematopoietic cells5. Were signatures of prior therapy detected by the authors? Their absence raises concerns regarding the overall assay sensitivity and/or suggests that sampling strategy was suboptimal for detecting mutagenic exposures.
6. Interpretation of model data<br />
While critical of our findings in cultured human cells as “not adequately representative of physiological human tissue” – a limitation we explicitly acknowledged in our manuscript’s title and discussion – the authors cited a C. elegans study6 in support of their argument of “low non-selective mutagenic potential of CX-5461”. This interpretation is incorrect: the worm study reported high copy-number aberrations, high SNV burdens, and a distinct A>T/T>A-rich signature after CX-5461 exposure, with survival requiring multiple repair pathways (homology-directed repair, microhomology-mediated end joining, nucleotide excision repair, and translesion synthesis). If anything, these cross-species findings reinforce rather than contradict our observations that CX-5461 is highly mutagenic. The concentrations used in that study were chosen to promote viability in the worms, not to minimise mutagenicity. Selective viability does not equate to selective mutagenicity.
7. Clinical mutagenicity testing<br />
We agree that clinical safety assessments must be rigorous and physiologically relevant. The authors dismissed our experiments as not rivalling the “GLP-compliant, non-mutagenic” results of the CX-5461 drug development pathway. However, those mutagenicity data are not available in the public domain and have neither been shared by the authors nor the company that distributes CX-5461.
We urge the authors to reconsider and not simply dismiss our findings. First, the primary clinical quality mutagenicity assay (required by agencies such as the US Food and Drug Administration (FDA), European Medicines Agency, and UK Medicines and Healthcare Regulatory Agency (MHRA)) referred to by the authors comprises the Ames test – a reverse gene mutation test performed in prokaryotes (e.g., E.coli, Salmonella).
Second, according to the FDA’s ICH S2(R1) guidance for a standard battery of mutagenicity assays (Safety Implementation Working Group of the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use), additional genotoxic assays should be performed in mammalian cells in vitro (where some of the more common assays include metaphase chromosome aberration assays, the micronucleus assay, and the mouse lymphoma L5178Y cell Tk (thymidine kinase) gene mutation assay (MLA)) or in in vivo studies as necessary.
Third, the FDA guidance acknowledges that “no single test is capable of detecting all genotoxic mechanisms relevant in tumorigenesis” and that the standard battery serves primarily for hazard identification rather than comprehensive assessment of mutagenic potential. For negative in vivo results, the ICH S2(R1) guidance requires evidence of adequate target-tissue exposure (e.g., toxicity in the tissue, TK/PK data, or direct tissue concentrations) to validate interpretability. Without such data, negative findings have limited meaning, especially where in vitro systems demonstrate strong mutagenicity.
Fourth, while the Ames test served its purpose for decades, there are well-described problems including false positives, false negatives and critically, a lack of human metabolism that even supplementation with rodent S9 mix cannot always overcome.
Finally, a point also raised by the accompanying commentary to our publication is that perhaps the time has come to re-evaluate how mutagenicity assays are performed. Current assays cannot capture the genome-wide mutation patterns revealed by whole-genome sequencing in human cells, and as a community we should consider using unbiased, agnostic, modern genomic approaches capable of detecting all classes of mutational changes in human cells. This is not an attack on CX-5461; rather, it is a call to the community to consider re-evaluation of mutagenicity assays in drug development.
8. Unsubstantiated claims<br />
The claim of potential contaminants accounting for the mutagenic outcomes we and others have observed is speculative and unsupported. The fact that multiple studies1,6 observed the same mutagenic outcomes using CX-5461 from independent sources suggests that this is unlikely. The authors showed no analytic chemistry (LC-MS/MS) and/or spiking experiments to substantiate this claim.
9. Inadequate supporting material throughout <br />
There were many gaps in the methods/supporting information, including adequate clinical annotation, precise sampling times/total treatment cycle, and basic quality-control metrics. Experimental details (e.g., antibodies used for MACS sorting, essential for interpreting analysed subpopulations) were not provided. These omissions limit transparency, reproducibility, and the interpretability of the findings.
10. Beneficence, non-maleficence, autonomy, justice<br />
First, in academia and medicine, we are guided by the principle of doing no harm. In identifying mutagenesis in experimental systems (an incidental finding), we acted in the best interest of the community – reporting an observation that could have an impact on patients and acknowledging the limitations of our system. We have no role in the (dis)continuation of clinical trials; we simply presented our data transparently and highlighted potential risk. <br />
Second, while the authors chose to discontinue their trial, several others remained active (e.g., NCT04890613, NCT06606990, NCT07069699, NCT07147231, NCT07137416). Their decision was conservative, and in our view, scientifically prudent. We commend their caution. However, it does not justify criticism of those of us reporting safety concerns in good faith.<br />
Third, as a community, we serve society better by being aware of issues, addressing the problems with robust experiments rather than polarising into groups “for” or “against” a compound, so that truly beneficial compounds can get to patients as quickly as possible. <br />
Finally, safety concerns may extend beyond mutagenesis and include tumour promotion effects. CX-5461’s interaction with TOP2B, for example, has been linked to serious, late-emerging toxicities, including therapy-induced leukaemia and cardiotoxicity7-10.
Concluding remarks<br />
Given the experimental and analytical shortcomings outlined above, definitive conclusions regarding CX-5461’s in vivo mutagenicity cannot yet be drawn. The absence of evidence should not be taken as evidence of absence. Rigorous, longitudinal studies with appropriate controls and independent oversight are required to assess true medium- to long-term risks.
We share the authors’ view that thorough, transparent evaluation of anticancer agents is essential. Given the authors’ vested interest in finding a negative result, we suggest independent individuals be involved in performing the analysis/interpretation of their studies to negate potential conflicts of interest. We remain open to collaboration in this effort, in the shared interest of patient safety and scientific integrity.
- Koh, G.C.C., Boushaki, S., Zhao, S.J., Pregnall, A.M., Sadiyah, F., Badja, C., Memari, Y., Georgakopoulos-Soares, I., and Nik-Zainal, S. (2024). The chemotherapeutic drug CX-5461 is a potent mutagen in cultured human cells. Nat Genet 56, 23-26. 10.1038/s41588-023-01602-9.
- Abascal, F., Harvey, L.M.R., Mitchell, E., Lawson, A.R.J., Lensing, S.V., Ellis, P., Russell, A.J.C., Alcantara, R.E., Baez-Ortega, A., Wang, Y., et al. (2021). Somatic mutation landscapes at single-molecule resolution. Nature 593, 405-410. 10.1038/s41586-021-03477-4.
- Machado, H.E., Mitchell, E., Obro, N.F., Kubler, K., Davies, M., Leongamornlert, D., Cull, A., Maura, F., Sanders, M.A., Cagan, A.T.J., et al. (2022). Diverse mutational landscapes in human lymphocytes. Nature 608, 724-732. 10.1038/s41586-022-05072-7.
- Khot, A., Brajanovski, N., Cameron, D.P., Hein, N., Maclachlan, K.H., Sanij, E., Lim, J., Soong, J., Link, E., Blombery, P., et al. (2019). First-in-Human RNA Polymerase I Transcription Inhibitor CX-5461 in Patients with Advanced Hematologic Cancers: Results of a Phase I Dose-Escalation Study. Cancer Discov 9, 1036-1049. 10.1158/2159-8290.CD-18-1455.
- Mitchell, E., Pham, M.H., Clay, A., Sanghvi, R., Williams, N., Pietsch, S., Hsu, J.I., Obro, N.F., Jung, H., Vedi, A., et al. (2025). The long-term effects of chemotherapy on normal blood cells. Nat Genet 57, 1684-1694. 10.1038/s41588-025-02234-x.
- Ye, F.B., Hamza, A., Singh, T., Flibotte, S., Hieter, P., and O'Neil, N.J. (2020). A Multimodal Genotoxic Anticancer Drug Characterized by Pharmacogenetic Analysis in Caenorhabditis elegans. Genetics 215, 609-621. 10.1534/genetics.120.303169.
- Pan, M., Wright, W.C., Chapple, R.H., Zubair, A., Sandhu, M., Batchelder, J.E., Huddle, B.C., Low, J., Blankenship, K.B., Wang, Y., et al. (2021). The chemotherapeutic CX-5461 primarily targets TOP2B and exhibits selective activity in high-risk neuroblastoma. Nat Commun 12, 6468. 10.1038/s41467-021-26640-x.
- Zhang, W., Gou, P., Dupret, J.M., Chomienne, C., and Rodrigues-Lima, F. (2021). Etoposide, an anticancer drug involved in therapy-related secondary leukemia: Enzymes at play. Transl Oncol 14, 101169. 10.1016/j.tranon.2021.101169.
- Cowell, I.G., Sondka, Z., Smith, K., Lee, K.C., Manville, C.M., Sidorczuk-Lesthuruge, M., Rance, H.A., Padget, K., Jackson, G.H., Adachi, N., and Austin, C.A. (2012). Model for MLL translocations in therapy-related leukemia involving topoisomerase IIbeta-mediated DNA strand breaks and gene proximity. Proc Natl Acad Sci U S A 109, 8989-8994. 10.1073/pnas.1204406109.
- Zhang, S., Liu, X., Bawa-Khalfe, T., Lu, L.S., Lyu, Y.L., Liu, L.F., and Yeh, E.T. (2012). Identification of the molecular basis of doxorubicin-induced cardiotoxicity. Nat Med 18, 1639-1642. 10.1038/nm.2919.