Prognostic Impact of Wilms’ Tumour 1 Mutation in Patients with Acute Myeloid Leukaemia

By Maratib Ali¹, Manzar Bozdar², Sarah Fatimah¹, Rafia Mahmood², Tanveer Ahmed¹, Hatim Khalid¹

Affiliations

1. Department of Haematology, Combined Military Hospital, Peshawar, Pakistan
2. Department of Haematology, Armed Forces Institute of Pathology, Rawalpindi, Pakistan

doi: 10.29271/jcpsp.2026.03.341

ABSTRACT
Objective: To detect WT1 gene alterations among individuals diagnosed with acute myeloid leukaemia (AML) and investigate their relation to the response of induction therapy.
Study Design: A descriptive study.
Place and Duration of the Study: Department of Haematology, Armed Forces Institute of Pathology, Rawalpindi, Pakistan, from June to December 2023.
Methodology: The study enrolled all freshly diagnosed AML patients who underwent clinical, haematological, and molecular testing. Based on their WT1 mutation status, participants were categorised into distinct groups and assessed after four weeks of induction therapy. Independent t-test and chi-square tests were used to analyse the variables, while odds ratios (ORs) with 95% confidence intervals (CIs) were computed using cross-tabulation.
Results: Within the cohort of 98 newly diagnosed AML cases, patients had a mean age of 36.5 years, showing a male predominance with a male-to-female ratio of 1.22:1. WT1 mutations were detected in 12 (12.2%) patients. These patients showed significantly lower haemoglobin, higher leucocyte counts, reduced platelet counts, and higher bone marrow blast percentage (p <0.05). Complete remission occurred in 75% of WT1-mutated versus 62.8% of wild-type patients (p = 0.408). Although not statistically significant, WT1 mutations demonstrated a trend towards a more aggressive presentation and poorer therapeutic response.
Conclusion: WT1 mutation in AML is associated with aggressive disease and less differentiated French-American-British (FAB) subtypes. Although remission rates were lower in WT1-mutated cases, the difference was not statistically significant. Larger prospective studies are needed to establish its prognostic significance and guide individualised therapy.

Key Words: Acute myeloid leukaemia, WT1 mutation, Induction therapy, Prognosis, FAB classification.

INTRODUCTION

Acute myeloid leukaemia (AML) represents a diverse group of haematological cancers characterised by uncontrolled proli-feration of immature myeloid precursors within bone marrow and peripheral circulation.¹ Despite progress in mole- cular characterisation and treatment options, overall survival remains limited, especially among high-risk groups.² Consequently, there is an ongoing need to identify reliable prognostic biomarkers that can aid in risk stratification and inform treatment  decisions.

Located on the short arm of chromosome 11 at position 11p13, the Wilms’ tumour 1 (WT1) gene produces a transcription factor essential for proper cellular development and has been linked  to  multiple  cancer  types,  including  AML.³

WT1 is highly expressed in leukaemic blasts and has been shown to promote leukaemogenesis through various mechanisms, such as inhibition of apoptosis, regulation of cell cycle progression, and modulation of cellular differentiation.⁴ Notably, approximately 10% of patients with AML harbour WT1 mutations, which have  been  connected  to  diverse  clinical  trajectories.⁵

Multiple research groups have examined how WT1 mutations influence AML prognosis, yielding contradictory findings. Some studies have reported an association between WT1 mutations and adverse clinical outcomes, such as lower complete remission (CR) rates, shorter disease-free survival (DFS), and shorter overall survival (OS).^6,7 However, other studies have failed to demonstrate a significant prognostic impact of WT1 mutations or  have   reported   conflicting   findings.^8,9

These discrepancies in published literature may stem from numerous factors, including variations in patient demographics, treatment protocols, and molecular detection techniques. Furthermore, the prognostic significance of WT1 alterations might be modified by coexisting molecular abnormalities, parti- cularly FLT3-ITD and NPM1 mutations, which are recognised as having  a  substantial  impact  on  AML  patient  outcomes.¹⁰

Considering the potential importance of WT1 mutations as prognostic indicators in AML, a thorough understanding of their clinical ramifications is essential. This study sought to examine the prognostic implications of WT1 mutations within an AML patient cohort. This study aimed to specifically assess how these mutations influence treatment response following induction chemotherapy while considering other pertinent clinical and molecular variables.

By elucidating the prognostic role of WT1 mutations in AML, this study may contribute to improving risk stratification and personalised treatment approaches, ultimately improving patient outcomes.

METHODOLOGY

This investigation employed a prospective cohort study design and was executed within the Department of Haematology, Armed Forces Institute of Pathology, Rawalpindi, Pakistan, from June to December 2023. Ethical approval was obtained from the Institutional Review Board of Armed Forces Institute of Pathology, Rawalpindi, Pakistan (Ref. No. FC-HEM22-17/READ-IRB/23/2189). The sample size for this study was determined using the WHO calculator, considering a previously observed probability of WT1 gene mutation to be 6.8%.¹¹ These calculations determined an optimal sample size of 98 participants, allowing for a 5% error margin with 95% confidence level. Participant recruit-ment employed a non-probability convenience sampling technique. All patients received comprehensive study information and provided informed consent. The maximum available participants (98) during the study timeframe were enrolled.

The study enrolled patients with recent cytogenetically normal AML diagnoses, irrespective of age or gender. Patients with secondary AML transformed from other blood disorders, and those with previous chemotherapy or radiation exposure, were excluded.

Clinical information, including age, gender, presenting symptoms, and physical examination findings, was recorded at presentation. Haematological evaluation was performed on 3 mL of EDTA-anticoagulated blood for complete blood counts, including total leucocyte count (TLC), haemoglobin concentration, platelet count, and peripheral blood blast percentage. Bone marrow examination findings were docu-mented, and cases were segregated with morphological classification according to French-American-British (FAB) AML subtype criteria.

For molecular analysis, 5 mL of EDTA-anticoagulated bone marrow aspirate was obtained from each patient at the time of diagnosis. Genomic DNA was extracted using the QIAamp DNA Blood Mini Kit according to the manufacturer's instructions. WT1 mutation screening targeted exons 7 and 9, which harbour the majority of clinically significant muta-tions. Real-time fluorescent qualitative Polymerase Chain Reaction (PCR) utilised specific primers targeting WT1 muta-tion on the ABI-7500 real-time PCR system to amplify these critical regions. Each PCR run included positive controls, negative controls, and internal quality control samples. PCR results stratified patients into WT1-positive and WT1- negative groups.

All patients received standardised induction therapy according to the institutional protocol. After the completion of induction therapy, bone marrow aspiration was repeated to assess remission status. CR required bone marrow blasts <5%, absolute neutrophil count >1x10⁹/L, and platelet count >100x10⁹/L.

Data entry utilised Microsoft Excel with analysis through the Statistical Package for Social Sciences (SPSS) version 26.0. Normality of quantitative data was assessed using the Shapiro-Wilk test. Continuous variables were expressed as mean ± standard deviation (SD), while categorical variables were presented as frequencies and percentages. An independent t-test was utilised to compare the two variables. Chi-square testing determined associations between WT1 mutation status and clinical-haematological parameters, encompassing TLC, haemoglobin, platelet count, blast percentage, and FAB subtype. The odds ratio (OR) with 95% confidence interval (CI) was calculated using a cross-tabulation analysis. Statistical significance was set at p ≤0.05.

RESULTS

The study cohort included 98 patients diagnosed with de novo AML who met the inclusion criteria. The mean age at diagnosis was 36.5 years, with ages spanning from 7 to 68 years, reflecting a relatively young patient population. Gender distribution revealed a slight male predominance, with 54 (55.1%) males and 44 (44.9%) females, thereby a male-to-female ratio of approximately 1.22:1.

Molecular analysis revealed WT1 gene mutations in 12 (12.2%) patients, while the remaining 86 patients (87.8%) harboured wild-type WT1. Comparative analysis of baseline haematological parameters between WT1-mutated and wild-type patients revealed clinically relevant differences, although statistical significance was not achieved for all the variables (Table I). Patients with WT1 mutations exhibited tendencies toward more aggressive disease characteristics, manifested by reduced haemoglobin concentrations (mean 7.7 ± 0.5 g/dL vs. 8.7 ± 0.9 g/dL, p = 0.001), raised TLC (mean 46.4 ± 13.2 × 10⁹/L vs. 35.3 ± 11.4 × 10⁹/L, p = 0.003), diminished platelet counts (mean 38.8 ± 12.9 × 10⁹/L vs. 55.3 ± 17.0 × 10⁹/L, p = 0.002), and elevated bone marrow blast proportions (mean 73.0 ± 5.7% vs. 62.0 ± 7.3%, p = 0.001).

Table I: Baseline clinico-haematological characteristics of AML patients stratified by WT1 mutation status.

Characteristics	WT1-mutated (n = 12)	WT1-wild type (n = 86)	p-values*
Mean age, years	34.6 ± 18.1	36.7 ± 17.4	0.697
Male gender, n (%)	7 (13.0)	47 (87.0)	0.813
Female gender, n (%)	5 (11.4)	39 (88.6)	0.819
Mean haemoglobin, g/dL	7.7 ± 0.5	8.7 ± 0.9	0.001
Mean total leucocyte count, ×109/L	46.4 ± 13.2	35.3 ± 11.4	0.003
Mean platelet count, ×109/L	38.8 ± 12.9	55.3 ± 17.0	0.002
Mean bone marrow blasts, %	73.0 ± 5.7	62.0 ± 7.3	0.001
*Independent samples t-test. WT1: Wilms’ tumour 1.

Table II: Distribution of FAB morphological subtypes according to WT1 mutation status.

FAB subtypes	WT1-mutated (n = 12) n (%)	WT1-wild type (n = 86) n (%)
M0 (Minimally differentiated)	0 (0.0)	8 (9.3)
M1 (Without maturation)	5 (41.7)	18 (20.9)
M2 (With maturation)	6 (50.0)	22 (25.6)
M3 (Acute promyelocytic)	0 (0.0)	12 (14.0)
M4 (Acute myelomonocytic)	1 (8.3)	15 (17.4)
M5 (Acute monocytic)	0 (0.0)	8 (9.3)
M6 (Acute erythroid)	0 (0.0)	2 (2.3)
M7 (Acute megakaryoblastic)	0 (0.0)	1 (1.2)
FAB: French-American-British; WT1: Wilms’ tumour 1.

Table III: Induction chemotherapy response according to WT1 mutation status.

Response category	WT1-mutated (n = 12) n (%)	WT1-wild type (n = 86) n (%)	OR (95% CI)	p-values*
CR	9 (75.0)	54 (62.8)	0.84 (0.58–1.21)	0.408
Refractory disease	3 (25.0)	32 (37.2)	1.48 (0.54–4.12)	0.408
*Chi-square test. CR: Complete remission; WT1: Wilms’ tumour 1; OR: Odds ratio.

According to the FAB classification criteria, the distribution of AML subtypes differed between WT1-mutated and wild- type patients (Table II). WT1 mutations were predominantly associated with morphologically less differentiated subtypes, particularly M1 (acute myeloblastic leukaemia without maturation) and M2 (acute myeloblastic leukaemia with maturation). In contrast, the wild-type cohort showed a more heterogeneous distribution across various FAB subtypes.

Treatment response to standard induction chemotherapy represented a critical endpoint for assessing the prognostic impact of WT1 mutations (Table III). CR was achieved in 9 out of 12 WT1-mutated patients (75.0%), compared to 54 out of 86 wild-type patients (62.8%).

On the contrary, refractory disease was observed in 3 out of 12 WT1-mutated patients (25.0%), compared to 32 of 86 wild-type patients (62.8%).

The OR for achieving CR in WT1-mutated patients was 0.84 (95% CI: 0.58–1.21), indicating approximately half the odds of remission compared to wild-type patients, although CIs crossed unity. Conversely, the OR for refractory disease was 1.48 (95% CI: 0.54–4.12), suggesting nearly twice the odds of treatment failure, although without statistical significance.

DISCUSSION

The prognostic impact of WT1 mutations in AML has generated extensive research interest, with contradictory results emerging across numerous studies. In this study, the observed trends suggested an association between WT1 mutations and adverse clinical outcomes, although the differences did not reach statistical significance, potentially due to the limited sample size.

This study investigated the prognostic significance of WT1 gene mutation among 98 patients with de novo acute myeloid leukaemia, revealing a mutation frequency of 12.2%. This frequency aligns closely with recent large-scale genomic studies. Bhatnagar et al. reported WT1 mutations in approximately 5-12% of adults with AML in their Alliance for Clinical Trials in Oncology study, demonstrating remarkable consistency across diverse patient populations.¹² Similarly, a thorough 2024 review by Debnath et al. documented WT1 mutation frequencies of approximately 10% in adult AML cohorts across multiple studies.¹³ These concordant findings validate the molecular screening methodology of the study and confirm that WT1 mutations, while relatively uncommon, represent a recurrent genetic aberration in AML patho- genesis.

The data revealed that WT1-mutated patients presented with markedly more aggressive haematological parameters at diagnosis. These cases demonstrated notably reduced haemoglobin levels (7.7 vs. 8.7 g/dL, p = 0.001), substantially elevated TLC (46.4 vs. 35.3 × 10⁹/L, p = 0.003), diminished platelet counts (38.8 vs. 55.3 × 10⁹/L, p = 0.002), and considerably elevated bone marrow blast proportions (73.0% vs. 62.0%, p = 0.001). Goel et al. demonstrated through using RNA sequencing that elevated WT1 expression at diagnosis was associated with worse haematological parameters and adverse outcomes in AML patients.¹⁴ Furthermore, Baranwal et al. in their 2024 transplantation study reported that WT1-mutated patients consistently exhibited more aggressive disease features, including higher blast counts and reduced platelet levels at presentation.¹⁵ The biological basis for these aggressive features likely relates to WT1's critical function in regulating haematopoietic stem cell self-renewal and differentiation. When loss-of-function mutations occur, they appear to promote uncontrolled proliferation while simultaneously impairing normal maturation pathways.¹⁶

When examining treatment outcomes, this study revealed that 9 (75%) of WT1-mutated patients achieved CR, compared with 54 (62.8%) of wild-type patients, yielding an OR of 0.84 (95% CI: 0.58-1.21). The limited number of WT1-mutated patients (n = 12) necessitates cautious interpretation of treatment response data. While trends were observed, the study was underpowered to definitively establish significant differences in clinical outcomes. The CIs for ORs are consequently wide, reflecting this statistical uncertainty. The trend toward lower remission rates in the WT1-mutated group is consistent with several recent investigations. Wang et al., in their comprehensive 2021 analysis, reported that WT1-mutated patients had significantly lower CR rates following standard induction chemotherapy, particularly when co-occurring mutations were present.¹⁷ Similarly, Yu T et al. demonstrated in their 2023 study that high WT1 expression in non-M3 AML was associated with reduced rates of CR and higher relapse rates.¹⁸ It is worth noting that the modest effect size observed in the study may reflect the relatively small number of WT1-mutated patients (n = 12), which inherently limits the statistical power to detect significant differences even when clinically meaningful trends exist.

Interestingly, the observed refractory disease rate of 3 (25.0%) in WT1-mutated patients versus 32 (37.2%) in wild- type patients (OR 1.48, 95% CI: 0.54-4.12) diverges somewhat from several previous reports. Atluri et al., in their 2023 Alliance study, found that WT1-mutated patients with concurrent FLT3-ITD mutations had significantly higher treatment failure rates and markedly inferior survival outcomes compared to those without FLT3-ITD.¹⁹ However, the findings of the study are also supported by local genomic data from Pakistan, where Shahid et al. reported substantial genetic heterogeneity in AML, emphasising the need for population-specific molecular characterisation of prognostic markers, including WT1.²⁰ This finding highlights an important point: the absence of statistical significance in refractory disease analysis may be explained by the complex genetic architecture of AML. In reality, WT1 mutations rarely occur in isolation, and their prognostic impact appears highly context- dependent, influenced by the specific constellation of accompanying genetic alterations present in each patient.

It is important to acknowledge that the prognostic significance of WT1 mutations remains somewhat debatable in contemporary AML research. While this study and several others suggest adverse associations, the picture is far from clear-cut. Similar findings were reported by Tien et al., who showed that concomitant WT1 mutations adversely affected prognosis in specific molecular subgroups of AML, particularly in patients with double-mutant CEBPA.²¹ Burd et al., in the landmark Beat AML Master Trial, demonstrated that dominant WT1 mutations (those with a variant allele fraction ≥0.30) had distinctly different clinical implications compared to subclonal mutations. This suggests that simply categorising patients as WT1-mutated or wild-type may be an oversimplification that fails to capture the underlying bio- logical complexity.²²

This study was carried out at a single tertiary care centre with a modest sample size, which may restrict the extent to which the findings can be generalised to broader populations. Follow-up duration was short, restricting assessment of long-term survival and relapse outcomes. Extended molecular profiling, including co-mutation analysis, was not performed, which could have provided deeper insights into the prognostic effect of WT1 mutations. Additionally, treatment response was evaluated after only one induction cycle, without long-term monitoring of minimal residual disease.

CONCLUSION

Mutation of WT1 in AML is associated with distinct clinical and morphological characteristics, including more aggressive disease at presentation and a predominance in less differentiated FAB subtypes (M1 and M2). However, the limited sample size and absence of long-term outcome data prevent definitive conclusions about their independent prognostic value. Future prospective studies with comprehensive molecular profiling— evaluating co-mutations, allelic burden, and mutation sites—are needed to clarify the independent prognostic role of WT1 mutations. Such research will enhance understanding of disease biology and support more individualised therapeutic approaches for AML patients.

ETHICAL APPROVAL:
Ethical approval was obtained from the Institutional Review Board of Armed Forces Institute of Pathology, Rawalpindi, Pakistan (Ref. No. FC-HEM22-17/READ-IRB/23/2189).

PATIENTS’ CONSENT:
Informed consent of the patients was obtained before the initiation of the study.

COMPETING INTEREST:
The authors declared no conflict of interest.

AUTHORS’ CONTRIBUTION:
MA: Conception and design of the study; acquisition, analysis, and interpretation of data.
MB, SF, RM, TA: Conception and design of the study; critical review of the manuscript for important intellectual content.
HK: Acquisition, analysis, and interpretation of data.
All authors approved the final version of the manuscript to be published.

REFERENCES

1. Dohner H, Wei AH, Appelbaum FR, Craddock C, DiNardo CD, Dombret H. Diagnosis and management of AML in adults: 2022 recommendations from an international expert panel on behalf of the ELN. Blood 2022; 140(12):1345-77. doi: 10.1182/blood.2022016867.
2. Kantarjian H, Borthakur G, Daver N, DiNardo CD, Issa G, Jabbour E, et al. Current status and research directions in acute myeloid leukemia. Blood Cancer J 2024; 14(1):163. doi: 10.1038/ s41408-024-01143-2.
3. Othman J, Potter N, Ivey A, Tazi Y, Papaemmanuil E, Jovanovic J, et al. Molecular, clinical, and therapeutic determinants of outcome in NPM1-mutated AML. Blood 2024; 144(7):714-28. doi: 10.1182/blood.2024024310.
4. Quattrocchi A, Cappelli LV, De Simone G, De Marinis E, Gentile M, Gasperi T, et al. Biomarkers in acute myeloid leukemia: From state of the art in risk classification to future challenges of RNA editing as disease predictor and therapy target. Aspects Mol Med 2023; 2:100023. doi: 10.1016/j.amolm.2023.100023.
5. Gaidzik VI, Schlenk RF, Moschny S, Becker A, Bullinger L, Corbacioglu A, et al. Prognostic impact of WT1 mutations in cytogenetically normal acute myeloid leukaemia: A study of the German-Austrian AML Study Group. Blood 2009; 113(19):4505-11. doi: 10.1182/blood-2008-10-183392.
6. Wakita S, Marumo A, Morita K, Kako S, Toya T, Najima Y, et al. Mutational analysis of DNMT3A improves prognostic stratification in acute myeloid leukaemia. Cancer Sci 2023; 114(4): 1297-308. doi: 10.1111/cas.15720.
7. Ma S, Tang L, Tang H, Wu C, Pu X, Yang J, et al. WT1 and DNMT3A mutations in prognostic significance of acute myeloid leukaemia: A meta-analysis. Cancer Biother Radiopharm 2025; 40(1):22-30. doi: 10.1089/cbr.2024.0093.
8. Hou HA, Huang TC, Lin LI, Liu CY, Chen CY, Chou WC, et al. WT1 mutation in 470 adult patients with acute myeloid leukaemia: Stability during disease evolution and implication of its incorporation into a survival scoring system. Blood 2010; 115(25): 5222-31. doi: 10.1182/blood-2009-12-259390.
9. Sato H, Kobayashi T, Kameoka Y, Teshima K, Watanabe A, Yamada M, et al. WT1 expression in peripheral blood at diagnosis and during the course of early consolidation treatment correlates with survival in patients with intermediate and poor-risk acute myeloid leukaemia. Int J Clin Oncol 2024; 29(4):481-92. doi: 10.1007/s10147-024-02480-9.
10. Gale RE, Green C, Allen C, Mead AJ, Burnett AK, Hills RK, et al. The impact of FLT3 internal tandem duplication mutant level, number, size, and interaction with NPM1 mutations in a large cohort of young adult patients with acute myeloid leukemia. Blood 2008; 111(5):2776-84. doi: 10.1182/blood-2007-08-109090.
11. Xinan Pan, Gao Mengge, Ke Wang, Yu Wang, Jun Kong, Yuqian Sun, et al. Prognostic Impact of WT1 mutation on AML of different risk groups based on 2022 European Leukaemia net (ELN) risk classification. Blood 2022; 140(Suppl 1):3216-7. doi: 10.1182/blood-2022-166323.
12. Bhatnagar B, Kohlschmidt J, Orwick SJ, Buelow DR, Fobare S, Oakes CC, et al. Framework of clonal mutations concurrent with WT1 mutations in adults with acute myeloid leukaemia: Alliance for clinical trials in oncology study. Blood Adv 2023; 7(18):5299-303. doi: 10.1182/bloodadvances.2023010482.
13. Debnath, A., Nath, S. Prognosis and treatment in acute myeloid leukemia: a comprehensive review. Egypt J Med Hum Genet 25, 91 (2024). httpss://doi.org/10.1186/s43042- 024-00563-w.
14. Goel H, Pandey AK, Kumar R, Ningombam SS, Naz F, Makkar H, et al. RNA sequencing identifies WT1 overexpression as a predictor of poor outcomes in acute myeloid leukemia. Cancers 2025; 17(11):1818. doi: 10.3390/cancers17111818.
15. Baranwal A, Basmaci R, He R, Viswanatha D, Greipp P, Murthy HS, et al. Genetic features and outcomes of allogeneic transplantation in patients with WT1-mutated myeloid neoplasms. Blood Adv 2024; 8(3):562-70. doi: 10. 1182/bloodadvances.2023010960.
16. Kongtim P, Cao K, Ciurea SO. Donor specific anti-HLA antibody and risk of graft failure in haploidentical stem cell transplantation. Adv Hematol 2019; 2016:2016:4025073. doi: 10.1155/2016/4025073.
17. Wang Y, Weng WJ, Zhou DH, Fang JP, Mishra S, Chai LH, et al. Wilms tumor 1 mutations are independent poor prognostic factors in pediatric acute myeloid leukemia. Front Oncol 2021; 11:632094. doi: 10.3389/fonc.2021.632094.
18. Yu T, Zhan Q, Yan X, Luo X, Wang X, Tang X, et al. Clinical significance of WT1 in the evaluation of therapeutic effect and prognosis of non-M3 acute myeloid leukemia. Cancer Biol Ther 2023; 24(1):2285801. doi: 10.1080/15384047. 2023.2285801.
19. Atluri H, DiGennaro J, Patel KP, Routbort M, Oran B, Isaacs GC, et al. Clinical and prognostic implications of WT1 mutations and co‑occurring mutations in de novo and relapsed acute myeloid leukemia. Blood 2023; 142(S1):959. doi: 10.1182/blood-2023-188319.
20. Shahid S, Shakeel M, Siddiqui S, Ahmed S, Sohail M, Khan IA, et al. Novel genetic variations in acute myeloid leukemia in Pakistani population. Front Genet 2020; 11:560. doi: 10. 3389/fgene.2020.00560.
21. Tien FM, Hou HA, Tang JL, Kuo YY, Chen CY, Tsai CH, et al. Concomitant WT1 mutations predict poor prognosis in acute myeloid leukaemia patients with double mutant CEBPA. Haematologica 2018; 103(11):e510-3. doi: 10.3324/haematol.2018.189043.
22. Burd A, Levine RL, Ruppert AS, Mims AS, Borate U, Stein EM, et al. Precision medicine treatment in acute myeloid leukaemia using prospective genomic profiling: Feasibility and preliminary efficacy of the beat AML master trial. Nat Med 2020; 26(12):1852-8. doi: 10.1038/s41591-020- 1089-8.