Decoding mitochondrial genes in pediatric AML and development of a novel prognostic mitochondrial gene signature

  1. Shilpi Chaudhary
  2. Shuvadeep Ganguly
  3. Jayanth Kumar Palanichamy
  4. Archna Singh
  5. Dibyabhaba Pradhan
  6. Radhika Bakhshi
  7. Anita Chopra
  8. Sameer Bakhshi

  • Curated by eLife

    eLife logo

    Evaluation Summary:

    Chaudhary and colleagues follow up their preliminary study on mitochondrial genome copy number in AML with this current study by looking if the expression of specific genes encoding mitochondrial components could provide further insight into AML prognosis. Multivariate analysis was used to identify those genes whose expression was prognostic of patient outcome, which led to the identification of three mitochondrial genes whose expression was used to build a multivariate risk model for childhood AML patients. Altogether, the work by Chaudhary and colleagues interestingly builds on their previous work and suggests that mitochondria may influence AML outcomes, and measuring mitochondrial parameters may help assess patient risk. However, the authors will need to identify the novelty of their findings over the previous reports from their own group.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. The reviewers remained anonymous to the authors.)

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

Background

Gene expression profile of mitochondrial-related genes is not well deciphered in pediatric acute myeloid leukaemia (AML). We aimed to identify mitochondria-related differentially expressed genes (DEGs) in pediatric AML with their prognostic significance.

Methods

Children with de novo AML were included prospectively between July 2016-December 2019. Transcriptomic profiling was done for a subset of samples, stratified by mtDNA copy number. Top mitochondria-related DEGs were identified and validated by real-time PCR. A prognostic gene signature risk score was formulated using DEGs independently predictive of overall survival (OS) in multivariable analysis. Predictive ability of the risk score was estimated along with external validation in The Tumor Genome Atlas (TCGA) AML dataset.

Results

In 143 children with AML, twenty mitochondria-related DEGs were selected for validation, of which 16 were found to be significantly dysregulated. Upregulation of SDHC (p<0.001), CLIC1 (p = 0.013) and downregulation of SLC25A29 (p<0.001) were independently predictive of inferior OS, and included for developing prognostic risk score. The risk score model was independently predictive of survival over and above ELN risk categorization (Harrell’s c-index: 0.675). High-risk patients (risk score above median) had significantly inferior OS (p<0.001) and event free survival (p<0.001); they were associated with poor-risk cytogenetics (p=0.021), ELN intermediate/poor risk group (p=0.016), absence of RUNX1-RUNX1T1 (p=0.027), and not attaining remission (p=0.016). On external validation, the risk score also predicted OS (p=0.019) in TCGA dataset.

Conclusion

We identified and validated mitochondria-related DEGs with prognostic impact in pediatric AML and also developed a novel 3-gene based externally validated gene signature predictive of survival.

Article activity feed

  1. eLife

    Evaluation Summary:

    Chaudhary and colleagues follow up their preliminary study on mitochondrial genome copy number in AML with this current study by looking if the expression of specific genes encoding mitochondrial components could provide further insight into AML prognosis. Multivariate analysis was used to identify those genes whose expression was prognostic of patient outcome, which led to the identification of three mitochondrial genes whose expression was used to build a multivariate risk model for childhood AML patients. Altogether, the work by Chaudhary and colleagues interestingly builds on their previous work and suggests that mitochondria may influence AML outcomes, and measuring mitochondrial parameters may help assess patient risk. However, the authors will need to identify the novelty of their findings over the previous reports from their own group.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. The reviewers remained anonymous to the authors.)

  2. eLife

    Reviewer #1 (Public Review):

    In this paper, Chaudhary et al assessed 143 children with AML, and out of 20 mitochondria-related DEGs that were chosen for validation, 16 were found to be significantly dysregulated. They show that upregulation of SDHC and CLIC1 and downregulation of SLC25A29 are independently predictive of lower survival, which was included in developing a prognostic risk score. They also show that this risk score model is independently predictive of survival better than ELN risk categorization, and high-risk patients had significantly inferior OS and event-free survival. The authors demonstrated that high-risk patients are associated with poor-risk cytogenetics, ELN intermediate/poor risk group, absence of RUNX1-RUNX1T1, and not attaining remission (p=0.016). The risk score also predicted survival in the TCGA dataset. They concluded that they have "identified and validated mitochondria-related DEGs with prognostic impact in pediatric AML and also developed a novel 3-gene based externally validated gene signature predictive of survival."

    Although this paper is interesting, it lacks novelty and does not advance the field significantly. The authors have used a similar approach in their recent paper in Mitochondrion where they showed that PGC1A driven increased mitochondrial DNA copy number predicts outcome in pediatric AML patients. Additionally, the authors have a small number of patients and chose only 20 genes for their analysis.

  3. eLife

    Reviewer #2 (Public Review):

    The design of the study is good. One of the strengths/novelties of the paper is that the authors stratified patients into 3 groups based on mitochondrial DNA (mtDNA) copy number and performed RNA-seq, where previous studies have no/little information about mtDNA copy number. Out of 143 patients, they sequenced only 3, 4, and 5 patients from 3 groups, along with 3 controls. High heterogeneity among cancer patients is unlikely to be accounted for by 3-5 samples per group, thus having limited statistical power. The authors should discuss the limitations of the small sample size and possible outcomes. The authors validated their differentially expressed genes with TCGA LAML which are mostly adult patients. A correct comparison will be with pediatric AML from other larger studies that had not stratified patients based on mtDNA.

  4. eLife

    Reviewer #3 (Public Review):

    Childhood acute myeloid leukemia (AML) is a heterogeneous disease with different outcomes for different patients, making identifying patients with different prognoses for clinical management. A variety of approaches have been used to stratify AML patients' risk, including molecular and clinical measurements to build prognostic risk scores. Previously, Chaudhary et al found that mitochondrial genome copy number per AML cell could stratify patients who would have good and poor outcomes and survival. This interesting finding suggested that mitochondrial amount and/or function alter AML disease course and suggested a further in-depth study of mitochondria in AML.

    Chaudhary and colleagues follow up their preliminary study on mitochondrial genome copy number in AML with this current study by looking if the expression of specific genes encoding mitochondrial components could provide further insight into AML prognosis. The authors collected childhood AML patient samples and grouped them based on mitochondrial genome copy number. They then performed transcriptomic analysis and identified a number of nuclear-encoded mitochondrial component genes whose expression was correlated or anticorrelated with mitochondrial genome copy number and this was confirmed with targeted analysis of identified transcripts in validation cohorts. Multivariate analysis was used to identify those genes whose expression was prognostic of patient outcome. This led to the identification of three mitochondrial genes (SDHC, CLIC1, SLC25A29) whose expression was used to build a multivariate risk model for childhood AML patients. The risk model based on the expression of these genes outperformed currently used ELN risk stratification and could be combined with ELN to increase prognostic power. Lastly, the authors used publically available data from adult AML patients and found that their risk score also had prognostic power in adult AML patients as well.

    Altogether, the work by Chaudhary and colleagues interestingly builds on their previous work and suggests that mitochondria may influence AML outcomes, and measuring mitochondrial parameters may help assess patient risk. Numerous exciting questions remain: what outputs of the mitochondria influence AML disease course and how? Why are some mitochondrial genes but not others correlated with mitochondrial DNA copy number in AML cells and how does this influence mitochondrial properties? Outside of predicting patient risk, can the mitochondrial phenotype of AML cells predict effective therapies? How does the mitochondrial risk model perform compared to and when utilized with other transcriptional-based risk stratification models proposed in the literature?

  5. eLife

    Author Response

    Reviewer #1 (Public Review):

    In this paper, Chaudhary et al assessed 143 children with AML, and out of 20 mitochondria-related DEGs that were chosen for validation, 16 were found to be significantly dysregulated. They show that upregulation of SDHC and CLIC1 and downregulation of SLC25A29 are independently predictive of lower survival, which was included in developing a prognostic risk score. They also show that this risk score model is independently predictive of survival better than ELN risk categorization, and high-risk patients had significantly inferior OS and event-free survival. The authors demonstrated that high-risk patients are associated with poor-risk cytogenetics, ELN intermediate/poor risk group, absence of RUNX1-RUNX1T1, and not attaining remission (p=0.016). The risk score also predicted survival in the TCGA dataset. They concluded that they have "identified and validated mitochondria-related DEGs with prognostic impact in pediatric AML and also developed a novel 3-gene based externally validated gene signature predictive of survival."

    Although this paper is interesting, it lacks novelty and does not advance the field significantly. The authors have used a similar approach in their recent paper in Mitochondrion where they showed that PGC1A driven increased mitochondrial DNA copy number predicts outcome in pediatric AML patients. Additionally, the authors have a small number of patients and chose only 20 genes for their analysis.

    We appreciate that the reviewer found our paper interesting and read our recently published article in Mitochondrion.

    In our previous work, the key finding was the predictive impact of mitochondrial DNA copy number on patients’ survival outcome in pediatric AML. Hence in the current paper, we deciphered it further to explore the heterogeneity associated with altered mitochondrial DNA copy number by identifying dysregulated genes in patients stratified by mitochondrial DNA copy number. We have identified several dysregulated mitochondria-related genes in pediatric AML for the first time hence we believe there is novelty in the work. Furthermore, developed novel mitochondria related gene signature that can predict survival of the pediatric patients with AML by analysing the prognostic impact of the dysregulated genes. These genes were identified to be dysregulated in pediatric AML for the first time and the gene signature model had predictive ability over and above ELN. Hence, we believe there is a novelty in the current work.

    The initial transcriptomic analysis was done in limited number of patients which remains a limitation of the study and mentioned in discussion line 313-315. However, the validation of selected genes by RT-PCR was carried out in a consecutively recruited cohort of pediatric AML patients over more than 3 years (total 143 patients between July 2016 to December 2019), with a median follow up of 36 months.

    As the main aim of the study was to decipher and validate mitochondria-related DEGs, the 20 genes were chosen based their mitochondrial localization as per mitochondrial compartment score. We have further validated these 20 genes in the external cohort from TCGA dataset (n=179), and the external validation adds to the strength of the study.

    Reviewer #2 (Public Review):

    The design of the study is good. One of the strengths/novelties of the paper is that the authors stratified patients into 3 groups based on mitochondrial DNA (mtDNA) copy number and performed RNA-seq, where previous studies have no/little information about mtDNA copy number. Out of 143 patients, they sequenced only 3, 4, and 5 patients from 3 groups, along with 3 controls. High heterogeneity among cancer patients is unlikely to be accounted for by 3-5 samples per group, thus having limited statistical power. The authors should discuss the limitations of the small sample size and possible outcomes. The authors validated their differentially expressed genes with TCGA LAML which are mostly adult patients. A correct comparison will be with pediatric AML from other larger studies that had not stratified patients based on mtDNA.

    We appreciate that reviewer liked our study design. We agree that the initial transcriptomic sequencing was done in only few patients out of 143 patients recruited for this study.

    Small sample size in the sequencing cohort is a limitation of the study considering the heterogeneity of AML (mentioned in the revised manuscript discussion line 313-315).

    However, the validation of selected genes by RT-PCR was carried out in a consecutively recruited cohort of pediatric AML patients over more than 3 years (total 143 patients between July 2016 to December 2019), with a median follow up of 36 months. Only validated genes were used for further analysis and developing risk score.

    We have validated the differentially expressed genes in the publicly available TCGA adult AML dataset and we agree that the better comparison would be with the larger cohort of pediatric AML patients. However, we could not find publicly available pediatric AML patients DEGs database, which additionally reports clinical outcomes for analysing prognostic impact. Hence the TCGA dataset was chosen.

    External validation of our gene signature risk score further adds to the strength of the study as the gene signature appears valid for both pediatric as well as adult AML patients (this is mentioned in revised manuscript result line 205-207).

  10. Version published to 10.1101/2022.04.01.22273235v1 on medRxiv