Mitochondrial heteroplasmy improves risk prediction for myeloid neoplasms

Abstract

Clonal hematopoiesis of indeterminate potential is the primary pathogenic risk factor for myeloid neoplasms, while heteroplasmy (mutations in a subset of cellular mitochondrial DNA) is another marker of clonal expansion associated with hematological malignancies. We explore how these two markers relate and influence myeloid neoplasms incidence, and their role in risk stratification. We find that heteroplasmy is more common in individuals with clonal hematopoiesis of indeterminate potential, particularly those with higher variant allele fractions, multiple mutations, or spliceosome machinery mutations. Individuals with both markers have a higher risk of myeloid neoplasms than those with either alone. Furthermore, heteroplasmic variants with higher predicted deleteriousness increase the risk of myeloid neoplasms. Incorporating heteroplasmy in an existing risk score model for individuals with clonal hematopoiesis of indeterminate potential significantly improves sensitivity and better identifies high-risk groups. This suggests heteroplasmy as a clonal expansion marker and potentially as a biomarker for myeloid neoplasms development.

Fig. 1. Description of CHIP mutations.

Distribution of the number of individuals carrying mutations stratified by genes in (A) UKB and (B) ARIC. Distribution of VAF stratified by gene in (C) UKB (number of mutations used to derive statistics; median VAF [Q1, Q3] DNMT3A: n = 11,648; 8.7% [5.6%, 15.0%]; TET2: n = 6923; 9.2% [5.7%, 18.6%]; ASXL1: n = 1734; 12.0% [5.7%, 22.1%]; PPM1D: n = 658; 8.9% [5.4%, 15.6%]; TP53: n = 699; 7.1% [5.2%, 12.4%]; SRSF2: n = 317; 14.3% [7.7%, 29.0%]; SF3B1: n = 239; 13.5% [9.4%, 20.4%]; U2AF1: n = 123; 9.5% [6.1%, 17.6%]) and in (D) ARIC (number of mutations used to derive statistics; median VAF [Q1, Q3] DNMT3A: n = 113; 14.1% [8.5%, 19.2%]; TET2: n = 85; 10.2% [4.6%, 20.1%]; ASXL1: n = 19; 18.8% [13.6%, 25.4%]; PPM1D: n = 3; 11.7% [11.3%, 12.6%]; TP53: n = 10; 7.9% [5.4%, 10.9%]; SRSF2: n = 0; NA [NA, NA]; SF3B1: n = 8; 8.6% [6.6%, 11.3%]; U2AF1: n = 4; 21.2% [15.8%, 23.9%]). Two-sided Mann-Whitney-Wilcoxon rank sum test with continuity correction was used to assess the difference between VAF of spliceosome mutations compared to that of other mutations in both UKB (P < 2.2e-16) and ARIC (P = 0.7234). Number of mutations per individual in (E) UKB and (F) ARIC. Only the DTA, DDR and classic spliceosome mutations are presented. Abbreviations: VAF, variant allele fraction; and DDR, DNA damage response. DTA includes DNMT3A, TET2, ASXL1. * Denotes the P value for a comparison between spliceosome mutations and other CHIP mutations. Source data are provided as a Source Data file.

Fig. 2. Description of heteroplasmy.

Distribution of the number of individuals with heteroplasmy stratified by complex/region in (A) UKB and (B) ARIC. Distribution of mMSS stratified by complex/region in (C) UKB (number of individuals used to derive statistics; median mMSS [Q1, Q3] Complex I: n = 38,931; 0.000 [0.000, 0.140]; Complex III: n = 11,700; 0.037 [0.000, 0.109]; Complex IV: n = 22,632; 0.043 [0.000, 0.228]; Complex V: n = 9416; 0.012 [0.000, 0.029]; D-loop: n = 36,829; 0.015 [0.007, 0.017]; rRNA: n = 14,891; 0.258 [0.121, 0.589]; tRNA: n = 8876; 0.146 [0.081, 0.296]) and (D) ARIC (number of individuals used to derive statistics; median mMSS [Q1, Q3] Complex I: n = 694; 0.000 [0.000, 0.174]; Complex III: n = 236; 0.032 [0.000, 0.087]; Complex IV: n = 426; 0.069 [0.000, 0.243]; Complex V: n = 175; 0.015 [0.002, 0.028]; D-loop: n = 559; 0.015 [0.007, 0.017]; rRNA: n = 267; 0.320 [0.127, 0.633]; tRNA: n = 179; 0.154 [0.079, 0.306]). Graphical representation of mMSS was truncated at mMSS = 2. All values of mMSS were included for the analyses. Two-sided Mann-Whitney-Wilcoxon rank sum test with continuity correction was used to assess the difference between mMSS of rRNA/tRNA compared to mMSS of other complexes/regions in both UKB (P < 2.2e-16) and ARIC (P < 2.2e-16). Number of heteroplasmies per individual in (E) UKB and (F) ARIC. Abbreviations: mMSS, modified mitochondrial local constraint (MLC) score sum. * Denotes comparison between mMSS in rRNA/tRNA and mMSS occurring in other complexes/regions. Source data are provided as a Source Data file.

Fig. 3. Association between CHIP and heteroplasmy.

The percentage of individuals presenting with both CHIP and heteroplasmy, only with CHIP, or only with heteroplasmy in (A) UKB and (B) ARIC. The prevalence of heteroplasmy in different CHIP gene subsets was evaluated using multivariable logistic regression without multiple testing correction in (C) UKB (CHIP vs. no CHIP, P < 2e-16; VAF ≥ 20% vs. 2% ≤ VAF < 20%, P < 2e-16; multiple mutations vs. single mutation, P = 1.05e-7; and spliceosome mutated CHIP vs. other CHIP, P < 2e-16) and in (D) ARIC (CHIP vs. no CHIP, P = 5.5e-3; VAF ≥ 20% vs. 2% ≤ VAF < 20%, P = 0.0887; multiple mutations vs. single mutation, P = 0.0376; and spliceosome mutated CHIP vs. other CHIP, P = 0.0528). The absolute number of individuals with heteroplasmy is indicated by the numbers on the left side of the bars. The prevalence of multiple heteroplasmies within individuals with heteroplasmy across different CHIP gene subsets was evaluated using multivariable logistic regression without multiple testing correction in (E) UKB (CHIP vs. no CHIP, P < 2e-16; VAF ≥ 20% vs. 2% ≤ VAF < 20%, P = 3.11e-13; multiple mutations vs. single mutation, P = 3.22e-5; and spliceosome mutated CHIP vs. other CHIP, P = 1.86e-8) and in (F) ARIC (CHIP vs. no CHIP, P = 0.621; VAF ≥ 20% vs. 2% ≤ VAF < 20%, P = 0.116; multiple mutations vs. single mutation, P = 0.407; and spliceosome mutated CHIP vs. other CHIP, P = 0.125). The absolute number of individuals with multiple heteroplasmies is indicated by the numbers on the left side of the bars. mMSS of different CHIP subsets in individuals with heteroplasmy was compared using multivariable linear regression without multiple testing correction in (G) UKB (number of individuals used to derive statistics; median mMSS [Q1, Q3] no CHIP: n = 111,524; 0.020 [0.000, 0.157]; CHIP: n = 11,445; 0.052 [0.006, 0.293]; 2%; ≤VAF < 20% : n = 9131; 0.040 [0.005, 0.248]; VAF ≥ 20%: n = 2314; 0.128 [0.014, 0.545]; single mutation: n = 10,311; 0.046 [0.005, 0.275]; multiple mutations: n = 1134; 0.123 [0.013, 0.541]; DNMT3A: n = 3963; 0.061 [0.007, 0.306]; TET2: n = 2434; 0.068 [0.007, 0.376]; ASXL1: n = 761; 0.111 [0.013, 0.529]; PPM1D: n = 255; 0.071 [0.008, 0.328]; TP53: n = 232; 0.071 [0.007, 0.360]; SRSF2: n = 183; 0.334 [0.016, 0.763]; SF3B1: n = 130; 0.183 [0.016, 0.522]; U2AF1: n = 62; 0.309 [0.036, 0.717]. CHIP vs. no CHIP, P < 2e-16; VAF ≥ 20% vs. 2% ≤ VAF < 20%, P < 2e-16; multiple mutations vs. single mutation, P = 9.36e-16; and spliceosome mutated CHIP vs. other CHIP, P = 9.28e-16) and in (H) ARIC (number of individuals used to derive statistics; median mMSS [Q1, Q3] no CHIP: n = 2027; 0.033 [0.000, 0.191]; CHIP: n = 132; 0.074 [0.009, 0.502]; 2% ≤ VAF < 20%: n = 96; 0.056 [0.008, 0.436]; VAF ≥ 20%: n = 36; 0.103 [0.010, 0.587]; single mutation: n = 110; 0.067 [0.007, 0.506]; multiple mutations: n = 22; 0.162 [0.013, 0.444]; DNMT3A: n = 43; 0.154 [0.009, 0.405]; TET2: n = 38; 0.065 [0.012, 0.687]; ASXL1: n = 8; 0.461 [0.068, 0.773]; PPM1D: n = 0; NA [NA, NA]; TP53: n = 2; 0.127 [0.065, 0.188]; SRSF2: n = 0; NA [NA, NA]; SF3B1: n = 5; 0.066 [0.000, 1.040]; U2AF1: n = 3; 0.013 [0.010, 0.440]. CHIP vs. no CHIP, P = 2.3e-7; VAF ≥ 20% vs. 2% ≤ VAF < 20%, P = 0.368; multiple mutations vs. single mutation, P = 0.603; and spliceosome mutated CHIP vs. other CHIP, P = 0.976). Graphical representation of mMSS was truncated at mMSS = 2. All values of mMSS were included for the analyses. For all analysis, two-sided tests were used. Abbreviations: CHIP, clonal hematopoiesis of indeterminate potential; DDR, DNA damage response; mMSS, modified mitochondrial local constraint (MLC) score sum; and VAF, variant allele fraction. DTA includes DNMT3A, TET2, ASXL1. * Indicates the comparison between spliceosome mutations and other CHIP genes. In the UKB, these analyses were adjusted for age modeled as a restricted cubic spline, sex, smoking status, and a history of cancer. In ARIC, these analyses were adjusted for age modeled as a restricted cubic spline, sex, and smoking status. For the analyses including mMSS, the models were also adjusted for heteroplasmy count. Source data are provided as a Source Data file.

Fig. 4. Risk of MN incidence based on CHIP and heteroplasmy status.

Kaplan-Meier curves and hazard ratios from the adjusted Cox proportional hazards models comparing the risk of MN development between individuals with heteroplasmy and those without heteroplasmy in (A) UKB (n = 434,304) and (B) ARIC (n = 7632), and those with heteroplasmy only, those with CHIP only and those with both CHIP and heteroplasmy in (C) UKB and (D) ARIC. In the UKB, these analyses were adjusted for age modeled as a restricted cubic spline, sex, smoking status and the presence of prevalent cancer. In ARIC, these analyses were adjusted for age modeled as a restricted cubic spline, sex and smoking status. Abbreviations: CHIP, clonal hematopoiesis of indeterminate potential; CI, confidence interval; HR, hazard ratio; and ref, reference. Source data are provided as a Source Data file.

Fig. 5. Distribution of heteroplasmy and risk of MN by CHRS-M categories.

A Proportions of heteroplasmy count (0, 1, 2, 3, and 4 + ) within each CHRS-M category (low: n = 29,224; intermediate: n = 4009; and high: n = 364; P < 2e-16 using Chi-square test, two-sided). B Boxplots and scatterplots of the distribution of mMSS by CHRS-M category (low: n = 29,224; intermediate: n = 4,009; and high: n = 364; P < 2e-16 using ANOVA, two sided). The vertical lines within the box indicate the 25th (Q1), 50th (Q2, center), and 75th (Q3) percentiles of the distribution. The whiskers indicate the Q1 – 1.5 × (Q3 – Q1) and Q1 + 1.5 x (Q3 – Q1). Values outside the range of the whiskers (outliers) are displayed as dots. C A Sankey diagram of the reclassification of CHRS to CHRS-M. D Kaplan-Meier curves for the risk of MN development by CHRS-M category (n = 33,597). The dotted lines indicate the cumulative incidence of individuals who remain in the same category. The solid lines indicate the cumulative incidence of individuals who are recategorized from low to intermediate (green) risk and from intermediate to high (dark blue) risk categories using the CHRS-M. E Kaplan-Meier curves for the risk of MN development in the CHRS low-risk group (n = 30,542) with the same data as (B) on an enlarged y-axis. Abbreviations: CHRS, clonal hematopoiesis risk score; CHRS-M, clonal hematopoiesis risk score with mitochondrial heteroplasmy; and mMSS, modified mitochondrial local constraint (MLC) score sum. Source data are provided as a Source Data file.

Fig. 6. Hazard ratios (95% confidence intervals) for the associations of CHRS-M categories with subtypes of MN.

Hazard ratios (95% confidence intervals) for the associations of CHRS-M categories with subtypes of MN. The analysis was restricted to UKB participants with CHIP. Hazard ratios (the point) and 95% confidence intervals (error bars) were estimated using Cox proportional hazards models with the low-risk group as the reference category (low: n = 29,224; intermediate: n = 4,009; and high: n = 364). All models are adjusted for age modeled as a restricted cubic spline, sex, smoking status, and a history of cancer. Abbreviations: AML, acute myeloid leukemia; CI, confidence interval; CHRS-M, clonal hematopoiesis risk score with mitochondrial heteroplasmy; HR, hazard ratio; MDS, myelodysplastic syndrome; MN, myeloid neoplasms; and MPN, myeloproliferative neoplasms. Source data are provided as a Source Data file.