Outlier handling: No datapoints are removed. Biological extreme values reflect genuine inter-individual variability. Statistical outliers detected by 1.5×IQR are ⚠ flagged in tables; sensitivity analyses confirm no qualitative conclusion changes. Non-parametric tests (Mann-Whitney U, Kruskal-Wallis) are inherently robust to extremes through rank transformation.
Small subgroup policy:n = 3–4 = limited power, cautious interpretation. n = 1–2 = statistically non-viable, descriptive data only, no inferential testing.
Multiple comparisons: All disease groups compared individually to Healthy reference. No correction applied — exploratory analysis. P-values facilitate hypothesis generation.
Section 1
Study Design & Sample Description
Left ventricular snap-frozen biopsy tissue obtained at LVAD implantation (Cardiotoxic, DCM groups), at heart transplantation/donation (Healthy), and from an in vitro iPSC-derived cardiomyocyte model. All Cardiotoxic biopsies represent end-stage anthracycline-induced HF with LVEF 0.10–10% at implantation.
Study groups
Group
n
Tissue
Clinical context
Cardiotoxic
21
LV biopsy (LVAD)
End-stage HF after doxorubicin; LVEF 0.10–10%
Healthy
11
LV biopsy
Non-failing: 4 HTX biopsies + 7 DCD donors
DCM nonmutant
10
LV biopsy (LVAD)
DCM, no pathogenic mutation identified
DCM gen. mutation
10
LV biopsy (LVAD)
DCM with causative genetic variant
Cells (CTP2-C5)
21
In vitro
iPSC-CM from Cardiotoxic Patient 2; control vs dox; mitochondrial transfer subgroups
Cardiotoxic cohort — clinical details (n=21)
Recruited from UMCU (n=18) + 2 additional centres (n=3). Median time chemotherapy → LVAD: 132 months (11 years), range 12–456 months.
Patient
Age
Sex
Cancer
Dose mg/m²
Months → LVAD
LVEF pre-LVAD
Genetics
1
60
F
NHL
400
288
—
—
2
56
M
NHL
400
96
—
—
3
67
M
NHL
400
168
10%
—
4
36
F
Osteosarcoma
400
276
10%
—
5
64
F
Breast
360
180
0.10%
VUS FLNC
6
64
M
Lymphoma
400
156
0.12%
—
7
40
M
Sarcoma
540
432
0.25%
VUS FLNC
8
57
F
Breast
300
72
0.20%
—
9
44
M
NHL
311*
456
—
—
10
44
M
Lymphoma
400
72
0.12%
—
11
44
F
Breast
300
12
—
—
12
47
F
Breast
366†
36
—
VUS MYL3‡
13
52
F
NHL
400
228
—
—
14
48
F
Breast
240
12
—
—
15 ⚠ outlier
52
F
Breast
240
15
10%
—
16
54
F
Breast
240
132
—
—
17 ⚠ outlier
26
M
Sarcoma
240
48
0.10%
—
18
64
F
Breast
240
12
10%
—
CTP1
57
F
Breast
240
—
—
—
CTP3
45
F
Breast
270
—
—
—
CL046
63
F
Breast
240
—
—
—
* Epirubicin 510 mg/m² = 311 dox-equivalent. † Epirubicin 600 mg/m² = 366 dox-equivalent. ‡ VUS MYL3 assessed likely non-pathogenic. Pt15: copy number IQR outlier (20.85). Pt17: indels/1000 IQR outlier (6.31). Sensitivity analyses: removing these does not change any qualitative conclusion.
Outlier sensitivity
Removing Patient 9 (lowest mutation burden, 456 months to LVAD) actually strengthens the Cardiotoxic vs Healthy SNV result from p = 0.039 to p = 0.020 — confirming results are conservative, not inflated.
DCM genetic mutations
Sample
Gene(s)
Age
Sex
Pathway
CL025
MYBPC3
52
M
Sarcomeric DCM
CL026
TTN, TNNC1
34
M
Compound sarcomeric
CL030
DSC2
57
M
Arrhythmogenic DCM
CL042
LAMP2
50
F
Danon disease (lysosomal)
CL045
PLN R14del
59
M
Dutch founder mutation
CL047 ⚠
PLN R14del
63
F
Dutch founder mutation
CL056
TTN
54
M
Sarcomeric DCM
CL060
PLN R14del
59
F
Dutch founder mutation
CL064
MYH7, PRDM16
17
F
Compound sarcomeric
CL071
PLN R14del
70
F
Dutch founder mutation
Section 2
Sequencing Quality Control
Dedicated targeted mtDNA sequencing (CD Genomics, capture-probe, Illumina). QC data available for 64/73 samples.
Mean target depth
12,821×
Range 4,819× – 27,990×
Coverage ≥30× (all samples)
100%
Complete mtDNA genome
Mean capture rate
91.5%
Range 81.8 – 96.1%
Mean duplication rate
32.5%
Range 19.3 – 44.1%
Interpretation
Complete mtDNA genome coverage at extreme depth in all samples. All mutation rates are reported normalised per 1,000 bases to account for inter-sample depth variation.
All groups compared to Healthy (Mann-Whitney U, two-sided). * p<0.05** p<0.01trend p<0.10
Copy number — Kruskal-Wallis H = 25.84, p < 0.0001
Group vs Healthy
n
Mean
Median
U
p-value
Cardiotoxic
21
13.56
12.84
172.0
0.026
* ↑
DCM nonmutant
10
12.13
11.33
55.0
1.000
ns
DCM gen. mutation
10
10.37
10.29
28.0
0.029
* ↓
Key finding
Cardiotoxic: elevated copy number = compensatory biogenesis (PGC-1α) response to anthracycline oxidative damage. DCM gen. mutation: opposite direction — depleted copy number = primary mitochondrial depletion from sarcomeric/channel dysfunction. Two mechanistically distinct failure modes visible at the mtDNA level.
SNV burden — Kruskal-Wallis H = 8.39, p = 0.078
Group vs Healthy
n
Mean
Median
U
p-value
Cardiotoxic
21
14.13
15.33
168.0
0.039
*
DCM nonmutant
10
15.36
15.29
21.0
0.018
*
DCM gen. mutation
10
13.35
12.78
34.0
0.149
ns
Indel burden — Kruskal-Wallis H = 14.44, p = 0.006
Group vs Healthy
n
Mean
Median
U
p-value
Cardiotoxic
21
2.69
1.94
172.0
0.026
*
DCM nonmutant
10
2.03
1.90
30.0
0.085
trend
DCM gen. mutation
10
2.65
2.26
14.0
0.004
**
Convergence finding
Indel accumulation is significantly elevated in both Cardiotoxic and DCM gen. mutation vs Healthy, despite different upstream mechanisms (POLG disruption vs replication fidelity failure). These two groups are statistically indistinguishable from each other — suggesting converging mtDNA genome instability as a final common pathway in end-stage HF.
Total mutation burden — Kruskal-Wallis H = 8.31, p = 0.081
Group vs Healthy
n
Mean
Median
U
p-value
Cardiotoxic
21
16.82
16.34
168.0
0.039
*
DCM nonmutant
10
17.40
16.96
22.0
0.022
*
DCM gen. mutation
10
15.99
14.65
31.0
0.098
trend
Section 9
Insertion / Deletion Balance
Kruskal-Wallis on insertion fraction: H = 1.45, p = 0.694 — not significant. All groups trend toward slight insertion predominance. High within-group variance reflects individual mtDNA haplogroup effects; not useful as a disease discriminator.
Group
n
Ins. fraction (%)
Del. fraction (%)
Ins/Del ratio
Healthy
11
56.8
43.2
1.60
Cardiotoxic
21
51.4
48.6
1.62
DCM nonmutant
10
51.3
48.7
1.35
DCM gen. mutation
10
60.3
39.7
2.46
Section 10
Cardiotoxic Group — Internal Stratifications
10.1 By cancer type
Statistical viability warning
NHL n=3 (limited power), Sarcoma n=2 and Osteosarcoma n=1 are not testable. Breast cancer (n=11) is the only subgroup with reasonable power.
Cancer
n
Copy number
SNVs/1000
Indels/1000
Mut. burden
Status
Breast
11
13.98
15.01
2.46
17.47
Interpretable with caution
Lymphoma
4
12.06
14.36
2.71
17.07
n=4 · limited
NHL
3
14.29
10.55
3.10
13.65
n=3 · limited
Sarcoma
2
13.94
15.53
4.05
19.58
n=2 · descriptive
Osteosarcoma
1
11.89
11.39
1.20
12.59
n=1 · not testable
10.2 High vs low cumulative doxorubicin dose (vs Healthy)
Dose group
n
Copy number (p)
SNVs/1000 (p)
Mut. burden (p)
High (≥ 300 mg/m²)
13
13.28 0.018*
13.64 0.132 ns
16.27 0.105 ns
Low (< 300 mg/m²)
5
15.99 0.009**
15.83 0.038*
19.03 0.052 trend
Apparent paradox — dose confounded by progression speed
Low-dose patients show higher copy number, SNVs, and mutation burden. This is a confound: dose and time-to-LVAD are strongly correlated (ρ = 0.667, p = 0.003). High-dose patients took a median of 180 months to LVAD; low-dose only 24 months. See Sensitive Phenotype Analysis below.
10.3 Time from chemotherapy to LVAD
Time group vs Healthy
n
Copy number (p)
Indels/1000 (p)
Mut. burden (p)
Rapid (< 36 months)
4
16.79 0.026*
2.32 0.343 ns
16.63 0.138 ns
Slow (≥ 36 months)
14
13.24 0.011*
2.92 0.035*
17.15 0.067 trend
Indel accumulation reaches significance only in the long-interval group — consistent with progressive mtDNA instability continuing for years after the anthracycline insult.
Section 11
DCM Mutation Type Stratification
Statistical viability — all gene-level subgroups
Only PLN R14del (n=4) has limited-power inference. All others (n=1) are not testable. All values descriptive only.
Gene
n
Copy number
SNVs/1000
Indels/1000
Mut. burden
Status
PLN R14del
4
10.84
11.54
2.21
13.74
n=4 · limited
TTN
1
10.59
13.07
1.83
14.91
n=1
LAMP2
1
10.02
14.57
2.52
17.10
n=1
MYBPC3
1
9.65
18.82
1.83
20.65
n=1
TTN / TNNC1
1
10.41
16.90
3.25
20.14
n=1
DSC2
1
9.54
16.93
4.56
21.49
n=1
MYH7 / PRDM16
1
10.18
7.02
3.68
10.70
n=1
PLN R14del — hypothesis-generating observation
The 4 PLN R14del patients show near-Healthy SNV rates (11.54 vs 10.67) and the lowest mutation burden of any DCM subgroup (13.74). PLN R14del causes DCM via SERCA2a calcium overload rather than primary mitochondrial dysfunction — this may explain relatively preserved mtDNA integrity despite end-stage HF. Requires replication.
Section 12
In Vitro CTP2-C5 Cardiomyocytes
iPSC-CM line from Cardiotoxic Patient 2 (NHL, 400 mg/m², 96 months to LVAD). MTT15/MTT50: received mitochondrial transfer from Resilient donor line (15 µg and 50 µg per 6-well plate). Seahorse confirms improved energy production after transfer.
Control vs dox (n=11 vs 10, Mann-Whitney)
Treatment
n
Copy number
SNVs/1000
Indels/1000
Mut. burden
All p-values
Control
11
10.29
15.09
2.74
18.26
ns for all metrics
Dox-treated
10
9.77
15.20
2.70
17.92
Subgroup statistical viability
Each MTT × treatment cell has n = 2–3. No inferential testing possible. Overall ctrl vs dox (n=11 vs 10) is the only valid comparison in this section.
Critical negative finding
Doxorubicin at cytotoxic concentrations (MTT50 ≈ 50% cell death) produces no detectable change in any mtDNA parameter in CTP2-C5 cardiomyocytes. This contrasts sharply with human Cardiotoxic tissue. Possible explanations: (1) acute vs chronic exposure — tissue reflects decades of damage; (2) absent systemic co-factors (neurohormonal, inflammatory); (3) iPSC-CM metabolic immaturity. This dissociation indicates the human LVAD tissue captures unique biology inaccessible by current in vitro models.
Section 14
Integrated Summary & Cardio-Oncology Relevance
Principal findings ranked by statistical strength
↑ Copy number — Cardiotoxic vs Healthy (p=0.026*). Compensatory biogenesis response to anthracycline oxidative damage.
↓ Copy number — DCM gen. mutation vs Healthy (p=0.029*). Primary mitochondrial depletion. Opposite direction to Cardiotoxic.
↑↑ Indels — DCM gen. mutation vs Healthy (p=0.004**). Strongest individual finding. Elevated indels despite low copy number.
↑ SNVs + Mut. burden — Cardiotoxic vs Healthy (p=0.039* both). ROS-driven mtDNA point mutations in anthracycline-damaged myocardium.
↑ SNVs + Mut. burden — DCM nonmutant vs Healthy (p=0.018*, 0.022*). Oxidative accumulation in non-genetic DCM.
Progressive indels over time — Cardiotoxic: slow-interval patients (≥36 months) have significantly elevated indels vs Healthy (p=0.035*). mtDNA instability continues years after chemotherapy.
No dose-response. After controlling for time-to-LVAD (confounded with dose, ρ=0.667), doxorubicin dose has no independent effect on any mtDNA parameter.
In vitro ≠ in vivo. No mtDNA changes in CTP2-C5 cells at any dox dose. The human LVAD tissue captures unique biology requiring systemic in vivo context.
Two-axis mechanistic model
Group
Biogenesis axis (copy number)
Mutational axis (SNVs + indels)
Interpretation
Cardiotoxic
↑↑ Elevated
↑ Both elevated
Compensatory biogenesis + oxidative + POLG damage
DCM gen. mutation
↓ Depleted
↑↑ Indels strongly elevated
Mitochondrial depletion + replication infidelity
DCM nonmutant
→ No change
↑ SNVs elevated
ROS-driven point mutations, no biogenesis response
Healthy
→ Reference
→ Reference (lowest)
Biological baseline
Limitations
Sample size (n=10–21): Findings are hypothesis-generating; replication required before clinical translation.
Single timepoint (end-stage LVAD): Cannot determine trajectory or causality without longitudinal sampling.
No heteroplasmy quantification: Cannot distinguish homoplasmic vs heteroplasmic variants — critical for mechanistic interpretation.
NuMT contamination: MT/autosomal read ratio may be influenced by nuclear mitochondrial DNA segments. ddPCR recommended for confirmation.
Cardiotoxicity defined post hoc: LVAD need captures only the most severe end of the anthracycline cardiotoxicity spectrum.
VUS reclassification risk: Patients 5 and 7 carry FLNC VUS. Reclassification as pathogenic would change group assignment.
The Dose Paradox: Sensitive vs Resilient Phenotypes
Low-dose patients showed higher mtDNA copy number, SNVs, and mutation burden than high-dose patients — counter to a simple dose-damage model. Partial correlation analysis reveals this is a confound driven by progression speed, not dose.
Dose × time correlation
ρ = 0.667
Spearman, p = 0.003
High dose → median latency
180 mo
15 years to LVAD
Low dose → median latency
24 mo
2 years to LVAD
Dose effect (partial r)
−0.45
Controlling for time, p=0.063
Partial correlations (n=18)
Controlling for time → Dose vs mtDNA
Outcome
Partial r
p
Copy number
−0.447
0.063
trend
SNVs/1000
+0.124
0.624
ns
Indels/1000
−0.055
0.829
ns
Mut. burden
+0.070
0.784
ns
Controlling for dose → Time vs mtDNA
Outcome
Partial r
p
Copy number
−0.010
0.969
ns
SNVs/1000
−0.433
0.073
trend
Indels/1000
−0.068
0.789
ns
Mut. burden
−0.345
0.161
ns
2 × 2 breakdown: dose × progression speed
Statistical viability
Cells with n=2 are descriptive only.
Dose × Progression
n
Copy number
SNVs/1000
Mut. burden
Status
High dose + Rapid (≤36 mo)
2
15.63
17.70
21.55
n=2
Low dose + Rapid (≤36 mo)
3
16.33
13.97
15.60
n=3
High dose + Slow (≥36 mo)
11
12.85
12.91
15.31
Main estimable cell
Low dose + Slow (≥36 mo)
2
15.47
18.60
24.18
n=2
Revised hypothesis
The "sensitive phenotype" is better characterised as individuals with a low heteroplasmy threshold who reach pathogenic mtDNA dysfunction rapidly after any anthracycline exposure, regardless of dose. They present as rapid progressors (LVAD within 1–4 years). Their elevated copy number is an acute biogenesis stress response. High-dose patients are the resilient phenotype: sufficient mitochondrial reserve to tolerate higher doses while delaying decompensation by 5–38 years. Individual susceptibility — not dose — is the primary determinant.
Interactive Model
Heteroplasmy Threshold Simulator
Models how baseline heteroplasmy, pathogenic threshold, dox dose, and annual drift interact to predict clinical cardiotoxicity timing. Dox dose = immediate heteroplasmy jump proportional to dose (400 mg/m² ≈ +18 percentage points). Annual drift = ongoing stochastic accumulation post-treatment.
Baseline heteroplasmy (%)20%
Pathogenic threshold (%)70%
Dox dose (mg/m²)300
Annual drift rate (%/yr)2.0%
Predicted phenotype
—
Post-dox heteroplasmy
—
Years to threshold
—
Predicted LVAD latency
—
Adjust sliders to run the simulation.
Scatter: individual patients — dose vs time to LVAD (bubble size = copy number)