Review Papers Claim by Claim

Visual paper analysis — Reznik et al., "Mitochondrial DNA copy number variation across human cancers" (eLife 2016)

Figure: illustrative reconstruction of Reznik et al.'s main observation — multiple cancer types (notably bladder and clear‑cell kidney) show tumor mtDNA depletion vs matched normal; lung adenocarcinoma showed accumulation. These are summary/illustrative numbers derived from their reported distributions and text (see Reznik et al. 2016) and should be treated as schematic, not raw data plots. All primary claims below link to original sources.

High-level strengths

• Scale: multi-center TCGA WXS/WGS across 22 tumor types and 1090 matched tumor/normal pairs — gives statistical power rare at the time Reznik et al. TCGA-scale study
• Method transparency: authors supply formula, purity/ploidy correction (ASCAT), plate correction, and make copy-number estimates available in Supplementary File 1 for reuse — good reproducibility practice Methods & data availability in Reznik et al.
• Cross-modal checks: correlation between mtCN and mitochondrial transcription, survival associations in several cancers, and focused IHC (mitochondrial protein) to support biological conclusions in subsets.

Primary weaknesses, biases and blindspots

• Bulk tissue confounding: bulk DNA includes stromal/immune cells with distinct mtCN; ESTIMATE-derived immune/stromal scores show weak correlations but deconvolution was not performed — so tumor-cell–intrinsic mtCN is uncertain Bulk tissue confounding noted by Reznik et al.
• WXS batch effects: exome library capture differences cause plate/center-dependent mt read depth; authors corrected plate effects for WXS but residual bias remains plausible — WGS less affected (authors' own analysis) Batch effects and correction (Reznik et al.)
• Correlative claims: associations with somatic mutations (IDH1 in LGG, PTEN lower mtCN, UCEC CNA-high higher mtCN) are statistical signals that need functional validation; direction of causality not established.
• Limited functional follow-up: IHC was performed on only 3 tumor:normal pairs for three tumor types — too small to generalize compensation vs transcription regulation hypotheses.
• NUMT and mapping artefacts: authors used strict MAPQ and properly-paired filters to reduce NUMT contamination, but WXS capture biases and NUMTs remain potential confounders in read-depth-based mtCN estimation — a recognized community pitfall JGG review on mtDNA analysis pitfalls

Context since 2016 — how this paper sits in the field

• PCAWG/TCMA work (Nature Genetics 2019) greatly expanded and deepened mtDNA analyses with 2,658 genomes and ultra‑deep mtDNA coverage, confirming tissue‑specific mtDNA copy number distributions, cataloguing somatic mtDNA mutations, and offering a dedicated portal (TCMA) — this later work complements and largely validates the primary observations in Reznik et al. while providing more resolution and filtering for NUMTs and artefacts PCAWG/TCMA 2019 (Nat Genet)
• Methodological consensus has shifted toward WGS, rigorous NUMT filtering, and single-cell or enrichment-based assays for per-cell mtCN/heteroplasmy resolution; single‑cell mtDNA tools and mtscATAC / Cryo‑mtscATAC approaches now address many bulk confounders (see recent single‑cell reviews and protocols) Single-cell mtDNA technologies review (2025)

Concrete recommendations (for researchers building on Reznik et al.)

1. Re-run mtCN estimates using WGS where possible; when using WXS, always model plate/center and provide uncorrected/corrected values (Reznik did this — good) but validate with orthogonal assays (qPCR, ddPCR) on the same samples Reznik methods transparency
2. Use single-cell mtDNA genotyping or enriched mtRNA expression (MAESTER, mtscATAC, Cryo‑mtscATAC) to deconvolve tumor vs immune/stromal mtCN contributions (recent protocols enable this) Cryo-mtscATAC-seq protocol (2025)
3. For functional follow-up: manipulate mtCN (e.g., TFAM knockdown, POLG perturbation, or transmitochondrial cybrids) and measure OXPHOS, ROS, immune signaling (cGAS/STING/interferon) and tumor phenotype in vitro/in vivo; correlate with mtCN and heteroplasmy load (papers show compensation effects are possible) Functional follow-up recommendations review

Where Reznik et al. could be falsified

Direct falsification paths:

• High-coverage WGS of the same tumor/normal pairs with orthogonal mtCN quantification (qPCR/ddPCR) showing no systematic tumor depletion across the reported cancer types would contradict Reznik's main claim.
• Single-cell mtDNA genotyping revealing that apparent bulk mtDNA depletion arises entirely from loss/gain of infiltrating immune/stroma cells rather than tumor-cell intrinsic changes would overturn tumor-intrinsic conclusions.

Short checklist for reuse / meta-analysis

• Pull both corrected and uncorrected mtCN from Supplementary File 1 (Reznik 2016) and match sequencing platform & plate IDs.
• Prefer WGS-derived estimates when available; if using WXS, include plate correction covariates and validate with qPCR in subsets.
• Account for tumor purity (ASCAT) and include ESTIMATE immune/stroma as covariates; where possible use single-cell data to partition signals.

Selected evidence citations (key papers referenced in this review)

• Reznik et al. 2016
• PCAWG/TCMA 2019
• JGG review on pitfalls
• Oral SCC mtDNA profiling (2019)
• Review on mtDNA copy number in disease (2020)

Actionable next steps (if you are the investigator)

1. Re-analyse a subset of Reznik's paired samples using available WGS (where present) and orthogonal qPCR/ddPCR on DNA aliquots to quantify per‑cell mtCN concordance.
2. Perform single-cell mtDNA genotyping (MAESTER/mtscATAC or Cryo‑mtscATAC) on tumors showing depletion (e.g., KIRC, BLCA) to partition tumor vs infiltrate contributions and measure per‑cell heteroplasmy and mtCN distributions.
3. In model systems, perturb mtCN (TFAM/POLG knockdown or controlled repletion) and measure OXPHOS, ROS, cGAS/STING, and immune signaling to test mechanistic consequences for tumor growth and immune microenvironment.