Nuclear Mitochondrial DNA Segment (NUMT)

A nuclear mitochondrial DNA segment (NUMT, pronounced "new-might", a term coined by Jose Lopez and colleagues in 1994) is a transposed fragment of cytoplasmic mitochondrial DNA (mtDNA) that has been physically integrated into the linear nuclear chromosomes of a eukaryote. Across the human reference genome (GRCh38 / T2T-CHM13), more than 750 to 1,000 fixed and polymorphic NUMT insertions have been cataloged, ranging in size from short 30 bp fragments up to complete, concatenated 16.6 kb full-mitogenome insertions ("mega-NUMTs") and collectively spanning over 600 kilobases (~0.02% of the human nuclear genome). Because vertebrates and most metazoans utilize a modified mitochondrial genetic code (where UGA encodes tryptophan and AUA encodes methionine) and lack the nuclear promoter and splicing signals required to express mitochondrial genes inside the nucleoplasm, inserted NUMTs are immediately rendered non-functional nuclear pseudogenes upon integration.

In evolutionary genomics, molecular forensics, and biodiversity metabarcoding, NUMTs play a dual role: on one hand, because the neutral substitution rate of nuclear DNA (~2 × 10⁻⁹ per site per year) is roughly 10-fold slower than that of cytoplasmic vertebrate mtDNA (~2 × 10⁻⁸ per site per year), ancient NUMTs persist as "molecular fossils" of ancestral mitogenomes that diverged millions of years ago. On the other hand, when universal PCR primers targeting mitochondrial cytochrome c oxidase subunit I (COX1) or MT-RNR2 inadvertently co-amplify nuclear NUMTs alongside cytoplasmic mtDNA, the resulting paralogous reads masquerade as cryptic species in phylogenetics or false heteroplasmy in clinical mitochondrial disease diagnostics.

Mitochondrial Leakage and Non-Homologous End Joining (NHEJ) at Double-Strand Breaks

Unlike retrotransposons (such as Alu or LINE-1 elements), which transpose via a reverse-transcribed RNA intermediate, sequence analysis of NUMT junctions demonstrates that NUMTs integrate directly as double-stranded DNA fragments without an RNA intermediate (lacking 3′ poly(A) tails or intron splicing). The transfer of mtDNA from the mitochondrial matrix to the nuclear chromosomes proceeds through a three-step pathway:

1. Mitophagy and Organellar Membrane Rupture: Oxidative stress, mitophagic turnover, apoptosis-abortive pore formation (BAX/BAK macropores), or gametogenic mitochondrial degradation releases linear fragments of degraded mtDNA into the cytosol, which subsequently traverse the nuclear pore complex or enter the nucleoplasm during mitotic nuclear envelope breakdown.
2. Capture at Nuclear Double-Strand Breaks (DSBs): When ionizing radiation, replication fork stalling, or topoisomerase poisons create a chromosomal double-strand break (DSB) in a germline or somatic cell, the non-homologous end joining (NHEJ) and microhomology-mediated end joining (MMEJ) repair complexes (Ku70/Ku80, DNA-PKcs, XRCC4, and DNA Ligase IV) capture free cytoplasmic mtDNA fragments as "filler DNA" to bridge the broken chromosomal ends.
3. Junction Signatures: High-resolution sequencing of NUMT–nuclear boundaries reveals 1 to 7 bp microhomologies in ~55% of insertions, blunt-end ligations in ~35%, and small target-site deletions prevention—showing that NUMT insertion actively prevents larger deleterious deletions during DSB repair.

Ongoing Germline vs. Oncogenic Somatic NUMTogenesis

Comparative primate genomics shows that NUMT integration is an ongoing evolutionary process: approximately 15% to 20% of human NUMTs are human-specific (inserted after the Homo–Pan split ~6 million years ago), while large-scale whole-genome sequencing of >60,000 human parent–offspring trios demonstrates that de novo germline NUMTs arise at a rate of roughly 1 new insertion per 4,000 to 10,000 live births. Furthermore, somatic cancer genomes (particularly breast, lung, and colorectal adenocarcinomas with homologous recombination deficiency) exhibit elevated somatic NUMTogenesis, occasionally inserting mtDNA fragments into tumor-suppressor exons (such as TP53 or APC).

In cats (Felis catus), a tandemly repeated 7.9 kb "Lopez mega-NUMT" duplicated 38 to 76 times on Chromosome D2, whereas in domestic horses, gorillas, and certain subterranean rodents, dozens of distinct NUMT paralogs exist for every mitochondrial gene.

Neutral Nuclear Drift (ω ≈ 0.89) vs. Mitochondrial Purifying Selection (ω ≈ 0.031)

Once an mtDNA fragment is ligated into a nuclear chromosome, its evolutionary trajectory diverges sharply from its cytoplasmic parent molecule:

Because functional cytoplasmic COX1 must fold into 12 transmembrane helices to pump protons in Complex IV, non-synonymous mutations at first and second codon positions are eliminated by purifying selection (ω = d_N/d_S ≈ 0.031, second-position Shannon entropy H_pos2 = 0.000 bits), while third-codon positions accumulate rapid synonymous transitions (H_pos3 = 1.842 bits).

By contrast, a nuclear NUMT experiences no selective constraint on its amino acid sequence: mutations occur randomly across all three codon positions (ω ≈ 0.890, with H_pos2 rising to 0.945 bits), frameshift insertions and deletions (indels) accumulate without penalty, and spontaneous deamination of methylated nuclear 5-methylcytosine (5mC) at CpG dinucleotides generates C→T (G→A) transitions that shift the base composition toward the ~41% GC average of nuclear chromatin.

How NUMTs Masquerade as Cryptic Species and Clinical Heteroplasmy

In PCR-based DNA barcoding, environmental DNA (eDNA) metabarcoding, and clinical mitochondrial sequencing, universal primers annealing to conserved regions of COX1, MT-CYB, or MT-RNR2 can simultaneously amplify both the cytoplasmic target and one or more nuclear NUMTs. In fact, because ancient NUMTs evolve 10 times more slowly than cytoplasmic mtDNA, universal primers designed across broad taxonomic groups sometimes match the ancestral sequence preserved inside the nuclear NUMT with fewer mismatches than they match the rapidly diverged cytoplasmic mitogenome ("preferential NUMT amplification").

Unfiltered NUMT co-amplification produces two systematic errors:

1. Taxonomic Over-Splitting (False Cryptic Species): In Neighbor-Joining, Maximum Likelihood, or Automatic Barcode Gap Discovery (ABGD) pipelines, a NUMT that integrated 3 to 8 million years ago diverges from the host's modern cytoplasmic COX1 by 3% to 9% Kimura two-parameter (K2P) distance—falling squarely inside the interspecific barcoding range and splitting a single biological species into two or more artificial "cryptic species" clades.
2. False Clinical Heteroplasmy and Paternal Inheritance Artifacts: In next-generation sequencing (NGS) panels for mitochondrial diseases, short sequencing reads (150 bp) mapped from nuclear NUMTs onto the 16,569 bp human mitochondrial reference sequence (rCRS) create apparent low-frequency variant alleles (1%–15% variant allele fraction) that mimic somatic or inherited mtDNA heteroplasmy, or falsely suggest biparental transmission when a child inherits a rare paternal nuclear mega-NUMT.

Five-Step Bioinformatic and Experimental Rescue Protocol

To eliminate NUMT paralog contamination in biodiversity and clinical pipelines, rigorous diagnostic workflows enforce five quantitative filters:

1. Codon-Position Entropy & Reading-Frame Translation Gate: Translate every protein-coding COX1 or MT-CYB amplicon using the taxon-appropriate mitochondrial genetic code (e.g., NCBI Translation Table 2 for vertebrates, Table 5 for invertebrates) and verify zero in-frame stop codons, zero non-triplet frameshift indels, second-codon entropy H_pos2 = 0.000 bits, and ω = d_N/d_S < 0.10.
2. Paired COX1 + MT-RNR2 Concordance: Amplify both protein-coding COX1 (658 bp) and structural MT-RNR2 (1,559 bp, 16S mtDNA) and check for topological phylogenetic concordance (5.82× coupled barcoding gap) plus conservation of Watson–Crick stem base pairing in MT-RNR2.
3. Long-Range PCR and Whole-Mitogenome Circularization: Amplify mtDNA in overlapping long-range amplicons (8 to 16.6 kb) or perform long-read sequencing with topological circularity verification, which excludes linear nuclear fragments shorter than the full mitogenome.
4. RT-PCR from Polyadenylated cDNA or Alkaline/Exonuclease V Enrichment: Treat genomic DNA extracts with Exonuclease V (RecBCD) (which digests linear nuclear chromosomes while sparing closed circular mtDNA), or reverse-transcribe polyadenylated mitochondrial mRNA, as nuclear NUMTs are transcriptionally silent.
5. Nuclear-Masked Reference Alignment: Map clinical short-read sequencing libraries simultaneously against a joint nuclear + mitochondrial reference (GRCh38 + rCRS) rather than rCRS alone, discarding reads that map with higher or equal quality to known nuclear NUMT coordinates.

How did a nuclear NUMT cause early researchers to think they had sequenced DNA from a Cretaceous dinosaur?

In 1994, a widely publicized study reported the amplification of an 80-base-pair mitochondrial cytochrome b sequence from an 80-million-year-old Cretaceous bone fragment, which clustered separately from living birds, reptiles, and mammals on phylogenetic trees. Subsequent re-analysis by Zischler, Hedges, Pääbo, and colleagues proved that the "dinosaur" sequence was actually a human nuclear mitochondrial pseudogene (NUMT) located on human Chromosome 11 that had contaminated the PCR reaction: because the human NUMT had integrated into the primate nuclear genome tens of millions of years ago and evolved 10 times more slowly than modern human cytoplasmic mtDNA, it branched near the base of amniotes and mimicked an ancient extinct lineage.

Why do universal PCR primers sometimes amplify a nuclear NUMT even though cells have thousands of times more cytoplasmic mtDNA?

Although a typical somatic cell contains 1,000 to 10,000 copies of cytoplasmic mtDNA versus only 2 diploid copies of a given nuclear NUMT, cytoplasmic mtDNA accumulates point mutations 10 times faster than nuclear DNA. If a species acquires a single third-codon substitution near the 3′ end of a universal COX1 primer binding site on its cytoplasmic mtDNA, Taq polymerase extension efficiency drops by 100- to 1,000-fold, whereas the ancient, unmutated primer binding site preserved inside the nuclear NUMT (or inside a 50-copy tandemly duplicated mega-NUMT) amplifies during the exponential growth phase of PCR without mismatch penalty.

Can a newly formed NUMT ever become a functional nuclear gene?

In animals, newly inserted NUMTs are almost universally dead-on-arrival pseudogenes because the vertebrate mitochondrial genetic code reassigns UGA from a stop codon to tryptophan (Trp); if a mitochondrial open reading frame is transcribed and translated by cytoplasmic 80S ribosomes, every internal UGA codon causes premature termination. By contrast, in flowering plants (which use the standard universal genetic code inside their mitochondria) and certain fungi, ancient mitochondrial gene transfers into the nucleus occasionally acquired nuclear promoters and mitochondrial targeting presequences over evolutionary time, completing functional endosymbiotic gene transfer.

1. Lopez, J. V., Yuhki, N., Masuda, R., Modi, W., & O'Brien, S. J. (1994). "Numt, a recent transfer and tandem amplification of mitochondrial DNA to the nuclear genome of the domestic cat." Journal of Molecular Evolution, 39(2), 174–190. https://doi.org/10.1007/BF00163806
2. Bensasson, D., Zhang, D. X., Hartl, D. L., & Hewitt, G. M. (2001). "Mitochondrial pseudogenes: evolution's misplaced witnesses." Trends in Ecology & Evolution, 16(6), 314–321. https://doi.org/10.1016/S0169-5347(01)02151-602151-6)
3. Hazkani-Covo, E., Zeller, R. M., & Martin, W. (2010). "Molecular poltergeists: mitochondrial DNA copies (numts) in sequenced nuclear genomes." PLOS Genetics, 6(2), e1000834. https://doi.org/10.1371/journal.pgen.1000834
4. Wei, W., Schon, K. R., Elgar, G., Orioli, A., Tanguy, M., Giess, A., Tischkowitz, M., Caulfield, M. J., & Chinnery, P. F. (2022). "Nuclear-embedded mitochondrial DNA sequences in 66,083 human genomes." Nature, 611(7934), 105–114. https://doi.org/10.1038/s41586-022-05288-7