Chromosome 2 Fusion

"The data we present here demonstrate that a telomere-to-telomere fusion of ancestral chromosomes occurred, leaving a pathognomonic relic at band 2q13." — IJdo, Baldini, Ward, Reeders and Wells, 1991¹

"Based on the predicted model, thousands of intact TTAGGG motifs in tandem should exist." — Jeffrey Tomkins and Jerry Bergman, 2011²

"Just one fusion event of the total 298 sequenced revealed telomere repeats on each side of the fusion point" — Letsolo, Rowson and Baird, on real telomere fusions in human cells, 2010³

Summary

Humans have 46 chromosomes.  Chimpanzees, gorillas and orangutans have 48.  In 1982 Jorge Yunis and Om Prakash compared banded chromosomes across the four species and proposed that human chromosome 2 is two ape chromosomes joined end to end.⁴  That was a prediction.  It said that somewhere on the long arm of chromosome 2 there should be two telomeres stuck together facing each other, and further along there should be a centromere that stopped working.  This article asks whether it came true, what the creationist replies get right, and what a fusion does and does not show.

What both sides agree on.

1. The junction exists, and everyone measures it the same way. In the hg38 reference the head-to-head junction is at chr2:113,602,928. Computed for this article from the reference bytes, the array is about 834 bp long and parses into about 148 degenerate repeat units, of which 39% are perfect. Fan and colleagues counted 127 to 158 units at 46 to 48% perfect in 2002;⁵ Tomkins and Bergman printed a 798 bp core in 2011 that matches hg38 base for base; the 2026 telomere-to-telomere assembly gives 875 bp.⁶ Nobody disputes the data.
2. The dead centromere exists, and the creationist and mainstream measurements agree to two base pairs. Tomkins in 2017: 41,608 bases, containing an L1PA3 insertion of 5,957 bases and an SVA-E element of 2,571 bases.⁷ Yang and colleagues in 2026: chr2:132,644,386 to 132,685,996 in the T2T assembly, which is 41,610 bases, three alpha-satellite arrays "interrupted by various simple repeats (CCTCTC) and retrotransposon elements (SVA and L1PA3) in the human lineage."⁶ Same structure, same two named insertions, from two different assemblies.
3. The array sits inside an intron of DDX11L2, and the locus is transcribed. Tomkins said so in 2013. Checked here against hg38 RefSeq, the array lies wholly between two exons of transcript NR_024004.2, and Yang confirms expression from GTEx.⁶

Where the disagreement is, and who has checked.

4. The creationist case says a real fusion should have left thousands of intact repeats. Real fusions do not. Duncan Baird's lab sequenced 298 telomere-to-telomere fusions in living human cells. Between 38% and 60% had no telomere repeat at the junction at all. Those that did averaged 5.8 repeats. One of the 298 had repeats on both sides.³ Chromosome 2 has about 148 units and repeats on both sides. Against the only real yardstick, 2q13 has about 25 times too much telomeric repeat for a fresh fusion, not too little: 834 bp against a mean of about 35 bp. And a human telomere stays capped and unfusable until it has eroded below about thirteen repeats, so a junction with thousands of intact repeats is not a fusion missing something. It is a fusion that could not have happened.⁸ Seven sequenced ancestral fusions in the Indian muntjac retain 168 to 837 bp.⁹ Jeffrey Tomkins compares the site with natural fusions in other mammals, which join at centromeres and carry satellite DNA, and sets aside telomere fusions in cancer cells as products of failed genome maintenance.¹⁰¹⁰ No creationist treatment has compared the amount of repeat at 2q13 with any measured fusion junction. That is the largest gap on the creationist side.
5. Tomkins's mechanism for why the array should be preserved rests on a false premise. He argues that recombination is suppressed near the centromere, so the array should have decayed slowly.² The junction is 17.6 Mb from the centromeric bands, in 2q14.1, in ordinary recombining DNA at 0.85 cM/Mb, 33 times the rate of a genuinely pericentromeric window on the same chromosome. Computed here from the UCSC cytoband and deCODE tracks.
6. Tomkins's chromosome-wide count of telomere hexamers is right and cuts against him. He counted 45,450 TTAGGG on chromosome 2 to show the motif is everywhere. An independent count here gives 45,648. Chance alone predicts 52,539. Chromosome 2 has 13% fewer telomere hexamers than random sequence; the 2q13 array is about 900 times enriched at one spot.
7. DDX11L2 is not a gene a fusion could not land in. It is a three-exon fragment of the 26-exon DDX11 gene, with 1,286 bp of exons 19 to 21 deleted and replaced by 40 unrelated bases.¹¹ Every transcript is annotated non-coding. And the DDX11L family lives at chromosome ends: every member maps to a terminus except the one at the fusion site, and Costa's own figure legend calls that locus "one of the two ancestral telomeres fused head-to-head at 2q13-14."¹²
8. The dead centromere is homologous to the chimpanzee 2B centromere, and it was already dead in Neanderthals and Denisovans. Yang finds 96.7 to 97.5% of its unique k-mers in both archaic genomes.⁶ Karen Miga finds both the active and the degenerate chromosome-2 centromere profiles in both archaic genomes,¹³ and Pääbo's team found 12 Denisovan sequence fragments carrying the fusion junction, and none in chimpanzees or bonobos.¹⁴

The date, and what a fusion proves.

9. The mainstream disagrees with itself about when, and the smallest answer is still 66 times a 6,000-year earth. A substitution clock gives 0.74 million years, revised to 0.9 with a 95% interval of 0.4 to 1.5.¹⁵ Ventura's subterminal-satellite argument needs more than 4 million.¹⁶ Yang's incomplete-lineage-sorting analysis needs more than 5 million.⁶ The range runs from 400,000 to about 7 million years. Each date is computed on the mainstream timescale, and the lowest comes from a clock whose bonobo control returns a date where no fusion happened.¹⁵
10. A fusion in the human line does not, by itself, show descent from a chimp-human ancestor, and the site's condition does not date it. The young-earth biologist Todd Wood wrote in 2006 that the chromosome differences "could have occurred after creation", and that the site is hard to explain any other way than by fusion.¹⁷ The array's degeneracy does not refute that view: real fusions join at the worn inner edge of a telomere, and the matching stretch on an unfused human chromosome end is nearly as degenerate today.¹⁸ Nor does the 12% difference between two human alleles of the array.⁵ Finished genomes confirm it, and show that it runs into the plain DNA beside the array at 1 to 5%, which is more than human and chimpanzee differ, so slow mutation since the alleles split did not make it.¹⁹ What the young view still has to explain is the dead centromere's match to chimpanzee 2B, and that is a question about common ancestry, not about age.

Part 1: What is at the site

The prediction was made from pictures of stained chromosomes.  The test was done with DNA sequence, and it was done four times.

1991.  Johan IJdo and colleagues screened a human cosmid library with a (TTAGGG)n probe and pulled out two overlapping clones that mapped to band 2q13 and nowhere else interstitially.  Sequencing showed "two degenerate arrays of TTAGGG in an inverted (head-to-head) arrangement".¹  The flanks on either side of the array are themselves inverted relative to each other, and each matches sequence next to a real cloned telomere:  95% on the left, 90% on the right.¹  IJdo printed the word "degenerate" in 1991 and gave a reason for the shortness on the same page:  a long inverted repeat is structurally unstable, folds into cruciforms, and gets trimmed.  "We suggest such a phenomenon could have resulted in progressive shortening of this inverted sequence until relative stability was reached."¹  The mainstream discoverers described the site as short and degenerate in the paper that announced it.  The shortness objection was answered before it was raised.

2002.  Yuxin Fan, in Barbara Trask's lab, sequenced over 600 kb around the site from a BAC clone.  The result is the architecture that any explanation has to account for.  There are two head-to-head arrays "with no other sequence between the arrays."⁵  A 168 kb block on the centromere side matches 9pter, and also 9p11.2 and 9q13, at 98 to 99%.  A 67 kb block on the far side matches the end of chromosome 22.⁵  The match to those two chromosome ends runs to within 21 kb and 1.4 kb of the array.⁵  Fan also reports a physical gene-order test by Kasai and colleagues:  38 cosmids from human 2q12 to 2q14 hybridised in the same order on human chromosome 2 and on the short arms of chimpanzee chromosomes 12 and 13.⁵  Kasai's own paper says the same and adds a control that Fan does not report:  in the cynomolgus monkey the order of the clones on each homologue was inverted, so the method can detect a difference in order when there is one.²⁰

2005.  The finished chromosome 2 sequence placed the junction in a 15 bp interval and found the second half of the prediction:  a 2.6 Mb region at 2q21 enriched for duplications shared with the pericentromeric regions of ten other chromosomes, containing "three tracts totalling 31,198 bp of alpha-satellite sequence over 36,696 bp".²¹

2026.  Zikun Yang, Yafei Mao, Evan Eichler and colleagues rebuilt the site from finished assemblies of ten great-ape chromosomes and 94 human haplotypes.  The 109 kb site is three segmental duplications:  50 kb from ancestral chromosome 9, 36 kb from ancestral chromosome 12 carrying three exons of DDX11, and 22 kb related to chromosome 22.  The interstitial telomere sequences sit at chr2:114,027,333 to 114,027,658 and 114,027,659 to 114,028,207 in T2T-CHM13v2.0, adjacent and head to head.⁶

Measured here

Everything below was computed for this article from public reference data with two short scripts kept beside the research files.

The first script fetches chr2:113,600,000 to 113,606,000 from the UCSC sequence API and finds the point where G-rich telomere runs give way to C-rich ones.  Output, run 6 September 2026:

hg38 chr2  head-to-head junction at 113602928
  G-rich arm  113602652-113602928  (276 bp)
  C-rich arm  113602928-113603486  (558 bp)
  whole array 113602652-113603486  (834 bp)
  G-rich arm: 50 repeat units parsed, 19 perfect (38%)
  C-rich arm: 98 repeat units parsed, 39 perfect (40%)

The reference reads TTGGGGTTAGGGTTAG then CTAAACCTAACCCTAACCCC across the junction.  Tomkins and Bergman's printed 798 bp core sequence matches this reference base for base from position 113,602,679.

The second script counts exact TTAGGG and CCCTAA hexamers across the whole hg38 chromosome 2 FASTA and compares the count with what independent bases at the chromosome's own composition would produce.  Output:

length 242,193,529  non-N 240,548,228
GC 40.2%
TTAGGG observed 45,648  expected(iid) 52,539  obs/exp 0.87
CCCTAA observed 45,937  expected(iid) 51,966  obs/exp 0.88

Two things follow.  Tomkins's 2011 count of 45,450 and 45,770 on build 37.1 is right to within half a percent.²  And chromosome 2 carries fewer telomere hexamers than chance.  The background rate is one per 5,270 bp.  The array has one unit per 5.6 bp.  That is an enrichment of roughly 900-fold at one place on the chromosome, and nowhere else.

A third check, from hg38 RefSeq:  DDX11L2 transcript NR024004.2 runs chr2:113,599,027 to 113,603,717 on the minus strand, with an exon ending at 113,601,429 and the next beginning at 113,603,594.  The array, at 113,602,652 to 113,603,486, lies wholly between them.  Tomkins is right that the site is in an intron.  Two further things are also true.  Both DDX11L2 accessions are NR, the RefSeq prefix for a non-coding transcript.  And the shorter transcript, NR_024005.3, ends at 113,601,576 and does not contain the array at all.

Part 2: What a real fusion leaves behind

The creationist objection, in Tomkins and Bergman's words, is that "thousands of intact TTAGGG motifs in tandem should exist" if two telomeres really joined.²  Human telomeres are thousands of base pairs of TTAGGG, so joining two should leave a long, clean, inverted run.  The site has about 800 bp of damaged repeat, which Tomkins and Bergman call a "vague shadow of what should be present given the model proposed."²

Tomkins did ask what a real fusion leaves, and answered from natural fusions in other mammals, which join at centromeres (Part 3).  No creationist source has asked how much telomeric repeat a telomere-to-telomere junction keeps.  The measurement exists, and it has since 2007.

298 real fusions.  Bo Letsolo, Jan Rowson and Duncan Baird drove human cells into telomere crisis and sequenced 298 telomere-to-telomere fusion junctions base by base.  Every one of the 298 deleted at least one telomere, and the deletion ran on into the neighbouring DNA, with a mean of 2.4 kb lost.³  Then:

"The TTAGGG telomere repeat content at the fusion junctions was limited, with 38% to 60% of fusions in the HEK293 and MRC5 cells, respectively containing no telomere repeats.  Of the fusions that did contain TTAGGG repeats immediately adjacent to the fusion point, the mean length was 5.8 ± 0.7 (SE) repeats."³

The largest surviving block in the whole data set was 43 repeats.  And the configuration at 2q13, repeats facing each other on both sides of the join, occurred once:  "Just one fusion event of the total 298 sequenced revealed telomere repeats on each side of the fusion point, in this case 6.5 repeats from 17p and 3.5 repeats from 10q."³

A telomere with thousands of repeats cannot fuse.  Tomkins's expectation has a hidden step in it.  It assumes two full-length telomeres met and joined.  Baird's lab had already measured what it takes for a human telomere to become fusable at all.  Rebecca Capper and colleagues sequenced fusion junctions in fibroblast and epithelial cells and found the pattern Letsolo found again three years later:  47% of junctions carried no TTAGGG, and the rest carried between 0.33 and 12.8 repeats, mean 4.7.⁸  From that they drew a threshold:  "we conclude that telomeres of <13 pure TTAGGG repeats distal to the TVR region are fusogenic, and that in the absence of telomerase fusogenic telomeres >13, repeats are likely to be rare."⁸  A long telomere is bound by TRF1 and TRF2 and is capped.  It becomes fusable only after it has eroded below about thirteen repeats.  Thousands of intact repeats at a junction is not what a fusion is missing.  It is what would have prevented the fusion.

There is one known way to get a junction with kilobases of intact repeat, and Capper names it.  Knock out TRF2 in the laboratory and long, healthy telomeres will fuse;  those junctions "usually contain many kilobases of telomeric repeat DNA."⁸  That is the picture the creationist argument is drawn from.  It is a picture of an experiment, not of anything that happens on its own.

There is a second finding in the same paper that matters as much.  Fusions did not happen in the clean TTAGGG array.  They happened inside the degenerate telomere-variant-repeat zone that sits just inside every human telomere.  Of 201 fusions where that zone could be read, only 29 had a fusion point outside it.³  A fusion junction is expected to be degenerate on the day it forms, before a single year of decay.

And the 2q14 array itself takes part.  In the same experiments the authors observed fusions between the 17p and XpYp telomeres and "the interstitial ancestral telomere–telomere fusion at 2q14".³  An old interstitial telomere behaves like a telomere.

Seven ancestral fusions in a deer.  The Indian muntjac has 6 chromosomes in the female and 7 in the male.  The Chinese muntjac has 46.  Eric Green's group at NHGRI sequenced seven of the ancestral fusion sites that built the Indian karyotype, 1.88 Mb of finished sequence with "no gaps in the regions spanning the telomeric and satellite I repeats."⁹  The retained (TTAGGG)n blocks, from their Table 1:  367, 616, 341, 413, 168, 261 and 837 bp, with a few smaller fragments beside them.⁹  The largest, 837 bp, is the size of the human array.

Put side by side.

The expectation of thousands of intact repeats is not a measurement.  It is an assumption, and the only measurements of what fusions do leave put it wrong by two orders of magnitude.  Chromosome 2 has more telomeric repeat at the junction than any real fusion anyone has sequenced, and it has the one feature that turned up once in 298.

The caveats.  Letsolo's and Capper's cells were somatic and driven into crisis, which the Objections section answers.  The muntjac sites are telomere-to-centromere fusions, so only the retained-repeat column transfers.  What transfers is the order of magnitude:  hundreds of base pairs, not tens of kilobases.

Fan's own puzzle, and its answer.  Fan asked the question the creationist literature quotes:  if the fusion happened within the last 6 million years, why are the arrays 14% diverged from canonical repeats when ordinary non-coding DNA has diverged less than 1.5% between human and chimp in the same time?⁵  The creationist literature usually stops there.  Fan did not.  Her first explanation was that the chromosomes joined at degenerate interstitial arrays near their ends rather than at the ends themselves, which Letsolo later showed is what fusions do, and which Yang 2026 demonstrated directly at this site.  Her second was rapid decay after fusion, and she supplied the evidence:  the array in her BAC and the array in IJdo's clone, two human alleles of the same locus, differ from each other by 12%, while the flanking non-repeat sequence differs by 2.4%.⁵  Fan's conclusion:  "Considering the high mutability of interstitial telomere repeat arrays, the fusion partners could have joined either within terminal or subterminal arrays to form chromosome 2."⁵

Finished genomes now show that the 12% is real and not sequencing error, and that it is not confined to the array.  Part 5 sets out what that number can and cannot date.

Part 3: The creationist case, taken on its own terms

The best-argued creationist treatment is Jeffrey Tomkins's, in four papers:  with Jerry Bergman in the Journal of Creation in 2011, then in the Answers Research Journal in 2013 and 2017, and in a 2018 paper to the International Conference on Creationism.  The 2018 paper restates the case and proposes that telomere-like repeats inside chromosomes are designed regulatory elements, after finding about 5,000 of them overlapping annotated genes and as many overlapping enhancer start sites.¹⁰  Nathaniel Jeanson's 2013 ICR piece is the popular form.  This section takes each argument at its strongest.

Argument 1:  the array should have been preserved, because recombination is suppressed near the centromere.  This carries the weight of the 2011 paper.  Tomkins and Bergman write that a high decay rate "is an unreasonable expectation because meiotic recombination is suppressed in pericentric DNA due to its close proximity to the centromere", and list as their first summarised finding that "The reputed fusion site is located in a peri-centric region with suppressed recombination".²

The premise is checkable in two minutes and it is wrong.  In hg38 the centromeric bands of chromosome 2 end at 96,000,000.  The junction is at 113,602,928, in band 2q14.1, 17.6 Mb away.  Measured against the deCODE genetic map for this article:

Computed from the UCSC recombAvg bigWig, which is the sex-averaged deCODE map of Halldorsson and colleagues, by the third script beside this article.  The fusion site recombines at four-fifths of the chromosome average and 33 times the rate of a genuinely pericentromeric window on the same chromosome.  It is ordinary euchromatin, and the argument that the array should have been preserved has no mechanism left under it.

Argument 2:  telomere motifs are everywhere on chromosome 2, so an interstitial array means nothing.  The count is right and the inference is not.  Chromosome 2 has fewer TTAGGG hexamers than chance.  What is unusual about 2q13 is 148 units in 834 bp, in two arrays facing each other, flanked by two different chromosome ends in opposite orientations.  No creationist source has addressed that architecture, and no other place in the genome is known to have it.

Argument 3:  the core sequence has no match on chimpanzee 2A or 2B.  Tomkins searched the 2011 chimpanzee assembly, panTro 2.1, and reported "The most startling outcome of this analysis is that the fusion site did not align with chimp chromosome 2A".²  That was a result about an assembly.  Yang reports that in the pre-T2T ape assemblies the duplications homologous to the fusion site were collapsed in 63 of 74 cases, and that the subtelomeric repeats were left unresolved.⁶  With finished chromosomes the homologous sequences are found on Pan 2a and 2b and on bonobo chromosomes 20 and 22.  Tomkins himself reported a hit at "the telomere region of its long arm" on chimpanzee 2B and rejected it because the model predicts the short arm.²  That objection assumes a subtelomeric block stays where it started.  Yang's reconstruction shows it does not.  The duplications homologous to the fusion site sit at different chromosome ends in different species:  human chr9, chr12 and chr22, and bonobo chr20 and chr22.⁶

Argument 4:  the site is a working part of a gene.  This is the 2013 paper and the strongest thing on the creationist side.  Tomkins showed that the array sits in the first intron of DDX11L2, that CTCF, c-Myc and Pol II bind over it in ENCODE data, and that DDX11L2 is expressed in 255 cell and tissue types.²²  His inference:  "Chromosome fusions would not be expected to form complex multi-exon, alternatively spliced functional genes."²²

The observations are correct and were confirmed here.  The inference fails on three facts, all in the literature Tomkins cites.

First, DDX11L2 is a fragment.  Fan's companion 2002 paper, under the gene's older name CHLR1, had already shown that the copies at 2qFus, 15qter, 16qter, 19pter and Xqter share only "CHLR1 exons 18 and 22–25 and the 3′ UTR" and that "In all these copies, 1286 bp containing exons 19–21 have been deleted and replaced with a 40-bp segment not present in any of the full-length copies."¹¹  The intact 26-exon gene is on chromosome 12.  Yang 2026 puts the fusion-site copy at three exons of ancestral DDX11 inside a 36 kb duplicate.⁶  Tomkins's own 2013 Figure 3 caption calls it "the DDX11L2 pseudogene" while his abstract calls it "a functional RNA helicase gene."  By 2017 he had moved to "the DDX11L2 long noncoding RNA gene".⁷  That correction should be credited.  It also concedes the point:  every DDX11L2 transcript in hg38 is annotated non-coding.

Second, the family lives at chromosome ends, and Costa says so in as many words:  "we defined a novel gene family – the most telomeric ever identified – named DDX11L."¹²  Valerio Costa's 2009 paper, which named DDX11L2 and which Tomkins cites for it, maps the family to "1pter, 2q13/14.1, 2qter, 3qter, 6pter, 9pter/9qter, 11pter, 12pter, 15qter, 16pter, 17pter, 19pter, 20pter/20qter, Xpter/Xqter and Yqter."¹²  Every address in that list is a chromosome terminus except one.  Costa's Figure 4 legend explains the exception:  "2qFS_I represents one of the two ancestral telomeres fused head-to-head at 2q13-14".¹²  Tomkins wrote that Costa mentioned nothing of the fusion sequence.  The legend names it.  A gene family found only at telomeres, with one interstitial member at the exact band the 1982 banding predicted, is not evidence against a fusion.  It is what a fused-in subtelomere looks like.

Third, transcription and transcription-factor binding do not discriminate.  CTCF binds tens of thousands of sites, many in repeats.  Pol II occupancy is found over most of the transcribed genome.  Subtelomeric DNA is transcribed at chromosome ends, and there is no reason it would stop when moved inland.  Yang's CRISPR deletion of the whole 109 kb site changed expression of hundreds of genes;  that removed four annotated loci from one cell line and says nothing about the 834 bp junction itself.

Argument 5:  there is no satellite DNA at the site, and real mammalian fusions have it.  Tomkins in 2018:  "all known chromosome fusion events in extant mammals involve satellite DNA and breaks at or near centromeres".¹⁰  He is on firm ground that known fusion sites carry satellite:  every muntjac site has a large block of centromeric satellite I next to the telomeric block.⁹  The answer is geometric.  Muntjac fusions are telomere-to-centromere, so the satellite belongs to the centromere that was consumed.  The human event is telomere-to-telomere between two subtelomeres, and neither partner brings centromeric satellite to the junction.  The human satellite is 19 Mb further down the arm at the dead centromere, where the model puts it, and Hillier measured 31 kb of it there.

Argument 6:  the gene-order match with chimpanzee was never independently determined.  Jeanson wrote that "when the chimpanzee DNA sequence was elucidated, the researchers did not independently determine the gene order along all of the chimpanzee chromosomes."²³  That was a fair worry about the 2005 draft chimpanzee assembly, which was scaffolded onto the human genome.  Kasai and colleagues had already answered it in 2000 by hybridising 38 cosmids onto chimpanzee chromosomes and reading their order on the chromosomes themselves:  "no difference was detected in the relative order of the cosmid clones between human and chimpanzee chromosomes."²⁰  FISH cannot be scaffolded onto a reference.  And Yang's ape chromosomes are gap-free assemblies built without the human reference.

Argument 7:  the array is too degenerate for its supposed age.  Tomkins in 2018:  "the fusion site should be about 98.5% similar to pristine fused repeats based on standard evolutionary predictions.  However, the 798-base fusion signature is only 70% identical to the sequence of a hypothetical pristine fusion of the same size".¹⁰  The measurement holds.  Aligned for this article against a perfect head-to-head array of the same length, the hg38, T2T-CHM13 and HG002 arrays are 84 to 85% identical under standard DNA scoring, with 8 to 10% of aligned bases mismatched.  Tomkins's 70% is the same finding under harsher gap penalties.¹⁹

The 98.5% does not hold.  It assumes the fusion joined two pristine telomere ends.  Real fusions join inside the worn inner zone of the telomere (Part 2).  And the repeat at 2q13 was there before the fusion:  Yang found the same telomere-like motif at the matching place in the related duplications on human chromosome 12 and bonobo chromosome 20.⁶  In the T2T genome that chromosome-12 stretch is the inner edge of a living, unfused telomere.  Its outer kilobase is near-perfect repeat, and the inner two kilobases grow steadily more irregular.  The stretch Yang names parses as 53% perfect repeat units, against 38 to 43% for the two arms at 2q13.¹⁸  A junction made there is degenerate on its first day.  The degeneracy is no evidence against a fusion, and Part 6 shows that it is no evidence of age either.

Part 4: The dead centromere

A fusion of two chromosomes that each had a centromere leaves a chromosome with two.  One has to be switched off, and its alpha satellite should still be sitting somewhere on the arm.  This is the second half of the 1982 prediction, and it has a clean history.

IJdo's 1991 paper mentions it as unpublished data:  "Hybridization studies suggest that there is a remnant of an ancestral centromere at band 2q21, which is consistent with the telomeric fusion proposed here (A.B., unpublished data)."¹  In 1992 Avarello and colleagues in Pavia hybridised alphoid probes to the human karyotype under low stringency.  Every centromere lit up, and one other place did:  the long arm of chromosome 2 at about q21.3 to q22.1.²⁴  Hillier sequenced 31 kb of alpha satellite there in 2005.  Yang resolved it in 2026 into three arrays at chr2:132,644,386 to 132,685,996, "derived from divergent monomeric α-satellite regions ancestral to human and chimpanzee rather than from higher-order repeats (HORs)", broken up by SVA and L1PA3 insertions in the human lineage.⁶

Tomkins argued in 2011 that the 2q21 alphoid clones "are clearly a diverse mixture of alphoid monomers, forming three separate groups and not distinctly representative of functional centromeric DNA."²  In 2017 he added that the site "is only 41,608 bases in length, but this also includes two insertions of a LPA3/LINE repeat (5957 bases) and a SVA-E element (2571 bases)", and that it lies inside the gene ANKRD30BL.⁷  In 2018 he subtracted the two insertions to leave 33,080 bases of satellite, "a fraction of the size of normal human centromeres that range in length between 250,000 and 5,000,000 bases".¹⁰

Set that beside Yang:  three groups of monomeric alpha satellite, about 41.6 kb, interrupted by an L1PA3 and an SVA.  The creationist and the T2T consortium describe the same object in the same terms.  Tomkins reads short, broken and not higher-order as not a centromere;  Yang reads it as a centromere that stopped working and decayed, which is what a switched-off centromere is expected to become.  Yang's figure comparing it with chimpanzee 2B marks "Potential α-satellite deletion breakpoints".⁶  On that reading the missing part is the higher-order core that makes a working centromere hundreds of kilobases long, and what remains is monomeric satellite older than the human-chimpanzee split.  A relic of 33 kb is what deleting the core would leave, so its small size is no argument against it.  The description cannot break the tie.  Two other things can.

The first is homology.  Tomkins raises it and argues the reverse:  "its human alphoid repeat DNA sequence does not closely match homologous chimpanzee centromeres and chromosomes".⁷  His evidence is hybridisation studies from 1995 and 1997 and a BLASTN search of one 171-base consensus monomer against the chimpanzee assembly of the time.  On the same page he says that assembly was unfit to test the region:  "over 48% of chimpanzee chr2B which allegedly contributed the cryptic centromere is nothing but meaningless N in thousands of gaps".  Yang compared the relic directly with the finished chimpanzee 2B centromere and found the three arrays have "homology to chimpanzee", derived from monomeric satellite "ancestral to human and chimpanzee".⁶  A block of alpha satellite at one interstitial spot on a human arm, inside a region enriched for pericentromeric duplications, matching the centromere of the ape chromosome the fusion model says it came from, is not answered by calling it diverse.

The second is the archaic genomes, and this constraint uses no molecular clock.  Karen Miga in 2016 used chromosome-assigned satellite sequences to read centromere profiles in the Denisovan and Neanderthal genomes and found "both active and degenerate centromeric satellite profiles on chromosome 2 in these archaic genomes".¹³  Yang made the same test with unique k-mers from the dead centromere:  "∼97.5% (8,246/8,460) and ∼96.7% (8,181/8,460) are also identified in Neanderthal and Denisovan genomes, respectively, confirming that the centromeric degeneration occurred long before the divergence of modern and archaic humans."⁶  And the Denisovan genome carries the fusion junction itself:  Meyer and colleagues searched every DNA fragment sequenced from the Denisovan individual, "identified 12 fragments containing joined repeats", and found none in chimpanzee or bonobo reads.¹⁴  Twelve fragments is a small number and the paper claims no more.¹⁵

The ANKRD30BL point fails the same way as the DDX11L2 point, and one part of it fails on the annotation.  Tomkins writes that the relic "extends across intron and exon regions of the gene with part of it actually coding for amino acids in the resulting protein".¹⁰  The current hg38 gene tracks do not show that.  Checked for this article against the GENCODE and NCBI RefSeq annotations, the satellite at chr2:132,208,801 to 132,250,410 contains no exon of any ANKRD30BL transcript.  It lies inside one long intron of transcripts annotated as non-coding.  The one protein-coding transcript, NM_001358416.1, ends at 132,161,952, about 47 kb short of it.  A block of alpha satellite inside an intron shows that a gene was annotated over it.  It does not make alpha satellite a gene part.

Part 5: When, and how it spread

The mainstream does not agree on a date, and the article says so.  Four arguments give four answers.

Poszewiecka's abstract:  "we estimated its formation time at 0.9 Mya with a 95% confidence interval of 0.4-1.5 Mya."¹⁵  Ventura's discussion:  "Since the phylogenetic analyses of both human and chimpanzee suggest the spread of these segments occurred >4 mya, it follows that the fusion of chromosome IIp and IIq occurred early during evolution".¹⁶  Yang's highlight:  "The fusion event is driven by ILS and occurred ∼5–7 million years ago".⁶  Poszewiecka's upper bound and Ventura's lower bound cannot both be right, and the paper reports a bonobo control that gives a date where none should exist.¹⁵  These methods measure different things:  a substitution deficit in flanking DNA, a copy-number contrast between species, and the sorting of ancestral polymorphism.  One of them is miscalibrated.

None of this helps a young earth.  The lowest bound in the table is 400,000 years.  That is 66 times a 6,000-year earth.  The highest is about 7 million years, more than a thousand times.  A dispute about whether a thing is 0.9 or 5 million years old is not evidence that it is 6,000 years old, and a reader should not let "the scientists disagree" carry that weight.

The state of the array, without a clock.  Two features of the site have been read as signs of age, on both sides of the debate.  Finished genomes now let both be checked, and neither works as a clock.

The first is degeneracy.  The array is 38 to 48% perfect TTAGGG.  Part 3 showed why that does not date it.  Fusions join inside the worn inner zone of a telomere, the repeat at 2q13 was already in the duplicated blocks before the fusion,⁶ and the matching stretch at the inner edge of the unfused chromosome-12 telomere is 53% perfect today.¹⁸  A fusion made yesterday at that spot would look degenerate.

The second is Fan's 12% between two human alleles.  For this article five human copies of the site were compared:  the hg38 reference, which at this spot is Fan's own BAC;  the T2T-CHM13 genome;  the two haplotypes of the near-error-free HG002 genome;  and IJdo's 1991 cosmid.¹⁹

Three things follow.  First, Fan's difference is real.  The HG002 paternal haplotype reproduces IJdo's 1991 array to 0.1%, so sequencing error does not explain it.  Second, the five copies fall into two groups, and one person, HG002, carries one of each.  Third, the gap between the groups is not confined to the array.  The plain DNA beside it differs by 1 to 5%, and Fan measured 2.4%.  Human and chimpanzee non-coding DNA differs by less than 1.5% on Fan's own figure.⁵  For two human alleles to differ more than human and chimpanzee do by slow point mutation, they would have to have been kept apart since before the two species split, which happens only under special selection, as at some immune genes.  The simpler reading is an exchange with one of the related subtelomeric copies on other chromosomes, which sit 95 to 98% similar to these flanks.  No single donor copy turned up in the T2T genome, so that reading is a hypothesis.

The consequence cuts against both sides' use of the number.  The 12% cannot be turned into a mutation rate for the array, so it does not date the fusion, and it is no evidence that the fusion is too young or too old for anyone.  It shows that this locus trades sequence with other chromosome ends, which a young account and an old one must both allow.

Fixation.  A chromosome fusion starts in one person with 45 chromosomes.  Nobody else carries it.  How does it reach everyone?  The creationist side is right to press this, and the mainstream answer is thin.  Paweł Stankiewicz, a clinical cytogeneticist at Baylor, proposes that "the HSA2 fusion was a single nonrecurrent event that spread through a small polygamous clan population bottleneck."²⁵  No population-genetic model accompanies it.  What the data do show is that the endpoint is viable.  Only 13 families homozygous for a Robertsonian fusion have been reported, and among them is a healthy 44-chromosome man with normal spermatogenesis.²⁵  Domestic sheep carry centric fusions in many combinations.  Neil Bruère and Peter Ellis followed a New Zealand flock for ten years and reported that fertility "was not significantly different from New Zealand national sheep breeding data" and that "Less than 1% of phenotypically abnormal lambs were recorded in a total of 1995 progeny born over 10 years."²⁶  They also bred sheep down to 48 chromosomes from the normal 54:  "The 2n = 48 category represents a triple homozygous ewe and a triple homozygous ram and is the first report of the viable evolution of such domestic animals."²⁶  A fusion is not lethal and not sterilising, and an animal homozygous for three of them can be healthy and fertile.  Whether it can fix in a large population is a computation nobody on either side has published.

Part 6: What a fusion shows, and what it does not

The fusion is real.  Miga's abstract sums up what everyone measures:  "Genomic signatures of this event are limited to inverted telomeric repeats at the precise site of chromosomal fusion and to the small amount of relic centromeric sequences that remain on 2q21.2."¹³  Limited, and sufficient.  No creationist source explains the architecture without a fusion;  every objection raised concerns the site's condition, which real fusions reproduce.

A fusion in the human line does not by itself show descent from a chimp-human common ancestor.  This is the creationist side's best defensible position, and it is worth stating exactly.  Two ape chromosomes fused in some ancestor of every living human.  A created human kind with 48 chromosomes, in which the fusion happened after creation, is logically possible.  Chromosome number varies within kinds:  sheep and muntjacs carry fusions with little effect on the animal.  The case for common ancestry rests on other evidence, argued in Human and Chimpanzee DNA Similarity, and the fusion should not be presented as if it settled it.

What a within-kind fusion would predict, and how the data treat it.  Todd Wood, a young-earth baraminologist, stated the position in 2006:  the chromosome differences between humans and chimpanzees "could have occurred after creation.  In fact, it is difficult to imagine a scenario other than chromosomal fusion to explain the inverted array of telomere and subtelomere repeats at the putative fusion site on chromosome 2".¹⁷  A fusion within a human kind created a few thousand years ago makes testable predictions, and two of the obvious ones turn out not to discriminate.  The array need not be near-perfect, because a junction made at the worn inner edge of a telomere is degenerate from its first day.  Two human alleles need not be near-identical, because the 12% between them reaches into flanking DNA at a level slow mutation does not produce on any ordinary timescale (Part 5).  The dead centromere fits the loss of its core, which is one deletion, plus an L1PA3 and an SVA insertion.  The predictions that do discriminate concern likeness, not age.  The satellite should not match a chimpanzee centromere better than any other, unless designed to.  It matches chimpanzee 2B.  The subterminal satellite that fills chimpanzee and gorilla chromosome ends, derived from the sequence beside the human fusion site, should have no particular relationship to the fusion.  Ventura finds the source of that satellite is the fusion neighbourhood, and that humans carry it in fewer than 14 copies where chimpanzees carry about 798.¹⁶

The Neanderthal and Denisovan data deserve a separate word, because they cut differently for the two sides.  For the mainstream they are a clock-free floor:  the fusion and the dead centromere predate the common ancestor of modern and archaic humans, which every published estimate puts at several hundred thousand years.  For a young-earth model in which Neanderthals and Denisovans are post-Flood descendants of Noah, their sharing the fusion is expected and costs nothing.  The archaic data therefore do not decide between the models.  What they add is that about 97% of the dead centromere's unique sequence was already present in those genomes.⁶  A within-kind fusion has to put the fusion and most of the centromere's present state before the archaic and modern lines separated.  None of that needs a slow, steady clock, because deletions and insertions are single events.  But no creationist paper has said when or how they happened, or how the fusion reached every human.  And the match of the relic to chimpanzee 2B, and of the flanks to ape chromosome ends, have to be put down to a created likeness.  That is a design claim, not a data claim, and it should be labelled as one.

Objections and responses

"The telomere repeats were there before the fusion, so the site is not a fused telomere at all."  Yang's 2026 highlight reads "The fusion was formed by SDs and subtelomeric repeats rather than telomeric sequences."⁶  A reader who stops at the highlight gets the paper backwards.  It is about how the fusion happened.  Its model:  "each SD contained a TAR1 element and telomeric sequences, arranged in an inverted orientation on ancestral chr2a and chr2b.  The end-to-end fusion of these two distinct SDs may have facilitated the fusion of ancestral chr2a and chr2b".⁶  That is Fan's first explanation from 2002, now demonstrated, and it is what Letsolo's real fusions do 86% of the time.  It does concede something real:  the array is not the intact ancestral chromosome ends.  It is not a concession that there was no fusion.

"Interstitial telomere sequences are common in mammalian genomes, so 2q13 is nothing special."  True, and Ventura shows one way they arise:  subtelomeric material moved inland to chimpanzee chromosomes VII and XIII without any fusion.  That is exactly why the 2q13 case does not rest on the presence of an interstitial repeat.  It rests on architecture:  two arrays head to head with nothing between them, flanked by sequence matching 9pter on one side and 22qter on the other in opposite orientations, and a dead centromere down the arm.  The Issues section says how to test whether another such site exists.

"Fan herself said the arrays are too degenerate for a 6 million year old fusion."  She asked the question and answered it on the same page.  Her first answer, a join at arrays that were already degenerate, is the one Yang has since confirmed.  Her second, rapid decay shown by the 12% allelic divergence, has not held up as she framed it:  finished genomes show the divergence reaches into the flanking DNA, so it is not a property of the array alone (Part 5).  Quoting the question without the answer is the commonest misuse of this paper.

"The fusion site is a functional promoter, and function rules out a fusion."  It is transcribed, and it sits in an intron of a non-coding three-exon fragment of a subtelomeric gene family whose every other member is at a chromosome end.  Transcription of subtelomeric DNA is normal.  Function in the sense of a measured phenotype for the 834 bp junction has not been shown by anyone.  Even if it were, an old degenerate telomere acquiring a CpG island and transcription over a million years is expected, not surprising.

"Telomere-to-telomere fusions are seen only in the damaged genomes of cancer cells."  Tomkins makes this point:  they are "the products of the failure of mechanisms maintaining genomic integrity in cells".¹⁰  Correct, and the article says so.  Crisis is the most permissive condition for fusion, so if even then 38 to 60% of junctions keep no repeat and the rest keep about six, a germline fusion that then had time to decay has no reason to keep thousands.  Capper also looked in ordinary cells and found fusions there, at about four per million in proliferating and senescent normal fibroblasts, at the same short-repeat junctions.⁸  The objection would need a measurement of a germline telomere-telomere fusion junction showing kilobases of intact repeat outside a TRF2 knockout.  None is known.

"The mainstream cannot even agree whether it was 0.9 or 5 million years ago."  It cannot, and the table above says so.  The disagreement is a factor of about six inside a range whose floor is 66 times the age of a young earth.  It is a calibration problem for one of those methods, not a problem for the fusion.

"A fusion in one individual could never reach fixation in a whole species."  This is the best creationist objection to the mainstream account and it should be pressed.  Stankiewicz's answer is a small polygamous clan and a bottleneck, with no model behind it.  Two things temper it.  Homozygous fusion carriers exist and are healthy, in humans and in sheep, and Bruère bred sheep homozygous for three different fusions at once.²⁶  And the objection is symmetrical, though not equal.  A within-kind fusion also had to reach every human, in far fewer generations.  But a young-earth history has a bottleneck built into it, the eight people of the Flood, and a new chromosome form spreads by chance far more easily among a handful of founders than in a large population.  That is why Stankiewicz reaches for a bottleneck at all.  Whichever side proposes the fusion owes the calculation, and neither has published it.

"Then the fusion proves humans and chimpanzees share an ancestor."  This is the objection to our own verdict from the other side, and it also fails.  A fusion shows that an ancestor of all living humans had 48 chromosomes and that two of them joined.  It does not say who that ancestor was.  Whether the 48-chromosome ancestor was also the ancestor of chimpanzees is decided by other evidence, and the dead centromere's homology to chimpanzee 2B and the subterminal-satellite history lean that way without settling it.  Anyone who says the fusion alone settles common ancestry is claiming more than the site can carry.

Issues and unanswered questions

These computations would settle what is still open, and none has been run.

1. Is 2q13 unique? Scan hg38 and T2T-CHM13v2.0 for every interstitial run of 200 bp or more of telomere-like repeat, classify each by whether its two flanks match different subtelomeres in opposite orientations, and report the count. The prediction is that 2q13 is unique or nearly so, and this is the test of the "interstitial repeats are common" objection. The scan script beside this article is the starting point.
2. Is the 2q21 satellite closer to chimpanzee 2B than to any other ape centromere? Yang says it is homologous; a published distance to every ape centromere would make the tie-breaker quantitative. Computable from the public T2T assemblies.
3. Where did the two haplotype groups come from, and what is the array's own mutation rate? Five human copies of the site fall into two groups that differ by about 4% in the array and 1 to 5% in the DNA beside it (Part 5). A BLAT search of the T2T genome finds related subtelomeric copies at 95 to 98% but no near-exact donor. Sorting the 94 HPRC haplotypes into the two groups, and searching their subtelomeres for a donor, would show how common each group is and whether an exchange made it. Yang reports no large structural variants at the site but does not report base-level diversity there. A per-generation mutation rate for the array, from long-read parent-offspring trios across chr2:113,602,600 to 113,603,500, is still missing.
4. Can a fusion fix? A Wright-Fisher simulation of a fusion allele with a stated heterozygote cost in a population of stated size, asking what combinations admit fixation in the available time. Both sides need this and neither has published it.
5. Why do the dating methods disagree? Poszewiecka's bonobo control gives 0.67 Myr where no fusion happened, which the authors report as a contradiction. Until that is explained, the substitution clock's 0.9 Myr should be read as the least secure of the four dates.
6. Does Capper's threshold hold in the germ line? Every measurement of what a real fusion leaves comes from cultured somatic cells. The thirteen-repeat threshold is a property of TRF1 and TRF2 binding, which is not cell-type specific, so there is no reason to expect it to differ. But nobody has sequenced a germline telomere-telomere fusion junction in any mammal. That is the weakest joint in Part 2, and it is weak for both sides: the "thousands of repeats" expectation needs the same missing measurement.

Sources not read in full.  Four sources here are cited from their abstracts, and in each case the abstract carries the claim cited.  Miga 2016 is closed at Oxford Academic;  ten routes were tried, including the publisher PDF listed as free by Unpaywall, and all failed, so the archaic-centromere result rests on the abstract's own words.  Yunis and Prakash 1982 is closed at Science, and is cited only as the origin of the prediction.  Avarello 1992 is closed at Springer.  Kasai 2000 is closed at Springer;  it is now cited directly rather than through Fan, from an abstract that states the 38-cosmid result in full.  Bruère and Ellis 1979 is closed at Oxford University Press, which now publishes the journal.  Capper 2007 was obtained in full from PubMed Central after this article was drafted, and its threshold result is now in Part 2.

Where this article's own side is weakest.  No creationist treatment has compared the repeat at 2q13 with a measured fusion junction.  Tomkins's two arguments from function, the promoter in DDX11L2 and the satellite inside ANKRD30BL, fail on the annotation, and his 98.5% expectation assumes a pristine join that real fusions do not make.  The within-kind fusion view that Todd Wood stated in 2006 survives the site's condition:  neither the degeneracy nor the 12% dates it.  What that view lacks is any published account of the centromere's match to chimpanzee 2B, of when the centromere's core was lost, and of how the fusion reached every human.  Its answer to the match is a created likeness, which cannot be tested.

Verdict

Human chromosome 2 is a fusion of two ancestral chromosomes, and the evidence is strong.  The junction, the inverted head-to-head arrays, the flanks matching two different chromosome ends, the dead centromere 19 Mb away that matches the chimpanzee 2B centromere, the gene-order match, and the presence of all of it in Neanderthal and Denisovan genomes are each what the 1982 prediction required, and each was found afterward.

The creationist objections concern the site's condition, and real fusions reproduce that condition.  Tomkins's observations are correct:  the array is short, degenerate, in an intron, and transcribed.  His inferences fail against measurements he did not cite.  Real telomere fusions keep about six repeats or none, and a telomere long enough to keep thousands does not fuse at all.  The site is 17.6 Mb from the centromere in ordinary recombining DNA.  Chromosome 2 has fewer telomere hexamers than chance.  DDX11L2 is a non-coding fragment of a family that otherwise lives only at chromosome ends.  The satellite at 2q21 holds no exon of ANKRD30BL.  The relic's homology to chimpanzee 2B, which he denies from 1990s probes and a gapped assembly, is there in the finished genomes.  And his own measurement of the dead centromere matches the T2T consortium's to two base pairs.

The site's condition does not date the fusion.  The mainstream dates run from 0.4 to 7 million years and disagree with each other by a factor of six, and the article reports that plainly.  Each is computed on the mainstream timescale, and the lowest comes from a clock whose bonobo control returns a date where no fusion happened.  The site itself carries no independent clock.  Its degeneracy was there on the day it formed.  The 12% between two human alleles is real, but it reaches into the flanking DNA at a level slow mutation does not make, so it cannot be turned into a rate.  What is certain without any clock is that the fusion and the dead centromere were already in the Neanderthal and Denisovan genomes.

What an honest young-earth position looks like here.  Say that the fusion is real, as Todd Wood did in 2006.  Say that a fusion in the human line does not by itself show descent from a chimp-human ancestor, because chromosome number varies within kinds.  Say that the site's degeneracy and its allelic divergence do not date it, which is true, and drop the arguments that the site is too degenerate to be a fusion or sits inside working genes, which fail.  Then say plainly that the dead centromere's match to chimpanzee 2B, and the flanks' match to ape chromosome ends, fit common descent without extra assumptions and fit separate creation only as a created likeness, which cannot be tested.  Anything stronger than that is not supported by the data on the table.

Companion pages

• The literature — every source consulted, with what it measured, whether it was read in full, and which way it cuts.
• Quarantine — the one quotation held back, and why the verdict on it was reversed.
• Human and Chimpanzee DNA Similarity — where the common-ancestry question the fusion cannot settle is argued.
• Adam, Eve, and Human Genetic Diversity — the single-couple bottleneck, which the fixation problem here touches.

Sources