Nylonase

"I propose that this was the original coding sequence contained in the 1295-base-long stretch of pOAD2 plasmid harbored by Flavobacterium Sp. KI72."  (Susumu Ohno, 1984)¹

Summary

Nylon is man-made.  Bacteria that live on the waste of nylon factories were found in Japan in the 1960s and 1970s.  For forty years this has been told as the cleanest case of a new gene appearing in front of scientists:  one inserted DNA letter shifted a reading frame and a new enzyme was born.  This article checks that story against the papers it rests on.

• Nobody watched the enzyme arise. The famous strain, KI72, was fished out of sludge in 1974 already able to eat the nylon by-product.² Another had been reported ten years earlier by a researcher at a nylon company.³ No sample of either from before nylon exists.
• One laboratory experiment did watch the ability appear. In 1995 a strain of Pseudomonas aeruginosa that could not use nylon oligomers was selected until it could. It took nine days to get growth on the nylon monomer and about three more months to get growth on the oligomers.⁴⁴ The "nine-day nylonase" of popular accounts is a misreading. The genetic change was never published.⁴
• Three hypotheses were offered for the main enzyme, NylB. Hirosuke Okada's group in 1983 proposed a duplicated gene that then diverged. Susumu Ohno in 1984 proposed a frameshift that created a new reading frame. Tetsuya Yomo and colleagues in 1992 showed that NylB belongs to a gene family much older than nylon.
• Ohno proposed. His readers reported. Ohno wrote "I suggest", "I assume" and "I propose", and his hypothetical ancestor was built by deleting one letter from a published sequence.¹¹¹ A National Center for Science Education journal, TalkOrigins, New Mexicans for Science and Reason, Kenneth Miller, BioLogos and Wikipedia have all stated the frameshift as something that was found.
• The evidence is against the frameshift. A second NylB in another bacterium shares only about 37% of its amino acids with the first, which one recent frameshift cannot explain.⁵ The 2005 crystal structure shows NylB has the fold of an old and widespread enzyme family and still works as an esterase.⁶ Of the 47 amino acids that separate NylB from its weak twin NylB', two account for the difference in activity.⁷
• Checks run for this article found two errors in Ohno's paper. His abstract prints the ancestor as 472 residues long. The length he meant is 427. And the stop codon that ends his ancestor at 427 comes from a dropped letter in his copy of the sequence. In the deposited sequence the frame runs on.⁸
• Scored by the taxonomy of Measuring Biological Information, the NylB change is a modification of function. An existing enzyme kept its old activity and became 200 to 1,000 times better at a new substrate. That is a real increase in capability and must not be scored as a loss. It is not a new gene. The 1995 laboratory result cannot be scored, because nobody knows what changed.
• Both sides got this wrong at different times. Evolutionary writers presented a proposal as an observation. Creationist writers first argued that a designed plasmid system built new enzymes in nine days, and later scored the change as degeneration. Neither claim survives the primary papers. Royal Truman's 2015 series came closest, scoring NylB as an easy modification of an existing enzyme.

Verdict:  the nylon-eating bacteria are real and the change in NylB is real, small and beneficial.  The frameshift was never observed, was never tested by its author, and is contradicted by the structure of the enzyme.  It was inferred and then reported as observed.  The case does not show that mutation cannot improve an enzyme, and the young-earth side has no tested account of the one event that was actually watched.

Part 1: What was watched

Two polymers and three dates

The polymer invented by Wallace Carothers at DuPont in 1935 is nylon 6,6, made from a diamine and a dicarboxylic acid.⁹  The bacteria in this story eat something else:  the by-products of nylon-6, which is made from a single building block, 6-aminohexanoic acid.  Older papers call the same molecule ε-aminocaproic acid.  Nylon-6 factories release the monomer, short linear chains of it, and rings of two or more units.²  These short chains and rings are the "nylon oligomers".  No bacterium in this article is reported to eat nylon fibre itself.

In Japan, where every strain in this story was isolated, Toyo Rayon first made nylon-6 in 1941 and began full-scale production in 1951.¹⁰  So the clock for these bacteria starts later than 1935.

The isolates

The first report came from inside the industry.  Takashi Fukumura of Toyo Rayon's research laboratories described Corynebacterium aurantiacum B-2, a bacterium that grew on a ring of four nylon units as its only source of carbon and nitrogen.  He presented the work in Tokyo in June 1964 and published it in 1966.³  That is about thirteen years after full-scale production began in Japan.

The famous strain came ten years later.  Shinichi Kinoshita and colleagues at Osaka University put sludge samples into a medium whose only food was the ring of two nylon units, the cyclic dimer.  They kept the cultures that grew and isolated one colony.  They named it Achromobacter guttatus KI72.²  It has since been renamed Flavobacterium and then Arthrobacter, and UniProt now files its enzyme under Paenarthrobacter ureafaciens.⁶  A third strain, Pseudomonas sp. NK87, was isolated from the waste water of a different nylon-6 factory and published in 1989.¹¹

This is what was watched:  bacteria that could already eat nylon oligomers were pulled out of sludge and waste water.  Nobody watched them acquire the ability, and nobody had the same strain from 1930 to compare.  That the ability was new was inferred from the fact that the food was new.

Three enzymes with three histories

Strain KI72 uses three enzymes, all coded on a plasmid named pOAD2.  A plasmid is a small loop of DNA that bacteria carry beside their chromosome and can pass to other bacteria.

The 98.6% figure, the lack of any likeness between EII and EIII, and the family assignments come from the Osaka group and from Cordova and Sanford's database searches.¹¹⁵¹²  The three enzymes must be kept apart.  Ohno's frameshift concerned EII only.  Wikipedia's current article attaches it to EIII.¹³

None of the three is a new kind of protein.  Seiji Negoro's group, by then at the University of Hyogo, later solved the crystal structures of all three, and each turned out to belong to an old and widespread enzyme superfamily.  NylA folds like the amidase signature enzymes, and its closest structural match is subunit A of glutamyl-tRNA amidotransferase, an enzyme of protein synthesis.  It also cuts ordinary carboxyl esters about as fast as it cuts the nylon ring.¹⁴  Its authors concluded that its nylon activity came from adding substrate-binding residues around a catalytic triad "conserved in the AS superfamily".¹⁴  NylC belongs to the N-terminal nucleophile hydrolases, and a structural search matched 259 of its residues to a known aminopeptidase.¹⁵¹⁵  The NylC enzymes of Agromyces and Kocuria differ from the KI72 copy at only 5 and 15 of 355 positions.¹⁵  NylB, as Part 4 shows, has the fold of the β-lactamase family.  So the question "did a new fold arise?" has the same answer for all three enzymes:  no.

The 1995 experiment

One event in this story was watched from start to finish.  Irfan Prijambada and colleagues at Osaka took Pseudomonas aeruginosa PAO1, a hospital isolate from New Zealand with no history in a nylon factory.  They wrote that they wanted to create "a new metabolic activity that would degrade the Ahx oligomer in a strain that is not inherently capable of such degradation."⁴  Ahx is their abbreviation for the monomer.

The experiment had two steps.

1. PAO1 was spread on plates whose only food was the monomer. "After 9 days of incubation" colonies appeared, at a rate of one per thousand cells plated.⁴ One was kept and named PAO5501. The monomer is a plain amino acid with its amino group at the far end. It is not a nylon oligomer.
2. PAO5501 was moved into liquid medium whose only food was the linear dimer. Slight growth appeared in the third week. After a fresh transfer, "It was only in the third month that the turbidity of the culture broth reached 1 at A600."⁴ The strain isolated from that flask, PAO5502, grew on the linear dimer or the cyclic dimer in four days.

Extracts of PAO5502 cells turned both dimers into monomer.  Extracts of PAO1 did not.⁴  So the ability to eat nylon oligomers arose in a flask in about three months.

The authors did not find the gene.  They wrote, "Though a molecular basis for the emergence of nylon oligomer metabolism in PAO5502 is still unknown, it is probable that the basic mechanisms acting during environmental stress are involved in this adaptation."⁴  They listed three known routes by which bacteria adapt to strange foods:  a change in what an enzyme accepts, the switching on of a silent gene, and a change in a regulator.  Their closing sentence leans to the second:  "After the cells accumulated the required genetic alteration to make a cryptic region active, cells grew in the nylon oligomer medium."⁴  Five years later the group leader's review still said "the molecular basis of nylon oligomer metabolism in strain PAO5502 is still unknown".¹⁶  No later paper naming the gene was found for this article.

Part 2: Three hypotheses for NylB

Okada 1983: a duplicate that diverged

Okada's group sequenced the region of plasmid pOAD2 that holds nylB and found a second, similar gene elsewhere on the same plasmid.  They named it nylB'.  Their abstract calls the enzymes "newly evolved" in its first sentence and ends:  "we now conclude that EII enzyme is newly evolved by gene duplication followed by base substitutions on the same plasmid."¹⁷

The only evidence of newness was the substrate.  The enzymes were "active towards man-made compounds" and "not towards any of the natural amide bonds tested."¹⁷  The body of this paper is behind a paywall and was not read, so this article does not say what date, if any, Okada put on the duplication.

Ohno 1984: a frameshift

Susumu Ohno of the City of Hope in California was the author of Evolution by Gene Duplication.  By 1984 he thought duplication had a limit:  a copy of an old gene gives "a mere variation of the preexisted theme."  He was looking for a source of truly new proteins, and his abstract names it:  "I suggest an unused open reading frame of the existing coding sequence."¹

DNA is read three letters at a time.  The same stretch can be read in three different frames, depending on where the reading starts.  Usually only one frame makes sense.  The others hit a stop signal within a few dozen letters.  Ohno reasoned that a gene built from a short repeated unit could stay open in more than one frame.

He did no experiment.  He took the 1,295-letter sequence Okada had published and examined it on paper.  He noticed a second long frame, two letters out of step with the enzyme's frame.  Then he made one edit:

"It should be further noted that if T marked by an arrow in the 3rd row of Fig. 1a is deleted, this longer open reading frame identified as PR.C. (the abbreviation of the preexisted coding sequence) in Fig. 1 a and b starts from the first A-T-G in 1st row of Fig. 1a and ends in T-G-A that are the last three bases in the last row of Fig. 1b, thus specifying a polypeptide chain 427 amino acid residues long.  I propose that this was the original coding sequence contained in the 1295-base-long stretch of pOAD2 plasmid harbored by Flavobacterium Sp. KI72."¹

The story he told from this runs in the other direction.  An arginine-rich protein of no known use, PR.C., was the original gene.  A few decades ago one T was inserted.  The insertion put a stop signal into the old frame and a start signal into the new one, and the new frame happened to code for an enzyme that cuts nylon oligomers.

His wording is that of a proposal throughout.  The section heading says the longer frame "Might Have Been the Original Coding Sequence."¹  The figure caption says "I assume that the longer open reading frame identified as PR.C. was the original coding sequence of this stretch of plasmid DNA until several decades ago."¹  He had two reasons.  First, the two NylB genes differ too much for one to have come from the other since nylon:  "so extensive an amino acid sequence divergence is not expected to occur in so short a time span".¹  Second, a sequence that stays open in two long frames at once is very unlikely unless it began as a repeat.¹

The second observation holds up.  In the genome of KI72 itself, only 2 of 1,481 genes of 350 codons or more have a second frame on the same strand that stays open for nearly the whole gene, as the nylB gene does.¹⁸  Royal Truman, writing in the creationist Journal of Creation, argued in 2015 that the long open frame "can be largely explained as coincidence" of the sequence's high G+C content.¹⁹  The host genome does not bear that out.  Its genes are rich in G and C too, and the feature is still rare there.  Ohno's observation was real.  His explanation of it is what Part 4 tests.

Ohno also conceded that his ancestor did not look like a working protein.  It "is not likely to function as an enzyme of any sort".¹  So the hypothesis needs a gene that was kept intact for no known reason until the day it was destroyed.

Two errors in Ohno's paper

The series memory recorded that Ohno's paper prints the length of PR.C. as 427 in some places and 472 in others.  This was settled for this article in three ways.⁸

• The scanned pages print 472 in the abstract and in the footnote to Table 1, and 427 four times in the text of page 2424.
• Figure 1b numbers the last residue 427. The percentages in Table 1 divide by 427: 58 alanines is 13.58% of 427 and 12.29% of 472, and the table prints 13.58.
• The whole segment is 1,295 letters long. A protein of 472 residues needs 1,416.

So 427 is the number Ohno meant, and 472 is a misprint.  The misprint is in the abstract, which is the part most readers see.  Twenty-one years later the crystal structure paper still described Ohno's ancestor as "472-residue-long".⁶

The second error matters more.  Cordova and Sanford reported that Ohno's copy of the sequence differs from the deposited one at its last few letters.¹²  That is correct.  Ohno's figure ends TAGGCGGCTGA.  The sequence deposited in GenBank from Okada's paper reads TAGGCGGCGTGA.  Ohno's copy dropped a G, and the TGA stop signal that ends PR.C. at 427 residues is the result.  In the deposited sequence that frame has no stop at that point.  In the complete plasmid sequence it runs 14 codons further.⁸  This does not by itself refute the frameshift.  It shows that the hypothetical gene was never checked against the data it was built from.

Yomo 1992: a family older than nylon

By 1992 the Osaka group had sequenced the NylB gene of the second bacterium, Pseudomonas NK87.  Yomo, Urabe and Okada compared the three genes.  The two on the KI72 plasmid are 88% identical.  The NK87 gene and the KI72 gene are 37% identical.⁵  The authors concluded that the three are relatives:  "This strongly suggests that the nylB gene family has diverged from a common ancestral gene."⁵

Then they dated the split with a molecular clock, using the fastest protein clock they knew so that the answer would be a minimum:

"The time of the divergence of F-nylB and P-nylB is estimated to be at least 1.4 × 10^8 years ago, using a very high amino acid-substitution rate of 9 × 10^-9 per site per year for the fibrinopeptide (13).  Therefore, most of the amino acid substitutions from the ancestor of the nylB gene family to its descendants of today might have occurred before the beginning of nylon manufacture."⁵

Two things about this paper are easy to miss.  It was communicated to the journal by Ohno himself, and it does not cite his 1984 paper.⁵  And it does not claim that the activity is old.  The authors wrote that "the time at which their gene products obtained the new enzyme activity is still unknown".⁵

The paper's title names its main finding, which concerns the opposite strand.  On the opposite strand of all three nylB genes, the frame that lines up with the gene's own codons has no stop signal.  Yomo and colleagues calculated that such a frame, if nothing protected it, had only a 0.007 chance of surviving the mutations that separate the genes, and they proposed "an unknown force that is preserving these nonessential NSFs".⁵  Their calculation borrowed the codon usage of E. coli.⁵  The genes of KI72 use the three codons that read as stops on the opposite strand only about a third as often, and 195 of its 1,481 long genes, 13.2%, have the same open frame on the opposite strand.  Rerun with KI72's own codon usage, Yomo's equation gives 0.17 instead of 0.007.¹⁸  In this genome the open opposite frames are ordinary and need no special force.

The 140-million-year figure is an inference from an assumed rate.  The measurement under it is not.  Two enzymes doing the same job in two bacteria differ at nearly two positions in three, and on any timescale that is not the product of one letter inserted forty years before.

Part 3: How a proposal became a finding

Ohno's paper appeared in April 1984.  The next year it was in a popular journal as a result.  The table lists six later accounts, oldest first, with the words each used.

The sources are Thwaites,²⁰ Thomas,²¹ Musgrave,²² Miller,⁹ Venema²³ and Wikipedia.¹³

Thwaites was a geneticist at San Diego State University.  His sentence about the frameshift is hedged:  "The DNA sequence suggests that a simple 'frame-shift' mutation could have brought about the chance formation of at least this one enzyme."  His conclusions are not hedged.  New proteins "have been discovered", and "we know that the new enzymes have formed" since 1940.²⁰²⁰  The shift from "could have" to "we know" happens inside three pages.

Thomas wrote that "Detailed examination of the DNA sequences of the original bacterium and of the nylon-ingesting version show identical versions in the gene for a key metabolic enzyme, with only one difference in over 400 nucleotides."²¹  No original bacterium was ever sequenced.  Ohno compared a published sequence with an edited copy of itself.  The Japanese scientists did not find a frameshift.  They proposed duplication, and the frameshift came from California.

Musgrave, a pharmacologist at the University of Adelaide, added that the enzymes "act on no known amide bond other than the nylon beta amide bond, and have no relationship to any known protease."²²  The crystal structure published the next year showed a relationship to a large enzyme family.

Miller is a cell biologist at Brown University and the co-author of widely used school textbooks.  He wrote that the nylon enzyme "was derived by duplicating a gene that controlled another, quite different function, and then inserting a single extra base into its DNA code."⁹  Ohno ran no study of these bacteria.  Miller's book appeared three years after the crystal structure.  He did report the 1995 experiment's two steps with the right times, nine days and three months.⁹

Venema, then a biologist at Trinity Western University, called the enzyme "truly 'de novo'".  His post lists two references.  One is Ohno 1984.  The other is the 2005 crystal structure paper, whose abstract proposes that the enzyme came from "a preexisting esterase".⁶²³  The post does not mention this.

Venema's post had a target, and the target explains why the frameshift mattered so much.  The design argument he was answering is Stephen Meyer's, built on experiments by Douglas Axe.  Axe mutated the working domain of a β-lactamase and estimated that the share of sequences performing a specific function "by any domain-sized fold may be as low as 1 in 10^77".[axe-2004:prevalence]  Meyer drew the conclusion for evolution:  "random changes to existing genes inevitably efface or destroy function before they generate fundamentally new folds or functions".²⁴  His claim is about new folds, and it covers new genes made from "(a) preexisting genes or from (b) nonfunctional sections of the genome".²⁴  Venema answered with nylonase:  "if only one in 10 to the 77th proteins are functional, there should be no way that this sort of thing could happen in billions and billions of years, let alone 40.  Either this was a stupendous fluke … or evolution is in fact capable of generating the information required to form new protein folds."²³

Neither horn of that choice is what happened.  NylB has the fold of the β-lactamase family, the same fold family Axe used for his estimate.  Nothing in the case required a new fold to be found.  So nylonase is not a counterexample to Meyer and Axe.  It is not evidence for them either.  It sits outside the question they asked, which is how often a random sequence folds and works.  That question, and the large disagreement about Axe's figure, are taken up in How Rare Is a Working Protein?

The first writers to say in public that Ohno had only proposed were on the other side.  Don Batten wrote in 2003 that Thwaites had relied on "a research paper published the previous year where this was suggested."²⁵  Ann Gauger of the Discovery Institute wrote in 2017:  "It was a hypothesis only, however, as a careful reading of his paper shows."²⁶  She granted that the last step was ordinary evolution, "a completely plausible case of gene duplication, mutation, and selection operating on a pre-existing enzyme to improve a pre-existing low-level activity".²⁶  She also noted that the Osaka group, who did the laboratory work, had not proclaimed the frameshift as fact.  That matches what was found here.  Negoro's 2000 review and 2005 paper both list the frameshift among hypotheses that "have been proposed" and then argue for something else, without saying Ohno was wrong.⁶

Between Batten and Gauger, Royal Truman devoted the second of four 2015 papers in the creationist Journal of Creation to testing Ohno's proposal against the sequence data and the protein database.¹⁹  He overstated one point.  He wrote that new research "has led to universal rejection of Ohno's claim by the experts in the field".²⁷  The Osaka papers just cited do not say that.

Part 4: The evidence against the frameshift

A second NylB that is too different

Ohno's title calls the enzyme "unique".  Five years later Kazuo Kanagawa and colleagues isolated Pseudomonas NK87 and found the same two enzyme activities in it.  Its ring-opening enzyme, NylA, was nearly identical to the one in KI72.  Its NylB was not.  Antibodies to one did not recognise the other, and the two genes would not pair in a hybridisation test.¹¹  The genes sit on different plasmids in NK87:  nylA on pNAD2 and nylB on pNAD6.¹¹  The authors drew the conclusion themselves:

"DNA sequencing of the cloned P-EII gene revealed that the F-EII and P-EII enzymes have 35% homology in their amino acid sequences (unpublished results).  Thus, F-EII and P-EII enzymes appear to be evolutionarily related.  However, these two probably branched far earlier from their ancestral enzymes than the F-EI and P-EI enzymes did in their evolution."¹¹

The 35% and Yomo's 37% were both reproduced for this article from the GenBank sequences.  The KI72 NylB matches the NK87 NylB at 130 of 352 aligned positions, 36.9%.  The weak twin NylB' matches it at 123 of 352, 34.9%.²⁸

A frameshift forty years ago makes one new reading frame in one plasmid.  It cannot make two.

Two substitutions out of forty-seven

NylB and NylB' are both 392 amino acids long.  Kato and colleagues counted 47 differences between them, and a direct comparison of the GenBank sequences for this article also gives 47.⁷²⁸  Ohno and the 2005 structure paper print 46.  NylB' has between 0.1% and 1% of NylB's activity on the nylon dimer.

In 1991 Ko Kato and colleagues asked which of the 47 differences matter.  They swapped blocks of sequence between the two cloned genes and measured each hybrid.  Two positions carried the whole effect.  They then changed only those two residues in NylB', glycine to aspartate at position 181 and histidine to asparagine at position 266, and its activity rose to match NylB.

"Our studies demonstrated that among the 47 amino acids altered between the EII and EII' proteins, a single amino acid substitution at position 181 was essential for the activity of 6-aminohexanoate-dimer hydrolase and substitution at position 266 enhanced the effect.  The remaining 45 amino acid substitutions had no effect on the activity, although these 45 alterations included many non-synonymous substitutions"⁷

Each change is a single DNA letter.⁷  Kato gave the gain as about 1,000-fold.  The 2005 paper, using purified enzymes, gave about 10-fold for the first substitution alone and about 200-fold for both.⁶

This removed the premise that Okada and Ohno shared, that a working nylon enzyme was far from anything that existed before.  It is two letters from a nearly inactive relative.

Later work by the same group showed that NylB is neither the only nor the best answer within reach.  Adding a third substitution, D370Y, to the two natural ones raised the weak parent's activity on the nylon dimer 160-fold, and the triple mutant handles the dimer more efficiently than natural NylB does:  1.58 against 0.43 in the standard measure of catalytic efficiency.²⁹  A separate run of random mutation and selection in the laboratory found a different route.  A different set of three substitutions, R187S, F264C and D370Y, raised the same activity about 80-fold, and the authors concluded that there are "at least two alternative modes" for improving it.³⁰  So the nylon activity of this esterase sits two or three substitutions away by more than one path.  That is a short distance for mutation and selection to cover.  It is also the kind of small change inside an existing fold that design advocates say they expect.  Truman wrote in 2015 that "the reaction being catalyzed is not very challenging, as Behe already pointed out … and could be attained from many starting points and by many variants."²⁷

One limit must be stated.  Kato's experiment ran from NylB' toward NylB.  It does not show that history ran that way.  Kato wrote only that the two genes "have diverged from their common antecedent".⁷

The fold of an old enzyme family

In 2005 Seiji Negoro and colleagues solved the crystal structure of a NylB and NylB' hybrid.  The chain folds like the family of enzymes that includes the β-lactamases, which break down penicillin, and the enzymes that build bacterial cell walls.  The closest structural matches were a cell-wall enzyme from Streptomyces and a carboxylesterase from Burkholderia.⁶  The catalytic serine and lysine sit where the family keeps them.

They then tested what the enzyme can do.  NylB cuts short carboxyl esters, the ordinary work of an esterase.  The weak hybrid cuts them about as well.  On one ester NylB scored 6.39 units per milligram and the hybrid 4.81.  On the nylon dimer NylB scored 4.16 and the hybrid 0.023.⁶  So both enzymes are esterases, and one of them is also good at nylon.  NylB has no detectable activity against penicillin-type antibiotics, so it is not a β-lactamase.  It has that family's fold.⁶

Their conclusion:  "the nylon oligomer-degrading enzyme (EII) is considered to have evolved from preexisting esterases with β-lactamase folds."⁶

This is the strongest evidence against the frameshift.  A reading frame that came into use by accident forty years ago has no reason to fold like a family of enzymes found across the bacteria, with its working residues in the family's positions.  The frame that codes for NylB is an old one.

Trypsin cuts nylon oligomers

Trypsin is a digestive enzyme of animals.  It cuts proteins after the amino acids lysine and arginine.  The nylon-6 monomer is lysine without one of its two amino groups, so it fits trypsin's pocket.

In 1959 Mituo Ebata and Kenji Morita of Kanazawa University reported that trypsin slowly hydrolyses nylon-6 compounds.  In one hour it cut 2.2% of the linear dimer, the same compound NylB cuts.  In three hours it cut 0.26% of a nylon-6 polymer about 18 units long.³¹  They wrote, "It is interesting that poly-ε-aminocaproic acid (polycaproamide) was also slowly hydrolysed in an amorphous state though the polymer was almost insoluble in water."³¹

The argument from this is not about the date of the paper.  Trypsin's structure was not shaped by a polymer that animals never met.  It cuts the nylon-6 bond anyway, because that bond is an ordinary amide bond of the kind found in every protein.  So the chemistry needed no invention.  An enzyme built for another job can do it weakly, and NylB' does it weakly.

The activity is a trace.  Trypsin is not a working nylonase, and this result does not show that any bacterium could live on nylon waste before the two substitutions.

Cordova and Sanford's database test

In 2019 Salvador Cordova and John Sanford posted a manuscript that tests Ohno's hypothesis against protein databases.  Both authors are creationists, and Sanford is a retired Cornell plant geneticist.  The manuscript is a preprint.  It was posted on three preprint servers between March 2019 and January 2020 and has not appeared in a journal.  It has not been peer reviewed, and no published reply to it was found.¹²

Their reasoning is simple.  If Ohno was right, his ancestor PR.C. was a real gene with a history, and relatives of it should turn up in other organisms, while NylB should have no relatives.  If he was wrong, the reverse should hold.  They reported the reverse:  NylB named in 201 organisms with tens of thousands of possible relatives, and nothing for PR.C.¹²  They were not the first to look.  Royal Truman had reported in 2015 that a BLASTP search of the whole NCBI database with the PR.C. protein found "no candidate matches".¹⁹

Their central claims were checked for this article on 27 September 2026.

• NylB has many relatives. Confirmed. UniProt lists 279 entries under the gene name nylB and files NylB in the β-lactamase family.³² Nearly all of those names were assigned by computer from similarity. They show that the family is widespread. They do not show that those proteins cut nylon. Cordova and Sanford say so themselves.¹²
• PR.C. has no relatives. Not contradicted, and not fully tested here. The searches of the full protein database submitted for this article had not returned after 55 minutes. A scan of the 575,748 reviewed proteins in Swiss-Prot, run locally, found no close relative of PR.C. Its best matches were arginine-rich proteins that scored little better than scrambled copies of PR.C. did. That scan sees close relatives only. It missed the distant relatives of NylB as well, so it cannot confirm the preprint's zero.³²
• Ohno's stop codon is a copying error. Confirmed, as described in Part 2.
• P. aeruginosa PAO1 has a NylB relative. Confirmed, with a caution. The protein they name, PA4347, matches NylB at 27.8% of aligned positions. It is a distant member of the same family, less like NylB than the two known NylB enzymes are like each other.²⁸

Where the preprint misstates its sources, the errors are of care and not of substance.  It dates Ebata and Morita to 1957 in the text, which is the year the paper was received.  It dates Yomo to 1999.  It gives two references the page numbers of other papers.  It says Kato mutated "the 47 amino acids", which he did not.  It says Ohno "presented his assertions very forcefully, as if they were facts", which does not describe a paper that says "I propose".¹²  The defect it describes is real, but it belongs to Ohno's readers.

The test also proves less than the authors say.  A gene that was destroyed need not have left relatives in a database, so silence about PR.C. is weak evidence.  The strong evidence is on the other side of their table:  NylB belongs to an old family.

Part 5: Scoring what happened

The series sorts every observed change into four kinds.  A gain adds a working coded element that was not there, such as a gene, a promoter or a binding site.  A loss breaks or degrades one.  A modification changes what an existing element does, in degree or in range.  A transfer brings in an existing element from another organism.  The definitions are argued in Measuring Biological Information.

NylB: a modification of function

The modification is a real improvement.  The enzyme lost nothing that was measured.  Its esterase activity on the short ester is slightly higher than the weak hybrid's.⁶  Kevin Anderson and Georgia Purdom of Answers in Genesis scored the change as "biochemically degenerative".³³  The measurements do not support that.  An enzyme that does what it did before and also does something more has gained range.

Three things would each make this a gain, and none is supported.

1. Ohno's frameshift. A new reading frame coding for a new protein is a new gene by any definition. The structure rules it out.
2. A gene that arose after nylon. The 37% identity between the two NylB genes rules it out.
3. An enzyme with no activity at all on nylon before the change. Then the substitutions would have built a binding site from nothing. But NylB' already cuts the nylon dimer at about half a percent of NylB's rate, and even trypsin cuts it.³¹⁶ The activity rose. It did not appear.

The date and the direction are not known

Every account of this case, creationist accounts included, says that the two substitutions happened after nylon.  Nobody has shown it.  There are two readings of the same two genes.

• NylB' is the older form. After nylon waste appeared, a copy of it picked up the two substitutions and became NylB. This is a modification that improves function.
• NylB is the older form. NylB' is a spare copy that lost the two residues and most of its nylon activity. This is a loss in the copy, and nothing new happened after nylon except that bacteria carrying NylB prospered where nylon waste was dumped.

One check made for this article bears on the choice.  The NylB of Pseudomonas NK87 is an active nylon enzyme that parted from the KI72 genes long before the two KI72 genes parted from each other.  When it is aligned with KI72's NylB, the residue that lines up with position 181 is aspartate, the same residue the active KI72 enzyme has, inside a similar stretch of sequence.  At position 266 it has aspartate, where NylB has asparagine and NylB' has histidine.²⁸  If the alignment is right, the key residue at 181 is shared by both active enzymes and missing only from the weak twin.  The simplest reading is that aspartate at 181 is the older state and NylB' lost it.

This is an inference from one pairwise alignment at 37% identity.  It has not been published or reviewed, and the same residue could have arisen twice.  It is reported here as a question the published record leaves open.  If it holds, the best-known half of the nylonase story did not happen after nylon at all.

The 1995 experiment is a separate event

PAO5502 must not be folded into the NylB story.  PAO1 carries no nylon plasmid.  Its chromosome does carry distant relatives of NylB, and Cordova and Sanford suggest that point mutations in one of them produced the new activity.¹²  That is a guess.  So is the authors' own suggestion of a silent region switched on.  Two facts limit the guesses.

• Colonies able to use the monomer appeared at one per thousand cells.⁴ A given single-letter change arises about once in several billion cell divisions.³⁴ A rate a million times higher points to a change with many possible routes, such as the loss of a repressor, or to a response to starvation.
• The new enzyme activities were higher in cells grown on the nylon dimer than in cells grown on glucose.⁴ So they are regulated.

If a silent gene was switched on by breaking its repressor, the event is a loss of a control element with a useful result.  If a working enzyme changed what it accepts, it is a modification.  If a new promoter was built, it is a gain of the same kind as the citrate result in the Long-Term Evolution Experiment.  The experiment cannot be scored until someone sequences the strain.

The population arithmetic

A working note written for this site estimates how many bacteria were exposed to nylon waste before the enzyme was found:  about 1.5 × 10^26 bacteria-generations.  The inputs were checked for this article.

The arithmetic in the note is correct for its inputs.³⁴  The inputs are the problem.  Two of the five have no source, and a third is an illustration.  A smaller guess is as easy to defend.  One factory pond of a thousand cubic metres over 13 years, with cells dividing once a day, gives about 5 × 10^18 bacteria-generations.  The pond size and the daily division are assumptions made here for scale, with no source.  The note's figure is 30 million times larger.  The number is an order-of-magnitude guess, and the guesses differ by seven to eight orders of magnitude.  It cannot be used as a measurement.

Two conclusions survive the uncertainty.

First, the search was not the hard part.  A particular single-letter change turns up about once in 5.6 × 10^9 cell divisions.³⁴  The first substitution alone raises activity about tenfold, so selection can keep it while the second is awaited.⁶  Even the low guess above supplies several hundred million times more divisions than that.  Two such changes in forty years are what ordinary mutation and selection predict.

Second, the output was small.  At most two nucleotides of new functional sequence were produced, and if NylB is the older form, none.  The series reports output as unique functional sequence created per organism-generation.  For this case that cell is left empty.  The numerator is zero or two, and the denominator is not known to within eight orders of magnitude.

Part 6: Both sides got it wrong

The evolutionary side

The error was one of reporting.  A paper that said "I propose" was cited as a paper that found.  What that side got right is substantial.  Bacteria do adapt to man-made compounds.  Mutation and selection can improve an enzyme, and the improvement in NylB is large.  The claim Dave Thomas set out to answer, that mutations can only scramble what is there, is wrong.²¹  The 1995 experiment is a real observation of a new ability arising in a flask.

The creationist side

Don Batten, 2003.  Batten, a plant physiologist with Creation Ministries International, wrote the first detailed creationist reply.  He accepted that the enzymes were new.  He wrote that the lack of activity on natural compounds suggested "that the enzymes are completely new, not just modified existing enzymes."²⁵  His argument was that a designed system produced them:  "The presence of the transposases in such numbers on the plasmid suggests that the plasmid is designed to adapt when the bacterium is under stress."²⁵

Three of his supports fail against the primary papers.

• He wrote that the Pseudomonas result "was achieved in a mere nine days!"²⁵ Nine days produced growth on the monomer. The oligomers took about three more months.
• He wrote that Yomo and colleagues "show that it is highly unlikely that any of these genes arose through a frame shift mutation".²⁵ Yomo's paper does not discuss the frameshift. Ian Musgrave was right to say so.²²
• He predicted that the new activities in the laboratory strain would prove to be on plasmids. No such result was found for this article.

Batten was right that the frameshift had only been suggested, and right that a change appearing in one cell per thousand does not look like a rare point mutation.

Anderson and Purdom, 2008.  This paper has the origin right.  It says the altered enzyme "still possesses the esterase function".³³  Then it calls the result degenerative, which the measurements in Part 5 do not support.

Royal Truman, 2015.  Truman's four-part series in the Journal of Creation is the most thorough creationist treatment found.  He accepted the esterase origin:  "Negoro and others provided compelling evidence the enzymes involved in degrading the synthetic materials created during manufacture of nylon-6 were derived easily from existing precursors."²⁷  He scored the result as modification, not invention, and said what an evolutionary example would have to show:  such examples "must deal with the creation of complex proteins, molecular machines, organs and organisms, not merely the modification of existing proteins to utilize new sources of nutrients."²⁷  He also stated the design reading as a prediction:  "we predict focused research will show that more than random processes are involved."³⁸

That prediction is the right form for the claim, and nobody has tested it.  No one has measured how the two substitutions arose.  Two other parts of the series fail checks made for this article.  His explanation of Ohno's long open frame by G+C content does not hold in the host genome, as Part 2 shows.  And he suggested that the open frames Yomo found on the opposite strand might mean "a few thousand years of mutations was not long enough to generate harmless Stops".¹⁹  In KI72's genome such frames are common, so they say nothing about age.¹⁸

Batten, 2017.  Batten's update accepts the esterase origin and makes a concession the 2008 paper did not:  "In this case, a new function has indeed been added.  Even if the direction of change is towards less specificity, overall there is an addition of genetic information."³⁹  He keeps his 2003 conclusion that the bacteria have "a designed ability to adapt", and he does not correct the nine days.³⁹

Neither side noticed that the discovery dates from 1964, or that nobody has dated the two substitutions.  Both took a new food to mean a new enzyme and argued about the mechanism.

Objections and responses

"The frameshift was wrong.  The enzyme still evolved.  What is the difference?"  The difference is what the case can be used to show.  It was used for forty years to show that a working gene can appear from a scrambled reading frame in one step.  It does not show that.  It shows that an existing enzyme can be tuned to a related substrate by two single-letter changes.  Both are evolution in the broad sense.  Only the first bears on where new genes come from.

"Nylonase shows that a new protein fold can arise in forty years, so the design argument from fold rarity fails."  This was Dennis Venema's argument in 2016, aimed at Stephen Meyer and Douglas Axe.²³  It rests on the frameshift.  NylB has the fold of an old enzyme family, the same family Axe used for his estimate, and the other two nylon enzymes have old folds too.  No new fold arose, so the case does not test how rare new folds are.  It does not support the design argument either.  Whether Axe's estimate is right is a separate question with its own measurements, and they disagree widely.  See How Rare Is a Working Protein?

"Cordova and Sanford are creationists and their paper was never reviewed."  Both statements are true.  Nothing in this article rests on their authority.  Each claim of theirs that is used here was checked against the database or the primary paper, and their citation errors are listed.  The main evidence against the frameshift comes from the Osaka group's own peer-reviewed papers.

"A gene family 140 million years old contradicts a young earth."  On the conventional clock it does, and the authors chose a fast clock to get a minimum.  The young-earth position has to reject the clock and explain the difference between the two NylB genes another way, either as enzymes that were distinct from the start or as the result of much faster change.  No creationist paper found for this article models either.  Truman offered one clue in 2015, the open frames on the opposite strand of the genes, as a sign that they had not been mutating for long.¹⁹  The check in Part 2 removes it.  Such frames are common in the host genome, and Yomo's survival odds depend on how many substitutions separate the genes, not on how many years they took.¹⁸  That is a gap in the young-earth account and it should be stated as one.  The argument in this article does not depend on the date.  It depends on the 37%, which is a measurement.

"The 1995 experiment shows a nylon enzyme evolving in three months.  That is evolution observed."  It is.  A new ability arose under observation in a strain that lacked it, and this is the strongest part of the case for the evolutionary reading.  What arose is unknown.  It may have been a gain.  Kenneth Miller wrote that genes already present "turns out not to have been a possible explanation for the Pseudomonas experiments", because the strain's inability was documented first.⁹  The experimenters themselves thought a silent region had been switched on.  A strain that cannot grow on a compound may still carry a gene that is off or weak.

"Plasmids are designed adaptation systems, so the change was directed."  This is Batten's position.  Truman put it in testable form in 2015:  "we predict focused research will show that more than random processes are involved."³⁸  It could be true and it has not been tested.  No experiment shows that the insertion elements on pOAD2 produced either substitution.  The strain in the 1995 experiment is not known to carry a plasmid.  Until someone shows the mechanism acting, it is a proposal, and this article cannot treat a creationist proposal more gently than Ohno's.

"Two mutations is within Behe's limit, so this supports Behe."  It is consistent with the limit and does not test it.  Two changes that can be selected one at a time are within reach of any large bacterial population.  The case says nothing about what harder changes cost.  That question belongs to the organism articles of Benchmarking Evolution.

"Refuting one frameshift does not refute new genes."  Correct.  The series concedes in its own voice that the antifreeze gene of northern cod arose from DNA that was not a gene, and that the citrate result in the Long-Term Evolution Experiment is a gain.  See Antifreeze Proteins and Measuring Biological Information.

Issues and unanswered questions

• Okada 1983 was read in abstract only. The body is behind the publisher's paywall and no open copy was found. Whether Okada dated the duplication to after nylon is not verified.
• The searches for Ohno's PR.C. in the full protein database did not return. The local scan that replaced them is a weak test. The request numbers are in the note on database checks.
• PAO5502 was never sequenced, as far as the published record shows. Whether the strain still exists is not known. Its genome against that of PAO1 would settle how the 1995 result should be scored.
• The direction of change between NylB and NylB' is open. The alignment reported in Part 5 needs a proper family tree with more members and outside relatives before it can carry weight.
• The genes of Fukumura's strain B-2 are unknown. It may or may not carry a NylB.
• Read in part or not at all: only the first page of Kinoshita 1975 and of Fukumura 1966 was read, and it does not say where the sludge came from. Kawashima 2009 and Ohki 2009 were read in abstract only. Kanagawa 1993, Kato 1995, Negoro 2007 and Venema and McKnight's Adam and the Genome were not read.
• The histories of NylA and NylC are not examined here. Their folds are known and both are old, but their dates and the direction of their changes may differ from NylB's.
• Ohno's second open frame is still unexplained. A frame shifted by two letters that stays open across the whole nylB gene is rare in the host genome, 2 of 1,481 long genes. The fold rules out Ohno's account of the gene's origin, but nobody has explained why that frame is open.
• No measurement exists of how many bacteria were exposed to nylon waste. The ledger cell for this case stays empty.

Verdict

The observed core of the nylonase story is modest and solid.  Bacteria able to live on nylon-6 by-products were isolated from sludge and factory waste water in Japan from 1964 on.  In one laboratory experiment, a strain that could not use nylon oligomers acquired the ability in about three months.

The famous part of the story was never observed.  Ohno proposed in 1984 that a single inserted letter created NylB from the scrambled frame of another gene.  He tested nothing, his hypothetical gene ends at a stop codon that is a copying error, and the enzyme has the fold and the side activity of an old enzyme family.  His proposal was reported as a finding, and it is still stated as fact in the most-read account.

What the evidence supports is that an esterase of the β-lactamase fold differs from its nearly inactive twin at two positions that matter, and that changing those two raises its activity on nylon oligomers 200 to 1,000 times.  By the series taxonomy that is a modification of function.  It is a real improvement and it is not a new gene.  The other two nylon enzymes also belong to old enzyme superfamilies, so no new fold arose anywhere in the case.  Whether it happened after nylon is not known.

The young-earth side was right to doubt the frameshift and wrong in most of what it put in its place.  The exception is Royal Truman's 2015 series, which scored NylB as an easy modification of an existing enzyme, as the evidence does.  It has no tested account of the 1995 experiment, and it has not explained the distance between the two NylB genes.

Sources