Every source the main article rests on, grouped by role, with what each one says and which way it cuts. Everything below was fetched and read except where marked not obtained; those are collected at the end, because an argument resting on a source nobody on this side has read should be flagged rather than buried. Library keys refer to sources/index.tsv.
Verbatim quotations from both sides, with page locators, are on the quotes page.
A note on method. No quotation in this article or on the quotes page was established through an AI page-summarising tool. Every quotation was read out of bytes downloaded locally — publisher PDFs converted with pdftotext -layout, HTML stripped locally with a tag-removal script, and one source (Lenski et al. 1991, a scanned JSTOR offprint with no text layer) rendered to 300 dpi page images and OCR'd, then checked by eye against the images.
The founding design
Lenski, Rose, Simpson & Tadler 1991, "Long-Term Experimental Evolution in Escherichia coli. I. Adaptation and Divergence During 2,000 Generations," American Naturalist 138:1315-1341. The source of every design parameter in the article's table: Davis minimal broth with 0.025 g/L glucose "which permits a stationary phase bacterial density of about 5 × 107 cells per mL," 10 mL in a 50-mL flask at 37 °C and 120 rpm, 0.1 mL into 9.9 mL daily, "~6.64 generations of binary fission," and the first fitness result of +37% over 2,000 generations. All the cell-count arithmetic in the article derives from this page. Essential; everything else is downstream of it. Key lenski-1991.
Lenski 2023, "Revisiting the Design of the Long-Term Evolution Experiment with Escherichia coli," Journal of Molecular Evolution 91:241-253. Not obtained — closed access, no repository copy, Unpaywall reports no OA location. Used only for its abstract, which states the LTEE "began in 1988" and had "recently reached 75,000 generations." Would have been the best single source on why each design choice was made; worth buying or requesting if the article is revised.
GenBank NC_012967.1, Escherichia coli B str. REL606. The ancestral genome, 4,629,812 bp, read from the NCBI record. This is the denominator in every "how many mutations were tried" calculation.
The Lenski corpus: what changed over 50,000-60,000 generations
Sniegowski, Gerrish & Lenski 1997 (Nature 387:703-705). Three of twelve populations hypermutable by 10,000 generations, rates "two orders of magnitude higher," caused by defects in mutS, uvrD and mutL, and rising by hitch-hiking on the beneficial mutations they generate. The first documented adaptive loss of function in the experiment, and the origin of Behe's mutator argument. Key sniegowski-1997.
Cooper & Lenski 2000 (Nature 407:736-739). Catabolic breadth across 64 informative carbon sources: nine substrates decayed in parallel by 2,000 generations, sixteen by 10,000, and total catabolic function fell from 1.00 to 0.68 by 10,000 generations and 0.58 by 20,000. Argues the decay is antagonistic pleiotropy rather than mutation accumulation, on the grounds that it was early, parallel, and pre-dated any mutator. The single best-quantified loss result in the corpus, and it cuts for the creationist reading. Key cooper-2000.
Cooper, Schneider, Blot & Lenski 2001 (J. Bacteriol. 183:2834-2841). The ribose deletion in all twelve populations: IS150-mediated, occurring at ~5 × 10-5 per cell generation, worth 1-2% fitness, and confirmed by transducing the operon back in and by building the deletion independently without IS150. The cleanest demonstration anywhere that deleting a working operon can be adaptive. It is Behe's best evidence and it comes from Lenski's lab. Key cooper-2001.
Woods, Schneider, Winkworth, Riley & Lenski 2006 (PNAS 103:9107-9112). pykF and nadR substituted in all twelve populations; four "blind" candidate genes compared against 36 random ones. Cited in the article for parallel evolution that is not obviously contingent — Lenski's own counterexample to over-generalising the citrate story. Key woods-2006.
Wielgoss et al. 2011 (G3 1:183-186). The point-mutation rate, 8.9 × 10-11 per base pair per generation, inferred from synonymous substitutions in 19 sequenced genomes. Multiplied by the REL606 genome this is 4.1 × 10-4 per genome per generation, which is the input to every mutation-supply calculation in the article. Key wielgoss-2011.
Maddamsetti, Lenski & Barrick 2015 (Genetics 200:619-631). Clonal interference and frequency-dependent interactions in Ara–5 and Ara–6. Read for background on ecological coexistence; not cited directly, because Good et al. 2017 covers the same ground for all twelve populations with better data. Key maddamsetti-2015.
Lenski et al. 2015 (Proc. R. Soc. B 282:20152292). More than 1,100 new competition assays testing whether the power-law fitness model predicted the next 10,000 generations. It did. Gives both model equations verbatim and the predicted values at 40,000, 50,000 and 60,000 generations. The load-bearing source for the claim that fitness has not declined and shows no ceiling — the hardest LTEE result for the young-earth side. Key lenski-2015.
Tenaillon et al. 2016 (Nature 536:165-170). 264 complete genomes. The full mutation census (14,572 point mutations, 500 IS insertions, 726 small deletions, 1,132 small insertions, 267 large deletions, 45 large duplications); the 63-kbp genome shrinkage; the six hypermutators holding 96.5% of point mutations; the IS150 hypermutator Ara+1; the antimutator reversal; the dN/dS excess of 17.1× early and 3.4× over 50,000 generations; and the "57 genes with two or more mutations" set that both sides of the Behe dispute argue over. Read as the Europe PMC author manuscript. Cuts both ways and is quoted by both sides. Key tenaillon-2016.
Maddamsetti et al. 2017 (Genome Biol. Evol. 9:1072-1083). The fine-tuning analysis of those 57 genes: KEIO essentiality scores, the near-absence of nonsense mutations in essential genes, and the ten genes showing amino-acid-level parallelism at protein interfaces. This is the paper Behe cites in his own endnote to dismiss it, and reading it is the fastest way to see that his "all beneficial mutations are degradative" claim overshoots his source. Key maddamsetti-2017.
Couce et al. 2017 (PNAS 114:E9026-E9035), "Mutator genomes decay, despite sustained fitness gains, in a long-term experiment with bacteria." Asks whether selection or mutational bias better predicts which sites actually changed in the LTEE's mutator genomes, using a direct-coupling-analysis fitness predictor. Bias wins, and it still wins when the analysis is restricted to genes essential under LTEE conditions. Explains it by genomic draft: in an asexual population, strong linkage carries passengers along with a sweep. The strongest pro-decay evidence in the corpus, and it is Lenski's and Tenaillon's own. Any article that cites the rising fitness trajectory without citing this one is telling half the story. Key couce-2017.
Good, McDonald, Barrick, Lenski & Desai 2017 (Nature 551:45-50). Whole-population metagenomic sequencing at 500-generation intervals to 60,000 generations. Source of the non-mutator accumulation rate (~20 mutations per 10,000 generations early, ~10 late), the effective population size of ~107, the drift-timescale argument that the observed mutations "must have reached observable frequencies through the direct or indirect action of natural selection," and the finding that nine of twelve populations carry clades coexisting for more than 10,000 generations. Read as the Europe PMC author manuscript. Key good-2017.
Barrick et al. 2009, "Genome evolution and adaptation in a long-term experiment with Escherichia coli," Nature 461:1243-1247. Not obtained — closed access; Unpaywall reports no OA location; the Wayback copies of Lenski's own reprint archive stop at 2008 and the Barrick lab path returns HTML. Its 20,000-generation mutation counts are superseded by Tenaillon 2016 and Good 2017, both of which were obtained, so nothing in the article depends on it.
Wiser, Ribeck & Lenski 2013, "Long-term dynamics of adaptation in asexual populations," Science 342:1364-1367. Not obtained — closed access. Both of its model equations and all six of its predicted fitness values are quoted verbatim inside Lenski et al. 2015, which was obtained, so the article cites the 2015 paper for them rather than relaying an unread source.
Lenski 2017, "Convergence and Divergence in a Long-Term Experiment with Bacteria," American Naturalist 190:S57-S68. Not obtained — closed access. A review; nothing in the article rests on it.
The citrate story
Blount, Borland & Lenski 2008 (PNAS 105:7899-7906). The discovery paper. Cit+ first detected between generations 31,000 and 31,500; turbidity noticed at ~33,127; three replay experiments totalling 3,212 replicate cultures and 17 independent Cit+ mutants, all from clones sampled at 20,000 generations or later; the ancestral mutation-rate upper bound of 3.6 × 10-13 and the potentiated estimate of 6.6 × 10-13; the Ara−→Ara+ control rate showing the potentiation was not general hypermutability; and the stable ~1% Cit− minority maintained by negative frequency-dependent selection. Read the statistics carefully: the third and largest replay gave P = 0.0823, and a 20,000-generation clone produced two Cit+ mutants against a model expecting none. The paper is honest about both. Key blount-2008.
Blount, Barrick, Davidson & Lenski 2012 (Nature 489:513-518). 29 sequenced clones; the 2,933-bp tandem duplication; the rnk-citT module and the luciferase-reporter test showing the native citT control region is silent in oxygen; the engineered single-copy construct with its >60-hour lag and its 1.0% competitive advantage; copy-number escalation to 2, 3, 4 and 9; the mutS mutator arising after Cit+; and the plasmid experiments showing epistasis with the C3 background. The 1.0% figure is the empirical foundation of the Discovery Institute's stepwise claim, and Lenski's own lab measured it. Read as the Europe PMC author manuscript. Key blount-2012.
Quandt, Deatherage, Ellington, Georgiou & Barrick 2014 (PNAS 111:2217-2222). The REGRES backcrossing method, the dctA refinement mutation, and the finding that "strains containing just the citT and dctA mutations fully use citrate, indicating that earlier mutations thought to have potentiated the initial evolution of Cit+ are not required." Also identifies the dcuS defect in E. coli B that Minnich later builds an argument on. A major concession to the Minnich position, published two years before Minnich's paper, by Lenski's former postdoc. Key quandt-2014.
Quandt, Gollihar, Blount, Ellington, Georgiou & Barrick 2015 (eLife 4:e09696). The gltA1 citrate-synthase mutation: it disrupts NADH inhibition, raising flux; it is beneficial on acetate; and without it the citT duplication goes from roughly neutral to "highly deleterious." Later gltA2 alleles cut activity below the ancestral level, one of them by reducing transcription tenfold. The best-measured single case in the whole corpus, and the only one where enzyme kinetics were actually assayed. It also destroys the tidy version of the stepwise claim. Key quandt-2015.
Van Hofwegen, Hovde & Minnich 2016 (J. Bacteriol. 198:1022-1034). 46 independent Cit+ mutants under direct and modified-direct selection; the "12 to 100 generations" figure and the weekly-transfer protocol that produces it; the DV130 / DV133 / DV133T / DV159 genomes; the 10-14 joint-probability estimate; and the "no new genetic information" conclusion. The paper the whole Discovery Institute case rests on. Its experiments are good and its interpretive claims are separable from them. Key vanhofwegen-2016.
Roth & Maisnier-Patin 2016 (J. Bacteriol. 198:1009-1012). The invited commentary published beside Van Hofwegen et al. Roth is a UC Davis Salmonella geneticist with no connection to the design movement and he agrees with the reinterpretation: duplication instability plus intermittent selection, not a neutral potentiation requirement. Also supplies the low-oxygen-in-stationary-phase hypothesis that makes plain citT amplification plausible as a first step. The most important source for the claim that Minnich's reinterpretation is mainstream-supported rather than partisan. Key roth-2016.
Lenski & Blount 2016, "On the Evolution of Citrate Use," Telliamed Revisited, 20 February 2016. The reply. Concedes the science ("looks fine and interesting"), disputes the interpretation, supplies the 19-day figure read out of Van Hofwegen's own Table 1, and gives the Origin of Species analogy against "no new information." A blog post, not a peer-reviewed comment; no journal reply was ever published, and that absence is itself worth knowing. Key lenski-2016.
Leon, D'Alton, Quandt & Barrick 2018 (PLoS Genetics 14:e1007348). The decisive measurement: the rnk-citT module knocked into twenty-five backgrounds (23 evolved clones plus REL606 and the revertant ZDB706) spanning Ara–3's history, with competitive fitness measured against isogenic Cit− partners. Note that the paper lists its group-wise effects backward in time, which is easy to mis-transcribe. +1.0% in the 1988 ancestor, +2.4% in the earliest evolved group, −5.4% group-wise in the anti-potentiated interval with five individual strains worse than −20%, +1.7% in the final group, +0.9% measured inside the whole 31,000-generation population. Also explicit that direct-selection conditions "dramatically increase the fitness benefit" and would make citT activation "universally beneficial in any genetic background." This paper answers the stepwise question with numbers. Anyone arguing either side without it is arguing without the data. Key leon-2018.
Turner, Blount, Mitchell & Lenski 2023 (Microbiology 169:001390). Four things this article needs. (i) The peer-reviewed, dated statement that Cit+ "has so far evolved in only one of the twelve LTEE populations ... even after 75 000 generations" — replacing a blog post at 64,000. (ii) The real mechanism of Cit−/Cit+ coexistence, which is cross-feeding on the succinate, fumarate and malate a Cit+ cell must export in order to import citrate, not simply glucose specialisation. (iii) The LTEE spontaneously repaired the broken dcuS gene in the Cit− clade, by a five-base deletion that restored the reading frame, proved by rebuilding that allele in the ancestor's own genome — the direct answer to Minnich & Hovde's 2025 strain argument, published two years before it and not cited in it. (iv) A measured fitness decline in the Cit− ecotype when its cross-feeding partner is removed. Obtained late, after the first draft was written; it changed four sentences. Key turner-2023.
uz-Zaman, D'Alton, Barrick & Ochman 2024 (PLoS Biology 22:e3002418). Mines the LTEE's RNA-seq and Ribo-seq data for proto-genes — previously non-coding sequence that has begun to be transcribed. Nine confirmed, five of them created by an IS150 element parking its outward-facing promoter upstream, one with ribosome-profiling evidence of translation into a 117-amino-acid ORF, at an estimated rate of about one per 60,000 generations. The authors are explicit that no function or fitness effect has been shown for any of them. This falsifies the flat claim, made in the first draft of this article and made routinely in creationist writing, that nothing has ever arisen from non-coding DNA in the LTEE. It is also another instance of promoter capture, which is this article's central mechanism. Key uz-zaman-2024.
Couce, Limdi, Magnan, Owen, Herren, Lenski, Tenaillon & Baym 2024 (Science 383:eadd1417). Transposon-insertion libraries in the ancestor and in evolved clones, measuring the fitness effect of hundreds of thousands of insertions in every gene. Two numbers the Behe dispute needed and did not have: presumptive inactivating mutations "contribute most (>50%) of the early adaptive mutations in the LTEE, and they continue to be a sizable fraction over the long run (~25%)"; and the beneficial fraction of all possible gene disruptions falls 6.8% → 4.3% → 3.2% from ancestor to 2,000 to 15,000 generations while the deleterious fraction holds flat near 20%. Read as the NIH author manuscript through Europe PMC; the direct PMC PDF URL returns a challenge page. Concedes half of Behe's case and refutes the other half, with a number, in the same paper. Key couce-2024.
Izutsu, Lake, Matson, Dodson & Lenski 2024 (Microbiology 170:001494). 48 populations at 2-, 8-, 100- and 1000-fold daily dilution for 150 days, against theory and simulation. Adaptation started earlier and went further at the larger dilutions, the signature of a regime where beneficial mutations are common enough to interfere with one another. Cited for one clause: the LTEE's own 100-fold transfer sits in the clonal-interference regime. Key izutsu-2024.
Chihoub, Pintard, Lenski, Tenaillon & Couce 2025 (PNAS 122:e2501901122). Fitness effects of RNA-polymerase mutations across a 15,000-generation trajectory: "Fitness effects of both beneficial and deleterious mutations were attenuated in fitter backgrounds. In contrast, pleiotropic effects became more severe and widespread with greater adaptation." Obtained and read; not cited, because it is a background-dependence result rather than an LTEE headline. Its relevance, if the article is expanded, is to the specialisation argument: adapting to a simple environment makes an organism more fragile elsewhere. Key chihoub-2025.
Blount et al. 2020 (eLife 9:e55414). 2,500 generations of Cit+ populations evolved in citrate-only (DM0) and glucose-plus-citrate (DM25) media; maeA amplifications; the live/dead staining that gives Behe his death figures (10.7% ancestral, 29.6% and 39.9% for Cit+ clones in DM25, 33.1% and 44.2% in DM0); and the candid admissions that mortality "varied considerably even between replicate assays for reasons that we do not understand" and that "most evolved clones show no significant increase in fitness, and some are less fit than their ancestor." Cuts toward Behe on the phenotype and against him on the framing. Key blount-2020.
Minnich & Hovde 2025 (J. Physiol. 603:1325-1328). A letter to the editor replying to Root-Bernstein & Bernstein's 2024 agent-based model. Restates the 10-14 probability argument with its two component estimates, and makes the strongest version of the strain argument: E. coli B carries a five-base-pair dcuS deletion that K-12 does not, so REL606 needs an extra mutation K-12 does not need. Concludes that with K-12 "it is likely that all 12 flasks would rapidly have become Cit+." Open access, checkable, and unrefuted. Key minnich-2025.
Root-Bernstein & Bernstein 2024, "'Evolutionary poker': An agent-based model of interactome emergence and epistasis tested against Lenski's long-term E. coli experiments," J. Physiol. 602:2511-2535. Not obtained. It is the target of Minnich & Hovde's letter, and the letter quotes the two passages it disputes with page references, so the article's use of the exchange does not depend on reading the original. Worth obtaining before any revision that discusses the model itself.
The Discovery Institute's case
Behe 2010, "Experimental Evolution, Loss-of-Function Mutations, and 'The First Rule of Adaptive Evolution'," Quarterly Review of Biology 85:419-445. The gain-of-FCT / loss-of-FCT / modification-of-function scheme, and the survey of four decades of microbial evolution experiments behind the rule. Retrieved from a Wayback snapshot of the author's Lehigh copy; the journal version is closed access. The classification scheme is the whole argument. Read it before disputing anything downstream of it. Key behe-2010.
Behe 2019, Darwin Devolves, chapter 7. The LTEE chapter. Source of "it's very likely that all of the identified beneficial mutations worked by degrading or outright breaking the respective ancestor genes," of the "sideshow" verdict on citrate, of the antimutator example, and of the endnote conceding the Maddamsetti fine-tuning result while dismissing it as "probably" degradation. Read from the EPUB in the library. Key behe-2019.
Lents, Swamidass & Lenski 2019, "The end of evolution?", Science 363:590. The review of Darwin Devolves. The journal version is paywalled; the full text as posted on the Science "Books et al." blog was retrieved through the Wayback Machine and matches the passages Behe quotes back at it. Key lents-2019.
Lenski 2019, "Is the LTEE breaking bad?", Telliamed Revisited, 26 February 2019. Lenski's own longer answer on the LTEE specifically. Concedes the losses explicitly, states that the LTEE "was not intended to mimic the complexities of nature," and identifies the assumption behind Behe's "probably degrading" — that the ancestral protein was already optimal. The most useful single statement of the mainstream position on this experiment, and unusually candid about what the experiment cannot show. Key lenski-2019.
Behe 2020, "Citrate Death Spiral," Discovery Institute. Behe on Blount et al. 2020. Every number in it checks out against the paper; the framing does not. Key behe-2020.
McLatchie 2026, "Answering Dave Farina on the Lenski Experiment," Science and Culture Today, 18 August 2026. The current statement of the stepwise-citrate claim, and the article this project was asked to examine. Its four-sentence summary of the path is quoted in full in the main article. Note that its endnote attributes the stepwise-with-advantage description to Van Hofwegen et al. 2016, which reports the steps but measures no fitness effect for any of them. Key mclatchie-2026b.
Behe 2007, The Edge of Evolution, chapter 7, subsection "Lab Star." Model: a limit on what mutation and selection can build, estimated from malaria, HIV and the LTEE (at 10^13 cells). Direction: complicates. Verified: full (library EPUB, no printed pages). Uses the LTEE as one of four data points and says "After a while, beneficial changes from the experiment petered out," citing de Visser & Lenski 2002, which reported a declining rate. Lenski 2015's power law contradicts "petered out"; the article names this as the creationist-side forecast the fitness data refuted. Key behe-2007b.
Behe 2012, "Rose-Colored Glasses: Lenski, Citrate, and BioLogos," Evolution News, 13 November 2012. Model: Behe's own application of the Edge counting rule to Blount 2012. Direction: complicates (it places Cit+ inside the edge, so the result neither confirms nor refutes the Edge). Verified: full (page downloaded, text extracted). "Only required deleterious and neutral mutations count against the limit"; counts "perhaps three to four mutations"; reports Venema's "roughly a half-dozen"; says he would have expected "a single step could achieve it," which Leon 2018 later measured. Key behe-2012d.
Young-earth creationist treatments
Purdom 2008, "A Poke in the Eye?", Answers in Depth 3, Answers in Genesis, 30 June 2008. Model: designed adaptive capacity; Cit+ as loss of regulation or specificity in an existing transporter. Direction: supports in part (the existing-transporter prediction held) and undercuts in part (the "loss of regulation" reading did not). Verified: full. Quotes Blount 2008's own co-option sentence, so the prediction was not distinctive. Key purdom-2008b.
Purdom 2012, "Bacteria Evolve 'Key Innovation' or Not?", Answers in Genesis blog, 8 November 2012. Direction: undercuts the creationist case where it is quoted. Verified: full. Says "There was a definite loss of regulation" (contradicted by Blount 2012's reporter data: the native citT control region stays silent in a Cit+ cell) and that she was "amazed at just how many changes were necessary" (overtaken by Quandt 2014 and Leon 2018). Key purdom-2012.
Batten 2008, with a March 2016 addendum, "Bacteria 'evolving in the lab'?", Creation Ministries International. Direction: complicates. Verified: full. The addendum concedes "the existing controls and the citrate transporter gene were not actually broken" and then calls control "indeed broken"; says Cit+ is not beyond Behe's edge. Its claim that Cit+ cells have "about a 20% reduced ability to feed on glucose" has no citation and is not used. Key batten-2008b.
Carter 2011, updated 2025, "Can mutations create new information?", Journal of Creation 25(2):92-98 and creation.com. Direction: complicates. Verified: full. Files aerobic citrate use under "degraded information" as "loss of control," citing Batten 2008; also holds that a function from duplication "must necessarily be related to the function of the old," which Cit+ fits. Key carter-2025.
Sanford 2014, Genetic Entropy, 4th edition, chapter 11, section 2b on the LTEE. Model: genetic entropy. Direction: complicates. Verified: full (library EPUB). Expects the LTEE's fitness rise ("not unexpected or remarkable in any way"), claims every adaptation was loss of function or regulation, and forecasts extinction "over hundreds of years." The article quotes all three and prices the forecast by drift at LTEE population size. Key sanford-2014.
Wile 2016, "Another Creationist Prediction Confirmed," Jay Wile's blog (Goodreads mirror), 16 May 2016. Direction: supports, weakly. Verified: full. Reads Van Hofwegen 2016 as confirming Purdom 2008; the article notes that "easy" is Wile's gloss on Purdom's "conceivable." Key wile-2016b.
Fabich on citrate: searched for (two web searches, the library) and not found. The library's fabich-2016 is on infectious disease and the Ark. Considered and not used; "not found" is not "does not exist."
Background and current status
Mauger 2025, "'Evolution under a microscope' going strong at MSU," MSUToday, 11 November 2025. A university press release, used only for the generation count (82,000) and the custody history (Michigan State → UT Austin in 2022 at 75,000 generations → back to Michigan State with Barrick in 2025). Not a scientific source and not treated as one. Key msu-2025.
Lenski 2020, "Time to restart the LTEE, this virus be damned," Telliamed Revisited, 1 April 2020. An April Fools' post. Cited for exactly one thing: the design figure "each population produces around half a billion new cells each day," which is independently derivable from Lenski et al. 1991 and is used here only as corroboration. It also records the 9 March 2020 freeze, the experiment's only significant interruption.
Leiby & Marx 2014 (PLoS Biology 12:e1001789). Argues metabolic erosion in E. coli runs mainly through mutation accumulation rather than trade-offs — the opposite conclusion to Cooper & Lenski 2000 on the mechanism, though not on the fact of the decay. Included so a reader knows the mechanism is contested. Key leiby-2014.
Lenski, Ofria, Pennock & Adami 2003 (Nature 423:139-144), "The evolutionary origin of complex features." The Avida digital-organism paper. Obtained and read, not cited in the article: it is a simulation, not the LTEE, and belongs to the evolutionary-computation topic rather than this one. Retained in the library for that article.
Rozen & Lenski 2000 (American Naturalist 155:24-35), "Long-term experimental evolution in Escherichia coli. VIII. Dynamics of a balanced polymorphism." The Ara–2 S/L cross-feeding polymorphism. Obtained, not cited: Good et al. 2017 covers coexistence across all twelve populations, which is the stronger form of the same point.
Wielgoss et al. 2013 (PNAS 110:222-227). The primary source for the mutator dynamics in population Ara–1: the mutT frameshift raised the point-mutation rate ~150-fold and the later mutY alleles brought it back down 40-60%. Cited late. An earlier draft took these facts from Behe's trade book, which swaps the two gene descriptions — it is mutT that sanitises oxidised nucleotides and mutY that is the repair glycosylase — and which omits the authors' own hedge that the mutY mutations may not have been adaptive. Key wielgoss-2013.
Hottes, Freddolino, Khare, Donnell, Liu & Tavazoie 2013 (PLoS Genetics 9:e1003617). A meta-analysis of genome-wide fitness data across 144 growth conditions: "at least one beneficial null mutation was identified in all but five of the 144 conditions considered." This is the paper Behe cites for the general form of his first rule, and an earlier draft of this article asserted that no general argument existed. Key hottes-2013.
What was checked and found not to matter
- The negative claims — "no new protein fold," "no new enzymatic chemistry," "no named species." These were checked against the LTEE publication list (the BibTeX behind
the-ltee.org/publications/, which is the site's one working page), against the Europe PMC citing-works lists for Blount 2012 and Tenaillon 2016, and by targeted searches for de novo gene birth, novel folds and speciation in the LTEE. The one hit was uz-Zaman et al. 2024 on proto-genes, which is now in the article. A negative claim over an eighty-paper corpus can never be closed; this is how far it was pushed. - The
the-ltee.orgofficial site. Its History and Introduction pages return no content and the templates carry injected spam links. Not usable as a source; the MSU press release was used for the generation count instead. - A peer-reviewed reply to Van Hofwegen et al. 2016. There is none. Lenski and Blount answered on a blog three days after the accepted manuscript posted, and no journal comment followed. The scientific answer came instead in the form of new data — Leon et al. 2018.
- A published statement that no second LTEE population has evolved Cit+ since 2016. An earlier draft of this literature review recorded that no such statement existed and that the article had to rest on Lenski's blog at 64,000 generations. That was wrong. Turner et al. 2023 open their Discussion with it: Cit+ "has so far evolved in only one of the twelve LTEE populations ... even after 75 000 generations." The lesson is the standing one — "I could not find it" is not "it does not exist," and a second search route found it.
Open questions this literature does not answer
- Would twelve K-12 populations have gone Cit+? Minnich's dcuS argument predicts yes, and nobody has run the experiment. It is checkable in principle and would settle more of this dispute than any further argument about existing data. Note what has already narrowed it: Turner et al. 2023 showed the LTEE repaired the dcuS frameshift by itself, in the Cit− clade, by a five-base deletion in a three-copy tandem array — so the "extra mutation" REL606 needs is a slipped-strand event at repeat rates, not the point mutation Minnich prices it as.
- What is the actual rate of citT activation under LTEE conditions? Blount measured it on minimal-citrate agar over 45 days of incubation, which gives a multi-step process far more opportunity per nominal generation than a daily transfer cycle does. The number that matters — the per-generation rate in a shaken flask with a 1:100 daily dilution — has never been measured, and both the "it should have happened many times" and "it happened exactly on schedule" arguments turn on it.
- What fraction of LTEE beneficial mutations degrade protein function? Bounded from below by Couce et al. 2024, not answered: more than 50% of the early adaptive mutations and about 25% of the long-run ones are presumptive inactivations. The other ~75% are unclassified, not shown to be non-degrading, so the count neither confirms nor refutes Behe's "all." "Presumptive inactivating" is an inference from mutation class — nonsense, frameshift, indel — not an enzyme assay, and a missense change at a binding interface, counted here as non-inactivating, may still cut activity. So 25% is a floor for long-run degradation, not a ceiling. The only enzyme kinetics anyone has actually run on an LTEE mutation are Quandt's, on citrate synthase.
- Why has no other population evolved Cit+ in 49,000 further generations? Anti-potentiation was specific to Ara–3's history, so the other eleven should not share it. Either clonal interference is a much stronger barrier than the Leon analysis implies, or something about the shared REL606 background is the real constraint — which would favour Minnich.