Abstract
Adaptive laboratory evolution (ALE) is used extensively in industrial microbiology and metabolic engineering for organism improvement. The extensive amount of data available for evolved Escherichia coli – its alleleome - provides an opportunity to assess the effectiveness of this technique in sampling this organism’s available amino acid space. Consistent with the underlying nature of the biological code, among approximately 25,000 amino acid substitutions less than half of the possible amino acid shifts are observed. The directionality of amino acid shifts associated with high growth temperature is not observed in E. coli strains evolved for thermotolerance, other than a possibly significant enrichment of aspartate to glutamate shifts. These results reinforce the relatively constrained nature of short-term E. coli evolutionary experiments, highlighting the need for consideration of complementary approaches for organism improvement.
Introduction
Humanity has long relied on evolution in the development of systems for meeting our material and energy needs. Such examples include, but are not limited to, the domestication of crops (), livestock () and medicinal plants (). This dependence extends to a reliance on yeast and other microbes for the production of fermented food and beverages, with this long relationship history documented in the genetic history of organisms such as Saccharomyces cerevisiae ().
Our long history of shaping the evolution of organisms that we find useful is also represented in our use of adaptive laboratory evolution (ALE) for both understanding biological systems (; ) and improving biomanufacturing platforms (; ). ALE can compensate for our imperfect knowledge of biological systems, as summarized in Orgel’s 2nd rule “evolution is cleverer than you are” () and it has proven especially useful for improving bioproduction when maintenance of redox or ATP generation is associated with production of the target molecule (). Reverse engineering and retrospective analysis of mutations acquired during ALE provide learning opportunities and support the goal of being able to ‘predictably program biology in the same way in which we write software’ ().
There is a substantial amount of accumulated ALE data for the model organism Escherichia coli. Analysis of the E. coli alleleome () featured a compilation of data from the ALE database (ALEdb) (), including long-term evolutionary experiments (LTEE). The roughly 33,000 mutations documented in the E. coli alleleome were compared to naturally-occurring sequence variation, with the conclusion that the mutations acquired during ALE are more severe and thus more likely to impact protein properties (). This compiled data also provides an excellent opportunity for the assessment of the effectiveness of ALE of E. coli in sampling the available biological space, as presented here.
Evolution’s reach is relatively short
This analysis focuses on non-synonymous amino acid substitutions and thus excludes mutations outside of coding regions, synonymous mutations, stop-gains and nonstops. These substitutions are unique in terms of their location within the genome, meaning multiple independent observations of the same amino acid shift at the same position within the same gene are counted as a single substitution. This reduction of the data leaves 25,286 observed amino acid shifts from 278 unique E. coli experiments.
While there are 190 possible non-synonymous exchanges, only 88 (46%) of these are observed within this data (Figure 1A). The fact that less than half of the possible amino acid substitution space is being sampled demonstrates the relatively short reach of the ALE approach. Additionally, interconversion of alanine and valine and alanine and threonine dominate the dataset, accounting for more than 18% of the observed amino acid shifts.
Figure 1
Catoiu et al. used the Grantham Score to quantify the severity of amino acid substitutions in ALE strains relative to sequence variation within wild-type strains. The Grantham Score is one of several scores based on amino acid physicochemical properties. Alternatively, Yampolsky and Stoltzfus estimated amino acid exchangeability scores from experimental data, where a high exchangeability trends with a low severity of effect on protein function () (Figure 1B). For example, lysine → alanine has a high exchangeability score (600), consistent with a reported lysine → alanine substitution that increased the half-life of β-1,3-1,4-glucanase at 70 °C by more than 5-fold (). Similarly, glutamine → valine has a score of 603 and this substitution increased NADH-dependent catalytic efficiency of ketol-acid reductoisomerase (IlvC) by more than 80-fold (). However, these two shifts, along with several other high-exchangeability substitutions, are not observed in the E. coli alleleome data (Figure 1B). This suggests that at least some of the under-sampled amino acid shifts could possibly improve protein or enzyme function, though this does not necessarily imply that such shifts could result in resistance to the applied selective pressure.
The observed frequency of amino acid sampling is consistent with the genetic code and genome sequence
The observed sparsity of amino acid sampling is consistent with the underlying nature of the biological code. Some amino acid substitutions require only a single nucleotide change (‘hop’), while others require two or even three, as shown in Figure 1B. Thus, there is a difference in the degree of mutation accessibility. The type of base pair change is also relevant. Transition mutations involve the interconversion of purines (A and G) or the interconversion of pyrimidines (C and T). The interconversion between purines and pyrimidines is a transversion mutation. Transition mutations are generally known to occur more frequently relative to transversions () and thus there is an expected difference in amino acid shifts according to the type of mutation required. Again, the analysis presented here is restricted to mutations leading to amino acid shifts. For example, while there are 192 possible 1-transition codon shifts, only 116 of these result in a non-synonymous amino acid substitution in E. coli. For example, the shift from AAA to AAG is synonymous (lysine).
Amino acid shifts enabled by a single transition mutation account for 14,225 (56.26%) of the observed substitutions studied here, with single transversion mutations accounting for 10,766 (42.58%). Only 295 of the 25,286 (1.17%) of the observed substitutions were incurred with two hops and none of the codon shifts requiring three mutations were observed. Thus, the rarity of multi-hop mutations corresponds to the sparse observation of some amino acid substitutions in this E. coli alleleome.
However, not all poorly sampled amino acid substitutions can be attributed to the requirement of multiple hops. For example, while there are two single-transversion mutations that result in an arginine → threonine shift (AGA → ACA, AGG → ACG), there are no observations of this substitution. The low occurrence of these codon shifts can be attributed to the low frequency of these codons in the E. coli genome, 2.05 and 1.22 out of 1000, respectively. The observed frequency of 1-hop mutations generally trends with source codon frequency for both mutation types (Figure 1C).
Thus, the apparently relatively short reach of ALE within this E. coli dataset is consistent with the underlying nature of the genetic code. Some shifts are rare because they require multiple mutations within a single codon, or because of the type of nucleotide mutation required, while others are rare because the corresponding source codon has a low frequency. Others may be rare due to the lack of selective advantage provided by the shift.
Evolution for thermotolerance: single strain ALE vs. organismal differences
Thermotolerance has been a popular proving ground for E. coli evolutionary studies, including six studies featured here published between 2012 and 2021 (; ; ; ; ; ). Together, these six papers have observed 685 unique amino acid shifts that are expected, on average, to support increased thermotolerance over up to approximately 5,000 generations. In contrast to the relatively short-term nature of ALE of a single organism, Pinney et al. leveraged eons-deep organism temperature tolerance data to perform a deep mechanistic study of temperature adaptation. Specifically, they analyzed 14,399 site-specific amino acid changes in order to identify shifts associated with growth at high temperature vs growth at low temperature (). The alignment between these two datasets – amino acid shifts associated with ALE of E. coli for thermotolerance and the identity of amino acid shifts associated with organism temperature tolerance – can be informative regarding the similarity of short-term and long-term evolutionary strategies.
Pinney et al. analyzed the reciprocity of amino acid shifts in relation to the encoding organisms’ growth temperature. For example, while there were 387 instances of a high growth temperature homolog encoding isoleucine relative to a low growth temperature homolog encoding a leucine, there were only 122 instances of the reverse trend (p = 7x10–32 by Fisher’s exact test). Pinney et al. identified 26 amino acid shifts with this significant enrichment of directionality (Table 1). Thirteen of these shifts require two or three hops and were not observed in these E. coli thermotolerance ALE studies; their absence can probably be attributed to the low frequency of such mutations. Pinney et al. used a p-value of 5.3x10–5 as criteria for significance, based on the Bonferonni correction factor and comparison of 190 amino acid pairs. No amino acids shifts in the thermotolerance ALE dataset met this criterion for significance. Three of the shifts could be considered significantly enriched with application of the Benjamini-Hochberg correction (α = 0.05), but only the D/E shift is enriched in the same direction as the organism-wide comparison, the other two (Q/K and F/I) are enriched in the opposite direction. Thus, alignment of these E. coli thermotolerance evolutionary studies with Pinney’s analysis is poor.
Table 1
| Residue in low Tgrowth/high Tgrowth homologs | # observed by Pinney et al. (forward, reverse) | # observed in thermotolerance ALE strains (p-value) |
|---|---|---|
| A/E | 147, 35 | 4, 14 (0.018) |
| A/I | 133, 63 | 0, 0 |
| A/K | 202, 28 | 0, 0 |
| A/L | 162, 88 | 0, 0 |
| A/R | 86, 30 | 0, 0 |
| A/V | 178, 93 | 34, 20 (0.056) |
| A/Y | 84, 21 | 0, 0 |
| D/E | 190, 71 | 17, 3 (0.0018)a |
| D/K | 91, 27 | 0, 0 |
| F/I | 114, 42 | 1, 9 (0.011)b |
| G/K | 97, 20 | 0, 0 |
| G/N | 81, 31 | 0, 0 |
| H/Y | 105, 16 | 7, 6 (0.78) |
| L/I | 387, 122 | 1, 7 (0.034) |
| M/I | 87, 35 | 10, 6 (0.32) |
| P/K | 92, 26 | 0, 0 |
| P/Y | 25, 3 | 0, 0 |
| Q/K | 65, 13 | 1, 10 (0.0067)b |
| R/E | 106, 33 | 0, 0 |
| R/I | 54, 11 | 0, 0 |
| R/K | 225, 71 | 0, 4 (0.046) |
| T/I | 84, 30 | 7, 10 (0.47) |
| T/K | 55, 19 | 1, 4 (0.18) |
| V/I | 315, 152 | 7, 5 (0.56) |
| W/I | 28, 4 | 0, 0 |
| W/Y | 61, 20 | 0, 0 |
Amino acid shifts determined by Pinney et al. to be associated with growth at high temperature (n=14,399, p<5.3x10-5) are not enriched in thermotolerant strains generated by ALE (n=685).
p-values were determined by chi-squared tests relative to the null hypothesis of equal occurrence in both the forward and reverse direction. a,bindicate significance (α=0.05) according to the Benjamini-Hochberg correction, considering only the amino acid shifts with non-zero observances. aindicates the same directionality as Pinney et al.; bindicates the opposite directionality.
The absence of significant reciprocity for these amino acid pairs in the E. coli thermotolerance ALE dataset could be due to the relatively small nature of the dataset. The requirement of p < 5.3x10–5 could be satisfied with an observation of 15 shifts in one direction and zero shifts in the other. The E. coli ALE thermotolerance dataset has five pairs of amino acids with at least 15 total observed exchanges and yet none meet the requirement for significance. Thus, the alignment between these two datasets – one that compares distinct organisms separated by eons of evolutions and one that compares strains of the same species subjected to relatively few generations of evolution - is low.
Discussion
The analysis presented here demonstrates that the available biological space in E. coli is being under-sampled by ALE, in a manner that is consistent with the underlying nature of the biological code. The dramatic increase in accessibility of genome sequencing and analysis has enabled this data collection and analysis. This issue of sparse sampling within the natural set of amino acids needs more attention and discussion. For example, the data analysis pipeline should be primed to recognize rarely-sampled substitutions.
While there have been advances in directed evolution of individual enzyme targets in order to improve amino acid sampling (Pines, ; ; ), these negate the ‘clever’ nature of evolution by requiring the a priori determination of protein targets, as opposed to the organism-wide scope of ALE. A strategy for refactoring of the E. coli genetic code in order to increase access to these under-sampled substitutions has been presented, but requires recoding 60 of the 64 codons (). Recoding of E. coli codons has been demonstrated, but at a more limited scale (; ). But until there is an experimental strategy for compensating for the short reach of ALE, it is important to use multiple strategies for organism improvement and not rely exclusively on an evolutionary approach.
Limitations
This analysis is restricted to the E. coli alleleome, as assembled by (). LTEE-type experiments have not been considered, only those described as ALE. The redacted nature of the alleleome dataset means that the selective pressure(s) used and associated number of generations is not publicly available.
Statements
Data availability statement
The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author.
Author contributions
LJ: Writing – original draft, Writing – review & editing.
Funding
The author(s) declared that financial support was received for this work and/or its publication. This work was supported by the Department of Energy (DOE) Office of Biological and Environmental Research under award #DE-SC0022090. The funding body had no role in the design of the study, data collection and analysis, decision to publish, or preparation of the manuscript.
Acknowledgments
I am thankful to Adam Feist (University of California, San Diego), Robert Jernigan (Iowa State University), and Thomas Mansell (Iowa State University) for helpful conversations.
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that generative AI was not used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
References
1
BastianS.LiuX.MeyerowitzJ. T.SnowC. D.ChenM. M. Y.ArnoldF. H. (2011). Engineered ketol-acid reductoisomerase and alcohol dehydrogenase enable anaerobic 2-methylpropan-1-ol production at theoretical yield in Escherichia coli. Metab. Eng.13, 345–352. doi: 10.1016/j.ymben.2011.02.004. PMID:
2
BidenJ. R. (2022). Advancing biotechnology and biomanufacturing innovation for a sustainable, safe, and secure American bioeconomy. Fed. Regist.87, 12.
3
BlabyI.LyonsB.Wroclawska-HughesE.PhillipsG.PyleT.ChamberlinS.et al. (2012). Experimental evolution of a facultative thermophile from a mesophilic ancestor. Appl. Environ. Microbiol.78, 144–155. doi: 10.1128/aem.05773-11. PMID:
4
CaspetaL.ChenY.GhiaciP.FeiziA.BuskovS.HallstromB. M.et al. (2014). Altered sterol composition renders yeast thermotolerant. Science346, 75–78. doi: 10.1126/science.1258137. PMID:
5
CatoiuE. A.PhaneufP.MonkJ.PalssonB. O. (2023). Whole-genome sequences from wild-type and laboratory-evolved strains define the alleleome and establish its hallmarks. PNAS120, e2218835120–e2218835120. doi: 10.1073/pnas.2218835120. PMID:
6
DeatherageD. E.KepnerJ. L.BennettA. F.LenskiR. E.BarrickJ. E. (2017). Specificity of genome evolution in experimental populations of Escherichia coli evolved at different temperatures. PNAS114, E1904–E1912. doi: 10.1073/pnas.1616132114. PMID:
7
DunitzJ. D.JoyceG. F. (2013). “ Leslie E. Orgel,” in Biographical memoirs ( Washington, DC, USA: National Academy of Science).
8
FangY.YuW.LiJ.LaoL.YuanY.ChenY.et al. (2026). Recoding multiple rare codons enables the simultaneous incorporation of up to five distinct noncanonical amino acids. Nat. Chem. 18, 695-706. doi: 10.1038/s41557-026-02084-y. PMID:
9
GoodB. H.McDonaldM. J.BarrickJ. E.LenskiR. E.DesaiM. M. (2017). The dynamics of molecular evolution over 60,000 generations. Nature551, 45–4+. doi: 10.1038/nature24287. PMID:
10
HortonJ.TaylorT. (2023). Mutation bias and adaptation in bacteria. Microbiology-SGM169, 001404. doi: 10.1099/mic.0.001404. PMID:
11
JesusA.BonhommeV.EvinA.IvorraS.SoterasR.SalavertA.et al. (2021). A morphometric approach to track opium poppy domestication. Sci. Rep.11, 9778. doi: 10.1038/s41598-021-88964-4. PMID:
12
JinT.ChenY.JarboeL. R. (2016). “ Evolutionary methods for improving the production of biorenewable fuels and chemicals,” in Biotechnology for biofuel production and optimization. Ed. TrinhC. E. C. ( Elsevier).
13
KosakaT.NakajimaY.IshiiA.YamashitaM.YoshidaS.MurataM.et al. (2019). Capacity for survival in global warming: Adaptation of mesophiles to the temperature upper limit. PLoS One14, e0215614. doi: 10.1371/journal.pone.0215614. PMID:
14
LambrosM.Pechuan-JorgeX.BiroD.YeK.BergmanA. (2021). Emerging adaptive strategies under temperature fluctuations in a laboratory evolution experiment of Escherichia coli. Front. Microbiol.12. doi: 10.3389/fmicb.2021.724982. PMID:
15
LarsonG.FullerD. (2014). The evolution of animal domestication. Annu. Rev. Ecol. Evol. Syst.45, 115–136. doi: 10.1146/annurev-ecolsys-110512-135813. PMID:
16
LeeJ. M.MoonS. Y.KimY. R.KimK. W.LeeS. J.KongI. S. (2017). Improvement of thermostability and halostability of beta-1,3-1,4-glucanase by substituting hydrophobic residue for Lys(48). Int. J. Biol. Macromol.94, 594–602. doi: 10.1016/j.ijbiomac.2016.10.043. PMID:
17
LegrasJ.MerdinogluD.CornuetJ.KarstF. (2007). Bread, beer and wine:: Saccharomyces cerevisiae diversity reflects human history. Mol. Ecol.16, 2091–2102. doi: 10.1111/j.1365-294X.2007.03266.x
18
MeyerR.DuValA.JensenH. (2012). Patterns and processes in crop domestication: an historical review and quantitative analysis of 203 global food crops. New Phytol.196, 29–48. doi: 10.1111/j.1469-8137.2012.04253.x. PMID:
19
OstrovN.LandonM.GuellM.KuznetsovG.TeramotoJ.CervantesN.et al. (2016). Design, synthesis, and testing toward a 57-codon genome. Science353, 819–822. doi: 10.1126/science.aaf3639. PMID:
20
PhaneufP. V.GostingD.PalssonB. O.FeistA. M. (2019). ALEdb 1.0: a database of mutations from adaptive laboratory evolution experimentation. Nucleic Acids Res.47, D1164–D1171. doi: 10.1093/nar/gky983. PMID:
21
PinesG.PinesA.GarstA. D.ZeitounR. I.LynchS. A.GillR. T. (2015). Codon compression algorithms for saturation mutagenesis. ACS Synth. Biol.4, 604–614. doi: 10.1021/sb500282v. PMID:
22
PinesG.WinklerJ. D.PinesA.GillR. T. (2017). Refactoring the genetic code for increased evolvability. MBIO8, e01654-17. doi: 10.1128/mbio.01654-17. PMID:
23
PinneyM. M.MokhtariD. A.AkivaE.YabukarskiF.SanchezD. M.LiangR.et al. (2021). Parallel molecular mechanisms for enzyme temperature adaptation. Science371, 1010–101+. doi: 10.1126/science.aay2784. PMID:
24
SandbergT.PedersenM.LaCroixR.EbrahimA.BondeM.HerrgardM.et al. (2014). Evolution of Escherichia coli to 42 °C and subsequent genetic engineering reveals adaptive mechanisms and novel mutations. Mol. Biol. Evol.31, 2647–2662. doi: 10.1093/molbev/msu209. PMID:
25
SayousV.LubranoP.LiY. Y.Acevedo-RochaC. G. (2020). Unbiased libraries in protein directed evolution. Biochim. Biophys. Acta Proteins Proteomics1868, 140321. doi: 10.1016/j.bbapap.2019.140321. PMID:
26
TenaillonO.Rodriguez-VerdugoA.GautR. L.McDonaldP.BennettA. F.LongA. D.et al. (2012). The molecular diversity of adaptive convergence. Science335, 457–461. doi: 10.1126/science.1212986. PMID:
27
WannierT. M.KunjapurA. M.RiceD. P.McDonaldM. J.DesaiM. M.ChurchG. M. (2018). Adaptive evolution of genomically recoded Escherichia coli. PNAS115, 3090–3095. doi: 10.1073/pnas.1715530115. PMID:
28
YampolskyL. Y.StoltzfusA. (2005). The exchangeability of amino acids in proteins. Genetics170, 1459–1472. doi: 10.1534/genetics.104.039107. PMID:
29
YuH. R.MaS.LiY. W.DalbyP. A. (2022). Hot spots-making directed evolution https://loop.frontiersin.org/people/353085easier. Biotechnol. Adv.56, 107926. doi: 10.1016/j.bioteChadv.2022.107926. PMID:
Summary
Keywords
adaptive, evolution, mutation, transition, transversion
Citation
Jarboe LR (2026) Adaptive laboratory evolution of Escherichia coli is clever, but its reach is also relatively short. Front. Ind. Microbiol. 4:1840830. doi: 10.3389/finmi.2026.1840830
Received
27 March 2026
Revised
18 April 2026
Accepted
23 April 2026
Published
08 May 2026
Volume
4 - 2026
Edited by
Kuppam Chandrasekhar, Kyungpook National University, Republic of Korea
Updates
Copyright
© 2026 Jarboe.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Laura R. Jarboe, ljarboe@iastate.edu
Disclaimer
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.