In the cell, most tasks are performed by big molecules called proteins, which behave like molecular machines. Although proteins are often described as having one job each, this is not always true, and many proteins can perform different roles. Enzymes are a type of protein that facilitate chemical reactions. They are often specialised to one reaction, but they can also accelerate other side-reactions. During evolution, these side-reactions can become more useful and, as a result, the role of the enzyme may change over time.

The main role of the enzyme called NfsA in Escherichia coli bacteria is thought to be to convert molecules called quinones into hydroquinones, which can protect the cell from toxic molecules produced in oxidation reactions. As a side-reaction, NfsA has the potential to protect bacteria from an antibiotic called chloramphenicol, but it generally does this with such low efficacy that the effects are negligible. Producing hydroquinones is helpful to the cell in some situations, but if bacteria are regularly exposed to chloramphenicol, NfsA’s role aiding antibiotic resistance could become more important. Over time, the enzyme could evolve to become better at neutralising chloramphenicol. Therefore, NfsA provides an opportunity to study the evolution of proteins and how bacteria adapt to antibiotics.

To see how evolution might affect the activity of NfsA, Hall et al. generated 250 million E. coli with either random or targeted changes to the gene that codes for the NfsA enzyme. The resulting variants of NfsA that were most effective against chloramphenicol all had a change that eliminated the enzyme’s ability to convert quinones. This result demonstrates a key trade-off between roles for NfsA, where one must be lost for the other to improve.

These results demonstrate the interplay between a protein’s different roles and provide insight into bacterial drug resistance. Additionally, the experiments showed that the bacteria with improved resistance to chloramphenicol also became more sensitive to another antibiotic, metronidazole. These findings could inform the fight against drug-resistant bacterial infections and may also be helpful in guiding the design of proteins with different roles.

Many (if not all) enzymes are promiscuous, meaning that in addition to their primary biological role(s) they can catalyse minor side reactions that have no apparent physiological relevance, either because they are too inefficient or because the substrate is not naturally encountered (Copley, 2015). From an evolutionary perspective, promiscuity can play an important role in contingency, providing a reservoir of potential functions that a cell can tap in response to changing circumstances (O'Brien and Herschlag, 1999; Copley, 2015). As demonstrated by the emergence of resistance to xenobiotic pollutants or clinical antibiotics (O'Brien and Herschlag, 1999; Hall, 2004; Ramos et al., 2005; Copley, 2009; Khersonsky and Tawfik, 2010), a strong selection pressure can cause latent promiscuous activities to be rapidly amplified to physiologically relevant levels (Newton et al., 2015).

Catalytic transitions to an alternate substrate have been modelled experimentally using iterative rounds of random mutagenesis (e.g. error-prone PCR (epPCR)), a powerful directed evolution strategy that enables adaptive landscapes to be explored under defined laboratory conditions (Kaltenbach et al., 2015; Kaltenbach et al., 2016). These laboratory evolution studies have indicated that selection for substantial increases in a promiscuous activity typically results in only weak trade-offs against the native activity; and therefore, the transition from one primary function to another tends to progress via generalist enzyme intermediates (Kaltenbach et al., 2016). Two leading teams have offered contrasting hypotheses to explain this phenomenon. In 2005, Tawfik and co-workers proposed that enzymes possess an innate ‘robustness’ and stability that buffers them against the potentially detrimental effects of novel mutations, coupled with a ‘plasticity’ that can amplify promiscuous functions with relatively few mutations (Aharoni et al., 2005). More recently, Tokuriki and co-workers demonstrated that a robust native activity is not a prerequisite for weak trade-offs, and suggested that the predominance of these in the literature may instead be artefactual; a consequence of laboratory evolution studies being highly biased towards strong selection for a new function without any selection against the native activity (Kaltenbach et al., 2016). They argue that it is unclear how specialisation can occur in this manner, and that in nature, selection might frequently exist to erode the original function. By way of example, they offer a scenario where the native and new substrate compete for the same active site (Kaltenbach et al., 2016).

In addition to their exclusive emphasis on positive selection, we note that the studies overviewed by the Tokuriki team were also heavily biased towards heterologous enzyme expression and/or a transition in activity from one exogenously applied substrate to another (Kaltenbach et al., 2016). Thus, there has been little consideration of how the native substrate might influence the evolutionary trajectory. We were therefore motivated to study the selection of a promiscuous function within the native host environment, with particular focus on the key catalytic changes driving the transition. Recognising that the stochastic nature of iterative random mutagenesis is unlikely to yield the most efficient pathway to a selected outcome, we sought to implement simultaneous mass-mutagenesis on a massive scale that would allow us to retrospectively assess all possible intermediates of our top variants, and infer the most plausible stepwise evolutionary trajectories. We were able to achieve both these goals by employing the Escherichia coli nitro/quinone reductase NfsA as a new model system that offers several key advantages. NfsA is a member of a large bacterial superfamily comprising highly promiscuous FMN-dependent oxidoreductases that accept electrons from NAD(P)H and transfer them to a diverse range of substrates (Williams et al., 2015; Akiva et al., 2017). Expression of nfsA is governed by the soxRS regulon, and NfsA is thought to guard against oxidative stress through reduction of water-soluble quinones such as 1,4-benzoquinone (Liochev et al., 1999; Paterson et al., 2002; Copp et al., 2017). Although most efficient with quinone substrates, NfsA is also able to reduce a wide diversity of nitroaromatic compounds (Valiauga et al., 2017). This is generally believed to represent non-physiological substrate ambiguity, as there are relatively few nitroaromatic natural products, and in many cases nitro-reduction yields a more toxic derivative (Winkler and Hertweck, 2007; Parry et al., 2011; Williams et al., 2015). An important exception is that nitro-reduction of chloramphenicol transforms this antibiotic to a product that is not discernibly toxic to bacteria (Yunis, 1988; Erwin et al., 2007; Crofts et al., 2019). We have observed that over-expressed native NfsA confers only slight chloramphenicol protection to E. coli host cells, but reasoned that we could select for improved detoxification in an extremely high-throughput manner by plating variant libraries on chloramphenicol-amended media. Because members of the bacterial nitroreductase superfamily appear to have unusually plastic active sites (Akiva et al., 2017), we considered that simultaneous mass-mutagenesis of up to eight active-site residues should be possible. In effect, we aimed to strip NfsA of its engine, and then select for a superior configuration of parts assembled within the empty chassis. By leaping directly to a new fitness peak, we considered that we might arrive at synergistic combinations of substitutions that would be difficult to achieve by iterative random mutagenesis approaches.

By exploiting a powerful selection for antibiotic resistance, we were able to implement simultaneous mass-mutagenesis on an unprecedented scale, to amplify a promiscuous functionality. We acknowledge that this approach of focusing exclusively on eight key active-site residues means the reconstructed NfsA ‘engine’ is unlikely to be an optimal fit within the pre-existing chassis, and that further gains in activity would undoubtedly result by selecting residue substitutions in the second shell, or beyond. Nevertheless, we reasoned that our approach would allow us to gain comprehensive insight into key catalytic changes driving improved chloramphenicol detoxification, without being subject to the stochastic vagaries of epPCR, or the well-established phenomenon that it can only access a limited and unbalanced repertoire of residues (on average, only 5.7 of the 19 alternative amino acids per codon position, with a bias towards similar residues [Hermes et al., 1990]). We also initially considered that this approach might allow us to leap to a fitness peak that iterative random mutagenesis strategies would be unable to scale. However, when we examined every hypothetical evolutionary intermediate, we discovered this was not the case; although stepwise evolution would have been greatly constrained in the progression from wild-type NfsA to either of our top two variants, there were plausible trajectories to achieve these outcomes. Notably, the critical first step in any of these trajectories was the substitution of R225, leading to near-total abrogation of the native quinone reductase activity, which was never restored and appears incompatible with the selected activity. The importance of eliminating R225 was emphasised by epPCR mutagenesis of nfsA, when 50/50 unique variants selected for improved chloramphenicol detoxification were found to have lost R225, while 20/20 unselected variants retained it (Supplementary files 1e,f). The strong trade-off between quinone and chloramphenicol reduction is a very different scenario to the predominantly weak trade-offs observed in previous laboratory evolution studies (e.g. those reviewed by Kaltenbach et al., 2016). Even the more recent work of Ben-David et al., 2020 who encountered an abrupt activity trade-off when they evolved the calcium-dependent lactonase mammalian paraoxonase-1 into an efficient organophosphate hydrolase, found that the native functionality could subsequently be restored and was not incompatible with the evolved one (Ben-David et al., 2013; Ben-David et al., 2020).

By choosing to select for a novel enzymatic activity within the native cellular environment, we deliberately set out to explore the additional complexities of metabolic interference, which have potential to play dominant roles in shaping natural evolutionary outcomes. Copley, 2020 recently described an equation that succinctly summarises how the rate of a promiscuous reaction in the presence of a native substrate might be improved by (1) increasing the concentration of the enzyme; (2) increasing the ratio of promiscuous to native substrate; and/or (3) altering the active site to diminish substrate competition, by enhancing binding or turnover of the promiscuous substrate, or decreasing binding of the native substrate. In a landmark 2015 study she and co-workers experimentally demonstrated the importance of diminished substrate competition, focusing on a single key Glu to Ala substitution that enabled several orthologs of ProA (L-gamma-glutamyl phosphate reductase, a key enzyme in proline synthesis) to replace E. coli ArgC (an N-acetyl glutamyl phosphate reductase required for arginine synthesis) (Khanal et al., 2015). Where measurable, all of the substituted variants showed decreased affinity (increased K_M) for the native substrate; and in all but one case there was substantial improvement in k_cat/K_M for the promiscuous substrate as well (Khanal et al., 2015). These findings are similar to our observation that elimination of quinone reductase activity from NfsA via substitution of R225 provided a platform for successive improvements in chloramphenicol affinity to amplify host cell resistance. Together, these examples support the proposal of Kaltenbach et al., 2016 that during natural evolution of a promiscuous activity there is likely to be active selection against the original function, as well as our own supposition that most previous laboratory evolution studies have evaded this phenomenon by focusing on heterologous enzymes and/or exogenously applied substrates. Moreover, our observations that the unselected promiscuous activities of NfsA (reduction of a structurally diverse panel of prodrugs) were mostly unaffected is consistent with their central thesis, that positive selection alone does not lead to specialisation. The emerging picture is that evolution in the natural intracellular milieu involves both selection for the new function, and selection against the old.

An interesting difference between our scenario and that of the Copley team is that NfsA is less essential to the fitness of its host cell than ProA (e.g. deletion of nfsA does not impair E. coli growth even under oxidative stress from heavy metal challenge [Ackerley et al., 2004]). This means that when a new stress is encountered and a promiscuous function becomes essential, as we have modelled here, the enzyme can potentially evolve without necessitating gene duplication to preserve the original function. An apparent ‘freedom to operate’ is manifest in the vast diversity of primary functionalities observed in the superfamily of nitroreductases that NfsA belongs to (which spans activities as divergent as quinone reduction, flavin reduction to power bioluminescence, flavin fragmentation, dehalogenation and dehydrogenation [Akiva et al., 2017]). Although this contrasts with the prevailing Innovation–Amplification–Divergence (IAD) model for natural enzyme evolution (Bergthorsson et al., 2007), it may not be an exceptional scenario – rather, as previously argued by Newton et al., 2015 it is likely that only a minority of enzymes in a cell are under active selection for improved catalytic activity at any time, and redundancy in metabolic networks means that there is latent evolutionary potential that can be immediately tapped to adapt to stress without the requirement of rare and costly gene duplication events. That a single mutation may suffice to rapidly amplify a desirable promiscuous activity simply by eliminating native substrate competition confers substantial ‘robustness’ at a cellular level, even if it means that individual enzymes may not be as robust as previously considered.

All data generated or analysed during this study are included in the manuscript and supporting files.

1. 1. Ackerley DF
   2. Gonzalez CF
   3. Keyhan M
   4. Blake R
   5. Matin A

   (2004) Mechanism of chromate reduction by the Escherichia coli protein, NfsA, and the role of different chromate reductases in minimizing oxidative stress during chromate reduction

   Environmental Microbiology 6:851–860.

   https://doi.org/10.1111/j.1462-2920.2004.00639.x

   • PubMed
   • Google Scholar
2. 1. Aharoni A
   2. Gaidukov L
   3. Khersonsky O
   4. McQ Gould S
   5. Roodveldt C
   6. Tawfik DS

   (2005) The 'evolvability' of promiscuous protein functions

   Nature Genetics 37:73–76.

   https://doi.org/10.1038/ng1482

   • PubMed
   • Google Scholar
3. 1. Akiva E
   2. Copp JN
   3. Tokuriki N
   4. Babbitt PC

   (2017) Evolutionary and molecular foundations of multiple contemporary functions of the nitroreductase superfamily

   PNAS 114:E9549–E9558.

   https://doi.org/10.1073/pnas.1706849114

   • PubMed
   • Google Scholar
4. 1. Ben-David M
   2. Wieczorek G
   3. Elias M
   4. Silman I
   5. Sussman JL
   6. Tawfik DS

   (2013) Catalytic metal ion rearrangements underline promiscuity and evolvability of a metalloenzyme

   Journal of Molecular Biology 425:1028–1038.

   https://doi.org/10.1016/j.jmb.2013.01.009

   • PubMed
   • Google Scholar
5. 1. Ben-David M
   2. Soskine M
   3. Dubovetskyi A
   4. Cherukuri KP
   5. Dym O
   6. Sussman JL
   7. Liao Q
   8. Szeler K
   9. Kamerlin SCL
   10. Tawfik DS

   (2020) Enzyme evolution: an epistatic ratchet versus a smooth reversible transition

   Molecular Biology and Evolution 37:1133–1147.

   https://doi.org/10.1093/molbev/msz298

   • PubMed
   • Google Scholar
6. 1. Bergthorsson U
   2. Andersson DI
   3. Roth JR

   (2007) Ohno's dilemma: evolution of new genes under continuous selection

   PNAS 104:17004–17009.

   https://doi.org/10.1073/pnas.0707158104

   • PubMed
   • Google Scholar
7. Software

   1. Calcott MJ

   (2020) Analyse_epistatic_interactions, version swh:1:rev:62624a66324e230bf273b17591a537c691fa316f

   Software Heritage.

   https://archive.softwareheritage.org/swh:1:dir:b7eb8d2c7f4db9ef5198ef99e80cb5ada45ecf42/
8. 1. Chan-Hyams JVE
   2. Copp JN
   3. Smaill JB
   4. Patterson AV
   5. Ackerley DF

   (2018) Evaluating the abilities of diverse nitroaromatic prodrug metabolites to exit a model gram negative vector for bacterial-directed enzyme-prodrug therapy

   Biochemical Pharmacology 158:192–200.

   https://doi.org/10.1016/j.bcp.2018.10.020

   • PubMed
   • Google Scholar
9. 1. Copley SD

   (2009) Evolution of efficient pathways for degradation of anthropogenic chemicals

   Nature Chemical Biology 5:559–566.

   https://doi.org/10.1038/nchembio.197

   • PubMed
   • Google Scholar
10. 1. Copley SD

    (2015) An evolutionary biochemist's perspective on promiscuity

    Trends in Biochemical Sciences 40:72–78.

    https://doi.org/10.1016/j.tibs.2014.12.004

    • PubMed
    • Google Scholar
11. 1. Copley SD

    (2020) Evolution of new enzymes by gene duplication and divergence

    The FEBS Journal 287:1262–1283.

    https://doi.org/10.1111/febs.15299

    • PubMed
    • Google Scholar
12. 1. Copp JN
    2. Williams EM
    3. Rich MH
    4. Patterson AV
    5. Smaill JB
    6. Ackerley DF

    (2014a) Toward a high-throughput screening platform for directed evolution of enzymes that activate genotoxic prodrugs

    Protein Engineering, Design and Selection 27:399–403.

    https://doi.org/10.1093/protein/gzu025

    • Google Scholar
13. 1. Copp JN
    2. Hanson-Manful P
    3. Ackerley DF
    4. Patrick WM

    (2014b) Error-prone PCR and effective generation of gene variant libraries for directed evolution

    Methods in Molecular Biology 1179:3–22.

    https://doi.org/10.1007/978-1-4939-1053-3_1

    • PubMed
    • Google Scholar
14. 1. Copp JN
    2. Mowday AM
    3. Williams EM
    4. Guise CP
    5. Ashoorzadeh A
    6. Sharrock AV
    7. Flanagan JU
    8. Smaill JB
    9. Patterson AV
    10. Ackerley DF

    (2017) Engineering a multifunctional nitroreductase for improved activation of prodrugs and PET probes for Cancer gene therapy

    Cell Chemical Biology 24:391–403.

    https://doi.org/10.1016/j.chembiol.2017.02.005

    • PubMed
    • Google Scholar
15. 1. Copp JN
    2. Pletzer D
    3. Brown AS
    4. Van der Heijden J
    5. Miton CM
    6. Edgar RJ
    7. Rich MH
    8. Little RF
    9. Williams EM
    10. Hancock REW
    11. Tokuriki N
    12. Ackerley DF

    (2020) Mechanistic understanding enables the rational design of salicylanilide combination therapies for Gram-Negative infections

    mBio 11:e02068.

    https://doi.org/10.1128/mBio.02068-20

    • PubMed
    • Google Scholar
16. 1. Crofts TS
    2. Sontha P
    3. King AO
    4. Wang B
    5. Biddy BA
    6. Zanolli N
    7. Gaumnitz J
    8. Dantas G

    (2019) Discovery and characterization of a nitroreductase capable of conferring bacterial resistance to chloramphenicol

    Cell Chemical Biology 26:559–570.

    https://doi.org/10.1016/j.chembiol.2019.01.007

    • PubMed
    • Google Scholar
17. Book

    1. Day MA

    (2013)

    Structural Comparisons of the E. coli Prodrug-Activating Nitroreductase Enzymes, NfsA and NfsB

    Birmingham, UK: University of Birmingham.

    • Google Scholar
18. 1. Erwin AL
    2. Kline T
    3. Unrath WCT
    4. Nelson K
    5. Weber A
    6. Howald WN

    (2007) Chloramphenicol is a substrate for a novel nitroreductase pathway in Haemophilus influenzae

    Antimicrobial Agents and Chemotherapy 51:2820–2829.

    https://doi.org/10.1128/AAC.00087-07

    • Google Scholar
19. 1. Firth AE
    2. Patrick WM

    (2008) GLUE-IT and PEDEL-AA: new programmes for analyzing protein diversity in randomized libraries

    Nucleic Acids Research 36:W281–W285.

    https://doi.org/10.1093/nar/gkn226

    • PubMed
    • Google Scholar
20. 1. Hall BG

    (2004) Predicting the evolution of antibiotic resistance genes

    Nature Reviews Microbiology 2:430–435.

    https://doi.org/10.1038/nrmicro888

    • PubMed
    • Google Scholar
21. 1. Hermes JD
    2. Blacklow SC
    3. Knowles JR

    (1990) Searching sequence space by definably random mutagenesis: improving the catalytic potency of an enzyme

    PNAS 87:696–700.

    https://doi.org/10.1073/pnas.87.2.696

    • PubMed
    • Google Scholar
22. 1. Kaltenbach M

    (2014) Dynamics and constraints of enzyme evolution

    Journal of Experimental Zoology Part B: Molecular and Developmental Evolution 322:468–487.

    https://doi.org/10.1002/jez.b.22562

    • Google Scholar
23. 1. Kaltenbach M
    2. Jackson CJ
    3. Campbell EC
    4. Hollfelder F
    5. Tokuriki N

    (2015) Reverse evolution leads to genotypic incompatibility despite functional and active site convergence

    eLife 4:e06492.

    https://doi.org/10.7554/eLife.06492

    • Google Scholar
24. 1. Kaltenbach M
    2. Emond S
    3. Hollfelder F
    4. Tokuriki N

    (2016) Functional Trade-Offs in promiscuous enzymes Cannot be explained by intrinsic mutational robustness of the native activity

    PLOS Genetics 12:e1006305.

    https://doi.org/10.1371/journal.pgen.1006305

    • PubMed
    • Google Scholar
25. 1. Khanal A
    2. Yu McLoughlin S
    3. Kershner JP
    4. Copley SD

    (2015) Differential effects of a mutation on the normal and promiscuous activities of orthologs: implications for natural and directed evolution

    Molecular Biology and Evolution 32:100–108.

    https://doi.org/10.1093/molbev/msu271

    • PubMed
    • Google Scholar
26. 1. Khersonsky O
    2. Tawfik DS

    (2010) Enzyme promiscuity: a mechanistic and evolutionary perspective

    Annual Review of Biochemistry 79:471–505.

    https://doi.org/10.1146/annurev-biochem-030409-143718

    • PubMed
    • Google Scholar
27. 1. Kobori T
    2. Sasaki H
    3. Lee WC
    4. Zenno S
    5. Saigo K
    6. Murphy ME
    7. Tanokura M

    (2001) Structure and site-directed mutagenesis of a flavoprotein from Escherichia coli that reduces nitrocompounds: alteration of pyridine nucleotide binding by a single amino acid substitution

    The Journal of Biological Chemistry 276:2816–2823.

    https://doi.org/10.1074/jbc.M002617200

    • PubMed
    • Google Scholar
28. 1. Liochev SI
    2. Hausladen A
    3. Fridovich I

    (1999) Nitroreductase A is regulated as a member of the soxRS regulon of Escherichia coli

    PNAS 96:3537–3539.

    https://doi.org/10.1073/pnas.96.7.3537

    • Google Scholar
29. 1. Newton MS
    2. Arcus VL
    3. Patrick WM

    (2015) Rapid bursts and slow declines: on the possible evolutionary trajectories of enzymes

    Journal of the Royal Society Interface 12:20150036.

    https://doi.org/10.1098/rsif.2015.0036

    • Google Scholar
30. 1. O'Brien PJ
    2. Herschlag D

    (1999) Catalytic promiscuity and the evolution of new enzymatic activities

    Chemistry & Biology 6:R91–R105.

    https://doi.org/10.1016/S1074-5521(99)80033-7

    • PubMed
    • Google Scholar
31. 1. Parry R
    2. Nishino S
    3. Spain J

    (2011) Naturally-occurring nitro compounds

    Nat. Prod. Rep. 28:152–167.

    https://doi.org/10.1039/C0NP00024H

    • PubMed
    • Google Scholar
32. 1. Paterson ES
    2. Boucher SE
    3. Lambert IB

    (2002) Regulation of the nfsA gene in Escherichia coli by SoxS

    Journal of Bacteriology 184:51–58.

    https://doi.org/10.1128/JB.184.1.51-58.2002

    • PubMed
    • Google Scholar
33. 1. Poelwijk FJ
    2. Tănase-Nicola S
    3. Kiviet DJ
    4. Tans SJ

    (2011) Reciprocal sign epistasis is a necessary condition for multi-peaked fitness landscapes

    Journal of Theoretical Biology 272:141–144.

    https://doi.org/10.1016/j.jtbi.2010.12.015

    • PubMed
    • Google Scholar
34. 1. Prosser GA
    2. Copp JN
    3. Mowday AM
    4. Guise CP
    5. Syddall SP
    6. Williams EM
    7. Horvat CN
    8. Swe PM
    9. Ashoorzadeh A
    10. Denny WA
    11. Smaill JB
    12. Patterson AV
    13. Ackerley DF

    (2013) Creation and screening of a multi-family bacterial oxidoreductase library to discover novel nitroreductases that efficiently activate the bioreductive prodrugs CB1954 and PR-104A

    Biochemical Pharmacology 85:1091–1103.

    https://doi.org/10.1016/j.bcp.2013.01.029

    • PubMed
    • Google Scholar
35. 1. Ramos JL
    2. González-Pérez MM
    3. Caballero A
    4. van Dillewijn P

    (2005) Bioremediation of polynitrated aromatic compounds: plants and microbes put up a fight

    Current Opinion in Biotechnology 16:275–281.

    https://doi.org/10.1016/j.copbio.2005.03.010

    • PubMed
    • Google Scholar
36. 1. Reetz MT
    2. Kahakeaw D
    3. Lohmer R

    (2008) Addressing the numbers problem in directed evolution

    Chembiochem : A European Journal of Chemical Biology 9:1797–1804.

    https://doi.org/10.1002/cbic.200800298

    • PubMed
    • Google Scholar
37. Book

    1. Rich MH

    (2017)

    Discovery and Directed Evolution of Nitroreductase Enzymes for Activation of Prodrufs and PET Imaging Compounds

    Victoria University of Wellington, Wellington, New Zealand.

    • Google Scholar
38. Book

    1. Sambrook JF
    2. Russell DW

    (2001)

    Molecular Cloning: A Laboratory Manual

    Cold Spring Harbor Laboratory Press .

    • Google Scholar
39. 1. Stibitz S

    (1994) Use of conditionally counterselectable suicide vectors for allelic exchange

    Methods in Enzymology 235:458–465.

    https://doi.org/10.1016/0076-6879(94)35161-9

    • PubMed
    • Google Scholar
40. 1. Valiauga B
    2. Williams EM
    3. Ackerley DF
    4. Čėnas N

    (2017) Reduction of quinones and nitroaromatic compounds by Escherichia coli nitroreductase A (NfsA): Characterization of kinetics and substrate specificity

    Archives of Biochemistry and Biophysics 614:14–22.

    https://doi.org/10.1016/j.abb.2016.12.005

    • PubMed
    • Google Scholar
41. 1. Weinreich DM
    2. Watson RA
    3. Chao L

    (2005) Perspective: sign epistasis and genetic costraint on evolutionary trajectories

    Evolution 59:1165–1174.

    https://doi.org/10.1111/j.0014-3820.2005.tb01768.x

    • Google Scholar
42. Website

    1. Williams EM

    (2013) Development of bacterial nitroreductase enzymes for noninvasive imaging in Cancer gene therapy. Victoria university of Wellington, Wellington, New Zealand

    Accessed December 7, 2020.

    http://hdl.handle.net/10063/4994
43. 1. Williams EM
    2. Little RF
    3. Mowday AM
    4. Rich MH
    5. Chan-Hyams JV
    6. Copp JN
    7. Smaill JB
    8. Patterson AV
    9. Ackerley DF

    (2015) Nitroreductase gene-directed enzyme prodrug therapy: insights and advances toward clinical utility

    Biochemical Journal 471:131–153.

    https://doi.org/10.1042/BJ20150650

    • PubMed
    • Google Scholar
44. 1. Winkler R
    2. Hertweck C

    (2007) Biosynthesis of nitro compounds

    ChemBioChem 8:973–977.

    https://doi.org/10.1002/cbic.200700042

    • PubMed
    • Google Scholar
45. 1. Yang G
    2. Anderson DW
    3. Baier F
    4. Dohmen E
    5. Hong N
    6. Carr PD
    7. Kamerlin SCL
    8. Jackson CJ
    9. Bornberg-Bauer E
    10. Tokuriki N

    (2019) Higher-order epistasis shapes the fitness landscape of a xenobiotic-degrading enzyme

    Nature Chemical Biology 15:1120–1128.

    https://doi.org/10.1038/s41589-019-0386-3

    • PubMed
    • Google Scholar
46. 1. Yunis AA

    (1988) Chloramphenicol: relation of structure to activity and toxicity

    Annual Review of Pharmacology and Toxicology 28:83–100.

    https://doi.org/10.1146/annurev.pa.28.040188.000503

    • PubMed
    • Google Scholar
47. 1. Zhu H
    2. Dean RA

    (1999) A novel method for increasing the transformation efficiency of Escherichia coli-application forbacterial artificial chromosome library construction

    Nucleic Acids Research 27:910–911.

    https://doi.org/10.1093/nar/27.3.910

    • PubMed
    • Google Scholar

Acceptance summary:

The potential for endogenous native substrates to interfere with evolution of the new function has so far been mostly a matter of speculation and your paper now shows conclusively that in the quinone reductase NfsA, a promiscuous chloramphenicol-detoxifying nitroreductase activity can only be amplified if a key first mutation eliminates the native activity of NfsA, after which additional substitutions incrementally improve the affinity for chloramphenicol as a substrate. This example substantiates a model for enzyme evolution, in which a single mutation can open the path to the rapid amplification of a desirable promiscuous activity simply by eliminating native substrate competition.

Decision letter after peer review:

Thank you for submitting your article "A giant leap in sequence space reveals the intracellular complexities of evolving a new function" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by Andrei Lupas as the Reviewing Editor and Patricia Wittkopp as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

As the editors have judged that your manuscript is of interest, but as described below that additional experiments are required before it is published, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

Summary:

This manuscript describes a laboratory evolution experiment designed to explore effects that may shape evolutionary trajectories in a native host environment. The model system is E. coli nitro/quinone reductase NfsA, a promiscuous FMN-dependent oxidoreductase that reduces toxic compounds and has the basal ability to reduce the antibiotic chloramphenicol. This function was used to select for improved detoxification by mass-mutagenizing eight active-site residues and isolating variants with up to 10-fold higher tolerance against chloramphenicol. The five best variant proteins were purified and characterized, showing that their kcat/Km was only marginally improved, with worse kcat but improved Km, indicating that the improvements in detoxification were driven by enhanced substrate affinity. For the top two variants, all possible evolutionary trajectories were recreated and their EC₅₀'s tested to determine the most likely possible step-wise paths from NfsA to the final variants. The authors found that iterative evolutionary strategies could have generated similar variants, but that there were only few accessible pathways, indicating epistatic effects. The analysis also showed that for both variants, elimination of arginine at position 225 in the first step enabled further improvements to take hold and played a role in the loss of wildtype 1,4-benzoquinone activity. The sensitivity to four out of five tested prodrugs was however unchanged. Turnover of the fifth prodrug, namely reduction of metronidazole, which yields a toxic product, was on the other hand increased in the evolved variants, and could be used as a counter-selectable marker. This was briefly tested showing the potential of such an application.

Essential revisions:

This study presents a wealth of data, and is well reasoned, carefully executed and clearly laid out. However, although it states that its aim was to study the evolution of a promiscuous function within the native host environment and thus under metabolic interference of the native substrate, this was not the approach taken. Instead, a fitness peak for the promiscuous function was identified through mass mutagenesis at eight positions followed by selection, and then two potential evolutionary paths leading from the wild type to this peak were inferred based on an analysis of all possible mutant combinations at the mutagenized positions. The authors need to make clear throughout the paper that the variants able to detoxify chloramphenicol were not evolved and did not arise against metabolic interference of the native substrate. This is an important point as the considerable potential of endogenous metabolites to shape evolutionary outcomes (Abstract) is purely inferred from the observation that the first mutation in both reconstructed evolutionary paths appears to have been a mutation at R225, which led to a substantial drop in the turnover rate of the endogenous substrate. From this the authors conclude (very prominently throughout the paper) that the evolution of a new activity is only possible after loss of activity against the original substrate. From the data presented, it is however not clear to what extent this conclusion is supported.

1) According to the data in Figure 4 and Supplementary file 1B, mutation of R225 alone is accompanied by a ~2-fold increase in kcat/Km for chloramphenicol. This seems to be sufficient to explain the ~2-fold increase in EC₅₀ for chloramphenicol without invoking loss of quinone reductase activity. The control experiment in Figure 5, showing that substitution of R225 has no effect on most promiscuous activities of NfsA, also seems to indicate that the loss of native activity is not required for the evolution of chloramphenicol resistance. It would be important to determine the kinetic parameters of 1,4-benzoquinone reduction for NfsA and the purified R225V and R225D mutants in order to establish the loss of quinone reductase activity in the postulated first step of the evolutionary path. It would also be useful to study the effect of 1,4-benzoquinone competition on the chloramphenicol reductase activity of the mutants, at least the first ones along the proposed path, in order to show that they rapidly become insensitive to the native substrate.

2) After the initial screening of the transformed library of NfsA variants, 0.05% of gene variants are reported to be more effective in chloramphenicol detoxification than the wild type. In the next steps, this number is reduced to the top 30 variants, as characterized by their improvement of chloramphenicol EC₅₀ values (Figure 1D). However, it is not clear from the presentation whether these observations were controlled for the expression levels of the different NfsA mutants. Protein variants are often expressed at different levels in vivo, which can have a significant effect on the activity measured. Figure 1D was used for selection of the "best" variants for the rest of the study and to support this choice and the conclusions of the manuscript, relative enzyme expression levels should be reported (and if significantly different, should be corrected for). Such expression levels are reported later on for the 36_37 and 20_39 variants, but are missing at this early stage.

3) While mutation of R225 appeared to be required for improved chloramphenicol detoxification in this study, the authors only considered the effects of substitutions at eight positions. This is probably the main weakness of the combinatorial mutagenesis approach used here. It seems plausible that substitutions at other positions could also increase chloramphenicol tolerance, possibly opening a path without loss of quinone reductase activity. If the authors were able to perform one round of error-prone PCR on NfsA with selection for improved chloramphenicol resistance and obtain mainly variants with substitution of R225, this would substantially strengthen their claim that evolution of increased chloramphenicol resistance can only occur through loss of quinone reductase activity.

4) Even with additional experimental support for the main conclusion of the article, it seems fundamentally problematic to extrapolate from two instances to a general principle of evolution. The authors should tone done the claims that improved chloramphenicol detox activity is ONLY possible after elimination the native activity and instead comment on the two characterized mutant pathways as examples of this phenomenon, within the limitations of the experimental setup.

Essential revisions:

This study presents a wealth of data, and is well reasoned, carefully executed and clearly laid out. However, although it states that its aim was to study the evolution of a promiscuous function within the native host environment and thus under metabolic interference of the native substrate, this was not the approach taken. Instead, a fitness peak for the promiscuous function was identified through mass mutagenesis at eight positions followed by selection, and then two potential evolutionary paths leading from the wild type to this peak were inferred based on an analysis of all possible mutant combinations at the mutagenized positions. The authors need to make clear throughout the paper that the variants able to detoxify chloramphenicol were not evolved and did not arise against metabolic interference of the native substrate. This is an important point as the considerable potential of endogenous metabolites to shape evolutionary outcomes (Abstract) is purely inferred from the observation that the first mutation in both reconstructed evolutionary paths appears to have been a mutation at R225, which led to a substantial drop in the turnover rate of the endogenous substrate. From this the authors conclude (very prominently throughout the paper) that the evolution of a new activity is only possible after loss of activity against the original substrate. From the data presented, it is however not clear to what extent this conclusion is supported.

Thank you for raising this point. We recognise the distinction between our mass mutagenesis strategy and stepwise evolution, and have edited our manuscript to eliminate any unjustified claims regarding evolution of our variants (most commonly, we added the word “hypothetical” in front of “evolutionary trajectory” , and substituted the word “evolved” with “selected” where the former was not justified). We do however disagree with the statement that our variants “did not arise against metabolic interference of the native substrate”. A key aspect of our work that is different from many previous directed evolution studies is that we selected our improved variants within their native host environment, where they were very specifically exposed to their native substrates. We consider that this clearly shaped the outcomes.

With credit to the excellent suggestion from the reviewers that we perform epPCR on the native nfsA gene to see whether substitution of R225 predominates, we note that we now have much stronger evidence to support our conclusions around the need to eliminate the native activity to enable improved chloramphenicol detoxification (see response to reviewers’ point 3 below).

1) According to the data in Figure 4 and Supplementary file 1B, mutation of R225 alone is accompanied by a ~2-fold increase in kcat/Km for chloramphenicol. This seems to be sufficient to explain the ~2-fold increase in EC₅₀ for chloramphenicol without invoking loss of quinone reductase activity. The control experiment in Figure 5, showing that substitution of R225 has no effect on most promiscuous activities of NfsA, also seems to indicate that the loss of native activity is not required for the evolution of chloramphenicol resistance. It would be important to determine the kinetic parameters of 1,4-benzoquinone reduction for NfsA and the purified R225V and R225D mutants in order to establish the loss of quinone reductase activity in the postulated first step of the evolutionary path. It would also be useful to study the effect of 1,4-benzoquinone competition on the chloramphenicol reductase activity of the mutants, at least the first ones along the proposed path, in order to show that they rapidly become insensitive to the native substrate.

There are two suggestions here, which will be addressed in turn.

i) We had in fact reported our efforts to obtain 1,4-benzoquinone reduction kinetics for the purified R225V and R225D variants in addition to wild type NfsA, however these data were evidently buried a bit too deeply in the legend to Supplementary file 1B for them to be readily accessed by a reader. In brief, it was not possible for us to derive meaningful 1,4-benzoquinone kinetic parameters for the two variants (both R225V and R225D exhibit a just-detectable level of 1,4-benzoquinone activity, as was indicated in the “Rate with 100 µM p-benzoquinone” column of Supplementary file 1B and in Figures 4A and B; but the activity was essentially undetectable at lower concentrations). This is the basis for our conclusion that quinone reductase activity has indeed been lost in the postulated first step of the evolutionary path. To more overtly justify this conclusion to the reader, we have now referred to these data explicitly in the main text, as detailed in our response to the next point.

ii) We agree that it would be ideal to study the effect of 1,4-benzoquinone competition on the chloramphenicol reductase activity of the mutants, but unfortunately both chloramphenicol and 1,4-benzoquinone reduction assays rely on monitoring NADPH depletion at 340 nm, so it is not possible to deconvolute the relative turnover of each substrate by spectrophotometric assays in a mixed solution (in our previous study, we were able to independently monitor PR-104A reduction at 400 nm). However, it is important to note that in our previous study the top selected variant was still partially active with 1,4-benzoquinone in vitro, making it plausible that substrate competition could occur. Here we show that 1,4-benzoquinone activity had effectively been eliminated from our R225-substituted variants, and we therefore consider it reasonable to presume that 1,4-benzoquinone will no longer exhibit substrate competition with chloramphenicol. We have now clarified these points in the text, with the last two sentences of the relevant paragraph now reading:

“Here, we were unable to perform the same in vitro competition assays, as both 1,4-benzoquinone and chloramphenicol reduction are monitored by following NADPH depletion at 340 nm. […] In contrast, for wild type NfsA, 1,4-benzoquinone reduction (k_cat/K_M = 5.8 × 10⁶; Valiauga et al., 2017) is nearly 1000-fold more efficient than chloramphenicol reduction (Table 1). Our data are therefore consistent with in vivo chloramphenicol reduction having been amplified for the selected variants by the elimination of competitive quinone inhibition.”

We also note that the additional epPCR experiments (see point 3 below) strongly reinforce that evolving improved chloramphenicol detoxification necessitates elimination of R225, which is consistent with 1,4-benzoquinone outcompeting chloramphenicol as a substrate for wild type NfsA.

2) After the initial screening of the transformed library of NfsA variants, 0.05% of gene variants are reported to be more effective in chloramphenicol detoxification than the wild type. In the next steps, this number is reduced to the top 30 variants, as characterized by their improvement of chloramphenicol EC₅₀ values (Figure 1D). However, it is not clear from the presentation whether these observations were controlled for the expression levels of the different NfsA mutants. Protein variants are often expressed at different levels in vivo, which can have a significant effect on the activity measured. Figure 1D was used for selection of the "best" variants for the rest of the study and to support this choice and the conclusions of the manuscript, relative enzyme expression levels should be reported (and if significantly different, should be corrected for). Such expression levels are reported later on for the 36_37 and 20_39 variants, but are missing at this early stage.

We agree that this is an important variable that merits consideration. We have now performed SDS-PAGE analysis for the E. coli strains expressing each of the top 30 variants alongside wild type NfsA controls, and can confirm that there is no substantial variation in expression between these variants and wild type NfsA. We have provided these data as a supplementary figure (new Figure 1—figure supplement 1), and have added the following sentence to the main text: “SDS-PAGE analysis revealed that there was no substantial variation in expression between these variants and native nfsA, and hence the enhanced chloramphenicol detoxification was not a consequence of elevated expression levels Figure 1—figure supplement 1).”

3) While mutation of R225 appeared to be required for improved chloramphenicol detoxification in this study, the authors only considered the effects of substitutions at eight positions. This is probably the main weakness of the combinatorial mutagenesis approach used here. It seems plausible that substitutions at other positions could also increase chloramphenicol tolerance, possibly opening a path without loss of quinone reductase activity. If the authors were able to perform one round of error-prone PCR on NfsA with selection for improved chloramphenicol resistance and obtain mainly variants with substitution of R225, this would substantially strengthen their claim that evolution of increased chloramphenicol resistance can only occur through loss of quinone reductase activity.

This was a really fantastic suggestion, and we kicked ourselves for not having previously thought to run this experiment. Our previous conclusions regarding the need to eliminate the native quinone reductase activity have now received a compelling corroboration by our discovery that 100% of unique selected variants (50/50) after one round of epPCR on native nfsA were carrying an R225 substitution. We have added an additional section of text entitled “Directed evolution of NfsA via epPCR yields exclusively R225-substituted variants” and two additional supplementary tables (Supplementary file 1E and F) and a new supplementary figure to describe these findings; and have reworked our Abstract and Discussion slightly to recognise them.

4) Even with additional experimental support for the main conclusion of the article, it seems fundamentally problematic to extrapolate from two instances to a general principle of evolution. The authors should tone done the claims that improved chloramphenicol detox activity is ONLY possible after elimination the native activity and instead comment on the two characterized mutant pathways as examples of this phenomenon, within the limitations of the experimental setup.

Thanks to the excellent suggestion above (major point 3), we now have far stronger evidence to support our primary conclusion. Nevertheless, we have tried to tone down our previous claims, as detailed using track changes throughout the manuscript.

Author details

1. Kelsi R Hall

   1. School of Biological Sciences, Victoria University of Wellington, Wellington, Wellington, New Zealand
   2. Centre for Biodiscovery, Victoria University of Wellington, Wellington, New Zealand

   Contribution

   Conceptualization, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing

   Competing interests

   No competing interests declared

2. Katherine J Robins

   School of Biological Sciences, Victoria University of Wellington, Wellington, Wellington, New Zealand

   Present address

   The Manchester Institute of Biotechnology, The University of Manchester, Manchester, United Kingdom

   Contribution

   Conceptualization, Data curation, Formal analysis, Investigation, Methodology

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-5049-4246

3. Elsie M Williams

   1. School of Biological Sciences, Victoria University of Wellington, Wellington, Wellington, New Zealand
   2. Centre for Biodiscovery, Victoria University of Wellington, Wellington, New Zealand

   Contribution

   Validation, Methodology

   Competing interests

   No competing interests declared

4. Michelle H Rich

   School of Biological Sciences, Victoria University of Wellington, Wellington, Wellington, New Zealand

   Present address

   BiOrbic, SFI Bioeconomy Research Centre, University College Dublin, Belfield, Ireland

   Contribution

   Validation, Investigation, Methodology

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-4876-4029

5. Mark J Calcott

   1. School of Biological Sciences, Victoria University of Wellington, Wellington, Wellington, New Zealand
   2. Centre for Biodiscovery, Victoria University of Wellington, Wellington, New Zealand

   Contribution

   Data curation, Software

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-7736-8095

6. Janine N Copp

   School of Biological Sciences, Victoria University of Wellington, Wellington, Wellington, New Zealand

   Present address

   Michael Smith Laboratories, University of British Columbia, Vancouver, Canada

   Contribution

   Validation, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-6690-0480

7. Rory F Little

   School of Biological Sciences, Victoria University of Wellington, Wellington, Wellington, New Zealand

   Present address

   Leibniz Institute for Natural Product Research and Infection Biology, Hans Knöll Institute, Jena, Germany

   Contribution

   Validation, Methodology

   Competing interests

   No competing interests declared

8. Ralf Schwörer

   1. Centre for Biodiscovery, Victoria University of Wellington, Wellington, New Zealand
   2. Ferrier Institute, Victoria University of Wellington, Wellington, New Zealand

   Contribution

   Resources

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-9352-6559

9. Gary B Evans

   1. Centre for Biodiscovery, Victoria University of Wellington, Wellington, New Zealand
   2. Ferrier Institute, Victoria University of Wellington, Wellington, New Zealand

   Contribution

   Resources, Supervision

   Competing interests

   No competing interests declared

10. Wayne M Patrick

    1. School of Biological Sciences, Victoria University of Wellington, Wellington, Wellington, New Zealand
    2. Centre for Biodiscovery, Victoria University of Wellington, Wellington, New Zealand

    Contribution

    Conceptualization, Resources, Supervision, Funding acquisition, Writing - review and editing

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0002-2718-8053

11. David F Ackerley

    1. School of Biological Sciences, Victoria University of Wellington, Wellington, Wellington, New Zealand
    2. Centre for Biodiscovery, Victoria University of Wellington, Wellington, New Zealand

    Contribution

    Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Writing - original draft, Project administration, Writing - review and editing

    For correspondence

    david.ackerley@vuw.ac.nz

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0002-6188-9902

• 909

  Page views

• 115

  Downloads

• 3

  Citations

Article citation count generated by polling the highest count across the following sources: PubMed Central, Crossref, Scopus.