All life on our planet requires ribonucleotide reduction for de novo synthesis of deoxyribonucleotides, from ribonucleotides (Nordlund and Reichard, 2006; Lundin et al., 2009). The emergence of ribonucleotide reduction likely drove the transition from RNA to DNA early in the evolution of life (Lundin et al., 2015; Reichard, 1993; Torrents et al., 2002; Poole et al., 2002) and its centrality is reflected in its near ubiquity in organismal genomes (Lundin et al., 2009). Structural analyses indicate that the catalytic cores of ribonucleotide reductases (RNRs) are homologous (Lundin et al., 2015; Sintchak et al., 2002), despite sequence similarities between them being extremely limited (Tauer and Benner, 1997). RNRs catalyse deoxyribonucleotide synthesis via a common mechanism wherein a cysteinyl-free radical is generated in the active site. While this similarity in catalysis underscores a common evolutionary origin, RNRs have diverse mechanisms for radical generation (Nordlund and Reichard, 2006). Three broad, evolutionarily distinct, families exist which are also divided on the basis of their radical generation mechanism (Nordlund and Reichard, 2006; Lundin et al., 2015; Lundin et al., 2010). This divergence likely coincides with adaptation to varied environments (Lundin et al., 2010; Hofer et al., 2012). Class I enzymes are strictly aerobic in that they require oxygen or superoxide for activation (Rose et al., 2018). This typically, though not exclusively (Blaesi et al., 2018; Srinivas et al., 2018), occurs through oxidation of a dimetal ion centre from which a stable cysteine radical is generated, either directly, or indirectly through formation of a tyrosyl radical in the small subunit. Class III enzymes generate a stable glycine radical via cleavage of S-adenosyl methionine, and are strictly anaerobic, while class II enzymes generate a 5'-deoxyadenosyl radical through cleavage of adenosylcobalamin and operate irrespective of oxygen presence or absence.

To date, very few species have dispensed with ribonucleotide reduction, albeit indirectly, in that they instead rely on their hosts for deoxyribonucleotides (Lundin et al., 2009). Examples include a strain of Buchnera aphidicola str. Cc from the Cedar bark aphid, Cinara cedri (Pérez-Brocal et al., 2006). B. aphidicola are maternally inherited aphid endosymbionts that synthesise amino acids in short supply in the aphid diet, are obligately intracellular, and appear to be on the path to transitioning to an organelle (Andersson, 2000; Tamas et al., 2002). Two bacterial pathogens that live in very close association with their hosts—Ureaplasma urealyticum (Glass et al., 2000) and Borrelia burgdorferi (Fraser et al., 1997)—have also lost genes for ribonucleotide reduction, as have two eukaryotic parasites, Giardia lamblia and Entamoeba histolytica (Lundin et al., 2009).

While ribonucleotide reduction is essential, we reasoned that it should be possible to dispense with genes for ribonucleotide reduction if deoxyribonucleotides or their precursors are available via growth media. We report the successful elimination of all three RNR operons from the bacterium Escherichia coli. The knockout strain was created by introducing a broad-spectrum deoxyribonucleoside kinase from Mycoplasma mycoides (Wang et al., 2001), which we predicted would enable use of media-supplied deoxyribonucleosides for deoxyribonucleotide synthesis in the absence of ribonucleotide reduction. As predicted, the resulting strain is dependent on deoxyribonucleosides in the growth medium but cannot grow if the media are supplemented with deoxyribose (dR) plus the four nucleobases (A, G, C, T). To understand the impact of limited deoxyribonucleoside availability, we subjected our knockout line to experimental evolution followed by genome sequencing. We were interested in establishing whether, under such conditions, our lines would utilise the reverse deoxyriboaldolase (DERA) pathway for deoxyribonucleotide synthesis. This pathway is reversible in vitro (Horinouchi et al., 2006a; Ogawa et al., 2003), and has long been considered a plausible alternative to ribonucleotide reduction (Racker, 1951; Racker, 1952) for dNTP synthesis, with some considering it to be a plausible ancestral route for the origin of DNA (Benner et al., 1989; Poole et al., 2014). Our evolution experiments reveal that this pathway is not coopted for deoxyribonucleotide synthesis following loss of ribonucleotide reduction, with cells grown under limiting deoxyribonucleoside levels exhibiting a filamentous morphology indicative of stress. Instead, it appears that the DERA pathway is a liability under conditions of limited deoxyribonucleotide availability; we observe loss-of-function mutations in deoB, which are predicted to prevent catabolism of 2-deoxy-D-ribose-1-phosphate. Available genome sequence data from obligate intracellular bacteria that lack ribonucleotide reduction indicate that deoB has in fact been lost on multiple occasions, suggesting that recycling of dR is disadvantageous under conditions where this sugar is in limited supply; deletion of deoB would enable this sugar to be rerouted for deoxyribonucleotide synthesis but would also preclude de novo synthesis via the reverse DERA reaction. Our results thus illuminate a key adaptive step taken by obligate intracellular and pathogenic species to mitigate the loss of ribonucleotide reduction.

While all life is dependent on ribonucleotide reduction for the synthesis of deoxyribonucleotides, a handful of species are known to lack RNR genes (Lundin et al., 2009) and presumably depend on their host for a source of dNTPs. In this work, we report the successful deletion of all RNR operons from E. coli (Figure 1). The resulting line (∆RNR) was sequenced to confirm genomic deletion of RNR genes. Our ∆RNR knockout line grows on MOPS minimal media + 1% glucose supplemented with dNs (Figure 2), but does not grow when provided with their constituent components (dR and the four bases A, G, C, T). Growth experiments reveal that, of the four dNs, only dC is essential for growth (Figure 3). Under limiting levels of dNs, our lines exhibit a filamentous phenotype that can be reversed by increasing dN supplementation (Figure 4). Filamentous phenotypes have been documented under a range of stress conditions, including in the presence of β-lactam antibiotics (Stevenson et al., 2016; Miller et al., 2004; Yao et al., 2012). As cell length changes during growth in our experiment, it is not possible to use OD measurements for quantification of either doubling time or cell number (Stevenson et al., 2016). Our experiments should thus be taken as a qualitative indication of growth. Interestingly, we found that this phenotype reduced over the course of the experiment, suggesting that our lines were adapting to the scarcity of dNs. It will be interesting to determine whether this change is associated specifically with disruption of deoB, the effect of which is predicted to make dNs more readily available for DNA synthesis.

In order to understand how obligate intracellular bacteria adapted to the loss of ribonucleotide reduction, we established an evolution experiment where we serially passaged replicate lines in either 1 mg/mL or 0.25 mg/mL dNs. The resulting lines were sequenced after 30 transfers and one line evolved in 0.25 mg/mL dNs (ΔRNR_250_T30_1) was selected for a second round of evolution at 0.01 mg/mL dNs through a further 10 transfers (Figure 5).

Genome sequencing revealed independent mutations to two genes across our replicate lines. One gene, cdd, carried mutations in 8/10 lines by transfer 30. The cdd locus codes for cytidine deaminase, which converts dC to deoxyuridine, which in turn will be available for dTTP production. Our data reveal two independent mutations (a segmental deletion in seven lines, and an inactivating SNP in the eighth). The segmental deletion eliminated part of the cdd ORF and part of ORF of the upstream gene, yohK, which, together with yohJ, has been shown to code a 3-hydroxypropionate transporter (Nguyen-Vo et al., 2020), and was positionally identical across all seven lines. This deletion likely occurred prior to the establishment of our replicate lines, but appeared not to have gone to fixation. An independent ∆RNR line (∆RNR34) that lacked this deletion rapidly lost cdd when passaged in the experimental growth media (Figure 8—figure supplement 3). This, together with the independent truncation mutation in ∆RNR_T30_250_L5, which impacts cdd but not yohK, suggests there is selection against cdd but not necessarily yohK. Given the fact that our ∆RNR lines can grow with dC as sole dN supplement but do not grow if dC is omitted from the dN mix, it may be that eliminating cytidine deamination prevents loss of this key dN through deamination. We have not assessed growth on dU, but one possibility is that deoxyuridine is less amenable as a substrate for production of all four deoxyribonucleotides. One intriguing observation is that, despite the rapid loss of cdd, this loss of function appears not to have occurred across all replicates by transfer 30. This may be a consequence of the length of the evolution experiment. Comparing the genomes of species known to lack genes for ribonucleotide reduction (Lundin et al., 2009) indicates that absence of cdd frequently coincides with this state. Our genome analyses show that all confirmed members of the genus Ureaplasma lack nrd genes and cdd, suggesting that loss of cdd function is selectively advantageous. Establishing whether loss of one gene drives the loss of the other is not straightforward however. An analysis of strains of B. aphidicola revealed that 48/74 strains retain ribonucleotide reduction, but 0/74 possess cdd (Supplementary file 1N), indicating that the latter gene can be lost in lines capable of synthesising their own deoxyribonucleotides. The picture is also not clear in Borrelia, with cdd found in species variously possessing or lacking ribonucleotide reduction (Supplementary file 1M).

Deletion of yohK and cdd results in a small, but intact ORF (Figure 8—figure supplement 2). While this seems unlikely to be functional, the creation of an ORF, particularly where the upstream gene codes for a transmembrane protein, may be consistent with the ‘Car Trunk’ hypothesis for the avoidance of mutations that would create a toxic protein (Omer et al., 2017). The fact that deletion of these genes is reproducible under defined experimental conditions (Figure 8—figure supplement 3) suggests that this region might be suitable for testing this hypothesis.

All lines grown in 0.25 mg/mL dNs showed a marked mutational skew (A:T→G:C) that differs from the G:C→A:T mutational skew expected for E. coli (Lee et al., 2012), and more generally for bacteria (Hershberg and Petrov, 2010). Knockout of cdd does not appear linked to this skew; 4/5 lines grown at 1 mg/mL dNs do not show an obvious A:T→G:C skew (Figure 8). For lines grown at 0.25 mg/mL dNs, two (L2, L3) carry intact cdd genes and both genomes show evidence of A:T→G:C mutational skew. Such skew has been previously shown to be associated with defective mismatch repair (Lee et al., 2012). By transfer 30, we observe only two lines with SNPs in mismatch repair genes, with only one mutation being nonsynonymous (Supplementary file 1). Thus, the skew we observe cannot be explained by widespread mutation to the mismatch repair pathway. The skew may instead be driven in part by the growth dependency on dC (Figure 3D), with the other dNs unable to compensate for lack of dC (Figure 3E) but appearing to provide minor improvements to growth (Figure 2B). It is not currently clear why this should be so. The two major nucleoside transporters in E. coli, NupC and NupG, are able to transport dNs (Patching et al., 2005), so import does not appear to be a barrier.

We also observed loss of function of deoB in our experiments (Table 1). This observation is interesting for several reasons. First, it indicates that loss of ribonucleotide reduction does not lead to compensatory deoxyribonucleotide synthesis via the DERA pathway, which has been speculated to be an ancestral route for dNTP synthesis, predating the advent of ribonucleotide reduction (Benner et al., 1989; Poole et al., 2014). While the enzymes from this pathway can operate biosynthetically (Horinouchi et al., 2006a; Ogawa et al., 2003; Horinouchi et al., 2006b), loss of phosphopentomutase function via mutation of deoB suggests that this pathway is not readily accessible for synthesis of deoxyribonucleotides in vivo, even though there would have been a strong selective advantage to such an alternative synthetic route during our experiment. We note however that both acetaldehyde toxicity (to both DERA and cells more generally) and low intracellular availability may preclude a simple switch to using the DERA pathway for dNTP production (Ogawa et al., 2003; Poole et al., 2014).

Indeed, under the conditions of our experiment, where dN supplementation is reduced by 25-fold, we observe strong selection to dispense with deoB. This suggests that not only is the reverse DERA pathway inaccessible for biosynthesis, but salvage is selected against when dNTPs are scarce. In the second phase of our experiment, dN supplementation is dropped from 0.25 mg/mL to 0.01 mg/mL, and it is under these conditions of severe dN restriction that we observe parallel loss-of function mutations at the deoB locus (Table 1). We conclude that, when deoxyribonucleotides are in short supply, deoB loss of function ensures that dNTPs are not lost to glycolysis via salvage (Figure 9). The biochemistry of the DERA pathway supports mutation of deoB as the most plausible target, since this step is responsible for interconversion of deoxyribose-5-phosphate and deoxyribose-1-phosphate. The latter is the substrate for phosphorylase-mediated addition of nucleobases, generating dNs, whereas the former is the substrate for DERA-mediated salvage. Thus, deoB loss-of-function mutations serve to prevent dR loss to central metabolism while still permitting phosphorylase-mediated dN resynthesis (Figure 9). Thus, while conversion of DERA to a synthetic pathway might be a better long-term solution to loss of ribonucleotide reduction, loss of function of this pathway is advantageous in the short term, precluding its cooption to deoxyribonucleotide synthesis. This interpretation is consistent with a recent experiment where mutation of dihydrofolate reductase led to deoB loss of function to prevent loss of deoxyribose-5-phosphate to glycolysis (Rodrigues and Shakhnovich, 2019).

Analysis of species that lack nrd genes reveals frequent loss of deoB, suggesting that loss of ribonucleotide reduction might drive loss of deoB, as observed in our experiments. For members of Ureaplasma, the pattern is clear (Supplementary file 1L): all members of this genus completely lack genes for ribonucleotide reduction and show no evidence of coding for either deoB or cdd, the two most mutated genes in our experiments. For B. aphidicola cdd appears to have been completely lost from all sequenced strains. Close to half of sequenced Buchnera strains show co-occurrence of ribonucleotide reduction and phosphopentomutase genes (48/74 strains), with 30% (22/74 strains) lacking genes for both functions (Supplementary file 1N), suggesting that loss of one function frequently results in loss of the other. However, two strains appear to carry deoB while lacking nrd genes, and two further strains show the opposite pattern. If confirmed, this would indicate loss of one function may not always result in loss of the other.

We observe a very clear pattern in members of the Borreliaceae. The NCBI Taxonomy Browser (Schoch et al., 2020) lists two genera with genome data, Borrelia and Borreliella, following a recent proposal (Adeolu and Gupta, 2014; Barbour et al., 2017; Gupta and Bergström, 2019) to split the genus Borrelia into two (though this proposal has been contested; Margos et al., 2017; Margos et al., 2018; Margos et al., 2020). Our analyses indicate that Borrelia (sensu Adeolu and Gupta, 2014) code class Ib RNRs but lack deoB, while Borrelliela lack both deoB and RNR genes. This pattern could be taken to indicate that members of the Borrelia lost deoB but retained ribonucleotide reduction. However, phylogenetic analyses (Zhong et al., 2006) and location on linear plasmids (Miller et al., 2013) indicate that the class Ib RNR genes (nrdEF) in Borrelia has been acquired via horizontal gene transfer, presumably following initial loss of ribonucleotide reduction in the Borreliaceae.

It is noteworthy that class Ib RNR, carried by members of the Borrelia, is manganese-dependent (Cotruvo and Stubbe, 2010), whereas class Ia is an iron-dependent enzyme (Torrents, 2014). Interestingly, the most well-studied member of the Borreliaceae, Borreliella burgdorferi (sensu Adeolu and Gupta) does not require iron (Posey and Gherardini, 2000). It is tempting to speculate that initial loss of ribonucleotide reduction was driven by selection for iron independence, with subsequent reacquisition of ribonucleotide reduction favouring the manganese-dependent enzyme.

More generally, the switch to an intracellular existence may result in relaxed selection on ribonucleotide reduction if host-derived deoxyribonucleotides are sufficient to replace endogenous production. This appears to be an infrequent outcome as most intracellular bacteria retain ribonucleotide reduction. This rare event might be driven by additional pressures, as evolution of iron independence in B. burgdorferi suggests. Our experiments indicate that, in the absence of endogenous ribonucleotide reduction, disruption of the DERA pathway via mutations to deoB may serve to minimise loss of these building blocks when dNTP availability is limited. Given the necessity of genome replication, we conclude that life without ribonucleotide reduction drives the loss of dNTP catabolism via DERA, enabling building block reuse for DNA synthesis.

It is interesting to note that loss of ribonucleotide reduction appears extremely rare in the evolution of endosymbiosis and parasitism, despite the seeming ubiquity of host-derived deoxyribonucleotides. Loss of production in favour of acquisition from the host ought perhaps to be a more common outcome. Acquisition from the host would appear straightforward as many bacteria are known to possess mechanisms for uptake of dNs, deoxyribonucleotides, and even DNA (Huang et al., 2022). With respect to dN uptake, E. coli for instance codes for two broad-spectrum nucleoside transporters, NupC and NupG, which have been reported to take up dNs (Patching et al., 2005). These broad specificities are noteworthy in the context of our experimental results showing ΔRNR is viable when the media is supplemented with dC alone (Figure 3). It may be that imported dNs are not readily utilised in DNA synthesis as they must first be converted to deoxyribonucleotides for utilisation in genome synthesis. In our experiments, it was only possible to eliminate ribonucleotide reduction after first introducing a broad-spectrum deoxyadenosine kinase gene from M. mycoides, presumably because it improves the opportunity for imported dNs to be coopted for deoxyribonucleotide synthesis instead of being catabolised. Indeed, a recent study shows that dNs, deoxyribonucleotides, and DNA can all be used as a nitrogen source by E. coli (Huang et al., 2022). Those authors noted elevated expression of genes involved in catabolism, including deoD, which codes nucleoside phosphorylase, part of the DERA pathway (Figure 9). Thus, under conditions where DNA and its monomeric building blocks are important sources of nitrogen and phosphorus, loss of ribonucleotide reduction would presumably have a deleterious impact on growth. This trade-off may explain the small number of instances of loss of ribonucleotide reduction.

It remains to be seen whether there are conditions wherein the reverse DERA pathway (Poole et al., 2014) can operate synthetically in vivo as an alternative to ribonucleotide reduction, but one possibility is that this may be favoured under conditions where other nitrogen and phosphorous sources are abundant.

All genome data have been deposited to the Single Read Archive (https://www.ncbi.nlm.nih.gov/sra) under accessions SRX17677859-SRX17677886. The following biosamples are associated with this project: SAMN30957267-SAMN30957273. The data have been deposited with links to BioProject accession number PRJNA882995 in the NCBI BioProject database (https://www.ncbi.nlm.nih.gov/bioproject/).

1. 1. Arras SDM
   2. Sibaeva N
   3. Catchpole RJ
   4. Horinouchi N
   5. Si D
   6. Rickerby AM
   7. Deguchi K
   8. Hibi M
   9. Tanaka K
   10. Takeuchi M
   11. Ogawa J
   12. Poole AM

   (2022) NCBI BioProject

   ID PRJNA882995. Experimental evolution of an Escherichia coli line lacking ribonucleotide reduction.

   https://www.ncbi.nlm.nih.gov/bioproject/?term=PRJNA882995

1. 1. Adeolu M
   2. Gupta RS

   (2014) A phylogenomic and molecular marker based proposal for the division of the genus Borrelia into two genera: The emended genus Borrelia containing only the members of the relapsing fever Borrelia, and the genus borreliella gen nov containing the members of the Lyme disease Borrelia (Borrelia burgdorferi sensu lato complex)

   Antonie van Leeuwenhoek 105:1049–1072.

   https://doi.org/10.1007/s10482-014-0164-x

   • PubMed
   • Google Scholar
2. 1. Andersson JO

   (2000) Evolutionary genomics: Is buchnera a bacterium or an organelle?

   Current Biology 10:R866–R868.

   https://doi.org/10.1016/s0960-9822(00)00816-2

   • PubMed
   • Google Scholar
3. 1. Barbour AG
   2. Adeolu M
   3. Gupta RS

   (2017) Division of the genus borrelia into two genera (corresponding to lyme disease and relapsing fever groups) reflects their genetic and phenotypic distinctiveness and will lead to a better understanding of these two groups of microbes

   International Journal of Systematic and Evolutionary Microbiology 67:2058–2067.

   https://doi.org/10.1099/ijsem.0.001815

   • PubMed
   • Google Scholar
4. 1. Benner SA
   2. Ellington AD
   3. Tauer A

   (1989) Modern metabolism as a palimpsest of the rna world

   PNAS 86:7054–7058.

   https://doi.org/10.1073/pnas.86.18.7054

   • PubMed
   • Google Scholar
5. 1. Blaesi EJ
   2. Palowitch GM
   3. Hu K
   4. Kim AJ
   5. Rose HR
   6. Alapati R
   7. Lougee MG
   8. Kim HJ
   9. Taguchi AT
   10. Tan KO
   11. Laremore TN
   12. Griffin RG
   13. Krebs C
   14. Matthews ML
   15. Silakov A
   16. Bollinger JM
   17. Allen BD
   18. Boal AK

   (2018) Metal-free class ie ribonucleotide reductase from pathogens initiates catalysis with a tyrosine-derived dihydroxyphenylalanine radical

   PNAS 115:10022–10027.

   https://doi.org/10.1073/pnas.1811993115

   • PubMed
   • Google Scholar
6. Software

   1. Bushnell B

   (2014) BBTools

   Software Package.

   http://bbtools.jgi.doe.gov
7. 1. Cotruvo JA
   2. Stubbe J

   (2010) An active dimanganese (III) -tyrosyl radical cofactor in Escherichia coli class Ib ribonucleotide reductase

   Biochemistry 49:1297–1309.

   https://doi.org/10.1021/bi902106n

   • PubMed
   • Google Scholar
8. 1. Cotruvo JA
   2. Stubbe J

   (2011) Escherichia coli class Ib ribonucleotide reductase contains a dimanganese (III) -tyrosyl radical cofactor in vivo

   Biochemistry 50:1672–1681.

   https://doi.org/10.1021/bi101881d

   • PubMed
   • Google Scholar
9. 1. Edgar RC

   (2004) Muscle: A multiple sequence alignment method with reduced time and space complexity

   BMC Bioinformatics 5:113.

   https://doi.org/10.1186/1471-2105-5-113

   • PubMed
   • Google Scholar
10. Book

    1. Fehér T

    (2008)

    Microbial gene essentiality: Protocols and bioinformatics

    Humana.

    • Google Scholar
11. 1. Fraser CM
    2. Casjens S
    3. Huang WM
    4. Sutton GG
    5. Clayton R
    6. Lathigra R
    7. White O
    8. Ketchum KA
    9. Dodson R
    10. Hickey EK
    11. Gwinn M
    12. Dougherty B
    13. Tomb JF
    14. Fleischmann RD
    15. Richardson D
    16. Peterson J
    17. Kerlavage AR
    18. Quackenbush J
    19. Salzberg S
    20. Hanson M
    21. van Vugt R
    22. Palmer N
    23. Adams MD
    24. Gocayne J
    25. Weidman J
    26. Utterback T
    27. Watthey L
    28. McDonald L
    29. Artiach P
    30. Bowman C
    31. Garland S
    32. Fuji C
    33. Cotton MD
    34. Horst K
    35. Roberts K
    36. Hatch B
    37. Smith HO
    38. Venter JC

    (1997) Genomic sequence of a Lyme disease spirochaete, Borrelia burgdorferi

    Nature 390:580–586.

    https://doi.org/10.1038/37551

    • PubMed
    • Google Scholar
12. 1. Gilroy R
    2. Ravi A
    3. Getino M
    4. Pursley I
    5. Horton DL
    6. Alikhan N-F
    7. Baker D
    8. Gharbi K
    9. Hall N
    10. Watson M
    11. Adriaenssens EM
    12. Foster-Nyarko E
    13. Jarju S
    14. Secka A
    15. Antonio M
    16. Oren A
    17. Chaudhuri RR
    18. La Ragione R
    19. Hildebrand F
    20. Pallen MJ

    (2021) Extensive microbial diversity within the chicken gut microbiome revealed by metagenomics and culture

    PeerJ 9:e10941.

    https://doi.org/10.7717/peerj.10941

    • PubMed
    • Google Scholar
13. 1. Glass JI
    2. Lefkowitz EJ
    3. Glass JS
    4. Heiner CR
    5. Chen EY
    6. Cassell GH

    (2000) The complete sequence of the mucosal pathogen Ureaplasma urealyticum

    Nature 407:757–762.

    https://doi.org/10.1038/35037619

    • PubMed
    • Google Scholar
14. 1. Gupta RS
    2. Bergström S

    (2019) Distinction between borrelia and borreliella is more robustly supported by molecular and phenotypic characteristics than all other neighbouring prokaryotic genera: response to margos’ et al. “the genus borrelia reloaded” (plos one 13(12): e0208432)

    PLOS ONE 14:e0221397.

    https://doi.org/10.1371/journal.pone.0221397

    • PubMed
    • Google Scholar
15. 1. Hershberg R
    2. Petrov DA

    (2010) Evidence that mutation is universally biased towards at in bacteria

    PLOS Genetics 6:e1001115.

    https://doi.org/10.1371/journal.pgen.1001115

    • PubMed
    • Google Scholar
16. 1. Hofer A
    2. Crona M
    3. Logan DT
    4. Sjöberg B-M

    (2012) Dna building blocks: Keeping control of manufacture

    Critical Reviews in Biochemistry and Molecular Biology 47:50–63.

    https://doi.org/10.3109/10409238.2011.630372

    • PubMed
    • Google Scholar
17. 1. Horinouchi N
    2. Ogawa J
    3. Kawano T
    4. Sakai T
    5. Saito K
    6. Matsumoto S
    7. Sasaki M
    8. Mikami Y
    9. Shimizu S

    (2006a) Biochemical retrosynthesis of 2’-deoxyribonucleosides from glucose, acetaldehyde, and a nucleobase

    Applied Microbiology and Biotechnology 71:615–621.

    https://doi.org/10.1007/s00253-005-0205-5

    • PubMed
    • Google Scholar
18. 1. Horinouchi N
    2. Ogawa J
    3. Kawano T
    4. Sakai T
    5. Saito K
    6. Matsumoto S
    7. Sasaki M
    8. Mikami Y
    9. Shimizu S

    (2006b) One-Pot microbial synthesis of 2’-deoxyribonucleoside from glucose, acetaldehyde, and a nucleobase

    Biotechnology Letters 28:877–881.

    https://doi.org/10.1007/s10529-006-9019-5

    • PubMed
    • Google Scholar
19. 1. Huang L
    2. Zhang Y
    3. Du X
    4. An R
    5. Liang X

    (2022) Escherichia coli can eat DNA as an excellent nitrogen source to grow quickly

    Frontiers in Microbiology 13:894849.

    https://doi.org/10.3389/fmicb.2022.894849

    • Google Scholar
20. 1. Jeong H
    2. Barbe V
    3. Lee CH
    4. Vallenet D
    5. Yu DS
    6. Choi S-H
    7. Couloux A
    8. Lee S-W
    9. Yoon SH
    10. Cattolico L
    11. Hur C-G
    12. Park H-S
    13. Ségurens B
    14. Kim SC
    15. Oh TK
    16. Lenski RE
    17. Studier FW
    18. Daegelen P
    19. Kim JF

    (2009) Genome sequences of Escherichia coli B strains REL606 and BL21(DE3)

    Journal of Molecular Biology 394:644–652.

    https://doi.org/10.1016/j.jmb.2009.09.052

    • PubMed
    • Google Scholar
21. 1. Karlström HO

    (1970) Inability of Escherichia coli B to incorporate added deoxycytidine, deoxyandenosine, and deoxyguanosine into DNA

    European Journal of Biochemistry 17:68–71.

    https://doi.org/10.1111/j.1432-1033.1970.tb01135.x

    • PubMed
    • Google Scholar
22. 1. Langmead B
    2. Salzberg SL

    (2012) Fast gapped-read alignment with bowtie 2

    Nature Methods 9:357–359.

    https://doi.org/10.1038/nmeth.1923

    • PubMed
    • Google Scholar
23. 1. Lee H
    2. Popodi E
    3. Tang H
    4. Foster PL

    (2012) Rate and molecular spectrum of spontaneous mutations in the bacterium Escherichia coli as determined by whole-genome sequencing

    PNAS 109:E2774–E2783.

    https://doi.org/10.1073/pnas.1210309109

    • PubMed
    • Google Scholar
24. 1. Lenski RE
    2. Rose MR
    3. Simpson SC
    4. Tadler SC

    (1991) Long-Term experimental evolution in Escherichia coli. I. adaptation and divergence during 2,000 generations

    The American Naturalist 138:1315–1341.

    https://doi.org/10.1086/285289

    • Google Scholar
25. 1. Lundin D.
    2. Torrents E
    3. Poole AM
    4. Sjöberg BM

    (2009) RNRdb, a curated database of the universal enzyme family ribonucleotide reductase, reveals a high level of misannotation in sequences deposited to GenBank

    BMC Genomics 10:589.

    https://doi.org/10.1186/1471-2164-10-589

    • PubMed
    • Google Scholar
26. 1. Lundin D
    2. Gribaldo S
    3. Torrents E
    4. Sjöberg B-M
    5. Poole AM

    (2010) Ribonucleotide reduction-horizontal transfer of a required function spans all three domains

    BMC Evolutionary Biology 10:383.

    https://doi.org/10.1186/1471-2148-10-383

    • PubMed
    • Google Scholar
27. 1. Lundin D
    2. Berggren G
    3. Logan DT
    4. Sjöberg BM

    (2015) The origin and evolution of ribonucleotide reduction

    Life 5:604–636.

    https://doi.org/10.3390/life5010604

    • PubMed
    • Google Scholar
28. 1. Margos G
    2. Marosevic D
    3. Cutler S
    4. Derdakova M
    5. Diuk-Wasser M
    6. Emler S
    7. Fish D
    8. Gray J
    9. Hunfeldt K-P
    10. Jaulhac B
    11. Kahl O
    12. Kovalev S
    13. Kraiczy P
    14. Lane RS
    15. Lienhard R
    16. Lindgren PE
    17. Ogden N
    18. Ornstein K
    19. Rupprecht T
    20. Schwartz I
    21. Sing A
    22. Straubinger RK
    23. Strle F
    24. Voordouw M
    25. Rizzoli A
    26. Stevenson B
    27. Fingerle V

    (2017) There is inadequate evidence to support the division of the genus Borrelia

    International Journal of Systematic and Evolutionary Microbiology 67:1081–1084.

    https://doi.org/10.1099/ijsem.0.001717

    • Google Scholar
29. 1. Margos G
    2. Gofton A
    3. Wibberg D
    4. Dangel A
    5. Marosevic D
    6. Loh S-M
    7. Oskam C
    8. Fingerle V

    (2018) The genus Borrelia reloaded

    PLOS ONE 13:e0208432.

    https://doi.org/10.1371/journal.pone.0208432

    • PubMed
    • Google Scholar
30. 1. Margos G
    2. Fingerle V
    3. Oskam C
    4. Stevenson B
    5. Gofton A

    (2020) Comment on: gupta, 2019, distinction between borrelia and borreliella is more robustly supported by molecular and phenotypic characteristics than all other neighbouring prokaryotic genera: response to margos’ et al. “the genus borrelia reloaded” (plos one 13(12): e0208432). plos one 14(8):e0221397

    Ticks and Tick-Borne Diseases 11:101320.

    https://doi.org/10.1016/j.ttbdis.2019.101320

    • PubMed
    • Google Scholar
31. 1. Martin JE
    2. Imlay JA

    (2011) The alternative aerobic ribonucleotide reductase of Escherichia coli, NrdEF, is a manganese-dependent enzyme that enables cell replication during periods of iron starvation

    Molecular Microbiology 80:319–334.

    https://doi.org/10.1111/j.1365-2958.2011.07593.x

    • PubMed
    • Google Scholar
32. 1. Miller C
    2. Thomsen LE
    3. Gaggero C
    4. Mosseri R
    5. Ingmer H
    6. Cohen SN

    (2004) Sos response induction by ß-lactams and bacterial defense against antibiotic lethality

    Science 305:1629–1631.

    https://doi.org/10.1126/science.1101630

    • Google Scholar
33. 1. Miller SC
    2. Porcella SF
    3. Raffel SJ
    4. Schwan TG
    5. Barbour AG

    (2013) Large linear plasmids of Borrelia species that cause relapsing fever

    Journal of Bacteriology 195:3629–3639.

    https://doi.org/10.1128/JB.00347-13

    • PubMed
    • Google Scholar
34. 1. Monje-Casas F
    2. Jurado J
    3. Prieto-Alamo MJ
    4. Holmgren A
    5. Pueyo C

    (2001) Expression analysis of THE nrdhief operon from Escherichia coli conditions that trigger the transcript level in vivo

    The Journal of Biological Chemistry 276:18031–18037.

    https://doi.org/10.1074/jbc.M011728200

    • PubMed
    • Google Scholar
35. 1. Neidhardt FC
    2. Bloch PL
    3. Smith DF

    (1974) Culture medium for enterobacteria

    Journal of Bacteriology 119:736–747.

    https://doi.org/10.1128/jb.119.3.736-747.1974

    • PubMed
    • Google Scholar
36. 1. Nguyen-Vo TP
    2. Ko S
    3. Ryu H
    4. Kim JR
    5. Kim D
    6. Park S

    (2020) Systems evaluation reveals novel transporter yohjk renders 3-hydroxypropionate tolerance in Escherichia coli

    Scientific Reports 10:19064.

    https://doi.org/10.1038/s41598-020-76120-3

    • PubMed
    • Google Scholar
37. 1. Nordlund P
    2. Reichard P

    (2006) Ribonucleotide reductases

    Annual Review of Biochemistry 75:681–706.

    https://doi.org/10.1146/annurev.biochem.75.103004.142443

    • PubMed
    • Google Scholar
38. 1. Ogawa J
    2. Saito K
    3. Sakai T
    4. Horinouchi N
    5. Kawano T
    6. Matsumoto S
    7. Sasaki M
    8. Mikami Y
    9. Shimizu S

    (2003) Microbial production of 2-deoxyribose 5-phosphate from acetaldehyde and triosephosphate for the synthesis of 2’-deoxyribonucleosides

    Bioscience, Biotechnology, and Biochemistry 67:933–936.

    https://doi.org/10.1271/bbb.67.933

    • PubMed
    • Google Scholar
39. 1. Omer S
    2. Harlow TJ
    3. Gogarten JP

    (2017) Does sequence conservation provide evidence for biological function?

    Trends in Microbiology 25:11–18.

    https://doi.org/10.1016/j.tim.2016.09.010

    • PubMed
    • Google Scholar
40. 1. Panosian TD
    2. Nannemann DP
    3. Watkins GR
    4. Phelan VV
    5. McDonald WH
    6. Wadzinski BE
    7. Bachmann BO
    8. Iverson TM

    (2011) Bacillus cereus phosphopentomutase is an alkaline phosphatase family member that exhibits an altered entry point into the catalytic cycle

    The Journal of Biological Chemistry 286:8043–8054.

    https://doi.org/10.1074/jbc.M110.201350

    • PubMed
    • Google Scholar
41. 1. Patching SG
    2. Baldwin SA
    3. Baldwin AD
    4. Young JD
    5. Gallagher MP
    6. Henderson PJF
    7. Herbert RB

    (2005) The nucleoside transport proteins, nupc and nupg, from Escherichia coli: specific structural motifs necessary for the binding of ligands

    Organic & Biomolecular Chemistry 3:462.

    https://doi.org/10.1039/b414739a

    • Google Scholar
42. 1. Pérez-Brocal V
    2. Gil R
    3. Ramos S
    4. Lamelas A
    5. Postigo M
    6. Michelena JM
    7. Silva FJ
    8. Moya A
    9. Latorre A

    (2006) A small microbial genome: the end of a long symbiotic relationship?

    Science 314:312–313.

    https://doi.org/10.1126/science.1130441

    • PubMed
    • Google Scholar
43. 1. Poole AM
    2. Logan DT
    3. Sjöberg BM

    (2002) The evolution of the ribonucleotide reductases: much ado about oxygen

    Journal of Molecular Evolution 55:180–196.

    https://doi.org/10.1007/s00239-002-2315-3

    • PubMed
    • Google Scholar
44. 1. Poole AM
    2. Horinouchi N
    3. Catchpole RJ
    4. Si D
    5. Hibi M
    6. Tanaka K
    7. Ogawa J

    (2014) The case for an early biological origin of DNA

    Journal of Molecular Evolution 79:204–212.

    https://doi.org/10.1007/s00239-014-9656-6

    • PubMed
    • Google Scholar
45. 1. Posey JE
    2. Gherardini FC

    (2000) Lack of a role for iron in the Lyme disease pathogen

    Science 288:1651–1653.

    https://doi.org/10.1126/science.288.5471.1651

    • PubMed
    • Google Scholar
46. 1. Racker E

    (1951) Enzymatic synthesis of deoxypentose phosphate

    Nature 167:408–409.

    https://doi.org/10.1038/167408b0

    • PubMed
    • Google Scholar
47. 1. Racker E

    (1952) Enzymatic synthesis and breakdown of desoxyribose phosphate

    Journal of Biological Chemistry 196:347–365.

    https://doi.org/10.1016/S0021-9258(18)55739-2

    • Google Scholar
48. 1. Reichard P

    (1993) From RNA to DNA, why so many ribonucleotide reductases?

    Science 260:1773–1777.

    https://doi.org/10.1126/science.8511586

    • PubMed
    • Google Scholar
49. 1. Rodrigues JV
    2. Shakhnovich EI

    (2019) Adaptation to mutational inactivation of an essential gene converges to an accessible suboptimal fitness peak

    eLife 8:e50509.

    https://doi.org/10.7554/eLife.50509

    • PubMed
    • Google Scholar
50. 1. Rose HR
    2. Ghosh MK
    3. Maggiolo AO
    4. Pollock CJ
    5. Blaesi EJ
    6. Hajj V
    7. Wei Y
    8. Rajakovich LJ
    9. Chang WC
    10. Han Y
    11. Hajj M
    12. Krebs C
    13. Silakov A
    14. Pandelia ME
    15. Bollinger JM
    16. Boal AK

    (2018) Structural basis for superoxide activation of flavobacterium johnsoniae class i ribonucleotide reductase and for radical initiation by its dimanganese cofactor

    Biochemistry 57:2679–2693.

    https://doi.org/10.1021/acs.biochem.8b00247

    • PubMed
    • Google Scholar
51. 1. Schoch CL
    2. Ciufo S
    3. Domrachev M
    4. Hotton CL
    5. Kannan S
    6. Khovanskaya R
    7. Leipe D
    8. Mcveigh R
    9. O’Neill K
    10. Robbertse B
    11. Sharma S
    12. Soussov V
    13. Sullivan JP
    14. Sun L
    15. Turner S
    16. Karsch-Mizrachi I

    (2020) Ncbi taxonomy: a comprehensive update on curation, resources and tools

    Database 2020:baaa062.

    https://doi.org/10.1093/database/baaa062

    • PubMed
    • Google Scholar
52. 1. Silva F
    2. Lourenço O
    3. Pina-Vaz C
    4. Rodrigues AG
    5. Queiroz JA
    6. Domingues FC

    (2010) The use of DRAQ5 to monitor intracellular DNA in Escherichia coli by flow cytometry

    Journal of Fluorescence 20:907–914.

    https://doi.org/10.1007/s10895-010-0636-y

    • PubMed
    • Google Scholar
53. 1. Sintchak MD
    2. Arjara G
    3. Kellogg BA
    4. Stubbe J
    5. Drennan CL

    (2002) The crystal structure of class II ribonucleotide reductase reveals how an allosterically regulated monomer mimics a dimer

    Nature Structural Biology 9:293–300.

    https://doi.org/10.1038/nsb774

    • PubMed
    • Google Scholar
54. 1. Srinivas V
    2. Lebrette H
    3. Lundin D
    4. Kutin Y
    5. Sahlin M
    6. Lerche M
    7. Eirich J
    8. Branca RMM
    9. Cox N
    10. Sjöberg B-M
    11. Högbom M

    (2018) Metal-Free ribonucleotide reduction powered by a dopa radical in Mycoplasma pathogens

    Nature 563:416–420.

    https://doi.org/10.1038/s41586-018-0653-6

    • PubMed
    • Google Scholar
55. 1. Stevenson K
    2. McVey AF
    3. Clark IBN
    4. Swain PS
    5. Pilizota T

    (2016) General calibration of microbial growth in microplate readers

    Scientific Reports 6:38828.

    https://doi.org/10.1038/srep38828

    • PubMed
    • Google Scholar
56. 1. Tamas I
    2. Klasson L
    3. Canbäck B
    4. Näslund AK
    5. Eriksson AS
    6. Wernegreen JJ
    7. Sandström JP
    8. Moran NA
    9. Andersson SGE

    (2002) 50 million years of genomic stasis in endosymbiotic bacteria

    Science 296:2376–2379.

    https://doi.org/10.1126/science.1071278

    • PubMed
    • Google Scholar
57. 1. Tauer A
    2. Benner SA

    (1997) The B12-dependent ribonucleotide reductase from the archaebacterium Thermoplasma acidophila: an evolutionary solution to the ribonucleotide reductase conundrum

    PNAS 94:53–58.

    https://doi.org/10.1073/pnas.94.1.53

    • PubMed
    • Google Scholar
58. 1. Torrents E
    2. Aloy P
    3. Gibert I
    4. Rodríguez-Trelles F

    (2002) Ribonucleotide reductases: divergent evolution of an ancient enzyme

    Journal of Molecular Evolution 55:138–152.

    https://doi.org/10.1007/s00239-002-2311-7

    • PubMed
    • Google Scholar
59. 1. Torrents E

    (2014) Ribonucleotide reductases: essential enzymes for bacterial life

    Frontiers in Cellular and Infection Microbiology 4:52.

    https://doi.org/10.3389/fcimb.2014.00052

    • PubMed
    • Google Scholar
60. 1. Wang L
    2. Westberg J
    3. Bölske G
    4. Eriksson S

    (2001) Novel deoxynucleoside-phosphorylating enzymes in mycoplasmas: evidence for efficient utilization of deoxynucleosides

    Molecular Microbiology 42:1065–1073.

    https://doi.org/10.1046/j.1365-2958.2001.02700.x

    • PubMed
    • Google Scholar
61. 1. Yao Z
    2. Kahne D
    3. Kishony R

    (2012) Distinct single-cell morphological dynamics under beta-lactam antibiotics

    Molecular Cell 48:705–712.

    https://doi.org/10.1016/j.molcel.2012.09.016

    • PubMed
    • Google Scholar
62. 1. Zhong J
    2. Skouloubris S
    3. Dai Q
    4. Myllykallio H
    5. Barbour AG

    (2006) Function and evolution of plasmid-borne genes for pyrimidine biosynthesis in Borrelia spp

    Journal of Bacteriology 188:909–918.

    https://doi.org/10.1128/JB.188.3.909-918.2006

    • PubMed
    • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "Characterisation of an Escherichia coli line that completely lacks ribonucleotide reduction yields insights into the evolution of obligate intracellularity." for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Christian Landry as the Senior Editor. The following individual involved in the review of your submission has agreed to reveal their identity: Anders Hofer (Reviewer #2).

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

Essential revisions:

1) Revise the introductory text about cofactor usage by class I RNRs recommended by reviewer #1. Also, modify the discussion of pathogens as detailed by reviewer #2 and comment on the selection of the starting strain as suggested by reviewer #3.

2) Make the changes to Figure 9 that are described by reviewer #2.

3) Comment further on how observation of cell elongation would impact OD measurements (reviewer #1 comment) and address the concerns of reviewer #3 about cell morphology changes and analysis of evolutionary trajectory.

4) Modify the abstract and introduction to include a statement about the need to introduce a nucleotide kinase gene.

5) Describe the control experiment to identify selective pressures induced by the antibiotic.

6) Clarify whether population-level sequencing experiments were performed as detailed by reviewer #3.

7) Revise Figure 1 to include the entire gel and include more effective annotations.

Reviewer #1 (Recommendations for the authors):

Line 72-73 – revise terminology referring to "an oxygen-containing metal ion centre" to describe Cys radical generation by class I RNRs. The Cys oxidant cofactor employed by class I enzymes does not contain oxygen. Class I RNR cofactors are assembled from either oxygen or an oxygen-derived oxidant (superoxide). But the important part about the oxygen dependence during assembly is not the incorporation of oxygen per se – it has not been established whether that happens in every case anyway – but the fact that oxygen provides oxidizing equivalents that can convert the precursor form of the cofactor (Mn(II), Fe(II), etc) to an oxidant with sufficient potency to oxidize a Cys side chain. Also, class Ie RNRs do not contain metal in their activated form – so it is also not accurate to say that all class I RNRs use a metal center to accomplish Cys oxidation.

Line 284 – there are a number of more recent studies that support the idea that class Ib RNR is used under conditions of iron limitation in E. coli.

Line 643 – missing word.

Reviewer #2 (Recommendations for the authors):

The broad questions are already described in the Public Review section (major findings and possible future experiments). Generally, I think the findings are very interesting but I have a few things that I would like to be corrected.

I find it a bit misleading that it is mentioned in the manuscript that the few organisms that lack RNR are obligately intracellular and Borrelia burgdorferi is mentioned as one of them. B. burgdorferi is mainly extracellular in the mammalian host, although it has been suggested that it can have intracellular stages. Another extracellular pathogen that lacks RNR is Giardia (several species). This is a protozoan parasite (eukaryote) that is extracellular. I think this organism should be mentioned as well in the manuscript. I am also surprised that it is stated that Mycoplasma is known to be dependent on salvage synthesis for dNTP production (row 300). They actually encode a novel subclass of class I RNRs that was published a few years ago (DOPA radical-dependent RNRs). Remove the sentence or explain this part further. The statement of lacking RNR is valid for Ureaplasma but not Mycoplasma.

I also think some adjustments need to be made in Figure 9. First of all, I think it would be good to use "ribonucleotides" and "deoxyribonucleotides" instead of NTPs and dNTPs in the top part of the figure showing the RNR reaction. This will both solve the problem that many RNRs use diphosphates as substrates and also that dNKs phosphorylate deoxyribonucleosides to the monophosphate level and not the triphosphate level. Another problem with the figure is that PPase normally means pyrophosphatase and because of that I think it is better to write "nucleoside phosphorylase" or alternatively just "phosphorylase" if the space is limited. A third problem is that it looks like the cells are taking up dNTPs, which is usually not the case, at least not in prokaryotes that do not have endocytosis. The uptake arrow from the outside should instead lead to dNs.

It should be stated in the abstract and introduction that they needed to introduce an exogenous deoxynucleoside kinase gene to the E. coli cells in order to get the system to work (E. coli can normally only phosphorylate thymidine). I would also like them to discuss more what is known about the substrate specificity of the kinase and what transporters E. coli has for the uptake of the deoxynucleosides.

Reviewer #3 (Recommendations for the authors):

Specific questions/comments for the authors:

– A general question, why did the authors choose REL606 as opposed to other strains? Is this due to evolutionary experiment convenience? If so, please briefly discuss this.

– Please clarify if a clonal and/or population-level sequence was performed, and how.

– What is the control experiment (i.e. a strain evolved in parallel) to detect any selective pressures that could be introduced by the strep in the evolution media?

– What is the significance of the cell morphology analyses for the evolutionary trajectory hypotheses presented in the study? An earlier justification or a better-justified discussion would be helpful here. To my knowledge, REL606 does not particularly exhibit a morphologically distinct response to many environments the bug has been tested/evolved before, or is this not the case? If so, then why is this a key observation attributable to the hypothesis tested here?

– Is there a correlation between the number of mutations vs. the morphological modifications of the evolved cell?

– Please clarify what is meant by this sentence (line 414): Note however that elongate cell morphology and clumping preclude use of OD measurements for accurate estimation of doubling time or cell counts (reflected in the large standard error relative to REL606 controls), so can only be used to give a general indication of growth.

– I'm intrigued by the cross-contamination speculation (and lack thereof) for the T30.L1 line. Did the authors perform population-level sequencing? Shouldn't that reveal whether early contamination took place?

– As a general comment, the second half of the Results section is rather difficult to follow. Some arguments are emphasized in this part (especially the loss of deoB and cdd genes section) and others can move to Discussion; there is a lot of back/forthing going on in this section, making it harder to grasp the sequence of events. Similarly, some paragraphs emphasize results are in the Discussion section (see, for instance, section starting line 653).

– Please include what blue/red lines correspond to on Figures 2 and 3 directly, in addition to the legend.

– Figure 1 should include the entire gel, not a trimmed version. Include an arrow pointing to exactly what we are looking at here.

– Overall summary figure (indicating the "Sequence" would be a good place to specify what you mean by sequence: and how (clonal, genome, how many replicates, etc.)).

– Specify what you mean by length (uM) on the figure in addition to the caption. The colors here do not indicate the specific strain on the Figure – is red the WT? Blue evolved WT? Unclear.

Essential revisions:

1) Revise the introductory text about cofactor usage by class I RNRs recommended by reviewer #1. Also, modify the discussion of pathogens as detailed by reviewer #2 and comment on the selection of the starting strain as suggested by reviewer #3.

We have now reworded the introductory text for accuracy and added references to recent papers reporting on the advances that reviewer #1 notes (ll 72-76). We also added references regarding the recent studies showing a likely role for Ib in replication during iron starvation (ll 295-296). Thanks for alerting us to these.

Thanks to Reviewer #2 for noting some inaccuracies in our descriptions of the key lineages that do not have RNRs. We have corrected these errors. We had originally omitted the eukaryote cases as our work is focused on bacteria, but agree it is helpful to mention these, so we have now added these in for completeness (ll 81-89).

We have also corrected the statement regarding the presence of ribonucleotide reductase in Mycoplasma (l 312). We now cite the discovery of DOPA radical-dependent RNRs in the introductory text. (ll 73-74)

Regarding the choice of starting strain, we did indeed choose this because of some of the convenient features of this strain for experimental evolution, including genomically coded antibiotic resistance and presence of a neutral SNP in a gene involved in arabinose utilisation, which makes this lineage suitable for competition experiments. We made use of the former in the project, but not the latter.

We have now made a statement regarding choice of strain under the methods subsection, ‘Strains and growth conditions’. (ll 125-131)

2) Make the changes to Figure 9 that are described by reviewer #2.

Figure 9 has been updated for clarity, following advice from Reviewer #2. We have replaced NTPs/dNTPs with ribonucleotides and deoxyribonucleotides, respectively. Thanks for this suggestion. While we and others have used the abbreviation PPase elsewhere, we do appreciate the point, and have altered this to remove any ambiguity. PPase is now replaced with “nucleoside phosphorylase". The arrow from outside the cell has been corrected to lead to the deoxyribonucleotides label inside the cell, and the cell boundary has been redrawn for clarity. We have also updated the figure legend (ll 834-845).

3) Comment further on how observation of cell elongation would impact OD measurements (reviewer #1 comment) and address the concerns of reviewer #3 about cell morphology changes and analysis of evolutionary trajectory.

We mentioned this in the abstract, but we have now added brief mention of this in the final paragraph of the introduction, where we summarise key findings (ll 109-111).

For establishing if there is either growth or no growth under various conditions, as we have done, a qualitative assessment such as the one presented in Figure 3 is sufficient. The issue of whether OD is impacted by cell elongation has been documented previously (Stevenson et al. https://www.nature.com/articles/srep38828), and is only a problem if trying to quantify parameters such as doubling time or when trying to estimate cell counts. We do not do either of these, as calculation of both requires an assumption of normal cell morphology in E. coli. We have added a note to clarify this in the first paragraph of the Discussion section, as per the suggestion from Reviewer #1. (ll 606-613)

4) Modify the abstract and introduction to include a statement about the need to introduce a nucleotide kinase gene.

We have now added statements to this effect in the abstract (ll 38-39) and in the final paragraph of the introduction (ll 94-98).

5) Describe the control experiment to identify selective pressures induced by the antibiotic.

Both the control and the experimental lines are resistant to streptomycin. The basis of the resistance is a genomic mutation in the rpsL gene which occurred following selection for resistance in the creation of REL606 (see Studier et al. 2009; https://doi.org/10.1016/j.jmb.2009.09.021). As streptomycin is included in the growth media of both the control and experimental lines, it should not affect interpretation of results.

If additional mutations do appear as a result of antibiotic resistance or some other form of media adaptation, they would be expected to appear in both control and experimental lines as both are evolved in the same antibiotic-containing growth media. Our genome sequencing indicated only one gene is mutated in both control and experimental lines. This gene, kdbD, encodes a putative histidine kinase. Across the control and experimental lines, we observe a range of mutations, including synonymous and nonsynonymous substitutions, plus one apparent frameshift in kdbD. Mutation of this gene cannot be specifically linked to any selection pressures associated with antibiotic resistance, and we note that this gene is not well studied. However, we plan to revisit this matter in a follow-up paper where we plan to present detailed population-level genomic data (see response to point 6, below).

Given that antibiotic presence is a constant across all lines, and given that we do not see evidence of specific adaptation to this antibiotic (as the lines are already preadapted to its presence) we do not believe it is necessary to discuss this, so we have not made any changes to the manuscript.

6) Clarify whether population-level sequencing experiments were performed as detailed by reviewer #3.

As noted in the Methods section (under Sequencing and genome assembly), genome data are derived from single colonies (ll 235-237). We have since performed population-level sequencing which we will present as part of a subsequent study, where we have extended the evolution experiment through to 100 transfers, and examined population dynamics.

7) Revise Figure 1 to include the entire gel and include more effective annotations.

Figure now edited and figure legend updated.

Reviewer #1 (Recommendations for the authors):

Line 643 – missing word.

Corrected, thanks.

Reviewer #3 (Recommendations for the authors):

– Please clarify if a clonal and/or population-level sequence was performed, and how.

Clonal – see above.

– What is the control experiment (i.e. a strain evolved in parallel) to detect any selective pressures that could be introduced by the strep in the evolution media?

Note that a control experiment is presented, but not for Streptomycin, for reasons noted above.

– What is the significance of the cell morphology analyses for the evolutionary trajectory hypotheses presented in the study? An earlier justification or a better-justified discussion would be helpful here. To my knowledge, REL606 does not particularly exhibit a morphologically distinct response to many environments the bug has been tested/evolved before, or is this not the case? If so, then why is this a key observation attributable to the hypothesis tested here?

See above. The cell morphology is reversible, so it is not significant in evolutionary terms. As now explained in the revised text, this type of morphology has been seen in a range of conditions, and is associated with stress. We show it is specifically a response to the environment (i.e. low [dN]) and is alleviated when cells are grown in higher [dN]. Perhaps what is interesting is that this environmental phenotype appears to be less severe as the experiment proceeds. This might be due to improved capacity to utilise dNs, perhaps via reduction of loss to catabolism following mutation of deoB. We now make this suggestion in the discussion (l l613-617).

– Is there a correlation between the number of mutations vs. the morphological modifications of the evolved cell?

No. We show that the morphology is environmental, not genetic – our cell lines lose this phenotype when grown in higher [dNs], though note above comment about mutations that might influence the availability of dNs for DNA synthesis.

– Please clarify what is meant by this sentence (line 414): Note however that elongate cell morphology and clumping preclude use of OD measurements for accurate estimation of doubling time or cell counts (reflected in the large standard error relative to REL606 controls), so can only be used to give a general indication of growth.

This is now discussed, as noted above. We cite a key paper (Stevenson et al.) which explains why these types of cellular morphology/behaviour preclude use of OD measurements for estimating doubling time/cell counts.

– I'm intrigued by the cross-contamination speculation (and lack thereof) for the T30.L1 line. Did the authors perform population-level sequencing? Shouldn't that reveal whether early contamination took place?

We have not performed population sequencing for the present work. However, we are currently undertaking this for a follow-up study which takes our evolution experiment through to 100 transfers. As the reviewer notes, with population data, we should indeed be able to confirm the status of the SNP in deoB in lines 1 and 8, and we look forward to presenting that work in our next article.

– As a general comment, the second half of the Results section is rather difficult to follow. Some arguments are emphasized in this part (especially the loss of deoB and cdd genes section) and others can move to Discussion; there is a lot of back/forthing going on in this section, making it harder to grasp the sequence of events. Similarly, some paragraphs emphasize results are in the Discussion section (see, for instance, section starting line 653).

We presume that, by the ‘second half’ of the results, the reviewer is referring to the bioinformatics-based analyses presented. We presented the results of screens for several genes for three genera, and elected to present results by genus, rather than by gene. We believe this is the more logical way to follow for understanding the results in biological context, but it does inevitably result in some back and forth, since we are discussing the same gene sets in several genera.

Regarding the results from line 653, we felt it useful to reemphasise the key results, but we have now added some further discussion around uptake of deoxyribonucleosides, which we hope puts this in clearer context. This is at ll 680-682 in the revised manuscript.

– Please include what blue/red lines correspond to on Figures 2 and 3 directly, in addition to the legend.

Corrected.

– Figure 1 should include the entire gel, not a trimmed version. Include an arrow pointing to exactly what we are looking at here.

Corrected.

– Overall summary figure (indicating the "Sequence" would be a good place to specify what you mean by sequence: and how (clonal, genome, how many replicates, etc.)).

Corrected. The legend to Figure 5 now reads:

“Five lines of ∆RNR and three lines of wild type progenitor (REL606) were established at one of two conditions (1 mg/mL or 0.25 mg/mL dNs in MOPS+1% glucose), and serially passaged for 30 transfers. Genomic material from each line was then extracted from a single colony and sent for sequencing giving us clonal level genome information. To further investigate adaptation to low concentrations of dNs, the “fittest” ∆RNR line grown at 0.25 mg/mL dNs (∆RNR_250_T30_1) was used to seed a subsequent experiment. Ten replicate lines of ∆RNR_250_T30 _1 were serially passaged for an additional 10 transfers in MOPS+1% glucose and 0.01 mg/ml dNs. DNA from a single colony of each of these 10 lines (∆RNR_10_T40_1-10) was then extracted and sent for whole genome sequencing”.

We have also indicated this in methods and in the main text, and we have also altered the text in the figure itself to make it clear we sequenced from single colonies.

– Specify what you mean by length (uM) on the figure in addition to the caption. The colors here do not indicate the specific strain on the Figure – is red the WT? Blue evolved WT? Unclear.

Corrected. Figure 6 has been updated to include the axis label to read “cell length (uM)”. The missing colour indicators in the key have been updated.

Author details

1. Samantha DM Arras

   School of Biological Sciences, University of Auckland, Auckland, New Zealand

   Contribution

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

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0009-0003-7544-6007

2. Nellie Sibaeva

   School of Biological Sciences, University of Auckland, Auckland, New Zealand

   Contribution

   Conceptualization, Validation, Investigation

   Competing interests

   No competing interests declared

3. Ryan J Catchpole

   School of Biological Sciences, University of Canterbury, Christchurch, New Zealand

   Present address

   Department of Biochemistry and Molecular Biology, University of Georgia, Athens, United States

   Contribution

   Conceptualization, Supervision, Validation, Investigation

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-6641-647X

4. Nobuyuki Horinouchi

   Division of Applied Life Sciences,Graduate School of Agriculture, Kyoto University, Kyoto, Japan

   Present address

   Amano Enzyme Inc, Nagoya, Japan

   Contribution

   Conceptualization, Supervision, Validation, Investigation

   Competing interests

   No competing interests declared

5. Dayong Si

   Division of Applied Life Sciences,Graduate School of Agriculture, Kyoto University, Kyoto, Japan

   Present address

   State Key Laboratory of Animal Nutrition, Laboratory of Feed Biotechnology, College of Animal Science and Technology, China Agricultural University, Beijing, China

   Contribution

   Conceptualization, Validation, Investigation

   Competing interests

   No competing interests declared

6. Alannah M Rickerby

   School of Biological Sciences, University of Auckland, Auckland, New Zealand

   Contribution

   Conceptualization, Formal analysis, Validation, Investigation

   Competing interests

   No competing interests declared

7. Kengo Deguchi

   Division of Applied Life Sciences,Graduate School of Agriculture, Kyoto University, Kyoto, Japan

   Contribution

   Investigation

   Competing interests

   No competing interests declared

8. Makoto Hibi

   Division of Applied Life Sciences,Graduate School of Agriculture, Kyoto University, Kyoto, Japan

   Present address

   Department of Biotechnology, Biotechnology Research Center, Toyama Prefectural University, Toyama, Japan

   Contribution

   Conceptualization, Supervision, Investigation

   Competing interests

   No competing interests declared

9. Koichi Tanaka

   Division of Applied Life Sciences,Graduate School of Agriculture, Kyoto University, Kyoto, Japan

   Present address

   Department of Nutritional Science, Okayama Prefectural University, Okayama, Japan

   Contribution

   Conceptualization, Supervision, Investigation

   Competing interests

   No competing interests declared

10. Michiki Takeuchi

    Division of Applied Life Sciences,Graduate School of Agriculture, Kyoto University, Kyoto, Japan

    Present address

    Industrial Microbiology, Graduate School of Agriculture, Kyoto University, Kyoto, Japan

    Contribution

    Conceptualization, Supervision, Investigation

    Competing interests

    No competing interests declared

11. Jun Ogawa

    Division of Applied Life Sciences,Graduate School of Agriculture, Kyoto University, Kyoto, Japan

    Contribution

    Conceptualization, Resources, Supervision, Funding acquisition, Project administration, Writing – review and editing

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0003-2741-621X

12. Anthony M Poole

    School of Biological Sciences, University of Auckland, Auckland, New Zealand

    Contribution

    Conceptualization, Formal analysis, Supervision, Funding acquisition, Investigation, Visualization, Writing – original draft, Project administration, Writing – review and editing

    For correspondence

    a.poole@auckland.ac.nz

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0001-9940-2824

• 705

  Page views

• 99

  Downloads

• 0

  Citations

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