RNA replication errors and the evolution of viruspathogenicity and virulenceIsabel S Novella, John B Presloid and R Travis Taylor

Available online at www.sciencedirect.com

ScienceDirect

RNA viruses of plants and animals have polymerases that are

error-prone and produce complex populations of related, but

non-identical, genomes called quasispecies. While there are

vast variations in mutation rates among these viruses, selection

has optimized the exact error rate of each species to provide

maximum speed of replication and amount of variation without

losing the ability to replicate because of excessive mutation.

High mutation rates result in the selection of populations

increasingly robust, which means they are increasingly

resistant to show phenotypic changes after mutation. It is

possible to manipulate the mutation rate, either by the use of

mutagens or by selection (or genetic manipulation) of fidelity

mutants. These polymerases usually, but not always, perform

as well as wild type (wt) during cell infection, but show major

phenotypic changes during in vivo infection. Both high and low

fidelity variants are attenuated when the wt virus is virulent in

the host. Alternatively when wt infection is non-apparent, the

variants show major restrictions to spread in the infected host.

Manipulation of mutation rates may become a new strategy to

develop attenuated vaccines for humans and animals.

Addresses

Department of Medical Microbiology and Immunology, College of

Medicine and Life Sciences, University of Toledo, USA

Corresponding author: Novella, Isabel S (isabel.Novella@utoledo.edu)

Current Opinion in Virology 2014, 9:143–147

This review comes from a themed issue on Virus replication in

animals and plants

Edited by C Cheng Kao and Olve B Peersen

For a complete overview see the Issue and the Editorial

Available online 22nd October 2014

http://dx.doi.org/10.1016/j.coviro.2014.09.017

1879-6257/# 2014 Elsevier B.V. All rights reserved.

IntroductionLow polymerase fidelity has a major role shaping the

genetic architecture and evolution of RNA viruses.

Natural selection has targeted the accuracy of these

enzymes to strike a compromise between the need for

rapid replication and the advantage of extensive variation.

Too high of a mutation rate leads to mutational melt-

down, while insufficient mutation rate results to lack of

enough diversity. Once variation is available its survival or

extinction will depend on selection and random events.

www.sciencedirect.com

The evolution of RNA viruses fits the quasispecies

model developed by Manfred Eigen and colleagues

proposed the original quasispecies theory to describe

the behavior of error-prone RNA populations, which

have been proposed to constitute the origin of self-

replication [1–3]. Quasispecies is a special application

of mutation-selection balance that applies to populations

with mutation rates such that selection targets the popu-

lation as a whole by virtue of its ability to withstand

mutation. For a recent in-depth review of viral quasis-

pecies, see Ref [4��].

A quasispecies contains a complex combination of gen-

omes that are related but non-identical. In practical terms,

standard sequencing of viral quasispecies results in the

consensus sequence, with the average nucleotide at each

position (Figure 1). Note that the consensus sequence

may or may not exist in the population (Figure 1c). The

most common genome is the master sequence, which may

be the same as the consensus (Figure 1a) or not

(Figure 1b).

Other important concepts (see Table 1) for a quasipecies

are mutation rate and mutant frequency. The latter is the

result of selection and drift acting on a population with a

specific mutation rate. All other things being equal it is

possible to compare mutation rates by measuring mutant

frequencies. If the mutation rate is high enough a qua-

sispecies cannot be formed. The value at which this

happens is the error threshold and the population is said

to have undergone mutational meltdown. At this point

the minimal genetic information required for replication

is lost.

Sequence space is a representation of the possible com-

binations that a genome of N nucleotides can have. With

four possible nucleotides for each position, the total

number of combinations for a 10 kb genome is 410,000,

a number beyond the wildest imagination. Most sequence

space is empty, because the vast majority of combinations

will yield a non-replicating genome. The sequence space

occupied by a population describes the mutant cloud or

mutant swarm.

Selection for robustnessThe main consequence of high mutation rates under

selection is that populations, rather than individual vir-

ions, are the main targets of selection based on their

robustness. Genetic robustness is the ability of a genome

to remain unchanged despite mutation. More robust

Current Opinion in Virology 2014, 9:143–147

144 Virus replication in animals and plants

Figure 1

Q

M

C

(a) (b) (c)

Current Opinion in Virology

Structures of viral quasispecies. Each line represents a genome, and

each circle a mutation. Once a wt genome starts replicating and prior to

the action of selection there will be an average of one mutation per

genome. Because mutations follow a Poisson distribution some

genomes will have no mutations and other will have multiple mutations.

The collection of genomes is the quasispecies (Q), which can be

described in terms of its consensus sequence (C) and its master

sequence (M). In population a the consensus and master sequences are

the same, but in populations b and c there is a difference between the

two. In population c all genomes carry at least one mutation and, thus,

the consensus sequence is not present in the population.

Figure 2

(a) (b) (c) (d)

Current Opinion in Virology

Structure of viral quasispecies with differences in robustness. Each line

represents a genome and each circle a mutation. Black indicates sites

that can mutate into a positively selected (beneficial) allele; gray

indicates sites that can mutate into a neutral allele; white represents a

site that can mutate into a negatively selected (deleterious) allele.

Populations a, b, and c have the same consensus sequence (wt) and

mutant frequency (and presumably mutation rate). However, compared

to population a, population b has a larger fraction of mutations that are

neutral and fewer beneficial and deleterious mutations, so it is more

robust. In contrast, population c has a larger fraction of mutations that

are deleterious and, thus, it is less robust. In addition to increasing the

neutral mutation rate, phenotypic stability can be increased by

decreasing the mutation rate, as shown with population d.

genomes have a higher neutral mutation rate compared to

less robust genomes (Figure 2). The mechanism under-

lying this process is elimination of genomes that are too

sensitive to the overall negative effects of mutation or, in

other words, selection for high robustness. This is often

referred to as ‘survival of the flattest’ [5].

John Holland and colleagues provided the first evidence

that less fit and more robust populations of vesicular

Table 1

Basic concepts of quasispecies theory

Term Definition

Consensus sequence A genomic sequence that has the most

common nucleotide at each position.

Error threshold The mutation rate at which a quasispecies

cannot be formed.

Mutation frequency The rate of mutants after selection and drift

have operated on a mutating population.

Mutation rate The rate at which new mutations appear per

nucleotide and round of replication.

Mutational meltdown Loss of meaningful genetic information at

extremely high mutation rates.

Quasispecies A complex population of genomes related

but non-identical that is selected for

robustness.

Current Opinion in Virology 2014, 9:143–147

stomatitis virus (VSV) may outcompete more fit and less

robust populations [6]. Additional evidence came from

work with phage f6 [7]. More recently we have demon-

strated that, as predicted by quasispecies theory,

increased selective pressure leads to overall increased

robustness in VSV, while relaxed selection allows the

generation of populations with lower overall robustness

[8�]. However, relaxed selection also leads to overall

fitness loss, as expected from the operation of Muller’s

ratchet (reviewed in Ref [4��]) and once fitness is low

enough other mechanisms might become important and

result in long-term robust populations. The work of

Cristina Escarmis and colleagues showing resistance to

extinction and development of unexpected phenotypes

in bottlenecked foot-and-mouth disease virus (FMDV)

populations illustrates this point [9,10].

One common criticism of the work we have just described

is that these are artificial systems that will never be

reproduced in real life. However, there is evidence that

the evolution of RNA viruses outside the laboratory is

such that natural populations are increasing in robustness

[11,12].

Host effects on polymerase fidelityOne of the most interesting features of viral RNA poly-

merases is the paradox that while variation in fidelity

www.sciencedirect.com

Selection of RNA virus fidelity Novella, Presloid and Taylor 145

among viral species can reach about two orders of mag-

nitude (10�4 to 10�6 substitutions per nucleotide copied

and round of replication [13]), for individual species a

change of 50% in fidelity can have major phenotypic

consequences [14]. Furthermore, evidence is accumulat-

ing that suggests that the fidelity of viral polymerases is

host dependent. Several plant RNA viruses, including the

common model cucumber mosaic virus (CMV), have

different mutant cloud sizes depending on the host plant

[15]. Similar results were obtained for animal viruses

[16,17]. Additional evidence comes from work in which

chikungunya virus (CHIKV) mutator strains selected in

mammalian cells did not show increased mutant fre-

quency in mosquito cells [18�]. We must point out that

this result has other possible explanations that the authors

discuss though.

Host enzymes may have a substantial effect on viral

fidelity and potentially drive viral evolution. One

example is the synthesis of APOBEC3G, a deaminase

that targets cytidine and results in retroviral hypermuta-

genesis and inhibition of replication [19–21]. The effect

of this enzyme is cell specific [22], but the viral Vif protein

is always the one that counteracts the effect. Adenosine

deaminase is another editing enzyme that restricts

measles virus and other paramyxoviruses [23]. Other

modulators of fidelity include the concentration and

balance of dNTP pools [24] and oxidative stress [25].

Selection of error rateRNA viruses have evolved to combine a genomic size and

mutation rate that are favorable. The Coronaviridaefamily, which includes the viruses with the largest

RNA genomes (up to 32 kb), is the only one with 30–50

exonuclease activity that provides higher accuracy

(reviewed in Ref [26]). Smaller viruses lack such correc-

tion mechanisms. Mutation rates constitute a limit to the

maximum genomic size because the larger the genome

the higher probability that replication will result in

multiple mutations so an error threshold is crossed. As

such, exonuclease mutants are highly susceptible to RNA

mutagens [27].

Many researchers claim that selection has resulted in high

mutations rates because that provides the most opportu-

nities for adaptation to an ever-changing world [4��].Other authors have argued that instead selection has

targeted speed of replication and high mutation rates

are a consequence [28,29]. As we will see in the next

section, there is some evidence that adaptability rather

than speed of replication is the most important feature.

Analyses of high-fidelity mutantsPoliovirus replication in the presence of mutagens results

in the selection of strain with high polymerase fidelity

[30,31]. Other high-fidelity picornaviruses can be isolated

in a similar fashion [32–34]. Phenotypic analysis of these

www.sciencedirect.com

mutants typically fails to show major defects in replication

so growth curves are virtually identical to those of wild

type (wt). However, in vivo infection results in dramatic

changes in fitness, as measured in competition exper-

iments, virus spread and virus titers, as well as pathogen-

esis. High fidelity picornaviruses are attenuated in mice

(reviewed in Refs [34–36]).

One important result that is not always sufficiently

emphasized is that a high-fidelity mutant that is subjected

to mutagenesis, such that the mutant cloud is extended to

resemble that of wt, attenuation is lost and neuroviru-

lence is restored [30]. This result is important because it

suggests that the speed of replication is not the target of

selection. High-fidelity mutants lower their mutation

rates because the polymerases are slower so there is a

better chance to reject a nucleotide that pairs improperly

[37,38]. The mutagenized population represents a virus

that still replicates more slowly than wt because it main-

tains its genetic change, but despite this biochemical

defect the population behaves as wt and is capable of

reaching high titers in all the relevant tissues.

Similar analyses using other viral species have yielded

results that are generally similar, but with some interest-

ing surprises. High-fidelity CHIKV [14] has, as expected,

decreased genetic diversity and no obvious replication

problems in mammalian cells. However, there is a small

but measurable fitness defect in mosquito cells. The

mutant is attenuated in mice and has substantial defects

during replication in mosquitoes. Based on results with

other systems described in this article, including polio-

virus, it is likely that these changes are the results of

changes in fidelity. However, it would be important to

test pleotropic effects of the mutation, for instance inter-

actions with interferon stimulated gene products or other

restriction factors that could explain these results. It is

particularly remarkable that these dramatic phenotypic

changes correspond to a difference in error rate of only

30%. Attenuation of high fidelity mutants and protection

against subsequent challenge with wt has resulted in the

proposal that these might be useful life-attenuated

vaccines [39].

Treatment of HIV-infected patients with nucleotide/

nucleoside analogs can result in the selection of high

fidelity variants (reviewed in Ref [40�]). There is a

correlation between fitness and fidelity [41] and in vivothese mutants do not replicate well, and tend to fix

compensatory mutations in the presence of the drug or

revert in its absence. It is possible that the replication

defect is the result of slow polymerase activity due to

variation in the population. The literature shows an

extensive correlation between population diversity and

development of AIDS ([42] and references therein) which

is consistent with (but does not prove) the role of variation

in disease progression.

Current Opinion in Virology 2014, 9:143–147

146 Virus replication in animals and plants

High fidelity viruses typically generate fewer RNA gen-

omes, but the RNA has a higher specific infectivity.

Generally they do not show major replicative defects in

cell culture, but they fail to replicate well, spread, and

cause disease in vivo. The limited data currently available

suggest that the mechanism underlying these changes invivo is insufficient occupancy of sequence space.

Analyses of low fidelity mutantsHolland and coworkers demonstrated long ago that the

mutation rates of viral RNA polymerases cannot be

increased much over wt values without massive loss of

virus titers [43] and these viruses are replicating at the

edge of error threshold. Not surprisingly, selection of

lower fidelity mutants results in changes that are rela-

tively minor, between 50% and approximately threefold

[18�,44��].

Mutator strains of RNA viruses replicate faster but have

lower specific infectivities. A mutator CHIKV showed an

interesting phenotype [18�]. Although the strain did not

show major replicative defects in mammalian cells, repli-

cation in mosquito cells was limited and RNA synthesis

was impaired. The virus was attenuated in mice and

replication was virtually eliminated in mosquitoes,

suggesting the occurrence of pleiotropic effects. A mutant

engineered in Sindbis virus [18�] showed a similar beha-

vior, except that some replication took place in mosqui-

toes. The replicative defect was not observed in

Drosophila cells, suggesting that this phenotype is mos-

quito-specific. Once again, lack of defect in cell culture

did not match the results in vivo, and the virus did not

replicate well in fruit flies.

Like their high fidelity counterparts, mutator strains are

attenuated in mice, even when they do not have major

replicative defects. Structural data of the coxackievirus

polymerase helped design and engineer a collection of

mutator strains that were analyzed [44��]. While some

variants had severe fitness defects others behaved as wt in

cell culture. Not surprisingly, however, the mutator

strains were attenuated in mice. Furthermore, there

was a direct correlation between fitness and mutation

frequency. Many of the strains failed to spread to target

organs and some were unable to establish persistent

infections. Interestingly, the authors were unable to gen-

erate stable high fidelity variants. This may be explained

by the low robustness that coxackievirus seems to have

compared to poliovirus, as shown by a higher sensitivity to

mutagenic treatment [45]. This particular example

demonstrates that other factors have an effect on and

are affected by polymerase fidelity. Once again it would

be important to test potential pleiotropic effects.

ConclusionsFor many years high polymerase error has been con-

sidered one of the main problems in the management

Current Opinion in Virology 2014, 9:143–147

of RNA virus infections because of it confers a potential

for rapid viral evolution. This potential translates in the

development of resistance to drugs and natural antiviral

mechanisms and in the possibility of species transmission

from zoonotic reservoirs. While these problems are real,

new research demonstrated how high error rates can be

the Achilles heel of RNA viruses and we can target

polymerase fidelity to develop new antiviral drugs or to

design new attenuated vaccines.

References and recommended readingPapers of particular interest, published within the period of review,have been highlighted as:

� of special interest�� of outstanding interest

1. Eigen M: Steps Towards Life. New York, NY: Oxford UniversityPress; 1992, .

2. Eigen M: On the nature of virus quasispecies. Trends Microbiol1996, 4:216-218.

3. Swetina J, Schuster P: Self-replication with errors. A modelfor polynucleotide replication. Biophys Chem 1982, 16:329-345.

4.��

Domingo E, Sheldon J, Perales C: Viral quasispecies evolution.Microbiol Mol Biol Rev 2012, 76:159-216.

This is an outstanding, in-depth review of viral quasispecies and thebiological consequences of high mutation rates and includes sectionsdevoted to HIV, HCV and HBV.

5. Wilke CO, Wang JL, Ofria C, Lenski RE, Adami C: Evolution ofdigital organisms at high mutation rates leads to survival ofthe flattest. Nature 2001, 412:331-333.

6. de la Torre JC, Holland JJ: RNA virus quasispecies populationscan suppress vastly superior mutant progeny. J Virol 1990,64:6278-6281.

7. Burch CL, Chao L: Evolvability of an RNA virus is determined byits mutational neighbourhood. Nature 2000, 406:625-628.

8.�

Novella IS, Presloid JB, Beech C, Wilke CO: Congruent evolutionof fitness and genetic robustness in vesicular stomatitis virus.J Virol 2013, 87:4923-4928.

This is the first report describing overall increased robustness in popula-tions under selection and overall decreased robustness in populations theevolve under random drift.

9. Escarmis C, Lazaro E, Manrubia SC: Population bottlenecks inquasispecies dynamics. Curr Top Microbiol Immunol 2006,299:141-170.

10. Escarmis C, Lazaro E, Arias A, Domingo E: Repeated bottlenecktransfers can lead to non-cytocidal forms of a cytopathicvirus: implications for viral extinction. J Mol Biol 2008, 376:367-379.

11. Rolland M, Brander C, Nickle DC, Herbeck JT, Gottlieb GS,Campbell MS, Maust BS, Mullins JI: HIV-1 over time: fitness lossor robustness gain? Nat Rev Microbiol 2007, 5:C1.

12. Koelle K, Cobey S, Grenfell B, Pascual M: Epochal evolutionshapes the phylodynamics of interpandemic influenza A(H3N2) in humans. Science 2006, 314:1898-1903.

13. Sanjuan R, Nebot MR, Chirico N, Mansky LM, Belshaw R: Viralmutation rates. J Virol 2010, 84:9733-9748.

14. Coffey LL, Beeharry Y, Borderia AV, Blanc H, Vignuzzi M:Arbovirus high fidelity variant loses fitness in mosquitoesand mice. Proc Natl Acad Sci U S A 2011, 108:16038-16043.

15. Roossinck MJ, Schneider WL: Mutant clouds and occupation ofsequence space in plant RNA viruses. Curr Top MicrobiolImmunol 2006, 299:337-348.

www.sciencedirect.com

Selection of RNA virus fidelity Novella, Presloid and Taylor 147

16. Jerzak GV, Bernard K, Kramer LD, Shi PY, Ebel GD: The West Nilevirus mutant spectrum is host-dependant and a determinantof mortality in mice. Virology 2007, 360:469-476.

17. Ciota AT, Koch EM, Willsey GG, Davis LJ, Jerzak GV, Ehrbar DJ,Wilke CO, Kramer LD: Temporal and spatial alterations inmutant swarm size of St. Louis encephalitis virus in mosquitohosts. Infect Genet Evol 2011, 11:460-468.

18.�

Rozen-Gagnon K, Stapleford KA, Mongelli V, Blanc H, Failloux AB,Saleh MC, Vignuzzi M: Alphavirus mutator variants presenthost-specific defects and attenuation in mammalian andinsect models. PLoS Pathog 2014, 10:e1003877.

A careful study demonstrating the phenotypic consequences, both in cellculture and in vivo, of increasing the mutation rates in two arboviruses.

19. Lecossier D, Bouchonnet F, Clavel F, Hance AJ: Hypermutationof HIV-1 DNA in the absence of the Vif protein. Science 2003,300:1112.

20. Mangeat B, Turelli P, Caron G, Friedli M, Perrin L, Trono D: Broadantiretroviral defence by human APOBEC3G through lethalediting of nascent reverse transcripts. Nature 2003, 424:99-103.

21. Boi S, Kolokithas A, Shepard J, Linwood R, Rosenke K, Van Dis E,Malik F, Evans LH: Incorporation of mouse APOBEC3 intomurine leukemia virus virions decreases the activity andfidelity of reverse transcriptase. J Virol 2014, 88:7659-7662.

22. Fourati S, Lambert-Niclot S, Soulie C, Wirden M, Malet I,Valantin MA, Tubiana R, Simon A, Katlama C, Carcelain G et al.:Differential impact of APOBEC3-driven mutagenesis on HIVevolution in diverse anatomical compartments. AIDS 2014,28:487-491.

23. Ward SV, George CX, Welch MJ, Liou LY, Hahm B, Lewicki H, de laTorre JC, Samuel CE, Oldstone MB: RNA editing enzymeadenosine deaminase is a restriction factor for controllingmeasles virus replication that also is required forembryogenesis. Proc Natl Acad Sci U S A 2011, 108:331-336.

24. Diamond TL, Roshal M, Jamburuthugoda VK, Reynolds HM,Merriam AR, Lee KY, Balakrishnan M, Bambara RA, Planelles V,Dewhurst S et al.: Macrophage tropism of HIV-1 depends onefficient cellular dNTP utilization by reverse transcriptase. JBiol Chem 2004, 279:51545-51553.

25. Seronello S, Montanez J, Presleigh K, Barlow M, Park SB, Choi J:Ethanol and reactive species increase basal sequenceheterogeneity of hepatitis C virus and produce variants withreduced susceptibility to antivirals. PLoS ONE 2011, 6:e27436.

26. Smith EC, Denison MR: Implications of altered replicationfidelity on the evolution and pathogenesis of coronaviruses.Curr Opin Virol 2012, 2:519-524.

27. Smith EC, Blanc H, Vignuzzi M, Denison MR: Coronaviruseslacking exoribonuclease activity are susceptible to lethalmutagenesis: evidence for proofreading and potentialtherapeutics. PLoS Pathog 2013, 9:e1003565.

28. Elena SF, Sanjuan R: Adaptive value of high mutation rates ofRNA viruses: separating causes from consequences. J Virol2005, 79:11555-11558.

29. Holmes EC: Error thresholds and the constraints to RNA virusevolution. Trends Microbiol 2003, 11:543-546.

30. Vignuzzi M, Stone JK, Arnold JJ, Cameron CE, Andino R:Quasispecies diversity determines pathogenesis throughcooperative interactions in a viral population. Nature 2006,439:344-348.

www.sciencedirect.com

31. Pfeiffer JK, Kirkegaard K: A single mutation in poliovirus RNA-dependent RNA polymerase confers resistance to mutagenicnucleotide analogs via increased fidelity. Proc Natl Acad Sci U SA 2003, 100:7289-7294.

32. Sierra M, Airaksinen A, Gonzalez-Lopez C, Agudo R, Arias A,Domingo E: Foot-and-mouth disease virus mutant withdecreased sensitivity to ribavirin: implications for errorcatastrophe. J Virol 2007, 81:2012-2024.

33. Ferrer-Orta C, Sierra M, Agudo R, de la Higuera I, Arias A, Perez-Luque R, Escarmis C, Domingo E, Verdaguer N: Structure of foot-and-mouth disease virus mutant polymerases with reducedsensitivity to ribavirin. J Virol 2010, 84:6188-6199.

34. Meng T, Kwang J: Attenuation of human enterovirus 71 high-replication-fidelity variants in AG129 mice. J Virol 2014,88:5803-5815.

35. Lauring AS, Andino R: Quasispecies theory and the behavior ofRNA viruses. PLoS Pathog 2010, 6:e1001005.

36. Lancaster KZ, Pfeiffer JK: Viral population dynamics andvirulence thresholds. Curr Opin Microbiol 2012, 15:525-530.

37. Arias A, Arnold JJ, Sierra M, Smidansky ED, Domingo E,Cameron CE: Determinants of RNA-dependent RNApolymerase (in)fidelity revealed by kinetic analysis of thepolymerase encoded by a foot-and-mouth disease virusmutant with reduced sensitivity to ribavirin. J Virol 2008,82:12346-12355.

38. Arnold JJ, Vignuzzi M, Stone JK, Andino R, Cameron CE: Remotesite control of an active site fidelity checkpoint in a viral RNA-dependent RNA polymerase. J Biol Chem 2005, 280:25706-25716.

39. Vignuzzi M, Wendt E, Andino R: Engineering attenuated virusvaccines by controlling replication fidelity. Nat Med 2008,14:154-161.

40.�

Lloyd SB, Kent SJ, Winnall WR: The high cost of fidelity. AIDSRes Hum Retrovirus 2014, 30:8-16.

A detailed description of the mutants selection for higher fidelity duringantiviral drug treatment and the tradeoffs they undergo in terms of fitness,generic variation and virulence.

41. Dapp MJ, Heineman RH, Mansky LM: Interrelationship betweenHIV-1 fitness and mutation rate. J Mol Biol 2013, 425:41-53.

42. Williamson S, Perry SM, Bustamante CD, Orive ME, Stearns MN,Kelly JK: A statistical characterization of consistent patterns ofhuman immunodeficiency virus evolution within infectedpatients. Mol Biol Evol 2005, 22:456-468.

43. Holland JJ, Domingo E, de la Torre JC, Steinhauer DA: Mutationfrequencies at defined single codon sites in vesicularstomatitis virus and poliovirus can be increased only slightlyby chemical mutagenesis. J Virol 1990, 64:3960-3962.

44.��

Gnadig NF, Beaucourt S, Campagnola G, Borderia AV, Sanz-Ramos M, Gong P, Blanc H, Peersen OB, Vignuzzi M:Coxsackievirus B3 mutator strains are attenuated in vivo. ProcNatl Acad Sci U S A 2012, 109:E2294-E2303.

Detailed study of polymerase mutants of coxackievirus that haveincreased mutation rates and are attenuated in the natural host to anextent that correlates with polymerase fidelity.

45. Graci JD, Gnadig NF, Galarraga JE, Castro C, Vignuzzi M,Cameron CE: Mutational robustness of an RNA virus influencessensitivity to lethal mutagenesis. J Virol 2012, 86:2869-2873.

Current Opinion in Virology 2014, 9:143–147