Puzzles, challenges, and information reservoir of SARS-CoV-2 quasispecies

ABSTRACT

Upon the emergence of SARS-CoV-2 in the human population, it was conjectured that for this coronavirus the dynamic intra-host heterogeneity typical of RNA viruses would be toned down. Nothing of this sort is observed. Here we review the main observations on the complexity and diverse composition of SARS-CoV-2 mutant spectra sampled from infected patients, within the framework of quasispecies dynamics. The analyses suggest that the information provided by myriads of genomic sequences within infected individuals may have a predictive value of the genomic sequences that acquire epidemiological relevance. Possibilities to reconcile the presence of broad mutant spectra in the large RNA coronavirus genome with its encoding a 3’ to 5’ exonuclease proofreading-repair activity are considered. Indeterminations in the behavior of individual viral genomes provide a benefit for the survival of the ensemble. We propose that this concept falls in the domain of “stochastic thinking,” a notion that applies also to cellular processes, as a means for biological systems to face unexpected needs.

INTRODUCTION

Populations of replicating RNA viruses are composed of multitudes of variant genomes that individually have a transient existence. Such dynamic complexity is the result of average mutation rates of around 10⁻⁴ mutations introduced per nucleotide copied. This error rate is about 10⁵-fold higher than that exhibited by complex DNA viruses whose replication errors can be corrected by proofreading-repair activities and post-replicative DNA repair pathways. Other factors that may contribute to mutational input and population complexity of RNA viruses are cellular editing activities, large viral population sizes, rapid viral genome replication, and active virus particle turnover (production and degradation) in infected hosts. Additional variation-promoting factors, acting on dissimilar genomes of the same ensemble, are positive and negative selection, as well as population bottlenecks, all acting at virtually any step of virus infectious and transmission cycles. The heterogeneous dynamic genome collectivities are referred to as viral quasispecies. This term recognizes the seminal conceptual breakthrough that was opened by quasispecies theory in its mathematical description of the process of genome replication with the regular production of error copies. In quasispecies theory, mutant distributions represent a key ingredient for the self-organization and adaptability of simple replicons in the critical events that led to life’s origin (1–3).

Although the genomes of all biological forms undergo modifications, their consequences have a major and immediate impact on viruses. This is due to the combined weight of high error rate and reduced genome size, together with the small number of mutations (one or very few) that can alter some of their relevant phenotypic traits [i.e., cell receptor recognition, interaction with host factors, sensitivity to antibodies, etc., reviewed in (4)]. The viral mutant distributions (also termed mutant spectra, clouds-, or swarms)—which are parallel to those portrayed by quasispecies theory—have proven key to virus adaptability and survival. They are also the substrate for the expression of new viral phenotypes that emerge as a result of interactions that are established among components of a mutant spectrum through their expressed proteins. One of the consequences of such interactions is that replicative fitness (the capacity to produce progeny in a given environment) can be lower for viral populations amplified from individual biological clones from a mutant ensemble than for the parental mutant ensemble, both adjusted to comparable population size [examples described or reviewed in (5–12)]. The genome composition of an RNA virus population can only be defined statistically, which entails an indetermination in assigning a genomic nucleotide sequence to a virus behavior. This applies to each of the viral populations which replicate at different sites within an infected host.

The consensus (or average) sequence of a population of viral genomes (or their encoded proteins) includes at each position the residue (nucleotide or amino acid) that is most represented at that position in the sequences under study. It is the one generally recorded in data banks, and it is obtained by procedures that are not sensitive to differences among individual genomes that coexist in the same population. Detection of such differences can be achieved either by sequencing biological or molecular clones amplified from the sample under study, or by ultra-deep sequencing methods, using diverse sample preparation protocols and bioinformatics pipelines. These procedures have consistently indicated extreme complexity of the RNA viral populations that replicate in their hosts and that, therefore, participate in inter-host transmission that may culminate in epidemiological spread. Intra-host virus variation and inter-host virus evolution offer complementary views of the dynamic plasticity of viruses, and of its consequences for viral pathogenesis, and for the design of disease control measures (7, 13).

Briefly, the main concepts pertinent to quasispecies dynamics of viral populations can be summarized as follows: (i) Mutant spectra are a rich phenotypic reservoir, and virus adaptation is achieved via the replacement of dominant viral subpopulations by others. (ii) Quasispecies dynamics conditions the efficacy of antiviral interventions. Vaccines have to be multiepitopic (antigenically complex), and antiviral agents have to be administered in combination. This applies to current anti-COVID-19 vaccines (and vaccines intended for protection against antigenically variable RNA viruses) that prevent disease but not infection. For these viruses, vaccine breakthrough infections have been described, and the need to update the antigenic composition of vaccines is amply recognized. (iii) Quasispecies theory has inspired the exploitation of high error rates to establish lethal mutagenesis (extinction of a virus by an externally induced excess of mutations) as a broad-spectrum antiviral approach. Crossing an infectivity threshold to achieve viral extinction is the virological parallel with crossing the error threshold to lead to loss of inheritable information, as defined in quasispecies theory. Several licensed antiviral agents (some of them used to treat COVID-19) act partly by lethal mutagenesis. (iv) The degree of adaptation of a virus to a given environment is quantified by viral fitness, which can influence the extent of cellular modifications evoked by a virus during infection, and it is also a determinant of virus resistance to antiviral agents. The frequency of individual genomes within a mutant spectrum is (at least in part) determined by their relative fitness in that environment. (v) Interactions among components of a mutant spectrum (positive interactions of complementation or cooperation, or negative interactions of interference) may influence the behavior or the ensemble, rendering virus collectivities units of selection. Viral populations harbor the double potential of being either a substrate for individual selection or for group selection. Several recent publications (and previous articles quoted in them) include the experimental results on which the five points listed above have been established (11, 12, 14, 15). Based on these precedents, we examine the scenario that has been opened with the emergence of SARS-CoV-2 in the human population, and with the need to control COVID-19.

VIRAL DISEASE EMERGENCE UNDER THE QUASISPECIES FRAMEWORK

The adaptive capacity of the microbial world in general is due to several molecular mechanisms of genome variation: mutation, hypermutation, recombination (homologous, non-homologous, replicative, non-replicative), genome segment reassortment, and several modes of gene transfer (conjugation, fusion, assimilation, infection, transduction, transformation); some of them are shared by viral and cellular pathogens (16–18). These inherent sources of genome (and genome complement) instability are responsible of the coexistence of repertoires of variant genomes of any given pathogen. Environmental heterogeneities, fluctuations, and unidirectional alterations, act as selective sieves to promote multiplication of some variant forms in detriment of (or preferentially to) others. These events are facilitated by ecological modifications associated with climate change, as well as by adverse social, economic, and political circumstances. Inherent adaptability and alteration of viral traffic are major ingredients of viral survival and persistence. Our global world is prone to infectious disease emergence and reemergence (19–21). For viruses, the efficacy of such triggering factors has been evidenced by an average of about one viral disease emergence or reemergence per year over the last century. Emergences have been most dramatically exemplified by the 1918 influenza pandemic, the in-between-centuries AIDS pandemic, and, recently, by the COVID-19 pandemic. Together, these three viral diseases have taken the lives of about 100 million people.

A review of the literature shows that, curiously, in the beginning of their recognition as new pathogens, emergent human viruses are very often assumed to display limited genetic variation. This has also happened with SARS-CoV-2. The intra-host heterogeneity of this new human coronavirus is still presented as limited (22, 23), despite evidence to the contrary, as summarized in the present article. To understand and confront the challenges posed by RNA virus genome variation, it is counterproductive to ignore quasispecies on the grounds that it is a theoretical concept disconnected from the reality of viruses as disease agents. Notorious historical misconceptions, due to lack of focus on quasispecies implications (despite the latter having been explained already in the early 1980s), include predictions that creating vaccine based on the expression of the influenza hemagglutinin and neuraminidase subtypes matching those of the circulating virus would provide “a la carte” vaccines for each influenza season; or that antiviral vaccines formulated with a single peptide antigen with the amino acid sequence of an epitope at a virus surface could be effective in evoking solid protection, among other examples [reviewed in (4)]. An understanding of viral quasispecies dynamics makes us aware of the fact that complexity (of the viruses to be controlled) cannot be effectively combatted by simplicity (single antiviral agents, monoclonal antibodies, or peptide antigens). In this context, complexity refers not only to the multiple forms of a given microbial pathogen but also to the ecological and sociological factors that boost disease emergence in rather unpredictable ways.

In late twentieth century, opposition to viral quasispecies (perhaps because of incredulity of the high mutation rates it portrayed, or because it was perceived incorrectly as redundant with polymorphisms studied by classic genetics, or as rivalling the virus “species” concept used in virus taxonomy) was expressed by attributing the large number of mutations detected in viral population to amplification artifacts and sequencing errors [see (7)]. Arguments in favor of ignoring mutant spectra, and to focus the study of viral genome evolution on the consensus sequence of isolates, also include that very few of the mutations in minority genomes are relevant for virus survival or pathogenesis, and that, therefore, few are likely to reach epidemiological relevance. In that view, there is no use in worrying about mutations (and other genome modifications) since they occur frequently, randomly, and uncontrollably. None of the justifications to elude the analysis of mutant spectrum composition are sound and backed by evidence. This is documented with different viruses in multiple scenarios where mutant spectrum dissection has proven to be of theoretical and practical relevance in several facets of medical practice (11, 12, 14, 15, 24–28). A strong case in support of enquiring about mutant spectrum composition is being provided by SARS-CoV-2, paradoxically a coronavirus with the a priory tag of showing limited genetic variation.

HIGH COMPLEXITY AND INFORMATION POTENTIAL OF SARS-CoV-2 MUTANT SPECTRA

The inference of limited intra-host variation of SARS-CoV-2 (22, 23) was probably advanced because this coronavirus encodes a 3’ to 5’ exonuclease (ExoN) in protein nsp14 which should reduce the final error rate of the polymerase complex, if it behaved as described for other coronaviruses, notably SARS-CoV and murine hepatitis virus (29–32). To what extent ExoN reduces the error rate of SARS-CoV-2 is not known. Despite a possible reduction, ultra-deep sequencing analyses of SARS-CoV-2 isolates from different patient cohorts have documented the presence of very complex mutant clouds. We next examine two complementary aspects of SARS-CoV-2 mutant spectra: complexity and composition.

To measure SARS-CoV-2 mutant spectrum complexity, we adapted an experimental and bioinformatics pipeline to examine in depth the presence of point mutations and deletions in virus from nasopharyngeal samples of patients diagnosed with COVID-19 of different severity (33). The procedure allowed a comparison of the number of point mutations and deletions that were detected when the mutation or deletion frequency cut-off was computationally lowered from 20% (lowest resolution similar to that attained with molecular cloning and Sanger sequencing) to 0.1% [maximum resolution reached due to the high clean read coverage (average 110,074 clean reads per sample; range 89,201–129,807) with our experimental and bioinformatics procedures]. The reliability of the mutations detected with the 0.1% mutation frequency cut-off deduced from this coverage is critical for the quantification of diversity, and, in addition to clean read coverage, it was substantiated by the following observations: (i) The majority of mutations recorded with a 0.5% mutation frequency cut-off were also detected with the 0.1% cut-off. (ii) Mutation type preferences, and mapping of mutations in the first, second, or third codon position were statistically indistinguishable using the two cut-off values. (iii) The percentage of amino acid substitutions recorded in mutant spectra, and that are also represented in the consensus sequence of epidemiologically distant isolates, was very similar with the two cut-off values. Observations (i) to (iii) require that the majority of mutations scored with the 0.1% mutation cut-off are bona fide mutations present in viral RNA template molecules, and not random mutations artifactually introduced during the amplification and sequencing procedures [data and further details in (34)]. Such high-resolution mutant spectrum analysis has been used to determine the frequency of mutations, deletions, and deduced haplotypes, as well as to calculate several diversity indices whose combined values provide a quantification of quasispecies complexity (35–37). The procedure has been applied to diagnostic samples of SARS-CoV-2 from successive COVID-19 waves, as well as from cases of vaccine breakthrough COVID-19 (33, 34, 38).

The main conclusions of these studies are: (i) Upon lowering the mutation frequency cut-off from 0.5% to 0.1%, the number of different mutations [counted separately for patients with mild, moderate or severe disease (34)] detected in the mutant spectra increased 55-fold in the nsp12 (polymerase)-coding region, and 97-fold in the S (spike)-coding region. This is a remarkable increase that contrasts with the increase in the number of deletions that amounted to 5-fold and 4-fold for the nsp12 (polymerase)-and the spike (S)-coding region, respectively. (ii) SARS-CoV-2 quasispecies are mainly populated by mutations that are present within the 0.10%-0.49% frequency range; on average, 98.7% of the total number of mutations is found in that range (Fig. 1). The origin and biological relevance of the immense pool of low-frequency variant genomes are intriguing questions. The big repertoire of low-frequency mutations may belong to low-fitness viable genomes, or to defective genomes which are maintained in the population by complementation. With the advent of high-profundity deep sequencing, the distinction between a mutation (or group of mutations) causing lethality versus a profound fitness decrease (but whose genome does not reach fitness zero) is difficult to resolve (7). (iii) In the mutant spectra of SARS-CoV-2 examined so far there is a deficit of intermediate frequency mutations, as compared with mutant spectra from hepatitis C virus (HCV) from serum samples of chronically infected patients, which were analyzed using comparable methodology (34). (iv) In the complex SARS-CoV-2 mutant spectra there is an overwhelming dominance of transition over transversion mutations [for the number of different mutations, the transition to transversion ratio is 104 and 210 in the nsp12 (polymerase)-and the spike (S)-coding region, respectively (34)]. The preferred mutations are U → C > A → G > C → U, which is the same ranking than calculated with the 0.5% mutation frequency cut-off. The mutation types and the residues that are neighbors of the mutation sites (at least those that have been analyzed in the viral RNA of the diagnostic samples of our patient cohort) suggest that cellular editing activities APOBEC and ADAR did not play a major role in mutation introduction (39–43).

Fig 1.

The expanded capacity to detect minority mutations using high resolution ultra-deep sequencing. The scheme depicts the relative number of different mutations (number duplicated symmetrically for clarity in the bottom axis) with increasing mutant frequency cut-off (percentage given on the left and right sides of the scheme). The relative number of mutations in the nsp12-coding region is given by the left-most and right-most borders of the blue rectangles; the corresponding number in the spike-coding region is given by the left-most and right-most borders of the orange rectangles. The relative number of mutations is drawn with a scale according to the quantification of different mutations considering the entire patient cohort described in (34). The red arrows highlight the exponential increase in the number of mutations.

All indications are that quasispecies dynamics applies to SARS-CoV-2, with the qualification that the average fitness-decreasing effect of individual mutations may be enhanced due to the large genome size, in agreement with the inverse relationship between tolerated mutation rate and genome size (for a given superiority of the master sequence over its accompanying mutant spectrum) predicted by quasispecies theory (1–4, 7–9). This may explain the abundance of low-frequency mutations, although confirmation of this suggestion requires additional work. Regarding the assignment of a phenotype to a precise genomic sequence, the presence of myriads of low-frequency genomes introduces for SARS-CoV-2 the same indeterminations that have been previously evidenced with other RNA viruses; such viruses stand near a borderline between reproducible behavior and unpredictable phenotypic shifts.

These indeterminations are aggravated by the fact that the biological effect of a specific mutation (or amino acid substitution) may depend on the sequence context; for example, SARS-CoV-2 spike (S) amino acid substitutions Q498R and N501Y were more active in enhancing the binding to the cellular receptor angiotensin-converting enzyme 2 (ACE2) when they were located in the context of the S sequence of Omicron than in the context of S sequence of previous variants (44). Two or more mutations that fall on the same genome may exert no mutual influence, or may produce positive or negative epistasis (enhancement or diminution of replication, or of the efficiency of another step in the infectious cycle). A selected genome may hitchhike mutations that are not relevant to the response to the active selective constraint, but that may exert an influence on other phenotypic traits. Virus heterogeneity and dynamics accentuates the tremendous lottery that means that a particular mutant invades a new tissue or organ, or that it is transmitted to an uninfected host to either pursue or abort further replication.

The presence of a low-frequency but broad mutant cloud also raises the question of how the suppressive effects exerted by the mutant spectra on the replication of genomes, including high-fitness genomes hidden at low frequency within a mutant spectrum (45–48), will manifest for SARS-CoV-2. This question is pertinent to the expression of potentially valuable phenotypes encoded in low-frequency genomes. This is an issue that is likely to acquire relevance if antiviral treatments to combat severe COVID-19 are widely implemented. Comparative clinical and experimental evolution research will be needed to quantify the kinetics of selection of SARS-CoV-2 drug resistant mutants, and compare the values with those accumulated with the studies on HIV-1 and HCV treatments.

Regarding composition, SARS-CoV-2 mutant spectra may contain mutations (and clusters of mutations) that are typical of distant clades of the same virus [ (38, 49), and unpublished results]. The generation and selection of genomes with a precise cluster of mutations was described in an early study with poliovirus (50). Minority amino acids, which are not a signature of the clade to which the isolates belong, are present in the consensus sequences of SARS-CoV-2 from other (precedent or posterior) clades (33, 34, 38, 49, 51, 52). The observation has been made with standard isolates during the pandemic, as well as with virus from fully vaccinated individuals who nevertheless were infected with the Alpha variant in April 2021, and developed COVID-19. The vaccine-breakthrough viruses included amino acid substitutions which are a signature of Delta Plus, Iota, and Omicron variants, and not of the infecting Alpha variant (38). Mutations of virus within infected individuals which lead to residues that are also found in the consensus sequences of distant clades probably reflect the restricted regions of sequence space that SARS-CoV-2 mutant spectrum can occupy in its natural environment (see next section on “The space of the possible: an unexpected resource?”).

The availability of millions of consensus sequences of SARS-CoV-2 from successive epidemiological waves is unprecedented in molecular virology (i.e., NCBI, ENA, GISAID data banks). The analysis of such large data repositories is currently being compared with the sequences identified in mutant spectra from individual patients (Martínez-González, García-Crespo et al., unpublished results). The information retrieved to date is hinting at the intriguing possibility that mutant spectra may harbor information on the types of amino acid substitutions that are likely to reach epidemiological dominance at later times. The challenge is to quantify the connections between mutant spectrum compositions and the increasing number of consensus sequence repertoires contributed to data banks.

THE SPACE OF THE POSSIBLE: AN UNEXPLOITED RESOURCE?

The theoretical sequence space available to any virus is a number that defies our imagination (2). In the case of SARS-CoV-2 with a genome length of about 29,900 nucleotides, considering only point mutations, the theoretical sequence space is of 29,900⁴ different genomes (four being the number of nucleotides at any position; possible rare, nonstandard nucleotides, are ignored); 29,900⁴ is 7.99 × 10¹⁷, an unimaginably vast number. The theoretical sequence space is further expanded by the presence of deletions and insertions. However, the actual sequence space in which the virus may actually exist is far more limited. The first constraint that restricts the occupation of the theoretical sequence space is the genome organization itself. Regulatory regions, open reading frames, and encoded proteins all have a range of functional sequences that, in addition, must be coordinated to produce viable molecular solutions to yield infectious progeny. According to measurements with other RNA viruses, the number of lethal or highly detrimental mutations out of a number introduced randomly in a viral genome by site-directed mutagenesis may approximate 40% (53, 54).

Even accepting an ambiguity in the distinction between a severe fitness decreasing effect and strict lethality of mutations (fitness zero of the genomes carrying them), the space of tolerated point mutations in a viral genome remains enormous. Additional movements in sequence space are driven by deletions (and less frequently by additions) of widely different sizes. Deletions are present in consensus sequences (relative to the sequence of other isolates) or within mutant spectra (relative to the consensus sequence of the corresponding isolate). What the increased penetration into the mutant spectrum of SARS-CoV-2 isolates is showing is a repetition of a limited number of haplotypes [at least in the amplicons of the nsp12- and S-coding regions that we have analyzed (33, 34)], relative to what a free exploration of sequence space might have anticipated. Although additional comparisons between mutant spectra and consensus sequences are needed, the results suggest that mutant spectra may play the role of a watchtower of permitted genomic sequences. Such information may allow advanced computational procedures to predict which genomes may be more likely to become epidemiologically relevant at later stages of the pandemic.

THE POSSIBLE PARTICIPATION OF ExoN IN MUTANT SPECTRUM CONSTRUCTION

A question that arises from the ultra-deep sequencing analyses is whether complex mutant spectra can be generated during replication of a virus that expresses an active proofreading-repair activity. A major reason to anticipate that SARS-CoV-2 would display limited intra-host genetic variation (22, 23) is the presence in its protein nsp14 of an ExoN (3’ to 5’ exonuclease) activity. Yet, it is now quite evident that this virus is particularly adept in mutating, and it does so with functional consequences that reach even public health strategies. This has been illustrated by recent debates and decisions on the need to update the composition of anti-COVID-19 vaccines, as traditionally done for prevention of other RNA viral diseases such as the human influenza or the animal foot-and-mouth disease.

We have suggested several possibilities to explain how broad mutant spectra can develop despite the presence of an ExoN domain in nsp14 (55). They are the following: (i) Despite its activity in vitro (56–58), the ExoN may not be active (or less active than the ExoN of other coronaviruses) in infected cells. (ii) The error rate is only one of several parameters that participate in the display of a mutant spectrum at any point in time; other relevant values include the number of viral RNA molecules that contribute to the replication complexes, number of replication rounds, number of infected cells, viral yield per cell, or the participation of cellular editing activities in the mutational input (55, 59). Depending on such replicative features, an active ExoN may have a limited influence in the mutant spectrum broadness. More work is needed to calibrate how several interacting influences can end up in complex mutant spectra, and in elevated rates of evolution of consensus sequences in an epidemiological setting.

THE EXPANDING REALMS OF STOCHASTIC THINKING

Random events with biological consequences abound at all stages of the infectious cycle of viruses. Of particular relevance for the concepts discussed in the present article is the stochastic occurrence of mutations and recombination events (60, 61), which provide broad repertoires of mutant and recombinant genomes, as well as of intracellular recombination intermediates. The major mechanism by which specific mutant and recombinant genomes make their way to subsequent intracellular and extracellular replication rounds is selection, which is guided by the relative fitness of competing entities in the environment where the competition takes place (61, 62). Random genetic drift of genomes, due mainly to population bottlenecks of different intensity, can modify the outcomes that would be dictated based exclusively on the operation of negative and positive selection.

In this scenario, high mutation rates and quasispecies dynamics introduce an uncertainty in virus composition and behavior. The two major reasons are that the exact composition in genomic types of a viral population is unknown, and that any individual genome characterized by a nucleotide sequence has a transient existence (although we do not know the average half-life of the genomes). The quantum mechanical indeterminations that underlie the origin of mutations [due to variations in the electronic distributions in the bases of nucleotides (63)] are translated into biological uncertainties.

Molecular stochasticity with biological consequences has also been documented in cell biology. A known case is provided by parallel features of viral quasispecies with cancer cell heterogeneity and dynamics. Tumors [mainly those known as “hot” tumors due to the diversity of cellular types they contain (64)] consist of complex aggregates of phenotypically different cells that share some level of uncontrolled multiplication. Genetic, epigenetic, and stochastic factors fuel tumor heterogeneity in rather unpredictable ways, with consequences for cancer treatment failure (65). The basis of tumor adaptability is due to the coexistence (and unpredictable generation) of nonidentical cell types that may accommodate to different intra-host environments. Stochasticity plays a key role in tumor development and spread, just as observed for viruses expanding in infected hosts.

Despite not having a genetic basis, conformation heterogeneity of prions (proteinaceous infectious particles) offers another example of minority phenotypes in whose occurrence stochastic events most likely play a role. The large ensembles of a pathogenic prion protein (PrP) may include minority conformations─amidst the dominant conformation─that may be subjected to Darwinian selection, evidenced when the ensemble is transferred to another environment (66–70). The connection with the quasispecies concept is quite straightforward, with conformation playing the role of mutation in driving a phenotypic transition. These studies have documented selection of PrP conformation types as a host range determinant. Prions have been defined also as “proteinaceous nucleating particles” (71), and nucleation may trigger the conformational shift of the protein aggregate. Monitoring of single PrP fibrils detected the growth of fibril subpopulations that were hidden in the parental PrP ensemble (69). There is a quasispecies of structural isomorphs that can adapt to (cause pathology in) new hosts, as a parallel to individual genomes in a viral mutant spectrum that shows a phenotype that is different from the phenotype depicted by the viral population ensemble (4). Again, stochasticity plays a key role in the initial event of a conformational change.

Examination of single molecules carrying out their biological function has revealed additional examples of functional differences among individual molecules of an ensemble. Several studies have reported heterogeneity in catalytic rates of individual enzyme molecules due to several mechanisms [quoted and explained in (72)]. A study of individual molecules of the RecBCD helicase of Escherichia coli (a helicase/nuclease involved in the repair of double stranded DNA breaks via homologous recombination) showed variation in the DNA unwinding rates of the individual molecules. The unwinding reaction could be interrupted by depleting a required ligand from the reaction mixture. Upon reintroduction of the ligand, unwinding resumed at a rate that was different from that exhibited originally by the same molecule. The results suggest that a conformational change occurred during the unwinding interruption that lasted during the subsequent processive unwinding of the molecule (72), a form of conformation selection, with parallels to the observations with the prion protein PrP. Biochemical heterogeneity can have different origins (conformational fluctuations of a protein, presence of different stable sub-states in equilibrium, etc.). The changes in RecBCD were stochastic, as are the mutations that arise in a replicating viral RNA population. In both cases, the heterogeneity confers molecular plasticity to respond to environmental changes and unpredictable needs.

In a subsequent study, Kowalczykowski and colleagues reached parallel conclusions regarding the possible requirement of coordination of the rates of leading- and lagging-strand DNA synthesis in the replisome complex (macromolecular aggregate that replicates DNA) (73). Evidence was presented that the stochastic behavior of replisome components ensures complete DNA duplication without a requirement of coordination of leading- and lagging-strand synthesis. Again, a stochastic sampling from a distribution of different rates was key to completion of a biochemical event.

In viral quasispecies, as well as in critical cellular processes (such as cancer development or DNA synthesis) and protein-mediated pathological states (such as prion disease), stochastic events with a genetic basis (mutation-driven) or a biophysical basis (protein conformation states-driven) can procure a distribution of related elements to cope with environmental needs and unexpected events. Randomness, not only tightly controlled regulation, appears to be critical for survival. The so-called “stochastic thinking” becomes increasingly important to interpret molecular virology.