Despite enormous recent advances in molecular virology and immunology the pathobiology of chronic viral infections is still incompletely understood. In particular, the reason(s) some individuals clear virus, while others remain infected and ultimately develop disease is/are unknown. Many clinical features, including patient age and other demographic data, duration of disease 7, genetic background 8 metabolic factors 9 (including body mass index, glucose tolerance and iron overload) and so forth have some influence on the outcome of patients treated for HCV, though few of these factors are easily modified therapeutically. Significant research effort has therefore been directed at defining the principle genetic, biochemical and immunological characteristics of patients that predict either viral clearance or chronic viral infection in the hope these factors can be targeted therapeutically. Observations that stronger specific CD4/CD8 immune responses with CD4⁺ T-helper (T_H1) cytokine profiles, for example, are found more frequently in patients with self limiting viral infections than in those who develop chronic viral carriage 1011 has directed attention to the balance of CD4⁺ T_H1/T_H2 lymphocyte responses and resulted in attempts to enhance immune responsiveness therapeutically 1213, in the belief viral clearance will be enhanced long term by these therapies. This strategy has yet to prove beneficial.

Viruses, like other self-replicating molecules, are primarily concerned with producing more viruses. The survival of viruses, as obligate intracellular parasites, over a geological time scale, implies the development of strategies that allow them to effectively paralyse or to circumvent cellular defence mechanisms – including the innate immune system and those defences preventing cell entry – while maintaining those metabolic processes essential for viral replication; cell membrane integrity, cell homeostasis and the apparatus essential for protein synthesis. Long term, those viruses capable of inducing a chronic vegetative cellular state and subverting the cellular machinery necessary for their replication will be selected for, while those causing premature (that is, before viral replication is complete) lethal cell injury will not. Observed viral adaptation 14 and rapid development of antiviral drug resistance 5 over much shorter time-scales confirms their evolution is highly dynamic. As discussed previously 15, the ineluctable consequence of an RNA virus quasispecies is a protein quasispecies, and these proteins will possess a near-infinite spectrum of phenotypes, at least potentially. It is therefore entirely unsurprising that viruses with nucleic acid sequences coding for proteins that block apoptosis 16, interfere with the Toll-like receptors (TLRs) that mediate interferon signalling and expression of many other host genes 17 and interrupt other innate cell defense pathways, including interferon [IFN] regulatory factor [IRF], Janus activated kinase (Jak1), signal transducer and activator of transcription proteins 1 and 2 (STAT1/2), and inducible nitric oxide synthase [iNOS]) 18, protein kinase (PKR), and so forth, have been selected for and flourish. The nature of viral protein quasispecies make it entirely predictable that these viral anti-cell defence mechanisms would act at multiple levels against each pathway, as the actions of HCV proteins against interferon signalling confirm 192021, and in infinitely subtle ways against defence mechanisms as yet unidentified. While these viral defence mechanism(s) represent potentially important therapeutic targets – in particular the serine protease 21222324 – one would anticipate resistance to these therapies developing rapidly due to the multiple layers of redundancy of anti-cell-defence mechanisms viruses possess, as is seen with HIV and increasingly with HBV 5 therapy. Furthermore, it is entirely to be expected these resistant strains will be selected for by drug treatment.

At present, the only proven treatment for HCV is Interferon alpha (IFN-∝) (IFN-∝2a, IFN-∝2b and consensus interferon) or combination IFN-∝, now usually administered as long-acting pegylated 3 or albumin-conjugated 25 forms, with ribavirin, as combination therapy. Although the viral kinetics observed in individual patients may not be as clean as those represented schematically here (Figure 1), four main patterns of response are seen; i) Non-responders, in whom levels of HCV RNA appear completely unchanged by therapy, ii) the sustained viral response (SVR), defined by sustained absence of HCV RNA from serum long after therapy is withdrawn, iii) Relapsers, in whom virus is documented to be eradicated from serum during therapy, but in whom withdrawal of therapy results in recurrence and iv) a fourth group, often grouped with non-responders 26, but possibly better classified as partial responders; In these patients the HCV RNA clearly falls during treatment, sometimes to undetectable levels, indicating some response of virus to therapy, but rebounds back to pre-treatment levels (and sometimes beyond) despite adherence to ongoing therapy. This rebound in HCV RNA levels implies a compensatory viral homeostatic response – a potentially significant therapeutic target, as previously discussed 27 – suggesting differentiation of this group from true non-responders is conceptually important. The viral factors known to determine the likelihood of response to antiviral therapy include a) viral genotype b) viral load c) RNA quasispecies diversity and d) acute hepatitis.

A large number of clinical trials confirm that the single most important determinant of HCV clearance in response to treatment is viral genotype 28293031323334. Briefly, for all patients with HCV genotype 1 optimal combination pegylated interferon/ribavirin therapy for 48 weeks will result in a SVR of about 50% 335, while for genotype 2 about 90% of patients will clear virus long-term using the same treatment regimen for 24 weeks 363738. Patients with genotypes 2 or 3 respond more promptly and require shorter duration of treatment than patients infected with genotype 1 3339 or 4 40. As studies of the treatment of chronic viral hepatitis, by definition, are conducted in patients whose underlying innate, humoral and cellular (including T_H1/T_H2 profiles) antiviral responses do not permit spontaneous viral clearance, those patients with less effective immune responses – assuming, momentarily, normal population-based variations in those responses have any relevance whatsoever to whether or not viral clearance occurs – are preferentially selected and are over-represented in these trials.

Consider the following thought experiment: A large therapeutic trial is conducted in a population where the point-prevalence of genotypes 1 and 2 HCV is equal and no other genotypes are represented. If 2000 patients with chronic HCV are randomly selected, and treated with pegylated interferon/ribavirin for one year, at the completion of the study roughly 500/1000 patients with type 1 HCV will remain infected, while ~900 of the ~1000 patients with type 2 HCV will experience SVR. As the virus genotype with which a patient becomes infected is a random function of the point-prevalence of that genotype in the background population with which that individual interacts, and is completely independent of any underlying patient characteristic, the enormous differences in outcome seen in this trial will occur irrespective of any underlying human leukocyte antigen (HLA) type 41, cytokine polymorphisms 42, CD4⁺ T_H1/T_H2 lymphocyte responses 43, or any other genetic, metabolic or biochemical feature of the host, suggesting the genotype-specific outcome following treatment with interferon/ribavirin is almost purely a consequence of interaction(s) between the virus and the therapy, and that other factors are, comparatively, irrelevant for the vast majority of patients. As it is further likely that other viral factors such as pre-treatment viral load 3444546 and, possibly, quasispecies complexity and diversity present prior to therapy 474849 (vide infra) will account for at least some of patients with genotype 2 who do not experience SVR, it is clear viral factors rather than any host attributes, are the primary determinants of SVR and non-response. Why should the genotype of HCV (and probably other viruses 50) be important and determine response to antiviral therapy?

The primordial transition from simple chemical solutions to organic polymers through to the genesis of biological complexity and the origins of life as we now know it was contingent on the emergence of self-organizing and-self replicating molecules. Irrespective of whether Eigen 51 is correct that RNA was the original building block on which all subsequent biological complexity developed, the critical problem confronting self-replicating molecules, of any form, is replication, more specifically, replicative fidelity. Unless replication is near-perfect, accumulation of 'genetic' errors results in inexorable deterioration until all useful organization – "error catastrophe" – is lost. Eigen demonstrated that error catastrophe occurs if the product of the 'genome' size (N, strictly, the number of bits of information coded by the system) and the error rate of copying (ε, strictly, the rate information is lost during each round of replication) exceeds log S, where S is the selective advantage the replicative milieu imparts to error-free molecules over those containing errors, that is:

Nε < log S     Equation 1.

Meaning, if the rate information is lost during replication exceeds the rate at which it is concentrated by any selective advantage these molecules possess, error catastrophe occurs. Because the selective advantage of any single mutation is unlikely to be enormous, log S is unlikely to greatly exceed unity (1), therefore ε cannot be much larger than N^-1. That is, for a genome of 100 'bases' replicative stability requires any replicase (ribozyme, protein or whatever) cannot have less fidelity than 10^-2 errors/base synthesized, and molecules of greater length (as would be required to encode any biologically meaningful information) would need progressively more faithful polymerases to prevent genomic disorganization. Eigen suggested this problem could be circumvented by the emergence of molecular co-operation and hypercycles 5253. Clearly finding Eigen's hypercycles of co-operative RNA molecules conceptually troubling Niesert et al 54., dissected the hypercycle theory mathematically and demonstrated three other 'catastrophes' – selfish RNA (parasitism), hypercycle short circuiting and stochastic population collapse, any of which can result in molecular extinction – that beset populations of self replicating molecules, especially when molecular size and complexity increase. Another major problem exists; although greatly increased replicative fidelity is certainly necessary to ensure stable replication of biologically relevant macromolecules, it is insufficient; the problem is 'self'. By definition, self-replicating molecules need to replicate themselves, and not competitor molecules, hence mechanisms to distinguish 'self' from 'non-self' molecules are required, and this is increasingly problematic once polymer size and complexity increase and molecular co-operation and specialization is required and invoked; How does the replicase recognise which molecule to copy? Self-replicating molecules sail between the Scylla of lethal mutation and the Charybdis of population collapse 54 and, therefore, need to develop mechanism(s) that i) ensure adequate fidelity ii) allow differentiation of self from non-self molecules; that is, permit recognition of geno-'type' to ensure preferential self replication. iii) prevent molecular parasitism and selfish genetic replication (broadly, to ensure the 'genetic quality' of the molecules to be replicated) and iv) prevent stochastic population collapse. One imaginable mechanism – replicative homeostasis (RH) – links the functional output of the replicase epicyclically as both negative and positive feedback to modulate the replicase functions – processivity and fidelity (Figures 2, 3) – and in a manner that dictates effective replication requires co-operative interactions between multiple elements spatially separated on the genome. Replication contingent upon recognition of, and response to, complex three-dimensional complementarities between the polymerase and envelope proteins, and the polymerase and transcription initiation sequences of the 5'UTR of RNA molecules, constitutes a very sophisticated encryption technique that maximises the probability only 'self' (i.e. homotypic) molecules will be replicated, and effectively assays the functional integrity and quality of the entire viral genome, and its cognate proteins, as well as minimising the probability that either hypercycle short circuits or stochastic collapse occur. Once self-replicating molecules emerge, however, any replicative infidelity at all ensures multiple molecular species will arise, and these different species must compete for finite resources in the fitness landscape. As Spiegleman and Orgel suggested originally 55, molecules that develop reproductive strategies that maximise replication of "self" genes, while thwarting propagation of "rival" genes will proliferate at the expense of those that do not. Thus, the intense thermodynamically driven competition for survival causes self-replicating molecules to develop inexorably more complex, subtle and metabolically expensive strategies to ensure the genomes they are replicating (and with) are 'self', optimal quality (~consensus sequence) and fit, and to guard against rival genomes parasitizing or otherwise thwarting 'self'-replication. These incrementally more sophisticated strategies include acquisition of double RNA strands, DNA, protein-nucleic acid symbiosis, cell walls, receptor polymorphisms, motility, multicellularity, innate cellular defences, epistasis, lekking and other behaviours, language and money. Protein-nucleic acid symbiosis is clearly critical for HCV to function, but it does raise the question; Which of the competing self-replicating molecules, the polymerase or the RNA, conducts the HCV orchestra? Is it the RNA, directing synthesis of RNA_pol to produce more RNA, or is it, as seems more likely, the RNA polymerase that subtly manipulates and directs its RNA(s) to produce the viral shells necessary for production of more RNA_pol? In this light is it possible prions are just primitive RNA_pol or RNA_pol modulating proteins that simply highjack cellular RNAs coding for cellular RNA_pol cuckolding them into producing more prion protein?

The mechanism of RH has been described in detail previously 27 but, in brief, it is proposed to result from differential interactions of wild-type (wt) and variant (mt) envelope and envelope related proteins on RNA_pol in a series of feedback epicycles that link RNA_pol functions fidelity and processivity, RNA replication and viral protein synthesis, structure and function (Figures 2, 3), such that, in general terms, excess production of mutant envelope proteins, reflecting inadequate replicative fidelity, interact with RNA_pol to increase its fidelity and reduce processivity, while excess production of wild type (consensus sequence) envelope sequences, reflecting overly faithful replication (rendering the virus susceptible to immune-mediated clearance or destruction through attenuation and loss of replicative plasticity), interact with RNA_pol causing decreased fidelity. The ineluctable consequence of these interactions is the formation of highly stable, but reactive equilibria that permit viruses to respond rapidly to adverse changes to their conditions (e.g. immune recognition of dominant epitopes) and changing characteristics (e.g. evolving receptor polymorphisms) of their hosts.

Several independent lines of evidence strongly suggest that RH is mediated in HCV by interactions between the E2 protein, with probable contribution from P7 that likely 'fine-tunes' RNA_pol modulation in a manner similar to that proposed for HIVnef 56, and the interferon sensitivity region (ISDR) of NS5A and the thumb domain of the RNA-dependent RNA polymerase from NS5B. First, these regions are obviously important for genotype-specific virus related functions; HCV genotypes characteristically vary in length, with genotype 1 typically comprising 9030 to 9042 nucleotides, genotype 2 has 9099 and genotype 3 9063 nucleotides. The nucleotide insertions or deletions responsible for these genotype-specific differences are found within the E2 and NS5 portions of the genome 57. Second, while hypervariable regions within the E2 proteins evolve very rapidly 57, other regions of it are tightly genotype-constrained and contain several highly conserved amino acids and even in areas less obviously invariant, the amino acid substitutions appear non-random 585960. As we argued previously 61, this must indicate an important, and genotype-specific, viral function; different sequences of these regions are clearly adequate for virus-host interactions for other genotypes, hence these genotype-specific regions of E2 must have interact with other viral structures. Third, and while this is controversial 62, evidence suggests sequence variability in both the NS5A ISDR 63 and NS5B 64 and the HCV E2 65 is predictive of both response to interferon and viral load 64. Fourth, in chimpanzees persistent HCV infection develops

only

As it relates to control of HCV RNA quasispecies, RH is primarily a mechanism regulating transcription. However, accessory proteins known to alter the processivity and fidelity of cellular DNA-dependent DNA polymerases 72 and DNA-dependent RNA polymerase 73, as well as viral RNA-depended RNA and DNA polymerases (as reviewed 27), and the cellular reverse transcriptase, telomerase 74, are strongly conserved in evolution, implying that ability to vary the fidelity and processivity with which RNA and DNA are synthesized is a critical and normal cellular function. The advantages inherent to an ability to vary RNA sequence and hence protein function or immunogenicity are obvious, while those conferred by modulating DNA sequences are less so, at least for coding regions (in preparation). The cellular analogs of viral functions controlled by RH – replication, generation of antigenic diversity (envelope structure), quasispecies expansion and so forth – may therefore be modulated by similar mechanisms, and mediated by ancestrally-related molecules, acting on cellular polymerases to exert control over cellular quasispecies (e.g. the liver cells) by modulating cell division, differentiation and expression of cell-surface proteins (Figure 4). The common ancestral origins and possible structural similarities of the proteins that mediate RH in viruses and those that modulate polymerase functions in cells suggests an obvious mechanism by which the envelope proteins of HCV, an RNA virus incapable of integration within genomic DNA, HBsAg and other similar viral envelope proteins, might trigger malignant transformation in hepatocytes to cause HCC (in preparation), by interference with these accessory polymerase molecules, thus destabilizing cellular polymerases. Could it be malignancy is an incidental consequence of the competition between self-replicating molecules?

It is well established that patients with low initial viral load, commonly defined as <800,000 IU ml^-1 26, respond better to therapy than those with high viral loads 3444546. For example, when patients with HCV genotype 1 were treated with pegylated interferon 2∝ (PEG-IFN∝) and ribavirin for 48 weeks 3 56% of those with low baseline viraemia, defined in that study as <2 × 10⁶ copies ml^-1, developed SVR compared to 41% of those with high base-line viraemia, defined as >2 × 10⁶ copies ml^-1. These viral load-dependent differential responses are seen across genotypes 339 and have been reported in both treatment naïve patients and in patients receiving PEG-IFN∝ plus ribavirin (RBV) after failing treatment with combination standard IFN∝ plus RBV or IFN∝ monotherapy 75. Similarly, high genetic and quasispecies diversity predicts poorer outcome with therapy 4749.

However, the influence of initial viral load on response to therapy is extremely unlikely to be due to high absolute concentrations of virus per se and is much more likely due to the underlying reason(s) why viral load is high in the first place; After all, the distinction between a high viral load (>8 × 10⁵ IU ml^-1) and a low viral load (<8 × 10⁵ IU ml^-1) is entirely arbitrary, differentiation between responders and non-responders is not defined by any specific viral load, and the absolute difference between high and low levels is completely inconsequential when compared to the 4–5 log fall in HCV RNA that will occur if therapy is effective. What determines viral load?

Replicative homeostasis predicts, in general, that viral load will be high in patients with viral strains that have high affinity Env_wt:RNA_pol interactions that RH postulates increase polymerase processivity (while reducing fidelity), thus resulting in a high set-point of the replicative equilibrium. Patients with these viral strains, therefore, will have – and by the same mechanism – both high-level basal viral replication and increased quasispecies complexity and diversity compared to strains where Env_wt:RNA_pol interactions are less high-affinity. As RH predicts that mutant viral envelope proteins (Env_mt) normally act to decrease RNA_pol processivity (and increase fidelity), while interferon or/and interferon-induced endogenous cellular effector proteins like PKR act to reduce viral replication by interference with normal Env_wt:RNA_pol interactions, thus reducing RNA_pol processivity, and that these interactions with RNA_pol occur at the same binding site(s), probably the thumb domain of nonstructural (NS) region NS5B and the interferon sensitivity-determining region (ISDR) in NS5A. Replicative homeostasis predicts that those virus strains with poor prognostic characteristics (high basal load and broad quasispecies diversity) will be resistant to therapy because exogenously administered inhibitors (IFN) or/and their effector molecules like protein kinase (PKR) are less easily able to disrupt the high affinity Env_wt:RNA_pol interactions that cause these adverse viral characteristics to develop. As a corollary, RH would predict mutations within this region would reduce the affinity of any Env_wt:RNA_pol interactions, thus reducing viral load, and rendering them more susceptible to interferon therapy, a hypothesis confirmed experimentally 7677. Furthermore, and by the same mechanism, RH predicts some mutations involving Env:RNA_pol interaction sites may actually increase viral replication, thus explaining the usually transient increase in HCV RNA levels observed in some therapeutic trials 767879. This phenomenon has been reproduced in-vitro for HIV 8081, Semliki Forest virus 82 and probably HBV 83, where mutations within envelope sequences have been demonstrated to cause increased viral replication, presumably due to abnormal Env:RNA_pol interactions, as previously discussed 27. Finally, a relationship between viral load and HCV genotypes has been reported by some 8485 (and was apparent with small numbers in an early publication 34, (but not all 86) workers further suggesting the mechanism(s) that maintain genotype and determine viral load may be linked, as RH implies.

In other words, the adverse prognosis patients with high viral load and quasispecies diversity experience with currently available treatments is an innate consequence of the replicative equilibrium intrinsic to those viral strains, rather than absolute viral load or level of quasispecies diversity per se. Of course, as a secondary consequence, the associated broad antigenic diversity generated by the replicative equilibrium of these viral strains will also impair immune-mediated clearance of infected cells.

Accurate diagnostic tests for HCV have dramatically improved the safety of transfusion medicine and caused acute HCV to be an increasingly uncommon clinical entity, however, when it does occur, chronic viral persistence and liver disease develops in 50–80% of patients 26. Treatment of acute HCV is characterised by very high response rates and rapid clearance of HCV RNA from serum 878889, with Jaeckel et al., reporting a 98% end of treatment response (not SVR) and clearance of HCV RNA from serum by 3.2 weeks when treatment was begun, on average, 89 days after infection occurred 87. This same group (although slightly different patient cohort) reporting an SVR of 89% after 24 weeks off therapy 88, a significantly better outcome than one might predict considering the distribution of genotypes treated.

Although some of this apparent improvement in outcome with treatment of acute HCV is due to the ancient artefact of immediate intervention (tacitly acknowledged by Santantonio et al., 89) that physicians have benefited from for millennia – if the natural history of disease X is that 50% of acute cases resolve spontaneously, and the other 50% go on to develop chronic disease, and treatment Y cures 50% of chronic cases but has no impact on whether acute disease resolves, then administration of Y to all cases in the acute phase will result in "cure" of 75% of cases overall, of which only 25% are properly attributable to Y, with the other 50% expected anyway – there seems little doubt treatment of acute HCV does actually improve outcomes. The reasons for this are not clear, but one hypothesis put forward is that during acute HCV the virus has not yet fully deployed its impressive anti-cell-defence arsenal that Gale and co-workers 2090919293 and others 71 have so elegantly dissected. While this explanation certainly sounds plausible, (though it does raise the obvious question: "If so, why can't the innate cellular mechanisms clear virus during this window of opportunity?") and is possibly even true, it is not supported by viral kinetic data; During acute infection HCV viraemia, and therefore viral RNA and protein concentrations, rise rapidly to peak by about week 12 at about 10^7–8 copies HCV RNA ml^-1, then fall by ~2 logs to 10⁵ copies HCV RNA ml^-1 long-term (Figure 5) 9495. Therefore, during acute infection (e.g. point A, Figure 5), the viral anti-cell-defence molecules (presumably mostly proteins, but possibly also ribozymes and small interfering RNAs) are present in concentrations at least 2 logs higher than seen during chronic infection (e.g. points B to C, Figure 5) suggesting, in fact, that treatment of HCV during the acute phase should be less effective treatment of chronic infection. It obviously isn't. Of course, it may be that the "quality" of these viral anti-cell defence molecules isn't optimal during acute disease (Why?) or that they haven't had time to neutralize innate cellular defences (How long does cleavage of Toll-like receptor 3 by HCV NS3/4A proteases etc., take?), or some other reason, but the "failed cell neutralization" hypothesis, as it relates to acute infection at least, is unsupported by current data. A different explanation might be that the RNA polymerase is unstable early (it is certainly highly processive, and as argued previously, replicating less faithfully than it does long-term 15), because the Env:RNApol interactions RH predicts have not yet evolved to mature stability, thus rendering the replicative equilibria more susceptible to agents like interferon and ribavirin that RH postulates to act by further destabilizing them. Parenthetically, whatever the correct explanation, the apparently enhanced efficacy of therapy for acute HCV is strong evidence that the absolute level of viraemia, per se, is not an important determinant of treatment outcome, as suggested above.

Replicative homeostasis provides a rational mechanistic basis for the actions of both ribavirin and interferon that explains many phenomena observed during treatment of HCV, including the genotype-specific differences in response rates, and the adverse impact high pre-treatment viral load and quasispecies diversity has on treatment outcome. It also provides a rational explanation for other previously unexplained phenomena like rebound of HCV RNA levels seen during treatment with both interferon and ribavirin, and the transient increased viremia seen in some patients receiving ribavirin, and occasionally, interferon. Replicative homeostasis also suggests an explanation for the apparent immunostimulatory effects of ribavirin (and, by extension, other nucleos(t)ides) and interferon, and thus, elucidates the enhanced phase 2 clearance of HCV RNA seen during treatment ribavirin. Ockham's razor would suggest this mechanism makes it unnecessary to postulate a direct immunostimulatory mechanism for ribavirin, other nucleosides or interferon (though, clearly, interferon has other direct effects on innate intracellular responses), although we do not suggest such action(s) is/are excluded.

It is worth considering the likely characteristics of the envelope-polymerase interactions postulated to mediate RH and their implications for drug therapy. First, by definition, RH is a mechanism by which, in part, viral genotype is maintained by genotype-specific envelope-polymerase interactions. These interactions, therefore, will be virus genotype and probably virus sub-species-specific, hence therapeutic vaccines and other ligands capable of interaction with the viral polymerase and/or envelope at their RMEs and disrupting normal Env:RNA_pol interactions will probably be highly genotype-specific and, compared to nucleos(t)ide analogues, have relative lack of toxicity with a high therapeutic index. However, as a consequence of this specificity, their optimal use may require viral genotyping. Second, the reactions postulated to mediate RH result in profound inhibition of viral replication, at least initially during acute viral infection, making it likely therapies targeting this homeostatic function of viruses effectively will be extremely potent. Third, as discussed above, it is possible drugs or therapeutic vaccines that disrupt viral RH and reduce RNA_pol processivity and increase its fidelity, will restrict viral protein diversity and cause increased cellular expression of dominant viral epitopes beyond TIR, increasing the likelihood of an effective and neutralizing immune response and therefore the likelihood of permenant viral clearance. Fourth, and implicit in the above discussion, the nature of RH implies that the putative Env:RNA_pol RMEs must be highly conserved; any conformation other than this would result in the polymerase unable to recognise and distinguish between wild-type and mutant envelope motifs, rendering the virus unable to recognise, and respond to, deleterious unbalanced accumulation of excess mutant or wild-type virus, as we have suggested previously 61. It is therefore possible that viral resistance to drugs or therapeutic vaccines targeting these interactions would develop slowly, if at all. Finally, recombinant polypeptides are relatively cheap to produce, and their infrequent administration as therapeutic vaccines – as opposed to complex daily regimens of nucleoside analogues and protease inhibitors – is relatively simple and, hence, more likely to encourage compliance, an important therapeutic consideration for all patients, but especially relevant to third world populations.