Introduction

The existence of multipartite viruses, i.e. viruses with a genome consisting of several segments, each packaged in a different viral particle, is intriguing because they must find a way to keep their genomic information together during within-host colonization and between-host transmission. In addition, no clear general advantage specific to this genomic architecture has yet been identified1,2. A series of studies on faba bean necrotic stunt virus (FBNSV; family Nanoviridae, genus Nanovirus) aiming to understand how multipartite viruses function and potentially their existence revealed three surprising features: a genome formula, a multicellular lifestyle, and the possibility of non-concomitant transmission of genomic segments.

The genome of FBNSV consists of eight separately encapsidated circular single-stranded (ss) DNA segments, each of about 1 kb and encoding a single gene. The different segments of FBNSV do not accumulate at equal frequencies within hosts3. Instead, they reproducibly accumulate at a frequency distribution where some segments largely dominate while others are rare in an active infection3. This frequency distribution, termed genome formula (GF), is host-specific3, impacts gene expression within hosts4, and differs from that in insect vectors5. The fact that the GF changes ‘immediately’ upon host species switch suggests that it may represent a mechanism allowing rapid adaptation of the virus to variation in its environment. This notion was theoretically investigated and confirmed6, even though the adaptive nature of the genome formula has yet to be experimentally confirmed (but see further discussion below). The non-uniform frequency distribution nevertheless adds to the issue of genome integrity because it is a priori easier to lose rare segments.

FBNSV can produce functional infections even when its genomic segments do not cooccur within individual host cells; in fact, more often than not they do not. Instead, gene products traffic between host cells such that proteins produced by one segment often accumulate in cells containing other segments but where their cognate segment is missing. This viral system is thus complemented and functional at a multicellular scale7.

Lastly, the different segments of FBNSV do not need to be transmitted together but can be acquired and inoculated separately8. This can occur when different insect vector individuals acquire distinct segment subsets from different host individuals and together inoculate them in the same host. The complete genome can then be reconstituted in the recipient host. This can also occur when the same vector individual sequentially acquires distinct segment subsets from different hosts and inoculates them all at once in the recipient host. In this latter case, the complete genome is reconstituted in the vector.

Thus, the maintenance of genome integrity is challenged in FBNSV by the non-uniform GF, but this challenge is mitigated by the fact that genome integrity is not mandatory either at the individual cell level or during between-host transmission. Below, we rapidly review evidence on the generality of these intriguing processes, i.e. whether they occur in other multipartite viruses, in the monopartite segmented viruses, which package their genomic segments together within each viral particle, or even in monopartite nonsegmented viruses9.

Multipartite viruses

Among these processes, the one that has been most extensively investigated so far, perhaps because of its earlier discovery, is the GF (Table 1). Several studies on multipartite viruses, although still limited in number, cover a rather diverse set of viruses with respect to the type of nucleic acid bearing the genetic information (ssDNA, (+)ssRNA, and (−)ssRNA). It should be noted, however, that most of these studies bear on ssDNA viruses while the majority of multipartite viruses are RNA viruses1,10.

Table 1 Viruses for which the existence of a genome formula and its potential variation were investigated
Full size table

A first trend that appears from these studies is that with few exceptions, the GF corresponds to a non-uniform distribution of segment frequencies: as in FBNSV, some segments are rare, and others are frequent. The few cases of equifrequency have so far been observed exclusively in the genus Begomovirus11,12 (family Geminiviridae), some members of which are bipartite ssDNA viruses with each of their two segments encoding more than one gene. This observed equifrequency cannot be considered a characteristic of this genus since one begomovirus with a 5–6 ratio of its segments’ frequency has been identified11 while two other begomoviruses with equifrequency in the plant host show different segment ratios in their vector12. Moreover, equifrequency cannot be viewed as a feature of bipartite viruses either since the bipartite ssDNA bidensoviruses (family Bidnaviridae, genus Bidnavirus) exhibit a GF with unequal segment frequencies13. Finally, equifrequency cannot be explained by the fact that the distinct segments encode several genes, as the bidensoviruses13 and tripartite ssRNA positive sense (+) bromoviruses14,15,16,17 (family Bromoviridae, genus Bromovirus) also have segments which carry more than one ORF and a GF with unequal segment frequencies.

As mentioned above, GF was hypothesized to represent a mechanism enabling multipartite viruses to rapidly adjust their gene expression to environmental heterogeneity through gene copy number variation3,6. In agreement with this hypothesis, almost all studies asking whether GF varies across ‘environments’ found that it does. This is the case whether the ‘environment’ corresponds to different hosts, a passage from host to insect vector, different organs within an individual host, or coinfection with other genomic entities such as satellites (Table 1). The only exception is ssRNA(+) tripartite cucumber mosaic virus (CMV), a bromovirus, the mean GF of which does not seem to change across host plant species17 nor in the presence of coinfecting viruses18, though in the latter case, coinfection significantly decreases the GF variance. The latter study interestingly discusses that this lack of effect could be due to the quantification technique used. Indeed, in ssRNA(+) viruses, genomic and messenger RNAs are sometimes co-detected and it can thus be tricky (when possible at all) to distinguish gene copy number from gene expression variation. This is relevant because, as these authors note and further explained in the next paragraph, gene expression was shown to be much more similar across host species than GF in FBNSV4, the only virus where this comparison was explicitly performed.

The study by Gallet et al.4 showed that the GF indeed has functional implications as gene copy number variation correlates with mRNA variation in hosts of two species: plants of each species with a higher copy number of a gene also have higher titers of the corresponding mRNA, indicating that GF indeed affects gene expression4. Perhaps counterintuitively, the GF differed much more between hosts than the corresponding formulae based on mRNA distributions. This arose because the slopes of the mRNA-to-DNA relationships differed across segments and for some segments across hosts. These differing slopes indicate different mRNA/gene copy ratios, something also observed in other studies on ssDNA viruses11,19,20,21. Therefore, gene expression regulation may arise from processes occurring at two levels. One is acting at the individual segment level and is linked to segment-specific regulatory sequences. For example, the promotor strength or mRNA stability can depend on the segment4,21, the host species4, or even the organ19, leading to different mRNA/gene copy ratios. The second level is acting at the population of segments level through the observed gene copy number variation resulting in the differing GF. These two regulation levels must not be confused and are complementary.

The adaptive nature of the GF, although intuitively and theoretically appealing, remains yet to be demonstrated. No study has unequivocally shown that GF variation affects even a fitness component. In FBNSV, virus titer increases as the infection converges to setpoint GF, i.e. the GF at equilibrium, but this relationship was not statistically confirmed3. The only other virus where the question was investigated is a bromovirus, ssRNA (+) alfalfa mosaic virus (AMV)15. It was shown that total RNA titer but not encapsidated RNA was maximized at the setpoint GF pointing at a possible increased expression but no detectable higher viral accumulation.

These and many other questions beg for further investigation. For example, how does the GF arise? A recent study showed that processes at the individual segment and at the group of segment levels concur to establish the GF of FBNSV in a host plant22, but a lot remains to be elucidated on the mechanisms leading to the GF. It would also be important to study how the GF interacts with host physiology and resistance/tolerance mechanisms. Finally, it would be worth investigating whether there is polymorphism within viruses for their GF reaction norms which would suggest, assuming such polymorphism has a genetic basis, that GF can evolve.

Monopartite segmented viruses

It is intuitive to think that monopartite segmented viruses, because they package their segments in the same virion and because some of them are known to have specific packaging mechanisms, should package one copy of each segment in each viral particle and thus have equal frequencies for all their segments within hosts. These intuitions, however, proved wrong, at least in a number of cases and on several accounts. First, the fact that a virus possesses specific packaging mechanisms does not preclude the production of viral particles that do not contain all genomic segments. This is the case, for example, for the influenza A virus (IAV; family Orthomyxoviridae, genus Alphainfluenzavirus), which is known to package its eight genomic segments using specific packaging mechanisms in a “7 + 1” configuration23,24. Yet, this virus has been shown to produce very large proportions of ‘semi-infectious particles’25 or SIPs, which are particles unable to express at least one viral protein (for a review, see Diefenbacher and colleagues26). Several mechanisms could give rise to SIPs, and the lack of genomic segments has been demonstrated to be one of them27. A detailed investigation revealed that, on average, an IAV segment has a 58% probability of being replicated in a host cell when introduced by a single virion, this probability varying from 50% to 67% depending on the segment28. These probabilities, along with other evidence, imply that coinfection, allowing complementation within host cells, is common in IAV28,29,30. Such observations led the authors of some of the above-cited publications to hypothesize that SIP production could be a way to allow IAV to modulate its gene expression in a manner analogous to the GF. This hypothesis has been supported further by the observation that a mutation in the nucleoprotein (NP) segment of an IAV strain strongly decreased the packaging, genomic accumulation, and gene expression of the neuramidinase (NA) segment, thus modifying the IAV GF31. Intriguingly, presumably through complementation, while this mutation increases the production of SIPs, it also enhances viral load. This study provides another example where GF modification appears linked to increased viral fitness, though here, as well, the potential causal relation between GF modification and increased fitness remains to be demonstrated31. As a final note on IAV, we would like to remark that even though the very definition of SIPs implies the non-expression of specific genes in individual cells and that coinfections by complementing SIPs are very common within hosts, the hypothesis that IAV could potentially function in a multicellular way of life, at least in hosts where multiplicity of infection would be low, has not been experimentally evaluated and thus remains a possibility. Indeed, it was recently shown that IAV stimulates the formation of cell-connecting tunneling nanotubes which it can use for cell-to-cell transfer of genome segments32,33,34 and viral proteins32,33. Beyond the relevance of this cell-to-cell trafficking for immune evasion, it was shown that through this mechanism, complementation and reassortment do not necessitate coinfection at the individual host cell level and that genome segments can move across host cells in a virion-free way34. Thus, the features characterizing a multicellular way of life, i.e. the ability to complement viral functions across different host cells, are present in this monopartite segmented virus. What needs to be further characterized is the extent to which such complementation may involve exclusively trafficking of gene products and not that of genome components. Indeed, in the latter case, infections become cell-autonomous immediately after the transfer of the relevant genome segments.

Similarly to IAV, dsRNA orbiviruses in the order Reovirales also possess highly specific packaging mechanisms (reviewed by Borodavka and colleagues35). Nevertheless, a study on bluetongue virus (BTV) remarkably revealed not only that the 10 genomic segments accumulate at markedly different frequencies within hosts but, moreover, that they produce different GFs in hosts and vectors36.

It is perhaps less surprising that GFs were also found in viruses that are known for their less specific packaging such as members of the order Bunyavirales37,38. Indeed, a GF was found not only in multipartite rice stripe virus (RSV; family Phenuiviridae, genus Tenuivirus)39 but also in monopartite segmented tomato spotted wilt virus (TSWV; family Tospoviridae, genus Orthotospovirus)40 and Rift Valley fever virus (RVFV; family Phenuiviridae, genus Phlebovirus)37,41. The latter studies showed that bunyaviruses do not have a mechanism ensuring the incorporation of a complete set of segments in a single viral particle. They thus produce many viral particles, perhaps the majority, lacking genome segments. Most viruses of this order replicate in both hosts and vectors. A comparison between those that use plants as hosts and those that use animals could thus potentially help elucidate the still unresolved question of why multipartite viruses almost exclusively infect plants and fungi. Three examples illustrate this statement: (i) RVFV infects mammals and is transmitted by mosquitoes. It thus replicates solely in animals where it packages several segments together and clearly fits in the category of monopartite segmented viruses. Its packaging has been shown to be more efficient in insect (vector) cells than in mammalian (host) cells, resulting in a larger proportion of viral particles containing all genomic segments42. (ii) RSV infects plants and is transmitted by insect vectors. It does not form enveloped viral particles but propagates in both host plants and insect vectors in the form of three separate ribonucleoprotein filaments, each representing a genome segment. It fits in the category of multipartite viruses but, nevertheless, its GF appears more constant over time in the insect vector than in the host plant39. (iii) Finally, as a sort of intermediate stage, TSWV also infects plants and is transmitted by insects, but it does form enveloped viral particles. Interestingly, mutants lacking the viral envelope were nevertheless able to infect plants43, suggesting that the virus propagates within host plants as nucleoprotein filaments. A subsequent study showed that envelope-deficient mutants were unable to infect the vector midgut and vector primary cell cultures44, suggesting that the virus enters and propagates within insect vectors as fully assembled viral particles packaging several segments together. A definitive confirmation of such functioning would imply that TSWV adopts a multipartite lifestyle within the plant and switches to a monopartite segmented lifestyle within the insect vector. Further investigation of multipartite and monopartite segmented bunyaviruses transmitted by insects and infecting plants could reveal steps in the viral life cycle more permissive, and perhaps conducive, to multipartitism in plants and monopartite segmentation in insects.

Finally, a recent study on RVFV showed that complementation between viral particles containing incomplete genomes occurs beyond the within-cell level41. This study showed that coinfecting viral particles with incomplete genomes could complement each other and be successfully acquired by vectors in which they can produce viral particles containing genomes. Thus, particles with incomplete genomes may significantly contribute to the between-host transmission of these monopartite segmented viruses reminiscent of the non-concomitant transmission observed in the multipartite FBNSV8.

Monopartite nonsegmented viruses

The sections above discuss several features associated with the multipartite architecture of viral genomes and how these may be extended to at least some monopartite segmented viruses. While much more speculative at this stage, we wish to briefly mention potential analogies in monopartite nonsegmented viruses, in particular with respect to the possibility of gene expression regulation through gene copy number differential modification, and the mechanisms facilitating the maintenance of genomic integrity.

The question of the GF and gene copy number variation is an interesting prospect for monopartite nonsegmented viruses. Gene copy number variation has thus far been considered and studied solely through gene amplification and gene accordion processes45,46. In this case, a gene can be duplicated several times, up to tens of copies, within a monopartite nonsegmented genome, inducing the adaptive overexpression of the amplified gene and the selection of the corresponding genome variant. This process appears possible only in large dsDNA viruses, which have viral particles that can accommodate important genome length variations. Logically, gene copy number variation through gene amplification is considered impossible in small ssDNA or RNA monopartite nonsegmented viruses where the genome length is highly constrained within small icosahedral particles46. However, the production and accumulation of myriads of diverse defective genomes (DGs) in all virus infections47, each DG type containing only a subset of the viral genes, could result in variation of gene copy numbers and so variation in gene expression. To our knowledge, a situation where the production and regulation of the DG population would engender a phenomenon analogous to the GF in monopartite nonsegmented viruses has never been formally investigated. Nevertheless, it has been argued that DG-related variation of gene copy number may, at least in some cases, provide advantages to the viruses26,48,49,50. It is thus possible to imagine that even some monopartite nonsegmented viruses could generate a GF during cell and host infection. In this speculation, a possible though not exclusive mechanism allowing the maintenance of genome integrity could be the preferential packaging and transmission of non-defective full-length genomes linked to genome length constraints. In any case, if it were true that monopartite nonsegmented viruses can as well generate GFs through the extemporaneous production of DGs, this would indicate that any potential advantage incurred by GFs is not the driving force leading to genome segmentation.

In addition to DGs that are generated from the viral genome, monopartite nonsegmented viruses are sometimes associated with satellites51. Both satellites and DGs can be beneficial47, and they need complementation by the helper virus. The maintenance of these successful associations may profit from analogous mechanisms as those revealed for FBNSV: the within hosts multicellular way of life, and the non-concomitant between-hosts transmission of helper virus and satellite. While there is no theoretical hindrance to these mechanisms, none have been explicitly envisaged and investigated thus far in a monopartite nonsegmented viral system.

Altogether, the studies discussed above suggest that the frontier between multipartite, monopartite segmented, and perhaps even monopartite nonsegmented viruses may be much more flexible than anticipated. Even though the specific manifestations of phenomena such as GFs, non-concomitant gene transmission, and multicellular way of life, may depend on the viral packaging strategies, their essence in allowing flexible and rapidly modulable gene expression and complementation beyond the single host cell level could be much more shared than previously thought. This possibility opens avenues to future research, making it perhaps necessary to reconsider the extent of multicomponent viral systems in the virus world on how they function and evolve.