Cheating in the viral world

The success of many viruses depends upon cooperative interac- 1 tions between viral genomes. For example, viruses that coin- 2 fect the same cell can share essential gene products, such as 3 replicase, the enzyme that replicates the viral genome. How- 4 ever, when cooperation occurs, there is the potential for ‘cheats’ 5 to exploit that cooperation. We suggest that: (1) the biology 6 of viruses makes viral cooperation particularly susceptible to 7 cheating; (2) cheats are common across a wide range of viruses, 8 including viral entities that are already well studied, such as de- 9 fective interfering genomes, and satellite viruses. Consequently, 10 evolutionary theory developed to explain cheating offers a con- 11 ceptual framework for understanding and manipulating viral 12 dynamics. At the same time, viruses offer unique opportunities 13 to study how cheats evolve, because cheating is relatively com- 14 mon in viruses, compared with taxa where cooperation is more 15 usually studied, such as animals. 16


20
The existence of cooperation opens the door to cheats, which 21 exploit that cooperation 1,2 . Cooperation can be observed at  evolutionary theory tell us how viral cheats will evolve and 53 when they will spread? And can cheats be exploited or ma-54 nipulated to help control viral infections? 55 We synthesise the relevant evolutionary and virology liter-56 atures, showing how similar issues have been examined in 57 these two fields, but from very different perspectives.  ing these bodies of work together offers novel insights, both 59 for how cheating evolves, and for understanding viral dy-60 namics. Specifically, we: (1) define cheating, and provide 61 a detailed example of how this definition can be applied to 62 viruses; (2) identify and classify the different kinds of cheat-63 ing in the viral world; (3) discuss why cheating might be es-64 pecially common in viruses; (4) examine how the prevalence 65 of cheating in viruses poses novel problems for evolutionary 66 theory; (5) apply ideas from social evolution theory, both to 67 understand the evolution of viral cheats, and to assess their 68 Box 1: How do Viruses Cooperate? From an evolutionary perspective, cooperation is when a trait performed by one individual provides a benefit to another individual, and has evolved at least partly because of this benefit 9 . When multiple viral genomes infect the same host cell, there is an opportunity for cooperation between viral genomes 8 . This is because gene products encoded by one viral genome can produce a benefit that goes to all of the viral genomes inside the cell. For example, the replicase produced by one viral genome, to replicate that genome or its progeny genomes, can also replicate other (unrelated) viral genomes inside the same host cell (Fig. 1). Capsid proteins, required for building the capsid that transports viral progeny to new host cells, can produce a similar shared benefit, as can proteins that suppress the host cell's immune response, or indeed any viral gene product that provides a benefit to all of the viral genomes inside an infected cell (Fig. 1). In some cases, viral cooperation can also extend beyond the cell, to include cases where benefits are shared between viral genomes infecting different cells. One example is when viruses suppress the release of interferon from host cells 55 . This suppression is costly to the viral genome encoding the gene for suppression, but it provides a public benefit by keeping the local population of host cells susceptible to infection by neighbouring viruses. Evolutionary biologists sometimes use the term 'public goods' to refer to factors that provide a cooperative benefit to the local group of individuals 9 . Famous examples include iron-scavenging molecules produced by bacteria, acorns stored by cooperative woodpeckers, or humans contributing to some group task, such as hunting 5 . Virologists use the term 'trans-complementable' in an analogous way, to describe when gene products are shared between different viral genomes in the same cell 48 .

Fig. 1. Cooperation and cheating in viruses.
In viruses, cooperation can occur when gene products produced by one viral genome provide a benefit to a different viral genome. When this occurs, the gene can be called 'trans-complementable' and the gene product a 'public good'. (a) The replicase enzyme replicates the viral genome to produce more copies. Replicases encoded by one genome can potentially replicate other viral genomes, including those that did not encode the replicase. A viral genome that is shorter, such as a cheat, is likely to be replicated more quickly than the longer cooperator virus genome. (b) Viral capsids are required to transport viral progeny to new cells. Capsids produced by one viral genome can be used by other viral genomes, including those that did not produce capsid proteins. Cheat genomes can be more likely to be incorporated inside capsids than full-length cooperator genomes 21 . Fig. 2. Cheating throughout the natural world. Cheating occurs throughout the natural world, including in viruses: (1) the common cuckoo (Cuculus canorus) lays eggs in other birds' nests, here tricking a reed warbler (Acrocephalus scirpaceus) into taking care of a much larger cuckoo chick 107 ; (2) Noncooperative cheat individuals of the bacterial pathogen Pseudomonas aeruginosa (labelled in white) exploit iron-scavenging molecules produced by cooperators (labelled in green); (3) in Vesicular Stomatitis Virus (VSV), when a defective interfering cheat genome (labelled in green) is grown in a mixed infection with wild-type VSV (labelled in red), the defective interfering genome exploits replicase proteins encoded by the wild-type cooperator, resulting in a colony (a) that is dominated by the defective interfering genome, and grows less effectively than a colony consisting just of the cooperative wild-type (b) 108 ; (4) in cucumber mosaic virus (CMV) infections, a cheat satellite (satCMV) exploits gene products encoded by the wild-type cooperator, substantially reducing the overall viral load and leading to less severe infections in plants infected by both cheat and cooperator (a) compared to plants infected by just the cooperative wild-type virus (b) 109 . species into neglecting their own chicks and instead feeding  For two individuals to count as a cheat and cooperator respectively, three conditions must be met: (1) the cooperator must have a higher fitness than the cheat when each are alone; (2) the cheat must have a higher fitness than the cooperator when both are mixed; (3) the mixture containing both cheat and cooperator must have a lower fitness than when the cooperator was alone. 1). In contrast, a genome that does not encode replicase 108 represents a potential cheat, because it could exploit repli-109 cases produced by other genomes, without producing a repli-110 case itself. If these two genomes do represent a cooperator 111 and a cheat, then we would expect to see the three results 112 given above: (i) when grown on their own, the replicase-113 encoding genome (cooperator) will have a higher growth rate 114 than the genome that does not encode replicase (cheat); (ii) 115 the genome that does not encode replicase (cheat) will have a 116 higher growth rate than the replicase-encoding genome (co-117 operator) when both are grown together; (iii) the mixture 118 containing both genomes will have a lower growth rate than 119 when the replicase-encoding genome (cooperator) is grown 120 on its own (Fig. 3).

122
The mutant variant 'DI PV1' is a cheat of poliovirus. DI PV1 123 contains a large deletion that removes the entire capsid pro-124 tein region (Fig. 4) 20,21 . Therefore, when grown on its own, 125 DI PV1 produces no viral capsids, and so is unable to spread 126 between host cells. However, when wild-type poliovirus and 127 Fig. 4. DI PV1 is a model viral cheat. DI PV1, a defective interfering genome, is a cheat of poliovirus. (i) DI PV1 lacks the section of genome that encodes capsid proteins, resulting in a substantially shorter genome than the cooperative wild-type. (ii) Consequently, DI PV1 gains more than a 1,000-fold replication advantage over the wild-type cooperator when both coinfect the same cell (adapted from 21 Fig. 1).
DI PV1 are grown together, copies of DI PV1 can be incorpo-128 rated into viral capsids produced by the wild-type cooperator.

129
In coinfected cells, the shorter length of DI PV1 means that it 130 is replicated substantially faster than the wild-type, and it is 131 also able to enter virions more effectively than the wild-type.

132
Consequently, DI PV1 is able to achieve more than 1,000 133 times as many genomes inside viral capsids as the wild-type The success of a cheat depends upon its ability to interact 154 with and exploit cooperators. In viruses, this depends primar-155 ily on two factors related to population structure: the number 156 of viral genomes that infect each host cell; and the extent to 157 which these viral genomes originate from different cells. Vi-158 ral cheats will spread best when multiple viral genomes, that 159 come from different host cells, infect the same cells. In these 160 conditions, viral cheats are more likely to be coinfecting a 161 cell with a cooperator.

368
Is cheating in viruses the same as cheating elsewhere in the 369 living world (Fig. 2)? We argue that while it is clearly anal-370 ogous, viral biology leads to important differences. These 371 include:

372
(1) The high mutation rate and simple genome of viruses 373 means that mutations to cheating can happen relatively easily. 374 For example, defective interfering genomes regularly emerge 375 de novo in viral infections 83 .

376
(2) The short-term advantages of cheating in viruses can be 377 exceptionally high. Viral cheats can achieve more than a 378 1,000-fold replicative advantage over cooperators, which is 379 orders of magnitude higher than the fitness advantages seen 380 in cuckoos, non-producing bacteria, or other cheats 13,18,21 .

381
(3) The fitness advantage of cheats is often transient at a lo-382 cal scale. Cheats can emerge easily, and then spread rapidly, 383 for example within a host, but then show poor or even non-384 existent transmission to new hosts 31,77 .

385
Taken together, these three features mean that cheating is 386 both common and transient in many viruses. Viral cheats are 387 therefore special in the extent to which they are characterised 388 by 'boom and bust' dynamics. Cooperative viruses will con-389 sequently be selected to evolve mechanisms to either avoid 390 generating cheats, and/or reduce exploitation by cheats.