Animals constantly evolve to protect themselves against viruses, and in turn, viruses evolve to escape their host’s new defenses. As a result, genes involved in this arms’ race are some of the fastest evolving in nature. A better understanding of how host-virus evolution works could help in the search for treatments for many human and animal diseases.

Repetition is one of the gold standard requirements for biological experiments. Watching different groups of animals and viruses evolve under the same conditions makes it possible for researchers to work out whether certain changes are more likely than others. This is easy to do in the laboratory, where conditions can be controlled, but much more complicated to accomplish in the wild. Wild populations are rarely completely isolated, and often face different environmental conditions. One animal-virus pair for which this is not the case is made up of the fly Drosophila innubila, and its virus Drosophila innubila nudivirus. They live in the 'sky islands' of North America, patches of forests surrounded by hundreds of kilometers of desert. These islands are like natural test tubes, isolated ecosystems each with its own separate fly and virus populations and limited gene flow between populations.

To understand how this virus-host pair evolves, Hill and Unckless sequenced the genomes of flies and viruses from four different populations. While the fly genomes did not show evidence of strong differences between populations, the virus genomes did. There were two distinct types of virus, one of which was a lot more effective than the other at infecting flies, possibly because it was better at blocking the fly's immune defenses. Unexpectedly, this virus type had evolved more than once, emerging separately on at least four different occasions. Hill and Unckless suggest that the natural interactions between flies with similar genomes and the virus guide evolution down the same path time and time again.

This work on wild populations contributes to the understanding of the evolution of viruses and their hosts. One question left unanswered is why both types of virus (one more effective at infecting the flies and the other less so) persist in each population when one is better at blocking the fly's immune response? Future work using isolated populations like these could shed more light on the pressures that shape the evolution of viruses and their hosts, potentially helping in the study of human viruses, like HIV.

Antagonistic coevolution between hosts and their parasites is nearly ubiquitous across the diversity of life (Burt and Trivers, 2006). As a result, genes involved in immune defense are among the fastest evolving genes in host genomes (Nielsen et al., 2005; Sackton et al., 2007; Enard et al., 2016; Shultz and Sackton, 2019). Viruses are a particular fitness burden on hosts; for viruses to persist within populations, they must successfully invade the host organism, contend with the host-immune system, replicate and then transmit the newly produced particles to a new host (Holmes, 2007; Gifford, 2012). Once successfully established in a population, natural selection acts to modulate the rate the virus propagates relative to its virulence, optimizing the ratio of virulence to transmission (Williams and Nesse, 1991; May and Nowak, 1995; Lipsitch et al., 1996). Due in part to their elevated mutation rate and large population sizes, viruses can accomplish this, and often co-opt or manipulate host-pathways in the process (Burgyán and Havelda, 2011; Davey et al., 2011; Palmer et al., 2019).

Given the strong selective pressures viruses exert on their hosts (and hosts on viruses), it would be useful to assess how the two players behave in replicate evolutionary experiments. Experiments examining the evolution of pathogens and hosts are common in the lab (Perron et al., 2006; Paterson et al., 2010; Bull et al., 2011; Martins et al., 2014; Scanlan et al., 2015) and even in patients (Pennings, 2012; Pennings et al., 2014; Feder et al., 2019), but difficult in natural settings (Souza et al., 2002; Grubaugh et al., 2015), so we are rarely able to observe the coevolutionary dynamics between a virus and its host in replicate populations to examine natural coevolution and identify if evolution repeats itself. These studies are necessary as they frequently help characterize the pathways viruses have evolved to escape or suppress the host-immune system, and whether the same host pathways are common targets of the same virus (Sabin et al., 2010).

Several studies of viruses in natural populations highlight that the same initial virus can follow the same adaptive path in multiple replicate samples. This results in parallel or convergent evolution of a virus better optimized to the host, possible due to the relatively high viral mutation rate and viral population size (Bull et al., 1997; Casino et al., 1999; Crandall et al., 1999; van Mierlo et al., 2012). However, in other systems, the same initial virus can evolve divergently in a host or location-dependent manner. Specifically, virulence and the diversity of mutations can differ dramatically based on differences in the host environment, the extremely strong selective pressures acting on the virus, and the unpredictability of some viruses due to their elevated mutation rate (Martinez-Picado et al., 2002; Cuevas et al., 2003; Real et al., 2005; Grubaugh et al., 2015). Characterizing how viruses adapt to their long-term hosts in naturally structured populations will help broaden and expand our understanding of how hosts and pathogens evolve in response to each other, and how repeatable evolution is in the face of minor environmental differences.

We took advantage of a natural host/DNA virus infection model involving populations separated by hundreds of kilometers to study how viruses evolve with their hosts in replicate populations with limited gene flow. Drosophila innubila is a mycophagous species that lives in montane forests in the ‘Sky islands’ in southwestern North America. They are commonly infected with a double-stranded DNA virus, the Drosophila innubila Nudivirus (DiNV) (Hill and Unckless, 2018). A previous study examined the rates of evolution of DiNV (Hill and Unckless, 2018), finding the envelope and replication machinery to be rapidly evolving in DiNV, suggesting its importance in viral propagation (Hill and Unckless, 2018). Additionally, a viral suppressor of Toll was identified in the genome of a closely related virus and DiNV, suggesting that components of the Toll pathway (perhaps antimicrobial peptides – AMPs) interact with the virus, which these viruses have then evolved to suppress (Palmer et al., 2019). Consistent with this, genes in the Toll pathway and Toll-regulated AMPs are rapidly evolving in D. innubila (Hill et al., 2019). A more thorough examination of the natural evolution of both the host and virus is necessary to understand how DiNV and D. innubila interact with each other beyond this suppressor protein.

We surveyed natural genetic variation in DiNV in four replicate populations to infer its co-evolutionary history with D. innubila. To this end, we (a) estimated how long DiNV has infected D. innubila, (b) inferred the evolutionary history of the host and virus among the populations, including signatures of selection and recombination (c) examined host and virus genetic variation associated with viral titer within individual hosts, and (d) characterized gene expression differences associated with that genetic variation. We identified two viral multilocus genotypes that differ by 11 focal SNPs and found that these viral types are maintained within the same host population and across multiple isolated host populations. These SNPs are tightly linked, likely brought together by a combination of recurrent mutation and recombination, and this linkage appears to be maintained by strong selection. One viral type is associated with 100-fold higher viral titer and increased virulence compared to the other. Further, we found evidence that the SNPs associated with the high titer type have evolved independently in at least three geographically-isolated viral populations infecting the sympatric D. innubila and D. azteca, and a geographically separate population of viruses infecting D. falleni. Together, these results suggest rapid evolutionary dynamics of host–virus interactions, due to recurrent evolution of a highly virulent haplotype that interacts with multiple host pathways.

Viruses are constantly evolving not just to propagate within a host, but also to optimize their infection across hosts. This optimization involves the relationship between their ability to infect an individual and to transmit to others, tempered by the pathogenic effects on infected hosts caused by viral activity (May and Nowak, 1995; Lipsitch et al., 1996; Alizon and van Baalen, 2008). Since the host is also evolving in response to the virus, an evolutionary arms-race often ensues (Dawkins and Krebs, 1979; Kaltz and Shykoff, 1998; Daugherty and Malik, 2012). DNA viruses have large genomes and often recombination, placing them as a somewhat transitionary pathogen between RNA viruses, bacteria and eukaryotic pathogens and parasites. Here, to work towards expanding our understanding of the co-evolution of viruses and their hosts, we examine the population dynamics of Drosophila innubila Nudivirus (DiNV), a DNA virus infecting D. innubila (Unckless, 2011). Within our set of viral samples, we found two DiNV types which differ by 11 SNPs (Figure 1, named High and Low types). One haplotype (the High type) is associated with higher viral titer, likely due to an increased manipulation of the host-immune system and increased expression of viral factors (Figure 2).

The derived High type has likely recurrently evolved in each population since the last glacial maximum (~10,000 years ago). The two types appear to not co-infect individuals, and mutations appear in a similar order as if navigating an epistatic fitness landscape (Dobzhansky, 1937; Kondrashov et al., 2002; Gavrilets, 2004). Finally, despite the higher titer and transmission rate of the High type compared to the Low type, we found the two types in all sampled populations. Thus, a fundamental question is: why are there two haplotypes? Possible explanations are (a) that the two haplotypes are neutral and coexist due to genetic drift, (b) we caught High type amid a selective sweep, or (c) the two haplotypes are adaptively maintained due to a tradeoff.

Given the immense differences in titer and survival between High and Low types and the rest of the preponderance of evidence presented above, we find it implausible that the two haplotypes are associated with equal viral fitness and therefore evolving under a model of genetic drift (Gillespie, 2004).

It is possible we caught the High type DiNV in the middle of a selective sweep in each population. Differing basic reproduction numbers (R₀) would result in changes in the ratio of types over time (as seen between 2001 and 2017, Figure 1D). As we only have two time points, we could be witnessing a selective sweep of the High type spreading to fixation (Nielsen, 2005), with recombination causing the observed differences in the background. As we found the High type appears to have evolved recurrently, it would be unlikely that we have caught intermediate sweeps in all four populations sampled (Figure 1D). Using the frequency of the High type between 2001 and 2017 CH samples, we can calculate the selection coefficient for the High type if increasing at an exponential rate (which is likely if mid-selective sweep given the intermediate frequency at both time points). D. innubila is active during the monsoon season if Arizona (late July to early September, 10 weeks with a generation time of 2–3 weeks) (Patterson and Stone, 1949). If we assume five generations per year (the minimum number of viral generations per year, limited by the host generation time) and an increase among infected individuals from 52% on 2001 to 71% in 2017, the selection coefficient = 0.0055, which suggests the High type would take ~350 years maximum to fix in a population once it has arisen (from a frequency 1/N_e to one in the viral population). Given the coalescence time of the two types is close to the expansion time of the two viruses (2–30 thousand years), this does not fit with our results, suggesting the two types are more likely to be adaptively maintained and not amid a selective sweep. Further, if the High type was sweeping, it would be remarkable for us to catch these sweeps occurring in all four populations sampled, given the lack of gene flow between populations, our estimated time to fixation of 350 years, and estimated initial infection ~30 thousand years ago.

Finally, two viral types could be maintained because of a tradeoff between the two virus types. We can imagine several scenarios for such a tradeoff. First, within a single population of genetically identical hosts, two viral types could coexist if they had equal basic reproduction numbers (R₀) (Nowak and May, 1994; May and Nowak, 1995). The basic reproduction number in a simple epidemiology model is the ratio of the transmission rate of the pathogen to its virulence. So, while we’ve observed that virulence is higher for the High type virus compared to the Low type, it may be also true that transmission is also higher in the High type (inferred from Figure 1D) and those differences are perfectly balanced creating equal R₀. We find this unlikely but a useful starting point for considering the adaptive maintenance of the two types. A more likely second scenario is one where in a single population of a given species, different host genotypes favor different viral type and that the interaction between optimal R₀ among the viral genotypes and susceptibility of the host genotypes creates a stable equilibrium for both host genotypes and viral genotypes (Anderson and May, 1981). These host genotypes could also be different host species. In fact, in the Chiricahuas, D. innubila was more likely to be infected with the High type virus and D. azteca was more likely to be infected with the Low type virus (Figure 6) even though the two species are sympatric and the differences in viral titer between the High and Low types were qualitatively similar. Therefore, R₀ might be higher for the High type in D. innubila and higher for the Low type in D. azteca with some transmission between host species. These community dynamics could maintain both viruses in both species. The second broad category of tradeoffs involves environmental heterogeneity, which could also create the conditions for the maintenance of both viral types. Furthermore, differing environmental conditions (such as differences in population densities) could affect the frequencies in each location (Anderson and May, 1981). Finally, different viral types could also preferentially infect different tissues, and have different titers due to limitations of the given infected tissues. Previous work in nudiviruses identified nudivurses that infect different tissues including the gut, fat body and ovaries (Jackson et al., 2005; Burand et al., 2012; Palmer et al., 2018). Two DiNV strains infecting different tissues may also help explain the lack of recombination between High and Low types, as viral particles cannot exchange information if they segregate to different tissues. However, this would ignore the fact that we found no evidence of coinfection and viruses infecting two tissues could more easily coinfect.

It is possible that the High type has not recurrently evolved in each population and instead the two viral types are maintained by selection in the face of ongoing gene conversion. However, this would not explain the shared background variants which are fixed in the High type and population exclusive, or the evidence of recurrent evolution in the assembled phylogeny (Figure 5). It is entirely possible however that within each population, selection is maintaining the High type due to the fitness deficits of intermediate types, while gene conversion erodes the association of the eleven haplotype SNPs, which is possible given the recombination events found around the significantly associated SNPs (Figure 3—figure supplement 1).

Studies of different viruses have found that they can undergo speciation-like events through the accumulation of genetic incompatibilities (Matsubara and Otsuji, 1978; Rokyta and Wichman, 2009; Meyer et al., 2016). In these cases, the two viral types cannot recombine and produce viable viral progeny. This may be occurring in DiNV if infection by one viral type limits the ability of the other type to infect the same cell – akin to prezygotic reproductive isolation (Coyne and Orr, 2004). Alternatively, coinfection could happen but generate inviable recombinant particles (Meyer et al., 2016) – akin to postzygotic reproductive isolation. For example, the fourth mutation might only be favored when in the presence of the derived allele of the third mutation because the third mutation causes a conformational change in protein A that allows the conformational change in protein B caused by the fourth mutation to produce infective viral particles (Orr, 1995). The fact that 19K is known to form a complex with other PIF proteins and other membrane proteins suggests it could play the central role in the formation of this incompatibility system (Wang et al., 2007; Rohrmann, 2013). If incompatibilities are present between types, they are likely in rapidly evolving vital nudivirus genes (Hill and Unckless, 2017; Hill and Unckless, 2018). As well as being rapidly evolving across baculoviruses and nudivurses, we found these genes also have high rates of adaptive evolution (Figure 4) and are associated with SNPs that distinguish the High and Low DiNV strains (Figure 1), further supporting the idea that the few genes are key to nudivirus replication are targets of host-suppression and locked in arms-races across the entire nudivirus/baculovirus phylogeny (Hill and Unckless, 2017).

DNA viruses such as DiNV have complicated replication cycles and large genomes. This makes them a sort of evolutionary intermediate between RNA viruses (small genomes, high mutation rates) and eukaryotes (large genomes, low mutation rates) and tangential to bacteria and archaea (intermediate genomes, low recombination rates) (Rohrmann, 2013). However, adaptation appears to occur through changes in a few key proteins (Hill and Unckless, 2017; Hill and Unckless, 2018). Here we found the evolution of two competing viral types that differ in variants near these few key genes, which cause the strains to differ in virulence and titer. Overall our results suggest that the high mutation rates and extremely high levels of selection can result in the repeated and convergent evolution of novel host-virus interactions. Additionally, we found that these host-virus interactions for large DNA viruses can be much more complicated than previous models suggest (Dolan et al., 2018; Feder et al., 2019).

Sequencing data have been deposited on the NCBI SRA under the study accession: SRP187240 Genomes used in this study are available at the following accessions: Drosophila innubila - GCF_004354385.1 Drosophila innubila Nudivirus - GCF_004132165.1 Drosophila azteca - GCA_005876895.1.

1. 1. Hill T
   2. Unckless RL

   (2020) NCBI Sequence Read Archive

   ID SRP187240. Drosophila Sky Island sequencing.

   https://www.ncbi.nlm.nih.gov/sra/SRX5449484[accn]
2. 1. Hill T
   2. Unckless RL

   (2020) Dryad Digital Repository

   Drosophila Sky Island data analysis.

   https://doi.org/10.5061/dryad.2fqz612mh
3. 1. Hill T
   2. Unckless RL

   (2020) figshare

   ID innubila_population_genomics/87662. Innubila population genomics.

   https://figshare.com/projects/innubila_population_genomics/87662

1. 1. Alizon S
   2. van Baalen M

   (2008) Transmission-virulence trade-offs in vector-borne diseases

   Theoretical Population Biology 74:6–15.

   https://doi.org/10.1016/j.tpb.2008.04.003

   • PubMed
   • Google Scholar
2. 1. Altschul SF
   2. Gish W
   3. Miller W
   4. Myers EW
   5. Lipman DJ

   (1990) Basic local alignment search tool

   Journal of Molecular Biology 215:403–410.

   https://doi.org/10.1016/S0022-2836(05)80360-2

   • PubMed
   • Google Scholar
3. 1. Anders S
   2. Pyl PT
   3. Huber W

   (2015) HTSeq--a Python framework to work with high-throughput sequencing data

   Bioinformatics 31:166–169.

   https://doi.org/10.1093/bioinformatics/btu638

   • PubMed
   • Google Scholar
4. 1. Anderson RM
   2. May RM

   (1981) The population dynamics of microparasites and their invertebrate hosts

   Philosophical Transactions of the Royal Society of London 291:451–524.

   https://doi.org/10.1098/rstb.1981.0005

   • Google Scholar
5. 1. Bouckaert R
   2. Heled J
   3. Kühnert D
   4. Vaughan T
   5. Wu CH
   6. Xie D
   7. Suchard MA
   8. Rambaut A
   9. Drummond AJ

   (2014) BEAST 2: a software platform for bayesian evolutionary analysis

   PLOS Computational Biology 10:e1003537.

   https://doi.org/10.1371/journal.pcbi.1003537

   • PubMed
   • Google Scholar
6. Software

   1. Buffalo V

   (2018)

   Scythe

   Scythe.
7. 1. Bull JJ
   2. Badgett MR
   3. Wichman HA
   4. Huelsenbeck JP
   5. Hillis DM
   6. Gulati A
   7. Ho C
   8. Molineux IJ

   (1997)

   Exceptional convergent evolution in a virus

   Genetics 147:1497–1507.

   • Google Scholar
8. 1. Bull JJ
   2. Heineman RH
   3. Wilke CO

   (2011) The phenotype-fitness map in experimental evolution of phages

   PLOS ONE 6:e27796.

   https://doi.org/10.1371/journal.pone.0027796

   • PubMed
   • Google Scholar
9. 1. Burand JP
   2. Kim W
   3. Afonso CL
   4. Tulman ER
   5. Kutish GF
   6. Lu Z
   7. Rock DL

   (2012) Analysis of the genome of the sexually transmitted insect virus Helicoverpa zea Nudivirus 2

   Viruses 4:28–61.

   https://doi.org/10.3390/v4010028

   • Google Scholar
10. 1. Burgyán J
    2. Havelda Z

    (2011) Viral suppressors of RNA silencing

    Trends in Plant Science 16:265–272.

    https://doi.org/10.1016/j.tplants.2011.02.010

    • PubMed
    • Google Scholar
11. Book

    1. Burt A
    2. Trivers R

    (2006)

    Genes in Conflict

    Harvard University Press.

    • Google Scholar
12. 1. Casino C
    2. McAllister J
    3. Davidson F
    4. Power J
    5. Lawlor E
    6. Yap PL
    7. Simmonds P
    8. Smith DB

    (1999) Variation of hepatitis C virus following serial transmission: multiple mechanisms of diversification of the hypervariable region and evidence for convergent genome evolution

    Journal of General Virology 80:717–725.

    https://doi.org/10.1099/0022-1317-80-3-717

    • PubMed
    • Google Scholar
13. 1. Chateigner A
    2. Bézier A
    3. Labrousse C
    4. Jiolle D
    5. Barbe V
    6. Herniou EA

    (2015) Ultra deep sequencing of a baculovirus population reveals widespread genomic variations

    Viruses 7:3625–3646.

    https://doi.org/10.3390/v7072788

    • PubMed
    • Google Scholar
14. 1. Cingolani P
    2. Platts A
    3. Wang leL
    4. Coon M
    5. Nguyen T
    6. Wang L
    7. Land SJ
    8. Lu X
    9. Ruden DM

    (2012) A program for annotating and predicting the effects of single Nucleotide Polymorphisms, SnpEff: snps in the genome of Drosophila melanogaster strain w1118; iso-2; iso-3

    Fly 6:80–92.

    https://doi.org/10.4161/fly.19695

    • PubMed
    • Google Scholar
15. 1. Costa A
    2. Jan E
    3. Sarnow P
    4. Schneider D

    (2009) The imd pathway is involved in antiviral immune responses in Drosophila

    PLOS ONE 4:e7436.

    https://doi.org/10.1371/journal.pone.0007436

    • PubMed
    • Google Scholar
16. Software

    1. Coyne J
    2. Orr HA

    (2004)

    Speciation

    Speciation.
17. 1. Crandall KA
    2. Kelsey CR
    3. Imamichi H
    4. Lane HC
    5. Salzman NP

    (1999) Parallel evolution of drug resistance in HIV: failure of nonsynonymous/synonymous substitution rate ratio to detect selection

    Molecular Biology and Evolution 16:372–382.

    https://doi.org/10.1093/oxfordjournals.molbev.a026118

    • PubMed
    • Google Scholar
18. 1. Cuevas JM
    2. Moya A
    3. Elena SF

    (2003) Evolution of RNA virus in spatially structured heterogeneous environments

    Journal of Evolutionary Biology 16:456–466.

    https://doi.org/10.1046/j.1420-9101.2003.00547.x

    • PubMed
    • Google Scholar
19. 1. Daugherty MD
    2. Malik HS

    (2012) Rules of engagement: molecular insights from host-virus arms races

    Annual Review of Genetics 46:677–700.

    https://doi.org/10.1146/annurev-genet-110711-155522

    • PubMed
    • Google Scholar
20. 1. Davey NE
    2. Travé G
    3. Gibson TJ

    (2011) How viruses hijack cell regulation

    Trends in Biochemical Sciences 36:159–169.

    https://doi.org/10.1016/j.tibs.2010.10.002

    • PubMed
    • Google Scholar
21. 1. Dawkins R
    2. Krebs JR

    (1979) Arms races between and within species

    Proceedings of the Royal Society of London B: Biological Sciences 205:489–511.

    https://doi.org/10.1098/rspb.1979.0081

    • Google Scholar
22. 1. DePristo MA
    2. Banks E
    3. Poplin R
    4. Garimella KV
    5. Maguire JR
    6. Hartl C
    7. Philippakis AA
    8. del Angel G
    9. Rivas MA
    10. Hanna M
    11. McKenna A
    12. Fennell TJ
    13. Kernytsky AM
    14. Sivachenko AY
    15. Cibulskis K
    16. Gabriel SB
    17. Altshuler D
    18. Daly MJ

    (2011) A framework for variation discovery and genotyping using next-generation DNA sequencing data

    Nature Genetics 43:491–498.

    https://doi.org/10.1038/ng.806

    • Google Scholar
23. Software

    1. Dobzhansky T

    (1937)

    Genetics and the origin of species

    Genetics and the origin of species.
24. 1. Dolan PT
    2. Whitfield ZJ
    3. Andino R

    (2018) Mechanisms and concepts in RNA virus population dynamics and evolution

    Annual Review of Virology 5:69–92.

    https://doi.org/10.1146/annurev-virology-101416-041718

    • PubMed
    • Google Scholar
25. 1. Eden E
    2. Navon R
    3. Steinfeld I
    4. Lipson D
    5. Yakhini Z

    (2009) GOrilla: a tool for discovery and visualization of enriched GO terms in ranked gene lists

    BMC Bioinformatics 10:1–7.

    https://doi.org/10.1186/1471-2105-10-48

    • Google Scholar
26. 1. Eilertson KE
    2. Booth JG
    3. Bustamante CD

    (2012) SnIPRE: selection inference using a poisson random effects model

    PLOS Computational Biology 8:e1002806.

    https://doi.org/10.1371/journal.pcbi.1002806

    • PubMed
    • Google Scholar
27. 1. Enard D
    2. Cai L
    3. Gwennap C
    4. Petrov DA

    (2016) Viruses are a dominant driver of protein adaptation in mammals

    eLife 5:e12469.

    https://doi.org/10.7554/eLife.12469

    • PubMed
    • Google Scholar
28. 1. Feder AF
    2. Pennings PS
    3. Hermisson J
    4. Petrov DA

    (2019) Evolutionary dynamics in structured populations under strong population genetic forces

    G3: Genes, Genomes, Genetics 9:3395–3407.

    https://doi.org/10.1534/g3.119.400605

    • Google Scholar
29. 1. Ferreira ÁG
    2. Naylor H
    3. Esteves SS
    4. Pais IS
    5. Martins NE
    6. Teixeira L

    (2014) The Toll-Dorsal pathway is required for resistance to viral oral infection in Drosophila

    PLOS Pathogens 10:e1004507.

    https://doi.org/10.1371/journal.ppat.1004507

    • PubMed
    • Google Scholar
30. Book

    1. Gavrilets S

    (2004)

    Fitness Landscapes and the Origin of Species

    Princeton University Press.

    • Google Scholar
31. 1. Ghosh S
    2. Chan CK

    (2016) Analysis of RNA-Seq data using TopHat and cufflinks

    Methods Mol Biol 1374:339–361.

    https://doi.org/10.1007/978-1-4939-3167-5_18

    • Google Scholar
32. 1. Gifford RJ

    (2012) Viral evolution in deep time: lentiviruses and mammals

    Trends in Genetics 28:89–100.

    https://doi.org/10.1016/j.tig.2011.11.003

    • Google Scholar
33. Software

    1. Gillespie J

    (2004)

    Population Genetics: A Concise Guide

    Population Genetics: A Concise Guide.
34. 1. Grubaugh ND
    2. Smith DR
    3. Brackney DE
    4. Bosco-Lauth AM
    5. Fauver JR
    6. Campbell CL
    7. Felix TA
    8. Romo H
    9. Duggal NK
    10. Dietrich EA
    11. Eike T
    12. Beane JE
    13. Bowen RA
    14. Black WC
    15. Brault AC
    16. Ebel GD

    (2015) Experimental evolution of an RNA virus in wild birds: evidence for host-dependent impacts on population structure and competitive fitness

    PLOS Pathogens 11:e1004874.

    https://doi.org/10.1371/journal.ppat.1004874

    • PubMed
    • Google Scholar
35. 1. Haas BJ
    2. Papanicolaou A
    3. Yassour M
    4. Grabherr M
    5. Blood PD
    6. Bowden J
    7. Couger MB
    8. Eccles D
    9. Li B
    10. Lieber M
    11. MacManes MD
    12. Ott M
    13. Orvis J
    14. Pochet N
    15. Strozzi F
    16. Weeks N
    17. Westerman R
    18. William T
    19. Dewey CN
    20. Henschel R
    21. LeDuc RD
    22. Friedman N
    23. Regev A

    (2013) De novo transcript sequence reconstruction from RNA-seq using the trinity platform for reference generation and analysis

    Nature Protocols 8:1494–1512.

    https://doi.org/10.1038/nprot.2013.084

    • PubMed
    • Google Scholar
36. 1. Hill T
    2. Koseva BS
    3. Unckless RL

    (2019) The genome of Drosophila innubila Reveals Lineage-Specific Patterns of Selection in Immune Genes

    Molecular Biology and Evolution 36:1405–1417.

    https://doi.org/10.1093/molbev/msz059

    • PubMed
    • Google Scholar
37. 1. Hill T
    2. Unckless RL

    (2017) Baculovirus molecular evolution via gene turnover and recurrent positive selection of key genes

    Journal of Virology 91:e01319-17.

    https://doi.org/10.1128/JVI.01319-17

    • PubMed
    • Google Scholar
38. 1. Hill T
    2. Unckless RL

    (2018) The dynamic evolution of Drosophila innubila Nudivirus

    Infection, Genetics and Evolution 57:151–157.

    https://doi.org/10.1016/j.meegid.2017.11.013

    • Google Scholar
39. Preprint

    1. Hill T
    2. Unckless R

    (2020) Selection and demography shape genomic variation in a ‘Sky Island’ species

    bioRxiv.

    https://doi.org/10.1101/2020.05.14.096008

    • Google Scholar
40. 1. Holmes EC

    (2007) Viral evolution in the genomic age

    PLOS Biology 5:e278.

    https://doi.org/10.1371/journal.pbio.0050278

    • PubMed
    • Google Scholar
41. 1. Jackson TA
    2. Crawford AM
    3. Glare TR

    (2005) Oryctes virus--time for a new look at a useful biocontrol agent

    Journal of Invertebrate Pathology 89:91–94.

    https://doi.org/10.1016/j.jip.2005.03.009

    • PubMed
    • Google Scholar
42. Software

    1. Joshi N
    2. Fass J

    (2011)

    Sickle: A sliding window, adaptive, quality-based trimming tool for fastQ files

    Sickle: A sliding window, adaptive, quality-based trimming tool for fastQ files.
43. 1. Kaltz O
    2. Shykoff JA

    (1998) Local adaptation in host–parasite systems

    Heredity 81:361–370.

    https://doi.org/10.1046/j.1365-2540.1998.00435.x

    • Google Scholar
44. 1. Kamita SG
    2. Maeda S
    3. Hammock BD

    (2003) High-frequency homologous recombination between baculoviruses involves DNA replication

    Journal of Virology 77:13053–13061.

    https://doi.org/10.1128/JVI.77.24.13053-13061.2003

    • PubMed
    • Google Scholar
45. Software

    1. Kassambara A
    2. Kosinski M
    3. Biecek P

    (2017)

    Survminer: Drawing Survival Curves Using'ggplot2'

    Survminer: Drawing Survival Curves Using'ggplot2'.
46. 1. Kelly DC

    (1982) Baculovirus Replication

    Journal of General Virology 63:1–13.

    https://doi.org/10.1099/0022-1317-63-1-1

    • Google Scholar
47. 1. Kondrashov AS
    2. Sunyaev S
    3. Kondrashov FA

    (2002) Dobzhansky-Muller incompatibilities in protein evolution

    PNAS 99:14878–14883.

    https://doi.org/10.1073/pnas.232565499

    • Google Scholar
48. 1. Kosakovsky Pond SL
    2. Posada D
    3. Gravenor MB
    4. Woelk CH
    5. Frost SD

    (2006) GARD: a genetic algorithm for recombination detection

    Bioinformatics 22:3096–3098.

    https://doi.org/10.1093/bioinformatics/btl474

    • PubMed
    • Google Scholar
49. 1. Lamiable O
    2. Kellenberger C
    3. Kemp C
    4. Troxler L
    5. Pelte N
    6. Boutros M
    7. Marques JT
    8. Daeffler L
    9. Hoffmann JA
    10. Roussel A
    11. Imler JL

    (2016) Cytokine diedel and a viral homologue suppress the IMD pathway in Drosophila

    PNAS 113:698–703.

    https://doi.org/10.1073/pnas.1516122113

    • PubMed
    • Google Scholar
50. 1. Lewis SH
    2. Quarles KA
    3. Yang Y
    4. Tanguy M
    5. Frézal L
    6. Smith SA
    7. Sharma PP
    8. Cordaux R
    9. Gilbert C
    10. Giraud I
    11. Collins DH
    12. Zamore PD
    13. Miska EA
    14. Sarkies P
    15. Jiggins FM

    (2018) Pan-arthropod analysis reveals somatic piRNAs as an ancestral defence against transposable elements

    Nature Ecology & Evolution 2:174–181.

    https://doi.org/10.1038/s41559-017-0403-4

    • Google Scholar
51. 1. Li H
    2. Handsaker B
    3. Wysoker A
    4. Fennell T
    5. Ruan J
    6. Homer N
    7. Marth G
    8. Abecasis G
    9. Durbin R
    10. 1000 Genome Project Data Processing Subgroup

    (2009) The sequence alignment/Map format and SAMtools

    Bioinformatics 25:2078–2079.

    https://doi.org/10.1093/bioinformatics/btp352

    • Google Scholar
52. 1. Li H
    2. Durbin R

    (2009) Fast and accurate short read alignment with Burrows-Wheeler transform

    Bioinformatics 25:1754–1760.

    https://doi.org/10.1093/bioinformatics/btp324

    • PubMed
    • Google Scholar
53. 1. Lipsitch M
    2. Siller S
    3. Nowak MA

    (1996) The evolution of virulence in pathogens with vertical and horizontal transmission

    Evolution 50:1729–1741.

    https://doi.org/10.1111/j.1558-5646.1996.tb03560.x

    • Google Scholar
54. 1. Liu X
    2. Fu YX

    (2015) Exploring population size changes using SNP frequency spectra

    Nature Genetics 47:555–559.

    https://doi.org/10.1038/ng.3254

    • PubMed
    • Google Scholar
55. 1. Maeda S
    2. Kamita SG
    3. Kondo A

    (1993) Host range expansion of Autographa californica nuclear polyhedrosis virus (NPV) following recombination of a 0.6-kilobase-pair DNA fragment originating from Bombyx mori NPV

    Journal of Virology 67:6234–6238.

    https://doi.org/10.1128/JVI.67.10.6234-6238.1993

    • PubMed
    • Google Scholar
56. 1. Martin M

    (2011) Cutadapt removes adapter sequences from high-throughput sequencing reads

    Technical Notes 1:12.

    https://doi.org/10.14806/ej.17.1.200

    • Google Scholar
57. 1. Martinez-Picado J
    2. Frost SD
    3. Izquierdo N
    4. Morales-Lopetegi K
    5. Marfil S
    6. Puig T
    7. Cabrera C
    8. Clotet B
    9. Ruiz L

    (2002) Viral evolution during structured treatment interruptions in chronically human immunodeficiency virus-infected individuals

    Journal of Virology 76:12344–12348.

    https://doi.org/10.1128/JVI.76.23.12344-12348.2002

    • PubMed
    • Google Scholar
58. 1. Martins NE
    2. Faria VG
    3. Nolte V
    4. Schlötterer C
    5. Teixeira L
    6. Sucena É
    7. Magalhães S

    (2014) Host adaptation to viruses relies on few genes with different cross-resistance properties

    PNAS 111:5938–5943.

    https://doi.org/10.1073/pnas.1400378111

    • PubMed
    • Google Scholar
59. 1. Matsubara K
    2. Otsuji Y

    (1978) Preparation of plasmids from lambdoid phages and studies on their incompatibilities

    Plasmid 1:284–296.

    https://doi.org/10.1016/0147-619X(78)90046-X

    • PubMed
    • Google Scholar
60. 1. May RM
    2. Nowak MA

    (1995) Coinfection and the evolution of parasite virulence

    Proceedings. Biological Sciences 261:209–215.

    https://doi.org/10.1098/rspb.1995.0138

    • PubMed
    • Google Scholar
61. 1. McDonald JH
    2. Kreitman M

    (1991) Adaptive protein evolution at the adh locus in Drosophila

    Nature 351:652–654.

    https://doi.org/10.1038/351652a0

    • PubMed
    • Google Scholar
62. 1. McKenna A
    2. Hanna M
    3. Banks E
    4. Sivachenko A
    5. Cibulskis K
    6. Kernytsky A
    7. Garimella K
    8. Altshuler D
    9. Gabriel S
    10. Daly M
    11. DePristo MA

    (2010) The genome analysis toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data

    Genome Research 20:1297–1303.

    https://doi.org/10.1101/gr.107524.110

    • PubMed
    • Google Scholar
63. 1. Merkling SH
    2. van Rij RP

    (2013) Beyond RNAi: antiviral defense strategies in Drosophila and mosquito

    Journal of Insect Physiology 59:159–170.

    https://doi.org/10.1016/j.jinsphys.2012.07.004

    • PubMed
    • Google Scholar
64. 1. Meyer JR
    2. Dobias DT
    3. Medina SJ
    4. Servilio L
    5. Gupta A
    6. Lenski RE

    (2016) Ecological speciation of bacteriophage lambda in allopatry and sympatry

    Science 354:1301–1304.

    https://doi.org/10.1126/science.aai8446

    • PubMed
    • Google Scholar
65. 1. Nielsen R

    (2005) Molecular signatures of natural selection

    Annual Review of Genetics 39:197–218.

    https://doi.org/10.1146/annurev.genet.39.073003.112420

    • PubMed
    • Google Scholar
66. 1. Nielsen R
    2. Bustamante C
    3. Clark AG
    4. Glanowski S
    5. Sackton TB
    6. Hubisz MJ
    7. Fledel-Alon A
    8. Tanenbaum DM
    9. Civello D
    10. White TJ
    11. J Sninsky J
    12. Adams MD
    13. Cargill M

    (2005) A scan for positively selected genes in the genomes of humans and chimpanzees

    PLOS Biology 3:e170.

    https://doi.org/10.1371/journal.pbio.0030170

    • PubMed
    • Google Scholar
67. 1. Nowak MA
    2. May RM

    (1994) Superinfection and the evolution of parasite virulence

    Proceedings R. Soc 255:81–89.

    https://doi.org/10.1098/rspb.1994.0012

    • Google Scholar
68. 1. Orr HA

    (1995)

    Population genetics of speciation: the evolution of hybrid incompatibilities

    Genetics 139:1805–1813.

    • PubMed
    • Google Scholar
69. 1. Palmer WH
    2. Medd NC
    3. Beard PM
    4. Obbard DJ

    (2018) Isolation of a natural DNA virus of Drosophila melanogaster, and characterisation of host resistance and immune responses

    PLOS Pathogens 14:e1007050.

    https://doi.org/10.1371/journal.ppat.1007050

    • PubMed
    • Google Scholar
70. 1. Palmer WH
    2. Joosten J
    3. Overheul GJ
    4. Jansen PW
    5. Vermeulen M
    6. Obbard DJ
    7. Van Rij RP

    (2019) Induction and suppression of NF- κB signalling by a DNA virus of Drosophila

    Journal of Virology 93:e01443.

    https://doi.org/10.1128/JVI.01443-18

    • PubMed
    • Google Scholar
71. 1. Paradis E
    2. Claude J
    3. Strimmer K

    (2004) APE: analyses of phylogenetics and evolution in R language

    Bioinformatics 20:289–290.

    https://doi.org/10.1093/bioinformatics/btg412

    • PubMed
    • Google Scholar
72. 1. Paterson S
    2. Vogwill T
    3. Buckling A
    4. Benmayor R
    5. Spiers AJ
    6. Thomson NR
    7. Quail M
    8. Smith F
    9. Walker D
    10. Libberton B
    11. Fenton A
    12. Hall N
    13. Brockhurst MA

    (2010) Antagonistic coevolution accelerates molecular evolution

    Nature 464:275–278.

    https://doi.org/10.1038/nature08798

    • PubMed
    • Google Scholar
73. Book

    1. Patterson JT
    2. Stone WS

    (1949)

    Studies in the genetics of Drosophila

    University of Texas Publications.

    • Google Scholar
74. 1. Pennings PS

    (2012) Standing genetic variation and the evolution of drug resistance in HIV

    PLOS Computational Biology 8:e1002527.

    https://doi.org/10.1371/journal.pcbi.1002527

    • PubMed
    • Google Scholar
75. 1. Pennings PS
    2. Kryazhimskiy S
    3. Wakeley J

    (2014) Loss and recovery of genetic diversity in adapting populations of HIV

    PLOS Genetics 10:e1004000.

    https://doi.org/10.1371/journal.pgen.1004000

    • PubMed
    • Google Scholar
76. 1. Perron GG
    2. Zasloff M
    3. Bell G

    (2006) Experimental evolution of resistance to an antimicrobial peptide

    Proceedings of the Royal Society B: Biological Sciences 273:251–256.

    https://doi.org/10.1098/rspb.2005.3301

    • Google Scholar
77. Software

    1. Picard

    (2020) Picard

    Picard.

    http://broadinstitute.github.io/picard
78. 1. Purcell S
    2. Neale B
    3. Todd-Brown K
    4. Thomas L
    5. Ferreira MA
    6. Bender D
    7. Maller J
    8. Sklar P
    9. de Bakker PI
    10. Daly MJ
    11. Sham PC

    (2007) PLINK: a tool set for whole-genome association and population-based linkage analyses

    The American Journal of Human Genetics 81:559–575.

    https://doi.org/10.1086/519795

    • PubMed
    • Google Scholar
79. 1. Quast C
    2. Pruesse E
    3. Yilmaz P
    4. Gerken J
    5. Schweer T
    6. Yarza P
    7. Peplies J
    8. Glöckner FO

    (2013) The SILVA ribosomal RNA gene database project: improved data processing and web-based tools

    Nucleic Acids Research 41:D590–D596.

    https://doi.org/10.1093/nar/gks1219

    • PubMed
    • Google Scholar
80. Software

    1. R Development Core Team

    (2013) R: A Language and Environment for Statistical Computing

    R Foundation for Statistical Computing, Vienna, Austria.

    http://www.r-project.org
81. 1. Real LA
    2. Henderson JC
    3. Biek R
    4. Snaman J
    5. Jack TL
    6. Childs JE
    7. Stahl E
    8. Waller L
    9. Tinline R
    10. Nadin-Davis S

    (2005) Unifying the spatial population dynamics and molecular evolution of epidemic Rabies virus

    PNAS 102:12107–12111.

    https://doi.org/10.1073/pnas.0500057102

    • PubMed
    • Google Scholar
82. 1. Robinson MD
    2. McCarthy DJ
    3. Smyth GK

    (2010) edgeR: a bioconductor package for differential expression analysis of digital gene expression data

    Bioinformatics 26:139–140.

    https://doi.org/10.1093/bioinformatics/btp616

    • PubMed
    • Google Scholar
83. Software

    1. Rohrmann GF

    (2013)

    Baculovirus Molecular Biology

    Baculovirus Molecular Biology.
84. 1. Rokyta DR
    2. Wichman HA

    (2009) Genic incompatibilities in two hybrid bacteriophages

    Molecular Biology and Evolution 26:2831–2839.

    https://doi.org/10.1093/molbev/msp199

    • PubMed
    • Google Scholar
85. 1. Sabin LR
    2. Hanna SL
    3. Cherry S

    (2010) Innate antiviral immunity in Drosophila

    Current Opinion in Immunology 22:4–9.

    https://doi.org/10.1016/j.coi.2010.01.007

    • Google Scholar
86. 1. Sackton TB
    2. Lazzaro BP
    3. Schlenke TA
    4. Evans JD
    5. Hultmark D
    6. Clark AG

    (2007) Dynamic evolution of the innate immune system in Drosophila

    Nature Genetics 39:1461–1468.

    https://doi.org/10.1038/ng.2007.60

    • Google Scholar
87. 1. Sagulenko P
    2. Puller V
    3. Neher RA

    (2018) TreeTime: maximum-likelihood phylodynamic analysis

    Virus Evolution 4:vex042.

    https://doi.org/10.1093/ve/vex042

    • PubMed
    • Google Scholar
88. 1. Scanlan PD
    2. Buckling A
    3. Hall AR

    (2015) Experimental evolution and bacterial resistance: (co)evolutionary costs and trade-offs as opportunities in phage therapy research

    Bacteriophage 5:e1050153.

    https://doi.org/10.1080/21597081.2015.1050153

    • PubMed
    • Google Scholar
89. 1. Schulz MH
    2. Zerbino DR
    3. Vingron M
    4. Birney E

    (2012) Oases: robust de novo RNA-seq assembly across the dynamic range of expression levels

    Bioinformatics 28:1086–1092.

    https://doi.org/10.1093/bioinformatics/bts094

    • PubMed
    • Google Scholar
90. 1. Shin J-H
    2. Blay S
    3. Graham J
    4. McNeney B

    (2006) LDheatmap : An R Function for Graphical Display of Pairwise Linkage Disequilibria Between Single Nucleotide Polymorphisms 

    Journal of Statistical Software 16:jss.v016.c03.

    https://doi.org/10.18637/jss.v016.c03

    • Google Scholar
91. 1. Shultz AJ
    2. Sackton TB

    (2019) Immune genes are hotspots of shared positive selection across birds and mammals

    eLife 8:e41815.

    https://doi.org/10.7554/eLife.41815

    • Google Scholar
92. Software

    1. Smit AFA
    2. Hubley R

    (2008)

    RepeatModeler Open-1.0

    RepeatModeler Open-1.0.
93. Software

    1. Smit AFA
    2. Hubley R

    (2013)

    RepeatMasker Open-4.0

    RepeatMasker Open-4.0.
94. 1. Smith NGC
    2. Eyre-Walker A

    (2002) Adaptive protein evolution in Drosophila

    Nature 415:1022–1024.

    https://doi.org/10.1038/4151022a

    • Google Scholar
95. 1. Soetaert K
    2. Petzoldt T
    3. Setzer RW

    (2010) Solving differential equations in R : package deSolve

    Journal of Statistical Software 33:1–24.

    https://doi.org/10.18637/jss.v033.i09

    • Google Scholar
96. 1. Souza V
    2. Travisano M
    3. Turner PE
    4. Eguiarte LE

    (2002) Does experimental evolution reflect patterns in natural populations? E. coli strains from long-term studies compared with wild isolates

    Antonie Van Leeuwenhoek 81:143–153.

    https://doi.org/10.1023/a:1020594013195

    • PubMed
    • Google Scholar
97. 1. Stadler T
    2. Kuhnert D
    3. Bonhoeffer S
    4. Drummond AJ

    (2013) Birth-death skyline plot reveals temporal changes of epidemic spread in HIV and hepatitis C virus (HCV)

    PNAS 110:228–233.

    https://doi.org/10.1073/pnas.1207965110

    • Google Scholar
98. 1. Stoletzki N
    2. Eyre-Walker A

    (2011) Estimation of the neutrality index

    Molecular Biology and Evolution 28:63–70.

    https://doi.org/10.1093/molbev/msq249

    • PubMed
    • Google Scholar
99. 1. Unckless RL

    (2011) A DNA virus of Drosophila

    PLOS ONE 6:e26564.

    https://doi.org/10.1371/journal.pone.0026564

    • PubMed
    • Google Scholar
100. 1. van Mierlo JT
     2. Bronkhorst AW
     3. Overheul GJ
     4. Sadanandan SA
     5. Ekström JO
     6. Heestermans M
     7. Hultmark D
     8. Antoniewski C
     9. van Rij RP

     (2012) Convergent evolution of argonaute-2 slicer antagonism in two distinct insect RNA viruses

     PLOS Pathogens 8:e1002872.

     https://doi.org/10.1371/journal.ppat.1002872

     • PubMed
     • Google Scholar
101. 1. Wang Y
     2. Kleespies RG
     3. Huger AM
     4. Jehle JA

     (2007) The genome of Gryllus bimaculatus Nudivirus indicates an ancient diversification of baculovirus-related nonoccluded nudiviruses of insects

     Journal of Virology 81:5395–5406.

     https://doi.org/10.1128/JVI.02781-06

     • PubMed
     • Google Scholar
102. 1. Wang Y
     2. Kleespies RG
     3. Ramle MB
     4. Jehle JA

     (2008) Sequencing of the large dsDNA genome of oryctes Rhinoceros nudivirus using multiple displacement amplification of nanogram amounts of virus DNA

     Journal of Virological Methods 152:106–108.

     https://doi.org/10.1016/j.jviromet.2008.06.003

     • PubMed
     • Google Scholar
103. 1. Webster CL
     2. Waldron FM
     3. Robertson S
     4. Crowson D
     5. Ferrari G
     6. Quintana JF
     7. Brouqui J-M
     8. Bayne EH
     9. Longdon B
     10. Buck AH
     11. Lazzaro BP
     12. Akorli J
     13. Haddrill PR
     14. Obbard DJ

     (2015) The discovery, distribution, and evolution of viruses associated with Drosophila melanogaster

     PLOS Biology 13:e1002210.

     https://doi.org/10.1371/journal.pbio.1002210

     • PubMed
     • Google Scholar
104. 1. Williams GC
     2. Nesse RM

     (1991) The dawn of darwinian medicine

     The Quarterly Review of Biology 66:1–22.

     https://doi.org/10.1086/417048

     • PubMed
     • Google Scholar
105. 1. Wilm A
     2. Aw PP
     3. Bertrand D
     4. Yeo GH
     5. Ong SH
     6. Wong CH
     7. Khor CC
     8. Petric R
     9. Hibberd ML
     10. Nagarajan N

     (2012) LoFreq: a sequence-quality aware, ultra-sensitive variant caller for uncovering cell-population heterogeneity from high-throughput sequencing datasets

     Nucleic Acids Research 40:11189–11201.

     https://doi.org/10.1093/nar/gks918

     • PubMed
     • Google Scholar
106. 1. Wu TD
     2. Nacu S

     (2010) Fast and SNP-tolerant detection of complex variants and splicing in short reads

     Bioinformatics 26:873–881.

     https://doi.org/10.1093/bioinformatics/btq057

     • PubMed
     • Google Scholar
107. 1. Zambon RA
     2. Nandakumar M
     3. Vakharia VN
     4. Wu LP

     (2005) The toll pathway is important for an antiviral response in Drosophila

     PNAS 102:7257–7262.

     https://doi.org/10.1073/pnas.0409181102

     • PubMed
     • Google Scholar

Acceptance summary:

This study reports on a phenotypic and genetic polymorphism of DiNVirus in natural populations of Drosophila. The authors present a series of experiments and assessments to understand how the polymorphism evolved and what implication it has for the host and conclude that the observed polymorphism arose multiple times independently and that it is maintained in a polymorphic state. The results are very clear and convincing and provide an excellent example for the power of natural selection in shaping host-parasite interactions.

Decision letter after peer review:

Thank you for submitting your article "Recurrent evolution of two competing haplotypes in an insect DNA virus" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Diethard Tautz as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Sebastian Gagneux (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

As the editors have judged that your manuscript is of interest, but as described below that additional experiments are required before it is published, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

Your manuscript presents an exciting finding of the evolution and biology of a natural viral pathogen of Drosophila. You suggest that two distinct haplotypes of the virus evolved multiple times in isolated host populations. These two haplotypes are characterised by 11 SNPs and associated phenotypic traits (differences in viral titer). The independent evolution of such a complex trait seems rather unusual, making this nice example of evolution of host-parasite interactions.

The three reviewers of this manuscripts have a lot of praise for the study. However, they also raise a number of important points that needs to be addressed. Most important, the substantive technical points raised by reviewer 2 need careful consideration. This includes explaining and resolving the inconsistencies in the analysis and the right choice of the methods. Reviewer 1 pointed out problems with the simulation. Below are the detailed reviews. A final decision will only be possible when the problems in the analysis are solved.

Reviewer #1:

This study reports on a phenotypic and genetic polymorphism of DiNVirus in Drosophila. The study is very appealing as this study system is a natural system (unlike D. melanogaster) with a well understood ecology and biogeography. The authors go through a series of experiments and assessments to understand how the polymorphism evolved and what implication it has for the host. The main conclusion is that the observed polymorphism arose multiple times independently and that it is maintained in a polymorphic state. The mechanism for this maintenance is not entirely clear. A simulation model is used to bring some light into this puzzle. For the most part, the study is solid and well carried out. Below are a number of points that may help the authors to present their material more clearly.

The Introduction is not to the point. The Introduction moves among various aspects of host-parasite interactions without giving the reader an idea where this is going. At various places topics are raised that then later dismissed. For example: the first two paragraphs are about host response to viruses. In the third paragraph it become system specific, but now switches mainly to the DiNV system. It is not clear where one is going here. The Introduction (sixth paragraph) is summed up with a very generic phrase without much perspective on what is going to follow. At this place I still do not know what this paper will be about. A much more targeted Introduction is necessary. What are the questions? What was driving this research? What are the hypotheses?

With 11 SNPs spread across the chromosome and obligate recombination every round of replication, the proportion of viral offspring with sub-optimal multi-SNP-genotypes must be huge. How can the right genotypes be maintained? Discuss.

The authors run computer simulations (called SIR models, but they seem actually to be SI models) to support their ideas about virus evolution. The simulation regarding the accumulation of mutation is fine and gives quantitative support for presented evolutionary scenario.

A second simulation is about the competition of the two viral types. This is by far the weakest part of the manuscript. I am not convinced about the value of this simulation. The outcome can be predicted from the assumptions of the model, several of them are very speculative. Without data on transmission over the course of the infection, the simulations are not very helpful. I suggest to leave this out. It is ok to speculate about this in the Discussion, but it makes the Results part heavy and less strong. Also, the associated figure (Figure 7) is hard to understand.

The authors stress at multiple place that it is likely that the increased virulence of the high type is traded-off against transmission. The evidence for this is much weaker than the strength of these statements suggests. I strongly suggest to tone this down.

Reviewer #2:

This manuscript presents an exciting analysis of the evolution and biology of a natural viral pathogen of Drosophila. In outline: the authors claim that two distinct haplotypes of the virus co-occur in multiple host populations, that these 'evolved' independently (i.e. separate origins) through convergence, and that they have alternative 'life histories' (high titer and virulence, versus low titer and virulence), which may permit their coexistence.

The practical experiments and sequencing data are substantial and appear sound, and the work is likely to be of interest to a very broad host-pathogen audience. However, while some of these headline claims are well-supported, I have a number of serious concerns about the analyses and interpretation of others, and a few key methodological details are missing. The analyses would need substantial revision, or at least additional checks, before I would be convinced by the story.

1) Is there any possibility of circularity in defining the 'high' and 'low' haplotypes?

Two 'types' are defined based on 11 linked SNPs that are described as having 'high' or 'low' titer phenotypes. First, this requires a more robust approach to define 'significance' as (if there is any LD) tests are not independent, and sample sizes relative to predictive SNPs are small. Phenotypes (titers) should be permuted across genotypes (within populations) a thousand times and the analyses re-run for each permutation, then the tails of this distribution used to define significance, as commonly done for e.g. the DGRP. Second, this feels like it is dangerously circular: supposing all 11 SNPs were false positives, and arbitrary haplotypes erroneously defined based on them? Post hoc analysis of the difference between the two haplotypes could still show Figure 1B, because those differences were what drove the (erroneous) detection of the SNPs. So would Figure 1C, since that result necessarily follows from the definition of the haplotypes.

To convince me that the 'two haplotypes' interpretation is real, for each of the randomisation replicates one would need to define a 'high' and a 'low' haplotype based on the any randomisation spurious 'significant' SNPs and re-run the rest of Figure 1 to demonstrate that the separation between the real haplotypes is greater than that between those that result from permutation tests. I appreciate that this will require some computing time, but the authors already have all of the code, so it shouldn't be more than an afternoon of 'hands on' time.

Since the haplotypes overlap in phenotype space (Figure 1B) and not all have all SNPs (Figure 1C), how were intermediate haplotypes (with <11 SNPs) assigned to high or low?

2) Did the haplotypes evolve independently three or four times?

Of the 11 SNPs defining the haplotype, three are non-synonymous, five are in the UTRs of known virulence genes, and three are intergenic SNPs. Is it really credible that a specific base change has arisen and been selected for independently in each of four populations, at each of the 11 sites? i.e. always A->C being beneficial at a site, but never A->G at that site? Even for the non-coding ones? This is an extraordinary claim that would require extraordinary evidence, for which the rates of recombination and gene conversion seem at the heart of the argument.

In some places the authors appear to assume (or assert) that recombination is frequent, while in others they assume it is completely absent, and nowhere do they explicitly test for it. If it is common, then none of the tree-based analyses can be used. If it is absent, then it is very hard to explain the patterns of diversity or LD in Figure 1—figure supplement 3, and some of the simulations may be inappropriate.

The tree-based analyses seem to be the basis of the major claim that these haplotypes arose independently multiple times, and that the order of the mutations arising was similar. Any tree analysis absolutely requires the absence of recombination or gene conversion, and this needs to be explicitly tested for here (e.g. using GARD, or possibly if diversity is low by inferring the ARG using tsinfer). The failure of LD to decay with distance hints that recombination is absent, but gene conversion over very short distances could still shuffle mutations between haplotypes. However, assuming its branch lengths are in mutations, the shape of the tree in Figure 6A (many short tips below large crowns) shouts 'recombination' to me. Where did this tree come from? It is not ultrametric, so if it was inferred with BEAST this is not standard output.

If the authors do have evidence that recombination is absent, can they confirm that the other analyses (population size history from the SFS; simulations) also assumed zero recombination? Although zero recombination would make me worry even more about the p-values in the GWAS discovery of the SNPs that define the haplotypes – but permutation tests would help deal with that.

I am really confused about their view of recombination, because a couple of times they imply recombination may be common, for example by noting that recombination is required in Nudivirus replication – but then this would invalidate the tree-based analyses? Plus, mechanistic recombination is irrelevant without coinfection infection, since recombination only between identical haplotypes does not have any effect. The observation that no co-infections occur suggests recombination should be very rare – but that's not what the tree looks like. The sixth paragraph of the Discussion suggests that not enough thought has been given to the likelihood, or implications, of recombination.

Finally, if there is no recombination, then 'fully derived' haplotypes (those with all 11 SNPs) can only be as old as the youngest of the 11 SNPs, unless the SNPs arose multiple time within populations as well. Is that compatible with the patterns of diversity and the timescale? One check would be to mask those 11, re-infer the tree, and check that the clades are still monophyletic. If these arose by re-current mutations due to strong selection, they could be warping the tree – this might be akin to the problem of recurrent drug or MHC selected mutations in HIV trees, where the known sites are excluded before analysis.

If we believe the two haplotypes are real, a much more credible 'story' to me would be two distinct and potentially old haplotypes maintained by selection in the face of a low level of ongoing gene conversion (or recombination). Is there some aspect of the data that this does not fit? This seems just as good a 'story', just as interesting, and just as publishable.

3) The MK-like analyses

Although a relatively minor part of the story, the MK analysis is extremely unclear. In part this is because it doesn't appear in the Materials and methods.

i) What software was used? The supporting data implies VCFtools can do this (which surprised me) but the method it uses was not explained. Was SNIpre fitted with the original code? The text seem to imply that raw counts were used, which would not be suitable unless Ks was <0.3

ii) Although divergence is required, we're not told how (or from what species) it was estimated. If Ks>0.3, then something more than raw counts should certainly be used. If Ks>0.8, I would strongly advise against doing any sort of MK analysis.

iii) Figure 4 gives 'difference from background'. What is meant by difference from background? But if the background is a signal of strong constraint, could a positive signal here mean relaxed constraint rather than positive selection. What is the evidence that it's positive selection rather than relaxed constraint?

iv) In Figure 4—figure supplement 1 the populations are presented separately, and differences among them are interpreted as differences in positive selection. But surely the divergence number must be massive, larger than the polymorphism number, so almost all of this variation is due to differences in constraint affecting the Pn/Ps ratio. Or is the method one that uses high frequency derived SNPs as evidence of positive selection? If so, how was ancestral state identified? This is probably not possible to do reliably if Ks>0.3. We need to see some raw numbers, and a lot more detail on the methods.

Reviewer #3:

This is an interesting piece of work on the co-evolution of the Drosophilainnubila Nudivirus (DiNV) and its host. The authors analyzed several natural populations of D. innubila, some of which were partially infected with DiNV, and simultaneously characterized both the host and the infecting viral pathogen using a combination of DNA and RNA sequencing and various complementary analytical approaches. Using a GWAS approach, they discovered a viral variant with high virulence (High Type) that differed by 11 strongly linked SNPs from low virulence variants. They demonstrate that the High Type associates with a higher viral titer and increased host mortality, and also validate these findings using experimental infection assays. Using a transcriptomic approach, they show that the High Type overexpresses genes known to be linked to viral virulence, and this correlated with the under expression of host genes involved in antiviral immunity, indicating that the increased virulence of the High Type is at least partially due to the inhibition of host defense mechanisms. They further show that loci associated with differences in virulence were under strong selection for adaptation, particularly genes involved in the viral envelope and virulence proteins. Based on their reconstruction of the most likely evolutionary histories of the High Type in the different host populations, they further conclude that the High Type emerged multiple times independently, and yet, the High Type did not outcompete the Low virulence variants in these populations. This might indicate varying trade-offs between virulence and transmission in the High and Low Types that vary across these populations. Finally, they show that the same phenomenon for High Type evolution of DiNV can also be observed in other Drosophila species. I have just a few comments:

1) Throughout the manuscript, the authors switch back and forth between the present and past tense, which seems awkward from a stylistic point of, I'd suggest to stick to the past tense through-out.

2) The fact that the exact same 11 SNPs evolve multiple times independently in mostly the same order is interesting. The authors note that an expected alternative could be different SNPs emerging in the same genes, but they don't discuss the potential mechanism, by which the exact same SNPs seem to be preferred instead.

3) The authors found little evidence of mixed infection with both the High and Low Types and conclude potential in-compatibility. Please expand on the potential mechanisms of this.

4) Related to the above comment, the authors observe no "hybrids" of High and Low Types again, suggesting "incompatibility". Please discuss the difference between this genetic/genomic versus ecological incompatibility referred to above, as well as the potential link between these two types of incompatibilities.

5) Please rephrase the first sentence of the Discussion (what is the deference between "to better infect" and "optimizing the infection"?).

Reviewer #1:

[…] Below are a number of points that may help the authors to present their material more clearly.

The Introduction is not to the point. The Introduction moves among various aspects of host-parasite interactions without giving the reader an idea where this is going. At various places topics are raised that then later dismissed. For example: the first two paragraphs are about host response to viruses. In the third paragraph it become system specific, but now switches mainly to the DiNV system. It is not clear where one is going here. The Introduction (sixth paragraph) is summed up with a very generic phrase without much perspective on what is going to follow. At this place I still do not know what this paper will be about. A much more targeted Introduction is necessary. What are the questions? What was driving this research? What are the hypotheses?

We have rewritten the Introduction following your suggestions, thank you for the comments regarding this. We have moved sections around and rewritten parts so there is a better flow between paragraphs. We have also better provided our initial hypotheses and how these led to our findings. Specifically, we use the first paragraph to describe the host/virus coevolution and how viruses suppress the host, we then outline why people study viruses and why the DiNV system is ideal for viral study (with details of the system). Finally, we use the last paragraph to discuss the experiments we performed and a basic outline of our results.

With 11 SNPs spread across the chromosome and obligate recombination every round of replication, the proportion of viral offspring with sub-optimal multi-SNP-genotypes must be huge. How can the right genotypes be maintained? Discuss.

We have further expanded upon this in both the Results and the Discussion. In the Results we highlight that though there is obligate recombination, the chance that two strains with differing genotypes infect the same cell in an actual organism is very low, so we cannot know the actual rate or recombination, but it is likely lower than the high rate suggested and would lead to a lower proportion of intermediate strains. In the Discussion we also cover this and highlight that one possible difference between the High and Low strains is that they could infect different tissues (as is seen between other nudivirus strains). If the two types preferentially infect different tissues, then the chance of them co-infecting the same cell is impossible and would result in our observed absence of recombination between the High and Low type. We also discuss incompatibilities that could occur in suboptimal genotypes which could be the cause of their absence in our survey, or other factors which could drive the maintenance of the full haplotypes.

The authors run computer simulations (called SIR models, but they seem actually to be SI models) to support their ideas about virus evolution. The simulation regarding the accumulation of mutation is fine and gives quantitative support for presented evolutionary scenario.

A second simulation is about the competition of the two viral types. This is by far the weakest part of the manuscript. I am not convinced about the value of this simulation. The outcome can be predicted from the assumptions of the model, several of them are very speculative. Without data on transmission over the course of the infection, the simulations are not very helpful. I suggest to leave this out. It is ok to speculate about this in the Discussion, but it makes the Results part heavy and less strong. Also, the associated figure (Figure 7) is hard to understand.

We have now removed the second set of simulations and associated figures regarding the possibility of virulence/transmission equilibrium, as we agree it is the weakest part of the manuscript and is unlikely, as it is an unstable equilibrium. We now discuss the idea of a virulence/transmission equilibrium in the Discussion, however, we also discuss all the possible things that could have resulted in us finding two viral types (incomplete sweep, trade-off, optimized to different hosts).

The authors stress at multiple place that it is likely that the increased virulence of the high type is traded-off against transmission. The evidence for this is much weaker than the strength of these statements suggests. I strongly suggest to tone this down.

We have rewritten our Discussion to tone this down, in our Results we describe a relationship we find between increasing titer and increasing virulence. Beyond this, in the Discussion we examine multiple possible causes for the two maintained strains. We hypothesize that the two strains could be neutral and are drifting, that the High type is partway through a sweep, that a trade-off allows them to exist alongside each other or that each virus type is optimized to different host species/genotypes (with migration between two types). Overall, we think we have removed the majority of our Results and Discussion where talk about the trade-off, as it is not the only suggested explanation.

Reviewer #2:

[…] The practical experiments and sequencing data are substantial and appear sound, and the work is likely to be of interest to a very broad host-pathogen audience. However, while some of these headline claims are well-supported, I have a number of serious concerns about the analyses and interpretation of others, and a few key methodological details are missing. The analyses would need substantial revision, or at least additional checks, before I would be convinced by the story.

1) Is there any possibility of circularity in defining the 'high' and 'low' haplotypes?

Two 'types' are defined based on 11 linked SNPs that are described as having 'high' or 'low' titer phenotypes. First, this requires a more robust approach to define 'significance' as (if there is any LD) tests are not independent, and sample sizes relative to predictive SNPs are small. Phenotypes (titers) should be permuted across genotypes (within populations) a thousand times and the analyses re-run for each permutation, then the tails of this distribution used to define significance, as commonly done for e.g. the DGRP. Second, this feels like it is dangerously circular: supposing all 11 SNPs were false positives, and arbitrary haplotypes erroneously defined based on them? Post hoc analysis of the difference between the two haplotypes could still show Figure 1B, because those differences were what drove the (erroneous) detection of the SNPs. So would Figure 1C, since that result necessarily follows from the definition of the haplotypes.

There are two important considerations here so we will address them separately. We address the false negative problem below.

Defining significance since tests are not independent: Based on reviewer comments, we have significantly revised our approach at examining linkage disequilibrium and hope this clarifies some of the concerns. Given the signatures of LD, we did perform permutations and now include the permutation threshold in Figure 1A (slightly less stringent than the corrected P-value with a FDR of 0.01). So this was an important improvement but did not qualitatively change the result.

To convince me that the 'two haplotypes' interpretation is real, for each of the randomisation replicates one would need to define a 'high' and a 'low' haplotype based on the any randomisation spurious 'significant' SNPs and re-run the rest of Figure 1 to demonstrate that the separation between the real haplotypes is greater than that between those that result from permutation tests. I appreciate that this will require some computing time, but the authors already have all of the code, so it shouldn't be more than an afternoon of 'hands on' time.

The haplotype association problem: This is an important concern – if the first SNP is a false positive, they are all false positives. We think we have three convincing lines of evidence that these are not false positive SNPs. First, we have now performed the association study in the five populations (3 innubila, 1 azteca and 1 falleni) independently. Given the viral population structure among these populations (except maybe the sympatric innubila/azteca), these represent 5 independent tests and the haplotypes are significantly associated in all five populations. Thus a false negative would have to be a false positive 5 (or 4 if we don’t consider sympatric innubila/azteca as separate) times. Even if we consider our FDR threshold of 0.01 (these SNPs have lower P values), the likelihood of a false positive in 5 independent tests is 0.01^5 = 10^-10 or 1 in 10 billion (if 4 independent tests, that value drops to 10^-8 or 1 in 100 million). Second, we have followed the reviewer’s suggestions (we think) and permuted the titer associated with each strain 100,000 times, and each time binned strains by the most significant SNP and its 10 most closely linked SNPs, then find the difference in titer between two types. We find that the difference between the high and low type is significantly higher than by random chance for all SNPs, and no random combinations of alleles show the titer increase seen in Figure 1B. We have included the distribution of permuted differences between the artificial high and low types in Figure 1—figure supplement 2, with the true difference shown as a dotted line.

Since the haplotypes overlap in phenotype space (Figure 1B) and not all have all SNPs (Figure 1C), how were intermediate haplotypes (with <11 SNPs) assigned to high or low?

We have now changed the manuscript to consider intermediate strains to be a third subset of DiNV and have excluded them from analyses comparing High to Low type, including Figure 1B.

2) Did the haplotypes evolve independently three or four times?

As we have clarified in the manuscript, the haplotype has recurrently evolved three times to our knowledge in the Sky Islands, and at least once more in another location.

Of the 11 SNPs defining the haplotype, three are non-synonymous, five are in the UTRs of known virulence genes, and three are intergenic SNPs. Is it really credible that a specific base change has arisen and been selected for independently in each of four populations, at each of the 11 sites? i.e. always A->C being beneficial at a site, but never A->G at that site? Even for the non-coding ones? This is an extraordinary claim that would require extraordinary evidence, for which the rates of recombination and gene conversion seem at the heart of the argument.

We agree it is extraordinary that specific base changes always arising is strange, though this is not uncommon for viruses, especially given the low, but ubiquitous levels of codon usage bias seen in viruses. Additionally, it is possible that specific nucleotide changes are necessary for regulatory differences. We have discussed this in the manuscript and have also searched for gene conversion and find no evidence of it between the significant SNPs and surrounding SNPs (Figure 5—figure supplement 3). Our argument for increased effective mutation rate coupled to titer also increases the likelihood of this occurring, as we discuss in the manuscript.

In some places the authors appear to assume (or assert) that recombination is frequent, while in others they assume it is completely absent, and nowhere do they explicitly test for it. If it is common, then none of the tree-based analyses can be used. If it is absent, then it is very hard to explain the patterns of diversity or LD in Figure 1—figure supplement 3, and some of the simulations may be inappropriate.

We now address recombination in the second section of the manuscript (after the associations). The reviewer is correct that establishing and being clear about the recombination landscape is crucial for the rest of the manuscript. We have inserted also clarifications in the manuscript in the Discussion. In summary, while recombination events occur anywhere in the genome and frequently, the effective recombination rate is not as extremely high as we may have intimated, as most recombination events are between identical genomes. Recombination between two distinct haplotypes (including the low and high titer haplotypes) would require that the two types infect the same cell and this may be rare even if a host is superinfected. Even when recombination occurs between the two haplotypes, we hypothesize that recombinant genomes may have incompatibilities and so may not be able to leave the host cell. Based on these comments we have also changed our Discussion in the manuscript and have more thoroughly searched for evidence of recombination events between different SNPs. We find no evidence of recombination between populations, meaning each SNP must recurrently evolve in each population, however the final haplotype may have formed by bringing the SNPs onto the same background.

The tree-based analyses seem to be the basis of the major claim that these haplotypes arose independently multiple times, and that the order of the mutations arising was similar. Any tree analysis absolutely requires the absence of recombination or gene conversion, and this needs to be explicitly tested for here (e.g. using GARD, or possibly if diversity is low by inferring the ARG using tsinfer). The failure of LD to decay with distance hints that recombination is absent, but gene conversion over very short distances could still shuffle mutations between haplotypes. However, assuming its branch lengths are in mutations, the shape of the tree in Figure 6A (many short tips below large crowns) shouts 'recombination' to me. Where did this tree come from? It is not ultrametric, so if it was inferred with BEAST this is not standard output.

The reviewer is correct, we have now used GARD to identify recombination events and find evidence of recombination events between significantly associated SNPs and background SNPs. We also see evidence of recombination in our r² analyses between background SNPs but not our significant SNPs. We have also regenerated the phylogeny in BEAST2 using BDsky/skyline (now in Figure 5, appropriate given the recombination findings). We have also generated a phylogeny in BEAST2 considering recombination and get a very similar phylogeny to that seen now in Figure 5. In all cases we find recurrent evolution of each high type SNP in each population, but the combinations of SNPs may either be due to recurrent mutation or recombination. As we find no evidence of recombination between populations, even if the full complement of 11 SNPs has not recurrently evolved on the same background, we still find evidence of recurrent recombination of each mutation in each population.

If the authors do have evidence that recombination is absent, can they confirm that the other analyses (population size history from the SFS; simulations) also assumed zero recombination? Although zero recombination would make me worry even more about the p-values in the GWAS discovery of the SNPs that define the haplotypes – but permutation tests would help deal with that.

To identify recombination events and measure associations of alleles, we have performed the following tests/used the following tools:

– GARD

– R2 in r

– 4 allele test between SNP combinations

We have also factored in recombination in the following analyses and still find support for recurrent evolution of each SNP (though SNPs may have recombined to form the final High Type):

– BEAST2

– TreeTimes

– DeSolve Simulations

– StairwayPlot

We find no evidence of recombination between populations, e.g. we find no recombinant haplotypes which are ½ CH and ½ PR ; however we find recombination within populations. The results shown in Figure 1—figure supplements 2 and 3, suggests recombination is on average quite common. Based on the GARD results and BEAST2 analysis, recombination could have occurred between intermediate types in each population, which may have generated the full High type. However, for this to occur each mutation must recurrently occur in each population on the intermediate type.

I am really confused about their view of recombination, because a couple of times they imply recombination may be common, for example by noting that recombination is required in Nudivirus replication – but then this would invalidate the tree-based analyses? Plus, mechanistic recombination is irrelevant without coinfection infection, since recombination only between identical haplotypes does not have any effect. The observation that no co-infections occur suggests recombination should be very rare – but that's not what the tree looks like. The sixth paragraph of the Discussion suggests that not enough thought has been given to the likelihood, or implications, of recombination.

We think the confusion is the difference between actual recombination (the physical process) and effective recombination (the genetic signature of recombination between two distinct haplotypes). Coinfection is required for effective recombination but not actual recombination. We expect recombination between neutral genotypes to be common if coinfections are common, but do not see recombination between the High and Low types, suggesting either they do not co-infect or recombinants between the two types are inviable. We have included this in the Discussion and reworked our discussion of recombination throughout.

Finally, if there is no recombination, then 'fully derived' haplotypes (those with all 11 SNPs) can only be as old as the youngest of the 11 SNPs, unless the SNPs arose multiple time within populations as well. Is that compatible with the patterns of diversity and the timescale? One check would be to mask those 11, re-infer the tree, and check that the clades are still monophyletic. If these arose by re-current mutations due to strong selection, they could be warping the tree – this might be akin to the problem of recurrent drug or MHC selected mutations in HIV trees, where the known sites are excluded before analysis.

If we believe the two haplotypes are real, a much more credible 'story' to me would be two distinct and potentially old haplotypes maintained by selection in the face of a low level of ongoing gene conversion (or recombination). Is there some aspect of the data that this does not fit? This seems just as good a 'story', just as interesting, and just as publishable.

In all cases, the phylogeny has been generated without the 11 focal SNPs but was also generated with the SNPs and is nearly identical in both cases, we have now mentioned this in the manuscript. The idea of the two old haplotypes being maintained in the face of gene conversion is interesting, and we now discuss this in the manuscript. However, the idea of long-term maintenance does not fit with any analyses we have performed, as High types do not cluster together. Additionally, we find the High type is present in the geographically distinct Drosophila falleni population and clusters completely separately from the innubila samples.

We do find some evidence of recombination events occurring around the significantly associated SNPs, so it is possible that multiple intermediate strains (with different complements of SNPs) have recurrently evolved in each population and have recombined to generate the full High type, or that the High type has evolved once in each population and is eroded as you described. We address this possibility in both the Results and the Discussion. We now have a dedicated section of our Results regarding detecting recombination and inferring if the High type SNPs were brought together by recombination. We are unable to conclude if the High type was formed exclusively by recombination (likely both sequential mutation and recombination play a role), but find no evidence of recombination between populations, which suggests that even if recombination brings the SNPs together onto the same background, the SNPs must have recurrently evolved in each population.

3) The MK-like analyses

Although a relatively minor part of the story, the MK analysis is extremely unclear. In part this is because it doesn't appear in the Materials and methods.

i) What software was used? The supporting data implies VCFtools can do this (which surprised me) but the method it uses was not explained. Was SNIpre fitted with the original code? The text seem to imply that raw counts were used, which would not be suitable unless Ks was <0.3

ii) Although divergence is required, we're not told how (or from what species) it was estimated. If Ks>0.3, then something more than raw counts should certainly be used. If Ks>0.8, I would strongly advise against doing any sort of MK analysis.

iii) Figure 4 gives 'difference from background'. What is meant by difference from background? But if the background is a signal of strong constraint, could a positive signal here mean relaxed constraint rather than positive selection. What is the evidence that it's positive selection rather than relaxed constraint?

iv) In Figure 4—figure supplement 1 the populations are presented separately, and differences among them are interpreted as differences in positive selection. But surely the divergence number must be massive, larger than the polymorphism number, so almost all of this variation is due to differences in constraint affecting the Pn/Ps ratio. Or is the method one that uses high frequency derived SNPs as evidence of positive selection? If so, how was ancestral state identified? This is probably not possible to do reliably if Ks>0.3. We need to see some raw numbers, and a lot more detail on the methods.

We apologise, the MK-based analyses methods were removed during the writing of the manuscript accidentally, as our results have been reconfigured as 2 manuscripts from one large manuscript, we have added them back. VCFtools was not used, we instead used the original SnIPRE code and SNPeff SNP assignments. We have also explained what we mean by difference from the background, specifically that we found the average statistical measure for nearby genes and compared each focal gene to that, to see how much the focal gene deferred from its surrounding average. We have attempted to answer all your points and questions in this new section of the Materials and methods. Average dS from the outgroups used was 0.133 and 0.279 between DiNV-Kallithea and DiNV-OrNV respectively, which we feel is enough divergence to adequately perform the analysis and is not too much to result in all differences being due to the Pn/Ps ratio. We also used these comparisons to identify the ancestral state, based on the Kallithea and OrNV variant. We have now included our measures of Dn/Ds and Pn/Ps as well as the estimated Selection Effect for our genes of interest as a supplementary figure. In this case we find that envelope proteins and AMPs have an excess of Dn/Ds per site compared to other genes, which is driving their elevated selection effect, while an excess of Pn/Ps and a deficit of Dn/Ds in unknown function DiNV genes is likely driving most other differences. We have chosen to use MK-based tests as opposed to Dn/Ds to identify if the changes are due to positive selection rather than relaxed constraint, as Dn/Ds is weighted by Pn/Ps to identify the proportion of substitutions fixed by selection.

Reviewer #3:

[…] 1) Throughout the manuscript, the authors switch back and forth between the present and past tense, which seems awkward from a stylistic point of, I'd suggest to stick to the past tense through-out.

We apologise for this, we have corrected this inconsistency.

2) The fact that the exact same 11 SNPs evolve multiple times independently in mostly the same order is interesting. The authors note that an expected alternative could be different SNPs emerging in the same genes, but they don't discuss the potential mechanism, by which the exact same SNPs seem to be preferred instead.

We have expanded our discussion of the compatibility, both from an evolution perspective and from a functional perspective. First, we hypothesize that the significantly associated SNPs have epistatic effects and thus must occur in a particular order to traverse the fitness landscape form Low type to the High type. In line with this, very few genes show signatures of adaptation in DiNV, which could further limit what beneficial mutations fix. We also discuss the different conditions which could lead to two types of DiNV being found in each population. We also hypothesize some methods these incompatibilities could occur, such as a change in the 19K/PIF-6 protein structure which limits membrane access for viral particles without this variant and could be rescued by a matching change in another protein. Or these two changes together are neutral but in combination prevent membrane access for the other viral type. We also find evidence of recombination events around the significantly associated SNPs, which could easily facilitate the recombination of variants onto the same background, without needing to traverse a lower fitness genotype combination.

3) The authors found little evidence of mixed infection with both the High and Low Types and conclude potential in-compatibility. Please expand on the potential mechanisms of this.

We have expanded our discussion of the compatibility, both from an evolution perspective and from a functional perspective, we have also highlighted similar studies where diverged viruses are unable to coninfect similar systems or are unable to produce viable recombinant particles.

4) Related to the above comment, the authors observe no "hybrids" of High and Low Types again, suggesting "incompatibility". Please discuss the difference between this genetic/genomic versus ecological incompatibility referred to above, as well as the potential link between these two types of incompatibilities.

We have included this in our expanded section in the Discussion. Briefly we discuss how the incompatibility may not be caused by a functional effect, as most SNPs are upstream of genes, but could be due to an effect of the expression of the proteins in combination on viral particle formation, host fitness or even viral ability to transmit between hosts. We do not have zero intermediate types, but we have explained that these could be steps between two fitness peaks as the few intermediates without low fitness.

5) Please rephrase the first sentence of the Discussion (what is the deference between "to better infect" and "optimizing the infection"?).

We have attempted to clarify that we mean that viruses are evolving according to the transmission/virulence trade-off and so are not just evolving to propagate within hosts but to transmit to others, which usually requires the virus not kill the host before transmission can occur.

Author details

1. Tom Hill

   The Department of Molecular Biosciences, University of Kansas, Lawrence, United States

   Contribution

   Conceptualization, Data curation, Software, Formal analysis, Funding acquisition, Investigation, Visualization, Methodology, Writing - original draft, Project administration, Writing - review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-4661-6391

2. Robert L Unckless

   The Department of Molecular Biosciences, University of Kansas, Lawrence, United States

   Contribution

   Conceptualization, Supervision, Validation, Methodology, Writing - original draft, Project administration, Writing - review and editing

   For correspondence

   unckless@ku.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-8586-7137

• 1,430

  Page views

• 172

  Downloads

• 7

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.