Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Intraspecific host variation plays a key role in virus community assembly

Abstract

Infection by multiple pathogens of the same host is ubiquitous in both natural and managed habitats. While intraspecific variation in disease resistance is known to affect pathogen occurrence, how differences among host genotypes affect the assembly of pathogen communities remains untested. In our experiment using cloned replicates of naive Plantago lanceolata plants as sentinels during a seasonal virus epidemic, we find non-random co-occurrence patterns of five focal viruses. Using joint species distribution modelling, we attribute the non-random virus occurrence patterns primarily to differences among host genotypes and local population context. Our results show that intraspecific variation among host genotypes may play a large, previously unquantified role in pathogen community structure.

Introduction

Parasites constitute the majority of biological diversity on our planet1,2,3,4, and they influence both the demography and evolution of their host populations5,6,7. Host susceptibility, pathogen infectivity, and environmental favourability have been identified as the corner stones of disease within the disease triangle framework8. However, it is becoming increasingly clear that multiple infections within individuals are abundant3, and have the potential to change the evolutionary and epidemiological trajectories of pathogens9. Consequently, accounting for the diversity of infection is necessary to understand and predict disease dynamics and costs of infection for the host.

Understanding the determinants of the assembly and composition of pathogen communities is one of the key challenges in disease biology today. As a challenge it is analogous to the long-standing debate on the relative importance of biotic interactions versus external drivers of community dynamics. While some theories suggest species interactions to structure biological communities10,11, others highlight the importance of environmental drivers, including stress and disturbance on community dynamics12,13. To date, disentangling biotic processes from the abiotic ones has remained challenging14. In recent years, pathogens are increasingly studied within a community ecological framework15,16,17,18,19. Environmental variables and wider landscape context, such as human management, are linked to infection load, parasite diversity, and coinfection prevalence across multiple spatial scales18,20,21,22,23,24,25,26. The composition of parasite communities has also been linked to pathogen transmission mode, degree of host specialty, and life-cycle complexity27,28,29, as well as host history, phylogeny, geographical range, longevity, and growth strategy30,31,32,33,34,35,36,37,38. High parasite prevalence itself is a strong predictor of coinfections9,39. For vector-borne diseases, positive co-occurrence is common for pathogens that share a vector or transmission site, or when vectors show preference for already infected individuals15,40,41.

Co-occurrence of pathogens among host individuals is often non-random and coinfections can reach unexpectedly high levels15,20,23,42,43,44. One of the key challenges is to determine how biotic interactions between hosts and their pathogens themselves shape these distributions. Under the community ecological framework, a host can be viewed as a resource patch and its resistance as a local filter that determines the pathogen community within that host18. Hosts are resistant against most pathogen species they encounter45, and even for pathogens capable of infecting a host species, there is often considerable variation among individuals in their susceptibility7,46,47,48,49,50. The effect of intraspecific variation in disease resistance on the dynamics of individual pathogens is well described51,52,53,54. However, the importance of intraspecific host resistance variation for community assembly and diversity of species that exploit the host is only beginning to gain attention55. Due to allocation costs associated with genetically-based resistance, a host resistant against a particular pathogen may be susceptible to others56,57. On the other hand, limited evidence suggests that the same resistance loci may provide protection against several different pathogens58. Pathogens attacking the same host may also compete for host resources (resource-mediated interaction), or interact via elicited host immune responses59. Induced immunity by a first arriving pathogen may change the resistance phenotype, as immuno-suppression of the host by the first arriving pathogen may facilitate establishment and replication of later arriving pathogens60,61,62,63. On the other hand, cross-reactive immune responses elicited by the first parasite have the potential to suppress the success of later arriving parasites63,64. These biotic interactions could result in non-random pathogen co-occurrence patterns across host genotypes. Variation in host resistance may be spatially structured with pronounced differences in resistance observed among host populations53 and regions65. Such spatially structured resistance variation may also drive spatially structured co-occurrence patterns of pathogens exploiting the same host. Whether the host genotype is indeed a strong determinant of within-host parasite communities in the wild, and what the consequences of these within-host parasite community assembly processes are for host populations, remain unanswered18,66.

Here, we study the importance of the host genotype in determining the structure of within-host virus communities. Viruses are in principle obligate parasites as they require a host for reproduction. A growing body of evidence has demonstrated that consequences of virus infection can shift along the pathogenic–mutualistic-continuum, even for the same interaction67,68, and visually asymptomatic infections are common in wild plants3. Using cloned replicates of naive Plantago lanceolata plants as sentinel traps placed in natural populations during a seasonal epidemic of viruses, we can tease apart the role of the host genotype from drivers that affect the distribution of viruses within the local population context, which may include environmental variation, the local disease pool, host population structure and history, as well as local vector communities. Moreover, we aim to understand how biotic interactions among the viruses59,60,61,62,63,64 influence their community assembly.

We characterize the establishing virus communities using PCR detection69. We first test whether the viruses occur in the same sentinel plant more often than would be expected based on their frequencies alone. In other words, we test whether virus co-occurrence patterns differ from expectations of a random distribution. We then employ a joint species distribution modelling (JSDM) framework70, that allows us to tease apart the effect of local population context (consisting of unmeasured environmental variation as well as host population structure and history) on virus (co-)occurrences from host plant characteristics and host genotype. We can account for the shared environmental responses of the target species, which makes the model a robust method also for sparse data71. Using this approach, we are also able to capture signals of possible biotic community assembly processes from virus‐to‐virus association matrices after controlling for shared environmental responses of the viruses. The performance of JSDMs in relation to traditional, single-species distribution modelling (SDM) methods has recently been validated72. The application of these kinds of multivariate statistical tools— typically used in community ecological analysis—to parasite data has the potential to reveal new insights of the determinants of parasite community assembly and composition73,74.

In this study, we ask: (1) Do we see more (or less, respectively) co-occurrences between the viruses than what would be expected solely based on their frequencies?; (2) Does the local population context affect the virus community composition?; (3) Do host genotypes differ in the virus communities they acquire, suggesting genotype-level variation in overall sensitivity to infection?; (4) After accounting for the aforementioned effects (2–3) of the local population context and plant host characteristics (including the host genotype), is there evidence of residual virus co-occurrence patterns across the entire data indicative of competitive or facilitative virus interactions?; and (5) Do these residual co-occurrence patterns vary among host genotypes indicating genotype-specific resistance responses affecting virus community structure? Our results indicate that while the population context also drives virus community assembly, host genotypes vary in the virus communities they acquire.

Results

Detection of viruses in the field experiment

Out of the 320 sentinel host plants, 68% were hosts to at least one virus over the study period. Three viruses were clearly more common in the sentinel plants: closterovirus in 120 individuals, betapartitivirus in 102 individuals, and capulavirus in 84 individuals; while caulimovirus and enamovirus were rare: in 10 and 5 individuals, respectively (Figs. 1a and 2). Out of the 217 infected individuals, 49 (23%) hosted more than one virus, and in total, we found 17 virus combinations, ranging from single infections to four of the five viruses found in the same plant (Fig. 2). Both overall virus prevalence and the composition of virus communities varied among plant genotypes and plant populations (Figs. 1a, b and 2).

Fig. 1: Virus infections in sentinel plants.
figure1

Infections are plotted by population (a) and genotype (b), and the locations of the study populations in the field experiment in the Åland Islands (c). The genotypes and populations are ordered from left to right according to decreasing overall number of infections. ‘Clo’ refers to Plantago closterovirus, ‘Be’ to Plantago betapartitivirus, ‘Cap’ to Plantago lanceolate latent virus, ‘Cau’ to Plantago latent caulimovirus, and ‘En’ refers to Plantago enamovirus.

Fig. 2: (Co-)infections in the original data (upper panel) and predicted coinfections (lower panel) based on the model variant 2, ordered by host genotypes and population, as indicated by the X-axis.
figure2

Both the genotypes and populations are ordered with respect to frequency, so that the bars on the left-hand side show the population and genotype with the highest total amount of virus infection. ‘Clo’ refers to Plantago closterovirus, ‘Be’ to Plantago betapartitivirus, ‘Cap’ to Plantago lanceolate latent virus, ‘Cau’ to Plantago latent caulimovirus, and ‘En’ refers to Plantago enamovirus. The total number of plants in the upper panel is 20, whereas in the lower panel the total number is simulated plants is 20 (original number of plants) times 2000 (number of MCMC iterations used for the simulation), resulting in 40,000 simulated plants.

Analysis of virus co-occurrence

We found significant non-random positive co-occurrences between species pairs capulavirus and caulimovirus as well as betapartitivirus and caulimovirus, when we analysed the complete data set (Fig. 3). When we analysed the co-occurrences separately for each host plant genotype, we found positive co-occurrences between betapartitivirus and caulimovirus on genotype 609_19, as well as between betapartitivirus and capulavirus on genotype 2818_6. We also found negative co-occurrences between betapartitivirus and closterovirus, as well as capulavirus and closterovirus on plant genotype 2818_6 (Fig. 3). When analysing co-occurrence patterns within each population, we found a significant positive association between capulavirus and closterovirus, and negative association between closterovirus and betapartitivirus in plant population 433. The expected and observed numbers of co-occurrence, as well as the exact probabilities for a greater or smaller number co-occurrences than expected for these species pairs, are provided in Supplementary Table 3.

Fig. 3: Co-occurrences between virus species.
figure3

Co-occurrences are shown either in the whole data set (left, with total number of sentinel plants 320), or per plant genotype (upper panels, 80 plants per genotype), or by population (lower panels, 80 plants per population) as denoted by the horizontal axis. The genotypes and populations are ordered from left to right according to decreasing the overall frequency of disease. The plus (and minus) signs denote the pairs, for which the observed values were higher (or lower, respectively) than what would be expected based on their overall frequencies, and for which the probability of this difference was <0.1. The line colours denote the true numbers of co-occurrences between the species, as shown in the legend. ‘Clo’ refers to Plantago closterovirus, ‘Be’ to Plantago betapartitivirus, ‘Cap’ to Plantago lanceolate latent virus, ‘Cau’ to Plantago latent caulimovirus, and ‘En’ refers to Plantago enamovirus. The exact probabilities for the focal pairs are provided in the Supplementary Table 3.

Joint species distribution models of virus communities

The model variants 2 and 3 performed almost equally well, as seen from their performance (Table 1). Model variant 1, excluding host plant genotype as a covariate, was clearly inferior. Model variant 2 also resulted in the smallest WAIC value, implying best predictive power. We did not detect any significant residual co-occurrence patterns between viruses after accounting for the effect of the local population context and host-related variables. We looked into this with sentinel plant level latent variables that are uniform across sentinel plant genotypes (model variant 2) as well as sentinel plant level latent variables that covary with sentinel plant genotype (model variant 3), and neither of these model variants captured virus co-occurrences with strong statistical support and their explanatory performances did not differ. Based on these results, we decided to consider the simpler model variant 2 as our best model.

Table 1 Joint species distribution model variants and their explanatory performance and predictive performance (based on cross-validation), measured by the Tjur R2 coefficient of determination110 (see ‘Methods’ and Supplementary information).

The variance partitioning conducted for the model variant 2 revealed sentinel plant genotype to be the most important determinant for virus community composition (42% of variance explained, averaged over species; Fig. 4), followed by the local population context (29%; Fig. 4). The importance of variables differed between the viruses. Plant genotype explained most of the variation for capula- and caulimoviruses, while for enamovirus the sentinel plant genotype and the population context were almost equally important. For clostero- and betapartitiviruses the population context explained more variation in the data than plant genotype.

Fig. 4: Partitioning of the variance explained by model variant 2 (Table 1).
figure4

The diagram overlays the average proportions (over species) of variance explained by different groups of explanatory variables (out of the total variation explained by the model) and the concept of the disease triangle. The legend labels denote the different variables for which the partitioning is calculated, and the percentages indicate the mean values for the whole community. The barplot gives these results separately for each virus: the horizontal axis shows the focal five viruses (ordered from left to right according to their decreasing overall infection rate) and the vertical axis shows the proportion of variance explained. ‘Clo’ refers to Plantago closterovirus, ‘Be’ to Plantago betapartitivirus, ‘Cap’ to Plantago lanceolate latent virus, ‘Cau’ to Plantago latent caulimovirus, and ‘En’ refers to Plantago enamovirus.

The importance of the random effect at the level of sentinel plant individuals differed between the viruses, but followed roughly the same pattern: For capula-, caulimo- and enamovirus, the sentinel plant individual random effect was minor, but for clostero and betapartitivirus, its effect was slightly more pronounced (resulting in a total average effect of 14%). However, further inspection revealed that none of the residual correlations between virus species gained strong statistical support. Hence, we see no signal of potential biotic interactions between viruses after taking into account the effects of fixed explanatory variables, i.e. the sentinel plant genotype, size, signs of herbivory and local population context.

As expected, the predicted coinfections based on model variant 2 show similar patterns to what we can see in the raw data (Fig. 2). When examining both the coinfection profiles (Fig. 2), and the posterior mean estimates for the regression coefficients (Table 2), we see that capula- and caulimovirus are much more likely to occur on sentinel plant genotype 609_19 (with posterior mean estimate 2.47 for capula- and 0.67 for caulimovirus, Cap and Cau in Table 2, respectively, that gained strong statistical support based on the 90% central credible interval). Other sentinel plant genotypes were more dominated by single infections of closterovirus and betapartitivirus as well as their co-occurrences. Thus, the overall structure of the virus communities among plant genotypes was similar regarding the two most prevalent species closterovirus and betapartitivirus, but sentinel plant genotype 609_19 hosted significantly more capulavirus, which consequently also increases the probability of coinfections between capulavirus and other viruses. Regarding caulimovirus, six out of the total ten of its occurrences were together with capulavirus, and all of these co-occurrences were on sentinel plant genotype 609_19. Closterovirus, betapartitivirus and capulavirus are tenfold more prevalent in our data in comparison to caulimovirus and enamovirus, which can be seen in their dominance of the co-occurrence patterns in the community.

Table 2 Regression coefficients for model variant 2 for each virus species.

Sentinel plant size had a more minor effect on the community structure, as did signs of herbivory (Fig. 4), although both sentinelt plant size and herbivory did have a minor positive effect with strong statistical support on the probability of occurrence of closterovirus (Table 2).

Our result for the same set of model variants fitted with less conservative priors for the latent part of the model show corresponding results to our main variants: model variant 1 is clearly inferior, whereas there is no big difference between variants 2 and 3. With model variants 2 and 3, we are able to detect one association with strong statistical support, between betapartitivirus and caulimovirus. For more details, see our Supplementary information on the joint species distribution modelling.

Discussion

Understanding how pathogen communities are formed is a key challenge in understanding disease dynamics, as multiple infections can be significant drivers of epidemics as well as pathogen virulence and evolution9,18,19,75. The host is expected to be a strong determinant in the formation of pathogen communities, as both theory and controlled experiments have demonstrated host resistance to be a key determinant of disease dynamics76,77,78,79,80. Indeed, diversity of resistance in host populations could partly explain non-random co-occurrence patterns of pathogens detected in wild plants15,20,23,44,46. In our field experiment using sentinel plants of four genotypes, we found that most of the model-explained variation in virus occurrences was explained by the local population context and sentinel genotype (Fig. 4). Some viruses occurred significantly more or less together than would be expected based on their frequencies in both the full data set as well as when sentinel plant genotypes and local population context were analysed separately.

However, the results of our JSDM modelling (Table 1) indicate that the patterns evident in the co-occurrence analysis (Fig. 3) are influenced more clearly by the local population context and host genotype variation than by direct or indirect biotic interactions among the viruses. While disentangling host genotypic effects from other factors affecting pathogen communities has remained challenging, we were able to uncover the roles of these determinants of virus communities in wild hosts using naive sentinel plants in wild plant populations.

Of the total amount of variation explained with our best model variant, the population context explained within-host virus communities to a large extent, although the proportion of explained variation varied among the viruses (Fig. 4). Drivers that could vary among our plant population include abiotic variables which we did not explicitly record as many more plant populations would be needed to tease apart relevant variation in local population context for virus communities. These drivers are often found to filter parasites according to their niche preferences from the regional disease pool into local populations, thereby playing a major role in how within-population and -host-parasite communities are formed20,22,26. In addition to abiotic variables, the local P. lanceolata populations are likely to differ in biotic factors including plant species community composition and abundance of suitable vectors which may be linked to virus prevalence and diversity15,20. The local population context further includes any differences in population dynamics and trajectories, such as historical pathogen pressure, which may vary among these populations81. Albeit non-significant, the effect of sentinel plant individual on the (co-)occurrences of the viruses can be attributed to some unmeasured abiotic or individual-related variables, which may influence the (co-)occurrences of the viruses.

While there are multiple studies investigating within-host parasite communities44,69,73,74,82, to our knowledge the effect of host genotype on the assembly has rarely been tested experimentally in wild systems, or with multiple parasites simultaneously. In our data, sentinel plant genotype accounted for most of the variation in virus occurrences of the total variation explained in the JSDM model. Indeed, both virus occurrence, and the acquired virus communities varied among the four P. lanceolata genotypes. In particular, sentinel plant genotype 609_19 had greater infection prevalence and diversity of viruses than the other genotypes (Fig. 2). As our model controlled for the effect of sentinel plant size and level of herbivore damage, such host genotype-level differences may reflect variation in constitutive resistance, such as resistance genes, among the plant genotypes. The natural P. lanceolata populations in the Åland Islands contain considerable phenotypic variation in resistance against powdery mildew P. plantaginis83,84, and while resistance against viruses in this system is not well understood, an exceptionally diverse repertoire of candidate loci (Nucleotide-binding leucine-rich repeat; NLRs) that confer resistance against a broad range of pathogens, have been characterized in P. lanceolata (Laine, personal communication). Uncovering both phenotypic and molecular level virus resistance in this system is an important avenue of future research. Spatially structured variation in resistance is characteristic of natural host-parasite systems53,85,86,87, and based on our findings, intraspecific variation in disease resistance in a host population may play a large, previously unquantified role, in the non-random distribution of co-occurring pathogens that have been detected in the previous studies15,20,23,44,46.

Intraspecific variation in traits other than resistance could also generate the differences we observe. To confirm which traits are involved, future studies should explore in more detail the ecological outcomes of these interactions, and their molecular underpinning. It is highly plausible that the host genotype could indirectly affect virus occurrences via their attractiveness or resistance against vector herbivores88,89. Vector preference for infected hosts41,90 could also influence virus co-occurrence patterns. Transmission mode is often found to be critical for how pathogen communities are formed15,40,91,92, and reciprocally, the amount of genotypic variation within a host population may explain the abundance and composition of herbivory community present89.

A community of pathogens could be shaped by both direct and indirect pathogen–pathogen associations: reaction triggered by an earlier arrival could either induce or suppress resistance against later arriving pathogens, or within-host competition could favour one pathogen over the other. Evidence for both negative and positive pathogen–pathogen interactions have been reported in studies of multiple infections19,59,62,63,93,94. Although we find both positive and negative co-occurrence patterns among the viruses, these are largely explained by local population context and host genotype. After controlling for these in our model, we do not find strong statistical support for signals of associations among the viruses, as would be expected if arrival by one would decrease or increase the arrival probability of another. Hence, our results do not support the hypothesis that virus–virus interactions—either direct or those mediated by host immunity—would be the key drivers of virus community assembly at the within-host level in this system. However, our sample size could be insufficient to detect such interactions as some of our viruses are rare, and their arrival probability to the sentinel plants is also subject to random processes. In addition, we only accounted for a subset of all possible pathogens infecting plants in this system, thereby potentially missing some influential members of the community. Furthermore, the effects of induced immunity triggered by a first arriving pathogen may be short-lasting63,95 and, therefore, undetectable with the timescale of this experiment. Induced immunity could play a more important role among viruses of the same genus or strains of the same virus species, where the famous phenomenon of cross-protection is more often recorded63,96 and as is predicted by theory75. Given that the variants with less conservative priors detected a significant positive interaction between betapartitivirus and caulimovirus, we conclude that our study design was successful in capturing the effects of the host genotype, but larger-scale investigations would be required to detect signals of virus–virus interactions.

In our experimental design we kept the plants in their pots which meant these plants experienced different rooting environments than the wild plants but allowed us to standardize some factors (e.g., soil medium). However, this approached allowed us to control for this level of variation in our data. Our approach may have affected vector preferences as visual presentation of the plants, in addition to other cues, is important for vector dynamics90. Nonetheless, transmission of all five focal viruses to the sentinel plants did occur. Whether the virus prevalences we detected with our approach are in line with infections of wild plants is difficult to assess, given that virus prevalences vary greatly among populations in the Åland Islands (0–64%)69. Overall, our study does not only highlight the importance of the host genotype, but also the need for further research on other aspects of virus ecology. Although we have placed the current work into a context of pathogens, viruses may also have neutral or positive effects on the hosts despite their parasitic lifestyle67. While knowledge of virus diversity and roles of viruses in wild populations is increasing3,44,67, research at the community scale remains scarce18.

Here, we have quantified the importance of intraspecific host plant variation on how within-host virus communities assemble by using sentinel plants in natural populations during a seasonal epidemic, which allows teasing this factor apart from other drivers of virus occurrence. Applying JSDMs to interpret the effects of host genotype and local population context, we find that while the population context has a strong influence on virus communities within individual hosts, not accounting for the host genotype might underestimate the role host genotypes have in generating variation in pathogen communities. Such variation in within-host pathogen diversity may have far reaching implications for all key aspects of disease: transmission, virulence suffered by the host, and pathogen evolution. With these results, we are one step closer to binding together the different spatial scales and processes that underpin pathogen metacommunities.

Methods

Study species

Plantago lanceolata is a globally occurring perennial herbaceous plant97. It is an obligate outcrosser with wind-dispersed pollen, also capable of vegetative reproduction97. In the Åland Islands, SW of Finland, it typically grows on dry meadows, forming a network of approximately 4000 small connected populations81. The size and location of the populations have been monitored since the early 1990s as a part of the metapopulation studies of the Glanville fritillary butterfly and powdery mildew Podosphaera plantaginis81,98. In the Åland Islands, P. lanceolata also hosts a diverse community of viruses that vary in their occurrence among P. lanceolata populations and among the individuals within populations69. We used five recently characterized viruses from the Åland Islands, to study within-host viral communities69: Plantago lanceolata latent virus in genus Capulavirus and Plantago lanceolata caulimovirus in genus Caulimovirus with DNA-genomes, and Plantago betapartitivirus in genus Betapartitivirus, Plantago enamovirus in genus Enamovirus, and Plantago closterovirus in genus Closterovirus with RNA-genomes. The viruses are hereafter referred to by their genus for understandability. These viruses were initially identified from P. lanceolata in the Åland Islands by sequencing plant small RNAs69. Plants use RNA-silencing mechanism and produce short interfering RNA (SiRNA) molecules in a defense response against viral infection99. Hence, these viruses trigger an active defense response in P. lanceolata. Also, although not directly demonstrating their pathogenic nature, Susi et al.69 found that plants with virotic symptoms (necrotic spots/yellow colour) are more likely to carry a virus infection. Currently, the detailed transmission dynamics and vector species, as well as the viruses’ distribution outside the Åland Islands remain unknown. More detailed information of the virus families is compiled in Supplementary Table 1.

Field experiment with sentinel plants of different genotypes

To study the effect of plant host genotype on the variation of within-host virus communities, we set up an experiment using sentinel trap plants in natural populations of P. lanceolata in the Åland Islands. To obtain genetically uniform plant material, we cloned four greenhouse-grown maternal P. lanceolata plants into 80 replicates each. The maternal plants originate from natural P. lanceolata populations in the Åland Islands, and were grown from seeds in an insect free greenhouse at the University of Helsinki. The plants are expected to represent four different genotypes (ID:s 609_19, 4_13, 511_14, 2929_6), as their maternal plants originated from distant populations 7–40 kilometres apart. Their resistance against viruses is currently unknown, but they represent different mildew resistance phenotypes as has been confirmed during laboratory maintenance of P. plantaginis. The maternal plant individuals used in the experiment were confirmed to be free of target viruses, that would have been the result of seed-borne infection, by PCR-testing using specific primers. Each maternal plant was cloned into 80 replicates by placing maternal plants on pots containing vermiculate and kept on a tray containing fertilized water.

After one month, the roots grown from the maternal plant’s pot through to the vermiculate were cut. After another month, new plants shooting from the cut roots in the vermiculate were separated and individually planted into 10 cm × 10 cm pots containing an equal amount of sand and potting soil. After two additional months in the greenhouse, during the last week of May 2017, the plants were taken to the Åland Islands and placed into four P. lanceolata populations (ID:s 877, 9031, 433, 3302; Fig. 1c). The populations were selected for the study as they represent different parts of the Åland Islands, were remote to humans, and large enough to host a field-experiment. These populations were different from the ones the maternal plants used for cloning originated from. These four populations were included in the analyses as a categorical variable to capture ‘local population context’ (local temperature, vectors, plant communities, etc.) that may influence virus distributions among P. lanceolata populations in the Åland Islands.

Twenty replicates of each sentinel plant genotype were placed into each of the four P. lanceolata populations resulting in 80 plants per population, and 320 plants altogether. The plants were kept in their pots for the duration of the experiment, and they were placed in a random order among natural vegetation and reshuffled three times per week to avoid within-population spatial effects. The plants were kept separated from the local soil on plastic freezer boxes and watered when necessary. Signs of herbivory (holes, bitemarks, and thrip damage) were recorded after two weeks of exposure, and again after seven weeks of exposure. Plant size was measured during the first week of exposure by counting the number of leaves, and by measuring the length and width of the longest leaf. Based on these measurements we calculated plant size by using the equation n × A, where n is the number of leaves, and leaf area A is calculated using the equation of ellipse area: A = πab, where a is a half axis of the width of the longest leaf, and b is the half axis of the length of the longest leaf. For those 13 plants missing measurement data, an average over all recorded values for all plants was used, in order to not to lose any virus occurrence data from the analysis.

Nucleic acid extractions and virus detections with PCR

To detect the viruses infecting plants during the growing season, leaf samples were collected for nuclear acid extractions after two weeks and again after 7 weeks of exposure to the natural virus and vector communities. Samples were collected from a single leaf of similar age (young but large enough for sampling) from each plant. For DNA extraction, we collected a 1 cm² piece of leaf from each plant. Samples were stored in −20 °C until DNA extraction with E.Z.N.A. Plant Kit (Omega Biotek, USA) at the Institute of Biotechnology at University of Helsinki. For RNA extractions, 3 cm² leaf samples were collected, immediately deep-frozen in liquid nitrogen, and stored in −80 °C before RNA-extraction. Total RNA was extracted using phenol-chloroform extraction with a modified method from Chang et al.100. Two additional phenol cleaning steps prior chloroform cleaning of the RNA were performed. In the additional cleaning steps, we used 800 ml of equal volumes of phenol solution (pH 4.5) and chloroform-isoamylalcohol, mixed with isolation buffer containing the sample, vortexed, and centrifuged for phase-separation in 14 800 rpm for 15 min. For the PCR detection of the RNA viruses, RNA was translated into cDNA. For reverse transcription, we used 2 ng of total RNA, mixed with 2 µl random hexamer primers (Promega) and sterile nuclease free water in 17,125 µL volume incubated for 5 min in 70 °C. Subsequently, 1 µL Moloney Murine Leukemia Virus Reverse Transcriptase (M-MLV RT; Promega Corporation, USA), 5 µL M-MLV RT buffer, 1.25 µL of dNTP (10 mM) mix, and 0.625 µL of RiboLock RNaseinhibitor were added and the 37.41 µL reaction mix was incubated in 37 °C for 60 min. For virus detection PCR, we used specific primers69,101 as well as two additional primer pairs for capulavirus (PiLVi2_forward_1 5′-GTGTTTAACAATGAAGTGAGCC-3′ and PiLVi2_reverse_4 5′-AATCCATCCACACATCCAATC-3′) and caulimovirus (forward primer 5′-AGGAGATGCCCATACTTTACC-3′ and reverse primer 5′-GACTTGCCAGAACCTGATTTAC-3′). PCR reactions to detect viruses were performed in final volume of 10 µL containing of 1–3 µL of DNA or cDNA, and GoTaq Green® polymerase 5x Mastermix (Promega Corporation, USA) according to manufacturer’s instructions. Samples were subjected to initial denaturation in 95 °C for 2 min, following 35 cycles of denaturation in 95 °C for 40 s, annealing 53–60 °C for 40 s, and extension 72 °C for 1 min with a final extension step of 72 °C for 5 min. The full protocol with virus specific PCR conditions is described in the Supplement (section ‘PCR-detection of viruses’). The amplicons were resolved on a 1.2–1.5% agarose gel and visualized using Gel Doc XR System (Bio-Rad Laboratories, Inc., USA).

Statistical analysis

For all the statistical analysis, we pooled the detected occurrences of the five focal viruses over the two timepoints of sampling by collapsing the occurrence data so that each sentinel plant had one observed virus community. Only when a sentinel plant had not been infected by a certain virus in either of the timepoints accounted as an absence of the virus while infection in one or both timepoints was accounted for as virus presence. To understand whether the co-occurrence of viruses differs from expected co-occurrences calculated solely from the prevalences of these viruses, we first analysed the co-occurrence patterns both in the full data set as well as separately for each sentinel host genotype and plant population (Fig. 3). We used the R package ‘cooccur’102 and its identically named function, and applied a probabilistic model103 which calculates expected frequencies of species co-occurrences based on a distribution of random, independent species. By comparing the expected and observed co-occurrences the applied algorithm gives the probabilities of co-occurrence greater than or less than what is observed in the data analytically, without relying on randomisations or test statistics, under the condition that the probability of occurrence for a species at each sentinel plant is equal to its observed frequency among all the sentinel plants, i.e. in this case the prevalence of the virus103.

For addressing our study questions about the effects of host genotype and characteristics as well as local population context on the (co-)occurrence patterns of the viruses, as well as the possible signals of biotic interactions between the viruses on virus community assembly, we applied a joint species distribution modelling (JSDM) framework ‘Hierarchical Modelling of Species Communities’ (HMSC104), which is a multivariate Bayesian hierarchical generalised linear latent variable model. Essentially, HMSC is a multivariate generalised linear model, enabling the modelling of the whole community of viruses as opposed to fitting individual single-species distribution models105. In addition, HMSC is a latent variable model70. Latent variable models include unobserved, i.e. latent predictors, which are typically included to model correlation, or to account for missing predictors70. Hence, in this context, the latent variables are random effects that model the co-occurrences between species due to either biotic interactions or some other effects not included in the fixed part of the model, such as unmeasured effects of the environment. For a more detailed description of JSDMs and latent variable models, please see the comprehensive review by Warton et al.70.

The structure of the HMSC modelling framework is described in detail by Ovaskainen et al.104,106, with connections to community ecological theory and case studies. In our study, we modelled the virus community, denoted by the n × ns matrix Y of virus occurrences, comprising of individual components yij, denoting virus j = 1 ,…, ns, where ns = 5, on host plant i = 1 ,…, n, where n = 320, with probit regression

$$y_{{{ij}}} = 1_{L_{{{ij}}} + \varepsilon _{{{ij}}} > 0}$$
$$L_{{{ij}}} = L_{{{ij}}}^F + L_{{{ij}}}^R$$

where εijN(0,1), Lij is the linear predictor for the occurrence of virus j on sentinel plant i, which is further divided to fixed (\(L_{{{ij}}}^F\)) and random (\(L_{{{ij}}}^R\)) parts. The fixed effects F model the influence of the local population context and the influence of the sentinel plant characteristics. The random effect R models the residual variation in virus occurrences at the level of individual sentinel plants, that cannot be attributed to the above-described responses of the viruses to the fixed covariates. For exact formulation how the different components are modelled, with corresponding notation, please see Ovaskainen et al.106.

Briefly, following the compact matrix notation of Chapter 7.3.2 in Ovaskainen et al.106, we model the n × ns community matrix of viruses Y with a n × ns matrix L of all linear predictors Lij for all species and sentinel plants, as L = LF + LR. The matrix of fixed effects can be further decomposed as LF = XB, where X is the n × nc matrix of environmental covariates, and B is the nc × ns matrix of regression coefficients, i.e. species responses to the covariates, and nc is the total amount of covariates included in the model. Because the environmental covariates X are known and given as input for the model (Table 1), only the species responses B are estimated. Analogously, the matrix of random effects can be decomposed as LR = . Here, H is the n × nf matrix of latent factors, or site loadings, and Λ is the nf × ns matrix of latent factor loadings, or, where nf is the number of latent factors. Both the site H and Λ are estimated, as is the number of latent factors nf. The species loadings Λ can then be translated into residual associations between virus species by transforming them into covariation between species as Ω = ΛTΛ, and further into correlations.

We fitted three JSDM variants to the data by varying the way the sentinelt plant genotype was included in the model (Table 1). As explanatory variables (denoted by matrix X in ref. 71) we used the local plant population context (categorical variable with four classes), which is a proxy for the plant population-level effects, such as variation in abiotic conditions, vector communities, and disease pool (categorical variable with four classes); and at the level of the sentinel host plants, we include the plant size (a continuous variable), signs of herbivory (a categorical variable with two classes; yes/no), as well the genotype of the sentinel host plant (a categorical variable with four classes). To examine the residual co-occurrence patterns among hosts, we also included the sentinel plant individual as a latent variable random effect.

First, we fitted a model with only the local population context, plant size and signs of herbivory (variant 1) as fixed explanatory variables X. Then, we fitted a model including also the sentinel plant genotype, i.e. the full set of fixed explanatory variables (variant 2). With both of these model variants (1 and 2), we included random effects at the level of sentinel plants individuals. Finally, we fitted a model with the same full set of fixed explanatory variables X as with model variant 2, but we modified the random effects by allowing these residual patterns to covary with the sentinel plant genotype (variant 3), details of which are explained by Tikhonov et al.107. In this case, the latent factor loadings Λ are furthermore modelled as a linear regression of the selected fixed explanatory variables, which in this case was the sentinel plant genotype. Hence, as a summary, our model variants vary in terms of what is included in the matrix X of explanatory variables, namely if sentinel plant genotype is included (variant 2) or not (variant 1), and do we allow the residual associations between viruses to covary among the sentinel genotype (variant 3) or not (variant 2).

We used the default priors of the package ‘Hmsc’108, except that for the parameter Λ of species loadings, of the random part of the model. While the HMSC framework is usually not very sensitive to the choices of priors, when data is sufficient, they can be sensitive to the prior chosen for Λ. The multiplicative gamma process shrinking prior109 for the species loadings Λ has several prior parameters, but out of those, the user is advised to pay attention to the choice of α, a vector of two values, which can be used to adjust the level of shrinkage that the prior implies for the matrix Ω of species associations106. Hence, we used two alternative priors. First, we used the default of α = (50, 50), which imposes a lot of shrinkage. We refer to this group of model variants as our main model variants. Second, we used α = (3,3), which imposes much less shrinkage, but as a trade-off, also increases the risk of overfitting.

The model variant comparison approach allows us to examine the relevance of sentinel plant genotype as a predictor of virus community composition (comparison of model variants 1 and 2), as well as to see whether the residual co-occurrences between the viruses differ between the sentinel plant genotypes (variant 3). The comparison of different priors enables us to examine how sensitive our models were for these choices. We compared the model variants in terms of their explanatory and predictive performance, where the first tells us how well the model predicts the data used to fit it, whereas the latter illustrates how well the model predicts independent data which has not been used for model fitting. We calculated the Tjur R2 coefficient of determination, a statistic that has been recommended to be used as a standard measure of explanatory power for binary outcomes110. The coefficient is obtained by calculating the mean of the predicted probabilities of presences and absences, and then taking the difference between those two means. Hence, a high coefficient value implies high predicted probabilities for presences and low probabilities for absences. When interpreting it, it is good to note that with sparse data, the probabilities of presence tend to be low in the first place, and thus the Tjur R2 coefficient can remain rather low as well. Nevertheless, if the model is completely uninformative and predicts a 50% probability for both presence and absence, the coefficient value will be zero, thus revealing a poor model fit. For examining explanatory power, we fit the model to the full data set and base our comparison on predictions made for the same data. To examine the predictive power of the model, we conducted a 10-fold cross-validation and compared the model variants based on the same Tjur R2 coefficient as with explanatory power, but calculated from the predictions made to new, unknown host plants. To complement our comparison based on model accuracy, we calculated the widely applicable information criterion (WAIC)111 for all the variants.

We also conducted a partitioning of the variance explained by the best-performing model variant, to assess how different (groups of) variables are contributing to the overall variance explained by the model at the level of the linear predictor. Finally, we used the best-performing model variant to simulate predicted coinfections profiles.

We implemented our analyses with the R package ‘Hmsc’ (version 3.0-7108). The performance comparison, variance partitioning and predictions were conducted with the tools provided in the package. For a full formal description of the structure of the modelling framework, please see Ovaskainen et al.104,106, and for the covariate-dependent latent variables used in model variant 3, please see Tikhonov et al.107. The analytical pipeline and an R package along with the data used is available in Zenodo (https://doi.org/10.5281/zenodo.4117739). For all the statistical analysis, we used R version 4.0.0112. For more details on the statistical analysis, please see Supplementary information (section ‘Supplementary information on the joint species distribution modelling’).

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

The data supporting our results along with the analytical pipeline implemented as an R package are archived in Zenodo (https://doi.org/10.5281/zenodo.4117739).

References

  1. 1.

    Dobson, A., Lafferty, K. D., Kuris, A. M., Hechinger, R. F. & Jetz, W. Homage to Linnaeus: How many parasites? How many hosts? Proc. Natl Acad. Sci. USA 105, 11482–11489 (2008).

    ADS  CAS  Google Scholar 

  2. 2.

    Kreuze, J. F. et al. Complete viral genome sequence and discovery of novel viruses by deep sequencing of small RNAs: a generic method for diagnosis, discovery and sequencing of viruses. Virology 388, 1–7 (2009).

    CAS  Google Scholar 

  3. 3.

    Prendeville, H. R., Ye, X., Jack Morris, T. & Pilson, D. Virus infections in wild plant populations are both frequent and often unapparent. Am. J. Bot. 99, 1033–1042 (2012).

    Google Scholar 

  4. 4.

    Treena, I. B. et al. Distribution and diversity of phytophthora across Australia distribution and diversity of phytophthora across Australia. Pac. Conserv. Biol. 23, 1–13 (2017).

    Google Scholar 

  5. 5.

    Anderson, R. M. & May, R. M. Infectious Diseases of Humans; Dynamics and Control (1991).

  6. 6.

    Hudson, P. J., Dobson, A. P. & Newborn, D. Prevention of population cycles by parasite removal. Science 282, 2256–2258 (1998).

    ADS  CAS  Google Scholar 

  7. 7.

    Thrall, P. H. et al. Rapid genetic change underpins antagonistic coevolution in a natural host-pathogen metapopulation. Ecol. Lett. 15, 425–435 (2012).

    Google Scholar 

  8. 8.

    Stevens, R. B. in Plant Pathology, an advanced treatise (eds Horsfall, J. G. & Dimond, A. E.) 357–429 (1960).

  9. 9.

    Susi, H., Barrès, B., Vale, P. F. & Laine, A.-L. Co-infection alters population dynamics of infectious disease. Nat. Commun. 6, 5975 (2015).

    ADS  CAS  Google Scholar 

  10. 10.

    Connell, J. H. Diversity in tropical rain forests and coral reefs. Science 199, 1302–1310 (1978).

    ADS  CAS  Google Scholar 

  11. 11.

    Tilman, D. Resource competition and community structure. Monogr. Popul. Biol. 17, 1–296 (1982).

  12. 12.

    Chesson, P. Mechanisms of maintenance of species diversity. Annu. Rev. Ecol. Syst. 31, 343–366 (2000).

    Google Scholar 

  13. 13.

    Grime, P. J. Plant Strategies, Vegetation Processes, and Ecosystem Properties (2001).

  14. 14.

    Ovaskainen, O., Rybicki, J. & Abrego, N. What can observational data reveal about metacommunity processes? Ecography 42, 1877–1886 (2019).

    Google Scholar 

  15. 15.

    Seabloom, E. W., Hosseini, P. R., Power, A. G. & Borer, E. T. Diversity and composition of viral communities: coinfection of barley and cereal yellow dwarf viruses in California grasslands. Am. Nat. 173, E79–E98 (2009).

    Google Scholar 

  16. 16.

    Mihaljevic, J. R. Linking metacommunity theory and symbiont evolutionary ecology. Trends Ecol. Evol. 27, 323–329 (2012).

    Google Scholar 

  17. 17.

    Johnson, P. T. J., de Roode, J. C. & Fenton, A. Why infectious disease research needs community ecology. Science 349, 1259504 (2015).

    Google Scholar 

  18. 18.

    Borer, E. T., Laine, A.-L. & Seabloom, E. W. A Multiscale approach to plant disease using the metacommunity concept. Annu. Rev. Phytopathol. 54, 397–418 (2016).

    CAS  Google Scholar 

  19. 19.

    Tollenaere, C., Susi, H. & Laine, A.-L. Evolutionary and epidemiological implications of multiple infection in plants. Trends Plant Sci. 21, 80–90 (2016).

    CAS  Google Scholar 

  20. 20.

    Seabloom, E. W., Borer, E. T., Lacroix, C., Mitchell, C. E. & Power, A. G. Richness and composition of niche-assembled viral pathogen communities. PLoS ONE 8, 1–9 (2013).

    Google Scholar 

  21. 21.

    Borer, T., Seabloom, E. W., Mitchell, C. E. & Power, A. G. Local context drives infection of grasses by vector-borne generalist viruses. Ecol. Lett. 13, 810–818 (2010).

    Google Scholar 

  22. 22.

    Richgels, K. L. D., Hoverman, J. T. & Johnson, P. T. J. Evaluating the role of regional and local processes in structuring a larval trematode metacommunity of Helisoma trivolvis. Ecography 36, 854–863 (2013).

    Google Scholar 

  23. 23.

    Rodelo-Urrego, M. et al. Landscape heterogeneity shapes host-parasite interactions and results in apparent plant-virus codivergence. Mol. Ecol. 22, 2325–2340 (2013).

    CAS  Google Scholar 

  24. 24.

    Bernardo, P. et al. Geometagenomics illuminates the impact of agriculture on the distribution and prevalence of plant viruses at the ecosystem scale. ISME J. 12, 173–184 (2018).

    Google Scholar 

  25. 25.

    Makiola, A. et al. Land use is a determinant of plant pathogen alpha‐ but not beta‐diversity. Mol. Ecol. 28, 3786–3798 (2019).

    Google Scholar 

  26. 26.

    Cottenie, K. Integrating environmental and spatial processes in ecological community dynamics. Ecol. Lett. 8, 1175–1182 (2005).

    Google Scholar 

  27. 27.

    Dobson, A. Population dynamics of pathogens with multiple host species. Am. Nat. 164, 64–78 (2004).

    Google Scholar 

  28. 28.

    Malpica, M. & Sacrista, S. Association and host selectivity in multi-host pathogens. PLoS ONE 1, e41 (2006).

    ADS  Google Scholar 

  29. 29.

    Poulin, R., Blanar, C. A., Thieltges, D. W. & Marcogliese, D. J. Scaling up from epidemiology to biogeography: Local infection patterns predict geographical distribution in fish parasites. J. Biogeogr. 39, 1157–1166 (2012).

    Google Scholar 

  30. 30.

    Nunn, C. L., Altizer, S., Jones, K. E. & Sechrest, W. Comparative tests of parasite species richness in primates. Am. Nat. 162, 687–614 (2003).

    Article  Google Scholar 

  31. 31.

    Ezenwa, V. O., Price, S. A., Altizer, S., Vitone, N. D. & Cook, K. C. Host traits and parasite species richness in even and odd-toed hoofed mammals, Artiodactyla and Perissodactyla. Oikos 115, 526–536 (2006).

  32. 32.

    Gilbert, G. S. & Webb, C. O. Phylogenetic signal in plant pathogen – host range. Proc. Natl Acad. Sci. USA 104, 4979–4983 (2007).

    ADS  CAS  Article  Google Scholar 

  33. 33.

    Lindenfors, P. et al. Parasite species richness in carnivores: effects of host body mass, latitude, geographical range and population density. Glob. Ecol. Biogeogr. 16, 496–509 (2007).

    Article  Google Scholar 

  34. 34.

    Mitchell, C. E., Blumenthal, D., Jarošík, V., Puckett, E. E. & Pyšek, P. Controls on pathogen species richness in plants’ introduced and native ranges: roles of residence time, range size and host traits. Ecol. Lett. 13, 1525–1535 (2010).

    Article  Google Scholar 

  35. 35.

    Endara, M.-J. & Coley, P. D. The resource availability hypothesis revisited: a meta-analysis. Funct. Ecol. 25, 389–398 (2011).

    Article  Google Scholar 

  36. 36.

    Miller, Z. J. Notes and comments fungal pathogen species richness: why do some plant species have more pathogens than others? Am. Nat. 179, 282–292 (2012).

    ADS  Article  Google Scholar 

  37. 37.

    Dallas, T. A. & Presley, S. J. Relative importance of host environment, transmission potential and host phylogeny to the structure of parasite metacommunities. Oikos 123, 866–874 (2014).

    Article  Google Scholar 

  38. 38.

    Strong, D. R. & Levin, D. A. Species richness of plant parasites and growth form of their hosts. Am. Nat. Nat. 114, 1–22 (2019).

  39. 39.

    Louhi, K. R., Karvonen, A., Rellstab, C., Louhi, R. & Jokela, J. Prevalence of infection as a predictor of multiple genotype infection frequency in parasites with multiple-host life cycle. J. Anim. Ecol. 82, 191–200 (2013).

    Article  Google Scholar 

  40. 40.

    Raybould, A. F., Maskell, L. C., Edwards, M. L., Cooper, J. I. & Gray, A. J. The prevalence and spatial distribution of viruses in natural populations of Brassica oleracea. N. Phytol. 141, 265–275 (1999).

    Article  Google Scholar 

  41. 41.

    Shoemaker, L. G. et al. Pathogens manipulate the preference of vectors, slowing disease spread in a multi-host system. Ecol. Lett. 22, 1115–1125 (2019).

    Article  Google Scholar 

  42. 42.

    Shaw, D. J. & Dobson, A. P. Patterns of macroparasite abundance and aggregation in wildlife populations: a quantitative review. Parasitology 111, S111–S127 (1995).

    Article  Google Scholar 

  43. 43.

    Klein, J., Satta, Y. & Uigin, O. The molecular descent of the major histocompatibility complex. Annu. Rev. Immunol. 11, 269–295 (1993).

    CAS  Article  Google Scholar 

  44. 44.

    Kamitani, M., Nagano, A. J., Honjo, M. N. & Kudoh, H. RNA-Seq reveals virus-virus and virus-plant interactions in nature. FEMS Microbiol. Ecol. 92, 1–11 (2016).

    Article  CAS  Google Scholar 

  45. 45.

    Mysore, K. S. & Ryu, C. M. Nonhost resistance: How much do we know? Trends Plant Sci. 9, 97–104 (2004).

    CAS  Article  Google Scholar 

  46. 46.

    Remold, S. K. Unapparent virus infection and host fitness in three weedy grass species. J. Ecol. 90, 967–977 (2002).

  47. 47.

    Bergelson, J., Kreitman, M., Stahl, E. A. & Tian, D. Evolutionary dynamics of plant R-Genes. Plant Pathol. 292, 2281–2286 (2001).

    CAS  Google Scholar 

  48. 48.

    Laine, A.-L. Detecting local adaptation in a natural plant-pathogen metapopulation: a laboratory vs. field transplant approach. J. Evol. Biol. 20, 1665–1673 (2007).

    Article  Google Scholar 

  49. 49.

    Mandadi, K. K. & Scholthof, K. B. G. Plant immune responses against viruses: How does a virus cause disease? Plant Cell 25, 1489–1505 (2013).

    CAS  Article  Google Scholar 

  50. 50.

    Bruns, E., Carson, M. & May, G. Pathogen and host genotype differently affect pathogen fitness through their effects on different life-history stages. BMC Evol. Biol. 12, 135 (2012).

    Google Scholar 

  51. 51.

    Strauss, A. T., Bowling, A. M., Duffy, M. A., Cáceres, C. E. & Hall, S. R. Linking host traits, interactions with competitors and disease: mechanistic foundations for disease dilution. Funct. Ecol. 32, 1271–1279 (2018).

    Google Scholar 

  52. 52.

    Mundt, C. C. Use of multiline cultivars and cultivar mixtures for disease management. Annu. Rev. Phytopathol. 40, 381–410 (2002).

    CAS  Google Scholar 

  53. 53.

    Jousimo, J. et al. Disease ecology. Ecological and evolutionary effects of fragmentation on infectious disease dynamics. Science 344, 1289–1293 (2014).

    ADS  CAS  Google Scholar 

  54. 54.

    Ekroth, A. K. E., Rafaluk-Mohr, C. & King, K. C. Host genetic diversity limits parasite success beyond agricultural systems: a meta-analysis. Proc. R. Soc. B Biol. Sci. 286, 20191811 (2019).

    Google Scholar 

  55. 55.

    Bolnick, D. I. et al. Why intraspecific trait variation matters in community ecology. Trends Ecol. Evol. 26, 183–192 (2011).

    Google Scholar 

  56. 56.

    Rausher, M. D. Tradeoffs in performance on different hosts: evidence from within- and between-site variation in the beetle Deloyala guttata. Evolution 38, 582–595 (1984).

    Article  Google Scholar 

  57. 57.

    Stearns, S. C. Trade-offs in life-history evolution. Funct. Ecol. 3, 259–268 (1989).

    Article  Google Scholar 

  58. 58.

    Zhou, X. et al. Loss of function of a rice TPR-domain RNA-binding protein confers broad-spectrum disease resistance. Proc. Natl Acad. Sci. USA 115, 3174–3179 (2018).

    CAS  Article  Google Scholar 

  59. 59.

    Karvonen, A., Jokela, J. & Laine, A. L. Importance of sequence and timing in parasite coinfections. Trends Parasitol. 35, 109–118 (2019).

    Article  Google Scholar 

  60. 60.

    Van Hulten, M., Pelser, M., Van Loon, L. C., Pieterse, C. M. J. & Ton, J. Costs and benefits of priming for defense in Arabidopsis. Proc. Natl Acad. Sci. USA 103, 5602–5607 (2006).

    ADS  Google Scholar 

  61. 61.

    Graham, A. L. Ecological rules governing helminth-microparasite coinfection. Proc. Natl Acad. Sci. USA 105, 566–570 (2008).

    ADS  CAS  Google Scholar 

  62. 62.

    Laine, A.-L. Context-dependent effects of induced resistance under co-infection in a plant-pathogen interaction. Evol. Appl. 4, 696–707 (2011).

    Google Scholar 

  63. 63.

    Syller, J. Facilitative and antagonistic interactions between plant viruses in mixed infections. Mol. Plant Pathol. 13, 204–216 (2012).

    Google Scholar 

  64. 64.

    Porrozzi, R., Teva, A., Amaral, V. F., Santos Da Costa, M. V. & Grimaldi, G. Cross-immunity experiments between different species or strains of Leishmania in rhesus macaques (Macaca mulatta). Am. J. Trop. Med. Hyg. 71, 297–305 (2004).

    Google Scholar 

  65. 65.

    Burdon, A. J. J., Thrall, P. H. & Brown, A. H. D. Resistance and virulence structure in two linum marginale-melampsora lini host-pathogen metapopulations with different mating systems. Evolution 53, 704–716 (1999).

    CAS  Google Scholar 

  66. 66.

    Lively, C. M., de Roode, J. C., Duffy, M. A., Graham, A. L. & Koskella, B. Interesting open questions in disease ecology and evolution. Am. Nat. 184, S1–S8 (2014).

  67. 67.

    Roossinck, M. J. & Bazán, E. R. Symbiosis: viruses as intimate partners. Annu. Rev. Virol. 4, 123–139 (2017).

    CAS  Google Scholar 

  68. 68.

    Hily, J. M., Poulicard, N., Mora, M. Á., Pagán, I. & García-Arenal, F. Environment and host genotype determine the outcome of a plant-virus interaction: From antagonism to mutualism. N. Phytol. 209, 812–822 (2016).

    Google Scholar 

  69. 69.

    Susi, H., Filloux, D., Frilander, M. J., Roumagnac, P. & Laine, A.-L. Diverse and variable virus communities in wild plant populations revealed by metagenomic tools. PeerJ 2019, e6140 (2019).

    Google Scholar 

  70. 70.

    Warton, D. I. et al. So many variables: joint modeling in community ecology. Trends Ecol. Evol. 30, 766–779 (2015).

    Google Scholar 

  71. 71.

    Ovaskainen, O. & Soininen, J. Making more out of sparse data: hierarchical modeling of species communities. Ecology 92, 289–295 (2011).

    Google Scholar 

  72. 72.

    Norberg, A. et al. A comprehensive evaluation of predictive performance of 33 species distribution models at species and community levels. Ecol. Monogr. 89, e01370 (2019).

    Google Scholar 

  73. 73.

    Aivelo, T. & Norberg, A. Parasite–microbiota interactions potentially affect intestinal communities in wild mammals. J. Anim. Ecol. 87, 438–447 (2018).

  74. 74.

    Dallas, T. A., Laine, A., Ovaskainen, O. & Dallas, T. A. Detecting parasite associations within multi-species host and parasite communities. Proc. R. Soc. B Biol. Sci. 286, 20191109 (2019).

    Google Scholar 

  75. 75.

    Alizon, S., de Roode, J. C. & Michalakis, Y. Multiple infections and the evolution of virulence. Ecol. Lett. 16, 556–567 (2013).

    Google Scholar 

  76. 76.

    Thrall, P. H. & Burdon, J. J. Effect of resistance variation in a natural plant host-pathogen metapopulation on disease dynamics. Plant Pathol. 49, 767–773 (2000).

    Google Scholar 

  77. 77.

    Thrall, P. H., Burdon, J. J. & Young, A. Variation in resistance and virulence among demes of a plant host-pathogen metapopulation. J. Ecol. 89, 736–748 (2001).

    Google Scholar 

  78. 78.

    Lively, C. M. The effect of host genetic diversity on disease spread. Am. Nat. 175, E149–E152 (2010).

  79. 79.

    Benavides, J. A. et al. From parasite encounter to infection: multiple-scale drivers of parasite richness in a wild social primate population. Am. J. Phys. Anthropol. 147, 52–63 (2012).

    Google Scholar 

  80. 80.

    Susi, H. & Laine, A. L. The effectiveness and costs of pathogen resistance strategies in a perennial plant. J. Ecol. 103, 303–315 (2015).

    Google Scholar 

  81. 81.

    Ojanen, S. P., Nieminen, M., Meyke, E., Pöyry, J. & Hanski, I. Long-term metapopulation study of the Glanville fritillary butterfly (Melitaea cinxia): survey methods, data management, and long-term population trends. Ecol. Evol. 3, 3713–3737 (2013).

    Google Scholar 

  82. 82.

    Jackson, J. A., Pleass, R. J., Cable, J., Bradley, J. E. & Tinsley, R. C. Heterogenous interspecific interactions in a host–parasite system. Int. J. Parasitol. 36, 1341–1349 (2006).

    CAS  Google Scholar 

  83. 83.

    Susi, H., Vale, P. F. & Laine, A.-L. Host genotype and coinfection modify the relationship of within and between host transmission. Am. Nat. 186, 000–000 (2015).

    Google Scholar 

  84. 84.

    Laine, A. Resistance variation within and among host populations in a plant–pathogen metapopulation: implications for. J. Ecol. 92, 990–1000 (2004).

    Google Scholar 

  85. 85.

    Greischar, M. A. & Koskella, B. A synthesis of experimental work on parasite local adaptation. Ecol. Lett. 10, 418–434 (2007).

    Google Scholar 

  86. 86.

    Hoeksema, J. D. & Forde, S. E. A meta-analysis of factors affecting local adaptation between interacting species. Am. Nat. 171, 275–290 (2008).

    Google Scholar 

  87. 87.

    Laine, A.-L., Burdon, J. J., Dodds, P. N. & Thrall, P. H. Spatial variation in disease resistance: from molecules to metapopulations. J. Ecol. 99, 96–112 (2011).

    Google Scholar 

  88. 88.

    Smith, C. M. & Boyko, E. V. The molecular bases of plant resistance and defense responses to aphid feeding: current status. Entomol. Exp. Appl. 122, 1–16 (2007).

    CAS  Google Scholar 

  89. 89.

    Crutsinger, G. M. et al. Plant genotypic diversity predicts community structure and governs an ecosystem process. Science 647, 966–968 (2006).

    ADS  Google Scholar 

  90. 90.

    Mauck, K., Bosque-Pérez, N. A., Eigenbrode, S. D., De Moraes, C. M. & Mescher, M. C. Transmission mechanisms shape pathogen effects on host-vector interactions: Evidence from plant viruses. Funct. Ecol. 26, 1162–1175 (2012).

    Google Scholar 

  91. 91.

    Susi, H. & Laine, A. L. Host resistance and pathogen aggressiveness are key determinants of coinfection in the wild. Evolution (N. Y). 71, 2110–2119 (2017).

    Google Scholar 

  92. 92.

    Bergner, L. M. et al. Demographic and environmental drivers of metagenomic viral diversity in vampire bats. Mol. Ecol. 29, 26–39 (2019).

    Google Scholar 

  93. 93.

    Telfer, S. et al. Species interactions ina parasite community drive infection risk in a wildlife population. Science 330, 243–247 (2010).

    ADS  CAS  Google Scholar 

  94. 94.

    Halliday, F. W. et al. Facilitative priority effects drive parasite assembly under coinfection. Nat. Ecol. Evol. https://doi.org/10.1038/s41559-020-01289-9 (2020).

  95. 95.

    Mauch-Mani, B., Baccelli, I., Luna, E. & Flors, V. Defense priming: an adaptive part of induced resistance. Annu. Rev. Plant Biol. 68, 485–512 (2017).

    CAS  Google Scholar 

  96. 96.

    Cassells, A. C. & Herrick, C. C. Cross protection between mild and severe strains of tobacco mosaic virus in doubly inoculated tomato plants. Virology 78, 253–260 (1977).

    CAS  Google Scholar 

  97. 97.

    Sagar, A. G. R. & Harper, J. L. Plantago Major L., P. Media L. and P. Lancoeolata L. J. Ecol. 52, 189–221 (1964).

    Google Scholar 

  98. 98.

    Laine, A.-L. & Hanski, I. Large-scale spatial dynamics of a specialist plant pathogen in a fragmented landscape. J. Ecol. 94, 217–226 (2006).

    Google Scholar 

  99. 99.

    Baulcombe, D. RNA silencing in plants. Nature 431, 356–363 (2004).

    ADS  CAS  Google Scholar 

  100. 100.

    Chang, S., Puryear, J. & Cairny, J. A simple and efficient method for isolating RNA from pine trees. Plant Mol. Biol. Report. 11, 113–116 (1993).

    CAS  Google Scholar 

  101. 101.

    Susi, H. et al. Genome sequences of a capulavirus infecting Plantago lanceolata in the Åland archipelago of Finland. Arch. Virol. 162, 2041–2045 (2017).

    CAS  Google Scholar 

  102. 102.

    Griffith, D. M., Veech, J. A. & Marsh, C. J. cooccur: probabilistic species co-occurrence analysis in R. J. Stat. Softw. 69, 1–17 (2016).

    Google Scholar 

  103. 103.

    Veech, J. A. A probabilistic model for analysing species co-occurrence. Glob. Ecol. Biogeogr. 22, 252–260 (2013).

    Google Scholar 

  104. 104.

    Ovaskainen, O. et al. How to make more out of community data? A conceptual framework and its implementation as models and software. Ecol. Lett. 2, 561–576 (2017).

    Google Scholar 

  105. 105.

    Elith, J. & Leathwick, J. R. Species distribution models: ecological explanation and prediction across space and time. Annu. Rev. Ecol. Evol. Syst. 40, 677–697 (2009).

    Google Scholar 

  106. 106.

    Ovaskainen, O. & Abrego, N. Joint Species Distribution Modelling: With Applications in R. Ecology, Biodiversity and Conservation (Cambridge University Press, 2020).

  107. 107.

    Tikhonov, G., Abrego, N., Dunson, D. & Ovaskainen, O. Using joint species distribution models for evaluating how species-to-species associations depend on the environmental context. Methods Ecol. Evol. 8, 443–452 (2017).

    Google Scholar 

  108. 108.

    Tikhonov, G. et al. Hmsc: hierarchical model of species communities. R. package version 3, 0–7 (2020).

    Google Scholar 

  109. 109.

    Bhattacharya, A. & Dunson, D. B. Sparse Bayesian infinite factor models. Biometrika 98, 291–306 (2011).

    MathSciNet  CAS  MATH  Google Scholar 

  110. 110.

    Tjur, T. Coefficients of determination in logistic regression models—a new proposal: the coefficient of discrimination. Am. Stat. 63, 366–372 (2009).

    MathSciNet  MATH  Google Scholar 

  111. 111.

    Watanabe, S. A widely applicable bayesian information criterion. J. Mach. Learn. Res. 14, 867–897 (2013).

    MathSciNet  MATH  Google Scholar 

  112. 112.

    R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org (2020).

Download references

Acknowledgements

We thank Mikko Jalo, Pauliina Hyttinen, and Vanja Milenkovic for helping with data collection in the field. We thank Pauliina Hyttinen and Laura Häkkinen for help with RNA-extractions. We thank Krista Raveala for help in cloning and caring for the plants. The Institute of Biotechnology at University of Helsinki is acknowledged for carrying out DNA-extractions. This research was supported by funding from the Academy of Finland (296686) and European Research Council (4100097 RESISTANCE) to A.-L.L., and Luova Doctoral Programme Fellowship to S.S.

Author information

Affiliations

Authors

Contributions

S.S., A.-L.L., and H.S. designed the study. S.S. performed the experiment and data collection. A.N. performed the statistical analysis. S.S., A.N., and A.-L.L. wrote the paper and H.S. contributed substantially with comments on the paper.

Corresponding author

Correspondence to Suvi Sallinen.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Peer review information Nature Communications thanks Carolyn Malmstrom, Alex Strauss and the other, anonymous, reviewer for their contribution to the peer review of this work. Peer reviewer reports are available.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Sallinen, S., Norberg, A., Susi, H. et al. Intraspecific host variation plays a key role in virus community assembly. Nat Commun 11, 5610 (2020). https://doi.org/10.1038/s41467-020-19273-z

Download citation

Further reading

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing