Skip to main content
  • Loading metrics

Effect of Biodiversity Changes in Disease Risk: Exploring Disease Emergence in a Plant-Virus System

  • Israel Pagán,

    Affiliation Centro de Biotecnología y Genómica de Plantas (UPM-INIA), and E.T.S.I. Agrónomos, Campus de Montegancedo, Universidad Politécnica de Madrid, Pozuelo de Alarcón, Madrid, Spain

  • Pablo González-Jara,

    Affiliation Centro de Biotecnología y Genómica de Plantas (UPM-INIA), and E.T.S.I. Agrónomos, Campus de Montegancedo, Universidad Politécnica de Madrid, Pozuelo de Alarcón, Madrid, Spain

  • Alejandra Moreno-Letelier,

    Affiliation Departamento de Ecología Evolutiva, Instituto de Ecología, Universidad Nacional Autónoma de México, Ciudad de México, México

  • Manuel Rodelo-Urrego,

    Affiliation Centro de Biotecnología y Genómica de Plantas (UPM-INIA), and E.T.S.I. Agrónomos, Campus de Montegancedo, Universidad Politécnica de Madrid, Pozuelo de Alarcón, Madrid, Spain

  • Aurora Fraile,

    Affiliation Centro de Biotecnología y Genómica de Plantas (UPM-INIA), and E.T.S.I. Agrónomos, Campus de Montegancedo, Universidad Politécnica de Madrid, Pozuelo de Alarcón, Madrid, Spain

  • Daniel Piñero,

    Affiliation Departamento de Ecología Evolutiva, Instituto de Ecología, Universidad Nacional Autónoma de México, Ciudad de México, México

  • Fernando García-Arenal

    Affiliation Centro de Biotecnología y Genómica de Plantas (UPM-INIA), and E.T.S.I. Agrónomos, Campus de Montegancedo, Universidad Politécnica de Madrid, Pozuelo de Alarcón, Madrid, Spain


The effect of biodiversity on the ability of parasites to infect their host and cause disease (i.e. disease risk) is a major question in pathology, which is central to understand the emergence of infectious diseases, and to develop strategies for their management. Two hypotheses, which can be considered as extremes of a continuum, relate biodiversity to disease risk: One states that biodiversity is positively correlated with disease risk (Amplification Effect), and the second predicts a negative correlation between biodiversity and disease risk (Dilution Effect). Which of them applies better to different host-parasite systems is still a source of debate, due to limited experimental or empirical data. This is especially the case for viral diseases of plants. To address this subject, we have monitored for three years the prevalence of several viruses, and virus-associated symptoms, in populations of wild pepper (chiltepin) under different levels of human management. For each population, we also measured the habitat species diversity, host plant genetic diversity and host plant density. Results indicate that disease and infection risk increased with the level of human management, which was associated with decreased species diversity and host genetic diversity, and with increased host plant density. Importantly, species diversity of the habitat was the primary predictor of disease risk for wild chiltepin populations. This changed in managed populations where host genetic diversity was the primary predictor. Host density was generally a poorer predictor of disease and infection risk. These results support the dilution effect hypothesis, and underline the relevance of different ecological factors in determining disease/infection risk in host plant populations under different levels of anthropic influence. These results are relevant for managing plant diseases and for establishing conservation policies for endangered plant species.

Author Summary

Biodiversity has been proposed as a major ecological factor determining disease prevalence. However, the relationship between biodiversity and disease risk remains underexplored. Few studies focus on host-virus systems and, particularly on plant viruses. To address this subject the prevalence of virus infection and disease symptoms was monitored in wild-pepper (chiltepin) populations under different levels of human management. For these populations, species diversity, host genetic diversity and host plant density were determined. Higher levels of human management resulted in increased disease and virus infection risk, which was associated with decreased habitat species diversity and host genetic diversity, and with increased host plant density. More specifically, for wild chiltepin populations, species diversity of the habitat was the primary predictor of disease risk; and host genetic diversity was the primary predictor in managed populations, with host density being generally a poorer predictor of disease risk. These results support a dilution effect of biodiversity on disease risk, and underline the relevance of different ecological factors in determining disease risk in wild and in human-managed habitats.


Understanding the relationship between the risk of infectious diseases and host ecology is a long-standing goal of biological research, central for the management of current infectious diseases and for preventing the emergence of new ones. Indeed, changes in host ecology are among the most frequently identified causes of disease emergence (i.e. the increase of disease incidence following its appearance in a new, or previously existing, host population) [1][3]. Because infectious diseases involve interactions between at least two species, it has been proposed for a long time that ecosystem biodiversity will play a key role in disease risk. Current declines in biodiversity have been proposed to be linked with the emergence of infectious diseases, which have fueled a renewed interest on this subject [4]. Two major hypotheses with different predictions relate biodiversity to disease risk. The “Amplification Effect” hypothesis predicts that diversity will be positively correlated with disease risk, as it will result in increased abundance of inoculum sources for a focal host. The “Dilution Effect” hypothesis predicts a negative correlation between biodiversity and disease risk, as a reduction in diversity could result in an increased abundance of the focal host species facilitating disease transmission [5]. These two hypotheses can be considered to represent extremes of a continuum, as the effects of diversity on disease risk would be related to the host range of the pathogen: an Amplification Effect would require a generalist pathogen, while the more restricted the host range of the pathogen, or the higher the differences between shared hosts in their ability to amplify or transmit the pathogen, the higher the Dilution Effect. Increasing evidence derived from pathogens with broadly different life-styles indicates that biodiversity reductions most often result in increased disease risk [4].

The idea linking biodiversity with disease risk is not new in animal or plant pathology. Two classical hypotheses in plant pathology state that the high impact of plant diseases in crops is associated with: i) the reduced species diversity, and higher host density, of agroecosystems as compared to wild ecosystems [6]; ii) the reduced genetic diversity of crops as compared to their wild ancestors or relatives [7]. However, despite that a number of recent studies on the ecology of plant diseases have been added to those dating from the 1980s, support for these hypotheses is still often circumstantial [8]. Attention has focused on analyses of foliar diseases caused by fungi, which mostly indicate that increased biodiversity reduces disease risk [9][14]. Remarkably, there are fewer reports referring to viral diseases, which represent a large fraction of emergent plant pathogens [15], and may differ from fungal ones in their relationship to biodiversity. While most plant pathogenic fungi are directly transmitted specialists [16], most plant-infecting viruses are vector transmitted, and are host generalists but often vector specialists [17]. Most studies with plant viral diseases have focused on generalist viruses infecting grasses, generally finding an amplification effect [18][21]. Interestingly, work on plant diseases largely failed to assess the role of various possible mechanisms by which reduced biodiversity may affect disease risk (but see [10], [11], [22]). Particularly, it is often difficult to differentiate the effects of increased host density and of reduced species diversity [4]. Hence, there is a need of research aimed at analyzing the effects of biodiversity on plant disease risk and, specifically, at disentangling the role of the various factors associated to ecosystem diversity. This is the goal of the present work.

The focal host in this study is the wild pepper Capsicum annuum var. glabriusculum (Dunal) Heiser and Pickersgill [23], also known as “chiltepin”. Chiltepin is found in Mexico in a variety of habitats from the Yucatan peninsula and the Gulf of Mexico to the Sonoran desert [24], [25]. Chiltepin is a deciduous, perennial bush that grows for 5–8 years and vegetates and reproduces during the rainy season. Birds disperse the seeds from its red pungent fruits [24]. Human harvesting of fruits from wild chiltepin plants is a common practice in central and northern Mexico [26], [27]. A second level of human exploitation involves tolerance or favoring the growth of spontaneously dispersed chiltepin plants in anthropic habitats, such as pastures and living fences (i.e., let-standing plants, sensu [28]). Last, chiltepin cultivation in home gardens or in small traditional plots has started in the recent past [25]. Cultivation has not yet lead to domestication, and cultivated chiltepin populations, which are managed as annual crops, do not show obvious phenotypic differences with wild ones [25]. Wild chiltepin populations show a large genetic variation and a strong spatial structure associated with the biogeographical province of origin, and human management results in a significant loss of both spatial structure and genetic diversity [25]. This habitat diversity makes chiltepin a uniquely good system to analyze the relationship between biodiversity and disease risk.

We focused on two contrasting pepper-infecting virus groups. The first involves two species of the genus Begomovirus (Geminiviridae): Pepper golden mosaic virus (PepGMV), and Pepper huasteco yellow vein virus (PHYVV), here treated collectively as “begomoviruses”. These species have a two-segmented single-stranded (ss) DNA genome; narrow host ranges limited in nature mostly to species of the genera Capsicum, Solanum and Datura (Solanaceae), and are transmitted in a persistent manner by the whitefly Bemisia tabaci Gennadius (Homoptera, Aleyrodidae) [29][31]. The B biotype of B. tabaci, characterized by a broad plant host range, a high reproductive potential, and a high efficiency as a vector for begomoviruses, is prevalent in Central and North America [32]. The second virus is Cucumber mosaic virus (CMV), genus Cucumovirus (Bromoviridae), with a tripartite, ssRNA genome, and a typical generalist, infecting more than 1000 species of both mono- and dicotyledonous plant families. CMV is transmitted in a non-persistent manner by more than 80 species of aphids, thus being also a vector generalist [33].

Utilizing these host-pathogen systems we specifically addressed if: i) modification of chiltepin habitat associated with different levels of human management resulted in changes in disease or infection risk, ii) reduction of species diversity increases disease or infection risk, iii) decreased host genetic diversity had an effect on disease or infection risk, iv) increased host plant density resulted in increased disease or infection risk and v) the above effects were different for viruses with different life-histories.

To answer these questions, we visited over three years neighboring wild, and human managed (i.e., of let standing and cultivated plants) chiltepin populations from different biogeographic provinces in Mexico. For each population, species diversity, host density and the prevalence of plants showing symptoms of virus infection were quantified in the field as an estimate of disease risk. Plants were collected at each population and their status (infected/non-infected by several viruses) was determined in the laboratory in order to estimate infection risk. Results indicate that disease and begomovirus infection risks, but not CMV infection risk, decrease with increasing biodiversity. We propose that observed differences between begomovirus and CMV infection risk can be due to different transmission modes.

Materials and Methods

Field Surveys and Plant Collections

Chiltepin populations were visited during the summers of 2007 to 2009 at different sites over the species distribution range in Mexico (Figure 1 and Table 1). A total of 26 populations were localized in different habitats representing three levels of human management: i) ten wild populations (W) in which fruit gathering by local people may occur; ii) six populations of let-standing plants (here from called “let-standing populations”), in anthropic habitats, either pastures (LSP) or live fences (LSF), in which chiltepin plants are tolerated or favored, and iii) ten cultivated populations (C) either at home gardens (CHG) or at small monocultures (CMC). Population sites were assigned to 6 biogeographical provinces: Yucatan (YUC), Eastern side of the Sierra Madre Oriental (SMO), Altiplano Zacatecano-Potosino (AZP), Costa del Pacífico (CPA), Costa del Pacífico Sur (CPS) and Sonora (SON) [34]. A total of 14 populations, located in YUC, SMO, AZP, CPA and CPS, were visited during 2007 and 2008. The 2009 survey was extended to other populations of these five biogeographical provinces and SON, to a total of 26 populations (Table 1). Populations were visited between the 15 of July and the 30 of August, in an attempt to homogenize plant phenology among locations at the stage of flowering and beginning of fruit setting. Due to the highly unpredictable rain regime at some regions, or to extinction, not all populations could be surveyed for the three years.

Figure 1. Geographic location of chiltepin populations, and prevalence of symptomatic plants, begomoviruses and CMV.

Map shows the location of populations from wild (W), let standing (LSP, LSF), and cultivated (CMC, CHG) populations within six biogeographical provinces in Mexico. Bar graphics show the average prevalence of symptomatic (grey) and asymptomatic (black) plants, as well as the prevalence of begomovirus (green) and CMV (blue) infection, for each chiltepin population. Boxes group populations from the same biogeographical province, and are colored accordingly.

Table 1. Prevalence of virus-like symptoms (i.e., mosaic, leaf curl, leaf lamina reduction, and/or stunting) in the analyzed Mexican chiltepin populations monitored in July–August of 2007 to 2009.

At each location, the following information was collected: 1) The census of the chiltepin population. 2) The status of each censused plant: asymptomatic or showing symptoms commonly related to virus infection (i.e., mosaic, leaf curl, leaf lamina reduction, and/or stunting). 3) The area (m2) occupied by the chiltepin population. 4) The inventory of the non-herbaceous vegetation, determined as the number of individuals of each bushy or arboreal species, in the same area of the chiltepin population, to estimate species richness, and evenness according to the Shannon index [35]. Populations BER-w, PEL-w, MOC-w and MAU-w (Table 1) were too large – i.e., more than 200 plants – to census all plants, and both chiltepin censuses and biodiversity inventories were limited to a fixed transect. In this case, the area occupied by the chiltepin population was calculated by prospecting a width of 4 m along the fixed itinerary.

At each population and visit, plants were systematically sampled for laboratory analyses. Plants were sampled regardless of their showing or not symptoms: One plant out of every x plants was sampled along fixed itineraries, with itinerary length and x (0<x≤4) depending on population size, 1–3 young branches with fresh leaves were collected per plant.

Virus Detection and Plant Genotyping

Infection by CMV and by Potyvirus species was analyzed by DAS-ELISA, using commercial antisera against CMV or a monoclonal antibody against a highly conserved motif in the coat protein of potyviruses (Agdia Biofords), according to the manufacturer's instructions. Infection by Chiltepin yellow mottle virus (ChYMV, Tymoviridae) was analyzed by molecular hybridization using a 32P-labeled RNA probe complementary to nucleotides 5365–5777 of ChYMV genomic RNA (Accession No. FN563124) [36]. Infection by species of the genus Begomovirus was detected by PCR using degenerate primers designed on the alignment of DNA-A sequences of 43 begomovirus species from the New World: BAOPsp (5′-GCGCCCTGCAGGGGCCYATGTAYAGGAAGCC-3′) and BAONsp (5′-GCGCGCGGCCGCGANGCATGNGTACATGCCAT-3′), which amplify a region in the coat protein gene located between nucleotide positions 392 and 884 in the genome of PepGMV (Accession No. AY928512).

Molecular hybridizations and PCR were performed on total nucleic acid preparations from chiltepin leaves extracted by grinding 200 mg of fresh leaf tissues in three volumes of 200 mM Tris-HCl pH 9, 25 mM EDTA, 1% SDS, 400 mM LiCl, followed by phenol-chloroform extraction [25]. Plants genotyped using the set of 9 nuclear microsatellites markers described in [25] were used to estimate genetic diversity of the 26 chiltepin populations.

Statistical Analyses

Generalized linear mixed models (GLMM) were used to analyze the difference in the prevalence of virus infection (Begomovirus and CMV), and in the frequency of symptomatic plants, according to chiltepin population, biogeographical province and level of human management of the population, considering these factors as fixed effects. The rationale for considering population as a fixed effect is that all the chiltepin populations that we were able to find were included in the analyses, rather than using a random representation of them. The symptom and virus prevalence values determined for each population in the different years were considered as dependent measures; thus, they were treated as repeated measures in the GLMM. This seems the correct approach for wild and let-standing populations, in which at least a subset of the plants sampled over the years were the same, since chiltepin plants live for several years. This might not be so for cultivated populations, in which plants are managed as an annual crop and may change plots over the years. However, we considered that plots from different years were close enough to be spatially correlated, and therefore repeated measures are warranted. In addition, results did not differ when data from cultivated populations were analyzed as independent measures (not shown). To determine whether values of analyzed traits were significantly different among classes within each factor, Bonferroni analyses were employed in all cases using the GLMM marginal means calculated for each class [37]. GLMM accommodates missing data, so that the 26 chiltepin populations sampled could be included in the analysis. Parallel analyses using only the 8 populations for which data on the 3 years of sampling were available yielded comparable results (data not shown).

The contributions of each ecological factor to the variation in virus and symptom prevalence were estimated using a Principal Component Analysis (PCA). Host plant density (d), host plant genetic diversity estimated as expected heterozygosity (He), species diversity estimated as species richness (number of species, SR) and Shannon index (Sh), of 24 chiltepin populations were scaled to zero mean and unit variance, inserted in a regression matrix and rotated to obtain the principal components (PCs). Importantly, species diversity was not measured in TLA-w and HER-cmc, so that these populations were excluded from the analysis. Significance thresholds for the load of each ecological factor on a PC were determined using a broken-stick model [38]. Bivariate analyses, considering both linear and non-linear models, of begomoviruses, CMV and symptom prevalence onto the ecological factors and their corresponding PCs, yielded the proportion of the variance in each of these variables explained by each factor and each principal component (R2), and the significance of these correlations. For these bivariate analyses, we utilized the GLMM marginal means of begomoviruses, CMV and symptom prevalence calculated for each population. Statistical analyses were performed using the statistical software package SPSS 17.0 (SPSS Inc., Chicago, USA).

Information theory was used to determine the relative importance of the ecological factors in the variation of symptom and virus prevalence [39]. This approach was chosen because it allows making inferences across a set of causal model structures for symptom and virus prevalence [39]. To do so, a set of models that included a global model, which contained all ecological factors (species richness, Shannon index, expected heterozygosity and host plant density), and nested models, which contained different combinations of the predictor variables was fitted. Since species richness and Shannon index always loaded in the same PC, different selection model analyses in which the nested models considered SR, or Sh, or both variables together were performed. The three approaches gave similar results. For simplicity, only results considering species richness are shown. We ranked the models according to second order Akaike's Information Criteria (AICc) to account for small sample size (R library: AICcmodavg) [39]. The model with the lowest AICc score was selected as the best-ranked model. We calculated AICc Delta (Δi), as the difference between the AICc of a given model and that of the best-ranked model. Delta quantifies how strongly models compete (Δi = 0 for best-ranked model; Δi = 1–2 indicates substantial empirical support; Δi = 4–7 indicates considerable less support; and Δi>10 indicates no support [39]). Finally, the Akaike relative weight (ωi) of each model was calculated following the expression: ωi = exp(Δi)/Σexp(Δi).


Prevalence of Virus Infection

The status of a total of 1820 censused plants was recorded during the summers of 2007–2009. The prevalence of plants showing symptoms of virus infection (symptomatic plants) (Table 1) marginally varied among year (χ2 = 5.86, P = 0.060), ranging between 16.2% and 21.6% of the census.

A subset of 1081 plants, either symptomatic or asymptomatic, was analyzed for infection by ChYMV (a chiltepin-infecting tymovirus, see [36]), CMV, begomoviruses or potyviruses. Low prevalence of potyvirus infection (2.87%) precluded further analyses. ChYMV infection was limited to locations around Tula, AZP (Table 1) where its prevalence was high (42.86%). Infection by CMV and by begomoviruses was detected during the three years of the study in all biogeographical provinces and under different levels of human management (Figure 1 and Table 2). PepGMV and PHYVV were the only begomovirus species detected infecting chiltepin, and their relative prevalence did not depend on the level of human management of the chiltepin population (data not shown). Therefore, from here on these two species will be considered together and referred to as “begomoviruses”. CMV prevalence remained stable (≈7%) among years (χ2 = 0.06, P = 0.970), while begomovirus prevalence was about 3–5 times higher (19–36%) and varied largely according to year (χ2 = 58.25, P<1×10−5) (Table 2).

Table 2. Laboratory-determined prevalence of begomoviruses and CMV infection in Mexican chiltepin populations.

Begomoviruses, CMV or ChYMV infection explained the symptoms of 212/281 (78.7%, of these 81% being infected by begomoviruses) laboratory-analyzed symptomatic plants from all populations and years. This fraction did not differ according to the level of human management of the population (59/76 analyzed symptomatic plants for wild populations; 50/70 for let-standing populations, and 103/135 for cultivated populations) (χ2 = 0.86, P = 0.651). The fraction of infected plants showing symptoms (212/369, i.e., 57.4% in total) was lower in wild populations (59/133, 44.4%) than in cultivated (103/153, 67.3%) or in let-standing populations (50/83, 60.2%) (χ2≥4.54, P≤0.033). This fraction did not differ between the later two levels of human management (χ2 = 0.89, P = 0.345).

Human Management and Virus Prevalence

The effect of geography (biogeographical province and chiltepin population), and level of human management in the prevalence of symptomatic plants, begomoviruses or CMV, was analyzed. GLMM analyses using biogeographical province as a fixed effect showed that neither the prevalence of symptomatic plants, begomoviruses or CMV did depend on this factor (F5,47>0.797, P<0.557). Similarly, the prevalence of CMV infection did not vary among chiltepin populations (F25,47 = 1.512, P = 0.108). However, population was a factor determining the prevalence of symptomatic plants and begomovirus infection (F25,47>4.369, P<1×10−4). Bonferroni-corrected multiple comparisons showed that this was solely due to the higher prevalence in populations HUJ-chg and LIB-cmc (P<0.046 in 21/25 populations in both cases), and when these populations were removed from the analysis, population was no longer a factor in the prevalence of symptomatic plants and begomovirus infection (F23,45 = 1.984, P = 0.183). Populations HUJ-chg and LIB-cmc were not excluded from further analyses in order to consider as much of the variability in the analyzed factors as possible. These results show that the biogeographical factors analyzed largely do not affect viral and symptom prevalence. Consequently, the populations corresponding to each level of human management could be analyzed together.

The level of human management was a factor in determining the prevalence of symptomatic plants, (F2,47 = 7.619, P<1×10−4), which was significantly lower in wild than in cultivated populations (P<1×10−3), with let-standing populations showing an intermediate value (P≥0.276). Similarly, human management also affected the prevalence of begomovirus infection (F2,47 = 5.774, P = 6×10−4). Values were higher in cultivated populations than in wild (P = 6×10−4) and intermediate in let-standing populations (P≥0.076). However, CMV prevalence did not vary depending on this factor (F2,47 = 1.459, P = 0.243). Thus, increased levels of human management are associated with higher prevalence of symptomatic plants and begomoviruses, but not to prevalence of CMV. We therefore explored which ecological factors varying between populations with different levels of human management were linked to these differences in disease and virus infection risk.

Ecological Factors and Human Management

The relative importance of focal host plant density (d), host genetic diversity (expected heterozygosity, He) and species diversity of the habitat expressed either as species richness (SR) or considering also species evenness (Shannon index, Sh), on chiltepin populations was analyzed by including these variables in a PCA. Relevant statistical parameters of these ecological factors are provided in Table S1 of the Supporting Information. Parallel PCAs were performed considering all populations together and individually for each level of human management (Table 3). Importantly, SR and Sh loaded in the same PC in all cases. Since both variables represent the same ecological factor, we performed separate PCAs considering either SR or Sh, but the choice of index did not alter the results (data not shown).

Table 3. Principal component analysis of four ecological factors, and their association with symptom, begomovirus and CMV prevalence in chiltepin populations with different levels of human management in Mexico.

The PCA using the data set that included all the populations (All) yielded three main PCs collectively explaining 95.7 percent of the total variance. Species diversity (SR and Sh) was highly associated with PC1, d with PC2, and He with PC3 (Squared loadings>81.9) (Table 3, All column). The PCA restricted to wild populations, largely mirrored the results obtained with the All data set. However, the fraction of total variance explained by PC1 was higher, and that explained by PC2 and PC3 lower than in the All analysis (Table 3, Wild column).

PCAs considering either let-standing or cultivated populations separately yielded PCs explaining similar percentages of the variance than in the All data set. However, variables loading in each PC differed from All and wild data sets. For let-standing populations, He was now associated with PC1, SR and Sh with PC2, and d with PC3 (Squared loadings>82.7) (Table 3, Let-standing column). Similarly, in cultivated populations He was associated with PC1, SR/Sh with PC2, and d with PC3 (Squared loadings>85.0) (Table 3, Cultivated column). Importantly, SR, Sh and d loaded positively into their respective PCs, but the loading of He was always negative (not shown). The results above indicate that the relative importance of the ecological factors considered in this study vary depending on the level of human management of the chiltepin population.

To determine how human management affects species and genetic diversity, and plant density, we performed GLMM analyses on each PC obtained with the All data set using level of human management as a factor. The three major PCs significantly differed depending on the level of human management (F2,24≥4.995, P≤0.015). Values of PC1 (species diversity) were significantly higher in wild than in cultivated populations (P = 0.045), with intermediate values in let-standing populations. The opposite trend was observed for PC2 (plant density). For PC3 (host genetic diversity), values for cultivated populations were lower than in let-standing and wild populations (P≤0.015), the later two types not differing (P = 0.952).

Association between Ecological Factors and Virus Prevalence

To further explore the association between the considered ecological factors and disease risk, the influence of each ecological factor on symptom, begomovirus and CMV prevalence was studied using model selection analyses. For the All data set, symptom, begomovirus and CMV prevalence were chiefly determined by species diversity, either measured as SR (Table 4) or Sh (not shown). The model including only species richness was unambiguously the best (ω>0.79) (Table 4). In wild populations, symptom prevalence was mainly associated with species richness and host density (ω = 0.45 and 0.33, respectively), and begomovirus prevalence was chiefly associated with host density (ω = 0.40). However the best-ranked model included also host genetic diversity (ω = 0.55). Although the best-ranked model explaining CMV prevalence included all the ecological factors, the single factor that best explained the variable was host genetic diversity (ω = 0.12) (Table 4). Model selection analyses were largely similar for let-standing and cultivated populations. In both types of populations, host genetic diversity best explained symptom and begomovirus prevalence (ω = 0.51 in both cases), with the model including all the ecological factors showing slightly lower weight. Finally, host density chiefly determined CMV prevalence (ω = 0.57 and 0.35), but the model considering all factors showed slightly higher weight in cultivated populations (Table 4). Thus, these analyses are in agreement with the PCA.

Table 4. Model selection results. For virus and symptoms prevalence, model structures included species diversity (Species richness, SR), genetic diversity (expected heterozygosity, He), and host plant density (d). Best-ranked models are bolded and have the lowest AICc.

Effect of Ecological Factors in Virus Prevalence

The effect of ecological factors in symptom and virus prevalence was analyzed by bivariate analyses of each factor onto the prevalence of symptomatic plants, begomoviruses and CMV. For the All data set, SR was negatively associated with symptom and begomovirus prevalence (P<0.050), explaining 31.1% and 20.2% of the variance in these variables, respectively (Figure 2 and Table S2). Therefore, species diversity was the primary predictor of symptom and begomovirus prevalence. In wild populations, SR explained 20.2% of the variance being negatively associated with symptom prevalence (P = 0.026) (Figure 2 and Table S2), and d explained 27.7% (P = 0.039) of the variance in symptom prevalence, and 44.4% (P = 0.038) of the variance in begomovirus prevalence. Thus, in wild populations species diversity had also a principal role in determining symptom prevalence, with a lesser effect of plant density in symptom and begomovirus prevalence. In contrast, in let-standing populations He was negatively associated with symptom and begomovirus prevalence (P≤0.025), explaining 85.7% and 65.1% of the variance in these two traits, respectively. In addition, a negative correlation between SR and CMV prevalence was found (P = 0.041, 45.8% of the variance explained), and d showed a positive association with CMV prevalence (P = 0.048, 31.1% of the variance explained) (Figure 2 and Table S2). Finally, He in cultivated populations was also negatively associated with symptom prevalence (P≤0.015, 55.2% of the variance explained), and d explained 28.5% of the variance in CMV prevalence (P = 0.022) (Figure 2 and Table S2). Parallel bivariate analyses using the PCs associated with species richness, Shannon index, expected heterozygosity and host density of each PCA, instead of the original variables yielded similar results (Table 3 and Figure S1).

Figure 2. Bivariate relationships between ecological factors and disease/infection risk.

Significant regressions of each ecological factor and the prevalence of symptomatic plants (A), and of begomovirus (B) and CMV infection (C), are represented according to the level of human management: Wild (green triangles), let-standing (red squares), and cultivated (blue dots). PCs with the highest association with each ecological factor are shown in parenthesis. SR = Species richness expressed as number of species, He = Host genetic diversity expressed as expected heterozygosity, d = Host plant density. Note the different scales in the X-axis depending on the ecological factor. The Y-axis represents marginal mean prevalence values for each population over the monitored period.


We have analyzed the prevalence of virus disease and virus infection in populations of a wild plant to test whether increased host density and decreased host genetic diversity in agroecosystems, as compared with wild ecosystems, favors disease risk. These two classical hypotheses of plant pathology [6], [7] are particular cases of a more general one, which is receiving much attention recently, stating that habitat biodiversity is a determinant of disease risk [4] and may be at the root of disease emergence [3], [4]. The wild pepper or chiltepin was the focal host for this study, taking advantage of some unique characteristics of this species. First, wild populations of chiltepin are found in a large variety of habitats in different biogeographical provinces of Mexico [24], [25], which anticipated large differences in species diversity among habitats, as was indeed the case (Table S1). Second, the genetic diversity of wild chiltepin populations differs according to their geographical origin as shown for the 10 wild populations analyzed here [25]. Last, chiltepin populations show different levels of human management, including populations of let-standing plants, which are not sown or planted, but are tolerated or protected in anthropic habitats; and cultivated populations, in which plants are sown in home gardens or in small traditional plots.

The risk of virus disease was estimated as the prevalence of symptomatic plants. Although unapparent virus infection may affect plant fitness [40], we call here diseased plants those showing macroscopic symptoms. This is grounded in our (unpublished) observations of a fecundity reduction in symptomatic plants as compared to both infected or non-infected asymptomatic ones. However, we are aware that prevalence of symptomatic plants may underestimate disease risk, if symptom development were correlated with increased host mortality. Considering these caveats, the risk of disease was positively correlated with the level of human management of the population, being higher in cultivated than in wild populations, and intermediate in let-standing populations. Hence, results support the concept that transition of host habitat from wild ecosystems to agricultural ones results in an increase of disease risk.

A GLMM analysis of the variation of three PCs – associated with species diversity, host genetic diversity and host density – according to the level of human management, strongly suggested that the higher disease risk associated with increased human management is determined by a reduction of biodiversity, both as species diversity and host genetic diversity, and/or by an increased host density. It should be noted that habitat biodiversity and host genetic diversity, estimated as SR and He, respectively, vary along a continuum over the three levels of human management of chiltepin populations, which should avoid spurious associations. However, we cannot discard that other factors structured according to the level of human management, not specifically addressed in this work, may influence disease risk. Examples could be time of exposure to virus infection, and nutrient availability. Nevertheless, the three PCs associated to the analyzed ecological factors explained more than 95% of the variance of the analyzed variables, regardless that all populations were considered together or differentiating between wild, let-standing and cultivated populations. Therefore, although other ecological factors might have minor effects on disease risk, those here considered accounted for most of the variance within and between levels of human management, and may largely explain the emergence of viral disease associated with human management of chiltepin populations.

More specifically, both PC and model selection analyses showed that, for the All and wild data sets, species diversity of the habitat was the major predictor of disease risk (Figure 2, Tables 3 and 4). For let-standing and cultivated populations host genetic diversity was the major predictor of disease risk (Figure 2, Tables 3 and 4). The risk of infection by begomoviruses, which mostly explained symptoms, followed a largely similar pattern, except for a noticeable role of host density in determining virus prevalence in wild populations. These results support the dilution effect hypothesis for a plant-virus system. Moreover, they stress the importance of preserving biodiversity to maintain ecosystem services, a key concept in conservation biology [4]. Results also agree with most analyses of a variety of animal [5], [41] and plant systems [9][14], contributing to extend this hypothesis to plant virus diseases. The relationship between biodiversity and disease risk has received comparatively little attention in wild plant-infecting viruses. To our knowledge, Cereal- and Barley yellow dwarf luteoviruses (C/BYDV), which infect many species of grasses and are transmitted in a persistent manner by aphids in a highly species-specific way, is the best characterized system. In this case, most results are compatible with the amplification effect hypothesis, although the complex relationships between grass species, vector multiplication, and virus multiplication/transmission, make the effect of biodiversity on disease risk largely dependent on species composition [18][21], [42], [43]. Differences in life histories between luteoviruses and begomoviruses, which cause most symptoms of virus disease in chiltepin (Tables 1 and 2), could explain why effects of biodiversity vary between both systems. Begomoviruses have a narrow host range [29][31], and are persistently transmitted by B. tabaci, which has a wide host range [32]. Consequently, the larger the species diversity of the habitat, the larger the number of plant species in which B. tabaci can feed, and the lower fraction of meals resulting in begomovirus transmission to the focal host, resulting in host encounter reduction (sensu [5]). Interestingly, other reports of a dilution effect of biodiversity also refer to persistently transmitted viruses in which the host range of the virus is narrower than that of the vector [44], [45].

Importantly, which of the two different components of biodiversity was the primary predictor of disease and begomovirus infection risk depended on the level of human management. The reduced weight of species diversity in anthropic habitats could be explained by species diversity being largely reduced in cultivated vs. wild populations, and not varying largely among let-standing populations (Table S1). Host genetic diversity has been shown to have a negative effect on the risk of fungal diseases in crops [46][48]. Results from fungal pathogens were interpreted as due to differences in resistance-susceptibility among host genotypes, resulting in decreased transmission efficiency [46][48]. This mechanism could be also invoked to explain our results as differences in resistance to begomovirus infection have been reported among chiltepin genotypes [49]. However, genotype diversity might also reduce pathogen transmission by other mechanisms, for instance, microenvironment changes [13] or modification of the behavior of insect vectors [50].

The reduction of disease and begomovirus-infection risk with higher biodiversity was not coupled to a lower chiltepin density, as host plant density always loaded into a different PC than species or host diversity. An accepted axiom in plant pathology is that higher host density leads to higher disease risk. However, data are scarce and mostly inconclusive [6], [8], [9], and the effects of biodiversity and host density on disease risk are often difficult to differentiate [4]. The few works that attempted to differentiate these effects yielded contrasting results: density was the primary factor determining disease risk [10], [11] or there were independent and complex interactions between the effects of both factors [22]. The methodology used here avoids artificial correlations [51] and allowed disentangling the effects of these two ecological factors on disease and begomovirus infection risk.

Interestingly, infection by CMV followed a different pattern: infection risk did not depend on the level of human management, and host plant density was a relevant parameter in managed populations, but not in wild ones. The different pattern of infection risk found for begomoviruses and CMV could be due to differences in their life histories. At odds with begomoviruses, CMV is a generalist regarding the host and the vector and, perhaps more importantly, it is transmitted in a non-persistent manner. While persistent transmission is effected during feeding periods among plants that are hosts of the aphid vector, non-persistent transmission occurs during probing visits to plants that need not be hosts of the aphids, which remain viruliferous for short periods of time [52]. Thus, proximity of plants susceptible to the virus could be more important than biodiversity in determining CMV infection risk, similarly to directly transmitted fungi infecting leaves [10], [11]. Consequently, the mechanisms of transmission, in addition to the host range of the pathogen and/or its vectors, could be a primary factor in determining the relationship between biodiversity and disease risk, an unexplored issue, to our knowledge.

Finally, a larger fraction of begomovirus- or CMV-infected plants showed symptoms in managed populations than in wild ones, strongly suggesting a higher virulence of virus infection in the former, perhaps due to a higher susceptibility of plants in human-managed populations to virus infection and its effects. The relationship between host physiological condition and disease susceptibility is an underexplored subject [53]. However, we could speculate that plants of managed populations, which benefit from higher levels of water and/or nutrients than those from wild habitats, as shown by their production of about five times as much fruits (our unpublished observations), would be more competent hosts for virus vectors [20], [54]. This would encourage more frequent and longer meals, thus being under higher inoculum pressure of persistently transmitted viruses. Also, a more favorable host condition could result in higher levels of virus multiplication [21], [54], [55]. If this were the case, in addition to suffering more virulent virus infections, plants in cultivated and let-standing populations would be more competent hosts for virus vectors, virus multiplication and transmission. These factors would contribute to the higher disease risk, and thus to disease emergence, in human managed populations, regardless of the ecological factors here analyzed. However, we cannot exclude that the larger proportion of symptomatic plants in managed habitats would be the result of increased life span of infected plants due to the enhanced availability of resources in cultivated and let-standing populations, which could contribute to explain our observations.

In summary, our results show the important role of biodiversity reduction in the emergence of viral diseases associated to human management of plant populations. Our work also suggest that other ecological and genetic factors, perhaps resulting in increased virulence in anthropic habitats, need to be considered in order to fully understand the dynamics of emergence, which should be the subject of future research.

Supporting Information

Figure S1.

Bivariate relationships between principal components (PCs) associated to ecological factors and disease/infection risk. Significant regressions of PC and the prevalence of symptomatic plants (A), and of begomovirus (B) and CMV infection (C), are represented according to the level of human management. The X-axis represents the PC as a continuous variable comprised of the principal component scores for each population. Ecological factors with the highest loading on each PC are shown in parenthesis. SR = Species richness expressed as number of species, He = Host genetic diversity expressed as expected heterozygosity, d = Host plant density. The Y-axis represents marginal mean prevalence values for each population over the monitored period.


Table S1.

Relevant statistical parameters of each ecological factor analyzed in Mexican chiltepin populations.


Table S2.

Analysis of association between symptoms, begomovirus and CMV prevalence, and ecological factors.



Antolín López-Quirós and Miguel Ángel Mora provided excellent technical assistance. We thank two anonymous reviewers for useful comments.

Author Contributions

Conceived and designed the experiments: FGA. Performed the experiments: PGJ MRU. Analyzed the data: IP AF FGA. Wrote the paper: IP AML AF DP FGA. Collected field data: PGJ AML DP FGA.


  1. 1. Morse SS, Schluederberg A (1990) Emerging viruses: the evolution of viruses and viral diseases. J Infect Dis 162: 1–7.
  2. 2. Weiss RA, McMichael AJ (2004) Social and environmental risk factors in the emergence of infectious diseases. Nat Med 10: S70–S76.
  3. 3. Jones RAC (2009) Plant virus emergence and evolution: Origins, new encounter scenarios, factors driving emergence, effects of changing world conditions, and prospects for control. Virus Res 141: 113–130.
  4. 4. Keesing F, Belden LK, Daszk P, Dobson A, Harwell CD, et al. (2010) Impacts of biodiversity on the emergence and transmission of infectious diseases. Nature 468: 647–652.
  5. 5. Keesing F, Holt RD, Ostfeld RS (2006) Effects of species diversity on disease risk. Ecol Lett 9: 485–498.
  6. 6. Burdon JJ, Chilvers GA (1982) Host density as a factor in plant-disease ecology. Annu Rev Phytopathol 20: 143–166.
  7. 7. Day PR (1978) The genetic basis of epidemics. In: Horsfall JG, Cowling EB, editors. Plant Disease, an Advanced Treatise, Vol. II. New York: Academic Press. pp. 263–285.
  8. 8. Stukenbrock EH, McDonald BA (2008) The origin of plant pathogens in agro-ecosystems. Annu Rev Phytopathol 46: 75–100.
  9. 9. Pfleeger TG, Mundt CC (1998) Wheat leaf rust severity as affected by plant density and species proportions in simple communities of wheat and wild oats. Phytopathology 88: 708–714.
  10. 10. Knops JMH, Tilman D, Haddad NM, Naeem S, Mitchell CE, et al. (1999) Effects of plant species richness on invasion dynamics, disease outbreaks, insect abundance and diversity. Ecol Lett 2: 286–293.
  11. 11. Mitchell CE, Tilman D, Groth JV (2002) Effects of grassland species diversity, abundance and composition on foliar fungal disease. Ecology 83: 1713–1726.
  12. 12. Mitchell CE, Reich PB, Tilman D, Groth JV (2003) Effects of elevated CO2, nitrogen deposition, and decreased species diversity on foliar fungal plant disease. Global Change Biol 9: 438–451.
  13. 13. Zhu YY, Fang H, Wang YY, Fan JX, Yang SS, et al. (2005) Panicle and canopy moisture in rice cultivar mixtures. Phytopathology 95: 433–438.
  14. 14. Roscher C, Schumacher J, Foitzik O, Schulze E-D (2007) Resistance to rust fungi in Lolium perenne depends on within-species variation and performance of the host species in grasslands of different plant diversity. Oecologia 153: 171–183.
  15. 15. Woolhouse M, Gaunt E (2007) Ecological origins of novel human pathogens. Crit Rev Microbiol 33: 231–244.
  16. 16. Farr DF, Bills GF, Chamuris GP, Rossman AY (1989) Fungi on plants and plant products in the United States. St Paul: APS Press. 1252 p.
  17. 17. Power AG, Flecker AS (2003) Virus specificity in disease systems: are species redundant? In: Kareiva P, Levin SA, editors. The importance of species. Perspectives on expendability and triage. Princeton: Princeton University Press. pp. 330–351.
  18. 18. Power AG, Mitchell CE (2004) Pathogen spillover in disease epidemics. Am Nat 164: S79–S89.
  19. 19. Malmstrom CM, Hughes CC, Newton LA, Stoner CJ (2005a) Virus infection in remnant native bunchgrasses from invaded California grasslands. New Phytol 168: 217–230.
  20. 20. Borer ET, Adams VT, Engler GA, Adams AL, Schumann CB, et al. (2009) Aphid fecundity and grassland invasion: invader life history is the key. Ecol Applic 19: 1187–1196.
  21. 21. Borer ET, Seabloom EW, Mitchell CE, Power AG (2010) Local context drives infection of grasses by vector-borne generalist viruses. Ecol Lett 13: 810–818.
  22. 22. Power AG (1987) Plant community diversity, herbivore movement and an insect-transmitted disease of maize. Ecology 68: 1658–1669.
  23. 23. D'Arcy WG, Eshbaugh WH (1974) New World peppers (Capsicum, Solanaceae) north of Colombia: a resume. Baileya 19: 93–105.
  24. 24. Tewksbury JJ, Nabhan GP, Norman D, Suzan H, Tuxill J, et al. (1999) In situ conservation of wild chiles and their biotic associates. Conserv Biol 13: 98–107.
  25. 25. González-Jara P, Moreno-Letelier A, Fraile A, Piñero D, García-Arenal F (2011) Impact of human management on the genetic variation of wild pepper, Capsicum annuum var. glabriusculum. PLoS ONE 6: e28715.
  26. 26. Nabhan GP (1990) Conservationists and Forest Service join forces to save wild chiles. Diversity 6: 47–48.
  27. 27. Votava EJ, Nabham GP, Bosland PW (2002) Genetic diversity and similarity revealed via molecular analysis among and within an in situ population and ex situ accessions of chiltepin (Capsicum annuum var. glabriusculum). Conserv Genet 3: 123–129.
  28. 28. Casas A, Otero-Arnaiz A, Pérez-Negrón E, Valiente-Banuet A (2007) In situ management and domestication of plants in Mesoamerica. Ann Bot 100: 1101–1115.
  29. 29. Brown JK, Poulos BT (1990) Serrano golden mosaic virus: a newly identified whitefly-transmitted geminivirus of pepper and tomato in the United States and Mexico. Plant Dis 74: 720.
  30. 30. Garzón-Tiznado JA, Torres-Pacheco I, Ascencio-Ibañez JT, Herrera-Estrella L, Rivera-Bustamante RF (1993) Inoculation of peppers with infectious clones of a new geminivirus by biolistic procedure. Phytopathology 83: 514–521.
  31. 31. Torres-Pacheco I, Garzón-Tiznado A, Brown JK, Becerra-Flora A, Rivera-Bustamante RF (1996) Detection and distribution of geminiviruses in Mexico and the Southern United States. Phytopathology 86: 1186–1192.
  32. 32. Brown JK, Frolich DR, Rosell RC (1995) The sweet-potato or silverleaf whiteflies: biotypes of Bemisia tabaci or a species complex? Annu Rev Entomol 40: 511–534.
  33. 33. Palukaitis P, García-Arenal F (2003) Cucumber mosaic virus. Descriptions of Plant Viruses, No. 400. Harpenden, UK: Association of Applied Biologists and Rothamsted Research. Available:
  34. 34. CONABIO (1997) Provincias biogeográficas de México. Escala 1:4000000. Comisión Nacional para el Conocimiento y Uso de la Biodiversidad, México, D. F. Available:
  35. 35. Shannon C (1948) A mathematical theory of communication. Bell Syst Technol J 27: 379–423.
  36. 36. Pagán I, Betancourt M, de Miguel J, Piñero D, Fraile A, et al. (2010) Genomic and biological characterization of Chiltepín yellow mosaic virus, a new tymovirus infecting Capsicum annuum var. aviculare in Mexico. Arch Virol 155: 675–684.
  37. 37. Sokal RR, Rohlf FJ (1995) Biometry. 3rd ed. New York: W.H. Freeman and Company.
  38. 38. Peres-Neto PR, Jackson DA, Somers KM (2003) Giving meaningful interpretation to ordination axes: assessing loading significance in principal component analysis. Ecology 84: 2347–2363.
  39. 39. Burnham KP, Anderson DR (2002) Model Selection and Multi- model Inference: A Practical Information Theoretic Approach, 2nd ed. New York: Springer-Verlag.
  40. 40. Remold SK (2002) Unapparent virus infection and host fitness in three weedy grass species. J Ecol 90: 967–977.
  41. 41. Roche B, Guégan J-F (2011) Ecosystem dynamics, biological diversity and emerging infectious diseases. Comp Rend Biol 334: 385–392.
  42. 42. Malmstrom CM, McCullough AJ, Johnson HA, Newton LA, Borer ET (2005) Invasive annual grasses indirectly increase virus incidence in California native perennial bunchgrasses. Oecologia 145: 153–164.
  43. 43. Hall GS, Peters JS, Little DP, Power AG (2010) Plant community diversity influences vector behavior and Barley yellow dwarf virus population structure. Plant Pathol 59: 152–1158.
  44. 44. Pitre HN, Boyd FJ (1970) A study of the role of weeds in corn fields in the epidemiology of corn stunt disease. J Econ Entomol 63: 195–197.
  45. 45. Farrell JAK (1976) Effects of intersowing with beans on the spread of groundnut rosette virus by Aphis craccivora Koch (Hemiptera, Aphididiae). Malawi Bull Entomol Res 66: 331–333.
  46. 46. Wolfe MS (1985) The current status and prospects of multiline cultivars and variety mixtures for disease resistance. Annu Rev Phytopathol 23: 251–173.
  47. 47. Mundt CC (2002) Use of multiline cultivars and cultivar mixtures for disease management. Annu Rev Phytopathol 40: 381–410.
  48. 48. Burdon JJ, Thrall PH, Ericson L (2006) The current and future dynamics of disease in plant communities. Annu Rev Phytopathol 44: 19–39.
  49. 49. Hernández-Verdugo S, Guevara-González RG, Rivera-Bustamante S, Oyama K (2001) Screening wild plants of Capsicum annuum for resistance to Pepper huasteco virus: Presence of viral DNA and differentiation among population. Euphytica 122: 31–36.
  50. 50. Power AG (1991) Virus spread and vector dynamics in genetically diverse plant populations. Ecology 72: 232–241.
  51. 51. Drummond DA, Raval A, Wilke CO (2006) A single determinant dominates the rate of yeast protein evolution. Mol Biol Evol 23: 327–337.
  52. 52. Hull R (2002) Matthews' Plant Virology. 4th ed. San Diego, CA: Academic Press.
  53. 53. Beldomenico PM, Begon M (2010) Disease spread, susceptibility and infection intensity: vicious circles? Trends Ecol Evol 25: 21–27.
  54. 54. Cronin JP, Elsh ME, Dekkers MG, Abercrombie ST, Mitchell CE (2010) Host physiological phenotype explains pathogen reservoir potential. Ecol Lett 13: 1221–1232.
  55. 55. Schrotenboer AC, Allen MS, Malmstrom CM (2011) Modification of native grasses for biofuel production may increase virus susceptibility. GCB Bioenergy 3: 360–374.