Predicted distribution of Metaparasitylenchus hypothenemi (Tylenchida: Allantonematidae), parasite of the coffee berry borer

Sample collection: Adult H. hampei populations from coffee plantations across Soconusco, Chiapas, Mexico, and from one site in San Pablo, San Marcos Department, Guatemala, were examined for parasitism by the nematode M. hypothenemi. Sampling occurred in 2003 and 2019 at 59 sites distributed over the study area. In 2003, adult coffee berry borers were captured in 23 coffee plantations using traps baited with a methanol/ethanol mixture (1:1 by volume), placed in each plantation, and suspended from the middle part of a coffee shrub during the rainy season (April–July 2003). In 2019, remaining robusta coffee berries were collected directly from the plant during the dry season (January–May 2019) in 36 coffee plantations, and adult berry borers were extracted from them. Sites were chosen based on various criteria: high shade density, low crop management practices, soil with high organic matter content, and high relative humidity (Perez et al., 2015). In both samples, nematode infection was determined by dissection of the captured borers under a stereo microscope.

Influence of altitude, temperature, and precipitation on nematode presence: Sample site altitude was recorded using a GPS device (Garmin Oregon 350; Garmin, Inc., Taipei, Taiwan). Mean annual temperature and precipitation data were obtained from the climate database WorldClim (https://www.worldclim.org/, consulted 2020). The bioclimatic variables BIO₁ (mean annual temperature) and BIO₁₂ (annual precipitation) were updated for Mexico (30 arc sec resolution, approximately 90 m / pixel) (Cuervo-Robayo et al., 2013). Data were obtained for each sampling site using the point sampling tool in the software program QGIS 3.10.3 (https://qgis.org/es/site/) (QGIS, 2018).

A multivariate analysis of variance (MANOVA) was performed to analyze the effect of altitude, annual precipitation, and temperature, with and without presence/absence data from 2003 and 2019. The missForest package was used to estimate two missing data for temperature and annual precipitation (Stekhoven and Bühlmann, 2012). A canonical discriminant analysis was performed to identify the environmental variables that were most strongly associated with nematode presence. A Chi-square test was performed using the statistical software package R (version X) to determine whether the presence of M. hypothenemi varied across the four different annual rainfall distribution ranges of the Soconusco (R Core Team, 2019).

Species distribution modeling (SDM): A model for the geographical distribution of M. hypothenemi was calibrated for the Neotropical Biogeographical Region, which includes all the coffee-growing regions of the Americas. Unpublished nematode parasitism data from 2003 were incorporated to improve model accuracy. A species distribution model (SDM) was generated with the MaxEnt (Maximum Entropy-Species Distribution Modeling, Version 3.4.1) algorithm (Phillips et al., 2006), using data from 29 sites where the parasite was present, climate variables from WorldClim, and Shuttle Radar Topography Mission (SRTM) altitude data from the US Geological Survey (USGS, http://www.usgs.gov) with a resolution of 30 arc sec (~1 km₂). MaxEnt was used under linear constraints, with random selection of 30% of the presence sites, minimum training presence, and 500 iterations (Anderson et al., 2003). The first model (model A) was generated using 15 climate variables – BIO₁ (annual mean temperature), BIO₂ (mean diurnal range), BIO₃ (isothermality), BIO₄ (temperature seasonality), BIO₅ (max temperature of warmest month), BIO₆ (min temperature of coldest month), BIO₇ (temperature annual range), BIO₁₀ (mean temperature of warmest quarter), BIO₁₁ (mean temperature of coldest quarter), BIO₁₂ (annual precipitation), BIO₁₃ (precipitation of wettest month), BIO₁₄ (precipitation of driest month), BIO₁₅ (precipitation seasonality), BIO₁₆ (precipitation of wettest quarter), BIO₁₇ (precipitation of driest quarter) – and the altitude, without considering a possible correlation between the variables. The climate variables, BIO₈ (mean temperature of wettest quarter), BIO₉ (mean temperature of driest quarter), BIO₁₈ (precipitation of warmest quarter), and BIO₁₉ (precipitation of coldest quarter), were not included because of odd discontinuities between neighboring pixels (Escobar et al. 2014). The second model (model B) was constructed using eight climate variables (BIO₁, BIO₂, BIO₃, BIO₇, BIO₁₂, BIO₁₃, BIO₁₅, and BIO₁₆) and the altitude, which were selected after multicollinearity analysis because of their low correlation (r < 0.75). The third model (model C) included the six climate variables with the highest percentage of contribution according to the Jackknife test (> 0.95; BIO₂, BIO₃, BIO₁₂, BIO₁₃, BIO₁₅, BIO₁₆, and the altitude). To improve prediction accuracy, the summation of these three models was used to generate the final model (model D). Model precision was evaluated using partial ROC analysis, and the errors of omission and predicted proportional areas were considered satisfactory (error of omission (E) = 5%, 50% data for evaluation, bootstrap=1000) (Escobar et al., 2013). The models were evaluated with the PartialROC function (Barve, 2008), and the geographical accuracy of the model was assessed by projecting presence and absence data onto the final SDM (model D). A threshold was selected to include at least 90% of the presence data (27 records), which was projected onto the categories with the highest probability of occurrence (> 6) for the transformation of SDM to the binary model (presence = 1/absence = 0). The binary models were used to estimate the breadth of each SDM (1 pixel = 1 km²) (Moo-Llanes et al., 2013).

Influence of altitude, temperature, and precipitation on nematode presence: Parasitism was detected in 29 out of 59 investigated sites (Tables 1, 2). Multivariate analysis of variance (MANOVA) showed significant differences in the binomial response (presence-absence) to environmental variables (altitude, precipitation, and temperature) in data from 2003 (F = 6.1; g.l. = 3; p = 0.003), as well as 2019 (F = 10.5; g.l. = 3; P < 0.001). Canonical discriminant analysis revealed that in both years, the presence of M. hypothenemi was associated with precipitation, while its absence was associated with temperature and altitude (Figs. 1, 2).

Figure 1.

Figure 2.

Sites with nematode presence were located between 450 and 1020 masl, in areas where the annual mean temperature ranges between 22°C and 25.6°C. However, 86% of the sites with nematode presence were located in the area with the highest annual rainfall (4000–4700 mm), so the probability of locating sites with nematode presence is lower in areas where annual rainfall is less than 4000 mm (χ² = 73.4; g.l. = 4; P < 0.001) (Fig. 3). Because 10 out of 59 sampled sites were in the same locality, only 54 points are marked on the map. The predictive performance of the models was significant (partialROC > 1; Table 3) with a similar amplitude and geographical distribution in all four models.

Figure 3.

The final SDM model is shown in Fig. 4D. All models predicted areas with low probability of nematode presence in western Mexico, Central America (Costa Rica and Panama), and South America (Colombia, Ecuador, and Peru). The presence of parasitism appeared in areas of high probability of occurrence (86% in categories > 6), whereas its absence fell in areas of low occurrence probability (65% in categories < 4) (Fig. 5). The average SDM extent was 4692 km². Areas with high probability of occurrence were concentrated in southeastern Mexico and Guatemala (Fig. 5).

Figure 4.

Figure 5.

Our results demonstrate that natural populations of this nematode are adapted to the environmental conditions of Soconusco and have persisted for at least the last 20 years (Castillo et al., 2002; Poinar et al., 2004; Pérez et al., 2015). Precipitation, in particular, affects the presence of this parasite in the study region, whereas its absence is associated with altitude and temperature. Although no parasitism data are available for the period 2004–2018, our results indicate a direct relationship between the presence of this species and rainfall. Moreover, since the duration of the rainy season in the study region is six to ten months, our results also suggest temporal patterns of parasitism. The dynamics of these patterns, however, have remained a relatively ill-explored research topic, and further research in this field is still necessary. Sites with the highest probability of nematode occurrence fall in the areas with the highest annual precipitation (> 4000 mm), between 450 and 1020 masl. At these altitudes, the dry season lasts only two months (December–January), which reduces the risk of dehydration and allows the nematodes to maintain a permanent presence. The coffee berry borer depends on humidity for its survival during the critical phases of its life cycle, including emergence from the fruit, as well as colonization, penetration, and development inside the fruit (Baker et al., 1994). Borer populations accumulate in residual coffee berries, ensuring their survival and development during the dry season (Damon, 2000). These factors might also favor transmission and persistence of the parasite since nematodes are highly susceptible to dehydration and sunlight (Banu and Rajendran, 2003). However, it remains unknown whether H. hampei is the only natural host of M. hypothememi. The coffee berry borer is an exotic pest that invaded Mexican coffee plantations in 1978 (Baker, 1984), whereas parasite populations might be endemic (unpublished data, Simota et al.).

The model of the possible distribution of M. hypothenemi in the Neotropics predicted with a high degree of probability the presence of this nematode in southeastern Mexico and southwestern Guatemala. These regions include high-altitude and humid areas with particular abiotic conditions that allow for high levels of endemism (Morrone and Márquez, 2003). The present study is the first published investigation that applies this type of model to evaluate the geographical distribution of an insect-parasitizing nematode of agro-economic importance. However, its anecdotal presence in Honduras (Poinar et al., 2004) and present observation in Guatemala suggest that, as predicted by the model, nematode populations are distributed in plantations in Central America.

Species distribution is geographically constrained by environmental conditions and is ultimately determined by climatic tolerances and evolutionary processes (Wiens and Graham, 2005). Environmental and geographical space models were generated for the Neotropical region, which includes various large coffee-growing regions. The coffee berry borer is an important pest in these regions and is now known to complete its life cycle only in coffee berries (Johnson et al., 2020). The model predicts small areas in South America where M. hypothenemi can be encountered. However, because the abiotic conditions that are essential for the survival of this parasite are not adequately matched, the probability of occurrence in these regions is considered low. Finding other areas where the parasite is present would allow confirmation of the model’s predictions and, as such, continues to be a priority. The model is the first theoretical tool available, and its predictive validity is evidenced by the suitable projection of data from areas with high and especially low probabilities of occurrence, allowing us to predict the possible geographical distribution of M. hypothenemi with a high degree of reliability.

Although climate change is only one of the many factors that contribute to the decline of a species’ population, it has been identified as one of the main causes of species extinction (Román-Palacios and Wiens, 2020). The models presented here indicate that the presence and distribution of M. hypothenemi may be affected by climate change because the parasite is geographically restricted to areas of high precipitation rates in southeastern Mexico and Central America, and its potential distribution in other regions of the Neotropics is limited. Even though extinction has not been documented for any species of parasitic nematode, endemic species are usually more vulnerable to climate change (Carlson et al., 2017), which highlights the need to design conservation plans for this species.

The parasite M. hypothenemi plays a role in the natural regulation of coffee berry borer populations by significantly reducing the fecundity of infected hosts (Poinar et al., 2004; Castillo et al., 2019). Although there is no direct evidence that the parasite causes the mortality of its host, parasite-induced sterility in females of the coffee berry borer may lead to a population decrease in the following generations of this pest (Castillo et al., 2002; Castillo et al., 2019). Further research is needed to fully understand the bioecology of this parasite to chart its demography and explore host-parasite interactions under controlled conditions. Likewise, additional studies focusing on the parasite’s genetic diversity can shed light on its geographical origin and hence provide valuable input to establish priority conservation areas. Prediction models can be a useful tool to identify sites in other countries of the Americas where M. hypothenemi may be present and contribute to the understanding of the ecological relationships between the parasite and the environment.