To evaluate potential abiotic driven changes in parasite specialization, we projected ENMs to future climates. We considered Relative Concentration Pathways 4.5 and 8.5, and used climatologies for five CMIP5 models (MPI-ESM-MR, CESM1-BGC, ACCESS-1-0, MIROC5, CMCC-CM) chosen to reduce interdependency (Sanderson et al. 2015). ENMs were projected to 2050 using updated data layers for annual rainfall, mean temperature of the wettest quarter, and isothermality. Soil and ENVIREM variables, for which future projections were not available, remained constant. We calculated the average habitat suitability projected under the five different models for 2050. Considering all pixels within 200 km of any occurrence record of S. hermonthica, we tested whether the habitat suitability score was significantly different in future compared to 1979-2013 using a Wilcoxon signed rank test.
Results:
Complete host specialization is rare.
We first characterized levels of host specialization observed in S. hermonthica populations across Africa. Degree of host specialization was highly variable across the 27 populations studied (Fig. 2). Populations were characterized by high emergence on either millet or sorghum only, high emergence on two hosts, or high emergence on all three hosts (Fig. 2). Hierarchical clustering with a cut-point at 0.5 showed five groups: millet specialists (Group 1; n=2), millet/sorghum generalists (Group 2; n=3), sorghum specialists (Group 3; n=7), virulent generalists (Group 4; n=3), and maize/sorghum specialists (Group 5; n=10; Fig. 2). Besides the ‘virulent generalists’, all groups included populations from at least two studies, suggesting that differences among studies did not strongly bias clustering. With the exception of millet specialists from Bambey, Senegal, most populations had modest or low levels of specialization as indicated by Pairwise Differentiation Index values (Fig. 2B). These results suggest that although parasites may become locally adapted to commonly cultivated host species, complete specialization is rare.
Distribution of host communities shapes specialization.
We then investigated whether the distribution of host communities across environments might shape the evolution of host specialization (Q1). S. hermonthica-prone regions covered approximately 628 million hectares, of which 80% was estimated to have nonzero production of maize, millet, or sorghum. Crop production patterns generally followed rainfall gradients, with greater land area planted to millet in more arid regions of the Sahel, sorghum dominant at lower latitudes and in eastern Sudan, and maize most common in more mesic regions (Fig. 2). In 2000, sorghum or maize were the dominant crop in most S. hermonthica-prone areas (38% sorghum; 38% maize), with 24% of areas characterized by millet as the major cereal crop.
Specialization observed in experiments was strongly associated with spatial variation in host crop communities (Fig. 3A-C). For parasites on all three hosts, relative host crop area harvested within a 50 km radius was a strong predictor of mean relative emergence in experimental studies. This finding was especially pronounced for millet (pmillet<0.001;bharvest_area=1.02; LMM; Table S2-S3). Relative crop area harvested was also a statistically significant predictor of mean relative emergence on sorghum (psorghum=0.04; bharvest_area=0.22; LMM) and maize (pmaize=0.04;bharvest_area=0.23; LMM; Table S2-S3). These results are consistent with the conclusion that parasites adapt to the most abundant host in a particular region and also suggest that relative parasite emergence is a reasonable proxy for fitness (Fig. 1).
Host specialization may structure parasite distributions.
Our analyses provided little support for a positive relationship between specialization and parasite occurrence on different hosts (Q2; Fig. 3D-F). Within a 50-km radius of all 27 locations with parasite emergence data, 15 locations had at least one parasite occurrence record with information about host of origin. Proportion of records on a particular host for these 15 Striga populations was not a statistically significant predictor of emergence on maize (pmaize = 0.07;boccurrence= 0.18; LMM), millet (pmillet =0.06; boccurrence=0.21; LMM), or sorghum (psorghum=0.57; boccurrence= 0.05; LMM; Table S2). Availability of additional parasite occurrence and performance data could reveal a stronger relationship between specialization and patterns of parasite occurrence on different hosts.
Parasite abiotic niche reflects host environmental tolerance.
To characterize the abiotic basis of patterns of host specialization, we built environmental niche models (ENMs) for potentially host-specialized parasite populations by excluding all occurrences with unknown host or with a non-focal host (Fig. S2). Despite a smaller number of observations compared to the all-occurrence model, ENMs had good predictive accuracy for each of the host-specific models. Area under the receiver operating characteristic curve (AUC) for the test set was 0.848 for the all-occurrence model (n = 1049) compared to 0.850 for the sorghum-only model (n = 262), 0.908 for the millet-only model (n = 157), and 0.841 for the maize-only model (n = 74).
Modeled niches generally reflected known environmental tolerances of hosts, with millet-specific models predicting high habitat suitability in low nitrogen (N), low rainfall environments but maize-specific models predicting higher habitat suitability in environments characterized by more nitrogen-rich soils and higher rainfall (Fig. 4A-D). Across all host species, annual rainfall and soil N were generally among the most important predictors of parasite occurrence (Table S4). Annual rainfall was the most common factor limiting habitat suitability for maize-parasitizing S. hermonthica (59% of all grid cells; 10th-90th percentile for grid cells with habitat suitability ≥ 0.5: 854-1537 mm/yr) but was also strongly limiting for parasites of sorghum (41% of cells; 10th-90th percentile: 509-1197 mm/yr) and millet (32% of cells; 10th-90th percentile: 441-1164 mm/yr) (Fig. S4). Mean temperature of the wettest quarter was also an important predictor of S. hermonthica parasitizing millet (Table S4) and frequently limited habitat suitability for millet-parasitizing populations (Fig. S4). Soil clay content was an important predictor for S. hermonthica occurrence on sorghum (Table S4), but was only limiting for millet-parasitizing populations in eastern Sudan and sorghum-parasitizing populations in western Senegal (Fig. S4). Higher soil clay content has been anecdotally associated with parasitism on sorghum in general (Mohamed et al. 2001) and in Sudan (Wilson-Jones 1955).
In agricultural as well as natural ecosystems, more productive environments may be associated with increased availability of alternate hosts, favoring generalists (Thrall et al. 2007). Parasitism on maize was associated with increased environmental quality, with habitat suitability peaking in locations of higher crop productivity (Fig. 4E). Combined crop yield per area harvested was significantly greater where S. hermonthica parasitizes maize compared to millet (p < 0.001, Wilcoxon rank sum test) but not compared to sorghum (p = 0.2), consistent with greater niche overlap between maize and sorghum (Table S5). We observed a weak trend towards reduced parasite specialization with increasing environmental productivity (Fig. 4F; p = 0.15, Chi-square goodness-of-fit test for linear regression model). Taken together, parasite ENMs are highly sensitive to differences in multivariate environmental tolerance of hosts, consistent with host cultivation on a gradient from marginal, warm environments (pearl millet) to more productive, cooler environments (maize).
Abiotic gradients shape distribution of millet and sorghum specialists.
ENM contrasts show where host niches differ most in multivariate environmental space. If abiotic environment is associated with host specialization, specialists may be most likely where environmental axes most strongly differentiate hosts in niche space (Futuyma and Moreno 1988). Consistent with this idea, ENM contrasts were strongly predictive of variation in S. hermonthica performance on host species for which specialist parasite populations have been previously reported (Fig. 3G-H). ENM contrasts were most strongly associated with mean relative emergence on pearl millet (pmillet < 0.001; bENM = -0.68; LMM) and were also significantly associated with mean relative emergence on sorghum (psorghum = 0.0067; bENM = -0.50; LMM) (Table S2, S6). In contrast, host specialization predicted by ENM contrasts was not strongly associated with mean relative emergence on maize (pmaize = 0.19, bENM = 0.15; LMM; Table S2, S6).
Future change in parasite distributions
To investigate potential changes in host-specialization over time, we projected ENMs to future climates (Fig. S5). By 2050, we predicted an overall increase in habitat suitability for S. hermonthica, with median increase in habitat suitability in its current range of 0.07 under RCP 4.5 (p < 0.001; Wilcoxon signed rank test comparing habitat suitability of grid cells in present vs. 2050) and 0.09 under RCP 8.5 (p < 0.001; Wilcoxon signed rank test; Table S7). Habitat suitability increased most for millet-parasitizing populations, followed by maize-parasitizing S. hermonthica (Table S7). Changes in habitat suitability were heterogenous across space, with many regions in the Sudano-Sahelian zone of west Africa becoming less suitable for S. hermonthica and its hosts (Sultan et al. 2013) but central and east Africa generally becoming more suitable (Fig. S5).