Megafauna decline have reduced pathogen dispersal which may have increased emergent infectious diseases

The Late Quaternary extinctions of megafauna (defined as animal species >44.5 kg) reduced the dispersal of seeds and nutrients, and likely also microbes and parasites. Here we use body-mass based scaling and range maps for extinct and extant mammal species to show that these extinctions led to an almost seven-fold reduction in the movement of gut-transported microbes, such as Escherichia coli (3.3 km2/day to 0.5 km2/day). Similarly, the extinctions led to a seven-fold reduction in the mean home ranges of vector-borne pathogens (7.8 km2 to 1.1 km2). To understand the impact of this, we created an individual-based model where an order of magnitude decrease in home range increased maximum aggregated microbial mutations 4-fold after 20,000 years. We hypothesize that pathogen speciation and hence endemism increased with isolation, as global dispersal distances decreased through a mechanism similar to the theory of island biogeography. To investigate if such an effect could be found, we analysed where 145 zoonotic diseases have emerged in human populations and found quantitative estimates of reduced dispersal of ectoparasites and fecal pathogens significantly improved our ability to predict the locations of outbreaks (increasing variance explained by 8%). There are limitations to this analysis which we discuss in detail, but if further studies support these results, they broadly suggest that reduced pathogen dispersal following megafauna extinctions may have increased the emergence of zoonotic pathogens moving into human populations.

2 Abstract -The Late Quaternary extinctions of megafauna (defined as animal species >44.5 kg) 22 reduced the dispersal of seeds and nutrients, and likely also microbes and parasites. Here we use 23 body-mass based scaling and range maps for extinct and extant mammal species to show that these 24 extinctions led to an almost seven-fold reduction in the movement of gut-transported microbes, 25 such as Escherichia coli (3.3 km 2 /day to 0.5 km 2 /day). Similarly, the extinctions led to a seven-26 fold reduction in the mean home ranges of vector-borne pathogens (7.8 km 2 to 1.1 km 2 ). To humanity's close association with agriculture and domestic animals (Dobson & Carper, 1996) 59 (Wolfe, Dunavan, & Diamond, 2007) . A closer proximity with animals and higher human 60 population densities increased the establishment and spread of EIDs. Infectious diseases existed 61 in hunter-gatherers but were subject to differing evolutionary pressures that allowed them to persist 62 in low population densities (versus high population densities of agrarian societies). A fast-acting, 63 4 highly virulent disease would quickly kill off the sparse hunter-gather population before the 64 disease had a chance to spread, thus also killing off the disease. 65 The rise of agriculture and settling of peoples into close-knit communities clearly impacted . This reduced movement may also impact 82 EID formation by increasing the immune-naivety of the remaining host species because they will 83 no longer regularly interact with as many pathogens.

84
In this paper, we first quantify the global change in pathogen dispersal through faeces 85 (dispersed through the gut) and obligate ectoparasites (e.g. ticks) before and after the Late  Where D is digestibility, which we set to 0.5 as a parsimonious assumption because the actual 144 value is unknown for many extant and extinct animals.

145
Distance between consumption and defecation or straight line fecal transmission distance is simply However, animals rarely move in a straight line, and without any additional information, we can 151 assume a random walk pattern with a probability density function governed by a random walk as: Here we define the mean fecal diffusivity as the mean range in any pixel a generalist microbe could 156 travel during its lifetime assuming an equal chance of colonizing any mammal species. EID modelling 185 We then tested whether these changes in pathogen dispersal distance could help explain the

189
We also divide our analysis into vector driven (Table S3), non-vector driven (Table S4) and all 190 diseases (Table 2). To control for spatial reporting bias, they estimated the mean annual per country  Table S1). In addition to the 145 known EID outbreaks, 199 we randomly generated ~five times more random points (>600 points) to compare them (all results

200
in the paper are the average of three separate runs where the control points vary randomly) (see 201 Figure S2 as an example distribution). 202 We then used the Ordinary Least Squares (OLS) multiple regression models to predict 203 the EID events. We used Akaike's Information Criterion (AIC) for model inter-comparison, 204 corrected for small sample size. Whenever spatial data are used there is a risk of autocorrelation 205 because points closer to each other will have more similar signals than points far from each 206 other. We therefore used Simultaneous Auto-Regressive (SAR err ) models (Table 2)  values. We will refer to this as pseudo-R 2 in the paper even though we are aware that several 217 different estimates of model fit are frequently referred to as pseudo-R 2 . We also did a VIF  its ecosystem) averages ~8 km 2 (Table 1). In parts of Eurasia and southern South America, that 225 had a particularly high pre-extinction diversity of large-mammals, the mean home range exceeded 226 25 km 2 (Figure 1). Following the extinctions, the mean global home range of a generalist blood 227 parasite has been reduced to 1.1 km 2 or 14% of the previous global average ( Table 1). The Outside of abiotic dispersal by wind or water, this is a potentially important way for microbes to 236 move across an ecosystem. Following the megafauna extinctions, the mean distance travelled by 237 microbes globally through biotic means decreased to 0.5 km 2 or ~15% of the non-extinction value.

238
The largest declines in distance travelled are in the Americas and Eurasia. 239 We estimated the approximate increase in time for fecal pathogens and obligate  Table 2). We started with 16 variables described in 293   Table S1, including our six maps from Figure 1, but the model that best predicted EIDs included 294 ΔFD plus reporting bias (estimated as log (Journal of Infectious Disease articles (JID)); richer 295 countries with more scientists will find more EIDs), human population density and rainfall. Both  diversity is strongly positively correlated with pathogen diversity, and we also found this on its 301 own (Table 2). However, adding SRcurrent to our best model increases AIC (Table 2 - Table S3 and S4). In a sensitivity study (SI Appendix, Table   306 S5 and S6) we tested the resilience of our results and found that our model results remained 307 significant under a wide range of scenarios. For instance, moving the EID location randomly by 308 one pixel to estimate the great uncertainty in knowing the exact EID emergence coordinates, did 309 not greatly change our results. 310 We then create a new EID prediction map based on model FD (Table 2 and    population density, SRcurrent current species richness, Δ MHRchange in mean home range ( Figure   560 1c and eq 2), ΔFDchange in fecal diffusivity (Figure 1f and eq 7), and Rainaverage rainfall.

561
In column one, we show the variables of interest, in column two, we show the individual model 562 coefficient, r 2 and significance using the Bonferroni correction to determine significance or

578
A sensitivity study for our parameters is show in Figure S7.