How changes in biodiversity affect disease, particularly in the
face of large-scale land-use change, is a contentious topic in disease
ecology that has implications for public health and conservation policy.
The ‘dilution effect’ hypothesis argues that declines in biodiversity
are associated with increased disease risk, but this can be challenging
to demonstrate because many pathogens have complex life cycles such that
changes to the species composition and abundance of hosts can influence
the density and infection prevalence of vectors via multiple mechanisms.
Key to addressing this debate is a quantification of interactions
between hosts, vectors, and pathogens. In their recent study published
in Molecular Ecology, Kocher et al. (2022) captured thousands of
sandflies, some species of which are vectors for the Leishmaniaprotozoan that causes Leishmaniasis, across a human footprint gradient
in French Guiana (Fig. 1). By implementing DNA metabarcoding of vectors
combined with an innovative modeling approach, they effectively
quantified the nuanced relationships between changes in land-use,
mammalian host diversity, vector abundance, and parasite prevalence. In
support of the dilution effect hypothesis, Kocher et al. found that
sites with higher mammal diversity were associated with lower relative
abundance of reservoir hosts and higher Leishmania infection prevalence
in sandflies. However, while infection prevalence was lower when mammal
diversity was high, the density of sandfly vectors was higher, which
resulted in a weak overall effect of mammal diversity on the density of
infected vectors, the most important indicator of Leishmaniatransmission risk.
Debate over the dilution effect hypothesis in disease ecology (Levi et
al., 2016; Wood et al., 2014) has yet to be settled in part due to the
challenge of directly measuring interactions between hosts, vectors, and
pathogens. Parallel to this debate has been the advent of DNA
metabarcoding, which allows for the sequencing of taxonomically
informative genetic regions from a bulk DNA sample. More recently,
emerging research has demonstrated that DNA metabarcoding of
invertebrate-derived environmental DNA (iDNA) from hematophagous
insects, including sandlies, can be used to inform locally-present
vertebrate diversity (Kocher et al., 2017; Massey et al., 2022). Kocher
and coauthors capitalize on the power of DNA metabarcoding to address
the dilution effect using sandfly vectors as a source of DNA from the
vector (the sandfly), hosts (the wildlife the sandflies feed upon), and
the parasites (the Leishmania protozoa contracted from a
vertebrate host). Kocher et al. captured and sorted 18,508 female
sandflies (Fig. 2) and 855 individual blood-fed dipterans, 755 of which
were sandflies. To profile sand fly diversity and estimate the
prevalence of Leishmania parasites from across this landscape, they
sequenced pools of sandflies and found Leishmania DNA in 26.3% of their
pools. Separately, they sequenced individual, blood-fed sandflies (Fig.
3) to link sandfly vector species with specific mammalian hosts.
Sandflies fed upon a diverse suite of mammal species (n=28 including 11
known host species of Leishmania) with host-feeding preferences varying
by the genus of sandfly. Typically, the inference from sequencing
individual blood meals would end there, but in an important contribution
to the field Kocher at al. went on to develop an analytical framework to
extract much more information from this iDNA data.
As the practice of using DNA metabarcoding of iDNA for monitoring
biodiversity increases, there is substantial need to develop analytical
frameworks to pull greater inference from the data. Kocher et al.
developed an innovative hierarchical Bayesian modeling framework for
data containing species of vertebrate host nested within sandfly genus
nested within a sampling site. Each blood-fed sandfly contains the DNA
from a random draw of vertebrates at a site contingent on both the
feeding preferences of that sandfly species and the relative abundance
of hosts at that site. The key challenge is distinguishing between
feeding preferences and host abundance. Kocher et al’s innovation was to
make feeding preferences and relative abundances distinguishable by
sampling across vectors with distinct preferences (including leveraging
information from dipterans other than sandflies) and across sites with
distinct host communities. They recognized that the number of vector
bloodmeals coming from each vertebrate host is a multinomial process
resulting in a joint likelihood composed of the product of multinomial
likelihoods across all vector and sites. The resulting posterior
distribution would typically describe a vector of probabilities, but
Kocher et al. made these probabilities a sensible function of relative
host abundances and sandfly feeding preferences. The posteriors for
relative host abundances could then be used to calculate the Shannon
diversity index at each site directly from metabarcoding of iDNA, which
could then be linked to the density and infection prevalence of
sandflies. A really wonderful and powerful approach!
This is an exciting moment for combining statistical and molecular
approaches to facilitate more efficient and effective biodiversity data
collection in remote environments. Kocher at al. showed that
vector-derived DNA metabarcoding can measure host, vector, and pathogen
communities directly, and that with a large enough sample size these
data can be used to statistically model species interactions,
biodiversity, and landscape epidemiology. This framework should have
great utility for future research both in landscape epidemiology and for
the growing field of biodiversity monitoring using iDNA. Because
identifying individually blood-fed sandflies under a dissecting
microscope is extremely laborious, perhaps further advances in
statistical modeling would make use of iDNA metabarcoding of sandfly
pools rather than individual blood-fed vectors. Such an approach would
substantially increase throughput and reduce costs but presents
additional statistical challenges.
Kocher et al. used their methodological advance to make an important
scientific contribution to our understanding of the dilution effect. Key
assumptions underlying the dilution effect in this system are that (1)
diversity declines with anthropogenic disturbance, (2) more competent
hosts for Leishmania have higher relative abundance at degraded
sites, and (3) the vectors are sufficiently generalist feeders that
bloodmeals can be diverted away from competent hosts. Kocher at al. find
support for all three of these assumptions. However, the overall risk of
contracting Leishmania also depends on the abundance of sandfly
vectors, which they found to increase with mammal diversity counter to
the dilution effect hypothesis. Vector abundance can increase with host
diversity either because sandflies are host-limited and higher diversity
sites have higher host abundance, or because sandflies respond
negatively to the changing environmental conditions in degraded
habitats. We find the latter to be plausible because edge effects in
degraded tropical forests increase desiccation, which may reduce the
availability of humid habitats with decaying organic matter in which
sandflies reproduce (Moncaz, Faiman, Kirstein, & Warburg, 2012). Thus
land-use change may be associated with increased risk of contractingLeishmania if bitten by a sandfly (higher infection prevalence),
but reduced risk of being bitten because sandflies are less common
(lower abundance). The result was that diluting and amplifying forces
combined to produce no apparent effect of mammal diversity on the
density of infected sandflies. This acts as an important reminder that
teasing apart mechanisms is critical to understanding the effects of
biodiversity and land use change on disease risk.