Serology and statistical analyses
Blood samples were pre-diluted 1:30 and were analyzed for the presence
of antibodies against EBLV-1 using a modified Rapid Fluorescent Focus
Inhibition Test (RFFIT), in which EBLV-1b was used as challenge virus
(Leopardi et al., 2018; Serra-Cobo, Amengual, Carlos Abellán, & Bourhy,
2002). Samples were analyzed on a three-fold dilution basis and titers
were calculated through the Reed-Muench method and expressed as
LogD50/ml. Samples were considered positive when able to inhibit viral
growth at a minimum dilution of LogD50/ml ≥1.95.
We then tested how individual and environmental parameters influence
both the likelihood of being seropositive and the serological titer of
neutralizing antibodies against LYSVs in the target species M.
myotis. In particular, the variables tested included geographic
(altitude and coordinates), seasonal (before and after the birth pulse),
demographic (age and sex) and genetic parameters (relative presence of
individuals showing haplotypes clustering to distinct clades, using as a
reference mitochondrial haplogroups defined by Ruedi et al, 2008). Full
statistical analyses are reported as supplemental material
(Supplementary Methods). Briefly, we applied a population averaged
models using Generalized Estimating Equations (GEE) and a linear mixed
model (LMM) to investigate the effects of all variables, respectively on
the likelihood of showing neutralizing antibodies and on their titer
(Dohoo, Wayne, & Stryhn, 2009; Liang & Zeger, 1986; West, Welch, &
Galecki, 2007). For both types of models, we treated individuals sampled
within the same colony as a cluster, thus assuming that all observations
obtained within the same colony are correlated. Accordingly, we
performed a robust estimation of the variances of the regression
coefficients for the qualitative analysis and included a
random-intercept for colony in the LMM (Rao et al., 2014; Ying & Liu,
2006). To assess the goodness of proposed models, we used the
Quasi-likelihood under the independence model Criterion (QIC) and the
Area under the ROC Curve (AUC) for GEE models, or the Akaike Information
Criterion (AIC) and residual analysis for LMM models (Dohoo et al.,
2009; Littell, Milliken, Stroup, Wolfinger, & Oliver, 2006; Samur,
Coskunfirat, & Saka, 2014) considering a p-value <0.05 as
significant.