Transbounding Emerging Dis - 2020 - Meza - Predicting The Presence and Titre of Rabies Virus Neutralizing Antibodies From

Received: 1 May 2020 | Revised: 3 August 2020 | Accepted: 3 September 2020
DOI: 10.1111/tbed.13826
ORIGINAL ARTICLE
Predicting the presence and titre of rabies virus-neutralizing

antibodies from low-volume serum samples in low-containment
facilities
Diana K. Meza1,2 | Alice Broos1,2 | Daniel J. Becker3 | Abdelkader Behdenna1 |

Brian J. Willett2 | Mafalda Viana1 | Daniel G. Streicker1,2
1
Institute of Biodiversity, Animal Health and
Comparative Medicine, College of Medical, Abstract
Veterinary and Life Sciences, University of Serology is a core component of the surveillance and management of viral zoon-
Glasgow, Glasgow, UK
2 oses. Virus neutralization tests are a gold standard serological diagnostic, but re-
Medical Research Council, University
of Glasgow Centre for Virus Research, quirements for large volumes of serum and high biosafety containment can limit
Glasgow, UK
widespread use. Here, focusing on Rabies lyssavirus, a globally important zoonosis,
3
Department of Biology, Indiana University,
Bloomington, IN, USA
we developed a pseudotype micro-neutralization rapid fluorescent focus inhibition
test (pmRFFIT) that overcomes these limitations. Specifically, we adapted an existing
Correspondence
Diana K. Meza, Institute of Biodiversity,
micro-neutralization test to use a green fluorescent protein-tagged murine leukaemia
Animal Health and Comparative Medicine, virus pseudotype in lieu of pathogenic rabies virus, reducing the need for specialized
College of Medical, Veterinary and Life
Sciences, University of Glasgow, Glasgow
reagents for antigen detection and enabling use in low-containment laboratories.
G12 8QQ, Scotland, UK. We further used statistical models to generate rapid, quantitative predictions of the
Email: [email protected]
probability and titre of rabies virus-neutralizing antibodies from microscopic imag-
Funding information ing of neutralization outcomes. Using 47 serum samples from domestic dogs with
Consejo Nacional de Ciencia y Tecnología,
Grant/Award Number: 334795 / 472296;
neutralizing antibody titres estimated using the fluorescent antibody virus neutrali-
Wellcome Trust and Royal Society, Grant/ zation test (FAVN), pmRFFIT showed moderate sensitivity (78.79%) and high speci-
Award Number: 102507/Z/13/Z; Human
Frontier Science Program, Grant/Award
ficity (84.62%). Despite small conflicts, titre predictions were correlated across tests
Number: RGP0013/2018; Wellcome repeated on different dates both for dog samples (r = 0.93) and in a second data set of
Senior Research Fellowship, Grant/Award
Number: 217221/Z/19/Z; Wellcome Trust
sera from wild common vampire bats (r = 0.72, N = 41), indicating repeatability. Our
& Royal society, Grant/Award Number: test uses a starting volume of 3.5 µl of serum, estimates titres from a single dilution
102507/Z/13/Z; Royal Society; National
Science Foundation, Grant/Award Number:
of serum rather than requiring multiple dilutions and end point titration, and may be
DEB-1020966; Leverhulme Trust, Grant/ adapted to target neutralizing antibodies against alternative lyssavirus species. The
Award Number: RPG-2015- 259
pmRFFIT enables high-throughput detection of rabies virus-neutralizing antibodies
in low-biocontainment settings and is suited to studies in wild or captive animals
where large serum volumes cannot be obtained.
KEYWORDS
biologic assay, Desmodus rotundus, generalized linear mixed models, immunofluorescence
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium,
provided the original work is properly cited.
© 2020 The Authors. Transboundary and Emerging Diseases published by Wiley-VCH GmbH
1564 |
wileyonlinelibrary.com/journal/tbed Transbound. Emerg. Dis. 2021;68:1564–1576.
|
18651682, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/tbed.13826 by Nat Prov Indonesia, Wiley Online Library on [22/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MEZA et al. 1565
1 | I NTRO D U C TI O N Obregón-Morales et al., 2017; Steece & Altenbach, 1989). As

such, antibody detection is commonly used for ecological and ep-
The last few decades have seen a surge in newly emerging human vi- idemiological studies of bat rabies (Blackwood et al., 2013; Costa
ruses that originate from wildlife (Cunningham et al., 2017; Daszak et al., 2013; de Thoisy et al., 2016; George et al., 2011; Horton
et al., 2000; Goodin et al., 2018). Key examples include Nipah virus et al., 2020; Streicker et al., 2012; Wright et al., 2010).
(Gurley et al., 2017), Seoul virus (Kerins et al., 2018), 2009 H1N1 Existing tests to detect RV antibodies include enzyme-linked
(Mena et al., 2016) and the recent SARS-CoV-2 (Zhou et al., 2020). immunosorbent assays (ELISAs) (Barton & Campbell, 1988; Cliquet
Understanding the epidemiological dynamics of such viruses et al., 2004; Wasniewski & Cliquet, 2012) and virus neutralization
within their natural host populations is a fundamental component tests (Cliquet et al., 1998; Smith et al., 1973). ELISAs are a rapid and
to discerning the spatiotemporal dynamics of past outbreaks and effective method for testing large numbers of samples, but they do
anticipating future emergence (Cunningham et al., 2017; Plowright not detect virus neutralization (Ma et al., 2012; Welch et al., 2009).
et al., 2016). Investigating the dynamics of zoonotic viruses within Moreover, since ELISAs ideally require a secondary antibody spe-
wildlife presents multiple challenges such as limited sample sizes, cific to the host species of the original sample (Reynes et al., 2004),
biased sampling and multiple diagnostic tests that are difficult to which is often non-existent, they are not widely used in wildlife
compare. Moreover, viruses themselves may be undetectable at rabies surveillance. Tests detecting virus-neutralizing antibodies
the moment of sampling when infectious periods are short or virus (VNAs) are more commonly adopted (De Benedictis et al., 2012; Irie
shedding is intermittent (Becker et al., 2019; Gilbert et al., 2013; & Kawai, 2002; Moore & Hanlon, 2010). Specifically, the fluorescent
Plowright et al., 2019). Consequently, key parameters needed to antibody virus neutralization test (FAVN) and the rapid fluorescent
inform population-level disease dynamics (e.g., incidence, infec- focus inhibition test (RFFIT) are considered the gold standard for
tion and incubation periods) are difficult or impossible to measure measuring vaccination response by the World Health Organization
directly (Borremans et al., 2016). Serological tests offer a powerful (WHO) (Briggs et al., 1998; De Benedictis et al., 2012). Both the
alternative to approaches that rely on pathogen detection (Gilbert FAVN and RFFIT have been modified to facilitate use with wildlife
et al., 2013). Since pathogen-specific antibodies generally persist samples. For instance, Kuzmin et al. (2008) modified the RFFIT into a
longer than the pathogen itself, serological data can inform indi- micro-neutralization test using 4-well Teflon-coated slides instead of
vidual exposure histories and population seroprevalence and, in 8-well chamber slides. This produced a sensitive and specific test that
some cases, help approximate the force of infection (Borremans only required 3.5 µl of serum (regular RFFIT and FAVN require ~50
et al., 2016; Gamble et al., 2020; Gilbert et al., 2013; Metcalf μl of serum) (Kuzmin et al., 2008). Several other laboratories have
et al., 2016). However, applying serology to longitudinal studies introduced the modification of pseudotype viruses (i.e., non-rabies
of wildlife presents distinct challenges from tests used in clinical viruses engineered to express heterologous envelope glycopro-
diagnostic settings, where accuracy takes precedence over scal- teins) to quantify VNA titres without highly pathogenic live lyssa-
ability. In particular, serological tests for studies of wildlife should viruses (Temperton et al., 2015), which in most countries involves
be amenable to the small volumes of serum that often character- biosafety-level (BSL)-3 laboratories to produce concentrated virus
ize collections from small-bodied hosts (e.g., rodents, bats, birds), stocks (WHO, 2018). Pseudotyped viruses based on vesicular sto-
scalable to large numbers of samples required for such studies, matitis virus (Moeschler et al., 2016), human immunodeficiency virus
and possible to implement in low-biocontainment laboratories. and murine leukaemia virus (MLV) (Wright et al., 2008) tend to be
Rabies lyssavirus (RV; Genus Lyssavirus, Family Rhabdoviridae) is highly sensitive and specific relative to live lyssavirus counterparts.
a zoonotic virus transmitted in saliva by the bite of infected mam- Viral pseudotypes have also been modified to incorporate molecu-
mals (mainly Carnivora and Chiroptera) (Rupprecht et al., 2017). lar biomarkers such as firefly luciferase, renilla luciferase or green
RV constitutes a substantial global health problem that is respon- fluorescent protein (GFP), eliminating the need to fix and stain cells
sible for over 59,000 human fatalities annually, mostly attributable with a fluorescein isothiocyanate (FITC)-conjugated rabies antibody,
to domestic dogs (WHO, 2017). In Latin America, vampire bat RV which reduces test cost (Bentley et al., 2015; Mather et al., 2013;
causes more cases of human and domestic animal rabies mortality Moeschler et al., 2016; Moore & Hanlon, 2010; Wright et al., 2008).
than RV transmitted by dogs. Similarly, bats are consistently among Despite these improvements, existing variations of RV neutral-
the most frequent sources of human rabies exposure in the United ization tests have several limitations. First, all tests require serial di-
States (Ma et al., 2020; Schneider et al., 2009). Bat lyssaviruses ex- lutions of serum samples, each of which requires manual microscope
emplify the challenges of studying zoonoses using pathogen detec- scoring that increases personnel time and risks of error (Moeschler
tion and the potential advantages of serological inference (Turmelle et al., 2016; Péharpré et al., 1999). Second, testing is commonly split
et al., 2010). Active infections are rarely detected due to low popu- into screening and end point titration phases, requiring tests to be
lation-level incidence and the short infectious period that precedes run in duplicate or sequentially. Third, pseudotypes have not yet
death (Jackson et al., 2008; Turmelle et al., 2010). However, abortive been integrated into micro-neutralization tests, implying that tests
infections (i.e., bats that are exposed to RV but do not become in- to detect VNAs against most lyssaviruses in their natural bat reser-
fectious) are routinely observed through antibody presence in ap- voirs can only be carried out in BSL-3 or higher facilities. The grow-
parently healthy bats (Constantine et al., 1968; Jackson et al., 2008; ing demand to study wildlife on large spatial and temporal scales
|
1566 MEZA et al.
would benefit from minimizing these logistical constraints (Becker and, when appropriate, titres of RVNAs in test sera. Our approach,
et al., 2019). which we refer to as a pseudotype micro-neutralization RFFIT (here-
The ideal serological diagnostic would be a scalable virus neu- after, pmRFFIT, Figure 1), provides a safe, low-cost alternative to
tralization test that requires low-volume samples and can be car- standard RV neutralization tests that is suitable for large-scale, pop-
ried out in any standard microbiology/cell culture laboratory. Here, ulation-level studies or laboratory studies where small animals are
the micro-neutralization RFFIT is adapted to use an MLV-based viral longitudinally sampled.
pseudotype bearing the RV glycoprotein and carrying a GFP marker
gene. Further, we introduce a novel quantitative approach that com-
bines digital image analysis of infected cells with statistical analy- 2 | M E TH O DS
sis, which allows us to estimate rabies virus-neutralizing antibody
(RVNA) titres without multiple serum dilutions or rounds of testing. 2.1 | Laboratory work
Our approach relies on the expectation of a negative relationship
between RVNA concentrations in serum and the number of cells that 2.1.1 | Cell lines
become infected upon viral challenge in the presence of that serum.
Defining that relationship quantitatively using known titres of stan- For all the neutralization tests, mouse neuroblastoma cells N2A
dard rabies immune globulin (SRIG) allows predicting the presence (Neuro-2A, ATCC® CCL131™) were cultured in Minimum Essential
F I G U R E 1 Workflow of the pmRFFIT approach. (a) Set-up of the neutralization test using an MLV(RG) pseudotype and 2 dilutions of
either SRIG or animal serum. (b) Microscopy phase and imaging to perform the cell count of the fluorescent cells to construct a database to
fit the statistical models. (c) Construction of the statistical models with two different types of prediction [Colour figure can be viewed at
wileyonlinelibrary.com]
|
MEZA et al. 1567
Media (MEM) supplemented with 10% foetal bovine serum (FBS), 1% six different titre concentrations of SRIG in lieu of serum samples:
100× non-essential amino acid (NEAA), 1% 200mM L-glutamine and 0.01, 0.025, 0.05, 0.1, 0.15 and 0.2 IU/ml (reconstituted at 30 IU/ml
1% antibiotic–antimycotic. For viral pseudotype production, human in 1.0 ml of nuclease-free water, 2nd WHO International Standard,
embryonic kidney 293T cells (293T, ATCC® CRL-3216™) were cul- Human; NIBSC, UK) (Figure 1a).
tured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented
with 10% FBS, 1% 200 mM L-glutamine, 100 units/ml penicillin,
1 µg/ml streptomycin and 400 µg/ml G418 (Geneticin) antibiotic. 2.1.4 | Imaging and data processing
After incubation, slides were photographed at 4× magnification

2.1.2 | Pseudotype production under a fluorescence microscope (EVOS FL Cell Imaging System).
As representative images of the whole well, five equally sized fields
MLV pseudotypes expressing the challenge virus standard-11 (four corners and centre) were selected and photographed clockwise
(CVS-11) rabies virus glycoprotein (MLV(RG)) were generated by from the top left corner, such that the fifth photograph was always
co-transfection of pCMVi (MLV gag-pol expression vector), pCNCG the centre field. Each photograph was processed through the freely
(MLV viral origin and GFP reporter gene) and pI.18-CVS-11 (plasmid available software, ImageJ (version 1.52k (Rasband, 2018)) using the
encoding the CVS-11 RG) (Bock et al., 2000; Towers et al., 2000; Autolocal Threshold Phansalkar plugin (size radius 15) (part of the Fiji
Wright et al., 2008). Plasmids were transfected using polyethyl- distribution (Schindelin et al., 2012)). Each image was transformed
enimine (Polysciences, Inc.) into 293T cells at a ratio of 1:1.5:1 of into a binary representation (every pixel stored as a single bit), indi-
pCMVi/pCNCG/pl.18-CVS-11. After 72 hr, the supernatant contain- cating the presence (white) and absence (black in the background) of
ing the viral pseudotypes was harvested, aliquoted and stored at GFP fluorescence. Next, the command “Analyze Particles” was used
−80°C until further use. Virus concentration was determined in N2A to count the total number of fluorescent cells per field (i.e., infected
cells by calculating the TCID50 (50% tissue culture infective dose) cells). This command grouped and counted the white neighbouring
using the end point method and using the Spearman–Kärber formula pixels with a pre-determined size area and circularity to be a single
(Condit, 2001; Hierholzer & Killington, 1996). cell (size area: 5–50 circularity: 0.80–1.0), so counts corresponded
to the number of infected cells. Cell count outputs were converted
into a standardized spreadsheet using a Python version 3.7.2 script
2.1.3 | Neutralization test (Python Core Team, 2019) (script available in supplementary infor-
mation). At the end of the image processing step, each serum sam-
A two-dilution micro-neutralization test was established follow- ple was described by 10 data points consisting of the number of the
ing Kuzmin et al. (2008); however, in place of 4-well Teflon-coated fluorescent cells in each of 5 fields (photographs) in the 1:10 and
slides, 6-well Teflon-coated slides (Tekdon Inc.) were used, enabling 1:25 dilutions (Figure 1b).
3 samples to be analysed per slide. Each serum sample (starting vol-
ume of 3.5 µl) was screened at a 1:10 and 1:25 dilution (a total of 2
wells per sample). Serum dilutions were inoculated with 12.5 µl of 2.2 | Statistical analysis
MLV(RG) (viral input: 300–400 TCID50) and incubated in humidified
square petri dishes at 37°C, 5% CO2 for 90 min. After incubation, All statistical analyses were executed in R (R Core Team, 2020).
25 µl of N2A cells (2 × 106 cells/ml) was added to the serum–virus To predict the probability of RVNA presence, a generalized linear
mixture and slides were incubated at 37°C, 5% CO2 for 66 to 72 hr mixed model (GLMM) with a binomial distribution was fit to the bi-
in the humidified square petri dishes. This extended incubation pe- nary outcome of the cell counts from the SRIG concentration series.
riod (compared to the FAVN (48 hr) or the RFFIT (20 to 24 hr)) was Titres ≤ 0.1 IU/ml were considered RVNA-negative and > 0.1 IU/ml
required for the MLV(RG) to generate sufficient GFP expression for were considered RVNA-positive. Other serology studies in wildlife
later microscopy. Since our pseudotype virus used GFP to indicate have used similar thresholds to detect RVNA (Araujo et al., 2014;
viral entry into cells, neither acetone fixation nor staining with anti- Campos et al., 2019; Marcelo Azevedo de Paula Antunes et al., 2017;
rabies monoclonal globulin was required. Every neutralization test Silva et al., 2010). To predict RVNA titres, a GLMM with a log-normal
included three internal controls: (1) a cell-only control where no distribution was fit to the infected cell counts across the SRIG con-
viral pseudotype or serum was added to the well; (2) a viral pseudo- centration series. For both the binomial and log-normal models, a
type control, comprising back titrations of the MLV(RG) with (a) the model was constructed including all data (i.e., from the 1:10 and 1:25
original viral pseudotype concentration (300–400 TCID50), (b) a 1:10 dilutions). For this model, two fixed effects were considered: (1) the
dilution and (c) a 1:100 dilution; and (3) a negative control for neu- count of virus-infected N2A cells (scaled to improve model conver-
tralization that included two wells with 300–400 TCID50 of MLV(RG) gence) and (2) the serum dilution level (two factors: 1:10 and 1:25
without any serum. A standard curve was generated with each neu- dilution). Random slope and intercept terms were considered for the
tralization test describing the expected number of infected cells ver- date the test was run (“test date”) to account for observed varia-
sus titre concentration. For this, the same protocol was used with tion in the relationships between SRIG titres and infected cell counts
|
1568 MEZA et al.
across dates (Figure 2). A random slope term was also considered titres). For the binomial models, the sensitivity and specificity with
for the field number (1 to 5) within each microscope well (“field”) a threshold of > 0.1 IU/ml for seropositivity were used. For the log-
to account for variation in cell counts between fields (the middle normal models, the Spearman correlation coefficients of predicted
field, field 5, had more agglomerated cells in particular). To evaluate and known SRIG titres were compared.
whether a simpler, single dilution test produced comparable results,
the full data set was then subset to the 1:10 dilution only. The bino-
mial and log-normal models fit to this data subset included only the 2.2.2 | Validation of the pmRFFIT
fixed effect of the virus-infected N2A cell counts, but the random
effects were identical to those explained above (i.e., test date and To understand the variability of the pmRFFIT, replicate SRIG titre
field). Models were fit using the lme4 package (Bates et al., 2015). concentration curves were produced on 6 different dates between
The predict function was used to generate the predicted probability 30/05/2019 and 27/06/2019 (hereafter “Test 1” through “Test 6”).
that a SRIG concentration or serum sample was seropositive (bino- We evaluated dispersion in the counts (and hence potential predic-
mial model) and its corresponding RVNA titre (log-normal model). tive power) for the 1:10 versus the 1:25 dilution series of different
Predictions per field were averaged to obtain results per sample SRIG concentrations by calculating the interquartile range (IQR) of
(Figure 1c). the infected cell count.
Dog (N = 47) and bat (N = 41) sera were used to assess pmRFFIT
repeatability. Dog sera had previously been tested using FAVN at
2.2.1 | Model selection the Animal Health and Veterinary Laboratories Agency, Weybridge
(Wright et al., 2009). Bat sera were collected from common vampire
To select the best-performing models, a top-down model selection bats (Desmodus rotundus) as part of an ongoing field study in Peru
strategy was implemented (Zuur et al., 2009). First, we identified the (collection permit: 0142–2015 SERFOR/DGGSPFFS; exportation
optimal random effects by comparing models with alternative ran- permit: 003327-SERFOR). Specific and individual sample informa-
dom-effect structures, but the same fixed effects, using the Akaike tion is available in the supplementary information (SI, Table S1).
information criterion corrected for small sample size (AICc). This was To quantify repeatability, dog and bat sera were processed
calculated using the MuMIn package (Bartoń, 2020). A lower AICc using the pmRFFIT as described above on two separate dates. For
with a difference of at least two values (ΔAICc > 2) was considered the binomial model, repeatability was measured as the proportion
as evidence of improved model fit. Next, keeping the chosen ran- of samples with the same predicted serological status on both test
dom effects, models with the two different data sets were evalu- dates. For the log-normal model, repeatability was calculated as the
ated (simpler model: 1:10 data set, and more complex model: 1:10 Spearman correlation coefficient of predicted RVNA titres from dif-
and 1:25 data set) (Table 1). Since models fit to different data sets ferent test dates.
cannot be compared through AICc, the simpler and more complex We measured the accuracy of the pmRFFIT as the overall propor-
models were compared using predictions for SRIG (sera with known tion of positive and negative samples correctly detected relative to
F I G U R E 2 Raw counts of fluorescent

cells for the SRIG concentration curves
from 6 different pmRFFIT from different
test dates and for each dilution. The
dilutions were performed on the same
date, with one well next to the other. Each
colour represents a different test date.
The loess curves denote differences in the
counts from test to test [Colour figure can
be viewed at wileyonlinelibrary.com]
|
MEZA et al. 1569
TA B L E 1 Random-effect structures
Model Random effects AICc ΔAICc W
investigated for the binomial (B) and
log-normal (L) models, and respective B 1. Test date as random intercept and 91.62 0.0 0.71
model fit measures: AICc, difference in random slope
AICc from best model (ΔAICc) and weight 2. Test date as random intercept and 93.49 1.87 0.28
(W). The fixed effects for all these models random slope + number of field as
were identical and included the scaled random intercept
count of the fluorescent cells and the
3. Test date as random intercept 102.32 10.70 0.0
dilution effect
4. Test date as random 104.28 12.66 0.0
intercept + number of field as random
intercept
L 1. Test date as random intercept and −1746.59 0.0 0.7
random slope + number of field as
random intercept
2. Test date as random intercept and −1744.89 1.70 0.3
random slope
3. Test date as random −1637.32 109.27 0.0
intercept + number of field as random
intercept
4. Test date as random intercept −1637.30 109.29 0.0
F I G U R E 3 Results of the models fit to the SRIG control data. All graphs compare the one-dilution (1:10) and the two-dilution (1:10
and 1:25) models. (a) Model fit of the titres of the SRIG concentration curve versus the predicted probability from the binomial model.
The horizontal dash line represents the 95% probability, and the vertical dash line represents the previously set threshold at 0.1 IU/ml.
(b) Sensitivity and specificity of the binomial prediction per test date. (c) Correlation of the titres of the observed SRIG values versus the
predicted values with the log-normal models. (d) Comparison of the Spearman coefficients between the correlations of the observed titres
versus the predicted titres per test date [Colour figure can be viewed at wileyonlinelibrary.com]
the FAVN, using binomial predictions from dog sera. The same data the assumption that future applications would preferably use a single
were used to estimate the sensitivity (accurately predicted positive/ test. To compare pmRFFIT accuracy in different FAVN titre ranges, the
(accurately predicted positive + inaccurately predicted negative)) and RVNA values obtained through FAVN were categorized through the
specificity (accurately predicted negative/(accurately predicted neg- cut function in R. The number of categories was obtained by dividing
ative + inaccurately predicted positive)) of the pmRFFIT. Sera with the range of the RVNA titre values (maximum minus minimum value
FAVN titres > 0.1 IU/ml were classified as positive and ≤ 0.1IU/ml as obtained through FAVN) by the optimal category width, calculated
negative. Only data from the first pmRFFIT test date were used under with the Freedman–Diaconis rule (Freedman & Diaconis, 1981).
|
1570 MEZA et al.
F I G U R E 4 Repeatability of the
predictions in bat (N = 41) and dog
(N = 47) samples. (a) Percentage of
repeatability of the seropositive outcome
of the bat and dog serum samples with
the pmRFFIT test. (b) Linear correlation
of the predicted titres between the first
and second tests for dog serum and
bat serum. Diamonds indicate samples
that were contradictory in the binomial
prediction [Colour figure can be viewed at
wileyonlinelibrary.com]
We later re-evaluated accuracy for weak positives using only the one-dilution model was more sensitive (100% versus 58.33%,
the lowest titre FAVN samples. Since we designed the pmRFFIT Figure 3a,b). Furthermore, the two-dilution binomial model failed
to quantify low RVNA titres and many of the available dog sera to correctly predict the seropositive controls on 4 out of the 6 test
had RVNA titres above the range of our test, direct comparison of dates, confirming improved performance of the one-dilution model
continuous titre predictions for the whole range was not feasible. (Figure 3b).
Instead, we compared predicted pmRFFIT titres using the relevant
FAVN range (0–0.22 IU/ml). To evaluate the sensitivity and spec-
ificity of different threshold values in the log-normal predictions, 3.2 | Prediction of titre values using log-
we applied a receiver operating characteristic (ROC) curve analy- normal GLMM
sis. The ROC curve was produced across all the range of evaluated
titres, using a threshold of > 0.1 IU/ml (positives) as the bench- The log-normal GLMMs accurately predicted RVNA titres from the
mark reference. data sets generated through our protocol on different test dates
(Figure 3). The best log-normal model included a random intercept
and slope for test date. Although the most complex model had the
3 | R E S U LT S lowest AICc, the simpler model (without the random intercept of
field) had a ΔAICc < 2 (Table 1). Observed and predicted SRIG titres
We assessed the variability of the pmRFFIT using the SRIG titre were highly correlated for both the one- and two-dilution models
concentration curves generated in 6 different dates (“Test 1” (r = 0.95, Figure 3c). When comparing test dates (i.e., one-to-one
through “Test 6”). As expected, the number of infected cells de- comparison between correlations of the one-dilution and the two-
clined at higher SRIG titres in all replicates; however, the shape dilution model from the same test dates), the correlation coefficients
of the antibody decay curve varied across test dates (Figure 2). At were similar, suggesting the simpler one-dilution model is sufficient
the 0.1 IU/ml SRIG concentration, infected cell counts were more for titre prediction (Figure 3d).
dispersed in the 1:25 dilution than in the 1:10 dilution, as indicated
by higher IQR within each of the 6 test dates. Across all the SRIG
concentrations in all test dates (N = 36), 77.78% of the count com- 3.3 | Repeatability, accuracy, sensitivity and
parisons were less dispersed in the 1:10 dilution suggesting this specificity with animal sera
dilution could be more precise for downstream statistical analysis
(SI, Figure S2). Among all the animal serum samples (N = 88), 95.45% were con-
sistently predicted as seropositive or seronegative (92.68% for bats
and 97.87% for dogs). Bat samples (N = 41) had a repeatability of
3.1 | Prediction of seropositivity using 66.7% for positive samples (6/9 positive samples were positive in the
binomial GLMM second round) and 100% for negative samples (N = 32). Dog sam-
ples (N = 47) had 100% repeatability for positive samples (N = 28)
Binomial GLMMs accurately predicted seropositive and seronega- and 94.74% for negative samples (18/19 negative samples remained
tive SRIG concentrations (Figure 3). The best random effects for negative in the second round) (Figure 4a, SI, Figure S4). Predicted
the binomial model included a random slope and intercept for test titres between test repetitions were correlated for both sera (both
date (Table 1). The models built with the 1:10 dilution data only r = 0.81, bat r = 0.72, dog r = 0.93, Figure 4b).
(“one-dilution model”) and from both the 1:10 and 1:25 dilution Relative to FAVN, pmRFFIT predictions using the binomial
data (“two-dilution model”) had equivalent specificity (100%), but model on dog sera showed overall moderate accuracy (80.43%
|
MEZA et al. 1571
F I G U R E 5 Validation of pmRFFIT
with the FAVN on the dog serum
samples. (a) Seropositivity comparison
for all FAVN titres: sensitivity = 78.79%;
specificity = 85.71%; overall
accuracy = 80.85% (N = 47). (b)
Seropositive comparison of lower FAVN
titres: overall accuracy = 72.73% (N = 33).
(c) ROC curve of the log-normal titre
prediction of the first test compared to
the FAVN at threshold > 0.1 IU/ml for
seropositivity. The continuous thresholds
for the pmRFFIT are in green [Colour
figure can be viewed at wileyonlinelibrary.
com]
of samples in agreement with FAVN; sensitivity = 78.79%; spec- 4 | D I S CU S S I O N

ificity = 84.62%). Dog samples categorized as positive according
to the binomial prediction (N = 28) had a higher correct predic- The most commonly applied serological tests to detect RVNA titres
tion. Only 2 positive samples failed in the pmRFFIT prediction challenge a range of serial dilutions of serum with infectious RV.
compared to the FAVN. For dog samples categorized as negative This process is labour-intensive and requires laboratory capacity to
by the binomial pmRFFIT (N = 9), 7 samples were contradictory. grow large quantities of pathogenic RV. Here, we provide an alterna-
These misclassified samples were consistently predicted as neg- tive serological framework that uses a combination of digital image
ative in the pmRFFIT, and titres displayed a strong correlation analysis and statistical analysis to estimate the presence and titre of
between tests (r = 0.93, SI, Table S3). The binomial model was RVNA from a single dilution using only 3.5 µl of serum.
most accurate for samples with high RVNA titres (>1.04 IU/ml, The pmRFFIT differs from other lyssavirus neutralization tests in
N = 13; 100% predicted to be seropositive; Figure 5a). At lower several key aspects. It uses an MLV(RG) pseudotype rather than patho-
FAVN titres (<1.04 IU/ml, N = 33), 72.73% of samples were pre- genic RV, allowing the pmRFFIT to be performed in any low-contain-
dicted correctly by the pmRFFIT. To evaluate the accuracy of ment laboratory with appropriate cell culture and microscopy facilities.
the binomial prediction in this lower range, the titre categoriza- The addition of GFP expression is significant, since it removes the
tion process was repeated using the subset of data with FAVN need for FITC-conjugated antibody (reducing reagent costs) and the
titres ≤ 0.871 IU/ml (N = 33). Among the higher titre samples fixation and staining steps used in traditional RFFIT or FAVN. One po-
in this particular set (>0.27 to ≤ 0.871 IU/ml, N = 11), 90.91% tential drawback of using GFP expression to measure infectivity is the
were correctly predicted as seropositive (Figure 5b). From the prolonged neutralization period (66-hr versus 24-hr RFFIT and 48-hr
samples with titres ≤ 0.27 IU/ml, only 63.64% were correctly FAVN) required to gain sufficient fluorescence for image processing
predicted (N = 22, Figure 5b). Accuracy rose to 83.33% in the (Aubert, 1992; Smith et al., 1973). Longer neutralization requirements
lowest FAVN titre range (≤0.07 IU/ml, N = 12). The pmRFFIT ti- (60 hr) were also required in a FAVN modification using a GFP express-
tres from the log-normal prediction were correlated with FAVN ing recombinant CVS-11-eGFP but did not alter results relative to the
titres (r = 0.51–0.85, depending on the data subset, SI Figure S3). test run with CVS-11 (Xue et al., 2014). Fortunately, extended incu-
The ROC curve relative to the FAVN titre values showed that bations are unlikely to alter neutralization outcomes since MLV(RG)
threshold values around 0.166 IU/ml maximized sensitivity and pseudotype is replication incompetent. This prevents infection of ad-
specificity. ditional cells during incubation (Temperton et al., 2015).
|
1572 MEZA et al.
The pmRFFIT also uses an imaging pipeline that combines sys- Cell-based experiments are expected to be more variable than
tematic photography of microscope fields with automated digital other serological diagnostics, and often the allowed variation for
image processing to count infected cells. Microscopy in neutralization test precision is up to 30% (Kostense et al., 2012; Timiryasova
tests is time-consuming and presents challenges for interlaboratory et al., 2019). For example, Kostense et al. (2012) evaluated RFFIT
comparisons due to multiple sources of variation, especially those repeatability by testing 3 different types of immunoglobulins (HRIG,
that affect the manual readout (e.g., laboratory user, manual pipet- CL184 and SRIG), and intratest variation was 26%, 18% and 25%,
ting, uneven cell monolayer) (Briggs et al., 1998; Cliquet et al., 1998; respectively. Similarly, Timiryasova et al. (2019) validated a RFFIT
Eschbaumer et al., 2014; Péharpré et al., 1999; Timiryasova protocol with 15.7% variation among intratest repetitions. In com-
et al., 2019). The pmRFFIT standardized approach minimizes these parison, our test showed higher repeatability for serum samples
sources of error while potentially reducing microscope operator time. from dogs. Repeatability was slightly lower in bat samples; however,
Moreover, the imaging process generates traceable and permanent two out of the 3 conflicting bat samples were explainable as thresh-
electronic records of the raw data, eliminating the need to manually old effects. Specifically, predicted titres in both tests occurred near
digitize records of field counts. Several investigators have previously the threshold defining seropositivity and varied by < 0.05 IU/mL be-
incorporated image processing into RV neutralization tests. For ex- tween test replicates, indicating only trivial variation between tests
ample, Pérhapré et al. (1999) modified the FAVN and RFFIT readout (Figure 4b, SI, Table S2). Nevertheless, quantitative titre predictions
by using an automated image analysis; however, one important draw- across replicate runs were less consistent in bat samples relative to
back was the high cost of the equipment (a motorized stage and the dogs (Figure 4b). One explanation for this finding could be linked to
software required for this). The pmRFFIT uses freely available image serum quality in bat samples arising from the collection of whole
processing program and plugins (ImageJ) making the approach acces- blood using peripheral venipuncture and capillary collection, rather
sible for any user. Streicker et al. (2012) also used photography and than cephalic or jugular venipuncture used for dog samples. Higher
image processing to count pixels using a microRFFIT but did not make haemolysis levels in bat samples could have increased variation in
full use of the quantitative nature of imaging data to obtain RVNA test outcomes (Neumann & Bonistalli, 2009). As a second possibility,
titres and used pathogenic RV rather than a viral pseudotype. Finally, predictions of lower titres tended to have a higher variability and
instead of scoring microscope fields or wells as virus positive or neg- titres in bats were on average lower than dogs, many of which were
ative across many dilutions to obtain serological status and titre, the previously vaccinated. Indeed, 26 out of 47 dogs compared to 5 out
pmRFFIT achieves this by quantifying levels of cellular infection in of 41 bat samples had predicted titres ≥ 0.15 IU/ml. Consistent with
a single serum dilution. The efficacy of this statistical modelling ap- our findings, Kostenze et al. (2012) also detected that the RFFIT had
proach highlights the value of historically underutilized quantitative prediction limitations at lower titre levels (limit of quantification:
data on cellular infectivity for lyssavirus serology. 0.2 IU/ml, limit of detection: 0.118 IU/ml). Moreover, they observed
Model selection indicated substantial day-to-day variation in the antibody concentration curves had an initial and evident bend in
SRIG dilution series. This effect was unsurprising since virus neutral- the titre concentration at 0.2 IU/ml, but the ability to observe slope
ization tests are biologically dynamic systems that can be influenced changes in the lowest concentrations of serum (<0.1 IU/ml) was
by many factors (e.g., variability in the humidity of the incubator, challenging. Overall, the pmRFFIT was most precise in discriminat-
technical manipulation, light condition of the microscope, variabil- ing strong positives from strong negatives, where the highest and
ity in GFP expression in the cells) (Briggs et al., 1998; Hammami lowest titres were constantly predicted as such. In consideration of
et al., 1999; Kostense et al., 2012). Since our statistical approach these points, standard serological methods might have limitations
handles this variability through the random effect of test date, the measuring low RVNA titres (<0.5 IU/ml), while the pmRFFIT grants a
pmRFFIT is best suited for large numbers of serum samples that highly repeatable approach for these titre values.
require testing to be carried out across multiple batches. However, Using the external validation set of dog samples with known
performance is only marginally reduced when running single models RVNA titres, the overall accuracy of the pmRFFIT was moderately
for each test date, implying the pmRFFIT may still be useful when high (Figure 5). The pmRFFIT was also able to estimate RVNA titres
fewer samples are available for testing (see SI, Figures S4 and S5). from a single serum dilution which correlated with known FAVN titres
Surprisingly, fitting the GLMMs to data from a single 1:10 dilution of (SI Figure S3). Discrepancies occurred in samples with FAVN titres
SRIG predicted both seropositivity and RVNA titre more accurately ranging from 0.07 to 0.29 IU/ml but were not observed for samples
than models fit to both the 1:10 and 1:25 dilutions. The reduced with higher titres (N = 23). In the previous analysis of these samples,
performance of the two-dilution model reflected higher variability sera that did not reach the 0.5 IU/ml threshold were considered neg-
in the 1:25 dilution compared to the 1:10 dilution, as evidenced by ative and were not re-tested (Wright et al., 2009). We therefore can-
greater IQR values (SI, Figure S2). Ultimately, this variability likely not disregard the possibility of inaccurate titre assessment through
reflects both higher stochasticity in infected cell counts at lower FAVN, considering low-titre detection is even more challenging
serum concentrations and pipetting error. Regardless, the ability to through classical approaches and that titre prediction through FAVN
detect low titres (<0.2 IU/ml) with just one dilution and without the tests can vary even for higher titres (i.e., samples scoring 0.5 IU/ml
need to conduct separate tests for screening and titration is advan- can range from 0.38 to 0.66 IU/ml when repeated) (Liu et al., 2012;
tageous since manual effort and materials are reduced. Wright et al., 2009). Although we did not carry out a formal validation
|
MEZA et al. 1573
for the bat samples, the estimated seroprevalence was 0.21 (CI: 0.09, funding was provided by the National Science Foundation (grant:
0.35), consistent with several previous studies using the RFFIT (de DEB-1020966). DKM was funded by the Human Frontier Science
Thoisy et al., 2016; Steece & Altenbach, 1989; Streicker et al., 2012, Program (grant: RGP0013/2018) and the Mexican National Council
2013). Nevertheless, the ability of our approach to estimate RVNA for Science and Technology (CONACYT, 334795/472296). AB was
titres for all samples tested without additional laboratory work pro- supported by the Leverhulme Trust (grant RPG-2015- 259). We
vides an additional layer of information. The titre continuous levels thank Carlos Tello for sampling coordination. We thank Carlos
can be further used to gauge confidence in positive or negative pre- Shiva, Nestor Falcon, William Valderrama, Sergio Recuenco and
dictions. For example, predicted titres abutting the threshold for se- John Claxton for assistance with sample storage and permits. We
ropositivity might be viewed with caution. thank Pablo Murcia, Megan Griffiths, Nardus Mollentze and Laura
Selecting a different threshold (i.e., 0.166 from the ROC curve) Bergner for feedback on experimental design and earlier versions of
would have increased the assessment values of the pmRFFIT (e.g., this manuscript. SERFOR authorized the collection and exportation
repeatability in bats) but would have predicted low-titre sera as of samples (0142-2015 SERFOR/DGGSPFFS, №003327-SERFOR).
negative. Detection of low RVNA titres is important for epidemio-
logical studies of rabies exposure (Gold et al., 2020). For example, ETHICS
bats sometimes produce low levels of RVNA after viral exposures, The authors confirm that the ethical policies of the journal, as noted
and the antibody response can wane to undetectable levels within on the journal’s author guidelines page, have been adhered to, and
months after the first exposure (Jackson et al., 2008; Turmelle the appropriate ethical review committee approval has been re-
et al., 2010). Indeed, the predicted titres among our predicted se- ceived. Bat capture and sampling methods were approved by the
ropositive bat samples were consistently less than < 0.12 IU/ml. Research Ethics Committee of the University of Glasgow School of
For this reason, we designed our test to detect low RVNA titres, Medical Veterinary and Life Sciences (Ref 08a/15).
with seropositivity defined as > 0.1 IU/ml. The ROC analysis for
the dog sera showed this was close to the optimal range of thresh- C O N FL I C T S O F I N T E R E S T
olds for sensitivity and specificity. Titres > 0.2 IU/ml, as might be The authors declare no conflict of interest.
required for studies concerned with protective levels following
immunization, cannot currently be estimated accurately. Using DATA AVA I L A B I L I T Y S TAT E M E N T
pmRFFIT, such samples would be classified as seropositive with The data that support the findings of this study will be openly availa-
a predicted titre of 0.2 IU/ml. If desired, studies aiming to pre- ble in Zenodo. A doi has been reserved for scripts and data at http://
dict RVNA titres higher than the focus of this study would require doi.org/10.5281/zenodo.3765714.
higher additional serum dilutions in conjunction with a greater
range of SRIG concentrations. ORCID
In summary, the pmRFFIT quantifies RVNA titres using a single, Diana K. Meza https://orcid.org/0000-0001-9796-6706
low-biocontainment test that requires only a single dilution of test Alice Broos https://orcid.org/0000-0001-7593-1000
sera. We recommend first employing the binomial model at a pre-de- Daniel J. Becker https://orcid.org/0000-0003-4315-8628
termined threshold for seropositivity and then using the log-normal Brian J. Willett https://orcid.org/0000-0001-8912-3266
model to predict titres for the putatively positive samples (considered Mafalda Viana https://orcid.org/0000-0001-5975-6505
positive by the binomial model). If benchmark data are available, it Daniel G. Streicker https://orcid.org/0000-0001-7475-2705
would be possible to select the most sensitive and specific threshold
after predicting titres for all available samples using the ROC curve REFERENCES
analysis. Downstream analyses could use binomial and continuous Araujo, D. B., Martorelli, L. A., Kataoka, A. P. G. A., Campos, A. C. A.,
data as needed for specific project objectives. The pmRFFIT could be Rodrigues, C. S., Sanfilippo, L. F., Cunha, E. S., Durigon, E. L., &
Favoretto, S. R. (2014). Antibodies to rabies virus in terrestrial
extended to other lyssaviruses for similarly low-biocontainment test-
wild mammals in native rainforest on the north coast of São Paulo
ing of neutralizing antibodies. The ability to perform the test using the state, Brazil. Journal of Wildlife Diseases, 50(3), 469–477. https://doi.
low-volume sera that are often available for studies of wild bats or org/10.7589/2013-04-099
longitudinal studies of captive bats is a particularly useful feature of Aubert, M. (1992). Practical significance of rabies antibodies in cats and
dogs. Revue Scientifique et Technique de l'OIE, 11(3), 760. https://doi.
our approach. Such studies are increasingly desirable to investigate
org/10.20506/rst.11.3.622
the transmission dynamics of poorly understood bat viruses. Bartoń, K. (2020). MuMIn: Multi-model inference. R Package Version
1.43.17. https://CRAN.R-projec t.org/package=MuMIn
Barton, L. D., & Campbell, J. B. (1988). Measurement of rabies-spe-
cific antibodies in carnivores by an enzyme-linked immunosor-
AC K N OW L E D G M E N T S
bent assay. Journal of Wildlife Diseases, 24(2), 246–258. https://doi.
org/10.7589/0090-3558-24.2.246
This work was funded by a Sir Henry Dale Fellowship jointly funded Bates, D., Maechler, M., Bolker, B. M., Walker, S. C., & Mächler, M. (2015).
by the Wellcome Trust and Royal Society (102507/Z/13/Z) and a Fitting linear mixed-effects models using lme4. Journal of Statistical
Wellcome Senior Research Fellowship (217221/Z/19/Z). Additional Software, 67(1), 1–48. https://doi.org/10.18637/jss.v067.i01.
|
1574 MEZA et al.
Becker, D. J., Crowley, D. E., Washburne, A. D., & Plowright, R. K. (2019). De Benedictis, P., Mancin, M., Cattoli, G., Capua, I., & Terrregino, C.
Temporal and spatial limitations in global surveillance for bat filo- (2012). Serological methods used for rabies post vaccination surveys:
viruses and henipaviruses. Biology Letters, 15(12), 1–7. https://doi. An analysis. Vaccine, 123(7–8), 278–285. https://doi.org/10.1016/j.
org/10.1098/rsbl.2019.0423 vaccine.2012.06.089
Bentley, E. M., Mather, S. T., & Temperton, N. J. (2015). The use of Marcelo Azevedo de Paula Antunes, J., de Castro Demoner, L., Morosini
pseudotypes to study viruses, virus sero-epidemiology and vacci- de Andrade Cruvinel, T., Paula Kataoka, A., Fátima Alves Martorelli,
nation. Vaccine, 33(26), 2955–2962. https://doi.org/10.1016/j.vacci L., Puglia Machado, G., & Megid, J. (2017). Rabies virus exposure of
ne.2015.04.071 Brazilian free-ranging wildlife from municipalities without clinical
Blackwood, J. C., Streicker, D. G., Altizer, S. M., & Rohani, P. (2013). cases in humans or in terrestrial wildlife. Journal of Wildlife Diseases,
Resolving the roles of immunity, pathogenesis and immigration for ra- 53(3), 662–666. https://doi.org/10.7589/2016-09-204
bies persistence in vampire bats. Proceedings of the National Academy de Thoisy, B., Bourhy, H., Delaval, M., Pontier, D., Dacheux, L., Darcissac,
of Sciences, 110(51), 20837–20842. https://doi.org/10.1073/ E., Donato, D., Guidez, A., Larrous, F., Lavenir, R., Salmier, A., Lacoste,
pnas.1308817110 V., & Lavergne, A. (2016). Bioecological drivers of rabies virus circula-
Bock, M., Bishop, K. N., Towers, G., & Stoye, J. P. (2000). Use of a tran- tion in a neotropical bat community. PLoS Neglected Tropical Diseases,
sient assay for studying the genetic determinants of Fv1 restriction. 10(1), 1–21. https://doi.org/10.1371/journal.pntd.0004378
Journal of Virology, 74(16), 7422–7430. https://doi.org/10.1128/ Eschbaumer, M., Law, S., Solis, C., Chernick, A., van der Meer, F., & Czub,
JVI.74.16.7422-7430.2000 M. (2014). Rapid detection of neutralizing antibodies against bo-
Borremans, B., Hens, N., Beutels, P., Leirs, H., & Reijniers, J. (2016). vine viral diarrhoea virus using quantitative high-content screening.
Estimating time of infection using prior serological and individual in- Journal of Virological Methods, 198, 56–63. https://doi.org/10.1016/j.
formation can greatly improve incidence estimation of human and jviromet.2013.12.017
wildlife infections. PLOS Computational Biology, 12(5), e1004882. Freedman, D., & Diaconis, P. (1981). On the histogram as a density es-
https://doi.org/10.1371/journal.pcbi.1004882 timator:L2 theory. Zeitschrift Für Wahrscheinlichkeitstheorie Und
Briggs, D. J., Smith, J. S., Mueller, F. L., Schwenke, J., Davis, R. D., Gordon, Verwandte Gebiete, 57(4), 453–476. https://doi.org/10.1007/BF010
C. R., Schweitzer, K., Orciari, L. A., Yager, P. A., & Rupprecht, C. E. 25868
(1998). A comparison of two serological methods for detecting the Gamble, A., Garnier, R., Chambert, T., Gimenez, O., & Boulinier, T. (2020).
immune response after rabies vaccination in dogs and cats being ex- Next-generation serology: Integrating cross-sectional and capture–
ported to rabies-free areas. Biologicals, 26(4), 347–355. https://doi. recapture approaches to infer disease dynamics. Ecology, 101(2),
org/10.1006/biol.1998.0162 https://doi.org/10.1002/ecy.2923
Campos, A. A. S., Santos, R. N., Benavides, J. A., Carvalho Ruthner Batista, George, D. B., Webb, C. T., Farnsworth, M. L., O'Shea, T. J., Bowen, R.
H. B., Finoketti, F., Wagner, P. G. C., Zafalon-Silva, B., Alievi, M., Silva, A., Smith, D. L., Stanley, T. R., Ellison, L. E., & Rupprecht, C. E. (2011).
F. B., Witt, A., Tartarotti, A., Cassia Rodrigues da Silva, A., Ferreira, Host and viral ecology determine bat rabies seasonality and main-
K. C. S., Frazzon, A. P. G., Roehe, P. M., & Franco, A. C. (2019). Rabies tenance. Proceedings of the National Academy of Sciences, 108(25),
surveillance in wild mammals in South Of Brazil. Transboundary and 10208–10213. https://doi.org/10.1073/pnas.1010875108
Emerging Diseases, October, 1–8. https://doi.org/10.1111/tbed.13415 Gilbert, A. T., Fooks, A. R., Hayman, D. T. S., Horton, D. L., Müller,
Cliquet, F., Aubert, M., & Sagné, L. (1998). Development of a fluorescent T., Plowright, R., Peel, A. J., Bowen, R., Wood, J. L. N., Mills, J.,
antibody virus neutralisation test (FAVN test) for the quantitation Cunningham, A. A., & Rupprecht, C. E. (2013). Deciphering serology to
of rabies-neutralising antibody. Journal of Immunological Methods, understand the ecology of infectious diseases in wildlife. EcoHealth,
212(1), 79–87. https://doi.org/10.1016/S0022-1759(97)00212-3 10(3), 298–313. https://doi.org/10.1007/s10393-013-0856-0
Cliquet, F., McElhinney, L. M., Servat, A., Boucher, J. M., Lowings, J. P., Gold, S., Donnelly, C. A., Nouvellet, P., & Woodroffe, R. (2020). Rabies
Goddard, T., Mansfield, K. L., & Fooks, A. R. (2004). Development virus-neutralising antibodies in healthy, unvaccinated individu-
of a qualitative indirect ELISA for the measurement of rabies vi- als: What do they mean for rabies epidemiology? PLOS Neglected
rus-specific antibodies from vaccinated dogs and cats. Journal of Tropical Diseases, 14(2), e0007933. https://doi.org/10.1371/journ
Virological Methods, 117(1), 1–8. https://doi.org/10.1016/j.jviro al.pntd.0007933
met.2003.12.001 Goodin, D. G., Jonsson, C. B., Allen, L. J. S., & Owen, R. D. (2018).
Condit, R. C. (2001). Principles of virology. In D. M. Knipe & P. M. Howley Integrating landscape hierarchies in the discovery and model-
(Eds.), Fields Virology 4th edn. (Vol. 1, 19–51). Philadelphia: Lippincott ing of ecological drivers of zoonotically transmitted disease from
Williams & Wilkins. wildlife. In The Connections Between Ecology and Infectious Disease,
Constantine, D. G., Tierkel, E. S., Kleckner, M. D., & Hawkins, D. M. (1968). Advances in Environmental Microbiology, 5, 299–317. Tx. https://doi.
Rabies in New Mexico cavern bats. Public Health Reports (Washington, org/10.1007/978-3-319-92373- 4_9.
DC: 1896), 83(4), 303–316. https://www.ncbi.nlm.nih.gov/pmc/artic Gurley, E. S., Hegde, S. T., Hossain, K., Sazzad, H. M. S., Hossain, M.
les/PMC1891046/pdf/pubhealthreporig 00004-0041.pdf. J., Rahman, M., Sharker, M. A. Y., Salje, H., Islam, M. S., Epstein, J.
Costa, L. J. C., Andrade, F. A. G., Uieda, W., Martorelli, L. F. A., Kataoka, H., Khan, S. U., Kilpatrick, A. M., Daszak, P., & Luby, S. P. (2017).
A. P. A. G., & Fernandes, M. E. B. (2013). Serological investigation Convergence of humans, bats, trees, and culture in Nipah virus trans-
of rabies virus neutralizing antibodies in bats captured in the east- mission, Bangladesh. Emerging Infectious Diseases, 23(9), 1446–1453.
ern Brazilian Amazon. Transactions of the Royal Society of Tropical https://doi.org/10.3201/eid2309.161922
Medicine and Hygiene, 107(11), 684–689. https://doi.org/10.1093/ Hammami, S., Schumacher, C., Cliquet, F., Tlatli, A., Aubert, A., & Aubert,
trstmh/trt080 M. (1999). Vaccination of Tunisian dogs with the lyophilised SAG2
Cunningham, A. A., Daszak, P., & Wood, J. L. N. (2017). One Health, oral rabies vaccine incorporated into the DBL2 dog bait. Veterinary
emerging infectious diseases and wildlife: Two decades of progress? Research, 30(6), 607–613. https://doi.org/hal-00902600.
Philosophical Transactions of the Royal Society B: Biological Sciences, Hierholzer, J. C., & Killington, R. A. (1996). Virus isolation and quan-
372(1725), 20160167. https://doi.org/10.1098/rstb.2016.0167 titation. In Brian W. J. Mahy & Hillar O. Kangro ((Eds.), Virology
Daszak, P., Cunningham, A. A., & Hyatt, A. D. (2000). Emerging in- Methods Manual (pp. 25–46). London: Academic Press. https://doi.
fectious diseases of wildlife - threats to biodiversity and human org/10.1016/B978-012465330-6/50003-8
health. Science, 287(5459), 443–449. https://doi.org/10.1126/scien Horton, D. L., Breed, A. C., Arnold, M. E., Smith, G. C., Aegerter, J. N.,
ce.287.5452.443 McElhinney, L. M., Johnson, N., Banyard, A. C., Raynor, R., Mackie,
|
MEZA et al. 1575
I., Denwood, M. J., Mellor, D. J., Swift, S., Racey, P. A., & Fooks, A. Neglected Tropical Diseases, 4(3), e595. https://doi.org/10.1371/journ
R. (2020). Between roost contact is essential for maintenance of al.pntd.0000595
European bat lyssavirus type-2 in Myotis daubentonii bat reservoir: Neumann, E. J., & Bonistalli, K. N. (2009). Effect of blood sample han-
‘The Swarming Hypothesis’. Scientific Reports, 10(1), 1740. https:// dling post-collection on Erysipelothrix rhusiopathiae antibody ti-
doi.org/10.1038/s41598-020-58521-6 tres. Veterinary Journal, 180(3), 325–329. https://doi.org/10.1016/j.
Irie, T., & Kawai, A. (2002). Studies on the different conditions for ra- tvjl.2008.07.020
bies virus neutralization by monoclonal antibodies #1–46-12 and Obregón-Morales, C., Aguilar-Setién, Á., Perea Martínez, L., Galvez-
#7–1-9. Journal of General Virology, 83(12), 3045–3053. https://doi. Romero, G., Martínez-Martínez, F. O., & Aréchiga-Ceballos, N. (2017).
org/10.1099/0022-1317-83-12-3045 Experimental infection of Artibeus intermedius with a vampire bat
Jackson, F. R., Turmelle, A. S., Farino, D. M., Franka, R., McCracken, G. rabies virus. Comparative Immunology, Microbiology and Infectious
F., & Rupprecht, C. E. (2008). Experimental rabies virus infection of Diseases, 52, 43–47. https://doi.org/10.1016/J.CIMID.2017.05.008
big brown bats (eptesicus fuscus). Journal of Wildlife Diseases, 44(3), Péharpré, D., Cliquet, F., Sagné, E., Renders, C., Costy, F., & Aubert, M.
612–621. https://doi.org/10.7589/0090-3558-44.3.612 (1999). Comparison of visual microscopic and computer-automated
Jorge, R. S. P., Pereira, M. S., Morato, R. G., Scheffer, K. C., Carnieli, P., fluorescence detection of rabies virus neutralizing antibodies.
Ferreira, F., Furtado, M. M., Kashivakura, C. K., Silveira, L., Jacomo, Journal of Veterinary Diagnostic Investigation, 11(4), 330–333. https://
A. T. A., Lima, E. S., de Paula, R. C., & May-Junior, J. A. (2010). doi.org/10.1177/104063879901100406
Detection of rabies virus antibodies in brazilian free-ranging wild Plowright, R. K., Becker, D. J., McCallum, H., & Manlove, K. R. (2019).
carnivores. Journal of Wildlife Diseases, 46(4), 1310–1315. https://doi. Sampling to elucidate the dynamics of infections in reservoir hosts.
org/10.7589/0090-3558-46.4.1310 Philosophical Transactions of the Royal Society B: Biological Sciences,
Kerins, J., Koske, S., Kazmierczak, J., Austin, C., Gowdy, K., & Dibernardo, 374(1782), 20180336. https://doi.org/10.1098/rstb.2018.0336
A. (2018). Outbreak of Seoul virus among rats and rat owners-United Plowright, R. K., Peel, A. J., Streicker, D. G., Gilbert, A. T., McCallum,
States and Canada, 2017. Canada Communicable Disease Report, H., Wood, J., Baker, M. L., & Restif, O. (2016). Transmission or with-
44(2), 71–74. https://doi.org/10.14745/ccdr.v44i02a07 in-host dynamics driving pulses of zoonotic viruses in reservoir-host
Kostense, S., Moore, S., Companjen, A., Bakker, A. B. H., Marissen, W. populations. PLOS Neglected Tropical Diseases, 10(8), e0004796.
E., von Eyben, R., Weverling, G. J., Hanlon, C., & Goudsmit, J. (2012). https://doi.org/10.1371/journal.pntd.0004796
Validation of the rapid fluorescent focus inhibition test for rabies vi- Python Core Team. (2019). Python: A dynamic, open source programming
rus-neutralizing antibodies in clinical Samples. Antimicrobial Agents language. https://www.python.org/.
and Chemotherapy, 56(7), 3524–3530. https://doi.org/10.1128/ R Core Team. (2020). R: A language and environment for statistical comput-
AAC.06179-11 ing. Vienna, Austria: R Foundation for Statistical Computing. https://
Kuzmin, I. V., Niezgoda, M., Franka, R., Agwanda, B., Markotter, W., www.r-projec t.org/
Beagley, J. C., Urazova, O. Y., Breiman, R. F., & Rupprecht, C. E. Rasband, W. S. (2018). ImageJ. National Institutes of Health. https://im-
(2008). Lagos bat virus in Kenya. Journal of Clinical Microbiology, agej.nih.gov/ij/.
46(4), 1451–1461. https://doi.org/10.1128/JCM.00016- 08 Reynes, J.-M., Molia, S., Audry, L., Hout, S., Ngin, S., Walston, J., &
Liu, Y., Zhang, S., Zhang, F., & Hu, R. (2012). A semi-quantitative sero- Bourhy, H. (2004). Serologic evidence of lyssavirus infection in bats,
logical method to assess the potency of inactivated rabies vaccine Cambodia. Emerging Infectious Diseases, 10(12), 2231–2234. https://
for veterinary use. Virologica Sinica, 27(4), 259–264. https://doi. doi.org/10.3201/eid1012.040459
org/10.1007/s12250 -012-3260-y Rupprecht, C. E., Kuzmin, I. V., & Meslin, F. (2017). Lyssaviruses and ra-
Ma, X., Monroe, B. P., Cleaton, J. M., Orciari, L. A., Gigante, C. M., Kirby, bies: Current conundrums, concerns, contradictions and controver-
J. D., & Wallace, R. M. (2020). Rabies surveillance in the United States sies. F1000Research, 6, 184. https://doi.org/10.12688/f1000resea
during 2018. Journal of the American Veterinary Medical Association, rch.10416.1.
256(2), 195–208. https://doi.org/10.2460/javma.256.2.195 Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V., Longair, M.,
Ma, X., Niezgoda, M., Blanton, J. D., Recuenco, S. E., & Rupprecht, C. E. Pietzsch, T., Preibisch, S., Rueden, C., Saalfeld, S., Schmid, B.,
(2012). Evaluation of a new serological technique for detecting ra- Tinevez, J.-Y., White, D. J., Hartenstein, V., Eliceiri, K., Tomancak, P.,
bies virus antibodies following vaccination. Vaccine, 30, 5358–5362. & Cardona, A. (2012, July 28). Fiji: An open-source platform for bi-
https://doi.org/10.1016/j.vaccine.2012.06.037 ological-image analysis. Nature Methods, 9(7), 676–682. https://doi.
Mather, S. T., Scott, S. D., Temperton, N. J., Wright, E., King, B., & Daly, org/10.1038/nmeth.2019
J. (2013). Current progress with serological assays for exotic emerg- Schneider, M. C., Romijn, P. C., Uieda, W., Tamayo, H., Silva, D. F. D.,
ing/re-emerging viruses. Future Virology, 8(8), 745–755. https://doi. Belotto, A., Silva, J. B. D., & Leanes, L. F. (2009). Rabies transmitted
org/10.2217/fvl.13.60 by vampire bats to humans: An emerging zoonotic disease in Latin
Mena, I., Nelson, M. I., Quezada-Monroy, F., Dutta, J., Cortes-Fernández, America? Revista Panamericana De Salud Pública, 25(3), 260–269.
R., Lara-Puente, J. H., Castro-Peralta, F., Cunha, L. F., Trovão, N. S., https://doi.org/10.1590/S1020- 498920 09000300010
Lozano-Dubernard, B., Rambaut, A., van Bakel, H., & García-Sastre, Smith, J. S., Yager, P. A., & Baer, G. M. (1973). A rapid reproducible test for
A. (2016). Origins of the 2009 H1N1 influenza pandemic in swine in determining rabies neutralizing antibody. Bulletin of the World Health
Mexico. Elife, 5(JUN2016). https://doi.org/10.7554/eLife.16777 Organization, 48(5), 535–541. https://www.ncbi.nlm.nih.gov/pmc/
Metcalf, C. J. E., Farrar, J., Cutts, F. T., Basta, N. E., Graham, A. L., Lessler, articles/PMC2482941/pdf/bullwho00178-0027.pdf.
J., Ferguson, N. M., Burke, D. S., & Grenfell, B. T. (2016). Use of se- Steece, R., & Altenbach, J. S. (1989). Prevalence of rabies specific anti-
rological surveys to generate key insights into the changing global bodies in the mexican free-tailed bat (tadarida brasiliensis mexicana)
landscape of infectious disease. The Lancet, 388(10045), 728–730. at lava cave, New Mexico. Journal of Wildlife Diseases, 25(4), 490–
https://doi.org/10.1016/S0140-6736(16)30164-7 496. https://doi.org/10.7589/0090-3558-25.4.490
Moeschler, S., Locher, S., Conzelmann, K.-K., Krämer, B., & Zimmer, G. Streicker, D. G., Franka, R., Jackson, F. R., & Rupprecht, C. E. (2013).
(2016). Quantification of lyssavirus-neutralizing antibodies using Anthropogenic roost switching and rabies virus dynamics in
vesicular stomatitis virus pseudotype particles. Viruses, 8(9), 254. house-roosting big brown bats. Vector-Borne and Zoonotic Diseases,
https://doi.org/10.3390/v8090254 13(7), 498–504. https://doi.org/10.1089/vbz.2012.1113
Moore, S. M., & Hanlon, C. A. (2010). Rabies-specific antibodies: Streicker, D. G., Recuenco, S., Valderrama, W., Gomez Benavides, J.,
Measuring surrogates of protection against a fatal disease. PLoS Vargas, I., Pacheco, V., Condori Condori, R. E., Montgomery, J.,
|
1576 MEZA et al.
Rupprecht, C. E., Rohani, P., & Altizer, S. (2012). Ecological and an- against emerging lyssaviruses. Virology, 408(2), 183–189. https://doi.
thropogenic drivers of rabies exposure in vampire bats: Implications org/10.1016/j.virol.2010.09.014
for transmission and control. Proceedings of the Royal Society B: Wright, E., McNabb, S., Goddard, T., Horton, D. L., Lembo, T., Nel, L. H.,
Biological Sciences, 279(1742), 3384–3392. https://doi.org/10.1098/ Weiss, R. A., Cleaveland, S., & Fooks, A. R. (2009). A robust lentiviral
rspb.2012.0538 pseudotype neutralisation assay for in-field serosurveillance of ra-
Temperton, N. J., Wright, E., & Scott, S. D. (2015). Retroviral pseudo- bies and lyssaviruses in Africa. Vaccine, 27, 7178–7186. https://doi.
types - From scientific tools to clinical utility. In eLS (pp. 1–11). : John org/10.1016/j.vaccine.2009.09.024
Wiley & Sons, Ltd. https://doi.org/10.1002/9780470015902.a0021 Wright, E., Temperton, N. J., Marston, D. A., McElhinney, L. M., Fooks,
549.pub2. A. R., & Weiss, R. A. (2008). Investigating antibody neutralization
Timiryasova, T. M., Luo, P., Zheng, L., Singer, A., Zedar, R., Garg, S., Petit, of lyssaviruses using lentiviral pseudotypes: A cross-species com-
C., Moore, S., Hu, B. T., & Brown, M. (2019). Rapid fluorescent focus parison. Journal of General Virology, 89(9), 2204–2213. https://doi.
inhibition test optimization and validation: Improved detection org/10.1099/vir.0.2008/000349-0
of neutralizing antibodies to rabies virus. Journal of Immunological Xue, X., Zheng, X., Liang, H., Feng, N. A., Zhao, Y., Gao, Y., Wang, H.,
Methods, 474, 112626. https://doi.org/10.1016/j.jim.2019.06.017 Yang, S., & Xia, X. (2014). Generation of recombinant rabies virus
Towers, G., Bock, M., Martin, S., Takeuchi, Y., Stoye, J. P., & Danos, O. CVS-11 expressing eGFP applied to the rapid virus neutralization
(2000). A conserved mechanism of retrovirus restriction in mam- test. Viruses, 6(4), 1578–1589. https://doi.org/10.3390/v6041578
mals. Proceedings of the National Academy of Sciences, 97(22), 12295– Zhou, P., Yang, X.-L., Wang, X.-G., Hu, B., Zhang, L., Zhang, W., & Shi,
12299. https://doi.org/10.1073/pnas.200286297 Z.-L. (2020). A pneumonia outbreak associated with a new coronavi-
Turmelle, A. S., Jackson, F. R., Green, D., McCracken, G. F., & Rupprecht, rus of probable bat origin. Nature, 579(7798), 270–273. https://doi.
C. E. (2010). Host immunity to repeated rabies virus infection in big org/10.1038/s41586-020-2012-7
brown bats. Journal of General Virology, 91(9), 2360–2366. https:// Zuur, A. F., Ieno, E. N., Walker, N. J., Saveliev, A. A., & Smith, G. M. (2009).
doi.org/10.1099/vir.0.020073-0 Mixed effects modelling for nested data. In A. F. Zuur E. N. Ieno N. J.
Wasniewski, M., & Cliquet, F. (2012). Evaluation of ELISA for de- Walker A. A. Saveliev & G. M. Smith (Eds.), Mixed effects models and
tection of rabies antibodies in domestic carnivores. Journal of extensions in ecology with R (pp. 101–142). New York, NY: Springer.
Virological Methods, 179(1), 166–175. https://doi.org/10.1016/j.jviro https://doi.org/10.1007/978-0-387-87458-6_5
met.2011.10.019
Welch, R. J., Anderson, B. L., & Litwin, C. M. (2009). An evaluation of
two commercially available ELISAs and one in-house reference lab-
S U P P O R T I N G I N FO R M AT I O N
oratory ELISA for the determination of human anti-rabies virus anti-
bodies. Journal of Medical Microbiology, 58(6), 806–810. https://doi.
Additional supporting information may be found online in the
org/10.1099/jmm.0.006064-0 Supporting Information section.
WHO. (2017). Human rabies: 2016 updates and call for data. Weekly ep-
idemiological record, 92(7), 77–88. https://www.who.int/wer/2017/
wer9207/en/ How to cite this article: Meza DK, Broos A, Becker DJ, et al.
WHO. (2018). WHO expert consultation on rabies, third report, WHO
Predicting the presence and titre of rabies virus-neutralizing
Technical Report Series, No. 1012. (pp. 1–183). Geneva: World
Health Organization.. World Health Organization technical report antibodies from low-volume serum samples in low-
series. Retrieved from http://www.who.int/bookorders containment facilities. Transbound. Emerg. Dis.
Wright, E., Hayman, D. T. S., Vaughan, A., Temperton, N. J., Wood, J. L. 2021;68:1564–1576. https://doi.org/10.1111/tbed.13826
N., Cunningham, A. A., Suu-Ire, R., Weiss, R. A., & Fooks, A. R. (2010).
Virus neutralising activity of African fruit bat (Eidolon helvum) sera

Transbounding Emerging Dis - 2020 - Meza - Predicting The Presence and Titre of Rabies Virus Neutralizing Antibodies From

Uploaded by

Copyright:

Available Formats

Transbounding Emerging Dis - 2020 - Meza - Predicting The Presence and Titre of Rabies Virus Neutralizing Antibodies From

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Transbounding Emerging Dis - 2020 - Meza - Predicting The Presence and Titre of Rabies Virus Neutralizing Antibodies From

Uploaded by

Copyright:

Available Formats

Received: 1 May 2020 | Revised: 3 August 2020 | Accepted: 3 September 2020

Predicting the presence and titre of rabies virus-neutralizing

Diana K. Meza1,2 | Alice Broos1,2 | Daniel J. Becker3 | Abdelkader Behdenna1 |

biologic assay, Desmodus rotundus, generalized linear mixed models, immunofluorescence

1 | I NTRO D U C TI O N Obregón-Morales et al., 2017; Steece & Altenbach, 1989). As

After incubation, slides were photographed at 4× magnification

F I G U R E 2 Raw counts of fluorescent

of samples in agreement with FAVN; sensitivity = 78.79%; spec- 4 | D I S CU S S I O N

You might also like