Background
The Asian tiger mosquito, Aedes (Stegomyia) albopictus, is an invasive species that
has expanded its territory to over 40% of the earth’s terrestrial landmass in the
last 30 years [1]. Ae. albopictus is an efficient vector of all serotypes of dengue,
a disease that has increased in frequency over the past 30 years in the Americas [2],
where it represents an annual cost of 2,100,000,000 USD per year [3]. This mosquito
is also an efficient vector of the three genotypes of Chikungunya virus, a worldwide
emerging pathogen that causes fever, fatigue, and joint swelling in humans. Since
2006, Chikungunya outbreaks have been increasingly recorded outside the virus’s native
range in tropical Africa, perhaps because of a mutation in the virus’s envelope gene,
which increases the replication and dissemination capacity of the virus in Ae. albopictus
[4]. During the second quarter of 2014, Chikungunya has been detected throughout much
of the Americas, with major outbreaks occurring in several Caribbean nations, and
local transmission confirmed or suspected in the United States, Panama, Venezuela,
Peru, and Chile, creating an imminent threat for humans throughout the Americas, who
have no prior exposure to this infection [5].
The first cases of Chikungunya disease in Panama were reported in May 2014, occurring
in nonresidents who most likely picked up the virus in their Caribbean countries of
origin. On 23 July 2014, Panama’s health authority reported autochthonous transmission
of Chikungunya virus. Coincidentally, the earliest cases involved patients located
in Juan Diaz, an urban area on the eastern outskirts of Panama City, where the first
specimen of invasive Ae. albopictus was collected in 2002. Ae. albopictus has expanded
across much of Panama since that time, yet to date, no information exists about the
degree of expansion or about the factors contributing to the geographic expansion
of this important mosquito vector across Panama. Here, we map the temporal expansion
of Ae. albopictus, use species distribution models to determine the ecological and
nonecological factors associated with its expansion, and comment on the implications
for vector and disease control programs in Panama and elsewhere in the American tropics.
Tempo and Mode of Ae. albopictus Expansion in Panama
Panama’s Ministry of Health (MINSA) maintains a nationwide surveillance program for
Aedes mosquitoes (S1 Methods) and provided us with geographic coordinates and dates
for confirmed samples of Ae. albopictus collected between 2002 and 2013, which were
supplemented with Jose Loaiza’s surveys of mosquitoes across Panama. Mosquito occurrence
data were placed into three temporal pools: 2002–2005, 2006–2009, and 2010–2013. Between
2002 and 2005, Ae. albopictus was found only in the eastern portion of Panama City
(Fig. 1A). Between 2006 and 2009, mosquito density increased in Panama City and also
expanded to Colón, central Panama’s Caribbean port (Fig. 1B). Between 2010 and 2013,
Ae. albopictus expanded both eastward from Panama City and also into western Panama
between the Costa Rican border and Santiago, Veraguas (Fig. 1C). Although Ae. albopictus
appears to have expanded westward from Panama City along the Pan-American highway,
the lack of confirmed samples from the Azuero Peninsula east to Panama City’s western
edges raise the possibility that the 2010–2013 distribution of Ae. albopictus in western
Panama was the result of a separate colonization event from Costa Rica, as the species
has occurred in several locations in that country since at least 2009 [6].
10.1371/journal.pntd.0003383.g001
Figure 1
Occurrence points for Ae. albopictus for three time periods during its recent expansion
across the Republic of Panama: A) between 2002 and 2006, Ae. albopictus was found
only in the eastern metropolitan area of Panama City; B) during 2006 and 2009, Ae.
albopictus expanded to the Colón on the Caribbean coast; and C) between 2010 and 2013,
the species was found throughout much of western Panama as well as east of Panama
City.
Darker blue colors indicate political districts with higher human population densities.
Ae. albopictus apparently has not yet spread to the Bocas del Toro province in northwestern
Panama or the Azuero Peninsula, which includes the city of Chitré, nor to much of
the lightly inhabited Darién province.
Road Networks Alone Best Explain the Geographic Expansion of Ae. albopictus across
Panama
We created competing species distribution models (SDMs) via maximum entropy machine
learning algorithms using the Maxent software package (version 3.3) [7] to evaluate
the factors associated with Ae. albopictus expansion. SDMs predict the suitability
(i.e., probability of species occurrence, range: 0–1) of map cells based on the distribution
of known occurrence points and the environmental conditions of map cells. For environmental
conditions, we used all 19 WorldClim climate layers [8] as well as geographic information
system (GIS) layers of principal roads and population density [9], which we rasterized
and scaled to 2.5 arc minutes. SDMs were created using the 2006–2009 mosquito occurrence
data, and model fit was evaluated by comparing the later (2010–2013) occurrence data
against the model-predicted suitability of those points. We created seven SDMs including
“only climate,” “only human density,” and “only roads,” as well as all possible combinations
of those three datasets. All SDMs were generated based on ten replicates using cross
validation and 10,000 background points, and we set our threshold for habitat suitability
at 10% of all observed occurrences. We used the 2010–2013 sample data to compare the
fit of each SDM in two ways: first, we calculated the mean modeled suitability of
all 2010–2013 occurrence points, and second, we calculated the percentage of those
occurrence points having a predicted suitability above the 10% minimum suitability
threshold.
An SDM based only on the road network best predicted the 2010–2013 distribution of
Ae. albopictus in Panama, compared to SDMs based on climate or human population density
or even to models that included roads and other factors (Table 1; Fig. 1). The average
suitability of 2010–2013 occurrences based on roads alone was 0.487, compared to average
suitability of other models that ranged from 0.285–0.319. Likewise, 80% of the 2010–2013
samples occurred in areas predicted to be suitable habitats for Ae. albopictus in
the roads-only model, compared to frequencies ranging from 34%–53% in other models.
Interestingly, climate alone was the poorest predictor of suitability, predicting
habitat suitability for only 34% of the points at which Ae. albopictus was actually
sampled between 2010–2013 (Table 1). Our findings appear to be unbiased by mosquito
sampling effort or by the relative intensity of sampling along versus off the principal
road network (S1 Fig.).
10.1371/journal.pntd.0003383.t001
Table 1
Performance of various geographic species distribution models to predict the expansion
of Ae. albopictus in Panama.
Model
Area under the ROC Curve (AUC)
Average Suitability of 2010–2013 Occurrence Points
2006–2009 10% Occurrence Threshold
% of 2010–2013 Occurrence Points above Threshold
Roads Only
0.881
0.487
0.600
80%
Population Density Only
0.957
0.302
0.239
53%
Roads and Population Density
0.975
0.319
0.329
49%
Climate, Roads, and Population Density
0.986
0.291
0.369
39%
Roads and Climate
0.894
0.306
0.353
36%
Climate and Population Density
0.982
0.285
0.383
35%
Climate Only
0.979
0.309
0.438
34%
Models were parameterized using occurrence points sampled between 2006 and 2009. Area
under the curve (AUC) measures the efficiency of the model to discriminate occurrences
from random background points; AUC ranks did not correlate with model predictive performance.
Model performance was evaluated using two criteria based on 2010–2013 occurrence points:
first, by averaging the predicted suitability of all 110 occurrence points, and second,
by calculating the frequency of those occurrence points having a predicted suitability
above the 10% model threshold.
Global versus Local Scales of Ae. albopictus Expansion
In general, our results agree with the global pattern of rapid expansion for Ae. albopictus,
which is mainly attributed to human-aided dispersal [1]. Earlier studies have accurately
predicted the global expansion of Ae. albopictus using climate-based SDMs [10,11].
Our results should not be seen as in conflict with those findings; rather, they demonstrate
the dynamics of Aedes invasions on differing scales of time and space. At global scales,
all of Panama is within the climate threshold for Ae. albopictus [10,11]; therefore,
the immediate geographic spread across Panama is likely to be determined by factors
other than ecology. Likewise, international expansion of Ae. albopictus has occurred
principally via oceanic container vessels and/or international air traffic [1], yet
our results confirm the primacy of road networks for determining patterns of Aedes
expansion and distributional limits at local scales [12,13].
Interactions with Ae. aegypti and Implications for Dengue and Chikungunya Control
Panama’s current urban mosquito control programs focus primarily on Ae. aegypti, yet
both this species and Ae. albopictus are vectors of Chikungunya and dengue viruses
[4,5]. Some evidence from outside the Americas suggests that reducing Ae. aegypti
populations may be less effective at reducing Chikungunya and dengue outbreaks in
Panama if simultaneous efforts to reduce the population of Ae. albopictus are not
undertaken [14,15]. At the same time, these efforts might facilitate the ecological
replacement of Ae. aegypti by Ae. albopictus, which could have both favorable and
unfavorable consequences that are difficult to predict a priori. For example, there
is evidence that Ae. aegypti is a more efficient vector of dengue virus than Ae. albopictus
[16], which may be the result of a greater preference for human bite targets among
Ae. aegypti than among Ae. albopictus [17]. Additionally, the particular strain of
Chikungunya virus currently circulating in the Americas lacks the mutation allowing
for selectively enhanced transmission efficiency in Ae. albopictus [5]. On the other
hand, current vector control programs include the indoor application of insecticide
in urban areas of Panama, taking advantage of the fact that Ae. aegypti tends to rest
inside dwellings rather than in vegetation outside homes [18], while the latter is
the preferred resting habitat of Ae. albopictus [17]. However, Ae. albopictus may
be ecologically more plastic than Ae. aegypti [17], and it is likely only a matter
of time until the mutations favoring Chikungunya transmission in Ae. albopictus migrate
to the Americas.
Our model presents implications for the control of dengue and Chikungunya disease.
The road-only model predicts future expansion of Ae. albopictus into northwestern
and eastern Panama as well as in the Azuero Peninsula, which includes Chitré, Panama’s
third largest urban area (Fig. 2). This presents an immediate opportunity for Panama’s
Ministry of Health to control the expansion of Ae. albopictus. Evidence from Europe
suggests that passive transport of larvae occurs in items in which open water accumulates,
such as used tires, while adults can be passively transported inside the cabin of
cars and trucks [19]. Specifically, we recommend the fumigation of vehicles at transportation
checkpoints (see suggested checkpoints in Fig. 2), which could stop the movement of
adults and immature stages of Ae. albopictus across Panama.
10.1371/journal.pntd.0003383.g002
Figure 2
Geographic model predicting future range expansion of Ae. albopictus in Panama.
This model is based on the best-performing-species distribution model (highway network
model). Blue pixels represent locations predicted to be likely areas of Ae. albopictus
expansion, whereas gray pixels represent areas that had a model suitability that was
below the minimum threshold and therefore were unlikely to harbor mosquitoes. Orange
points represent species occurrences sampled between 2010 and 2013. A series of surveillance
and fumigation chokepoints at strategic locations on the highway network (e.g., points
A, B, C, and D) could limit the continued expansion of Ae. albopictus as a first step
to reduce the epidemiological risk posed by this invasive vector.
Finally, our results present a cautionary tale in the face of proposals to release
genetically modified Ae. aegypti (GM programs); trial GM program releases began in
Panama in May 2014. Given that Ae. aegypti has similar demographic and dispersal patterns
as Ae. albopictus [13], Ae. aegypti populations may quickly rebound via recolonization
after cessation of GM programs. Thus, GM strategies might have only short-term effects
on vector population size and may commit Panama to a repeated and costly program for
long-term arbovirus control [20]. Additionally such programs could increase the chance
that Ae. albopictus displaces Ae. aegypti, making the GM program less relevant. We
encourage health authorities in Panama and elsewhere in tropical America to fully
consider the ecology of Ae. albopictus alongside Ae. aegypti when developing dengue
and Chikungunya disease control programs.
Supporting Information
S1 Fig
A map of 2010–2013 Ae. albopictus occurrences (yellow points) compared to Ministry
of Health (MINSA) occurrence points for Ae. aegypti (gray points).
MINSA surveys exhaustively across Panama for mosquitoes of medical importance, recording
positive species occurrences, but they do not tabulate negative samples. In order
to estimate sampling intensity and the proportion of sampling effort along the principle
road network (gray lines), we plotted Ae. aegypti data that were provided to us by
MINSA for the years 2007–2010. These points serve as a proxy for MINSA sampling effort.
Comparing these points to the 2010–2013 Ae. albopictus occurrences and the road network
demonstrates that MINSA intensively sampled for mosquitoes in areas such as Bocas
del Toro and the eastern Azuero Peninsula where Ae. albopictus was not recorded and
also routinely sampled in areas such as much of eastern Panama where no roads occur.
(PDF)
Click here for additional data file.
S1 Methods
Sampling strategies for adult and immature stages of Aedes mosquitoes and other medically
important mosquito species in Panama.
(DOCX)
Click here for additional data file.