Conflicting Effects of Climate and Vector Behavior on the Spread of a Plant Pathogen

Abstract

Local climatic conditions are important determinants of disease dynamics through effects on vector population performance or distribution. Yet, climate may also be epidemiologically significant due to effects on host-pathogen infection dynamics. We developed a model to explore interactive effects between climate-mediated acceleration in disease phenology (i.e. faster incubation or symptom onset) and vector preference based on host symptom status. Higher incubation rates favored pathogen outbreaks, but more rapid symptom onset may constrain spread if vectors avoid symptomatic hosts. Next, we tested whether warmer conditions favored greater spread of the plant pathogen, Xylella fastidiosa, by its leafhopper vector, Graphocephala atropunctata. Inoculated and healthy plants were reared in two temperature-controlled greenhouses. At 6 times post inoculation a healthy and inoculated plant were exposed to non-infective vectors, after which pathogen spread was evaluated. Incubation rate and symptom onset in infected hosts was significantly accelerated at higher temperature. Although there was a tendency for greater pathogen spread at higher temperature, the effect depended on time since inoculation. In later introductions, after disease symptoms manifest, vectors were more likely to be found on healthy hosts. Vector avoidance of symptoms, particularly for hosts reared at higher temperature, constrained pathogen spread at later introductions. These results indicate that climate and vector behavior may mediate interactively pathogen spread. Further consideration of such epidemiological complexities is needed to predict adequately the consequences of climate change for disease dynamics.

Full text

Phytobiomes •XXXX •XX:X-X http://dx.doi.org/10.1094/PBIOMES-01-17-0004-R
RESEARCH
Conflicting Effects of Climate and Vector Behavior on the Spread
of a Plant Pathogen
Matthew P. Daugherty, Department of Entomology, University of California, Riverside, CA 92521; and Adam R.
Zeilinger and Rodrigo P. P. Almeida, Department of Environmental Science, Policy and Management, University
of California, Berkeley, CA 94720
Accepted for publication 31 March 2017.
ABSTRACT
Local climatic conditions are important determinants of disease
dynamics through effects on vector population performance or
distribution. Yet, climate may also be epidemiologically significant
due to effects on host2pathogen infection dynamics. We
developed a model to explore interactive effects between climate-
mediated acceleration in disease phenology (i.e., faster incubation
or symptom onset) and vector preference based on host symptom
status. Higher incubation rates favored pathogen outbreaks, but
more rapid symptom onset may constrain spread if vectors avoid
symptomatic hosts. Next, we tested whether warmer conditions
favored greater spread of the plant pathogen, Xylella fastidiosa,
by its leafhopper vector, Graphocephala atropunctata. Inoculated
and healthy plants were reared in two temperature-controlled
greenhouses. At six times postinoculation, a healthy and
inoculated plant were exposed to noninfective vectors, after which
pathogen spread was evaluated. Incubation rate and symptom
onset in infected hosts was significantly accelerated at higher
temperature. Although there was a tendency for greater pathogen
spread at higher temperature, the effect depended on time since
inoculation. In later introductions, after disease symptoms
manifest, vectors were more likely to be found on healthy hosts.
Vector avoidance of symptoms, particularly for hosts reared at
higher temperature, constrained pathogen spread at later
introductions. These results indicate that climate and vector
behavior may mediate interactively pathogen spread. Further
consideration of such epidemiological complexities is needed to
predict adequately the consequences of climate change for
disease dynamics.
Evidence linking climate to vector2pathogen2host interactions
(Lopes et al. 2009; Richards et al. 2007; Savage et al. 2011) has led
to the general perception that climate change will alter the distri-
bution and dynamics of infectious diseases. Yet, notable gaps
remain in our understanding of the precise manner in which
pathosystems will respond to a changing climate, undermining our
ability to generate specific predictions (Lafferty 2009; Rohr et al.
2011).
The literature is replete with examples of links between climate
and transmission efficiency (i.e., vector competence), particularly
for the inoculation phase of transmission (Daugherty et al. 2009;
Richards et al. 2007), which is often positively related to tem-
perature because of increased vector feeding rate or biting fre-
quency (Son et al. 2010) or increased extrinsic incubation rate (i.e.,
reduced time to vector infectiveness) (Carpenter et al. 2011). The
efficiency with which vectors acquire a pathogen from infected
hosts might also covary with temperature due to differences in
vector feeding activity or, indirectly, by affecting the intrinsic in-
cubation rate (i.e., host incubation rate). Climate is well known to
influence transmission dynamics (Canto et al. 2009) due to path-
ogen multiplication rates (Feil and Purcell 2001; Munyaneza et al.
2012) and infection persistence (Lieth et al. 2012; Lopes et al. 2009)
that depend on temperature. Yet, studies are lacking that evaluate
explicitly whether such effects of climate alter disease dynamics.
Moreover, other ecological contingencies exist that may mediate
the influence of climate on host2pathogen dynamics and spread.
One such contingency for vector-borne pathogens concerns the
potential for climatic conditions to influence simultaneously disease
onset and vector behavior. Climate can affect the pace of symptom
onset (i.e., host latent period) and disease symptom severity in
infected hosts (Lovell et al. 2004). Effects on host symptomology
may be important because vectors often exhibit orientation or
feeding preferences for individual hosts based on host infection
status. Such preferences are well documented for arthropod vectors
of plant pathogens (Blua and Perring 1992; Ingwell et al. 2012;
Mauck et al. 2010), with visual cues (Daugherty et al. 2011) or other
host phenotypic changes (i.e., chemical volatiles) (Eigenbrode et al.
Corresponding author: M. P. Daugherty; E-mail address: matt.daugherty@ucr.edu
*The e-Xtra logo stands for “electronic extra”and indicates that one supplementary
appendix, one supplementary figure, and three supplementary files are published
online.
© 2017 The American Phytopathological Society
1

2002) determining the attractiveness of infected hosts. Preference
for infected hosts tends to favor disease outbreaks, whereas
avoidance of infected hosts should constrain outbreaks (McElhany
et al. 1995; Sisterson 2008). Yet host phenotype following in-
fection need not be consistent over time (Blua and Perring 1992),
leading to concomitant changes in the magnitude or even trajectory
(i.e., switches between preference and avoidance) of vector
preference as disease progresses (Werner et al. 2009). Thus, if host
incubation period and symptom onset depend on local environ-
mental conditions, climate may interact with vector behavior to
determine disease incidence. We evaluated this hypothesis using
a combination of epidemiological modeling and an empirical test
with a generalist plant pathogen and its leafhopper vector.
Xylella fastidiosa is a xylem-limited bacterium that is pathogenic to
a wide variety of plants, including grapevines, in which it causes
Pierce’s disease (Purcell 1997). Multiplication of the bacterium plugs
xylem vessels, which leads to leaf scorch symptoms, shoot dieback,
and plant death (Purcell 1997). In parts of California the most
dominant vector is the blue-green sharpshooter, Graphocephala
atropunctata (Purcell 1975). This native xylem-sap feeding insect is
more efficient at transmitting X. fastidiosa to grapevines than are
other vectors, including the invasive glassy-winged sharpshooter
(Homalodisca vitripennis) (Daugherty and Almeida 2009).
Multiple lines of evidence suggest that climate, particularly
temperature, plays an important role in X. fastidiosa epidemiology.
There are direct effects on vector performance (Son et al. 2009),
density (Gruber and Daugherty 2013), feeding rate (Son et al.
2010), and transmission, which is generally more efficient at
higher temperatures (Daugherty et al. 2009). In addition, temper-
ature governs X. fastidiosa host infection dynamics. A greater
proportion of grapevines recover from X. fastidiosa infection under
colder overwintering conditions (Lieth et al. 2012), whereas
X. fastidiosa multiplication rates are generally higher under warmer
conditions (Feil and Purcell 2001). This last result is notable be-
cause G. atropunctata acquisition efficiency depends on host in-
fection level, with a threshold of approximately 10
4
CFU/g of plant
tissue required for efficient acquisition (Hill and Purcell 1997).
Collectively, these results suggest that warmer conditions may
exacerbate Pierce’s disease incidence.
G. atropunctata host preference may also be epidemiologically
significant. Sharpshooters exhibit preferences for host species and
grapevine variety (Purcell 1981) and strong within-host feeding-site
preferences that may underlie acquisition efficiency (Daugherty
et al. 2010). In addition, research has documented sharpshooter
preferences based on host disease symptom status. Sharpshooters
avoid symptomatic hosts (Marucci et al. 2005), but not asymp-
tomatic infected hosts, using visual cues associated with leaf scorch
symptoms (Daugherty et al. 2011). Avoidance of symptomatic
hosts may constrain pathogen acquisition and, therefore, temper
disease incidence compared with what it would be in the absence of
symptoms. However, to date no studies have measured pathogen
transmission and spread over a gradient of disease symptoms.
Despite ample evidence that climate affects host2pathogen in-
fection dynamics and host symptomology, and further evidence of
vector response to host symptomology, the net epidemiological
effect is not well understood. Therefore, we used a combination of
modeling and experiments to clarify potential interactions between
climate and vector behavior with respect to disease dynamics. First,
we modeled the effects of host incubation rate, symptom onset,
and vector preference to understand the dynamic consequences
of climate for disease incidence. Next, to test the qualitative pre-
dictions from the model, we conducted an experiment with
G. atropunctata and X. fastidiosa to estimate pathogen spread over
a temperature-associated gradient in disease.
MATERIALS AND METHODS
Epidemiological modeling. To better understand the individual
and combined effects of climate and vector behavior on disease
dynamics, we used a variation on a vector-borne epidemiological
model (Zeilinger and Daugherty 2014):
dS
dt
5mðE1C1IÞ2bSV
ðpI 1S1E1CÞ
dE
dt
5bSV
ðpI 1S1E1CÞ2ðg1mÞE
dC
dt
5gE2ðd1mÞC(1)
dI
dt
5dC2mI
dU
dt
5nV2apIU
ðpI 1S1E1CÞ2aCU
ðpI 1S1E1CÞ
dV
dt
5apIU
ðpI 1S1E1CÞ1aCU
ðpI 1S1E1CÞ2nV
where Sdenotes uninfected “healthy”or “susceptible”hosts, E
denotes latently “exposed”hosts that have been infected with the
pathogen but are neither infectious nor showing symptoms, C
denotes “asymptomatic-infectious”hosts that are an acquisition
source but are not yet showing symptoms, Idenotes “symptomatic-
infectious”hosts that are both infectious and diseased, Udenotes
“noninfective”vectors (i.e., those that have not acquired the
pathogen), and Vdenotes “infective”vectors. The parameters mand
ndenote host and vector turnover rates (i.e., loss of infection or
mortality), respectively, and aand bcorrespond with acquisition
and inoculation rates, respectively. Acquisition was assumed to be
similar for asymptomatic- and symptomatic-infectious hosts, which
may not always be true (Zeilinger and Daugherty 2014). The pa-
rameter pdescribes vector preference based on host infection
phenotype. When p51, vector contact with symptomatic hosts
(i.e., I) versus any of the asymptomatic host categories (i.e., S,E,or
C) is proportional to the relative density of the two phenotypes,
when p.1 vectors are disproportionately more likely to contact
symptomatic hosts, and when p,1 vectors are disproportionately
less likely to contact symptomatic hosts. We assumed that vector
preference does not differ based on vector infectivity (but see
Ingwell et al. 2012) nor that vector preference is affected directly by
warming but rather is indirectly affected by changes in host plant
quality (e.g., disease symptoms). Finally, gdescribes the rate at
which exposed hosts develop to become infectious (i.e., asymp-
tomatic incubation rate), and dis the rate at which asymptomatic-
infectious hosts become symptomatic (i.e., rate of symptom onset).
We investigated the epidemiological consequences of climate over
gradients in the rates at which exposed and asymptomatic-infectious
hosts transitioned to asymptomatic-infectious and symptomatic-
infectious states, respectively. In other words, g(incubation rate)
and/or d(rate of symptom onset) were assumed to be positively
related to temperature, as appears to be the general case for
X. fastidiosa incubation and Pierce’s disease symptom onset in
grapevines (Feil and Purcell 2001; Lieth et al. 2012).
Model behavior was evaluated primarily relative to whether an
outbreak was favored and the initial outbreak dynamics, rather than
2

investigating effects on equilibrium prevalence, to more directly
match up with the experiment conducted in the second part of the
study. Specifically, transient dynamics were evaluated using nu-
merical simulations throughout much of parameter space, especially
for g,d, and p, but with generally low initial infection prevalence in
the vector and host (i.e., E
0
,C
0
,I
0
, and V
0
near zero). In addition, we
explored effects on the pathogen reproduction number, R
0
,as
a convenient metric of the relative potential for disease outbreak to
occur (Supplementary Appendix):
R05abgTðm1dpÞ
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
abgmnNTðg1mÞðd1mÞðm1dpÞ
p(2)
where Nis the total density of hosts, and Tis the total density of
vectors, both of which were assumed to be constants. In a de-
terministic setting, as is the case here, disease outbreak occurs if R
0
is greater than 1.
Experiment with G. atropunctata and X. fastidiosa.To eval-
uate the qualitative predictions of the model, we tested the effect of
incubation temperature on the likelihood of pathogen spread over
time in the X. fastidiosa2grapevine2sharpshooter system. Early in
the spring we propagated approximately 200 own-rooted grapevine
cuttings (‘Cabernet Franc’) in 10-cm-square pots filled with
Supersoil potting soil (Rod Mclellan Company, San Mateo, CA).
After 1 month of growth, we mechanically inoculated half of the
plants with two 10-ml droplets of a turbid suspension of X. fas-
tidiosa (STL isolate, originally isolated from a symptomatic
grapevine in Napa Valley, CA) and SCP buffer. The other half of
the plants were inoculated only with SCP buffer. One week later,
half of the pathogen-inoculated and half of the healthy plants were
moved into two greenhouses at UC Berkeley’s Jane Gray Research
Greenhouse (Berkeley, CA) set to different temperatures on a 16:8 h
(day/night) photoperiod. In a high temperature room, day tem-
perature was set to 32°C and night was 21°C. Corresponding
conditions in a low temperature room were set to 21°C day and
night. Plants remained in these rooms until use in transmission
trials, and we regularly repositioned the plants within the rooms to
ensure they all were exposed to similar average conditions.
In order to estimate treatment effects on incubation rate, symptom
onset, and pathogen spread, we evaluated inoculated plants and
conducted transmission trials at six successive dates, approximately
3, 4, 8, 10, 14, and 16 weeks postinoculation. The earliest dates were
believed to be less than the time required for symptoms to develop
(i.e., less than the latent period) and for high pathogen populations to
develop (i.e., less than the incubation period) in grapevines raised in
the greenhouse (Hill and Purcell 1995). Conversely, the later dates
should allow sufficient time for strong symptoms and high in-
fectiousness to manifest in the inoculated vines. Trials were con-
ducted in a common intermediate temperature room (day 527°C,
night 521°C) toensure that pathogen spread was not confounded by
direct effects of temperature on sharpshooter behavior and trans-
mission (Daugherty et al. 2009; Son et al. 2010).
At each of the six dates, one X. fastidiosa-inoculated plant and
one healthy plant from a given temperature treatment were placed
into a 60 360 360 cm mesh and clear plastic cage (Bugdorm 2,
Megaview Sciences Inc., Taiwan). In establishing these pairs, care
was taken to choose plants of similar size. Afterward, five
G. atropunctata adults were introduced into each cage, placed in the
center of the cage on the floor. Sharpshooters were from a colony
established from insects collected earlier that summer along the
Russian River near Forestville, CA, and then raised on healthy
sweet basil plants (Ocimum basilicum L.). Previous tests indicated
that the vectors were free from X. fastidiosa infection. Vector
groups remained in the cages for 7 days.
At the end of each trial we counted the number of vectors on the
inoculated and healthy plants, and then we removed all vectors. We
then inspected each of the inoculated plants for evidence of Pierce’s
disease symptoms. For these inspections, we noted only whether
plants showed any of the characteristic leaf scorch symptoms (i.e.,
symptom presence/absence); we did not attempt to score relative
levels of disease severity. Next we collected petiole samples from
each of the inoculated plants. These petioles were plate cultured
with dilution to estimate X. fastidiosa infection level in inoculated
plants at approximately the time vectors were present (Hill and
Purcell 1995). Finally, the initially healthy plants remained in the
medium temperature room for up to 3 months after all trials were
completed. At the end of this period, we noted whether they showed
any disease symptoms and plate cultured petiole samples to de-
termine if the initially healthy plants had become infected with
X. fastidiosa during the trial.
Between the six successive dates we replaced plants with a new
pair, such that among dates the cages were independent of each
other (i.e., nonrepeated measures). A total of 96 cages were used,
with eight replicates of each temperature treatment at each of the six
dates. Those inoculated plants that did not test positive at the end of
a given trial were plate cultured again at the end of the study to
determine whether they were successfully infected. Five cages for
which the inoculated plant was not successfully infected with
X. fastidiosa were excluded from analysis.
Data analysis. It is worth noting that we were limited to
a single greenhouse for each of the temperature treatment levels.
This logistical constraint introduces the potential for autocorrelation
in some the response variables regarding symptomology and in-
fectiousness in the inoculated plants—though measures of vector
preference and pathogen spread, which were applied at the cage
level, would be arguably more truly independent. Regardless, for
the purposes of analysis, we necessarily treated all plants as in-
dependent replicates.
We analyzed effects on X. fastidiosa incubation rate by com-
paring the proportion of inoculated vines whose infection level,
based on plate culturing, was more than 10
4
CFU/g of plant material
at the end of the transmission trial. In other words, plant infectivity
was treated categorically based on whether it was above the level
that is known to be required for efficient acquisition by these vectors
(Hill and Purcell 1997). We tested for effects of temperature
treatment as a fixed effect and day (i.e., number of days post-
inoculation) as a covariate using a generalized linear model with
binomial error (Crawley 2007). We used stepwise deletion of
nonsignificant interactions from the full model, with goodness-of-fit
Ftests, to determine the minimum adequate model (Crawley 2007).
Effects on Pierce’s disease symptom onset were analyzed by
comparing the proportion of inoculated vines showing disease
symptoms at the end of each trial. As with plant infection, we used
a generalized linear model with binomial error to test for effects of
temperature as a fixed effect and day as a covariate.
The number of insects on each plant was used as a snapshot
estimate of vector preference for plants differing in infection status.
In each cage, we calculated the proportion of insects found on the
inoculated plant, ignoring the few insects that were on the cage
itself. Thus, 0.5 would indicate no vector preference, less than 0.5
would indicate preference for the healthy plant, and greater than 0.5
would indicate preference for pathogen-inoculated plants. This
proportion was compared using a two-way analysis of variance
(ANOVA), in which temperature was a fixed effect and day was
a covariate (Crawley 2007). Although it is not uncommon for
proportion data to violate the assumptions of such linear models,
requiring transformation or other modeling approaches, we in-
spected the data for normality and homogeneous variances and
3

found no evidence that either assumption was violated. Thus, the
ANOVA was run on the raw, untransformed, proportions. We also
tested whether the intercepts for high and low temperatures differed
significantly from 0.5 with separate ttests.
The final set of analyses relate to the potential for pathogen
spread, which we estimated by evaluating whether those initially
healthy plants became infected with X. fastidiosa during a given
trial. We used a generalized linear model with binomial error to test
for effects of temperature and day, both as fixed effects. A sig-
nificant interaction was followed-up with pairwise Fisher’s exact
tests.
RESULTS
Epidemiological modeling. Again, analysis of the model con-
centrated largely on the transient dynamics and the likelihood of
disease outbreak (i.e., R
0
.1) to more closely relate to the ex-
periment in the second half of the study (i.e., under what conditions
is pathogen spread most/least likely). Results from the model
showed that the incubation rate of exposed hosts and the rate of
symptom onset (gand d, respectively) can both strongly affect
disease dynamics, but not necessarily in an equivalent manner.
More rapid transition from the exposed to asymptomatic-infectious
host state (i.e., higher g) consistently increased incidence of both
infectious host categories (Fig. 1). More rapid symptom onset (i.e.,
higher d) favored a greater relative prevalence of symptomatic-
infectious compared with asymptomatic-infectious hosts (Fig. 1),
but did not necessarily affect the total prevalence of all infectious
hosts (i.e., C 1I). The prevalence of infection in vectors was highly
sensitive to the combination of dand vector preference. Higher
values of dincreased the prevalence of infective vectors if vectors
prefer symptomatic hosts, but had no effect if vectors do not exhibit
preference based on host infection phenotype, and generally re-
duced the prevalence of infective vectors if vectors avoid (i.e.,
p,1) symptoms (Supplementary Fig. S1).
An interaction between vector preference and hypothesized ef-
fects of climate is also apparent in the outcomes for R
0
(Figs. 2 and
3). Vector preference for symptoms generally increased the po-
tential for disease outbreak whereas avoidance of symptomatic
hosts constrained it (Figs. 2 and 3). More interesting were the
divergent effects of accelerated disease phenology based on dif-
ferent assumptions for vector preference. High incubation rates or
more rapid symptom onset promoted higher values of R
0
if vectors
Fig. 1. Simulated phase-plane dynamics for the prevalence of
symptomatic-infectious hosts versus asymptomatic-infectious hosts for
four combinations of incubation rate of exposed hosts (g) and rate of
symptom onset (d). Prevalence at t
0
is denoted by the open symbol and
final prevalence at t
100
by the closed symbols. Other parameter values are
a50.5, b50.4, m50.15, n50.15, p51, N5100, T5100, S
0
5100,
E
0
50, I
0
50, C
0
50, U
0
599, and V
0
50.
Fig. 2. Pathogen net reproduction rate (R
0
) as a function of incubation rate
of exposed hosts and rate of symptom onset under scenarios where
vectors A, strongly avoid (p50.05); B, show no preference (p51); or
C, strongly prefer (p55) symptomatic hosts. Other parameter values
are a50.4, b50.3, m50.2, n50.2, N5100, and T5150.
4

show no preference or prefer symptomatic hosts (Fig. 2B and C).
Yet, if vectors avoid symptoms, higher incubation rate favored
pathogen outbreak whereas more rapid symptom onset constrained
it (Fig. 2A). For pathosystems in which vectors avoid symptoms,
the magnitude of avoidance interacts strongly with climate (Fig. 3).
A warmer climate favors disease in the absence of vector preference
but is also more sensitive to vector avoidance of symptoms.
Specifically, for a given incremental reduction in the value of p,R
0
decreases proportionately more in the warmer conditions compared
with cooler conditions, leading to at least as strong constraints on
disease outbreaks if vectors strongly avoid symptoms (Fig. 3).
Experiment with G. atropunctata and X. fastidiosa.Over the
15 weeks that plants were in the low and high temperature
rooms, mean (6SD) day and night temperatures were 22.6 61.9
and 17.5 61.1°C and 30.0 63.1 and 21.6 61.4°C, respectively.
Temperatures in the common room in which transmission trials
took place were intermediate at 24.7 62.8 and 17.9 61.4°C for day
and night, respectively.
As noted, 91 of the 96 grapevine cuttings inoculated with
X. fastidiosa had detectable infections by the end of the study.
However, bacterial populations (CFU/g of plant) in these infected
plants at the time vectors were exposed to them varied dramatically,
from 0 to such high populations that individual colonies could not
be counted even after 10
6
-fold dilution. The best fit model for the
proportion of inoculated plants that were infectious (i.e., .10
4
CFU/g) included significant effects of temperature (x
2
1
514.563,
P50.0001) and number of days postinoculation (x
2
1
513.231, P5
0.0003). Infectivity increased rapidly at the higher temperature with
approximately 10% of plants infectious at the second date, 31 days
after inoculation, and up to approximately 80% infectious at later
dates (Fig. 4A). Infectivity in the low temperature plants progressed
much more slowly, with the first infections greater than 10
4
CFU/g
occurring in the fifth trial, 97 days after inoculation.
Symptom expression in the inoculated plants showed patterns
that were similar to plant infectivity, although symptoms manifest
slightly more slowly than infections. The best-fit model included
significant effects of temperature (x
2
1
514.740, P50.0001) and
days postinoculation (x
2
1
514.797, P50.0001) on the proportion
of inoculated plants that showed any disease symptoms. Symp-
toms manifest quickly in the high temperature plants, with plants
showing symptoms by the third trial, 53 days after inoculation (Fig.
4B). The cold temperature plants only started to show Pierce’s
disease at the last trial, 113 days after inoculation with X. fastidiosa.
The proportion of G. atropunctata found on the infected plant
was used as an indicator of vector preference based on plant in-
fection status. This proportion was not affected significantly by the
main effects of temperature (F
1,86
51.571, P50.2135) or days
postinoculation (F
1,86
51.221, P50.2723), but the interaction was
significant (F
1,86
55.773, P50.0184). The intercepts did not
differ significantly from 0.5 for vectors in low (b6SE 520.632 6
0.101; t
1
521.311, P50.197) or high temperatures (b6
SE 520.417 60.091; t
1
51.030, P50.3088). Vectors in both
treatments started in the first dates near 50% on the inoculated plant,
indicating no preference (Fig. 5). However, although the proportion
of vectors on the inoculated plant did not change appreciably in the
low temperature (m6SE520.001 60.001), the proportion
declined over time in the high temperature (m6SE 520.003 6
0.001). Between the first and last vector introduction date, based on
Fig. 3. Model prediction for the effect of vector avoidance of symptoms
(i.e., p,1) on the pathogen net reproduction rate in cool (solid line; g5
0.2, d50.2) versus warm conditions (dashed line; g51, d51). Open
versus closed symbols reflect approximate changes in vector preference
observed in the experiment between early versus late vector introduction
dates, which are predicted to lead to a substantially greater reduction in
the potential for pathogen spread over time in high temperature versus
low temperature conditions. Other parameter values are a50.4, b50.3,
m50.25, n50.25, N5100, and T5150.
Fig. 4. Proportion of inoculated plants over time that A, have high enough
Xylella fastidiosa infection levels to be adequate pathogen sources (i.e.,
.10
4
CFU/g of plant) or B, are showing Pierce’s disease symptoms.
Open and closed symbols denote the overall proportions for plants raised
at low temperature and at higher temperature, respectively. Lines denote
fit of the generalized linear model. Points are offset slightly for clarity.
5

the fit of the linear model, observed numbers of vectors on the
infected plants equated to pchanging from 1 (i.e., 1:1, 50% on
infected; no preference) to less than 0.25 (i.e., 1:4, 20% on infected;
avoidance) in the high temperature and did not change significantly
from 1 in the cold temperature. According to the modeling, the
disproportionately greater increase in vector avoidance of symp-
toms observed over time in the high but not the low temperatures
may constrain substantially pathogen spread over time (note
overlaid points in Figure 3), a prediction that is supported quali-
tatively by the empirical estimates of spread.
The proportion of initially healthy plants that became infected
with X. fastidiosa was used as a metric of pathogen spread. Overall,
vectors spread the pathogen from the inoculated to the initially
healthy plant in approximately 16% of replicates (14 of 91 cages),
with spread occurring in all but the initial trial. Main effects of
temperature (x
2
1
50.419, P50.5173) and days postinoculation
(x
2
5
59.581, P50.088) were not significant, but the interaction
was significant (x
2
5
511.438, P50.0434). If the first census (in
which the vast majority of inoculated plants were not yet infectious)
is ignored, the significance of the interaction is more pronounced
(x
2
4
511.438, P50.0221). In the low temperature treatment, the
first cases of pathogen spread occurred in the fourth trial, 73 days
postinoculation, with approximately 25% of replicates showing
spread at that date and afterward (Fig. 6). Pathogen spread occurred
much earlier in the high temperature treatment, at 31 days post-
inoculation. However, after an early peak of 60% of cages showing
spread in the high temperature, there was a strong decline over time
to zero in the final date.
DISCUSSION
Mitigating the effect of climate change on disease dynamics
requires knowledge of the manner in which climate affects different
aspects of a pathosystem. We used a combination of epidemio-
logical modeling and an empirical test to investigate links among
temperature, infection dynamics, symptom onset, and pathogen
spread in a system where prior research suggests both an important
role for climate (Daugherty et al. 2009; Feil and Purcell 2001; Lieth
et al. 2012; Son et al. 2009) and strong response by vectors to
host symptomology (Daugherty et al. 2011; Marucci et al. 2005).
The results indicate that warmer conditions can facilitate pathogen
spread, but with effects that are contingent on vector response to
disease symptoms.
In the experiment, warmer conditions accentuated the develop-
ment of high X. fastidiosa populations within infected hosts, which
is consistent with previous studies of X. fastidiosa that showed
a generally positive effect of temperature on pathogen multipli-
cation over the same range of temperatures (Feil and Purcell 2001).
Other pathosystems have documented relationships between tem-
perature and pathogen incubation rate, though the effect need not be
positive and is frequently not monotonic. For example, huang-
longbing prevalence in citrus trees varies regionally in part be-
cause one of the pathogens associated with the disease is heat
sensitive (Lopes et al. 2009). Further studies are needed of intrinsic
incubation rate over climate gradients, and corresponding effects on
pathogen acquisition, especially given that more rapid incubation
should favor higher disease incidence (Zeilinger and Daugherty
2014). Therefore, it is notable that in the current study we found
only a tendency for higher likelihood of pathogen spread at higher
incubation temperatures.
In addition to increasing incubation rate, warmer conditions
favored more rapid onset of disease symptoms. Similar effects of
temperature have been noted for the onset or severity of directly
transmitted plant pathogens (Lovell et al. 2004). Although we did
not quantify symptom severity, anecdotally, infected hosts showed
more severe disease symptoms at later dates, particularly at the
higher temperature. Regardless, disease symptom onset was not
coincident with the onset of host infectiousness. Intrinsic incubation
period and host latency are often used synonymously in the de-
scription of some pathogens where they coincide temporally or
where symptom severity is used to define infectivity (e.g., number
of lesions). Yet, infectiousness can often predate symptomology.
For example, in huanglongbing, pathogen acquisition by vec-
tors may predate the first disease symptoms by several months
or more (Coletta-Filho et al. 2014). A substantially delayed
symptom onset compared with incubation period may undercut
the effective application of interventions, such as roguing of
infected hosts, if relying solely on symptom expression (Coletta-
Filho et al. 2014). Additionally, the relative duration of in-
cubation periods and asymptomatic phases may be an important
Fig. 5. Mean proportion (6SE) of Graphocephala atropunctata found on
the inoculated plant over time. Means of each treatment are shown for
clarity.
Fig. 6. Proportion of cages over time in which infection spread to the
initially healthy plant. Different lowercase and uppercase letters represent
significant or marginally significant difference among time points at the
low and high temperatures, respectively. * denotes days for which there
was a significant difference between the temperatures.
6

determinant of disease dynamics when vectors respond strongly
to host disease status.
Vectors of animal (Cornet et al. 2013; O’Shea et al. 2002) and
plant (Eigenbrode et al. 2002; Ingwell et al. 2012; Mauck et al.
2010) pathogens may respond broadly to changes in host phenotype
following infection. In the current study, we quantified vector
preference in an admittedly simplistic manner by counting the
number of vectors on each host at the end of a trial. This approach
was taken to minimize disturbing the vectors in a way that might
encourage artifactual pathogen spread. A prior field experiment
(Daugherty et al. 2011) suggested that, although there appears to be
some individual variation, sharpshooters typically do not move very
often, particularly after moving onto healthy grapevines. Thus, our
snapshot measure should provide a reasonable proxy for vector
preference. We observed that vectors were progressively less likely
to be found on inoculated hosts that had been infected longer, at
least in the warmer conditions. Vectors showed no preference for
infection status for hosts raised in the cooler conditions, which
showed very delayed symptom onset. These results are not only
congruent with prior studies of vector behavior (Daugherty et al.
2011; Marucci et al. 2005), but also a more recent study of vector
feeding based on citrus variegated chlorosis infection status (De
Miranda et al. 2013). In the latter study, vectors were less likely to
feed on symptomatic hosts and engaged in truncated feeding bouts
on symptomatic hosts, whose vascular function is compromised,
relative to asymptomatic-infected hosts. Thus, disease symptom
onset can trigger a strong response by vectors in a manner that may
influence pathogen acquisition.
Our analysis expanded on previous theory (McElhany et al. 1995;
Sisterson 2008; Zeilinger and Daugherty 2014) by investigating the
extent to which vector behavior might interact with presumed
climate-mediated changes in host incubation rate and symptom
onset. The results suggest that if vectors either show no preference
or prefer symptomatic hosts, warmer conditions should consistently
favor a greater potential for disease outbreaks. Yet, if a vector
avoids symptomatic hosts, disease dynamics will depend strongly
on the nature with which climate affects the pathosystem. For such
vectors, conditions that accelerate the incubation rate but not
symptom onset maximize disease incidence whereas accelerated
symptom onset but not incubation rate minimize it. In other words,
vector avoidance of symptoms may constrain the extent to which
climate can facilitate disease outbreaks, by limiting the role
symptomatic hosts play as acquisition sources. This prediction was
supported generally by the results of our experiment, for which
disease spread was not consistently greater under warmer condi-
tions. Later on, as symptoms manifest in the warmer conditions,
vectors increasingly avoided infected hosts, and pathogen spread
declined. Collectively these results provide novel evidence that not
only may climate alter vector preference for hosts in a dynamic
manner (Werner et al. 2009), but that such changes are epidemi-
ologically significant. Importantly, this body of literature and our
present work only considers changes in vector preference from
warming mediated through changes in host plant quality. Warming
could also influence vector feeding behavior directly, for example,
through increasing movement rates; the epidemiological conse-
quences of such direct effects on vector feeding deserve further
consideration.
The generality of the conclusion that climate and vector behavior
interact to determine disease dynamics is likely to depend on key
attributes of a given pathosystem, including the amount of time
between the onset of host infectiousness and onset of symptoms.
As noted, some pathosystems can have protracted asymptomatic-
infectious periods (i.e., several months; Coletta-Filho et al. 2014).
Yet for others, especially those in which vectors respond to rapidly
induced volatile cues from infected hosts (Eigenbrode et al. 2002),
this gap is likely to be much shorter (i.e., days to a week). Another
important consideration is the nature of vector preference. In
many systems, unlike sharpshooters and X. fastidiosa, vectors are
attracted to infected hosts (Eigenbrode et al. 2002; Mauck et al.
2010). For such systems, more rapid symptom onset should favor
greater disease incidence, with presumably a greater potential for
warmer conditions that favor pathogen multiplication to promote,
unchecked, disease outbreaks. A better understanding of the in-
terplay among these factors is needed to predict specifically how
pathosystems will respond to a changing climate.
ACKNOWLEDGMENTS
We thank N. Killiny for providing insects, S. Purcell for helpful
discussion throughout this project, and K. Anderson for comments
on an earlier draft of this manuscript. R. P. P. Almeida was sup-
ported by a grant from the Pierce’s Disease Control Program. M. P.
Daugherty was supported by funds from USDA-CSREES and from
the Pierce’s Disease Control Program.
LITERATURE CITED
Blua, M. J., and Perring, T. M. 1992. Effects of zucchini yellow mosaic virus on
colonization and feeding behavior of Aphis gossypii (Homoptera: Aphididae)
alatae. Environ. Entomol. 21:578-585.
Canto, T., Aranda, M. A., and Fereres, A. 2009. Climate change effects on
physiology and population processes of hosts and vectors that influence the
spread of hemipteran-borne plant viruses. Glob. Change Biol. 15:1884-1894.
Carpenter, S., Wilson, A., Barber, J., Veronesi, E., Mellor, P., Venter, G., and
Gubbins, S. 2011. Temperature dependence of the extrinsic incubation period
of orbiviruses in Culicoides biting midges. PLoS One 6:e27987.
Coletta-Filho, H. D., Daugherty, M. P., Ferreira, C., and Lopes, J. R. S. 2014.
Temporal progression of ‘Candidatus Liberibacter asiaticus’infection in
citrus and acquisition efficiency by Diaphorina citri. Phytopathology 104:
416-421.
Cornet, S., Nicot, A., Rivero, A., and Gandon, S. 2013. Malaria infection
increases bird attractiveness to uninfected mosquitoes. Ecol. Lett. 16:
323-329.
Crawley, M. J. 2007. The R Book. John Wiley & Sons, England. doi:10.1002/
9780470515075
Daugherty, M. P., and Almeida, R. P. P. 2009. Quantifying the transmission
process of a plant pathogen and its vectors: Xylella fastidiosa and
sharpshooters in grapevines. Entomol. Exp. Appl. 132:84-92.
Daugherty, M. P., Bosco, D., and Almeida, R. P. P. 2009. Temperature mediates
vector transmission efficiency: Inoculum supply and plant infection dynamics.
Ann. Appl. Biol. 155:361-369.
Daugherty, M. P., Lopes, J. R. S., and Almeida, R. P. P. 2010. Vector within-host
feeding preference mediates transmission of a heterogeneously distributed
pathogen. Ecol. Entomol. 35:360-366.
Daugherty, M. P., Rashed, A., Almeida, R. P. P., and Perring, T. M. 2011. Vector
preference for host infection status: Sharpshooter movement and Xylella
fastidiosa transmission. Ecol. Entomol. 36:654-662.
De Miranda, M. P., Villada, E. S., Lopes, S. A., Fereres, A., and Lopes, J. R. S.
2013. Influence of citrus plants infected with Xylella fastidiosa on stylet
penetration activities of Bucephalogonia xanthophis (Hemiptera:
Cicadellidae). Ann. Entomol. Soc. Am. 106:610-618.
Eigenbrode, S. D., Ding, H., Shiel, P., and Berger, P. H. 2002. Volatiles from
potato plants infected with potato leafroll virus attract and arrest the virus
vector, Myzus persicae (Homoptera: Aphididae). Proc. R. Soc. Lond. B 269:
455-460.
Feil, H., and Purcell, A. H. 2001. Temperature-dependent growth and survival of
Xylella fastidiosa in vitro and in potted grapevines. Plant Dis. 85:1230-1234.
Gruber, B. R., and Daugherty, M. P. 2013. Predicting the effects of seasonality
on the risk of pathogen spread in vineyards: Vector pressure, natural
infectivity, and host recovery. Plant Pathol. 62:194-204.
Hill, B. L., and Purcell, A. H. 1995. Acquisition and retention of Xylella
fastidiosa by an efficient vector, Graphocephala atropunctata.
Phytopathology 85:209-212.
7

Hill, B. L., and Purcell, A. H. 1997. Populations of Xylella fastidiosa in plants
required for transmission by an efficient vector. Phytopathology 87:1197-1201.
Ingwell, L. L., Eigenbrode, S. D., and Bosque-P´
erez, N. A. 2012. Plant viruses
alter insect behavior to enhance their spread. Sci. Rep. 2:578.
Lafferty, K. D. 2009. The ecology of climate change and infectious disease.
Ecology 90:888-900.
Lieth, J. H., Meyer, M. M., Yeo, K.-H., and Kirkpatrick, B. C. 2012. Modeling
cold curing of Pierce’s disease in Vitis vinifera ‘Pinot Noir’and ‘Cabernet
Sauvignon’grapevines in California. Phytopathology 101:1492-1500.
Lopes, S. A., Frare, G. F., Bertolini, E., Cambra, M., Fernandes, N. G., Ayres,
A. J., Marin, D. R., and Bov´
e, J. M. 2009. Liberibacters associated with citrus
Huanglongbing in Brazil: ‘Candidatus Liberibacter asiaticus’is heat tolerant,
‘Ca. L. americanus’is heat sensitive. Plant Dis. 93:257-262.
Lovell, D. J., Hunter, T., Powers, S. J., Parker, S. R., and Van den Bosch, F.
2004. Effect of temperature on latent period of Septoria leaf blotch on winter
wheat under outdoor conditions. Plant Pathol. 53:170-181.
Marucci, R. C., Lopes, J. R. S., Vendramim, J. D., and Corrente, J. E. 2005.
Influence of Xylella fastidiosa infection of citrus on host selection by
leafhopper vectors. Entomol. Exp. Appl. 117:95-103.
Mauck, K. E., De Moraes, C. M., and Mescher, M. C. 2010. Deceptive chemical
signals induced by a plant virus attract insect vectors to inferior hosts. Proc.
Natl. Acad. Sci. USA 107:3600-3605.
McElhany, P., Real, L. A., and Power, A. G. 1995. Vector preference and disease
dynamics: A study of Barley yellow dwarf virus. Ecology 76:444-457.
Munyaneza, J. E., Sengoda, V. G., Buchman, J. L., and Fisher, T. W. 2012.
Effects of temperature on ‘Candidatus Liberibacter solanacearum’and zebra
chip potato disease symptom development. Plant Dis. 96:18-23.
O’Shea, B., Rebollar-Tellez, E., Ward, R. D., Hamilton, J. G. C., El Naiem, D.,
and Polwart, A. 2002. Enhanced sandfly attraction to Leishmania-infected
hosts. Trans. R. Soc. Trop. Med. Hyg. 96:117-118.
Purcell, A. H. 1975. Role of the blue-green sharpshooter, Hordnia circellata,inthe
epidemiology of Pierce’s disease of grapevines. Environ. Entomol. 4:745-752.
Purcell, A. H. 1981. Vector preference and inoculation efficiency as components
of resistance to Pierce’s disease in European grape cultivars. Phytopathology
71:429-435.
Purcell, A. H. 1997. Xylella fastidiosa, a regional problem or a global threat?
J. Plant Pathol. 79:99-105.
Richards, S. L., Mores, C. N., Lord, C. C., and Tabachnick, W. J. 2007. Impact of
extrinsic incubation temperature on virus exposure on vector competence of
Culex pipiens quinquefasciatus Say (Diptera: Culicidae) for West Nile virus.
Vector Borne and Zoonotics Dis. 7:629-636.
Rohr, J. R., Dobson, A. P., Johnson, P. T. J., Kilpatrick, A. M., Paull, S. H.,
Raffel, T. R., Ruiz-Moreno, D., and Thomas, M. B. 2011. Frontiers in climate
change-disease research. Trends Ecol. Evol. 26:270-277.
Savage, L. T., Reich, R. M., Hartley, L. M., Stapp, P., and Antolin, M. F. 2011.
Climate, soils, and connectivity predict plague epizootics in black-tailed
prairie dogs (Cynomys ludovicianus). Ecol. Appl. 21:2933-2943.
Sisterson, M. S. 2008. Effects of insect-vector preference for healthy or
infected plants on pathogen spread: insights from a model. J. Econ.
Entomol. 101:1-8.
Son,Y.,Groves,R.L.,Daane,K.M.,Morgan,D.J.W.,andJohnson,M.W.2009.
Influence of temperature on Homalodisca vitripennis (Hemiptera: Cicadellidae)
survival under various feeding conditions. Environ. Entomol. 38:1485-1495.
Son, Y., Groves, R. L., Daane, K. M., Morgan, D. J. W., Krugner, R., and
Johnson, M. W. 2010. Estimation of feeding thresholds for Homalodisca
vitripennis (Hemiptera: Cicadellidae) and its application to prediction of
overwinter mortality. Environ. Entomol. 39:1264-1275.
Werner, B. J., Mowry, T. M., Bosque-Perez, N. A., Ding, H., and Eigenbrode,
S. D. 2009. Changes in green peach aphid responses to potato leafroll
virus-induced volatiles emitted during disease progression. Environ. Entomol.
38:1429-1438.
Zeilinger, A., and Daugherty, M. P. 2014. Vector preference and host
defense against infection interact to determine disease dynamics. Oikos
123:613-622.
8