O R I G I N A L A RT I C L Edoi10.1111evo.13631Two d.docx

O R I G I N A L A RT I C L E
doi:10.1111/evo.13631
Two decades of evolutionary changes in
Brassica rapa in response to fluctuations in
precipitation and severe drought
Elena Hamann,1,2 Arthur E. Weis,3 and Steven J. Franks1
1Department of Biological Sciences, Fordham University,
Bronx, New York 10458
2E-mail: [email protected]
3Department of Ecology and Evolutionary Biology, University
of Toronto, Toronto, ON M5S 3B2, Canada
Received March 27, 2018
Accepted October 5, 2018
As climate changes at unprecedented rates, understanding
population responses is a major challenge. Resurrection studies
can
provide crucial insights into the contemporary evolution of
species to climate change. We used a seed collection of two
Californian
populations of the annual plant Brassica rapa made over two
decades of dramatic precipitation fluctuations, including
increasingly
severe droughts. We compared flowering phenology, other

drought response traits, and seed production among four
generations,
grown under drought and control conditions, to test for
evolutionary change and to characterize the strength and
direction of
selection. Postdrought generations flowered earlier, with a
reduced stem diameter, and lower water-use efficiency (WUE),
while
intervening wet seasons reversed these adaptations. There was
selection for earlier flowering, which was adaptive, but delayed
flowering after wet years resulted in reduced total seed mass,
indicating a maladaptive response caused by brief wet periods.
Furthermore, evolutionary changes and plastic responses often
differed in magnitude between populations and drought periods,
suggesting independent adaptive pathways. While B. rapa
rapidly evolved a drought escape strategy, plant fitness was
reduced
in contemporary generations, suggesting that rapid shifts in
flowering time may no longer keep up with the increasing
severity
of drought periods, especially when drought adaptation is
slowed by occasional wet seasons.
K E Y W O R D S : Drought escape, global change, phenotypic
plasticity, phenology, rapid evolution, resurrection study.
There is now abundant evidence that climate change and altered

precipitation patterns (IPCC 2014) trigger large-scale species
losses, shifts in vegetation communities, and evolutionary plant
responses (Parmesan and Yohe 2003; Jump and Penuelas 2005;
Parmesan 2006; Franks et al. 2014). Particularly well
documented
are worldwide shifts in flowering time following advanced
spring-
time (Menzel et al. 2006; Miller-Rushing and Primack 2008;
Cleland et al. 2012). While much of the shift in this trait may be
due to the direct effects of temperature on developmental rate,
some could be due to an evolutionary response to selection im-
posed by a warmer environment (Nicotra et al. 2010; Hoffmann
and Sgro 2011; Merila and Hendry 2014; Gugger et al. 2015;
Stoks et al. 2016). Phenotypic plasticity, in which organisms re-
spond to changes in environmental conditions, is itself
genetically
mediated (Bradshaw 1965) and so can also evolve in responses
to selection imposed by a variable environment (Sultan 2000;
Hairston et al. 2001; Alpert and Simms 2002; Nussey et al.
2005;

Baythavong 2011; Sultan et al. 2013; Hamann et al. 2016). Key
goals in understanding the impact of global change on plants are
determining the relative contributions to population persistence
made by existing phenotypic plasticity, evolutionary change in
phenotypes, and the evolution of the plastic response in func-
tional traits like flowering time (Gienapp et al. 2008).
The prospects for population persistence depend on the rates
at which the mean and optimal phenotypes shift. Environmental
shifts can leave the population maladapted to the new condi-
tions causing a drop in net reproductive rate, possibly below
replacement level. Existing plasticity can promote persistence if
it shifts mean phenotype in the adaptive direction (Charmantier
et al. 2008). However, with sustained shifts in the optimum,
2 6 8 2
C© 2018 The Author(s). Evolution C© 2018 The Society for the
Study of Evolution.
Evolution 72-12: 2682–2696
http://orcid.org/0000-0003-2888-6440
http://orcid.org/0000-0002-7056-4886
http://orcid.org/0000-0001-9681-3038

L O N G - T E R M E VO L U T I O N I N C A L I F O R N I A
N B . r a p a
the population may reach the limit of adaptive plastic response
(Anderson 2012). Evolutionary responses to selection in key
functional traits can restore adaptation when optimal
phenotypes
shift. Theoretical models indicate that if sufficient genetic vari-
ation is maintained in key traits, the population can be
“rescued”
by adaptive evolution (Burger and Lynch 1995; Gomulkiewicz
and Shaw 2013; Gonzalez et al. 2013; Shaw and Shaw 2014).
Evolutionary rescue models generally make the reasonable
simplifying assumption that the environment shifts at a constant
rate. However, climate change has not proceeded at a steady
rate.
Variability between and within years has been and will remain
the
rule, and in fact may even increase (IPCC 2014). The
evolutionary
impact of this variation will depend on its temporal scale. At
one
extreme, fluctuations running for more than a generation could

cause temporary selective reversals in mean phenotype, while
those occurring over shorter time scales may select for
physiolog-
ical/developmental flexibility and plasticity in functional traits.
Although the classic Darwinian perspective viewed evolution
via natural selection as a very slow process, it is now clear that
measureable evolutionary change in natural populations can oc-
cur on decadal timescale or less (Bone and Farres 2001; Reznick
and Ghalambor 2001; Grant and Grant 2002; Réale et al. 2003;
Thompson 2013; Hendry 2016). While it is not certain how
com-
monly natural selection will ultimately keep pace with ongoing
environmental change (Etterson and Shaw 2001; Visser 2008;
Shaw and Etterson 2012), this question is ripe for study. The
“res-
urrection approach” provides a particularly powerful tool for
this
purpose for those species that produce dormant propagules, such
as the seeds of flowering plants or the resting eggs of aquatic
crustaceans (Hairston et al. 1999; Franks et al. 2008; Franks

et al. 2018). In the resurrection protocol, ancestor and
descendant
propagules from natural populations are grown under common
conditions. Because the environment is held constant,
phenotypic
differences between the generations can be attributed
confidently
to genetic (evolutionary) differences, rather than plastic devel-
opmental responses (Franks et al. 2008). Incidences of adaptive
response to environmental perturbations have been documented
for several cladoceran populations (Hairston et al. 1999; Geerts
et al. 2015; Stoks et al. 2016). For these studies, resting eggs
were exhumed from lake sediments, hatched, and compared to
individuals from contemporary generations. Franks et al. (2007)
were the first to take this approach with plants using
fortuitously
stored seed, and found the evolution of earlier flowering in
Bras-
sica rapa over the course of a multiyear drought. The
resurrection
approach has been applied to plants in only a few instances (re-

viewed in Franks et al. 2018); a lack of appropriately stored
seed
material has hampered its further application. However, a recent
effort, called Project Baseline, has secured suitable seed stocks
from multiple species from multiple locations for future
resurrec-
tion studies; exciting studies can be expected in the near future,
filling critical gaps in our understanding of evolution (Etterson
et al. 2016).
With their initial resurrection experiment, Franks et al. (2007)
documented a rapid evolutionary change toward earlier
flowering
time over the course of a five-year (1999–2004) drought in
Cali-
fornian populations of the annual Brassica rapa (field mustard).
In Mediterranean climates, such as found in southern Califor-
nia, drought abbreviates the growing season. Follow-up studies
established two important points: first, selection favors early
flow-
ering under abbreviated growing seasons, but late flowering
under

extended seasons (Weis et al. 2014); second, early-flowering in-
dividuals had higher survival, smaller stem diameter, fewer leaf
nodes, and lower water-use efficiency (WUE) than late-
flowering
plants. This second point indicates that drought favors plants
that
develop rapidly, and thus flower at an earlier age (Franks and
Weis 2008; Franks 2011). It thus appears that over a few
genera-
tions B. rapa evolved an adaptive drought escape strategy
through
earlier flowering rather than drought tolerance, which would oc-
cur through increased WUE (Heschel and Riginos 2005; Franks
2011).
Having documented the initial evolutionary change over the
1999–2004 California drought (Franks et al. 2007), here we
carry
the investigation forward an additional 10 years. Drought
episodes
were more frequent and severe in California over that time
(Fig. 1; U.S. DroughtMonitor 2017; Swain et al. 2018), with

seven
of 10 years showing abbreviated growing seasons. Using seeds
collected in 1997 and 2004, and two more recent generations
col-
lected in 2011 and 2014, we performed resurrection experiments
addressing several issues on the adaptive response to environ-
mental change. As drought severity and frequency increased in
the past decade, plants may have continued to advance flower-
ing time. Alternatively, further fitness gains through phenology
may have been only marginal, transferring the thrust of selec-
tion toward drought tolerance (i.e., increased WUE). But there
is a potential trade-off between escape and tolerance: high WUE
may increase tolerance, but retard development rates and
thereby
constrain drought escape (Heschel and Riginos 2005; Franks
2011). Furthermore, although drought predominated over the re-
cent decade, there were two consecutive wet years. This raises
the
question whether these intermediate wet years reversed the di-
rection of selection, stalled, or temporarily reversed adaptation

to
drought. Alternatively, fluctuations in environmental conditions
could favor genotypes with higher phenotypic plasticity, allow-
ing plants to produce an optimal phenotype in each environment
(Via and Lande 1985; Alpert and Simms 2002). Finally, this
study
was replicated in two geographically distinct populations
(Franke
et al. 2006). Previous studies showed that evolutionary shifts in
flowering phenology in two populations were similar in direc-
tion but differed in magnitude (Franks et al. 2007; Franks
2011),
and that the majority of genetic changes in the populations were
EVOLUTION DECEMBER 2018 2 6 8 3
E . H A M A N N E T A L .
Figure 1. Early and late winter precipitation deviations from the
mean (in mm) in Santa Ana, Orange County, CA, for the past
two
decades (original data acquired from NOAA: station #
USC00047888). Cumulative precipitation during the growing
season was calculated

(100 days following the first rainfall �5 mm, which initiate
germination). Mean cumulative precipitation for days 1–50 and
days 51–100
was then calculated for each growing season leading up to
collection years (marked with a dashed box for predrought
generations, and
full box for postdrought generations), and values were plotted
as deviations from the mean for the first 50 days after rainfall
(in white)
and the following 50 days (in gray).
independent rather than shared by both populations (Franks et
al.
2016). Following multiple populations over consecutive drought
events enabled us to further test for the consistency (i.e., paral-
lelism vs independence) of evolutionary responses to selection
by
drought.
By resurrecting ancestral and descendant lines from two pop-
ulations over an 18 generation span, and growing them in a
recip-
rocal transplantation setup (i.e., in control and drought
conditions
mimicking their respective environments), this study examined

how drought events shaped selective pressures acting on B. rapa
and whether evolutionary changes are consistent between pop-
ulations and drought events. Specifically, we ask (1) whether
evolutionary phenotypic changes occurred between generations
collected at regular intervals between the wet-dry transitions,
(2)
if these evolutionary changes are adaptive and consistent in
direc-
tion and magnitude between the two populations, and (3)
whether
plastic responses to an experimental drought treatment align
with
evolutionary changes observed in nature.
Material and Methods
STUDY SPECIES
Brassica rapa (L.) Brassicaceae, commonly known as field mus-
tard, is an annual, self-incompatible, herbaceous plant
introduced
to North America about 300 years ago. In coastal California,
the growing season begins with the arrival of the winter rains,
which trigger plant germination (i.e., from late October to

January). The rains continue until early to late spring, during
which time plants complete their life cycles. The growing
season is terminated by the onset of annual summer drought, the
timing of which varies among years (Franke et al. 2006). As in
previous studies on this system (Franke et al. 2006; Franks et
al.
2007; Franks and Weis 2008; Franks 2011), we sampled two
populations: Arboretum (ARB) and Back Bay (BB). The two
sites, located in Orange County, California, are about 3 km
apart.
The soil at the BB site is sandier and more drained, resulting in
a consistently drier site than ARB, which in turn is more
variable
in soil water availability (Franke et al. 2006). We collected
seeds
from >200 plants per site at four points in time, representing
four
generations for each population. Precipitation data for the last
two
decades was obtained from the closest weather station located
at ca. 5 km from the sites (Santa Ana weather station # 121 in

Orange County, California; NOAA 2017). Ancestral predrought
lines were collected in 1997, after a series of normal wet years,
where especially the late winter precipitation was above
average,
resulting in long growing seasons (Fig. 1). The descendant post-
drought lines were collected in 2004 after a series of
abnormally
dry years (Fig. 1). Although the early winter precipitation was
above average in 1999–2000 and 2002–2003, the late winter
2 6 8 4 EVOLUTION DECEMBER 2018
N B . r a p a
precipitation was below average for a five-year period before
seed
collection (Fig. 1). Another generation was collected in 2011
after a short rainy episode recorded between 2009 and 2011,
which received above average late-winter precipitation (Fig. 1).
A final generation was collected in 2014, after three
consistently

and extremely dry years, which received very little rainfall
during
the entire growing season (Fig. 1; Swain et al. 2014).
EXPERIMENTAL DESIGN
For each of the eight groups (ARB’97, BB’97, ARB’04, BB’04,
ARB’11, BB’11, ARB’14, BB’14), we grew 120 randomly se-
lected seeds for one generation (i.e., refresher generation) in the
greenhouse to reduce maternal and storage effects, and to gener-
ate maternal family lines for the experiment (Franks et al. 2007;
Franks et al. 2018). In September 2016, seeds were directly
sown
into a growing medium (Sunshine RSI #1 Mix, Sun Gro
Horticul-
ture, Vancouver, BC, Canada) and grown in container trays
(cones
of 4 cm × 17 cm) under a 16 h light: 8 h dark photoperiod. Os-
mocote Smart-Release R© 14-14-14 fertilizer (Scotts,
Marysville,
OH, USA) was added a week after germination and plants were
watered daily to soil capacity. For each plant, date of
germination
and onset of flowering were recorded. Once flowering had
started,

plants were hand-pollinated in bulk within the same groups
every
three days. Seeds were collected at maturity for each individual,
stored separately in coin envelopes and kept refrigerated at 4°C.
Germination rates were high (>90%) for all groups, except for
BB 2004 (>65 %), ensuring that an unbiased sample of the gene
pool from the original populations was grown (Franks et al.
2018;
Weis 2018).
In January 2017, we randomly selected 60 family lines from
the 120 grown in the refresher generation. Four seeds per
maternal
line were sown into individual 9.3 × 9.3 × 8 cm pots placed in
square carry trays fitting 15 pots (T.O. Plastics, Clearwater,
MN,
USA). The pots were filled with the same growing medium, and
plants were grown in the greenhouse under the same light con-
ditions as previously. Osmocote Smart-Release R© fertilizer
was
again added a week after germination and plants were watered
every day to soil capacity to ensure seedling establishment. The

drought treatment was initiated ca. 15 days after germination,
once seedlings had produced three true leaves. Half of the
plants
from each group (i.e., two replicates per maternal line from
each
group: 60 lines × 2 replicates = 120 plants per group) were
grown
under a control treatment, where plants were watered every day
to
soil capacity, while the other half were given a drought
treatment
in which they were watered every five days. Volumetric water
content was measured every week with a probe (FieldScout TDR
100 Soil Moisture Sensor, Spectrum Technologies Inc., Texas,
USA) in a random subset within each treatment. At the start of
the watering treatments all trays were randomized in a split-
block
design (see Fig. S1). Each tray contained 15 maternal lines, so
that four trays contained one replicate of all 60 maternal lines
of
one group. In total, 1920 plants were grown for this experiment
(8 groups × 60 lines × 2 replicates × 2 treatments; Fig. S1).
Plants were monitored daily, and the date of germination

and flowering recorded. Onset of flowering was then calculated
as the number of days between germination and first flowering.
The stem diameter at first flowering was measured just above
the cotyledon node using a caliper. Once plants started
flowering,
they were hand-pollinated within groups every three days to
allow
seed set.
Ten weeks after the initiation of watering treatments, we mea-
sured two traits related to drought tolerance, specific leaf area
(SLA) and water use efficiency (WUE), on a subset of plants.
Leaf disks of 8 mm in diameter were taken from three new but
fully developed leaves from 40 plants per group and treatment.
Leaf disks were stored in individual coin envelopes, dried at
60°C
for 48 h, and weighed together. SLA was calculated by divid-
ing the fresh leaf coring area by their mean dry mass (Perez-
Harguindeguy et al. 2013). WUE was measured by stable
isotope
analyses (Farquhar et al. 1989) on 16 random plants per group.

One new but fully developed leaf per plant was collected, stored
in individual coin envelopes and dried at 80°C for 48 h. Sam-
ples were then finely ground using a FastPrep R©-24 tissue
lyser
(MP Biomedicals, Solon, OH, USA) and 1–2 mg were weighed
into 5 × 9 mm tin capsules. The Stable Isotope Ecology Labora-
tory at the University of Georgia, USA, analyzed samples using
isotope ratio mass spectrometry. The results are reported as
δ13C
(‰) relative to PDB standard (Perez-Harguindeguy et al. 2013).
Upon senescence, siliques were collected for each individual
plant and stored in coin envelopes in a dry environment. Seeds
were then separated from silique shavings and weighed to
obtain
aggregate seed mass per individual.
STATISTICAL ANALYSES
All functional traits were analyzed with linear mixed-effect
mod-
els (Crawley 2007), using Type III sums of squares with the
lmerTest package (Kuznetsova et al. 2017) for R (R Develop-
ment Core Team 2008). To test for differences between the pop-

ulations, generations, and effects of the drought treatment, we
specified separate models for each variable with the fixed
factors
generation (1997, 2004, 2011, 2014), population (ARB, BB),
and treatment (control, drought), and their respective two-way
and three-way interactions. A significant generation effect is in-
dicative of differences between ancestor and descendants lines,
implying an evolutionary change in response to natural drought
events. A significant population effect shows that populations
differ in functional traits, and a significant treatment effect in-
dicates plastic responses to experimental drought. Furthermore,
a generation × population interaction shows that the popula-
tions evolved differently, while a population × treatment
implies
that plastic responses to drought differ between populations,
and
year × treatment interaction indicates that plasticity differs be-
tween generations (evolutionary changes in plasticity). To

account
for potential differences between maternal lines, we included
ma-
ternal lines nested within their respective population and gener-
ation as a random factor. Blocks were also accounted for in our
models as a random factor. All variables were analyzed using a
Gaussian distribution with an identity link function, and data
were
log-transformed when needed to satisfy normality. Using
lmerTest
and its “rand” function, we report F-values and P-values for
fixed
effects and χ2-values and P-values for random effects after Bon-
ferroni correction (α < 0.01). Contrasts for fixed effects were
tested using differences of least squares means (diff lsmeans) as
implemented in the “step” function of lmerTest, and using the
“pairs” function of the lsmeans package (Lenth 2016). P-values
for diff lsmeans are reported after Tukey adjustment for
multiple
comparisons.
Selection analyses (Lande and Arnold 1983) were performed

to test whether changes in flowering time and stem diameter
were adaptive and followed the direction of selection. No se-
lection analyses were performed for SLA and WUE because of
the reduced statistical power stemming from measuring these
traits on a small subset of plants. Standardized linear (β) and
nonlinear (γ) selection gradients were estimated as the regres-
sion coefficients of relative fitness on the standardized mean
trait values of genotypes within each group (Conner and Hartl
2004). Our goal was to estimate the impact of a fitness func-
tion likely to vary between populations, and shifting over time,
and so we relativize fitness and standardize trait values within
each generation and population (De Lisle and Svensson 2017).
Relative fitness was calculated by dividing the seed mass of
genotypes (averaged for the two half-siblings grown under each
treatment) by the mean seed mass within each group (popula-
tion, generation, and treatment). Standardized mean trait values
were also calculated within each group. Separate linear and
non-

linear models were performed for each group (population, gen-
eration, and treatment) to retrieve selection gradients (linear β
and quadratic γ) and P-values, which were corrected for
multiple
testing (α < 0.003). The parameter estimate from the quadratic
re-
gressions were doubled to obtain the quadratic selection
gradients
(Stinchcombe et al. 2008).
To investigate potential changes in the degree of phenotypic
plasticity between generations and populations in response to
the experimental drought treatment, a phenotypic plasticity
index
(Piv) was calculated following Valladares et al. (2006). This in-
dex was calculated as the difference between the maximum and
minimum mean value of onset of flowering and stem diameter
at flowering for each genotype divided by the maximum mean
(standardized index ranging from 0 to 1). The mean Piv was
then
compared between generations and populations using Wilcoxon
signed-rank tests. No corrections for multiple testing were

applied
to avoid being overly conservative with these nonparametric
tests.
All analyses were performed on R version 3.3.3 software (R
Development Core Team 2013).
Results
Over the course of 18 generations of fluctuating precipitation,
we
found evidence for evolutionary changes in our natural Brassica
rapa populations, with several traits showing significant shifts
between ancestors and descendants. The evolutionary responses
generally differed between populations but followed the
direction
of selection. Furthermore, the experimental drought treatment
in-
duced plastic responses in B. rapa lines, which also often
differed
between populations.
EVOLUTIONARY CHANGES ACROSS 18
GENERATIONS AND CONSISTENCY ACROSS
POPULATIONS

We here describe evolutionary responses revealed by ancestral-
descendant comparisons under common conditions, focusing on
the high-watering treatment but including the drought treatment
when appropriate. We found an evolutionary shift in flowering
time, with descendants flowering earlier than ancestors (P <
10−4;
Table 1). While both populations generally advanced flowering
in response to drought, the evolutionary change varied between
populations, as indicated by a significant population ×
generation
interaction (P < 10−4; Table 1). For the ARB (wet site) popula-
tion, descendants from 2004 started flowering 2 days earlier
than
the 1997 ancestors, yet this shift was statistically significant
only
before p-value adjustment (P = 0.02 before adjustment, P = 0.28
after adjustment; Fig. 2A). The generation collected in 2011, af-
ter two intermediate wet years, started flowering at a similar
time
to lines collected in 2004, and lines collected in 2014 started
flowering 1 day earlier, although this difference was not statis-
tically significant (Fig. 2A). The accumulated long-term evolu-
tionary shift in flowering time between generations from 1997

and 2014 was significant, with descendants flowering 3 days
ear-
lier (P = 0.03). The BB (dry site) population always flowered
about a week earlier than the ARB population except in 2011
(Fig. 2A). Furthermore, the generation collected in 2004
flowered
1 day earlier than the 1997 ancestors, though the shift was not
sta-
tistically significant. However, the generation collected after
the
intermediate wet years in 2011 significantly delayed flowering,
compared to 2004, by 6 days (P < 10−4), and the descendant
generation collected in 2014 subsequently advanced flowering
time by 8 days (P < 10−4). In the long-term, flowering time
was significantly advanced by 3 days between generations from
1997 and 2014 (P = 0.006). Similarly as under well-watered
N B . r a p a
30
35

40
45
1997 2004 2011 2014
Collection year
O
ns
et
o
f f
lo
w
er
in
g
(d
ay
s
af
te
r
ge
rm

in
at
io
n)
4.0
4.5
5.0
5.5
6.0
1997 2004 2011 2014
S
te
m
d
ia
m
et
er
a
t f
lo
w

er
in
g
(m
m
)
Collection year
A B
Figure 2. Mean ± SE for traits related to a drought escape
strategy: (A) onset of flowering, and (B) stem diameter at
flowering. Blue
lines represent the ARB population (the wetter and more
variable site), red lines the BB population (the sandier and drier
site). Full
lines with circles represent accessions grown under control
treatment and dashed lines with triangles represent plants grown
under
drought treatment. Gray-shaded zones represent generations
collected after consecutive dry years. Multiple contrasts for
fixed effects
are reported in the text as differences in least square means.
conditions, ARB lines from 2004 had an earlier onset of flow-
ering than 1997 ancestors when grown under drought conditions
(P = 0.01), and BB descendants from 2014 flowered

significantly
earlier than ancestors from 2011 (P < 10−4).
The stem diameter at flowering, which is an indicator of
whether plants flower at an earlier developmental stage (i.e., es-
cape strategy), varied across generations (P < 10−4; Table 1),
suggesting evolutionary changes, and differed between popula-
tions (P < 10−4; Table 1). Under well-watered conditions, no
significant differences in stem diameter were detected among
generations for the ARB population (Fig. 2B). However, diam-
eter in the BB population varied across generations (Fig. 2B).
A significant increase in stem diameter was seen between 2004
and 2011 (P < 10−4), and a subsequent decrease was recorded in
lines from 2014 (P < 10−4; Fig. 2B). When grown under
drought
conditions, stem diameter increased for both ARB (P = 0.005)
and BB (P < 10−4) between 2004 and 2011, after a wet pe-
riod and subsequently decreased in 2014 for ARB (P = 0.04)
and BB (P < 10−4) after the severe three-year drought (Fig.
2B). Furthermore, a significant population × generation interac-
tion was found for the stem diameter at flowering (P < 10−4;
Table 1), indicating differences in evolutionary changes
between
populations. The BB population always had a smaller stem di-
ameter at flowering compared to the ARB population, except in
2011 (Fig. 2B). Furthermore, a positive relationship was found

between the time to flowering and the stem diameter at first
flowering (r = 0.35, P < 10−4), suggesting that individuals
that flowered early also had a smaller stem diameter at first
flowering (Fig. S2A).
Water use efficiency (WUE) and specific leaf area (SLA),
both traits that relate to drought stress tolerance, showed sig-
nificant generation × population interactions (both P < 10−4,
Table 1), indicating that evolutionary changes between genera-
tions differed between populations. In the dry site (BB) popu-
lation, WUE peaked in 2011, after the wet years, with lower
values found in 2014 (P = 0.048; Fig. 3A). In contrast, the wet
site (ARB) population showed very little variation in WUE, but
generally had a higher WUE compared to the BB population, es-
pecially in 2014 (P = 0.01; Fig. 3A). When grown under drought
conditions, WUE also peaked in 2011 for BB, with WUE greater
in 2011 than in 2004 (P = 0.02) and 2014 (P = 0.0002; Fig. 3A).
For the Arb population, WUE was lower in 2004 than in 1997
(P = 0.04) and 2014 (P < 10−4, Fig. 3A). Additionally, WUE
was positively correlated with time to first flowering (r = 0.31,
P < 10−4). Individuals that flowered rapidly after germination
generally had a low WUE (Fig. S2B). For the other drought

response trait, SLA, there was little change over time for the
Arb population (Fig. 3B). However, in the dry site (BB) pop-
ulation, SLA showed a substantial increase between 2011 and
2014 under well-watered (P = 0.004) and drought (P < 10−4)
conditions (Fig. 3B).
Aggregate seed mass per plant, a component of fitness, dif-
fered between generations when grown under well-watered
condi-
tions (P < 10−4; Table 1). Both populations tended to have a
higher
-32.5
-32.0
-31.5
-31.0
-30.5
1997 2004 2011 2014
W

at
er
u
se
e
ffi
ci
en
cy
(
13
Collection year
36
40
44
48
1997 2004 2011 2014
S
pe
ci
fic
le

af
a
re
a
(m
m
2
m
g-
1 )
Collection year
A B
Figure 3. Mean ± SE for traits related to a drought tolerance
strategy: (A) WUE, and (B) SLA. Blue lines represent the ARB
population (the
wetter and more variable site), red lines the BB population (the
sandier and drier site). Full lines with circles represent
accessions grown
under control treatment and dashed lines with triangles
represent plants grown under drought treatment. Gray-shaded
zones represent
generations collected after consecutive dry years. Note that the
y-scale for panel (A) WUE is in negative values. Multiple
contrasts for

fixed effects are reported in the text as differences in least
square means.
seed mass in the 2004 generation, after the five-year drought
episode, compared to their ancestral generation from 1997,
indicating that evolutionary changes in phenotypic traits in re-
sponse to the first drought episode increased plant fitness.
However, this difference was statistically significant for ARB
(P = 0.005), but only before P-value adjustment for BB (P =
0.04
before adjustment, P = 0.48 after adjustment; Fig. 4). After the
intermediate wet years in 2011, both populations had a lower
seed
mass and reduced fitness compared to 2004 (P = 0.04 for ARB,
P = 0.001 for BB; Fig. 4). No significant differences in seed
mass were detected between generations collected in 2011 and
2014 (Fig. 4). When grown under drought conditions, seeds
mass
showed no evolutionary changes between generations (Fig. 4).
Finally, seed mass also differed between populations (P < 10−4;
Table 1). The BB population produced significantly more seeds
than the ARB population in 1997 (P < 10−4), 2004 (P = 0.01),
and 2014 (P = 0.006), but not in 2011 (Fig. 4).
SELECTION GRADIENTS

Linear selection gradients were always negative for onset of
flow-
ering, indicating that selection generally favored earlier flower-
ing. Significant directional selection for earlier flowering was
detected for all generations of the ARB population and for BB
2011 when grown under control conditions (Table 2). Accord-
ingly, the significant shift toward earlier flowering in ARB
2004,
after the first drought episode, followed the direction of se-
lection, and confirms the adaptive nature of a drought escape
strategy in postdrought lines. However, the delayed flowering
seen in BB 2011 after the intermediate wet years (Fig. 2A) op-
posed the direction of selection. Stabilizing selection was also
de-
tected, especially when plants were grown under the
experimental
drought treatment, as seen for ARB 2011, BB 1997, and BB
2011
(Table 2), indicating that while earlier flowering is generally fa-
vored, there is an optimum flowering time, and that flowering
too early reduced plant fitness. The stem diameter at flowering

was also under directional selection, with thicker stem
diameters
being favored in BB 2004 and 2014 when grown under drought
conditions (Table 2). However, we saw a strong reduction in
stem
diameter in response to the drought treatment in 2014 (Fig. 2B),
which exceeded optimal stem diameter at first flowering and led
to reduced fitness.
PLASTIC RESPONSES TO THE EXPERIMENTAL
DROUGHT TREATMENT
By decreasing the frequency of watering, the volumetric water
content (%) of the growing medium was significantly reduced
fivefold (P < 10−4). This drought treatment induced important
plastic responses (i.e., treatment effect) and revealed
evolutionary
changes in plasticity between generations in certain traits (i.e.,
generation × treatment interaction).
The drought treatment did not affect the onset of flowering
(P = 0.73; Table 1), indicating a lack of plasticity in this trait in
response to the drought treatment. For the stem diameter at
flow-

ering, we found a significant treatment effect (P = 0.001), and
N B . r a p a
T
a
b
le
1
.
Li
n
e
a
r-
m
ix
e
d
e
ff
e
ct

m
o
d
e
l
fo
r
th
e
co
m
p
a
ri
so
n
o
f
e
v
o
lu
ti
o
n

a
ry
re
sp
o
n
se
s
a
cr
o
ss
fo
u
r
g
e
n
e
ra
ti
o
n
s,

tw
o
p
o
p
u
la
ti
o
n
s,
a
n
d
tw
o
w
a
te
ri
n
g
tr
e

a
tm
e
n
ts
,
a
n
d
th
e
ir
re
sp
e
ct
iv
e
in
te
ra
ct
io
n

s.
O
ns
et
of
fl
ow
er
in
g
S
te
m
di
am
et
er
at
fl
ow
er
in
g
W

U
E
S
L
A
S
ee
d
m
as
s
N
D
f
D
D
f
F
/X
2
P
D
D
f

F
/X
2
P
D
D
f
F
/X
2
P
D
D
f
F
/X
2
P
D
D
f
F
/X
2

P
G
en
er
at
io
n
3
46
5
39
.1
6
<
10
−4
46
3
32
.9
9
<
10
−4

15
3
2.
72
0
.0
4
36
0
20
.0
4
<
10
−4
46
3
10
.1
9
<
10
−4
P

op
ul
at
io
n
1
46
5
28
3.
42
<
10
−4
46
3
10
8.
88
<
10
−4
15
3

19
.3
8
<
10
−4
35
8
1.
16
0.
28
46
3
36
.4
6
<
10
−4
T
re
at
m

en
t
1
13
93
0.
11
0.
73
5.
13
41
.0
4
0.
00
1
18
9
69
.5
6
<
10

−4
46
4
0.
07
0.
89
14
05
14
0.
69
<
10
−4
P
op
×
G
en
er
at
io

n
3
46
5
21
.8
6
<
10
−4
46
3
16
.8
1
<
10
−4
15
3
7.
57
<
10

−4
36
0
14
.6
1
<
10
−4
46
3
0.
95
0.
41
P
op
×
T
re
at
1
13
93

1.
76
0.
18
14
00
8.
41
0.
00
3
18
9
0.
35
0.
55
46
4
4.
18
0
.0
4
14

05
14
.8
7
0.
00
01
G
en
er
at
io
n
×
T
re
at
3
13
93
1.
82
0.
14

14
00
6.
51
0.
00
02
18
8
0.
74
0.
53
47
4
8.
28
<
10
−4
14
05
3.
41

0.
01
P
op
×
G
en
er
at
io
n
×
T
re
at
3
13
93
1.
22
0.
31
14
00

0.
67
0.
56
18
8
4.
5
0.
00
4
47
4
2.
64
0
.0
4
14
05
1.
07
0.
35

M
L
:P
op
:G
en
er
at
io
n
1
N
A
27
8.
28
<
10
−4
N
A
77
.2
2

<
10
−4
N
A
5.
12
0
.0
2
N
A
31
.5
<
10
−4
N
A
9.
73
0.
00
2

B
lo
ck
1
N
A
1.
15
0.
33
N
A
19
.3
2
<
10
−4
N
A
N
A
N
A

N
A
N
A
N
A
N
A
33
.3
1
<
10
−4
N
o
n
si
g
n
ifi
ca
n
t

in
te
ra
ct
io
n
s
w
e
re
re
m
o
v
e
d
fr
o
m
th
e
fi
n
a
lm

o
d
e
l.
W
e
re
p
o
rt
o
n
P
-v
a
lu
e
s
fo
r
fi
x
e
d

e
ff
e
ct
s
a
n
d
o
n
χ
2
v
a
lu
e
s
fo
r
ra
n
d
o
m
e

ff
e
ct
s
(i
.e
.,
b
lo
ck
a
n
d
m
a
te
rn
a
ll
in
e
n
e
st
e

d
in
p
o
p
u
la
ti
o
n
a
n
d
g
e
n
e
ra
ti
o
n
).
S
ig

n
ifi
ca
n
t
P
-v
a
lu
e
s
a
ft
e
r
B
o
n
fe
rr
o
n
i
co
rr

e
ct
io
n
(α
=
0
.0
1
)
a
re
sh
o
w
n
in
b
o
ld
,
a
n
d

in
it
a
li
c
w
h
e
n
si
g
n
ifi
ca
n
t
b
e
fo
re
co
rr
e
ct
io

n
.
D
e
n
o
m
in
a
to
r
d
e
g
re
e
s
o
f
fr
e
e
d
o
m

(D
D
f)
a
re
ca
lc
u
la
te
d
u
si
n
g
S
a
tt
e
rt
h
w
a
it
e

’s
a
p
p
ro
x
im
a
ti
o
n
a
s
im
p
le
m
e
n
te
d
in
lm
e

rT
e
st
.O
n
se
t
o
f
fl
o
w
e
ri
n
g
,s
te
m
d
ia
m
e
te

r
a
t
fl
o
w
e
ri
n
g
,a
n
d
se
e
d
m
a
ss
w
e
re
m
e
a
su

re
d
o
n
a
ll
1
9
2
0
in
d
iv
id
u
a
ls
(1
2
0
p
e
r
p
o
p

u
la
ti
o
n
/g
e
n
e
ra
ti
o
n
/t
re
a
tm
e
n
t)
,w
h
il
e
S

LA
w
a
s
m
e
a
su
re
d
o
n
a
su
b
se
t
o
f
6
4
0
in
d
iv
id

u
a
ls
(4
0
p
e
r
p
o
p
u
la
ti
o
n
/g
e
n
e
ra
ti
o
n
/t

re
a
tm
e
n
t)
,
a
n
d
W
U
E
o
n
a
su
b
se
t
o
f
2
5
6

in
d
iv
id
u
a
ls
(1
6
p
e
r
p
o
p
u
la
ti
o
n
/g
e
n
e
ra

ti
o
n
/t
re
a
tm
e
n
t)
.
P
o
p
:
p
o
p
u
la
ti
o
n
,

Tr
e
a
t:
tr
e
a
tm
e
n
t,
M
L:
m
a
te
rn
a
l
li
n
e
s.
900
1100

1300
1997 2004 2011 2014
To
ta
l s
ee
d
m
as
s
(m
g)
Collection year
Figure 4. Mean ± SE for total seed mass used as fitness. Blue
lines represent the ARB population (the wetter and more vari-
able site), red lines the BB population (the sandier and drier
site).
Full lines with circles represent accessions grown under control
treatment and dashed lines with triangles represent plants grown
under drought treatment. Gray-shaded zones represent genera-
tions collected after consecutive dry years. Multiple contrasts

for
fixed effects are reported in the text as differences in least
square
means.
significant population × treatment (P = 0.003) and generation ×
treatment (P = 0.0002) interactions (Table 1), indicating plastic-
ity and evolutionary shifts in plasticity across generations, as
well
as differences in plastic responses between populations. Stem
di-
ameter was reduced for both populations in all generations when
grown under drought conditions compared to control conditions,
but the reduction was more pronounced for the BB population,
and less pronounced for lines from 2011 (Fig. 2B). WUE also
varied in response to the drought treatment (P < 10−4; Table 1),
and evolutionary changes in plasticity were found for this trait
that
differed between populations as indicated by a significant three-
way interaction (P = 0.004; Table 1). In general, plants tended
to have a higher WUE (i.e., less negative δ13C) when grown un-
der experimental drought conditions (Fig. 3A). WUE was
greater
under drought compared to well-watered conditions for ARB in

1997 (P = 0.0003) and 2014 (P = 0.003), and for BB genera-
tions in 2004 (P = 0.04) and 2011 (P < 10−4; Fig. 3A).
Moreover,
while SLA rarely differed between plants grown under control
and
drought conditions, a significant generation × treatment interac-
tion was found (P < 10−4; Table 1). Only BB lines from 2014
had a
higher SLA under dry compared to control conditions (P <
10−4;
Fig. 3B). Finally, seed mass was affected by the drought treat-
ment, and this effect differed between populations and
generation,
as indicated by a significant population × treatment interaction
and a significant generation × treatment interaction (P = 0.0001,
P = 0.01, respectively; Table 1). Seed mass was generally
reduced
by the experimental drought treatment, and this reduction was
significant in all generations of the BB population (all P < 0.02;
Fig. 4), yet only for the 2004 generation in the ARB population
(P < 10−4; Fig. 4).
We also compared plasticity between generations and popu-

lations. This analysis revealed that both populations had a
similar
and rather low degree of phenotypic plasticity for onset of
flower-
ing (Fig. 5A). However, the strong drought selection that
advanced
flowering time in the BB population in 2014 also acted to
reduce
the plasticity index (Piv) compared to 2011, and led to BB
having
significantly lower plasticity compared to ARB in the last gen-
eration (Fig. 5A, Table S3). The stem diameter at flowering was
more plastic than onset of flowering for both populations, and
the BB population was more plastic compared to ARB in the
two
first generations (Fig. 5B, Table S3). The wet years
significantly
reduced the degree of plasticity in stem diameter in both
popula-
tions, and postdrought lines tended to have increased plasticity
in
stem diameter (significant for ARB 2004, and BB 2014; Fig.
5B,

Table S3).
Discussion
Using a resurrection approach, we detected rapid evolutionary
re-
sponses to drought in California populations of Brassica rapa.
Over the past 20 years these populations have been exposed
to more drought-abbreviated growing seasons than historically.
Within this time span they also experienced two consecutive
years
of above-average precipitation. Over this period of time, we de-
tected rapid evolutionary changes in traits related to drought es-
cape (flowering phenology and stem diameter), drought
tolerance
(WUE and SLA) and reproductive fitness (seed mass). Given
that the four collection generations were reared simultaneously
in
common environments, the phenotypic differences among them
can be attributed to genetic change over time, directly demon-
strating evolution. In addition to evolutionary changes in traits,
we also saw evolutionary shifts in trait plasticity. These shifts

in phenotype are generally consistent with adaptation to fluc-
tuations in precipitation, but we also found evidence for both
parallel and nonparallel responses to repeated bouts of selection
by drought. Here, we consider the observed changes in light of
known selection patterns, and discuss differences between popu-
lations and generations across the past two decades of
fluctuating
precipitation.
EVOLUTIONARY SHIFTS IN FLOWERING TIME OVER
20 YEARS
Life-history theory predicts that the optimal time for first
flower-
ing in annual plants is set by a trade-off between time allocated
to vegetative growth and time allocated to reproduction (Cohen
1976; Fox 1992; Eckardt 2005; Johansson et al. 2013). Under
short growing seasons, plants must flower early in order to com-
plete flower production, pollination, and seed maturation before
conditions turn lethal. When growing seasons are longer, plants
have the luxury of extending vegetative growth, allowing them

to flower at a larger size, and mature more seeds in the allot-
ted time. In the Mediterranean climate of southern California,
the growing season begins with the arrival of the winter rains in
late November to early January. This period lasts until early to
late spring, followed by the annual summer dry period (Franke
et al. 2006). Drought years are characterized by short growing
seasons, while wet years have longer growing seasons. Optimal
flowering time shifts with growing season length. Long seasons
favor extended vegetative growth, which allow plants to flower
at a larger size, and so have increased seed yield. Short seasons
favor rapid flowering; even though faster plants are smaller,
they
are more successful than slower ones because they complete
seed
maturation before the soil water is depleted (Cohen 1976; King
and Roughgarden 1983; Fox 1992; Kozłowski 1992; Ejsmond
et al. 2010; Johansson et al. 2013; Weis et al. 2014). Given
these
predictions, we expected directional selection for advanced
flow-

ering time over the drought intervals, and a rebound to longer
flowering times in the wetter, intervening intervals. While these
predictions have been tested for single drought episodes (Franks
et al. 2007; Weis et al. 2014), we lack studies of whether re-
peated fluctuations in soil moisture conditions would cause
shifts
in selection and repeated changes in the direction of evolution
in
natural plant populations. Our study showed that these
predictions
were generally supported over an extended period of
fluctuations
in precipitation. The overall pattern was for B. rapa to show
shifts
to earlier flowering time following drought periods, and shifts
to
later flowering time following wet periods. However, there were
interesting differences between drought periods and populations
that provide some novel insights into how populations respond
to
fluctuating conditions.
The ARB population, which occurs in an area of greater soil

moisture than the BB population, showed a strong shift to ear-
lier flowering following the first drought period that occurred
between 1997 and 2004. However, the ARB population then
showed relatively little response to the wet period preceding
2011
or the dry period preceding 2014. In contrast, the BB population
showed a relatively modest shift to earlier flowering during the
first drought period, but a large shift to later flowering after the
second wet period, followed by a substantial shift back to
earlier
flowering following the last drought period. Theory predicts an
asymmetrical fitness function, with more negative consequences
of flowering too late than too early (Weis et al. 2014; Austen et
al.
2017), and this was corroborated by the consistently negative
selection gradients indicating that selection always favors ear-
lier flowering. However, populations did not always follow
these

N B . r a p a
0.06
0.09
0.12
1997 2004 2011 2014
Collection year
0.15
0.20
0.25
0.30
0.35
1997 2004 2011 2014
Collection year
Pi
v
O
ns
et
of
flo
w

er
in
g
Pi
v
St
em
di
am
et
er
at
flo
w
er
in
g
*
* **
*
* **
***
*
Figure 5. Mean ± SE phenotypic plasticity index (Piv) for (A)
onset of flowering, and (B) stem diameter at flowering. Blue
lines represent

the ARB population (the wetter and more variable site), red
lines the BB population (the sandier and drier site). Gray-
shaded zones
represent generations collected after consecutive dry years.
Asterisks represent statistically significant differences between
generations
or populations:
∗
p < 0.05,
∗ ∗
p < 0.01,
∗ ∗ ∗
p < 10−4 as revealed by Wilcoxon signed-rank tests.
Table 2. Linear (β) and quadratic (γ) selection gradients
analysis on mean trait values for onset of flowering time and
stem diameter at
flowering.
Onset of
flowering
Stem
diameter
Pop Year Treatment Linear (β) Quadratic (γ) Linear (β)
Quadratic (γ)
ARB 1997 Control −1.59∗ ∗ −13.84(∗ ) −0.02 −0.10
Drought −1.13(∗ ) −2.36 −0.05 −0.44(∗ )

2004 Control −0.96∗ −9.07(∗ ) 0.08 −0.07
Drought −0.42 −7.76(∗ ) −0.03 −0.01
2011 Control −1.14∗ ∗ ∗ −2.38 0.13(∗ ) −0.04
Drought −1.29(∗ ) −23.79∗ ∗ 0.07 0.07
2014 Control −1.11∗ ∗ ∗ −8.83(∗ ) 0.03 0.10
Drought −0.28 −11.19∗ 0.12(∗ ) 0.01
BB 1997 Control −0.15 −3.99(∗ ) 0.06 −0.10
Drought −0.12 −15.23∗ ∗ ∗ 0.12 −0.42(∗ )
2004 Control −0.10 −2.72(∗ ) 0.03 −0.07
Drought −0.66∗ −2.50(∗ ) 0.15∗ 0.02
2011 Control −1.77∗ ∗ ∗ −0.64 0.01 0.06
Drought −1.03(∗ ) −17.36∗ ∗ ∗ −0.01 −0.08
2014 Control −0.01 −0.04 0.17(∗ ) −0.05
Drought 0.19 −0.11(∗ ) 0.18∗ ∗ ∗ −0.03
Traits were standardized within each group (Pop: population,
Year: collection year, Treatment: watering treatment), and
relative fitness calculated for each
group. Linear models were computed for each group separately.
A Bonferroni correction was applied (α = 0.003): (∗ )p < 0.05:
significant before correction,
∗
p< 0.003,
∗ ∗
p < 0.0001,
∗ ∗ ∗
p < 10-4.

predictions and selection directions. The patterns are consistent
with the following scenario: the optimal flowering time is later
at
the ARB site because of generally greater soil moisture at this
site
compared to the BB site (Franke et al. 2006), and thus the ARB
population, which is adapted to conditions at this site (Franks
2011), generally shows later flowering than the BB population.
Drought shifts the optimum flowering time to earlier at both
sites,
but there is a greater shift at the previously wetter ARB site;
conversely, wet periods cause a shift to later flowering, but a
greater shift in the optimal flowering time at the previously
drier
BB site. Because the ARB population is already later flowering,
the wet period does not induce much of a shift to later flowering
in the population, but it does induce a large shift to later

flower-
ing in the early-flowering BB population. However, the delay in
flowering time in the BB population ran counter to the direction
of selection, indicating that this substantial reversal in flower-
ing time induced considerable fitness reductions. Thus the two
populations have different flowering time optima under wet and
dry conditions, as well as different phenotypic distributions.
This
follows the general idea that the response to selection depends
on both the pattern of selection as well as the phenotypic distri-
bution of the population (Weis et al. 2014). Thus in predicting
responses to climatic changes, it will be important to determine
how phenotypic optima and distributions change in response to
new conditions.
It is also important to note that despite rapid evolutionary re-
sponses to fluctuations in precipitation, plant fitness was rarely
in-
creased and barely maintained in more recent generations, at
least
when the plants were reared under greenhouse conditions. While

seed set and seed mass could differ under greenhouse
conditions,
which includes hand-pollination, compared to natural conditions
in the field, plant fitness in the greenhouse appears at least rep-
resentative of natural field conditions, as plants reached similar
size, number of siliques and seed mass as in a prior field study
of the same populations (Franke et al. 2006). Furthermore, even
if the correlation between fitness measures in the greenhouse
and
field was not as strong as assumed, growing plants under
common
greenhouse conditions allows comparing plant fitness across
gen-
erations to infer on the adaptive nature of evolutionary changes
in phenotypic traits (Franks et al. 2018). After the initial five-
year drought between 1999 and 2004, descendant generations of
both populations produced a higher seed mass relative to their
ancestral lines, and changes in onset of flowering followed the
direction of selection patterns for ARB, indicating adaptive evo-
lutionary changes. However, the BB population, which delayed

flowering after the intermediate wet years in 2011 and opposed
the direction of selection gradients incurred fitness reductions,
indicating that this shift was maladaptive. Furthermore, the sub-
sequent evolutionary shifts toward advanced flowering after the
record-breaking three-year drought episode in 2014 did not
suffice
to increase seed mass production compared to that of the ances-
tral pre-drought generations (1997 or 2011). This further
indicates
that advances in flowering time, which are inherently limited by
plant development, may no longer suffice to offset the negative
effects imposed by increasingly severe drought episodes, even
when they follow the direction of selection. Moreover, it also
seems that the intermediate wet years recorded between 2009
and 2011, which reversed previous adaptation patterns and led
to
delays in flowering time, slowed down drought adaptation and
subsequently reduced plant fitness. While the general trend to-
ward increasing severity of drought creates important selective

pressures, the stochastic occurrence of wet seasons creates
coun-
terproductive (over the long-term) selective spells. Overall, this
combination leading to reduced or barely maintained plant
fitness
would suggest that rapid evolutionary changes in flowering time
might not be able to keep pace with changes in environmental
conditions (Etterson and Shaw 2001; Visser 2008; Shaw and Et-
terson 2012), especially if drought episodes become more
severe
(Mann and Gleick 2015; Swain et al. 2018) or if fluctuations in
conditions become more extreme.
CONSISTENCY OF EVOLUTIONARY RESPONSES
A major debate in evolutionary biology is to what extent evo-
lutionary responses to environmental changes are consistent, re-
peatable, and predictable across populations and over time
(Grant
and Grant 2002). However, very few previous studies
(Kettlewell
1956; Grant and Grant 2014) have been able to study
evolutionary

changes to environmental conditions fluctuating over a period
of
decades. Our long-term study, examining phenotypic changes in
two populations over 18 generations, allows investigating the
con-
sistency of evolutionary responses to repeated selective drought
spells.
We found that the direction of evolutionary responses to
changes in precipitation was generally consistent across popu-
lations and over time, but the magnitude of the responses varied
greatly. As with a previous study (Franks et al. 2007; Franks
2011), the two populations, ARB and BB, both evolved earlier
flowering and smaller stem diameter at time of first flowering
fol-
lowing drought events, indicating a drought escape strategy, but
the populations differed in the amount of change. We also found
that both populations responded to subsequent periods of in-
creased precipitation by evolving later flowering, and
subsequent
drought periods by evolutionary reversal to earlier flowering.

But again the magnitude of the changes differed among popula-
tions, and also differed over time. The differences among
popula-
tions are likely due, at least in part, to differences in soil
moisture
available at the different sites, as well as differences in the
popula-
tions that have been shaped by these different conditions
(Franke
et al. 2006; Franks 2011). Furthermore, differences in responses
over time are likely due to differences in the temporal pattern of
precipitation as well as the existing phenotypic distribution of
the
populations (Etterson and Shaw 2001; Jump and Penuelas
2005).
Before the first collection in 1997, there was an extended period
of four years of above-average precipitation, while the 2004
col-
lection was made after six years where there were generally
drier
than average conditions in the later half of the growing season.
In
contrast, the 2011 collection was made after two wet years, and

the 2014 collection after three years of severe drought (Fig. 1).
As
N B . r a p a
droughts continue to become more extensive and severe (Swain
et al. 2018), it is likely that the ability of populations to
respond
to either increases or decreases in precipitation will become
depleted (Shaw and Etterson 2012; Shaw and Shaw 2014).
Future
studies under experimentally controlled conditions are needed
to determine the repeatability of evolution to environmental
changes.
PHENOTYPIC PLASTICITY AND EVOLUTIONARY
CHANGES IN PLASTICITY
In addition to evolutionary changes, plants can also respond to
climatic changes through plasticity or through evolutionary
shifts

in plasticity (Price et al. 2003; Parmesan 2006; Nicotra et al.
2010; Richter et al. 2012; Sultan et al. 2013). By combining the
resurrection approach with experimental manipulations of water
availability in the greenhouse, we could examine both plastic
re-
sponses, as well as evolutionary shifts in plasticity, by
comparing
ancestors and descendants in their degree of drought response.
While experimental drought did not induce plastic changes in
flowering time, in accord with a previous study (Franks 2011),
traits correlated with flowering time varied substantially.
Drought
generally reduced the stem diameter at flowering, but also in-
creased WUE. While advanced flowering time in nature has
been
generally associated with accelerated developmental rates and
lower WUE (Franks 2011), the experimental drought treatment
in
this study induced a more conservative water use strategy.
These
responses may seem conflicting, yet they can be explained by

opposing selection patterns depending on the timing of drought
(Heschel and Riginos 2005). A series of studies on Impatiens
capensis (Meerb.) showed that early-season drought is more
likely
to select for drought escape via low WUE and early
reproduction,
while late-season drought tends to select for increased tolerance
via high WUE (Heschel et al. 2002; Heschel and Riginos 2005).
In southern California, late-season drought regulates the length
of the growing season, but as evidenced by the precipitation
data
(Fig. 1), early-season precipitation was generally below average
as well during dry years. In contrast, the experimental drought
treatment started a few weeks after germination and is thus
more
representative of late-season drought. Hence, it is likely that the
late initiation of the experimental drought led to increased
WUE,
while selection imposed by early-season drought in nature fa-
vored lower WUE in association with earlier flowering (Fig.
3A).

The contrasting strategies displayed by plants under early
versus
late season drought indicate the importance of drought timing
on plant responses, which has important implications for our
understanding of plant responses to changes in climatic condi-
tions, and indicates that experimental drought conditions need
to
be carefully calibrated to accurately reflect predicted conditions
under climate change (Jentsch et al. 2007). Additionally, the
antagonistic responses may also reflect a trade-off between
escaping drought through earlier flowering and avoiding
drought
by having a more conservative water-use strategy, which may
reflect selection for different drought-coping mechanisms
(Heschel et al. 2002; Franks 2011). These negative correlations
between multiple traits may further constrain adaptive evolution
to climate change (Etterson and Shaw 2001; Etterson 2004).
Furthermore, the experimental drought treatment considerably
reduced seed set in all generations of both populations (Fig. 4).
While a previous study demonstrated increased survival of

postdrought lines under dry conditions, suggesting adaptive
shifts in flowering time (Franks et al. 2007), here we found no
evidence that the evolutionary shifts were adaptive, at least for
the seed set component of fitness under the experimental
drought
conditions in the greenhouse in this study. While we recognize
that our experimental drought treatment may not exactly mimic
natural drought episodes, postdrought generations should still
have a relatively higher seed mass under experimental drought
conditions compared to pre-drought generations if adaptive
evolution has occurred. However, our results do not follow such
a trend and rather suggest that evolutionary responses to
drought
did not suffice to increase plant fitness.
To assess evolutionary changes in plasticity, we examined the
generation by treatment interaction terms in ANOVAs. We
found
evidence for evolutionary changes in plasticity of some but not
all
traits. Stem diameter, SLA, and seed mass all showed evidence

for
evolutionary changes in plasticity, while flowering time and
WUE
did not. We thus have some indication that as environmental
condi-
tions continue to fluctuate, some traits will evolve changes in
their
plasticity. We further compared phenotypic plasticity indices
for
onset of flowering and stem diameter at flowering between gen-
erations and populations, and found noteworthy patterns. While
plasticity in onset of flowering was relatively constrained in
both
populations, as shown in previous studies (Gugger et al. 2015),
the degree of plasticity decreased in the most recent generation
of
the BB population. This pattern is consistent with important
evo-
lutionary changes toward advanced flowering time after drought
and may suggest genetic assimilation (Pigliucci et al. 2006) for
earlier flowering in increasingly dry climates. In contrast, the
ARB population, which did not show a significant evolutionary

shift toward earlier flowering after the last drought episode
could
gain in having increased plasticity in this trait to respond to
climate fluctuations (Alpert and Simms 2002). Furthermore, the
stem diameter at flowering was comparatively more plastic than
onset of flowering, especially in the BB population and in post-
drought lines, which may suggest that plasticity in stem
diameter
could allow the accommodation of earlier flowering and the
evolution of an escape strategy. To our knowledge, only one
other
study has used the resurrection approach to document the evolu-
tion of phenotypic plasticity in functional traits during the
range
expansion of an invasive plant (Sultan et al. 2013). However, it
is
currently unclear to what extent such shifts in plasticity will
help

populations adapt to changing conditions (Horgan-Kobelski et
al.
2016).
To conclude, this resurrection study assessed ongoing evo-
lutionary changes in two populations of B. rapa in response to
the drying southern California climate over the past two
decades.
We observed significant advances in flowering time in descen-
dant lines relative to ancestral lines, which were associated with
reduced WUE and stem diameter at flowering, indicating the
evolution of an escape strategy, which generally followed the
direction of selection patterns. WUE and stem diameter also re-
sponded plastically to the experimental drought treatment, yet
plastic responses in WUE did not follow the same pattern as the
evolutionary response to natural drought episodes. Overall,
evolu-
tionary changes followed the same direction in both
populations,
but the magnitude of these changes was population specific. The
more recent drought episode also appeared to impose stronger

selective pressures, leading to further advances in flowering
time.
However, the pronounced shifts in flowering time did not
always
allow the maintenance of plant fitness, leading to the conclusion
that the increasing severity of the drought episodes may outpace
plant adaptation, which may be additionally hindered by rare
wet seasons, which reversed advances in phenology. Future
field
studies should follow plant fitness measures in situ to provide
a complement to experiments under controlled common condi-
tions in the greenhouse and to provide additional inferences
about
adaptive evolution, population dynamics, and persistence in the
face of climate change.
AUTHOR CONTRIBUTIONS
E.H. performed the experiment, analyzed the data and wrote the
manuscript. S.F. contributed the seed material and both S.F and
A.W.
helped write the manuscript.
ACKNOWLEDGMENTS
We would like to thank Conor Gilligan, Hansol Lee, Stephen
Johnson,

Richard Rizzitello, and Mike Lambros for help with data
collection and
technical support at the Louis Calder Center. This research was
supported
by the Swiss National Science Foundation (# P2BSP3 168833)
to E.H.,
the National Science Foundation (DEB-1142784 and IOS-
1546218) to
S.F., and by an NSERC Discover Grant to A.W. We are also
grateful
to the editors and three anonymous reviewers for the
suggestions made
which greatly improved the manuscript
DATA ARCHIVING
The data that support the findings of this study are available
from the cor-
responding author upon request, and are archived in the Dryad
repository
under https://doi.org/10.5061/dryad.s03n4d1.
CONFLICT OF INTEREST
The authors have no conflict of interest to declare.
LITERATURE CITED
Alpert, P., and E. L. Simms. 2002. The relative advantages of
plasticity and
fixity in different environments: when is it good for a plant to
adjust?
Evol. Ecol. 16:285–297.
Anderson, J. T., Inouye, D. W., McKinney, A. M., Colautti, R.
I., Mitchell-
Olds, T. 2012. Phenotypic plasticity and adaptive evolution
contribute

to advancing flowering phenology in response to climate
change. Pro-
ceedings of the Royal Society B- Biological Sciences 279:
3843–3852.
Austen, E. J., L. Rowe, J. R. Stinchcombe, and J. R. K. Forrest.
2017. Ex-
plaining the apparent paradox of persistent selection for early
flowering.
New Phytol. 215:929–934.
Baythavong, B. S. 2011. Linking the spatial scale of
environmental variation
and the evolution of phenotypic plasticity: selection favors
adaptive
plasticity in fine-grained environments. Am. Nat. 178:75–87.
Bone, E., and A. Farres. 2001. Trends and rates of
microevolution in plants.
Genetica 112–113:165–182.
Bradshaw, A. D. 1965. Evolutionary significance of phenotypic
plasticity in
plants. Adv. Genet. 13:115–155.
Burger, R., and M. Lynch. 1995. Evolution and extinction in a
changing
environment: a quantitative-genetic analysis. Evolution 49:151–
163.
Charmantier, A., R. H. McCleery, L. R. Cole, C. Perrins, L. E.
B. Kruuk,
and B. C. Sheldon. 2008. Adaptive phenotypic plasticity in
response to
climate change in a wild bird population. Science 320:800–803.

Cleland, E. E., J. M. Allen, T. M. Crimmins, J. A. Dunne, S.
Pau, S. E. Travers,
E. S. Zavaleta, and E. M. Wolkovich. 2012. Phenological
tracking en-
ables positive species responses to climate change. Ecology
93:1765–
1771.
Cohen, D. 1976. The optimal timing of reproduction. Am. Nat.
110:801–807.
Conner, J. K., and D. L. Hartl. 2004. A primer of ecological
genetics. Sinauer
Associates, MA, USA.
Crawley, M. J. 2007. The R Book. John Wiley & Sons, West
Sussex, England.
De Lisle, S. P., and E. I. Svensson. 2017. On the standardization
of fitness
and traits in comparative studies of phenotypic selection.
Evolution
71:2313–2326.
Eckardt, N. A. 2005. A time to grow, a time to flower. Plant
Cell 17:2615–
2617.
Ejsmond M. J., C. M., Kapustka F., and K. J. 2010. How to time
growth and
reproduction during the vegetative season: an evolutionary
choice for
indeterminate growers in seasonal environments. Am. Nat.
175:551–
563.
Etterson, J. R. 2004. Evolutionary potential of Chamaecrista

fasciculata in
relation to climate change. I. Clinal patterns of selection along
an envi-
ronemntal gradient in the Great Plains. Evolution 58:1446–
1456.
Etterson, J. R., S. J. Franks, S. J. Mazer, R. G. Shaw, N. L. S.
Gorden, H.
E. Schneider, J. J. Weber, K. J. Winkler, and A. E. Weis. 2016.
Project
baseline: an unprecedented resource to study plant evolution
across space
and time. Am. J. Bot. 103:164–173.
Etterson, J. R., and R. G. Shaw. 2001. Constraint to adaptive
evolution in
response to global warming. Science 294:151–154.
Farquhar, G. D., K. T. Hubick, A. G. Condon, and R. A.
Richards. 1989.
Carbon isotope fractionation and plant water-use efficiency.
Springer
New York, New York, NY, pp. 21–40.
Fox, G. A. 1992. Annual plant life histories and the paradigm of
resource
allocation. Evol. Ecol. 6:482–499.
Franke, D. M., A. G. Ellis, M. Dharjwa, M. Freshwater, M.
Fujikawa, A.
Padron, and A. E. Weis. 2006. A steep cline in flowering time
for
Brassica rapa in southern California: population-level variation
in the
field and the greenhouse. Int. J. Plant Sci. 167:83–92.

Franks, S. J. 2011. Plasticity and evolution in drought
avoidance and escape
in the annual plant Brassica rapa. New Phytol. 190:249–257.
Franks, S. J., J. C. Avise, W. E. Bradshaw, J. K. Conner, J. R.
Etterson, S. J.
Mazer, R. G. Shaw, and A. E. Weis. 2008. The resurrection
initiative:
https://doi.org/10.5061/dryad.s03n4d1
N B . r a p a
storing ancestral genotypes to capture evolution in action.
Bioscience
58:870–873.
Franks, S. J., E. Hamann, and A. E. Weis. 2018. Using the
resurrection ap-
proach to understand contemporary evolution in changing
environments.
Evol. Appl. 11:17–28.
Franks, S. J., N. C. Kane, N. B. O’Hara, S. Tittes, and J. S.
Rest. 2016. Rapid
genome-wide evolution in Brassica rapa populations following
drought
revealed by sequencing of ancestral and descendant gene pools.
Mol.
Ecol. 25:3622–3631.
Franks, S. J., S. Sim, and A. E. Weis. 2007. Rapid evolution of

flowering
time by an annual plant in response to a climate fluctuation.
PNAS
104:1278–1282.
Franks, S. J., J. J. Weber, and S. N. Aitken. 2014. Evolutionary
and plastic
responses to climate change in terrestrial plant populations.
Evol. Appl.
7:123–139.
Franks, S. J., and A. E. Weis. 2008. A change in climate causes
rapid evolution
of multiple life-history traits and their interactions in an annual
plant. J.
Evol. Biol. 21:1321–1334.
Geerts, A. N., J. Vanoverbeke, B. Vanschoenwinkel, W. Van
Doorslaer, H.
Feuchtmayr, D. Atkinson, B. Moss, T. A. Davidson, C. D.
Sayer, and
L. De Meester. 2015. Rapid evolution of thermal tolerance in
the water
flea Daphnia. Nat. Climate Change 5:665.
Gienapp, P., C. Teplitsky, J. S. Alho, J. A. Mills, and J. Merila.
2008. Cli-
mate change and evolution: disentangling environmental and
genetic
responses. Mol. Ecol. 17:167–178.
Gomulkiewicz, R., and R. G. Shaw. 2013. Evolutionary rescue
beyond the
models. Philos. Trans. R Soc. B Biol. Sci. 368:20120093.
Gonzalez, A., O. Ronce, R. Ferriere, and M. E. Hochberg. 2013.

Evolution-
ary rescue: an emerging focus at the intersection between
ecology and
evolution. Philos. Trans. R Soc. B Biol. Sci. 368:20120404.
Grant, P. R., and B. R. Grant. 2002. Unpredictable evolution in
a 30-year study
of Darwin’s finches. Science 296:707–711.
———. 2014. 40 years of evolution: Darwin’s finches on
Daphne Major
Island. Princeton Univ. Press, Princeton.
Gugger, S., H. Kesselring, J. Stöcklin, and E. Hamann. 2015.
Lower plas-
ticity exhibited by high- versus mid-elevation species in their
pheno-
logical responses to manipulated temperature and drought.
Annal. Bot.
116:953–962.
Hairston, N. G., C. L. Holtmeier, W. Lampert, L. J. Weider, D.
M. Post, J. M.
Fischer, C. E. Cáceres, J. A. Fox, and U. Gaedke. 2001. Natural
selection
for grazer resistance to toxic cyanobacteria: evolution of
phenotypic
plasticity? Evolution 55:2203–2214.
Hairston, N. G., W. Lampert, C. E. Caceres, C. L. Holtmeier, L.
J. Weider, U.
Gaedke, J. M. Fischer, J. A. Fox, and D. M. Post. 1999. Lake
ecosystems:
rapid evolution revealed by dormant eggs. Nature 401:446–446.
Hamann, E., H. Kesselring, J. F. Scheepens, G. F. J.

Armbruster, and J. Stoeck-
lin. 2016. Evidence of local adaptation to fine- and coarse-
grained envi-
ronmental variability in Poa alpina in the Swiss Alps. J. Ecol.
104:1627–
1637.
Hendry, A. P. 2016. Eco-evolutionary dynamics. Princeton
Univ. Press, Prince-
ton and Oxford.
Heschel, M. S., K. Donohue, N. Hausmann, and J. Schmitt.
2002. Population
differentiation and natural selection for water-use efficiency in
Impatiens
capensis (Balsaminaceae). Int. J. Plant Sci. 163:907–912.
Heschel, M. S., and C. Riginos. 2005. Mechanisms of selection
for drought
stress tolerance and avoidance in Impatiens capensis
(Balsaminacea).
Am. J. Bot. 92:37–44.
Hoffmann, A. A., and C. M. Sgro. 2011. Climate change and
evolutionary
adaptation. Nature 470:479–485.
Horgan-Kobelski, T., S. Matesanz, and S. E. Sultan. 2016.
Limits to future
adaptation in the invasive plant Polygonum cespitosum:
expression of
functional and fitness traits at elevated CO2. J Heredity 107:42–
50.
IPCC. 2014. Climate change 2013: the physical science basis.
Contribution of

working group I to the fifth assessment report of the
intergovernmental
panem on climate change [Stocker, T. F., D. Qin, G.-K.
Plattner, M.
Tignor, S. K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex,
and P. M.
Midgley (eds.)]. Cambridge Univ. Press, Cambridge, United
Kingdom
and New York, NY, USA.
Jentsch, K., J. Kreyling, and C. Beierkuhnlein. 2007. A new
generation of
climate change experiments: events not trends. Front. Ecol.
Environ.
5:365–374.
Johansson, J., K. Bolmgren, and N. Jonzén. 2013. Climate
change and the
optimal flowering time of annual plants in seasonal
environments. Global
Change Biol. 19:197–207.
Jump, A. S., and J. Penuelas. 2005. Running to stand still:
adaptation and the
response of plants to rapid climate change. Ecol. Lett. 8:1010–
1020.
Kettlewell, H. B. D. 1956. Further selection experiments on
industrial
melanism in the Lepidoptera. Heredity 10:287.
King, D., and J. Roughgarden. 1983. Energy allocation patterns
of the Califor-
nia grassland annuals Plantago erecta and Clarkia rubicunda.
Ecology
64:16–24.

Kozłowski, J. 1992. Optimal allocation of resources to growth
and repro-
duction: implications for age and size at maturity. Trends Ecol.
Evol.
7:15–19.
Kuznetsova, A., P. B. Brockhoff, and R. H. B. Christensen.
2017. LmerTest:
tests in linear mixed effect models. J. Stat. Software 82:1–26.
Lande, R., and S. J. Arnold. 1983. The measurement of
selection on correlated
characters. Evolution 37:1210–1226.
Lenth, R. V. 2016. Least-squares means: the R package lsmeans.
J. Stat.
Software 69:1–33.
Mann, M. E., and P. H. Gleick. 2015. Climate change and
California drought
in the 21st century. Proc. Natl. Acad. Sci. 112:3858–3859.
Menzel, A., T. H. Sparks, N. Estrella, E. Koch, A. Aasa, R.
Ahas, K. Alm-
Kubler, P. Bissolli, O. Braslavska, A. Briede, et al. 2006.
European
phenological response to climate change matches the warming
pattern.
Global Change Biol. 12:1969–1976.
Merila, J., and A. P. Hendry. 2014. Climate change, adaptation,
and phenotypic
plasticity: the problem and the evidence. Evol. Appl. 7:1–14.
Miller-Rushing, A. J., and R. B. Primack. 2008. Global warming

and flowering
times in Thoreau’s concord: a community perspective. Ecology
89:332–
341.
Nicotra, A. B., O. K. Atkin, S. P. Bonser, A. M. Davidson, E. J.
Finnegan,
U. Mathesius, P. Poot, M. D. Purugganan, C. L. Richards, F.
Valladares,
and M. van Kleunen. 2010. Plant phenotypic plasticity in a
changing
climate. Trends Plant Sci. 15:684–692.
NOAA. 2017. National oceanic and atmospheric administration:
climate data
online.
Nussey, D. H., E. Postma, P. Gienapp, and M. E. Visser. 2005.
Selection
on heritable phenotypic plasticity in a wild bird population.
Science
310:304–306.
Parmesan, C. 2006. Ecological and evolutionary responses to
recent climate
change. Ann. Rev. Ecol. Evol. Syst. 37:637–669.
Parmesan, C., and G. Yohe. 2003. A globally coherent
fingerprint of climate
change impacts across natural systems. Nature 421:37–42.
Perez-Harguindeguy, N., S. Diaz, E. Garnier, S. Lavorel, H.
Poorter, P. Jau-
reguiberry, M. S. Bret-Harte, W. K. Cornwell, J. M. Craine, D.
E. Gur-
vich, et al. 2013. New handbook for standardised measurement

of plant
functional traits worldwide. Aust. J. Bot. 61:167–234.
Pigliucci, M., C. J. Murren, and C. D. Schlichting. 2006.
Phenotypic
plasticity and evolution by genetic assimilation. J. Exp. Biol.
209:
2362–2367.
Price, T. D., A. Qvarnstrom, and D. E. Irwin. 2003. The role of
phenotypic
plasticity in driving genetic evolution. Proc. R Soc. Lond. Ser.
B Biol.
Sci. 270:1433–1440.
R Development Core Team. 2008. R: a language and
environment for statistical
computing. R foundation for Statistical Computing, Vienna,
Austria.
Réale, D., A. G. McAdam, S. Boutin, and D. Berteaux. 2003.
Genetic and
plastic responses of a northern mammal to climate change. Proc.
R Soc.
B Biol. Sci. 270:591–596.
Reznick, D. N., and C. K. Ghalambor. 2001. The population
ecology of
contemporary adaptations: what empirical studies reveal about
the con-

ditions that promote adaptive evolution. Genetica 112:183–198.
Richter, S., T. Kipfer, T. Wohlgemuth, C. C. Guerrero, J.
Ghazoul, and
B. Moser. 2012. Phenotypic plasticity facilitates resistance to
climate
change in a highly variable environment. Oecologia 169:269–
279.
Shaw, R. G., and J. R. Etterson. 2012. Rapid climate change and
the rate of
adaptation: insight from experimental quantitative genetics.
New Phytol.
195:752–765.
Shaw, R. G., and F. H. Shaw. 2014. Quantitative genetic study
of the adaptive
process. Heredity 112:13–20.
Stinchcombe, J. R., A. F. Agrawal, P. A. Hohenlohe, S. J.
Arnold, and
M. W. Blows. 2008. Estimating nonlinear selection gradients
using
quadratic regression coefficients: double or nothing? Evolution
62:
2435–2440.
Stoks, R., L. Govaert, K. Pauwels, B. Jansen, and L. De
Meester. 2016.
Resurrecting complexity: the interplay of plasticity and rapid
evolution
in the multiple trait response to strong changes in predation
pressure in
the water flea Daphnia magna. Ecol. Lett. 19:180–190.
Sultan, S. E. 2000. Phenotypic plasticity for plant development,

function and
life history. Trends Plant Sci. 5:537–542.
Sultan, S. E., T. Horgan-Kobelski, L. M. Nichols, C. E. Riggs,
and R. K.
Waples. 2013. A resurrection study reveals rapid adaptive
evolution
within populations of an invasive plant. Evol. Appl. 6:266–278.
Swain, D. L., B. Langenbrunner, J. D. Neelin, and A. Hall.
2018. Increasing
precipitation volatility in twenty-first-century California. Nat.
Climate
Change 8:427–433.
Swain, D. L., M. Tsiang, M. Haugen, D. Singh, A. Charland, B.
Rajaratnam,
and N. S. Diffenbaugh. 2014. The extraordinary California
drought of
2013/2014: character, context, and the role of climate change.
Bull. Am
Meteorol. Soc. 95:3–7.
Thompson, J. N. 2013. Relentless evolution. Chicago Univ.
Press, Chicago,
USA.
U.S. DroughtMonitor. 2017. California.
Valladares, F., D. Sanchez-Gomez, and M. A. Zavala. 2006.
Quantitative
estimation of phenotypic plasticity: bridging the gap between
the evolu-
tionary concept and its ecological applications. J. Ecol.
94:1103–1116.

Via, S., and R. Lande. 1985. Genotype-environment interaction
and the evo-
lution of phenotypic plasticity. Evolution 39:505–522.
Visser, M. E. 2008. Keeping up with a warming world;
assessing the rate of
adaptation to climate change. Proc. R Soc. B Biol. Sci.
275:649–659.
Weis, A. E. 2018. Detecting the “invisible fraction” bias in
resurrection ex-
periments. Evol. Appl. 11:88–95.
Weis, A. E., S. M. Wadgymar, M. Sekor, and S. J. Franks. 2014.
The shape
of selection: using alternative fitness functions to test
predictions for
selection on flowering time. Evol. Ecol. 28:885–904.
Associate Editor: J. Anderson
Handling Editor: M. Servedio
Supporting Information
Additional supporting information may be found online in the
Supporting Information section at the end of the article.
Figure S1. Schematic overview of the split-plot design of the
resurrection experiment.
Figure S2. Relationship between time to first flowering and a)
stem diameter at first flowering, and b) WUE.
Table S3. Wilcoxon signed-rank tests comparing the Piv of
onset of flowering and stem diameter at flowering between
populations and generations
(PopYear).

B R I E F C O M M U N I C AT I O N
doi:10.1111/evo.12833
Increased susceptibility to fungal disease
accompanies adaptation to drought
in Brassica rapa
Niamh B. O’Hara,1,2,3,4 Joshua S. Rest,2 and Steven J.
Franks3
1Jacobs Technion-Cornell Institute, Cornell Tech, New York,
New York 10011
2Department of Ecology and Evolution, Stony Brook
University, Stony Brook, New York 11794
3Department of Biology, Fordham University, Bronx, New York
10458
4E-mail: [email protected]
Received November 20, 2014
Accepted November 23, 2015
Recent studies have demonstrated adaptive evolutionary
responses to climate change, but little is known about how these
re-
sponses may influence ecological interactions with other
organisms, including natural enemies. We used a resurrection
experiment
in the greenhouse to examine the effect of evolutionary
responses to drought on the susceptibility of Brassica rapa

plants to a
fungal pathogen, Alternaria brassicae. In agreement with
previous studies in this population, we found an evolutionary
shift to
earlier flowering postdrought, which was previously shown to
be adaptive. Here, we report the novel finding that postdrought
descendant plants were also more susceptible to disease,
indicating a rapid evolutionary shift to increased susceptibility.
This was
accompanied by an evolutionary shift to increased specific leaf
area (thinner leaves) following drought. We found that
flowering
time and disease susceptibility displayed plastic responses to
experimental drought treatments, but that this plasticity did not
match the direction of evolution, indicating that plastic and
evolutionary responses to changes in climate can be opposed.
The
observed evolutionary shift to increased disease susceptibility
accompanying adaptation to drought provides evidence that
even
if populations can rapidly adapt in response to climate change,
evolution in other traits may have ecological effects that could
make species more vulnerable.
K E Y W O R D S : Alternaria brassicae, drought, flowering
time, rapid evolution, resurrection approach.

Ongoing changes in climate, including warming and altered pre-
cipitation, have been increasing drought severity and frequency
over the past 50 years (IPCC 2014). These changes in climate
are
having widespread effects in many populations, including rapid
evolutionary shifts in traits such as phenology (Parmesan and
Yohe 2003). Studies in the emerging field of eco-evolutionary
dynamics suggest that such rapid evolutionary responses may
po-
tentially influence ecological interactions (Pelletier et al. 2009;
Dargent et al. 2013), but our understanding of the ecological ef-
fects of evolutionary responses to climate change remains
limited.
Although adaptive evolutionary responses could potentially help
populations cope with climate change (Penuelas and Filella
2001;
Hoffmann and Sgro 2011; Franks et al. 2014), these benefits
could
be lost if the evolutionary shifts also result in increased
suscepti-
bility to predation or disease. This would be particularly likely
if

there are trade-offs between responses to increased abiotic
stress
and the ability to defend against natural enemies. Assessing the
ecological effects and potential costs of climate change
adaptation
is critical for making predictions about the full effects of
climate
change.
Here, we examine the evolution and plasticity of multiple
plant traits and the potential ecological effects of adaptation in
a
unique system in which a rapid evolutionary response to a
change
in climate has been demonstrated. Prior research found a rapid
adaptive evolutionary shift to earlier flowering in response to a
change in climate (drought) in southern California populations
of the annual plant Brassica rapa (Franks et al. 2007; Franks
and Weis 2008). In the current study, we investigate whether
2 4 1
C© 2015 The Author(s). Evolution C© 2015 The Society for the
Study of Evolution.
Evolution 70-1: 241–248

this adaptive evolutionary change may have ecological effects
by exploring plant susceptibility to a fungal pathogen. Altered
susceptibility is particularly likely because the early flowering
plants, which are able to escape drought, allocate resources to
rapid growth and development (Franks 2011), potentially
leaving
fewer resources available for defense. We hypothesized that
post-
drought descendant plants would show greater disease suscepti-
bility than predrought ancestral plants, with the drought causing
the evolution of increased susceptibility as a byproduct of selec-
tion for earlier flowering. We focused on defense against a
fungal
pathogen, Alternaria brassicae, which was commonly observed
in our field sites with 21.9% (±21.1) to 35.2% (±32.7) of B.
rapa tissue in quadrats sampled displaying symptoms (O’Hara
et al. 2016). This fungus is also known to have important effects
on both wild populations and agricultural varieties of crucifers

(Tewari and Conn 1993). We used a resurrection approach
(Franks
et al. 2008), measuring pathogen response in ancestral
predrought
(seeds field-collected in 1997) and descendant postdrought
(seeds
field-collected in 2004) B. rapa populations grown under the
same
conditions in the greenhouse. We conducted a full-factorial ex-
periment with the two plant populations (ancestral and descen-
dant), two levels of drought treatment (well watered and
drought
stressed), and two levels of fungal inoculation (inoculated and
noninoculated control). We assessed phenotypes including flow-
ering time, disease susceptibility, and specific leaf area (SLA)
to
determine how adaptive evolution in response to drought in B.
rapa affected disease susceptibility.
Methods
STUDY SYSTEM
The plant-pathogen system we used is the foliar fungal pathogen

A. brassicae, which causes Alternaria black spot in its host B.
rapa
L. (Brassicaceae, field mustard) (Conn et al. 1990). A. brassicae
is a necrotrophic fungus that causes damping off, leaf spots, de-
foliation, and reduced seed yield in B. rapa (Tewari 1991;
Koike
et al. 2006). Brassicas have multiple lines of defense against
Al-
ternaria fungi, including a waxy cuticle that forms a barrier to
invasion (Tewari and Skoropad 1976) and induced defenses
upon
successful invasion, governed by multiple genes, including phy-
toalexins, which may impart partial, but not total resistance to
the
disease in B. rapa (Nowicki et al. 2012). Because B. rapa is an
important crop species (bok choi, napa cabbage, oilseed, turnip,
polish canola), its response to this costly and destructive
pathogen
has been extensively studied in agriculture (Rotem 1994; Meena
et al. 2010).
B. rapa PROPAGATION

Previous to this study, a large number of seeds (>10,000) were
collected from ripened seedpods (siliques) along a transect in a
natural population of B. rapa located on the University of Cal-
ifornia Irvine campus in May of 1997 (ancestors) and June of
2004 (descendants). The temporally distinct ancestral and de-
scendant populations are hereafter referred to as our
populations.
Plants were grown for a generation (about 90 days) and crossed
within-population under greenhouse conditions to reduce mater-
nal effects (Franks et al. 2007). These F1 plants were crossed
within-population prior to this study to reduce storage effects
and
the F2 (refreshed) generation was used in the current study. For
all
crosses, at least 500 plants per population were crossed at
random
once they started flowering, using a feather to transfer pollen,
and
visiting each plant at least two times every 3 days.
A. brassicae CULTIVATION
B. rapa tissue infected with A. brassicae was collected from

Bodega Bay, California. This collection site is distant (702 km)
from our natural Brassica populations to avoid the potential
issue
of coevolution affecting differential disease susceptibility in an-
cestral versus descendant plants. A. brassicae fungal spores
were
isolated from the plant tissue and identified by the Oregon State
University Plant Clinic. Spore plugs were grown on carrot dex-
trose agar plates for one week followed by a week on carrot
agar
plates under 12 hours of light and 12 of dark to encourage
sporula-
tion. Fresh spores were collected the day of inoculation,
strained
through gauze to remove hyphae, and adjusted to a concentra-
tion of 1 × 106 spores/ml in 0.05% Tween. All fungal work was
conducted in sterile conditions, and was permitted under APHIS
license #P526P-11-00130.
EXPERIMENTAL DESIGN
Using B. rapa F2 seeds, we conducted a greenhouse experiment
(from February 18th to June 17th, 2012) growing 288 ances-

tral and 288 descendant plants from seed. Both populations were
subjected to a full-factorial combination of a pathogen
treatment
(mock inoculated or inoculated with spores) and a drought treat-
ment (well watered or drought stressed).
For cultivation, seeds were planted individually in separate 8
× 8 × 13 cm pots filled with Sunshine Mix #1 growth media
(Sun
Gro Horticulture, Vancouver, BC, Canada), with 1.4 g of slow
re-
lease 14-14-14 Osmocote fertilizer and supplemented with Mir-
acle Gro All Purpose 20-20-20 fertilizer weekly during watering
(3.0 g/l) (Scotts, Marysville, OH, USA). To avoid room position
effects, plants were moved in blocks among randomized coordi-
nates in the greenhouse every 5 days. Blocks were small (about
eight plants) and included both populations. Inoculated plants
were kept separate from control plants to avoid cross-infection.
Light hours were gradually lengthened from 12 to 14 hours to
mimic the growing season. Because B. rapa is self-
incompatible,

plants were hand pollinated between randomized pairs of plants
every three days, once they started flowering. All open flowers
2 4 2 EVOLUTION JANUARY 2016
were pollinated. All plants were watered daily to saturation for
two weeks to allow establishment. After two weeks, we began
the
drought and inoculation treatments.
Plants that received a drought treatment were watered to
saturation every 4 days. Plants that did not receive the drought
treatment continued to be watered to saturation daily. Soil mois-
ture was monitored using a Field Scout TDR 100 Soil Moisture
Meter (Spectrum Technologies). The moisture level in the soil
of
drought treated plants was significantly lower than well-watered
plants (wet = 28.57% soil moisture (±0.30), dry = 21.55% soil
moisture (±0.22), F = 351.7, p < 0.001). This drought treatment
was designed to mimic field conditions based on field observa-
tions and precipitation records of the study population, which
was

characterized by a wet period pre- and immediately
postgermina-
tion followed by limited precipitation, rather than a sudden stop
in rain (Franks et al. 2007). Using this study design, plants were
kept alive but also experienced a drought treatment by
infrequent
watering.
Plants were inoculated with A. brassicae by wounding 2-
week-old leaves (one leaf per plant and two wounds per leaf)
with
a sterile pipette tip and placing 10 µl of a fresh spore solution
on
each wound. Control plants were wounded and treated with 10
µl
of 0.05% Tween. We wounded the leaves to inoculate our plants
because A. brassicae enters leaves through wounds as well as
through stomata and by enzymatically degrading the cuticle and
cell wall and forming specialized penetration structures
(Tsuneda
and Skoropad 1978). Immediately following inoculation, plants
were kept at 90% humidity for 3 days and then placed at

ambient
humidity and either well watered or drought treated (as
described
above). High humidity following inoculation is standard
protocol
in plant pathology studies because it is known to encourage
spore
germination. These conditions also mimic field conditions for
our study population that experience more moisture early in the
growing season and a high incidence of A. brassicae infection
(O’Hara et al. 2016).
TRAIT MEASUREMENTS
Host susceptibility was assessed in terms of disease severity,
with
plants showing greater damage scored as more susceptible. The
disease severities of the leaves for a subset of 277 randomly se-
lected plants, including both noninoculated control and
inoculated
plants, were scored 21 days postinoculation, using a visual
index
(Fig. S1) that ranged from 1 to 10 based on the amount of
chloro-

sis and necrosis (Buchwald and Green 1992). Generally, disease
severity scores were independently verified by two researchers
who were blinded to whether they were assessing ancestral or
descendant plants. Infected leaves displayed a highly significant
increase in disease severity (one-way ANOVA comparing inoc-
ulated vs. control plants: inoculated mean = 4.62 (±0.20), non-
inoculated mean = 3.64 (±0.16), F1,117 = 41.34, p < 0.001),
demonstrating the efficacy of this treatment.
We quantitatively validated our visual index and the efficacy
of the inoculation with a detached leaf assay of 50 leaves. Prior
to
inoculation, fully expanded leaves were detached from plants
and
placed in petri dishes on filter paper premoistened with distilled
water and inoculated following the same procedure previously
described. Four days postinoculation, leaves were cleared,
stained,
and visualized through a microscope. Leaves were cleared using
a 1:3 acetic acid to ethanol solution and shaken overnight at a
low speed, followed by a 1:5:1 acetic acid, ethanol, and glycerol

solution. After rinsing in water, leaves were boiled for 3
minutes
in a solution of 5% Parker black ink and distilled white vinegar,
and then destained using water that was acidified with a few
drops of vinegar, followed by a 5% vinegar wash (Vierheilig et
al.
1998). The number of spores invading leaf tissue was counted at
100× magnification. Infected, stained leaves had an average of
9.5 (±8.7) spores per wound, while uninfected plants were free
of symptoms and spores. We also found that spore counts were
correlated with the disease severity scores (Pearson correlation:
r
= 0.784, p = < 0.001).
Plants were monitored daily and the date of flowering was
recorded for all 527 plants that germinated. The experiment was
conducted until all plants senesced. Specific leaf area (SLA),
the
ratio of the light capturing surface area of a leaf per unit of dry
leaf mass (Milla and Reich 2007), was also measured. SLA is
of-
ten altered in response to stress and is informative about
resource

allocation (Cornelissen et al. 2003). To calculate SLA, the
newest
fully expanded leaf was collected from a randomly selected
subset
of 280 plants 58 days postplanting, scanned, desiccated with
silica
beads, and then weighed. Leaf area in scanned images was mea-
sured using ImageJ (Schneider et al. 2012). SLA was calculated
by dividing the area of each leaf by its dry weight.
DATA ANALYSIS
To determine if evolutionary shifts in plant traits (flowering
time,
disease susceptibility, and SLA) occurred, we compared
ancestral
and descendant plants, following the resurrection protocol
(Franks
et al. 2008). We tested for differences in trait means under all
treatments using a three-way ANOVA, with population
(ancestor
or descendant), pathogen treatment (fungal or mock inoculated),
drought treatment (well watered or drought stressed) and their
in-

teractions as fixed effects. A two-way ANOVA was used to test
for
an effect of population and drought treatment on disease suscep-
tibility since only inoculated plants display disease
susceptibility.
We then tested two hypotheses using one-way ANOVAs: (1)
traits
(flowering time, disease susceptibility, and SLA) evolved in
this
population, which we tested by comparing trait means of
ancestors
to descendants for each trait within inoculation and drought
treat-
ments, and (2) traits displayed plasticity in response to a
drought
EVOLUTION JANUARY 2016 2 4 3
treatment in the greenhouse, which we tested by comparing trait
means of wet to dry treated plants within inoculation treatment
and
temporal populations. One-way ANOVAs were used because
they

are direct tests of these a priori hypotheses. Trait values
(flowering
time, disease susceptibility, and SLA) were dependent
variables,
with each treatment tested (population, inoculation, and drought
treatment) modeled as fixed effects in specific analyses. For all
models of disease susceptibility, only data collected for
inoculated
plants were used. All analyses were conducted on transformed
data (Table S1) using R 3.0.1 stats package (R Core Team
2013).
Results
EVOLUTION OF FLOWERING TIME AND DISEASE
SUSCEPTIBILITY
We found evidence for the rapid evolution of earlier flowering,
with population significantly affecting flowering time (three-
way
ANOVA; Table 1). We also found the same pattern of descen-
dants flowering earlier than ancestors in each treatment, but this
shift was only significant under the well-watered/noninoculated
condition (ANOVA: F1, 131 = 5.278, p = 0.023; Fig. 1A).

We found evidence for an evolutionary shift to greater
pathogen susceptibility (two-way ANOVA; Table 1). When we
analyzed each treatment separately, we also found that descen-
dants were more susceptible to the pathogen than the ancestors
under both well watered/inoculated (ANOVA: F1, 67 = 16.25,
p < 0.001; Fig. 1B) and drought treated/inoculated (ANOVA:
F1, 65 = 5.302, p = 0.025; Fig. 1B) conditions. Thus the
evolution-
ary shift to earlier flowering was accompanied by an
evolutionary
increase in disease susceptibility.
We found evidence for the rapid evolution of increased
SLA (three-way ANOVA; Table 1). We also found the same
pattern of descendants having greater SLA than ancestors in
each treatment, although this was only significant under the
well
watered/noninoculated condition (ANOVA: F1, 69 = 4.160, p =
0.045; Fig. 1C).
PLASTIC RESPONSES TO DROUGHT
We found a plastic response in flowering time to water avail-
ability, based on a significant effect of the drought treatment

O R I G I N A L A RT I C L Edoi10.1111evo.13631Two d.docx

O R I G I N A L A RT I C L Edoi10.1111evo.13631Two d.docx

Recommended

Recommended

More Related Content

Similar to O R I G I N A L A RT I C L Edoi10.1111evo.13631Two d.docx

Similar to O R I G I N A L A RT I C L Edoi10.1111evo.13631Two d.docx (20)

More from vannagoforth

More from vannagoforth (20)

Recently uploaded

Recently uploaded (20)

O R I G I N A L A RT I C L Edoi10.1111evo.13631Two d.docx