Biology Research

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

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

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

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

EVOLUTION DECEMBER 2018 2 6 8 5

E . H A M A N N E T A L .

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

2 6 8 6 EVOLUTION DECEMBER 2018

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

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

EVOLUTION DECEMBER 2018 2 6 8 7

E . H A M A N N E T A L .

-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

2 6 8 8 EVOLUTION DECEMBER 2018

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

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,

EVOLUTION DECEMBER 2018 2 6 8 9

E . H A M A N N E T A L .

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

2 6 9 0 EVOLUTION DECEMBER 2018

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

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;

EVOLUTION DECEMBER 2018 2 6 9 1

E . H A M A N N E T A L .

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

2 6 9 2 EVOLUTION DECEMBER 2018

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

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

EVOLUTION DECEMBER 2018 2 6 9 3

E . H A M A N N E T A L .

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:

2 6 9 4 EVOLUTION DECEMBER 2018

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

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.

EVOLUTION DECEMBER 2018 2 6 9 5

E . H A M A N N E T A L .

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).

2 6 9 6 EVOLUTION DECEMBER 2018