Journal of Experimental Botany, Vol. 48, No. 306, pp. 101-111, January 1997
Journal of
Experimental
Botany
Response of ovule development and post-pollen
production processes in male-sterile tomatoes to
chronic, sub-acute high temperature stress
Mary M. Peet 13 , D.H. Willits2 and R. Gardner1
1
Department of Horticultural Science, North Carolina State University, Raleigh, NC 27695-7625, USA
Department of Biological and Agricultural Engineering, North Carolina State University, Raleigh,
NC 27695-7625, USA
2
Received 3 June 1996; Accepted 14 August 1996
Abstract
In order to determine the effects of high temperature
on ovule development and reproductive processes
subsequent to pollen production, nine day/night temperature combinations were imposed over a 9 month
period as four separate experiments, each with three
treatments, including one common treatment. In order
to eliminate known effects of high temperatures on
pollen production and stylar position, high temperature
treatments were applied only to male-sterile tomatoes
(Lycopersicon esculentum Mill.). Pollen was obtained
from male-fertile plants given optimal growth conditions. This allowed comparison of mean daily temperatures from 25-29 °C; day/night temperature
differentials (DIFs) of 2, 6, and 10 °C; day temperatures
of 28, 30 and 32 °C at night temperatures of 22, 24,
and 26 C; and night temperatures of 22, 24 and 26 °C
at day temperatures of 28, 30 and 32 °C. Average
weight per fruit and flower number did not demonstrate a consistent pattern of response to high temperature. Other reproductive characteristics (% fruitset,
total number and weight of fruit per plant, and seediness index) decreased as mean daily temperature rose
from 25 °C to 26 °C and from 28 °C to 29 °C. The primary parameter affecting these variables was mean
daily temperature, with day temperature having a secondary role. Thus, in determining reproductive
responses of tomato to temperatures within this range,
day temperature, night temperature and DIFs do not
need to be considered independently of their effect on
mean daily temperature. If this relationship holds true
in other species, and for pre-pollen production pro3
To whom correspondence should be addressed. Fax: +1 919 515 2505.
© Oxford University Press 1997
cesses as well, modelling the effects of projected climate change should be simplified.
Key words: Global climate change, thermotolerance, stress
resistance, ovule development, pollination.
Introduction
Global warming has already affected populations of
alpine species, some of which are not reproducing
(Grabherr, 1994). As discussed by Hall (1992), high
temperatures also reduce yields in many crops, but the
genetics and physiology of heat tolerance in reproductive
tissues have received comparatively little attention.
Temperature increases of 4°C are projected over the next
few decades, potentially reducing the reproductive fitness
of plants in both natural and agro-ecosystems, especially
in tropical and sub-tropical regions. The effects of global
climate change on food availability have been studied
using a number of assumptions, both as to the extent of
climate change and the effect on agriculture. Adams et al.
(1990) concluded that climate change is not a food
security issue. A more recent analysis, however, using
additional data and assumptions (Adams et al., 1995),
while confirming that moderate warming is not a threat
to US agriculture, concluded that specific regions, particularly the southern USA, will experience economic losses.
For the South East, Delta States and Southern Plains,
these loss predictions ranged from 0.54% to 5.35%,
depending on location and climate model used.
Tomato {Lycopersicon esculentum Mill.) is an example
of a species whose productivity in warm-summer areas is
102 Peefetal.
likely to be adversely affected by even slight temperature
increases. Fruitset in tomatoes is reduced when maximal
daily temperatures average above 30 °C and minimal daily
temperatures average above 21 °C (Moore and Thomas,
1952). Tomatoes have the additional disadvantage, in
terms of projected global change scenarios, of being
relatively insensitive to CO 2 enrichment (Behboudian and
Lai, 1994), a component in the models discussed above
which, for most crops, partially compensates for the
predicted adverse effect of higher temperatures.
A number of explanations have been offered for the
poor reproductive performance of tomatoes at high temperatures. These include reduced or abnormal pollen
production, abnormal development of the female reproductive tissues, reduced supply of photosynthates, and
poor production of growth regulators in sink tissues.
Evidence for each is summarized below.
Effects on reproductive structure and function
Reductions in pollen production at high temperatures
have been noted in many crops, including tomatoes. In a
comparison of high temperature-resistant and -susceptible
tomato cultivars, Dane et al. (1991) found that prolonged
periods of high temperature caused drastic reductions in
pollen fertility in most genotypes. In both greenhouse
and phytotron experiments, Peet and Bartholomew (1996)
found that total pollen production and the percentage of
normal grains were reduced in plants grown at high night
temperatures. In tomatoes, in vivo pollen germination is
optimal near 20 °C (Charles and Harris, 1972), but the
decrease in pollen germination from 20-27 °C is much
less dramatic than the decrease in fruitset. At temperatures
above 30 °C, flower formation, pollen grain and ovule
formation, style elongation, pollen germination, fertilization, and seed formation are all adversely affected and it
was not possible to identify a single process as causing
poor fruitset (Dane et al., 1991; Kuo et al., 1979;
El-Ahmadi and Stevens, 1979; Hanna and Hernandez,
1982).
In a series of experiments in Israel, Levy et al. (1978)
attempted to distinguish between effects on male and
female parents. By pollinating low temperature-grown
plants with pollen from high temperature-grown plants
and vice versa, they were able to show that exposing
either parent to high temperatures reduced fruitset, but
overall, microspores were more affected than macrospores. Comparing high temperature-resistant
and
-susceptible cultivars in the field under high temperature
days (maximum 36-39 °C), they also found cultivar
differences in susceptibility, but within each cultivar, the
amount of flower abscission was strongly correlated with
the amount of style exsertion. No fruitset was ever
observed when the style protruded more than 1 mm out
of the antheridial cone. Flower abscission also occurred
in flowers with normal styles, however. In controlled
environment experiments at 33/23 °C, bud drop, reduced
viable pollen and style exsertion occurred in both susceptible and resistant cultivars, but responses were much more
pronounced in the susceptible cultivar. The antheridial
cone split, followed by flower abscission, in the susceptible, but not the resistant cultivar.
Because of this uncertainty as to the causes of reproductive failure at high temperatures, it has been difficult
to breed resistant cultivars. In a comparison of resistant
and susceptible cultivars, Dane et al. (1991) concluded
that breeding for improved fruitset at high temperatures
has been difficult to achieve and suggested more research
was needed on the causes of the wide genotypic variation
in heat tolerance.
Effect of carbohydrate supply on fruitset
High temperatures may reduce the amount of carbohydrates available to support ovule growth. Daytime temperatures over the optima for photosynthesis may decrease
carbohydrates fixed, while at the same time, high nighttime temperatures may increase respiration. Stephenson
(1981) suggests that the rate of abortion of flowers or
fruit represents the plant's assessment of its ability to
support subsequent fruit development. If conditions are
favourable, more fruit will be retained, and if unfavourable, less.
The carbohydrate balance theory of high temperature
abscission is also supported by interactions of light and
temperature effects. As light (and presumably photosynthesis) increases, the deleterious effects of high temperatures decrease (Atherton and Harris, 1986). In summer,
when light is high, buds develop over a wider range of
temperatures than in winter, when light level is reduced
(Calvert, 1969). Kinet (1977) studied the effects of different irradiances and daylengths on the incidence of flower
abortion. Under 8 h photoperiods with 0.26 MJ m " 2 d" 1
(400-700 nm), all flowers in the first inflorescence aborted
at 20 °C. The percentage of flowers that aborted was
reduced by doubling the daily irradiance at the same
daylength or doubling the daylength at the same
irradiance.
Hormonal imbalances
Whatever the reproductive component which most limits
fruitset at high temperature, a secondary, if not primary,
effect is reduced or altered production of growthregulating compounds. Hormonal imbalances have been
suggested as the cause of poor fruitset in tomatoes at
high temperature (Kuo and Tsai, 1984).
It is not clear, however, whether altered hormone levels
are the cause or result of reduced seed set. The developing
seeds are the source of auxins and gibberellins needed for
fruit development (Varga and Bruinsma, 1976) so where
High temperature tomato fruitset
normal pollen and ovule development is inhibited by high
temperatures, it is not surprising that hormone levels are
altered. Kuo et al. (1989) suggested that at high temperatures hormonal changes in the pistil prior to anthesis
may reduce pollen germination and growth. If this were
the case, however, it seems logical to assume that spraying
with synthetic auxin and gibberellin would result in
seeded, rather than parthenocarpic fruit.
Summary and objectives
Although there is a significant literature concerning temperature effects on fruit and seed development, there is
no consensus as to which processes or tissues are critical
to heat tolerance either in tomatoes or in other crops.
This is at least partly because it is not clear how heat
stress should be imposed experimentally. For example,
Iwahori (1965) and Sugiyama et al. (1966) exposed 20 °C
grown plants to 3 h periods of 40 °C temperatures. This
would represent an acute temperature stress, but one of
short duration, an unlikely event under natural conditions. One purpose of the present experiment was to
impose conditions more realistically representing the
stress that plants in warm temperate regions may experience after global warming, specifically a 4°C increase in
mean temperature from a base of 25 °C.
An additional objective was to answer some basic
questions about how temperature stresses are perceived
by plants. These include whether day temperature, night
temperature or mean temperatures are the most important
limitations to reproductive development. Answering this
question will not only standardize subsequent work, but
may also offer some clues as to the mechanisms of high
temperature effects. A further aim of this research was to
determine the importance of high temperature effects
when experienced only by female reproductive tissue. This
was done by applying pollen from male-fertile plants
grown under optimal conditions to male-sterile plants
grown under a range of high temperature stress
conditions.
Materials and methods
Overview
All experiments were conducted in growth chambers located in
the NCSU phytotron. The data presented here were compiled
as a result of nine experimental treatments imposed over a
9 month period as four separate experiments, each with three
treatments (Fig. 1). All possible efforts were made to standardize
conditions in the four experiments, but as an added precaution,
a common 30/24 °C day/night temperature treatment was
included in all four experiments. Data in each experiment could
thus be expressed as a percentage of the 30/24 °C treatment for
that experiment. Taken together, the four experiments allow
comparison of mean daily temperatures of 25, 26, 27, 28, and
29 °C. The effects of day/night temperature differentials of 2, 6,
and 10 °C can also be compared at mean temperatures of 27 °C.
Temperature
103
Treatment*
Day/Night ° C
A
A
32/26
30/26
28/26
32/24
30/24
28/24
A
30/22
( 2 9 ) Mean Daily
Temperature
A
28/22
Day/ night
differential
Expenment #
Fig. 1. Diagrammatic representation of the nine temperature treatments.
Each square represents a day/night temperature treatment combination
used in one of the experiments except for the 30/24 °C treatment (centre
of grid) which was included in all experiments. The number of the
experiment in which the treatment was imposed is shown in a triangle
within the square. Because of the experimental design, it was also
possible to compare different daily mean temperatures (shown in circles)
and day/night differentials (shown in squares).
In addition, the effects of day temperatures of 28, 30 and 32 °C
can be compared at a night temperature of 24 °C, and the
effects of night temperatures of 22, 24, and 26 °C can be
compared at a day temperature of 30 °C.
Standard conditions for all experiments
In all experiments, seeds of a male-sterile line of tomatoes
('NC8288', a mixed population of 50% homozygous male-sterile
plants and 50% heterozygous male-fertile plants, provided by
R Gardner) were germinated at 22 °C for approximately 7 d.
Seedlings were then transplanted into 6 cm diameter peat
Jiffystrips® with holes (Jiffy Products, Shippagan, Canada)
containing a gravel: vermiculite mix ( 2 : 1 , v/v). Jiffystrips were
placed in Jiffy trays No. 625 with holes (Jiffy Products, Batavia,
II) and maintained at 26/22 °C in the NC State Phytotron
greenhouse for 5 weeks or until flowering. In addition to natural
night, greenhouse plants were provided with a long-day lighting
regime consisting of an incandescent light break from 11 p.m.
to 2 a.m. At the start of each treatment, chambers were serviced
and light levels adjusted by replacing bulbs to provide irradiance
of 500 pE m~ 2 s" 1 . Chamber relative humidity varied somewhat
during the course of the day, peaking at night, dropping slightly
as temperatures were raised in the morning, then increasing
again to a daytime peak in the afternoon. In the lowest
temperature treatment (28/22 °C) daytime levels ranged from
86-92% and night-time levels from 94-95%. In the highest
temperature treatment (32/26 °C) daytime levels ranged from
82-90% and night-time levels from 90-96%.
Male-sterile seedlings, which have the 'green-stem' characteristic were separated at the first transplanting from male-fertile
plants which have the normal 'purple-stem' characteristic. Four
104
Peet et al.
to six weeks after the first transplanting (depending on plant
size), male-sterile plants were transplanted into 9 I containers
holding the same mix as the smaller strips. Sixteen male-steriles
were placed into each of the three chambers. Temperature
treatments were initiated after flowering, except for experiment
3 (see below). In all cases, daytime temperatures began at 7
a.m. and night-time temperatures at 7 p.m. Ten fertile plants
were transplanted into 20.3 cm diameter pots filled with the
same mix as the stenles and put in separate chambers which
were maintained at 26/22 °C. Thus no pollen transfer from
fertiles to stenles should have been possible in addition to the
controlled pollinations described below. For fertile plants
'standard' phytotron conditions of cool white fluorescent and
incandescent lighting from 8 a.m. to 5 p.m. were followed by a
light break of incandescent light only from 11 p.m. to 2 a.m.,
simulating a photomorphogenic tong-day photopenod. All
plants were pruned to two main stems, staked, and watered
twice daily with standard phytotron nutrient solution (Thomas
and Downs, 1991). Axillaries were removed periodically. Most
plants were sprayed at least once for aphids and in some
experiments, a calcium chloride spray was applied to reduce
blossom-end rot.
Pollen from the fertile plants was collected on to a glass slide
by vibrating open flower clusters with a commercial pollinating
rod. Pollen was applied to steriles by gently pressing or dragging
the slide against the stigmas of open flowers. After pollinating
one chamber, additional pollen was collected on the slide before
pollinating in the next chamber. Each flower was tagged by day
of pollination. No flower was pollinated more than once. For
most experiments, pollination took place over a 2 week period
and involved at least six pollination dates. In tomatoes, flowers
within a cluster open sequentially, a process that can continue
for 5 or more days so it is not unusual for the last-opened
flowers on lower clusters to be pollinated simultaneously with
the first-developing flowers on upper clusters. Pollination was
continued until at least all flowers on the first and second
clusters of the most-slowly developing plants, i.e. those in the
25 ~C mean daily temperature treatment, had been pollinated,
during which time flowers on the upper clusters (up to the 6th
cluster in the 29 °C temperature treatment) were also pollinated
in order to maintain a normal fruitload on the plant.
After the end of the pollination period, plants were left in
the chamber to continue to develop for several additional weeks
or until it was clear how many flowers on the first two clusters
would develop into fruit. For all experiments, there was a single
harvest which included both green and ripe fruit. Each fruit
was weighed individually and date of pollination recorded from
the tag originally placed on the flower. Dates of pollination of
flowers that had abscised or had not yet developed into fruit
were also recorded. Fruit were cut along the median and relative
seed content evaluated visually on a scale of 0 to 5, with 0
indicating no visible seeds and 5 indicating a large amount of
seed and gel. Even very small fruit were harvested and weighed,
but almost all fruit were over 40 g at the time of harvest.
Since 10% of 'green-stem' plants initially placed in the
treatment chambers are fertile (RG Gardner, personal communication), continued monitoring of the male-sterile population
was necessary during flowering. Anthers are shorter in malesterile plants, and they release little or no pollen when vibrated.
Purported 'male-steriles' observed to produce pollen during the
course of the experiment were not included in the final harvest
data, but were left in the chamber to maintain equal plant
numbers in all treatments.
Experiment 1
Transplants seeded on 24 August were moved into 'Jiffystrips'
on I September 1994. The second transplanting and placement
in growth chambers occurred on 26 September, but temperatures
of 26/22 °C were maintained in all chambers until flowenng
(10 October) at which time, temperature treatments of 28/22
°C, 30/24 °C, or 32/26 °C were imposed in chambers containing
male-sterile plants. Ten male-fertile plants were left at 26/22 °C.
Dates pollinated were October 14, 17, 19, 21, 24, and 27.
Harvest started November 8 and continued through
November 11.
Experiment 2
Transplants seeded on 23 September were moved into
'Jiffystrips' on 3 October 1994. The second transplanting and
placement in growth chambers occurred on 21 November The
next day 10 male-fertile plants were placed in a 26/22 °C growth
chamber and treatments of 32/22, 30/24 or 28/26 °C, were
imposed on the remaining male-sterile plants, representing DIFs
of 10, 6, and 2°C. Male-sterile plants in the highest DIF were
pollinated 28 and 30 November, 7, 9, 12, 13, 16, and
19 December. Plants in the lower two DIF treatments were first
pollinated on December 7. Fruit were harvested on January 12.
Experiment 3
Transplants seeded on 16 November were moved into
'Jiffystrips' on 23 November. On 20 December steriles were
transplanted into large pots and moved into phytotron chambers. Temperature treatments of 32/24, 30/24 or 28/24 °C were
started the next day even though plants did not flower until
January 5. Fertiles were moved into phytotron chambers
maintained at 26/22 °C. Dehumidifiers were also set up in all of
the treatment chambers in this experiment. Some blossom-end
rot was observed in this experiment and the number of daily
waterings was raised to three. Pollination dates were January
11, 13, 17, and 20. Harvest started February 23 and continued
through February 24.
Experiment 4
Transplants seeded on 11 January were moved into peat liners
on 17 January. On 6 March, after transplanting into larger
containers, male-sterile plants were moved into treatments of
30/22, 30/24, or 30/26 °C and male-fertile plants into chambers
set at 26/22 °C. Steriles were pollinated on March 7, 9, 10, 13,
14, 15, 17, 20, 21, and 22. All plants were sprayed with calcium
chloride on 21 and 22 March to reduce the symptoms of
blossom-end rot. Fruit were harvested 19-26 April.
Analysis
Within each experiment, data were analysed using the JMP
statistical package (SAS Institute, Cary, NC). JMP procedures
included linear and quadratic regression, one-way analysis of
variance (ANOVA) and Student's r-tests. Individually harvested
fruit and flower data were combined by cluster. For purposes
of comparing older and younger clusters, statistical analyses
were conducted separately for the first two clusters and for all
clusters. Per cluster data was then averaged over the whole
plant and either regressed against mean temperature or used
for the one-way ANOVA. Percentage fruitset was calculated by
dividing the number of fruit present per cluster by the total
number of fruit, flowers and aborted flowers. Flower number
shown on figures, however, represents only flowers which had
either aborted or were present on the plant at the final harvest,
but does not include fruit. A response surface was created using
Deltagraph™ (Deltapoint) by expressing means generated in
the separate experiments as a percentage of the common
temperature (30/24 = C). Thus, information from all four
High temperature tomato fruitset
experiments could be summarized in a single figure for
each variable.
In addition to this summary information, PROC RGR was
used to compare single parameter regressions of the response
surfaces against average temperature, day temperature, night
temperature, and DIF. In this procedure each dependent
variable was regressed in an independent linear model, i.e. four
separate models were used for each dependent variable. The
probability that the variables did not explain a significant
amount of the variation is given by the T statistic.
Results
Experiment 1
All treatments in this experiment had a day/night temperature differential of 6 °C. Mean temperatures were 25, 27
or 29 °C. Similar trends and levels of significance were
seen for all variables whether all clusters or the first two
clusters were considered. Data for all clusters are presented here. Flower number per plant (Fig. 2) increased
from 4.4 at 25 °C to 7.5 at 27 °C and 8.5 at 29 °C
(differences between 25 °C and 27 °C not significant using
Student's /-test). Both linear and quadratic regressions of
flower number on temperature were significant.
All other characteristics (Figs 2, 3) associated with
fruit production were either not significantly affected by
high temperature (average fruit weight) or decreased with
increasing mean temperature. The number of fruit per
plant decreased from 6.75 at 25 °C to 1.25 at 29 °C
(differences between 25 °C and 27 °C not significant).
27
105
Predictably, since the number of flowers was increasing
and the number of fruit decreased, fruitset decreased from
48% at 25 °C to 11% at 29 °C (differences between 27 °C
and 29 °C not significant). Total fruit weight per plant
decreased from 679.42 g at 25 °C to 117.32 g at 29 °C,
mainly because of decreased fruit number (Fig. 2). The
fruit also contained fewer seeds, based on a visual rating
(Fig. 3). The seediness index decreased from 0.93 at 25 "C
to 0.21 at 29 °C (differences between 25 °C and 27 °C not
significant). Both linear and quadratic regressions of fruit
number, fruit weight and seediness index on temperature
were significant.
Experiment 2
All treatments in this experiment had a mean temperature
of 27°C with a day/night temperature differential (DIF)
of 2, 6 or 10 °C. Using only the first two clusters, none
of the characteristics analysed differed at the P = 0.05
level of significance for either regression or one way
ANOVA. Considering all clusters (Fig. 4), the effect of
DIF on total fruit weight was significant (P = 0.006), with
lowest fruit weight per plant at a 10 °C DIF (469.1 versus
651.5 and 738.1 for 2°C and 6°C, respectively) The
differences between 2 °C and 6 °C DIF treatments were
not significant. Linear and quadratic regressions of total
fruit weight on DIF were also significant. None of the
other variables were significantly affected by DIF,
although there was a tendency for highest fruit (9.12
versus 7.5 and 6.75) and flower number (13.1 versus 9.25
and 10.62) at 6°C DIF compared to 2°C and 10 °C,
respectively. Average fruit weight also decreased with
increasing DIF (Fig. 4). Thus, lower fruit numbers and
lower average fruit weight both contributed to the significantly lower total fruit weights per plant at 10 °C DIF.
29
Mean Dally Temperature (C)
Fig. 2. Experiment 1: effect of daily mean temperature ( ± standard
error of the mean) on number of flowers, number of fruit, and
percentage fruitset. Flower number did not include flowers that had
developed into fruit, but did include positions where flowers were
formed but subsequently aborted. Means determined by averaging over
eight plants per treatment and over four clusters (25 °C and 29 °C) or
five clusters (27 °C). Linear Rsquare for number of fruit = 0.45,
P^O.0003 [/, = 41.3-1.38.*]; quadratic Rsquare = 0.46, P^0.0017 [/„ =
-72.156 + 7.96.V-O.I56X2). Linear
Rsquare
for
fruitset = 0.43,
/>^0.0005 [/, = 276.3-9.28*]; quadratic Rsquare = 0.47, P = 0.0012
2
[/, = 2249.6-156.x + 2.72.v ).
25
27
29
Mean Datty Temperature (C)
Fig. 3. Experiment 1- effect of mean temperature (±standard error of
the mean) on total fruit weight per cluster, average weight per fruit
(differences not significant) and seediness index. All clusters considered
in analysis. Means determined by averaging over eight plants per
treatment and over four clusters (25 °C and 29 "C) or five clusters
(27°C). Linear Rsquare for total fruit production = 0.53, /> = 0.0001
[/„ = 4195-140.5*]; quadratic Rsquare = 0.55, P = 0.0003
[/,=
2557—18.7*-2.26x!). The linear regression for seediness index was
significant at Rsquare = 0.66, P = 0.0001 [/, = 5.26-0.17*] and the
quadratic Rsquare = 0.72, /> = 0.0001 [ / , = - 2 4 . 5 + 2.04*-0.04* 2 ].
106 Peefetal.
1000
30
32
Day Temperature (C)
Fig. 4. Experiment 2: effect of day/night temperature differential (DIF)
at a mean temperature of 27 °C (± standard error of the mean) on total
fruit weight per plant (g), average weight per fruit (g) and number of
fruit per plant. Regression and ANOVA did not show significant effects
of DIF on the latter two variables, but trends in fruit number per plant
and average weight per fruit contributed to the significant effect on
total fruit weight per plant. The effect of DIF on total fruit weight was
significant overall (/> = O.OO55), but differences between 2°C and 6°C
DIF were not significant, according to Student's (-comparison. Linear
and quadratic regressions of total fruit weight on DIF were significant:
Rsquare = 0 l 7 , P = 0.04 [/, = 756.4-22.8*] and Rsquare = 0.39, P =
0.005 [/, = 475 10.6.t-l l.l.v2] for the linear and quadratic regressions,
respectively.
Experiment 3
In this experiment, all night temperatures were 24 °C, and
day temperatures were 28, 30 or 32 °C, resulting in mean
temperatures of 26, 27, or 28 °C. Although the same
general trend seen in Experiment 1 for more flowers, but
fewer fruit, at higher mean temperatures was seen in
Experiment 3, there were relatively few significant differences, especially considering only the first two clusters.
Because of the tendency (also seen in Experiment 1) for
more flowers and fewer fruit at higher mean temperatures,
percentage fruitset was significantly affected by day temperature (.P = 0.055), with significantly lower fruitset at
32 °C day temperature (25.94%) compared to 54.27% at
30 °C or 45.2% at 28 °C day temperatures which did not
differ significantly in fruitset. The quadratic regression of
fruitset on temperature was significant, but not the linear.
Using data from all clusters, more variables were
significantly affected by temperature (Fig. 5). The linear
regression of flowers per plant on mean temperature was
significant. Comparing means with a one-way ANOVA,
there were significantly more flowers per cluster at 32 °C
day temperature (/> = 0.02) than at the other two temperatures (11.37 versus 7.37 and 7.0 for 32, 30 and 28 °C,
respectively). Fruit number, total fruit weight per plant,
fruitset, seediness index, and average weight per fruit all
tended to decline at the highest day temperature, but only
temperature effects on fruitset were significant. For
fruitset, the mean at 32 °C day temperature (32.3%) was
Fig. 5. Experiment 3: effect of day temperatures of 28, 30 or 32 °C at a
night temperature of 24 °C ( ± standard error of the mean) on flower
and fruit number and per cent fruitset The linear regression of flowers
per plant on mean temperature was significant (Rsquare = 0.19, P < 0.03,
/ , = -24.23+ 1.09.v). Comparing means with a one-way ANOVA, there
were significantly more flowers per cluster at 32 °C day temperature
(P = 0 02) than at the other two temperatures. For fruitset, Student's ttest (/> = 0.05) indicated a significantly lower mean at 32 °C day
temperature (32.3%) than at the other two temperatures. Overall,
however, a one-way ANOVA only indicated a significance level of /> =
0.08 for the effect of day temperature on percentage fruitset.
significantly lower than at the other two temperatures
(56.35% and 51.84% for 30°C and 28 °C, respectively).
Overall, however, a one-way ANOVA only indicated a
significance level of P — 0.0% for the effect of day temperature on percentage fruitset.
Unlike Experiments 1, 2 and 4, in Experiment 3 treatments were imposed several weeks before flowering. This
should only have affected flowers in the first cluster,
however, because in all the experiments once treatments
were started, subsequently formed flowers, i.e. all clusters
above 1, experienced treatment temperatures both during
and after flower development, so really the only difference
in Experiment 3 was that cluster 1 also experienced
treatment temperatures during flower development. More
significant treatment effects were found in the analysis
using all clusters than using only the first two clusters, so
evidently the effects of starting temperature treatments
prematurely were not great. If temperature treatment
during ovule development of cluster 1 was having a strong
effect, more significant differences should have occurred
when the first two clusters were analysed separately. In
any case, since each experiment had an internal control
(the 30/24 °C treatment) and response surface data was
expressed as a percentage of the internal control, the
longer treatment exposure would have affected all treatments equally.
Experiment 4
In this experiment, day temperatures were 30 "C and night
temperatures were 22, 24 and 26 °C. Thus, mean temper-
High temperature tomato fruitset
atures were the same as in Experiment 3, but night rather
than day temperatures were varied. In this temperature
range, night temperature did not affect any of the variables measured (/> = 0.05) when all clusters were considered. However, when only the first two clusters were
considered, similar effects to those in Experiments 1 and
3 were seen (Fig. 6).
When only the first two clusters were considered, the
number of flowers increased significantly as night temperature increased (Fig. 6). This relationship was significant
for both the linear, the quadratic and the one-way
ANOVA. The greatest increase in flower number came
as night temperature increased from 24 to 26 °C, with
plants grown at 26 °C night temperature averaging 6.25
flowers per plant compared to 3.87 and 3.625 for the
24 °C and 22 °C night temperature treatments, respectively, which were not significantly different (P = 0.05).
Again considering only the first two clusters, fruit
number decreased as night temperature increased. This
relationship was significant for both the linear and quadratic regression (and for the one-way ANOVA). The
greatest decrease came as night temperature increased
from 22 °C to 24 °C, with 6.87 fruit per plant at 22 °C,
compared to 4.87 and 4.37 at 24 °C and 26 °C, which
were not significantly different (P = 0.05).
Because flower number was greater and fruit number
smaller, fruitset per cluster was also reduced by higher
1000
24
26
Night Temperature
Fig. 6. Experiment 4: When only the first two clusters were considered,
effect of night temperatures of 22, 24 or 26 °C at a day temperature of
30 °C (±standard error of the mean) on flower and fruit number, per
cent fruitset and total fruit weight per plant. The relationship between
flower number and night temperature was significant for both the linear
( / , = — 11.2 + 0.656.Y) and quadratic ( / „ = l^A-U.W.x
+ OieS.x1)
(Rsquarc = 0.26 and 0.31; P<00\
and 0.02, respectively). The
relationship between night temperature and fruit number was
significant for both the linear (/, = 20.375-O.635.v) and quadratic ( / , =
127.875-9.65.V + 0.1875X2) regression (Rsquare = 0.27 and 0.31; P 0.008
and 0.02, respectively). The linear (Rsquare = 0.32, / , = 211.1 l-6.5.v),
quadratic (Rsquare = 0.33,/,= -404.24 + 45.02.\-1.07.r ! ) and one-way
ANOVA were all significant for the regression of fruitset on night
temperature (/><0.004, / > <0.00l5 and /><0.0015, respectively). Total
fruit weight per plant declined linearly with increasing temperature
(Rsquare = 0.21, /><0.023,/ x = 2425.95-71/l l.v).
107
temperatures. The linear and one-way ANOVA were both
significant for this variable. Percentage fruitset at 26 °C
(40.7%) was significantly lower than at either 24 °C (58%)
or 22 °C (66.7%), which did not differ (P = 0.05).
Total fruit weight per plant declined linearly with
increasing temperature. Student's /-comparison (P = 0.05)
showed a lower mean at 26 °C than 22 °C (585 g plant" 1
versus 870 g plant" 1 ). Total fruit weight at 24°C (700)
did not differ significantly (Z' = 0.05) from either of the
other treatments. The quadratic and one-way ANOVA
were not significant for this variable, however. Average
fruit weight did not differ significantly with treatment, so
the decline in total fruit weight was mainly a result of
fewer fruit at higher temperatures. Seediness index was
also not affected by night temperature. Data from
Experiments 3 and 4 taken together suggest that regardless
of whether day or night temperatures are varied, fruitset
in tomatoes is not very sensitive to temperature increases
within the range of daily mean temperatures from
26-28 °C.
Combined data from Experiments 1-4
When data within each of the four experiments were
expressed as a percentage of the value for the 30/24 °C
treatment, response surfaces could be created for each
variable. Representing these data as part of a response
surface allows easier visualization of the trends in the
data discussed previously. It can easily be seen, for
example, that as mean temperatures increase from 25 to
29 °C, fruit number, total fruit weight, seediness index,
and percentage fruitset all decrease dramatically (Fig. 7af). The sharpest declines were from 25 °C to 26°C mean
temperature and from 28 °C to 29 °C mean temperatures,
with fairly similar values for treatments with daily means
of 26-28 °C. Differences between treatments with the
same mean daily temperature, but different day and night
temperatures were inconsistent and fairly small for these
variables. There seemed to be no particular advantage or
disadvantage to having high day, compared to high night
temperatures, assuming the means were the same. As with
the data in the individual experiments, there tended to be
more flowers at higher temperatures, but the overall trend
was less clear (Fig. 7a). Average weight per fruit showed
no clear trend at all with temperature (Fig. 7e), as was
also seen in the individual experiments. Similar trends
were seen whether all data or only the first two clusters
were considered. Data shown in Figs 7a-f include all
clusters.
PROC RGR confirmed the overriding importance of
daily mean temperatures, but for some of the variables
day temperature was also significant (Table 1). Night
temperature and DIF were not significant for any of the
variables (P<;0.10). For fruit number, fruit weight, seediness index, and fruitset, daily mean temperature explained
o
00
3"
Average Fruit Weight as % 30/24
£t
8
et
SI
8
5/
8
g
SI
8
&
ro
i/y
Total Fruit Wmght as % 30/24
8
5? 8
SI
8
<* 8
&
~*i
en
cp
o
—.
_.
_
/y
tf»
oo
Iff
&
Flower Number as % 3O/24
% Frultset as % 30/24
s
Seediness Index as % 30/24
201
$
SeeOlnass Index as % 30/24
8
Flower Number as % 30/24
$
A verage Fruit Weight as % 30/24
&
% Fruitset as % 3O/24
8
B 8
Total Fruit Weight as % 3O/24
_*
^y
en
Fruit Number as % 3O/24
Sf
8!
8
SI
8
8
B(
g
Fruit Number as % 30/24
8
g
High temperature tomato fruitset
109
Table 1. Probability of significance (Pr>\T\) for the response surfaces for all dependent variables against daily mean temperature,
day temperature, night temperature and DIF
Dependent variable
Daily mean
temperature
Dav
temperature
Night
temperature
Day/night
differential (DIF)
Fruit number
Total fruit weight
Average fruit weight
Seediness Index
% Fruitset
Flower number
0.0020
0.0155
0.1266
0.0079
0.0114
0.1012
0.0423
0 0762
0.2483
0 0582
0.0839
0.1460
0.1226
0.2027
0.3627
0.1725
0.1603
0.4404
0.8128
0.7876
0.8788
0.7832
0.8634
0.6753
the most variation (P<0.02). The second most significant
parameter was day temperature (P<0.085). For average
fruit weight and flower number, none of the parameters
explained a significant amount of variation at the 0.1
level. The order of significance for flower number was,
however, about the same as for most of the of the other
dependent variables (with the sign reversed on the parameter estimates). For average fruit weight, however, the
order of significance was different. If the significance
criteria were dropped to 0.15 for average fruit weight,
day temperature would have been the most significant
parameter.
Discussion
In this experiment, fruit were exposed to high temperatures for both the actual period of setting and until
harvest. Fruit exposed to higher temperatures will develop
faster, making it more difficult to compare fruit weight
than fruit number. For example, Peet and Bartholomew
(1996), noted that seed content and fruit number declined
at the highest night temperature (26 °C), but fruit weight
increased with high night temperatures because of a more
rapid rate of development and the fact that fruit were
harvested before the fruit reached full size. In the present
experiment, mean temperatures were higher than in Peet
and Bartholomew (1996) and harvest was delayed until
most fruit on the lower clusters had reached full size. To
try to isolate rate of development effects from final yield,
data were analysed separated for only the first two clusters
and all clusters. Both sets of data showed the same trends
in fruitset, fruit number and total fruit weight, but in
some cases, different or no trends were seen in average
fruit weight presumably because of the differential effects
of high temperatures on rate of development, length of
development and seedset.
Except for flower production and average fruit weight,
the higher the daily mean temperatures in the range
25-29 °C, the lower all the various indices of reproductive
fitness. In most cases, the decreases were linear. Flower
production, however, increased in some of the higher
temperature treatments, and average fruit weight, was
generally lower at high temperatures, but differences were
not always significant. As discussed above, average weight
per fruit is a difficult variable to interpret in experiments
lasting several weeks because fruit develop more rapidly
at high temperatures, but have fewer seeds. Thus, weight
per fruit during the period of linear increase in fruit
weight is greater, but final fruit size may be less. Decreased
seed content was associated with lower weight per fruit
at maturity in tomatoes (Dempsey and Boynton, 1965).
Abdul-Baki and Stommel (1995) suggest that at high
temperatures, tomato fruit have a tendency to be parthenocarpic. Fligh temperature-resistant lines may be those
that can set fruit even though few, if any, viable seed are
present. The data support this conclusion since the seediness index was much more sensitive to high temperature
than average fruit weight.
The single-factor ANOVAs and regressions did not
allow direct comparison of the importance of day temperature, night temperature, mean daily temperature and
day/night differential, although examination of the
response surface suggested mean daily temperature was
most important, as discussed above. To provide a more
quantitative description of the relative importance of
these factors, a PROC RGR analysis was run. This
analysis also indicated that mean temperatures were most
important, but day temperatures also emerged as significant. Night temperature and DIF did not significantly
affect any of the variables.
Extrapolating mean temperature data from plants
grown at relatively high densities and low light in growth
chambers to field conditions may be unwarranted. Such
an extrapolation might, however, explain why tomatoes
sometimes set fruit in dry climates which have very high
day temperatures and low night temperatures, but experi-
Fig. 7. Response surface for data from all four experiments produced by expressing the means from each treatment in each experiment as a
percentage of the value for the 30/24 °C treatment in that experiment. Thus the 30/24 °C treatment has a value of 100% in all experiments. All data
were calculated on a per cluster basis averaged over all clusters present on the plant, (a) Flower number (value of columns hidden indicated by
parentheses around mean daily temperature), (b) fruit number, (c) percentage fruitset, (d) total fruit weight per plant, (e) weight per fruit (value of
hidden columns indicated by parentheses around mean daily temperature), (f) seediness index.
110 Peefetal.
ence difficulty in climates with more moderate day temperatures and wanner nights. Higher light levels and lower
humidity in desert climates may also be a factor. It is
difficult to separate the two factors experimentally, and
humidity levels were not controlled in these experiments,
varying somewhat between the chambers. Generally, the
effects of relative humidity are small between 50-90%
(Kretchman, 1968) so it is unlikely that there was a
significant effect on experimental results, especially since
all pollen was produced in a separate chamber and handapplied. The role of humidity does, however, deserve
more attention in future studies.
The shape of the response curve to daily mean temperature is also interesting. There was an abrupt drop-off in
most of the characteristics associated with reproductive
fitness as daily mean temperatures increased from 25 °C
to 26 °C. From the range of temperatures 26-28 °C, there
was relatively little effect. From 28 °C to 29 °C, however,
the drop-off was again sharp. Further work should concentrate on physiological changes taking place at these
cardinal temperatures, including reductions in transpiration that might result in disproportionate increases in
leaf temperature. Overall, however, these data suggest
that modelling the effects of global climate change on
yield of heat-sensitive crops can be simplified by the use
of daily means.
The response surfaces represent effects on post-pollen
production processes only in terms of the male gamete.
Probably the affected processes are also post-pollen germination, since a previous study (Peet and Bartholomew,
1996) showed that per cent pollen germination at 26 °C
was not affected by development under night-time temperatures in the range 18-26°C when day temperature
was 26 °C. Abdul-Baki and Stommel (1995) found that
in vitro germination and tube growth of pollen subjected
to 45 °C for several hours differed between genotypes, but
was not related to the overall fruit-setting ability of the
genotype at high temperature. These experiments with
male-steriles suggest that fruitset is reduced at mean
temperatures much above 25 °C even when pollen supply
is adequate. This work supports the conclusion of Kuo
et al. (1979) that poor pollen production is not the sole
cause of poor fruitset at high temperature, but does not
distinguish between effects on ovule development, pollen
germination or the early stages of pollen tube growth,
fertilization, and subsequent development of the embryo
into a seed.
Other physiological factors which may have been affected by treatments include changes in style and stigma
from high temperatures which reduce the chances for
successful pollination. Charles and Harris (1972) reported
reduced stigma receptivity after exposure to high temperature. At least at very high temperatures, post-pollen
germination processes can also be affected by high temperatures. Ovules subjected to a temperature of 40 °C 18 h
after pollination aborted, perhaps due to inhibition of
pollen tube growth and endosperm degeneration
(Iwahori, 1966).
The importance of mean temperature suggests that
carbohydrate reserves may have been depleted over a
24 h period. Dinar et al. (1983) demonstrated that export
of assimilated carbon from a tomato leaf was reduced
under high temperature regimes. A tomato cultivar sensitive to heat stress had lower net photosynthetic rates under
high temperatures than more resistant cultivars (Bar-Tsur
et al, 1985). In potatoes, a heat-tolerant genotype produced a greater amount of sucrose during light-saturated
photosynthesis at 28 °C than a heat-sensitive cultivar
(Basu and Minhas, 1991) and a cultivar susceptible to
low light/high temperature stress had a lower net photosynthetic rate than a more tolerant line in both high and
low light growth environments (Midmore and Prange,
1992).
However, photosynthesis does not seem to explain heat
sensitivity in all cases. Bhatt and Srinivasa Rao (1993)
found that net photosynthesis and growth were higher in
pepper at 27/22 °C than 27/17 °C, but flower and fruit
numbers were higher at 27/17 °C. Similarly, in a study of
cotton boll retention (Heitholt and Schmidt, 1994), assimilate levels (soluble sugars and starch) did not explain
variation in boll retention among eight cotton genotypes.
As suggested by studies in peppers (Turner and Wien,
1994a, b) as well as tomatoes (Dinar and Rudich, 1985a,
b), heat sensitivity may be related to sink strength of the
buds relative to competing growing points rather than
assimilate supply per se.
Conclusions
It has been demonstrated here that an adequate pollen
supply is not sufficient for tomato fruitset at high temperatures. A decrease has been demonstrated in both total
fruit weight and number and seed content as daily mean
temperatures increase from 25 °C to 29 °C which can be
attributed to ovule development and post-pollen production processes. Increases from 25 °C to 26 °C and from
28 °C to 29 °C were particularly detrimental. Day/night
temperature differentials and the specific day and night
temperatures did not seem as important as the mean
terriperatures. The response surface model suggested in
this study needs further testing with a male-fertile system,
as well as with other crops, and in the field. Should the
conclusion of the importance of mean temperature on
reproductive processes be confirmed in this broader context, modelling the effects of global temperature increases
on plant reproduction should be simplified.
Acknowledgements
The North Carolina Agricultural Research Service supported
this study. The hard work and careful attention of Watson Hall
High temperature tomato fruitset
and the Director and staff of the NC State Phytotron are also
gratefully acknowledged.
References
Abdul-Baki AA, Stommel JR. 1995. Pollen viability and fruitset
of tomato genotypes under optimum and high-temperature
regimes. HortScience 30, 115-17.
Adams RM, Rosenzweig C, Peart RM, Ritchie JT, McCarl BA,
Glyer JD, Curry RB, Jones JW, Boote KJ, Allen Jr LH.
1990. Global climate change and US agriculture. Nature
345, 219-24.
Adams RM, Fleming RA, Chang C-C, McCarl BA, Rosenzweig
C. 1995. Economic effects of global climate change on US
agriculture. Climatic Change 30, 147-67.
Atherton JG, Harris GP. 1986. Flowering. In: Atherton JG,
Rudich J, eds. The tomato crop. A scientific basis for
improvement. New York: Chapman and Hall, 167-200.
Bar-Tsur A, Rudich J, Bravdo B. 1985. High temperature effects
on CO 2 gas exchange in heat-tolerant and sensitive tomatoes.
Journal of the American Society for Horticultural Science
110, 582-6.
Basu PS, Minhas JS. 1991. Heat tolerance and assimilate
transport in different potato genotypes. Journal of
Experimental Botany 42, 861-6.
Behboudlan MH, Lai R. 1994. Carbon dioxide enrichment in
'Virosa' tomato plant: Responses to enrichment duration and
to temperature. Hort Science 29, 1456-9.
Bhatt RM, Srinivasa Rao NK. 1993. Response of bell-pepper
(Capsicum annuum L.) photosynthesis, growth and flower and
fruitsetting to night temperature. Photosynthetica 28, 127-32.
Calvert A. 1969. Studies on the post-initiation development of
flower buds of tomato (Lycopersicon esculenlurri). Journal of
Horticultural Science 44, 117-26.
Charles WB, Harris RE. 1972. Tomato fruit-set at high and low
temperatures. Canadian Journal of Plant Science 52, 497-506.
Dane F, Hunter AG, Chambliss OL. 1991. Fruit set, pollen
fertility and combining ability of selected tomato genotypes
under high-temperature field conditions. Journal of the
American Society for Horticultural Science 116, 906-10.
Dempsey WH, Boynton JE. 1965. Effect of seed number on
tomato fruit size and maturity. Journal of the American
Society for Horticultural Science 86, 575-81.
Dinar M, Rudich J. 1985a. Effect of heat stress on assimilate
partitioning in tomato. Annals of Botany 56, 239—48.
Dinar M, Rudich J. 19856. Effect of heat stress on assimilate
metabolism in tomato flower buds. Annals of Botany
56, 249-57.
Dinar M, Rudich J, Zamski E. 1983. Effect of heat stress on
carbon transport from tomato leaves. Annals of Botany
51, 97-103.
El-Ahmadi AB, Stevens MA. 1979. Reproductive responses of
heat-tolerant tomatoes to high temperatures. Journal of the
American Society for Horticultural Science 104, 686-91.
Grabherr G. 1994. Climate effects on mountain plants. Nature
369, 448.
Hall AE. 1992. Breeding for heat tolerance. In: Janick J, ed.
Plant breeding reviews, Vol. 10, 129-68.
Hanna HY, Hernandez TP. 1982. Response of six tomato
genotypes under summer and spring weather conditions in
Louisiana. HortScience 17, 758-9.
Heitbolt JJ, Schmidt JH. 1994. Receptacle and ovary assimilate
111
concentration and subsequent boll retention in cotton. Crop
Science 34, 125-31.
Iwahori S. 1966. High temperature injuries in tomato. V.
Fertilization and development of embryo with special reference to the abnormalities caused by high temperature. Journal
of the Japanese Society for Horticultural Science 35, 379-86.
Kinet JM. 1977. Effects of light conditions on the development
of the inflorescence in tomato. Scientia Horticulturae 6, 15-26.
Kretchman DW. 1968. A preliminary report of several aspects
of fruit setting of greenhouse tomatoes. Research Summary
of the Ohio Agricultural Research and Development Center,
26, 5-8.
Kuo CG, Chen BW, Chou MH, Tsai CL, Tsay TS. 1979.
Tomato fruit-set at high temperatures. In: AVRDC.
Proceedings of the 1st International Symposium on Tropical
Tomato, Oct. 23-27, 1978 at Shanhua, Taiwan, Republic of
China. Asian Vegetable Research Development Center
Publication No. 78-59, 94-108.
Kuo CG, Tsai CT. 1984. Alternation by high temperature of
auxin and gibberellin concentrations in the floral buds,
flowers, and young fruit of tomato. HortScience 19, 870-72.
Kuo CG, Chen HM, Shen BJ, Chen HC. 1989. Relationship
between hormonal levels in pistils and tomato fruit-set in hot
and cool seasons. In: Green SK, Griggs TD, McLean BT,
eds. Tomato and pepper production in the tropics. Proceedings
of the International Symposium on Integrated Management
Practices at Tainan, Taiwan, 21-26 March 1988. Asian
Vegetable Research Development Center Publication
No. 89-317. Shanhua, Tainan, Taiwan, 138^*9.
Levy A, Rabinowitch HD, Kedar N. 1978. Morphological and
physiological characters affecting flower drop and fruit set of
tomatoes at high temperatures. Euphytica 27, 211-18.
Midmore DJ, Prange RK. 1992. Growth responses of two
Solanum species to contrasting temperatures and irradiance
levels: relations to photosynthesis, dark respiration and
chlorophyll fluorescence. Annals of Botany 69, 13-20.
Moore EL, Thomas WO. 1952. Some effects of shading and
para-chlorophenoxy acetic acid on fruitfulness of tomatoes.
Proceedings of the American Society for Horticultural Science
60, 289-94.
Peet MM, Bartholomew M. 1996. Effect of night temperature
on pollen characteristics, growth, and fruitset in tomato
(Lycopersicon esculentum Mill). Journal of the American
Society for Horticultural Science 121, 514-19.
Stephenson AG. 1981. Flower and fruit abortion: proximate
causes and ultimate functions. Annual Review of Ecology and
Systematics 12, 253-79.
Sugiyama T, Iwahori S, Takahashi K. 1966. Effect of high
temperature on fruit setting of tomato under cover. Ada
Horticulturae 4, 63-9.
Thomas JF, Downs RJ. 1991. Phytotron procedural manual.
North Carolina Agricultural Research Service Technical
Bulletin, 244 (revised).
Turner AD, Wien HC. 1994a. Dry matter assimilation and
partitioning in pepper cultivars differing in susceptibility to
stress-induced bud and flower abscission. Annals of Botany
73, 617-22.
Turner AD, Wien HC. 19946. Photosynthesis, dark respiration
and bud sugar concentrations in pepper cultivars differing in
susceptibility to stress-induced bud abscission. Annals of
Botany 73, 623-8.
Varga A, Bruinsma J. 1976. Role of seeds and auxins in tomato
fruit growth. Zeitschrift fur Pflanzenphysiologie 80, 95-104.