Following pollination and fertilization, flowers, or more accurately ovaries, are transformed into fruit in a process known as fruit set. In fruit crops, adequate fruit set is key to attaining high yields for economically profitable production. However, the proportion of flowers that set fruit on a single inflorescence or on a whole plant varies widely. Fruit set in perennial crops such as grapevines can be reduced for many reasons including, for example, abnormal flower development, water or nutrient deficit or excess, pathogen infection, physical injury, insufficient sunlight, or temperatures outside the optimal range (May, 2004; Keller, 2020). Supply of photosynthates, either de novo assimilated in the leaves or remobilized from stored reserves in the perennial organs, is often thought to be limiting for fruit set (Vasconcelos et al., 2009; Ruan et al., 2012). In grape flowers, eggs that are not fertilized within 3–4 days after anthesis will degenerate (Kassemeyer and Staudt, 1981). Successful fertilization requires the sequential events of pollination, pollen germination, and pollen tube growth. In grapevines, pollen germination and pollen tube growth are maximal in the temperature range 25°C–30°C, and temperatures <10°C and >35°C inhibit germination, but even temperatures <15°C slow pollen tube growth too much to permit fertilization before eggs degenerate (Staudt, 1982). Excessively high temperatures also decrease pollen tube growth by reducing auxin production in the pistil (Zhu et al., 2021). Therefore, poor fruit set may result from slow pollen tube growth under both cool and hot conditions.

Earlier work on temperature effects on grapevine reproductive behavior was typically conducted under controlled-environment conditions, where whole plants were exposed to different temperature regimes. For example, in a growth chamber study conducted with small, fruiting cuttings, vines exposed to day/night temperatures of 14/9°C from 4 days after budbreak failed to fully differentiate their inflorescences and did not set fruit (Buttrose and Hale, 1973). Though the time to anthesis decreased as the temperature increased, fruit set was highest in vines exposed to 20/15°C and decreased at higher temperatures down to zero at 38/33°C. In a similar experiment, flowering was delayed and fruit set was lower in fruiting cuttings exposed to 12/9°C for 1 week during early bloom compared with vines held at 17/14°C or 25/20°C (Ebadi et al., 1995a). A more recent growth-chamber study with fruiting cuttings concluded that temperature did not influence bloom phenology, but temperatures of 37/31°C (day/night) and higher induced flower abortion (Merrill et al., 2020). Similarly, fruit set decreased in pot-grown grapevines as the growth chamber day temperature during the bloom period increased from 25°C to 40°C, with all night temperatures set at 20°C (Kliewer, 1977). Exposure of pot-grown vines in growth chambers to 40/25°C (day/night) for 4 days during bloom led to complete flower abortion, but the same treatment imposed at fruit set did not alter reproductive growth compared with the 25/15°C control (Greer and Weston, 2010). However, because the growth-chamber approach simultaneously manipulates both sink and source temperatures, it cannot answer the question whether any differences in fruit set or subsequent berry growth arise as a consequence of temperature-induced differences in sink strength of flowers or berries. Sink strength, defined as sink size × sink activity, describes the capacity of a plant organ to use imported resources for growth, metabolism, or storage (Yu et al., 2015).

Previously, we demonstrated that brief periods of shoot apex (i.e., sink) temperature differences during the budswell, budbreak, and early post-budbreak period of grapevines induce persistent, season-long differences in shoot growth and architecture, and reproductive growth (Keller et al., 2010; Keller and Tarara, 2010). At least some of the temperature-induced differences in shoot growth can be attributed to differences in vascular tissue differentiation in the emerging shoots (Galat Giorgi et al., 2020). Here we tested the effects of differences in inflorescence temperature during the bloom period leading up to fruit set. Our main goal was to determine whether differences in inflorescence temperature, rather than whole-plant temperature, were associated with differences in the degree of fruit set and other reproductive traits. We used a forced-convection, free-air cooling and heating system (Tarara et al., 2000) to manipulate the inflorescence temperature of field-grown grapevines without altering the temperature of nearby source leaves or other plant organs. We hypothesized that cooling an inflorescence would reduce its sink strength and heating an inflorescence would enhance its sink strength. Consequently, we expected the cooled inflorescences to show slower development and lower fruit set. But because lower fruit set results in fewer berries per cluster, it was unclear whether a temporary reduction in sink strength at the onset of berry growth would lead to smaller berries by the time of harvest, or whether there would be compensatory berry growth leading to larger berries. Compensatory berry growth was previously observed with the same grape cultivar grown in the same region in which this experiment was conducted, but that study did not test temperature-induced differences in berry numbers (Keller et al., 2008). In addition, sinks and sources are linked through feedback mechanisms, and a decrease in sink activity might result in sugar accumulation and thus lower photosynthesis in source leaves (Körner, 2003; Yu et al., 2015). Assuming that an alteration in the sink strength would lead to a modification in source activity, we also hypothesized that inflorescence cooling would result in a transient decrease in leaf photosynthesis or an increase in shoot growth. We worked with Cabernet Sauvignon (Vitis vinifera L.), which is the world’s most widely planted wine grape cultivar, and its reproductive performance is thought not to be especially sensitive to unfavorable environmental conditions (May, 2004). We reasoned that if temperature effects were significant in this cultivar, they would likely also be significant in other, more responsive cultivars.

The present study, conducted in a vineyard in arid southeastern Washington, United States, found that the temperature of Cabernet Sauvignon inflorescences during the bloom period may alter flowering phenology, fruit set, and subsequent cluster development independently of the temperature of the canopy. Using targeted manipulation of inflorescence temperatures in canopies exposed to ambient temperatures, we found that warm microclimatic conditions were favorable for fruit set and seed development, which in turn advanced berry development and increased final berry weight. Flowers and berries on cooled inflorescences reached the subsequent phenological stages significantly later than those on heated inflorescences and produced smaller berries. Thus, manipulating the inflorescence temperature not only altered the duration of the bloom period (i.e., time to fruit set) but also had a lasting effect on berry development after fruit set. The time from fruit set to veraison varied relatively little among treatments, suggesting that differences in later phenology arose mostly from differences in the duration of bloom. By contrast, the temperature treatments had no effect on the number and size of the flowers, since the treatments were applied only after the flowers had been fully differentiated (Keller, 2020).

Cooling inflorescences to an average T_max/T_min of ~24/9°C not only lengthened the bloom period by 7 days but also reduced fruit set. However, despite the lower number of berries per cluster resulting from reduced fruit set, those clusters also had smaller berries with lighter seeds and lower sugar content at harvest, indicating that there was no compensatory increase in berry growth or sugar accumulation. In a growth chamber study, fruiting cuttings exposed to 14/9°C (day/night) failed to fully differentiate their inflorescences and did not set fruit (Buttrose and Hale, 1973). Fruit set in that study was highest in plantlets exposed to 20/15°C and decreased at higher temperatures down to zero at 38/33°C. Elsewhere, fruit set, seed number, and subsequent cell division and berry growth were highest when vines were held during bloom at 25/20°C, and decreased at higher temperatures (Kliewer, 1977). Another growth chamber study found no effect on fruit set but a reduction in seed number per berry in vines grown during bloom at 15/10°C compared with 25/10°C or 25/20°C (Ewart and Kliewer, 1977). In our study, inflorescence heating to an average T_max/T_min of ~32/14°C shortened the bloom period by 7 days but did not reduce fruit set. Except in the cooled inflorescences, fruit set in our study was similar to that observed previously for Cabernet Sauvignon in the same vineyard (Keller et al., 2010) and in different Australian wine regions but lower than in eight of the 10 cultivars tested in the Australian study (Dry et al., 2010). Differences in reproductive performance and environmental plasticity among grape cultivars is one possibility for differences among experiments. However, contrary to the cooling treatment, our heating treatment also decreased the variability in subsequent seed and berry development, suggesting that the high-temperature regime used here was close to the optimum for reproductive growth of Cabernet Sauvignon.

The fact that our experiment was conducted with a single grape cultivar in a single season limits the conclusions that can be drawn from our results. For example, this study could not identify an upper-temperature limit during bloom time, likely because the ambient T_max of inflorescences exceeded 35°C on only three consecutive days (peaking at 36.3°C) starting 4 days after the heating regime was terminated. A recent review of high-temperature effects on plant performance concluded that shortening the time to fruit set due to moderately high temperatures reduces seed and fruit weight (Zhu et al., 2021). Evidently, the temperature regime imposed by our heating treatment was not high enough to impair these reproductive traits. Despite shortening the bloom period, the moderately high temperatures applied in this study did not conclusively reduce seed weight and had no effect on final berry weight. The temperature response of Cabernet Sauvignon reproductive growth observed here resembles that of the subtropical crop tomato (Solanum lycopersicum L.) rather than that of the temperate (cereal) crops reviewed in Zhu et al. (2021). Even higher temperatures, however, may be detrimental to grapevine reproductive performance. For example, temperatures above 33/27°C (day/night) led to flower abortion and poor fruit set in growth chamber experiments (Buttrose and Hale, 1973; Greer and Weston, 2010; Merrill et al., 2020). Increasing the canopy temperature by 2°C–4°C above ambient during bloom in the field also reduced fruit set, and fruit set was lower in the warmer (T_max ≈ 33°C) of the 2 years studied (Pagay and Collins, 2017). The diurnal temperature range reached under field conditions in our study (15°C–20°C) was greater than that imposed in earlier growth-chamber experiments (3°C or 5°C in Ebadi et al., 1995a,b, 5°C or 12.5°C in Buttrose and Hale, 1973; 6°C in Merrill et al., 2020; 10°C or 15°C in Greer and Weston, 2010) but similar to other field experiments (Pagay and Collins, 2017).

Our forced-convection, free-air cooling and heating system worked well in that the desired temperature patterns were achieved, set points were not exceeded, and differences between the ambient (i.e., non-treated) control and the convective (i.e., blower) control were insignificant. The system thus served as an effective tool to manipulate the inflorescence microclimate, just as it had previously been used to manipulate the cluster microclimate during fruit development and ripening (Spayd et al., 2002). Moreover, the temperature regimes applied to inflorescences did not alter shoot growth, leaf gas exchange, and leaf non-structural carbohydrates. Gas exchange and carbohydrate measurements were conducted on the leaf above each treated inflorescence, because during bloom time this leaf is a main source of photosynthates for the inflorescence below it, and is thus most likely to be influenced by source–sink feedback networks (Keller, 2020). Consequently, the manipulation of sink strength by our temperature regimes did not modify the strength of nearby sources or the strength of other sinks (e.g., growing shoot tips). In addition, the reduction in fruit set, as well as seed and berry size following inflorescence cooling was not a result of limited photosynthate supply from the leaves but instead was a consequence of temperature acting on the flowers themselves. In growth chambers, low temperatures (12/9°C) just before and during bloom were found to impair embryo sac development in grape ovaries, possibly leading to an increase in floater seeds with aborted embryo and endosperm (Ebadi et al., 1995a,b). In our field experiment, however, there was no evidence that the number of seeds per berry or their viability was influenced by inflorescence temperature; the proportion of floater seeds was the same across all treatments. The proportion of floaters was similar to that reported for Chardonnay seeds by Ebadi et al. (1995a, 1996) but much higher than in a previous study with Cabernet Sauvignon, yet the seed number and seed weight were similar to that study (Hall et al., 2011). Both pollen germination on the stigma and the rate of pollen tube growth are also temperature-dependent; at <15°C these processes often become too slow for fertilization to occur (Staudt, 1982; Ebadi et al., 1995b). In some cases, pollen was sterile when inflorescences experienced temperatures <15°C at the beginning of bloom, even when photosynthate supply was abundant (Koblet, 1966). On the other hand, pollen tube growth is readily reactivated when warm temperatures follow a brief (e.g., 2 days) episode of cool temperatures (Staudt, 1982). In our experiment, the temperature of cooled inflorescences fluctuated on a diurnal and longer-term basis but was not consistently below 15°C, and thus, fruit set was reduced but not prevented entirely. In addition to the processes related to fruit set, inflorescence cooling likely also reduced cell division and cell expansion in the resulting berries, thus decreasing their final size. Both of these processes occur simultaneously during the first few weeks after anthesis (Harris et al., 1968), and both are limited by cold stress (<15°C) and heat stress (>35°C; Hale and Buttrose, 1974; Kliewer, 1977). We did not evaluate the number and size of mesocarp cells here, but because flowering occurs asynchronously across an inflorescence, the extended bloom period in the cooling treatment implies that berries resulting from early fruit set within a cooled inflorescence continued to be exposed to low temperatures after fruit set until the last flowers set fruit and cooling was terminated.

While the impact of inflorescence cooling on berry growth (i.e., cell division and cell expansion) decreased berry weight independently of berry number per cluster and seed number per berry, the final berry weight also correlated with both seed number and total seed weight across treatments. The correlation between seed number or weight and berry size is well established and is related to the seeds releasing growth-promoting auxin and gibberellin after fruit set (Winkler and Williams, 1935; Olmo, 1946; Keller, 2020). While only a single seed is required for a grape berry to develop, the total seed weight correlates strongly with the growth rate and final size (i.e., sink strength) of mesocarp cells and hence berries (Friend et al., 2009). Additionally, the correlation between final rachis length and flower but not berry number observed here supports the idea that developing flowers, much more so than berries, determine rachis growth and development (Theiler and Coombe, 1985; Gourieroux et al., 2016). Whether the connection between flower number and rachis growth is causally related to the occurrence of a ripening disorder named bunch-stem necrosis (BSN) remains unknown. An inverse correlation between the average bloom-time T_max and the incidence of late-season BSN has been reported for several grape cultivars (Theiler and Müller, 1987). However, although Cabernet Sauvignon is generally susceptible to BSN (Hall et al., 2011), we found no symptomatic clusters in the present study.

Differences in phenology, berry numbers, and berry size following inflorescence temperature manipulation did not alter shoot growth. Apparently, these shoots were not sink-limited. In earlier work with field-grown grapevines, reproductive growth was positively correlated with vegetative growth within shoots (Keller et al., 2015). Nonetheless, many wine industry members believe that late harvest or delayed fruit ripening will slow down shoot maturation and compromise the cold hardiness of overwintering buds. Our results do not support this assumption. Counting the number of brown internodes per shoot showed that the degree of periderm formation (often incorrectly termed “lignification”) was independent of fruit development. The timing of veraison (i.e., of ripening initiation) did not coincide with the formation of periderm across treatments. The former was strongly altered by the temperature regimes applied during the bloom period, whereas the latter occurred simultaneously across treatments, which shows that ripening initiation and periderm formation are independent processes. This result is similar to the finding that veraison does not coincide with the transition of grapevine buds from paradormancy to endodormancy (Camargo-Alvarez et al., 2020). In both studies, the timing of veraison was highly responsive to temperature, whereas periderm formation and bud dormancy seemed to be driven mostly by photoperiod.

Grape growers have several options for influencing the microclimate of an inflorescence or cluster. Examples include the choice of training system, trunk height, vineyard floor management (e.g., control of weeds or cover crops), and canopy management (e.g., manipulation of the number of leaves around the clusters). Different training systems result in differences in canopy density and shoot orientation, which influences the microclimate around the clusters (Reynolds and Vanden Heuvel, 2009). Under the climatic conditions of the present study, a vertical temperature gradient of up to +4°C m⁻¹ may exist upwards from the cooler trunk base (Peña Quiñones et al., 2019). An inflorescence that grows close to the ground is therefore likely to be exposed to cool air more often than is an inflorescence at a greater distance from the ground. Additionally, bare soil can be a source of heat transfer to the inflorescences by convection, especially during sunny days and when the soil surface is dry. Thus inflorescences are likely to be cooler in vineyards with a floor cover of weeds or cover crops. Removing leaves to expose clusters to solar radiation will also result in higher daytime (but somewhat lower nighttime) temperatures of inflorescences. However, leaf removal at or before the time of bloom eliminates source area and generally reduces fruit set (Vander Weide et al., 2021). Shoot thinning (i.e., removal of entire shoots) early in the growing season also reduces the density of the canopy and thus probably contributes to a warmer microclimate around the clusters (Wang et al., 2019).

The forced-convection, free-air cooling and heating system is a useful research tool to investigate sink/source interactions and to study effects of organ temperature independently from those of light. By manipulating grape inflorescence temperatures, rather than whole-plant temperatures, we found that sink strength and hence fruit development is dependent on the temperature acting directly on the sinks themselves. Our results also showed that temperature differences perceived by flowering structures before fruit set can alter the subsequent phenological development of the fruit. Cooling inflorescences from the beginning of bloom lengthened, and heating inflorescences shortened, the bloom period, thus delaying (cooling) or accelerating (heating) the time to fruit set as well as the time to ripening onset. Among grapevine yield components, both berry number per cluster and final berry weight were reduced when the inflorescences experienced cool temperatures. At harvest, berries developing from cooled inflorescences also had lower total seed weight and sugar content, and higher acidity. Whereas daily average temperatures above 16°C appear to be necessary during the bloom period to maximize fruit set and subsequent berry development in Cabernet Sauvignon grapevines, no upper-temperature limit was found in this study. Though our highest temperature regime (daily average 23.4°C) did not modify fruit set and berry traits compared with the ambient regime (daily average 21°C), it did reduce the time to fruit set and veraison, and decreased the variability in subsequent berry development and seed weight, suggesting that this regime was close to the optimum for reproductive performance. The applicability of these results to other grape cultivars, especially cultivars that differ in their sensitivity to temperature variation, remains to be tested.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

MK conceived of the study, obtained the funding, and wrote the manuscript. JF and JT designed and operated the free-air cooling and heating system. RS-B, JF, and LM conducted the experiment. RS-B and MK analyzed the data. All authors contributed to the article and approved the submitted version.

This research was funded by the USDA Northwest Center for Small Fruits Research (grant number 5953582718), Washington State Grape and Wine Research Program, and Walter Hochstrasser Stiftung.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.