Ovule abortion and seed set of field pea (Pisum sativum L.) grown under high temperature

Abstract In pea, high temperatures during reproductive development lead to severe yield loss. Although the ovule is the seed precursor, studies elucidating the effect of heat on this plant structure are scarce. We investigated the impact of heat in the field and growth chamber on ovules 4 days after the open flower (4DOF) stage. Objectives were to identify associations between ovaries and plant performance, and to evaluate seed set and ovule abortion of heat-treated plants for six cultivars from a diverse range of seed-to-ovule ratios. In the field, plants were seeded at early (control, [early seeded pea, ESP]) and late (stress plant [late seeded pea, LSP]) periods in the season. In growth chambers, plants were exposed to heat (35/18 °C) at early flowering for 4 days and then evaluated at maturity. Stressed plants (LSP) displayed twice as many aborted ovules than ESP during early embryo growth (pro-embryo to globular stage) in synchrony with reduced ovaries, ovules, and embryo sac size. Cultivars with reduced ovary size at 4DOF were related to a high number of reproductive nodes and pods in LSP (r = −0.44 to −0.48). Similarly, under growth chamber conditions, heat caused seed reduction by increasing the abortion of immature ovules (early embryonic stages) at various reproductive nodes. Collectively, our results indicated that pea seed loss from heat in the field is largely due to early embryo abortion, a novel finding, rather than disruption of pre-fertilization events. Compensatory effects on plant performance infer plant resource adjustment. Our findings contribute to the assessment and selection of high-yielding pea cultivars for future warming seasons.


Introduction
An increase in temperature in the environment is a prominent effect of climate change that leads to stress conditions in many agricultural systems worldwide Sun et al. 2019). Indeed, above-normal temperatures cause severe yield reduction in various crops, including chickpea (Cicer arietinum L.; Wang et al. 2006), soybean (Glycine max [L.] Merr.; Djanaguiraman et al. 2013), pea (Pisum sativum L.; Sadras et al. 2013), and wheat (Triticum aestivum L.; Gibson and Paulsen 1999). According to Demming-Adams et al. (2008), plants can partially withstand stress conditions by establishing morphological, anatomical, and physiological changes. However, their productivity is highly compromised in the process, particularly when stress is caused by unexpected elevated temperature (Hedhly et al. 2009) and increased frequency of heat waves. Understanding how rising temperatures affect biological processes such as plant reproduction is crucial to finding new alternatives and to overcome limitations in agricultural production under extreme temperatures.
In field pea (P. sativum L.), a cool-season legume, temperatures above 25 • C can harm its early and late physiological development (Guilioni et al. 2003;Bueckert et al. 2015;Tafesse et al. 2019). For example, the growth rates of pea plants decreased 22% on average under temperature cycles over 30 • C in greenhouse conditions (Guilioni et al. 2003). Certainly, this effect is important in the early physiological processes of plants; however, a major effect is noted in late physiological development, where the reproductive phase takes place (Prasad et al. 2008a;Sánchez et al. 2014). Various reports indicate that an increase in temperatures over 27 • C during reproductive development of pea triggers severe abortion of buds, flowers, young pods, and seeds, and, therefore, yield reduction of the crop (Lambert and Linck 1958;Guilioni et al. 2003). Early studies have suggested that early seed growth (∼6 days after open flower) was one of the most susceptible stages and embryonic development failed between 27 and 31 • C (Lambert and Linck 1958;Jeuffroy et al. 1990). Guilioni et al. (1997) evaluated various temperature stress levels and found that severe stress (33/30 • C) triggered abortion of young flowers and buds. Although various investigations have confirmed the heat susceptibility of pea plants during the reproductive phase, it is still unclear precisely how high temperatures disturb the seed production of the crop.
Male and female gametophytes are the key components of successful seed production during sexual reproduction of flowering plants. The male gametophyte is the pollen grain, and the female gametophyte is the embryo sac within an ovule. Adverse conditions during reproduction can threaten the reproductive process by affecting one or both gametophytes at any developmental stage (Hauser et al. 2006;Sakata and Higashitani 2008). Studies involving abiotic stresses, such as drought, high salinity, and metal toxicity, demonstrate that the female gametophyte can be impaired and aborted in soybean, common bean, and Arabidopsis thaliana (Kokubun et al. 2001;Hauser et al. 2006). For example, salt-stressed plants of Arabidopsis displayed ovule failure prior to fertilization and during embryo formation (Sun et al. 2004). Plenty of research has suggested a detrimental effect of high temperatures on male gametophyte performance in various crops, including legumes (Sakata and Higashitani 2008;Nikolova et al. 2012;Jiang et al. 2015), but unfortunately, the female component under heat stress has been overlooked (Sage et al. 2015). For example, in pea, Petkova et al. (2008) and Jiang et al. (2015) reported that performance of the male component of reproduction can be diminished under high temperatures (>35-45 • C); however, there is absence of research on the ovule, which would allow a thorough understanding of how reproductive development fails in heat stressed crops. Therefore, an assessment of female reproductive structures leading to seed maturation, such as the ovule and its embryo sac, during and after heat can provide insights and more comprehensive information about the susceptibility or robustness of pea plants experiencing high temperatures.
In the field, a common and practical technique to test cultivar performance under high temperatures is to use normal and delayed seeding times within a season. Plants purposefully seeded late are more likely to have their flowering phase disrupted (or affected) when frequent waves of high temperature are expected in summer (Bhandari et al. 2016). Our study aimed to evaluate the effect of high temperature on young ovaries at 4 days after the open flower (4DOF) stage from six pea cultivars that were seeded at early and late periods in the season. The cultivars represented a range of genetically inherited seed sizes and seed-to-ovule ratio (SOR; proportion of ovules developing into seeds per pod). A second objective was to evaluate seed set and ovule abortion at plant physiological maturity in the first four reproductive nodes (RN) of the same cultivars by exposing plants to heat at early flowering under growth chamber conditions. Correlations between traits at plant maturity (including number of pods and RN, and canopy temperature [CT]), measurements from young ovules (4DOF), and the assessments of abortive ovule stages were carried out to detect key associations under field and growth chamber conditions.

Plant material
We used six cultivars of field pea, namely 40-10, Naparnyk, CDC Meadow, CDC Sage, Carneval, and MFR043. These cultivars originated from Europe (40-10, Naparnyk, Carneval) and Canada (CDC Meadow, CDC Sage,and MFR043). Specifically, these cultivars were selected for their range of SOR observed in previous studies (Jiang et al. 2020), where 40-10 and Naparnyk showed high SOR; CDC Meadow and CDC Sage displayed medium SOR; and Carneval and MFR043 exhibited low SOR. Also, the selected cultivars corresponded to two leaf types: the normal or entire leaf (40-10, Naparnyk, and MFR043) and the semi-leafless (CDC Meadow, CDC Sage, and Carneval The six pea cultivars were grown in the field during springsummer 2017 at Sutherland (52 • 10 N, 106 • 30 W), Saskatoon, Saskatchewan. The experiment was implemented with two seeding dates, identified as early (normal conditions) and late (stressed) seeded plots. Early seeded pea (ESP) was sown at the end of April, whereas late seeded pea (LSP) was sown four weeks later. The ESP treatment was considered as the control. In LSP, plant flowering phases were expected to be under greater heat stress since they occurred in mid-July, when daytime air temperatures exceeded 30 • C in the field. The cultivar plots were arranged in a randomized complete block design (RCBD), with four replications for each seeding date. For each cultivar, 360 seeds were sown in plots 0.91 m wide and 4.6 m long. Although fertilizer was not applied, the seeds were inoculated with commercial rhizobia (Rhizobium sp.) to ensure nitrogen fixation by the experimental plants. Weed control was achieved by application of the herbicide Edge (ethalflurafin) plus Pursuit (imazethapyr) prior to seeding (fall), herbicides Viper (imazamox and benzon) 4 weeks after seeding, and Axial (pinoxaden) plus Centurion (clethodim) around 6 weeks after seeding.

Study 2: seed set and abortion under growth chamber conditions
In the growth chamber, we sowed four seeds per pot of each of the same six pea cultivars, using cylindrical pots of 7.6 L filled with peat base mix (Sunshine , RR. Horticulture Canada Ltd., AB., Canada), and 20 g of a slow-release fertilizer 14-14-14 (Nutricote , Brampton, ON, Canada). We set up the photoperiod at 16 h light/8 h dark, the temperature at 24 • C during light (day) time, and 18 • C during dark (night) time, designated the control or normal regime. Light was supplied by banks of cool fluorescent tubes that provided an irradiance of ∼ 450 ± 5 μmol photons m −2 s −1 . After germination, we thinned the plants from four to two plants per pot at the 3-to 4-leaf stage (Jeuffroy et al. 1990). We maintained two plants per pot, but only one was considered the experimental unit and was therefore evaluated at the end of the experiment. We provided complementary fertilization to the plants by applying a half-strength Hoagland nutrient solution (Hewitt 1952) on alternate days, starting from 3 weeks and finishing 6 weeks after sowing. Plant watering was carefully monitored and provided as required to avoid any conditions of drought stress.
To impose the heat treatment, half of the plants growing at 24 • C day/18 • C night that had reached the early flowering stage were transferred to a chamber where the temperature was set at 35 • C day/18 • C night in cycles, for four consecutive days. Plants were considered at the early flowering stage when they displayed opened flowers on RN 1 (stage 0.5, Maurer et al. 1966), flower buds with petals tightly closed on RN 2 (stage 0.2, Maurer et al. 1966), and flower buds covered by their sepals on RN 3. Cycles of temperature treatment were established according to a photoperiod of 16 h light (day) and 8 h dark (night), where 18 • C was maintained during the night and the temperature rose in steps of 3 • C every hour during the day to achieve 35 • C. This high temperature was maintained for 6 h and then dropped by 3 • C every hour until the chamber returned to 18 • C. After four days of heat treatment, the plants were returned to control conditions (24 • C day/18 • C night), where they were kept until the physiological maturity stage. Plants that remained at 24 • C day/18 • C night (the other half of the group) during the whole experiment were considered as controls or plants grown at normal conditions. The temperature of 24 • C during the day is below the stress-threshold of 28-32 • C for pea production in the field and it has been employed as a standard control temperature in prior studies of pea (Lambert and Linck 1958;Jiang et al. 2015;Liu et al. 2019). We established the experiment in an RCBD with a total of 48 pots corresponding to six cultivars, two temperature treatments, and four replications.

Study 1: young ovule failure and plant performance under field conditions
In the field, once plants were in the middle of flowering stage and the air temperature was > 28 • C, two to three flowers at open flower stage were randomly selected and tagged per plot from ESP and LSP. Four days following this open flower stage (4DOF), we collected the pistils from these tagged flowers and immediately fixed them in formalin-acetic acidalcohol (FAA50). On these samples, we measured ovary width and length, ovule length, and embryo sac area. The ovary length and width were determined with a caliper, and then the ovaries were carefully dissected by removing one of the ovary walls but keeping the ovules attached to the suture of the other ovary wall. We assessed the internal structure of the ovules by applying a clearing-staining procedure using Mayer's Hemalum stain, according to Enugutti et al. (2013). In these samples, ovule length, embryo sac area, and status (normal or aborted) were evaluated using a Zeiss Axioplan microscope. The resulting digital images were analyzed with Image J software. We designated ovules as aborted when they displayed embryo sacs with signs of damage, such as a loss of the boundary lining the embryo sac, and collapsed endosperm. Lastly, the proportion of aborted ovules per ovary was determined by dividing the number of aborted ovules by the total number of ovules within the ovary.
To complement our short-cycle stress measurements, we evaluated plant traits, CT, days to flowering (DTF), and flower-ing duration from each plot in the field. We also took end-ofseason yield measurements such as thousand kernel weight (TKW) and seed yield from combine-harvested plots. For the evaluation of plant performance traits, we randomly selected two plants (subsamples) at the end of the season, and their number of vegetative nodes, RN, reproductive nodes with fruit (RwF), aborted fruit nodes (AbFN), and number of pods (NP) were recorded. We measured CT on each plot using a handheld infrared thermometer (Model 6110.4 ZL, Everest Interscience Inc., Tucson, AZ, USA.). CT was measured when plants were flowering and the temperature in the environment exceeded 28 • C. We calculated the number of DTF from the seeding date to when 50% of the plants in each plot displayed opened flowers at the first RN. We also estimated flower duration (days) in a plot from when 50% of plants showed at least one opened flower to when 50% of plants reached terminal flowering. Average seed size was assessed from the mass of 100 seeds per plot, and the average weight of a seed (mg) was multiplied by 1000 for TKW in grams. Finally, we determined the seed yield in g m −2 after each plot was harvested.

Study 2: seed set and abortion under growth chamber conditions
When plants reached the physiological maturity stage (canopy turned green-yellow) in the growth chamber, we collected and measured pods from RN 1 to RN 4. In these samples, the number of seeds, the SOR, the seed diameter, and the frequency of ovule abortion, were assessed. Immediately after collection, we carefully opened the pods and recorded the number of seeds and aborted ovules per pod. The SOR was obtained by dividing the number of seeds by the total number of ovules in each pod. We measured the seed diameter from the hilum to the opposite side of the seed using a digital caliper. Ovules and early seeds that failed to reach a mature seed stage, that is, the embryo did not fill the seed coat, were considered aborted. The proportion of aborted ovules was determined by dividing the number of aborted ovules by the number of total ovules within the pod. We carefully defined the aborted stage of an ovule under a dissecting microscope according to the morphological characteristics of the embryo sac and embryo stage displayed by their aborted structures (Marinos 1970;Briggs et al. 1987). Although the six cultivars evaluated in the field were grown under the growth chamber, we could assess the seed set and abortion on only five of them (40-10, Naparnyk, CDC Meadow, CDC Sage, and Carneval). In fact, one cultivar (MFR043) showed floral variations that limited its seed formation under the growth chamber (data not shown).

Data analysis
We employed a linear mixed model with SAS statistical software (version 9.4, SAS Institute Inc., Cary, NC, USA) for both growth chamber and field data sets. For Study 1, the model included seeding dates and cultivar as fixed effects, whereas replications and their interactions with treatment Bars represent means of four replications (n = 11-12 ovaries) with their standard error. Symbols * , * * , and * * * indicate significant differences at P < 0.05, 0.01, and 0.001, respectively. factors (seeding dates, cultivar) were considered as random effects. For Study 2, the model included treatment, cultivar, RN, pod position, ovule position, and their interactions as fixed effects, whereas replication and its interaction terms were considered as random effects. The Kenwardroger option was used to approximate the degrees of freedom for unbalanced data. Since pea contained a varied number of ovules aligned on the suture within the ovary/pod, ovule position effect within ovary/pod was standardized across cultivars. Three ovule positions were considered: stylar, ovules localized closest to the style; medial, ovules at the medial area within the ovary/pod; and basal, ovules closest to the pedicel end of the ovary/pod (Gutiérrez et al. 1996;Jiang et al. 2017). Therefore, the total number of ovules within an ovary/pod was divided into three. When the number of ovules could not be divided evenly, the maximum difference in the number of ovules between categories was one (Gutiérrez et al. 1996). Associations among variables were determined by employing the Pearson correlation procedure, and the significance of each association was established at P < 0.05. Finally, the relationships between seeding date, young pistil assessments (4DOF), and plant performance in the field were analyzed using principal component analysis (PCA).

Study 1: young ovule failure and plant
performance under field conditions

Weather description
As anticipated, during the growing season from May to August, late seeded pea, the LSP treatment, experienced higher mean temperatures than ESP. The mean seasonal daily temperature for ESP and LSP was 16.4 and 17.8 • C, respectively. Particularly during flowering, the mean daily maximum temperature was 26.6 • C for ESP and 27.1 • C for LSP. Although both ESP and LSP experienced days with temperatures above 28 • C during flowering, the number of days where the temperature exceeded 30 • C was higher for LSP (4 days) than ESP (1 day). Similarly, the cumulative precipitation during flowering differed between both seeding dates. The total rainfall during flowering was 26.5 mm in ESP and 13 mm in LSP.

Ovary length and width at 4DOF
In the field, ovaries 4DOF showed variation in size (Figs. 1A and 1B) across cultivars depending on seeding date (P = 0.04). Ovaries from ESP were longer and wider compared to ovaries of the same age from LSP. Particularly, ovaries on cultivars identified as low and medium SOR (Carneval, CDC Meadow, and CDC Sage) were smaller under LSP (Figs. 1A and 1B). Ovaries of 40-10 and Naparnyk (high SOR) and those of MFR043 (low SOR) did not differ between seeding dates (Figs. 1A and 1B). Regardless of seeding date, 40-10 and Naparnyk had the largest ovaries, whereas CDC Meadow and Carneval the smallest ovaries in this group of cultivars.

Ovule length, embryo sac area, and status at 4DOF
Ovule length and embryo sac area 4DOF were influenced by seeding date (P < 0.0001) and the interaction between seeding date and ovule position (P < 0.05). Ovule lengths from LSP were 27% shorter than those from ESP (Fig. 1D), and the average area of the embryo sac was twice as small in LSP compared with ESP (Fig. 1E). In the ESP, ovules at the medial position within the ovary were larger than those at stylar and basal positions. In contrast, in the LSP, ovules at medial and stylar positions were similar in size but larger than basal ovules (Fig. 1D). Regardless of seeding date, 40-10 had longer ovules and larger embryo sac areas compared to the other cultivars. In general, ovules from the medial position within an ovary were significantly larger, followed by ovules at the stylar and basal positions.
The internal evaluation of the embryo sac of ovules from both ESP and LSP revealed that most (>0.90 per ovary) contained embryos at early growth stages (pro-embryo or globular stages), inferring that successful pollination and fertilization had occurred. Remarkably, the proportion of ovules showing embryo growth varied only among cultivars (P = 0.0053) and not by seeding date or their interactions. In particular, the proportion of ovules displaying embryos ranged from 0.94 to 0.99 per ovary in five out of six cultivars (40-10, Naparnyk, CDC Meadow, CDC Sage, and Carneval).
However, cultivar MFR043 showed a lower proportion of ovules containing embryos (0.79 ovules per ovary), regardless of seeding date.

Ovule failure at 4DOF
A careful assessment of the embryo sac per ovule typically revealed normal growth ( Fig. 2A). However, internal disruption of the embryo sac was evident in some cultivars. The signs of embryo sac damage included collapse of the embryo sac lining (Figs. 2B, 2C), endosperm shrinkage, lack of embryo sac expansion (Fig. 2C), and embryo vestiges at the dome of the ovule in extreme cases (Fig. 2C). The proportion of ovules with these early signs of abortion per ovary varied significantly between seeding dates (P < 0.001) and the interaction of seeding date by cultivar (P = 0.0098). As predicted, the proportion of aborted ovules was significantly greater at LSP, displaying an average of 0.47 aborted ovules per ovary whereas the average in ESP was 0.20 aborted ovules per ovary. Cultivars identified as having low and medium SOR (Carneval, CDC Meadow, and CDC Sage) had greater proportions of ovules with signs of abortion in LSP (Fig. 1C). In general, cultivars identified as having high SOR (40-10 and Naparnyk) had the smallest proportion of aborted ovules per ovary (0.04 to 0.26) in the entire experiment. The proportion of ovules with signs of abortion was more likely at the stylar and basal positions within the ovaries, although this effect was unrelated to seeding date.

Correlation between variables of pistils 4DOF and plant performance
Variables from pistils collected at 4 DOF showed both inverse and positive associations with plant performance in ESP and LSP. For example, ovary size (length and width), ovule length, and embryo sac area were inversely related to the  number of aborted fruiting nodes on plants at maturity, in both ESP and LSP (Table 1).
Furthermore, we observed specific associations in ESP and LSP. The proportion of aborted ovules per ovary (4DOF) was positively related to the number of pods on mature plants from ESP. In contrast, for mature LSP plants, the proportion of aborted ovules per ovary was positively correlated with the number of aborted fruiting nodes. These results indicated that cultivars exhibiting a high proportion of aborted ovules per ovary (4DOF) could produce a high number of pods in ESP but a high number of aborted fruiting nodes in LSP by the end of the season. In contrast, ovary size (length and width) of young pistils (4DOF) was inversely related to the number of RN and pods on mature plants on LSP. As such, cultivars displaying a reduction in ovary size were likely to produce more RN and, therefore, pods in LSP. Finally, ovary size (length and width), ovule length, and embryo sac area were positively associated with CT in LSP (Table 1). Overall, whereas variables evaluated on young ovaries provided relative information about plant reproductive abortion, diverse associations also revealed compensatory effects between young ovary and ovule growth and plant performance traits such as number of RN and mature pods under field conditions.

Relationship between seeding date, pistils 4DOF, and plant performance under field conditions
The relationships between seeding date, young pistils (4DOF), and plant performance variables were also analyzed by ordination analysis (PCA). The first two principal components (PC1 and PC2) selected from the analysis explained 58.8% of the variance in the data. Fig. 3 revealed two main clusters representing ESP and LSP. Variables such as ovary width and length, ovule length, embryo sac area, and proportion of aborted ovules per ovary (4DOF) strongly contributed to the variance explained in PC1 (36.7%). Particularly, ovary width and length and embryo sac area were associated with ESP, whereas the proportion of aborted ovules per ovary was associated with LSP (Fig. 3).
Variables related to plant performance contributed to the variance explained in PC2 (22.1%). In this case, variables such as number of RN, RwF, NP, and seed yield (Y) were highly associated with ESP, which revealed a high plant performance in these plots (Fig. 3). In contrast, CT, located in the opposite direction of plant yield, was associated with LSP. Additionally, two subgroups of cultivars could be differentiated within the LSP cluster. One subgroup identified as high SOR cultivars was associated with CT. The other subgroup represented medium and low SOR cultivars and was related to a high proportion of aborted ovules per ovary (PAO) at 4DOF, and AbFN, in LSP (Fig. 3).

Number of seeds and seed-to-ovule ratio at maturity
High temperature affected the number of seeds per pod and the SOR per pod on some cultivars, depending on their RN (P < 0.0001). Heat reduced seed number and SOR per pod to 30%-40% of the control on Carneval (nodes 1-3), 10%-20% on CDC Sage (nodes 1 and 2), and 20% on CDC Meadow (nodes 2 and 3). In contrast, a 20% increase in seed number and SOR per pod was detected in 40-10 (nodes 3 and 4). As anticipated, high temperature did not affect seed number or SOR compared to the control in Naparnyk (Table 2).

Seed diameter at maturity
Exposure to high temperature for 4 days at the early flowering stage also influenced seed diameter; the effect varied among cultivars (P < 0.0001). Specifically, seeds of 40-10, Naparnyk, CDC Meadow, and CDC Sage were 1.0%-2.7% smaller in diameter on plants exposed to heat stress than control plants. In contrast, seeds of Carneval were 2.7% larger in diameter on heat-treated plants compared to the control (Fig. 4B). The greatest reduction (2.7%) in seed diameter was observed in Naparnyk and the smallest reduction (1.0%) occurred in CDC Sage (Fig. 4B). In general, and regardless of treatment, nodes 1 and 2 had larger seeds than nodes 3 and 4. Among cultivars, 40-10 had the smallest seeds in the group, independent of temperature treatment.

Ovule and early seed abortion at maturity
With regard to failed ovule development, three main categories of abortion were detected. The first category corresponded to ovules < 1 mm, estimated to fail right before or after fertilization (BAF). The second group held ovules of 1 mm to < 3 mm that contained embryos which had aborted either at the globular or heart stages (GH). The third category corresponded to failed ovules of 3-5 mm which had their embryos abort at early or late cotyledonary stages (ELC).
The proportion of aborted ovules BAF per pod increased significantly under high temperature depending alone on cultivar (P < 0.0001) and regardless of RN (P = 0.3740) or ovule position within the pod (P = 0.6913). Among the five cultivars, CDC Meadow and Carneval exhibited up to a 2-to 12fold higher aborted ovules BAF per pod on plants exposed to high temperature compared to plants under control conditions (Fig. 4B). In contrast, the proportion of aborted ovules GH per pod was influenced by the interaction of cultivar, temperature, and node position (P = 0.0249) and the interaction of cultivar, temperature, and ovule position (P = 0.0234). High temperature increased the proportion of aborted ovules GH per pod over threefold in Carneval at nodes 2 and 3 and over twofold in CDC Sage at node 1 ( Table 2). In terms of ovule position within the pod, heat-treated CDC Meadow exhibited more aborted ovules GH at the pod's stylar position (Fig. 5C), heat-treated Carneval at the pod's medial position (Fig. 5E), and CDC Sage at the pod's basal position (Fig. 5D). Interestingly, high temperature did not increase the abortion of ovules ELC (P = 0.970). Regardless of temperature, a pod's stylar and basal positions had the highest proportion of aborted ovules of BAF and GH (0.07-0.31) per pod compared to ovules at a pod's medial position (0.03-0.09).

Correlation of seed set variables
SOR per pod was inversely associated with the three categories of ovule abortion per pod on plants under heat and control conditions (Table 3). We found a stronger association between SOR and aborted ovules BAF in plants exposed to heat stress, compared with plants under control conditions (Table 3). Finally, seed diameter was inversely associated with SOR and positively associated with the three types of ovule abortion per pod on heat-treated plants (Table 3). These last associations indicated that plants showing a reduction in seed size tended to have a high SOR and low ovule abortion in the study.

Discussion
High temperature during reproductive development is a major factor that constrains seed yield in field pea (Guilioni et al. 2003;Bueckert et al. 2015). To investigate the vulnerability of pea to high temperature, our study focused on the female reproductive component (ovules and embryo sacs) of young flowers (4DOF) and mature pods in terms of abortion and seed set after heat exposure. Additionally, plant performance was also investigated. Our results consistently showed that high temperature disturbed particularly various post-fertilization events during ovule development, especially during the early stages of embryo formation.

Effects on ovule and plant performance under field conditions
Seeding later than the recommended period in a field season is a helpful technique for evaluating cultivars under extreme environmental conditions, such as temperature (Kaur et al. 2015). In our study, young ovules and ovaries (4DOF) of cultivars selected for their low and medium SOR Table 2. Effect of four consecutive days of high temperature (35/18 • C) at early flowering stage on seed number, seed-to-ovule ratio (SOR), and proportion of aborted ovules with a globular-or heart-stage embryo (GH) per pod at the first four reproductive nodes of five pea cultivars grown under growth chamber conditions.   (Carneval, CDC Meadow, and CDC Sage) showed more deficient development from LSP compared to those features from ESP plants. Higher temperatures in LSP likely contributed to the disrupted growth of these female reproductive structures, from just 4 d with temperatures exceeding 30 • C compared with 1 day for the ESP treatment in our study. Correspondingly, the proportion of ovules with signs of early abortion, such as damage to the embryo sac perimeter, reduced embryo sac size, and disrupted maturation of endosperm, was increased in ovaries of the cultivars affected by LSP. Variation in precipitation between ESP and LSP could have acted as a confounding factor influencing ovary development in our study, because total rainfall during flowering was double (26.5 mm in ESP; 13 mm in LSP), although both amounts were low. According to Prasad et al. (2008b), when both high temperature and water limitations (drought) occur in the field, as in the LSP treatment here, various photosynthetic pathways and metabolites can be severely impaired, substantially diminishing yield quality and quantity. For example, studies investigating heat and drought stress during seed development in chickpea and lentil found that Rubisco activity, chlorophyll concentration, starch, sucrose and invertase activity in leaves and seeds were moderately compromised under each individual stress but effects intensified under combined stresses, especially in sensitive cultivars (Awasthi et al. 2014;Sehgal et al. 2019). Alternatively, the difference observed in ovary and ovule development between seeding dates could result from the higher precipitation that may have mitigated the effect of high temperatures (28 • C) on ESP Fig. 5. Effect of high temperature (35/18 • C) on the proportion of aborted ovules containing an embryo between the globular to heart stages (GH). Panels show values by ovule position within pods on five field pea cultivars grown in growth chamber conditions. Bars represent means of four replications (n = 16-32 pods) with their standard error. Symbols * , * * , and * * * indicate significant differences at P < 0.05, 0.01, and 0.001, respectively.
performance, resulting in reduced ovule and ovary damage (Kutcher et al. 2010). However, water supply can only mitigate heat stress in pea to a small extent. Tafesse (2018) studying field pea under stress, found that even when plants exposed to heat were kept well watered, they still exhibited a 9 • C increase in leaf temperature, an indication of plant stress, which was only 2.2 • C lower than plants under combined drought and heat stress conditions. Furthermore, we propose that other factors such as light intensity and UB-V radiation outside, not measured in this study but existing during clear sky conditions later in the season (Suzuki et al. 2014;Bal and Minhas 2017;Minhas et al. 2017), could have intensified stress on LSP. Studies using light and UV-B radiation at high intensity have impaired photosynthesis, damaged DNA, and degraded photosynthetic pigments and protein synthesis leading to lower plant growth and productivity in pea, bean, and soybean crops (Deckmyn et al. 1994;Ambasht and Agrawal 2003;Choudhary and Agrawal 2014). Therefore, additional factors such as lack of precipitation, light intensity and UV-B radiation likely exacerbated the stress conditions in our LSP trial. The reduction in seed yield under heat stress mainly has been attributed to a lack of fertilization in various crops due to pollen failure (Sakata and Higashitani 2008;Devasirvatham et al. 2013); however, ovules of pea collected at 4DOF after heat waves (>28 • C) in our field study showed few or no signs of fertilization failure. Indeed, more than 90% of the ovules displayed embryos at various stages of growth (pro-embryo to globular stage) on both ESP and LSP. Interestingly, the inverse correlation detected between ovary size, RN, and pod number on plants at LSP may indicate a mechanism of resource availability adjustment in these plants. Studies in mung bean, lentils, and chickpea show that plants seeded at a later period in the season tend to exhibit lower chlorophyll content, stomatal conductance, water content, and sucrose concentration compared to plants seeded at typical times (Kaushal et al. 2013;Kaur et al. 2015). Similarly, studies on tomato show that high temperature can change assimilate movement in plants, resulting in lower translocation toward young fruits (Dinar and Rudich 1985). Therefore, LSP exposed to higher temperatures (>30 • C) experienced a severe lack of assimilate availability, and consequently, suffered and produced higher embryo abortion despite prior successful fertilization of the megagametophyte.
Our results also demonstrated that some cultivars under high CT had larger ovaries, ovules, and embryo sacs in LSP (Table 1). Commonly, CT is a valuable measurement that indicates plant heat stress; the higher the plant's CT, the greater the level of plant stress (Ayeneh et al. 2002). Curiously, in our study, high CT was related to great expansion of key fruit components (ovaries, ovules, and embryo sacs) which may imply that some cultivars under high stress advanced their fruit formation. Indeed, pea fruits expand when ovary growth is stimulated by ovule development following successful fertilization and embryo sac growth (García-Martínez et al. 1991). In contrast, pea ovary growth ceases in the absence of fertilized ovules or seeds unless a hormonal treatment is provided, but those instances result in parthenocarpic fruits (i.e., ovary growth without seeds; Ozga et al. 2002). In line with this, our study also showed that CT was mainly associated with cultivars featuring a high SOR (40-10 and Naparnyk). Thus, these larger fruit components were likely related to accelerated growth as a strategy to escape heat but still be able to maintain seeds, as seen in certain cultivars. This condition may be attributed to the accelerated phenology in the pea plant under heat stress, also observed in other legumes such as lentil and chickpea (Kaushal et al. 2013;Bhandari et al. 2016). Altogether, this finding showed that fruit growth was closely associated with disturbance of plant performance in the field.

Effects on seed set at various reproductive nodes
Under growth chamber conditions, seed set was also evaluated at the level of the RN. Reduction in seed number occurred mainly in the first two to three RN on the plants, where opened flowers (node 1) and buds (nodes 2 and 3) were present during treatment. Comparably, Jeuffroy et al. (1990) evaluated various RN on the pea cultivar Solara and identified that high temperature (31 • C) reduced seed number at the first two plant RN. They attributed the effect to a critical stage of the flower (>six days after open flower) localized at the affected nodes, where heat stress could affect early embryo Table 3. Pearson correlation matrix among pea reproductive parameters, including seed-to-ovule ratio (SOR), seed diameter (SD), proportion of aborted ovules right before or after fertilization (BAF), proportion of aborted ovules at globular to heart embryo stages (GH), and the proportion of aborted ovules at early to late cotyledon stages (ELC) per pod. Prop. of aborted ovules ELC −0.50 * * * 0.02 −0.02 0.27 * -- * , * * , and * * * in bold denote significance level of the correlation coefficient at P < 0.05, 0.01, and 0.001, respectively. Note: Data were averaged over the first four reproductive nodes on five field pea cultivars exposed to heat stress for four consecutive days (35/18 • C, upper right side) and control temperatures (24/18 • C, lower left side) in growth chamber conditions. development. In a related study, Guilioni et al. (2003) found that high temperatures applied before and after flowering of cultivar Messire caused seed reduction by changing the pattern of seed production along a plant's main stem regardless of flower age at a specific node. Although Jeuffroy et al. (1990) and Guilioni et al. (2003) showed different responses of pea plants to heat stress, they agreed that assimilate availability could drive the fate of the seeds of heat-exposed plants.
In our study, although the reduction of seed yield on the first RN may be attributed to heat stress during sensitive young floral stages (open flowers and buds), the variation of node responses among cultivars confirms some kind of dependence on resource availability. Indeed, cultivar Carneval displayed a reduction of seeds at the lower three RN of the plants, whereas cultivars CDC Meadow and CDC Sage showed a reduction of seeds at just two of these RN. The effect could be related to a physiological mechanism in which sinks (seeds) during development are adjusted to the plant's access to photo-assimilates under stress (Aloni et al. 1991;Guilioni et al. 2003). Accordingly, evaluation of legume crops such as chickpea and lentil with diverse stress-tolerant cultivars has shown that yield and reproductive failure were associated with anticipated physiological responses; for example, alteration of chlorophyll content, sucrose availability, and photosynthetic efficiency (Bhandari et al. 2016). Therefore, the reduction in seed number and SOR in medium-and low-heattolerant cultivars, from abortion, is likely because assimilate supply was constrained, and that each cultivar's tolerance level in heat stress is really about the plant maintaining assimilate supply to developing embryos.

Effects on ovule and early seed abortion at different plant reproductive nodes
Under high temperatures, increased abortion of young ovules at fertilization or globular/heart stages (designated aborted ovules BAF and GH) suggests a high susceptibility of ovules to stress at early developmental stages. Aborted ovules of BAF (about 1 mm in length) could be associated with some level of fertilization malfunction; however, this possibility may apply to only a small portion of these aborted ovules.
Indeed, prior experiments (Osorio et al. 2022) following similar heat treatment showed that more than 90% of ovules of minute size had signs of fertilization success at early stages (e.g., a zygote and embryo). Jahnke et al. (1989), studying assimilate partitioning in pea, explained that ovaries (containing ovules) start a stage of active growth right after fertilization, and they become sinks for assimilates that compete against other developing organs on the plant. Under these conditions, embryo starvation could plausibly occur as a consequence of any adjustment or constraint in assimilate supply by the mother plant (Dinar and Rudich 1985;Aloni et al. 1991). Therefore, during development of offspring (ovules), those that received only a minor investment from the mother plant could be the ovules that would end up aborted (Lloyd 1980;Diggle 1995), as seen in our study.
Furthermore, the increase of aborted ovules containing embryos between globular to heart stages, GH, occurred at different RN of heat-treated plants, indicating vulnerability at early seed development. Following fertilization, the consecutive embryo, suspensor, and endosperm formation requires active cell division in the ovule, commonly under parental control (Ruan et al. 2010). In our study, high temperature at the early flower stage could have disrupted the plant's metabolic and hormonal activity and consequently interrupted young embryo formation Liu et al. 2019). For example, in common bean, young fruit abortion in heat-sensitive cultivars was associated with a low level of indole-3-acetic acid transport that accumulated at pedicels of aborted reproductive structures (Ofir et al. 1993). In a similar context, research in wheat showed that high temperature induced high ethylene levels in structures such as developing embryos and kernels in a heat-sensitive cultivar with low seed set (Hays et al. 2007). Savada et al. (2017), studying pre-and post-fertilized pea ovaries, found that heat stress modified ethylene evolution by increasing concentrations in pre-fertilized ovaries and decreasing concentrations in postfertilized ovaries. Also, a recent study on pea showed that developing seeds under heat stress showed higher concentrations of auxin, abscisic acid but lower concentrations of gibberellin, in combination with ethylene signalling gene expression (Kaur et al. 2021). Thus, ovule abortion at the early Fig. 6. Flowchart representing seed number adjustment within pods of pea plants exposed to heat stress and non-stress conditions.
embryo growth stage could be the result of disruption or adjustment of hormonal and metabolic activity in plants under heat stress, possibilities that require further research in pea.
Ovule abortion at the early embryo stage (globular to heart stage) in our study increased at various positions within the pod in heat-treated plants. Although ovule abortion in pea can normally occur at stylar and basal ends of pods growing on unstressed plants (Linck 1961), the increase of early embryo abortion at all ovule positions of stressed plants confirms the reduction of resource availability in the plants. Nakamura (1988) studied seed abortion in Phaseolus vulgaris and identified that reduced resource availability within the maternal plant impacted embryo survival, especially during early seed formation. Given that the linear ovule arrangement within a pod plays a crucial role in seeds' success in legumes, various factors add complexity to embryo growth. For example, Hedley and Ambrose (1981) stated that ovules within the same pod could exhibit different growth rates, where seeds with higher growth rates can be localized in the pod's middle position in some legume species. Alternatively, O'Donnell and Bawa (1993), following the pattern of ovule and seed abortion in Sophora japonica, suggested that abortion at specific ovule positions may be attributed to the sequence of fertilization and degree of ovule maturity, where the first fertilized ovules were not necessarily the ovules closer to the stylar end of the pod. Hence, in field pea, a combination of maternal supply adjustment, degree of ovule develop-ment, embryo competition, and sequence of fertilization may all interact with the probability of seed success under heat stress.

Effects on seed diameter
The increase in temperature reduced seed diameter on high and medium SOR cultivars. Consistently, seed size reduction has been observed in various legume crops, such as soybean and lentil, in response to heat stress (Awasthi et al. 2014;Kumar et al. 2016). In lentil, evaluation of several accessions showed that high temperature in the field caused a 6%-28% reduction in seed size (Kumar et al. 2016). In our study, although the effect was less than in lentil, heat-treated plants of high and medium heat-tolerant cultivars had a 1 to 3% seed size reduction. Two studies have revealed that the effect of high temperature on seed size can be attributed to disturbance of the plant's reproductive development (Wang et al. 2006;Tacarindua et al. 2012). For example, work on legumes and cereal crops showed that the increase of temperature at stages of pre-anthesis, full bloom, and seed filling affected final seed size mainly due to a shorter seed filling duration (Wang et al. 2006;Prasad et al. 2008a). In fact, under a short period of seed filling, seeds cannot reach satisfactory development, and an insufficient accumulation of photosynthates can reduce seed size (Munier-Jolain et al. 1998). In addition, high temperature disrupted the metabolic activity of cytokinin and invertase on various crops (Banowetz et al. 1999;Bhandari et al. 2016), which could lead to reduced cell division and limited assimilate partitioning for embryos and endosperms in development (Tacarindua et al. 2012).
In contrast to the reduced seed size identified on high and medium SOR cultivars, the low SOR cultivar Carneval increased seed size under high temperature. Although this increase in size was low (2.7%), we ascribed it to a compensatory effect for the high seed loss observed on heat-treated plants, as observed in related studies under field and control conditions in pea and chickpea, respectively (Wang et al. 2006). In our study, because the variation in seed size (1%-2.7%) was relatively low compared to the reduction in seed number (4%-43%), seed number appeared to be the trait most affected by high temperature. This observation is consistent with Sadras's (2007) remarks when reviewing the trade-off effect between seed size and number in various crops and concluded that seed number could be the most plastic trait related to plant resource allocation under adverse environmental conditions. Similarly, research on cereal and legume crop adaptation has shown that seed number could explain most of the variation in seed yield in response to adverse thermal conditions (Sadras and Dreccer 2015). In our study, high SOR cultivars always managed to maintain seed number at the expense of a slight reduction of seed size, whereas the low SOR cultivar had a high reduction of seed number and only a slight increase in seed size under heat stress conditions (Fig. 6).

Conclusions
As a critical element of reproduction in field pea, ovules and their failure to develop upon plant exposure to heat stress, were investigated in young pistils (4DOF) and mature pods under field and growth chamber conditions, respectively.
Inspection of embryo sacs within ovules 4DOF and aborted ovules in mature pods both revealed that high temperature directly increased ovule failure during early embryo formation, rather than ovule failure being collateral damage from disrupted pre-fertilization events (such as pollination failure). This is a novel finding from both controlled temperature and field-imposed heat stress.
Evaluation of ovaries, ovules, and embryo sacs at 4DOF provided related information on the reproductive abortion of plants on ESP and LSP. Interestingly a compensatory effect between young ovary growth and plant performance was detected in LSP, wherein cultivars that showed reduced ovary size 4DOF were related to a high number of RN and pods at the end of the season. Furthermore, the effect of high temperature on medium and low SOR cultivars was consistent with a high reduction in seed number on the first three out of four RN of these cultivars. The decrease in seed number under heat stress apparently occurred in response to resource availability rather than disruption of flowers at a specific stage. Indeed, although young ovule abortion increased on heattreated plants, this abortion occurred randomly at RN for aborted ovules < 1 mm, or was only detected at the three first RN (aborted ovules with embryos at globular to heart stages) of the plants.
We also observed a compensatory effect in terms of seed diameter, where cultivars retaining seed number showed a slight reduction of seed diameter under heat stress. Overall, high temperature reduced the SOR in cultivars such as CDC Meadow, CDC Sage, and Carneval, as expected, and young ovule abortion at 4DOF, and also measured at plant physiological maturity, inferring susceptibility of early embryo growth under heat stress.
The diverse compensatory effects detected in pods on heattreated plants point to a resource adjustment mechanism that occurs during plant reproductive development under stress. As such, this discovery opens avenues to screen and identify genotypes to develop heat tolerant cultivars that maintain more of their fertilized ovules by maintaining assimilate supply in stress.
Data generated or analyzed during this study are available from the corresponding author upon reasonable request.