Developmental patterns of flowers and pods and the effect on seed number in French serradella (Ornithopus sativus) and yellow serradella (Ornithopus compressus) cultivars

Seed was dehulled and scarified as required to ensure germination of hard seed (impermeable to water). The seed of Serratas were entirely soft (permeable to water) and did not require scarification. The seeds of all cultivars exhibited high germination rates (>90%) except for Erica (germination rate 50%).

Ten seeds were sown in 23-cm-diameter pots containing 5.8 L of a nutrient-rich, pasteurised, potting soil known to be able to support the growth of the plants for the duration of the experiment. At sowing, the soil was moistened, and the pots were covered with black plastic and kept at 4°C for 3 days in a controlled environment to promote even germination. The plants were then transferred to a naturally lit glasshouse for 35 days with partial temperature control that permitted daytime temperatures to rise to, typically, ~22°C and night-time temperatures to decline no lower than ~5°C. In practice, mean daily temperatures ranged between 10°C and 18°C, with an average of 15°C for the 35-day period (Fig. 1a). By 38 days after sowing, the plants had developed five or six leaves and the pots were transferred to a controlled-environment plant growth chamber set to a constant vernalising temperature (5°C) with a long photoperiod (20 h). The vernalisation and long photoperiod treatments were intended to reduce the time that it would take for first flowers to appear on all of the cultivars (Goward et al. 2023). After 3 weeks under these conditions, plants were thinned to one uniform seedling per pot, each with six or seven leaves per plant (i.e. Stage 36–37, BBCH scale; Enriquez-Hidalgo et al. 2020). Each pot represented one replicate of each cultivar. Five replicate pots had been allocated previously, at random, to the Control treatment (i.e. no flowers removed) and five to the EFR treatment. Two pots of Erica seedlings designated for the EFR treatment did not survive the vernalising temperatures and were removed from the experiment. All plants were moved to a glasshouse on 14 September 2020 (59 days after sowing), when natural daylengths were lengthening (starting at 12.8 h, Fig. 1a) and temperatures were rising (average daily temperature starting at ~17°C and ending at ~19°C, Fig. 1a). Natural photoperiods were calculated using the civil twilight method (Goodspeed 1975).

Fig. 1.

(a) Mean daily temperatures and photoperiods in the glasshouse and in the controlled environments used during the experiment. (b) Mean number of days from sowing to appearance of first flower for yellow serradella (O. compressus) cultivars (diamond pattern bars) and French serradella (O. sativus) cultivars (solid bars). Bar indicates l.s.d. (P = 0.05).

Temperature data were collected for the duration of the experiment by placing a Thermocron iButton temperature sensor (model number DS1921G-F50; Maxim Integrated, San Jose, CA, USA), on the soil surface in two pots selected at random. Temperature data were recorded every 30 min (accuracy ±0.5°C, resolution 0.5°C) and mean daily temperatures were calculated from these data. Thermal time (degree-days) estimates were made by summing mean daily temperatures.

Plants were monitored regularly and were managed to avoid potential confounding effects of inadequate nutrient supply, insect pests or plant disease issues. Pots were hand-watered daily to ensure that plants were neither waterlogged nor water-stressed. The decision to cease the experiment was made based on the duration of flowering exceeding ~3 months after first flower appearance for most cultivars, a timeframe which is rarely exceeded, prior to the onset of hot and dry conditions that determine the end of a typical growing season for these plants. Nodulation was supported by inoculating the plants at 9 days after sowing with a peat slurry containing rhizobial strain Bradyrhizobium lupini (WSM471; New Edge Microbials, Albury, NSW, Australia), immediately prior to watering. No pesticides were required during the experiment.

Serradellas often have multiple reproductive axes. For this experiment, flowering dates, flower numbers and pod numbers were recorded along two reproductive axes (either a main and a primary axis, or along two primary axes; Goward et al. 2023; Supplementary Table S1) chosen as a consequence of being the first axis to bear a flower (denoted ‘axis A’) and the second stem axis on which a flower was born (‘axis B’). These axes were trained on vertical stakes to facilitate easy monitoring of each RN. Flower development was monitored daily and small, paper ‘jewellers’ tags were marked with the date of the appearance of the first flower (i.e. Stage 53, BBCH scale) and hung on the peduncle of each umbel when the first flower appeared at each node. Flower appearance was defined by petals being visible (~2–3 mm in length). At this time, the total number of flower buds that comprised the umbel developing at the RN was also counted and noted on the tag. The number of flowers and number of pods at each RN were recorded on both stem axes. The average of the two stem axes was used to represent the plant. However, it was impractical to count the number of seeds per pod on both stem axes owing to the very high numbers of pods. Therefore, seeds per pod at each RN was recorded only for axis A. Seeds were counted when the pod colour began to turn yellow because, at this stage in their development, it was possible to count seed numbers accurately without dissecting the pods. At this stage, there was visible ‘segmentation’ along each pod, indicating segments containing individual seeds. An unfortunate consequence of counting seed numbers late in pod development was that values are missing for the Serratas EFR treatment at RN 15–RN 17 due to early dehiscence of pods at these nodes. Observations of flowers and pods were collected from the monitored axes of each plant for 20 RNs, or until flowering ceased (if this occurred before RN 20).

For the EFR treatment, plants were grown until the appearance of the first flower at RN 5 of axis B. All inflorescences and immature pods on the plant were removed by cutting through the peduncle at this time (e.g. Fig. S1). Control plants were left intact.

The time of appearance of first flower and the cultivar responses to flower removal treatments were evaluated using a two-way analysis of variance (ANOVA) that considered cultivar and flower removal treatments as factors. Following the ANOVA, Fischer’s least significant difference values were determined.

It was desirable to express rate of RN development in units of time per RN. Consequently, for each tracked stem axis, Excel (ver. 2208; Microsoft, Redmond, WA, USA) was used to form linear regressions between time of first flower appearance at each RN (the independent variable; y-axis) and RN number (RN 1, RN 2, etc., the dependent variable; x-axis). Time was measured in days or degree-days from the appearance of the first flower on RN 1. Rates of RN development were determined from these regressions. Values for duplicate axes were averaged to determine the replicate values used for ANOVA.

The duration of flowering for each cultivar was determined by the proportions of plants that continued to flower through time. Split-line regression analysis was used to assess each cultivar’s pattern of flowering by regressing the proportion of stems continuing to flower against RN number. Data beyond the RN at which a zero percentage of stems continuing to flower was recorded were excluded from the analysis to avoid biasing the regression. The break point of the split-line regression represented the RN beyond which there was a decline in the proportion of stems bearing new flowers. The standard errors associated with the break point were used to test for differences between flower removal treatments within cultivars (e.g. Dyson and Conyers 2013). Genstat did not recognise the very low RN at which the break point occurred in the Margurita EFR treatment, so the standard error for a predicted y-value was determined using the LINEST function in Excel to circumvent this issue.

Although the planned experiment was established with a fully balanced design (i.e. eight cultivars × two flower removal treatments × five replicates × 20 RNs), the flowering patterns observed in the serradellas resulted in a decreasing number of plants continuing to flower and set pods over time. As a result, the statistical design became unbalanced, with decreasing replication for ‘nodes bearing pods’ as the RN number increased. Separate ANOVAs were conducted for RN 1–RN 5 (2-way ANOVA (cultivar × RN) of the Control treatment), and for RN 6–RN 20 (3-way unbalanced ANOVA (cultivar × flower removal treatment × RN)) for flower number per RN, pod number per RN, and average number of seeds per pod at each RN. The use of separate ANOVAs was necessary because the removal of flowers and pods in the EFR treatment disrupted the development of pods and seeds on RN 1–RN 5.

The potential for seed production by each cultivar and the possible impacts of environmental stressors on seed production were explored by calculating the average total number of seeds initiated per stem axis:

where RN_i is the first (Control treatment) or first ‘surviving’ (EFR treatment) RN, and RN_j is the last node to produce a flower within a 500 or 700 degree-day window (as described below).

The total number of seeds per axis was determined for each replicate (i.e. plant) of each treatment (i.e. cultivar × flower removal treatment). Thus, the average value (and its standard error) derived for each treatment from these data also reflected the progressive declines in the proportion of plants that continued to flower.

Seed numbers produced in the Control treatment were assumed to represent a plant flowering at an optimal date with no loss of flowers or pods through frost. Seed numbers in the EFR treatment were intended to represent a plant that had flowered earlier than the optimal date, with loss of flowers and immature pods on the first five RNs through frost. The calculations explored two contrasting growing-season conditions for seed development: (1) a typical growing-season length in which the duration of the flowering window is largely unlimited (i.e. up to 700 degree-days); and (2) a shorter seed-development period (500 degree-days) to mimic a premature end to flowering and pod production due to heat stress or terminal drought. Time (degree-days) for development of seed numbers was counted from the appearance of the first flower (i.e. at RN 1) in the Control treatment, or from the appearance of the first ‘surviving’ flower (i.e. at RN 6) in the EFR treatment. For the purpose of these calculations, where 700 degree-days after the first surviving flower would have occurred at RNs >20 (i.e. after the flower and pod measurements had ceased), it was assumed that flowers and pods continued to be produced at RN 21 and above until the 700 degree-day window had elapsed. This extrapolation was necessary for cvv. Erica, Rosa and Margurita. For each of these cultivars, pods per RN and seed number per pod at RNs >20 were estimated using the average of pod and seed numbers recorded for RN 15–RN 20, and the number of nodes continuing to flower was estimated from the equations that describe the decline in podding duration and the rate of RN development for each cultivar, as reported in the Results. Missing data for seed number per pod for RN 15–RN 17 in the Serratas EFR treatment were substituted using the average seed number per pod derived from the other RNs in this treatment.

Standard errors associated with these estimates were calculated using all observations within the relevant period after the first or first-surviving flower had appeared (i.e. either 500 or 700 degree-days).

The total number of seeds initiated per stem in the Control treatment was compared with that in the EFR treatment under four ‘growing-season scenarios’ to test the ability of the cultivars to: (1) compensate for early flower loss (e.g. as may occur through frost), and/or (2) initiate seeds when the seed-production period was shortened prematurely.

The physiology of flowering and pod formation was examined in relation to RN number. All serradella cultivars followed a similar pattern of first flower appearance whereby there was an initial period over which all plants representing the cultivar were flowering on successive RNs (hereafter referred to as ‘full flowering’; Table 2, Fig. 2). At each node, all flowers in the umbel opened in quick succession after the first flower had emerged (usually over a 1–2-day period). Thereafter, the proportion of plants that continued to produce flowers at a higher RN declined, and ultimately each cultivar ceased flowering or was projected to do so. The RN number at which full flowering ceased differed among cultivars (P < 0.05; Table 2, Fig. 2). For example, in the Control treatment, Rosa was in full flower until RN 14, and at the other extreme, Margurita and Erica remained in full flower until only RN 7–9. Most cultivars exhibited a similar rate of decline in the proportion of plants that remained in flower after their full flowering period and were observed or projected to cease flowering by between six and 12 RNs after the end of their full-flowering period. Margurita and Erica were exceptions to this, with very similar and substantially slower declines in proportion of plants continuing to flower at successive RNs (Fig. 2, Table 2). Although Margurita and Erica completed their full-flowering periods at a relatively low RN, their slower decline in plants continuing to flower meant that they were expected to continue flowering on a substantially higher RN than occurred for any of the other cultivars (Table 2).

Fig. 2.

Split-line regressions describing the proportion of plants that represented each serradella cultivar that continued to produce flowers at each reproductive node (RN) in the: (a) Control treatment (i.e. no flowers removed), and (b) EFR treatment (i.e. all flowers removed from plant when the first flower appeared at RN 5). There were significant differences (P < 0.05) in the highest RN number at which full flowering occurred between the Control and EFR treatments for all cultivars, as reported in Table 2. Solid symbols and red lines, data for yellow serradella (O. compressus) cultivars; open symbols and black lines, French serradella (O. sativus).

Removing flowers lengthened the duration of the full flowering period (P = 0.05) for five of the seven cultivars assessed (Fig. 2, Table 2). SerraMax and Yellotas were the exceptions with no significant change to their full flowering periods. For King, Rosa and Serratas, the response to removal of flowers was to extend full flowering by two RNs. However, Erica and Margurita over-compensated for the removal of inflorescences and young pods on the first five RNs and extended full flowering by seven and nine RNs, respectively.

The pattern and, consequently, duration of pod formation largely reflected the pattern and duration of flowering among the cultivars (Fig. 3, Table 2). The duration of full flowering largely determined the highest RN at which all plants in each cultivar population formed pods. With only a few exceptions, the last RN at which pods were developed on all plants in a cultivar population was typically within ~2 nodes of the last RN for full flowering. The rate of decline in the proportion of plants forming pods for each cultivar was also similar to the rate of decline in the proportion of plants flowering after the period of full flowering.

Fig. 3.

Split-line regressions describing the proportion of plants representing each serradella cultivar that continued to produce pods at each reproductive node (RN) in the: (a) Control treatment (i.e. no flowers removed), and (b) EFR treatment (i.e. all flowers removed from plant when the first flower appeared at RN 5). Significant differences (P < 0.05) in the highest RN at which full podding occurred between the Control and EFR treatments as reported in Table 2 and shown on the figure by asterisks. Solid symbols and red lines, data for yellow serradella (O. compressus) cultivars; open symbols and black lines, French serradella (O. sativus) data.

Four patterns in the numbers of flowers produced at each RN were evident among the cultivars (Fig. 4). First, for Erica and Margurita, there were four flowers at RN 1 but flower number increased to a maximum of six per RN by RN 5 or RN 6 and this was maintained until about RN 10. Thereafter, flower numbers per RN declined gradually. Second, cvv. Rosa and Serratas exemplified a pattern in which flowering commenced with the maximum number of flowers (five or six flowers per RN) and this presentation was maintained up to ~RN 15. Flower number per RN then declined gradually with successive RNs. Third, King and SerraMax started with two flowers at RN 1 but increased to a maximum of four flowers per RN by RN 5 of RN 6 and this was maintained until ~RN 10. Thereafter, flower numbers per RN declined rapidly with no flowers or only one flower per RN from RN 15 onwards. Finally, Yellotas commenced flowering with four flowers per RN, slowly increased flower number to five per RN by RN 5, and then experienced an immediate decline in flower number until there was only one flower per RN at RN 17 and above.

Fig. 4.

Mean number of flowers (left panels) and pods (right panels) per reproductive node (RN) for each cultivar. Solid symbols, Control treatment (i.e. no flowers removed); open symbols, EFR treatment (i.e. all flowers and immature pods on RN 1–RN 5 removed from plants when first flowers appeared at RN 5). The dashed line at RN 5 indicates the point at which flower removal occurred. Bars associated with each symbol show 2 × s.e. The l.s.d. bars (P = 0.05) shown to the left of the dashed lines represent the results from 2-way ANOVAs (cultivar × RN) of flower numbers and of pod numbers on RN 1–RN 5. The l.s.d. bars (P = 0.05) to the right of the dashed line represent the results from 3-way ANOVAs (cultivar × flower removal × RN) of flower numbers and of pod numbers at RN 6–RN 20. French serradella (O. sativus) cultivar names are presented as standard text, whereas yellow serradella (O. compressus) cultivar names are in bold text.

In most cases, the EFR treatment did not change subsequent flower numbers per RN compared with the Control. However, a significant cultivar × flower removal treatment interaction (P < 0.05) existed because King maintained up to two more flowers per RN at RNs >14 in the EFR treatment. Likewise, there were a few other instances where there was one or two extra flowers per RN for cultivars (i.e. Rosa, SerraMax and Yellotas) in the EFR treatment compared with the Control.

The patterns of pod number per RN produced by the cultivars usually reflected the patterns of flower number per RN. The highest flower and pod numbers were observed by about RN 5 or RN 6. Thereafter, the numbers of pods produced were either similar to, or marginally lower than (typically by one or two pods per RN), the number of flowers per RN (Fig. 4).

A significant (P < 0.05) cultivar × flower removal treatment interaction was observed for pod numbers. Following a peak in pod numbers per RN, they typically declined in the Control. Removing flowers led to a delay in this decline in pod numbers for cvv. Erica, Rosa’ and King. Compensatory production of pods by this means was only clearly associated with production of additional flowers per node by one cultivar (King). For Erica and Rosa, the additional pods in the EFR treatment were due to more flowers (at RN 15–RN 20) developing into a pod. For most other cultivars, there was no consistent delay in the decline in pod number per RN due to removal of flower. Serratas was an anomaly in that, at RN 17–RN 20, pod number declined more rapidly and then ceased prematurely in the EFR treatment compared with the Control.

For the Control treatment, the average number of seeds per pod produced by each cultivar was stable for the first five RNs (no significant effect of RN, P = 0.47) but there was a main effect of cultivar (P < 0.001) (Fig. 5). The yellow serradella cultivars tended to produce more seeds per pod (average 8 seeds/pod, range 5–10 seeds/pod) than the French serradella cultivars (average 6 seeds/pod, range 5–7 seeds/pod). For both the Control and EFR treatments, stability of seed numbers per pod was maintained until at least RN 10, after which, marginal declines in the number of seeds per pod occurred for a few cultivars (i.e. one or two fewer seeds per pod). Although there was a significant main effect of flower removal (P = 0.03) on the average number of seeds per pod, the difference was marginal, with an average increase (over all cultivars) of only 0.24 seeds/pod in the EFR treatment compared with the Control.

Fig. 5.

Mean number of seeds per pod at each reproductive node (RN) for each cultivar. Bars associated with each symbol show 2 × s.e. Solid symbols with dashed lines, yellow serradella cultivars; open symbols with solid lines, French serradella cultivars. l.s.d. bars (P = 0.05) resulting from 2-way ANOVA (cultivar × RN) of seed numbers per pod on RN 1–RN 5 are shown to the left of the dashed line, which indicates the RN at which EFR treatments were imposed. l.s.d. bars (P = 0.05) resulting from a separate 3-way ANOVA (cultivar × flower removal × RN) of seed number per pod on RN 6–RN 20 are shown to the right side of the dashed line. Missed observations for Serratas EFR treatment: RN 15, RN 16 and RN 17. Solid symbols, data for yellow serradella (O. compressus) cultivars; open symbols, French serradella (O. sativus) data.

Total number of seeds initiated per stem in the Control and EFR treatments were compared at 700 degree-days after the appearance of the first flower on RN 1 (Fig. 6a). This tested whether the cultivars were able to compensate for removal of early flowers and pods by initiating additional seeds within the same time frame. Cvv. King, Rosa and Erica initiated a similar total number of seeds per stem in both treatments, indicating that additional seeds had been produced at RNs >RN 5 in response to removal of early flowers and immature pods (RN 1–RN 5). For cv. King, compensation for early loss of flowers occurred as a result of an increase in the production of flowers in the EFR treatment. However, for Rosa and Erica, the compensation occurred because more of the existing flowers (at higher RNs) produced a pod. Margurita, Yellotas and SerraMax did not compensate adequately for the early flower loss by producing enough additional pods per RN (Fig. 4). Of these cultivars, only Margurita mitigated against removal of flowers by extending its period of full flowering (Fig. 2). However, because pod numbers per RN were still increasing between RN 1 and RN 5 in all three cultivars (Fig. 4), the impact of losing flowers and immature pods at these nodes was comparatively modest (i.e. 24–33% fewer seeds initiated). Removing flowers had greater impact on Serratas (i.e. 51% fewer seeds initiated) than on Margurita, Yellotas and SerraMax, and Serratas exhibited little capacity to compensate for early flower loss: (1) it produced its highest numbers of flowers and pods per RN between RN 1 and RN 5 (which were then removed) (Fig. 4); (2) it did not extend its full flowering period substantially (Fig. 2); and (3) it appeared to cease pod formation prematurely (at RN 18) in the EFR treatment (Fig. 4).

Fig. 6.

Seed numbers initiated by each cultivar on main and/or primary stems under different growing-season scenarios. (a) Scenario 1: no flowers or pods removed (Control treatment) compared with all flowers and pods removed from the plant at appearance of the first flower on RN 5 (EFR treatment) to simulate the effect of early flower loss such as due to frost, up to 700 degree-days from first flower appearance at RN 1. (b) Scenario 2: Control compared with EFR treatment, with seed number initiation measured from the first surviving flower, i.e. RN 1 for the Control and RN 6 for EFR, to simulate the effect of purposefully flowering early/in a high-frost-risk window. *For cvv. Erica, Rosa and Margurita, extrapolated values were used for the EFR treatment because data collection ended prior to 700 degree-days after first flowers appeared at RN 6. (c) Scenario 3: seed number initiated in the Control up to 700 degree-days compared with the same treatment but under a more restricted time frame of 500 degree-days (e.g. to simulate the effect of a drought season). (d) Scenario 4: seed number initiation restricted to 500 degree-days (e.g. drought), with (EFR) or without (Control) flower loss (e.g. frost and drought).

Total seed numbers per stem were assessed at 700 degree-days from the first flower (i.e. from RN 1 for the Control) or the first surviving flower (i.e. RN 6 for the EFR treatment) (Fig. 6b). This was intended to mimic a situation where the cultivar had flowered early and been frosted, losing all flowers and young pods on RN 1–RN 5. This scenario differs from Scenario 1 in that the time available for seed initiation from appearance of the first or first surviving flowers was the same (700 degree-days). This permitted similar numbers of seeds to be initiated in the EFR treatment compared with the Control for most cultivars; the exceptions were Serratas and SerraMax (Fig. 6b). The routes by which compensation for flower loss occurred among the cultivars differed. Margurita and Erica extended their period of full flowering substantially in response to removal of flowers (Fig. 2). King, Rosa and Erica produced additional pods at nodes above ~RN 12, (Fig. 4). Yellotas appeared to benefit from a marginally slower decline in the proportion of plants continuing to flower and set pods in the EFR treatment. By contrast, seed initiation by Serratas was limited by its poor capacity for any compensatory pod production (see above). SerraMax was one of the few cultivars that did not extend full flowering in response to early flower loss, and although it did slow the decline in plants producing flowers and pods, this was not sufficient to compensate fully for the early loss of flowers and immature pods.

Seed initiation was then examined in the Control treatment over a seed production period that was restricted to 500 degree-days from first flower, in order to mimic the potential effect of a shorter ‘spring’. Fewer seeds (~15–26% less seed) were initiated by most cultivars (i.e. Margurita, Rosa, Yellotas and King) compared with the achievement within the longer, 700 degree-day seed initiation window (Fig. 6c). However, SerraMax and Serratas, which are characterised by having short periods of flowering and podding (Figs 2 and 3), largely completed these activities within the 500 degree-day window.

The cultivars were finally compared by assessing the potential effects of early flowering that had resulted in flower loss due to frost in a short spring (500 degree-days from first surviving flower; Fig. 6d). Within this very restricted timeframe, the number of seeds initiated was relatively low in both the Control and EFR treatments. For the cultivars that had previously exhibited capacity to compensate for removal of flowers, there was no further reduction in the number of seeds initiated as a result of this ‘early flowering with flower loss’ scenario. The exceptions were again SerraMax and Serratas, which had previously exhibited little or no ability to compensate for flower removal.