Four field trials were conducted in Ottava (SS) (40.8°N 8.5°E, 80 m above sea level) in Sardinia, Italy, in the 2016/17 (‘SS17’), 2017/18 (‘SS18’), 2020/21 (‘SS21’) and 2022/23 (‘SS23’) seasons, and two field trials were conducted in Fiorenzuola d’Arda (PC) (44.9°N, 9.9°E, 82 m asl) in Northern Italy, in the 2016/17 (‘FI17’) and 2017/18 (‘FI18’) seasons.

The soil in Ottava consisted of a sandy clay loam of a maximum depth of about 0.6–0.7 m overlying a limestone bedrock (Xerochrepts). The climate is typically Mediterranean, with an annual rainfall of 552 mm, mainly concentrated between October and April (with reference to the past 60 years), and a seasonal rainfall (Oct–May) of 480 mm. The annual mean temperature is 16.3°C, with minimum winter (Jan–Mar) temperatures of 6.6°C and maximum spring (Apr–Jun) temperatures of 22.5°C. The soil in Fiorenzuola d’Arda is a deep silt clay loam, mesic Udic Ustochrepts, and the climate is temperate, with an annual average rainfall of 786 mm (with reference to the past 37 years) and a seasonal rainfall (Oct-May) of 574 mm. The annual mean temperature is 12.5°C, with minimum winter (Jan–Mar) temperatures of -1.9°C and maximum spring (Apr–Jun) temperatures of 23.0°C.

In all the six environments, we compared two near-isogenic pairs of lines (NILs) (commercial cultivars and two backcross oligoculm selections containing the tin gene of the donor line 492) of bread wheat (Triticum aestivum L.), namely Janz ± tin and Kite ± tin. The two NILs, kindly provided by Dr. R.A. Richards, possess the tin (tiller inhibition) gene, in linkage with the Hg (hairy glume) gene at 10 ± 3 map units (Richards, 1988). Kite ± tin is awnless while Janz ± tin is awned.

Fields were prepared by chisel-ploughing to a depth of 0.25 m, followed by surface cultivation. Sowing was performed at the rate of 350 germinable seeds m^-2 between 31 October and 26 November in all environments except SS23, where the unfavorable rainfall pattern moved sowing to the 3 January. The preceding crop was barley at Fiorenzuola, and faba bean or Alexandrian clover at Ottava. Plots consisted of 8 rows 8.4 m long, with a between-row distance of 0.15 m, for a total area of 10 m². Nitrogen fertilization only was applied in the field trials in Fiorenzuola, split between sowing and April, for a total of 177 kg N ha_-1 in 2017 and 108 kg N ha_-1 in 2018. Both N and P were applied in Ottava, with the N rate varying from 49 to 119 kg ha_-1, and P from 13 to 40 kg ha_-1. Weeds, pests and diseases were chemically controlled.

Treatments were arranged in a split-plot design with four replications. Cultivars were assigned to the main plots and the presence/absence of the tin gene to the sub-plots.

Emergence, booting (DC 39, Zadoks et al., 1974), anthesis (DC 61) and physiological maturity (DC 92) were recorded by periodical inspections of the plots when more than 50% of plants in the plot had reached that phenological stage.

Two biomass samplings were carried out at the stages of anthesis and physiological maturity. Four 0.5 m long samples of uprooted plants per plot, roots excluded, were hand-cut at each stage for a total of 0.3 m². Anthesis and maturity samples were divided into stems plus leaves (indicated as ‘stems’) and spikes, and the number of stem and spikes ascertained and expressed on a square meter basis. All samples were oven-dried at 80°C for 48 hours before weighing. Spikes from the maturity samples were threshed, and grain weight, number of grains spike^-1 and moisture content determined.

Nitrogen percentage was determined on each Ottava biomass subsample by means of a Carbon/Hydrogen/Nitrogen Analyzer (628 Series, LECO Corporation, St. Joseph, MI, USA).

The number of grains m^-2 was calculated as the product of the number of spikes m^-2 and the number of grains spike^-1. Grain yield was obtained on a plot basis with mechanical harvesting. Grain weight and grain yield were expressed at 0% humidity.

Fruiting efficiency was calculated according to Fischer (2011) as:

Fruiting efficiency (number of grains set per unit of spike dry weight at anthesis) = (number of grains m^-2)/(spike weight m^-2 at anthesis)

The grain filling rate was estimated by dividing grain weight by the number of days between anthesis and maturity, considered to be roughly representative of the grain filling duration. Translocation was estimated on both a single stem basis and on a unit surface basis:

Translocation on individual stem basis (g stem^-1) = dry weight one stem at maturity – dry weight one stem at anthesis

Translocation on a surface basis (g m^-2) = Translocation on one stem basis x number of stems m^-2 at physiological maturity

The maximum fraction of photosynthetically active radiation intercepted (FIPAR) by the leaves was measured at booting, i.e. once all leaves had emerged and the maximum leaf interception reached. Measurements were made at noon using a portable probe (Sun-Scan Canopy Analysis System SS1-UM-1.05. Delta-T Devices Ltd., Burwell, Cambridge, UK) allowing the simultaneous measurement of photosynthetically active radiation above (using an external sensor) and below (using the probe) the canopy. The same instrument estimated the green area index (GAI). Using these data, we calculated the KPAR (extinction coefficient for photosynthetically active radiation) of each plot according to the formula:

KPAR = −[LN(1-FIPAR)]/GAI

The daily photothermal quotient (PTQ) for the booting-anthesis period was calculated following Fisher (1985) by dividing the daily intercepted PAR (calculated by multiplying the maximum FIPAR by half the daily total solar radiation recorded at the site) by the daily mean temperature minus 4.5°C. We also calculated the mean PTQ value for the whole booting-anthesis period.

Weather data (maximum and minimum temperature, rainfall, solar radiation, wind speed and air humidity) were recorded in meteorological stations located approx. 300 m from the fields. Reference evapostranspiration (ETo) was calculated from those data according to the Penman-Monteith equation (Allen et al., 1998). Cumulative growing degree days (°Cd) from sowing were calculated assuming a base temperature of 0 °C (Ritchie, 1991).

Linear regressions were used to explore the relationships between different agronomic traits, and to analyze trait plasticity (Finlay and Wilkinson, 1963; Lin et al., 1986; Becker and Leon, 1988). Plasticity was quantified separately for tin and free-tillering lines as the slope of the regression relating the line means of the two groups to the environment means (i.e. a response is non-plastic if it is in parallel to the mean environment response, as indicated by a regression coefficient of zero). We calculated plasticity for each group of lines using the 12 ‘environment x cultivar’ means. We used t-tests to calculate the probability that differences in the slopes for tin and free-tillering lines occurred by chance.

After assessing the homogeneity of variances by means of the Bartlett Test, we performed a combined analysis of variance (ANOVA) by superimposing the year as the main plot factor on the original design. The resulting split-split-plot design is an extension of the split-plot design, and can be adopted to accommodate a third factor (Gomez and Gomez, 1984; Quinn and Keough, 2002). Year was the whole plot factor (A), cultivar was the sub plot factor (B), and +/- tin the sub-sub plot factor (C). Error (a) (Block x A) was used to test the significance of A; error (b) (Block x B(A)) was used to test A and A x B; error (c) (Block x C x (A X B)) was used to test C, A x C, B x C and A x B x C. Statistical analyses were conducted using R software (R Core Team, 2017), package ‘agricolae’, ssp.plot procedure. Following a significant F test, means were compared through the least significant difference (LSD) test, considering a probability level of 0.05, calculated using the appropriate standard errors of the mean and t values (Gomez and Gomez, 1984).

Temperatures. From October to March, Fiorenzuola recorded lower temperatures than Ottava, with December showing the greatest difference: maximum temperatures averaged 16.8°C at Ottava and only 6.6°C at Fiorenzuola, and minimum temperatures averaged 8.5°C at Ottava and -1.8°C at Fiorenzuola ( Figure 1 ). Mean monthly minimum temperatures at Fiorenzuola dropped to around or below 0°C between December and January in both the 17/18 and 18/19 seasons. During this early part of the growing cycle, Ottava in 22/23 experienced the warmest temperatures between October and January.

In contrast, during the final three months of the growing cycle, both sites experienced similarly increasing temperatures. SS23 was the coolest environment during this period, particularly in terms of maximum temperatures, averaging 18.2°C compared with 24.4°C in the other environments. In April and May, the grain filling phase, maximum temperatures ranged from 18.7°C in SS23 to 22.6°C in FI17.

Evapotranspiration. Reference evapotranspiration (ETo) was also lower at Fiorenzuola than at Ottava during the October–March period, with values ranging from 0.3 to 1.9 mm d^-1 at Fiorenzuola and 1.2 to 2.7 mm d^-1 at Ottava. At Ottava, SS17 showed the highest ETo throughout the season, peaking at 6.5 mm d^-1 in June, while SS23 recorded the lowest values, with 3.7 mm d^-1 in May and 5.4 mm d^-1 in June.

Rainfall. SS18 recorded the highest total seasonal rainfall (784 mm from October to June), partly due to an exceptional 200 mm in May. SS23, the second-wettest environment (649 mm), received most rainfall during winter, with 200 mm in December, 180 mm in January, and 150 mm in February. The driest season was 2017, with totals of 311 mm at Ottava and 329 mm at Fiorenzuola, although spring rainfall was higher at Fiorenzuola (117 mm from March to May) than at Ottava (25 mm). Total rainfall in FI18 was 518 mm, including a notable 136 mm in May.

The overall performance of tin and free-tillering lines across these environmental conditions was assessed at both anthesis and maturity.

Depending on the environment and sowing date, anthesis occurred between 9 April (SS18) and 16 May (FI18). In terms of cumulative growing degree days from sowing, anthesis ranged from 1318°Cd (FI18) to 1718°Cd (SS21). All lines flowered at approximately the same time, indicating that the tin gene did not affect developmental rate. Compared with the Italian commercial cultivar Bologna, grown in an adjacent field and sown on the same day in 2017 and 2018 at both SS and FI, the Australian lines flowered 8 days earlier at FI and 17 days earlier at SS. Cultivar Bologna can be considered representative of the leading Italian bread wheat cultivars, as it ranked among the top five Italian cultivars in 2018 based on certified seed production (https://www.crea.gov.it/web/difesa-e-certificazione/-/statistiche).

A wide variation in total biomass at anthesis (558–1174 g m^-2) and related traits was observed across environments ( Table 1 ). The presence of the tin gene led to greater biomass, heavier stems, and higher spike dry weight per unit area, without affecting the proportion of fertile stems (about 0.90, regardless of tin presence) or the spike-to-total dry weight ratio (about 0.20). Cultivar Kite exhibited lower spike dry weight than Janz, both in absolute terms and relative to total dry weight.

Any effect of reduced tillering or cultivar on radiation interception by anthesis was ruled out based on data for the fraction of intercepted PAR ( Supplementary Table S1 ), which ranged from a minimum of 0.55 at SS23—where late sowing limited GAI to 1.42—to a maximum of 0.96 at SS18, where GAI reached 6.31, but were not different between NILs (0.82 and 0.81 on average, se= 0.01). This fraction of intercepted radiation was used to calculate the mean daily photothermal quotient (PTQ) for the period between booting and anthesis. We observed a wide variation in mean PTQ across environments, ranging from 0.6 MJ°C^-1 in SS23 (late sowing) to 1.4 MJ°C^-1 in FI17, along with inconsistent and difficult-to-interpret differences between tin and free-tillering lines. As a result, NILs did not differ significantly in mean PTQ values.

Data on nitrogen uptake at anthesis and its partitioning between the spike and vegetative tissues (“stems”) were available only for the four Ottava environments. These data showed no significant effect of the tin gene on N uptake or partitioning ( Supplementary Table S2 ), despite large variation in these traits across environments. Cultivar Janz exhibited greater N uptake than Kite.

The presence of the tin gene had no effect on grain yield, and no interaction was detected between the tin gene and either the environment or parental cultivar for this trait ( Table 2 ). Grain yield ranged from 3.5 ± 0.22 t ha^-1 at SS23, likely due to late sowing, to 6.7 ± 0.22 t ha^-1 at FI17. Yields were comparable to that of the Italian commercial cultivar Bologna, which produced 5.2 ± 0.7 t ha^-1 when grown in adjacent fields and sown on the same day in 2017 and 2018 at both sites.

Plasticity in yield and related traits was assessed by calculating the slope of the regression between trait values and a gradient of environmental conditions ranging from ‘unfavorable’ to ‘favorable’, separately for the two groups of NILs. According to the determination coefficients R², the environmental indexes explained from 76 to 96% of the trait variation. The plasticity of spikes m^-2 was significantly greater in free-tillering lines, indicating that this trait was less responsive (i.e., less plastic) in tin-containing genotypes ( Table 5 ). Conversely, tin lines exhibited greater plasticity for grain weight.

No other traits showed significant differences in plasticity due to the presence of the tin gene, although the difference in spike dry matter at anthesis approached significance (P ≤ 0.055).

GNO was the main determinant of grain yield, and the generally lower GNOs produced by tin lines derived from the inability of their spike fertility to compensate for the lower number of spikes m^-2. But in contrast with previous studies reporting a variable but significantly greater grain number spike^-1 in tin vs free-tillering lines at lower plant population densities (Duggan et al., 2005; Mitchell et al., 2012), we observed no difference in the average grain number spike^-1, signalling that the sowing rate adopted did not allow the full expression of the ‘gigas’ phenotype, i.e. limited tillering plus large and proliferous spikes and robust and vigorous vegetative parts (Atsmon and Jacobs, 1977). Duncan et al. (2005) similarly observed a decrease in tin spike fertility at higher sowing densities.

Analyzing GNO through the Fischer’s (2011) approach (GNO = spike weight m^-2 at anthesis x number of grains g of spike^-1 or ‘fruiting efficiency’) revealed that tin lines were able to build a greater biomass by anthesis thanks to their heavier stems, more than compensating for their lower number of spikes m^-2, and to translate this greater biomass into a greater spike weight at anthesis. The reduced tillering did not compromise their ability to intercept radiation in the critical period for grain number determination (here, roughly considered to coincide with the booting to anthesis period). In fact, although Moeller et al. (2014) observed both a lower leaf area index and lower radiation interception in tin lines compared with free-tillering lines, the relationship between tillering, leaf area index and radiation interception is not straightforward (Duggan et al., 2005; Sadras and Rebetzke, 2013). Therefore, what limited the production of higher grain numbers in tin lines was their lower fruiting efficiency, which prevented any advantage being gained from the greater biomass and spike weight at anthesis. As already observed (Gaju et al., 2009; Motzo et al., 2004), the increase in chaff weight induced by the tin gene is disproportionately greater than the increase in grains spike^-1, resulting in lower fruiting efficiencies. The earlier cessation of bud outgrowth associated with the tin gene during the transition of the shoot apex from the vegetative to the reproductive stage (Kebrom et al., 2012) likely favored the allocation of the diverted assimilates to organs that were growing in that period, while also lengthening the duration of their growth. Rachis and glumes start growing before the florets (McMaster et al., 1992); hence, this earlier availability of resources may have resulted in a greater growth and weight of the spike’s structural tissues and, in turn, a heavier chaff weight, but not necessarily in more fertile florets. A possible avenue to overcome this limitation could be an increase in the number of competent florets per spike weight, and/or in the proportion of competent florets that progress through pollination, fertilization and early grain survival to bear grains at maturity (Fischer, 2011). The association between floral abortion and the 7Ag.7DL translocation in wheat (Reynolds et al., 2005) could be useful in breeding for a higher fruiting efficiency.

The greater grain weight of tin lines was the yield component accounting for both the similar grain yields of tin and free-tillering lines, despite the lower number of grains m^-2 in the former, and the lack of any decrease in grain yield plasticity in tin compared with free-tillering lines.

The mean grain weight recorded here for tin lines (45 mg) was significantly higher than the values reported for most of the Australian environments at lower plant population densities, despite the negative effect of an increase in plant population density on the grain weight in tin lines observed by Mitchell et al. (2013). Moeller and Rebetzke (2017) quoted a mean grain weight of 38 mg for tin lines grown across different water stress environments, whereas the higher grain weight recorded by Mitchell et al. (2013), for their higher plant density treatment (about 200 plants m^-2) cultivated under irrigation, was 31 mg.

The greater anthesis biomass, translocation and grain filling rate associated with the higher grain weight of tin lines was in line with previous observations (Mitchell et al., 2013). Higher rates of grain filling and greater translocation are both important for environments with terminal water stress (Mitchell et al., 2013; Asseng et al., 2019). That is why the SS17 environment, characterized by the most severe water stress, was the one in which the greatest difference between the two NIL groups was observed. On the other hand, the lack of any increase in spike fertility meant that the potentially higher amount of water-soluble carbohydrates present in the denser stems by anthesis (Motzo et al., 2004; Dreccer et al., 2013) were shared between a lower number of grains per spike, allowing them to grow at a higher rate, resulting in higher final grain weights also in the more favorable environments with abundant spring rainfall (FI18, SS18).

Therefore, we can at least partly attribute the higher plasticity in grain weight observed in the tin lines to the exploitation of greater quantities of assimilates and resources per spike in the more favorable environments, which compensated for their lower plasticity in spikes m^-2, as shown by the similar levels of grain yield plasticity in tin and free tillering lines, in contrast with what was observed under Australian agricultural systems (Moeller and Rebetzke, 2017).

Grain protein percentage is the most important component of wheat grain quality, and low protein grains are better suited for feed and not for human consumption (Shewry, 2009). The range in grain protein percentage recorded in the four environments for which nitrogen traits were available lay within those reported by Duggan et al. (2005). Interestingly, the increase in grain weight did not cause a ‘dilution effect’ on grain protein content, and by consequence did not result in a decrease in grain protein percentage, confirming the findings of Duggan et al. (2005). GNO is generally considered the sink for grain nitrogen (Martre et al., 2006). In this sense, the lower GNO of tin lines meant that the similar amounts of nitrogen taken up by anthesis by tin and free-tillering lines were shared by a lower, although heavier, number of grains at maturity.