Selection for yield over five decades favored anisohydric and phenological adaptations to early-season drought in Australian wheat

Climate change in the Mediterranean-climate region of Australia is reducing growing season rainfall and delaying first autumn rain or the onset of ‘autumn break’. We tested the hypothesis that selection for yield and agronomic traits has favored adaptation to early season drought in Australian wheat (Triticum aestivum L.). Ten wheat varieties released between 1958 and 2012 were grown in a glasshouse. After sowing in dry soil, the equivalent of 25 mm rainfall was supplied, with no subsequent watering provided for 32 days to induce an early season drought treatment (ESD) while a well-watered treatment (WW) was planted on a wet soil that was water-saturated 48 h before sowing. We measured soil and plant water status, gas exchange, shoot and root traits at the end of drought (32 days after sowing) and at anthesis, and grain yield per plant at maturity. Grain yield increased with year of release at 0.43% yr–1 under well-watered conditions and at 0.35% yr–1 under drought. The improved yield under drought was associated with a shorter time to flowering, and a change from isohydric behavior (maintained Ψleaf, reduced gs, leaf photosynthesis and transpiration rates in response to drought) in older varieties to anisohydric behavior (decreased Ψleaf and increased gs, leaf photosynthesis and transpiration in response to drought) in newer varieties that reduced leaf area and maintained higher gs, and higher photosynthesis per unit leaf area. Direct selection for yield and agronomic traits between 1958 and 2012 has improved adaptation to early-season drought. Our collection of varieties is an interesting model to probe for variation in drought tolerance.


Introduction
Wheat will remain critical for food security in the foreseeable future, but drought limits its production in major mega-environments worldwide (Shiferaw et al. 2013;Fischer et al. 2014). Adaptive traits need to be tailored to the specific timing, intensity, and duration of water stress (Jordan & Miller 1980;Tardieu 2012), hence the importance of probabilistic, spatial quantification of drought patterns (Chenu 2015). Modeling the seasonal dynamics of water supply and demand using historic climate series and climate projections (2030,2070) returned four types of drought for wheat in Australia (Chenu et al. 2013;Watson et al. 2017) linked to non-stress condition, extreme drought with an onset at 500 o Cd before flowering, and increasing stress as the season progresses. However, modeling water supply, demand and water use at seedling stage, early in the season is unreliable as crop canopies are small, impeding the quantification of early-season drought (K. Chenu, personal communication, March 2021).
Climate change in the Mediterranean-climate region of Australia is increasing maximum temperature, reducing minimum temperature and growing season rainfall, and delaying the onset of 'autumn break' or the first good rains of the southern winter cropping season (Pook et al. 2009;Hochman et al. 2017;Cann et al. 2020). In response to the delayed autumn break, many Australian wheat growers are sowing their entire cropping land before the autumn break, further increasing the risk of early season drought (Fletcher et al. 2015(Fletcher et al. , 2016. In Western Australia, wheat crops sown into dry soil germinate and emerge after the first 20-25 mm rainfall, potentially leaving crops vulnerable to 20-23 days drought after emergence in two of four years and 30-32 days drought in one of four years (Chapman and Asseng 2001). While seedlings can survive early season drought, their growth and development are slow, reducing root and shoot biomass at anthesis and grain yield (Armstrong et al. 1996;French and Palta 2014).
Comparisons of varieties in historic collections reveal changes in the crop phenotype associated with selective pressure for yield and agronomic performance (Austin et al. 1980;Slafer 1994;Roche 2015). A collection of wheat varieties adapted to winter-rainfall in Australia showed that five decades of breeding favored (i) a communal phenotype with reduced competitive ability (Sadras and Lawson 2011;Cossani and Sadras 2021), (ii) higher crop photosynthesis associated with more erect canopies that favor light penetration, higher nitrogen uptake and greener leaves (Sadras et al. 2012a;Cossani and Sadras 2021), (iii) higher nitrogen-water co-limitation associated with a higher N uptake per unit of evapotranspiration (Sadras and Lawson 2013;Cossani and Sadras 2019), and (iv) a smaller root system with enhanced nitrogen uptake per unit root length (Aziz et al. 2017). In contrast to the adaptive role of a smaller root system (Aziz et al. 2017), a comparison of three varieties with exceptionally large and small root systems showed that larger root systems associated with late flowering adapted to severe early season drought allowing greater recovery in leaf area and shoot and root biomass by anthesis (Figueroa-Bustos et al. 2019).
Here, we used our historical collection of Australian wheats to test the hypothesis that selection for yield and agronomic traits in winter-rainfall environments has improved early-season drought resistance in Australian wheat varieties selected in the 1958-2012 period. We measured phenological and physiological traits to identify shifts associated with the putative improvement in drought tolerance.

3
Vol.: (0123456789) Plants were grown in 1.0 m deep PVC columns (0.15 m diameter) filled with soil, with 4 cm of gravel at the bottom to facilitate drainage. The soil was a reddishbrown sandy clay loam, Red Calcic Dermosal (Isbell 1993), collected from the top 0-15 cm of soil in a field at Cunderdin (31°64' S, 117°24' E), Western Australia. It comprised 63.5% brown sand, 8.3% silt and 28.3% clay with pH, 6.0 measured in a 1:5 suspension of soil in 0.01 M CaCl 2 . The soil contained 6 μg g −1 nitrate-N, 4 μg g −1 ammonium-N, 46 μg g −1 Colwell P, and 691 μg g −1 Colwell K. Air-dried soil was sieved to 2 mm and mixed with coarse yellow sand (200-2000 µm particle size) in a 4:1 ratio using a cement mixer for uniformity. The soil was packed slowly into each column from the bottom to the top by sections of 0.20 m length to a bulk density of approximately 1.50 g cm -3 . At sowing, the equivalent of 60 kg N ha -1 as urea; 72 kg P ha -1 as amended superphosphate (with Cu, Zn, Mo, S), and 70 kg K ha -1 as potash was mixed into the top 0.1 m of soil in each column. These rates are recommended for wheat on Cunderdin soil (Flower et al. 2012). Four uniform seeds were sown along a center row in each column and thinned to two per column at the two-leaf stage.

Treatments
We established an experiment with 10 varieties and two watering regimes [well-watered (WW) and early season drought (ESD)] in a completely randomized design with four replicates for each variety and treatment. The columns were rotated weekly to minimize spatial variability.
Forty-eight hours before sowing, the well-watered columns were slowly watered by hand to saturation, and the soil surface was covered with aluminum foil to prevent evaporation. From sowing to 32 days after sowing (DAS), the well-watered treatment received 200 mL water per week to maintain the soil water content to 90% field capacity. The columns of the drought treatment were maintained dry until sowing. Immediately after sowing, the ESD columns were watered with the equivalent of 25 mm rainfall (~ 441 mL water), which wet the top 0.12-0.15 m soil to a volumetric water content ~ 20%; no further water was applied for 32 days. After this time and until physiological maturity (Z91) (Zadoks et al. 1974), both the well-watered and droughted plants were equally watered by hand twice per week to maintain the soil water content close to 90% field capacity and to avoid drainage of excess water. Phenology, soil water content, leaf water potential, stomatal conductance and leaf photosynthesis Phenological development was monitored regularly using Zadok's scale (Zadoks et al. 1974). At 32 DAS for both WW and ESD treatments, before terminating the drought treatment, we measured volumetric soil water content, leaf water potential (Ψ leaf ), stomatal conductance (g s ), and leaf net photosynthetic rate. Volumetric soil water content in the top 0.15 m of the soil profile was measured with a 15 cm soil moisture probe connected to a water content sensor (CS658 HydroSense II Water Content Sensor, Campbell Scientific, Australia). Measurements of Ψ leaf , g s , and leaf net photosynthesis rate were made on the last expanded leaf of the main stem on four replicate columns (two plants for each column), between 10:30 and 14:00 on a day with clear sky. Rates of leaf net photosynthesis and g s were measured with a LI-COR gas-exchange system (LI-6400XT, LI-COR Bioscience, Nebraska, USA) with an LED light source on the leaf chamber. The cuvette CO 2 concentration was set to 380 µmol -1 and LED light intensity to 900 µmol m -2 s -1 , the average saturation intensity for photosynthesis in wheat (Austin 1990). Immediately after these measurements, Ψ leaf was measured using a Scholander pressure chamber (model 1000, PMS Instrument Co., Oregon, USA). The last expanded leaf of the main stem was loosely covered with a plastic sheath before excision and during the measurement to avoid evaporation (Turner 1988).

Shoot traits
We measured leaf area, leaf biomass, specific leaf area (SLA), tiller number, and shoot biomass for each variety and treatment at 32 DAS, before terminating the drought treatment, and at anthesis (Z61), and shoot biomass, yield and yield components at final harvest. At each sampling time, four columns (8 plants) per variety and treatment (4C x 2 T) were harvested and each column (two plants) served as a replicate. Shoots were cut at the crown and leaf area was measured using a portable leaf area meter (LI-3000, Li-COR Biosciences, Lincoln, NE, USA). The number of tillers was recorded, and stems and leaves were dried in an oven at 70 °C for 48 h and then weighed to determine dry weights. Shoot biomass, grain yield and yield components were measured at maturity. Spikes 1 3 Vol:. (1234567890) per plant were counted before separated from shoots, oven-dried at 60 °C for 48 h, and then threshed by hand. The number and weight of grains per plant were recorded. Harvest index (HI) was calculated as the ratio of grain yield to total shoot biomass.

Root traits
We measured total root length, total root biomass, total root length density (root length per unit of soil volume; RLD), and specific root length (SRL, root length per unit of biomass), which is an indirect measure of root thickness. Measurements were made at 32 DAS, just before terminating the drought treatment, and at anthesis (Z61) for each variety. Immediately after harvesting the shoots, the soil profile in each column was sampled from top to bottom. The roots in each column were recovered from the soil by washing and repeated sieving through a 1.4 mm sieve to produce a clean sample as described by Palta and Fillery (1993). After the roots were recovered from a section of the soil profile, they were placed in plastic bags and stored at 4 °C until measurement, often two days later. The root length in each root sample was measured as described by Liao et al. (2004). Briefly, roots were stained for 30 min with 0.1% (w/v) methyl blue prior to analysis, placed in a 0.2 × 0.3 m glass tray in about 3 mm water, and untangled to avoid any overlap. The glass tray was placed on the scanner, and the roots were scanned at 600 pixels per mm using an Epson scanner (ES2200) connected to a computer. The root material was then dried at 70 °C for 48 h and weighed. The images were analyzed for root length using WinRHIZO 2009 (model Pro, second version, Regent Instruments, Québec, Canada). Total root length density was calculated as the total root length to 1.0 m of the soil profile divided by the soil volume in column (0.0176 m 3 ).

Statistical analysis
The data were analyzed using SPSS 21.0 Statistical Analysis System, Origin Lab 10.0 and Microsoft Excel 2013. Chronological trends of phenotypic traits were tested using least-square regressions of trait deviation vs year of release. Absolute (g plant -1 yr -1 ) and relative (% yr -1 ) rates of change for each trait were calculated as the slope of the least-square regression between the trait and year of release. The relative rate was calculated in relation to the newest variety (Fischer et al. 2014). We report p-value as a continuous quantity (Greenland 2019).
We calculated the drought susceptibility index (DSI) (Fischer and Maurer 1978) as follows: where Y is yield, d and w are drought and wellwatered respectively, and D is drought intensity = 1 -(mean Y d of all varieties/(mean Y w of all varieties).

Results
Soil water content, leaf water potential, stomatal conductance and leaf gas-exchange at the end of the drought period Figure 1 and Table 1 summarize soil water content and water-related traits at 32 DAS, just before terminating the drought treatment. Under well-watered conditions, topsoil (0-0.15 m) volumetric water content was high (14.5-16.2%) and did not vary with year of release of the varieties (Fig. 1a). Consistently, leaf water potential (Ψ leaf ), stomatal conductance (g s ), leaf photosynthetic rate and leaf transpiration rate were maintained high under well-watered conditions ( Fig. 1b-e, Table 1). While Ψ leaf and leaf transpiration rate did not vary with year of release, g s and leaf photosynthetic rate varied non-linearly, increasing until mid-80 s with an apparent decline afterwards ( Fig. 1c-d).
Under drought, the volumetric soil water content ranged from 7.6-9.0% and decreased with year of release. Newer varieties dried the soil more than their older counterparts ( Fig. 1a; Table 1). Ψ leaf decreased linearly with the year of release from -1.29 to -1.80 MPa at a rate of -0.008 ± 0.002 MPa yr -1 (Fig. 1b, Table 1). Stomatal conductance, leaf net photosynthetic rate, and leaf transpiration rate increased linearly with year of release ( Fig. 1c,d,e; Table 1).
Leaf area, tiller number, and shoot biomass at the end of the drought period Under well-watered conditions, leaf area varied nonlinearly, increasing initially until 1969, with a subsequent decline until 1990, with an apparent increase afterwards (Fig. 1f, Table 1). Under drought, leaf area decreased linearly with the year of release from 63.2 to 30.6 cm 2 (1) DSI = 1 − Y d ∕Y w ∕D plant −1 at a rate of 0.58 ± 0.11 cm 2 plant -1 yr -1 ( Fig. 1f; Table 1). Tiller number decreased linearly with year of release from 2.8 to 1.0 tillers plant −1 in wellwatered plants and 1.12 to 0.30 tillers plant -1 under drought (Fig. 1g, Table 1). The rate of decline in tiller number with year of release was 2.8-fold higher under well-watered conditions than under drought (Fig. 1g, Table 1). Under well-watered conditions, shoot biomass decreased with year of release from 0.34 to 0.20 g plant −1 g plant -1 yr -1 ( Fig. 1h; Table 1), while under drought, shoot biomass ranged from 0.15 and 0.20 g plant -1 , and did not vary with year of release (Fig. 1h, Table 1).
Root traits at the end of the drought period Figure 2 and Table 1 summarize the variation in root traits under the two watering regimes at the end of the drought treatment. Total root length decreased linearly with year of release under well-watered and drought conditions (Fig. 2a). Under well-watered conditions, total root length decreased from 9.8 m plant -1 in Heron, the earliest variety in the series, to 5.1 m plant -1 in Mace, the most recent variety, while under drought, total root length decreased from 7.3 to 3.8 m plant −1 . The rate at which total root length declined with year of release was faster under well-watered conditions than under drought  Table 1 ( Table 1). Root biomass and root length density similarly declined with year of release under well-watered and drought conditions (Fig. 2 b, c). The decline in RLD with year of release was greater under well-watered conditions than under drought (Table 1). The root: shoot ratio declined linearly with year of release under drought from 0.28 in Heron to 0.17 in Mace ( Fig. 2d; Table 1). Under well-watered conditions the root: shoot ratio showed no trend (Fig. 2d, Table 1).

Shoot and root traits at anthesis
At anthesis of each variety, leaf area declined linearly with year of release in well-watered and droughted plants (Fig. 3a). Leaf area of well-watered plants declined from 826 cm 2 plant -1 in Heron to 655 cm 2 plant −1 in Mace at a rate of 3.9 ± 0.62 cm 2 plant -1 yr -1 . Leaf area of droughted plants declined from 476 to 332 cm 2 plant -1 at a rate of 2.9 ± 0.49 cm 2 plant -1 yr -1 ( Table 2). The number of tillers and shoot biomass similarly declined linearly with year of release in well-watered and droughted plants ( Fig. 3b; Table2). The decline could be associated with the fact that modern genotypes showed an early anthesis time respect to the older ones. The decline in tiller number with year of release was sharper in well-watered plants than droughted plants (Table 2).
Total root length, root biomass, root length density, and the root: shoot ratio at anthesis declined  linearly with year of release in both well-watered and droughted plants ( Fig. 3d-g, Table 2).

Phenology
Wheat varieties released between 1958 and 2012 emerged at 7-10 DAS under both watering regimes, with 1-3 d difference among the varieties. Time to 50% anthesis declined linearly with year of release in well-watered and droughted plants (Fig. 4a). Time to anthesis in well-watered plants decreased by 23 days, from 103 d in Heron to 80 d in Mace, while in droughted plants declined from 91 to 77 d. The rate at which time to anthesis declined was faster in well-watered than in droughted plants (Table 3). Time to physiological maturity (Z91) declined with year of release at similar rates in both watering regimes (Fig. 4b, Table 3). Duration of grain filling did not vary with year of release in well-watered plants but ranged from 39 d in Condor to 42 d in Krichauff. In droughted plants, duration of grain filling decreased linearly with year of release from 42 to 33 d at a rate of 0.13 ± 0.009 d yr -1 ( Fig. 4c; Table 3).

Grain yield and yield components
Grain yield increased with year of release under both well-watered and early season droughted conditions (Fig. 5a). The rate of increasing grain yield with year of release was faster under well-watered conditions than under drought (Table 3). Shoot biomass at final harvest showed no trend with year of release under both watering regimes ( Fig. 5b; Table 3). Harvest index increased linearly with year of release at similar rates for well-watered and droughted plants (Fig. 5c, Table 3). Ear number, grain number, and grain weight did not vary with year of release irrespective of watering regime  Table 3). There was a strong correlation between the mean grain yield of nine of the ten cultivars used in this study, measured in field plots in three sites in South Australia (Sadras and Lawson 2011) and the grain yield from plants grown in 1.0 m deep PVC columns under early season drought (ESD) and well-watered (WW) conditions in the glasshouse in this this study (Fig. 5g).
The yield-based DSI (Eq. 1) decreased linearly with year of release from 0.43 in Heron to 0.26 in Mace, at a rate of 0.003 ± 0.0003 yr −1 (Fig. 5g, Table 3).
Relative rates of change for traits of Australian wheat varieties released from 1958 to 2012 are presented in Tables 1, 2, and 3, together with the absolute rates of change with year of release of the varieties. While the absolute rates of change are the regression coefficients, which have agronomy relevance, the relative rates involve the analysis of change relative to the latest cultivar released, which have physiological importance. The relative rate is a standardized measure of trait change that allows the performance of cultivars to be compared on an impartial basis.  Table 2 1 3 Vol.: (0123456789)

Discussion
Wheat breeding focuses primarily on yield, agronomic adaptation, grain quality, and disease resistance (Richards et al. 2014). This selective pressure leads to predictable changes in phenotype, such as increased HI (Slafer et al. 2021), and some unexpected changes, such as reduction in root biomass compensated by increased nitrogen uptake per unit root length (Aziz et al. 2016). In our study, the phenotyping of a collection of wheat cultivars representing breeding efforts from 1958 to 2012 showed improved grain yield with year of release in plants under early season drought associated with shifts in stomatal conductance, leaf net photosynthesis and transpiration rate and phenological development. Grain yield in this study was measured in 8 plants per variety and treatment (4 pots of two plants each) grown in 1.0 m deep PVC columns in a glasshouse. Although these growing conditions may have resulted in a more severe early season drought than for field crops (Turner 2019), the mean grain yield of nine of ten varieties grown in this study positively and strongly correlated with that of the same varieties grown in the field at three sites in South Australia (Sadras and Lawson 2011). Despite the different growing conditions, the correlation confirms an improved grain yield in wheat varieties released between 1958 and 2007 in Australia (Sadras and Lawson 2011).

Selection for yield favored an anisohydric phenotype
The response to early season drought revealed a change in the continuum from isohydric to anisohydric phenotypes (Tardieu and Simonneau 1998;Blum 2015). Under the early season drought, the increase in soil water use at the top 0.12-0.15 m soil surface with the year of release, decreased Ψ leaf and increased g s , leaf photosynthesis and transpiration rates (anisohydric behavior), but shoot biomass was unchanged with the year of release. Old and modern cultivars in the series have almost similar shoot biomass despite it was expected modern cultivars to have higher shoot biomass as their leaf photosynthesis was higher Jensen et al 1989) than the older ones. However, leaf area and tiller number, two contributors to shoot biomass, decreased with year of release and likely the benefit in the modern cultivars of having a higher leaf photosynthesis rate for gaining shoot biomass was offset by the reduction in leaf area and tiller number. Despite the reduction in leaf area and tiller number with the year of release, the relative reduction of shoot biomass in recently released cultivars (drought/well irrigated) was smaller than in old cultivars, indicating that high leaf photosynthesis rate with high stomatal conductance and transpiration rate would contribute to maintenance of shoot biomass under early season drought in recently released cultivars. Older phenotypes reduced g s and maintained Ψ leaf at the expense of leaf photosynthesis and transpiration per unit leaf area (isohydric behavior) and did not reduce leaf area and tiller number, maintaining shoot biomass. Newer phenotypes featured a more anisohydric behavior maintaining g s , leaf photosynthesis and transpiration per unit leaf area at the expense of Ψ leaf . Similarly, Pima cotton (Gossypium barbadense L.) bred for higher yield potential had higher g s and smaller leaf area than earlier lines .
We did not measure osmotic adjustment but hypothesize that osmotic adjustment might have contributed to the maintenance of g s and leaf photosynthesis under decreasing Ψ leaf in newer phenotypes. Osmotic adjustment, which results from an active accumulation of solutes in response to decreasing Ψ leaf under water deficit (Turner and Jones 1980), varies among wheat genotypes (Morgan and Condon 1986;Blum et al. 1999;Moinuddin et al. 2005). However, the marked reduction in leaf area of newer phenotypes under drought is evidence against superior osmoregulation (Blum 1989;Nio et al. 2011Nio et al. , 2018. Further experiments that measure osmotic potential are needed to solve these questions. The higher g s of newer varieties under drought is a widespread response of crops to selection for yield (Roche 2015). In some crop-environment settings, high g s has been interpreted as an adaptation to elevated temperature at the expense of water use efficiency Lu et al. 1994;Sadras et al. 2012b;Galat Giorgi et al. 2020).
Selection for yield favored smaller root system Consistent with Aziz et al. (2016), we found a decline in root length, root biomass, and root length density with year of release in well-watered plants and it is likely that the decline is a consequence of shorter time to anthesis in modern cultivars. However, early season drought reduced root traits more markedly in older varieties than newer ones. Since drying soil often triggers seminal root growth that favors water uptake in some wheat genotypes (Palta and Gregory 1997;Whitmore and Whalley 2009;Hodgkinson et al. 2017), it is likely that, seminal roots of recent released varieties grew deeper in dry topsoil and slowed the decreasing rate with year of release. This is consistent with the early idea that phenotype must match the environment, and drought adaptation is associated with smaller root systems (Jordan and Miller 1980;Tardieu 2013). Although plants exposed to early-season drought were well-watered after 32 DAS, the recovery of root length, root biomass and RLD at anthesis was unremarkable for all varieties. This was likely because  Table 3 root system growth, which usually peaks at flowering (Gregory and Atwell 1991;Gregory et al. 1995;Palta and Gregory 1997), run-off time as time to anthesis shortened. Moreover, wheat genotypes with longer time to anthesis have larger root growth (total root length and root biomass) at maturity than those with shorter time to anthesis (Motzo et al., 1993;Aziz et al. 2016;Figueroa-Bustos et al. 2019.

Selection for yield reduced phenological response to drought
Early field studies with a similar collection of varieties revealed no trend in time to flowering with year of release Lawson 2011, 2013). In contrast, our glasshouse study showed earlier flowering in newer varieties in the well-watered treatment. The regressions between time to flowering and year of release converged between well-watered and droughted plants. Older varieties had a marked shortening of time to flowering in response to drought, whereas newer varieties were less responsive, despite being more stressed. In contrast, two independent studies under field and glasshouse conditions have reported delayed anthesis of wheat in response to early-season drought (French and Palta 2014;Figueroa-Bustos et al. 2019). The non-linear response of wheat phenology to water stress could explain these contradictory responses. Angus and Moncur (1977) found that time from flower initiation to anthesis shortened under mild water stress (Ψ leaf -1.5 MPa) and delayed with severe stress (Ψ leaf -2.5 MPa). In our study, a mild stress (Ψ leaf -1.8) shortened time to flowering, whereas more severe stress (Ψ leaf -2.5 MPa) delayed flowering in Figueroa-Bustos et al. (2019). Our findings highlight the need for simulation modeling to account for genotype-dependent phenological response to water deficit (Chauhan et al. 2019;McMaster et al. 2013) because mainstream crop models do not (Wallach et al. 2021). Overlooking the effect of plant water status on phenology can therefore bias predictions of crop adaptation to future climates, including early-season drought.
Interesting, shoot biomass at maturity did not change with year of release, but grain yield increased with year of release and the increase was associated to HI. The ratio of shoot biomass under well-watered to early season droughted was similar (~ 0.66) regardless of year of released, likely indicating that an increase in the potential yield through an increase in grain number in the recent released cultivars was a reason of high yielding even if they experienced early season drought. The shortening of grain filling with drought often reduces HI and yield (Kobata et al. 1992;He and Rajaram 1993;Wheeler et al. 1996). In our experiment, plants exposed to early season drought, maintained duration of grain filling in older varieties, but shortened it in newer ones. Drought did not affect HI since the proportional reduction in grain yield and shoot biomass was similar under well-watered and drought conditions. Modern cultivars had higher grain yield associated with higher HI than the old cultivars, presumably because of a higher amount of reserve carbohydrates and/or faster rates of remobilization to the grain in newer varieties (Kobata et al., 1992;Palta et al. 1994).
As reported in Aziz et al. (2017), the root system of wheat varieties released in Australia between 1958 and 2012 progressively reduced their root length, root biomass, and RLD with year of release. The rate of  Table 3 Plant Soil (2022) 476:511-526 522 decrease was 0.059 ± 0.005 m pl −1 yr −1 for total root length and 0.0017 ± 0.0002 cm cm 3 yr −1 for RLD. The reduction in root length, root biomass, and RLD with year of release was likely associated with the reduction in time to anthesis in the modern cultivars compared to the older ones. However, early season drought slowed down the rate of decline in these root traits. Since drying soil often triggers seminal root growth in some wheat genotypes (Palta and Gregory 1997;Whitmore and Whalley 2009;Hodgkinson et al. 2017), seminal root tips of recent varieties likely grew deeper than older varieties with decreasing soil water content in the topsoil layer, and this additional root growth slow down the rate of declining in root length, root biomass and RLD with year of release.
Early season drought reduced root length, root biomass, and RLD, less so in the varieties released after 1997 (Wyalkatchem, Gladius, Mace) than earlier varieties (Heron, Halberd, and Condor). This was likely because the recent varieties kept leaf photosynthesis and transpiration rate and reduced leaf area during early drought, with more available carbon assimilates for root growth, when the demand of daily carbon assimilates for wheat root system growth and proliferation was high (Gregory and Atwell 1991;Palta and Gregory 1997).
Under early drought, the recent released varieties had a higher capacity to maintain shoot biomass than older varieties, despite their shoot biomass potentiality decreased. It is unknown whether the higher capacity to maintain shoot biomass came from the high capacity of g s and leaf photosynthesis rate, as leaf area, one of factors determining whole plant photosynthesis decreased. It is likely that root biomass and length did not contribute to the high capacity to maintain shoot biomass, despite it is thought that a low root biomass gives a source for an increase in shoot biomass. At anthesis shoot biomass, root biomass and root length in plants exposed to early season drought was lower in recent released varieties than in older ones, associated with the differences in time to anthesis between recent and older released varieties (Aziz et al. 2017).
Under early season drought, the root: shoot ratio decreased linearly with year of release, and it is likely that the change in root: shoot ratio from high in the older cultivars to low in the recent released ones, increased the transpiration rate in recent released cultivars. This is because a high root: shoot ratio can decrease transpiration rate as a large root system provides more water to the plant while a less large shoot will provide less evaporation. A low root: shoot ratio could increase transpiration rate as the supply of water through the root system will be not enough due to increased evaporation by a larger shoot system (Saidi et al. 2010;Vadez et al. 2021). Indeed, the linear decrease in root: shoot ratio with year of release under early season drought was consistent with the linear increase in leaf transpiration rate with year of release. Under drought the root: shoot ratio of old varieties was higher as shoot biomass was more negatively affected than root biomass. This trend was different at anthesis as the root: shoot ratio decreased similarly in both modern and old varieties with the year of release and at similar rate under drought ad well-watered conditions.
Terminal drought or drought after anthesis often occurs in the Mediterranean climate zone of Australia. An interesting question is what the contributions from the performance under early season drought are to yield in the case terminal drought occurs?. Clearly this question cannot be answer with the results from this study, but the less effect of early season drought in shoot biomass at anthesis is likely to give a benefit on the yield under terminal drought. There is no unique and obvious answer to this question, as the contributions from the performance under early season drought to yield depends on the time, intensity and rate of development of the terminal drought. This can be only speculated that grain yield might be reduced if duration of grain filling is further reduced in the recent released cultivars, but if there is not room for further reductions in duration of grain filling, grain yield might be maintained. We may also speculate that grain yield under terminal drought could be less affected if biomass accumulation between the end of the early season drought and anthesis is high.

Conclusions
Wheat varieties released in Australia between 1958 and 2012, exposed to early-season drought after leaf emergence, progressively increased grain yield with year of release. The decreasing susceptibility index to early season drought with year of release was associated with a change from isohydric to anisohydric phenotypes and distinct phenological adaptation to early season drought. The higher g s and leaf photosynthesis rate of anisohydric varieties contributed to the increased grain yield with year of release since the early water deficit was moderate with Ψ leaf down to -1.8 (Alvarez et al. 2007;Sade et al. 2009). Selection for grain yield and agronomic adaptation has favored traits for early season drought tolerance in Australian wheat. The wheat collection used in this study is a suitable model to further investigate the phenotypic response to drought, including the dual role of labile carbohydrates in osmoregulation and as a buffer for grain fill and its genetic basis.