The response of soybean biomass and seed yield to variable seasonal water is linear. Discerning the genetic basis of water use efficiency on a seed yield basis (i.e., WUEy, but termed yield beta in this report) has long been a key research objective. Yield beta data have been collected on five RIL mapping populations of 250-300 members each, but RIL marker-genotyping is incomplete in three. A few yield beta QTLs have been detected to date, but their map positions have coincided with QTLs governing genotypic mean yield, and/or with genes governing plant growth habit and maturity. A carbon isotopic composition (δ13C) estimation of transpiration efficiency (TE), conducted in two populations, identified a strong δ13C QTL, but its map position was coincident with that of the growth habit gene. In all five populations, the yield of a genotype in a water-abundant environment was highly correlated with its yield in a water-scarce environment, which is expected when genotypic yield beta and mean yield are governed in a coupled manner by either linked or pleiotropic QTLs, as seems to be the case. If so, then genetic yield advances in water-abundant environments are expected to generate concomitant yield advances in water-scarce environments (though to a lesser degree). Thus, genetic yield improvement for rainfall-variable soybean production areas (i.e., Midwestern USA), where extremes in water-abundance and water-scarcity are equally probable, is probably best attained by elevating the slope of the yield-to-water response (i.e., WUEy), while concurrently limiting the potential for such elevation to cause a downward slippage of the low end of the response slope.
Irrigation, water stress, low rainfall, harvest index.
Water, when it is scarce, is often viewed as being an abiotic stressor. However, water should really be treated as just another resource (i.e., light, CO2, etc.) that must be efficiently exploited – even more so when crop productivity is constrained by its scarcity. Crop biomass (BM) is a highly linear function of the amount of water transpired (T) by the crop. Simply put, plants must exchange water for carbon. The proportionality of that exchange is determined by a biomass-based water-use efficiency coefficient (WUEb, defined as ΔBM / ΔT), to wit: BM = T x WUEb. Seed yield (SY) is similarly related to T, if harvest index (HI) - the seed fraction of total biomass - is incorporated into the equation, to wit: SY = T x WUEb x HI (Fig. 1, left). This equation was first postulated by Passioura (1977). Because WUEb x HI = WUEy (a yield-based WUE, defined as ΔSY / ΔT), the equation translates into: SY = T x WUEy. This equation clearly shows that if soybean breeders are to increase SY (beyond any genetic improvement still possible in the asymptotically limited HI parameter), genotypes must be selected that will (1) transpire more of the available annually rechargeable soil water, and/or (2) exhibit a greater WUEy, which means acquiring more carbon per unit of exchanged water (Purcell and Specht, 2004). Specht et al (1986) evaluated the seed yield response of 16 soybean cultivars to six seasonal water amounts by replenishing potential crop evapotranspiratory (ET) loss on a weekly basis at levels ranging from 0% to 100%. A severe drought fortuitously occurred in both test seasons. Regression of final cultivar seed yield on seasonal water amount demonstrated rather convincingly that seed yield was a linear function of seasonal water amount (E+T). The cultivar regression coefficients (termed genotypic yield betas by the author) were season-specific estimators of the cultivar WUEy. Notably, the cultivars A3127 and Williams 82 had the most extreme yield beta values (Fig. 1, right), and this was consistent with anecdotal perceptions relayed to agronomists by producers as to the performance of these two cultivars in water-abundant and water-scarce production environments. Discerning the genetic basis of yield beta – a quantitatively inherited trait – was of great interest, but applying classical quantitative genetics tools was unappealing. Instead, the focus shifted to the development of a soybean molecular marker linkage map and mapping populations of recombinant inbred lines (RILs). With these tools, the genetics of yield beta could be examined at a sub-genomic level, via marker-facilitated, quantitative trait locus (QTL) analyses.
Figure 1. Left Panel: A graphical depiction of the Passioura (1977) thesis that seed yield (SY) is a function of the amount of water transpired (T), a biomass (BM)-based water-use efficiency (WUEb) coefficient, and a harvest index (HI) coefficient reflecting the seed fraction of total crop BM. Note that WUEb x HI = WUEy. The WUEb and WUEy linear regression coefficients for the BM and SY trend lines reflect what a C4 or C3 photosynthetic crop species might display given theoretical ci/ca values of 0.7 and 0.3 respectively (Sinclair, 1994). The SY trend lines reflect an HI = 0.45 (solid) or HI = 0.35 (dashed). The SY (right) axis is scaled to be 0.5 of the BM (left) axis. Right Panel: Measured yield responses of soybean cultivars to a 0, 20, 40, 60, 80, or 100% weekly replenishment of all potential evapotranspiratory (ET) water loss. Regression data shown only for cultivars A3127 (highest yield beta) and Williams 82 (lowest yield beta); both had near-similar HI values. The 16-cultivar mean value and bracketing max/min range values are plotted at each of the six water levels. The June-September rainfall of 24.9 cm at the test site was ca. 66% of the rainfall norm (vertical line). The plotted data are from Specht et al (1986).
When the first RIL mapping population became available, the 267 RILs, two parents, and 19 elite (check) cultivars were evaluated for yield beta. A severe drought occurred in one of the test seasons (Specht et al, 2001). A coordinate graph of yields of the 290 genotypes in the 0%ET vs. 100% ET treatments provided an uncluttered means of visualizing the differences among 290 genotypes in terms of their yield response to water-scarce and water-abundant conditions (Fig. 2, left). The following observations were derivable from the yield beta data. (1) The genotypic yield range in the 0% ET treatment (i.e., ca. 0 to 2000 kg/ha) was significantly constrained compared to that of the 100% ET treatment (i.e., ca. 250 to 4500 kg/ha). Rossielle and Hamblin (1981) noted some 20 years ago that a given set of genotypes will ordinarily exhibit a smaller genetic variance in stress environment than that same set will exhibit in a non-stress environment. Breeders know well that if genetic yield variance is restricted, genetic yield gains from selection will also be lessened. (2) Genotypes occupied graph positions farther from the 1:1 diagonal line the greater their yield potential, which is consistent with the Rossielle and Hamblin theorization that selection aimed at genetically reducing the difference between non-stress and stress yield would almost certainly lead to lower rather than higher yielding genotypes. (3) Genotypic yields in the water-abundant and in the water-scarce treatments were highly correlated (r = 0.85), indicating that direct selection for yield improvement in the former will certainly not be incompatible with a concordant indirect improvement of yield in the latter. (4) The yield beta values of 40-65 kg/ha cm that were estimated for the early-maturing elite cultivars in this test (Fig 2, right), were somewhat lower than the 50-89 kg/ha cm values detected for the later-maturing elite cultivars in the earlier test (Fig. 1, right), indicating that shortening the days to maturation lessens yield beta. (5) As shown in the Fig. 2 (right) graph, the yield beta of a genotype correlated well with its mean yield performance over all six environments (r = 0.71). Though not graphed here, the yield beta correlation with yield in the 100% ET treatment (r = 0.84) was higher than that found in the 0% ET treatment (r = 0.48) (Specht et al, 2001), because the magnitude of genotypic yield-to-water slope was dependent upon the genetic yield variance at the slope ends, and (as already noted above) that variance was smaller at the lower end of the slope. The foregoing generalizations have been applicable to the yield beta data collected in four other mapping populations (data not yet published because marker genotyping is not yet complete). A preliminary analysis of yield beta data collected in Nebraska in 2003 (a severe drought year) on 42 elite cultivar checks and 350 germplasm bank accessions (from drought-prone or highly productive regions of China) now suggests that those generalizations may, in fact, extend to populations of unrelated genotypes.
Figure 2. Left Panel: A graph of non-irrigated (0% ET) and fully irrigated (100% ET) seed yields of the Minsoy and Noir 1 cultivars, their (267) RIL progeny of the determinate (D) or indeterminate (I) stem growth habit type, and 19 unrelated commercial cultivars. Trend lines provided for the two RIL progeny types and cultivar checks. Right Panel: A graph of mean yield versus yield beta. See Specht et al (2001) for additional details.
A carbon isotopic composition (δ13C) analysis procedure was used to evaluate transpiration efficiency (TE) in the first soybean RIL population (Specht et al, 2001), and also in a second population (manuscript in preparation). Both data sets resulted in the same generalizations, so only the first set of leaf δ13C data are presented (Fig. 3). Note that that leaf δ13C values were more positive (i.e., indicative of greater TE) in the 0% ET treatment, suggesting that soybean TE increases naturally in response to an increasing soil water deficit. Still, the correlation between the leaf δ13C value of a genotype and its yield was low in both water regimes. The δ13C values for the determinate RILs were likely artefacts, since juvenile (carbon-importing) leaflets were not ontologically available on determinate stem tips on the sampling date (i.e., older carbon-exporting leaflets were inadvertently sampled instead). Leaf δ13C (and thus TE) was positively correlated with yield beta, but not to the extent (r = 0.26) of enthusing soybean breeders to use it as a selection proxy for elevating TE or yield beta, despite the successful use of Δ13C for TE in wheat (Condon et al, 2002; Richards et al, 2002).
Figure 3. Carbon isotope composition (δ13C) values for leaf tissue collected (mid-July) from stem tips of the Minsoy and Noir 1 cultivars, their (267) RIL progeny of the determinate (D) or indeterminate (I) stem growth habit type, and 19 unrelated check cultivars. Leaf δ13C values plotted against the seed yield of those genotypes. The more positive the δ13C value, the greater the TE (Condon et al, 2002). Left Panel: 0% ET (non-irrigated) data. Right Panel: 100% ET (fully irrigated) data (check data lost). See Specht et al (2001) for other details.
To improve soybean WUEy, breeders need to identify genotypes whose yield-to-water response (i.e., yield beta) is steeper. However, selection for a higher yield beta is difficult, not only because of its quantitatively inherited nature, but also (and more importantly) because of the low probability of encountering drought seasons routinely enough in the specific breeder-targeted regions to effectively practice such selection. The identification of marker-flanked favourable alleles at QTLs for yield beta QTLs would bypass those difficulties, since marker-assisted selection (MAS) could be used to rapidly introgress those favourable alleles into elite cultivars. Still, one must take care to know just exactly what is being introgressed. In the Specht et al (2001) report, QTLs detected as controlling significant amounts of variation in yield beta and yield mean were in fact pleiotropic expressions of existing QTLs (or genes) known to control days to maturation and stem habit. Indeed, a QTL detected on soybean linkage group (LG) L controlled 25% of the phenotypic variation in leaf δ13C and its additive effect on leaf δ13C was a huge 0.405‰, but it mapped to the genomic position of a stem determinancy locus. As noted in a prior section, this leaf δ13C QTL was effectively an artefact arising from the fact that appropriate leaf tissue for the δ13C analyses could not be effectively sampled from the determinate RILs. Although a leaf δ13C QTL was detected on LG C2, and it had a lesser additive allelic effect of 0.204‰, this QTL accounted for only 7% of the leaf δ13C phenotypic variation. In QTL mapping experiments, selection bias (i.e., the overestimation of the effects attributable to moderate or minor QTLs) is common for QTLs controlling less than 10% of the phenotypic variation making it necessary for the researcher to confirm the QTL effect in an independently derived population. Moreover, the true additive effect really needs to be large enough to warrant the time, effort, and cost of manipulating the favourable alleles at the given QTL in a MAS program. So far, we have not yet observed a yield beta or leaf δ13C QTL in the currently genotyped RIL populations that would be worthy of such attention.
Our published (Specht et al, 1986, 2001) and unpublished results have confirmed a close association of soybean seed yield with transpiration. Moreover, the close association between genotypic yield beta (a field-based proxy for WUEy) and genotypic mean yield provides credence to the postulation that soybean WUEy (and thus drought tolerance per se) is improved coincidently by simply selecting for higher genotypic mean yield. It was also clear in each mapping population that the highest yielding entries when water was abundant (100% ET) were also the highest yielding entries when water was scarce (0% ET). This was consistent with the observation that most of the QTLs detected for yield beta WUEy in the now marker-genotyped RIL populations were positionally coincident (in terms of genome position) with the QTLs that we detected for yield per se. Whether this coincidence turns out to be linkage or pleiotropy is probably irrelevant, given that the highly positive correlation between two desirable traits is probably not going to limit the success of breeders trying to improve both traits when developing cultivars targeted for use in the Midwestern USA. Future attempts to wring out a large increase in soybean WUEb or WUEy would have to focus on upping the ΔBM or ΔSY numerators of these WUE parameters. Indeed, the evolution of some plant species from a C3 to a C4 form of photosynthesis provides an ecological precedent for WUE improvement efforts (Fig. 1, left).
Condon, AG, Richards, RA, Rebetzke, GJ and Farquhar, GD (2002) Improving intrinsic water-use efficiency and crop yield. Crop Science 42, 122-131.
Passioura, JB (1977) Grain yield, harvest index and water use of wheat. J. Aust. Inst. Agric. Sci. 43,117-120.
Purcell, LC and Specht, JE (2004) Chapter 12. Physiological traits for ameliorating drought stress. pp. 569-620. In Boerma, HR and Specht JE (eds.) Soybean: Improvement, production and uses. 3rd ed. Agron. Mongr. 16. ASA and CSSA, Madison, WI, USA.
Richards, RA, Rebetzke, GJ, Condon, AG and van Herwaarden, AF (2002) Breeding opportunities for increasing the efficiency of water use and crop yield in temperate cereals. Crop Science 42, 111-121.
Rosielle, AA and Hamblin, J (1981) Theoretical aspects of selection in for yield in stress and non-stress environments. Crop Science 21, 943-945.
Sinclair, TR (1994) Limits to crop yield? pp. 509-532. In Boote KJ et al. (eds.) Physiology and determination of yield. ASA, Madison, WI, USA.
Specht, JE, Chase, K, Macrander, M, Graef, GL, Chung, J, Markwell, JP, German, M, Orf, JH and Lark, KG (2001) Soybean response to water: A QTL analysis of drought tolerance. Crop Science 41, 493-509.
Specht, JE, Williams, JH and Weidenbenner, CJ (1986) Differential responses of soybean genotypes subjected to a seasonal soil water gradient. Crop Science 26, 922-934.