Effects of photoperiod sensitivity genes Ppd-B1 and Ppd-D1 on spike fertility and related traits in bread wheat

Unidad Integrada Balcarce (Estaci on Experimental Agropecuaria Balcarce, Instituto Nacional de Tecnolog ıa Agropecuaria and Facultad de Ciencias Agrarias, Universidad Nacional de Mar del Plata), Balcarce, Buenos Aires, Argentina Monsanto’s Beachell-Borlaug International Scholars Program, Soil & Crop Sciences Department, AgriLife Research – Texas A&M University System, College Station, Texas, USA Consejo Nacional de Investigaciones Cient ıficas y T ecnicas, Balcarce, Buenos Aires, Argentina


| INTRODUCTION
Bread wheat (Triticum aestivum L.) is the most widely grown crop at a global scale and a major source of carbohydrates and proteins in human nutrition (Mahjourimajd et al., 2016), and current and future prospects indicate that its demand will continue to grow. This implies that breeding efforts need to focus on increasing yield, because the opportunities for adding new arable land to the cultivated area are limited. Thus, grain yield improvement has been, and it continues to be, one of the central goals in wheat breeding programmes worldwide (Dixon et al., 2009;Foulkes & Reynolds, 2014;Mirabella et al., 2016;Rajaram, 2005). During the last 30-40 years, genetic and agronomic progress in wheat grain yield has been largely attributed to an increase in grain number (GN)/m 2 (Fischer, 2007;Foulkes et al., 2011;Reynolds et al., 2009;Shearman et al., 2005;Slafer et al., 1990). However, despite GN/m 2 is the trait that usually best explains yield and several authors have shown that it can be further increased (Abbate et al., 1998;Acreche et al., 2009;Foulkes et al., 2011;Parry et al., 2011;Reynolds et al., 2009), it is difficult to accurately determine it at early stages of a breeding programme, in which not enough seed is available to measure variables as per unit area. Therefore, it is necessary to use alternative, related traits as selection criteria.
According to Fischer's assimilate-based approach (Fischer, 1983), under optimal growing conditions (i.e., without water or nutrient limitations and in the absence of pests and diseases), GN in wheat can be considered as the product of (i) the duration of the spike growth period (SGP), (ii) the crop growth rate during the SGP, (iii) the dry weight partitioning to spikes during the SGP and (iv) the number of grains per unit of spike chaff dry weight (SCDW), that is a spike fertility (SF) index, also termed "fruiting efficiency" . Abbate et al. (1998), working on Argentinean high-yielding cultivars, observed that GN/m 2 was mainly related to SF. Since then, many authors have shown the existence of conspicuous variation for this trait among cultivars of diverse origin (Abbate et al., 2013;Acreche, Briceño-F elix, Mart ın S anchez, & Slafer, 2008;Fischer, 2007;Gonz alez et al., 2011;Gonz alez-Navarro et al., 2015;L azaro & Abbate, 2012;Martino et al., 2015;Mirabella et al., 2016;Shearman et al., 2005). Interestingly, recent work has evidenced that SF is a moderately heritable trait, with low genotype 9 environment interaction (Gonz alez-Navarro et al., 2015;Martino et al., 2015;Mirabella et al., 2016); moreover, a simple and fast methodology (Abbate et al., 2013) has been developed for high-throughput assessment of SF at maturity (SFm) in early breeding material using the dry weight of the chaff at maturity (spike weight after removing the grains) to provide an estimate of the spike dry weight at anthesis. Thus, it has been suggested that spike fertility can be used as a selection criterion in breeding programmes to develop high-yielding cultivars (Fischer, 2007(Fischer, , 2011Foulkes et al., 2011;Gonz alez et al., 2011;L azaro & Abbate, 2012;Abbate et al., 2013;Gonz alez et al., 2014;Slafer et al., 2015;among others).
Despite SF appears to be under a relatively simple genetic control as compared to yield (Martino et al., 2015;Mirabella et al., 2016), virtually no candidate genes have been proposed specifically for this trait to date . In this regard, several studies (Fischer, 2016;Gonz alez et al., 2003Gonz alez et al., , 2005Whitechurch & Slafer, 2002) have shown that modifications in day length during the spike growth phase, or changes in photoperiod sensitivity, affected fertile floret number at anthesis. In wheat, photoperiod response is mainly regulated by Ppd-A1, Ppd-B1 and Ppd-D1 genes (Law et al., 1978;Scarth & Law, 1983;Welsh et al., 1973); "sensitive" alleles confer increase in thermal time to anthesis when the photoperiod is reduced, whereas "insensitive" ones make the plant insensitive to day length. Slafer et al. (1996), Slafer et al. (2001) and Gonz alez et al. (2005), among others, have suggested increasing photoperiod sensitivity as a way to increase GNSm through a longer duration of the spike growth period before anthesis. It would be interesting to establish the degree to which photoperiod sensitivity genes affect GNSm, SCDWm and SFm. Thus, the aim of this study was to analyse the effects of

| Crop management
Three field experiments (termed Expt 1, Expt 2 and Expt 3) were Heading date information collected for each cultivar at each sowing date in Expt 1 was used for grouping cultivars into three groups of similar heading date (see Supporting information). In Expt 2 and Expt 3, each cultivar was sown at one of each three sowing dates, in order for all cultivars to have similar heading date (around the first week of November). Expt 2 and Expt 3 were conducted under a randomized complete block design with two replicates. The experimental unit consisted of a 5.5-m-long seven-row plot, with a 0.2 m inter-row distance.
GNSm, CDWm and SFm were determined as in Expt 1.

| Statistical analysis
Phenotypic information obtained from Expt 1, Expt 2 and Expt 3 was analysed along with the genetic information (i.e., the allelic constitution for Ppd-B1 and Ppd-D1 genes) obtained from previously published work (Vanzetti et al., 2013) using mixed models. The variance of data from Expt 1 was first analysed separately to assess sowing date effects, according to the following statistical model (model 1): where l is the general mean of the trait (SFm, SCDWm or GNSm), a i is the effect of the blocks, q j is the effect of the sowing date, c l is the effect of the Ppd-B1 gene, d m is the effect of the Ppd-D1 gene, qc jl is the interaction effect between sowing date and Ppd-B1, qd jm is the interaction effect between the sowing date and Ppd-D1, cd lm is the interaction effect between Ppd-B1 and Ppd-D1, qcd jlm is the interaction effect between sowing date, Ppd-B1 and Ppd-D1, and e ijklm is the error (variance not explained by the model).
The SFm, SCDWm and GNSm values of each cultivar at its optimal sowing date in Expt 1 were combined with data from Expt 2 and Expt 3 and analysed together. The statistical model used in the analysis of variance of combined data from Expt 1, Expt 2 and Expt 3 was as follows (model 2): where a q is the effect of year, q k(q) is the block effect within each year (random effect), ac ql is the interaction effect between year and Ppd-B1, ad qm is the interaction effect between year and Ppd-D1, acd qlm is the interaction effect between year, Ppd-B1 and Ppd-D1, e qiklm is the error (variance not explained by the model). The remaining terms were already defined in the previous equation.
Data were checked for ANOVA assumptions (normal distribution, homoscedasticity and independence of errors) before analysis. A 0.05 significance level was used for all tests. Before ANOVA, a regression analysis between SFm and year of cultivar release was performed to rule out confounding effects (e.g., a selection bias). As a result, no significant effect was found (R 2 =.0342). The relative frequency of the different allelic combinations in "older" vs. "recent" cultivars (i.e., released before and after 2000, respectively) was also determined and found to be fairly similar (Table S2). In addition, no effect of the population structure, as determined by genome-wide molecular marker analysis (Vanzetti et al., 2013), was found on the variables analysed in this study. Therefore, it was not included as a factor in the model.
The proportion of variance explained by the genes was estimated with the residual variances of models that included the cultivar and gene effects vs. those which did not include such effects.
Repeatability (Piepho & M€ ohring, 2007) for each trait was estimated as broad-sense heritability using the standard least squares method, as proposed by Holland et al. (2003).

| RESULTS
The mean, maximum, minimum and genetic coefficient of variation for SFm, SCDWm and GNSm are shown in Table 1  SCDWm and GNSm values of each cultivar at its optimal heading date in Expt 1 (i.e., around the first week of November) were included in further combined analyses together with data from Expt 2 and Expt 3.
The combined statistical analysis of Expt 1, Expt 2 and Expt 3 showed that the insensitive allele at both Ppd-B1 and Ppd-D1 genes was associated with greater SFm values, with no gene 9 gene interaction (Tables 3 and 4).
No statistically significant difference was detected between genes on SFm (Table 4). Cultivars with insensitive alleles at both genes showed the highest SFm mean (9.2% greater than SFm mean of cultivars with sensitive alleles at both genes). The difference between the insensitive and sensitive allele was 4.7 and 4.3% for Ppd-B1 and Ppd-D1, respectively (Table 4).
When analysing SCDWm, a significant effect of the Ppd-B1 gene was observed, as the insensitive allele decreased SCDWm by 9.2% as compared with the sensitive allele (Table 4) thereby increasing SFm. In the case of Ppd-D1, the allelic constitution did not affect SCDWm (p > .05).
A significant Ppd-B1 9 Ppd-D1 interaction was observed on GNSm in Expt 1 (  Tables 2   and 3). In this case, the insensitive allele at this gene decreased GNSm by 3.4% as compared with the sensitive one (Table 4).
The proportion of variance explained by the genes was 3.5% for SFm (both genes), 4.2% for SCDWm (Ppd-B1 gene) and 4.7% for GNSm (both genes). The broad-sense heritability for SFm, SCDWm and GNSm was 0.56, 0.47 and 0.39, respectively, indicating that the genotype effect was mildly repeatable even though there was a highly significant year effect in all variables (Table 3).

| DISCUSSION
The use of SF as a selection criterion in wheat breeding programmes aimed at increasing grain yield has been profusely discussed in the literature (Abbate et al., 1998(Abbate et al., , 2013Fischer, 2007Fischer, , 2011Foulkes et al., 2011;Gonz alez et al., 2011;L azaro & Abbate, 2012). Martino et al. (2015) and Mirabella et al. (2016) found that this trait is moderately heritable with low genetic 9 environmental interaction and that it is  . Several authors did suggest the use of Ppd-B1 (Worland et al., 1998) and Ppd-D1 (B€ orner et al., 1993;Worland, 1996;Worland et al., 1988;Worland et al., 1998) genes to directly increase grain number per unit area and yield in European wheat cultivars.
The present study shows that the insensitive alleles of the Ppd-B1 and Ppd-D1 genes are associated with an increase in SFm, independently of Ppd genes' well-known and widely reported effect on determining heading date in wheat and other cereals. Several authors (Gonz alez et al., 2005;Worland, 1996;Worland et al., 1998) have suggested that Ppd-D1 shows stronger effects than Ppd-B1 in traits such as spikelet number, grains per spikelet and heading date. Contrarily, our results show no differences between genes when SFm is analysed.
The absence of interaction between Ppd-B1 and Ppd-D1 on SFm indicates that the effects of these genes are additive and that the insensitive alleles can be combined to achieve higher SFm values. In addition, the absence of interaction between the genes and the environment and their repeatability add evidence in support for their manipulation as a means of increasing SFm in the context of a breeding programme.
Moreover, these results could partially explain the differences in SF found by Fischer (2016) between two closely related cultivars, 'Yecora' and 'Cajeme', which differ in their Ppd-D1 constitution.
Surprisingly, when GNSm was analysed, a differential effect of Ppd-D1 was observed as compared with that of Ppd-B1 ( Figure S1).
Ppd-D1 showed a highly significant effect on GNSm, as this variable In the case of Ppd-B1, and under certain environmental conditions, Worland et al. (1998)  in Figure S1, higher SFm values would be expected when combining insensitive alleles at both genes. to be validated at the crop level (i.e., through the assessment of GN/ m 2 ). Also, eventual trade-offs of this selection strategy should be investigated, such as a possible negative association between SF (or grain number per unit area) and grain weight (Martino et al., 2015) or other yield-related traits.
Spike chaff dry weight only showed association with Ppd-B1, as lower average SCDWm values were related to the presence of the insensitive allele. This is the first report, which we are aware of, that describes a gene related to this trait.
The relatively low percentage of phenotypic variance explained by the genes (3.5-4.7% depending on the trait and gene) suggests that, although the effects of both Ppd-B1 and Ppd-D1 on the assessed variables were repeatable even under different environmental conditions (as reflected by a highly significant year effect on all variables), many additional genes are possibly involved in the control of SF, GNS and SCDW. Further work on the identification of such genes would help design breeding strategies for increasing SF through the concurrent selection of the best allelic combinations at several loci.

| CONCLUSION
The results from this work show that the presence of photoperiodinsensitive alleles at both Ppd-B1 and Ppd-D1 genes is associated with higher SF, showing similar and additive effects. The allele effects, however, differed in their origin: insensitivity at Ppd-B1 reduced SCDW more than GNS, while insensitivity at Ppd-D1 increased GNS and left SCDW unchanged.
Although more research needs to be carried out in order to ascertain the physiological pathway by which these genes affect SF, our study represents a first approximation in order to elucidate the molecular and genetic basis underlying SF and physiological traits related to SF.