Organic agriculture, including the growing of organic wheat, is one of the fastest-growing sectors in the food industries of Europe and the United States (Greene & Dobbs 2001). Strict rules apply to the growing of organic crops including a ban on the use of artificial nitrogenous fertilizers; nitrogen inputs must come from animal manures and slurries, green manures, crop residues, legumes and residual soil organic matter. Advocates for organic farming point to savings in fossil fuels, reduced environmental pollution, greater biodiversity and more nutritious and better tasting food as positive reasons for farming organically. Opponents counter with reduced yields and quality, and question the claimed environmental and health benefits (Trewavas 2004). Despite the increasing interest in organic wheat, nothing is known at the molecular level about differences between grain grown solely with organic fertilizer compared to grain grown under conventional inorganic conditions. Grain nitrogen levels depend on two factors: continued nitrate uptake by roots and the remobilization of nitrogenous compounds from leaves and stems during grain filling (Masclaux et al. 2001). The proportion of nitrogen from these two sources is greatly influenced by environmental conditions and plant genotype (Purcino et al. 1998). In addition to the readily available ammonium and nitrate ions, organic fertilizers can contain a wide range of organic-nitrogen compounds such as amino acids, amino sugars and nitrogen heterocyclic compounds which may influence the soil microflora (Mäder et al. 2002; Shannon et al. 2002; Pernes-Debuyser & Tessier 2004). We examined the gene expression profiles of developing wheat grain grown under different nitrogen fertilizer regimes on Broadbalk field at Rothamsted in South East England (figure 1) to determine whether these profiles might be diagnostic for these regimes and prognostic for yield. On Broadbalk, winter wheat is grown under well defined nutrient supply conditions including a range of inorganic nitrogen treatments and two treatments based on farmyard cattle manure (FYM; Poulton 1996; Richards & Webster 1999; http://www.rothamsted.bbsrc.ac.uk/resources/ClassicalExperiments.html).

Figure 1 Historical data describing the Broadbalk winter wheat experiment which was established at Rothamsted, England in 1843 by Sir John Bennet Lawes and Sir Henry Gilbert to investigate the influence of inorganic fertilizers and organic manures on crop yield. (more ...)

All cDNA clones used for the microarray experiments were from the University of Bristol wheat unigene set (Wilson et al. 2004). All sequences within the unigene set have been placed in GenBank. Inserts were prepared and arrayed onto CodeLink microarray slides (Amersham BioSciences/GE Healthcare, Chalfont, UK) using standard procedures in conjunction with a BioRobotics MicroGrid II Microarrayer (Genomic Solutions Ltd, Huntingdon, UK). Fluorescently labelled cDNA probes were prepared from the various total RNA samples using an amino allyl-labelled first-strand reverse transcriptase reaction (Wilson et al. 2004).

Figure 2 Gene expression in response to different amounts and forms of nitrogen as determined by microarray analysis. Grain samples were taken from duplicate plots of continuous winter wheat on seven nitrogen treatments: N0, N1, N1:2:1, (more ...)

Table 1 Organic carbon and total nitrogen in Broadbalk and Woburn soils.

The 12 diagnostic genes in figure 2b had expression profiles up-regulated in the presence of FYM. For the purpose of this analysis only those genes that showed greater than fivefold up-regulation in the FYM compared to the N4 treatment are presented. Of the 12 genes in this category, nine could not be assigned a function. Of the three sequences for which a function could be assigned no overall link could be identified. One gene (Al818695; phosphoglycerate mutase) inter-converts 3-phosphoglycerate and 2-phosphoglycerate and is one of the many steps in glycolysis (Grana et al. 1992), while a second (AL821423) shows sequence similarities to a stress induced farnesylated protein isolated from barley (Barth et al. 2004). A third sequence, AL821213, shows similarities to a progesterone 5-beta-reductase (BAC20032), which catalyses the reduction of progesterone to beta-pregnane-3, 20-dione.

The cluster analysis enabled the classification of groups of genes with similar expression profiles across the various treatments. In figure 2c we present three representative profiles. The first group (figure 2c(i) and Electronic Appendix supporting table 1) contained 474 genes with highest expression in the FYM treatment and included the diagnostic genes. Curiously, for this group the FYM+N2 treatments often resulted in a decrease in expression. One hundred and fifty-five of the genes within this group had insufficient sequence similarity with other sequences to assign a putative function. However, the remaining 319 genes encoded a wide variety of proteins including proteins known to be involved in amino acid, lignin and flavonoid biosynthesis and glycolysis. The second group of 190 genes (figure 2c(ii) and Electronic Appendix supporting table 1) showed highest expression in N0 and N1 and lowest expression in the FYM treatment. A third group consisting of twenty genes showed low expression in the low nitrogen treatments (N0 and N1) and uniformly high expression at higher nitrogen treatments irrespective of the form of nitrogen (figure 2c(iii) and Electronic Appendix supporting table 1).

The expression profiles of the 12 diagnostic genes were further verified by quantitative-RT-PCR (Q-PCR) using RNA derived from material grown on Broadbalk during 2002, 2003 or 2004 (figure 3). The expression of the 12 diagnostic genes during 2002 and 2003 is shown (figure 3a). Of these, six genes showed greater than 10 fold change in the level of expression in the FYM versus N4 treatments on Broadbalk in 2002. The expression of these six diagnostic genes was further examined on four separate farms during 2004. In all four locations these six genes showed higher expression in response to a variety of organic inputs compared to controls that had received inorganic nitrogen (figure 3b).

Figure 3 Expression profiles of the ‘diagnostic genes’ determined by Q-PCR. (a) Analysis of the 12 sequences with the highest ratio of expression following FYM treatment when compared to N4 treatment (data from figure 2). (b) Q-PCR analysis of (more ...)

To examine if these six genes were more highly expressed in FYM versus N4 treatments throughout grain development, we determined their expression profile on Broadbalk in the five-week period following anthesis (figure 3c). All six genes showed a continuously elevated level of expression in the FYM treatment compared to the N4 treatment with the maximum level of differential expression observed at 14 and 21dpa. This time period is significant as it represents the early/middle stage of grain filling and reserve synthesis. Finally, we examined the specificity of expression of these six genes across the Broadbalk treatments (figure 3d). Our Q-PCR results showed that the six diagnostic genes were highly expressed in the FYM treatment, moderately expressed in the FYM+N2 treatment, and had lower expression in all other treatments, confirming the microarray data (figure 2). Taken together, the data indicates that these six genes were expressed most highly when the crop had received FYM or other organic inputs.

Figure 4 Q-PCR results for various storage protein sequences. Q-PCR results for the storage-protein sequences including six gliadins (a), one high molecular weight and one low molecular weight glutenin (b) and two globulin related sequences (c). In each case the (more ...)

The most important factor for farmers is grain yield. The relatively low nitrogen availability in organic farming systems limits yields and grain nitrogen and it is therefore difficult to produce high-protein grain under organic conditions (Gooding et al. 1993). To examine this further we compared the various expression profiles seen in figure 2 with the final grain yields and the percentage of nitrogen in grain dry matter for the corresponding treatments on Broadbalk (figure 4d). In the N0 and N1 treatments, whose gene expression profiles clustered together, total nitrogen uptake, yield and grain nitrogen were all relatively low. Increasing the nitrogen rate (N4, N1:2:1 and N6) resulted in similar gene expression profiles and increased nitrogen uptake, yield and grain nitrogen. The highest levels of grain nitrogen followed the N6 treatment and this treatment showed the greatest degree of divergence at the expression level from the other inorganic nitrogen treatments. The global gene expression clustering suggested a close relationship between the FYM and FYM+N2 treatments, discrete from the inorganic treatments alone or the N0 treatment, which could be regarded as an organic-N plot since it had not received additional inputs of organic or inorganic nitrogen.

Our results demonstrate that the amount and form of nitrogen available to wheat influences gene expression and grain yield. It is also important to consider what additional factors in the FYM treatments other than the form of nitrogen fertilizer may have contributed to the expression patterns observed. For instance, the soil of the FYM plots had higher organic matter content (table 1), which will have lead to improved water retention properties (Mäder et al. 2002; Pernes-Debuyser & Tessier 2004). However, in the FYM treatments described here inorganic nitrogen application (FYM+N2) down-regulated many of the genes highly expressed in the FYM treatment alone. This suggests that the differences observed between the organic and inorganic nitrogen treatments are not simply due to the increased organic matter content and associated changes in properties such as water retention or soil microbial populations. It is interesting to note that even though many of the genes, including some involved in nitrogen metabolism, were down-regulated in FYM+N2, yields were increased substantially compared to the FYM treatments and final grain nitrogen increased slightly (figure 4d). Our observations suggest that although many genes are expressed at higher levels in fertile organic conditions, their synergistic effect does not result in higher yields or increased grain nitrogen content. It is possible that in both the organic-N treatments and at the higher inorganic nitrogen inputs (N4, N6), the nitrogen in the soil during seedling establishment and the early tillering stages is sufficient to create a high yield potential for the crop. However, in the treatments with solely organic-N inputs (FYM), but not in the inorganic-N treatments or supplemented organic-N treatments (FYM+N2), nitrogen subsequently becomes limiting and in this situation many genes involved in nitrogen uptake, nitrogen metabolism and storage are up-regulated in an attempt to fulfil the initial yield potential. Such a scenario would be consistent with our observation that the addition of inorganic nitrogen to the FYM treated plots resulted in a slightly higher yield and a decrease in the expression level of many of the genes involved in nitrogen metabolism.

Numerous reports have discussed the environmental implications and the benefits of consuming organically grown food. Clearly the optimal use of nitrogenous fertilizers will reduce environmental pollution, however there is little evidence that nitrate leaching differs between organic and conventional production systems when following best practice (Stopes et al. 2002; Trewavas 2002). Our results suggest that at the level of gene expression, wheat grain from plants supplied with either organic or inorganic nitrogen is reproducibly different. However, our results also confirm that the limited availability of nitrogen in organic systems tends to limit yields and total grain nitrogen. In this study we have identified several genes whose expression appears to be indicative of the ‘organic status’ of the wheat grain in relation to the type of nitrogenous fertilizer. Such genes, if shown to be causally related to utilisation of organically originating nitrogen, might represent useful starting points for the targeted breeding of varieties that perform better under organic growing conditions. However, we believe that the large number of genes shown to be up-regulated in response to organic nitrogen fertilizer and induced by increased nitrogen supply indicate the complexity of the physiological response and require further experiments to dissect the effects that both the amount and form of nitrogenous fertilizer have on the developing grain.

The sequences reported in this paper have been deposited in the GenBank database (accession nos. AL808216–AL831324). The gene expression data have been deposited in the ArrayExpress database in accordance with Microarray Gene Expression Data Society recommendations (accession numbers A-MEXP-151 and E-MEXP-261; http://www.ebi.ac.uk/arrayexpress/Submissions/ index.html).

We thank W. Zhou, S. Parmar, S. Shepherd, C. Shepherd, J. Jones, C. Knight, K. Yu, S. White and M. Bond for technical assistance. We thank Howard Roberts for providing access to the Harpenden spring wheat crop during 2004. Rothamsted Research receives grant-aided support from the Biotechnology and Biological Science Research Council (BBSRC) of the UK. This project was supported by a grant from the BBSRC to M.J.H. and K.J.E.