The development of improved rice varieties with high tolerance of salt stress has been a major and long-standing goal of rice breeding efforts. Salinity is a significant constraint to rice productivity in many inland and coastal rice-growing areas and, furthermore, is contributing to the loss of arable lands in many countries due to salt accumulation as a result of excessive use of irrigation water with poor or improper drainage, a fact that is likely to be aggravated by sea level rise in coastal areas caused by climate change [1–3]. Therefore, efforts to improve the salinity tolerance of rice and many other crops are intensifying. Significant bodies of work have been accomplished on the characterisation of physiological responses affected by salt stress. These studies highlighted the complexity of the mechanisms involved in rice in which tolerance varies with the stage of development, with the crop being relatively more tolerant during germination and active tillering as well as during late grain filling and maturity, but sensitive during the early vegetative and reproductive stages [4, 5], and with weak association between the degree of tolerance at the two sensitive stages [6].

Numerous physiological studies on the mechanisms of tolerance during the vegetative stage have been published [5, 7–9], most of which showed an inverse relation between shoot Na⁺ content and/or Na⁺/K⁺ ratio and plant survival, injury scores and grain yield [10, 11]. Other traits suggested to be associated with salt tolerance in various studies are compartmentation of Na⁺ in older leaves and leaf sheaths and in the vacuoles, maintenance of mineral nutrient homeostasis, especially K⁺ and Ca²⁺, high selectivity for K⁺ and/or Ca²⁺ uptake over that of Na⁺, limiting effects of reactive oxygen species (ROS scavenging), accumulation of compatible solutes to offset osmotic effects (osmotic adjustment), maintenance of leaf area index and maintenance of tiller number [5, 9–13]. The importance of the apoplastic bypass flow in delivering Na⁺ to the xylem, thus reducing leaf Na⁺ concentration and improving tolerance, has also been noted [14–17]. During the reproductive stage, tolerant genotypes strongly exclude salt from flag leaves and developing panicles [6, 18]. The complexity of tolerance highlighted in these studies suggests the need for combining tolerance mechanisms at each stage as well as at the two most sensitive stages to develop varieties that are widely adapted to salt-affected areas.

Efforts also focused on the mapping of QTL loci controlling these various component traits, and a few major loci and numerous minor loci controlling various aspects related to salinity tolerance were subsequently identified. The best known and seemingly most robust QTL is Saltol/SKC1 on the short arm of chromosome 1 [19, 20]. QTLs have been identified in this region in a number of populations derived from several donors ([21–23], A. Ismail unpublished results), and the gene has been identified to a high degree of confidence as OsHKT1;5[24] (see [25] for nomenclatural clarification). A very recent association mapping effort using varieties from the japonica cultivar-group [21] has also identified the Saltol genomic region as controlling important aspects of salinity tolerance, as well as validating many other QTLs. In wheat, two members of the HKT gene family (including the wheat HKT1;5 orthologue) have also been shown to co-localise with major QTLs [26–28], and the HvCBL4 gene, a homologue of the Arabidopsis SOS3 known to confer salt tolerance, mapped to a genomic region similar to that of a barley salt tolerance QTL [29].

In addition to the Saltol locus, many other QTLs have been identified in rice (e.g. see [8] for a recent review). Several of these appear to be common among multiple mapping populations, though they seem to be derived from the same or genetically similar donors. Examples include the long arms of chromosomes 1, 3 and 6 [20, 22, 30–35]. Although numerous studies have identified hundreds of genes involved in salt stress responses, many of which lead to improved tolerance when over-/underexpressed, and some co-localise with QTL regions, studies on the cloning of other QTLs in cereals are yet to be published. This might be in part due to these additional QTLs typically controlling much smaller portions of the total variance than does Saltol, and thus being more difficult to work with.

Despite the long history of salinity tolerance research and breeding efforts, very few large-scale screening efforts have been undertaken. A number of authors published studies involving small numbers of accessions [10, 36–41]. Only four studies appear to have examined more than a few dozen accessions [9, 42–44], and even these have focused on breeding lines and improved cultivars, which tend to stem from a small donor pool typically involving Pokkali and/or Nona Bokra as donors. In screens of 21 and 38 genotypes of wheat (Triticum aestivum L.) [45], the authors concluded that leaf Na⁺ concentration showed little correlation with performance, and that Na⁺ exclusion and tissue tolerance were equally important, and segregating independently. However, even in this case, the germplasm examined consisted of breeding lines and improved cultivars stemming from a very restricted geographic area.

Thus, there seems to be a lack of large-scale screening efforts specifically aimed at identifying significant new donor germplasm, particularly with regard to traditional varieties/landraces. It would thus appear difficult to generalise about the level of salinity tolerance displayed or the mechanisms possessed by these species as a whole, factors that are important when considering breeding approaches. In addition, a common feature of nearly all these studies is screening under relatively mild salt stress, typically of electrical conductivity (EC) of 6 dS m^-1 (approx. 50 mM NaCl) to 12 dS m^-1 (approx. 100 mM NaCl), and there seems to be a lack of screening efforts specifically aimed at identifying highly tolerant germplasm that might therefore contain additional major QTLs as effective as or more effective than Saltol. The objectives of this work are to (1) screen known and novel germplasm under high salinity to identify new highly tolerant lines, with particular emphasis on traditional landraces that may have novel alleles/mechanisms of tolerance unrelated to Pokkali and Nona Bokra; (2) characterize the tolerant lines, specifically with reference to Na⁺ and K⁺ uptake, to classify lines based on tolerance mechanisms; (3) supplement phenotyping results with allele mining of the OsHKT1;5 gene, and relate alleles to function; and (4) integrate this information with respect to Saltol/OsHKT1;5 activity, and identify novel donors for use in breeding.

Screening of landraces from coastal and saline inland regions identified a number of accessions showing modest to significant salinity tolerance that are distinct from traditionally used donors such as Pokkali and Nona Bokra. These accessions are from diverse backgrounds, including nearly all cultivar-groups of O. sativa and also O. glaberrima. To our knowledge this is the first report of significant salinity tolerance from O. glaberrima, and also from the aromatic cultivar-group of O. sativa. Salinity tolerance in rice thus appears to be widespread both geographically and phylogenetically, or, put in another way, tolerance is not well associated with either geographic or cultivar-group origin. Together with the fact that these are landraces and not expected to show relationships apart from gene flow inherent in the species’ history, this suggests that many of these have probably gained tolerance independently and that multiple mechanisms may thus exist.

On the other hand, tolerance is quite well correlated with leaf Na⁺ concentration across almost all accessions of both O. sativa and O. glaberrima. Despite the diverse origins and relationships of the accessions, tolerance could in almost all cases be explained largely with reference to lower Na⁺ concentrations in the photosynthetically active leaves. This further suggests that processes controlling this are the predominant mechanisms of tolerance in O. sativa and O. glaberrima, and that tissue tolerance mechanisms (vacuolar sequestration, ROS scavenging, osmotic adjustment, certain hormonal responses) play secondary roles. Na⁺ exclusion from roots, sequestration of Na⁺ in roots, stems and basal portions of the leaf (sheath), partitioning of Na⁺ from leaf to leaf and dilution of Na⁺ content in a large biomass are mechanisms proposed to influence leaf Na⁺ concentration. Na⁺ sequestration is one such mechanism known to operate in a number of species from both the dicots and monocots (e.g., wild and cultivated barleys: [51, 52]; durum and bread wheat [53, 54]). However, the relationship is not universal. For example, it has not been observed in studies on maize and sorghum ([55–57], although see [58]) and it is an important [54] but not a universal determinant in wheat [45]. In some cases this may be due to a lack of genotypic variability [57]. The fact that total leaf and shoot Na⁺ content (not just concentration) also shows a very strong correlation indicates that Na⁺ sequestration from the leaf blade is a very important contributor to maintaining low tissue Na⁺ concentrations.

Dilution of Na⁺ concentrations through a large biomass is also a well-accepted mechanism for maintaining low tissue Na⁺ concentrations, and Yeo et al. [9] concluded that Na⁺ accumulation (content) showed only a poor correlation with performance in rice, being significantly confounded with plant height; tall varieties showed better tolerance and lower Na⁺ concentrations due simply to dilution of Na⁺ in the larger volume of tissue produced. The data presented provide an apparent contradiction to the latter, but this may be due to the screening conditions: the latter study conducted screening at relatively low salinity (60 mM) for short periods (10 days). Salt concentration of 150 mM NaCl was used for the physiological characterisation presented here, which is higher than that used in most previous screening studies; the higher salt concentration causes a much greater influx of Na⁺, which may overwhelm other mechanisms, notably the effect of plant vigour [9]. Under these conditions, growth effectively ceases in all varieties after the application of the salinity treatment. A few of the most highly tolerant varieties will resume growth after some time, but at a greatly reduced rate; over the lifetime of an experiment, even the most highly tolerant variety will produce only about half a new leaf. This growth arrest actually appears to be an adaptive feature, and lines that try to keep growing show a different type of growth arrest – the youngest leaves expand, but soon yellow and die, presumably due to excessive Na⁺ accumulation. Thus, screening at higher salinity levels may help to reduce the contribution of biomass to tolerance, and so “simplify” the response in this respect.

It is interesting to note that the correlation of SES scores with plant vigour is highest for leaf sheath biomass (r² = 0.54), followed by total harvested tissue and root biomass (r² = 0.47 and 0.46, respectively), but much lower for leaf biomass (r² = 0.14, 0.20 and 0.36 for leaf 4, 5 and 6, respectively). The leaf sheaths and roots are the main tissues known to act as reservoirs for Na⁺ sequestration, such as that mediated by OsHKT1;5[24]. Thus, the contribution of biomass may be partly to dilute the Na⁺ taken up, but also to provide a reservoir for sequestration in non-photosynthetic portions.

It should be noted that the most highly active allele, found in traditional donors such as Pokkali, Nona Bokra and others, has almost certainly originated within the aromatic cultivar-group, despite these being indica types. Indeed, although the sample size and fold changes are small, it seems that the most highly tolerant lines are those from the indica cultivar-group that also possess this Aromatic allele; these are often more tolerant than lines from the aromatic cultivar-group. It may be that some feature of the indica cultivar-group genetic background is in some way synergistic with the action of the Aromatic allele. Alternatively, it has been noted that many aromatic lines (according to the functional definition) have lower salt tolerance due to their inability to produce gamma aminobutyric acid – the same mutation that confers their aromaticity [59]. Although the aromaticity of most of the lines in this study hasn’t been tested, it may be that the HKT1;5 allele from traditional aromatic lines has evolved higher activity to compensate for this deficiency and, when transferred into other genetic backgrounds, its full effect is seen.

In many cases low tissue Na⁺ concentrations (and therefore tolerance) can be largely explained by the apparent relative activity of the particular HKT1;5 allele present in a line. This suggests that it is not sufficient to declare a line as a major new donor of tolerance without first determining the HKT1;5 allele present, and this should be a component of future screening efforts. However, several exceptions do exist and the association of low leaf Na⁺ concentration with the HKT1;5 allele is not as tight as that for SES score. Examples of these exceptions include accessions such as Carolina Gold (from Peru, tropical japonica cultivar-group, Japonica allele of HKT1;5), Gachia (Bangladesh, aromatic cultivar-group, Japonica allele) and several accessions from the Philippines and China (indica cultivar-group, Hasawi allele). These all possess much lower tissue Na⁺ concentrations and higher tolerance than would be predicted from their HKT1;5 allele. Likewise, tolerant O. glaberrima lines showed very low leaf Na⁺ concentrations, yet all share an OgHKT1;5 allele with several accessions that are manifestly not tolerant and have quite high leaf Na⁺ concentrations (data not shown); thus, it seems likely that these are also using some other mechanisms apart from OgHKT1;5 that are, nonetheless, highly effective. Also, varieties from Iran and Turkey would fit in this category. Although these mostly possess the Aromatic allele of OsHKT1;5, they appear to possess an additional mechanism that limits the amount of Na⁺ entering the root (as opposed to reducing the amount of Na⁺ translocated to the shoot) and so, unlike varieties such as Pokkali and FL478, they possess both low shoot and low root Na⁺ concentrations (Figure 8). Thus, in all these cases it seems likely that alternative mechanisms besides Saltol/OsHKT1;5 (for example, reduced transpirational bypass flow, alternative sequestration mechanisms) are contributing to a reduction in shoot Na⁺ content and concentration.

Thus, while maintaining low tissue Na⁺ concentrations appears to be the predominant trait conferring tolerance in most rice genotypes, the actual mechanisms conferring low tissue Na⁺ concentration may be quite diverse. Genetic evidence from multiple QTL studies (e.g. [20, 22]) shows that while HKT1;5 contributes a major QTL for Na⁺ exclusion, a number of other minor QTLs also exist. The FL478 × Hasawi F₂ population presented here also suggests that these mechanisms can be alternately separated and combined genetically using molecular markers. Hasawi is a landrace from Saudi Arabia that shows intermediate tolerance and tissue Na⁺ concentrations (Wei et al. in preparation, current data). It is in the aus cultivar-group, and is expected to contain QTLs/mechanisms distinct from those found in traditional donors from India and Bangladesh, such as Pokkali, the presumed tolerant donor for FL478 [60, 61]. Examination of an F₂ population derived from these parents showed transgressive segregation in both the tolerant and sensitive directions. This strongly suggests that the mechanisms present in the two parental lines are distinct and can be combined to produce plants with even higher tolerance. Thus, although maintenance of low tissue Na⁺ concentrations appears to be the predominant mechanism of tolerance in O. sativa and probably O. glaberrima, there appear to be many mechanisms by which this can be achieved, and these mechanisms are possibly additive.

Despite their reputations as salt-sensitive species, both O. sativa and O. glaberrima show a wide range of diversity in salinity tolerance. This is well distributed geographically and phylogenetically, yet, at this high salt concentration, tolerance appears to be mostly related to the ability to maintain low Na⁺ concentrations in the most sensitive tissues such as the blades of active leaves. Much of the variation in this trait can be explained in reference to the HKT1;5 allele, and future screening efforts should include genotyping for this gene to determine the novelty of the germplasm being evaluated. However, despite the uniformity of mechanisms, it appears that multiple highly effective QTLs/genes/pathways are contributing to this tolerance in different accessions, and these genes show at least some, and probably considerable, potential for pyramiding. Furthermore, some lines appear to show evidence of additional mechanisms of tolerance, such as the putative tissue tolerance in Daw Hawm from Thailand and Eratio from Senegal. The number of accessions screened and characterised in this work, although specifically chosen from areas likely to produce tolerant donors, and much larger than examined in similar studies, is nonetheless relatively modest and a more extensive screening may identify additional rare donors that have complementary mechanisms. Thus, there would seem to be considerable scope for additional exploration of genetic resources apart from traditional donors frequently used in breeding, to further improve salinity tolerance of rice and ensure higher productivity of salt-affected marginal soils.