Fate of selenium in biofortification of wheat on calcareous soil: an isotopic study

Selenium (Se) biofortification of staple cereal crops can improve the Se nutritional status of populations. A field trial employing an enriched stable isotope of Se (77Se) was undertaken over three consecutive cropping seasons in a coarse-textured, calcareous soil in Gilgit-Baltistan, Pakistan. The objectives were to (1) assess the feasibility and efficiency of Se biofortification, (2) determine the fate of residual Se, and (3) assess the consequences for dietary Se intake. Isotopically enriched 77Se (77SeFert) was applied, either as selenate or as selenite, at three levels (0, 10, and 20 g ha−1) to a wheat crop. Residual 77SeFert availability was assessed in subsequent crops of maize and wheat without further 77SeFert addition. Loss of 77SeFert was c.35% by the first (wheat) harvest, for both selenium species, attributable to the practice of flood irrigation and low adsorption capacity of the soil. No 77SeFert was detectable in subsequent maize or wheat crops. The remaining 77SeFert in soil was almost entirely organically bound and diminished with time following a reversible (pseudo-)first-order trend. Thus, repeat applications of Se would be required to adequately biofortify grain each year. In contrast to native soil Se, there was no transfer of 77SeFert to a recalcitrant form. Grain from control plots would provide only 0.5 µg person−1 day−1 of Se. By contrast, a single application of 20 g ha−1 SeVI could provide c. 47 µg person−1 day−1 Se in wheat, sufficient to avoid deficiency when combined with dietary Se intake from other sources (c. 25 µg day−1). Supplementary Information The online version contains supplementary material available at 10.1007/s10653-021-00841-1.


Background
Selenium (Se) is a crucial dietary micronutrient for human health, but half a billion people worldwide are at risk of Se deficiency (Combs 2001;Fairweather-Tait et al. 2011;Ligowe et al. 2020a). It is a vital component of many selenoproteins (Brown and Arthur 2001;Antonyak et al. 2018) which play an important role in regulating various body functions, such as metabolism of thyroid hormones and protecting cells from damage by free radicals (Rayman 2000(Rayman , 2012Yang et al. 2017). Its deficiency is associated with various health disorders such as cardiovascular diseases, cancer, and reduced fertility (Tinggi 2008;Fairweather-Tait et al. 2011;Zhang et al. 2020). The main sources of Se for humans and animals are foods such as cereals, poultry, meat, and fish; contributions from drinking water and other non-food sources are nominal (Rayman 2008;Joy et al. 2015a, b).
In intermediate and low-income countries, cereals provide a large proportion (50-80%) of daily calorific intake (WHO 2019). In Pakistan, cereals, primarily wheat, account for 75% of the energy supply in an average daily diet (Zia et al. 2014). However, it is suspected that the Se concentration in wheat from Gilgit-Baltistan is normally insufficient to meet the WHO-recommended daily allowance (RDA) of Se (50-70 lg day -1 ) for an adult (Ahmad 2020). The average concentration of Se in locally grown wheat has been reported to be 29 lg kg -1 (Ahmad 2020) which would supply \ 20% (8.67 lg Se day -1 ) of the Se RDA. Deficiency of Se in human populations can be addressed in multiple ways, such as taking Se supplements, dietary diversification, food fortification, and crop biofortification through agronomic or genetic interventions (White and Broadley 2009;Broadley et al. 2006Broadley et al. , 2010Chilimba et al. 2011). Most of these strategies have various shortcomings associated with them, while the efficacy of others, such as crop improvement and genetic modification, is not yet clear (White and Broadley 2009;White 2016). However, in the case of Pakistan, particularly Gilgit-Baltistan where the population is largely dependent on indigenous agricultural produce (Rasul and Hussain 2015), crop biofortification is the most feasible approach. Furthermore, crop biofortification with fertilisers can easily enhance plant Se content Mao et al. 2014) and has been tested successfully in other countries, such as the UK, Finland, and Malawi Alfthan et al. 2015;Ligowe et al. 2019). Finland adopted crop biofortification in the 1980s with Se-enriched fertilisers and successfully enhanced Se concentration in the Finnish food supply (Broadley et al. 2006;Alfthan et al. 2015). The efficiency of Se biofortification is likely to vary with climatic conditions, agricultural practices, and soil type (Ebrahimi et al. 2019).
Therefore, experience gained in other countries may not be applicable to the study area of this project.
Inorganic Se species, selenite (Se IV ) and selenate (Se VI ), are both available for plant uptake (Broadley et al. 2006;Li et al. 2008). Selenate is normally used for biofortification because it is more soluble and hence more bioavailable (Chilimba et al. 2012a;Ligowe et al. 2020b), but it is also more prone to leaching, particularly in coarse-textured soils at high pH. It is recognised that soils in Gilgit-Baltistan are largely coarse-textured and calcareous with a high pH ([ 7) (Hashmi and Shafiullah 2003). Selenite is normally less efficient in biofortification as it is sorbed strongly by soil Fe oxides and rapidly transferred to humus-bound forms (Li et al. 2008;Ligowe et al. 2019). However, adsorption on Fe oxides in calcareous soils is likely to be weak considering the likely trend in the Se IV adsorption envelope (H 2 SeO 3 , HSeO 3 -, SeO 3 2-; pKa 1 = 2.47, pKa 2 = 7.31). It is possible that high pH might ensure continued bioavailability of 'residual' Se-i.e. fertiliser-derived Se (Se Fert ) remaining in the soil for following crops in subsequent growing seasons.
This study aimed to understand the efficiency of Se biofortification in cereal crops with a single application of Se inorganic species (Se IV and Se VI ). The objectives of this study were to: (1) study the feasibility and efficiency of Se biofortification using a single application to wheat of an enriched 77 Se stable isotope to discriminate between soil-derived and fertiliser-derived Se ( 77 Se Fert ); (2) evaluate the fate of residual 77 Se Fert in a cereal rotation (wheatmaize-wheat), as practised in Gilgit-Baltistan; and (3) assess the consequences of biofortification for dietary Se intake.

Overview
A rotational field trial was undertaken over three consecutive cropping seasons (2017)(2018)(2018)(2019) in Gilgit-Baltistan. The crops chosen were wheat followed by maize and then wheat in the third growing season. The inorganic Se species, selenite (Se IV ) and selenate (Se VI ), were applied as isotopically enriched 77 Se; each species was applied at three different levels (0, 10, and 20 g ha -1 ) to the first wheat crop in March 2018. After wheat harvest, a maize crop was planted on the site (June 2018), grown, and subsequently sampled (November 2018); a second wheat crop was then planted (December 2018) and harvested (June 2019) as before. Soil was sampled at each harvest and analysed for soluble (Se sol ), adsorbed (Se ads ), organic (Se TMAH ), and total 77 Se ( 77 Se T ); plant analysis included grain and other parts (Mathers et al. 2017;Ligowe et al. 2020b).

Site selection and management
An agricultural field at the Mountain Agriculture Research Centre (MARC) Gilgit station (35.68157 N, 74.62981 E) was selected (supplementary material Fig. A1). The total area of the experimental field was 268 m 2 (17.6 9 15.2 m). Twenty plots (2 m 9 2 m) were established with 0.4 m between consecutive plots in the central 178 m 2 (14.6 9 12.2 m) of the experimental field. A discard area of 3 m was established on all four sides of the selected area which was cultivated in the same way as the rest of the field. Four replicates of each treatment were randomly distributed in a randomised block design. All plots received the same irrigation water and basal fertiliser of 140 kg ha -1 nitrogen and 80 kg ha -1 phosphorus as per local practice. Before sowing the first wheat crop, the soil was ploughed and seeds were sown by hand in straight lines in each plot according to the local agriculture practice. However, for sowing the second (maize) and third (wheat) crop the soil was not ploughed and seeds were planted with minimum tillage in all the selected areas.
Preparation and application of 77 Se IV and 77 Se VI solutions Enriched 77 Se treatment solutions of both 77 Se IV and 77 Se VI were prepared from an isotopically enriched stock of elemental 77 Se (150 mg; 99.66% atom % of 77 Se), purchased from Isoflex, San Francisco, USA, according to the methods described in Mathers et al. (2017). The treatment solution for each field was applied at the early stem extension stage (Zadoks stage 31). The experimental field was flood-irrigated, and two days later when the field was still moist, Se treatment solutions were applied to each plot. For 77 Se application, each treatment solution was separately mixed with 6 L irrigation water in a watering can and then sprayed evenly over the plot (1.5 L m -2 ). Each treatment application was followed by spraying with 6 L (1.5 L m -2 ) irrigation water from the same container to wash the treatment solution off the crop leaves.
Soil and plant sampling and processing Soil sampling was undertaken before the experiment started (H0, November 2017) and at the end of each growing season, i.e. following the first wheat harvest (H1, June 2018), the maize harvest (H2, November 2018), and the second wheat harvest (H3, June 2019). A five-point composite sample of topsoil (0-20 cm) was collected from each plot using a stainless steel auger. The soil was air-dried and sieved (\ 2 mm), and 10 g of each soil sample was finely ground in an agate ball mill (Retsch PM 400, Haan, Germany) for elemental analysis.
At the end of each growing seasons (H1, H2, and H3), plants were harvested from the central 1 m 2 of each plot (2 m 9 2 m) at 5 cm above the ground with a scythe. Plants were subsampled (10% of the total) and then air-dried at the MARC Laboratory in Gilgit. Wheat ears were hand-threshed to separate grain and chaff. Maize plants were divided into stems, leaves, husks, and grain. All of the crop samples were separately milled using an ultra-centrifugal mill fitted with a 0.5-mm stainless steel sieve.

Soil characterisation
Soil pH was measured using a pH meter (Hanna, model pH 209) with combined glass electrode on a soil-water suspension with a ratio (w/v) of 1:2.5 after shaking end-over-end for 30 min (Rowell 1994). Oxides of Fe, Mn, and Al in soil samples were determined in citrate-bicarbonate-dithionate (CBD) extracts of finely ground soil using a single quadrupole ICP-MS (model iCAP-Q, Thermo Scientific, Bremen, Germany). The milled soil was also used for measuring soil organic carbon in a Leco TruMac CN analyser (Stockport, UK). Acid digestion (HNO 3 -HClO 4 -HF) of finely ground soil was undertaken in PFA vessels using a teflon-coated graphite block digester (Model A3, Analysco Ltd.) controlled by a Eurotherm unit (Mather et al. 2017;Ligowe et al. 2020b). Total selenium concentration (Se T ), 77 Se isotopes, and other elemental analyses were undertaken using a triple-quadrupole ICP-MS (model iCAP-TQ, Thermo Scientific, Bremen, Germany).
Processing TQ-ICP-MS 77 Se and 80 Se intensity data The raw intensity data of Se isotopes ( 77 Se and 80 Se) were exported as intensity values (counts-per-second; CPS) from the triple quadrupole ICP-MS (iCAP TQ). Both isotopes were measured in O 2 cell mode as massshifted to m/z 93 ( 77 Se 16 O) and m/z 96 ( 80 Se 16 O) to reduce interferences from Se and Ge hydrides and the 40 Ar dimer. The 77 Se intensity signals were also corrected for minor interference at mass 93 ( 76 Ge 1-H 16 O) by running a Ge standard (5 lg L -1 ). Drift correction relied on Rh as an internal standard. Calibration slopes for both isotopes ( 77 Se and 80 Se) were derived from multi-isotope Se calibration standards (SPEX CertiPrep CLMS-2; 1, 2, 5, and 10 lg L -1 ). The concentration of native 77 Se in each sample was calculated from the natural isotopic abundance of 77 Se and the measured concentration of total (native) Se (from 80 Se); the 77 Se derived from fertiliser (Se Fert ) was then obtained by difference.

Modelling the loss of Se Fert
The loss in Se Fert concentration as a function of time was described using a reversible first-order equation (Eq. 1.), adapted from Crout et al. (2006), in which Se t is Se Fert remaining in soil at time t, Se 0 is the original concentration of Se Fert added to the soil (g ha -1 ), Se Eq is the 'equilibrium' Se Fert remaining at infinite time, and K is the sum of the forward and reverse rate constants (k 1 ? k 2 ).
Calculation of distribution coefficient The distribution coefficient (kd) is the ratio of Se ads to Se sol and was calculated for both Se Nat and Se Fert , respectively, from Eq. 2: where Se ads and Se sol are the concentrations of Se in soil (lg kg -1 ) in the 'adsorbed' and 'soluble' fractions.
Calculating bioconcentration factor and bioavailability ratio The bioconcentration factor (BCF) is formulated as the ratio of Se Nat or Se Fert in the plant to their respective concentrations in soil and is a convenient index of bioavailability. The BCF values for the wheat crop at H1 for both species of Se Fert (Se IV and Se VI ) and Se Nat were calculated from Eq. 3: In Eq. 3, Se plant is Se concentration (lg kg -1 ) in each fraction (straw, chaff and grain) of wheat plant, respectively, and Se soil is total Se Nat or Se Fert concentration (lg kg -1 ) in the corresponding soils. The total soil Se Nat concentration used in Eq. 3 was measured in the soil HF digests, while the total Se Fert concentrations in soil of 4 and 8 lg kg -1 were calculated from the application rates of 10 and 20 g ha -1 , respectively, assuming a topsoil mass of 2500 t ha -1 .
The relative bioavailability of Se Fert (Se IV and Se VI ) and Se Nat in the H1 wheat crop was calculated as a 'bioavailability ratio' (B R ) from Eq. 4: where BCF Fert represents the bioconcentration factor for fertiliser-derived Se ( 77 Se Fert ) and BCF Nat is the bioconcentration factor for soil-derived Se (Se Nat ) in the plant.

Statistical analysis
Basic statistical calculations including mean, median, standard deviation, and standard error were performed in Microsoft Excel 2016, while Minitab (version 18.1) was used for the ANOVA.

Soil characteristics
Basic soil characteristics including soil pH and the concentrations of organic carbon, CaCO 3, and metal oxides (Fe 2 O 3 , MnO 2, and Al(OH) 3 ) were similar across all the field plots and did not vary significantly (ANOVA, p [ 0.05). The mean soil pH of all plots was 7.84 ± 0.05. The soil organic carbon and CaCO 3 contents were 1.60% ± 0.103 and 1.73% ± 0.270, respectively; the mean concentrations of Fe, Mn, and Al oxides were 3.13, 0.151, and 0.738 g kg -1 , respectively. The mean total soil Se and TMAHextractable Se concentrations were 139 lg kg -1 and 94 lg kg -1 , respectively, suggesting a very low overall Se concentration and the presence of a nonorganic recalcitrant Se phase (c. 45 lg kg -1 ).

Fertiliser Se dynamics in soil
The concentration of fertiliser-derived Se (Se Fert ) in soil decreased with time following its application in March 2018 (Table 1). Compared to its original application of 10 and 20 g ha -1 , the average concentration of Se Fert had decreased by 30% and 42% at wheat harvest (H1; June 2018), by 51% and 62% at maize harvest (H2; November 2018) and by 60% and 82% at the second wheat harvest (H3; June 2019) (Fig. 1). There was no significant difference in the proportion of Se Fert lost between the four treatments at any of the three harvests (ANOVA, p [ 0.05).
The decrease in Se Fert concentration at H1 can be attributed to the combined effects of plant uptake and loss of Se from the soil through leaching or volatilisation. The losses due to plant uptake at H1 were low and varied between treatments and application levels. For Se IV , removal by the crop was 1.36% and 1.46% at 10 g ha -1 and 20 g ha -1 , respectively; equivalent figures for Se VI were greater, at 3.34% and 4.32%, which reflects the greater bioavailability of selenate. The minor contribution to 77 Se Fert loss by crop uptake at H1 suggests that 77 Se Fert was lost either through volatilisation or through leaching in irrigation water due to irrigation shortly after application of 77 Se Fert . It has been reported that volatilisation of Se accounts for a small proportion (6.1%) of Se loss from sediments (Karlson and Frankenberger 1990). Hence, it is reasonable to assume that flood irrigation (widely practiced locally) was the main cause of loss of 77 Se Fert , especially as the soil has a high pH which limits the retention of Se on Fe oxides. Surprisingly, the difference in total loss between Se IV and Se VI , mainly due to leaching, was quite small which again may reflect the high pH of the soil at which Se IV adsorption is comparatively weak. Further loss of 77 Se Fert , measured at H2 and H3, was not due to plant uptake because 77 Se Fert in H2 and H3 crops was below the detection limit for 77 Se (c. 0.25 lg kg -1 ). Continued losses of 77 Se Fert from the soil through leaching may be particularly likely in these soils because of their high pH, coarse texture, and low organic carbon content (1.6%); these are all characteristics which will reduce the ability of the soil to retain added Se (Gissel-Nielsen and Hamdy 1977; Moreno et al. 2013;Lopes et al. 2017).
Speciation analysis of post-H1 soils demonstrated that the only inorganic form of 77 Se Fert present in the soluble and adsorbed fractions in most of the samples was Se IV (supplementary material Table B1). It is reported that approximately 20% of the Se VI applied to Finnish soils is taken up by plants, whereas the remaining Se is reduced to Se IV and immobilised in the soil (Keskinen et al. 2009). The findings of this study are also consistent with Stroud et al. (2010) who studied the fate of Se added as Se VI to soils in the UK and reported that Se VI was not detectable in soil samples either before or after fertiliser application. It was initially thought that the calcareous nature of the soil might ensure the survival of some of the 77 Se Fert as Se VI , at least at H1, but this was not the case.
The added 77 Se Fert did not significantly increase total soil Se because the addition of 10 and 20 g ha -1 would only contribute concentrations of 4 and 8 lg kg -1 , respectively, assuming 2500 t ha -1 of topsoil. By contrast, the average concentration of the native soil Se was 139 ± 9.12 lg kg -1 .

Fractionation of selenium
Native soil Se (Se Nat ) The three-step extraction procedure demonstrated that the fractionation of Se Nat was fairly consistent across all four sampling events (H0, H1, H2, and H3) (supplementary material Fig. A2 and Table B2). The soluble and adsorbed fractions of Se Nat accounted for \ 2.5% (0.30-3.27 lg kg -1 ) and \ 2% (1.12-2.61 lg kg -1 ), respectively, across all samples. The organically bound Se (Se TMAH ) was typically constant at 60% (75-88 lg kg -1 ) for Se Nat in H0, H1, and H2 samples but was substantially reduced to 43% (60 lg kg -1 ) in H3 samples. The remaining Se Nat of  40-57% (47.4-78.2 lg kg -1 ) may be regarded as a 'recalcitrant' fraction of Se, locked up within mineral structures (Keskinen et al. 2009;Mathers et al. 2015). Possible chemical forms of recalcitrant Se are not known. It is likely that a proportion of Se added to the soil from rock weathering, atmospheric deposition, or irrigation water could be converted to a recalcitrant phase within CaCO 3 ; alternatively, it could simply represent Se within parent material, again possibly CaCO 3 , which slowly contributes more reactive fractions of Se due to weathering. Therefore, it is not clear whether the recalcitrant Se represents a sink or a source of bioavailable Se.
Fertiliser Se ( 77 Se Fert ) The soluble and adsorbed fractions of 77 Se Fert in post-H1 soils accounted for 8.5-8.6% (0.498-0.545 g ha -1 ) and 5.3-6.3% (0.340-0.365 g ha -1 ), respectively, in 10 g ha -1 application treatments, whereas for 20 g ha -1 applications the soluble and adsorbed fractions accounted for 7.3-7.9% (0.978-1.020 g ha -1 ) and 4.0-6.1% (0.501-0.849 g ha -1 ), respectively (supplementary material Fig. A3 and Table B3). There was no significant difference in the soluble fractions (Se sol ) of Se IV and Se VI treatments and the same was observed for the adsorbed fraction (Se ads ) (ANOVA, p [ 0.05). The average combined concentrations of soluble and adsorbed fractions for 10 and 20 g ha -1 treatments varied with time and decreased to 4.54% (0.297 ± 0.15 g ha -1 ) and 3.03% (0.201 ± 0.11 g ha -1 ) in H2 soils but slightly increased to 9.80% (0.369 ± 0.266 g ha -1 ) and 6.26% (0.254 ± 0.173 g ha -1 ) in H3 soils. As found for H1 soils, there was no significant difference in the native soil soluble or adsorbed fractions of Se IV and Se VI treatments in H2 and H3 soils, respectively (ANOVA, p [ 0.05). The results of sequential extraction for 77 Se Fert and Se Nat demonstrated that a comparatively larger proportion of freshly added Se was present in the bioavailable (soluble and adsorbed) fraction at H1 compared to H2 and H3 soils. However, available 77 Se Fert was too low to contribute to plant Se uptake in H2 and H3 crops which is consistent with the findings of Gissel-Nielsen et al. (1984) who found a minimal residual availability of Se in pasture systems. Similarly, Chilimba et al. (2012b) and Mathers et al. (2017) reported minimal recovery of residual Se in maize and wheat crops, respectively. Ligowe et al. (2019Ligowe et al. ( , 2020b also reported much lower recoveries of residual Se in maize and green vegetables. The remaining 77 Se Fert in H1, H2, and H3 soil samples was all present in an organically bound form; all the remaining 77 Se Fert was extractable with TMAH, and so there were insignificant concentrations present in a 'recalcitrant' pool. The sum of all extractable fractions (soluble, adsorbed, and organically bound) in each sample was equal to the total concentration of Se Fert in the soil after each harvest, but the total concentration decreased with time compared to the original application. An average decrease of 37% was observed in 77 Se Fert concentration after 72 days (H1) since the initial application; the remaining 77 Se Fert decreased further to 43% and 29% after 224 days (H2) and 439 days (H3), respectively. Figure 2 shows the measured soil Se concentrations and modelled data (calculated from Eq. 1). The trend approached a nonzero asymptote within the time frame of the study, suggesting long-term retention of some of the 77 Se Fert . However, as already discussed, the remaining 77 Se Fert was virtually all present as humus-bound residues (extractable with TMAH) and was not available for plant uptake beyond H1. With limited data (four time points), the trends shown must be interpreted with caution. Application of 20 g ha -1 showed the clearest contrast in Se Eq (Eq. 1) between Se IV and Se VI with 40% and 26% remaining in the soil;  equivalent values for 10 g ha -1 applications were 16% and 35%, but the overall trends were very similar over the 439-day period.

Risk assessment of water contamination
The continued loss of 77 Se Fert from soil to groundwater poses a potential risk of contamination to drinking water in the catchment area. Therefore, a simple risk assessment was undertaken to estimate the loss of Se to groundwater by considering the use of 6 million L ha -1 irrigation water applied as 12 irrigation events (50 mm depth). This is a common practice in the area due to the coarse-textured soil (Mountain Agriculture Research Centre, personal communication, August 2019). Approximately 30% (* 3 and 6 g ha -1 ) of 77 Se Fert was lost at H1 for the application rates of 10 g ha -1 and 20 g ha -1 which could therefore result in concentrations of 0.50 lg L -1 and 1.0 lg L -1 Se in drainage water, respectively, after the wheat-growing season. Alternatively, in a worst-case scenario, if it is assumed that all the 77 Se Fert applied (10 and 20 g ha -1 ) to the wheat crop was lost with the first irrigation of 0.5 million L ha -1 , this would result in 20 and 40 lg L -1 of Se in drainage water. Both of the above values, 20 and 40 lg L -1 , are still below the US EPA maximum contamination level of 50 lg L -1 for Se in drinking water, and neither of these scenarios allows for the dilution of drainage water that would occur following egress of Se-enriched water from the soil into surface water systems.

Distribution coefficient of Se Nat and 77 Se Fert in soil
The kd values for Se Nat and 77 Se Fert were very low with similar mean values of 0.76 ± 0.141 and 0.70 ± 0.199 demonstrating a very low absorption capacity in this soil to retain Se in an adsorbed reactive form. This is expected because the soil has a coarse texture, high pH, and low organic carbon content. The lack of a significant difference in the kd values for Se Nat and 77 Se Fert (ANOVA, p [ 0.05) suggests that the added 77 Se has achieved isotopic equilibrium within the 'reactive' Se fractions (Se sol and Se ads ). When Se is added to soil, some proportion of it will gradually transform into organically bound or recalcitrant phases, but the reactive pools should reach equilibrium rapidly. The low kd values and low retention ability of this soil also confirm the necessity for repeated seasonal applications of Se to each crop.

Selenium in the wheat crop at Harvest 1 (H1)
The application of different levels and species of Se had no effect on crop yield which is consistent with other investigations (Curtin et al. 2008;Broadley et al. 2010;Mathers et al. 2017). The average yields of straw ? chaff and grain, based on harvest of the central 1 m 2 of each plot, were 4.8 and 3.5 t ha -1 , respectively. The concentrations of Se Nat in the aboveground biomass (sum of straw, chaff, and grain) were very small compared to 77 Se Fert and similar across all plots ( Table 2). The concentration of Se Nat varied between plant tissues (straw, chaff, and grain) in all treatments (Fig. 3). Chaff had the highest average concentration of Se Nat at 7.15 lg kg -1 followed by straw (2.87 lg kg -1 ) and grain (1.14 lg kg -1 ) in all cases. The native Se concentration in chaff and grain was constant across all treatments. However, the concentration of Se Nat in straw revealed significant variation between different treatments (ANOVA, p \ 0.05) (Fig. 3); the reason for this is not clear as there is no reason to expect a difference in Se Nat in different plant tissues caused by the 77 Se Fert application. Furthermore, similar investigations (Chilimba et al. 2012a, b;Mathers et al. 2017;Ligowe et al. 2020b) in other type of soil have not shown changes in Se Nat concentration in plant tissues across different treatments. However, with Se Nat concentration being so small in all plant tissues it is possible that they may be subject to relatively substantial systematic analytical errors.
The concentration of Se was significantly greater in all plant tissues due to fertiliser Se application (Fig. 3). Both species of 77 Se Fert (Se IV Fert and Se VI Fert ) enhanced Se concentration in the wheat plant (grain, chaff, and straw) at harvest, but 77 Se VI Fert was more effective compared to 77 Se IV Fert . A single application of 10 and 20 g ha -1 of 77 Se IV Fert resulted in a 14-and 32-fold increase in grain Se compared to an extremely low grain Se concentration of 1.42 lg kg -1 in control plots. The same application rates of 77 Se VI Fert produced a 35-and 95-fold increase in grain Se over control plots.
The greater efficiency of Se VI in enhancing plant Se content observed in this study is consistent with other investigations, notwithstanding the calcareous nature  The numbers in 'Treatments' before Se species (selenite (Se IV ) and selenate (Se VI )) indicate the Se application rate (10 and 20 g ha -1 ) of the soils in Gilgit-Baltistan. Gupta and Winter (1989) reported that Se VI applications resulted in 5-18 times greater Se concentrations in forages and barley grain compared to the same application of Se IV . Chen et al. (2002) reported a ninefold increase in rice Se in China, compared to a 6.6-fold increase, as a result of 20 g ha -1 application of Se VI and Se IV , respectively. Boldrin et al. (2013) applied Se IV and Se VI to soil, as a foliar spray, and found that Se VI was more effective in raising rice Se concentrations in Brazil for both methods of application. Ros et al. (2016) compared a large set of data on Se biofortification and found that, on average, Se VI was 33 times more effective in increasing plant Se contents compared to Se IV at the same rate of application. However, in the current trial, for each g ha -1 of Se VI applied the grain Se concentration was increased by only 4.76-6.70 lg kg -1 , which is less efficient compared to other investigations. Broadley et al. (2010), Chilimba et al. (2012a), and Mathers et al. (2017), respectively, reported increases in grain Se concentration of 16-26 lg kg -1 , 15-21 lg kg -1 , and 12.1-17.3 lg kg -1 for each g ha -1 of Se VI applied.

Recovery of 77 Se Fert species in plants at H1
The recovery of 77 Se Fert by wheat plants (sum of straw, chaff, and grain) from soil was different for the two Se species (Table 3): selenate recovery was 33-40% greater than that of Se IV Fert at application rates of both 10 and 20 g ha -1 . The average recoveries of Se IV Fert by wheat plants were only 1.36% and 1.43% for 10 and 20 g ha -1 application, respectively (Table 3); the equivalent recoveries for Se VI Fert were 3.34% and 4.32%, respectively. For both species, the recovery was greater at the higher application rate (20 g ha -1 ).
The recovery of 77 Se Fert species also varied between plant tissues (Table 3); grain had a higher recovery in all cases followed by chaff and straw.
Compared to other investigations on field-grown wheat and other cereal crops, the recovery of both species was low. Mathers et al. (2017) reported 25.9-44.5% recovery in wheat plants grown on three contrasting sites in the UK with an application of 10 g ha -1 of 77 Se VI . Broadley et al. (2010) observed recoveries of 19.6-34.7% in a wheat crop (grain ? straw) following Se VI (aqueous Na 2 SeO 4 ) application at six different rates (ranging from 5 to 100 g ha -1 ) in the UK. However, Lyons et al. (2004) reported slightly smaller recoveries of 1.8-9.3% for foliar application of Se VI at four different rates (10, 30, 100, and 300 g ha -1 ) to a wheat crop in Australia. Similarly, Ducsay and Ložek (2006) reported plant uptake of 2.4-9.3% for foliar application of Se to a wheat crop in Slovakia.
The recovery of 77 Se Fert in wheat grain (first crop) observed for different treatments (Table 3) (Lyons et al. 2004), and 2-6% (Stephen et al. 1989). Recovery was also less than the 6.5-10.8% values reported by Chilimba et al. (2012a) for a maize crop.
The low 77 Se Fert recoveries in this study compared to other investigations conducted elsewhere could be due to low crop yield, soil characteristics, and possibly a limited period of 77 Se Fert availability for plant uptake in the Gilgit-Baltistan soils. Soil texture was sandy with low CaCO 3 and organic carbon contents, which may reduce nutrient retention and thereby render the 77 Se Fert more prone to leaching after irrigation. The values of BCF for 77 Se Fert (Se IV and Se VI ) were significantly larger than those of Se Nat (Fig. 4). This is consistent with findings from Mathers et al. (2017) who observed a higher BCF for wheat plants fertilised with Se VI compared to their control plots. Furthermore, as expected, the BCF of 77 Se VI Fert was greater than that of 77 Se IV Fert (Fig. 4) due to greater solubility . The other reason for differences in BCF between Se IV and Se VI may lie in the mechanisms governing root uptake from soil and subsequent translocation to aerial parts of the plant. Selenate is freely absorbed by plant roots via sulphate transporters and then transported through xylem vessels into plant stems and leaves, whereas a fraction of Se IV is likely to be converted to Se VI or organic Se before being transported to other parts of the plant (White 2016;Ali et al. 2017). The BCF also varied between different parts of the plant in the same treatment as the BCF values for 77 Se Fert in grain and chaff were similar but substantially greater than for straw due to different abilities of tissue in various parts of the plant to accumulate or retain Se. The other reason for the greater concentration of Se in grain is the transportation of Se in phloem from leaves to grains.
Differences in BCF between 77 Se Fert and Se Nat and between the applied 77 Se Fert species of Se IV and Se VI reflect differences in bioavailability. The average B R of Se Fert /Se Nat for the H1 wheat crop confirmed the large difference in bioavailability of 77 Se Fert compared to Se Nat to all tissues of the wheat (Table 4); in particular, the bioavailability of 77 Se VI Fert to wheat grain was 764 and 1261 times that of Se Nat at applications of 10 and 20 g ha -1 , respectively. The B R values of 77 Se Fert /Se Nat were comparable with other Se biofortification investigations (Stroud et al. 2010b;Chilimba et al. 2012a;Galinha et al. 2012) as cited in Mathers et al. (2017) and confirmed the greater bioavailability of 77 Se Fert VI (Table 4). Ali et al. (2017) studied the bioavailability of Se IV and Se VI to wheat plant in a pot experiment and found that mobility and availability of Se VI for plant uptake was 40-90% higher than that of Se IV . Similarly, Fan et al. (2015) from his studies on transformation and bioavailability of Se (Se IV and Se VI ) applied to tobacco plants reported that Se VI was 4.3-7.9 times more bioavailable than Se IV .
Values of B R varied between different parts of the plant (Table 4) and were greater in grain than in chaff and straw (ANOVA, p \ 0.001); there was no significant difference between chaff and straw (ANOVA, p [ 0.05). Possible reasons for a greater B R in grain lie in the timing, mode of 77 Se Fert application, and species of 77 Se Fert . The 77 Se Fert was applied at early stem extension stage, and its greater bioavailability compared to Se Nat would produce a greater concentration of 77 Se Fert in leaves which is then supplied to grain when its formation begins. The concentration of Se is generally greater in leaves before anthesis and then starts to decrease when its translocation from leaves to reproductive organs begins (White 2016). Furthermore, because 77 Se Fert was sprayed on the plant canopy it is likely that some of it may have been absorbed into leaves and translocated via phloem from leaves to grain. The B R values for Se VI were greater compared to Se IV because Se VI is readily translocated from roots to other parts of the plant, while Se IV tends to accumulate in roots (Li et al. 2008).

Consequences of biofortification for dietary Se intake
Wheat is the most important staple crop in Pakistan (Raza 2018) and forms an essential component of the daily diet for the majority (* 80%) of the local population (Zia et al. 2014). The average daily consumption of wheat-based food in Pakistan is 350 g person -1 and accounts for 75% of the daily calorific intake for an individual (Zia et al. 2014). Considering the average daily wheat consumption, the Se concentration in wheat grown on the control plots of this experiment would provide only 0.5 lg person -1 day -1 which accounts for \ 1% of the WHO-recommended (55-70 lg day -1 ) daily Se intake (Table 5). However, the biofortified wheat grain produced in this study with the application of 10 and 20 g ha -1 77 Se Fert VI would provide 17.2 and 47.4 lg day -1 accounting for 31.2% and 86.2% of the daily requirement. By contrast, the application of 77 Se Fert IV at the same level would provide only 7 and 16.1 lg person -1 day -1 contributing 12.7% and 29.3% towards the recommended daily Se intake. Based on Se concentrations in a typical Pakistani food basket from values reported in the literature, the Se contribution from other sources is around 25 lg day -1 (Hussain 2001;Iqbal et al. 2008;USDA 2019). Therefore, the consumption of wheat flour obtained from 10 and 20 ha -1 Se VI application would increase Se intake to 42.2 and 72.4 lg day -1 person -1 for individuals with access to diverse food sources, assuming that 25 lg day -1 has come from those other sources.
The results of this study suggest that wheat biofortification with just 20 g ha -1 of Se VI would be a successful strategy for boosting daily per-capita Se intake into the range required to avoid Se deficiency. Chilimba et al. (2012a) estimated that the addition of 5 g ha -1 of Se to Malawi soils represented a raw material cost of just 1.6-3.5 US cents ha -1 year -1 .

Residual Se in the crop at H2 and H3
No 77 Se Fert could be detected in the first rotation crop (maize) at H2 or the second wheat crop at H3. Thus, within analytical error, all the Se measured in all parts of the maize (grain, husk, leaf, and stem) and wheat (straw, chaff, and grain) was soil-derived. This is consistent with the observation that the major proportion of 77 Se Fert remaining in the soil had become organically bound (TMAH-extractable) and was unavailable for plant uptake. The TMAH extraction follows extraction with phosphate which should have dissolved 'adsorbed' forms of Se. Whatever 77 Se Fert was still available in the soluble and adsorbed fractions was at too low a concentration to measurably contribute to plant uptake. The negligible recovery of residual Se in the following crops is consistent with other investigations. Mathers et al. (2017) observed a negligible amount of residual Se in follow-up crops in his experiment on 77 Se application to winter wheat at sites in the UK with 10 g ha -1 Se VI applications. Stroud et al. (2010b) reported that no significant difference was found in Se concentration of wheat grain from control plots compared to those that were previously treated with 10 or 20 g ha -1 of Se as Na 2 SeO 4 . Similarly, Gupta et al. (1993) also reported no effect of residual Se on barley grain Se content in Canada with applications at three different rates (10, 20, and 40 g ha -1 ) of Se IV and Se VI , respectively. Chilimba et al. (2012b) reported a very small recovery (0.78% and 2%) of isotopically labelled residual 74 Se in maize grain at two different field sites in Malawi, treated with 10 g ha -1 74 Se VI the previous year. In addition, Wang et al. (2017) also reported a 77.9% and 91.2% reduction in Se concentration in wheat and maize grain from plots, which were treated with 30 and 60 g ha -1 of Se in previous cropping season in China. Almost all previous studies suggest that, regardless of soil type, virtually all applied Se retained by topsoil is rendered unavailable for uptake by the time of a second crop. Treatments below indicate the Se application rate (10 and 20 g ha -1 ) and the Se species (selenite (Se IV ) and selenate (Se VI ))

Conclusions
Both species of 77 Se Fert (Se IV and Se VI ) at application rates of 10 and 20 g ha -1 increased Se concentration in a wheat crop. However, Se VI was more efficient in increasing wheat grain Se at H1 and produced 33-40% greater recovery compared to Se IV . A single application of 20 g ha -1 Se VI increased Se concentration in wheat grain to 135 lg kg -1 compared to an extremely low Se concentration of 1.42 lg kg -1 from control plots. Considering an average per-capita wheat consumption of 350 g day -1 in Pakistan, 20 g ha -1 Se VI would provide c. 47 lg person -1 day -1 of dietary Se, which would be enough to avoid Se deficiency in the Gilgit-Baltistan population. There was no 77 Se Fert detected in subsequent harvests: H2 (maize) and H3 (wheat) crops suggesting continued annual applications of Se Fert would be required to sustain viable biofortification. On average, 71.1 ± 16.9% of 77 Se Fert was lost from the soil at the time of final soil sampling (H3-July 2019). There was no evidence of 77 Se Fert assimilation into a recalcitrant mineral phase within the timescale of the trial despite clear evidence of native recalcitrant Se. However, the 77 Se Fert which remained in soil after every harvest was largely (C 90%) present in an organically bound form (TMAH extractable) which rendered it completely unavailable for plant uptake so that 77 Se Fert was not detected in H2 and H3 crops. Therefore, it is clear that Se application to crops in every growing season would be required to obtain an increased concentration of Se in crops. Further research is required to more fully assess (1) the effect of Se lost from the soil on water quality immediately following application and (first) irrigation and (2) any longer-term residual effects from repeated applications as the capacity of these soils to continue fixing Se into organic forms is unknown.