Laboratory Microprobe X-Ray Fluorescence in Plant Science: Emerging Applications and Case Studies

In vivo and micro chemical analytical methods have the potential to improve our understanding of plant metabolism and development. Benchtop microprobe X-ray fluorescence spectroscopy (μ-XRF) presents a huge potential for facing this challenge. Excitation beams of 30 μm and 1 mm in diameter were employed to address questions in seed technology, phytopathology, plant physiology, and bioremediation. Different elements were analyzed in several situations of agronomic interest: (i) Examples of μ-XRF yielding quantitative maps that reveal the spatial distribution of zinc in common beans (Phaseolus vulgaris) primed seeds. (ii) Chemical images daily recorded at a soybean leaf (Glycine max) infected by anthracnose showed that phosphorus, sulfur, and calcium trended to concentrate in the disease spot. (iii) In vivo measurements at the stem of P. vulgaris showed that under root exposure, manganese is absorbed and transported nearly 10-fold faster than iron. (iv) Quantitative maps showed that the lead distribution in a leaf of Eucalyptus hybrid was not homogenous, this element accumulated mainly in the leaf border and midrib, the lead hotspots reached up to 13,400 mg lead kg-1 fresh tissue weight. These case studies highlight the ability of μ-XRF in performing qualitative and quantitative elemental analysis of fresh and living plant tissues. Thus, it can probe dynamic biological phenomena non-destructively and in real time.


INTRODUCTION
X-ray fluorescence (XRF) spectroscopy is a well-established analytical technique for qualitative and quantitative elemental evaluation. It is a multielemental, simultaneous technique and additionally a non-destructive tool, thus being suitable for in vivo plant analysis (Haschke, 2014). Particularly, the energy dispersive X-ray fluorescence microprobe (µ-XRF), as microanalysis technique, allows single point, 1D line and 2D mapping elemental determination in broad range of applications in agricultural and forestry science.
Conversely, other analytical techniques, which permit direct sample analysis, such as laser-induced breakdown spectroscopy (LIBS) and laser ablation inductively coupled plasma mass spectrometry/optical emission spectroscopy (LA-ICP-MS/OES), are destructive techniques. Therefore, they do not allow in vivo analysis. Hence, herein we present some µ-XRF cases studies in plant science exploring this unique advantage.
X-ray fluorescence is the emission of characteristic electromagnetic radiation resulted from a relaxation process. Figure 1 illustrates a series of events that take place during photon induced atomic excitation and further relaxation. Once an X-ray photon impinges upon matter with energy higher than the ionization energy of an inner shell electron, the latter particle may be ejected, producing a vacancy in the corresponding orbital. Subsequently, an electron from an upper orbital fills this vacancy, and the excess of energy can be emitted as a photon. If one ignores screening effects, the energy of this photon is approximately equal to the difference between the energy of the upper and inner orbitals. The emitted energy is characteristic for each chemical element, thus one can use this energy as a fingerprint that allows the elemental identification (Van Grieken and Markowicz, 1993). Moreover, the number of emitted photons is directly proportional to the amount of emitting atoms, thus the XRF peak area yields quantitative information.
There are several types of Energy Dispersive X-ray Fluorescence (EDXRF) systems available (Tsuji et al., 2005;Beckhoff et al., 2007;Margui, 2013). One can highlight the conventional benchtop EDXRF spectrometer, handheld EDXRF (McLaren et al., 2012;Kalcsits, 2016;Guerra et al., 2017) and µ-XRF ones (Beckhoff et al., 2006;Margui, 2013;Haschke, 2014). They can perform analysis of liquids and solids, normally detecting all elements with atomic number above Na in the mg kg −1 concentration range (Beckhoff et al., 2006;Margui, 2013;Navas et al., 2016). It is important mentioning that the sensitivity depends also on the chemical element, for example heavier atoms such as Fe or Zn present higher sensitivity, and therefore lower limits of detection (LOD), than P or K.
FIGURE 1 | The incoming photon is absorbed by an inner shell electron, the electron is ejected with certain kinetic energy that is equal to the difference between the energy of incoming photon and the binding energy. The departure of the electron excites the atom, then it relaxes while an outer electron fills the hole left in the inner shell. During the relaxation, the atoms can emit heat, another secondary electron (Auger process) or a characteristic photon (XRF).
The basic design of a µ-XRF spectrometer is presented in Figure 2. A similar µ-XRF equipment, specially designed for the in vivo analysis of plants, was built by Fittschen et al. (2017). The X-ray beam, usually produced by the collision of an electron beam against a metallic anode, is shaped and size defined by a primary optic element. This can be a simple collimator, an optical capillary or a focusing mirror. The sample is assembled in a positioning system and the X-ray fluorescence is analyzed by a detector able to discriminate both, photon yield and energy.
X-ray microprobes can be employed as a high throughput analytical system or explored when lateral resolution is required (Tian et al., 2015). In the framework of plant science, µ-XRF presents some complementary features compared to other microprobe and elemental imaging techniques such as scanning electron microscopy-energy dispersive spectroscopy (SEM-EDS) and transmission electron microscopy (TEM-EDS). One can highlight the simplicity of sample preparation for µ-XRF, since differently from SEM and TEM, it does not require conductive coating or thin slicing. Additionally, the µ-XRF LOD are at least one order of magnitude lower than those of SEM-EDS. It happens mainly due to the smaller spectral background found in XRF. Finally, µ-XRF does not require vacuum, although low pressure or He atmosphere can improve the ability of detection for light elements (Z < 22). Altogether, these features make µ-XRF an ideal tool for the analysis of fresh vegetal tissues or even in vivo plants. Table 1 shows the main µ-XRF advantages and drawbacks.
One of the major challenges in µ-XRF regards the matrix effects. The probability of the penetration of the excitation X-ray beam, as well as the probability of escape of an X-ray photon emitted by an analyte within the sample, depends on the matrix composition, density, and energies involved in excitation and emission. The XRF photon yield of a given element at a certain concentration in a light atom matrix, FIGURE 2 | Microprobe X-ray fluorescence spectrometer scheme. The X-ray beam coming from the excitation source is size defined by a focusing or collimating primary optic element. Once it excites the sample placed on a sample holder, the X-ray florescence energy is discriminated and counted by a detector.
Frontiers in Plant Science | www.frontiersin.org Allowing heterogeneous sample analysis carbon for instance, will be higher than the XRF yield for this same element at the same concentration in a heavier matrix such as silicon. Therefore, calibration curves must be built using standards whose matrices are as close as possible to those of the sample. Other possibility consists in calculating the photon attenuation caused by the matrix; this is called fundamental parameter approach. The LOD of µ-XRF in general are several orders of magnitude higher than those provided by LA-ICP-MS. The matrix effects can be neglected for samples considered "infinitely thin, " it means that the sample thickness does not break the linear relationship between the XRF intensity and the concentration (mass per unit area). In this case, the self-attenuation of the XRF is low enough to be ignored (Van Grieken and Markowicz, 1993;Duran et al., 2017). In case of "intermediate thickness, " i.e., the XRF intensity depends on the sample thickness, the matrix effects can be calculated and corrected (Van Grieken and Markowicz, 1993;Blonski et al., 2007). These two situations will be treated in detail below. There are also samples called "infinitely thick." In this case, the increasing the sample mass per unity of area does not influence the intensity of the XRF peak. For this type of sample, the quantitative analysis is usually accomplished by external analytical curves, standard addition, fundamental parameters or calibrating the equipment with a set of reference certified material whose matrix is similar to that of the sample.
One of main concerns regarding in vivo XRF measurements regards radiation-induced tissue damage. The radiation can damage biological tissues due to photolysis or heating. The dose necessary to cause such damages is ca. 10 7 Gy (1 Gray = 1 Joule kg −1 ) (Henderson, 1990). There was no evidence of radiation induced damage in the studies reported hereafter.
One the main limitations of XRF in the agronomic context concerns the difficulties involved in the detection of nitrogen and boron. Firstly, the XRF yield of such elements are low since they relaxes mainly via Auger emission (Van Grieken and Markowicz, 1993). Also, the low energy of the XRF photons require the use of vacuum.
Other fact that must be improved regards the lateral resolution, or magnification of XRF imaging. The beam size available in benchtop machines (tens of micrometers wide) (Janssens et al., 2000;Tsuji et al., 2005;Haschke, 2014) are relatively large if compared to synchrotron facilities (from micrometers to nanometers) (Castillo-Michel et al., 2017). This limits the application of benchtop µ-XRF to the investigation of a collection of cells while in synchrotrons a single cell can be mapped.
For infinitely thick and intermediate thickness samples, the elemental spatial correlation can become trick since the probability of photon escape depends on both energy and depth. For example, in a cellulose matrix 50 µm thick and density of 1 g cm −3 the transmission of Kα photons from Zn is 94% whereas that of Kα emitted by P is 0.8%. Thus, if a bunch of Zn atoms is beneath a bunch of P atoms, in a 2D projection they may appear in the same region indicating positive spatial correlation while they do not occupy the same volume in the space, this issue is illustrated in the Supplementary Figure 1.
Considering that under realistic conditions plants are exposed to micronutrients at concentration ranging from sub to few mg L −1 and normal concentration levels in tissues falls typically in tens of mg kg −1 , the main challenge concerning in vivo studies lies on the improving LOD. A strategy to achieve it could rely on customized equipment carrying optimized primary filters and higher solid angle detection. This latter one can be done by increasing the surface area of detectors (which also increases costs) or allowing experimentalist to tune the distance between sample and detector.
Our group has applied benchtop laboratory µ-XRF in plant science and this manuscript aimed at bringing up a set of µ-XRF methods and procedures which have been developed and used mainly for agronomical and environmental studies. Our goal is to highlight the application µ-XRF for the plant science community. In this paper, we feature the usage of µ-XRF to unravel the spatial distribution of nutrients in primed seeds, the dynamics of elemental redistribution on a fungi infected leaf, the in vivo monitoring of competitive uptake of Mn and Fe passing through the stem of common bean and finally the spatial distribution of Pb accumulated in the leaves of Eucalyptus.

MATERIALS AND METHODS
All experiments were carried out using a benchtop µ-XRF system (Orbis PC EDAX, United States) furnished with a Rh anode with max power rating at 50 kV and 1000 µA. The machine operates with 1 and 2 mm collimators, or 30 µm polycapillary optic. This facility is also equipped with 25 µm Al, 25 µm Ti, 25 µm Ni, 100 µm Rh, 127 µm Nb, and 250 µm Al optional primary filters. The detection was carried out by a 30 mm 2 silicon drift detector (140 eV FWHM at the 5.9 keV Mn-Kα line). The specific conditions such as X-ray beam size, tube current and voltage, dwell time, and number of points used in each of the examples are shown in Table 2. The pixels produced by the Orbis Vision software were linearly interpolated using Origin Lab 2016, the details on the algorithm can be found at Origin Lab user's manual.

Quantitative Zn Mapping in Primed Bean Seed
Phaseolus vulgaris (common bean) seeds were soaked for 20 min in a ZnSO 4 .7H 2 O solution at 1,000 mg L −1 , dried at room temperature for 24 h and carefully cut in the middle using a stainless-steel blade. Part of the seeds had their backs sliced yielding 1-2 mm sections while another part was preserved as shown in Supplementary Figure 2. The split seeds were placed in a sample holder with a polyimide tape with the cotyledon's inner side exposed for analysis. Supplementary Figure 3 presents the experimental setup used in this analysis (see Table 2 for details on the instrumental parameters).
To build up the quantitative Zn map of the treated seeds, we determined the Zn concentration using equation (1) (Leroux and Mahmud, 1966).
where, C is the Zn concentration in the sample (µg cm −2 ), I sample above threshold is the Zn XRF net intensity emitted by the analyte (cps, counts per second), Ab correction is the absorption correction factor (dimensionless) detailed in equation (3) and (4), and S is the elemental sensitivity for Zn (cps µg −1 cm 2 ). The threshold (cps) selects the analytical signal suitable for quantitative analysis, it is calculated according to equation 2.
where, BG (average) (cps) corresponds to the average background calculated from ten randomly selected points within the sample and t(s) is the dwell time per point. The Ab correction is given by: with: where I sample+irradiator is the Zn XRF net intensity from the sample plus irradiator (cps), whereas I 0 is the Zn XRF net intensity from the irradiator (cps), and I sample is the Zn XRF net intensity emitted by the sample (cps). The factor T expresses the sample transmittance to Zn Kα radiation. The instrumental sensitivity for Zn was calculated measuring a ZnTe standard thin film manufactured by Micromatter TM , Canada (Serial Number 6330). The equation (5) shows how the instrumental sensitivity was calculated.
where, I is the Zn XRF net intensity (cps) and C is the Zn concentration (16.2 µg Zn cm −2 ) of the ZnTe standard thin film.
The sample holder used for this emission-transmission analysis is shown in Figure 3a. It is regular XRF cuvette modified to allow inserting and removing the Zn disk irradiator just below the sample. This is important since I sample , I sample+irradiator and I o must be measured at a fixed distance from the X-ray source and detector.
The trueness of this method was checked measuring a set of five standards: 200, 400, 600, 800, and 1000 mg Zn kg −1 cellulose pressed pellet. This reference pellet was prepared using cellulose binder for XRF with particle size ≤20 microns PA (SPEX, United States) spiked with 1,000 mg of Zn L −1 (SpecSol, Brazil) standard solution. The pellet was made transferring 150 mg of the spiked cellulose to 15 mm diameter set die and pressed at 8 ton cm −2 for 3 min (Spex model 3625B X-Press). The assembling of the pellet on the top of the sample holder is shown in Figure 3b. In vivo Characterization of P, S, K, and Ca Distribution in Fungi Injured Soybean Leaves Leaves of a soybean plant were infected with spores of Colletotrichum truncatum. An agar plate containing the fungus was rinsed with deionized water to suspend the spores, then this suspension was filtered. Finally, the spores were spread on the adaxial face of the leaves using a Drigalski spatula.
The leaves were moistened by water spraying and the plants were incubated in plastic bags for 3 days. This procedure aimed at maintaining constant and adequate humidity for the inoculated fungus. Then, the plants were assembled in a homemade acrylic sample holder that was specially designed to keep the plant alive, in which the leaves were stretched for the analysis (see Supplementary Figure 4). The symptoms of the infection on leaves were daily monitored.
A region of 2 × 1.54 mm 2 , which comprised the fungi caused injury, was selected. The spatial distribution of the elements P, S, K, and Ca was in vivo determined using µ-XRF (see Table 2 for details on the instrumental parameters). The maps were recorded on the 3rd, 4th, and 5th day after the pathogen exposure.

In vivo Root Uptake and Transport of Fe and Mn
Phaseolus vulgaris plants were cultivated using vermiculite in a growth room at 27 • C and photoperiod of 12 h under LED lamps illumination of 6500 K, which supplied 250 µmol photons m −2 s −1 . At V3 stage, their roots were immersed in a solution containing both monohydrated FeSO 4 (9 × 10 −4 M) and MnSO 4 (9 × 10 −4 M).
The plants were assembled in a homemade sample holder. The flask containing the solution was covered with aluminum layers to avoid possible XRF fluorescence coming from the solution (see Figure 4). The measurements were carried out during 48 h of constant root exposure. After each measurement, the plants were returned to the growth chamber. The measurements were performed at the stem, approximately 25 mm above of the root crown. Two plants were examined, and no signs of radiation induced damage were observed during the analysis.   (Correia et al., 1995) for 50 days in a growth room. Roots were removed, and the shoots were transferred to the same culture medium supplemented with 0.5 g of 6BAP, 0.6% agar, 3% sucrose with the pH adjusted to 5.8. Every 21 days the plant material was transferred to a new bottle for nutritional supply. After four subcultures, the explants were transferred to a new culture media supplemented with 1,000 mg Pb (NO 3 ) 2 L −1 remaining exposed to this salt for 7 days.
A leaf from middle region of the shoot was cut from the plant and placed on a sample holder (see Supplementary  Figures 5A,B). The µ-XRF measurement was performed using a fresh leaf. The mean background (BG mean , cps) under the corresponding Pb Lα peak was determined through ten randomly points selected in the leaf mapped region. The analytical signal was separated from the background using the threshold equation (2).
The sensitivity, S (cps µg −1 cm 2 ) for Pb was calculated using a Pb thin film standard Micromatter (Serial Number 6331) containing 48.3 µg Pb cm −2 (like shown in equation 5). The quantification of Pb was carried out considering that the leaf was an infinitely thin sample. This assumption was verified measuring the attenuation of the Lα radiation emitted by a Pb disk (Ø 25 mm × 1 mm) by several leaf samples. The Tukey test at 95% confidence interval was performed to compare the obtained means.
The concentration of Pb in the leaves was determined using the equation (6) C(µg g −1 ) = I sample(cps) D(g cm −2 ) S cps µg −1 cm 2 where I sample (cps) is the XRF net intensity of Pb Lα in each point of the map and D is the leaf surface density (g cm −2 ). The trueness of the method was evaluated by measuring a thin cellulose pellet (0.0184 g cm −2 ) spiked at 1000 mg Pb kg −1 . Five spectra were recorded at the same instrumental conditions employed for the leaf maps.
In addition to the quantitative Pb spatial distribution maps, the chemical images for K and Ca are presented. Also, the correlation between Pb, Ca, and K was accessed through scatter plots.

Quantitative Zn Mapping in Primed Bean Seeds
The spatial determination of minerals present in seeds can strengthen biofortification efforts, improve the understanding of the elemental redistribution during seed processing, steps such as grain polishing and crushing. Altogether, this can ensure that nutrients will reach the final consumer (Gregory et al., 2017;Mathers et al., 2017;Zaman et al., 2018).
Synchrotron-based µ-XRF was used to investigate the localization of Zn in harvested grains of wheat that was biofortified through foliar Zn application (Ajiboye et al., 2015). µ-XRF has also been used to measure the mineral distribution of seeds during its development (Iwai et al., 2012), evaluating the elemental distribution between distinct genotypes (Singh et al., 2013), and determine the location of elements in rice grains (Kyriacou et al., 2014). Here we show the quantitative distribution of Zn in two bean seeds exposed to ZnSO 4 using a µ-XRF laboratory benchtop facility and the emission-transmission quantitative method.
The emission-transmission method requires the determination of three independent XRF net intensities: sample (cps), irradiator (cps), and sample+irradiator (cps). In the present study, two maps were enough to yield these three values, since in the map for the sample+irradiator condition part of the mapped area was not covered by the seed, and thus it gave the irradiation intensity (I 0 ). Figure 5 presents the steps involved in data processing. Figures 5A,B show the XRF maps recorded for I sample and I sample+irradiator , respectively. By applying equation 4 to each pixel of these maps, one obtains the sample transmittance shown in Figure 5C. The regions of transmittance equal to 1 corresponds to the part of the irradiator not covered by the seed sample.
Ideally, transmittance values above 1 should not be observed. Considering the 30 µm beam size, we believe that it was caused by micro heterogeneities of the irradiator. Another alternative to correct it consists in increasing the dwell time per pixel, which would improve the precision of the XRF counts. Figure 5D resulted from equation 3, it is the inverse of the well-known absorption factor found the fundamental equation of XRF. Figures 5C,D shows that despite the slicing, the sample thickness was not homogenous. Figure 5E presents Zn Kα XRF intensity above the quantitation threshold. These values were obtained by applying equation 2 in Figure 5A. Finally, Figure 5F results from applying equation 1 to Figures 5D,E. Figure 6 shows the mapped area (rectangle) for two primed bean seeds overlaid by the corresponding quantitative Zn chemical images. The seed shown in Figure 6a was sliced intending to avoid any spectral artifact coming from the seed back, whereas Figure 6b shows the image for a hemi-seed. Despite observing no spectral artifacts when analyzing a whole seed, the slicing procedure is still recommended. Since the X-ray escape depth increases as function of energy, the probability of having X-rays coming from the seed back would increase for a same line series (K or L, for instance) as function of the atomic number. This, in its turn, could lead to misinterpretation of the data, because the X-rays coming from the outer seed coat would yield a chemical image suggesting that the priming nutrient crossed the seed coat and penetrating in the endosperm. This process is illustrated in Supplementary  Figure 2. We previously observed this type or artifact for molybdenum (Kα 17,480 eV) soybean treated seeds (not shown here).
The trueness of this method was evaluated using a set of five standards (Supplementary Figure 6). The recovery ranged from 91 to 106%. Considering the pellet measurement, the limit of quantification (10σ) of the method was 1.6 µg Zn cm −2 .
Most of Zn remained trapped in the seed coat, even though the seed priming solution delivered a soluble Zn form. It was observed a Zn hotspot (reddish area) in the hilum region (spongy tissue that allows water uptake into the seed).
Measurements of intermediate thickness samples using the emission-transmission method can be interesting for environmental (Markowicz et al., 1996), geological (Funtua, 1999), biological (Funtua, 1999;Bamford et al., 2004;Bako et al., 2008), and drug (Mahawatte et al., 2006) samples. In plant science, this method was applied to quantify the elements in healthy and fumagine infected orange and lemon leaves. The quantification of Ti, Mn, Fe, Cu, and Zn gave limits of detection that ranged from 1 to 10 µg g −1 (Blonski et al., 2006). The same method was also employed to quantify inorganic elements in tobacco leaves (Oyewale et al., 2002) and radish plants (Gupta et al., 2007).
Regarding alternative techniques, simultaneous particleinduced X-ray emission using a focused ion beam (µ-PIXE), Rutherford backscattering (RBS) and scanning transmission ion microscopy (STIM) analyses yielded quantitative map of metal concentration in Arabidopsis thaliana seeds (Schnell Ramos et al., 2013).

In vivo Characterization of P, S, K, and Ca Distribution in Fungi Injured Soybean Leaves
X-ray fluorescence microprobe can support understanding of chemical composition of the infected tissues by bacteria, fungi, virus, and their proliferation in plants while alive. It support the understanding which elements plants may redistribute to oppose the invasion, or the nutrients that pathogens take up to develop. This information enables to find management strategies which mimics the self-defense mechanism of plants. In principle, following such spatial distribution patterns, one could recognize infections even before the appearance of visual symptoms and therefore avoid losses. Figure 7 shows disease development from the 3rd to the 5th day of fungi infection. The images present a picture of the soybean leaf, the infected region corresponds to the darker areas. The other figures are chemical images that unravel the spatial distribution of P, S, K, and Ca in the same of area shown by the picture. Additionally, we also show the Rh Kα Compton scattering map that indicate the presence of veins in the leaf. The dynamic chemical images showed that P, S, and Ca trended to accumulate in the injured area as the disease spreads. On the contrary, K seems to be depleted in the region attacked by the fungi.
Although there are few studies reporting the spatial distribution of elements during diseases, this behavior is probably correlated to a defense mechanism of the plant (Taiz and Zeiger, 2013). The optical image demonstrates the existence of a pre-damaged area around the pathogen where the calcium concentration was higher. Since the fungi is a necrotrophic pathogen (Ranathunge et al., 2012), the defense response of the plant consists in isolating the infected area avoiding the spread of the disease to the rest of the plant in tissue (Taiz and Zeiger, 2013). Nürnberger et al. (1994) showed that elicitors (molecule extrinsic coming from pathogens connect with plant proteins triggering defense mechanism) induces the influx of Ca 2+ and efflux of K + to cells (Nürnberger et al., 1994). Other study showed that Ca 2+ triggers the cell death program (Atkinson et al., 1990;Tavernier et al., 1995;Lamb and Dixon, 1997), then the increase of Ca XRF intensity might be linked to this mechanism.

In vivo Root Uptake and Transport of Fe and Mn
In this section, we show an example of how µ-XRF can be used to in vivo trace the absorption and transport of Mn and Fe in P. vulgaris. Although synchrotron XRF was previously used to monitor the uptake of Ni, Mn, and Cr (Hwang et al., 2016) and thallium (Scheckel et al., 2004), to the best of our knowledge, this is the first report monitoring the simultaneous uptake of two nutrients in a living plant using a benchtop XRF equipment.
Basically, three strategies are used to monitor the root to shoot uptake of elements in solution. Firstly, one can monitor the absorption using radiolabeled elements (von Wiren et al., 1995;Hart et al., 1998;Rengel et al., 1998). Additionally, it is possible to monitor the depletion of the target element in the solution to which roots are exposed (Hart et al., 1998). Finally, one can also collect the vegetal tissues from time to time and determine the concentration of the desired elements trough chemical or instrumental analysis (Avenue and Africa, 1973;Puccinelli et al., 2017).
Radioisotopes grant researchers the possibility of monitoring the uptake and transport of nutrients while the plant is alive, however, their manipulation involves many issues regarding safety and are not easy to acquire. The collection of aliquots from the solution in which roots are exposed allows tracing whether nutrients were absorbed, however, this procedure does not allow inferring about transport. Finally, the analysis of the tissues collected from several different individuals can introduce errors to the experiment due to intrinsic differences that they may present, moreover this procedure involves a large number of samples, it is also laborious and time consuming since several steps are involved for sample preparation and analysis.
In this context, µ-XRF can be useful to investigate the mechanisms of nutrient uptake and transport. The elements can be simultaneously monitored, and the X-ray beam can probe different parts of roots, stem, or leaves. Last but not least, the measurements can be carried out in vivo plants. The continuously increasing uptake tendency for Mn and Fe, in addition to previously published studies suggested that no measurable radiation damaged occurred Gomes et al., 2017).
In the present study the absorption and transport rates of Fe and Mn were monitored in vivo. Figure 8 shows the number of counts for Fe-Kα and Mn-Kα, normalized by the sensitivity of each element, as a function of exposure time (Figures 8A,B are biological replicate). The plants showed an initial content of Fe greater than Mn, but approximately after 250 min of exposure, Mn was higher than Fe. It was also observed during the 48 h that the increase of Mn content was much faster than for Fe, showing a preference of the plants for Mn uptake.
The observed results might be explained by the competitive inhibition reported in the literature (Malavolta et al., 1997;Alam et al., 2000), where Fe and Mn compete for the same transport sites due to their physical and chemical similarities, such as atomic masses, ionic radius and electronic structure (Madejczyk and Ballatori, 2012). IRT1 is a transporter protein member of the ZIP family, it is known as Fe 2+ transporter, however, assays carried out in yeast (Korshunova et al., 1999) and barley (Pedas et al., 2008) showed that this protein can also transport Mn by electrochemical active influx (Korshunova et al., 1999). Likewise, YSL is responsible to Fe uptake (Curie et al., 2001), but also translocates Mn complexed with nicotianamine (Andresen et al., 2018). Additionally, it was verified in rice that the transporter protein NRAMP5 also contributes to Mn and Fe transport, being another entry route for these micronutrients (Ishimaru et al., 2012). Table 3 presents the Fe and Mn absorption velocity expressed by counts min −1 . Therefore, it is possible to verify that the increasing content of Fe and Mn in the stem followed a linear function of the time. The slopes showed that absorption and transport of Mn was nearly 10-fold faster than for Fe. Manganese transport can occur by transporters or pumps, and these two mechanisms may coexist, while Fe uptake in beans only takes place trough transporters after Fe 2+ reduction (Andresen et al., 2018). The 9 × 10 −4 mol L −1 concentration is higher than commonly used in Hoagland's solution (Hoagland and Arnon, 1950), and the literature has shown that Mn toxicity decreased FIGURE 7 | µ-XRF monitoring the evolution of Colletotrichum truncatum on the surface of the soybean from the 3 rd to the 5 th day after the fungi inoculation. The images show pictures and chemical images for P, S, Compton scattering, K and Ca. The nutrient distribution pattern changes as the disease spreads. The unit of the scale the right side of the maps is counts per second (cps).
Frontiers in Plant Science | www.frontiersin.org  the translocation of absorbed Fe to shoots (Alam et al., 2000), and also the elevated availability of Mn induces Fe deficiency (Eroglu et al., 2016).

µ-XRF as Direct Determination of Pb in Eucalyptus Hybrid Leaf Cultivated in vitro
There is much expectation that fast growing plants might be used for bioremediation of soils (Couselo et al., 2012). Due to its high biomass rate production and stable organic tissue, these species may act as sinks extracting potentially toxic element from the soil. Several species were evaluated in phytoremediation studies, mostly from temperate climate as Populus (Di Lonardo et al., 2011), Brassica (Nehnevajova et al., 2007), and Salix (Lyyra et al., 2006) genus. The screening and selection of genotypes with potential for bioremediation can be carried out through in vitro studies (Nehnevajova et al., 2007;Doran, 2009;Di Lonardo et al., 2011). The determination of the spatial distribution of elements can show possible synergism or antagonism between the potentially toxic elements that one intends to remove from soil and the mineral nutrients in the plant (Bueno Guerra et al., 2013). Imaging techniques can be used to evaluate such interactions (Campos et al., 2015). The main techniques currently used are LA-ICP-MS (Tian et al., 2011), µ-XRF (Campos et al., 2015), synchrotron XRF (Hokura et al., 2006), and XRF microtomography (Terzano et al., 2008). Figure 9 shows that after 7 days of exposure to Pb(NO 3 ) 2 , Pb was absorbed and stored in the leaves. Fast absorption and translocation to the upper parts are important for the phytoremediation process since it reduces the bioremediation time (Nedelkoska and Doran, 2000). The highest intensities for K, Ca and Pb were found mainly at the midrib and leaf margins. Figure 10A indicates a weak positive correlation between Pb and K. Similar results with a higher correlation between Pb and S was founded in an accumulating ecotype (Tian et al., 2010). The Pb was also detected mostly within the vascular tissue, which suggested low mobility out of this region. This highlights the importance of K for Pb transport and toxicity control, in which the macronutrient is one of the mostly important element for osmotic cell wall and vacuole storage (Marschner, 1995). On the other hand, a strong positive correlation was found between K and Ca ( Figure 10B).
The XRF absorption factor was assessed for Pb in the Eucalyptus leaves. Table 4 presents the intensity of Pb Lα XRF peak for the irradiator (Pb disk) and irradiator covered by a leaf. The leaf blade and midrib attenuated 0.08 and 0.55% of the photons emitted by the disk, respectively. In most cases, attenuation values up to 10% can be neglected, since in this condition the concentration of the analyte and the XRF signal hold a linear relationship. Therefore, in the present situation the Eucalyptus leaf can be considered an infinitely thin sample, it means that the absorption correction (Ab correction ) factor is equal to one and the Pb concentration can determined directly from its XRF intensity and sensitivity, as shown in Equation 1. Figure 11 shows that regardless the approach chosen to transform the Pb Lα signal in concentration, one can clearly observe that the Pb storage is not homogenous along the leaf tissues. The accumulation was found mainly at the midrib and leaf margins, following the petiole direction toward the leaf tip.    Figure 11A expresses the Pb concentration in weight/area (µg cm −2 ) unity such as in Figure 6. Figure 11B presents an attempt of transforming the weigh/area concentration in a more common weight/weight concentration unity (mg kg −1 ). This step was accomplished by dividing the weight/area concentration by the average leaf surface density (g cm −2 ). However, Supplementary  Figure 7 and Table 4 show that the midrib is nearly two-fold thicker than the leaf blade. This fact makes the assumption of a homogenous surface density only an approximation. Aiming at accounting for this factor, Figure 11C presents the Pb concentration along the leaf sample considering the specific surface densities for midrib and leaf blade. Figure 11C shows that the Pb concentration varies from 4,880 to 13,400 mg Pb kg −1 fresh tissue weight. The median concentration of the Pb spots was 5,586 mg Pb kg −1 . These high concentrations values might be explained by the removal of the plant radicle. The stem was directly in contact with the spiked culture medium, which might have increased the phytoextraction potential. Anyway, it is curious to note that after 7 days of Pb exposure, the explants did not present any noticeable symptom of phytotoxicity. The recovery of the quantitative method was 108%.
One has to keep in mind that the concentration values supplied by analytical techniques such as atomic emission and absorption spectroscopies regards the average concentration of the element diluted in the whole sample tissue. It is likely that in these above-mentioned studies, Pb was not evenly distributed in leaves, but forming hotspots such as shown in the present paper.
This heterogeneous distribution in the leaf blade may suggest the presence tissue compartmentation of Pb. These maps can help to explain the mechanisms of tolerance presented by the Eucalyptus shoots (Nedelkoska and Doran, 2000), an important feature for maintaining growth, for the FIGURE 11 | µ-XRF quantitative map displaying the concentration of Pb in the leaf of Eucalyptus urophylla × E. grandis. Lead is not homogenously accumulated in the leaf tissue, panel (A) presents the Pb concentration in weight/area unity (mg Pb cm -2 leaf surface), panel (B) shows the Pb map in weight/weight (mg Pb kg -1 fresh tissue) unity considering a homogenous leaf surface density, and panel (C) shows the corrected map taking into account the difference between the thickness the leaf blade and midrib. Scale bar: 0.5 mm.

CONCLUSION
Benchtop µ-XRF is a versatile tool in plant science. It can be used to monitor the mineral elemental composition of a single point or produce lines and 2D images. As an elemental imaging tool, it is complementary to SEM-EDS and TEM-EDS, since it covers a larger sample area, from hundreds of µm 2 to cm 2 , and presents better limits of detection for most elements.
As a unique feature, µ-XRF allows for non-destructive analysis of fresh and living plant tissues bringing the possibility of dynamic and in vivo studies. Monitoring nutrient uptake kinetics while the plant is alive is possible using the technology described here. In biofortification or seed treatment effort, this tool allows to evaluate the spatial distribution of different elements and thus verifying the trueness of both, biofortification and seed treatments. In pathogen study, the chemical image allowed to verify changes in chemical distribution of elements such as K, P, S, and Ca in leaves infected. Bioremediation effort this technique allows to verify where is allocate the element and what concentration in plant.

AUTHOR CONTRIBUTIONS
ER organized and wrote most of the manuscript. ND and JC carried out the emission transmission quantitative analysis of Zn in primed seeds. MG and ER performed the in vivo characterization of P, S, K, and Ca distribution in injured soybean leaves. TdC monitored the root uptake and transport of Fe and Mn. ASN determined the content of Pb in the leaf of Eucalyptus hybrid. SS and EdA assisted all the experiments. HC designed the study, coordinated the work, supervised the students, and reviewed the whole manuscript.