Methods to visualize elements in plants

56 Understanding the distribution of elements in plants is important for researchers across a 57 broad range of fields, including plant molecular biology, agronomy, plant physiology, plant 58 nutrition, and ionomics. However, it is often challenging to evaluate the applicability of the 59 wide range of techniques available, with each having their own strengths and limitations. 60 Here, we compare scanning/transmission electron microscopy-based energy-dispersive X-ray 61 spectroscopy (SEM-EDS, TEM-EDS), X-ray fluorescence microscopy (XFM), particle- 62 induced X-ray emission (microPIXE), laser ablation ICP-MS (LA-ICP-MS), nanoscale 63 secondary ion mass spectroscopy (NanoSIMS), autoradiography, and confocal microscopy 64 with fluorophores. For these various techniques, we compare their accessibility, their ability 65 to analyze hydrated tissues (without sample preparation) and suitability for in vivo analyses, 66 as well as examining their most important analytical merits such as resolution, sensitivity, 67 depth of analysis, and the range of elements that can be analyzed. We hope that this 68 information will assist other researchers to select, access, and evaluate the approach that is 69 most useful in their particular research program or application. 70

Introduction concentrations were measured in vivo for up to 24 h (Doolette et al., 2018). Although some 145 translocation of the foliar-applied Zn was observed, it was found that the Zn had only limited 146 mobility regardless of the form of Zn applied. Similar results were also reported by Tian et al. 147 (2015). In a similar manner, autoradiography has been used to examine the translocation of 148 foliar-applied Zn over time in vivo in whole plants of wheat (Read et al., 2019). These authors 149 found that the use of 65 Zn-labelled compounds allowed for time-resolved analyses of Zn 150 distribution in live plants, reporting that 65 Zn was translocated throughout the plant (including 151 to the grain, where it is important for human nutrition) following its foliar application. 152 153 Improving human nutrition through foodstuffs 154 To improve human nutrition through higher quality foods, an understanding of nutrient 155 concentration and distribution within foodstuffs is essential. This is because the nutritive 156 value of foods depends not only upon the total elemental concentration, but also its 157 distribution and molecular speciation. Accordingly, biofortification strategies need to consider 158 both distribution and speciation of nutrients within the foodstuff tissue. As an example of a 159 study aiming to improve human nutrition, grains of buckwheat (Fagopyrum esculentum) were 160 examined using microPIXE (Pongrac et al., 2011). These authors found that the inner layers 161 of the pericarp were enriched in K, Mn, Ca, and Fe while the outer layer was enriched in Na, 162 Mg, P, S, and Al, and that by altering the milling approach it was possible to alter the 163 nutritional content of the grain. Furthermore, for both wheat and rice (Oryza sativa) grain, it is 164 known that whilst micronutrients tend to accumulate in the bran layers (i.e. the aleurone, the 165 tegument, and the pericarp), those elements which are generally more mobile within the 166 phloem (such as K, Mg, P, Fe, Zn, and Cu) tend to accumulate to higher concentrations in the 167 aleurone layer (de Brier et al., 2015). In this regard, Wang et al. (2011) used LA-ICP-MS to 168 examine the stable Zn isotope 70 Zn in wheat grain, finding that there are two barriers to Zn 169 transport in wheat grain: between the stem tissue rachis and the grain, and between the 170 maternal and filial tissues in the grain. 171 172 Not only is distribution important in influencing the nutritional value of foods, but it is also 173 necessary to understand how the co-localization of different elements within foods impacts on 174 nutrient availability to humans. For example, co-localization of micronutrients with P (often 175 present as phytate) likely reduces micronutrient availability in the human gut, with such co-176 localization observed in sweetcorn and maize (Zea mays) (Cheah et al., 2019) and wheat 177

Understanding toxic elements in plants and tolerance mechanisms 180
Understanding the behavior of toxic elements in plants, their impact on plant growth, their 181 translocation through the plant and accumulation in human foodstuffs, and the mechanisms 182 that plants use to tolerate these toxicants is of critical importance. First, to illustrate the 183 importance of understanding elemental distribution in crop plants, consider the problem of Al 184 toxicity. Soluble concentrations of Al are elevated in the acid soils that comprise ca. 3.95 185 billion ha of the global ice-free land (Eswaran et al., 1997). Although Al is highly toxic to 186 plant root growth, much remains unknown about how it exerts its toxic effects. In this regard, 187 NanoSIMS has been used to examine Al distribution in root tissues of soybean (Glycine max), 188 finding that Al accumulates almost entirely in the walls of cells in the rhizodermis and outer 189 cortex (Kopittke et al., 2015). These authors reported that this Al in the cell wall of young, 190 elongating roots was toxic and caused a rapid reduction in root elongation. Interestingly, in 191 tea (Camellia sinensis), a known accumulator of Al, much of the Al accumulated in the cell 192 walls of the leaves, representing a potential tolerance mechanism (Tolrà et al., 2011). As 193 another example, the accumulation of As in foods is of interest due to the consumption of this 194 carcinogen by humans. The distribution of As in roots of Arabidopsis was examined using 195 XFM, confirming the localization of a new arsenate reductase (HAC) which limits As 196 accumulation in the tissues (Chao et al., 2014). 197 198 Another area of major research interest has been in the use of imaging techniques to 199 understand how hyperaccumulating plants are able to tolerate high concentrations of 200 metal(loid)s in their tissues. Nickel hyperaccumulator plants (which make up the majority of 201 hyperaccumulator plants known globally) have been the most intensively studied (Reeves et 202 al., 2018). In most species studied to date, Ni is concentrated in the epidermal cell vacuoles of 203 the leaves (Küpper et al., 2001;Bhatia et al., 2004;Kachenko et al., 2008;van der Ent et al., 204 2017). Hyperaccumulation spans several length-scales, from whole plants down to organs, 205 tissues, individual cells, cellular organelles, and transporter molecules, and information at all 206 of these scales is important for understanding the mechanisms associated with 207 hyperaccumulation (van der Ent et al., 2017). 208 209 X-ray fluorescence-based approaches for visualization 210 With X-ray fluorescence-based approaches, elements are detected based upon their 211 characteristic fluorescent X-rays. These fluorescent X-ray are generated by passing the 212 specimen through a focused beam of high-energy X-rays (XFM), electrons  based EDS) or protons (PIXE). This beam excites a range of different elements (depending on 214 the energy of the incident X-rays, electrons, or protons) which are detected and quantified by 215 a detector to determine elemental concentrations in the specimen. The movement of the 216 specimen through the incident beam in x-y creates a raster map in which each point represents 217 a pixel with concentration data (or relative element intensity) for a range of elements. High-218 energy X-rays (> 15 keV) have great penetrative power, and will pass through plant 219 specimens (both sectioned tissues, and potentially even through entire, intact plant tissues), 220 whereas electrons and protons will only penetrate 5-50 µm into a specimen. In principle, the 221 incident X-ray beam does not destroy the sample, hence the method is typically considered 222 'non-destructive'. However, as X-rays are ionizing radiation, and hence depending on the 223 energy and dwell on the sample, damage might occur due to the formation of free radicals 224 which are highly reactive and damaging to the tissue being analyzed. Furthermore, the 225 incident X-ray beams do not generate heat in the specimen in contrast to electron and proton 226 beams, which consist of particles and have a far greater potential to damage the specimen 227 during scanning. Obtaining sufficient element sensitivity while keeping dwell low enough not 228 to cause beam-induced damage can be challenging in PIXE (Laird et al., 2019). Given that most SEM-and TEM-based EDS systems operate under a high vacuum, the plant 237 tissue specimen must be totally dehydrated (and coated with carbon to make it conductive for 238 electrons) prior to analysis ( Figure 2). However, where a cryo-SEM is available, it is possible 239 to examine frozen plant tissue specimens in the hydrated state. Appropriate specimen 240 preparation for cryo-SEM and cryotransfer remains extremely challenging technically, and 241 detection limits are poorer than for dehydrated specimens. In addition, it is also increasingly 242 possible to analyze living plants using environmental SEM (ESEMs), although there are 243 issues with sample size restrictions and beam damage (Danilatos, 1981;McGregor and 244 Donald, 2010). In the majority of studies, specimen dehydration (typically by freeze-drying or 245 lyophilization) is required, and this has the potential to cause artefacts due to elemental 246 redistribution (see van der Ent et al. (2018b) for a full discussion of considerations). 247 248 When imaging a specimen, SEM can typically achieve a resolution of ca. 1-50 nm. However, 249 when examining elemental composition using SEM-based EDS, the resolution is considerably 250 poorer due to the interaction of the electrons with the sample, typically being in the order of 251 2-5 µm and worsening with increasing accelerating voltage. For TEM-based EDS, the 252 resolution is better than for SEM-based EDS because the use of TEM requires the plant 253 tissues to be cut as ultrathin sections (ca. 60-100 nm in thickness), thereby greatly reducing 254 problems associated with the depth of penetration. Thus, for TEM-based EDS, it is possible to 255 achieve a resolution of ca. 100 nm. For XFM, given the penetrating nature of the X-rays (both the incident X-rays as well as the 307 fluorescent X-rays), elemental distribution can often be examined throughout the entire 308 thickness of plant tissues. However, the depth of analysis varies greatly depending upon the 309 element of interest, being determined by the energy of the corresponding fluorescent X-rays 310 (with these having a lower energy than the incident X-rays). This is best illustrated with the 311 following examples. For Ca, with a K-edge emission line of 3.7 keV, 50 % of the 312 fluorescence will be absorbed in a plant sample ca. 70 µm thick and 90 % in a sample ca. 200 313 µm thick. In contrast, for Se, with a K-edge emission line of 11.2 keV, 50 % of the 314 fluorescence will be absorbed in a sample ca. 2000 µm thick and 90% in a sample ca. 6500 315 µm thick. In other words, assuming a root with a thickness of 1000 µm, only the Ca in the 316 surface 100-200 µm can be detected (with the Ca in the vascular tissue being 'invisible') 317 while Se will be detected across the entire depth of the root cylinder. Thus, great care needs to 318 be taken when comparing the distribution of various elements, especially in thicker samples. 319 320 Most synchrotron-based XFM facilities tend to have a resolution in the order of 20 nm to 1 321 µm (Li et al., 2019b). The time required to conduct analyses also varies greatly. For 322 synchrotron-based systems with fast detector systems, the dwell is now routinely ≤ 1 ms, 323 meaning that a 1-megapixel image can be collected in ca. ≤ 17 min. The elements which can 324 be examined depend upon a wide range of factors. Often, elements can be accessed from P 325 (2.1 keV) to Ag (25 keV), while higher Z elements can potentially be examined using the L-326 edges. The detection limit varies widely depending upon the facility as well as the element 327 being analysed. For elements such as Mn, Fe, and Zn, the detection limit is excellent, being in 328 the order of ca. 1 mg/kg or even lower. However, the detection limit decreases for the lower Z 329 elements, including P, S, and K, often being ca. 10-1000 mg/kg, which is a function of the 330 smaller X-ray cross-sections (resulting in lower fluorescence yields) and operation of most 331 XFM beamline in air which absorbs low-energy X-rays. For synchrotron-based XFM, 332 analyses are potentially fully quantitative, for example using the GeoPIXE software package 333 which produces quantitative self-absorption corrected maps which are line overlap-resolved 334 and in which the background is subtracted (Ryan, 2000). constraints of availability and financial considerations) as required by experimental needs. In 383 addition, many laboratory-based systems provide vacuum and helium purge capabilities that 384 might not be available at synchrotron-based beamlines, thereby offering improved capability 385 for measurement of light elements such as Al, Si, S, and P. 386 387 However, these laboratory-based systems offer worse spatial resolution, often in the range of 388 5-50 µm (compared to 20 nm to 1 µm for synchrotron-based systems). In addition, laboratory-389 based systems typically use a concave focussed polychromatic X-ray source with 390 Bremsstrahlung background, with this having important differences to a monochromatic, 391 highly parallel X-ray source in synchrotron-based XFM. For example, there is no energy 392 tunability in laboratory-based systems, and hence X-ray absorption spectroscopy not possible. 393 Finally, the substantially lower X-ray flux for laboratory-based systems (typically 1000-to 394 10,000-times less bright) results in longer dwell times (50 to 100 ms per pixel) compared to 395 synchrotron-based systems (0.5 to 5 ms per pixel). Based on this discussion, it is clear that microPIXE is useful for examining elements in a wide 417 range of systems. Given that it can analyze a wider range of elements than many other 418 approaches (such as XFM), it is especially valuable for examining light elements (such as Na, 419 Mg, or Al) as well as heavier elements (such as Cd or the rare earth elements, as shown in 420 Figure 5A) which often cannot be analyzed with other approaches, all with a good detection 421 limit (Figure 1 and Figure 5A). Some recent studies using microPIXE include for the analysis 422 of Zn and Cd in a hyperaccumulator, Sedum plumbizincicola (Hu et al., 2015), and for the 423 study of Ni hyperaccumulation in Phyllanthus balgooyi (Mesjasz-Przybylowicz et al., 2016). 424 425

Mass spectrometry-based approaches for visualization 426
For mass spectrometry-based approaches, small portions of the sample are progressively 427 removed during scanning for analysis. Because analysis is by mass spectrometry, not only are 428 these generally highly sensitive techniques, but they also allow isotopic analyses. However, 429 given that small portions of the sample are removed for analysis during scanning, these mass 430 spectrometry based approaches are considered destructive.   (Salt et 457 al., 2008;Klug et al., 2011), and protocols for sample preparations which produce intact, dry 458 samples with unaltered ion distribution within tissue are available (Persson et al., 2016a). 459 460 LA-ICP-MS is therefore particularly useful in studies where high sensitivity is required with 461 access to an extremely broad range of elements with good sensitivity (Figure 1  NanoSIMS operates in an ultra-high vacuum, meaning that samples must first be dehydrated 489 before analysis. Furthermore, NanoSIMS requires a flat surface, and hence it is typically only 490 possible to examine sectioned tissues ( Figure 6). As for other techniques where sample 491 dehydration is required prior to analysis, extreme care must be taken to ensure that the method 492 used for sample processing does not cause experimental artefacts through redistribution of the 493 elements of interest. 494 495 NanoSIMS offers an excellent lateral resolution, with analyses routinely conducted at 496 resolutions as low as 100 nm (Figure 6). Using this technique, it is possible to analyse a very 497 wide range of elements of relevance to plant studies, from H to U. The sensitivity is also very 498 good, with the detection limit being low mg/kg range. The sensitivity for any given element 499 depends upon the primary beam selected, with either an Obeam or a Cs + beam available. The 500 negatively charged primary beam (i.e. O -) tends to favour the production of positively charged 501 secondary ions, while the positively charged primary beam (i.e. Cs + ) tends to favour the 502 production of negatively charged secondary ions. As a result, for elements such as Na, Mg, 503 Al, K, Ca, Mn, Fe, and Zn, the Obeam is generally preferred (Nuñez et al., 2017). In 504 contrast, for elements such as Si, P, S, Cl, As, and Se, the Cs + beam is generally preferred. 505

506
The main advantages of NanoSIMS are the excellent detection limit and spatial resolution, as 507 well as the wide range of elements that can be analyzed. As a result, this approach is 508 particularly suited to examining the sub-cellular distribution of elements within cross-sections 509 of plant tissues (Figure 1 and Figure 6). Isotopic analyses are also possible using this Autoradiography (Figure 1 and Figure 7A) is the oldest of the techniques discussed here, 520 having been used for plants since the 1920s (Hevesy, 1923). In autoradiography, radioactive 521 isotopes are supplied to a plant, which are taken up and redistributed throughout the plant 522 tissues. To then examine their distribution in the plant, an image is obtained of the decay 523 emissions from the various plant tissues ( Figure 7A). These decay emissions can be detected 524 using an X-ray film, or more recently, using digital autoradiography. 525 526 www.plantphysiol.org on May 5, 2020 -Published by Downloaded from Copyright © 2020 American Society of Plant Biologists. All rights reserved.
Compared to some other approaches, such as XFM or NanoSIMS, access to autoradiography 527 facilities is likely not too difficult. This approach also has a range of other advantages, 528 including being able to examine hydrated plant tissues, including for in vivo analyses. 529 Furthermore, autoradiography can be used to examine large samples, or even entire plants. 530 Another major advantage is the ability to separate background isotopes of an element (i.e. 531 those natively present in the plant tissues) from the radio-isotope of the same element added 532 exogenously. The resolution achievable with autoradiography varies between ca. 25-1000 µm 533 ( Figure 1) and depends upon a range of factors (Zhang et al., 2008). 534

535
The greatest challenge for using autoradiography are the highly restrictive and complicated 536 health and safety regulations in many jurisdictions for working with radio-isotopes. In 537 addition to this limitation, it is only possible to examine a single element at a time (Solon et 538 al., 2010 The main uses of autoradiography are for in vivo studies or studies in which sample 543 processing needs to be avoided. It also offers excellent detection limits and allows separation 544 of background elements from those added exogenously as radioisotopesthis being critical 545 when only a portion of the total element is of interest (Figure 1 and Figure 7A The final approach considered here is the use of laser confocal microscopy with element-552 selective fluorophores (Figure 1 and Figure 7B). 553 554 Given that many researchers are likely able to access laser confocal microscopy without 555 substantial difficulty, this approach is one of the easier ones in terms of facility access, as 556 most research institutions will have laser confocal microscopes, especially in medical 557 faculties where they are routinely used. This approach is also non-destructive, and can also be 558 used on hydrated samples, including for in vivo analyses of living plants. The maximum 559 resolution is similar to some other approaches, being ca. 1 µm. 560 561 www.plantphysiol.org on May 5, 2020 -Published by Downloaded from Copyright © 2020 American Society of Plant Biologists. All rights reserved.
All the approaches considered above have analyzed elemental composition directly. However, 562 laser confocal microscopy relies on the binding of ion-selective fluorophores to the element of 563 interest for their subsequent detection using excitation by specific wavelengths emitted by 564 lasers ( Figure 7B). This in itself represents a potential limitation of this technique, as 565 fluorophores will generally only bind to free ions not already bound strongly to other ligands 566 in the plant. Thus, the proportion of the total pool identified using the fluorophore can be 567 uncertain. In addition, issues with penetration into the plant tissue are largely unknown and 568 are hard to quantify. The range of elements that can be investigated using laser confocal 569 microscopy is entirely dependent on the commercial availability of fluorophores with specific 570 affinities for elements of interest. These include, but are not limited to: Zn (Zinpyr-1, FluoZin, 571 TSQ), Ni/Co (Newport Green), Cu (Phen Green) and Pb/Cd (Leadmium Green). 572

573
The use of laser confocal microscopy with element-selective fluorophores is particularly 574 suited to cases where in vivo analyses are required for an element where a suitable 575 fluorophore exists, with an excellent detection limit (Figure 1 and Figure 7B). However, 576 questions still remain regarding the binding of the fluorophores and their penetration into the 577 plant tissue. Recent studies using confocal microscopy with fluorophores include imaging the 578 distribution Ni 2+ with the dye Newport Green in Alyssum murale (Agrawal et al., 2013) andA. 579 lesbiacum (Ingle et al., 2008), the use of Zinpyr-1 for imaging the distribution of Zn 2+ in 580 Noccaea caerulescens (Kozhevnikova et al., 2017;Dinh et al., 2018) and Arabidopsis 581 (Sinclair et al., 2007), and Leadmium Green for imaging Zn 2+ and Cd 2+ in Sedum alfredii and 582 Picris divaricata (Lu et al., 2008;Hu et al., 2012). 583 584

Concluding remarks 585
Understanding the distribution of elements within plant tissues is critical for a range of 586 research programs within plant science, including for functional characterization in molecular 587 biology, improving plant nutrition and productivity, improving human nutrition, and 588 understanding toxic elements in plants and tolerance mechanisms. For analyzing plants, a 589 range of techniques are suitable but it can often be confusing as to which approach is best 590 given their range of advantages and limitations. It is clear from this review that there is no 591 single technique which is best. Rather, each of the techniques have their own strengths and 592 weaknesses. By comparing the accessibility, ability to analyze hydrated tissues (without 593 sample preparation) and conduct in vivo analyses, as well as comparing the resolution, 594 access the approach that is most useful for use in their particular research program. In 597 addition, it will be helpful to use correlative approaches in which the same sample is 598 examined with multiple techniques to exploit the advantages listed here. Of central 599 importance in the future will be the analyses of living plants (including in vivo analyses) with 600 minimal sample preparation at excellent resolution and with good detection limits across the 601 wide range of physiologically-relevant elementsthis requiring a strong correlative approach 602 (see Outstanding Questions). The use of such correlative approaches will enable important 603 research questions to be answered within the field of plant science. 1 Figure 1. Comparison of seven broad techniques used for examining element distribution in plants. All values are indicative. The color of the shading indicates a ranking: green is a potential advantage of the technique, red is a potential disadvantage, orange is neither an advantage nor disadvantage, and white is not ranked. Figure 3. Use of synchrotron-based X-ray fluorescence microscopy (XFM, Australian Synchrotron) for high-throughput screening of plant mutant libraries for arabidopsis. The image in (A) is an optical micrograph. The images show the distribution of (B) Fe, (C) Mn, (D) Zn, and (E) Se in ca. 6000 seeds, with each image having a resolution of ca. 20 megapixels when displayed at full resolution. The image in (F) shows a small portion of a detailed scan for Fe showing some seeds differing in their Fe concentration and distribution. The 'overview scans' (B-E) had 10 µm pixel size with a dwell of 1 ms per pixel, while the detailed scans [a small portion shown in (F)] had 1 µm pixel size with a dwell of 7 ms per pixel. In total, an estimated 40,000 seeds were examined, with only ca. 6000 seeds shown here. Note that the analyses are non-destructive.
www.plantphysiol.org on May 5, 2020 -Published by Downloaded from Copyright © 2020 American Society of Plant Biologists. All rights reserved.