Imaging trace element distributions in single organelles and subcellular features

The distributions of chemical elements within cells are of prime importance in a wide range of basic and applied biochemical research. An example is the role of the subcellular Zn distribution in Zn homeostasis in insulin producing pancreatic beta cells and the development of type 2 diabetes mellitus. We combined transmission electron microscopy with micro- and nano-synchrotron X-ray fluorescence to image unequivocally for the first time, to the best of our knowledge, the natural elemental distributions, including those of trace elements, in single organelles and other subcellular features. Detected elements include Cl, K, Ca, Co, Ni, Cu, Zn and Cd (which some cells were supplemented with). Cell samples were prepared by a technique that minimally affects the natural elemental concentrations and distributions, and without using fluorescent indicators. It could likely be applied to all cell types and provide new biochemical insights at the single organelle level not available from organelle population level studies.

. Complementary TEM and SXRF nanoprobe images of MIN 6 cell 5. The cell was grown in a medium not supplemented with Cd and it serves as a non-treated cell reference to treated-cell 4 (Fig. 2). As with cells 1-3 ( Fig. 1), the 350 nm thick cell 5 section was placed on a Au TEM finder grids coated with carbon and Formvar. (a) TEM image of the cell. The red rectangle outlines the area imaged with the SXRF nanoprobe. Scale bar = 1 m. (b) Enlargement of the high-resolution scan area in the TEM image a. (c -j) SXRF nanoprobe elemental distribution maps corresponding to b of, respectively, Cl, K, Ca, Co, Ni, Cu, Zn, Cd. All elemental spectra were collected simultaneously. With the exception of Cd, all elements were mapped by their K lines. Cadmium was mapped by its L lines. Scan step size = 0.05 m, dwell time = 4 s. One m scale bar (above b) applies to bj.
Spatial density scale bar with the measured range, in units of nmole/cm 2 , appears below each elemental map.
Contrast of the SXRF images (c-j) has been enhanced. Due to lack of a suitable Cd standard, there could be a systematic uncertainty of up to a factor of 2 in the inferred Cd concentrations (j, see Methods). Eedge of the cell, hheterochromatin, Nnucleus, arrows point to vesicles, dashed circle outlines a group of vesicles; all vesicles are potentially insulin producing. Additional concentration information, including uncertainties, appears in Table S3.
Three important observations emerge from the comparison between the elemental distribution maps in Figs. 2 and S1, and the additional information in Tables S2 (Fig. 2) and S3 ( Fig. S1): 1. With the exception of Cd, the spatial density ranges of each element are comparable in both figs. This indicates that supplementing cells with 1 mole/l CdCl 2 did not affect other elements' concentration ranges at the cellular level.
2. As expected, the maximum concentration of Cd in the non-supplemented cell 5 (Fig. S1j) is significantly lower than in the Cd-supplemented cell 4 (Fig. 2j). In addition, since Cd is a biologically non-essential element that is naturally present in cells at very low concentrations, cell 5's ultrastructure (Fig. S1b) cannot be identified in the Cd distribution map (Fig. S1j).
3. The small difference in each element's distribution map minimum values between cell 4 (    Due to enhanced contrast in the elemental maps in Figs. 1 and 2, minimum values (black pixels) represent a range of spatial densities in all figs., except in Fig. 1d. Similarly, in Figs. 2f, g, i, j, maximum values (red pixels) represent a range of spatial densities. In cases where only one min/max concentration value is reported (min for black pixels or max for red pixels), the black/red pixels correspond only to this value and not to a range of values.
Uncertainties are calculated from counting statistics. They do not include uncertainties due to background subtraction (which makes a small contribution to the total uncertainty of most elements in the table) and potential systematic uncertainties. Calculating uncertainty for spatial density = 0 is difficult. The uncertainties in these cases are calculated from the lowest measured non-zero spatial densities in the maps. This means that the reported values in these cases are lower limits on the uncertainties.
We did not have a suitable standard for normalizing Cd concentrations, measured in this study by the Cd L lines, to concentrations measured by K lines (all other elements reported here). As a consequence, there could be a systematic uncertainty of up to a factor of 2 in the calculated Cd concentrations (see Methods). Uncertainties are calculated from counting statistics. They do not include uncertainties due to background subtraction (which makes a small contribution to the total uncertainty of most elements above) and potential systematic uncertainties. Calculating uncertainty for spatial density = 0 is difficult. The uncertainties in these cases are calculated from the lowest measured non-zero spatial densities in the maps. This means that the reported values in these cases are lower limits on the uncertainties.
We did not have a suitable standard for normalizing Cd concentrations, measured in this study by the Cd L lines, to concentrations measured by K lines (all other elements reported here). As a consequence, there could be a systematic uncertainty of up to a factor of 2 in the calculated Cd concentrations (see Methods).