High-resolution complementary chemical imaging of bio-elements in Caenorhabditis elegans †

Here, we present a sub- l m multimodal approach to image essential elements in Caenorhabditiselegans . Acombinationof chemical imaging technologies reveals total metal concentration, chemical state and the protein to which an element is associated. This application of distinct yet complementary chemical imaging techniques provided unique insight into essential and trace elements at the subcellular level. Caenorhabditis elegans is a model system that elemental the sample to

Caenorhabditis elegans is a model system that displays highly compartmentalised elemental distribution ranging from high abundance species to ultra-trace elements. When imaging these elements, abundance within the sample does not necessarily equate to high sensitivity or the capacity for high spatial resolution. 1 Application of synchrotron-based X-ray fluorescence microscopy (XFM) is no exception; a number of factors determine whether an analyte can be detected and spatially mapped at subcellular resolution. These factors relate to both technical limitations and the nature of the sample itself. Considerations for imaging elements in C. elegans include: the energy of the incident beam, duration of exposure, the atomic mass of the element, the energy of the fluorescence, the characteristics of the detector used, the composition and thickness of the sample, and the environmental conditions in which the analysis takes place. 2 Fluorescence profiles of specific elements must be discerned from elastic (Rayleigh) and inelastic (Compton) scattering; 3 i.e. element-specific fluorescence must not be obscured by scatter peak tails. The incident X-ray energy determines which excitation events can occur, but also positions the scattering peaks within the X-ray spectrum. Lower mass (Z) elements have reduced cross-sections, fluorescent yields and are easily absorbed by the sample matrix, all limiting sensitivity. While self-absorption may be negligible for heavier elements in thinner specimens, it places real limitations on low-Z elements, even within single cells. 4,5 In addition, for specimens measured in air, argon (Ar) fluorescence causes a major interfering peak in the collected spectra (with associated tail) that can overwhelm the signal derived from lighter bio-elements, making their detection impossible. 6 Previously, within whole animals biologically important mid-Z elements, such as calcium (Ca), zinc (Zn) and redox-active metals have often been analysed without corresponding data on low-Z elements (e.g. phosphorus (P) and sulfur (S)).
C. elegans have been a successful test bed for pushing the boundaries of microscopy, 7,8 and are particularly well suited for whole-organism imaging of fundamental biochemistry. Examples span from Raman vibrational spectroscopy for imaging lipid metabolism 9 to scanning electron microscopy for profiling the C. elegans connectome. 10,11 We have used this nematode to extensively study metal metabolism via XFM, conducting population studies, 12 tomography 13 and X-ray absorption nearedge structure (XANES) spectroscopy 14 to appraise the complex biochemistry of metals in vivo. C. elegans are highly resistant to ionising radiation, 15 which permits analysis of hydrated and anesthetised samples for mid-Z elements using hard X-rays (410 keV).
While hydrated imaging of anesthetised samples is preferred in principle, the water content increases the absorption of low energy fluorescence, and therefore dehydration aids detection of elements with atomic masses below potassium (K; Z o 19). Preserving subcellular distribution of elements is challenging, particularly following chemical fixation. Even brief (o30 s) formalin fixation of thin tissue sections can cause redistribution and leaching of transition metals and electrolytes. 16 However, we previously demonstrated that cryofixation of C. elegans in liquid N 2 -cooled propane followed by lyophilisation does not cause significant variation in elemental content or subcellular distribution. 12 Evolving detector capabilities, such as those of the 384-channel Maia reduces sampling overheads by collecting spectra 'on-thefly' (i.e. continuously during a transit across the specimen). 18 The reduced overheads allow scanning of larger samples including whole C. elegans in a practical timeframe. Furthermore, this enables spatial oversampling, where data is collected at intervals less than the full width at half maximum (FWHM) of the beam profile. Although the modulation transfer function (MTF) supports the utility of sampling at half of the beam size, in most practical cases this oversampling is not performed due to detector time limitations. Recent upgrades to the Maia detector have improved low-energy sensitivity, thus mitigating this limitation.
Here, we have combined the Maia detector (Rev C), a N 2 environment for reduced Ar fluorescence, sample dehydration and incident energy selections to simultaneously image a broad range of endogenous biological elements from P to strontium (Sr) in wild type C. elegans. Specimens were mapped via XFM using 12.9 keV and 18.5 keV incident beams. Fluorescence emission was collected by the Maia detector mounted in the backscatter geometry (see ESI † for Experimental Methods). We present highly spatially resolved and correlated images of low-Z elements previously considered as 'difficult' analytes with respect to XFM in biology, whilst also using different modes of imaging to demonstrate the high complementarity possible using a unified imaging approach. 1 Mapping biological samples in a frozen hydrated state can reduce incident beam 'damage', permitting radiation exposure of up to 10 10 Gy, allowing the extended dwell time necessary for the collection of multiple spectra as in XANES analysis 19 or oversampling. However, cryogenic conditions can be prone to artifacts 4 and retain problematic water content. 20 At room temperature, cryofixed and lyophilised Vicia faba (fava bean) chromosomes showed no morphological change at radiation doses of up to 10 7 Gy, 21 suggesting cryogenic measurements may not always be necessary. Our dehydrated samples received a combined radiation dose of B10 6 Gy with no apparent morphological changes in Compton scattering used to identify microstructural features, 22 or in known elemental distributions previously described. 12 Derived Compton maps were used to relate C. elegans anatomy to all other measured elements in the hyperspectral image stack ( Fig. 1a; full resolution maps available as ESI †). Sulfur was ubiquitously distributed and best recapitulated the majority of anatomical structures (Fig. 1b). Phosphorus also reproduced structural definition of the specimen, though was found at comparatively lower concentrations in the head and full extent of the tail. Consistent with expectations, both elements showed high concentrations within embryos known to be rich in lipids, phospholipids and yolk proteins required for development.
As we have previously shown, Ca and manganese (Mn) were highly compartmentalised along the intestinal lumen, 13 as was Zn, in addition to being rich within the gonad and embryos, consistent with Zn finger transcription factors necessary for early development. Using the higher incident energy of 18.5 keV, we also found subcellular concentrations of Sr in the most anterior intestine. Strontium commonly substitutes for Ca in biological systems at a greatly reduced concentration; 23 higher sensitivity of XFM is needed to map to quantify distribution further along the intestinal tract.
Due to minimised time penalties from oversampling, the spatial resolution achieved approached that of light microscopy (approximately 200 nm 24 ), allowing application of statistical approaches to objectively determine if true co-localisation was occurring between specific elements that are both low-Z and highly mobile in the previously hydrated specimen. Potassium and chlorine (presumably as K + and Cl À ) appeared to have highly correlated cytoplasmic distribution (Fig. 2a). Pearson and Mander's correlation measures of the merged images showed strong association (Pearson's r = 0.708; Mander's R = 0.926). Using Li's method for intensity correlation analysis (ICA), 25 which overcomes several limitations of both Pearson and Mander's overlay comparisons, we determined the ICA quotient (Q) for the entire organism was 0.234 (+0.5 = perfect correlation; À0.5 = no correlation). Examination of both merged elements and the mapped product of difference from the mean (PDM) used to calculate the overall ICA Q showed distinct regions of K-enrichment within the gonad, as well as marked positive correlation in embryos. Interrogation of the pixel (concentration) histograms for each element showed a bimodal distribution indicative of highly enriched K in the gonad. Frequency distribution and PDM versus signal intensity (areal concentration) plots further demonstrated a skew towards areas of high K.
Li's ICA Q measure is particularly useful for visualising the degree of spatial co-localisation; to demonstrate this we assessed the correlation between Ca and Sr, which within the whole organism was less distinct due to low Sr concentration (ICA Q = 0.185). In the anterior intestinal cells where Sr was detectable, we observed high correlation between Ca and Sr in the resulting PDM image (Fig. 2b), consistent with intestinal co-localisation observed in other taxa. 26 These results illustrate that PDM imaging allows both spatial correlation at a sub mm level of detail, as well as the within the whole organism. The XFM methods used here significantly improved spatial resolution for in vivo mapping. Previously we have shown a distribution of iron (Fe) about the intestine at approximately 2 mm resolution. 12 Our sub-mm imaging approach permitted assessment of Fe revealing a level of detail comparable to histological staining and light microscopy. We found that punctate Fe deposits (Fig. 3a) resembled Fe distributions in formalinfixed, paraffin-embedded sections stained using the Perls method for non-heme Fe (Fig. 3b). 27 In addition to localised Fe deposits, XFM mapping also showed a more generalised Fe distribution not seen in Perls staining, indicative of heme. The correspondence between these two diverse imaging modalities suggests that both approaches accurately report on in vivo Fe. Although Perls staining is not quantitative, these results differ from those reported by Hackett et al., 16 who suggested formalin fixation alone alters Fe distribution in biological tissue. We suggest that neither radiation damage from XFM, nor extensive chemical processing for histological staining necessarily disturbs the distribution of non-heme Fe in C. elegans.
To further explore consistency between complementary imaging methods, we examined the in vivo localisation of the dominant Fe storage protein ferritin. 28 Using a green fluorescent protein (GFP) fusion to ferritin we compared high-resolution confocal fluorescence in vivo microscopy to XFM mapping and Perls staining ( Fig. 3c and d). Distribution of GFP fluorescence, and thereby ferritin localisation, was again remarkably similar to the Fe puncta previously imaged (see Movie, ESI †). Ferritin accounts for almost half of the Fe content of C. elegans and therefore represents a good proxy for non-heme Fe. The multiple imaging methods used serve as validation of each respective technique, providing the first consistent representation of subcellular Fe within a whole organism. Future directions of this complementary imaging approach could employ C. elegans with mutated genes that affect Fe metabolism, as well as ageing studies (such as those described in James et al. 14 ) to exploit the higher resolution mapping protocol described here. Here we focused on Fe to demonstrate complementary imaging of metal distribution; similar studies could employ the genetically encoded fluorescent calcium sensor GCaMP, 29 which has been used in the C. elegans model system, 30 to compare total body Ca concentration with cell-specific Ca 2+ content.
In summary, we have demonstrated sub-mm XFM mapping of bio-elements, both rare and ubiquitous, ranging from low-Z to highly abundant transition metals in a model organism ideally suited for studying metal metabolism. Mapping low-Z elements with confidence will facilitate new experimental paradigms. Fig. 2 (a) High abundance and low-Z elements K and Cl were analyzed for co-localisation in an individual C. elegans specimen. Images were merged and underwent whole organism correlation analysis using Pearson's, Mander's and Li's intensity correlation analysis (ICA; inset table). 25 Visualisation of the product of difference from the mean (PDM; presented on a black background) at high resolution improved interpretation by presenting the ICA quotient (Q) on a pixel-by-pixel basis. Using this method, both elements showed marked correlation in embryos (0 4 Q 4 +0.5; white arrowheads), as well as distinct potassium-rich regions within the distal and proximal gonad (À0.5 o Q o 0; white arrows). Scale bar = 50 mm. Histograms of pixel values revealed a bimodal distribution for K, consistent with K enrichment in the gonad (black arrow). (b) The advantages of visualising correlation are clear when comparing co-localisation of high abundance Ca with low abundance Sr, which shares similar biochemistry but is close to the XFM limit of detection. Co-localisation is less robust (lower Pearson's r, Mander's R and ICA Q) across the whole organism (inset table); though in the anterior intestine where both Ca and Sr are most concentrated shows high spatial correlation (0 o Q o +0.5).