The purpose of this research was to evaluate different protocols for performing both elemental and lipid imaging on a single tissue section using LA-ICP-MSI and AP-MALDI-PPI-MSI. A variety of protocol combinations were evaluated on a single tissue section via different MSI modalities to explore the effects of matrix application, matrix washing, and the order of acquisition. These workflows are summarised in Fig. 1 and are discussed further below.

Effects of 2,5-DHA matrix coating on elemental imaging using LA-ICP-MSI

Whilst a MALDI-like matrix is not necessarily needed for ionisation via AP-MALDI-PPI, the absence of a matrix would result in low desorption efficiency of biomolecules from the tissue section using a 355-nm laser. Thus, a UV absorbing matrix is used for AP-MALDI-PPI to allow for broad and soft desorption of lipids from the tissue section into the inlet capillary, where they are subsequently ionised by the cold plasma source. It also provides direct relevance to conventional MALDI-MSI workflows. In contrast, LA-ICP-MSI uses short wavelength UV lasers and higher pulse energies (266 nm used here) and combined with no requirements for enabling analyte desorption with minimal fragmentation, a MALDI-like matrix is not required for LA-ICP-MSI. Therefore, to first investigate the influence of matrix coating on elemental imaging, one mouse brain section was analysed by LA-ICP in the absence of matrix (Fig. 1a, reflective of a conventional LA-ICP-MSI experiment) and two brain sections coated in 2,5-DHA matrix, which was previously shown to perform well for lipid imaging using AP-MALDI-PPI [27, 30]. The tissue section in Fig. 1b was matrix coated via an automated sprayer and the tissue section in Fig. 1c was coated using sublimation. The investigation of both spraying and sublimation covers the two main application methodologies used for MALDI-MSI studies. To date, the effect of MALDI-like matrices on the imaging of metals in biological tissues has not been investigated in detail. However, assessing whether the matrix application influences elemental distribution in biological tissues is of utmost importance for protocol optimisation of multimodal analysis as the workflow can be simplified and time saved removing one extra step between MSI analysis. Moreover, assessing the advantages of imaging elements with the matrix still coated on the tissue might prevent the need for matrix washing as this process poses the risk of analyte delocalisation.

A broad range of elements were targeted for this study, including the distribution of 23Na, 24 Mg, 31P, 39 K, 44Ca, 55Mn, 56Fe, 63Cu, and 66Zn. Figure 2 shows the distribution of 23Na, 31P, 55Mn, 56Fe, and 66Zn across the native tissue (top row), spray-coated sample (middle row), and sublimation-coated sample (bottom row). The remaining elemental images for slides A, B, and C can be found in Electronic Supplementary Material Fig. S1.

Fig. 2

Elemental analysis acquired via LA-ICP-MSI of mouse brain tissue sections under different conditions. Slide A was analysed without matrix deposition. Slide B was analysed with matrix deposition via an automated sprayer. Slide C was analysed with matrix deposition via sublimation. 55Mn, 56Fe, and 66Zn are in mg/kg, and 23Na and 31P are in counts/second (see “Experimental methods”)

The native tissue (slide A) in Fig. 2 shows a higher signal intensity for many elements than tissues coated with matrix (slides B and C), this being more visually pronounced for the distribution of 31P, 39 K, and 66Zn when quantifying using external calibration standards not coated with matrix. This could be rationalised due to the influence that the matrix plays on the optimum quantity of molecules/ions ablated for each scan, suppressing elemental yields. High concentrations of non-analyte isotopes have been shown to suppress analyte signals in ICP-MS due to changes in the plasma temperature, aerosol transport efficiency of the sample, and MS conditions [53, 54].

Elemental distributions throughout the mouse brain tissues in Fig. 2 yielded specific distributions for each element selected. Whilst 23Na and 55Mn appear with a similar localisation within the tissue, 56Fe and 66Zn exhibit a more distinct distribution. 56Fe is localised more on the third ventricle and caudoputamen, whilst 66Zn is found at higher concentrations in the CA1, CA2, and CA3 field of the hippocampus, molecular layer dentate gyrus along with more intense regions in the cortical subplate and olfactory areas. There are several hot-spots observed for 56Fe that colocalise between slides A–C as well as those shown in Fig. 3. The colocalisation suggests these are not technical artefacts. We hypothesise this may be due to residual blood or blood vessels present within the tissue. The 31P distribution shows a relatively homogeneous distribution for all three samples. The spatial distributions of elements herein outlined are consistent with previous reports [55,56,57].

Fig. 3

Molecular and elemental distributions of slides D and E. a Selected lipid (top row) and elemental (bottom row) distribution images of slide D that were analysed using spray-based matrix coating. b Selected lipids (top row) and elemental that were analysed using sublimation-based matrix coating. In both cases, AP-MALDI-PPI-MSI was performed first followed by LA-ICP-MSI. 55Mn, 56Fe, and 66Zn are in mg/kg and 23Na and 31P are in counts/second (see “Experimental methods”). c Butterfly plot of AP-MALDI-PPI averaged mass spectra for spray-coated (top, orange trace) and sublimation-coated (bottom, blue trace) tissue sections. d Averaged LA-ICP MS mass spectrum of spray-coated (left) and sublimation-coated (right) tissue sections. Only the elements labelled were quadrupole selected and detected during LA-ICP-MS imaging. 20 × and 40 × signify the magnification of the given elemental signal so it is visually perceivable in the plots for comparison

The most prominent difference in elemental imaging between the three workflows is observed for 66Zn distribution, where the enriched area in the stratum radiatum and stratum lacunosum-moleculare region observed for slide A shifts to the outer pyramidal cell layer in the hippocampus in the matrix-coated samples (slides B and C, Fig. 1). One hypothesis for this shift is redistribution caused by matrix application. Extensive studies have shown the impact that sample preparation has had on the disruption and depletion of Zn distributions [58,59,60,61]. Another explanation for the different 66Zn distributions is due to the different coronal regions of the mouse brain, with the tissue section in slide A being approximately 100 µm away from slides B and C but closer in proximity from slides F and G, where the distribution of slide A seems to more closely resemble the distribution observed for slides F and G. This similarity with slides F and G suggests the distribution of 66Zn shown in slide A is reflective of the brain regions shown. With sections, optical images of H&E-stained tissues post acquisition are provided in Electronic Supplementary Material Fig. S5.

Despite the effect of the matrix on LA-ICP-MSI signal intensities, the relative elemental distributions observed for slides B and C remain comparable for most of the investigated elements, with Zn as a notable exception, that might be explained by factoring in possible differences due to the bregma coordinates of each section. No major differences in elemental distributions are observed when comparing different matrix applications [62]. These results demonstrate that tissue coated in MALDI matrices is still amenable to LA-ICP-MSI analysis, a critical observation if one is to perform biomolecular imaging using MSI prior to LA-ICP-MSI analysis.

Evaluation of sequential lipid and elemental imaging using 2,5-DHA-coated tissues

Next, we investigated sequential lipid and elemental imaging on matrix-coated tissues using AP-MALDI-PPI-MSI followed by LA-ICP-MSI. Slides D and E (Fig. 1) were sprayed and sublimated-coated, respectively, and then analysed via AP-MALDI-PPI-MSI followed by LA-ICP-MSI.

Lipid distributions were investigated in positive ion mode via AP-MALDI-PPI, which resulted in the generation of predominantly protonated lipid signals. Consistent with earlier reports, the dominant signals following AP-MALDI-PPI of brain tissue were from hexosylceramides (HexCer) and phosphatidylethanolamine (PE) lipid species as well as demethylated phosphatidylcholine (PC) [27, 30]. A range of well-known lipid species classified as [PE 34:1 + H]+ at m/z 718.5372, [HexCer 42:2;2 + H]+ at m/z 810.6792, [PE 36:1 + H]+/[PC 34:1-CH3 + H]+ at m/z 744.5518, [PE 38:4 + H]+ at m/z 768.5514, and [PE O-36:5 + H]+ at m/z 724.5259 were selected for visualisation and are shown for slides D (spray coated) and E (coated via sublimation) in the top rows of Fig. 3a and b, respectively. The corresponding averaged mass spectra within the lipid mass range for the spray and sublimation-coated samples are shown in Fig. 3c.

Following AP-MALDI-PPI-MSI, elemental imaging via LA-ICP-MSI was performed on the matrix-coated tissues. The elemental images for 23Na, 31P, 55Mn, 56Fe, and 66Zn for spray and sublimation-coated slides are shown in the bottom rows of Fig. 3a and b, respectively. The corresponding averaged LA-ICP-MSI spectra are shown in Fig. 3d. Critically, elemental distributions for both coating methods following AP-MALDI-PPI-MSI are similar for slides A–C (Fig. 2), with the distribution of 66Zn also being similar to that measured for slides B and C shown in Fig. 2. The LA-ICP-MSI signal intensities were also similar between the two coating methods. Higher elemental signal intensities were observed for slides that underwent prior AP-MALDI-PPI-MSI analysis (Fig. 3a, b) compared to matrix-coated slides that did not (Fig. 2). This is believed to have taken place due to the desorption of the matrix layer performed under AP-MALDI-PPI-MSI analysis, which competes with the total amount of molecules/clusters ablated, consequently suppressing elemental yields. Critically, though, these data demonstrate the feasibility of performing sequential lipid and elemental MSI on the same tissue section, even when using a MALDI matrix to enable efficient desorption and ionisation of lipids.

Effect of matrix washing on sequential lipid and elemental imaging

Whilst the above results show that LA-ICP-MSI can be performed on matrix-coated samples following biomolecular imaging, in some cases, it may be desirable to remove the matrix prior to analysis. Thus, we next explored the effects of adding a matrix-washing step between AP-MALDI-PPI-MSI and LA-ICP-MSI acquisition. The tissue sections were coated with 2,5-DHA, either by spray (slide F) or sublimation (slide G), analysed via AP-MALDI-PPI-MSI, matrix washed by rinsing the slide with methanol, and then analysed via LA-ICP-MSI.

As expected, the overall lipid profile obtained using both matrix coating methods was similar (see spectra in Electronic Supplementary Material Fig. S3). The spatial distribution of several selected lipids, namely, [PE 34:1 + H]+, [HexCer 42:2;2 + H]+, [PE36:1 + H]+/[PC 34:1-CH3 + H]+, [PE 38:4 + H]+, and [PE O-36:5 + H]+ species also remains similar throughout the biological tissue for both approaches (Fig. 4a and b), and also for data shown in Fig. 3.

Fig. 4

Molecular and elemental distributions of slides F and G. a Molecular and elemental distribution images of slide F that was coated using spray-based matrix deposition, analysed by AP-MALDI-PPI-MSI, matrix washed, and then analysed via LA-ICP-MSI. b Molecular and elemental distribution images of slide G that was matrix coated via sublimation, analysed via AP-MALDI-PPI-MSI, matrix washed, and then analysed via LA-ICP-MSI. Ion images are visualised with a 10 ppm tolerance, total ion current normalisation, and 99% quantile hot-spot removal. 55Mn, 56Fe, and 66Zn are in mg/kg and 23Na and.31P are in counts/second (see “Experimental methods”)

Interestingly, following matrix removal, the intensity and distribution images of 55Mn, 56Fe, and 66Zn for both datasets more closely resemble those obtained without matrix removal and from the native tissue sections shown in Figs. 2 and 3. These relatively water-insoluble elements are tightly bound into various proteins within the brain, the most relevant of which are: Mn is found in glutamine synthetase, an essential enzyme in astrocytes [63]; Fe is incorporated into neurons, microglia, and astrocytes via binding to transferrin [64]; and Zn is bound to metallothioneins with 3 of the 4 isoforms being expressed exclusively in the CNS, primarily in astrocytes [65].

In contrast, from the spray-coated slide, the spatial trace for 23Na, 31P (Fig. 4a), and 39 K (Electronic Supplementary Material Fig. S4) reveals distribution artefacts on the left side of the tissue on slide F. This likely arises due to the washing causing delocalisation of some elements, or biomolecules containing those elements. Element mobility after aqueous washing has been observed previously with lighter, water-soluble group I and II elements such as 23Na and 39 K either being leeched out of or redistributed within mouse brain samples [66]. The observation that some metals, like 23Na and 39 K, are delocalised upon solvent treatment whilst others such as 55Mn are not is also consistent with other studies investigating the influence of formalin fixation and paraffin embedding on elemental images obtained with LA-ICP-MS [67]. Interestingly, the sublimation-coated slide does not display the same elemental artefacts (Fig. 4b and S4), possibly suggesting that elemental delocalisation due to washing can be highly variable.

Despite delocalisation negatively affecting image quality, the 23Na and 31P signal intensities, after washing the matrix off for both coating modalities, remain relatively similar when compared to the protocol that does not include a washing step on slides D and E (Fig. 3), although it should be noted that delocalisation on scales smaller than the pixel used employed here could be present. In contrast, the lateral distribution images of 66Zn for slides F and G show an increase in intensity when adding a washing step compared to slides D and E. This increase in intensity for 66Zn distribution might have arisen due to the removal of the matrix that is believed to suppress and deplete the zinc signal as observed in Fig. 2, where the intensity for Zn was at its highest without matrix application. And the elemental distribution for 55Mn and 56Fe remains consistent for both protocols.

These results further demonstrate the possibilities for performing both lipid and elemental MSI sequentially on the same tissue section and that matrix washing prior to LA-ICP-MSI still preserves the distributions of many elements. However, analysis of matrix coated without washing appears to be most favourable for obtaining good-quality LA-ICP-MSI on the broadest range of elements. It should be noted that the exact nature of any delocalisation (e.g. the extent of delocalisation and the elements it impacts) will depend on the protocol deployed for washing.

Performing LA-ICP-MSI prior to AP-MALDI-PPI-MSI

Finally, we investigated the possibility of performing LA-ICP-MSI before lipid imaging by AP-MALDI-PPI-MSI. Standard LA-ICP-MSI conditions (i.e. those used to acquire data in Figs. 2, 3, and 4) completely ablated the tissue, as would be expected with the continuous raster approach and requirements for complete abaltion, and were therefore unsuitable for subsequent lipid imaging (refer to post-MSI stained tissue sections in Electronic Supplementary Material Fig. S5). Therefore, for this investigation, the laser power for LA-ICP-MSI was reduced to reduce the amount of tissue ablation. This was indeed confirmed by post-MSI staining and optical imaging (Electronic Supplementary Material Fig. S5). Following LA-ICP analysis, the tissue section was coated in 2,5-DHA via sublimation and then analysed by AP-MALDI-PPI-MSI. The results of this analysis are shown in Fig. 5. The top row of Fig. 5a shows the elemental images of 23Na, 31P, 55Mn, 56Fe, and 66Zn and the bottom row the ion distribution of [PE 34:1 + H]+, [HexCer 42:2;2 + H]+, [PE 36:1 + H]+/[PC 34:1-CH3 + H]+, [PE 38:4 + H]+, and [PE O-36:5 + H]+ species.

Fig. 5

Spatial and spectral data of elemental and molecular MSI workflow outlined on slide H. a Elemental (top row) and molecular (bottom row) distribution images of slide H acquired via LA-ICP-MSI analysis then matrix coated using sublimation and analysed by AP-MALDI-PPI-MSI. b Averaged mass spectra for slide H outlining the five selected ions shown on a. 55Mn, 56Fe, and 66Zn are in mg/kg and 23Na and.31P are in counts/second (see “Experimental methods”)

AP-MALDI-PPI-MSI following LA-ICP-MSI at reduced laser energy revealed similar ion images for lipids (Fig. 5a, top row) compared to the workflows discussed above where AP-MALDI-PPI-MSI was acquired prior to elemental analysis. Figure 5b shows the averaged mass spectra for the relevant lipid mass range for slide H (full mass range spectra are provided in Electronic Supplementary Material Fig. S6). Not only was the signal intensity similar to that acquired when lipid imaging was performed first but no notable evidence of significantly increased lipid fragmentation caused by irradiation with the LA-ICP laser was observed, thereby demonstrating the capabilities of profiling lipids on biological tissues post-laser ablation.

Although the lipid profiling acquired for this acquisition sequence was similar to earlier workflows, the elemental distribution in Fig. 5a shows notable differences in terms of both overall signal intensity and distribution. For example, distributions of 55Mn and 66Zn were overall similar to those shown in Fig. 2. However, for some, overall signal intensity and image contrast are reduced. The ion images of 31P and 23Na show a different preferential localisation. The distribution of 31P and 23 Na is relatively homogeneously distributed for the workflows shown in Figs. 2, 3, and 4. However, under lower laser energy conditions, both show a significant reduction in signal throughout the tissue, with signals preferentially localised more to the pyramidal cell layer in the hippocampus and in the third ventricle. Even though, the 56Fe distribution shows certain similarities, such as intensities being concentrated in the third ventricle for both workflows (Figs. 5 and 2), the high predominance of intensity in the hippocampus in Fig. 5 is not observed in the native tissue (Fig. 2). A possible rationalisation could be related to the only partial, and region-dependent, ablation efficiencies across the tissue under these reduced and non-ideal ablation laser energies (see post-MSI optical image of slide H in Electronic Supplementary Material Figure S5) [68]. Whilst this workflow demonstrates the feasibility of lipid imaging acquisition after LA-ICP, the above results support the idea that it is beneficial to perform lipid imaging prior to elemental imaging using LA-ICP.

Correlation analysis of endogenous elements and lipids

The ability to acquire elemental and lipid distributions from the same tissue section led us to investigate the spatial correlations between different elements and selected lipid species. For this, we used the workflow outlined for slide E (Figs. 1 and 3) where the tissue section was coated with 2,5-DHA via sublimation and then analysed sequentially by AP-MALDI-PPI and LA-ICP. We note that this is in contrast to previous studies that have performed co-registration of MALDI and LA-ICP-MS data using adjacent tissue sections [41].

The heatmap in Fig. 6 shows the correlation between a variety of lipid species representing classes well-detected by AP-MALDI-PPI, including ceramides, HexCer, and PE lipids, and the elements analysed in this study: 23Na, 24 Mg, 31P, 39 K, 44Ca, 55Mn, 56Fe, 63Cu, and 66Zn (note red indicates high spatial correlation). A table with their respective correlation numbers can be found in Electronic Supplementary Material Table S2.

Fig. 6

Heatmap of the spatial correlation (Pearson’s correlation) between molecular and elemental mass spectrometry imaging of workflow outlined on slide E. Hexosylceramide (HexCer) is used for glucosyl and galactosyl moieties, ceramides (Cer), phosphatidylethanolamines (PE), lysophosphatidylethanolamine (LPE), and phosphatidylserine (PS)

The resulting correlation heatmap shows a general correlation between lipids and the elemental images for traces of 24 Mg, 55Mn, 23Na, 39 K, and 31P. 56Fe was found to be the element with the lowest correlation for all lipid species across the board. The correlation between iron and the investigated lipid species could be explained due to the low intensity of 56Fe trace arising more from dorsal blood vessels and ventricles. The lipid species that presented the highest resemblance with the spatial distribution for the elements herein analysed is identified as [PE 36:1 + H]+/[PC 34:1-CH3 + 2H]+. The co-registration between this lipid ion at m/z 746.5672 and the element trace of 24 Mg is found to be the highest correlation across the heatmap showing a value of 0.82. [PE 36:1 + H]+/[PC 34:1-CH3 + 2H]+ and 24 Mg traces are distributed homogeneously throughout the mouse brain tissue with a highlight in the pyramidal cell layer of the hippocampus (Electronic Supplementary Material Fig. S7).

Figure 7a and b illustrates the spatial co-registration of the ions at m/z 792.5538 and m/z 810.6817 with 23Na, 24 Mg, 31P, 39 K, 44Ca, 55Mn, 56Fe, 63Cu, and 66Zn. The ion at m/z 792.5538 assigned as [PE 40:6 + H]+ is the highest intensity signal and is distributed within the gray matter with more pronounced intensities arising in the cortex. The elements that show the highest correlation with the distribution observed for [PE 40:6 + H]+ are 24 Mg and 23Na with a correlation coefficient value of 0.647 and 0.638, respectively. Whilst 63Cu and 66Zn show the lowest correlation values after 56Fe (0.066). The ion distribution shown in Fig. 7b at m/z 810.6817, assigned as [HexCer 42:2;2 + H]+, is observed to arise from the myelin-rich white matter. This species presents low correlation values for all elements analysed with the highest value being 0.489 for 31P trace. Both molecular distributions arising from [PE 40:6 + H]+ and [HexCer 42:2;2 + H]+ are in agreement with previous reports [27].

Fig. 7

AP-MALDI-PPI and LA-ICP image co-registration. a Image overlay of [PE 40:6 + H]+ (magenta) with 23Na, 24 Mg, 31P, 39 K, 44Ca, 55Mn, 56Fe, 63Cu, and 66Zn (cyan) and their respective correlation coefficient. b Image overlay of [HexCer 42:2;2 + H]+ (red) with 23Na, 24 Mg, 31P, 39 K, 44Ca, 55Mn, 56Fe, 63Cu, and 66Zn (blue) and their respective correlation coefficient