Characterization of metabolic compartmentalization in the liver using spatially resolved metabolomics

Cells adapt their metabolism to physiological stimuli, and metabolic heterogeneity exists between cell types, within tissues, and subcellular compartments. The liver plays an essential role in maintaining whole-body metabolic homeostasis and is structurally defined by metabolic zones. These zones are well-understood on the transcriptomic level, but have not been comprehensively characterized on the metabolomic level. Mass spectrometry imaging (MSI) can be used to map hundreds of metabolites directly from a tissue section, offering an important advance to investigate metabolic heterogeneity in tissues compared to extraction-based metabolomics methods that analyze tissue metabolite profiles in bulk. We established a workflow for the preparation of tissue specimens for matrix-assisted laser desorption/ionization (MALDI) MSI and achieved broad coverage of central carbon, nucleotide, and lipid metabolism pathways. We used this approach to visualize the effect of nutrient stress and excess on liver metabolism. Our data revealed a highly organized metabolic compartmentalization in livers, which becomes disrupted under nutrient stress conditions. Fasting caused changes in glucose metabolism and increased the levels of fatty acids in the circulation. In contrast, a prolonged high-fat diet (HFD) caused lipid accumulation within liver tissues with clear zonal patterns. Fatty livers had higher levels of purine and pentose phosphate related metabolites, which generates reducing equivalents to counteract oxidative stress. This MALDI MSI approach allowed the visualization of liver metabolic compartmentalization at high resolution and can be applied more broadly to yield new insights into metabolic heterogeneity in vivo.

nutrient abundance and restriction (9). In a satiated state, hepatocytes oxidize glucose to generate 23 energy and synthesize fatty acids (10). Fatty acids are then esterified into triacylglycerols (TAGs) 24 and transported to the adipose tissue for storage. In fasted conditions, the adipose tissue releases 25 fatty acids for oxidation by the liver to yield ketone bodies that can fuel distant organs (11).

26
Additionally, the liver performs glycogenolysis and gluconeogenesis to restore circulating glucose 27 levels upon fasting. In contrast, upon prolonged nutrient excess conditions, the liver acts as an 28 overflow depot for lipids when the endocrine and storage functions of the adipose tissue become 29 compromised (12). With rising rates of obesity, nonalcoholic fatty liver disease (NAFLD) is an 30 increasing cause of morbidity and mortality.

32
Despite the liver's central role in metabolic homeostasis, liver metabolism is characterized mostly 33 on the gene, protein, and signaling levels. However, as hepatocytes make up over 80% of liver

128
Imported peaks were moved to the local max using the mean spectra with a minimal interval width 129 of 5 mDa. Peaks were then normalized to total ion current (TIC), except for desiccation 130 experiments, as these two datasets were acquired from separate slides and runs. Ion images for 131 metabolites of interest were generated based on peak lists containing theoretical m/z and ppm

Results
Broad coverage of small metabolites with spatial resolution metabolism in situ to visualize regions of metabolism in the liver. As major concerns are residual 217 enzyme activity and non-enzymatic breakdown of labile metabolites, we evaluated whether enzyme 218 inactivation through desiccation or heat inactivation treatment would stabilize tissue metabolites for 219 MALDI MSI sample preparation. We compared procedures of storing cryosectioned tissue on slides 220 at -80 °C and thawing them in a vacuum desiccator to minimize rehydration due to condensation 221 (treatmentF, Fig. 1A) with desiccation immediately after tissue sectioning before storage 222 (treatmentDF). To assess tissue integrity, serial sections of liver were H&E stained for histological 223 analysis immediately after sectioning to evaluate the effects of freezing and desiccation. Minimal 224 differences were observed for gross tissue morphology (Fig. 1B). We used ATP stability as an

266
The tissue distribution of ATP, ADP, and AMP after heat treatment showed a relatively stable 267 distribution upon treatmentHF as compared to control tissues, but a loss of overall ATP levels upon 268 treatmentFHF ( Supplementary Fig. 2C). MSI ion images we used to visualize the relative spatial 269 distribution of metabolite intensities, but these images do not inform the total metabolite pools 270 unless the MS signal is calibrated for each metabolite. Thus, we excised representative tissue 271 slices from the same tissues that were analyzed using MSI and determined total metabolite levels 272 using LC-MS ( Supplementary Fig. 2D). As ATP use by enzymes will lead to increased levels of 273 AMP and ADP, successful heat stabilization of enzymes should lead to stable levels of ATP, ADP, 274 and AMP. This comparison showed that although heat stabilization of fresh tissue seemed to 275 maintain the spatial distribution of adenosine phosphate metabolites seen in control tissues, 276 absolute levels of ATP decreased due to thermal destabilization.

278
To visualize how heat treatment affected the abundance of all detected ions in an unbiased manner,

279
we constructed an ion segmentation map using bisecting k-means clustering on all identified 280 spectra. This showed that conductive heating applied with a commercial device (Denator,

281
Gothenburg, Sweden) set to optimze heat delivery based on frozen or fresh states and with 282 consideration to specimen dimensions led to an overall loss of spatial localization of metabolites 283 (Supplementary Fig. 2C and 2E). Since the whole tissues were processed, the heating profiles 284 needed to be optimized to provide uniform heating throughout the tissue; however, this was not 285 possible due to the tissue's thickness. Although regional clusters of metabolites in heat-treated 286 brains were largely maintained, they could not be accurately mapped to anatomical brain regions 287 due to the loss of tissue morphology. Together, these results indicate that the heat treatment 288 applied to the whole tissues prior to sectioning led to disruption of tissue structure and compromised 289 the integrity of anatomical regions. Further optimization of the heating profile for the denaturation 290 of enzymes and the preservation of metabolites is needed for uniform stabilization that would be 291 compatible with spatial metabolomics workflows.

293
Using the treatmentDF followed by MALDI MSI, several additional metabolites were detected. The 294 central carbon metabolism correlates with key energetic and biosynthetic pathways, including 295 glycolysis and the pentose phosphate pathway (PPP) ( Figure 1D). As expected, hexoses were 296 highly abundant within the vasculature of the tissue, whereas intracellular metabolites generated 297 from glucose were enriched in extravascular compartments rather than in the vasculature.

298
Together, these results suggest that optimized MALDI MSI sample preparation and data acquisition 299 workflow achieve broad coverage of small metabolites to generate reproducible spatial profiles of 300 biologically relevant metabolic pathways.

301
Distinct spatially-resolved metabolic signatures were observed in fed and fasted livers.

302
Regions of metabolism were investigated in the liver in response to fasting by generating spatially-303 resolved metabolic profiles. Livers from fasted mice showed marked histological differences in 304 hepatocyte shape due to the expected depletion of glycogen ( Fig. 2A). We evaluated whether 305 tissue metabolomes remained stable during cryosectioning, as this is a lengthy process for 306 experiments with multiple biological replicates that need to be mounted onto the same slide for data 307 acquisition. No significant difference was observed between total ATP, ADP, or AMP levels, 308 indicating that these labile metabolites remained stable during cryosectioning (Fig. 2B) Figure 3A). To visualize these differences in an unbiased manner, we constructed a segmentation map (Fig. 3D). This visualization showed distinct metabolic clusters within different anatomical regions of the liver and between control and fasted mice, while all biological replicates within each group clustered together ( Supplementary Fig. 3C). Metabolite clusters were observed 319 for the vasculature, hepatocytes, bile ducts, and the common bile duct. These clusters 320 corresponded with co-registered ion images of heme B, a cofactor of hemoglobin that is enriched

339
Furthermore, using spiked internal standards into the MALDI matrix, we observed that independent 340 mouse cohorts and replicates within each treatment group were highly reproducible in this study 341 (Supplementary Fig. 3A-C). Together, these results indicate that our workflow yielded consistent 342 and highly reproducible results to visualize metabolic compartments in the liver.

343
Visualization of fasted liver metabolism shows disruption of metabolism and fuel switching.

344
The liver acts as a metabolic rheostat to maintain whole-body energy homeostasis in times of 345 nutrient stress and excess. As MSI adds a spatial dimension to metabolomic analyses, we 346 dissected the metabolic compartmentalization in the fasting liver. We identified metabolic 347 differences using the unbiased UMAP approach, which showed separation of data clusters from  Fig. 4C). To explore 352 differences between liver and systemic metabolism, we extracted metabolite spectra from MSI data 353 on a pixel-by-pixel basis. As shown in Figure 3A, we used the segmentation map (Fig. 2D,

354
Supplementary Fig. 5A) to select regions-of-interest enriched for hepatocytes (extravascular tissue) 355 or heme B (intravascular tissue, circulating metabolites). In accordance with our previous 356 observations, ATP was significantly decreased and AMP significantly increased in extravascular 357 tissue upon fasting (Fig. 3B, left).

358
We also observed an increase in the fatty acid docosahexaenoic acid (DHA). This was 359 recapitulated in the UMAP distributions, where AMP and DHA were more abundant in fasted mice 360 ( Supplementary Fig. 4D). The metabolite profiles from intravascular regions did not show 361 differences in adenosine phosphate metabolites, but several fatty acids were significantly enriched 362 in the circulation upon fasting (Fig. 3B, right) whereas they were not significantly changed within distant organs, which is corroborated by these results and indicates that spatially-resolved 366 metabolomics can inform on metabolic compartmentalization within tissues.

394
Indeed, pathway analysis of the intravascular regions showed that several lipid metabolic pathways 395 were enriched (Fig. 3D). Interestingly, comparing the spatial distribution of fatty acids showed that 396 the abundance of DHA and ARA follow a specific and compartmentalization pattern in fed livers 397 (Fig. 3E, Supplementary Fig. 5B). DHA is a 22-carbon polyunsaturated omega-3 fatty acid (22:6), 398 whereas arachidonic acid (ARA) is a 20-carbon polyunsaturated omega-6 fatty acid (20:4). Both 399 can be synthesized from alpha-linolenic acid, which in turn is produced from the essential fatty acid 400 linoleic acid. These fatty acids can also be released from complex lipids through lipolysis. Relative 401 quantification of the metabolite intensity as a function of the distance between blood vessels 402 confirmed that DHA is enriched in proximity to the vasculature while ARA displayed the opposite 403 enrichment pattern (Fig. 3F, Supplementary Fig. 5D). Upon fasting, this distinct spatial

411
We also investigated how the liver's response to nutrient stress might contrast to its response to 412 nutrient excess by subjecting mice to a high-fat diet (HFD). Livers of HFD mice showed marked droplets, is associated with advanced fatty liver disease, inflammation, fibrosis, and poor clinical outcomes (24,25). We evaluated the changes in metabolite levels, and subsequent pathway analysis showed that several metabolic pathways were significantly enriched upon HFD feeding, including the pentose phosphate pathway and purine metabolism (Fig. 4B). Cells increase PPP decision-making (30). Advances in instrumentation and application have produced increased 468 molecular complexity and spatial resolution analyses leading to new insights into metabolic function 469 and heterogeneity at the single-cell scale (31,32). With increasing sensitivity and specificity in ion metabolic fidelity of the tissue during sample processing is essential to yielding meaningful 473 analyses, especially in comparison with chromatography-based mass spectrometry approaches 474 where metabolomes are stabilized by quenching steps and samples are maintained at low 475 temperatures until analysis while several sample preparation steps for MALDI MSI occur at ambient 476 conditions. Here, we demonstrate an approach to prepare tissue samples for MSI that minimizes 477 conversion or breakdown of labile metabolites while broadening the range of small metabolites 478 detected to more broadly cover metabolic pathways and yield new insights into tissue metabolism.

502
In contrast to fasting conditions, where glycolytic metabolism were low, fatty livers displayed higher 503 levels of glycolytic and PPP metabolites. Together with the marked increase in GSH levels,

504
indicates levels of oxidative stress, which is constant with high levels of NADPH levels. Additionally,

505
we observed an increase in purine metabolism, which may produce nucleotides needed to repair 506 DNA damage, generate essential energy carriers, or provide precursors for metabolic cofactors 507 such as NAD, which can all become disturbed by cellular redox stress. These results indicate that 508 although the lipid content of the liver increases upon HFD feeding, the lipid droplets act as an 509 overflow depot rather than being effectively metabolized by the liver to dissipate excess energy.