The dentate gyrus differentially metabolizes glucose and alternative fuels during rest and stimulation

Abstract

The metabolic demands of neuronal activity are both temporally and spatially dynamic, and neurons are particularly sensitive to disruptions in fuel and oxygen supply. Glucose is considered an obligate fuel for supporting brain metabolism. Although alternative fuels are often available, the extent of their contribution to central carbon metabolism remains debated. Differential fuel metabolism likely depends on cell type, location, and activity state, complicating its study. While biosensors provide excellent spatial and temporal information, they are limited to observations of only a few metabolites. On the other hand, mass spectrometry is rich in chemical information, but traditionally relies on cell culture or homogenized tissue samples. Here, we use mass spectrometry imaging (MALDI-MSI) to focus on fuel metabolism of the dentate granule cell (DGC) layer in murine hippocampal slices. Using stable isotopes, we explore labeling dynamics at baseline, as well as in response to brief stimulation or fuel competition. We find that at rest, glucose is the predominant fuel metabolized through glycolysis, with little to no measurable contribution from glycerol or fructose. However, lactate/pyruvate, β-hydroxybutyrate (βHB), octanoate, and glutamine can contribute to TCA metabolism to varying degrees. In response to brief depolarization with 50 mM KCl, glucose metabolism was preferentially increased relative to the metabolism of alternative fuels. With increased supply of alternative fuels, both lactate/pyruvate and βHB can outcompete glucose for TCA cycle entry. While lactate/pyruvate modestly reduced glucose contribution to glycolysis, βHB caused little change in glycolysis. This approach achieves broad metabolite coverage from a spatially defined region of physiological tissue, in which metabolic states are rapidly preserved following experimental manipulation. Using this powerful methodology, we investigated metabolism within the dentate gyrus not only at rest, but also in response to the energetic demand of activation, and in states of fuel competition.

Graphical Abstract

Stable isotope tracing in acute hippocampal slices followed by mass spectrometry imaging reveals DGC layer metabolism of alternative fuels, including lactate/pyruvate and β-hydroxybutyrate (βHB), into the TCA cycle under baseline conditions. Brief stimulation with 50 mM KCl causes a rapid decrease in DGC layer ATP, and a preferential increase in glucose metabolism relative to alternative fuel metabolism. Glucose preference occurs even under conditions of fuel competition. This work achieves broad metabolite coverage from a spatially defined region of physiological tissue, in which metabolic states are rapidly preserved following experimental manipulation.

Introduction:

Neurons require energy in the form of ATP to maintain ion gradients for action potential conduction, to transport cargo throughout the cell, and to reuptake, recycle, and repackage neurotransmitters, among many other cellular activities (Harris, Jolivet, & Attwell, 2012). Because these functions rely on a constant supply of ATP, neurons are particularly sensitive to energy disruptions, such as low oxygen or low blood sugar conditions found during stroke or severe diabetic hypoglycemia (Mariman, Lorca, Biancardi, Burgos, & Álvarez-Ruf, 2022; Rehni & Dave, 2018). While important, maintaining a consistent energy supply throughout the brain is a complicated task. Neuronal activity fluctuates rapidly both spatially and temporally, there is relatively little local fuel reserve for times of energy demand, and increased oxygen and nutrient supply can only happen reactively to demand via neurovascular coupling.

The brain relies on a constant macronutrient supply from the blood or from limited local generation of certain alternative fuels. Glucose is the primary carbon source for brain metabolism, but many other nutrients are available and the possibility of their metabolism by neurons is increasingly appreciated. Lactate is considered an end-product of glycolysis, but is also a metabolic fuel, and can act as an energy reservoir that is shared between organs (Hui et al., 2020, 2017; TeSlaa et al., 2021). On the cellular scale, an astrocyte-neuron lactate shuttle has been proposed in which astrocytes release lactate in times of energetic demand to fuel neurons (Bélanger, Allaman, & Magistretti, 2011). While this shuttle may occur during rest or in specific conditions (Ide, Schmalbruch, Quistorff, Horn, & Secher, 2000; Murphy-Royal et al., 2020), accumulating evidence now suggests that upon stimulation, neurons increase their use of glucose rather than importing and metabolizing lactate (Bak et al., 2009; Diaz-Garcia et al., 2017; Dienel, 2012, 2019; Lange et al., 2015; Li et al., 2023; Lundgaard et al., 2015; Rae, Nasrallah, & Bröer, 2009).

Another important fuel group for brain function are the ketone bodies, which can be produced either by the liver or locally by astrocytes (Guzman & Blazquez, 2004). Interest in neuronal ketolysis is particularly spurred by the finding that ketogenic diets can reduce seizure frequency in pharmacoresistant epilepsy and may also be beneficial in neurodegenerative disorders where glucose metabolism is reduced (Cunnane et al., 2016; Jensen, Wodschow, Nilsson, & Rungby, 2020; Ułamek-Kozioł, Czuczwar, Pluta, & Januszewski, 2019). Multiple studies support the direct metabolism of ketone bodies by neurons, which may help explain their beneficial functions (Achanta, Rowlands, Thomas, Housley, & Rae, 2017; Andersen, Westi, Neal, Aldana, & Borges, 2022; Edmond, Robbins, Bergstrom, Cole, & de Vellis, 1987). These findings are supported by the high affinity of neuronally-enriched monocarboxylate transporter 2 (MCT2) for ketone bodies (Broer et al., 1999), compared to the astrocytic MCT4 (Achanta & Rae, 2017; Dimmer, Friedrich, Lang, Deitmer, & Broer, 2000; Rafiki, Boulland, Halestrap, Ottersen, & Bergersen, 2003). Previous in vivo studies in rat brain estimated a 62% contribution of β-hydroxybutyrate to neuronal TCA cycle metabolism – a level sufficient to meet basal energy demands, but not to replace glucose in fueling increased activity during wakefulness (Chowdhury, Jiang, Rothman, & Behar, 2014). In patients observed during fasting, ketone bodies were able to replace glucose to support neurometabolism (Owen et al., 1967).

The contribution of lipid metabolism via β-oxidation has historically been disregarded in the brain, as it was assumed that there is little uptake of metabolizable lipid species across the blood-brain barrier, and that ATP production by β-oxidation is too slow and requires too much oxygen for the anoxia-sensitive neurons (Schönfeld & Reiser, 2013). Several early in vivo reports demonstrate ¹⁴C-palmitate uptake into the brain from circulation, although conclude most of the carbon is processed into stable lipid and protein compartments (Kimes, Sweeney, London, & Rapoport, 1983). While palmitate carbons are mainly incorporated into phospholipids, amino acids (particularly glutamate and glutamine) are substantially and rapidly labeled and may represent palmitate entry to the TCA cycle. Palmitate oxidation by the brain may be underestimated because of unlabeled oxidation in the liver (J. C. Miller, Gnaedinger, & Rapoport, 1987). Early in vitro work on cell-type specific metabolism found that while neurons and oligodendrocytes did not metabolize the medium chain fatty acid, octanoate, it was substantially oxidized by astrocytes, as measured by ¹⁴CO₂ release (Edmond et al., 1987). Recent studies are reexamining this question and have also found that lipid metabolism does contribute to brain energy, although astrocytes appear to be the major cellular compartment for β-oxidation (Andersen et al., 2022; Ioannou et al., 2019; Jernberg, Bowman, Wolfgang, & Scafidi, 2017).

Amino acids can also serve as additional fuel for neurometabolism, particularly glutamate, which enters the TCA cycle through αKG. As glutamate is the major excitatory neurotransmitter in the brain, its intracellular and extracellular concentrations, production, and metabolism are tightly regulated. Released glutamate is taken up by astrocytes and converted into glutamine before being returned to neurons and reconverted to glutamate in a process known as the glutamate-glutamine cycle. The formation of new glutamate or oxidation of existing pools may be a mechanism of regulating neuronal excitation (Andersen, Markussen, et al., 2021; Divakaruni et al., 2017; Schousboe, Scafidi, Bak, Waagepetersen, & McKenna, 2014).

Finally, glycogen is one of the few local energy reserves within the brain. Most glycogen is stored within astrocytes, and these glycogen stores are mobilized in response to energetic demand. It has been suggested that astrocytic metabolism of glycogen may spare glucose for neuronal use (Rothman et al., 2022). A smaller pool of neuronal glycogen may also be used directly by neurons (Duran, Gruart, López-Ramos, Delgado-García, & Guinovart, 2019; Saez et al., 2014).

Several discrepancies and inconsistencies in the field have arisen, likely due to both the methodological challenges of studying brain metabolism and the specific conditions being studied. In terms of methodology, mass spectrometry (often coupled with gas or liquid chromatography, GC- or LC-MS) provides rich chemical information, but usually relies on in vitro conditions or homogenized tissue samples. This compromises the detailed metabolic states and physiological interactions between cells. On the other hand, biosensor imaging in vivo or within acute brain slices is a powerful tool to investigate single cells within their native environment, and an increasing number of metabolic sensors are becoming available (Koveal, Díaz-García, & Yellen, 2020). However, sensors are currently available for only a minority of metabolites, and these studies are still limited to only a handful of sensors at a time.

A powerful technique recently applied to the field of metabolomics is mass spectrometry imaging (MSI), which combines the chemical resolution of mass spectrometry with the spatial advantages of imaging (Schwaiger-Haber et al., 2023; G. Wang et al., 2022; L. Wang et al., 2022) and can also monitor incorporation of ¹³C from supplied fuels into downstream metabolites. We use MALDI-MSI of acute hippocampal slices to define the metabolic activity of the hippocampal dentate granule cell (DGC) layer in resting conditions, following a brief chemical stimulation with KCl, and during fuel competition. Making use of stable isotope tracers and rapid thermal preservation, we determine the metabolic incorporation of glucose, lactate and pyruvate (lac/pyr), β-hydroxybutyrate (βHB), octanoate, and glutamine. We find that at rest, the DGC layer readily metabolized glucose through glycolysis, while there was little or no detectable contribution from glycerol or fructose. However, glucose as well as lac/pyr, βHB, octanoate, and glutamine contributed to the TCA cycle to varying degrees. Upon KCl stimulation, glucose metabolism through glycolysis and into the TCA cycle was preferentially increased relative to the alternative fuels. Finally, when either lac/pyr or βHB were supplied at increasing concentrations, they outcompete glucose for entry into the TCA cycle. While lac/pyr competition caused a modest decrease in glucose labeling of glycolytic intermediates, this effect was not observed with increasing βHB levels, suggesting different metabolic results depending on the alternative fuel provided. This work utilized the methodological advantages of MSI, stable isotope tracing, and rapid thermal preservation to measure metabolomics within the dentate gyrus of acute brain slices. In addition to reporting metabolic preferences at baseline, we also describe the metabolic response to acute stimulation and to competition with alternative fuels.

Methods:

Mice

Wildtype C57BL/6N mice were originally obtained from Charles River (RRID: IMSR_CRL:027), then bred in-house. Female and male mice between 4-6 weeks of age were included in this study. Animals were housed in ventilated cages within a barrier facility, which maintained 12 hr light/dark cycle, regulated cage temperature (24°C) and humidity (53%) and provided ad libitum access to water and food (Picolab Rodent Diet 5053). All experiments were performed in compliance with the NIH Guide for the Care and Use of Laboratory Animals and Animal Welfare Act. Specific protocols were approved by the Harvard Medical Area Standing Committee on Animals (institutional animal welfare assurance number: D16-00270 (A3431-01); protocol number: IS00001113-6).

Acute hippocampal slice preparation

To prepare acute hippocampal slices, mice were deeply anesthetized by open-drop isoflurane inhalation (300 μL/1L container volume until unresponsive to toe pinch), decapitated, and the brain was immediately extracted. After removal of the cerebellum and prefrontal cortex, the remaining tissue was submerged in ice-cold slicing solution: 87 mM NaCl, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 25 mM NaHCO₃, 7 mM MgCl₂, 0.5 mM CaCl₂, 25 mM D-glucose, and 75 mM sucrose (osmolality ~310 mmol/kg; pH ~7.4). The dorsal surface of the brain was then attached to the stage of the Compresstome tissue slicer (Precisionary Instruments) and embedded in 2% agarose. Horizontal hippocampal slices were cut at 450 μm thickness into cold slicing solution. Slices were recovered for a minimum of 30 minutes at 37 °C in artificial cerebrospinal fluid (aCSF): 120 mM NaCl, 2.5 mM KCl, 1 mM NaH₂PO₄, 26 mM NaHCO₃, 1 mM MgCl₂, 2 mM CaCl₂, 10 mM D-glucose (~300 mmol/kg, pH 7.4). Thereafter, slices were moved to room temperature until use. All solutions were continuously bubbled with 95% O₂ and 5% CO₂.

Hippocampal slice perfusion and stimulation

All solutions were kept at 38°C in a water bath, and slices were maintained at 32°C throughout the experiment using a perfusion-line heater (Braintree Scientific, Inc, BS-SYR-HSM). Slices were placed in a physiological chamber, in which the bottom is formed by a thermoelectric device and allowed to equilibrate for 5 minutes. At the desired timepoint, the perfusion solution was switched to 50 mM KCl aCSF (with reduced NaCl to 72.5 mM for osmotic balance) for 30 seconds (or 5 minutes where indicated). Stimulation time includes the period from the high KCl solution reaching the chamber until the tissue was thermally preserved. When stimulation was performed in isotopic tracer experiments, the labeled tracer remained present in the high KCl solution. All experimental conditions were performed on separate slices from the same mouse to achieve paired data, and biological replicates refer to the number of mice included (as indicated in each figure legend). In all experiments with presented data, a minimum biological replicate of 6 was chosen based on an a priori power analysis with preliminary results. To perform the power analysis, an initial group size of three mice was used to estimate anticipated means, using an alpha value of 0.05, a power value of 0.8, and an effect size of 2. Based on these parameters, a sample size of 6 was chosen. A total of 86 animals were used in this study. No exclusion criteria were predetermined, and no animals or slices were excluded from these experiments.

Stable isotope and 2DG tracing

To perform isotopic labeling experiments, uniformly-labeled fuels were purchased from Cambridge Isotope Laboratories (cat. no.: U-¹³C D-glucose CLM-1396; U-¹³C sodium L-lactate CLM-1579; U-¹³C sodium pyruvate CLM-2440; U-¹³C sodium D-3-hydroxybutyrate CLM-3853; U-¹³C sodium octanoate CLM-9617; U-¹³C L-glutamine CLM-1822-H; U-¹³C glycerol CLM-1510), or Sigma-Aldrich U-¹³C D-fructose cat. no. 587621. For ¹³C₆ glucose tracing, the glucose in the aCSF was exchanged with an equimolar amount of U-¹³C₆ glucose. For all other tracers, 2 mM was added, all in the continued presence of 10 mM unlabeled glucose. For the fuel competition experiments, unlabeled fuels were added at the indicated amounts, always with the continued presence of 10 mM labeled glucose. Perfusion with labeled tracers was applied for the indicated timepoints. For all carbon tracing experiments, a control slice was included in each round of experiments, in which no label was applied.

For 2-Deoxy-D-glucose (2DG) experiments, 1mM 2DG (cat. no. D6134, Millipore Sigma) was added to 9 mM glucose in either control aCSF or aCSF with 50 mM KCl. A control slice without 2DG was included in each round of experiments. A small background signal was detected at the m/z corresponding to 2DGP. In a separate group of experiments, we have confirmed that this signal remains unchanged following KCl stimulation. Therefore, the observed increase following KCl is a result of increased 2DGP.

Pharmacological manipulations

Pharmacological inhibition of glycogen phosphorylase was achieved with 1,4-dideoxy-1,4-imino-D-arabinitol (DAB; Cayman Chemical, cat. no. 20939). To apply, the perfusion solution was switched to U-¹³C₆ glucose aCSF in which 1 mM DAB was added, and perfusion continued for a total of 20 minutes. In cases where slices were stimulated, after 15 minutes perfusion with DAB, the solution was switched to high KCl U-¹³C₆ glucose aCSF, with the continued presence of 1 mM DAB for 5 minutes.

Glycogen measurements

For a quantitative comparison of glycogen levels across treatment conditions, a fluorometric assay was used (Sigma-Aldrich, cat. no. MAK016). Acute hippocampal slices were perfused with control or treatment aCSF (5 minutes in 50 mM KCl, ± 15 minutes pre-incubation with 1 mM DAB), dipped once in diH₂O to remove excess aCSF, and then rapidly frozen. Each slice had a mass of approximately 5 mg. Each sample was homogenized by dissociation and sonication in 100 μl diH₂O on ice, then boiled for 5 minutes and centrifuged at 13,000 g for 5 minutes. Supernatants were kept and the insoluble pellet discarded. The assay was performed following manufacturer’s instructions for fluorometric detection. All samples were run in triplicate. A blank of each sample was also included, which did not receive any hydrolysis enzyme, and served as a control for the background level of glucose present in samples. Fluorescence intensity was measured on a plate reader (Synergy H1 microplate reader, BioTek) at λ_ex = 535 nm / λ_em = 587 nm.

Protein content of each sample was measured to normalize for differences in sample size or preparation. A BCA assay was performed following manufacturer’s instructions (Millipore-Sigma, cat. no. 71285), using 5 μl of sample per well. Samples were run in triplicate and absorbance measured at 570 nm.

The average fluorescent intensity of triplicates was calculated, and the fluorescence intensity of the blank well was subtracted. BCA measurements of protein were used to correct for discrepancies in sample sizes and the amount of glycogen was then determined relative to the standard curve.

To convert glycogen content into μmol glycosyl units/g brain weight, we used the following equation:

$$Glycogencontent=\frac{\mu gglycogen}{mgprotein}\ast \frac{\mu \text{mol}\text{glycosyl units}}{162g∕mol}\ast \frac{100mgprotein}{gbrainweight}$$

This calculation assumes that the glycogen polymer is composed of individual glycosyl units (162 g/mol, accounting for a water loss from glucose during glycogen formation), and does not account for possible glucosamine units (Sun et al., 2021).

Sample preparation for MSI

At the desired experimental timepoint, solution was removed from the physiological chamber, 10 μl of 1% (W/V) carboxymethylcellulose (CMC) was added on top of the slice, and the tissue was rapidly thermally denatured at 80 °C for 2 seconds. Thereafter, the slice was rapidly frozen to −20 °C and remained frozen until sectioning. For further details on the thermal denaturation assembly, please refer to (Miller et al., 2023). Tissue was sectioned to 10 μm thickness on a cryostat (Microm HM 550, Thermo Scientific), and thaw-mounted onto indium tin oxide (ITO)-coated glass slides (MALDI IntelliSlides, Bruker Daltonics). Tissue sections were collected within the middle 350 μm of the acute slice, however no differences were observed in labeling dynamics throughout tissue depth ( Miller et al., 2023). If not used on the same day, slides were stored at −80°C before matrix application.

To acquire MSI data, the slides were scanned in a flatbed scanner (Epson Perfection 4490 Photo) at a resolution of 2400 dpi to acquire an optical image and were desiccated until being sprayed with matrix. To make matrix, 39.5 mg 1,5-diaminonaphthalene (DAN; Sigma-Aldrich, cat. no. D21200) was dissolved in 500 μL 1M HCl and 4 mL HPLC grade H₂O then vortexed until well mixed and sonicated before adding 4.5 ml 100% HPLC grade EtOH.¹⁵N-labeled compounds were added to the matrix prior to spraying to calibrate the instrument, and could then be used as reference points during data analysis. The compounds used were 100 μM ¹⁵N₁-glutamate ([M-H]⁻: 147.0418; Sigma-Aldrich, cat. no. 332143), 10 μM ¹⁵N₅-ATP ([M-H]⁻: 510.9692; Sigma-Aldrich, cat. no. 707783), and 1 μM ¹⁵N₅-AMP ([M-H]⁻: 351.044; Sigma-Aldrich, cat. no. 662658). [M-H]⁻ refers to the m/z of the negatively charged form of a molecule resulting from loss of a proton. For small m/z-range methods, standards were spotted onto the slide to calibrate in the low mass region. The standards used included 100 μM pyruvate ([M-H] ⁻: 87.0088), 100 μM lactate ([M-H]⁻: 89.0244), 100 μM ¹⁵N₁-glutamate ([M-H]⁻: 147.0418), and 1 μM ¹⁵N₅-AMP ([M-H]⁻: 351.044).

Once prepared, matrix was loaded into an HTX TM-Sprayer (HTX Technologies) and applied to the slide with a flow rate of 0.09 mL/min, spray nozzle velocity of 1200 mm/min, spray nozzle temperature of 75°C, nitrogen gas pressure of 10 psi, with 4 passes at a spacing of 2 mm. The coated ITO slide was secured into a Bruker slide adapter and inserted into the mass spectrometer.

MALDI imaging

Mass spectra were acquired on a timsTOF fleX (Bruker Daltonics, Billerica, MA) in negative ion mode. Using FlexImaging 5.1 software, hippocampal sections were outlined as individual ROIs with a pixel size of 20 μm. Larger metabolites were acquired in an imaging run (m/z range 50 - 1000), and smaller metabolites were acquired with a separate method (m/z range 50 - 400). In both methods, each MSI pixel consisted of 1000 laser shots. The laser Smartbeam parameter was set to single with a frequency of 10 kHz. Based on the mass range and selectivity of biomolecules for the two MALDI methods, the instrument parameters were optimized, e.g., quadrupole, ion transfer funnels, collision cell, and focus pre-TOF. Before each acquisition, the instrument was calibrated by MALDI using the three ¹⁵N standards spiked into the matrix, or the spotted small molecules, as described above.

MALDI imaging data analysis

All imaging files were imported to SCiLS Lab software (Bruker Daltonics, Billerica, MA) and regions of interest were drawn around the dentate granule cell (DGC) layer in each sample. The ion peak at m/z 241.011 was used to identify the DGC layer, and the region was conservatively drawn within this border. Total Ion Current (TIC) normalization was performed on all data. An overview spectrum of each region was then exported to a .csv file containing a list of ion intensities at all peaks.

This peak list was then imported to the free analysis software, mMass 5.5.0 (Strohalm, Kavan, Novák, Volný, & Havlíček, 2010) (updated version available at: github.com/gyellen/mMass). Peak picking was generally performed with a 0.05% relative intensity threshold (or 0.01% for the small m/z method), but adjusted per run if necessary. The resulting peak list was matched against an in-house library, typically using a ppm inclusion window of 10 ppm, which could be adjusted according to the ppm error of the spiked ¹⁵N compounds used for calibration. The resulting list of annotated metabolites was then exported to an Analysis Report containing the ion intensities at annotated m/z peaks. Occasionally, manual peak picking was necessary, as in the case of very low-intensity ions. No tests for outliers were conducted. Data points were only excluded in cases where the M+0 or expected M+X species remained below detection, and so an accurate fractional abundance could not be calculated for each isotopologue.

Natural abundance isotope correction and fractional calculation

To correct for naturally abundant isotope presence, a matrix method similar to that described by Millard et al. (Millard et al., 2019) was used to correct the apparent label fraction. Because we did not reliably detect every isotopologue species in each individual replicate, the values for the reliably detected species were used together with the correction matrix to calculate a non-negative least-squares estimate of the label fraction. All data presented has been corrected for natural isotope abundance. As we are interested in evaluating the fraction of the metabolite pool that receives labeled carbon (rather than to what extent each molecule is being labeled), we chose to represent our data as the fractional abundance of each isotopologue, rather than as atom% or molecular carbon labeling. Therefore, we calculate the [M+X] fraction as the intensity of signal detected for our isotopologue of interest divided by the sum of intensities of all isotopologues for that metabolite:

$$[M+X]fraction=\frac{[M+X]}{{\sum }_{i=0}^n[M+i]}$$

Where [M+X] is the pool size detected for the isotopologue of interest and n is the number of carbon atoms in the metabolite of interest.

In some cases, the total pool size is of interest rather than a labeling fraction. In these cases, relative abundance is plotted. Relative abundance is calculated as the peak intensity of the ion-of-interest divided by the average peak intensity of the control or reference samples:

$$Relativeabundance=\frac{peakionintensity}{{\mu }_{referencesampleionintensities}}$$

Where μ is the average ion intensity of all reference samples (unstimulated or unlabeled samples). Before normalization, the units of pool size are ion counts. However, metabolite pool size data is always shown as a normalized value since ion counts from MALDI MSI cannot be directly correlated with the quantity of that ion, as the intensity can be affected by tissue physicochemical properties or by ionization efficiencies.

Peak assignment and verification

Peak annotations were based on tandem-MS from tissue sections, accurate mass measurements, and ion mobility measurements. The timsTOF operating in negative ion mode was implemented for tandem-MS measurements. The precursor mass was selected for each metabolite based on MSI data within a 3 - 10 ppm mass error. Mass spectra were collected by rastering randomly over a 200 μm region and averaging 10 scans, of which each consisted of 1,000 laser shots. The precursor m/z isolation window had a range of 0.1 – 3, which means that in addition to the predicted/standard fragment peaks, some contaminating peaks from nearby compounds are expected. Collision energies were ramped between 7 - 55 eV, and the spectra were compared to experimental spectra from purified standards, or to reference fragmentation spectra from the Metlin or HMDB databases.

For calculation of cosine scores to evaluate the match between tissue MS-MS spectra and either experimental spectra for purified standards or predicted spectra from a database, the MS-MS spectra were peak-picked (SNR>5) and binned (<10 ppm). Experimental and standard MS-MS spectra were compared using cosine similarity. When computing the standard-specific cosine, only the peaks present in the standard MS-MS spectrum were considered.

Accurate mass measurements were performed on the 15 Tesla SolariX XR FTICR MS (Bruker Daltonics, Billerica, MA) by imaging tissue sections in negative ion mode. A defined pixel size consisted of 50 μm and 200 laser shots, covering m/z 46-3000. To improve the sensitivity of small molecules, a continuous accumulation of selected ions (CASI) window was set, Q1 mass was defined at m/z 150 with an isolation window of 200 Da. The ion distributions were then compared from the timsTOF and FT-ICR MSI runs. A mass error threshold of < 0.5 ppm using the FT-ICR was used for peak annotation. Both unlabeled and labeled (U-¹³C glucose and U-¹³C glutamine) tissue were analyzed to confirm M+0 as well as isotopologue species. To evaluate possible underlying isobaric species, we used ‘Mass to Formula’ conversion in the mMass software (Strohalm et al., 2010) (updated version available at: github.com/gyellen/mMass) to search for possible formulae within 1 ppm of our measured m/z. Any possible compounds were then searched in HMDB for likely identities, and if any were identified, they are listed in Supp Table 1.

For large compounds, ion mobility was measured using the trapped ion mobility separation of the timsTOF instrument, while for smaller compounds, the drift tube separation in an Agilent 6560 ion mobility Q-TOF instrument was used (detailed methods below). Ion mobilograms plot the ion counts for the target m/z peak and the major isotopologues (for ¹³C labeling) as a function of inverse ion mobility (1/K₀). These mobilities were compared to standard compounds run in the same session, or to predicted 1/K₀ values based on predicted CCS values from the CCSbase database (https://ccsbase.net/), using the calculation described by Gabelica et al. (Gabelica et al., 2019). See Supp Table 1.

Agilent Drift Tube Ion Mobility Liquid Chromatography QToF Mass Spectrometry

Analysis of small molecular mass compounds was performed on an Agilent 6560 drift tube ion mobility liquid chromatography (LC) coupled to a QToF (Quadrupole Time of Flight) mass spectrometer (LC-MS) (Santa Clara, CA). Both standard and tissue samples were prepared using published procedure (MacKay, Zheng, Van Den Broek, & Gottlieb, 2015; Packer et al., 2021) and analyzed using a Waters Acquity BEH130 C18 or Atlantis Premier BEH Z-HILIC (Hydrophilic Interaction Chromatography) 1.7 μm (2.1 x 100 mm) column with a flow rate of 0.3 mL/min (Aluri et al., 2021). The mobile phase consisted of a mixture of either: 10 mM ammonium carbonate, pH 9.0 (solvent A), and acetonitrile (solvent B); or 0.1% formic acid in water (solvent A), and 0.1% formic acid in acetonitrile (solvent B). Solvents A and B were combined in a gradient: 0-0.5 min: 95% A; 0.5-12 min: 50% A; 12-17 min: 30% A; 17.1-20 min: return to initial conditions. Agilent MassHunter (Santa Clara, CA) software was used for system control, calibrating both drift tube ion mobility and QToF MS prior to the sample run. All the drift tube and QToF MS parameters were kept constant between standards and tissue samples. Each sample was injected with 10 μL and the MS was operated in negative ion mode. The chromatographic window containing metabolites of interest was summed to extract both the mass spectra and mobilogram. A standard tuning mix (Agilent) was used to create a linear function for drift tube calibration, excluding ions that exhibited surfing mode propagation. CCS values were calculated from the fundamental low-field ion mobility Mason-Schamp equation. (Harrison, Kelso, Pukala, & Beck, 2019). The drift time and collisional cross section (CCS) were calculated for the metabolites of interest for both standards and tissue samples using the Agilent IM-MS Browser software.

Quantification and statistical analysis

The statistical details of each experiment are provided in the respective figure legend. In general, plots show mean ± standard error of the mean (SEM). All manipulations (e.g., KCl stimulation, stable isotope labeling, or drug treatments) were done on parallel samples (slices) from the same mouse (i.e., each replicate consisted of a matched control and experimental slice from the same mouse). When shown, peak intensities were normalized to the average of the untreated control sample. No randomization or blinding of data took place, but no subjective measurements were made on the data. Hypothesis testing for significant differences was performed by two-tailed unpaired Student's t-tests, or by unpaired One-way ANOVA with Dunnett’s correction for multiple comparisons, as implemented in the GraphPad Prism 7 software. Significance is indicated as p<0.001 (***), p<0.01 (**), or p<0.05 (*). Full statistical reports are provided in Supplementary Table 2 and Supplementary Table 3.

Results:

Spatial metabolomics of the DGC layer in physiological conditions

To investigate metabolism of the DGC layer during rest and activation within the native tissue architecture, we performed MALDI-MSI on acute hippocampal slices. The acute brain slice preparation (450 μm thickness) permits experimental manipulation of live tissue, including presentation of different fuels or stable isotope tracers. The base of our physiological chamber is a custom-built thermoelectric device, which allows for rapid heat inactivation of the tissue at the chosen experimental timepoint (Fig. 1a). This step is critical in preserving the metabolic state of the tissue during further sample manipulation (Dienel, 2021; Miller et al., 2023). Acute slices were then cryosectioned to 10 μm thickness, and the sections were thaw-mounted onto conductive glass slides. To perform MALDI imaging, 1,5-Diaminonaphthalene (DAN-HCl) matrix was sprayed onto the slide and a MALDI laser was repeatedly fired at specific locations across the tissue (Fig. 1b), generating an average mass spectrum and allowing ion image reconstruction (Fig. 1c; each ion image corresponds to the specific m/z value listed below). The ion at m/z: 241.0126 was found to display a reliably clear signal within the neuronal soma layer, and was used to define the region of interest around the DGC layer, from which average spectra were exported for further metabolite annotation and subsequent data analysis. We chose to investigate the DGC layer, as neuronal soma are densely packed in this region, and we could acquire a neuronally-enriched signal ( Miller et al., 2023); however, it is critical to note that at the current spatial resolution (20 μm), metabolic signals from astrocytes cannot be excluded, and we therefore cannot conclude absolute cell-type specificity of these results. Regardless, investigating a region as small as the DGC layer exhibits the advantage of MSI over GC-MS or LC-MS techniques, where it would be challenging to acquire data from such a defined region.

Figure 1: Glucose metabolism in the DGC layer at baseline.

a) Experiments are performed on acute hippocampal slices (450 μm thick) in a perfusion chamber, allowing for application of different substrates or stable isotope tracers. At the chosen experimental timepoint, the tissue is thermally denatured to preserve its metabolic state. b) Slices are cryosectioned to 10 μm thickness, thaw-mounted onto a conductive glass slide, and coated with matrix. Each laser location generates an individual mass spectrum. c) By repeating the process throughout the tissue, an image is generated, with one spectrum at each ‘pixel’ location. The shown ion images correspond to the m/z value indicated beneath each. Using these images, a region of interest outlining the dentate granule cell (DGC) layer is drawn (white dotted line). An average DGC layer spectrum is exported for further data analysis. d) Experimental design for isotopic fuel labeling time course. Slices are thermally preserved following 0, 3, 10, or 30 minutes of ¹³C₆-glucose perfusion. e) Time course of glucose metabolism into glycolytic and TCA cycle metabolites. Empty circles represent ¹²C molecules, filled circles represent ¹³C molecules, and position of carbon removal is indicated with orange line in (iso)citrate and αKG. ¹³C molecules from a second turn of the TCA cycle are shown in dashed boxes. Symmetry in succinate leads to two possible isotopomers in the second turn of the TCA cycle. N=6 biological replicates, defined as separate mice. *< 0.05; **<0.01; ***<0.001 by unpaired One-way ANOVA with Dunnett’s correction for multiple comparisons to the 0 timepoint condition. GAP/DHAP: glyceraldehyde-3-phosphate/dihydroxyacetone phosphate; bPG: bisphosphoglycerates; PG: phosphoglycerates; PEP: phosphoenolpyruvate; αKG: α-ketoglutarate.

Resting metabolism of glucose in the DGC layer

As glucose is considered an obligate fuel of the brain, we first investigated the extent and kinetics of glucose metabolism in the DGC layer by performing a labeling time course with 10 mM U-¹³C glucose. Slices were perfused for 0, 3, 10, or 30 minutes with U-¹³C glucose solution, in which U-¹³C glucose completely replaced unlabeled ¹²C glucose (Fig. 1d). As mass spectrometry separates ions based on mass and charge, it cannot distinguish isobaric species, which are unique compounds with the same chemical formula and therefore the same mass. For this reason, several isobaric species with likely similar behavior are plotted together, such as glyceraldehyde-3-phosphate and dihydroxyacetone phosphate (GAP/DHAP), as well as citrate and isocitrate ([iso]citrate). However, in the brain, some key metabolites have isobaric species which are not involved in metabolism, including hexose phosphate/inositol phosphate and hexose bisphosphate/ inositol bisphosphate (L. Wang et al., 2022). These metabolites were therefore excluded from our analysis. Despite these exclusions, we were able to achieve coverage of glycolysis downstream of GAP/DHAP and of several TCA cycle metabolites (Supp Table. 1).

Using this method, we detected a rapid entry of glucose carbons into glycolysis, reaching between 15-40% labeling after only 3 minutes of perfusion, and approaching a plateau after 30 minutes (at 47 ± 1.9% for GAP/DHAP, 61 ± 3.6% for bisphosphoglycerates [bPG], 76 ± 2.1% for phosphoglycerates [PG], and 46 ± 3.1% for phosphoenolpyruvate [PEP]; Fig. 1e). The lack of complete labeling of these metabolites is possibly due to metabolite channeling or pool segregation, such that not all molecules are actively involved in glycolytic metabolism, but may also be the result of influx of an unidentified endogenous fuel. While lactate and pyruvate are too small in the m/z range to be reliably detected by our method, we were able to follow the labeled carbon into the TCA cycle.

On the first turn of the TCA cycle, all intermediates will incorporate two heavy carbons from labeled acetyl-CoA, generating M+2-labeled isotopologues. On the second turn, M+4 citrate and isocitrate will be generated, and depending on the position of label incorporation, M+4 or M+3 α-ketoglutarate (αKG), followed by M+3 succinate and malate (Aldana et al., 2020; Buescher et al., 2015; McNair et al., 2017). After 30 minutes of perfusion, glucose carbons labeled 64 ± 0.7% of (iso)citrate as M+2 and 11.8 ± 1% as M+4 (Fig. 1e). While (iso)citrate was labeled quite rapidly, the remaining detected TCA metabolites showed slower labeling and did not label to as great an extent, possibly as a result of rapid exchange between αKG and the unlabeled glutamate pool (Mason et al., 1995). Due to low signal intensities, any M+4 or M+3 species in αKG or succinate remain below the limit of detection. Overall, we conclude that glucose is a readily metabolized carbon source for the DGC layer in resting conditions, which quickly passes through glycolysis and enters the TCA cycle. We next explored the metabolic dynamics of alternative fuels.

Resting metabolism of alternative fuels in the DGC layer

To investigate the metabolism of alternative fuels in the DGC layer, we repeated the labeling time course dynamics as outlined in Fig. 1d for each uniformly labeled alternative fuel. It is important to note that in these experiments, 10 mM unlabeled glucose remained present in all solutions. The pathway for metabolic incorporation of each fuel are shown in Fig. 2a. We first investigated the ability of another sugar, fructose, to supply carbon for glycolytic metabolites by entry at the level of GAP/DHAP. While we confirmed the presence of labeled hexose in our tissue (Fig. S1a), we were not able to detect label incorporation in any downstream metabolite, suggesting fructose is not a metabolic contributor to the DGC layer in these conditions. We next investigated the contribution of glycerol, which is converted to glycerol-3-phosphate by glycerol kinase and also enters glycolysis through GAP/DHAP. While labeled GAP/DHAP was detected after 3 minutes of incubation (8.7 ± 1.7%) and quickly plateaued, any labelling was below detection in downstream glycolytic intermediates (Fig. S1b). Notably, there was a very slow rise in the M+2 label of (iso)citrate, which reached only 3.4 ± 1% after 30 minutes (Fig. S1b). Therefore, we conclude that while glycerol is capable of being metabolized, it contributes only minimally to the resting metabolism of the DGC layer.

Figure 2: Alternative fuel metabolism in the DGC layer at baseline.

a) Schematic of metabolic incorporation of alternative fuels. Time course of (b) U-¹³C lactate and pyruvate, (c) U-¹³C β-hydroxybutyrate, (d) U-¹³C octanoate, or (e) U-¹³C glutamine metabolism into the TCA cycle. f) Total incorporation of all isotopologues from 10 mM ¹³C₆-glucose, or from 2 mM of each labeled species into the TCA cycle. Alternative fuels are provided in the continued presence of 10 mM unlabeled glucose. N=6 biological replicates. *< 0.05; **<0.01; ***<0.001 by unpaired One-way ANOVA with Dunnett’s correction for multiple comparisons to the 0 timepoint condition.

We next evaluated carbon sources that enter metabolism at the level of the TCA cycle. We first investigated the monocarboxylate lactate, which is in exchange with pyruvate via lactate dehydrogenase. Since adding lactate alone would drive the reaction toward pyruvate production and therefore increase cytosolic NADH:NAD⁺ ratio, we added pyruvate in a 1:10 ratio with lactate to reduce effects of disturbing the redox balance (2 mM U-¹³C lactate + 0.2 mM U-¹³C pyruvate: lac/pyr). While we observed no labeling in upstream glycolytic intermediates, arguing against gluconeogenesis by the DGC layer in these conditions, rapid labeling of the TCA cycle was detected. Lac/pyr was quickly incorporated into (iso)citrate and approached pseudo-steady state labeling after only about 10 minutes of perfusion. After 30 minutes, labeling was detected as 33 ± 3.6% in M+2 and 4.9 ± 2.1% in M+4 (iso)citrate (Fig. 2b). As observed with U-¹³C glucose labeling, there was again a disconnect in labeling kinetics and extent between (iso)citrate and the downstream TCA metabolites; second-turn labeling of (iso)citrate was also much slower than the first turn. Therefore, U-¹³C lac/pyr reached pseudo-steady state labeling dynamics in the TCA cycle more quickly than glucose, but at a lower fraction labeling, possibly reflecting a continued supply of unlabeled acetyl-CoA from the unlabeled glucose that was present.

Another monocarboxylate, the ketone body βHB, is converted into acetyl-CoA by βHB dehydrogenase and thiolase before entering the TCA cycle. βHB displayed a strikingly rapid and pronounced labeling pattern in (iso)citrate, reaching 38 ± 4.9% M+2 labeling after only 3 minutes of perfusion. After 30 minutes of perfusion, labeling reached 51 ± 2.9% M+2 and 18 ± 1.1% M+4 (iso)citrate (Fig. 2c). Note that the M+2 species begins to decrease as these molecules are converted into the M+4 species through a second turn of the TCA cycle. As previously observed, labeling dynamics and fraction again decrease in the downstream metabolites after (iso)citrate, although the fraction labeled from βHB is still high relative to the contribution from other fuels. We concluded that βHB is rapidly metabolized by the resting DGC layer to supply carbon for the TCA cycle.

While the brain is a lipid-rich organ, the ability of brain cells to metabolize these lipids for energy remains controversial. We therefore investigated the metabolism of an 8-carbon fatty acid, octanoate, which can enter mitochondrial metabolism through β-oxidation. We found that 2 mM U-¹³C octanoate is readily incorporated into the TCA cycle. While labeled carbons appeared quickly in (iso)citrate, pseudo-steady state labeling was reached by approximately 10 minutes at a smaller labeling fraction than observed for glucose, lac/pyr, or βHB. After 30 minutes perfusion with U-¹³C octanoate, (iso)citrate reached 22 ± 2% M+2 labeling and 4.3 ± 0.3 % M+4 labeling (Fig. 2d). Therefore, we conclude that the fatty acid octanoate can be metabolized to a limited extent by the resting DGC layer. Previous reports have suggested that most octanoate metabolism occurs in astrocytes (Andersen, Westi, et al., 2021; Andersen et al., 2022; Cremer, Teal, Heath, & Cavanagh, 1977), and again, we cannot definitively conclude the cell-type responsible for this metabolism in our region of interest, however our lower observed labeling fraction agrees with the finding that it is not a major carbon source for neurons.

Finally, we explored the ability of the DGC layer to incorporate carbon from the amino acid glutamine into the TCA cycle. After deamination to glutamate, these carbons are incorporated through conversion to αKG. Compared to the other alternative fuels tested, 2 mM U-¹³C glutamine entry into the TCA cycle was slower and contributed less to the labeling fraction. After 30 minutes, only 11 ± 1.9% of αKG was labeled as an M+5 species (Fig. 2e). Following similar dynamics, we detected only 6.2 ± 0.7% M+4 succinate, 5.8 ± 0.6% M+4 malate, and 6.6 ± 0.5% M+4 (iso)citrate after 30 minutes perfusion (Fig. 2e). Therefore, while glutamine can be metabolized into αKG, it is not a major carbon source for TCA cycle anaplerosis at rest. It remains possible that our exogenous labeled glutamine is being diluted by a large endogenous glutamine/glutamate pool, in which case the extent of labeling would be underestimated.

To investigate the incorporation of glucose and alternative fuels into amino acids, we calculated the labeled fraction of glutamate, glutamine, and aspartate. Label from both glucose and βHB were elevated in glutamate after 30 minutes (at 30% and 28%, respectively), while lac/pyr labeling reached 11% and octanoate labeled only 2% (Fig. S2a). Conversely, octanoate labeled 17% of glutamine. Glucose labeling reached 18%, while lac/pyr and βHB reached 9% and 8%, respectively (Fig. S2b). Finally, aspartate also showed the greatest labeling from glucose and βHB (at 22% and 21%, respectively), followed by lac/pyr at 14%, and octanoate at 1% (Fig. S2c). These results support the exchange between αKG and glutamate pools, which may lead to decreased labeling in TCA cycle metabolites after (iso)citrate.

As the calculated fraction of metabolite that becomes labeled ([M+X] fraction) can be affected by changes in total metabolite pool size, we also compared the relative abundance of each metabolite throughout perfusion with alternative fuels. There was a notable increase in αKG by 10 minutes of lac/pyr or βHB perfusion, and also a trend toward increased (iso)citrate pool size with βHB perfusion (Fig. S3).

From these experiments, we conclude that in addition to glucose, the resting DGC layer can readily metabolize alternative fuels, with a particularly strong contribution from βHB, followed by lac/pyr (Fig. 2f).

Metabolic response to KCl stimulation

After defining DGC-layer fuel utilization at baseline, we next asked how these cells might change their metabolic activity and fuel incorporation in response to stimulation. To trigger cellular depolarization, we chemically stimulated acute hippocampal slices for 30 seconds with 50 mM KCl solution (Fig. 3a). This created a reproducible and uniform activation throughout the slice, thereby limiting variability from our stimulation protocol. Brief KCl perfusion induced a large energetic demand, as seen in example ion images of the energetic storage molecule phosphocreatine (top), and the end-product of energy use AMP (bottom), in control (left) or stimulated (right) acute hippocampal slices (Fig. 3b,f).

Figure 3: Metabolic response to KCl stimulation.

a) Experimental design schematic. Slices were incubated for 5 minutes ± 30 seconds of 50 mM KCl stimulation immediately prior to thermal preservation. b) Representative ion images of acute hippocampal slices that were unstimulated (left) or stimulated (right). Top row shows ion intensities for phosphocreatine and bottom row shows ion intensities for AMP. The color bar represents the relative intensity of metabolites and is scaled separately for pCreatine and AMP. Total pool sizes are shown for metabolites in glycolysis (c), the TCA cycle (e), or energy supply (f), with or without KCl stimulation. For each metabolite, ion counts are normalized to the average of the unstimulated control condition. d) To evaluate glucose uptake and hexokinase activity, 2-Deoxy-D-glucose is used as a tracer to indicate changes in hexose phosphate pool size. g) Experimental design schematic to determine changes in metabolic flux following KCl stimulation. 30 seconds of KCl was applied either after 3 minutes of labeling (in the labeling dynamic phase), or after 30 minutes of labeling (nearing isotopic steady-state). h) Labeled fraction of glycolytic metabolites from U-¹³C glucose in each condition with or without KCl stimulation. (Iso)citrate labeled fraction in each condition, when supplied with (i) U-¹³C glucose, (j) U-¹³C lactate/pyruvate, (k) U-¹³C β-hydroxybutyrate, (l) U-¹³C octanoate, or (m) U-¹³C glutamine. In c, e, f, N = 19 biological replicates; In d, h-m, N=6 biological replicates. *< 0.05; **<0.01 by paired two-tailed Student’s t-tests. GAP/DHAP: glyceraldehyde-3-phosphate/dihydroxyacetone phosphate; bPG: bisphosphoglycerates; PG: phosphoglycerates; PEP: phosphoenolpyruvate.

To measure metabolic changes in response to KCl stimulation, we monitored the total pool size of metabolites throughout glycolysis (Fig. 3c) and the TCA cycle (Fig. 3e). While we detected few changes in total pool sizes, stimulation increased levels of phosphoglycerates and decreased (iso)citrate. As previously discussed, hexose phosphate and bisphosphate cannot be confidently annotated due to the isobaric inositol species. Therefore, to investigate the behavior of these metabolites after stimulation, we used 1 mM 2-deoxy-D-glucose (2DG) as a tracer molecule. After being taken up, 2DG can be phosphorylated by hexokinases to form 2DGP. We found that 30 seconds of KCl stimulation increased the pool size of 2DGP substantially (Fig. 3d), suggesting increased 2DG uptake and hexokinase activity.

Stimulation increases DGC-layer glucose metabolism

As there were few changes in total pool sizes of glycolytic and TCA cycle metabolites, we further investigated changes in metabolic activity by pairing stable isotope labeling with KCl stimulation. We chose to stimulate tissue for 30 seconds with 50 mM KCl after either 3 or 30 minutes of labeling (Fig. 3g). At 3 minutes, metabolites are in the dynamic/non-stationary phase of labeling (Fig. 1e; Fig. 2b-e), so an increase or decrease in label fraction at this point will inform about the rate of labeled carbon incorporation. For instance, if the labeled fuel is metabolized more rapidly, a higher labeled fraction will be detected at 3 minutes, as more ¹³C will have been incorporated into downstream metabolites. After 30 minutes of labeling, most metabolites are approaching or have reached isotopic steady state and changes in labeling fraction at this time point will reveal differences in the relative influx of carbon sources. For instance, if the total pool size of a given metabolite is unchanged, but more of that pool is labeled with ¹³C, it suggests that the incorporation of labeled fuel into that metabolite increased relative to that of unlabeled fuels. For a detailed discussion on isotope interpretations, we refer the reader to excellent reviews on this topic (Buescher et al., 2015; Jang, Chen, & Rabinowitz, 2018).

Performing 30 second KCl stimulation after 3 minutes of labeling with 10 mM U-¹³C glucose, we observed an increase in label fraction of several glycolytic intermediates (Fig. 3h). GAP/DHAP and bPG both increased substantially (from ~15% to ~26%, and from ~34% to ~51%, respectively), while PEP followed a similar trend, but did not reach statistical significance. After 30 minutes of labeling, KCl stimulation increased GAP/DHAP and PEP labeled fraction (from ~47 to ~56%, and from ~46% to ~59%, respectively), while bPG and PG labeled fraction remained unchanged (Fig. 3h). Therefore, the total contribution of carbon from glucose had increased in GAP/DHAP and PEP. These findings suggest that upon 30 seconds of KCl stimulation, the DGC layer increases glycolysis, thereby incorporating more labeled glucose into glycolytic intermediates.

Glucose supports TCA cycle metabolism during stimulation

As we observed that alternative fuels are metabolized into the TCA cycle of the resting DGC layer (Fig 2), we next tested the extent to which glucose or alternative fuels are used to meet an increased energy demand. Stimulation increased (iso)citrate labeling from U-¹³C glucose over 3 minutes (from ~11% to ~20%; Fig. 3i). Conversely, (iso)citrate showed a decreased label fraction from U-¹³C lac/pyr (from ~22% to ~14%; Fig. 3j) and a non-significant reduction in label fraction from U-¹³C βHB (Fig. 3k). This may be the result of either a decrease in incorporation of labeled alternative fuel, or a dilution of label by increased unlabeled glucose influx. To further explore the difference in label fraction between glucose and lac/pyr or βHB, we investigated the relative pool sizes of unlabeled (M+0) and labeled (M+2) species contributing to the label fraction calculation (Fig. S4). With U-¹³C glucose labeling, there is a decrease in the unlabeled species after KCl stimulation and also a non-significant rise in the labeled species (Fig. S4a). Conversely, a decrease in M+2 species from U-¹³C lac/pyr, and a trend toward a decrease from U-¹³C βHB (Fig. S4b,c) was observed.

KCl stimulation had no effect on U-¹³C octanoate incorporation (Fig. 3l), and a statistically significant but tiny increase in U-¹³C glutamine incorporation at 3 minutes (Fig. 3m). Only minimal changes were observed in the fraction labeled of other TCA cycle metabolites (Fig. S5).

Taken together, the increase in labeling of glycolytic intermediates and (iso)citrate from U-13C glucose relative to alternative fuels strongly suggests a rapid and preferential upregulation in glycolysis in response to the energetic demand of KCl activation.

Glycogen is a minor contributor during metabolic response of the DGC layer to stimulation

Glycogen is a known contributor to brain metabolism in times of energetic demand, and its metabolism may allow for glucose-sparing for neuronal metabolism (Dienel & Cruz, 2015; Rothman et al., 2022). Since glycogen is a large polymer of glucose molecules stored within cells, it is difficult to accurately label this pool with stable isotope tracers while not also saturating other metabolites. As glycogen is present in our acute slices, it is possible that this pool is contributing to the remaining unlabeled fraction of glycolytic intermediates, and its use may dilute label fractions after KCl stimulation. Therefore, to investigate the contribution of glycogen to DGC-layer metabolism, we modified our labeling and stimulation protocol to optimize detection of a decrease in label fraction, which may represent unlabeled glycogen contribution. Slices were labeled for a total of 35 minutes with 10 mM U-¹³C glucose to reach isotopic steady-state in glycolytic intermediates. Some slices were stimulated for the final 5 minutes with 50 mM KCl solution to induce a large energetic demand, in the presence or absence of 1 mM 1,4-dideoxy-1,4-imino-D-Arabinitol (DAB), a glycogen phosphorylase inhibitor (Fig. S6a,c) (Walls et al., 2008).

If glycogen were a large direct contributor to glycolysis, we would expect the unlabeled pool of glycolytic metabolites to increase as the unlabeled glycogen enters the pathway. With DAB application, this unlabeled contribution would be inhibited, and the labeled contribution could increase as additional U-¹³C glucose enters the pathway to compensate for the lack in glycogen carbons. Both by labeled fraction (Fig. S6d) and relative pool size (Fig. S6e), small effects of DAB were observed on DGC-layer glycolysis. After KCl treatment, GAP/DHAP label increased from ~56% to ~65% in control to DAB-treated conditions, and bPG label increased from ~58% to ~74% in control to DAB-treated conditions. An important caveat is the possibility of U-¹³C glucose incorporation into glycogen pools over the 35-minute labeling period. If this occurs, any labeled glycosyl units should be quickly depleted in the 5 min KCl stimulation protocol, followed by release of unlabeled glycosyl units. However, it is possible that the glycogen contribution to glycolytic metabolites is slightly underestimated due to labeled glycosyl units.

To verify that our DAB treatment protocol effectively reduces glycogen mobilization, a colorimetric glycogen assay was performed on homogenized whole hippocampal slices. It is noteworthy that our measured glycogen content (0.11 μmol/g in control) is lower than previously reported in the literature: ~12 μmol/g in the murine hippocampus, or ~7.8 μmol/g in the in vivo human brain (Oe, Baba, Ashida, Nakamura, & Hirase, 2016; Öz et al., 2015). This discrepancy may be due to the assay conditions (sample collection and preparation in cold water), which may allow continued glycogenolysis. Alternatively, the slicing and recovery protocols may result in lower glycogen concentrations in the slices than found in vivo. Methods for glycogen measurement by MALDI-MSI are currently in development, and it will be important to revisit the glycogen contribution to DGC layer metabolism once these are available. Despite the relatively low absolute values measured here, KCl stimulation caused a significant decrease in whole slice glycogen content, confirming that stimulation mobilizes this reserve fuel, as expected (Fig. S6b). In confirmation of the efficacy of DAB, this KCl-induced decrease in glycogen content was blocked by DAB application (Fig. S6b).

Therefore, while glycogen is certainly mobilized by the KCl-induced energetic challenge, relatively minimal contributions were observed in DGC-layer metabolism. It is possible that the glycogen is instead being metabolized by astrocytes in the whole-slice glycogen assay measurements, as this is the primary site of glycogen storage and has been proposed as a glucose-sparing mechanism. Estimating the precise contribution of glycogen remains challenging, and a more specialized investigation is warranted to ascertain the proportion of glycogen utilization under these specific conditions. Previous measurements suggest an increase in brain glycogenolysis in response to activation, and an increase in glucose metabolism to compensate for glycogenolysis inhibition (Dienel, Ball, & Cruz, 2007). However, our data provide no evidence that glycogen is the sole source of unlabeled metabolites or a major contributor to DGC layer metabolism during KCl stimulation.

Metabolic preference of the DGC layer during fuel competition at baseline

We have shown that the DGC layer can metabolize alternative fuels, including lac/pyr and βHB, but in response to stimulation, it specifically increases glucose metabolism. As certain physiological or pathological conditions can increase the presence of lactate or ketone bodies in the brain, we next asked if the presence of these alternative fuels might alter glucose metabolism. For instance, intense exercise may increase circulating levels of lactate from about 1 mM at rest to 15-25 mM and is readily taken up into the brain through monocarboxylate transporters (Goodwin, Harris, Hernández, & Gladden, 2007; Quistorff, Secher, & Van Lieshout, 2008). Lactate has been described as a shared energy storage molecule throughout the body (Hui et al., 2017; Ide et al., 2000). Additionally, circulating βHB is typically below 0.5 mM, but can increase in response to fasting, exercise, or in untreated diabetes, where it has been measured to reach up to 10 mM in blood and 6 mM in CSF (Ohman et al., 1994; Robinson & Williamson, 1980). In a ketogenic diet, which has attracted interest for its role in treatment of intractable epilepsy or diabetes (Lennerz et al., 2018; Ułamek-Kozioł et al., 2019), the level of circulating ketone bodies is elevated to around 1 - 4 mM, but can reach concentrations higher than 10 mM in extreme ketoacidosis (Laffel, 1999; van Delft, Lambrechts, Verschuure, Hulsman, & Majoie, 2010).

We therefore measured the effect of increasing unlabeled alternative fuels on label incorporation from U-¹³C glucose. Slices were incubated for 10 minutes with 10 mM U-¹³C glucose in the presence of 0, 2, 5, or 10 mM unlabeled lac/pyr or βHB, with or without stimulation by 50 mM KCl in the final 30 seconds (Fig. 4a). The presence of unlabeled alternative fuel could impact the fractional labeling through either feedback inhibition or label dilution (Fig. 4b). In the first case, additional supply or production of pyruvate or acetyl-CoA could produce feedback inhibition and slow glycolysis, thereby decreasing label from U-¹³C glucose in glycolytic intermediates. In the second case, incorporation of unlabeled carbon at the level of pyruvate or acetyl-CoA could increase the M+0 species of downstream metabolites and thereby decrease label fraction from U-¹³C glucose by label dilution.

Figure 4: Metabolic preference during fuel competition.

a) Schematic of experimental design. All slices were perfused for 10 minutes with 10 mM U-¹³C glucose in the presence of 0, 2, 5, or 10 mM unlabeled lac/pyr or βHB. Where indicated, 50 mM KCl stimulation was performed for 30 seconds (dark bars). b) Schematic of possible effects of unlabeled fuels on labeling fraction from U-¹³C glucose. Labeled fraction could decrease due to unlabeled carbon incorporation, or due to feedback inhibition and decreased metabolism of U-¹³C glucose. Glycolytic metabolites and (iso)citrate labeling fractions from U-¹³C glucose throughout increasing concentrations of (c) unlabeled lac/pyr or (d) unlabeled βHB. N=6 biological replicates. *< 0.05; **<0.01; ***<0.001. Statistical comparisons across fuel concentration (without KCl) were performed by unpaired One-way ANOVA with Dunnett’s correction for multiple comparisons to the 0 timepoint condition. Effects of KCl stimulation are compared by unpaired Student’s t-test within each concentration group. GAP/DHAP: glyceraldehyde-3-phosphate/dihydroxyacetone phosphate; bPG: bisphosphoglycerates; PG: phosphoglycerates; PEP: phosphoenolpyruvate.

With increasing concentrations of unlabeled lac/pyr, we observed a decrease in U-¹³C glucose labeling of glycolytic intermediates (Fig. 4c, light gray bars). Decreases were measured in GAP/DHAP (from ~41% with 0 mM lac/pyr to ~25% with 10 mM lac/pyr) and PG (from ~68% with 0 mM lac/pyr to ~44% with 10 mM lac/pyr), though any changes in bPG or PEP did not reach significance. The reduced labeling of glycolytic intermediates is most likely caused by feedback inhibition, as there was no evidence of gluconeogenesis from lac/pyr in our labeling time course experiments (Fig. 2). We further investigated the cause of the change in label fraction by plotting the relative intensities of both the unlabeled (M+0) and labeled (M+3) pools. Indeed, M+0 relative abundance did not increase in unstimulated conditions across concentrations, agreeing with the lack of gluconeogenesis previously observed (Fig. S7a-d; statistical results reported in Suppl. Table 2).

A large reduction in label fraction from U-¹³C glucose was also observed in (iso)citrate across lac/pyr concentrations (from ~37% with 0 mM lac/pyr to ~21%, ~15%, or ~10% with 2, 5, or 10 mM lac/pyr, respectively; Fig. 4c). Further investigation into the relative abundance revealed both increases in the M+0 and decreases in the labeled M+X species of (iso)citrate (Fig. S7e), and similar trends were measured in other TCA metabolites (Fig. S7f-h). These results suggest that in competition conditions, lac/pyr is preferentially incorporated into the TCA cycle relative to glucose.

Upon competition with unlabeled βHB, only one condition (5 mM βHB) decreased the label fraction of PEP from U-¹³C glucose (from ~48% with 0 mM βHB to ~38% with 5 mM βHB), while no other glycolytic metabolites were significantly altered (Fig. 4d and S7i-l), suggesting less feedback inhibition occurred from βHB than from lac/pyr competition. However, βHB at all concentrations had a striking effect on the label fraction of (iso)citrate (from ~40% with 0 mM βHB to ~11%, 10%, or 11% with 2, 5, or 10 mM βHB respectively; Fig. 4d). This decrease in label fraction was at least partially driven by a large increase in unlabeled (iso)citrate, which was also observed in αKG (Fig. S7m-p).

Therefore, in the DGC layer under resting conditions, both lac/pyr and βHB outcompete carbons from glucose for entry into the TCA cycle. However, only lac/pyr significantly decreased glucose labeling of glycolytic intermediates, while βHB had less obvious effects on glycolysis.

Metabolic preference of the DGC layer during fuel competition and KCl stimulation

We next tested if the presence of these alternative fuels would shift the metabolic response to 30 seconds of KCl stimulation (Fig. 4c,d, dark gray bars). In lac/pyr competition conditions, KCl stimulation increased labeling fractions from glucose in several glycolytic intermediates (Fig. 4c), overcoming the apparent reduction in labeling from glucose caused by fuel competition. Further investigation of pool sizes revealed this was due to an increase in the labeled pool fraction from U-¹³C glucose (Fig. S7a-d). Therefore, while lac/pyr reduces the labeling of glycolytic metabolites, this inhibition can be partially overcome by stimulation. Although βHB competition had less of an effect on glycolysis than lac/pyr, a similar result was observed, where KCl stimulation overcame the effects of βHB competition in PEP label fraction (Fig. 4d, S7i-l). While both lac/pyr and βHB competition increased the M+0 pool of (iso)citrate and αKG, this effect was reduced following KCl stimulation (Fig. S7e-h, m-p). Possibly the reduced unlabeled pools seen after KCl treatment reflect an increase in TCA cycling and thus metabolite consumption in response to the energy demands of activation.

The addition of either alternative carbon source for 10 minutes was not sufficient to increase energy stores or to buffer the DGC layer from the energetic challenge of stimulation. The expected decrease in ATP and pCr, and increase in AMP and creatine were still observed in all conditions, and the ATP:AMP or pCr:Cr ratios were not significantly altered (Fig. S8a-l).

Taken together, these results suggest that under baseline conditions in the DGC layer, lac/pyr and βHB can displace some glucose contribution to the TCA cycle. Lac/pyr competition can also decrease labeling of glycolytic intermediates from glucose, likely through product accumulation and feedback inhibition. In contrast, βHB metabolism had little effect on glycolytic activity. This suggests different mechanisms of action and potentially different outcomes of competition depending on the alternative fuel provided. In both cases, KCl stimulation drives an increase in glucose metabolism, partially overcoming any competition effects.

Discussion:

Here we used stable isotope labeling with rapid thermal preservation and MALDI-MSI in the physiological acute hippocampal slice to characterize the metabolism of glucose and alternative fuels by the DGC layer. We investigated metabolic activity in resting conditions and in response to brief KCl stimulation, as well as during fuel competition. Overall, our results show that at rest, the DGC layer primarily uses glucose to supply carbon for glycolytic intermediates, but it can readily use alternative fuels (including lac/pyr, βHB, and octanoate, or to a lesser extent, glutamine) to supply carbon for the TCA cycle. Brief chemical stimulation with 50 mM KCl induces a large energetic demand, and a preferential increase in glucose metabolism through glycolysis and TCA cycle relative to metabolism of alternative fuels. In this study, the application of MALDI-MSI to acute hippocampal slices offers a methodological advance, providing excellent coverage of central carbon metabolites within a small, well-defined region of an ex vivo tissue preparation. This preparation is well-suited for experimental manipulation followed by thermal preservation of metabolic state, allowing the investigation of several carbon sources, either in resting or activated conditions, or during fuel competition.

Glycolytic and TCA cycle substrates in the resting DGC layer

Of the fuels tested, only glucose supplied carbon for glycolysis, while neither fructose nor glycerol contributed substantially to glycolytic metabolites. Another potential point of carbon entry into glycolysis is at fructose-6-phosphate or GAP via the pentose phosphate pathway (PPP). Indeed, the role of endogenous inosine metabolism to support metabolic recovery of DGCs after strong activation has been investigated by our group. Inosine is first converted into pentose phosphates, then further into glycolytic intermediates, to supply carbon to metabolically stressed cells. Inhibition of this inosine metabolism led to prolonged energetic depletion in DGCs following strong stimulation ( Miller et al., 2023). Therefore, the PPP may act as another point of carbon entry into glycolysis, particularly in times of high energetic demand.

While glucose is the preferred carbon source for glycolysis, the DGC layer readily used alternative fuels for TCA cycle metabolism at rest. We found that the DGC layer was capable of metabolizing lac/pyr, βHB, octanoate, and glutamine into the TCA cycle under resting conditions, similar to findings in synaptic terminals (Mckenna, Tildon, Stevenson, & Hopkins, 1995). βHB was particularly rapid in its incorporation dynamics to (iso)citrate and reached the greatest labeling fraction after 30 minutes of incubation. This result is in agreement with recently published work showing that acute brain slices from both murine and human tissue displayed large incorporation of labeled glucose and βHB (Westi et al., 2022). Neurometabolism of βHB is also observed in in vivo studies using PET or NMR techniques, confirming uptake of circulating ketone bodies and metabolic incorporation (Pan et al., 2010; Svart et al., 2018).

It has previously been shown that octanoate is metabolized by astrocytes, but not by neurons or oligodendrocytes (Andersen et al., 2022; Edmond et al., 1987; Ioannou et al., 2019; Jernberg et al., 2017). Our results also show relatively little metabolism of octanoate in the neuronally-enriched DGC layer; however, ~22% labeling of (iso)citrate was detected after 30 minutes of U-¹³C octanoate perfusion. We cannot exclude the possibility that this reflects astrocyte uptake, metabolism, and release of labeled ketone bodies, which can then be incorporated by neurons, as has been demonstrated by others (Auestad, Korsak, Morrow, & Edmond, 1991; Guzman & Blazquez, 2001; Thevenet et al., 2016). Indeed, this would be in line with our observation that octanoate labels glutamine to a larger extent than glutamate, and so may indicate its metabolism in the astrocytic compartment (Fig S2b). The labeled glutamine may then be released by astrocytes and enter the DGC layer, where it may be incorporated into glutamate (Schousboe et al., 2014).

The observed variations in dynamics and magnitude of glucose or alternative fuel metabolism highlight the fact that each carbon source may be used differently and may lead to unique metabolic outcomes.

Metabolic response to stimulation

Stimulation by 30 seconds perfusion with 50 mM KCl caused a dramatic energy demand. This was measured by a decrease in the energy storage molecules ATP and phosphocreatine, with a concomitant increase in the end-products of energy use AMP and creatine. While this is a strong stimulation, it can be used as a model of the energetic demand of neuronal activation, and similar energetic responses were observed by our group in response to hilar electrical stimulation or picrotoxin-induced seizure-like events ( Miller et al., 2023). An advantage of our current system is the ability to rapidly preserve metabolic state of intact tissue using fast thermal preservation. This allowed us to investigate the rapid metabolic responses employed by the DGC layer to meet the energetic challenge of neuronal activation.

In response to KCl stimulation, our results suggest a preferential increase in glucose metabolism by the DGC layer. Labeling from ¹³C glucose increased in several metabolites of glycolysis, as well as in (iso)citrate, where it was favored over incorporation of the alternative fuels, lac/pyr, βHB, or octanoate. These results align with ¹³C NMR data in guinea pig cortical slice, where βHB metabolism occurs in neurons, however glucose metabolism is preferentially increased following KCl depolarization (Achanta et al., 2017). Acute and cultured hippocampal slice experiments from rats also found that glucose was necessary to maintain neuronal firing activity (Galow et al., 2014), and lactate was not able to replace glucose as a carbon source to fuel excitation, and may even decrease network activity (Hollnagel et al., 2020). While we cannot conclude neuronal specificity of these results, this data supports the proposal that neurons increase glycolysis upon stimulation rather than rely on astrocytic lactate (Diaz-Garcia et al., 2017; Li et al., 2023; Lundgaard et al., 2015; Rae et al., 2009). Metabolic support of neuronal activity may differ between cell types, for instance excitatory or inhibitory neurons, or neurons with different firing frequencies (Kann, 2016; Kann, Hollnagel, Elzoheiry, & Schneider, 2016). Again, the preferential metabolism of specific fuels emphasizes the unique purpose of each carbon source and may hint at additional functions of metabolic pathways beyond energy production.

Fuel competition in the DGC layer

High circulating concentrations of lactate or βHB may be encountered during strenuous exercise or strict diet regimens. These fuels are converted into acetyl-CoA and label TCA cycle metabolites more rapidly than glucose, which must first be metabolized through glycolysis. Our results revealed that both lac/pyr and βHB could outcompete glucose for entry into the TCA cycle. However, these fuels had differential effects on glycolysis. βHB caused very little change in glucose incorporation to glycolysis, while lactate caused a significant reduction in glucose labeling of several glycolytic intermediates.

This may be due to our specific experimental conditions and the mechanisms of action of these fuels. Lactate has been shown to decrease phosphofructokinase (PFK) activity by promoting dissociation of the active tetramer into less active dimers, and may decrease hexokinase activity in skeletal muscle through structural modification to promote localization to the soluble fraction (Costa Leite, Da Silva, Guimarães Coelho, Zancan, & Sola-Penna, 2007; Leite et al., 2011). In these ways, lactate could directly decrease glycolytic activity. However, we observed little effect of βHB on glycolytic metabolites, which is in contrast with previous studies (Cox et al., 2016; Lund, Ploug, Iversen, Jensen, & Jansen-Olesen, 2015; Shukla et al., 2014; Svart et al., 2018). One notable difference is that these studies treated cells or tissues with ketone bodies on the order of hours to days, and it seems that the main effect of βHB on glycolytic activity is through epigenetic modification and transcriptional regulation of glycolytic enzymes. It is likely that we would not observe any transcriptional changes in our 10-minute treatment period, but may observe a decrease in glycolysis if βHB were supplied for several hours.

The discrepancy between lactate and βHB effects on glycolysis could alternatively be due to the point of entry of these fuels. Lactate and pyruvate are direct end-products of glycolysis, and their cytoplasmic accumulation may cause feedback inhibition of glycolytic enzymes. In contrast, βHB enters the pathway at mitochondrial acetyl-CoA, which may allow glycolysis to continue and produce pyruvate and lactate as exportable endpoints, effectively uncoupling glycolysis and the TCA cycle. Alternatively, it is possible that the pathways to incorporate either ketones or pyruvate differentially alter intracellular conditions, such as redox state. This may in turn draw glycolytic metabolites into other pathways, such as the PPP. While this would explain a decrease in labeling of downstream metabolites, we saw no evidence for increased pentose phosphate labeling in our experiments. Additionally, we detected no change in the labeling of UDP-glucose, an intermediate metabolite in the formation of glycogen, so evidence for diversion of glucose for storage rather than through glycolysis is lacking. Further investigation will be required to determine the enzymes targeted or pathways involved in this discrepancy between lac/pyr and βHB.

While interpreting this data, it is important to remember that both lactate and βHB have signaling functions in addition to their roles as metabolic fuels. Lactate can signal through the G-protein coupled receptor GPR81, also known as the hydroxycarboxylic acid receptor 1 (HCAR1) to reduce cAMP levels (Clara Machado Colucci, D’Ávila Tassinari, da Silveira Loss, & Stürmer de Fraga, 2023). Through this pathway, lactate may be a signal of tissue metabolic state and lead to decreased neuronal firing activity (Briquet et al., 2022), and possibly also increased cerebral blood flow and epigenetic regulation (Morland et al., 2015). βHB is also a signaling molecule and binds to HCAR2 and FFAR3 (free fatty acid receptor 3). Through these receptors, it may influence lipid regulation and promote anti-inflammatory immune responses. Beyond direct receptor signaling, βHB can also inhibit histone deacetylases or can act as a histone modification to regulate gene expression (Newman & Verdin, 2017). The extent to which any metabolic adaptations took place as a result of the signaling functions of lactate or βHB during our 10-minute perfusion protocol was not directly investigated.

Labeling discrepancies throughout TCA metabolites

There is a pronounced difference in labeling among the TCA cycle metabolites reported both in our data and in the literature (Andersen et al., 2022; Arnold et al., 2022; Kanow et al., 2017; McNair et al., 2017). (Iso)citrate is rapidly labeled and reaches 60-80% labeling after 30 minutes with U-¹³C glucose or U-¹³C βHB. In contrast, αKG labeling increases very slowly and only reaches around half as much labeling as (iso)citrate. The difference in labeling rate between (iso)citrate and αKG is likely explained by label dilution from endogenous, unlabeled carbon sources. An exchange between αKG and endogenous glutamate would result in label dilution at this step and decreased labeling fractions in downstream TCA metabolites. We saw evidence of this exchange in our data, as glutamate incorporates label from glucose, lac/pyr, and βHB (Fig. S2). Glutamate is present in mM concentrations within neurons, providing a large endogenous unlabeled carbon pool. Glutamate dehydrogenase (GDH), alanine transaminase (ALT), and branched-chain amino acid transaminase (BCAT) can reversibly produce αKG from glutamate. Previous studies using in vivo NMR and mathematical modeling have estimated that the rate of cycling between αKG and glutamate is much higher (57 μmol/min·g) than that of TCA cycling (0.73 μmol/min·g) (Mason et al., 1995; Sibson et al., 2001). This interconversion can also be accomplished by aspartate aminotransferase (GOT1), which forms part of the malate-aspartate shuttle and simultaneously interconverts aspartate and oxaloacetate, creating another location of carbon exchange between amino acids and TCA metabolites.

It is also possible that the decline in TCA labeling is due to differences in enzyme kinetics of isocitrate dehydrogenase (IDH). IDH is responsible for converting isocitrate into αKG and is considered one of the rate-limiting steps of the TCA cycle (Hiltunen & Hassinen, 1977). The explanation of a further decline in labeling from αKG to succinate could again be explained by enzyme kinetics, and the regulation of αKG dehydrogenase (OGDH) activity by reactive oxygen species, and by NAD⁺ and CoA-SH availability (Tretter & Adam-Vizi, 2005). Another possible explanation of differences in labeling activity is the size and subcellular location of metabolite pools. With our current methodology, we are unable to distinguish cytoplasmic from mitochondrial regions. Many TCA metabolites can be transported across the inner mitochondrial membrane (for instance by the dicarboxylate or 2-oxoglutarate carrier, Slc25a10 and Slc25a11, respectively) and have non-metabolic roles outside of the mitochondria (Arnold et al., 2022; Haws, Leech, & Denu, 2020). Therefore, a large cytoplasmic metabolite pool, which is not actively involved in TCA cycle metabolism, would remain unlabeled and result in a smaller labeled fraction.

Regardless of location, the total pool size of each metabolite itself can also affect the calculated labeled fraction. If one metabolite has a small total pool, it will take a smaller absolute number of ¹³C-labeled molecules to reach a higher label fraction. In contrast, if an adjacent metabolite has a large total pool, the same number of ¹³C molecules will result in a smaller labeled fraction (Jang et al., 2018). With our current techniques, it would require further methodological development to assign an absolute quantification to the measured metabolites.

There have also been several accounts of ‘non-canonical’ or ‘broken’ TCA cycles, in which TCA metabolites do not follow the traditional circular pathway. In activated macrophage, breaks have been detected at aconitase (Palmieri et al., 2020) and at IDH (Jha et al., 2015; O’Neill, 2015), both of which would lead to an elevated labeling in (iso)citrate relative to downstream TCA metabolites. In differentiating pluripotent cells, an alternate route of citrate metabolism has been suggested, in which citrate is exported to the cytoplasm and converted back through OAA to malate. This malate can then enter the mitochondria and be metabolized back to citrate, completing the proposed citrate-malate shuttle (Arnold et al., 2022). This too could explain a relatively low label fraction in αKG and succinate, with larger metabolic involvement of (iso)citrate and malate. These modified TCA cycles are typically associated with a change in cell function, such as differentiation or immune polarization, but the extent to which something similar might occur in the DGC layer is currently unknown.

Limitations

Despite the technological advances outlined here, we are still limited by the spatial resolution of current MALDI-MSI instruments to study regional metabolism. We chose to investigate the dentate granule cell layer, as it represents a neuronally-enriched region where soma are tightly clustered. At this time, we cannot investigate the metabolism of single neurons scattered throughout cortical tissue without significant contribution from other cell types; nor can we investigate subcellular regions, including axons or dendrites, which may have unique metabolic behavior. Therefore, we limit the interpretation of our study to the neuronally-enriched DGC layer, but do not propose to categorically assign metabolic signals to neurons or astrocytes specifically, nor to extrapolate this behavior to other regions or conditions outside of those tested.

Interpretation of ion intensities from MSI spectra must also be done with caution. Regional variation in physicochemical properties (i.e., cellular density, lipid content, etc.) may alter matrix deposition/crystallization, effective laser fluence, or compound desorption. Further, the ionization efficiency of each compound is dependent on the matrix:analyte ratio, the type of matrix used, the competition for ionization by other compounds, and the likelihood of each metabolite to become ionized in positive or negative modes. For these reasons, the recovery of each metabolite is not equivalent and ion intensities cannot be directly correlated to absolute concentration. To overcome these limitations, we perform all of our comparisons within one region of interest and report data as values normalized to the control or reference sample.

Despite these limitations, MSI is an active field of development, and we expect future improvements will make cell-specific measurements and metabolite quantification possible. For instance, improvements in laser technology will allow more focused tissue desorption and possibly enhanced ion signal (Niehaus, Soltwisch, Belov, & Dreisewerd, 2019). As metabolomics studies by MALDI MSI increase, it is also likely that quantification techniques will become widely available (Stopka et al., 2023), and will expand to cover a large library of metabolites.

Finally, it is important to remember that the results presented here are performed in acute slice and use fuel concentrations that would be considered high within in vivo settings. Importantly, we have chosen only to investigate those fuels which readily cross the blood-brain-barrier and are presented to neurons in vivo. Although these experiments are not performed in vivo, the acute slice preparation allows key advantages, including rapid control of supplied fuels, neuronal excitation protocols, rapid heat denaturation, and spatially-resolved metabolomics.

Conclusion

In conclusion, we have used MALDI-MSI and stable isotope tracers to perform metabolomic studies of the DGC layer within a physiological tissue preparation. We conclude that in resting conditions, these cells metabolize glucose through glycolysis and the TCA cycle, but may also incorporate lac/pyr, βHB, octanoate, and to a lesser extent, glutamine into the TCA cycle. Following the energetic demand of brief KCl stimulation, glucose metabolism through glycolysis and into the TCA cycle preferentially increases relative to the alternative fuels studied. In conditions of high lac/pyr or βHB concentrations, either alternative fuel can outcompete glucose for entry to the TCA cycle, but lac/pyr caused a greater decrease in glucose labeling of glycolysis. This work combines several technical strengths to investigate the metabolism of glucose and alternative fuels within a small brain region of physiologically intact tissue across baseline conditions, KCl stimulation, and during fuel competition.

Abbreviations:

2DG

2-Deoxy-D-glucose

2DGP

2-Deoxy-D-glucose phosphate

αKG

α-ketoglutarate

βHB

β-hydroxybutyrate

bPG

bisphosphoglycerates

DAB

1,4-dideoxy-1,4-imino-D-Arabinitol

DAN

1,5-Diaminonaphthalene

DG

dentate gyrus

DGC

dentate granule cell

GAP/DHAP

glyceraldehyde-3-phosphate/dihydroxyacetone phosphate

KCl

potassium chloride

lac

lactate

MALDI

matrix-assisted laser desorption/ionization

MSI

mass spectrometry imaging

PG

phosphoglycerates

PEP

phosphoenolpyruvate

RRID

Research Resource Identifier

TCA

tricarboxylic acid cycle

pyr

pyruvate

Data availability:

The data that support the findings of this study are available from the corresponding author upon reasonable request.