Low‐Glucose Culture Conditions Bias Neuronal Energetics Towards Oxidative Phosphorylation

ABSTRACT

Neurons are almost exclusively cultured in media containing glucose at much higher concentrations than found in the brain. To test whether these “standard” hyperglycemic culture conditions affect neuronal respiration relative to near‐euglycemic conditions, we compared neuronal cultures grown with minimal glial contamination from the hippocampus and cortex of neonatal C57BL/6NCrl mice in standard commercially available media (25 mM Glucose) and in identical media with 5 mM glucose. Neuronal growth in both glucose concentrations proceeded until at least 14 days in vitro, with similar morphology and synaptogenesis. Neurons grown in high glucose were highly dependent on glycolysis as their primary source of ATP, measured using ATP luminescence and cellular respirometry assays. In contrast, neurons grown in 5 mM glucose showed a more balanced dependence on glycolysis and mitochondrial oxidative phosphorylation (OXPHOS), greater reserve mitochondrial respiration capacity, and increased mitochondrial population relative to standard media. Our results show that neurons cultured in artificially high glucose‐containing media preferentially use glycolysis, opposite to what is known for neurons in vivo as the primary pathway for ATP maintenance. Changes in gene and protein expression levels corroborate these changes in function and additionally suggest that high glucose culture media increases neuronal inflammation. We suggest using neuronal culture systems in 5 mM glucose to better represent physiologically relevant neuronal respiration.

Historically, in vitro primary neuronal cultures include 25 mM glucose, making it artificially hyperglycemic. In this article, Swain, et al., show that relatively pure forebrain neuronal cultures can be grown and maintained in 5 mM glucose media for up to 2 weeks, with negligible effects on morphology or synaptogenesis. However, oxidative phosphorylation and mitochondrial content are upregulated, and glycolysis is suppressed, relative to “standard” culture media. This report opens an experimental window to re‐test questions regarding neuronal respiration and function in disease states, using culture conditions that more closely mimic neuronal respiration in vivo.

Abbreviations

ABB

antibody‐blocking buffer

ATP

adenosine triphosphate

BP

BrainPhys culture media

DEGs

differentially expressed genes

DIV

days in vitro

DNP

2,4‐Dinitrophenol

ECAR

extracellular acidification rate

FDR

false‐discovery rate

GFAP

glial fibrillary acidic protein

GI

glycolytic index

GO

gene ontology

GSEA

gene set enrichment analysis

HG

high glucose (25 mM)

LCMS

Liquid chromatography mass spectrometry

LFG

log₂‐fold change

LG

low glucose (5 mM)

MAP 2

microtubule‐associated protein 2

NF‐H

neurofiliamin heavy chain

OCR

oxygen consumption rate

OXPHOS

mitochondrial oxidative phosphorylation

PBS

phosphate‐buffered saline

PCA

principal component analysis

PDL

poly‐d‐lysine

PFA

paraformaldehyde

PLL

poly‐l‐lysine

RIPA

radioimmunoprecipitation assay

RNA‐seq

RNA‐sequencing

RRID

Research Resource Identifier (see scicrunch.org)

TCA

tricarboxylic acid

vGAT

vesicular GABA transporter

vGluT1

vesicular glutamate transporter 1

1. Introduction

Brain functionality requires a steady flow of energy and consumes around 20% of the body's energy supply at rest (Mink et al. 1981). Neurons in the central nervous system consume a vast majority of this energy supply, primarily to support synaptic transmission (Hall et al. 2012; Harris et al. 2012; Hyder et al. 2013). To ensure energetic supply for such immense energy demand, neurons efficiently complete the oxidation of glucose via glycolysis and mitochondrial oxidative respiration to generate and maintain synaptic ATP levels, with a heavy reliance on OXPHOS in vivo and in acute brain slice experiments (Harris et al. 2012; Ivanov et al. 2014; Rolfe and Brown 1997; Schousboe et al. 2011). Since energy maintenance is critically important for brain and synaptic function and impaired in a variety of diseases, these pathways may serve as vital targets for drug discovery (Watts et al. 2018).

Glucose availability plays a critical role in neurogenesis and functionality during neural development in the CNS. During neurogenesis, neuronal stem cells (NCSs) undergo symmetric cell division and divide asymmetrically to form differentiated neurons (Rumpf et al. 2023). NSCs depend on glycolysis for energy production, limiting ROS production and differentiation (Khacho et al. 2016). On the other hand, in postmitotic neurons after metabolic remodeling, the energy production route changes to OXPHOS with decreased glycolysis (Rumpf et al. 2023). However, despite the predominant reliance on glycolysis, OXPHOS remains essential in NSCs. OXPHOS supports the proper timing of cell differentiation and contributes to generating neurons with appropriate identities (Park et al. 2025). Partial inhibition of OXPHOS in early NSCs prevents differentiation, causing prolonged cell divisions, while potent inhibition leads to early division defects and potential fate disruptions, as shown in invertebrates (Homem et al. 2014). A finely tuned neuronal metabolic balance ensures efficient localized ATP production during stem cell divisions, neuronal differentiation, and the energy‐intensive neurite growth and pruning processes (Jackson and Finley 2024; Khacho et al. 2016; Park et al. 2025). Elevated glucose levels tend to decrease neural stem cell proliferation while promoting differentiation in vivo (Fu et al. 2006; Ji et al. 2019), while changes in glucose in vitro are bidirectional: early reduction in glucose promotes neural stem cell proliferation, while during the differentiation period reduced glucose supports differentiation into neurons and astrocytes (Horie et al. 2004). In early embryonic stages (E8–E14 in mice), neurons do not rely on glucose to meet their energy needs, as glucose transporters are not yet fully expressed and mitochondria are still developing: the fetus is provided maternal ketones (like β‐hydroxybutyrate and acetoacetate) as a primary fuel source (Angelopoulos et al. 2022; Bronisz et al. 2018). By late gestation (E18–E20 in mice), glucose uptake is enhanced and glycolytic enzyme expression is elevated, and metabolism transitions to glucose as the primary fuel concomitant with upregulation of energy‐demanding processes like ion pumping and synaptogenesis (Angelopoulos et al. 2022; Béland‐Millar et al. 2017). Thus, the regulation of glucose is essential for maintaining the balance between proliferation and differentiation of the neuronal system (Molloy and Barry 2024).

As the primary metabolic fuel in the brain, glucose concentration is tightly regulated. Brain glucose level is estimated to be approximately 20% of blood plasma glucose (Dunn‐Meynell et al. 2009). Euglycemic blood glucose is 4–6 mM, and in the brain, it ranges between 1 and 3 mM (Díaz‐García et al. 2019; Dunn‐Meynell et al. 2009; Silver and Erecińska 1994). In clinically hyperglycemic conditions, blood glucose levels rise to 15 mM, and brain glucose levels rise to 4.5 mM (Kleman et al. 2008; Silver and Erecińska 1994). These glucose concentrations are far lower than those used in standard neuronal cultures.

The primary culture of rodent neurons is widely used as an approach to understand cellular metabolism (Agostini et al. 2016; Brewer and Cotman 1989; Mikrogeorgiou et al. 2018; Seibenhener and Wooten 2012). These in vitro neuronal cultures are routinely used as a platform for functional assays, which ideally should closely mimic the physiological environment. Plate‐based neuronal cultures provide researchers with a platform to investigate neuronal development and function such as signal transmission, cellular trafficking, and neuronal morphology. A minimal, completely defined media used in most culture systems contains all the necessary nutrients for cell survival and growth, and commercially available formulations are widely used. Surprisingly, the composition of this media used for primary neuronal culture for decades is highly non‐physiological, especially concerning glucose, which has a standard concentration of ~25 mM (Silver and Erecińska 1994). For context, to reach 25 mM glucose in brain parenchyma would require ~125 mM plasma glucose concentration, indicating conventional primary neuronal culture is impossibly hyperglycemic and fully saturating neuronal glucose uptake (Duarte and Gruetter 2012). As a case in point, a recent study showed that neurons need to be exposed acutely to low glucose conditions in order to generate an activity‐dependent upregulation of glycolysis, indicating that neuronal glycolysis is saturated in vitro when grown in standard high‐glucose media (Wang et al. 2024). This discrepancy in glucose concentration between standard culture media and that found in vivo may lead to misinterpretation when interrogating neuronal metabolism in culture and limits direct comparison between intact brain and primary neuronal cultures.

Here, we attempted to grow mouse embryonic neurons in more physiologically relevant glucose conditions, and compared morphology, metabolism, and gene expression to neurons grown in standard 25 mM commercial media. Our results show that neurons cultured in high glucose media preferentially use glycolysis and inhibit OXPHOS, which is the opposite of what is known for neurons in vivo as the primary pathway for ATP maintenance. We suggest studies on neuronal metabolism up to this point would benefit from critical review.

2. Materials and Methods

2.1. Animals

All animals were used according to animal welfare protocols approved by the University of Nevada, Reno (IACUC Protocol #20‐08‐1051) and US standards for animal research to minimize pain and distress (National Research Council (US) Committee for the Update of the Guide for the Care and Use of Laboratory Animals 2011). In vitro, neuronal culture was performed using embryonic mice of both sexes. We used a minimum of 8 embryos from each pregnant mouse to generate cultures. A total of 35 pregnant mice were used for the experiments described in this report. Mice were acquired from Charles River (Strain #027, C57BL/6NCrl, RRID: IMSR_CRL:027), and a colony was maintained locally. Animals were housed under a 12‐h light/12‐h dark cycle, provided ad libitum access to food and water, and maintained on standard mouse chow in a specific pathogen‐free facility.

2.2. Neuron Cultures

The isolation of the cortex and hippocampus, as well as the generation of neuronal cultures, followed approaches previously described (Swain et al. 2018; Xu et al. 2024). Neurons were obtained from the embryonic brain at 17 to 19 days of gestation. Briefly, pregnant female mice were deeply anesthetized with vaporized isoflurane in an airtight chamber until unresponsive and then killed by cervical dislocation. Embryos were removed from the uterus under sterile conditions and decapitated. The cortex and hippocampi were isolated from the embryonic brain, cut into fine pieces, and triturated with a plastic pipette to dissociate. Tissue and cells from multiple embryos in a litter were pooled. Dispersed cells were seeded onto high molecular weight PDL (70–150 kDa) coated cell culture plates at 80 000 cells/cm² for 96 well plates or 120 000 cells/cm² for coverslip‐containing 24 well plates and maintained in a sterile incubator at 37°C, 5% CO₂. Standard media (high glucose condition) was Neurobasal Plus medium (25 mM) (ThermoFisher, #A3582901) with B‐27 supplement Plus (ThermoFisher, #A3582801), GlutaMAX‐I (ThermoFisher, #35050061), and Antibiotic‐Antimycotic (ThermoFisher, #15240062). Low‐glucose media used Neurobasal A Medium (no D‐glucose, no sodium pyruvate; ThermoFisher, #A2477501) supplemented with 5 mM glucose, 0.22 mM Sodium Pyruvate, B‐27 supplement Plus, GlutaMAX‐I, and Antibiotic‐Antimycotic. Cultures were grown in one of the two media conditions throughout the experimental duration. One‐half of the media was replaced every 2 days. Experimental assays were performed at DIV 14 for cortical cultures and DIV 18 for hippocampal cultures. In a subset of experiments described below, we cultured neuronal cells in BrainPhys medium supplemented with SM1 (StemCell Technologies, #05792) following media preparation and media exchange as described in Faria‐Pereira et al. (2022).

2.3. Cell Viability Assessment

To measure cell viability, inferred from cellular reductive capacity, culture medium was replaced with fresh medium containing 10% (v/v, 1X) alamarBlue Cell Viability Reagent (ThermoFisher; #DAL1100) and stored in an incubator for 3 h. Then, 570 nm absorbance was measured on a multi‐well plate reader (RRID:SCR_020300, SpectraMax M5 Multi‐Mode Microplate Reader, Molecular Devices, San Jose, CA, US) and corrected for 600 nm absorbance, as per the protocol provided by the manufacturer. The data were computed using the formula provided by the manufacturer to calculate the reduction percentage. Viability of cell cultures, reported as a percentage of reduced alamarBlue reagent, was normalized against the absorbance of blank controls.

2.4. Immunohistochemistry

Neurons grown on glass coverslips in 24‐well plates were removed from the incubator at DIV 14, immediately fixed in 4% EM‐grade PFA (from 32% formalin stock, Electron Microscopy Sciences #15714) in 0.1 M PBS for 15 min, washed three times for 5 min each in freshly exchanged PBS, permeabilized in ice‐cold methanol for 10 min, and followed by another wash in PBS. All washes were done with PBS three times for 5 min each time. Cells were permeabilized with 0.1% Triton‐X in PBS for 20 min and then washed. Afterward, cells were incubated in an antibody‐blocking buffer (ABB: 1% fish gelatin, 0.5% Triton‐X, 0.025% sodium azide in PBS) for 1 h. Cells were incubated with primary antibodies diluted in ABB overnight at 4°C. Cells were then washed and incubated with secondary antibodies diluted in ABB for 2 h at room temperature. Antibodies used in each assay are listed in each respective section. After a final wash, coverslips were mounted with SlowFade Diamond (ThermoFisher Cat# S36967) and imaged on a Leica SP5 confocal microscope using the Lightning deconvolution package and Nyquist sampling with 40x oil‐immersion objective. Z‐stacks spanning the monolayer culture architecture were compressed and then analyzed in ImageJ (RRID: SCR_003070).

2.4.1. Excitatory/Inhibitory Content Assay

Primary antibodies against vGluT1 (guinea pig, Synaptic Systems Cat# 135304, RRID: AB_887878), vGAT (rabbit, Thermo Fisher Scientific Cat# PA5‐27569, RRID: AB_2545045), and bassoon (chicken, Synaptic Systems Cat# 141016, RRID: AB_2661779) were used at 1:5000, as well as MAP2 (mouse, Millipore Cat# MAB3418, RRID: AB_94856) at 1:2000. For secondaries, Alexa 405 α‐mouse (rat, R&D Systems Cat# FAB116V, RRID: AB_3646007), Alexa 488 donkey α‐rabbit (ThermoFisher Scientific Cat# A‐21206, RRID: AB_2535792), Alexa 594 goat α‐guinea pig (ThermoFisher Scientific Cat# A‐11076, RRID: AB_2534120), and Alexa 647 goat α‐chicken (ThermoFisher Scientific Cat# A‐21449, RRID: AB_2535866) were used at 1:1000. For each image, the MAP2 signal was isolated using the Huang threshold, and vGluT1, vGAT, and bassoon signals were detected using ImageJ/FIJI's default Analyze Particles algorithm. MAP2 and bassoon masks were dilated to identify synaptic contacts. Any vGluT1 and vGAT puncta that did not overlap with MAP2 or bassoon masks were then eliminated. vGluT1 and vGAT particles larger than 2 pixels wide were counted from the resulting image. The ratio of excitatory to inhibitory content was calculated as the ratio of vGluT1‐positive particles divided by vGAT‐positive particles.

2.4.2. Mitochondrial Labelling and Analysis

Cortical neurons were stained at DIV 14 with Mitotracker Orange (ThermoFisher #M7510) at 10 nM for 45 min in an incubator at 37°C, 5% CO₂. After incubation, the media was replaced, and neurons were kept in an incubator for an additional 5 min. The neurons were then fixed and stained for immunohistochemical analysis. Primary antibodies against MAP2 and the heavy chain of neurofilamin (NF‐H; chicken, ThermoFisher Scientific Cat# PA3‐16753, RRID: AB_2149632) were used at 1:2000. Secondary antibodies Alexa 488 donkey α‐mouse (ThermoFisher Scientific Cat# A‐21202, RRID: AB_141607) and Alexa 647 goat α‐chicken (ThermoFisher Scientific Cat# A‐21449, RRID: AB_2535866) were used at 1:1000. MAP2, NF‐H, and MitoTracker channels were split and thresholded; low‐intensity MAP2 staining in the cell body was used to identify and remove cell bodies from image analysis. MAP2 staining was used as a threshold mask to identify dendritic mitochondria, and NF‐H staining was used as a threshold mask to identify axonal mitochondria. Mitochondrial content was calculated as the area of mitochondria per a given area of MAP2 or NF‐H in dendrites and axons, respectively. Mitochondrial count and mean mitochondrion area were measured in ImageJ using the Mitochondria Analyzer plugin (Chaudhry et al. 2020). The total branch length was estimated using automatic tracing from the Simple Neurite Tracer plugin in ImageJ (Arshadi et al. 2021).

2.5. Cell Labelling and Sholl Analysis

Healthy cultured cortical neurons were patched on DIV 14 in whole‐cell current‐clamp mode at room temperature for 10 min to permit dye diffusion. Standard Na‐based extracellular solution and K‐based intracellular solution containing 0.25 mg/mL Dextran Alexa Fluor 568 were used. Borosilicate pipettes (1.5 mm OD) were pulled to a resistance of 4–6 MΩ. Access resistance during recordings was < 20 MΩ. After filling, fluorescent images of cells were acquired using a Zeiss Upright Axio Examiner microscope (RRID: SCR_025040) and 20x objective with a sCMOS camera (Photometrics Prime 95B‐25MM Camera, RRID: SCR_018464). Neuronal morphology was reconstructed manually with ImageJ Simple Neurite Tracer (Arshadi et al. 2021; Ferreira et al. 2014). Sholl analysis, mean branch length, total branch length, longest branch length, number of branches, and convex hull measurements were obtained using the Simple Neurite Tracer plugin.

2.6. Protein Profiling

2.6.1. Protein Extraction

Cortical neurons from five different pregnant mice were cultured, and lysates were harvested on DIV 14. Neuronal culture wells were washed three times gently with 0.1 M phosphate‐buffered saline and collected by scraping. The cells were pelleted at 4000 RCF (relative centrifugal force) at 4°C and lysed using 125 μL radioimmunoprecipitation assay (RIPA) buffer (ThermoFisher, Cat# 89900) per well. RIPA buffer included protease and phosphatase inhibitors (0.1% 2 mg/mL aprotinin, Millipore Sigma, Cat# A1153; 1% 1 mg/mL leupeptin, Cat# L8511; 1% 1 M sodium fluoride, Cat# 93482; 1% 0.1 M phenylmethylsulfonyl fluoride, Cat# 201154; 1% 0.5 M ethylenediaminetetraacetic acid, Cat# ED). Lysates were agitated for 30 min, centrifuged at 13 000 RCF for 20 min at 4°C, and their supernatants were stored at −80°C until ready for further protein analysis.

2.6.2. Protein Digestion

The protein content from samples was estimated using the fluorescence‐based protein assay EZQ (ThermoFisher Scientific, #R33200). Protein extracts (100ug per sample) were reduced, alkylated with iodoacetamide, and digested with a trypsin/Lys‐C protease mixture using ThermoScientific EasyPep Mini MS Sample prep kit (Cat #A40006) following kit instructions. The trypsin/Lys‐C was added at 1:10 (enzyme: protein) for the digestion. Samples were cleaned up for analysis using the column provided with the kit. Samples were reconstituted in 100 μL 0.1% formic acid in water, making a final concentration of 1 ug/μL for analysis.

2.6.3. Data‐Independent Acquisition of Protein Fragments

Liquid chromatography mass spectrometry (LCMS) was performed on an Evosep One LC platform (Evosep, RRID: SCR_024590, Odense, Denmark) interfaced with a Bruker timsTOF Pro 2 (RRID: SCR_026545, Billerica, MA) using DIA‐PASEF to characterize proteomic changes in serum. Peptides from serum digests were separated on the Evosep One LC using the 30 samples per day method on a PepSep C18 column (15 cm × 150 μm, 1.5μm packing). Separated peptides were then ionized using a CaptiveSpray (Bruker, Billerica, MA) ionization source with a 20 μm ZDV emitter prior to mass spectral analysis. Mass spectral analysis was performed using DIA‐PASEF (RRID: SCR_025304) with a staggered window scheme over a mass range of 100 to 1700 m/z and a mobility range from 0.70 to 1.30 1/K0 with a ramp time of 85 ms. Prior to sample analysis, a hybrid chromatographic library was generated using an aliquot from each sample to form a pool of peptides. This pool of peptides was analyzed by both Data‐Independent Acquisition (DIA) and Data‐Dependent Acquisition (DDA) using the Spectronaut v18.7 (Biognosys, Schlieren, Switzerland) integrated database search engine Pulsar to generate a hybrid library. The species‐specific FASTA database Mus musculus was downloaded from https://www.uniprot.org/proteomes/UP000000589 containing 54747 entries and used to generate the reference library. Biological samples were analyzed using the same LC gradient as that used for the generation of the library. Proteins were quantified and initial statistical analysis was performed using Spectronaut.

The protein‐level data including peptide counts and total raw intensities were exported from Spectronaut and further analyzed in R v4.4. Proteins with less than 2 peptides per protein group within one experimental group were removed from downstream analysis. The cyclic LOESS log2 normalization was used to transform the protein intensities, resulting in a normal distribution. Prior to performing statistical analysis, PCA and correlation matrices were plotted to assess data quality. The Linear Models for Microarray Data (limma v 3.6) Bioconductor package was used to calculate differential expression among the experimental conditions using the lmFit and eBayes options. Proteins were deemed to be significantly different with a fold change > 1 and a false‐discovery adjusted p‐value < 0.05.

2.7. RNA‐Sequencing and Differential Gene Expression Analysis

Cortical neuronal cells pooled from multiple embryos of both sexes, from 5 pregnant mice (constituting 5 independent biological replicates), were grown in 24‐well plates in 5 mM glucose or 25 mM glucose. Cells were lysed on DIV 14 and RNA was extracted with a ReliaPrep RNA Miniprep Systems kit (Promega #Z6011) using the manufacturer's protocol. The lysates from 8 wells with neurons from the same litter and grown in the same condition (5 mM or 25 mM) were run over the same column and constituted one sample. Five matched samples were collected for both the high (25 mM) and low (5 mM) glucose conditions. We prepared the RNA for quality assessment with an RNA 6000 Pico kit (Agilent # 5067‐1513) using the manufacturer's protocol. We then used a 2100 Bioanalyzer instrument (Agilent) to obtain RNA integrity numbers for all samples. RNA extracts were prepared for sequencing with Stranded mRNA Prep (Illumina #20040532) according to the manufacturer's protocol. The generated libraries were pooled and sequenced using a NextSeq P3 kit (Illumina #20040560). RNA‐sequencing (RNA‐seq) was performed on the NextSeq 2000 platform (Illumina) with 100 paired‐end reads.

To process RNA‐seq data from FASTQ files to gene‐level read counts, we performed quality control on the raw reads using FastQC v0.11.9 (RRID: SCR_014583; Andrews 2010), low‐quality bases and adapter sequences were trimmed with Fastp v0.4 (RRID: SCR_016962; Chen 2023). The cleaned reads were then aligned to the reference mouse genome GRCm39 from gencode using STAR v2.7 (RRID: SCR_004463; Dobin et al. 2012) and HISAT2 v2.2.1 (RRID: SCR_015530; Kim et al. 2019); STAR outperformed HISAT2. Finally, FeatureCounts v2.0 (RRID: SCR_012919; Liao et al. 2013) was used to assign STAR aligned reads to genes based on the reference annotation file gencode.vM33.primary_assembly .annotation.gff3, generating a table of read counts for each gene. Before continuing with the differential analysis, transcripts of genes with less than 15 counts across both experimental conditions were removed, and only protein‐coding transcripts were retained for the downstream analysis as additional quality controls. Differential analysis was conducted with the R Bioconductor package DESeq2 v1.44 (RRID: SCR_006442; Love et al. 2014). Differentially expressed genes were defined by having a > 0.6 ‐Log₂₋Fold change (LFC) and false‐discovery rate (FDR) adjusted p‐value (q) < 0.05.

2.8. Gene Set and Pathway Enrichment Analysis

After differential analysis for both transcript and protein datasets, differentially expressed genes (DEGs) were initially analyzed and concatenated using iPathwayGuide (Advaita BioCorporation; Donato et al. 2013; Draghici et al. 2007) in the Nevada Bioinformatics Center Core Facility (RRID:SCR_017802), at the University of Nevada, Reno. The full list of DEGs from both transcript and protein datasets was ranked by the largest fold change values and separately input into the GSEA (v4.3.3) software (Subramanian et al. 2005). The molecular signatures database (MSigDB) 3.0 hallmark pathways were used (Liberzon et al. 2015). Pathways were considered significant when p < 0.05 and q < 0.25 (Liberzon et al. 2011). iPathwayGuide was used to identify 144 genes that were differentially expressed at both the transcript and protein levels. Gene ontology (GO) enrichment was done using Metascape (v3.5.20240901; Zhou et al. 2019). Genes were split into two lists based on their fold change values in the transcript dataset, with the genes with positive fold changes being in one and negative fold changes in the other. Ontologies and pathways were considered significant at q < 0.25. The two gene lists generated a set of associated pathways on ontologies, pulling from Reactome Gene Sets (Fabregat et al. 2017), GO Terms (Ashburner et al. 2000), and KEGG Pathways (Kanehisa and Goto 2000). Graphs were generated in Metascape or using ggplot2 (Wickham 2016).

2.9. ATP Content Determination

Cortical neurons were maintained in either 25 mM glucose or 5 mM glucose media in 96‐well tissue culture flat bottom plates and used at DIV 14. A luciferase‐based luminescent ATP determination assay kit (ATPlite Luminescence Assay System 96‐well; Revvity #6016941) was used to determine the ATP content according to the manufacturer's protocol. In this assay, we used Tyrode's Buffer (NaCl 119 mM; KCl 2.5 mM; MgCl2 2 mM; CaCl2 2 mM; HEPES 25 mM) with four different energetic substrate conditions. The conditions were a positive control (added 5 mM or 25 glucose and 1.25 mM sodium pyruvate), “oxidative phosphorylation only” condition where glycolysis was inhibited (no glucose, added 1 mM iodoacetic acid and 1.25 mM sodium pyruvate), a “glycolysis only” condition where oxidative phosphorylation was inhibited (no pyruvate, added 5 mM glucose or 25 mM glucose and 1 μM oligomycin) and a negative control (no glucose, no pyruvate, added 1 mM iodoacetic acid and 1 μM oligomycin). Luminescence was measured using a multi‐well plate reader (SpectraMax M5 Multi‐Mode Microplate Reader, Molecular Devices, San Jose, CA, USA), and the ATP concentration was determined by plotting the results against an ATP standard curve. The data were normalized per well against total protein concentration evaluated from cell lysate using EZQ Protein Quantitation Kit (ThermoFisher Scientific; Waltham, MA, USA). On each plate, three technical replicates (separate wells of cultured neurons) were averaged per condition and time point to account for variability. Results include five (5) independent cultures (biological replicates) for each time point.

2.10. Cellular Respiration Measurements

Cellular oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured using a Seahorse XFe24 extracellular flux analyzer (Agilent Seahorse XFe24 Cell Analyzer, RRID: SCR_019539, SCR_024491). Embryonic cortical neurons were plated at 80 000 cells/well density in XF24 cell culture plates. These plates were coated with high molecular weight PLL (70–150 kDa) followed by 1% rat tail collagen I (ThermoFisher Scientific; Waltham, MA, USA) to increase cell adhesion (Lange et al. 2012). The assay was performed in Seahorse XF Media supplemented with 10 mM glucose, 2 mM l‐glutamine, and 1.25 mM pyruvate. The cell plates were gently washed and incubated for 1 h with prepared XF media in a non‐CO2 incubator. The OCR and ECAR were determined using XF media after the sequential injections of 1 μM oligomycin to inhibit complex V, 0.5 mM 2,4‐dinitrophenol (DNP) to uncouple the proton gradient, and 1 μM each of rotenone and antimycin to inhibit complex I and complex III, respectively (Sigma‐Aldrich; St. Louis, MO, USA). On each plate, three to five technical replicates (separate wells of same cultures) were averaged per media condition to account for variability, and results include five (5) independent cultures (biological replicates). Data from wells where uncorrected baseline values were low (< 100 counts), or that lacked a reponse to DNP were excluded form analysis. The data were normalized against total protein concentration per well. Non‐mitochondrial OCR value was subtracted from OCR to obtain basal respiration. The reserve respiration capacity was obtained by subtracting basal OCR from maximal respiration. Similarly, the basal glycolysis and reserve glycolytic capacity were obtained (Brand and Nicholls 2011). Furthermore, we calculated the total ATP production rate (J_ATP), the ATP production rate from glycolysis (J_ATP,glyc), the ATP production rate from oxidative phosphorylation (J_ATP,OX), and the glycolytic index (GI) using the formula described in Mookerjee et al. (2017). These values were derived from OCR and ECAR data and visualized in a bioenergetic space plot.

2.11. Calcium Imaging With Fluo‐4 Dye

Hippocampal neurons plated on coverslips in 24‐well plates on DIV 18 were incubated for 5 h with 10 μM bicuculline and 0.5 μM strychnine in neurobasal media to increase basal network activity (Samoilova et al. 2003), followed by a 30 min incubation with 0.5 μM fluo‐4 dye (Molecular Probes, Life Technologies, Grand Island, NY) at 37°C, 5% CO₂. After incubation, the media was replaced with fresh neurobasal media with 10 μM bicuculline and 0.5 μM strychnine until used for imaging. Coverslips were placed in a perfused bath chamber (~1 mL/min) on an upright on a Zeiss AxioExaminer microscope in HEPES‐buffered Tyrode's solution (plus bicuculline and strychnine) at 35°C bubbled with 95/5% O₂/CO₂. Fluo‐4 was illuminated during image acquisition constantly at 450–490 nm excitation by an LED light engine (Lumencor SOLA, attenuated 80% by ND filter), and emission was recorded at 500–550 nm. Cytosolic calcium oscillations were visualized using a 20x water immersion objective (0.50 NA). Images were captured at 50 fps (20 ms exposure) on a sCMOS camera (Photometrics Prime 95B) controlled by Visiview Software (RRID:SCR_022546).

Full‐frame images were analyzed using ImageJ/FIJI and custom scripts in MATLAB (RRID: SCR_001622, The MathWorks Inc., Natick, Massachusetts). The data were corrected for photobleaching using a second‐degree polynomial equation fitted to the data at each time point, following previously published methods (Bjarneson and Petersen 1991; Nathalie et al. 2007). Fitted values were then subtracted from the raw data. To facilitate subsequent data analysis, traces were normalized by feature scaling ($\frac{x_{\mathit{ij}}-x_{\min }}{x_{\max }-x_{\min }}$), where x _max and x _min are the maximum and the minimum amplitudes of fluo‐4 intensity. Feature scaling was employed to enhance the spiking features of the dataset and facilitate comparison between replicates. Normalizing the data improves data analysis accuracy by mitigating the influence of varying scales. A semi‐automated feature extraction method was used to estimate the spiking amplitude and time profile. Feature extraction was based on two criteria within a 100 ms sliding window. We considered the point a peak when (1) the value is higher than the set threshold value and (2) the value is greater than its preceding value and succeeding value. This approach reduced noise and captured relative peak data across populations with varying baseline (Swain et al. 2018; Xu et al. 2024). The threshold was fixed for the whole population in an imaging series.

2.12. Statistical Analysis

In plate‐based experiments, each biological replicate is defined as an independent reproduction of the experimental protocol, that is, the generation of a separate embryonic cell culture. Multiple technical replicates and readouts from individual wells of cultured neurons on the same assay plate were averaged to reduce variability and instrumentation error. The values for each set of biological replicates were subjected to statistical analysis, and the number of biological replicates per experiment (denoted as n) is reported. The technical replicate for each biological replicate was plotted in the graph as a superplot to illustrate data dispersion and reproducibility (Lord et al. 2020). In these plots (e.g., Figures 1, 3 and 9), individual biological replicates are indicated as dark‐colored symbols, while the technical replicates within each biological replicate are indicated as light‐colored symbols. Data are presented as mean ± SD unless otherwise indicated. Data presentation and statistical tests were performed in Prism (RRID: SCR_002798, GraphPad 10.3, La Jolla, CA). Normality tests used the Shapiro–Wilk test. If data were normally distributed, statistical comparisons were performed using two‐tailed unpaired t‐tests. Otherwise, the Mann–Whitney unpaired nonparametric test was used. Comparisons between three or more groups were performed using the nonparametric Kruskal‐Wallis test followed by Dunn's test for multiple comparisons. No outlier tests were used. Statistical results are reported as p‐values with significance at p < 0.05, indicated by an asterisk in the figures with p‐value type represented as *p < 0.05, **p < 0.01, and ***p < 0.001. Measured p‐values, t‐values, and degrees of freedom are provided in the results with precise information on statistical significance. The statistical details of comparisons for all the data elaborated in the results are summarized in Table S1.

FIGURE 1.

Primary neuronal cultures in a defined low glucose medium are healthy. (a) Procedure for generating mouse primary neuronal cultures. (b) Neuronal culture health was monitored at 10–18 DIV in high glucose (HG) and low glucose (LG) conditions using alamarBlue. (c) Example images for neuron (MAP2, Green) and astrocyte (GFAP, Red) immunohistochemical labelling of neuronal cultures in 25 mM or 5 mM glucose media (Scale bars 100 μm). (d) Quantification of GFAP‐positive glial cells versus MAP2‐positive neurons in culture. Dark symbols represent biological replicates (n), while lighter symbols are individual measurements within each biological replicate. (e) Example images of cultures stained for vGAT puncta (Red; inhibitory synapse marker) compared to vGluT1 puncta (Green; excitatory synapse marker) that colocalized with the synapse active zone marker bassoon (not shown) and neuronal MAP2 marker (Gray). Scale bars are 50 μm. (f) Quantification of the excitatory to inhibitory synaptic ratio. Three to five images were taken per culture (light symbols); average ratios of biological replicates (dark symbols) are also shown. All data shown consist of n = 5 biological replicates. Mean ± SD are shown. Statistical comparisons were determined using two‐tailed unpaired t‐tests, *p < 0.05.

FIGURE 3.

Mitochondrial content in neurons maintained in high‐ and low‐glucose media. (a, b) Example immunofluorescence images of neurons cultured in HG and LG media stained for dendrite marker MAP2 (green), axon marker NF‐H (blue), and mitochondria marker MitoTracker Orange (magenta). (c–h) Summary data of neurons cultured in HG and LG medium at DIV 14. (c) Total dendritic mitochondrial content calculated as MitoTracker area divided by MAP2 area. (d) Total axonal mitochondrial content calculated as the area of MitoTracker divided by the area of NF‐H. (e) Mean area of a single dendritic mitochondrion. (f) Mean area of a single axonal mitochondrion. (g) Quantified number of dendritic mitochondria per 100 μm branch. (h) Quantified number of axonal mitochondria per 100 μm branch. All scale bars represent 50 μm. Data are Mean ± SD. Two‐tailed unpaired t‐test was used for statistical comparisons. Data shown in panels (c–h) are n = 10 images from each biological replicate for HG and LG (light symbols), with 5 biological replicates per condition (dark symbols). Statistical comparisons were determined using two‐tailed unpaired t‐tests, *p < 0.05.

FIGURE 9.

Network activity in hippocampal cultures using fluo‐4 Ca²⁺ imaging. (a) Example raster plots of somatic Ca²⁺ spiking patterns illustrating event detection corresponding to different metabolic challenges for neurons grown in HG. Each plot illustrates events collected from multiple cells (shown per line) in one field of view in separate experiments. (b) Example raster plot of somatic Ca²⁺ spiking patterns illustrating event detection corresponding to different metabolic challenges for neurons grown in LG. Each plot illustrates events collected from multiple cells (shown per line) in one field of view in separate experiments. (c) Summary data of somatic spiking event frequency. (d) Summary data of mean spiking amplitudes. Summary data shown are individual replicates (symbols), with bar plots showing mean ± SD. Each experiment used 25–80 cells per field, and n = 5 biological replicates per condition. Kruskal‐Wallis test was used for statistical comparisons, *p < 0.05.

3. Results

3.1. Primary Neuronal Cultures in Defined Low and High Glucose Media Are Healthy With Few Glial Cells

We attempted preliminary experiments in low glucose media to see if neuronal cultures were viable. Cells were collected from the cortex of mouse embryos, and pairwise cultures were maintained in standard culture medium or in culture medium with glucose concentrations of 2.5 and 5 mM (Figure 1a). Cultures with chronic glucose concentrations below 5 mM were not viable. We assessed neuronal health at DIV 10–18 using an alamarBlue‐based cell viability assay, which measured cellular reductive capacity (Figure 1b) and found that cells were similarly healthy when grown in 25 mM glucose (HG) and 5 mM glucose (LG).

Our media had no serum, which limits glial differentiation and growth in standard conditions (Brewer et al. 1993; Meberg and Miller 2003). This is useful for plate‐based assays where neuronal cell cultures with limited glial cell contamination can be used to interrogate neuronal bioenergetics. Neuronal (MAP2) and astrocyte (GFAP) markers were used to evaluate the proportion of neurons to glia in HG and LG culture conditions (Figure 1c). We observed a nearly exclusive neuronal population in both HG and LG conditions with minimal glial contamination (Figure 1d, 1.81% ± 1.02% glia in HG and 3.99% ± 0.91% in LG, Two‐tailed unpaired t‐test, **p = 0.007, Table S1), with a 2.2‐fold increase in glia percentage in LG. We also checked for the presence of non‐neuronal cells (e.g., MAP2‐ and GFAP‐negative cells) with the nuclei marker DAPI and observed less than 1% cells in a subset of stained coverslips (data not shown). Thus, embryonic cultures in 5 mM glucose media are nearly purely neuronal, providing a suitable in vitro plate‐based platform for assessing neuronal metabolism. We also quantified the ratio between excitatory and inhibitory synapses in our neuronal culture systems using vGluT1 and vGAT antibody staining (Figure 1e). The ratio of excitatory to inhibitory synapses was similar between conditions (Figure 1f; LG: 2.75 ± 0.61 and HG: 2.50 ± 0.89; Two‐tailed unpaired t‐test, p = 0.294). The overall number of excitatory and inhibitory synapses was not altered by glucose concentration in media (p > 0.05, One‐way ANOVA, data not shown). Therefore, mouse primary neuronal cultures showed a healthy neuronal population with minimal glial contamination in both HG and LG media.

3.2. Neurons Maintained in a Low Glucose Medium Show Morphological Alteration and Decreased Connectivity

To investigate potential changes in neuron morphology, we used whole‐cell voltage recordings to fill individual cells with Alexa 568 dye to illuminate the neurite branching pattern. Healthy DIV 14 cortical neurons that showed action potential spiking activity in voltage recordings were filled with dye for > 10 min and then imaged after the pipette was carefully retracted. Multiple cells were filled from several separate cultures, and each filled cell counted as a separate biological replicate. The marked cells were traced using the ImageJ/FIJI Simple Neurite Tracer plugin (Figure 2a). Sholl analysis showed the neurons have a similar number of intersections with 10 μm concentric circles in HG and LG cultures (Figure 2b; n = 24 cells in HG, n = 20 cells in LG, from 6 biological replicates in each condition). Neurons cultured in HG show 1.43‐fold higher total branch length (Figure 2c, 4141 ± 1760 μm in HG versus 2901 ± 2159 μm in LG; Mann–Whitney test, *p = 0.012) and 1.68‐fold higher total neurite number (Figure 2d, 35.6 ± 24.4 in HG, versus 21.2 ± 17.4 in LG; Mann–Whitney test, *p = 0.021) as compared to culture in LG despite a similar average branch length (Figure 2e, 156.1 ± 67.1 μm in HG versus 164.2 ± 90.7 μm in LG; Mann–Whitney test, p = 0.853) and longest branch length (Figure 2f, 624.8 ± 188.7 μm in HG versus 555.4 ± 238.6 μm in LG; Two‐tailed unpaired t‐test, p = 0.283). To complement these observations, we also measured the convex hull of these neurons; the area of the smallest polygon that can be drawn around the neuron's branches. The convex hull was not changed by glucose concentration (Figure 2g; 0.31 ± 0.17 mm² in HG versus 0.23 ± 0.16 mm² in LG; Mann–Whitney test, p = 0.093). A similar effect of increased branching in high glucose was also seen in a Sholl analysis using DIV 18 hippocampal neurons (data not shown).

FIGURE 2.

Neurons in a low glucose medium show morphological alteration and decreased neuronal branching. (a) Representative neuronal morphology tracing images of dye‐filled cortical neurons used for Sholl analysis in HG and LG media. (b) Sholl analysis profile of cortical neurons in LG and HG media. Mean (dark lines) and SD boundaries (lighter lines) are shown for each condition. (c–g) Summary data of neuronal morphology indicators for (c) total branch length, (d) number of branches, (e) average branch length, (f) longest branch of neuronal trees, and (g) convex hull area, used to estimate the coverage of neurites. All graphs show mean ± SD. Statistical comparisons were determined using two‐tailed unpaired t‐tests, *p < 0.05. All scale bars represent 300 μm. Data shown in panels (c–g) are n = 24 cells in HG and n = 20 cells in LG (light symbols), from 6 biological replicates in each condition.

3.3. Neurons Maintained in a Low Glucose Medium Show an Increase in Mitochondrial Content

To assess changes in total mitochondrial content in neurites, we calculated total mitochondrial area divided by total dendritic area (MAP2 positive compartment) or axonal area (neurofilament heavy chain, NF‐H positive compartment; Figure 3a,b). One hundred images from five independent sets of cultures in HG and LG conditions were analyzed. There was a 25% increase in dendritic mitochondrial content (Figure 3c, 0.1528 ± 0.0378 in HG versus 0.1914 ± 0.0385 in LG; Two‐tailed unpaired t‐test, *p = 0.017) and a 19% increase in axonal mitochondrial content (Figure 3d, 0.1300 ± 0.035 in HG versus 0.155 ± 0.030 in LG; Two‐tailed unpaired t‐test, *p = 0.028) in LG‐maintained cultures. The estimated size of individual dendritic mitochondrial particles was ~23% larger in LG‐cultured neuronal dendrites than in HG‐cultured dendrites (Figure 3e, 0.31 ± 0.05 μm² in HG versus 0.38 ± 0.11 μm² in LG; Two‐tailed unpaired t‐test, *p = 0.039). Individual axonal mitochondria were not significantly increased (Figure 3f, 0.25 ± 0.06 μm² in HG versus 0.26 ± 0.04 μm² in LG; Two‐tailed unpaired t‐test, p = 0.452). When looking at the area and number of individual mitochondria, each measured as a single particle, we found that neurons cultured in HG medium show a slightly increased density of mitochondria in dendrites, which was not significant (Figure 3g, 39.1 ± 4 mitochondrial particles/100 μm in HG versus 36.1 ± 4.9 mitochondrial particles/100 μm in LG; Two‐tailed unpaired t‐test, p = 0.095) and axons (Figure 3h, 35.3 ± 7.3 mitochondria per 100 μm in HG versus 31.7 ± 7.8 mitochondria per 100 μm in LG; Two‐tailed unpaired t‐test, p = 0.264). An increase in mitochondrial content in LG suggests OXPHOS may be upregulated. Detailed statistical comparisons are provided in Table S1.

3.4. Neurons Grown in Low Glucose Media Upregulate Pathways Associated With Oxidative Phosphorylation

To investigate the alterations occurring in energy‐generating pathways during incubation in high‐ and low‐glucose environments, we set out to identify differentially expressed genes at the levels of the transcriptome and proteome. Beginning with the transcriptome, we aggregated and sequenced mRNA from neuronal cultures grown in high‐ and low‐glucose media. Although we initially sequenced mRNA from five matched cultures, one sample was removed due to a low RNA integrity number and low quantification. When comparing the two datasets, we identified a total of 14 846 genes with 1871 differentially expressed genes (DEGs). DEGs were defined by having a > 0.6 ‐Log₂Fold change (LFC) and false‐discovery rate (FDR) adjusted p‐value (q) < 0.05 (Figure 4a; Table S2). Genes with positive LFCs showed an increased number of transcripts in neurons grown in high‐glucose and those with negative LFCs had an increased number of transcripts in neurons grown in low‐glucose. Among the transcriptomic data, we found the DEGs with the highest LFCs were associated with inflammatory response, such as Serpinb2 (LFC = 6.679, q = 0.002) and Il12b (LFC = 7.280, q < 0.0001), and were upregulated in the high‐glucose condition (Figure 4b). GAD1, an enzyme that generates the inhibitory neurotransmitter GABA, was differentially expressed (LFC = −0.845, q < 0.0001) and may indicate greater inhibitory transmission in the low glucose condition. We did not find a unifying theme among the highest expressed genes in the low‐glucose condition. We specifically looked for DEGs that are associated with glycolysis and OXPHOS and found several, generally seeing transcripts associated with glycolysis to be more common in neurons grown in high‐glucose and transcripts related with OXPHOS to be more common in neurons grown in low‐glucose (Figure 4c). Seeing this trend, we conducted a gene set enrichment analysis (GSEA) to find overrepresented pathways. Using hallmark gene sets in the Molecular Signatures database, we only considered pathways with q < 0.25 and p < 0.05 and the gene list was ranked by the LFCs for all identified genes. We did not find any significant pathways that were correlated with the low‐glucose condition; however, we found several associated with the high‐glucose condition. Upregulated pathways associated with the high‐glucose condition include those associated with inflammatory responses (TNK‐α signaling via NF‐κB: Normalized enrichment score (NES) = 3.08, q < 0.001, p < 0.001; Inflammatory Response: NES = 1.98, q = 0.009, p = 0.002), hypoxia (NES = 2.39, q < 0.001, p = 0.001), pathways involving cell proliferation and differentiation (G2M Checkpoint: NES = 2.24, q = 0.001, p < 0.001; E2F Targets: NES = 2.22, q < 0.001, p < 0.001), and apoptosis (P53 Pathway: NES = 1.69, q = 0.057, p = 0.022). Identification of these pathways suggests culturing in higher glucose causes increased neuronal stress and may promote cell death. Surprisingly, hallmark pathways associated with energy generation were not significant (Glycolysis: NES = 1.39, q = 0.146, p = 0.124; Fatty Acid Metabolism: NES = −0.69, q = 1, p = 0.818).

FIGURE 4.

Transcriptomic and proteomic data show altered gene expression between high‐ and low‐glucose culture conditions. (a) Volcano plot depicting 1871 differentially expressed transcripts out of the 14,846 measured via RNA‐seq (Thresholds: 0.6 Log₂ (LFC); FDR adjusted p‐value (q) ≤ 0.05; n = 4). HG‐associated genes are shown in blue, and LG‐associated genes are shown in green. Genes with a ‐log (adjusted p‐value) > 20 appear at the top of the graph. (b) Top 10 upregulated and downregulated genes in the transcriptome. HG‐associated genes are shown in blue, and LG‐associated genes are shown in green. (c) A selected group of differentially expressed genes that are part of glycolysis and OXPHOS pathways. HG‐associated genes are shown in blue, and LG‐associated genes are shown in green. (d) Volcano plot depicting 955 differentially expressed proteins out of the 6546 detected via mass spectrometry (Thresholds: 0.6 Log₂(LFC); FDR adjusted p‐value (q) ≤ 0.05; n = 5). HG‐associated genes are shown in blue, and LG‐associated genes are shown in green. (e) Top 10 upregulated and downregulated genes in the proteome. HG‐associated genes are shown in blue, and LG‐associated genes are shown in green. (f) A selected group of differentially expressed genes that are part of glycolysis and OXPHOS pathways. HG‐associated genes are shown in blue, and LG‐associated genes are shown in green.

In separate experiments, we identified proteins in lysates from cultures grown in high and low glucose media using mass spectrometry. We identified a total of 6546 proteins with 955 DEGs (Figure 4d; Table S2). Among the top DEGs within the dataset were several genes with disparate functions. However, genes associated with signaling pathways were upregulated in the high glucose condition, and several genes related to differentiation and glial cell maturation seemed to be upregulated in the low glucose condition (Figure 4e). Genes associated with mitochondrial function (TFAM: LFC = −0.785, q = 0.023; SDHB: LFC = −0.639, q = 0.015) and excitatory and inhibitory signaling (GAD1; LFC = −0.990, q = 0.009; GLS1: LFC = 0.741, q = 0.015) were found. Like the RNA dataset, we found a few glycolytic proteins were more abundant in the high glucose condition, while OXPHOS and TCA cycle proteins were increased in the low glucose condition (Figure 4f). Pkfl (Transcript: LFC = 0.799, q < 0.001; Protein: LFC = 0.876, q = 0.028) and Pgm2l1 (Transcript: LFC = 0.779, q = 0.007; Protein: LFC = 1.323, q < 0.001) show upregulation in both protein and transcript datasets in high glucose. Unexpectedly, Hk2 (LFC = −1.332, q = 0.012) was seen at lower expression levels for neurons grown in high glucose. After conducting a GSEA on this dataset, we did not find any significant hallmark pathways associated with the high glucose condition. However, GSEA of proteins in low glucose showed significant association with OXPHOS (NES = −2.33, q = 0.006, p < 0.001), fatty acid metabolism (NES = −2.08, q = 0.014, p = 0.002), adipogenesis (NES = −1.95, q = 0.024, p < 0.001), and reactive oxygen species pathways (NES = −1.95, q = 0.018, p = 0.002). These results suggest an increased reliance on OXPHOS in the low glucose condition.

We found that there was some overlap between the DEGs that were seen between proteomic and expression analyses, with 144 DEGs differentially expressed in both datasets using iPathwayGuide (Figure 5a; Table S3). These genes show a weak but overall positive correlation between the LFCs in the transcript and protein datasets (R ² = 0.55), with 15 genes having opposite LFCs (Figure 5b). Using the LFCs from the transcriptomic data, we separated the genes into lists of upregulated in high glucose (76 genes) and upregulated in low glucose (68 genes) for analysis (Table S4). Several gene ontologies associated with hyperglycemic conditions were identified, affecting Ca²⁺ signaling (GO:0051592; q = 0.035) and GTPase signaling (GO:0007264; q = 0.045; Figure 5c). We also saw gene ontologies associated with neurotransmission in high‐glucose cultures, such as positive regulation of synaptic transmission (GO:0050806; q = 0.164), neurotransmitter receptor localization (GO:0099645; q = 0.113), and synapse organization (GO:0050808; q = 0.231). Gene ontologies and pathways associated with the low‐glucose condition affected neurotransmitter transport (GO:0006836; q = 0.161), amino acid transport (GO:0006865; q = 0.096), and GABAergic signaling (mmu04727; q = 0.232; Figure 5d).

FIGURE 5.

Gene Ontology analysis for genes differentially expressed in the transcriptome and the proteome. (a) 144 genes were differentially expressed in both the proteome and transcriptome. (b) Linear regression of the changes seen in the 144 differentially expressed genes in both datasets. (c) Gene ontology enrichment analyses showing biological processes associated with 76 genes upregulated in high glucose (LFC > 0) within the 144 differentially expressed genes in both datasets. Pathway and GO terms displayed are the top hits from term clusters created by Metascape. (d) Gene ontology enrichment analyses showing biological processes associated with 68 genes upregulated in low glucose (LFC < 0) within the 144 differentially expressed genes in both datasets. GO terms displayed are the top hits from term clusters created by Metascape.

3.5. Neurons Maintained in a Low Glucose Medium Show ATP Maintenance Biased Towards Oxidative Phosphorylation

Glucose is a primary energy source in the brain, and OXPHOS efficiently fulfills neuron energy requirements at rest and on demand during synaptic activity (Ashrafi and Ryan 2017). We anticipated that growing neurons in 25 mM versus 5 mM glucose may affect the bioenergetic profile underlying cellular ATP production and maintenance. Increased mitochondrial particle size in LG in Figure 3 and expression analysis described in Figure 4 suggest that mitochondrial OXPHOS may be increased in low glucose culture conditions. To address this, we measured the ATP content in mature neurons at DIV 14 with media containing four energy‐substrate conditions to assess the metabolic profile of individual pathways (Figure 6a). These conditions are described in Materials and Methods and include (I) positive control (glucose and pyruvate), (II) OXPHOS‐only condition (no glucose, plus iodoacetic acid and pyruvate), (III) Glycolysis‐only condition (no pyruvate, plus 5‐ or 25‐mM glucose and oligomycin), and (IV) negative control (no glucose, no pyruvate, plus iodoacetic acid and oligomycin).

FIGURE 6.

Neurons maintained in a low glucose medium show ATP maintenance biased towards oxidative phosphorylation. (a) Schematic diagram for the ATP assay performed to evaluate the metabolic activity in HG versus LG. (b) ATP content of neurons grown in HG, subjected to media for various periods before processing. Positive control had multiple energy substrates (yellow line and markers), and energetic pathways were limited to glycolysis‐only conditions (blue line and markers), mitochondrial OXPHOS only (brown line and markers), and compared to negative control (no energy substrates, black line and markers). Hatched box shows time point used for statistical comparison. (c) ATP content of neurons grown in LG. Markers are the same as defined in panel (b). All the data shown as Mean ± SD are from n = 5 matched biological replicates.

Our results show that ATP is maintained for up to 20 min in these end‐point assay conditions if energy substrates are available, but quickly dissipates if no substrates are present (negative controls in Figure 6b,c), suggesting rapid and continual ATP flux. When glycolysis and OXPHOS substrates were present (glucose and pyruvate) ATP was equally maintained in either HG or LG growth conditions, with ATP concentration slightly elevated in LG (15‐min time point, 8.29% increase in LG). However, ATP maintenance was biased towards glycolysis in neuronal cultures grown in HG, measured as a percentage of ATP measured in positive controls (Figure 6b; Glycolysis only: 68.18% and OXPHOS only: 32.56% at 15‐min time point). Conversely, neurons cultured in LG show increased OXPHOS‐dependent energy production (OXPHOS only: 92.26%, a 3.01‐fold increase compared to HG; Mann–Whitney test, **p = 0.008) and less glycolysis (Glycolysis only: 44.48% a 1.44‐fold decrease as compared to HG; Two‐tailed unpaired t‐test, **p = 0.002; Figure 6c). Statistical comparisons are presented in Table S1.

Even though OXPHOS produces 30–32 molecules of ATP and glycolysis yields only 2 molecules of ATP, glycolysis was the dominant pathway for energy maintenance in “standard” high‐glucose conditions, suggesting massive upregulation in glycolysis and/or inhibition of mitochondrial OXPHOS in these cultures. This result suggests that neuronal cultures based on 25 mM glucose do not accurately follow the in vivo bioenergetic profile (Hall et al. 2012; Harris et al. 2012; Hyder et al. 2013), which may complicate the interpretation of neurometabolic studies in these high glucose culture conditions.

3.6. Neurons Maintained in a Low Glucose Medium Show an Increase in Mitochondrial Respiration

Using the Seahorse extracellular flux analyzer, we assessed the mitochondrial and glycolytic respiration of neurons maintained in HG and LG media at DIV 14 as an orthogonal approach to monitor ATP production by these two pathways (Figure 7). The oxygen consumption rate (OCR) was used as a proxy for mitochondrial respiration. Baseline respiration and estimates of maximal and reserve respiratory capacity were measured as illustrated in Figure 7a. Summary data for oxygen consumption rate (OCR) are presented in Figure 7b. Neurons maintained in the LG medium exhibited a significant increase in basal mitochondrial respiration (Figure S1a) and ATP‐linked respiration (Figure S1b) compared to those in the HG medium. Moreover, neurons cultured in LG exhibit increased maximal mitochondrial respiration (Figure S1c) and a greater spare capacity for OXPHOS (Figure S1d) compared to those grown in HG. An increase in basal respiration, mitochondrial ATP‐linked respiration, and spare capacity in LG‐mediated culture indicates increased dependence on OXPHOS. Extracellular acidification rates (ECAR) were measured in the same experiments and used to track glycolysis (Figure 7c,d). Glycolysis rate and capacity in these experiments returned reciprocal results: basal glycolysis (Figure S1e), maximum glycolysis capacity (Figure S1f), and reserve glycolysis capacity (Figure S1g) were all significantly increased in cultures grown in HG compared to LG media. All the data for OCR and ECAR are provided in Table S5, and the statistical comparisons are presented in Table S1.

FIGURE 7.

Neurons maintained in a low glucose medium show an increase in mitochondrial respiration. (a) Experimental protocol for oxygen consumption rate (OCR). (b) Summary time‐course OCR graph of neurons cultured in HG and LG medium at DIV 14. (c) Experimental protocol illustrating extracellular acidification rate (ECAR) data obtained from the same experiments. (d) Summary time‐course ECAR graph of neurons cultured in HG and LG medium at DIV 14. (e) Visualization of bioenergetic phenotypes represented as stacked columns summing to total J_ATP production for neurons cultured in HG and LG media. (f) Bioenergetic space plot illustrating the metabolic difference between neurons cultured in high glucose (HG) and low glucose (LG) media. The slope of dotted lines through each point denotes J_ATP proportionality. The solid line indicates slope = 1 with GI index = 50%, indicating equal J_ATP production by glycolysis and OXPHOS. The shaded area (slope < 1) indicates a bias toward glycolytic respiration. All graphs show mean ± SD for n = 5 biological experiments.

The total rate of ATP production by glycolysis (J_ATP,glyc) and oxidative phosphorylation (J_ATP,ox) was calculated from the oxygen consumption rate (OCR) and extracellular acidification rate (ECAR), as described in Mookerjee et al. (2017). Consistent with the data above, aggregated ATP production by glycolysis at baseline was higher in HG, whereas ATP production by OXPHOS was higher in the LG condition (Figure 7e, Table S6). We assessed the maximum rate of ATP production in both HG and LG media cultures using DNP as an uncoupler for mitochondrial respiration. In HG media, glycolysis was responsible for most of the calculated ATP production. Conversely, in LG‐media cultures, oxidative phosphorylation (OXPHOS) contributed predominantly to maximum ATP production. The rate of ATP production by glycolysis and oxidative phosphorylation are provided in Table S5, and the statistical comparisons are presented in Table S1.

To better illustrate the relationship between the two bioenergetic pathways in both media conditions, we plotted the relative rate of ATP production from both pathways as glycolytic index (GI) in a bioenergetic space plot (Figure 7f), providing a summarized depiction of the cellular preference of the respiration pathway between glycolysis and OXPHOS in different conditions. The GI is derived from $\mathrm{GI}={\mathrm{J}}_{\mathrm{ATP},\text{glyc}}/{\mathrm{J}}_{\text{Total}\mathrm{ATP}}\times 100$, where GI = 0% when ATP is derived completely from OXPHOS, and GI = 100% when ATP is entirely derived from glycolysis. Neurons in LG medium prefer OXPHOS (GI = 23.5% at baseline and 30.8% in maximum respiration), whereas neurons grown in HG medium prefer glycolysis (GI = 57.8% at baseline and 65.1% in maximum respiration). These findings are consistent with expression data and the ATP luciferase assay above and suggest that mitochondrial respiration is suppressed in HG, whereas glycolytic activity is strongly enhanced.

3.7. Neurons Maintained in BrainPhys Medium Show Maximal Mitochondrial Respiration and Elevated Glycolysis

A defined BrainPhys media (BP) was developed recently to improve cultures of differentiated iPSCs and mouse primary neurons (Bardy et al. 2015; Faria‐Pereira et al. 2022). The media formulation was made to mimic physiological concentrations of salts and amino acids in the brain and spinal fluid. Further, it was designed to permit both cell culture and electrophysiological recording without the need to change media conditions. Germane to our study of the effects of glucose on neuronal respiration, BP contains 2.5 mM glucose at the time of assay, but cells are plated in a much higher glucose concentration (25 mM) and slowly weaned off hyperglycemic conditions over 2 weeks during media changes. Recently, BP was tested on primary neuronal mouse cultures to investigate metabolic activity and was shown to more closely mimic in vivo energetics (Faria‐Pereira et al. 2022; Rumpf et al. 2023). In the previous study by Pereira, et al., no differences were observed in neuronal health and glial content compared to standard media. Here, we cultured primary mouse neurons in BP following conditions described by Pereira, et al. to directly compare their metabolic profile with our results in LG and standard HG media conditions.

First, we performed the ATP assay at DIV 14 under four different energy‐substrate conditions to dissect ATP maintenance pathways, shown in Figure 8. ATP levels on BP with both substrates available were similar to HG and LG conditions (positive control; Kruskal‐Wallis test; HG versus BP: p = 0.29; LG vs. BP: p = 0.99). Interestingly, we found ATP levels were largely unperturbed in BP‐cultured cells when either single fuel source was available (Glycolysis only: 82.96% and OXPHOS only: 87.52% of BP ATP measured at 15‐min time point; Mann–Whitney Test; p = 0.151; Figure 8a). This profile differed from those in the HG and LG conditions shown in Figure 6 above. In glycolysis‐only conditions, BP was similar to HG media and elevated relative to LG media (HG vs. BP: p = 0.13, LG vs. BP: **p = 0.001, Kruskal‐Wallis test). In OXPHOS‐only conditions, BP was more similar to LG media (HG vs. BP: *p = 0.03, LG vs. BP: p = 0.29, Kruskal–Wallis test).

FIGURE 8.

Neurons maintained in BrainPhys medium show elevated OXPHOS and glycolytic activity. (a) ATP profile of neurons grown in BP shows either OXPHOS or glycolysis energy production can maintain ATP levels. (b) Summary time‐course OCR graph of neurons cultured in BP, HG, and LG medium at DIV 14. (c) Summary time‐course ECAR graph of neurons cultured in BP, HG, and LG medium at DIV 14. (d) Visualization of bioenergetic phenotypes represented as stacked columns summing to total J_ATP production for neurons cultured in HG, LG, and BP media. (e) Bioenergetic space plot illustrating the metabolic difference between neurons cultured in HG, LG, and BP media. The slope of dotted lines through each point denotes J_ATP proportionality. The solid line indicates slope = 1 (GI index = 50%), indicating equal J_ATP production by glycolysis and OXPHOS. The shaded area (slope < 1) indicates a bias towards glycolytic respiration. All graphs show mean ± SD for n = 5 biological experiments.

Second, mitochondrial respiration in BP was examined at DIV 14 using the Seahorse metabolic flux analyzer and compared to HG and LG growth media conditions (Figure 8b). The results for HG and LG media are replotted from Figure 7. Basal mitochondrial respiration rate (OCR) was elevated relative to HG and matched LG levels (Figure S2a). ATP‐linked respiration was also elevated, like LG (Figure S2b). However, BP exhibited maximum respiratory rates equivalent to HG when challenged with the proton uncoupler 2,4‐dinitrophenol (DNP; Figure S2c) which resulted in very limited (potentially negative) estimates for reserve respiratory capacity (Figure S2d). We also examined glycolytic respiration rate (ECAR) in neurons grown in BP media. Neurons maintained in BP show elevated glycolytic activity equivalent to HG‐cultured neurons (Figure 8c). Basal glycolysis was like HG culture conditions and elevated relative to LG (Figure S2e). Maximum glycolysis capacity (Figure S2f) and reserve glycolysis capacity (Figure S2g) were also like those of HG and elevated relative to those of LG.

J_ATP,glyc and J_ATP,ox were calculated from OCR and ECAR data (Figure 8d). Interestingly, we observed equivalent respiration in both glycolysis and OXPHOS, but glycolysis attained maximum ATP production in BP (Table S6). Subsequently, neuronal cultures in BP had GI = 51.5%, suggesting equal reliance on glycolysis and OXPHOS in a bioenergetic space plot (Figure 8e). Taken together, these results suggest that neurons maintained in BP have energy‐producing metabolism that functions at maximum capacity and does not discriminate between fuel sources. Neurons grown in BP may be losing function in the Seahorse media, complicating the interpretation of maximum mitochondrial respiratory capacity and mitochondrial reserve capacity, as respiration seems to decline during the assay (red line in Figure 8b, note respiration in DNP is not elevated to baseline), similar to prior reports of neuronal respiration in BP (Faria‐Pereira et al. 2022). All the data for OCR and ECAR and the rate of ATP production cultured in BP media were provided in Tables S5 and S6, and the statistical comparisons are presented in Table S1.

3.8. Neuronal Local Network Activity Parallels Energetic Profiles in Low Glucose Media

Intracellular Ca²⁺ is both an energy burden and a feedforward signal, coordinating rapid metabolic responses to increased neuronal activity. During activity, Na⁺ and Ca²⁺ enter the cell via voltage‐gated ion channels. ATP‐dependent pumping to extrude these ions dynamically and substantially increases ATP consumption. ATP is continually replenished by OXPHOS and glycolysis (Ashrafi and Ryan 2017; Hall et al. 2012; Rangaraju et al. 2014). Somatic calcium imaging was performed using fluo‐4 AM to assess neuronal activity when neurons were supplied with (I) both glucose and pyruvate (positive control), (II) OXPHOS‐only (no glucose, plus iodoacetic acid and pyruvate), (III) glycolysis‐only (no pyruvate, plus glucose and oligomycin) and (IV) negative control (no pyruvate or glucose, plus iodoacetic acid and oligomycin) conditions for hippocampal cultures maintained in HG and LG, at DIV 18 (Figure 9). Pre‐incubation with bicuculine and strychnine was used to block inhibitory activity and increase network activity, necessary to validate the presence of local synaptic networks (Samoilova et al. 2003; Xu et al. 2024). Thus, only spontaneous excitatory activity is present in these cultures during the imaging experiment.

Neurons grown in both HG and LG media show easily observable spiking events in positive control buffer conditions. Examples of raster plots for detected events (see Section 2) are shown for all four energy‐substrate conditions in HG and LG media (Figure 9a,b). Somatic spike events in positive control conditions, glycolysis only, and negative controlled conditions are not significantly altered between HG and LG conditions (Figure 9c, Kruskal‐Wallis test; positive control: p = 0.372, glycolysis only: p = 0.110, negative control: p = 0.745; Table S1). Interestingly, a significant increase in spiking frequency (Kruskal‐Wallis test; *p = 0.021) was observed in neurons grown in LG relative to HG when OXPHOS was the only major energy‐generating pathway present. No significant difference in mean event amplitude was observed between HG and LG across conditions (Figure 9d, Kruskal–Wallis test; Positive control: p = 0.203, OXPHOS Only: p = 0.093, Glycolysis Only: p = 0.358, Negative control: p = 0.646; Table S1), suggesting spike‐dependent Ca²⁺ dynamics were not altered.

Increased activity in OXPHOS‐only conditions for LG‐maintained neurons may be interpreted as reflecting higher energetic capability due to increased bias towards mitochondrial respiration and was mirrored by increased activity rates in glycolysis‐only conditions in HG, though not significant. These results of neuronal network function parallel ATP maintenance and respiration experiments described above, where neurons raised in culture media with low glucose show increased reliance on OXPHOS.

4. Discussion

We systematically compared the morphological and physiological environment of primary neurons cultured in either 5 mM or in 25 mM glucose with commercially available cell culture media (Thermofisher Neurobasal supplemented with B27) (Brewer et al. 1993). In both high‐ and low‐glucose media, we observed a healthy and almost exclusively neuronal population of cells, with limited glial cell contamination, up to DIV 14 in cortical cultures. Neurons maintained in low‐glucose media showed increased mitochondrial OXPHOS capacity and a more balanced reliance on OXPHOS relative to neurons grown in high‐glucose media, which heavily relied on glycolysis to maintain cytosolic ATP. This main finding suggests that low‐glucose media can be used to generate neuronal cultures with in vivo‐like respiratory profiles. In brief, we suggest that low‐glucose (LG) culture conditions provide a more physiologically relevant model for studying neuronal metabolism than the standard medium (Faria‐Pereira et al. 2022; Kleman et al. 2008). Our results challenge the validity of existing neuronal culture systems to represent physiologically relevant neuronal respiration and suggest that evidence on neuronal metabolism up to this point may require careful review, as hyperglycemic cultures may lead to aberrant interpretations. Specifically, that neurons grown in standard media grossly underestimate their endogenous reliance on mitochondrial respiration.

4.1. Standard Neuronal Cultures Mimic Hyperglycemia

When comparing our results in HG versus LG conditions, many HG phenotypes are reminiscent of hyperglycemia in vivo. Differences in routes of energy production were the result of modulating OXPHOS and the TCA cycle, supported by changes in pathways at the transcriptome and proteome. This is consistent with previous work showing hyperglycemia‐mediated alterations lead to decreased mitochondrial respiration and reduce oxidative phosphorylation efficiency in non‐neuronal cells (Audzeyenka et al. 2021). Like hyperglycemia, high‐glucose culture conditions suppressed mitochondrial activity and OXPHOS production of ATP, which was relieved when neurons were cultured in lower glucose conditions (Kleman et al. 2008; Wang et al. 2024). Our results suggest that the size of local cortical and hippocampal branching complexity may be aberrantly increased in high‐glucose culture conditions. Although it has been observed in many prior works that hyperglycemic environments decrease neuronal network complexity and outgrowth in adults (Ferreiro et al. 2020; Salazar‐García et al. 2024), other investigations reported an increase in neurite growth in high‐glucose conditions consistent with our results (Rao et al. 2018; Sharma et al. 2019). Our results are also consistent with previous results that showed decreased mitochondrial content and OXPHOS in HG, suggesting increased neuronal branching may be regulated by AMP‐activated protein kinase AMPK and the downstream target phosphoinositide 3‐kinase PI3K (Gioran et al. 2014; Kleman et al. 2008). Finally, our high glucose results also increase inflammation, as seen in hyperglycemia. Although the exact mechanism is unknown, high glucose is associated with cognitive decline and may increase the risk of neurodegenerative disease and poor outcomes after injury (Gupta et al. 2023; Madhusudhanan et al. 2020; Martín et al. 2006). Altering glucose levels may activate microglia and astrocytes, promoting degradation, or promote apoptosis in the neurons themselves, based on the significant association between high‐glucose culture conditions and the P53 apoptotic pathway (Gupta et al. 2023; Hsieh et al. 2019; Lee et al. 2024).

4.2. Control of Neuronal Respiration

Low‐glucose media leads to an upregulation of genes associated with OXPHOS and the TCA cycle, leading to the conclusion that higher prioritization of OXPHOS occurs when exposed to a more limited supply of glucose, upregulating genes involved with the TCA cycle primarily. The magnitude of changes in expression within these pathways is also relatively small, implying that OXPHOS and glycolysis output are sensitive to small changes in gene expression, and inherent tight regulation of enzyme activity in these pathways is sufficient to tune respiration. Additionally, in hyperglycemic conditions (which can reach up to 20 mM plasma glucose in vivo), neurons show an increase in small, dysfunctional mitochondria despite a higher mitochondrial count, a hallmark of diabetic neuropathy (Haigh et al. 2020).

The absence of glycolytic processes from our GO analysis and GSEA is conspicuous. This suggests that changes to glycolytic output are regulated downstream of transcription and translation, as neurons grown in high‐glucose media did show increased glycolysis rates. These changes could be post‐translational modifications, protein localization, or enzymatic inhibition or facilitation. Previous reports also suggest glycolysis can be rapidly accelerated, a clear benefit where activity rates are highly dynamic (Connolly et al. 2014; Díaz‐García et al. 2017). Put simply, this suggests that mitochondrial respiratory pathway gene expression is modulated in neurons by environmental factors, whereas glycolytic gene expression is not, and glycolysis is regulated by other means.

More investigation is required to understand the temporal effect of increased glucose concentration: do the effects we observed depend on an early and/or prolonged introduction to a specific glycemic environment, or can they be replicated acutely in mature neurons? Developmental studies suggest that a shift to increased glycolysis and increased mitochondrial biogenesis acts as a checkpoint in neuronal maturation; however, these studies were conducted in standard (hyperglycemic) culture conditions (Agostini et al. 2016). Recent work suggests that metabolism changes occur within 72 h (Wang et al. 2024). Our results suggest remodeling takes more than 2 h, as metabolism was not normalized in 10 mM glucose during incubation prior to Seahorse cellular respiration assays. A fine‐scale understanding of neuronal metabolic remodeling will be informative.

4.3. Glucose Controls Neuronal Differentiation and Morphology

To sustain ATP production to meet the energy demand for neuronal activity (Rangaraju et al. 2014) and to protect against oxidative damage (Wei et al. 2023) glycolysis plays a vital role in neuron development in HG culture conditions, at the expense of normal mitochondrial respiration and function. Previous studies have clearly indicated that a balanced metabolic regulation is necessary in neurogenesis (Rumpf et al. 2023). Our results suggest HG cultures suppress neuronal differentiation and maturation. Glycolysis byproducts may result in the accumulation of NAD+ and lower pH, which influence neuronal differentiation (Huang et al. 2022). Conversely, an increase in lactate as a glycolytic byproduct may enhance neuronal differentiation by shunting ATP production towards the pentose phosphate pathway (Pötzsch et al. 2021; Xu et al. 2023). In our study, DEGs associated with neuronal differentiation, like TFAM and SDHB, showed an increase in expression in neurons grown in low‐glucose media. In cultured cortical neurons, SDHB, a subunit of the electron transport chain, is upregulated during neuronal differentiation as neurons increase their usage of OXPHOS and generate more mitochondria (Agostini et al. 2016). Similarly, mitochondrial biogenesis during development is regulated by the mitochondrial transcription factor TFAM (Agostini et al. 2016). Other markers for neural differentiation, increased production of GABA and glutamate, are dysregulated (Agostini et al. 2016). GAD1 is upregulated in low glucose on both the transcript and protein levels, but GLS1 protein is more highly expressed in high glucose, suggesting increased excitatory signaling when neurons are exposed to high glucose. We did not see differences in inhibitory and excitatory synapses via immunostaining for vGAT and vGluT (Figure 1f); however, the immunostaining does not necessarily indicate there are no changes in excitatory and inhibitory signaling. Previous work has demonstrated that changes in GAD1 transcripts could serve as a proxy for altered GABAergic signaling, showing that low expression of GAD1 transcripts coincided with low GABA release due to a lack of GABA production (Dicken et al. 2015; Lau and Murthy 2012). Though not examined here, our data predict low‐glucose conditions induce a higher amount of inhibitory signaling in comparison to high‐glucose conditions. Conversely, we saw an increase in astrocytic markers in low‐glucose media, doubling the proportion of GFAP+ glial cells in low‐glucose media (to 4% of cell population). This was complemented by an increase in transcripts associated with oligodendrocyte precursor cells in low‐glucose conditions. Taken together, these results suggest that both glial survival and neuronal differentiation are enhanced by low‐glucose media. Whether an increased astrocytic population (though proportionally small) is directing neuronal survival in near‐euglycemic levels of glucose in culture can be pursued in future studies.

4.4. Alternative Culture Media

More recently, alternative physiological‐like culture media has been developed, like commercially available BrainPhys media (Bardy et al. 2015). Neurons maintained in this medium have been shown to grow as well as, or better than, standard media, with increased survival and synaptic contacts (Faria‐Pereira et al. 2022). We tested this media system in direct comparison to our HG and LG conditions. Neurons grown in BP media showed an increase in both OXPHOS and glycolysis, showing that these two pathways can be independently regulated, and elevation of one does not necessarily mean inhibition of the other. On the other hand, we note a lack of spare respiratory capacity for mitochondrial OXPHOS, in line with a previous investigation (Faria‐Pereira et al. 2022). This suggests that neurons grown in BP are functioning at their maximum respiratory capacity, providing limited availability of extra energy production, for example, during bouts of activity. Neurons running at peak mitochondrial respiratory capacity are likely to increase the formation of ROS and superoxides and potentially lead to longer‐term cellular damage (Kowalczyk et al. 2021; Turrens 2003). Moreover, an elevated pyruvate concentration in BP media (0.5 mM versus 0.22 mM in standard media) may enhance its conversion to lactate, potentially shifting cellular metabolism towards glycolysis, which could be a probable cause for the decrease in maximum respiratory capacity (Zheng et al. 2016). Neurons grown in BP were healthy in culture, but showed increased respiratory loss during seahorse assays, suggesting they may be less resilient to acute alterations in their environment.

4.5. Summary and Next Steps

This study set out to develop a plate‐based assay platform for primary neuronal culture that more closely mimics neuronal energetics found in vivo. We find that neurons can be cultured in near‐physiological glucose media, down to ~5 mM, with similar survival and morphology as standard hyperglycemic conditions. Lowered glucose conditions show changes in gene and protein expression to facilitate respiration dependent on mitochondrial OXPHOS and decreased inflammatory markers. We propose that these differences may open an experimental window to improve modeling and understanding of normal neuronal metabolism and diseases that impact neuronal mitochondrial function and respiration.

Author Contributions

Sarpras Swain: conceptualization, formal analysis, visualization, writing – original draft, methodology, investigation, supervision, writing – review and editing, software. David M. Roberts: formal analysis, visualization, investigation, methodology, data curation, conceptualization. Saad Chowdhry: conceptualization, formal analysis, data curation, visualization, methodology, investigation. Ryan Durbin: conceptualization, data curation, formal analysis, visualization, writing – original draft, writing – review and editing, methodology, investigation. Reece Boyd: conceptualization, formal analysis, visualization, methodology, investigation, data curation. Juli Petereit: data curation, formal analysis, writing – review and editing, supervision, software, resources. Robert Renden: visualization, writing – original draft, investigation, supervision, project administration, writing – review and editing, validation, funding acquisition.

Conflicts of Interest

The authors declare no conflicts of interest.

Peer Review

The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer‐review/10.1111/jnc.70125.

Supporting information

Table S1.

Table S5.

Table S6.

Figure S1.

Figure S2.

Table S2.

Table S3.

Table S4.

Swain, S. , Roberts D. M., Chowdhry S., et al. 2025. “Low‐Glucose Culture Conditions Bias Neuronal Energetics Towards Oxidative Phosphorylation.” Journal of Neurochemistry 169, no. 6: e70125. 10.1111/jnc.70125.

Data Availability Statement

The RNA‐sequencing data that support the findings of this study are openly available in BioProjectID at https://www.ncbi.nlm.nih.gov/bioproject/, reference number PRJNA1185784. The proteomics data that support the findings of this study are openly available in EMBL PRIDE database at https://www.ebi.ac.uk/pride/, reference number PXD059008. All other additional data are available from the authors on request.