Frataxin deficiency shifts metabolism to promote reactive microglia via glucose catabolism

FXN deficiency leads to a loss of homeostatic function in cerebellar microglia of knock-in knock-out (KIKO) mice

Although genetic deficiency of the mitochondrial protein FXN is causative of FRDA-related symptoms, the disease-specific cell types in cerebellum are unknown yet. Herein we used a comparative single-cell RNA sequencing (scRNA-seq) between cerebellum of WT and KIKO mice at the early stage of FRDA disease (6 mo old). ScRNA-seq analysis led to the identification of a total of 4,484 quality control (QC)-positive cells. Based on the expression levels of the most variable genes, we annotated homogeneous and robust cluster of cells from scRNA-seq data, resulting in 12 group of cells such as four cluster of endothelial cells (EC), endo-mural cells (EMC), choroid cells, microglia, T cells, pericytes, fibroblasts, astrocytes, and oligodendrocytes (Fig 1A and B). Next, we compared the numerosity of cell clusters between genotype and we found a contraction of choroid plexus cells, astrocytes, and oligodendrocytes (Fig 1C). Oppositely, increased markers of microglia/macrophage population in cerebellum of KIKO were observed (Figs 1B and S1A). Consistently, the gene ontology (GO) terms for biological processes revealed an enrichment in the inflammatory process and down-regulation of mitochondrial oxidative genes (Fig 1D).

Microglia are the primary innate immune cells of the central nervous system (CNS) that are sentinels participating in the inflammatory and cell clearance response (Norris & Kipnis, 2019). To corroborate the microglia dynamics of KIKO mice, we performed a high dimensional flow cytometry analysis (Fig 2A) and we observed that although the percentage of cerebellar CD45^lowCD11b⁺ microglia remained preserved, a higher expression of M1-like markers (CD86⁺ and MHC-II) and a tendency to a lower expression of M2-like marker CD206 was observed in microglial cells of KIKO compared with WT (Fig 2B). No changes were observed in the percentages of neutrophils (CD45⁺/CD11b⁻/CD3⁻/NK1.1⁺/CD90.2⁻ cells) and B cells (CD45⁺/CD11b⁻/BB220⁺/Ly6G⁺ cells), whereas a significant increase was detected in T cells (CD45⁺/CD11b⁻/BB220⁻/Ly6G⁻/CD3⁺/CD90.2⁺ cells) (Fig S1B). Next, to explore the molecular signatures of macrophage/microglia, cerebella CD45⁺/CD11b⁺ cells were isolated by magnetic cell sorting and their transcriptome was profiled by bulk RNA sequencing. We identified n = 1,275 differentially expressed genes (DEGs) (−0.75 > Log₂FC > +0.75; P < 0.05) between cerebella CD45+/CD11b⁺ of KIKO versus WT mice (Fig 2C). GO terms for biological processes of the top 200 down-regulated genes (orange bars) revealed a reduced fatty acid metabolism in microglia/macrophages of KIKO mice (Fig 2D) leading us to suppose a reduced mitochondrial functionality. In an opposite manner, the top 200 up-regulated genes (green bars) pertained to enhanced reactivity to inflammatory stimuli such as LPS, IL1, and TNFa (Fig 2E). Subsequently, we labeled macrophage/microglia in the cerebellum of KIKO mice with CD11b and analyzed cell morphology using the Sholl technique (Ristanovic et al, 2006). Morphological analysis revealed that CD11b⁺ cells in KIKO mice exhibited longer ramification lengths compared with WT mice (Fig S1C), suggesting an increased reactivity of microglia to inflammatory stimuli (Ziebell et al, 2015; Vidal-Itriago et al, 2022). Moreover, to assess mitochondrial functionality, CD45⁺/CD11b⁺ cells were isolated from the cerebellum, and mitochondrial membrane potential (ΔΨM) was measured. As expected, a reduced ΔΨM was detected in cerebellar macrophage/microglia of KIKO mice compared with WT mice (Fig S1D).

Loss of FXN enhances glucose catabolism in microglial cells

To further elucidate the molecular mechanisms underlying the observed inflammatory activation in microglia cells derived from the KIKO model, we established a cell model of FRDA by achieving stable down-regulation of FXN expression in a microglial cell line (BV2^FXN−). Although BV2^FXN− cells exhibited lower inflammation than control cells (BV2^SCR) under basal conditions, as shown in Fig 3A, the greatest reactivity to inflammatory stimuli was observed when BV2^FXN− cells were activated with LPS (Fig 3A). These results suggest that LPS-activated BV2^FXN− better phenocopy the inflammatory setting observed in cerebella microglia of KIKO mice.

It has been demonstrated that a sustained inflammatory status of macrophages is characterized by metabolic shift from oxidative phosphorylation to glycolysis. To test if a metabolic rearrangement occurs in FXN down-regulating microglia, we measured glycolysis- and TCA-related metabolites in BV2^FXN− cells. Notably, mitochondrial metabolites including citrate and oxaloacetate were reduced, whereas αKG was increased in LPS-activated BV2^FXN− cells (Fig 3B). On the contrary, glucose uptake (Fig 3C), accumulation of glycolysis and pentose phosphate shunt metabolites (Fig 3D) as well as lactate production (Fig 3E) were significantly increased. These results were consistent with the glucose avidity of inflammatory immune cells (Soto-Heredero et al, 2020). With the aim to explore if the inflammatory phenotype of BV2^FXN− cells was dependent on glycolysis, we inhibited glucose uptake by 2-deoxyglucose (2-DG) and as reported in the Fig S2A, a reduced inflammatory response to LPS was observed.

Itaconate decreases inflammatory responses in FXN-microglia via the Nrf2 pathway

Itaconate, a mitochondrial metabolite, is produced by macrophages in response to inflammatory stimuli to counteract inflammatory stress (Lampropoulou et al, 2016). To test if the highest inflammatory phenotype of LPS-activated BV2^FXN− cells was also associated with itaconate overproduction, we measured its levels, and a significant increase was detected compared to scramble conditions (with or without LPS) (Figs 4A and S2B). Itaconate overproduction observed in BV2^FXN− cells was in accordance with the increased expression levels the immune-responsive gene 1 (Irg1) (Fig 4B), the mitochondrial enzyme catalyzing the decarboxylation of cis-aconitate to synthesize itaconate (Lampropoulou et al, 2016). Remarkably, higher Irg1 expression levels were also detected in cerebellum-derived microglia (Fig 4C). It has been reported that itaconate production following inflammatory stimuli mitigates the inflammatory response by constraining Il1b induction and glycolysis (Lampropoulou et al, 2016). To test if itaconate exerts such\effect in LPS-activated BV2^FXN− cells, a cell permeable formulation of soluble itaconate (dimethylitaconate) was added to the culture medium. As expected, dimethylitaconate diminished Il1β, Il6, and Nos2 mRNA levels in BV2^FXN− cells (Fig 4D).

Butyrate reverts the immunometabolic signatures through Itaconate/Nrf2/GSH signaling

Butyrate (BUT) is a ubiquitous short-chain fatty acid principally derived from the enteric microbiome, which showed a neuroprotective role (Li et al, 2016; Lanza et al, 2019). Metabolomic analysis of BUT-treated macrophages revealed a substantial reduction in glycolysis (Flemming, 2019; Schulthess et al, 2019) as well as limited inflammatory response in microglia (Caetano-Silva et al, 2023). By virtue of the recently demonstrated anti-inflammatory effects of BUT on white adipocytes and BMDM of KIKO mice (Turchi et al, 2023), we asked if butyrate treatment was also effective in counteracting the changes of the immunometabolic profile in LPS-treated BV2^FXN−. As reported in Fig 5A, BUT reduced glucose uptake and lowered lactate production in LPS-treated BV2^FXN− cells, whereas a significant refill in the mitochondrial metabolites such as citrate, oxaloacetate and succinate were observed (Fig 5B). Notably, BUT further increased itaconate levels in BV2^FXN− (Fig 5C), leading us to suppose that BUT restored the homeostatic function by the itaconate-driven antioxidant protection. To test this hypothesis, we analyzed Nrf2 protein in BV2^FXN− cells treated with butyrate and expectedly an increased nuclear accumulation of Nrf2 was observed (Fig 5D). Nrf2 is the primary transcription factor protecting cells from oxidative stress by regulating the synthesis of glutathione (GSH) (Harvey et al, 2009). Interestingly, BUT increased GSH levels in BV2^FXN− cells (Fig 5E).

To decipher the molecular mechanism driving the protective effect of butyrate, we analyzed the transcriptomics responses to butyrate in BV2^FXN− (Fig 5F) and KIKO-derived microglial cells (Fig 5G), which showed a high purity as showed by positiveness to CD11b⁺ (Fig S2C). The genes that were significantly down-regulated (Log₂FC < −1.5) by butyrate were integrated by Venn diagram (Fig 5H) and their functional enrichment analysis suggested that BUT inhibits NfkB signaling pathway in microglia with FXN deficiency (Fig 5I). Consistently, a diminished level of the phospho-active form of Nf-κb was observed in LPS-activated BV2^FXN− treated with BUT (Fig 5J). To demonstrate the protective effects of BUT in vivo, asymptomatic 4-mo-old KIKO mice were fed with dietary BUT for 16 wk and at the end of dietary treatment, the transcriptome of CD11b⁺ microglial cells isolated from cerebellum, was profiled. In accordance with in vitro data, KIKO-derived CD11b⁺ microglial cells showed a reduced expression level of inflammatory genes following dietary BUT treatment (Fig S2D).

Butyrate improves the neuromotor abilities in KIKO mice

Next, we asked if the improvement of the neuroinflammatory status of BUT-treated mice was accompanied by improved neuromotor abilities. To this end, a battery of neuromotor tasks including accelerating rotarod test (Bohlen et al, 2009), pole tests (turning time and climb down time) (Que et al, 2021), and tightrope test (Miquel & Blasco, 1978) were conducted in KIKO mice at the end of dietary treatment. The rotarod test revealed lower neuromotor capacity in KIKO mice compared with the WT mice when the mice ran at maximum RPM (Fig 6A). Nicely, BUT treatment was effective in limiting KIKO falls (Fig 6A). Similar results were observed following pole test, in which KIKO mice showed a highest time to turn completely downward (Tturn) and to descend to the floor (Ttotal) than WT mice (Fig 6B). Although butyrate treatment was effective in improving Tturn (Fig 6B), no improvement was observed in Ttotal (data not shown). Restored neurobehavioral abilities were also observed at the tightrope test, in which butyrate reduced the higher walking time of KIKO than WT mice (Fig 6C). These results suggest that dietary BUT improves neuromotor abilities through neuroinflammatory limitation in FRDA mice.

Hcar2 mediates the anti-inflammatory effects of butyrate

BUT interacts with several G-protein–coupled receptors including GPR109A (encoded by Hcar2 gene), GPR43 (encoded by Ffar2 gene) and GPR41 (encoded by Ffar3 gene) leading to activation of anti-inflammatory signaling cascades (Parada Venegas et al, 2019; Deleu et al, 2021). According to what previously reported (Moutinho et al, 2022), our scRNAseq data revealed that, among these receptors, Hcar2 was expressed at the highest values in microglia (Fig 7A). Interestingly, the Hcar2 expression was higher in KIKO than WT mice (Fig 7B), suggesting an increased sensitivity to BUT in FRDA mice. Remarkably, Hcar2 levels were increased in primary microglia isolated from cerebellum of KIKO mice (Fig 7C), and BUT treatment was effective in restraining its up-regulation (Fig 7C). In line with these findings, BUT limited the expression levels of Hcar2 in LPS-activated BV2^FXN− cells (Fig 7D). To investigate if Hcar2 mediates the anti-inflammatory effects of BUT, we down-regulated Hcar2 by RNA interference in BV2^FXN− and the inflammatory genes expression was analyzed following LPS stimulation and BUT pretreatment. Consistent with our hypothesis, BUT was ineffective in limiting inflammatory response in Hcar2 down-regulating cells (Fig 7E).

Limitation of the study

In our investigations involving BV2 cells, we used a single shRNA construct. Although this approach allowed us to assess FXN down-regulation, it is essential to acknowledge the potential for off-target effects associated with shRNA techniques. Thus, a notable limitation of our study lies in the absence of additional shRNA sequences to validate the observed FXN down-regulation in BV2 cells. Moreover, an important aspect hindering the comprehensive understanding of itaconate accumulation in FXN-deficient BV2 cells relates to the observed imbalance in metabolites, notably the low levels of citrate juxtaposed with elevated levels of aKG. This metabolic perturbation complicates the interpretation of itaconate accumulation dynamics. To address this challenge and enhance our understanding, future studies could use metabolic flux analysis techniques alongside the assessment of itaconate derivatives. Such approaches would provide deeper insights into the metabolic alterations underlying itaconate accumulation in FXN-deficient BV2 cells.

Mice and treatments

Neurobehavioral tests

Before rotarod testing, mice were trained for 1 d on the rotarod (cat. n. 47600; Ugo Basile s.r.l.) at a constant speed of 7 rpm over 1 min, repeated four times. On the day of testing, each mouse was placed on a stationary rod which was then accelerated from 7 to 32 rpm over 5 min. The latency to fall was recorded. This was performed over three trials with 60-min inter-trial intervals.

Before pole testing, mice were acclimated to a wooden pole measuring 30 cm × 1 cm. For the turning time assessment, each mouse was placed head upwards at the top of the pole. The time taken for the mouse to turn 180° downward was recorded. The descent time, from turning to reaching the base of the pole, was subsequently documented.

For tightrope test, a rope measuring 60 cm in length and 1 cm in diameter was securely stretched between two platforms. Each mouse was placed at the center of the rope, and the time taken to reach either platform was noted. This procedure was repeated over three trials with 30-min intervals between trials.

Cells and treatments

Bulk RNA sequencing

Total RNAs were extracted from cells and tissues using the MiniPrep kit from ZYMO RESEARCH, in adherence to the manufacturer’s instructions. The quantification of total RNA was performed using the Qubit 4.0 Fluorimetric Assay from Thermo Fisher Scientific. Libraries were constructed from 50 ng of total RNA through the NEGEDIA Digital mRNA-seq research grade sequencing service provided by Next Generation Diagnostic srl. This service encompassed library preparation, quality assessment, and sequencing on an Illumina NovaSeq 6000 system using a single-end, 75-cycle strategy. The raw data underwent analysis with FastQC v0.12.0 and were subsequently subjected to quality filtering and trimming by Trimmomatic v0.39 (Bolger et al, 2014), using a Q30 threshold for both leading and trailing ends, with a minimum length of 15 nucleotides. The resultant reads were aligned to the reference genome (mm10) using HISAT2 v2.2.1 (Kim et al, 2019). Quantification of gene expression was accomplished using the featureCounts tool (Liao et al, 2014). Subsequently, the DESeq2 package v1.40.2 (Love et al, 2014) was used to calculate DEGs and normalize the expression count matrix. Functional enrichment analysis was carried out using EnrichR or FunRich 3.0 with the Biological Processes Gene Ontology (GO) database.

Immunophenotyping by flow cytometry

Cerebellum was immunophenotyped by high dimensional flow cytometry using a panel containing markers to identify cell types and to assess activation states. The use of these markers allowed us to exclude all cells of no interest based on physical parameters (side and forward scatter) and to gate on specific cells of interest. In particular, cerebellum of WT and KIKO mice was dissociated to single-cell suspension using adult brain dissociation kit from MACS Technology and using GentleMACS (Miltenyi Biotec), according to the manufacturer’s protocol. Cells were first gated on CD45⁺ cells and then on CD45^lowCD11b⁺ to identify microglial cells and to exclude infiltrated macrophages (CD45^highCD11b⁺) and non-myeloid leukocytes (CD45^highCD11b^low). Microglia were further stained for the expression of M1 (anti-CD86 and anti-MHC-II) or M2 (anti-CD206 and anti-Trem2) markers. Samples were acquired on a 13-color Cytoflex (Beckman Coulter) and for each analysis, at least 0.5 × 10⁶ live cells were acquired by gating on aqua Live/Dead negative cells (Sciarretta et al, 2023).

Sholl analysis

Confocal assay analysis was performed using the ImageJ tool (v1.54f). From each scan, two layers combining nuclei (cyan channel) and microglia (blue) were selected and merged. Next, the microglia channel was analyzed after a grey-scale conversion. A total of 10 cells per sample were selected for further analysis, among these four cells per sample were selected for visualization purposes. Each microglia cell was studied by the ImageJ plugin “Neuroanatomy”: more in detail, the “Simple Neurite Tracker” (SNT—v4.2.1) was used to analyze microglia cell structures. The original scan images were cropped to study single microglia cells. Each resulting image was segmented and simplified by the “skeletonize” function. The major branches of each selected cell were measured by the SNT tool. The main ramifications were also evaluated. For each cell studied, each major branch and/or ramification was measured, then the mean length of every branch was calculated and subsequently the total branch length of every cell was summed and averaged across the same sample cells. Then, the occupancy radius of each microglia cell was calculated using the Sholl tool of the Neuroanatomy plugin. The resulting data were analyzed and plotted by GraphPad Prism (v9.1.0).

Targeted metabolomics

All data were acquired on a Triple Quad API3500 (AB Sciex) with an HPLC system ExionLC AC System (AB Sciex). For targeted metabolomic analysis, cells were extracted using tissue lyser for 30 s at maximum speed in 250 μl of ice-cold methanol: water: acetonitrile (55:25:20) containing [U-¹³C₆]-glucose 1 ng/μl and [U-¹³C₅]-glutamine 1 ng/μl as internal standards (Merk Life Science). Lysates were spun at 15,000g for 15 min at 4°C, dried under N₂ flow at 40°C, and resuspended in 125 μl of ice-cold methanol/water 70:30 for subsequent analyses. Amino acids analysis was performed through the previous derivatization. Briefly, 50 μl of 5% phenyl isothiocyanate in 31.5% ethanol and 31.5% pyridine in water were added to 10 μl of each sample. Mixtures were then incubated with phenyl isothiocyanate solution for 20 min at RT, dried under N2 flow, and suspended in 100 μl of 5 mM ammonium acetate in methanol/H₂O 1:1. Quantification of different amino acids was performed by using a C18 column (Biocrates) maintained at 50°C. The mobile phases were phase A: 0.2% formic acid in water and phase B: 0.2% formic acid in acetonitrile. The gradient was T₀: 100% A, T_5.5: 5% A, and T₇: 100% A with a flow rate of 500 μl/min. Measurement of energy metabolites and cofactors was performed by using a cyano-phase LUNA column (50 mm × 4.6 mm, 5 μm; Phenomenex), maintained at 53°C, by a 5-min run in negative ion mode. The mobile phase A was water, whereas phase B was 2 mM ammonium acetate in MeOH, and the gradient was 50% A and 50% B for the whole analysis, with a flow rate of 500 μl/min. Acylcarnitines quantification was performed using a ZORBAX SB-CN 2.1 × 150 mm, 5 μm column (Agilent). Samples were analyzed by a 10-min run in positive ion mode. The mobile phases were phase A: 0.2% formic acid in water and phase B: 0.2% formic acid in acetonitrile. The gradient was T₀: 100% A, T_5.5: 5% A, and T₇: 100% A with a flow rate of 350 μl/min. All metabolites analyzed were previously validated by pure standards, and internal standards were used to check instrument sensitivity. MultiQuant software (version 3.0.3; AB Sciex) was used for data analysis and peak review of chromatograms. Data were normalized on the median of the areas and then used to perform the statistical analysis.

Lactate production and glucose uptake

Extracellular lactate levels were assessed in the culture medium via an enzyme-based spectrophotometric assay. The procedure involved the collection of cell media, followed by treatment with a 1:2 (vol/vol) solution of 30% trichloroacetic acid to precipitate proteins. Afterward, the resulting mixture was subjected to centrifugation at 14,000g for 20 min at 4°C, and the supernatant was carefully collected. Subsequently, the collected supernatant was incubated for 30 min at 37°C with a reaction buffer containing glycine, hydrazine, NAD+, and LDH (lactate dehydrogenase) enzyme. This incubation allowed for the conversion of lactate to pyruvate, while simultaneously reducing NAD+ to NADH. The concentration of NADH, which is stoichiometrically equivalent to the amount of lactate, was then determined at 340 nm using a spectrophotometer.

To monitor glucose uptake, 2-NBDG probes were used according to manufactures protocols. Flow cytometry analyses were performed by 13-color Cytoflex (Beckman Coulter) and the percentage of 2-NBDG–positive cells was calculated by FlowJo software.

Quantitative PCR (qPCR)

The total RNA was isolated using TRI Reagent (Sigma-Aldrich). Subsequently, 3 mg of RNA was reverse-transcribed with M-MLV (Promega). qPCR was performed in triplicate, using validated qPCR primers confirmed via BLAST searches. The Applied Biosystems Power SYBR Green Master Mix was used, along with the QuantStudio3 Real-Time PCR System (Thermo Fisher Scientific). mRNA levels were normalized to actin mRNA, and the relative mRNA levels were determined using the 2^−ΔΔCt method. The primers used for reverse transcription quantitative PCR (RT-qPCR) are as follows:

Fxn:

Forward: 5′-TCTCTTTTGGGGATGGCGTG-3′

Reverse: 5′-GCTTGTTTGGGGTCTGCTTG-3′

Il1b:

Forward: 5′-TGCACCTTTTGACAGTGATG-3′

Reverse: 5′-AAGGTCCACGGGAAAGACAC-3′

Il6:

Forward: 5′-GGATACCACTCCCAACAGA-3'

Reverse: 5′-GCCATTGCACAACTCTTTTCTCA-3'

Rpl8:

Forward: 5′-GGAGCGACACGGCTACATTA-3’

Reverse: 5′-CCGATATTCAGCTGGGCCTT-3’

Nos2:

Forward: 5′-GCCTTCAACACCAAGGTTGTC-3′

Reverse: 5′-ACCACCAGCAGTAGTTGCTC-3′

Hcar2:

Forward: 5′-GAGCAGTTTTGGTTGCGAGG-3′

Reverse: 5′-GGGTGCATCTGGGACTCAAA-3′

Irg1:

Forward: 5′-GCAACATGATGCTCAAGTC-3′

Reverse: 5′-TGCTCCTCCGAATGATACCA-3′

Tnfa:

Forward: 5′-ATGGCCTCCTCATCAGTTC-3′

Reverse: 5′-TTGGTTTGCTACGACGTG-3′

Immunoblotting

Tissues or cells were lysed in RIPA buffer containing 50 mM Tris–HCl (pH 8.0), 150 mM NaCl, 12 mM deoxycholic acid, 0.5% Nonidet P-40, as well as protease and phosphatase inhibitors. Next, 5 mg of proteins were loaded onto an SDS–PAGE gel and subjected to Western blotting. Nitrocellulose membranes were subsequently incubated with primary antibodies at a 1:1,000 dilution. Following this, the membranes were incubated with the appropriate horseradish peroxidase-conjugated secondary antibodies. Immunoreactive bands were detected using a FluorChem FC3 System (Protein-Simple) after the membranes were incubated with ECL Prime Western blotting Reagent (GE Healthcare). Densitometric analysis of the immunoreactive bands was performed using the FluorChem FC3 analysis software. The antibodies used for immunoblotting as are follows: NRF2 (cod.: SAB4501984; Sigma-Aldrich); pNfKb (cod.:; #30335; Cell Signalling); NfKb (cod.: #4764; Cell Signalling); Tubulin (cod.: 10094-1-AP; Proteintech).

Statistical analysis

The data were presented as the mean ± SD. The specific number of replicates for each dataset is provided in the corresponding figure legend. To evaluate the statistical significance between two groups, a two-tailed unpaired t test was conducted. For comparisons involving three or more groups, an ANOVA was performed, followed by either Dunnett’s test (for comparisons relative to controls) or Tukey’s test (for multiple comparisons among groups). These statistical analyses were carried out using GraphPad Prism 9 (GraphPad Software Inc.). In all instances, a significance threshold of P < 0.05 was set.

Author Contributions

• F Sciarretta: data curation, formal analysis, and investigation.

• F Zaccaria: data curation and formal analysis.

• A Ninni: data curation and formal analysis.

• V Ceci: formal analysis.

• R Turchi: formal analysis.

• S Apolloni: formal analysis.

• M Milani: formal analysis and methodology.

• I Della Valle: formal analysis and methodology.

• M Tiberi: formal analysis.

• V Chiurchiù: formal analysis and methodology.

• N D’Ambrosi: formal analysis and methodology.

• S Pedretti: formal analysis and methodology.

• N Mitro: formal analysis and methodology.

• C Volontè: formal analysis and methodology.

• S Amadio: formal analysis and methodology.

• K Aquilano: supervision, validation, and writing—review and editing.

• D Lettieri-Barbato: conceptualization, supervision, project administration, and writing—original draft, review, and editing.

Conflict of Interest Statement

The authors declare that they have no conflict of interest.