Prostaglandin D2 Synthase Controls Schwann Cells Metabolism and Peripheral Myelin Homeostasis

ABSTRACT

We previously reported that in the absence of Prostaglandin D2 synthase (L‐PGDS) peripheral nerves are hypomyelinated in development while in adulthood they present aberrant myelin sheaths. We now demonstrate that L‐PGDS expressed in Schwann cells is part of a coordinated program that controls myelin homeostasis, describing a new physiological pathway implicated in preserving peripheral myelin. In vivo and in vitro lipidomic, metabolomic, and transcriptomic analyses confirmed that myelin lipids composition, Schwann cells' energetic metabolism, and key enzymes controlling these processes are altered in the absence of L‐PGDS. Moreover, Schwann cells undergo a metabolic rewiring, turn to acetate as the main energetic source, and produce ketone bodies to ensure glial cell and neuronal survival. All these changes correlate with morphological myelin alterations. Collectively, we posit that myelin lipids serve as a reservoir to provide ketone bodies, which together with acetate represent the adaptive substrates Schwann cells can rely on to sustain the axo‐glial unit and preserve the integrity of the PNS.

• Prostaglandin D2 Synthase (L‐PGDS) controls PPNS myelin lipids’ composition and Schwann cells energetic metabolism

• In the absence of L‐PGDS Schwann cells turn to acetate as the main energetic source.

• These metabolic changes result in ketone bodies production to preserve neuronal survival.

1. Introduction

Maintenance of an intact communication between glial cells, Schwann cells in the peripheral nervous system (PNS) and oligodendrocytes in the central nervous system (CNS) is essential to ensure efficient and effective transmission of the nerve impulse. Myelin, the insulating plasma membrane made by myelinating Schwann cells in the PNS, is the main structure deputed to this function. In addition to facilitating the transmission of the electric impulses, myelin is also essential to preserve the communication between neurons and glial cells. Indeed, lack of myelin or alteration in its integrity results in permanent damage to the nervous system, eventually leading to neuronal cell death (Boucanova and Chrast 2020; Nave and Trapp 2008).

In the PNS, the interaction between glial cells and axons is composite and relies on an articulate set of molecules expressed both by Schwann cells and axons (Monk et al. 2015). Further, signaling coming from the extracellular matrix is also key to maintaining the correct structure of the axo‐glial unit (Wilson et al. 2021). While the identity of these molecules and pathways has been defined in development and after nerve injury (Monk et al. 2015; Pellegatta and Taveggia 2019), less is known about those implicated in the maintenance of this cross talk in adulthood to prevent myelin and/or axonal damage.

In addition, metabolism, and the transfer of key metabolites between glial cells and neurons are critical to maintain an intact nervous system (Asadollahi et al. 2024; Looser et al. 2024). Indeed, either constitutive or cell specific ablation of enzymes regulating mitochondrial metabolism and energy, among which Cox10 (Funfschilling et al. 2012), TFAM (Viader et al. 2011; Viader et al. 2013), and LKB1 (Beirowski et al. 2014), confirm that Schwann cells actively support axons. Further, it is becoming increasingly evident that not only glial cells' metabolism uphold axons, but that both oligodendrocytes and Schwann cells transfer energy metabolites, essentially lactate, which is eventually used by neurons to sustain their energy demand (Asadollahi et al. 2024; Funfschilling et al. 2012; Lee et al. 2012). In the PNS, it has been proposed that the excessive lactate production observed in LKB–1 deficient Schwann cell mutant, could be transferred to neurons to limit axonal loss (Beirowski et al. 2014). However, specific ablation of the monocarboxylate transporter MCT1 in Schwann cells did not prevent myelination, though mutant mice have thinner myelin sheath in sensory nerves and impaired regeneration after injury (Domenech‐Estevez et al. 2015; Jha et al. 2020; Morrison et al. 2015; Zochodne 2012). Further, it has been shown that immediately after injury, Schwann cells provide neurons with lactate to prevent accelerated axonal degeneration and promote the initiation of the nerve regeneration process (Babetto et al. 2020). Notably, acute nerve injury induces a metabolic adaptation in Schwann cells (Sundaram et al. 2023).

We previously reported that prostaglandin D‐2 synthase (L‐PGDS) modulates PNS myelination and maintenance by acting downstream of Neuregulin 1 type III (Trimarco et al. 2014), the main instructive signal for PNS myelination (Taveggia et al. 2005). In the absence of l‐pgds, nerves are hypomyelinated during development, but in adulthood the structure of their myelin is aberrant. We now extend these studies by showing that these morphological alterations correlate with a switch specifically in Schwann cells' metabolism. In vivo lipidomic analyses confirmed that the excessive myelin remodeling occurring in these mutants completely alters lipids' profile. Levels of lysophosphatidylcholines containing the omega‐6 fatty acids, which are part of the Arachidonic acid metabolism, are significantly reduced. Of note, these modifications are accompanied by changes in Schwann cell energetic metabolism, which ultimately rely on acetate and ketone bodies to sustain glial cell and neuronal survival. Importantly, we observed a direct correlation between myelin instability and energy metabolism rewiring, as it occurs in the CNS (Asadollahi et al. 2024). Our study indicates that similarly to the CNS (Puchalska and Crawford 2017) also in the PNS acetate metabolism could substitute for glycolytic activity.

To our knowledge this is the first study correlating prostaglandins to glial cell metabolism and myelin homeostasis identifying acetate and ketone bodies utilization as key metabolic pathways in the PNS. Collectively, we posit that Schwann cells adapt their energetic demand to respond to excessive myelin remodeling by producing and resorting onto ketone bodies to sustain the axo‐glial unit, as recently described in Drosophila (McMullen et al. 2023; Silva et al. 2022). In future studies, it could be of interest to investigate whether also in pathological conditions acetate is implemented to preserve nerve integrity and functionality.

2. Results

2.1. Loss of l‐pgds Does Not Alter Myelin Protein Expression

We previously showed that in the absence of the PGD2 prostanoid, sciatic nerves are hypomyelinated and characterized by aberrant myelin in aged mutant mice lacking h‐ and l‐pgds expression, though ablation of l‐pgds alone is sufficient to cause hypomyelination (Trimarco et al. 2014). To define the molecular components at the basis of this altered myelin structure, we analyzed myelin protein expression in sciatic nerves of null l‐pgds and littermate controls at different ages. Quantitative RT–PCR (Figure 1a,b) and Western Blotting analyses (Figure 1c,d) on sciatic nerves of 1‐ and 8‐months old l‐pgds ^−/− mice showed no differences in Myelin Basic Protein (MBP) and Protein Zero (MPZ) mRNA and protein levels both in l‐pgds ^−/− mice.

FIGURE 1.

In the absence of L‐PGDS, myelin protein expression is maintained. (a, b) qRT–PCR analyses on mRNA prepared from (a) 1 month old wild type littermate controls (WT) and l‐pgds ^−/− (KO) sciatic nerves and (b) 8 months old wild type littermate controls (WT) and l‐pgds ^−/− (KO) sciatic nerves. All samples were tested for Mbp or Mpz expression. In (a) Mbp expression is slightly increased in null mutants, whereas Mpz expression is comparable. Data have been normalized to 36B4 expression level and analyzed with the CFX Manager Software from Biorad. N = pool of 2 nerves/mice, 3 different mice/genotype. Error bars represent mean ± s.e.m. (Mbp Unpaired t‐test. p = 0.0502, t = 2.773, df = 4; MPZ: Unpaired t test analysis, p = 0.5796; t = 0.6020 df = 4). In (b) no differences in the expression levels were observed between the genotypes. Data have been normalized to 36B4 expression level and analyzed with the CFX Manager Software from Biorad. N = pool of 2 nerves/mice, 3 different mice/genotype. Error bars represent mean ± s.e.m. (Mbp Unpaired t‐test. p = 0.346, t = 1.067, df = 4; MPZ: Unpaired t test analysis, p = 0.1172; t = 1.992 df = 4). (c) Representative western blotting analyses on sciatic nerves prepared from 1 month old wild type littermate controls (WT) and l‐pgds ^−/− (KO) sciatic nerves. Lysates were tested for MBP, MPZ, and Tubulin (Tub) as a loading control. The graphs showed the quantification of the represented western blotting. Error bars represent mean ± s.e.m. N = 3 different mice/genotype. (d) Representative western blotting analyses on sciatic nerves prepared from 8 months old wild type littermate controls (WT) and l‐pgds ^−/− (KO) sciatic nerves. Lysates were tested for MBP, MPZ, and Tubulin (Tub) as a loading control. The graphs showed the quantification of the represented western blotting. Error bars represent mean ± s.e.m. N = 3 different mice/genotype. (e) Left panels: Representative immunofluorescence of organotypic mouse Schwann cell—neuronal cocultures myelinated for 14 days and then treated for an additional 7 days with 25 μM A‐T56. Middle panels: Representative immunofluorescence of mouse Schwann cells prepared from P2 sciatic nerves of l‐pgds ^−/− and wild type control littermate pups, seeded onto purified wild type DRG neurons. Cultures were allowed to myelinate for 35 days. Right panels: Representative immunofluorescence of primary rat Schwann cell cocultured with purified DRG neurons prepared E14.5 l‐pgds ^−/− and wild type control littermate allowed to myelinate for 7 days. At the end of the myelination, cultures were fixed and stained for MBP (fluorescein) and beta 3 tubulin (middle panels) or neurofilament (right and left panels) (rhodamine). Myelin degeneration was observed in AT‐56 treated cultures and in l‐pgds ^−/− mouse Schwann cells cocultured with wild type DRG neurons. Bar: 100 μm.

Since L‐PGDS is expressed both in neurons and Schwann cells, and neuronal L‐PGDS contributes to myelin formation (Trimarco et al. 2014), we next sought to determine the cell specific effect of l‐pgds in vitro by exploiting Schwann cells neuronal myelinating cocultures (Figure 1e).

We first blocked the enzymatic activity of L‐PDGS by adding AT‐56, a specific irreversible inhibitor of L‐PGDS (Irikura et al. 2009) onto wild–type organotypic myelinated mouse dorsal root ganglia (DRG) neurons—Schwann cells cocultures. Cultures were established as described in Taveggia et al. (2005) and allowed to myelinate for 14 days, after which they were treated with 25 μM AT‐56 or with the vehicle control (DMSO), for additional 7 days. As previously reported Trimarco et al. (2014), we observed myelin degeneration only in AT‐56 treated cocultures. To determine the contribution of neurons and/or Schwann cells to this phenotype, we established myelinated cocultures with either mouse Schwann cells prepared from l‐pgds ^−/− sciatic nerves and wild type mouse DRG neurons, or with wild type primary rat Schwann cells and l‐pgds ^−/− DRG neurons. In both conditions E14.5 DRG neurons were prepared, purified by anti‐mitotic treatment of endogenous Schwann cells and fibroblasts, as described in Taveggia et al. (2005). Primary rat Schwann cells were prepared from P2 sciatic nerves as described in Taveggia et al. (2005) and 200.000 cells were seeded per DRG neuronal explant. Mouse sciatic nerves were prepared from P2 l‐pgds ^−/− and control littermate pups. After enzymatic dissection, cells were grown onto 0.01 mg/mL poly‐L‐lysine and 10 ng/μl laminin in a defined media. To minimize the growth of fibroblasts, cultures were initially treated with 10 μM Cytosine β‐D‐arabinofuranoside hydrochloride. After expansion, we seeded 180,000–200,000 cells per DRG neuronal explant. Established cocultures were maintained in the seeding media for 3 days to allow Schwann cell proliferation. Myelination was then induced by supplementing the culture media with 50 μg/mL ascorbic acid and myelination achieved in 5 weeks. As shown in Figure 1e, myelination of primary mouse Schwann cells was overall inefficient. Of note, in the absence of glial l‐pgds myelination was even less efficient and the MBP⁺ segments were fragmented, similarly to those observed in AT‐56 treated cocultures. In turn, primary rat Schwann cells seeded onto l‐pgds ^−/− DRG neurons myelinate less than controls, but the MBP⁺ segments did not present any sign of degeneration.

To exclude that the myelin phenotype observed in cocultures of mouse l‐pgds ^−/− Schwann cells with wild type mouse DRG neurons and in AT‐56 treated cocultures, might be due to Schwann cells and/or neuronal cell death we performed TUNEL assays analyses on isolated mouse l‐pgds ^−/− Schwann cells (Figure S1a) and on myelinated cocultures treated with AT‐56 (Figure S1b). In both conditions, we did not observe any cell death or axonal swallowing, in agreement with our former studies (Trimarco et al. 2014).

Thus, despite the differences observed, myelination of mouse Schwann cells lacking l‐pgds ^−/− expression resembles more AT‐56 treated cocultures.

2.2. Myelin Alterations Accumulate in Vivo in the Absence of Glial L‐PGDS

We previously reported hypomyelination and myelin degeneration in l‐pgds ^−/− mice (Trimarco et al. 2014). Thus, we first determined if the number of myelin alterations progresses with time in sciatic nerves of 4‐, 6‐, and 8‐months old l‐pgds ^−/− mice. At 4 months we observed a slight increase in myelin degeneration, which is significantly increased at 6‐ (data not shown) and 8 months (Figure 2a–d, pink arrows).

FIGURE 2.

Glial L‐PGDS controls myelin maintenance. (a) Semi–thin sections of 4 months wild–type littermate controls (WT) and l‐pgds ^−/− sciatic nerves (KO). Pink arrows indicate myelin aberrations. Bar: 50 μm. (b) Graph, average of three different experiments, representing the percentage of myelin morphological aberrations in 4 months old wild type littermate controls (white bar) and l‐pgds ^−/− sciatic nerves (red bar). Alterations were determined as the number of fibers presenting myelin structural alterations over the total number of fibers in the entire reconstructed nerve cross section. Error bars represent mean ± s.e.m. N = 3 different mice/genotype (Unpaired t test analysis, p = 0.1322; t = 1.887 df = 4). (c) Semi‐thin sections of 8 month wild–type littermate controls (WT) and l‐pgds ^−/− sciatic nerves (KO). Pink arrows indicate myelin aberrations. Bar: 50 μm. (d) Graph, average of three different experiments, representing the percentage of myelin morphological aberrations in 8 months old wild type littermate controls (white bar) and l‐pgds ^−/− sciatic nerves (red bar). Alterations were determined as the number of fibers presenting myelin structural alterations over the total number of fibers in the entire reconstructed nerve cross section. Error bars represent mean ± s.e.m. N = 3 different mice/genotype (Unpaired t test analysis, p = 0.0180; t = 3.870 df = 4). (e) Semi–thin sections of 8 months wild–type littermate controls (WT) and ChAT‐Cre//l‐pgds ^flx/flx sciatic nerves (KO). Pink arrows indicate myelin aberrations. Bar: 50 μm. (f) Graph, average of three different experiments, showing the percentage of myelin morphological aberrations in 8 months old wild type littermate controls (white bar) and ChAT‐Cre//l‐pgds ^flx/flx sciatic nerves (red bar). Alterations were determined as the number of fibers presenting myelin structural alterations over the total number of fibers in the entire reconstructed nerve cross section. Error bars represent mean ± s.e.m. N = 3 different mice/genotype (Unpaired t test analysis, p = 0.8806 n.s.; t = 0.160 df = 4). (g) Semi‐thin sections of 8 month wild–type littermate controls (WT) and MPZ‐Cre//l‐pgds ^flx/flx . Pink arrows indicate myelin aberrations. Bar: 50 μm. (h) Graph, average of three different experiments, representing the percentage of myelin morphological aberrations in 8 months old wild type littermate controls (white bar) and MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves (red bar). Alterations were determined as the number of fibers presenting myelin structural alterations (pink arrows) over the total number of fibers in the entire reconstructed nerve cross section. Error bars represent mean ± s.e.m. N = 3 different mice/genotype. (Unpaired t test analysis, p = 0.0455; t = 2.868 df = 4).

Next, we assessed the in vivo contribution of glial and neuronal L‐PGDS. Thus, we crossed l‐pgds ^flx/flx mice (Kaneko et al. 2012) with mutant mice driving Cre recombinase expressed either in motor neurons (Choline O‐Acetyltransferase –Cre (ChAT‐Cre)) (Rossi et al. 2011) or in myelinating Schwann cells (MPZ‐Cre) (Feltri et al. 1999). We first confirmed specific recombination in ventral spinal cord of ChAT‐Cre (Figure S2a) and in sciatic nerves of MPZ‐Cre mice (Figure S3j). Ultrastructural analyses and g ratio measurements (Figure S2c–f), indicated that ChAT‐Cre//l‐pgds ^flx/flx sciatic nerves were hypomyelinated at P7 but myelin was morphologically normal in 1 adult mice (Figure 2e–f), suggesting that neuronal L‐PGDS might be implicated in regulating myelin formation, as formerly reported (Trimarco et al. 2014). Next, we performed a detailed characterization of MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves at different time points, from P7 to adulthood (Figure 2g,h and Figure S3). Morphological analyses and g ratio measurements indicated that, unlike ChAT‐Cre//l‐pgds ^flx/flx mutants, myelin formation was normal in the absence of glial l‐pgds, as confirmed by ultrastructural analyses at P7, 1 month and 10 months of age (Figure S3a–i). However, we observed an increase in the number of myelin alterations, starting at 8 months of age in MPZ‐Cre//l‐pgds ^flx/flx mice (Figure 2g–h, pink arrows), which is persistent in 10 months old mice (Figure S3g, pink arrows and S3k), similarly to complete l‐pgds ^−/− mice.

Collectively, in vitro and in vivo analyses indicate that myelin degeneration in adult mice is likely due to glial l‐pgds. Further, they also suggest that myelin lipid profile rather than proteins might be affected in the absence of a functional L‐PGDS.

2.3. Peripheral Nerve Lipid Metabolism Is Impaired in the Absence of Glial L‐PGDS

Since L‐PGDS is an enzyme of Arachidonic acid metabolism, which, together with omega‐6 fatty acids, is naturally present in myelin phospholipids, we retrieved phospholipid analyses on total extracts of 4‐, 6‐ and 8‐months old l‐pgds ^−/− sciatic nerves (Figure 3a–c), in 8‐, 10‐ and 14‐months MPZ‐Cre//l‐pgds ^flx/flx old mice (Figure 3f–h), as well as in 8 months old ChAT‐Cre//l‐pgds ^flx/flx old mice (Figure S2b).

FIGURE 3.

Targeted lipidomic analyses show alteration in phospholipids' profile in the absence of l‐pgds. (a) Graph showing phospholipids profile in sciatic nerves of 4 months old wild type littermate controls (WT) and l‐pgds ^−/− (KO) mice. Lipids were normalized over the total protein content. Error bars represent mean ± s.e.m. N = 6 different mice/genotype. (Unpaired t‐test * p = 0.0105; t = 3.141, df = 10, **p = 0.0057; t = 3.505, df = 10). (b) Graph showing phospholipids profile in sciatic nerves of 6 months old wild type littermate controls (WT) and l‐pgds ^−/− (KO) mice. Lipids were normalized over the total protein content. Error bars represent mean ± s.e.m. N = 5 different mice/genotype. (Unpaired t‐test. LPC 16:1 *p = 0.0157, t = 2.906, df = 10; LPC 17:0 ****p < 0.0001, t = 7.231, df = 10; LPC 18:2 ***p = 0.0007 t = 4.867, df = 10; LPC 20:4 ****p < 0.0001, t = 8.631, df = 10; LPC 20:3 ****p < 0.0001, t = 8.214, df = 10; PI 36:4 *p = 0.0294, t = 2.54, df = 10; PI 38:4 ****p < 0.0001, t = 7.169, df = 10; PC ae 36:5 **p = 0.0039 t = 3.739, df = 10; PE aa 32:2 *p = 0.029, t = 2.547, df = 10; PE aa *p = 0.0484, t = 2.247, df = 10). (c) Graph showing phospholipids profile in sciatic nerves of 8 months old wild type littermate controls (WT) and l‐pgds ^−/− (KO) mice. Lipids were normalized over the total protein content. N = 4 wild type and N = 3 null mice. Error bars represent mean ± s.e.m. (Unpaired t‐test. LPC 16:1 *p = 0.0373, t = 2.815, df = 5; LPC 17:0 *p = 0.0438, t = 2.68, df = 5; LPC 18:2 **p = 0.0074, t = 4.346, df = 5; LPC 20:4 *p = 0.0314, t = 2.962, df = 5; LPC 20:3 *p = 0.0197, t = 3.38, df = 5; PI 36:4 **p = 0.0036, t = 5.159, df = 5; PI 38:4 *p = 0.0396, t = 2.764, df = 5; PC ae 36:5 *p = 0.0434 t = 2.688, df = 5; PE aa 32:2 *p = 0.0452, t = 2.653, df = 5; PE aa *p = 0.0314, t = 2.963, df = 5). (d) Graph showing total and free cholesterol levels in sciatic nerves of 4 months old wild type littermate controls (WT) and l‐pgds ^−/− (KO) mice. Cholesterol levels were normalized over the total protein content. Variation in cholesterol amount is expressed as fold change to WT arbitrarily set as 1.0. N = 5 different mice/genotype. Error bars represent mean ± s.e.m. (Unpaired t‐test. Tot: P = not significant, t = 1.003, df = 8; Free: P = not significant, t = 0.292, df = 8). (e) Graph showing total and free cholesterol levels in sciatic nerves of 8 months old wild type littermate controls (WT) and l‐pgds ^−/− (KO) mice. Cholesterol levels were normalized over the total protein content. Variation in cholesterol amount is expressed as fold change to WT arbitrarily set as 1.0. N = 4 wild type and N = 3 null mice. Error bars represent mean ± s.e.m. (Unpaired t‐test. Tot: P = 0.0588, t = 2.438, df = 5; Free: P = 0.0346, t = 2.88, df = 5). (f) Graph showing phospholipids profile in sciatic nerves of 8 months old wild type littermate controls (WT) and MPZ‐Cre//l‐pgds ^flx/flx mice. Lipids were normalized over the total protein content. Error bars represent mean ± s.e.m. N = 3 different mice/genotype (Unpaired t test. No significant difference in any of the analyzed metabolites). (g) Graph showing phospholipids profile sciatic nerves of 10 months old wild type littermate controls (WT) and MPZ‐Cre//l‐pgds ^flx/flx mice. Lipids were normalized over the total protein content. Error bars represent mean ± s.e.m. N = 5 different mice/genotype. (Unpaired t‐test. LPC 16:1 **p = 0.0072, t = 5.066, df = 4; PC 24:0 *p = 0.0232, t = 3.578, df = 4; PC 30:2 *p = 0.0162, t = 3.998, df = 4; PE 40:1 *p = 0.028, t = 3.372, df = 4; PE 40:3 **p = 0.0037, t = 6.092, df = 4). (h) Graph showing phospholipids profile in sciatic nerves of 14 months old wild type littermate controls (WT) and MPZ‐Cre//l‐pgds ^flx/flx mice. Lipids were normalized over the total protein content. Error bars represent mean ± s.e.m. N = 4 MPZ‐Cre//l‐pgds ^flx/flx mice; N = 3 littermate controls (WT) mice. (Unpaired t‐test. LPC 16:1 *p = 0.0248, t = 3.171, df = 5; LPC 17:0 **p = 0.0055, t = 4.665, df = 5; LPC 18:2 **p = 0.0053, t = 4.697, df = 5; LPC 20:4 ***p = 0.0005, t = 8.018, df = 5; LPC 20:3 **p = 0.0012, t = 6.628, df = 5; PC 24:0 *p = 0.0214, t = 3.305, df = 5; PC 30:2 *p = 0.0341, t = 2.891, df = 5; PC 36:5 ***p = 0.0004, t = 8.208, df = 5; PE 40:1 *p = 0.0173, t = 3.501, df = 5; PE 40:3 *p = 0.0351, t = 2.868, df = 5). (i) Graph showing the phospholipids profile in Schwann cell—neuronal myelinated cocultures. After 14 days in myelinating conditions, cultures were treated for an additional 7 days with 25 μM AT‐56 or with DMSO as control vehicle. Levels of lysophosphatidylcholines containing omega–6 fatty acids are reduced in myelinated cocultures treated with AT‐56. Lipids were normalized over the total protein content. N = 3 different independent coculture experiments. Error bars represent mean ± s.e.m. (Unpaired t‐test. LPC 14:0 **p = 0.0073, t = 5.035, df = 4; LPC 16:1 *p = 0.0248, t = 3.503, df = 4; LPC 16:0 **p = 0.0026, t = 6.716, df = 4; LPC 17:0 *p = 0.0188, t = 3.818, df = 54; LPC 18:2 *p = 0.0132 t = 4.243, df = 4; LPC 18:1 **p = 0.0055 t = 5.447, df = 4; LPC 18:0 **p = 0.006 t = 5.316, df = 4; LPC 20:3 **p = 0.0011, t = 8.485, df = 4; LPC 20:4 **p = 0.0086, t = 4.802, df = 4).

Interestingly, phospholipids' profile was similar in 4‐months old l‐pgds ^−/− and control mice (Figure 3a). However, at 6‐ (Figure 3b) and at 8‐months of age (Figure 3c), l‐pgds ^−/− sciatic nerves were mostly altered as compared to littermate wild type controls. We observed a strong significant decrease (p < 0.001) in lysophosphatidylcholines containing Arachidonic acid (LPC 20:4) and in omega‐6 fatty acids precursors of Arachidonic acid, such as Linoleic acid (LPC 18:2) and di‐homo‐γ linoleic acid (LPC 20:3) (Figure S4a–d). Notably, in 8 months old l‐pgds ^−/− sciatic nerves we also observed a decrease in both total and free cholesterol, a critical component of myelin, whose amounts were not affected at 4 months (Figure 3d–e).

Targeted lipidomic analyses in MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves revealed similar alterations in phospholipid composition (Figure 3f–h), which were mostly significant in older mice. On the contrary, we did not observe significant differences in 8 months old ChAT‐Cre//l‐pgds ^flx/flx old mice (Figure S2b), further indicating a role of glial L‐PGDS in controlling myelin integrity.

To corroborate these results, we also performed targeted phospholipid analyses on organotypic wild–type myelinated DRG—Schwann cells cocultures treated with 25 μM AT‐56 for 7 days. Similarly to in vivo results, we observed a significant decrease in lysophosphatidylcholines containing arachidonic acid and in omega–6 fatty acids specifically in cocultures treated with AT‐56, namely LPC 20:4, LPC 18:2, and LPC 20:3 (Figure 3i).

Of note, the lipidomic profile of l‐pgds ^−/− and MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves does not completely overlap. This is not surprising as l‐pgds ^−/− mice are also hypomyelinated in development. Moreover, we cannot exclude that in l‐pgds ^−/− mice cells other than Schwann cells might contribute to the observed phenotype. Indeed, L‐PGDS is expressed also in macrophages (Joo et al. 2007) and in perineurial fibroblasts (Gerber et al. 2021).

Nevertheless, these in vivo analyses implicate glial L‐PGDS in controlling myelin homeostasis. Further, they also indicate that in the PNS, glial and neuronal L‐PGDS exert different roles, with the latter being more important for myelin formation.

2.4. Loss of l‐pgds in Sciatic Nerves From Adult Mice Decreases Glycolysis and Krebs Cycle Activity

Next, we determined whether also the energetic metabolism might be affected in sciatic nerves of l‐pgds ^−/− and of MPZ‐Cre//l‐pgds ^flx/flx mice. Thus, we examined the data relative to energetic metabolites from sciatic nerves of both mutants collected at 4‐, 6‐ and 8‐months (l‐pgds ^−/− ) and at 8‐, 10‐ and 14‐months (MPZ‐Cre//l‐pgds ^flx/flx ).

While at 4‐months, in l‐pgds ^−/− sciatic nerves we did not observe any significant alteration (Figure 4a), both at 6‐ (Figure 4b) and at 8‐months (Figure 4c), metabolites involved in glycolysis and Krebs cycle were overall reduced, and, as in the case of phospholipids, this effect was more significant at 6 months. Indeed, at this age we detected an increase in glucose‐6P/fructose‐6P accompanied by a decrease in glyceraldehyde‐3‐phosphate/dihydroxyacetone‐phosphate (DHAP/GAP) and pyruvate that are indicative of an overall slowdown in glycolysis (Figure 4b). Further, the analyses of the Krebs cycle revealed a reduction in citrate, succinyl–CoA, fumarate, malate and oxaloacetate in l‐pgds ^−/− nerves. Notably, we observed a decrease also in acetyl‐CoA in 8 months old mice (Figure 4c and Figure S4e). Since citrate and α‐ketoglutarate (Figure 4c and Figure S4f–g) were also reduced in 8 months old sciatic nerves, these data suggest that upon loss of L‐PGDS, there is a change in acetyl–CoA metabolic production/utilization in the Krebs cycle featuring a metabolic rewiring likely occurring in Schwann cells.

FIGURE 4.

L‐PGDS controls glycolysis and Krebs cycle metabolites in sciatic nerves. (a) Graph showing the metabolic profile in 4 months old wild type littermate controls (WT) and l‐pgds ^−/− (KO) sciatic nerves. Metabolites were normalized over the total protein content. N = 5 different mice/genotype. Error bars represent mean ± s.e.m. (Unpaired t‐test; no significant difference in any of the analyzed metabolites). (b) Graph showing the metabolic profile in 6 months old wild type littermate controls (WT) and l‐pgds ^−/− (KO) sciatic nerves. Metabolites were normalized over the total protein content. N = 6 different mice/genotype. Error bars represent mean ± s.e.m. (Unpaired t‐test. Glu6P/fru6P *p = 0.0196, t = 2.774, df = 10; DHAP–GAP ***p = 0.0004, t = 5.264, df = 10; pyruvate **p = 0.0098, t = 3.178, df = 10; lactate **p = 0.0066, t = 3.418, df = 10; ribose5P *p = 0.0272, t = 2.585, df = 10; citrate **p = 0.0047, t = 3.614, df = 10; succinyl–CoA *p = 0.0348, t = 2.441, df = 10; succinate *p = 0.0121, t = 3.057, df = 10; fumarate *p = 0.0136, t = 2.988, df = 10; malate *p = 0.0133, t = 3.001, df = 10; oxalacetate *p = 0.0194, t = 2.78, df = 10). (c) Graph showing the metabolic profile in 8 months old wild type littermate controls (WT) and l‐pgds ^−/− (KO) sciatic nerves. Metabolites were normalized over the total protein content. N = 4 wild type and N = 3 null mice. Error bars represent mean ± s.e.m. (Unpaired t‐test. DHAP–GAP **p = 0.0071, t = 4.392, df = 5; acetyl–CoA *p = 0.0367, t = 2.828, df = 5; citrate *p = 0.038, t = 2.799, df = 5; α–ketoglutarate *p = 0.0473, t = 2.617, df = 5; malate *p = 0.0433, t = 2.689, df = 5). (d) Graph showing the metabolic profile in 8 months old wild type littermate controls (WT) and MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves. Metabolites were normalized over the total protein content. N = 3 different mice/genotype. Error bars represent mean ± s.e.m. (Unpaired t‐test. DHAP–GAP *p = 0.0301, t = 2.631, df = 8; succinate **p = 0.0018, t = 4.58, df = 8). (e) Graph showing the metabolic profile in 10 months old wild type littermate controls (WT) and MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves. Metabolites were normalized over the total protein content. N = 5 different mice/genotype. Error bars represent mean ± s.e.m. (Unpaired t‐test. Acetyl–CoA *p = 0.0221, t = 3.632, df = 4). (f) Graph showing the metabolic profile in 14 months old wild type littermate controls (WT) and MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves. Metabolites were normalized over the total protein content. N = 4 MPZ‐Cre//l‐pgds ^flx/flx mice; N = 3 littermate controls (WT) mice. Error bars represent mean ± s.e.m. (Unpaired t‐test. Glucose **p = 0.0013, t = 6.532, df = 5; Glu6P/fru6P *p = 0.028, t = 3.063, df = 5; Acetyl–CoA ***p = 0.0009, t = 7.035, df = 5; succinate **p = 0.0013, t = 6.475, df = 5; oxalacetate *p = 0.014, t = 3.702, df = 5).

Previous studies have shown that both oligodendrocytes and Schwann cells, can convert pyruvate in lactate, which is then transferred to neurons via monocarboxylate transporters (Babetto et al. 2020; Boucanova and Chrast 2020; Lee et al. 2012; Philips et al. 2021; Saab and Nave 2017). To determine whether in the absence of L‐PGDS, glucose is converted to lactate, we measured this metabolite in 4‐, 6‐ and 8‐months old sciatic nerves of l‐pgds ^−/− mutants and littermate controls. To our surprise, both at 6‐ and 8‐months of age, lactate levels were reduced in mutant mice relative to controls (Figure 4b–c and Figure S4h).

Surprisingly, metabolites' changes in MPZ‐Cre//l‐pgds ^flx/flx mice at all analyzed time points (Figure 3d–f), were less pronounced and partly differ from complete null mice. In these mutants, we observed a reduction in glucose and a decrease in DHAP/GAP, that in addition to the overall reduction in glucose, could be explained by enhanced expression of Pdk4 (Figure 5d); whereas lactate concentration does not change. Strikingly, levels of acetyl‐CoA were also decreased at 14 months of age, along with those of succinate and oxaloacetate, indicating a relevant impairment in Schwann cells Krebs cycle. Nevertheless, these cells did not experience any survival impairment (Figure S1a).

FIGURE 5.

L‐PGDS regulates genes involved in lipid homeostasis and Schwann cell energetic metabolism. (a) Heat–map showing the main upregulated genes identified in RNAseq analyses of rat primary Schwann cells—mouse neuronal myelinating cocultures. After 14 days in myelinating conditions, cultures were treated for an additional 7 days with 25 μM AT‐56 or with DMSO as a control vehicle before processing (Alignments on rat genome with p < 0.001). (b) qRT–PCR analyses on mRNA prepared from rat Schwann cell–mouse neuronal cocultures myelinated for 14 days and then treated for an additional 7 days with 25 μM AT‐56 or with DMSO as vehicle control. The expression of Pdk4, Hmgcs2, Acsl3, and Angptl4 is increased in AT‐56 treated cocultures. Data have been normalized to gapdh expression level and analyzed with the StepOne Software v2.3 for Pdk4, Acsl3, and Angptl4 (Applied Biosystems). The expression of Hmgcs2 has been normalized to TBP expression level and analyzed with the CFX Manager Software from Biorad. N = pool of at least 10 coverslips/condition, from 3 different rat Schwann cell–mouse neuronal cocultures. Error bars represent mean ± s.e.m. (Two–way ANOVA; pdk4 ****p < 0.0001 F (2, 12) = 256.5; Hmgcs2: Unpaired t‐test. *p = 0.0268, t = 3.418, df = 4; Acsl3 ***p = 0.0004 F (2, 12) = 15.92; Angptl4 ****p < 0.0001 F (2, 12) = 315.1). (c) qRT–PCR analyses on mRNA prepared from 8 months old wild type littermate controls (WT) and l‐pgds ^−/− (KO) sciatic nerves tested for Hmgcs2 and Bdh1 expression. Data have been normalized to 36B4 expression level and analyzed with the CFX Manager Software from Biorad. The expression of both genes is increased in MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves. N = pool of 2 nerves/mice, 3 different mice/genotype. Error bars represent mean ± s.e.m. (Unpaired t‐test. Bdh1: **p = 0.0055, t = 5.456, df = 4). (d) qRT–PCR analyses on mRNA prepared from 10 months old wild type littermate controls (WT) and MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves tested for Pdk4 and Hmgcs2 expression. Data have been normalized to 36B4 expression level and analyzed with the CFX Manager Software from Biorad. The expression of both genes is increased in MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves. N = pool of 2 nerves/mice, 4 (Pdk4) N = 6 (Hmgcs2) different mice/genotype. Error bars represent mean ± s.e.m. (Unpaired t‐test. Pdk4 *p = 0.0308, t = 2.808, df = 6; Hmgcs2: *p = 0.0213, t = 2.854, df = 8). (e) STRING pathway analyses and functional enrichments identified genes involved in “cellular response to fatty acid” (GO:0071398) and in “regulation of lipid metabolic process” (GO:0019216) with a false discovery rate of 2.13e‐05. (f) STRING pathway analyses and functional enrichments identified genes involved in “mitochondrial part” (GO:0044429) with false discovery rate of 0.00094. (g) KEGG pathway analyses identify genes enriched in the PPARγ signaling pathway (false discovery rate 1.20e‐05). (h) Representative immunofluorescence of rat Schwann cell—mouse neuronal myelinated cocultures treated with 100 μM or 300 μM Arachidonic acid (complexed to BSA) for an additional 7 days after 14 days in myelinating conditions. At the end of the treatment, cultures were fixed and stained for MBP (rhodamine) and Neurofilament (fluorescein). Myelin degeneration was observed upon 300 μM Arachidonic acid treatment with no signs of axonal swelling. The same amount of BSA was added to control cocultures. Bar: 100 μm. (i, j) qRT‐PCR analyses on mRNA prepared from rat Schwann cell—mouse neuronal cocultures treated for additional 7 days with 100 μM or 300 μM Arachidonic acid after 14 days in myelinating conditions. Arachidonic acid induces the expression of Pdk4 (i ) and Hmgcs2 (j (KO)) mRNA. Data have been normalized to 36B4 (Pdk4) and TBP (Hmgcs2) expression level and analyzed with the CFX Manager Software from Biorad. N = pool of at least 10 coverslips/condition, from 3 different rat Schwann cell—mouse neuronal cocultures. Error bars represent mean ± s.e.m. (Unpaired t‐test. Pdk4: ****p < 0.0001, t = 38.16, df = 4; Hmgcs2: *p = 0.0429, t = 2.92, df = 4).

Collectively, these results indicate that in the absence of l‐pgds both lipidomic and energetic profiles are altered. Though these changes are more evident in l‐pgds ^−/− sciatic nerves and might be due to the presence of both hypomyelination and myelin instability, mice lacking glial l‐pgds have also a reduction in glycolysis and Krebs cycle, suggesting they might rely on alternative metabolic pathways.

2.5. In the Absence of l‐pgds, Genes Involved in Lipid Homeostasis and Cell Metabolism Are Upregulated

To characterize the molecular mechanism at the basis of these morphological alterations, we performed RNAseq analyses on wild–type myelinated mouse DRG neurons—rat Schwann cells cocultures treated with 25 μM AT‐56. Thus, we prepared and sequenced RNAs from cocultures treated with AT‐56 versus vehicle only treated cocultures, as control. Overall, the number of upregulated genes significantly outnumbered downregulated ones. To determine whether identified genes belonged to neurons (mouse) or Schwann cells (rat), we aligned identified sequences to both mouse and rat deposited exome sequences (derived from NCBI assembly databases Build 37/mm9 and RGSC_v3.4 respectively) and selected genes with a p value < 0.001. From the alignment on the mouse exome sequence, we detected a total of 30 differentially expressed genes, 18 of which were upregulated (FC > 1.7) and 12 downregulated genes (FC < 0.5). Similarly, from the alignment on rat exome sequence we identified a total of 281 genes, 24 of which upregulated genes (FC > 1.7) and 2 downregulated (FC < 0.5). Of note, only rat genes, whose identity differs from those aligning to the mouse sequence, showed functional relationships, indicating that L‐PGDS inhibition alters gene expression mainly in Schwann cells (Figure 5a).

To validate these results, we performed quantitative RT–PCR analyses for some of the identified genes. We confirmed upregulated expression of Pyruvate dehydrogenase kinase 4 (Pdk4) and of Hydroxymethylglutaryl‐CoA synthase (Hmgcs2) in myelinated cocultures treated with AT‐56 (Figure 5b) and in MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves (Figure 5d). Of note, in l‐pgds ^−/− mutants we observed upregulated expression of Hmgcs2 and of 3‐hydroxybutyrate dehydrogenase 1 (Bdh1) (Figure 5c), a crucial enzyne for the ketone bodies synthesis.

Further, also Acyl‐CoA Synthetase Long Chain Family Member 3 (Acsl3) and Angiopoietin‐like 4 (Angptl4), genes encoding for enzymes involved in lipids and Co‐enzyme A consumption or release, are specifically upregulated in rat Schwann cells (Figure 5b), indicating a possible rewiring in cell metabolism during myelin degeneration. Interestingly, STRING analysis revealed a strong inter–connection between proteins encoded by nine of these genes with a PPI enrichment p value < 1.0e–16, suggesting that these proteins are at least partially biologically associated in their function. Moreover, analysis of the functional enrichments, designated genes involved in “cellular response to fatty acid” (GO:0071398) and in “regulation of lipid metabolic process” (GO:0019216) both with a false discovery rate of 2.13e‐05 (Figure 5e). Analysis of cellular component associated genes identified “mitochondrial part” (GO:0044429) with false discovery rate of 0.00094 (Figure 5f). Finally, analysis of KEGG pathways revealed genes enriched in PPARγ signaling pathway with a false discovery rate of 1.20e–05 (Figure 5g).

2.6. Excess of Arachidonic Acid Is Sufficient to Induce Myelin Degeneration in Vitro

Lysophosphatidylcholines derive from the release of one fatty acid chain from phosphatidylcholines and contribute to plasma membrane lipids recycling. They can in fact be further cleaved to release the other fatty acid chain from the choline group; alternatively, they can bind a free fatty acid to create a new phosphatidylcholine that will eventually relocate in the plasma membrane. Since in our in vitro and in vivo lipidomic analyses both lysophosphatidylcholine containing long fatty acid chain and phosphatidylcholine were decreased, we reasoned unlikely their new insertion into the plasma membrane. Rather, we speculated that the decrease in lysophosphatidylcholine might indicate an intracellular accumulation of free omega‐6 fatty acids that could directly contribute to the morphological alterations observed in the absence of L‐PGDS. To corroborate this hypothesis, we exploited our in vitro culture system and treated already myelinated mouse DRG‐rat Schwann cell cocultures with Arachidonic acid, the omega 6 fatty acid direct precursor of prostanoid synthesis. We specifically added 100 μM or 300 μM Arachidonic acid bound to bovine serum albumin (BSA), as in this form, the Arachidonic acid crosses the plasma membrane and accumulates in the cytoplasm without exiting the cell. Although 100 μM Arachidonic acid–BSA is sufficient to slightly alter myelin integrity, addition of 300 μM Arachidonic acid–BSA causes extensive myelin degeneration, without signs of axonal swelling (Figure 5h) or Schwann cell death (Figure S1c), as observed when L‐PGDS enzymatic activity is inhibited. Though we did not formally prove arachidonic acid accumulation in vivo, these results indicate that myelin degeneration might be a direct consequence of this effect at least in vitro. Further, in myelinated cocultures treated with Arachidonic acid we observed a specific upregulation in the mRNA of Pdk4 (Figure 5i) and Hmgcs2 (Figure 5j).

Thus, the decrease in lysophosphatidylcholines and in phosphatidylcholines–containing long chain unsaturated fatty acid causes myelin degeneration in the absence of l‐pgds in vivo in Schwann cells and in vitro when L‐PGDS is not enzymatically active. Moreover, our in vitro data suggest that accumulation of Arachidonic acid, the substrate of L‐PGDS, is likely sufficient to cause myelin alteration, by possibly altering Schwann cell metabolism.

2.7. Loss of l‐pgds Enzymatic Activity Determines Metabolic Rewiring Causing Myelin Degeneration

In our in vivo analyses we observed impairment in glycolysis as well as in the Krebs cycles starting at 6 months of age in l‐pgds ^−/− mice. Interestingly, in MPZ‐Cre//l‐pgds ^flx/flx mice, acetyl‐CoA levels started to diminish at 10 months (ctrl: 0.650 ± 0.15 vs. mutant: 0.269 ± 0.077 p = 0.0171) and even more significantly at 14 months (ctrl: 0.508 ± 0.032 vs. mutant: 0.176 ± 0.033 p = 0.0009), indicating an impairment of Krebs cycle in sciatic nerves of mice lacking glial L‐PGDS expression.

To clarify the origins of these metabolic alterations, we performed metabolic flux analyses on Schwann cells—DRG neuronal myelinated cocultures in which we inhibited L‐PGDS enzymatic activity as above described. 24 h before cells' harvesting, we added either 1 mM [U‐¹³C₆]‐glucose, 2 mM [U‐¹³C₅]‐glutamine, 200 μM [U‐¹³C₁₆]‐Palmitate or 200 μM [U‐¹³C₂]‐Sodium acetate to regular culture medium, then we followed labeled metabolites by LC–MS/MS analyses. To avoid any confounding results, we performed these experiments in physiological conditions, thus maintaining unaltered both glucose and glutamine final concentrations.

2.7.1. Glucose Metabolism

We did not detect significant differences in glycolysis flux (Figure 6a); specifically, M + 6 glucose, M + 6 glucose–6P, M + 6 fructose–bis–P isotopologues of the preparatory stage of glycolysis were comparable to controls, as well as pay–off phase isotopomers M + 3 GAP/DHAP, M + 3 phosphoenolpyruvate (PEP) and M + 3 pyruvate. Of note, we observed a significant decrease in the acetyl‐CoA (Figure 6c) deriving from [U–¹³C₆]–glucose that could be either due to excessive consumption or limited production. Since citrate M + 2 and α–ketoglutarate M + 2 levels were comparable in AT‐56 and DMSO treated myelinated cocultures (Figure 6e), the decrease in acetyl–CoA could be caused by a limited production rather than excessive consumption.

FIGURE 6.

In vitro flux analyses: Glucose and L–glutamine metabolism. Schwann cell—neuronal myelinated cocultures were allowed to myelinate for 14 days then treated for an additional 7 days with 25 μM AT‐56 or with DMSO as vehicle control. Each labeled metabolite was added 24 h before processing the samples in physiological conditions. The acetyl‐CoA deriving from glucose and L–glutamine metabolism does not sustain the Krebs cycle in AT‐56 treated cocultures. (a) Metabolic fluxes scheme for [U–¹³C₆]‐glucose metabolism (glycolysis) showing the distribution of labeled carbons. (G6P: Glucose‐6‐phosphate; FBP: Fructose‐6‐phosphate; GAP: Glyceraldehyde 3‐phosphate; DHAP: Dihydroxyacetone phosphate; PEP: Phosphoenolpyruvate). (b) Metabolic fluxes scheme for [U–¹³C₆]–glucose metabolism (pentose phosphate) showing the distribution of labeled carbons. (c) Graph showing the relative abundance of the labeled metabolites deriving from [U–¹³C₆]‐glucose, presented as a mass distribution vector (MDV). N = 4 different independent coculture experiments. Error bars represent mean ± s.e.m. (Unpaired t‐test. M + 5 FBP *p = 0.0444, t = 2.534, df = 6; M + 2 pyruvate *p = 0.0198, t = 3.15, df = 6; M + 2 acetyl–CoA ***p = 0.0002, t = 8.218, df = 6; all other metabolites tested not significantly different). (d) Metabolic fluxes scheme for [U–¹³C₆]‐glucose metabolism (Krebs Cycle) showing the distribution of labeled carbons. (e) Graph showing the relative abundance of the labeled metabolites deriving from [U–¹³C₆]‐glucose and entering the Krebs cycle, presented as a mass distribution vector (MDV). N = 4 different independent coculture experiments. Error bars represent mean ± s.e.m. (Unpaired t‐test. M + 2 oxalacetate **p = 0.0025, t = 4.966, df = 6; all other metabolites tested not significantly different). (f) Metabolic fluxes scheme for [U–¹³C₅]‐glutamine metabolism showing the distribution of labeled carbons. (g) Graph showing the relative abundance of the labeled metabolites deriving from [U–¹³C₅]‐glutamine presented as a mass distribution vector (MDV). N = 4 different independent coculture experiments. Error bars represent mean ± s.e.m. (Unpaired t‐test. M + 5 α–ketoglutarate *p = 0.0478, t = 2.334, df = 8; M + 4 fumarate *p = 0.0403, t = 2.445, df = 8; M + 4 malate **p = 0.0082, t = 3.492, df = 8; M + 4 citrate ***p = 0.0006, t = 5.424, df = 8; M + 2 acetyl–CoA ***p = 0.0001, t = 7.112, df = 8; M + 3 oxalacetate *p = 0.0205, t = 2.879, df = 8; M + 3 malate ***p = 0.0005, t = 5.703, df = 8; M + 3 fumarate **p = 0.0057, t = 3.743, df = 8; all other metabolites tested not significantly different)

Analyses of different isotopomers of the same metabolites originating from the pentose phosphate pathway revealed an accumulation of M + 2 pyruvate (Figure 6b,c), which is consistent with the decrease in Acetyl CoA production (Figure 6c). These results are corroborated by a limited activity of PDH, which correlates with the observed transcriptomic increase in Pdk4 expression. PDK4, in fact, inhibits PDH activity and limits acetyl–CoA production from pyruvate, thereby regulating metabolite flux through the Krebs cycle. Interestingly, despite the reduction in acetyl–CoA, all other metabolic intermediates of the Krebs cycle were not altered (Figure 6d,e). Indeed, we observed an increase in M + 2 oxaloacetate (Figure 6e) that is synthetized from pyruvate through pyruvate carboxylase, an alternative metabolic strategy to replace oxaloacetate, occurring when PDH is blocked (Jitrapakdee et al. 2008).

2.7.2. L–Glutamine Metabolism

Next, we followed the metabolic fate of [U‐¹³C₅]‐glutamine in Schwann cells—neuronal myelinated cocultures treated with 25 μM AT‐56 as compared to DMSO vehicle only treated controls. Glutamine is catabolized by either the mitochondrial oxidative pathway or by the cytosolic reductive branch. In AT‐56 treated cocultures, we observed a significant decrease in M + 5 α‐ketoglutarate, an isotopologue common to both glutamine catabolic pathways. Similarly, we found a reduction in M + 4 fumarate, M + 4 malate, and M + 4 citrate isotopologues of the oxidative glutamine pathways (Figure 6f–g). These data indicate limited glutamine oxidative catabolism in the Krebs cycle. Further, we observed an overall decrease in the activity of the glutamine reductive pathway, as M + 3 fumarate, M + 3 malate, and M + 3 oxalacetate levels were also reduced, while M + 5 citrate was unaffected in the absence of L‐PGDS (Figure 6g). Of note, M + 2 acetyl‐CoA levels, resulting from both glutamine catabolic pathways, were reduced (Figure 6g). Thus, blocking L–PDGS enzymatic activity impairs both glutamine oxidative and reductive metabolism.

2.7.3. Fatty Acid β–Oxidation

The decrease in acetyl‐CoA production deriving from glucose and glutamine metabolism, coupled to in vivo lipids' dysregulation and Acsl3 and Carnitine Palmitoyltransferase 1A (Cpt1a) upregulation observed in transcriptomic analyses, would suggest an increase in fatty acids β–oxidation. Acsl3, in fact, binds specifically free long‐chain fatty acids to CoA determining either their degradation or incorporation in cellular lipids. On the contrary, Cpt1a catalyzes the transfer of the acyl group of long‐chain fatty acid‐CoA conjugates onto carnitine, an essential step for mitochondrial uptake of long‐chain fatty acids and their β‐oxidation. Thus, to determine whether in the absence of functional L‐PGDS β‐oxidation is hampered, we followed its metabolic flux by treating degenerating cocultures with 200 μM [U‐¹³C₁₆]‐palmitate (Figure 7a).

FIGURE 7.

Fatty acids β‐oxidation in the absence of L‐PGDS. (a) Metabolic fluxes scheme for [U–¹³C₁₆]‐palmitate metabolism showing the distribution of labeled carbons. (b) Graph showing the relative abundance of the labeled metabolites deriving from [U–¹³C₁₆]–palmitate presented as a mass distribution vector (MDV) in Schwann cell‐neuronal cocultures myelinated for 14 days and then treated for an additional 7 days with 25 μM AT‐56 or with DMSO as vehicle control. Labeled palmitate was added 24 h before processing the samples in physiological conditions. The acetyl‐CoA deriving from β‐oxidation metabolism does not sustain the Krebs cycle in AT‐56 treated cocultures. N = 3 different independent coculture experiments. Error bars represent mean ± s.e.m. (Unpaired t‐test. M + 2 Acetyl‐CoA *p = 0.0225, t = 3.614, df = 4; all other metabolites were not significantly different). (c) Graph showing levels of acyl‐carnitines in Schwann cell—neuronal myelinated cocultures treated with 25 μM AT‐56, normalized over the total protein content. N = 3 different independent coculture experiments. Error bars represent mean ± s.e.m. (Unpaired t‐test. C0 *p = 0.0413, t = 2.967, df = 4; C2: **p = 0.0085, t = 4.824, df = 4; C3: **p = 0.0082, t = 4.877, df = 4; C4: *p = 0.0168, t = 3.952, df = 4; C5: *p = 0.0189, t = 3.814, df = 4; C6: *p = 0.022, t = 3.638, df = 4; C12: **p = 0.0011, t = 8.485, df = 4; C16:1: *p = 0.0313, t = 3.254, df = 4). (d) Graph showing the ratio between C2/C0 acyl‐carnitines, a hallmark of fatty acid β oxidation, in Schwann cell–neuronal myelinated cocultures treated with 25 μM AT‐56 or with DMSO as vehicle control. N = 3 different independent coculture experiments. β oxidation is reduced when L‐PGDS is inactive. Error bars represent mean ± s.e.m. (Unpaired t‐test. *p = 0.0107, t = 4.518, df = 4). (e) Graph showing levels of acyl‐carnitines in 14 months old wild type littermate controls (WT) and MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves, normalized over the total protein content. N = 4 MPZ‐Cre//l‐pgds ^flx/flx mice; N = 3 littermate controls (WT) mice. Error bars represent mean ± s.e.m. (Unpaired t‐test. C0 **p = 0.003, t = 5.357, df = 5; C2: ***p = 0.0001, t = 10.8, df = 5; C3: *p = 0.0267, t = 3.106, df = 5; C5: *p = 0.0466, t = 2.628, df = 5; C6: **p = 0.0062, t = 4.535, df = 5; C8: *p = 0.0149, t = 3.639, df = 5; C10: *p = 0.0168, t = 3.529, df = 5; C12: **p = 0.0032, t = 5.298, df = 5; C14: ***p = 0.0005, t = 8.064, df = 5; C16: **p = 0.001, t = 6.81, df = 5; C16:1: ***p = 0.0002, t = 9.699, df = 5; C18: **p = 0.0031, t = 5.327, df = 5; C18:1 **p = 0.0026, t = 5.576, df = 5). (f) Graph showing the ratio between C2/C0 acyl‐carnitines, a hallmark of fatty acid β oxidation, in 14 months old wild type littermate controls (WT) and MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves. N = 4 MPZ‐Cre//l‐pgds ^flx/flx mice; N = 3 littermate controls (WT) mice. β oxidation does not change in the absence of L‐PGDS. Error bars represent mean ± s.e.m. (Unpaired t‐test, p = 0.1824 n.s., t = 1.548, df = 5). (g) Representative immunofluorescence of 14 days myelinated Schwann cell—neuronal cocultures treated for an additional 9 days with 25 μM AT‐56, in the presence or absence of 50 μM etomoxir, a specific Cpt1a inhibitor. At the end of the treatment, cultures were fixed and stained for MBP (white) and non‐phosphorylated neurofilament, SMI–32 (fluorescein). Unlike vehicle‐only treated cocultures (DMSO + 50 μM etomoxir), blocking l‐pgds and Cpt1a causes myelin and axonal degeneration. Bar: 100 μm. (h) Representative western blotting analyses of 14 days myelinated Schwann cell—neuronal cocultures treated for an additional 9 days with 25 μM AT‐56, in the presence or absence of 50 μM etomoxir, a specific Cpt1a inhibitor. Lysates were tested for phosphorylated neurofilament (SMI–31) and Calnexin (Clnx) as a loading control. Blocking l‐pgds and Cpt1a significantly impairs neurofilament expression.

Levels of acetyl‐CoA deriving from palmitate β‐oxidation were reduced (Figure 7b), suggesting that this metabolite could be either more consumed or less produced from fatty acids catabolism. Also in this case, we did not observe increased amounts in any of the Krebs cycle metabolites deriving from labeled palmitate (Figure 7b), making unlikely that the reduction in acetyl‐CoA is due to excessive consumption in this pathway. To corroborate these results, we quantified the acylcarnitine pool in vitro in Schwann cell–neuronal cocultures treated with AT‐56 (Figure 7c) and in vivo in MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves' lysates (Figure 7e). However, in both conditions, the ratio of acetyl‐carnitine (C2)/free carnitine (C0), a hallmark of overall fatty acid β‐oxidation, was unchanged or even decreased (Figure 7d,f).

Palmitate tracing experiment indicates that acetyl‐CoA levels were reduced in cocultures in the absence of functional L‐PGDS (Figure 7b). To understand the importance of fatty acid derived acetyl–CoA, we inhibited fatty acid entrance into mitochondria by blunting Cpt1a enzymatic activity, and hence reducing mitochondrial acetyl‐CoA production, in Schwann cells neuronal cocultures treated with AT‐56. Addition of 50 μM Etomoxir, a specific Cpt1a inhibitor (O'Connor et al. 2018), together with 25 μM AT‐56 to already myelinated cocultures, strikingly compromised axonal integrity as shown by immunohistochemical analyses for MBP and non‐phosphorylated neurofilament (SMI32) (Figure 7g). Notably, in these conditions, the expression level of phosphorylated neurofilament (SMI31) was also significantly hampered (Figure 7h).

Collectively, these results indicate that in the absence of a catalytically active L‐PGDS, fatty acids‐derived acetyl‐CoA is required by Schwann cells to support axonal integrity.

2.8. Krebs Cycle Is Sustained by Ketone Bodies and Acetate Consumption in the Absence of l‐pgds

The acetyl–CoA deriving from fatty acid β‐oxidation, which does not directly replenish the Krebs cycle (Figure 7a,b) is probably used in other mitochondrial metabolic pathways such as ketone bodies biosynthesis. All of this is supported by the upregulation in the expression of Hmgcs2 we observed in vitro and in vivo in MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves and of Bdh1 in l‐pgd ^−/− sciatic nerves (Figure 5b–d). Hmgcs2 catalyzes the condensation of acetyl‐CoA into acetoacetyl‐CoA to form HMG‐CoA, which, in mitochondria, drives β‐hydroxybutyrate synthesis, a ketone body carrier of energy to peripheral tissues in conditions of energy deprivation (Han et al. 2020; Newman and Verdin 2017). In turn, Bdh1, is a key enzyme for the synthesis of ketone bodies, as it catalyzes the reversible interconversion of Acetoacetate and β‐hydroxybutyrate (Ma et al. 2025).

To assess whether in the absence of functional L‐PGDS, Schwann cells and neurons might rely on ketone bodies as energy source, we measured β–hydroxybutyrate and acetoacetate levels in vivo in MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves and in AT‐56 treated cocultures. Interestingly, levels of β–hydroxybutyrate were significantly upregulated in 11 months MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves' lysates (Figure 8a), suggesting that the release of β–hydroxybutyrate might accumulate in sciatic nerves to sustain energy production in the absence of l‐pgds. Of note, in conditioned media of AT‐56 treated cocultures we observed a reduction of β–hydroxybutyrate (Figure 8b), suggesting that this metabolite might have been used as source of energy to maintain axonal integrity in vitro (Figure S1).

FIGURE 8.

In the absence of L‐PGDS myelinating Schwann cells increase ketone bodies synthesis and consumption. (a) Graph showing that β‐hydroxybutyrate levels are significantly increased in lysates of 11 months old MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves as compared to wild type littermate controls (WT). N = pool of 2 nerves/mice, 3 different mice/genotype. Error bars represent mean ± s.e.m. (Unpaired t‐test *p = 0.0157, t = 4.029, df = 4). (b) Graph showing that β‐Hydroxybutyrate levels are significantly reduced in the supernatant of Schwann cell—neuronal myelinated cocultures treated with 25 μM AT‐56 as compared to DMSO vehicle controls, but not detectable in corresponding cell lysates. N = 5 different cultures/conditions (cell media) N = 3 different cultures/conditions (cell lysates). Error bars represent mean ± s.e.m. (Two‐way Anova. **p = 0.0077, F(4,4) = 18.44, df = 4). (c) Graph showing that acetoacetate levels are significantly reduced in Schwann cell–neuronal myelinated cocultures treated with 25 μM AT‐56 as compared to DMSO vehicle controls. N = 3 different independent coculture experiments. Error bars represent mean ± s.e.m. (Unpaired t‐test. **p = 0.0019, t = 3.748, df = 4). (d) Graph showing that succinyl–CoA levels are significantly reduced in Schwann cell—neuronal myelinated cocultures treated with 25 μM AT‐56 as compared to DMSO vehicle controls. N = 3 different independent coculture experiments. Error bars represent mean ± s.e.m. (Unpaired t‐test. *p = 0.0189, t = 3.189, df = 4). (e) Graph showing that the energy balance is similar in 14 months old MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves as compared to wild type littermate controls (WT). N = 4 MPZ‐Cre//l‐pgds ^flx/flx mice; N = 3 littermate controls (WT) mice. Error bars represent mean ± s.e.m. (f). Graph showing that the energy balance is comparable in Schwann cell–neuronal myelinated cocultures treated with 25 μM AT‐56 or with DMSO, as vehicle control. N = 4 different independent coculture experiments. Error bars represent mean ± s.e.m. (g) Metabolic fluxes scheme for [U–¹³C₂]‐sodium acetate metabolism showing the distribution of labeled carbons. (h) Graph showing the relative abundance of the labeled metabolites deriving from [U–¹³C₂]‐acetate presented as a mass distribution vector (MDV) in Schwann cell—neuronal cocultures myelinated for 14 days treated for an additional 7 days with 25 μM AT‐56 or with DMSO as vehicle control. Labeled acetate was added 24 h before processing the samples in physiological conditions. N = 3 different independent coculture experiments. Error bars represent mean ± s.e.m. (Unpaired t‐test. M + 2 acetyl–CoA *p = 0.02, t = 3.748, df = 4; M + 2 α–ketoglutarate ***p = 0.0009, t = 8.953, df = 4; M + 2 succinate **p = 0.0034, t = 6.217, df = 4; M + 2 oxalacetate *p = 0.0216, t = 3.66, df = 4; M + 4 citrate *p = 0.0215, t = 3.666, df = 4; M + 3 α–ketoglutarate **p = 0.0027, t = 6.628, df = 4; M + 3 fumarate *p = 0.0496, t = 2.784, df = 4; M + 3 oxalacetate ***p = 0.001, t = 8.688, df = 4; M + 3 citrate **p = 0.0048, t = 5.657, df = 4; all other metabolites tested not significantly different). (i) Representative immunofluorescence of wild–type myelinated Schwann cells neuronal cocultures infected with Hmgcs2 sh372 or scramble shRNA (shscr). Three days post infection cocultures were supplemented with ascorbic acid for 10 days to allow myelination and then treated with 25 μM AT‐56 or vehicle only, as control, for an additional 7 days. At the end of the treatment, cultures were fixed and stained for GFP to visualize infected cells (fluorescein), non‐phosphorylated neurofilament SMI–32 (rhodamine), and MBP (far‐red). Ablation of Hmgcs2 coupled to l‐pgds block further hampers myelin integrity and determines axonal instability. N = 3 independent biologic replicates. Bar: 100 μm.

In our analyses, we also observed a significant decrease in intracellular acetoacetate (Figure 8c) along with reduced levels of succinyl–CoA, a key CoA donor in the ketone bodies utilization pathway (Figure 8d), suggesting that acetoacetate produces acetyl–CoA likely re‐entering the Krebs cycle, at least in vitro. In agreement, we found no alterations in ATP energy charge in MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves (Figure 8e) and in AT‐56 treated cocultures (Figure 8f). These results, along with the upregulation of Acsl3 and of Cpt1a expression and activity, strongly hint at a high rate of fatty acid β‐oxidation of myelin lipids to sustain ketone bodies biosynthesis and their utilization as energy source. Moreover, since the acetyl‐CoA deriving from glucose and L‐glutamine does not sustain the Krebs cycle (Figure 6), we hypothesized that the necessary acetyl‐CoA might derive from other sources. To determine whether the acetate metabolism (Figure 8g), an important metabolic route of the nervous system when glucose is not available (Bose et al. 2019), could contribute to Krebs cycle functioning, we supplemented Schwann cell neuronal cocultures treated with AT‐56 for 7 days, with 200 μM [U–¹³C₂]‐acetate using the same experimental paradigm above described. Although we detected a decrease in M + 2 acetyl–CoA, we surprisingly observed a strong increase in M + 3 citrate, M + 2 and M + 3 α–ketoglutarate, M + 3 fumarate, and M + 2 and M + 3 oxaloacetate (Figure 8h) deriving from labeled acetate, suggesting a more pronounced acetate consumption, which ensures, at least, Schwann cells and neuronal survival.

As above mentioned, one of the key enzymes for ketone bodies formation, is Hmgcs2, whose expression is upregulated in vitro and in vivo in the absence of L‐PGDS (Figure 5). One of the most striking results of this and our former study (Trimarco et al. 2014) is the morphological alteration in myelin not accompanied by axonal swallowing and/or neuronal cell death. To confirm that, in the absence of L‐PGDS, Schwann cells activate an alternative pathway relying on ketone bodies production to sustain the axo‐glial unit integrity, we ablated Hmgcs2 expression by lentiviral mediated knocked down. Established Schwann cells–DRG neuronal cocultures, were infected with three different shRNA specifically designed against Hmgcs2. Three days post infection, cocultures were supplemented with ascorbic acid for additional 10 days to allow myelination and then treated with 25 μM AT‐56 or vehicle only, as control, for additional 7 days. At the end of the treatment, cultures were fixed and stained for MBP, non‐phosphorylated neurofilament and GFP to visualize infection's levels. As expected, most cells infected by the shRNAs were Schwann cells (Figure 8i). Importantly, knock down of Hmgcs2, which was validated by qRT–PCR analyses (Figure S5a), not only enhances myelin breakdown, but also dramatically hampers axonal integrity (Figure 8i and Figure S5b).

Collectively, we posit that in the absence of glial L‐PGDS, myelin lipid composition is altered, and with time, glial cells undergo a rewiring in metabolism. Indeed, while in early phases of myelin degeneration Schwann cells rely on glucose to sustain their energy balance, when degeneration continues, Schwann cells switch to ketone bodies and acetate utilization to sustain the Krebs cycle and uphold the axo‐glial unit.

3. Discussion

Metabolic homeostasis is central to the maintenance of a healthy and functional organ. In the nervous system, this is particularly relevant, as the cross communication between glial cells and neurons is essential to integrate a continuous transmission of electric impulses from the center to the periphery and vice versa (Nave and Trapp 2008; Saab and Nave 2017).

To thoroughly investigate the complex regulation of metabolism in nerves, we combined multiple approaches. We described and analyzed metabolites and lipids in vivo in sciatic nerves of three mutant mice at different age: a complete null L‐PGDS, a Schwann cell specific and a motor neuron specific mutant. Next, we performed a systematic analysis of different metabolic pathways in vitro using the Schwann cell neuronal coculture system. Though in vitro we could partly reproduce the complexity of the interactions occurring in nerves, where in addition to glial cells and axons, other cells can also contribute to metabolic changes, these studies were instrumental to confirm the key changes in metabolism that we identified in sciatic nerves.

We also report that L‐PGDS expressed in Schwann cells or neurons exert different roles, with the neuronal enzyme regulating the thickness of the myelin sheath, in agreement with the model proposed in (Trimarco et al. 2014). In turn, glial L‐PGDS is key in controlling myelin homeostasis. Our results indicate that this enzyme in Schwann cells is crucial to maintain myelin lipid integrity, as, in MPZ‐Cre//l‐pgds ^flx/flx mice, but not in ChAT‐Cre//l‐pgds ^flx/flx mice, we observed a drastic change in myelin lipid composition. Thus, we propose that lysophosphatidylcholines, containing omega‐6 fatty acids, which are at the basis of the prostaglandin synthesis, are released from myelin and likely accumulate in Schwann cells cytoplasm. Accordingly, in vitro increased levels of Arachidonic acid are sufficient to cause myelin degeneration and to dysregulate the expression of Pdk4 and Hgmcs2, key enzymes in controlling cellular energy metabolism.

Notably, changes in myelin lipids' composition correlate with accumulation of morphological changes, as in oligodendrocytes (Asadollahi et al. 2024). Thus, we propose that L‐PGDS might be functional either to maintain intact the myelin structure or be part of a coordinated mechanism controlling myelin homeostasis and remodeling. Since we did not observe any significant overt morphological phenotype in mice before 6‐months of age in l‐pgds ^−/− mice and even later in MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves, we favor the last assumption, although we cannot exclude that lipids' remodeling could be more efficient in younger animals.

Based on the results of our study, we envisage two different hypotheses to explain how the absence of L‐PGDS alters myelin morphology. Although in adult mice myelin lipids' remodeling could be less efficient, it is also possible that the demand of PGD2 in Schwann cells induces a continuous release of lipids from the plasma membrane, reaching a level above which the structural integrity of myelin cannot be maintained. Regardless of the mechanisms, we have identified the synthase L‐PGDS as central in maintaining myelin homeostasis. In future studies it might be important to define whether L‐PGDS cooperates with other molecules in this task, and whether they could be pharmacologically targeted to minimize myelin alterations. Interestingly, dietary supplementation of phosphatidylcholine and phosphatidylethanolamine, the main class of lipids we observed reduced in our studies, improved myelination and nerve conduction velocity in a rat model of CMT1A (Fledrich et al. 2018).

A remaining open question is the final destiny of the lipids released from the plasma membrane. Accumulation of arachidonic acid in Schwann cells is sufficient to upregulate pdk4 and Hmgcs2 expression, whose expression is also upregulated in MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves. Upregulation of pdk4 might therefore be instrumental to limit acetyl‐CoA production by inactivating the pyruvate dehydrogenase, slowing down Schwann cells' glycolysis. Accordingly, we observed a general attenuation in the use of glucose as the main source to preserve the axo‐glial unit both in vitro and in vivo. In this context, it is important to note that previous studies have shown that the acetyl‐CoA deriving from pyruvate is dispensable for myelin maintenance and axon integrity (Della‐Flora Nunes et al. 2017).

Surprisingly, in l‐pgds ^−/− mice we also found a less efficient lactate production and Schwann cells rely more on acetate. Though lactate is generally transferred to axons to support axonal metabolic demand (Jha and Morrison 2020), its role in sustaining the energetic metabolism of peripheral nerves is complex and controversial (Deck et al. 2022; Jia et al. 2021). It is thus tempting to speculate that acetate might be a specific metabolite used to support neurons and glial cells as a substitute for lactate in adulthood (Jia et al. 2021). Whether this switch in metabolites' production occurs only in adulthood or it is present also in development and/or in pathological settings requires additional investigations.

Our results implicate that continuous remodeling of plasma membrane myelin lipids, due to the absence of L‐PGDS, drives Schwann cells towards a change in their metabolic activity. The decreased mitochondrial fatty acid‐derived acetyl‐CoA is likely used to generate ketone bodies to boost ketolysis (Puchalska and Crawford 2017), which serves as an alternative energetic route for lipid synthesis, especially in the nervous system (McMullen et al. 2023; Puchalska and Crawford 2017; Robinson and Williamson 1980; Silva et al. 2022). In both CNS (Koper et al. 1981) and PNS (Clouet and Bourre 1988), ketone bodies are preferentially used for myelin lipid synthesis. Although we did not observe any increase in malonyl‐CoA consumption in the absence of L‐PGDS in flux analyses, we cannot exclude that acetate might foster new lipid synthesis in vivo or might be preferentially activated in young mutant mice. Of note, metabolic incorporation of acetate into PNS myelin is prevalently used very early in development and after injury (Yao and Cannon 1983). Further, in Trembler mutant mice, which are characterized by a hypomyelinating peripheral neuropathy, acetate represents the main source for lipid synthesis (Clouet and Bourre 1988). Ketolysis and acetate might therefore be used also in our model to sustain myelin remodeling.

Since lipids can function as second messengers in the cells (Krey et al. 1997; Marion‐Letellier et al. 2016), it is also possible that those released from the plasma membrane in the absence of L‐PGDS act onto PPARγ to promote gene transcription. Many of the genes we found upregulated in our transcriptomic analyses are implicated in the regulation of lipid metabolic processes, which is critical to maintain cellular homeostasis and whose alterations are frequent hallmarks of several acquired and genetic neuropathies (Silva et al. 2025). Finally, we cannot exclude that both systems co‐exist. Hence, released lipids might on one side promote gene transcription to counteract continuous PNS myelin remodeling, and in parallel limit excessive production of acetyl–CoA in the mitochondria favoring ketolysis. In turn, in young mice, ketolysis could foster new lipid synthesis thus preventing excessive myelin remodeling.

It should be noted that the metabolic shift occurs essentially in glial cells. Accordingly, while in ChAT‐Cre//l‐pgds ^flx/flx sciatic nerves myelin does not degenerate and there are no alterations in lipid profile, MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves, instead, presented aberrant myelin as well as metabolic changes overlapping those observed in l‐pgds ^−/− mice. Though metabolic defects in MPZ‐Cre//l‐pgds ^flx/flx and l‐pgds ^−/− mice are similar, they also present important differences. In MPZ‐Cre//l‐pgds ^flx/flx nerves lipid alterations are significantly evident at 14 months, while they are already apparent at 6 months in l‐pgds ^−/− mice. At metabolic level, both mutants present a reduction in glycolysis, with an abrupt acetyl–CoA reduction in MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves. These differences are not surprising as L‐PGDS is expressed in several cells in addition to Schwann cells and neurons (Gerber et al. 2021; Joo et al. 2007). Alternatively, it is possible that the hypomyelinating phenotype present in l‐pgds ^−/− mice but not in MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves might render myelin more susceptible to changes in lipid composition, thus enhancing the overall process. Yet, our data indicate that ablation of glial L‐PGDS in vitro and in vivo is sufficient to trigger a metabolic shift in peripheral nerves. Based on these results, we posit that the metabolic rewiring occurs in Schwann cells, especially the one attributable to ketogenesis fueled by fatty acids derived from degenerating myelin, which might represent an adaptive response to ongoing morphological alterations that are ensuing in Schwann cells because of continuous myelin remodeling.

Our study also reveals that in the absence of L‐PGDS Schwann cells rely on acetate to sustain the Krebs cycle and that levels of acetoacetate, succinyl‐CoA, and β‐hydroxybutyrate are reduced in the absence of L‐PGDS, further supporting the notion that Schwann cells undergo enhanced ketolysis when L‐PGDS is not expressed or inactive. Notably, levels of β‐hydroxybutyrate were diminished in cell culture media, but upregulated in MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves (Figure 8), suggesting that the release of β‐hydroxybutyrate could accumulate in the nerve where it is used to produce ketone bodies and fuel the Krebs cycle in addition to indicating that the inter–conversion between acetoacetate and β‐hydroxybutyrate is shifted towards the latter. Although we do not have formal proof for the transfer of β‐hydroxybutyrate from Schwann cells to neurons to sustain their energy demand and the integrity of the axo‐glial unit, we have several pieces of evidence supporting this hypothesis: (i) in the absence of a functional L‐PGDS we did not observe any impairment in neuronal survival (Trimarco et al. 2014) (and this study); (ii) ablation of Hmgcs2, a key enzyme for ketone bodies synthesis, as well as (iii) block of lipid transfer into mitochondria, not only worsens myelin degeneration, but drastically compromises axonal stability. All these observations support our hypothesis that ketone bodies synthesis in Schwann cells in the absence of L‐PGDS is critical to sustain the axo‐glial unit. Interestingly, ketogenic diet reduces metabolic induced‐allodynia and promotes nerve growth in the epidermis (Cooper et al. 2018). Further, it improves CNS myelination and increases the number of oligodendrocytes in a model of Pelizaues‐Merzbacher disease (Stumpf et al. 2019).

In conclusion, we have unveiled a new pathway connecting prostaglandin activity to Schwann cell metabolism and myelin homeostasis in adulthood. Based on the results of our studies, we propose that glial L‐PGDS is part of a coordinated program aiming at preserving myelin integrity. Importantly, our study suggests that Schwann cell myelin lipids, particularly fatty acids, are a valuable reservoir to produce ketone bodies, which, together with acetate, could represent the adaptive substrates they can rely on under threatening circumstances to sustain the axo‐glial unit and preserve the integrity of the PNS.

4. Materials and Methods

4.1. Mice and Genotyping

All experiments were performed on male mutants and wild type mice in a C57/Bl6 congenic background following protocols approved by the Institutional Animal Care and Use Committee of San Raffaele Hospital and by the Italian Minister of Health (Protocol number 973). Generation of l‐pgds ^−/− , MPZ‐CRE, ChAT‐Cre mice and genotypes’ determination were previously described (Eguchi et al. 1999) (Feltri et al. 1999; Pellegatta et al. 2022; Rossi et al. 2011). l‐pgds ^flx/flx mice were genotyped by PCR analysis on genomic DNA, using the following primers: 5′‐GGG CAC TGT CAG CCT GTG TGC TTG TGC‐3′ (forward), 5′‐CCA CAC AGG TCC TAG CAG CAT GCC TC‐3′ (reverse). Cycling conditions were: 94°C for 60 s, 60°C for 60 s, and 72°C for 60 s (29 cycles), followed by a 10 min extension at 72°C.

For ChAT‐Cre//l‐pgds ^flx/flx mice, the presence of the null allele was determined by PCR on genomic DNA extracted from ventral and dorsal spinal cord regions using the following primers: 5′‐GGG CAC TGT CAG CCT GTG TGC TTG TGC‐3′ (forward LoxP1), 5′‐GGT GAG AGA AGT CAG TCA GAG GGC TGG‐3′ (forward Lox P2) and 5′‐CCT GGC TCC TTG GAG ACC CCT GCT GC–3′ (reverse Lox P2). Cycling conditions were: 94°C for 60 s, 61°C for 60 s and 72°C for 60 s (30 cycles), followed by a 3 min extension at 72°C. Amplified fragments were analyzed on a 2% agarose gel.

4.2. Morphological and Morphometric Analyses

Semi–thin and ultrathin sections were obtained as described (Quattrini et al. 1996). Sciatic nerves were removed and fixed with 2% glutaraldehyde (Electron Microscopy Science) in 0.12 M phosphate buffer, post fixed with 1% osmium tetroxide (Electron Microscopy Science), and embedded in Epon (Epoxy Embedding Medium kit, Sigma–Aldrich). Semi–thin transversal sections (0.5–1 μm thick) were stained with toluidine blue and examined by light microscopy (Olympus BX51). Digitized non‐overlapping images from corresponding levels of the sciatic nerve were obtained with a digital camera (Leica DFC300F) using a 100X objective. Images were merged and transversal sections analyzed using the NIH ImageJ v1.45s software (NIMH Image Library, RRID:SCR_005588). The percentage of myelin aberrations was determined as the number of fibers presenting myelin structural alterations over the total number of fibers in the analyzed section. g‐ratio measurements were performed on non‐overlapping digitalized electron micrographs images as reported in (La Marca et al. 2011) or using the Leica QWIN V3 software.

4.3. Cell Cultures and Treatments

Dorsal root ganglia (DRG) neurons from wild type or l‐pgds ^−/− mice, were isolated from E14.5 embryos and established on rat collagen I (Cultrex) coated glass coverslips or 6‐well plates as previously described (Taveggia et al. 2005). Cells were grown in Neuronal Basal media (Gibco), B27 (Gibco), 4 g/L D‐glucose (Fluka), 2 mM L‐glutamine (Invitrogen), supplemented with 50 ng/mL NGF (B.5017, Harlan Laboratories, Indianapolis, IN). In some experiments, explants were cycled with FUDR to eliminate all non‐neuronal cells. Primary rat Schwann cells were prepared as described in (Taveggia et al. 2005) and maintained in DMEM (Invitrogen), 10% FBS (vol/vol, Invitrogen), 2 mM L‐glutamine (Invitrogen), 2 μM Forskolin (Sigma–Aldrich) and 10 ng/mL rhNRG1 (R&D), until used. Rat Schwann cells were added (200.000 cells per coverslip) to established explant cultures of DRG neurons and cultured in MEM (Invitrogen), 10% FBS (vol/vol, Invitrogen), 2 mM L‐glutamine (Invitrogen), 4 g/L D‐glucose (Fluka) and NGF. Myelination was initiated by supplementing media with 50 μg/mL ascorbic acid (Sigma–Aldrich) and continued for 10–14 days.

Myelin degeneration was obtained by treating myelinated Schwann cell–DRG neuronal co‐cultures with 25 μM AT‐56 (Tocris) in myelinating medium for an additional 7 days, unless otherwise specified. In Arachidonic acid experiments, myelinated rat Schwann cell‐mouse DRG neuronal co‐cultures were treated with different concentrations of Arachidonic acid (Cayman Chemical Company) for 7 days. Briefly, ethanol was evaporated under nitrogen flux and Arachidonic acid was dissolved in DMSO to a final concentration of 164 mM. The fatty acid was bound to BSA with a molar ratio of 1:5 by heating the stock solution 7% BSA/5 mM Arachidonic acid at 37°C for 1 h. The compound was then diluted in myelinating medium.

In experiments in which we blocked Cpt1a activity, we treated already myelinated Schwann cell–DRG neuronal cocultures with 25 μM AT‐56 and 50 μM Etomoxir (Sigma–Aldrich) for 9 days. Cultures were then processed for immunofluorescence analyses or Western Blotting analyses as detailed below.

In metabolomics flux analyses, labeled metabolites were added to degenerating co‐cultures 24 h before collecting the cells. In detail, L–glutamine in the culture medium was completely substituted with 2 mM [U–¹³C₅]‐glutamine, 1 mM [U–¹³C₆]‐glucose was added in medium containing 21.2 mM glucose (final total glucose 22.2 mM), [U–¹³C₁₆]‐Palmitate was added to 200 μM final concentration (together with 1 mM L‐carnitine) and [U–¹³C₂]‐Sodium acetate to 2 mM.

4.4. Isolation and Culture of Mouse Schwann Cells

Sciatic nerves were dissected from postnatal day 3 C57BL/6 wild‐type and L‐PGDS KO mouse pups, by partly modifying the protocol described in (Honkanen et al. 2007), as follows. Groups of 4–6 nerve pairs were transferred for enzymatic digestion into 15‐ml Falcon tubes containing 1 mL of 0.25% trypsin without EDTA (Gibco) and 130 U/mL Collagenase (Sigma–Aldrich) and incubated at 37°C for 50 min. Dissociated cells were collected by centrifugation and resuspended in DMEM (Gibco) supplemented with 3% FBS (Gibco) and 1% penicillin/streptomycin (Gibco) and plated onto 35‐mm Petri dishes that were pre‐coated with 0.01 mg/mL poly‐L‐lysine (Sigma–Aldrich) and 10 ng/μl Laminin (Sigma–Aldrich). After cell adhesion, medium was replaced with a defined medium containing DMEM/F12 (Gibco), 1× N2 supplement (Gibco), 38 ng/mL dexamethasone (Sigma–Aldrich), 10 ng/mL 3,3′,5‐Triiodo‐L‐thyronine (Sigma–Aldrich), 400 ng/mL L‐Thyroxine (Sigma–Aldrich), 1% penicillin/streptomycin (Gibco), and 3% FBS (Gibco). For the first 72 h, cells were treated with the addition of 10 μM Cytosine β‐D‐arabinofuranoside hydrochloride (AraC, Sigma–Aldrich). After 72 h, AraC was removed and the cells grown in defined media for an additional 48 h. To promote Schwann cells growth, cultures were shifted to a growth media containing DMEM (Gibco), 10% FBS (Gibco), 1% penicillin/streptomycin (Gibco), 4 mM L‐glutamine (Gibco), 10 ng/mL rhNRG‐1 (R&D Systems), 2 μM forskolin (Sigma–Aldrich), 10 ng/mL hFGF‐10 (Peprotech), 20 μg/mL bovine pituitary extract (Gibco), and 1× ITS+premix (Corning).

4.5. Mouse Schwann Cells Seeding Onto DRG Neurons and Induction of Myelination

Mouse Schwann cells were incubated for 30 s at 37°C with 0.25% trypsin without EDTA (Gibco) to selectively detach Schwann cells from adherent fibroblasts. Cells were collected, centrifuged at 1000 rpm for 5 min at room temperature and resuspended in a seeding medium containing MEM (Gibco), 5% FBS (Gibco), 1% penicillin/streptomycin (Gibco), 4 mM L‐glutamine (Gibco), 4 g/L D‐glucose (Sigma–Aldrich), 50 ng/mL NGF (Envigo), 1× N2 supplement (Gibco), 2 μM forskolin (Sigma–Aldrich), and 20 μg/mL bovine pituitary extract (Gibco). Cells were then seeded onto previously isolated mouse DRG neurons at a density of 180,000–200,000 cells per DRG. Cocultures of isolated mouse Schwann cells and isolated DRG neurons were maintained in the seeding media for 3 days to allow Schwann cell proliferation. Myelination was then induced by supplementing the seeding media with 50 μg/mL ascorbic acid (Sigma–Aldrich). Medium was changed every other day for 5 weeks until the first myelin internodes were observed. Cultures were then fixed and processed for immunohistochemical analyses as described below.

4.6. Lentiviral Production and Infection

Individual shRNA clones (FE5V3SM11241‐231157720: sh372; FE5V3SM11242‐240857410: sh062; FE5V3SM11242‐243698710: sh362) specifically targeting mouse Hmgcs2 were obtained from Dharmacon (Horizon Discovery). Lentiviral vectors were transfected into HEK293T cells (Dharmacon) and viruses were produced according to the Dharmacon Trans‐Lentiviral Packaging System's instructions. Briefly, cells confluent at 80% were transfected using Ca₂P0₄ and viral particles were collected 3 days post transfection by ultracentrifugation at 4°C, 20.000 rpm for 2 h (Beckman). Collected viral particles were then suspended in DMEM and stored to −80°C, until used. To assess viruses' titer HEK293T cells were infected with different dilutions of each lentiviral vectors for 4 h and the number of turboGFP positive colonies was counted 3 days post infection. The viral titer was calculated as following: n colonies × dilution factor × 40 = TU/mL (transducing unit/mL) (Dharmacon's instructions). Established Schwann cells–DRG neuronal cocultures 1 week post dissection, were infected with lentiviruses at an MOI ranging between 5 and 9 and incubated for 16 h in MEM (Invitrogen), 5% FBS (vol/vol, Invitrogen), 4 mM L–glutamine (Invitrogen), 2 g/L D–glucose (Fluka) and 5 ng/mL NGF. Three days after infection we induced myelination by treating cocultures with 50 μg/mL ascorbic acid for additional 10 days. Cocultures were then supplemented with 25 μM AT56 and DMSO as control for additional 7 days.

4.7. RNA Extraction and Gene Expression Analyses

Cell cultures and sciatic nerves were homogenized in TriPure Reagent (Roche) using Precellys Lysing Kit CK14 and homogenizer (8000 rpm, 60 s at 0°C, twice). RNA extraction was performed according to the manufacturer's instruction.

For RNAseq analyses, we prepared RNA samples from rat Schwann cell–mouse DRG neuronal co‐cultures treated with 25 μM AT56 and DMSO as control (three different biological replicates/condition). Samples were processed at the Biotecnology Department at University of Verona as follows. Total RNA yield was determined using NanoDrop spectrophotometer (Thermo Fisher Scientific). Integrity of RNA samples was assessed using RNA 6000 Nano Kit (Agilent Technologies, Santa Clara, CA, USA) prior to library preparation. All samples showed an RNA integrity number (RIN) > 8. RNAseq libraries were prepared from 2500 ng total RNA using the TruSeq RNA Library Preparation Kit v2 (Illumina, San Diego, CA, USA) after poly(A) capture, according to manufacturer's instructions. Quality and size of RNAseq libraries were assessed by capillary electrophoretic analysis with the Agilent 4200 Tape station (Agilent). Libraries were quantified by real–time PCR against a standard curve with the KAPA Library Quantification Kit (KapaBiosystems, Wilmington, MA, USA). Libraries were then pooled at equimolar concentration and sequenced in 150PE mode on an Illumina NextSeq500 (Illumina). On average 42 million fragments were produced for each sample. Bioinformatic analysis was performed as follows. RNA sequences were aligned separately on mouse and rat deposited exome transcriptome sequences available at NCBI public database (mouse assembly GCF_000001635.24 and rat assembly GCF_000001895.5). We calculated relative gene expression in AT‐56 treated cocultures versus DMSO treated by DESeq2 (Love et al. 2014) and selected genes with a p value < 0.001. We then classified as upregulated differentially expressed genes with fold change (FC) > 1.7 and as down regulated genes with FC < 0.5. From mouse alignments, we identified 18 upregulated and 12 downregulated genes while from rat alignments we identified 24 upregulated and 2 downregulated genes. Transcriptomic data have been deposited to GEO. Data can be retrieved using the following tokens: GSE223246.

300 ng of RNA were reversed transcribed using SSIV Superscript (Invitrogen) following manufacturer's instructions. To assess genes’ expression specifically in Schwann cells, we analyzed RNA total extract from rat Schwann cells–mouse DRG neuronal cocultures using primers pairs specific to rat mRNA sequences. Primers sequences were as follows: Hmgcs2 forward 5′‐CGC AGT CTA CCC AAG TGG TA‐3′ and reverse 5′‐GTC ATC GAG GGT GAA AGG CT‐3′, Acsl3 forward 5′‐AGC AAA GGA GAC ACA TCC GTT‐3′ and reverse 5′‐CAG ATA TTC ATG AAT CGC TGT GTC‐3′, Pdk4 forward 5′‐AAA CCG CCC TTT CCT GAC A‐3′ and reverse 5′‐GTC CCA TAG CCT GAC ATG GAA‐3′, Angptl4 forward 5′‐ACC TTA AGA TAT GGC TGT TTT CTG CTG A–3′ and reverse 5′‐CTG GGA ACC CTA TCT CCA GTC G‐3′, Gapdh forward 5′‐GGT TAC CAG GGC TGC CTT CTC TTG TGA‐3′ and reverse 5′‐CGG AAG GGG CGG AGA TGA TGA CCC T‐3′, TBP forward 5′‐CCT GTT CAG AAC ACC AAT AGT TTA‐3′ and reverse 5′‐GTG GAT ACA ATA TTT TGG AGC TGT‐3′ Real Time–PCR (qRT–PCR) was performed using PowerUp SybrGreen master mix (Applied Biosystems), using 0.25 μM each primer and 5–15 ng cDNA for each reaction in triplicate. Cycling conditions were: 2 min at 50°C 2 min at 95°C followed by 39 cycles of 30 s 95°C, 30 s annealing temperature, 1 min 30 s at 72°C. Annealing temperatures were 60°C (Acsl3, Hmgcs2 and TBP), 61°C (Pdk4), 62°C (Angptl4) and 64°C (Gapdh). Results were analyzed using the StepOne software (Applied Biosystems) according to manufacturer's instructions.

Primers aligning on the mouse genome were used to detect the following transcripts: Mbp, Mpz, l‐pgds, gapdh, Hmgcs2, cpt1A, pdk4, and 36B4. qRT–PCR analyses were performed using Evagreen master mix (Biorad), using 0.5 μM each primer and 15 ng cDNA for each reaction in triplicate. Cycling conditions were: 5 min at 95°C followed by 39 cycles of 30 s at 95°C, 30 s at 60°C, 1 min at 72°C. Results were analyzed using the CFX Manager Software from Biorad according to the manufacturer's instructions. Primers sequences were as follows: Mbp forward 5′‐ACA CAC GAG AAC TAC CCA TTA TGG‐3′ and reverse 5′–GTT CGA GGT GTC ACA ATG TTC TTG‐3′, Mpz forward 5′‐CAC AAC CTA GAC TAC AGT GAC AAC G‐3′ and reverse 5′‐TTC GAG GAG TCC TTC GAA GAT TTG‐3′, l‐pgds forward 5′–GGA GAA GAA AGC TGT ATT GTA TAT GTG C‐3′ and reverse 5′‐TAA AGG TGG TGA ATT TCT CCT TCA G–3′, gapdh forward 5′‐TCA CCA GGG CTG CCA TTT GCA GTG G‐3′ and reverse 5′‐CGG AAG GGG CGG AGA TGA TGA CCC T‐3′, Hmgcs2 forward 5′‐CAG AAT CAG TGG AAG CAA GCT G‐3′ and reverse 5′‐CAG AGT GGT GAG AGA GAA GTG AG‐3′, cpt1A forward: 5′ AGG CCA CTG ATG ATG AAG GAG GG 3′ and reverse: 5′ GTT TGA GTT CCTCAC GGT CTA CC‐3′, pdk4 forward: 5′ GAA AAC CGT CCT TCC TTG ACC 3′ and reverse: 5′‐GTC TGT CCC ATA ACC TGA CAT AGA 3′, 36B4 forward 5′‐AGA TTC GGG ATA TGC TGT TGG‐3′ and reverse 5′‐AAA GCC TGG AAG AAG GAG GTC‐3′.

4.8. Immunofluorescence Analyses

Isolated mouse Schwann cells and isolated DRG neuronal cocultures, as well as organotypic Schwann cell–DRG neuronal cocultures treated with Arachidonic acid or with 25 μM AT‐56 (Cayman Chemicals) and 50 μM Etomoxir (Sigma–Aldrich) were fixed in 4% PFA, permeabilized in 100% methanol at −20°C for 15 min, blocked in 5% BSA (Sigma–Aldrich), 1% donkey serum (Jackson Immuno Research), 0.2% Triton–X (Sigma–Aldrich) in PBS 1X (Gibco).

All antibodies were previously validated for the applications used. Primary antibodies used in immunofluorescence studies were: mouse anti MBP (Covance Research Products Inc. Cat# SMI–94R–100 RRID:AB_510039 and Cat# SMI–99P–100 RRID:AB_10120129, 1:1.000), rat anti MBP hybridoma (diluted 1:2), chicken anti neurofilament M antibody (Covance PKC–593P, 1:1000)), chicken anti beta3 tubulin (Abcam Cat # 41489, RRID: AB_727049, 1:1000), mouse anti Neurofilament SMI32 (BioLegend Cat# 801701, RRID:AB_2315331, 1:1000), rabbit anti turbo GFP (Thermo Fisher Scientific Cat# PA5–22688 Lot# UG2799941 RRID:AB_2540616, 1:500). Secondary antibodies were: mouse anti Alexa Fluo 488 (Thermo Fisher Scientific Cat# A‐21202, RRID: AB 141607, 1:1000) rat anti Alexa Fluor 555 (Thermo Fisher Scientific Cat# A‐21434, Lot#1987272 RRID:AB_2535855, 1:1000)), chicken anti Alexa Fluor 488 (Thermo Fisher Scientific Cat# A‐11039, Lot#2304258, RRID:AB_2534096, 1:1000), goat anti chicken Alexa Fluo 555 (Thermo Fisher Scientific Cat# A‐21437 RRID: AB 2535858, 1:1000) and Rat Alexa fluor 647 (Thermo Fisher Scientific Cat# A21247, Lot#1921562, RRID AB_14778, 1:1000). Nuclei were stained by Hoechst. Tunel assay (Promega) was performed following manufacturer's instructions before staining for MBP and NF. Cocultures treated with DNAse I were used as assay positive control. Slides were examined by epifluorescence by confocal microscopy on a Leica SP5.

4.9. Preparation of Detergent Lysates and Immunoblotting

All sciatic nerves were lysed using the Precellys system (Bertin Instruments) in lysis buffer 2% (wt/vol) SDS, 25 mM Tris pH 7.4, 95 mM NaCl, 10 mM EDTA (all by Sigma–Aldrich), phosphatase inhibitor (PhoStop, Roche), and protease inhibitor (Complete Mini–EDTA free, Roche). Lysates from myelinated Schwann cell–neuronal cocultures were prepared and processed as described in (Trimarco et al. 2014). All lysates were boiled for 5 min at 100°C and centrifuged for 10 min at 16,800×g at 16°C. Supernatants were separated, and proteins were quantified using Pierce BCA protein assay (ThermoFisher Scientific) according to manufacturer's instructions.

Samples (8–25 μg protein extract) were run on 8%–10% acrylamide gel at 100 Volt in Running Buffer (1× Tris–Glycine‐SDS, Biorad) together with Precision Plus Protein Dual Color Standards (Biorad). Proteins were blotted on nitrocellulose membrane at 100 Volt at 4°C. Membranes were blocked in 5% milk in PBST (0.05% Tween in PBS 1X) and incubated for 16 h at 4°C with the following primary antibodies: mouse anti‐MBP (Covance Research Products Inc. Cat# SMI‐94R‐100 Lot# RRID:AB_510039 and Cat# SMI‐99P‐100 Lot# RRID:AB_10120129, 1:4.000), mouse anti‐NF (SMI 31 (BioLegend Cat# 801601, RRID:AB_10122491, 1:1000), chicken anti‐MPZ (Millipore Cat# AB9352 RRID:AB_571090, 1:500), mouse anti‐β‐Tubulin (Sigma‐Aldrich Cat# T4026, RRID:AB_477577), rabbit anti‐Calnexin (Sigma‐Aldrich, Cat# C4731, RRID:AB_476845, 1:2000). Membranes were incubated with secondary antibodies all used 1:10000 and included anti‐mouse IRdye 680 (LI‐COR Biosciences Cat# 926‐68,070, RRID:AB_10956588) and anti‐chicken IRdye 800 (LI‐COR Biosciences Cat# 926‐32,218, RRID:AB_1850023). Quantitative Western Blotting analyses were performed using the Odyssey Infrared Imaging System (LI‐COR Biosciences) according to the manufacturer's instructions. We used the integrated intensity of the fluorescent signal to quantify protein expression of each sample using ImageJ software. Samples were normalized for the housekeeping protein tubulin or vinculin.

4.10. Lipidomics and Metabolomics Analyses

Lipidomics and metabolomics data were obtained by liquid chromatography coupled to tandem mass spectrometry. We used an API–4000 triple quadrupole mass spectrometer (AB Sciex, Farmington, MA, US) coupled to a HPLC system (Agilent) and CTC PAL HTS autosampler (PAL System) for lipidomic analysis. Metabolomics was performed on an API–3500 triple quadrupole mass spectrometer (AB Sciex, Farmington, MA, US) coupled with an ExionLC AC System (AB Sciex, Farmington, MA, US).

4.10.1. Samples Preparation

For lipidomics, sciatic nerves or Schwann cell–DRG neuronal cocultures were homogenized by tissue lyser for 2 min in 250 μL of ice–cold methanol/acetonitrile 50:50. Lysates were spun at 20,000 g for 5 min at 4°C and supernatants passed through a 4 mm regenerated cellulose filter (200 nm Ø). Samples were finally dried under nitrogen flow at 40°C and resuspended in 100 μL in methanol/acetonitrile 50:50.

For metabolomics, sciatic nerves or cocultures homogenates were prepared by tissue lyser disruption for 2 min in 250 μL of ice‐cold methanol/water/acetonitrile 55:25:20 containing [U–¹³C₆]‐glucose (Sigma–Aldrich) 1 ng/μl and [U–¹³C₅]‐glutamine (Sigma–Aldrich) 1 ng/μl as internal standards (IS). Lysates were spun at 15,000×g for 5 min at 4°C and supernatants were passed through a 4 mm regenerated cellulose filter (4 mm Ø, Sartorius). Samples were then dried under nitrogen flow at 40°C and resuspended in 125 μL of methanol/water 70:30 for subsequent analyses.

Schwann cell–DRG neuronal cocultures were lysed in methanol/acetonitrile 50:50. LC–MS/MS runs were performed on an API–4000 triple quadrupole mass spectrometer (AB Sciex, Farmington, MA, US) coupled to a HPLC system (Agilent) and CTC PAL HTS autosampler (PAL System). Results were obtained after correction for natural abundance of ¹³C and expressed as Mass Isotopomer Distribution (MID). MultiQuant software (version 3.0.3, AB Sciex, Farmington, MA, US) was used for data analysis and peak review of chromatograms.

All results were normalized over the total protein content as determined by BCA assay on protein fractions after metabolites/lipids extraction. In vivo analyses were performed on sciatic nerves from at least 4 l‐pgds ^−/− and wild type littermate control mice, at 4, 6 and 8 months and on sciatic nerves from at least 3 MPZ‐Cre//l‐pgds ^flx/flx and littermates' control (l‐pgds ^flx/flx ) mice at 8 and 14 months and from at least 4 ChAT‐Cre//l‐pgds ^flx/flx and littermates' control (l‐pgds ^flx/flx ) mice at 8 months. In vitro analyses were performed on at least three different biological samples of Schwann cell‐DRG neuronal cocultures.

4.10.2. Phospholipids

Phospholipids were identified and evaluated by LC–MS/MS on an API–4000 triple quadrupole mass spectrometer (AB Sciex) coupled to an HPLC system (Agilent) and CTC PAL HTS autosampler (PAL System). Methanolic extracts were analyzed by using an XTerra Reverse Phase C18 column (3.5 μm 4.6 × 100 mm, Waters) and Methanol (MetOH) with 0.1% formic acid as isocratic mobile phase for positive ion mode in 5 min total run for each sample. Negative ion mode analysis was conducted with a cyano–phase LUNA column (50 mm × 4.6 mm, 5 μm; Phenomenex) and 5 mM ammonium acetate pH 7 in MetOH as isocratic mobile phase in 5 min total run for each sample. The identity of the different phospholipid families was confirmed using pure standards, namely one for each family, and lipidomics was quantified by external standard method. Collectively, we analyzed more than 200 phospholipids belonging to the following classes: phosphatidic acids (PA), lysophosphatidic acid (LPA), phosphatidylcholines (PC), lysophosphatidylcholines (LPC), phosphatidylethanolamines (PE), lysophosphatidylethanolamines (LPE), phosphatidylserines (PS), phosphatidylinositols (PI), lysophosphatidylinositols (LPI), phosphatidylglicerols (PG), sulfatides (Sul), ceramides (Cer), Galactosyl ceramide (Gal Cer), and sphingomyelins (SM).

Free and total cholesterol levels were determined in sciatic nerves methanolic extracts by cholesterol quantitation kit (MAK043‐1KT‐Sigma–Aldrich) following the manufacturer's instructions.

4.10.3. Metabolites

Quantification of energy metabolites (glycolysis, pentose phosphate pathway and Krebs cycle intermediates) was performed using a cyano–phase LUNA column (50 mm × 4.6 mm, 5 μm; Phenomenex) by a 5 min run in negative ion mode. We used water as stationary phase (A) and 2 mM ammonium acetate in MetOH as mobile phase (B). A 10% A and 90% B gradient and flow rate of 500 μL/min were applied for all analyses. Carnitine quantification was performed on acetonitrile/methanol extracts by using a Varian Pursuit XRs Ultra 2.8 Diphenyl column. Samples were analyzed by a 3 min run in positive ion mode using 0.1% formic acid in MetOH as mobile phase. All metabolites analyzed in the described protocols were previously validated by pure standards and quantified by standard curves and different internal standards (Audano et al. 2021; Cermenati et al. 2015).

4.11. Metabolites Flux Analyses

MultiQuant software (version 3.0.3, AB Sciex, Farmington, MA, US) was used for data analysis and peak review of chromatograms. Raw areas were normalized by the median of all metabolite areas in the same sample. Specifically, we defined the relative metabolite abundance ($m_a^N$) as:

$$m_a^N=\frac{x_n}{M_{a=1^a}^n}$$

where x _n represents the peak area of metabolite n for samples a, b, …, z, and $M_{a=1^a}^n$ represents the median of peak areas of metabolite n for samples a, b, …, z. Obtained data were then transformed by generalized log‐transformation and Pareto scaled to correct for heteroscedasticity, reduce the skewness of the data, and reduce mask effects (Ghaffari et al. 2019). In detail, obtained values were transformed by log10 and obtained values underwent Pareto scaling as follows:

$${\overline{x}}_{\mathit{ij}}=\frac{x_{\mathit{ij}}-{\overline{x}}_i}{\sqrt{s_i}}$$

where x _ij is the transformed value in the data matrix (i (metabolites), j (samples)) and s _i is the standard deviation of transformed metabolite values (van den Berg et al. 2006). Obtained values were considered as relative metabolite levels. Data processing and analysis were performed by MetaboAnalyst 5.0 web tool (Chong et al. 2019).

4.11.1. Acetate and Ketone Bodies Quantitation

Acetate's content was determined by Ketone Body Assay Kit (MAK134 Sigma–Aldrich). Schwann cell–DRG neuronal cocultures and nerves were homogenized using the Precellys System and tested for Acetoacetate content following manufacturer's instructions. β‐Hydroxybutyrate (BOH) content was assessed in coculture media supernatant and in sciatic nerves, following manufacturer's instructions. Results were normalized over the total protein content as determined by Pierce BCA assay.

4.12. Statistical Analyses

Data were collected randomly and assessed blindly on samples of comparable size, especially for all in vitro analyses. The data distribution was assumed to be normal, although we did not formally test it. All statistical analyses were performed on at least three different experiments. Statistical detailed analyses are reported in each figure legend and all assays were performed using the Prism 9 Software package (GraphPad).

A method checklist is available with the Supporting Information.

Author Contributions

A.T. designed the experimental plan, conducted most of the experiments, and wrote the manuscript. R.L.M., M.C., M.P., P.C., and M.F. contributed to in vitro and in vivo studies. M.A., S.P., G.I., D.C., and N.M. performed metabolomics and lipidomic studies, contributed to the experiment design, and helped in writing the manuscript. A.C. and L.M. performed bioinformatic studies. P.P., G.D., and A.Q. performed morphological analyses. C.T. designed the experimental plan, supervised the project, and wrote the manuscript. All authors commented on the manuscript.

Funding

This work was supported by Ministero dell'Università e della Ricerca, Progetto Eccellenza (2018–2022) to M.A., D.C. and N.M. and National Institute of Neurological Disorders and Stroke to C.T. (R01NS099102).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Lack of L‐PGDS does not affect in vitro neuronal and Schwann cell survival.

(a) Representative immunofluorescence of purified mouse Schwann cells prepared from P2 sciatic nerves of wild type littermate controls (WT) and l‐pgds ^−/− (KO) stained for Sox10⁺ coupled to Tunel assay. After 10 days in cultures, cells were fixed and stained for Sox10 (rhodamine) and processed for TUNEL assay. The graph shows the percentage of Tunel⁺/Sox10⁺ cells. No significant differences were observed among wild type and l‐pgds ^−/− Schwann cells. N = 4 different coverslip/condition experiments. Bar: 50 μm.

(b) Representative immunofluorescence coupled to Tunel assay of mouse DRG – rat Schwann cells myelinated cocultures. After 14 days in myelinating conditions, cultures were treated for additional 7 days with 25 μM AT‐56 or with DMSO as control vehicle. At the end of the treatment cultures were fixed and stained for MBP (white) and processed for TUNEL assay. Control cocultures were treated with 10 u/ml DNase1 as positive control for Tunel assay. AT‐56 treatments in already myelinated cultures induces myelin degeneration, but not cell death. N = 3 different independent coculture experiments. Bar: 100 μm.

(c) Representative immunofluorescence coupled to Tunel assay of mouse DRG – rat Schwann cells myelinated cocultures treated with 300 μM Arachidonic acid (complexed to BSA) for additional 7 days after 14 days in myelinating conditions. At the end of the treatment cultures were fixed and stained for MBP (white) and processed for TUNEL assay. Control cocultures were treated with 10 u/ml DNase1 as positive control for Tunel assay. Myelin degeneration was observed upon 300 μM Arachidonic acid treatment with no signs of Schwann cell death. The same amount of BSA was added to control cocultures. N = 3 different independent coculture experiments. Bar: 100 μm.

Figure S2: Morphological analyses of ChAT‐Cre//l‐pgds ^flx/flx nerves.

(a) Genotyping PCR for l‐pgds and for ChAT‐Cre alleles on genomic DNA prepared from P7 ventral (V) and dorsal (D) spinal cord regions of ChAT‐Cre//l‐pgds ^flx/flx mutants and littermate controls. The 444 bp l‐pgds null allele (*) is present only in the ventral spinal cord region of l‐pgds ^flx/flx mice expressing the Cre recombinase. The l‐pgds ^flx/flx allele (396 bp, **) is present in all spinal cord samples. To validate the results, we included control samples of genomic DNA prepared from tails of l‐pgds ^flx/flx (f/f), wild type (+/+) and heterozygous l‐pgds (+/−) mice. The wild type allele (290 bp, ***) is present only in wild type (+/+) and heterozygous l‐pgds (+/−) genomic DNA.

(b) Graph showing the metabolic profile in 8 months old wild type littermate controls (WT) and ChAT‐Cre//l‐pgds ^flx/flx sciatic nerves. Metabolites were normalized over the total protein content. N = 4 different mice/genotype. Error bars represent mean ± s.e.m. (Unpaired t‐test. PEaa 40:1 *p = 0.0256, t = 2.951, df = 6).

(c) Electron microscopy analyses of ChAT‐Cre//l‐pgds ^flx/flx and littermate controls sciatic nerves at P7. Scale bar: 2 μm.

(d) Graph showing g ratios as a function of axon diameter in P7 ChAT‐Cre//l‐pgds ^flx/flx nerves (red line) as compared to littermate controls (black line). (g ratio WT: 0.702 ± 0.0036; ChAT‐Cre//l‐pgds ^flx/flx : 0.723 ± 0.0036. Unpaired t test analyses; *p = 0.0107, t = 3.968, df = 5). N = 3 control mice; N = 4 ChAT‐Cre//l‐pgds ^flx/flx mice.

(e) Electron microscopy analyses of ChAT‐Cre//l‐pgds ^flx/flx and littermate controls sciatic nerves at P30. Scale bar: 2 μm.

(f) Graph showing g ratios as a function of axon diameter in P30 ChAT‐Cre//l‐pgds ^flx/flx nerves (red line) as compared to littermate controls (black line). (g ratio WT: 0.671 ± 0.0068; ChAT‐Cre//l‐pgds ^flx/flx : 0.682 ± 0.0074. Unpaired t test analyses; p = 0.324 n.s., t = 1.124, df = 4). N = 3 mice/genotype.

Figure S3: Developmental myelination is normal in MPZ‐Cre//l‐pgds ^flx/flx nerves.

(a) Electron microscopy analyses of MPZ‐CRE//l‐pgds ^flx/flx and littermate controls sciatic nerves at P7. Scale bar: 5 μm.

(b) Graph showing g ratios as a function of axon diameter in P7 MPZ‐CRE//l‐pgds ^flx/flx nerves (red line) as compared to littermate controls (black line). (g ratio WT: 0.6986 ± 0.0139; MPZ‐CRE//l‐pgds ^flx/flx : 0.7054 ± 0.0123. Unpaired t test analyses; p = 0.7343 n.s., t = 0.364, df = 4). N = 3 mice/genotype.

(c) Distribution of myelinated fibers is similar in 7 days MPZ‐CRE//l‐pgds ^flx/flx nerves and littermate controls. (Fishers exact test; p = 0.7379 (total versus 1–1.5 μm), p = 0.4094 (total versus 1.5–2 μm), p = 0.164 (total versus 2–2.5 μm), p = 0.1122 (total versus 2.5–3 μm), p = 0.4471 (total versus 3–3.5 μm)). Over 40 fibers for each genotype were counted. N = 3 mice/genotype.

(d) Electron microscopy analyses of MPZ‐CRE//l‐pgds ^flx/flx and littermate controls sciatic nerves at P30. Scale bar: 5 μm.

(e) Graph showing g ratios as a function of axon diameter in P30 MPZ‐CRE//l‐pgds ^flx/flx nerves (red line) as compared to littermate controls (black line). (g ratio WT: 0.6524 ± 0.0093; MPZ‐CRE//l‐pgds ^flx/flx : 0.659 ± 0.0024. Unpaired t test analyses; p = 0.5306 n.s., t = 0.6857, df = 4). N = 3 mice/genotype.

(f) Distribution of myelinated fibers is similar in 30 days MPZ‐CRE//l‐pgds ^flx/flx nerves and littermate controls. (Fishers exact test; p = 0.3402 (total versus 1–1.5 μm), p = 0.8216 (total versus 1.5–2 μm), p = 0.4637 (total versus 2–2.5 μm), p = 0.8875 (total versus 2.5–3 μm), p = 0.5962 (total versus 3–3.5 μm), p = 0.2178 (total versus > 3.5 μm)). Over 50 fibers for each genotype were counted. N = 3 mice/genotype.

(g) Semi–thin sections of 10 months wild–type littermate controls (WT) and MPZ‐CRE//l‐pgds ^flx/flx sciatic nerves (KO). Bar: 50 μm.

(h) Graph showing g ratios as a function of axon diameter in 10 months MPZ‐CRE//l‐pgds ^flx/flx nerves (red line) as compared to littermate controls (black line). (g ratio WT: 0.6631 ± 0.0058; MPZ‐CRE//l‐pgds ^flx/flx : 0.6685 ± 0.009. Unpaired t test analyses; p = 0.6499 n.s., t = 0.4898, df = 4). N = 3 mice/genotype.

(i) Distribution of myelinated fibers is similar in 10 months MPZ‐CRE//l‐pgds ^flx/flx nerves and littermate controls. (Fishers exact test; p = 0.9267 (total versus 1–2 μm), p = 0.7465 (total versus 2–3 μm), p = 0.1166 (total versus 3–4 μm), p = 0.246 (total versus 4–6 μm)). Over 250 fibers for each genotype were counted. N = 3 mice/genotype.

(j) qRT–PCR analyses on mRNA prepared from P7 wild type littermate controls (WT) and MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves tested for l‐pgds expression. l‐pgds is barely expressed in mutant nerves. Data have been normalized to gapdh expression level and analyzed with the CFX Manager Software from Biorad. N = pool of 2 nerves/mice, 3 different mice/genotype. Error bars represent mean ± s.e.m. (Unpaired t‐test. p = 0.0005, t = 10.35, df = 4).

(k) Graph, average of three different experiments, representing the percentage of myelin morphological aberrations in 10 months old wild type littermate controls (white bar) and MPZ‐Cre//l‐pgds ^flx/flx sciatic nerves (red bar). Alterations were determined as the number of fibers presenting myelin structural alterations over the total number of fibers in the entire reconstructed nerve cross section. Error bars represent mean ± s.e.m. N = 3 different mice/genotype. (Unpaired t‐test analysis, p = 0.0314; t = 3.248 df = 4).

Figure S4: Main dysregulated lipids and metabolites identified in l‐pgds ^−/− mutants.

(a) Scheme showing the omega‐6 fatty acid inter–conversions catalyzed by Δ6‐desaturase and elongase enzymes.

(b–d) Graph showing the differences in lysophosphatydilcholines contents in wild type littermate controls (WT) and l‐pgds ^−/− (KO) sciatic nerves at 4, 6 and 8 months. Values are relative to the analyses performed in Figure 3. Variation in lipids amount is expressed as fold change to WT arbitrarily set as 1.0. Error bars represent mean ± s.e.m. (Unpaired t‐test. LPC 18:2 6 months ***p = 0.0007, t = 4.867, df = 10; LPC 18:2 8 months **p = 0.0098, t = 3.178, df = 10; LPC 20:3 6 months **** p < 0.0001, t = 8.214, df = 10; LPC 20:3 8 months *p = 0.0197, t = 3.38, df = 5; LPC 20:4 6 months **** p < 0.0001, t = 8.631, df = 10; LPC 20:3 8 months *p = 0.0314, t = 2.962, df = 5).

(e) Graph showing the differences in acetyl–CoA contents in wild type littermate controls (WT) and l‐pgds ^−/− (KO) sciatic nerves at 4, 6 and 8 months. Values are relative to the analyses performed in Figure 4. Error bars represent mean ± s.e.m. (Unpaired t‐test. 8 months *p = 0.0367, t = 2.828, df = 5).

(f) Graph showing the differences in citrate contents in wild type littermate controls (WT) and l‐pgds ^−/− (KO) sciatic nerves at 4, 6 and 8 months. Values are relative to the analyses performed in Figure 4. Error bars represent mean ± s.e.m. (Unpaired t‐test. 6 months **p = 0.0047, t = 3.614, df = 10; 8 months *p = 0.038, t = 2.799, df = 5).

(g) Graph showing the differences in α–ketoglutarate contents in wild type littermate controls (WT) and l‐pgds ^−/− (KO) sciatic nerves at 4, 6 and 8 months. Values are relative to the analyses performed in Figure 4. Error bars represent mean ± s.e.m. (Unpaired t‐test. 8 months *p = 0.0473, t = 2.617, df = 5).

(h) Graph showing the differences in lactate contents in wild type littermate controls (WT) and l‐pgds ^−/− (KO) sciatic nerves at 4, 6 and 8 months. Values are relative to the analyses performed in Figure 4. Error bars represent mean ± s.e.m. (Unpaired t‐test. 6 months **p = 0.0066, t = 3.418, df = 10).

Figure S5: Ketone bodies synthesis is critical to maintain the integrity of the axo‐glial unit.

(a) qRT–PCR analyses on mRNA prepared from myelinated wild type Schwann cell – neuronal cocultures infected with shRNA lentiviruses against Hmgcs2 (sh062, sh362, sh372) or scramble sh RNA (shscr). Three days post infection, cocultures were supplemented with ascorbic acid to allow myelination for 10 days, then treated with 25 μM AT‐56 or vehicle only, as control, for additional 7 days. Hmgcs2 expression is significantly reduced specifically in AT‐56 treated cocultures infected with Hmgcs2 targeting lentiviruses. Data have been normalized to 36B4 expression level and analyzed with the StepOne Software v2.3 (Applied Biosystems). N = 3 different mRNA preparations and analyses. Error bars represent mean ± s.e.m. (Unpaired t test; shscr AT56–sh062AT‐56 ***p = 0.0001 (t = 14.33; df = 4); shscr AT56–sh362AT‐56 ***p = 0.0002 (t = 13.81; df = 4); shscr AT56–sh372AT‐56 ****p < 0.0001 F (t = 18.56; df = 4).

(b) Representative immunofluorescence of wild–type myelinated cocultures Schwann cells neuronal cocultures infected with Hmgcs2 sh362 or Hmgcs2 sh062. Three days post infection, cocultures were supplemented with ascorbic acid to allow myelination for 10 days, then treated with 25 μM AT‐56 or vehicle only, as control, for additional 7 days. At the end of the experiments cultures were fixed and stained for GFP to visualize infected cells (fluorescein), non‐phosphorylated neurofilament SMI–32 (rhodamine) and MBP (white). Ablation of Hmgcs2 coupled to L‐pgds block, causes axonal instability. N = 3 independent biologic replicates. Bar: 100 μm.

Trimarco, A. , Audano M., La Marca R., et al. 2026. “Prostaglandin D2 Synthase Controls Schwann Cells Metabolism and Peripheral Myelin Homeostasis.” Glia 74, no. 3: e70137. 10.1002/glia.70137.

Data Availability Statement

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