A mitochondrial–neuroinflammation–D-serine connection in epilepsy

Epilepsy is a neurological disorder marked by recurrent seizures caused by abnormal, synchronized electrical activity in the brain. Epilepsy can be caused or exacerbated by mitochondrial dysfunction, and children with mitochondrial diseases often exhibit treatment-resistant epilepsy (1). Because of the high prevalence of epilepsy and its diverse underlying mechanisms, there is a clear need for new and more effective treatments. The paper by Jiang et al. (2) established a new model of epilepsy that encompasses increased neuroinflammation and overproduction of the neuromodulator D-serine.

Jiang et al. (2) investigated the in vivo role of Mic19, a protein component of the MICOS (mitochondrial contact site and cristae organizing system) complex, which is essential for the formation of mitochondrial cristae junctions (3). A brain-specific Mic19 knockout (BKO) mouse led to impairment of MICOS function and mitochondrial damage in neurons. Mic19 BKO mice displayed epileptic seizures and increased postnatal lethality, presumably because of severe seizure episodes. Interestingly, the Mic19 BKO mice showed higher expression of neuroinflammatory genes, particularly interferon-alpha response genes, accompanied by increased microglial proliferation and activation. The corresponding increase in cytosolic DNA activated the cGAS–STING, driving the neuroinflammatory response.

To understand the underlying cause of microglial activation, the authors turned to the neuronal mitochondria, where Mic19 is strongly expressed. Mic19 BKO disrupted the cristae junctions in neurons in vivo, while its effects on glial mitochondria were minimal. Neurons showed increased mitochondrial ROS production and oxidative mtDNA damage. This was associated with the formation of Z-DNA, a special DNA conformation that activates the innate immune response (4).

The study showed that mtDNA was released from neurons, leading to microglial activation. The mechanism of mitochondrial release involved phosphoMLKL overexpression in neurons. This protein apparently formed pores in the outer mitochondrial membrane, and its knockdown in Mic19 BKO MEFS efficiently reduced mtDNA release in culture.

Is the abnormal release of mtDNAs connected to neuroinflammation? mtDNA is known to activate cGAS, triggering STING signaling. Activation of STING pathway was clearly detected in Mic19 BKO. The authors then hypothesized that mtDNAs derived from Mic19 BKO neurons are taken up by microglia, where they induce cGAS–STING activation and cytokine production (Fig. 1).

Fig. 1.

mtDNA–neuroinflammation–D-serine axis. Damage to mitochondrial cristae contacts in neurons leads to the release of mtDNA and the activation of microglia via the cGAS–STING pathway, a sensor of cytosolic DNA. Activation of this pathway leads to the release of cytokines, which in turn activate astrocytes to overproduce L-serine and glycine. L-serine shuttles to neurons, where it fuels the synthesis of D-serine by the serine racemase enzyme. Alternatively, activated astrocytes or microglia themselves may release D-serine (dashed arrows). The combination of neuroinflammation and increased NMDAR coagonists (D-serine and glycine) may cooperate to induce seizures. Inhibition of the STING pathway by H-151 in microglia or other cells ameliorates neuroinflammation and abrogates seizures. Created in BioRender Wolosker (2026), https://BioRender.com/ibw5ujg.

The activation of microglia in Mic19 BKO was accompanied by strong astrogliosis (proliferation and activation of astrocytes), presumably by microglial cytokines (Fig. 1). How does this affect brain metabolism or the function of astrocytes? ¹⁸F-fluorodeoxyglucose positron emission tomography (¹⁸F-FDG-PET) uncovered a general increase in ¹⁸F-FDG uptake. Although not specific to astrocytes, it suggested increased glycolysis. Astrocytes are involved in the production of L-serine from glucose, starting from the sequential activity of the enzymes 3-phosphoglycerate dehydrogenase, phosphoserine aminotransferase 1, and phosphoserine phosphatase (5). These enzymes were robustly upregulated in activated astrocytes, leading to a general increase in brain L-serine and glycine levels. In addition to being required for one-carbon metabolism and lipid synthesis, L-serine is the sole precursor of the transmitter/neuromodulator D-serine (6).

Jiang et al. (2) investigated the in vivo role of Mic19, a protein component of the MICOS (mitochondrial contact site and cristae organizing system) complex, which is essential for the formation of mitochondrial cristae junctions (3).

D-Serine is synthesized from L-serine by the enzyme serine racemase (SR), usually found in neurons under physiological conditions (7). Jiang et al. (1) demonstrated that the increase in brain D-serine is due to a combination of elevated L-serine, higher neuronal SR expression, and its stabilization by DISC-1. Therefore, the activation of a serine shuttle between astrocytes and neurons and neuronal SR levels ultimately caused D-serine overproduction (Fig. 1).

D-Serine is a physiologic coagonist of N-methyl D-aspartate-type of glutamate receptors (NMDARs) and is involved in both physiological and neurotoxic activation of these receptors (7). Mutations in NMDARs subunits that enhance agonist potency drive neuronal hyperexcitability, which contributes to epileptic phenotype (8). The receptor/channel requires simultaneous binding of both glutamate and D-serine to function, so it is conceivable that excess D-serine facilitates seizures by acting on NMDARs in the Mic19 BKO model. Although the levels of D-serine will possibly facilitate seizures by NMDAR stimulation, additional studies to demonstrate that increased extracellular D-serine leads to NMDAR-mediated epileptic activity are warranted. Furthermore, like D-serine, glycine is a bona fide coagonist of NMDARs and may also facilitate seizures (8), as increased glycine levels were observed in Mic19 BKO mice.

Interestingly, activated glia have been found to produce and release D-serine, as seen after traumatic brain injury (9). This would suggest that activated glia may also participate in the overall hyperexcitability of NMDARs, complicating the circular model to suggest that the entire tripartite neuron–glia complex may participate in the epileptic phenotype. In particular, both activated microglia and astrocytes release excessive amounts of D-serine after traumatic brain injury, and ameliorating this release led to improved recovery (10). Specific D-amino acid transporters, namely Slc7a10 for neurons and primarily Slc1a4 for glia, have been shown to regulate D-serine release (10), permitting future studies to examine the unique cellular contributions of D-serine in the Mic19 BKO model.

Another compelling question is whether there is a direct link between the neuroinflammation phenotype and seizures. To address this question, Jiang et al. (1) treated mice with H-151, a STING inhibitor, which strongly reduced seizure activity, ameliorated neuroinflammation, and lowered serine levels. Therefore, reducing neuroinflammation was sufficient to mitigate the pathology. It is also plausible that H-151 acts directly on astrocytes and neurons, since the cGAS–STING pathway is present in both cell types, albeit to a lesser extent than in microglia. This multicellular potential for H-151 actions may also regulate overall neuroinflammation, since astrocytes and neurons have also been shown to express several important cytokines, such as TNFα. Although these complex multicellular actions and responses may complicate the model, the actions of H-151 to reduce overall seizure activity are well supported in this study.

This study also raises several additional questions. mtDNA-elicited gliosis appears to stimulate L-serine synthesis and increase D-serine levels. How mtDNA stimulation differs from other stimuli with respect to its ability to stimulate microglia is unknown. Moreover, while the correlation between mtDNA release and microglial activation is compelling, other factors released from mitochondria may also contribute to microglial activation. Additional studies with selective inhibition of mtDNA release would clarify the role of released mtDNA in microglia activation.

This study does not rule out that astrocytes themselves are activated by neuron-derived mtDNA. Nestin-cre used to generate Mic19-BKO is also expressed by radial glia, so it may also affect astrocytes. Jiang et al. (2) did not see much mitochondrial damage in astrocytes, but astrocytic sensitivity to mtDNA was not assessed.

In summary, this study uncovered a novel and important mechanism connecting mtDNA release, neuroinflammation, serine metabolism, and seizures. It is conceivable that neuroinflammation also exacerbates epilepsy in mitochondrial diseases, some of which are characterized by intractable seizure episodes. Importantly, the study raises the possibility that treating neuroinflammation may be a potential therapeutic strategy for epilepsy.