Long-term use of benzodiazepines, commonly prescribed for anxiety and insomnia, may impair cognition and increase the risk of developing dementia. How? By activating microglia to prune synapses, according to researchers led by Jochen Herms and Mario Dorostkar at Ludwig-Maximilians University, Munich. In the February 28 Nature Neuroscience, they reported that diazepam, aka Valium, upregulated the mitochondrial receptor TSPO in microglia, triggering overzealous trimming of dendritic spines. Mice fed high doses of diazepam lost spines more quickly and had poorer memories than animals given low doses. Overall, this could explain the link between benzodiazepines and dementia.

  • Diazepam binds the mitochondrial translocator protein in microglia.
  • The glia ramped up synaptic pruning.
  • Given diazepam, mice did worse on learning tasks.

“This thorough study adds to a growing body of evidence that microglia are critical enactors of synaptic form and function,” Kim Green, University of California, Irvine, wrote to Alzforum.

Though largely replaced by selective serotonin reuptake inhibitors, benzodiazepines are still commonly prescribed for anxiety and insomnia, despite evidence tying them to cognitive impairment and higher risk of developing dementia (Sep 2014 news; Ferreira et al., 2021; Penninkilampi et al. 2018). 

To find out why these drugs weaken cognition, first author Yuan Shi fed 4-month-old mice 5 mg/kg diazepam once per day for one week or 1 mg/kg daily for eight weeks to mimic sedative and anxiolytic effects, respectively. The animals expressed green fluorescent protein in hippocampal and cortical neurons, allowing the researchers to detect synapse changes. Shi captured two-photon microscopy images from live mice through a cranial window, or confocal microscopy images from brain slices taken postmortem. Compared to controls, mice given the high dose had lost about 13 percent of dendritic spines in the somatosensory cortex one week later, which grew back over the next eight weeks. The animals that received the lower dose lost a similar fraction of their spines more slowly over the eight weeks of treatment and grew them back within three weeks.

This synaptic disruption manifested as a learning impairment. One day after their last sedative dose, mice showed no more interest in new objects or places than in familiar ones. Likewise, mice given the lower dose had trouble remembering familiar objects, though they did remember which arm of a Y-maze they had traversed.

How did diazepam cause synaptic pruning? Benzodiazepines work by binding GABA-A receptors; however, high-dose diazepam shriveled spines and impaired memory even when its binding to the GABA-A receptor was blocked by the competitive inhibitor flumazenil, or when diazepam was given to mice with mutated GABA-A receptors that don’t bind the drug.

The mitochondrial translocator protein, TSPO, also binds benzodiazepines. In fact, TSPO was identified first as a benzodiazepine receptor outside of the CNS. The protein also happens to be upregulated in the brains of people with Alzheimer’s (Kreisl et al., 2013). Could TSPO be involved in the synaptic pruning?

In TSPO knockout mice, diazepam did not alter synapses or cognition, hinting that the drug may work through TSPO. In support of this, synapses also withered in wild-type mice given the TSPO ligand XBD173. Indeed, diazepam-treated mice bound more 18FGE-180, a TSPO PET ligand, than did control animals, suggesting the drug upregulates the mitochondrial protein. All told, the researchers believe that benzodiazepines impair cognition through TSPO.

How could TSPO be involved in thinning synapses? The scientists suspected microglia were to blame because they upregulate TSPO upon activation and are known to engulf dendritic spines (Dec 2014 conference news; Apr 2016 news). Indeed, TSPO strongly co-localized with microglia in brain slices from diazepam-treated mice, and microglia isolated from these animals contained more TSPO than did microglia from controls. Though diazepam did not change the number of microglia in the mice, the cells were bigger and had longer processes with more branches (see image below). They accumulated 50 percent more puncta containing the synaptic marker PSD95, indicating they had swallowed chunks of synapses, and they contained more C1q, the complement protein that serves as a synaptic “eat me” signal for microglia.

Diazepam Distorts Microglia. Compared to microglia from wild-type mice (top left), those from mice given diazepam had longer, bigger, and more branched processes (top right). Microglia from TSPO knockout mice (bottom) remain unchanged with (right) or without (left) diazepam. [Courtesy of Shi et al., Nature Neuroscience, 2022.]

Further, after one week of diazepam treatment, the researchers saw GFP-labeled microglia flocking to dendritic spines. Supporting the idea that microglia mediate the cognitive effects, mice depleted of microglia remembered familiar objects and places just as well as control animals even after diazepam treatment.

What does all this mean for AD risk? “Alzheimer's symptoms appear after many synapses are gone, so chronic diazepam use may increase disease risk simply by reducing the number of synapses,” Herms told Alzforum. He is currently testing the effects of diazepam and TSPO ligands on dendritic spines in AD transgenic mouse models.

Makoto Higuchi, National Institute of Radiological Sciences, Chiba, Japan, noted contrasting results from the TSPO-specific diazepam derivative Ro5-4864. In mouse models of tauopathy, it was neuroprotective by lowering TSPO expression, mitigating gliosis, and alleviating cognitive deficits. “It is conceivable that diazepam, its analogs, and other TSPO ligands exert distinct effects on homeostatic and disease-associated microglial species,” he wrote (full comment below).—Chelsea Weidman Burke


  1. This excellent work has revealed that long-term treatment of wild-type mice with diazepam deteriorates cognitive function through its action on 18-kDa translocator protein (TSPO) in microglia. Notably, diazepam-stimulated microglia actively engulf dendritic spines in different brain regions, leading to the loss of post-synaptic structures in excitatory neurons. It should also be noted that these intriguing observations appear to be contrary to previous reports documenting neuroprotective effects of Ro5-4864, which is a derivative of diazepam that binds to TSPO without interacting with GABA-A receptors, in animal models of CNS injuries and Alzheimer’s and related diseases. Indeed, Ro5-4864 was shown to reduce soluble Aβ levels and mitigate gliosis and cognitive deficits in 3xTg mice (Barron et al., 2013). 

    We have also demonstrated that the long-term administration of Ro5-4864 to rTg4510 tau transgenic mice resulted in the suppression of hippocampal neuronal loss and brain atrophy, along with diminished neuroinflammation (Fairley et al., 2021). Our intravital multiphoton microscopic investigation has illustrated that live neurons burdened with tau aggregates are phagocytosed by activated microglia in rTg4510 tau transgenic mice, and that this phagocytic loss of neurons can be inhibited by treatment with Ro5-4864 (Takuwa et al., 2020). 

    Hence, it is conceivable that diazepam, its analogs, and other TSPO ligands exert distinct effects on homeostatic and disease-associated microglial species. Indeed, the present article showed that after treatment of mice with PLX5622, residual microglia exhibited an elevated TSPO level, were unresponsive to diazepam, and associated with slight increases of dendritic spines, in clear contrast to "aggressive" microglia emerging after the diazepam treatment in PLX-untreated mice.

    Expression of microglial TSPO is notably increased in the brains of tau transgenic mice and Alzheimer’s disease cases (Maeda et al., 2011; Ji et al., 2008), and these microglial cells enriched with TSPO may respond to diazepam by protecting neurons against toxic insults.

    Another issue is whether the modulation of microglial activity by diazepam can be monitored in the brains of living individuals using positron emission tomography (PET). TSPO-PET with a specific imaging agent, 18FGE-180, was conducted in the current study, indicating that the diazepam treatment enhances the abundance of TSPO in the brains of wild-type mice. This finding is also of great interest, though it needs to be clear whether therapeutic diazepam competes with 18FGE-180 on TSPO or the PET scan was performed after washing out diazepam from the brain. Although it is known that Ro5-4864 and other ligands bind to overlapping but not identical portions of a complex formed by TSPO and associated components (Kassiou et al., 2005), competition among Ro5-4864, GE-180, and a classical TSPO ligand, PK11195, was reported previously (Scarf and Kassiou, 2011; Cumming et al., 2018). A reduced level of TSPO as a consequence of Ro5-4864 treatment in rTg4510 mice was demonstrated by carrying out PET scans after a one-week washout period in our experiment (Fairley et al., 2021), which is again opposite to the changes noted in wild-type mice given diazepam.

    A longitudinal TSPO-PET assay will also be possible in clinical studies of healthy subjects and patients with neurodegenerative dementias to pursue the density of TSPO in the brain along the course of the treatment with diazepam and more selective TSPO ligands, such as XBD-173, following an adequate withdrawal of the drug. It will be informative to assess the differential responses of microglia to such an agent in intact and diseased brains.

    Finally, it should be considered that in the brain TSPO is expressed in non-microglial cells, including neurons and vascular endothelial cells. The neuronal TSPO might be of functional significance since behavioral alterations presented by TSPO-deficient mice could be partially reversed by neuronal expression of TSPO (Barron et al., 2021). The contribution of microglial versus neuronal TSPO to synaptic integrity will also be a fundamental topic for an in-depth understanding of this molecule in the maintenance and disruption of brain functions.


    . Ligand for translocator protein reverses pathology in a mouse model of Alzheimer's disease. J Neurosci. 2013 May 15;33(20):8891-7. PubMed.

    . Neuroprotective effect of mitochondrial translocator protein ligand in a mouse model of tauopathy. J Neuroinflammation. 2021 Mar 19;18(1):76. PubMed.

    . Tracking tau fibrillogenesis and consequent primary phagocytosis of neurons mediated by microglia in a living tauopathy model. bioRxiv, November 5, 2020 BioRxiv.

    . In vivo positron emission tomographic imaging of glial responses to amyloid-beta and tau pathologies in mouse models of Alzheimer's disease and related disorders. J Neurosci. 2011 Mar 23;31(12):4720-30. PubMed.

    . Imaging of peripheral benzodiazepine receptor expression as biomarkers of detrimental versus beneficial glial responses in mouse models of Alzheimer's and other CNS pathologies. J Neurosci. 2008 Nov 19;28(47):12255-67. PubMed.

    . Ligands for peripheral benzodiazepine binding sites in glial cells. Brain Res Brain Res Rev. 2005 Apr;48(2):207-10. Epub 2005 Jan 22 PubMed.

    . The translocator protein. J Nucl Med. 2011 May;52(5):677-80. Epub 2011 Apr 15 PubMed.

    . Sifting through the surfeit of neuroinflammation tracers. J Cereb Blood Flow Metab. 2018 Feb;38(2):204-224. Epub 2017 Dec 19 PubMed.

    . Regulation of Anxiety and Depression by Mitochondrial Translocator Protein-Mediated Steroidogenesis: the Role of Neurons. Mol Neurobiol. 2021 Feb;58(2):550-563. Epub 2020 Sep 29 PubMed.

  2. I really enjoyed the very systematic interrogation into the mechanisms underlying the increased risk of cognitive decline following benzodiazepine use. The methods used by the authors were innovative and very revealing, especially using 2-photon imaging in diazepam- and vehicle-treated Thy1-eGFP mice.

    First, the authors found that a short treatment with diazepam reduced dendritic formation and increased spine elimination. Although the effect lasted beyond treatment, spines were eventually recovered. Diazepam treatment also cause cognitive decline in the novel-object recognition and spontaneous alternation Y maze tests. Lower doses (anxiolytic) had similar but more muted effects. Changes in spine morphology were also evident after diazepam treatment.

    Elegant experiments were undertaken to prove that the effect of diazepam on dendritic spines was not mediated by GABA-A receptors. Instead, the authors found that TSPO mediated the diazepam effects on dendritic spines and cognition. TSPO KO mice crossed to Thy1-eGFP mice did not respond to diazepam like WT mice did: They did not lose dendritic spines or show cognitive impairment after treatment. And, the authors found that diazepam treatment for two weeks increased 18F-GE-180 TSPO PET tracer uptake in WT mice and increased TSPO expression in microglia. They ruled out a neurosteroid-related mechanism for this but interestingly, they found that diazepam altered the morphology of microglia, and increased their contact with dendritic spines, where C1q deposition was increased. Depletion of microglia with PLX5622 rescued diazepam effects in mice.

    Putting all this data together, the authors conclude that diazepam binds TSPO on microglia, which leads to increased C1q deposition on synapses, which in turn induces microglial phagocytosis of dendritic spines, resulting in cognitive decline.

    What is unclear is why diazepam binding to TSPO on microglia causes an increase in C1q deposition on dendritic spines. Intriguingly, TSPO KO mice treated with diazepam also had more C1q deposition on dendritic spines (similar to diazepam-treated WT mice) but they did not have more microglial engulfment of dendritic spines. The authors propose that this may be due to a compensatory mechanism. However, it would be interesting to know what happens to the complement receptor 3 (CR3), which is expressed on microglia and can bind C1q, C3b, and iC3b to induce phagocytosis. Are C3 or CR3 altered in TSPO KO mice? If CR3 is downregulated in the absence of TSPO, perhaps there might be less phagocytosis even in the presence of more C1q deposition. Understanding more about how diazepam affects microglial signaling pathways in the brain and macrophages in the periphery may open new avenues for therapeutics. The authors are to be congratulated for this excellent piece of work.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Benzodiazepines Boost Risk of Alzheimer’s Disease
  2. Meet GE180: A PET Ligand for Tracking Neuroinflammation?
  3. Paper Alert: Microglia Mediate Synaptic Loss in Early Alzheimer’s Disease

Paper Citations

  1. . Is there a link between the use of benzodiazepines and related drugs and dementia? A systematic review of reviews. Eur Geriatr Med. 2021 Aug 17; PubMed.
  2. . A Systematic Review and Meta-Analysis of the Risk of Dementia Associated with Benzodiazepine Use, After Controlling for Protopathic Bias. CNS Drugs. 2018 Jun;32(6):485-497. PubMed.
  3. . In vivo radioligand binding to translocator protein correlates with severity of Alzheimer's disease. Brain. 2013 Jul;136(Pt 7):2228-38. PubMed.

Further Reading

Primary Papers

  1. . Long-term diazepam treatment enhances microglial spine engulfment and impairs cognitive performance via the mitochondrial 18 kDa translocator protein (TSPO). Nat Neurosci. 2022 Mar;25(3):317-329. Epub 2022 Feb 28 PubMed. Nat Neurosci.