Bouts of white-matter destruction in the brain and spinal cord are a hallmark of multiple sclerosis, but the inflammatory disease also ravages gray matter. Using a mouse model of cortical MS, researchers led by Doron Merkler at the University of Geneva, Thomas Misgeld of the Technical University of Munich, and Martin Kerschensteiner of Ludwig-Maximilians University, Munich, report that cortical inflammation triggers the loss of dendritic spines, suppressing neuronal activity across the cortex. The study, published January 25 in Nature Neuroscience, found that calcium overload in a small fraction of spines marked them for destruction by resident microglia and infiltrating monocytes. Crucially, these spines and synapses rebounded once inflammation subsided. A drug that dampens microglial activation kept their appetite for the spines at bay. The findings add to mounting evidence implicating neuroinflammation in the destruction of synapses, and suggest that quelling it could even reverse damage already done.

  • A mouse model of cortical MS invokes inflammation throughout the brain.
  • Dendritic spines shrivel; neuronal activity wanes.
  • Calcium overload marks spines for removal by myeloid cells.

“These findings represent substantial promise for the clinical treatment of MS, as they suggest that early intervention with immunomodulatory therapies could work in combination with endogenous healing processes within the brain to prevent progressive MS-associated neurodegeneration,” wrote Marc Aurel Busche and Robert Ellingford of University College London. The findings may also have implications for other neurodegenerative diseases, they added.

Overzealous munching of synapses by ill-tempered microglia is a common scourge across neurodegenerative diseases, including Alzheimer’s, Huntington’s, and frontotemporal dementia (Nov 2015 conference news; Apr 2016 news; May 2016 news). In MS, an autoimmune attack on the fatty myelin sheath that surrounds axons is the predominant feature of the earlier, relapsing/remitting phase. However, as the disease progresses, it resists treatment, and gray matter in the cortex becomes increasingly damaged as well (Mahad et al., 2015; Eshaghi et al., 2018; Scalfari et al., 2018). Studies have found loss of synapses and dendritic spines in postmortem cortical samples of people who died with progressive MS, but what drives this degeneration is not understood (Jürgens et al., 2016; Albert et al., 2017). 

To investigate, co-first authors Mehrnoosh Jafari, Adrian-Minh Schumacher, and Nicolas Snaidero used a mouse model of cortical MS. The researchers injected interferon-γ and tumor necrosis factor-α (TNF-α)—proinflammatory cytokines detected in the meninges of patients with MS—into the somatosensory cortices of mice that previously had been immunized with myelin–oligodendrocyte glycoprotein (MOG). These animals mobilize an autoimmune response that attacks their own myelin, much as occurs in MS. Within three days of injecting the cytokines, widespread lesions emerged across the cortex. Akin to those found in people with progressive MS, the lesions were marked by extensive myelin loss and were inundated by phagocytes and, to a lesser extent, T cells.

The researchers next zoomed in on dendritic spines—the tiny protrusions that serve as a platform for synapses. Three days after the cytokine shock, the number of spines plummeted by a third in cortical projection neurons. Oddly, surviving and dying spines often sat next to each other on the same dendrite. In addition, the researchers detected widespread loss of excitatory, but not inhibitory, synapses across the cortex. This loss of excitatory input resulted in neuronal silencing. Strikingly, within two weeks after the cytokine injection, all of these deficits were gone: spines and synapses grew back, and neuronal activity was restored.

What caused some spines to perish, while others weathered the cerebral cytokine storm? Disturbances in calcium homeostasis are known to herald spine loss, so the researchers measured calcium levels within spines using cranial windows and confocal microscopy (Jul 2008 news). Three days after tripping off the inflammatory cascade, the scientists detected a calcium overload in about 6 percent of dendritic spines. Tracking the fate of these spines over time, they found that the calcium-loaded spines were more likely to be removed. They calculated that this inflammation-induced calcium overload accounts for the loss of about 1 percent of cortical spines per hour, resulting in a 30 percent loss over the course of the acute inflammatory episode.

Going, Going, GONE. In control mice (top rows), dendritic spines were not overloaded with calcium. In cortical MS mice (bottom rows), some spines had high calcium (marked by arrows) at 0 hours (left column). Two hours later (right column), one calcium-overloaded spine (dashed arrow) was gone; the other had filled with more calcium (yellow arrow). [Courtesy of Jafari et al., Nature Neuroscience 2020.]

The work supports the decades-old hypothesis that calcium dysregulation is a driver of neurodegeneration, noted Christopher Norris of the University of Kentucky in Lexington. “The present work … is yet another example of just how vital the ‘Ca2+ hypothesis’ is after all these years, and further cements the visionary status of its early proponents.”

How were the calcium-loaded spines destroyed? Using transgenic mice in which peripheral monocytes and resident microglia can be distinguished based on different fluorescent markers, the researchers caught both types of cells red-handed, i.e., stuffed with synaptic material.

Finally, the researchers asked whether an inhibitor of colony-stimulating factor 1 receptor might stem the synaptic carnage. High doses of CSF-1R inhibitors wipe out microglia completely, but low doses stifle microglial function. Such inhibitors have been widely used in AD studies. A recent report found that blocking CSF-1R with a newly developed, highly selective, and CNS-penetrant small molecule inhibitor called sCSF1Rinh lessened axonal damage in a mouse model of progressive MS (Hagan et al., 2020). 

In their model of cortical MS, Jafari and colleagues found that low doses of inhibitor halved the number of infiltrating monocytes, and reduced activation markers on brain-resident microglia. Using RNA sequencing, they found that CSF-1R inhibition kept the microglia in a homeostatic state; without it, the cells adopted a neurodegenerative signature similar to one reported in previous studies (Sep 2017 news). The treatment also dampened microglial engulfment of synapses, and prevented spine loss.

In all, the findings suggest that while inflammation goads innate immune cells to gorge on synapses, the brain has mechanisms in place to reverse the damage once inflammation subsides. This reversibility supports the idea that drugs such as CSF-1R inhibitors stand a chance at rejuvenating lost synapses, Kerschensteiner said. He plans to test the idea in a model of lower-grade, chronic cortical inflammation, which would be more akin to disease in humans, he said.

So far, the mechanisms that drive calcium overload in a subset of spines, and how that triggers their removal, remain unknown, Kerschensteiner said. He noted that the complement system could be involved in tagging the spines and marking them for removal, and/or calcium overload could lead to destabilization of the spines, beckoning phagocytes to the scene. Oleg Butovsky of Brigham and Women’s Hospital in Boston raised these possibilities as well, and added that the role of T cells in the synaptic destruction remains to be explored. He pointed out that while cytokine injection alone induced infiltration of monocytes and T cells, synaptic loss and demyelination only occurred in mice previously immunized with myelin–oligodendrocyte glycoprotein, suggesting an adaptive T cell response was somehow involved.

“This is a very exciting study, but many questions remain that require further investigation,” wrote Rebecca Gillani of Massachusetts General Hospital in Boston. Among them, Gillani wondered about the timing of synapse loss with respect to disease progression in people with MS, noting that it is possible that synapses are affected even before the transition from relapsing-remitting to progressive MS. Pinning down the beginning of this synaptic loss would help researchers know when to start treating it, she noted. She, along with other commentators, also wondered whether synaptic loss would be reversible in the context of the sustained, inflammatory assault that takes place in MS.—Jessica Shugart


  1. This interesting work raises questions such as how does the immune system surveil the CNS? One answer comes from a new paper in Cell led by Jonathan Kipnis, which shows immune cells in the meninge. When they study an MS mouse model, they find activated immune cells in the dural sinuses which appear to be a key neuroimmune interface (Rustenhoven et al., 2021). 

  2. This study strings together a series of enlightening and novel observations regarding cortical synapse decline in a model of cortical multiple sclerosis (c-MS), and aligns excitatory synapse loss, increased calcium load in dendrites and spines, phagocyte activation, and local inflammatory responses to provide an exciting mechanism underlying cognitive impairment in MS patients.

    A broader theme woven through the manuscript is the potential relevance of this mechanism to other neurodegenerative diseases or psychiatric disorders, and the potential for novel therapeutic approaches. While I am intrigued by the proposed relationship among these elements and find the overall concept exciting, the study also raises several key questions that should be addressed prior to making substantial conclusions. 

    For example, while the study employs elegant approaches to demonstrate decreased somatic calcium transients and increased calcium load within spines and dendritic shafts in the c-MS mice, the source of the calcium dysregulation remains unexplored. Some insight is inferred as extracellular calcium buffering reduces the degree of synapse loss. Yet the discussion reflects on the possibility of intracellular stores playing a role as well. In AD models, excess calcium release through intracellular ER stores is strongly implicated in spine loss, and mushroom spine loss specifically. Likewise, the observation of spine loss in a calcium-associated manner in the c-MS model is notable and can provide a strong structural basis for subsequent cognitive defects in MS patients. A deeper analysis into the type of spines affected would have provided interesting validation for this hypothesis, and the data are there awaiting analysis. 

    All spines are not created functionally or structurally equal, with stubby and thin spines typically considered highly labile, transient, and plastic, and more stable mushroom spines associated with sustained memory formation. It would be of interest to know which spine type is preferentially affected in this model. An additional question begging to be explored is how phagocytes identify the high calcium spines and target them for removal. Indeed, the presupposition is that the calcium dysregulation is upstream of the tagging and phagocyte digestion, and while this does make logical sense, it is not a foregone conclusion. At this stage in the study, these are correlated events, but the order of operations and functional relationship between these observations remains to be determined. Since the calcium buffering experiment affects the entire circuitry and cellular environment, it’s difficult to apply these findings to determine causality.

    Overall, this study may lay important groundwork describing the complex relationship among calcium dysregulation within synaptic compartments, selective loss of spines, phagocyte recruitment, and cognitive impairment—but the links among these observations are ripe for further exploration and validation of this hypothesis. We look forward to finding out more about these mechanisms.

  3. This study tackles one of the biggest challenges in treating people living with multiple sclerosis (MS), which is our limited ability to slow disability accumulation in progressive MS. Despite an immense effort, treatment options for progressive MS are limited. The currently FDA-approved medications for progressive MS can slow disability progression, but only in a subset of patients, tending to benefit people with active progressive MS most. This means that there are many people with progressive MS with no good treatment options to slow their disability progression.

    These authors hypothesize that targeting the widespread synaptic loss that occurs in MS may lead to new treatments for progressive MS. Despite the significant role that synapse loss likely plays in the accumulation of disability in MS, we lack a basic understanding of the mechanisms that drive synapse loss. I think that this study makes a significant step to advance the field by identifying specific mechanisms the underlie synapse loss in lesional acutely inflamed cortex, namely calcium overload in dendritic spines and activation of the innate immune system with synapse engulfment.

    A specific link between calcium overload in dendritic spines and synapse engulfment was not identified, but will be an important area for future research. The innate immune system is being targeted in the development of new disease-modifying therapies for progressive multiple sclerosis, with encouraging clinical trial results from Masitinib reported at the 2020 ACTRIMS-ECTRIMS meeting.

    This is a very exciting study, but many questions remain that require further investigation. Do similar mechanisms drive the widespread synapse loss that is seen in normal-appearing gray matter, or are these mechanisms specific to acutely inflamed lesional cortex? There is growing evidence that neurodegeneration, possibly including synapse loss, occurs very early in the course of multiple sclerosis. Therefore, synapse loss may be initiated years in advance of when patients transition from relapsing remitting disease to progressive MS.

    We need to better understand the timing of when synapse loss is initiated in the disease course, and at what time points it is amenable to treatment. These authors showed that synapse loss was transient in the setting of acute inflammation, so how does transient synapse loss become permanent widespread synapse loss in multiple sclerosis?

  4. This is a very interesting study that examines the structural and functional consequences of neuroinflammation within the mouse cortex to provide novel insights into the pathophysiology of multiple sclerosis (MS), an inflammatory disease of the central nervous system whose underlying mechanistic details are poorly defined.

    The authors utilize an experimental model of MS whereby proinflammatory cytokines known to be highly expressed in the meninges of MS patients are injected into the cortices of mice previously immunized with myelin-oligodendrocyte glycoprotein. They show that a single injection results in widespread synapse loss after three days, with a significant bias toward the reduction of excitatory connections. Analysis of the frequency of spontaneous calcium transients within cortical lesions revealed a consistent decrease in network activity in vivo. The authors confirm that this synaptic loss is driven by increased phagocytosis of synaptic components by both local microglia and infiltrating macrophages, with synapses showing elevated calcium concentrations being preferentially targeted for removal.   

    Most strikingly, the authors report that synapse density and neuronal activity recover to control levels 10 days post-injection, and that the pattern of spontaneous correlated firing within neuronal ensembles is largely preserved post-recovery. This suggests that endogenous homeostatic mechanisms within the cortex are unimpacted by MS-related neuroinflammation and that, upon the cessation of the inflammatory stimuli, are sufficient to repair significant structural and functional alterations. Additionally, the authors trialed a low-dose treatment with an inhibitor of colony stimulating factor 1 receptor, which greatly alleviated synapse loss and reduced phagocyte activation without diminishing overall microglia numbers.

    Together, these findings represent substantial promise for the clinical treatment of MS, as they suggest that early intervention with immunomodulatory therapies could work in combination with endogenous healing processes within the brain to prevent progressive MS-associated neurodegeneration, which is thought to be the major driver of progressive neurological disability in MS patients.

    However, a caveat is that the model used in this study relied on the induction of a single inflammatory episode by one injection of cytokines into the cortex. In contrast, clinical MS is driven by sustained proinflammatory signalling within the CNS. Therefore, whether such a striking recovery of neuronal function would be observed following the cessation of a sustained neuroinflammatory insult remains an open question.

    These findings also open several additional avenues for future research. Perhaps most intriguing is the mechanism underlying the preservation of connectivity patterns within neuronal ensembles following recovery from widespread synapse removal. Are there populations of resilient synapses that resist degradation and maintain a low level of connectivity, or do neurons maintain a “memory” of their previous connectivity patterns, allowing them to be precisely restored during recovery? However, it should be noted that in vivo recordings in this study were performed while animals were under light anaesthesia, a condition where neuronal activity is known to be highly synchronized to slow wave oscillations (Kerr, Greenberg, and Helmchen, 2005). Whether a similar recovery of coordinated activity is seen for subtler, more variable neuronal responses, such as sensory-evoked firing, would an interesting question to pursue.  

    Additionally, by what mechanism does high synaptic calcium concentration trigger phagocytes to target individual synapses for removal? In direct contrast, increased synaptic calcium concentration acts as the primary signal for maintaining individual synapses during synaptic pruning (Helias et al., 2008), a fundamental process of neurodevelopment. It will therefore be important to determine the differential mechanisms by which elevated calcium concentrations can induce synapse removal in the context of MS-related neuroinflammation while promoting synapse maintenance during development. Moreover, recent evidence indicates that neuronal hypoactivity activates microglia (Liu et al., 2019) and that microglia can in turn reduce neuronal firing (Badimon et al., 2020). It would therefore be interesting to determine the functional interactions between neurons and microglia more directly using the experimental paradigm utilized in this study.

    Finally, the neuropathological changes described in this study are not unique to MS. For example, our own lab and several others have also described network hypoactivity in tauopathy models associated with Alzheimer’s disease (Busche et al., 2019; Harris et al., 2020). Also, phagocytic removal of synapses is a hallmark of many neurodegenerative and neuropsychiatric disorders. It will therefore be important to assess whether convergent mechanisms exist between multiple conditions in order to give rise to this shared neuropathology. If so, deciphering such mechanisms would likely have widespread clinical impacts upon multiple neurological conditions.


    . Negative feedback control of neuronal activity by microglia. Nature. 2020 Oct;586(7829):417-423. Epub 2020 Sep 30 PubMed.

    . Tau impairs neural circuits, dominating amyloid-β effects, in Alzheimer models in vivo. Nat Neurosci. 2019 Jan;22(1):57-64. Epub 2018 Dec 17 PubMed.

    . Tipping the Scales: Peptide-Dependent Dysregulation of Neural Circuit Dynamics in Alzheimer's Disease. Neuron. 2020 Aug 5;107(3):417-435. Epub 2020 Jun 23 PubMed.

    . Structural plasticity controlled by calcium based correlation detection. Front Comput Neurosci. 2008;2:7. Epub 2008 Dec 24 PubMed.

    . Imaging input and output of neocortical networks in vivo. Proc Natl Acad Sci U S A. 2005 Sep 27;102(39):14063-8. Epub 2005 Sep 12 PubMed.

    . Neuronal network activity controls microglial process surveillance in awake mice via norepinephrine signaling. Nat Neurosci. 2019 Nov;22(11):1771-1781. Epub 2019 Oct 21 PubMed.

  5. Imaging technologies have advanced so much over the last 10-15 years and our knowledge of how different cell types interact with one another in the brain —within living animals—has sharply increased during this time. This is no coincidence. Along this line, these authors have really taken advantage of several powerful imaging technologies to address critical cellular/molecular mechanisms of neurodegeneration, while simultaneously offering translational approaches for improving neural function. I think this article will have a tremendous impact on the MS field and will be very influential in other research fields, e.g., Alzheimer’s and related disorders, as well.

    Several decades ago, Phil Landfield, Zaven Khatchaturian, John Disterhoft, Gary Gibson, and others hypothesized that Ca2+ dysregulation was a key mechanism for driving neural dysfunction and degeneration during aging and neurodegenerative diseases. The present work, which shows Ca2+ dysregulation in individual dendritic spines as a likely signal for microglia/macrophage-mediated removal of those spines, is yet another example of just how vital the Ca2+ hypothesis is after all these years, and further cements the visionary status of these early Ca2+ hypothesis proponents. 

    Nearly everyone who studies the role of neuroinflammation in the brain understands how complex this process is. Clearly, there are both beneficial and detrimental consequences of neuroimmune and inflammatory signaling. Much discussion has centered on how to promote the “good” while limiting the “bad” inflammation. The study by Jafari et al., along with other recent studies in the Alzheimer’s and related disorders field, further suggests that “premature” or aberrant synapse elimination is a prime example of a “bad” consequence of neuroinflammation that can be specifically targeted and curtailed. Drug developers should be paying close attention.

    Investigations in the Jafari et al. study were focused on multiple sclerosis, and the preclinical model they developed was appropriately limited to acute bouts of cortical inflammation that were associated with resolvable synapse loss. An interesting question that remains is whether net synapse loss, typical of many chronic neurodegenerative conditions, including Alzheimer’s, involves: (1) higher levels/rates of macrophage/microglia-mediated synapse removal; or 2) impaired synaptogenesis mechanisms (or synapse repair mechanisms); or 3) both. The answer to this question may help drive the development of new and distinct therapeutics for chronic neurodegenerative disorders.

  6. The authors developed a very valuable model to connect cortical inflammation, cortical circuit activity, and the loss of dendritic spines. Specifically, this effect is transient. When the inflammation receded, synaptic connectivity recovered, and cortical neurons re-established their original firing pattern. The authors also provided evidence that during inflammation it is the microglia and macrophages that are responsible for removing the spines. Blocking these cells protected synapses from removal, which suggests potential targets for immunomodulatory therapy.

    In this work, Jafari and collaborators investigated how the calcium signaling could modulate the microglia/macrophage capacity to phagocyte spine dendrites during a cortical inflammation induced in a new model of multiple sclerosis. This model was based on immunization with a small amount of myelin-oligodendrocyte glycoprotein followed by stereotaxic injection of IFNγ and TNFα in the somatosensory cortex. Utilizing the Thy1-GFP-M~BiozziABH mice, they showed that neuroinflammation of the cortex occurs during demyelination that happens three days after the cytokine stimulation. This is followed by the increase in the number of microglia/macrophages in this region, which correlates with a decrease in dendritic spines and excitatory synapses.

    To evaluate cortical activity, the authors quantified the calcium concentration before and after the cytokine stimulation, and they could observe that a decrease in the calcium concentration tracked with a decrease in excitatory activity. This phenomenon could be correlated with a decrease in NMDA signaling; this receptor is responsible for calcium influx through glutamate synapses promoting the expression of BDNF and new spines. It is noteworthy that this activity of the calcium pathway is restored 10 days after cytokine stimulation, and that neurons that previously expressed higher concentrations of calcium before injury typically averaged similarly high levels of activity after recovery. However, the most interesting part of this work was the demonstration that the elimination of dendritic spines occurs after the calcium overexpression at the same spine, a process that occurs each hour and signals microglia/macrophages to start the phagocytosis.

    To distinguish the microglial from the macrophage response in this process, the author utilized animals expressing CX3CR1-GFP for identification of microglia cells and CCR2-RFP to track infiltrated phagocytes. It is possible to see that both cells are capable of phagocytosing the synapses during the neuroinflammation. Moreover, the authors showed that blocking CSF-1R decreased the spinal loss, with activated microglia (CD45high, CD11bhigh, CC2-) being the population with the most genes affected (most of them correlated with MGnD profile). The CSF1-blocker also decreased the phagocytic capacity of both microglia and infiltrating macrophages.

    Despite the interesting results, some aspects could be better evaluated. The first is the source of calcium increase. Its already shown that pro-inflammatory cytokines such as IL6 and IL17 in the brain can induce ATP release, which can induce a high calcium influx by P2X7 receptors. Another question is why microglia/macrophages preferentially target excitatory synapses in this process. Is it a glutamate-NMDA receptor-based process? In this new cortical MS model, the presence of anti-MOG is important (because of the time between the immunization and cytokine administration); however, the complement activity was not evaluated, which could be important, since the genes that regulate C1q are altered in MGnD. In relation to the model, the role of T cells was little explored. The authors showed that cytokine administration itself can induce infiltration of monocytes and T cells to the cortex, however, without demyelination and spine loss. This reinforces the idea that the presence of T cells without MOG-immunization does not induce disease. Also, it will be interesting to evaluate how this cortex demyelination and spine dendritic loss impact cognition.

    It is very interesting that the neuronal loss was transient (figure 2D and 3A). The authors tracked the activity of individual neurons over time and found they resumed activity after recovery from the inflammation. Are the neurons fully recovered to the original status? And what is the mechanism? 

    A final note: We know that M-CSF (CSF-1) is involved in microglial and macrophage proliferation. MGnD increases the expression of the CSF-1 receptor after phagocytosing dead neurons. However, this receptor normally associates with an M0-homeostatic response. Thus, which factors distinguish the homeostatic function of microglia from its pathogenic role in neurodegeneration remains an open question. Despite the inhibition of CSF1r decreasing the microglia phagocytosis and the macrophage recruitment, it would be interesting to show whether blocking the CSF1 signaling also decreases the demyelination.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Microglia Control Synapse Number in Multiple Disease States
  2. Paper Alert: Microglia Mediate Synaptic Loss in Early Alzheimer’s Disease
  3. Microglia Prune Synapses in a Subtype of Frontotemporal Dementia
  4. More Calcium News: Plaques Cause Dendrite Damage via Ion Overload
  5. ApoE and Trem2 Flip a Microglial Switch in Neurodegenerative Disease

Paper Citations

  1. . Pathological mechanisms in progressive multiple sclerosis. Lancet Neurol. 2015 Feb;14(2):183-93. PubMed.
  2. . Deep gray matter volume loss drives disability worsening in multiple sclerosis. Ann Neurol. 2018 Feb;83(2):210-222. Epub 2018 Feb 6 PubMed.
  3. . The cortical damage, early relapses, and onset of the progressive phase in multiple sclerosis. Neurology. 2018 Jun 12;90(24):e2107-e2118. Epub 2018 May 16 PubMed.
  4. . Reconstruction of single cortical projection neurons reveals primary spine loss in multiple sclerosis. Brain. 2016 Jan;139(Pt 1):39-46. Epub 2015 Dec 14 PubMed.
  5. . Synaptic pathology in the cerebellar dentate nucleus in chronic multiple sclerosis. Brain Pathol. 2017 Nov;27(6):737-747. Epub 2017 Mar 6 PubMed.
  6. . CSF1R signaling is a regulator of pathogenesis in progressive MS. Cell Death Dis. 2020 Oct 23;11(10):904. PubMed.

Further Reading


  1. . Calcium Influx through Plasma-Membrane Nanoruptures Drives Axon Degeneration in a Model of Multiple Sclerosis. Neuron. 2019 Feb 20;101(4):615-624.e5. Epub 2019 Jan 24 PubMed.

Primary Papers

  1. . Phagocyte-mediated synapse removal in cortical neuroinflammation is promoted by local calcium accumulation. Nat Neurosci. 2021 Mar;24(3):355-367. Epub 2021 Jan 25 PubMed.