Mitochondria may be the universal powerhouses of eukaryotic cells, but there is more to these organelles than just ATP production. According to a paper in the September 9 Nature Neuroscience, mouse mitochondria express distinct subsets of proteins that imbue them with cell-specific functions. Researchers led by Thomas Misgeld of the German Center for Neurodegenerative Diseases in Munich reported that the mitochondrial proteome differed between astrocytes and neurons, and even between different types of neuron. Using these distinctive proteomes to identify cell-type-specific mitochondrial markers, the researchers spotted a dearth of neuronal mitochondria around amyloid plaques and in the spinal cord in mouse models of Alzheimer’s disease and amyotrophic lateral sclerosis (ALS), respectively. They saw the same in postmortem tissue from people with those diseases.

  • A new tagging system plucks intact mitochondria from the brain.
  • Their proteomes vary depending on cellular origin.
  • Around amyloid plaques, neuronal mitochondria are scarcer than astrocytic ones.

“We’ve known for over a decade that the mitochondrial proteome varies across organs. This paper is now reporting differences across even cell types,” commented Vamsi Mootha of Massachusetts General Hospital in Boston. “This is an exciting advance, one that will hopefully help the community unravel the role of mitochondrial dysfunction in neurodegeneration.”

Russell Swerdlow, University of Kansas Medical Center in Kansas City, agreed. “This is a very creative methods-development paper that will likely fill a much-needed niche in the mitochondrial research field,” he wrote to Alzforum. “As an investigator who labors over separating neurons, astrocytes, microglia, and endothelial cells from mouse brains in order to get a better grasp of what their mitochondria are up to, the potential of this approach to facilitate studies of brain mitochondria is obvious.”

Mitochondria Vacate Premises. In an amyloid plaque (left, dashed line), astrocyte mitochondria bearing the marker Sfxn5 (green, middle) and neuronal mitochondria expressing Ociad2 (purple, right) are relatively depleted from its core (solid white line). [Courtesy of Fecher et al., Nature Neuroscience, 2019.]

Mitochondrial function deteriorates in neurodegenerative diseases, including Alzheimer’s, Parkinson’s, and ALS. Still, it has been difficult to nail down exactly when malfunction begins, and if it is limited to specific cell types. For one thing, mitochondria often get damaged when they are purified from isolated cells; for another, their abundance makes it difficult to determine their origin.

To address these problems, co-first authors Caroline Fecher, Laura Trovò, and colleagues generated MitoTag mice. These animals harbor a Cre-activated gene that expresses a green fluorescent protein (GFP) targeted to the outer mitochondrial membrane. By crossing MitoTag mice with a strain that expresses Cre recombinase under the control of cell-specific promoters, the researchers were able to limit the green tag to those particular cells. This allowed them to isolate the tagged mitochondria using GFP antibodies, a far gentler approach than using mechanical homogenization and centrifugation to separate mitochondria from cell preparations.

In the resent study, the researchers used MitoTag to label mitochondria from three cell types in the cerebellum: Purkinje cells, a major inhibitory neuron; granule cells, a major excitatory neuron in this region; and astrocytes. Mitochondrial proteomes from all three cell types were 85 percent identical, and yet nearly 200 proteins were differentially expressed.

The proteome of astrocyte mitochondria most clearly diverged from those of neuronal ones, and pointed to functional differences. For example, astrocytic mitochondria abundantly expressed proteins involved in lipid metabolism that were absent or scarce in neuronal mitochondria, and they expressed higher levels of enzymes involved in fatty-acid metabolism. Though the brain primarily relies on glucose and lactate for energy, some studies have suggested it can metabolize fatty acids in a pinch. The researchers went on to confirm their proteomic results with functional assays, finding that astrocytic mitochondria more efficiently metabolized a long-chain fatty acid than did their neuronal counterparts.

The proteomes also laid bare differences between the two subtypes of neuronal mitochondria. Granule cell mitochondria were flush with the calcium channel Mcu, while Purkinje cell and astrocyte mitochondria had little of it. Functional assays confirmed that the granule cell organelles robustly took up calcium in a Mcu-dependent manner, while calcium uptake in Purkinje cell mitochondria was more sluggish. Maria Ankarcrona of Karolinska Institute in Stockholm noted that these findings mesh with granule cells being excitatory and using glutamate receptors and calcium influx for their signaling.

Moreover, the proteomes pointed to cell-type-specific mitochondrial markers. Sfxn5 marked astrocytic mitochondria, while Ociad2 and Nipsnap1 were neuronal. The German study used these markers to track different types of mitochondria in mouse models of disease and in human tissue, sans MitoTag.

In preliminary experiments using three APP/PS1 mice and three wild-type, the researchers found that the density of both Ociad2-positive neuronal mitochondria and Sfxn5-positive astrocytic mitochondria were less around Aβ plaques than in the normal parenchyma (see image above). In the spinal cords of three SOD-G93A mice, a model of motor neuron disease, neuronal mitochondria were morphologically altered, while astrocytic mitochondria appeared normal. Antibodies against these same markers revealed similar changes in distribution and shape of mitochondria in postmortem brain and spinal cord samples from four AD and four ALS patients, respectively. The researchers did not quantify any of these changes. Misgeld said these experiments merely served as a proof of concept that the markers could tag mitochondria from different cell types in disease models. Fecher said this initial study did not distinguish between loss of cells and loss of mitochondria, but that future studies could clarify how neuronal, astrocytic, and even microglial mitochondria are affected by pathology.

Mark Cookson of the National Institutes of Health thinks the models could help nail down in which cells mitochondria malfunction in Parkinson’s—where mutations in mitochondrial proteins PINK1 and parkin lead to early onset disease. “It would be of huge interest to express the MitoTag in different cells in PINK1 and parkin-deficient mice and see how the [mitochondrial] proteome reshapes under different conditions,” he wrote.

Fecher told Alzforum that MitoTag mice may also prove useful in examining differences in mitochondria that reside in axons and dendrites, versus those in the cell body. She pointed out that mitochondria located far from the soma rely heavily on local translation from nearby ribosomes, and may therefore have a distinct set of proteins and functions.

MitoTag mice are now available for purchase at The Jackson Laboratory.—Jessica Shugart


  1. In this very nice study, Misgeld’s group has engineered a set of MitoTag mice expressing GFP in the outer mitochondrial membrane (OMM) in a cell-type-specific manner. Importantly, they demonstrate that GFP-OMM-tagged mitochondria can be purified in separate fractions and that it is possible to, for example, study the mitochondrial proteome in specific cell populations. I think the most interesting is to see that mitochondrial proteins are differently expressed depending on cell type, revealing that mitochondria are not all the same but have diverse functions in different cell types.

    Using this method, it is also possible to determine cell-specific mitochondrial markers and changes in these markers related to neurodegeneration. In terms of function, mitochondria in cerebellar granule cells show a significantly higher expression of the mitochondria calcium uniporter (Mcu) as compared to Purkinje cell mitochondria. This is also reflected by the cell-type-specific mitochondria’s capacity to buffer calcium and their sensitivity to Mcu ablation. Mitochondria from granule cells buffer calcium better and are more sensitive to loss of Mcu as compared to inhibitory Purkinje cells (PC), very much in line with the fact that granule cells are excitatory and signal via glutamate receptors and calcium influx.

    We have for several years studied the endoplasmic reticulum (ER)-mitochondria interface and how the interplay between these two organelles is altered in Alzheimer’s disease (Hedskog et al., 2013; Leal et al., 2016; Leal et al., 2018; Schreiner et al., 2015; Filadi et al., 2018). Intriguingly, Misgeld and colleagues identify Rmdn3 (PTPIP51) in the proteomic profiling and show that Rmdn3 is enriched in PC mitochondria. Rmdn3 was first identified by Miller’s group as an OMM protein interacting with the ER protein VAPB. The Rmdn3 (PTPIP51)-VAPB complex is established as a scaffold between mitochondria and ER (DeVos et al., 2011) at specialized contact points referred to as mitochondria-associated membranes (MAM) (Vance, 1990). Here it is shown that PC have increased ER-mitochondria contacts, suggesting that such cells mainly handle calcium buffering via ER- mitochondria shuttling.

    Using human brain tissue, AD mouse models, and cells we have shown a correlation between Aβ levels and increased ER-mitochondria contact as well as increased calcium shuttling between the two organelles. Currently, partly based on RNA-Seq data, we hypothesize that in AD, increased ER-mitochondria contact is an initial stress response and a way for the cell to support mitochondrial functions and thus synaptic activity. As AD progresses, the dysregulation of organelle contact may turn into negative effects and thus negatively affect neuronal function and survival.


    . Alterations in mitochondria-endoplasmic reticulum connectivity in human brain biopsies from idiopathic normal pressure hydrocephalus patients. Acta Neuropathol Commun. 2018 Oct 1;6(1):102. PubMed.

    . TOM70 Sustains Cell Bioenergetics by Promoting IP3R3-Mediated ER to Mitochondria Ca2+ Transfer. Curr Biol. 2018 Feb 5;28(3):369-382.e6. Epub 2018 Jan 27 PubMed.

    . Amyloid-β peptides are generated in mitochondria-associated endoplasmic reticulum membranes. J Alzheimers Dis. 2015;43(2):369-74. PubMed.

    . Mitofusin-2 knockdown increases ER-mitochondria contact and decreases amyloid β-peptide production. J Cell Mol Med. 2016 Sep;20(9):1686-95. Epub 2016 May 20 PubMed.

    . Modulation of the endoplasmic reticulum-mitochondria interface in Alzheimer's disease and related models. Proc Natl Acad Sci U S A. 2013 May 7;110(19):7916-21. PubMed.

    . VAPB interacts with the mitochondrial protein PTPIP51 to regulate calcium homeostasis. Hum Mol Genet. 2011 Dec 13; PubMed.

    . Phospholipid synthesis in a membrane fraction associated with mitochondria. J Biol Chem. 1990 May 5;265(13):7248-56. PubMed.

  2. The work from Fecher et al. is a very innovative solution to the older problem of estimating if mitochondria in major cell types of the mammalian brain are different. Based on their differing physiology and energetic requirements, it is gratifying, if expected, that there were differentially expressed mitochondrial proteins between neurons and astrocytes. The additional differences in mitochondrial proteomes between neuronal subtypes also reflects important physiological differences, as the authors point out in their discussion, around Purkinje cells’ calcium handling.

    While there is some exploration of AD and ALS mice in this study, one can see a number of additional future applications in the context of disease modeling. For example, several groups have now proposed mechanisms by which PINK1 and parkin, both associated with early onset recessive parkinsonism, affect mitochondria in vivo but with different predicted effects on the mitochondrial proteome. Richard Youle’s lab has proposed that loss of parkin results in a failure of control of pathogenic mtDNA mutations (Pickrell et al., 2015), whereas more recently, Matheoud et al. showed that PINK1 is critical for mitochondrial antigen presentation in the context of bacterial infections (Matheoud et al., 2019). It would be of huge interest to express the MitoTag in different cells in PINK1 and parkin-deficient mice and see how the proteome reshapes under different conditions.


    . Intestinal infection triggers Parkinson's disease-like symptoms in Pink1-/- mice. Nature. 2019 Jul;571(7766):565-569. Epub 2019 Jul 17 PubMed.

    . Endogenous Parkin Preserves Dopaminergic Substantia Nigral Neurons following Mitochondrial DNA Mutagenic Stress. Neuron. 2015 Jul 15;87(2):371-81. PubMed.

  3. This is a very creative methods-development paper that will likely fill a much-needed niche in the mitochondrial research field. Over the past few decades it has increasingly become clear that mitochondria are not equivalent between cell types. For example, we’ve come to realize mitochondrial proteomes differ between different tissues. But what about mitochondria within different cell types within a single tissue? This issue is critical in studies of the brain, since the brain is not a homogeneous organ. It contains different cell types, and it has become increasingly appreciated that mitochondria in neurons differ functionally and even structurally from mitochondria in astrocytes.

    Unfortunately, the techniques used to isolate mitochondria from the different cell types within the brain are cumbersome and can disrupt mitochondrial integrity. In defining an approach that leverages a transgenic mouse that can strategically tag cell-type-specific mitochondria, with subsequent immunocapture of those mitochondria, the authors show they can effectively and efficiently enrich for intact, cell-type-specific brain mitochondria.

    As an investigator who labors over separating neurons, astrocytes, microglia, and endothelial cells from mouse brains in order to get a better grasp of what their mitochondria are up to, the potential of this approach to facilitate studies of brain mitochondria is obvious. From the perspective of someone who studies mitochondria in Alzheimer’s disease, I would hope this will allow for, and facilitate, new insights into how mitochondria and their function pertain to the disease. The one caveat is that while this technique will benefit studies of AD mouse model mitochondria, we will still need to not overlook the general limitations of these models.      

Make a Comment

To make a comment you must login or register.


Research Models Citations

  1. APPPS1

Further Reading

Primary Papers

  1. . Cell-type-specific profiling of brain mitochondria reveals functional and molecular diversity. Nat Neurosci. 2019 Oct;22(10):1731-1742. Epub 2019 Sep 9 PubMed.