Microglial autophagy–associated phagocytosis is essential for recovery from neuroinflammation

See allHide authors and affiliations

Science Immunology  16 Oct 2020:
Vol. 5, Issue 52, eabb5077
DOI: 10.1126/sciimmunol.abb5077

Microglia Matter

Multiple sclerosis (MS) is an autoimmune disease of the central nervous system, and microglial activation is linked to its progression. Berglund et al. show that microglia carry out a noncanonical form of autophagy that targets degradation and clearance of myelin. Deletion of the autophagy regulator Atg7, but not the canonical macroautophagy protein Ulk1, was associated with greater accumulation of phagocytosed myelin and lack of recovery from MS-like disease in mice. Microglia from aged wild-type mice with MS-like disease had functional and phenotypic similarities to Atg7-deficient microglia, but treatment of aged mice with trehalose promoted myelin clearance and disease remission. These results suggest that a noncanonical form of autophagy in microglia is critical to myelin degradation and clearance, thus providing insight into microglia function during MS.


Multiple sclerosis (MS) is a leading cause of incurable progressive disability in young adults caused by inflammation and neurodegeneration in the central nervous system (CNS). The capacity of microglia to clear tissue debris is essential for maintaining and restoring CNS homeostasis. This capacity diminishes with age, and age strongly associates with MS disease progression, although the underlying mechanisms are still largely elusive. Here, we demonstrate that the recovery from CNS inflammation in a murine model of MS is dependent on the ability of microglia to clear tissue debris. Microglia-specific deletion of the autophagy regulator Atg7, but not the canonical macroautophagy protein Ulk1, led to increased intracellular accumulation of phagocytosed myelin and progressive MS-like disease. This impairment correlated with a microglial phenotype previously associated with neurodegenerative pathologies. Moreover, Atg7-deficient microglia showed notable transcriptional and functional similarities to microglia from aged wild-type mice that were also unable to clear myelin and recover from disease. In contrast, induction of autophagy in aged mice using the disaccharide trehalose found in plants and fungi led to functional myelin clearance and disease remission. Our results demonstrate that a noncanonical form of autophagy in microglia is responsible for myelin degradation and clearance leading to recovery from MS-like disease and that boosting this process has a therapeutic potential for age-related neuroinflammatory conditions.


Multiple sclerosis (MS) is a chronic disease characterized by inflammation in the central nervous system (CNS) that triggers demyelination, glial cell dysfunction, and irreversible neuro-axonal damage (1). Although recent understanding of this immune dysfunction has led to the development of effective disease-modulatory treatments for the inflammatory-active relapsing-remitting phase of MS, there is only one recently approved treatment for the secondary progressive phase (2). Disease progression is not only the most clinically challenging aspect of MS, but it is also the least mechanistically explored. Microglia activation, mitochondrial damage, and ionic imbalance, among other mechanisms, have been associated with progressive neurodegeneration during MS (3).

Progressive MS exhibits a degenerative disease phenotype with glial cell dysfunction rather than infiltration of peripheral immune cells (3). In MS, microglia express an increased pro-inflammatory profile with both age and disease burden (4). Age is the strongest risk factor for developing progressive MS (5). Many age-associated neurodegenerative pathologies such as Alzheimer’s disease (AD) (68), as well as autoinflammatory diseases including Crohn’s disease and systemic lupus erythematosus (SLE) (9, 10), are characterized by impaired autophagy, a lysosomal degradation pathway used for removal of cellular constituents. Microglia surrounding MS lesions exhibit enhanced autophagy (11) and have the ability to phagocytose oligodendrocytes, as recently shown by single-nucleus RNA sequencing (12). Among CNS myeloid cells, microglia have been ascribed the highest phagocytic activity (13), but the underlying mechanisms and their association with MS progression or tissue repair remain to be characterized.

Canonical autophagosome formation is highly dependent on the Unc-51–like kinase 1 (ULK1)–complex formation, as well as on autophagy-related protein 7 (ATG7) lipidation of LC3, the key component of autophagosomes. In addition, LC3 can also be conjugated to membranes of phagosomes and endosomes, thereby facilitating the degradation of their cargo during processes termed LC3-associated phagocytosis and endocytosis, respectively (1416), broadly referred to as noncanonical autophagy hereafter. Noncanonical autophagy is dependent on the protein RUBICON (17), which also inhibits the canonical autophagy pathway (18). Mutations in the Rubicon gene are associated with a familial form of ataxia with impaired lysosomal degradation (19, 20). Furthermore, inhibition of noncanonical autophagy in macrophages elicited by deletion of either Atg7 or Rubicon causes an SLE-like disease in mice due to defective degradation of phagocytosed apoptotic cells (21). Similarly, specific impairment of noncanonical autophagy in microglia has been associated with reduced clearance of β-amyloid and progressive neurodegeneration in a murine model of AD (16), also evident in Atg7-deficient mice (22). Whereas targeting microglial autophagy in these diseases has been proposed to have great therapeutic potential (8), whether canonical or noncanonical autophagy affects disease progression in MS is currently unknown.

We previously established a link between Atg7 and disease severity in a common animal model of MS, experimental autoimmune encephalomyelitis (EAE) (23). Here, we pinpoint this effect to microglia, and we reveal how processing of myelin debris by microglia is dependent on ATG7 in a noncanonical form of autophagy. We further establish this process as a determinant of microglial phenotype in disease and aging and demonstrate how therapeutically inducing this pathway can restore CNS homeostasis.


Microglial Atg7 deficiency prevents recovery from EAE

To investigate the association between Atg7 and EAE, we deleted Atg7 from two compartments highly relevant for disease pathogenesis, T cells and myeloid cells, using CRE recombinase expressed under Lck and Lyz2 promoters, repectively (fig. S1). Deletion of Atg7 in T cells did not affect clinical disease, although we did observe a previously reported reduction in CD8+ T cell numbers (fig. S1, A and B) (24). In contrast, deletion of Atg7 in Lyz2-expressing myeloid cells led to a persistent disease state that lacked the recovery evident in wild-type control mice (fig. S1C). Lyz2Cre targets several myeloid cell types including microglia (fig. S1D) (22, 25, 26), and the effect of Atg7 deletion was restricted to the EAE recovery phase, suggesting CNS-intrinsic regulation. We detected the highest expression and most prominent Atg7 up-regulation during EAE in CD11b+ CD45INT cells (microglia) compared with other Lyz2-expressing CNS myeloid populations (Fig. 1A). Expression of the floxed Atg7 exon 14 was markedly reduced in microglia of Atg7fl/fl Lyz2Cre mice after EAE induction (fig. S1E). In addition, microglia from these mice displayed reduced lipidated membrane-bound LC3B (II) (fig. S1F), indicating less autophagosome formation and lysosomal loading of myelin after in vitro exposure (fig. S1G).

Fig. 1 Recovery from EAE requires microglial autophagy.

(A) Relative expression of Atg7 in microglia (CD45INT naïve, n = 4; and EAE, n = 6), bone marrow–derived monocytes/macrophages (CD45Hi naïve, n = 3; and EAE, n = 3) and neutrophils (Ly6G+ naïve, n = 3; and EAE, n = 3) from naïve and day 15 EAE animals, detected by qPCR and normalized to the geometric mean of two endogenous control genes, Gapdh and Hprt. (B) Disease course in Atg7wt/wt (n = 16), Atg7wt/fl (n = 20), and Atg7fl/fl (n = 19) mice and (C) in Ulk1wt/wt (n = 13), Ulk1fl/fl (n = 12), and Atg7fl/fl (n = 16) mice. Clinical course was compared using one-way analysis of variance (ANOVA) with Tukey’s post hoc test on area under curve. Error bars indicate confidence intervals. (D) Membrane-bound LC3B (II) detected by flow cytometry in naïve Atg7wt/wt (untreated, n = 7; starved, n = 7), Ulk1fl/fl (untreated, n = 4; starved, n = 8), Atg7fl/fl (untreated, n = 4; starved, n = 8), and Atg7fl/fl Ulk1fl/fl [“double knock-out” (DKO); untreated, n = 4; starved, n = 5] microglia after starvation in vitro. (E) Relative expression of Rubicon in microglia (CD45INT naïve, n = 6; and EAE, n = 4), bone marrow–derived monocytes/macrophages (CD45Hi naïve, n = 4; and EAE, n = 5) and neutrophils (Ly6G+ naïve, n = 6; and EAE, n = 5) from naïve and day 15 EAE animals, detected by qPCR and normalized to the geometric mean of two endogenous control genes, Gapdh and B-actin. (F) Phagocytosed myelin debris (dMBP) assessed in microglia from Atg7wt/wt, Ulk1fl/fl, and Atg7fl/fl mice 21 days p.i. by flow cytometry (all conditions, n = 5). (G) Membrane-bound LC3B (II) detected by flow cytometry after antibody labeling of ex vivo microglia 21 days p.i. in Atg7wt/wt (n = 5) and Atg7fl/fl (n = 6) mice. MFI, mean fluorescence intensity. (H) Example images of immunofluorescence and image analysis of fluoromyelin-stained myelin (FM, red), LC3B (II), and LC3B (II):myelin (overlapping pixels) of ex vivo microglia 5 days p.i. from Atg7wt/wt (n = 4), Ulk1fl/fl (n = 4), and Atg7fl/fl (n = 6) mice. (I) In vitro pulsing of microglia 5 days p.i. with CellVue- and pHrodo-stained myelin, assessed by flow cytometry (all conditions, n = 4; except pHrodo, n = 3). (J) In vitro clearance of PKH26-stained myelin from medium by microglia 5 days p.i. from Atg7wt/wt (n = 7) and Atg7fl/fl (n = 7) mice. Statistics: (A, E, and G) Mann-Whitney U test, (D, F, H, and J) Kruskal-Wallis test followed by Dunn’s post hoc test, and (I) ANOVA followed by Dunnet’s post hoc test (***P < 0.001, **P < 0.01, and *P < 0.05). Error bars indicate SEM. Experiments (A to D and E and H) were performed twice and (B, C, F, G, I, and J) three times.

To test the hypothesis that microglial ATG7 plays a role in EAE progression, we used Atg7 deletion under a tamoxifen (TAM)–inducible CRE recombinase expressed from the Cx3cr1 promoter. TAM-induced deletion is sustained in the self-renewing microglia population, whereas other Cx3cr1-expressing cells essential for EAE such as monocytes and dendritic cells are derived from bone marrow precursors and repopulated from the bone marrow after 2 to 4 weeks (fig. S2A) (27). In all experiments, EAE was induced 4 to 8 weeks after TAM treatment. Deletion of Atg7 exon 14 in Atg7fl/fl Cx3cr1CreERT2 microglia was confirmed at both 2 and 42 weeks after TAM administration (fig. S2B), as was reduced ATG7 protein expression (fig. S2C). Atg7fl/fl Cx3cr1CreERT2 mice exhibited a marked loss of recovery from EAE, confirming our hypothesis of a microglia-dependent phenotype (Fig. 1B).

Atg7 deficiency affects microglial tissue debris clearance through noncanonical autophagy

Because ATG7 is essential for both canonical and noncanonical autophagy (14, 17), we compared phenotypes of Atg7fl/fl Cx3cr1CreERT2 mice with mice in which microglia were deficient in ULK1, a protein only required for canonical autophagy (21). The Ulk1fl/fl Cx3cr1CreERT2 mice did not exhibit lack of recovery from EAE, suggesting that impaired noncanonical autophagy drives the disease phenotype in Atg7fl/fl Cx3cr1CreERT2 mice (Fig. 1C). This notion was further supported by the finding that microglia from both strains displayed reduced but comparable levels of starvation-induced canonical autophagy, as evident by decreased membrane-bound LC3B (II) (Fig. 1D) and increased interleukin-1β (IL-1β) secretion and mitochondrial membrane potential (fig. S2, D and E). This reflects an impairment in typical canonical autophagy functions in securing mitophagy and inflammasome stability (22, 2831). In addition, we determined the key regulator of noncanonical autophagy, Rubicon (17), to accompany elevated Atg7 expression in wild-type microglia after disease induction (Fig. 1E and table S1).

Deficiency in noncanonical autophagy has been associated with impaired degradation of bacterial proteins during infection and with impaired phagocytosis of apoptotic cells and β-amyloid, leading to SLE-like disease and AD, respectively (16, 2022, 32). Atg7fl/fl Cx3cr1CreERT2 microglia displayed a markedly increased load of intracellular myelin detected ex vivo during EAE (Fig. 1F). The ex vivo EAE microglia also exhibited reduced autophagosome formation detected as membrane-bound intracellular LC3B (II) after selective digitonin permeabilization (Fig. 1G). Immunofluorescence of microglia sorted 5 days after EAE induction, when microglia were activated but had not yet accumulated endogenous myelin load, confirmed the increased load of fluorescently labeled myelin and decreased LC3 colocalization to myelin-containing vesicles in Atg7fl/fl Cx3cr1CreERT2 compared with Ulk1fl/fl Cx3cr1CreERT2 and wild-type control mice (Fig. 1H).

To further validate the impaired degradation of the phagocytosed debris in Atg7-deficient microglia, we pulsed microglia with myelin stained with both a lipophilic dye (CellVue) and a pH-sensitive dye (pHrodo). Microglia were sorted ex vivo after immunization, circumventing the need for further stimulations, and we limited all incubations to ≤48 hours to minimize in vitro culture-induced changes of the microglial phenotype (33). We observed a decreased pHrodo signal in Atg7fl/fl Cx3cr1CreERT2 microglia accompanied by increased CellVue (Fig. 1I), indicating impaired degradation in Atg7fl/fl Cx3cr1CreERT2 microglia. This finding was further demonstrated using time-lapse imaging of microglia pulsed with labeled myelin, revealing an accumulation of phagocytosed myelin in Atg7fl/fl Cx3cr1CreERT2 microglia (movie S1).

The impaired loading to degradation vesicles was also observed when Atg7fl/fl Cx3cr1CreERT2 microglia were provided irradiated apoptotic CD171+ (neurons), glutamate/aspartate transporter–positive (GLAST+; astrocytes), or O4+ (oligodendrocytes) cells (fig. S2F). Further analysis demonstrated a markedly reduced clearance capacity of myelin debris from medium by microglia from Atg7fl/fl Cx3cr1CreERT2 compared with Ulk1fl/fl Cx3cr1CreERT2 and wild-type control mice after pulsed exposure to fluorescently labeled myelin (Fig. 1J). Together, our data suggest that an impairment in microglial autophagy–associated degradation of phagocytosed myelin compromises the clearance of myelin debris, leading to an inability to recover from MS-like disease.

Microglial Atg7 deficiency drives an altered transcriptional phenotype

To obtain a comprehensive overview of the consequences of Atg7 deletion during EAE, we performed RNA sequencing of microglia sorted from naïve mice and at 21 and 35 days postinduction (p.i.) of disease. We detected 467 and 147 differentially expressed genes (DEGs) (adjusted P value < 0.05; fold change > 1.5) between Atg7fl/fl Cx3cr1CreERT2 and littermate controls on days 21 and 35 p.i., respectively (table S1). Only 13 DEGs were detected (adjusted P value < 0.05; fold change > 1.5) between the genotypes in naïve microglia, suggesting that differences predominantly arise during EAE (table S1). We clustered DEGs (P < 0.01; fold change > 1.5) and then grouped the clusters on the basis of their expression patterns (Fig. 2A), performing functional annotation of the groups using ingenuity pathway analysis (IPA), over-representation analysis (ORA), and REViGO (table S2) (34, 35). The first group (orange) represented changes considerably more pronounced in Atg7fl/fl Cx3cr1CreERT2 microglia that occurred early in disease (day 21 p.i.) and returned close to levels in the naïve state by day 35 p.i. (Fig. 2A). These changes associated with pathways typical for activation of immune cells during EAE, such as interferon-γ (IFN-γ), signal transducers and activators of transcription 3 (STAT3), and granulocyte-macrophage colony-stimulating factor (GM-CSF) signaling, cell activation, expansion, and migration (Fig. 2B). The second group (blue) represented genes that remained down-regulated during disease, with Atg7fl/fl Cx3cr1CreERT2 microglia showing modest changes (Fig. 2A). Functionally, this group was enriched in pathways involved in myeloid cell function such as quantity, movement, and degranulation of myeloid cells, as well as activation of GATA2, which is important for the development of myeloid lineage cells (Fig. 2B). Genes in the third group (yellow) had a similar pattern to the first group in Atg7 Cx3cr1CreERT2 but not in wild-type microglia, which demosntrated the opposite pattern (Fig. 2A). These changes are predominantly associated with energy-related functions such as glycolysis and mitochondrial function, oxidative stress, and cell adhesion (Fig. 2B and fig. S3). The fourth group (purple) represented genes that gradually increased their expression during disease progression specifically in Atg7fl/fl Cx3cr1CreERT2 microglia (Fig. 2A). These genes are associated with cellular growth, including protein synthesis, eukaryotic translation initiation factor 2 (EIF2), and mammalian target of rapamycin (mTOR) signaling, and were closely related to microglial development and function in CNS pathology (Fig. 2B and fig. S3) (36). The functional differences between the groups translated into differential disease enrichment, with early changes associating with inflammatory and infectious diseases, whereas changes in the progressive stage of EAE demonstrated strong enrichment in neurodegenerative diseases (Fig. 2C). These patterns resemble the course of MS, with the initial phase being dominated by inflammatory processes followed by neurodegeneration that, in later disease stages, becomes decoupled from the initial inflammation (1).

Fig. 2 Atg7 deficiency induces pronounced and sustained alterations in microglial transcriptome during EAE.

Transcriptome analysis was performed using RNA sequencing of microglia sorted from the following groups of mice: naïve Atg7fl/fl (n = 4) and Atg7wt/wt (n = 3), day 21 p.i. Atg7fl/fl (n = 4, average score 3) and Atg7wt/wt (n = 3, average score 3), and day 35 p.i. Atg7fl/fl (n = 4, average score 2.5) and Atg7wt/wt (n = 4, average score 1.5) (table S1). (A) Heat map depicting gene clusters associated with genotype and EAE disease stage based on transcripts that displayed a P value <0.01 and fold change >1.5. The scale represents Z-score–transformed expression values (with red and blue indicating up-regulated and down-regulated genes, respectively, compared with the mean value of a gene from all samples). These gene clusters were further grouped according to their pattern of expression into the four groups that were analyzed using IPA to annotate significance: (B) Canonical pathways, upstream regulators, and diseases and functions [Benjamini-Hochberg (B-H)–adjusted P < 0.05], and (C) ORA for diseases using the GLAD4U database (false discovery rate < 0.05). Details regarding the differential expression analysis are presented in table S1, and a full list of significant functional annotations is provided in table S2. (A), (I), and (H) indicate activated, inhibited, and high, respectively.

Atg7fl/fl Cx3cr1CreERT2 microglia exhibited pronounced but transient changes during early disease (Fig. 2, A and B, orange and yellow) that translated into altered microglia function and EAE development. These transcriptional profiles associated with immune cell activation and migration reflected in increased microglia numbers and a more robust infiltration of bone marrow–derived macrophages (BMDMs), neutrophils, and T and B cells into the CNS parenchyma during early EAE (Fig. 3A). In accordance with IFN-γ and tumor necrosis factor (TNF) predicted to be activated upstream regulators (Fig. 2B and table S2), Atg7fl/fl Cx3cr1CreERT2 mice had an increased infiltration of IFN-γ+ T cells (fig. S4A). Atg7fl/fl Cx3cr1CreERT2 microglia themselves secreted markedly larger amounts of IFN-γ, TNF, and IL-1β (Fig. 3B), the latter suggesting activation of the inflammasome as previously reported to occur in response to myelin accumulation (37). Moreover, Atg7fl/fl Cx3cr1CreERT2 microglia were capable of stimulating T cell proliferation and expansion of pathogenic IL-17– and IFN-γ–producing cells in vitro (fig. S4, B to D). Atg7 deficiency promoted a microglial phenotype that can augment inflammatory responses during EAE (fig. S4).

Fig. 3 Atg7-deficient microglia have impaired myelin degradation and scavenger receptor recirculation associated with a pathogenic phenotype and increased inflammation in EAE.

(A) Spinal cord microglia and infiltrating immune cell counts from Atg7wt/wt (n = 4) and Atg7fl/fl (n = 4) mice at day 21 p.i. analyzed by flow cytometry. (B) ELISA of supernatants from microglia sorted 21 days p.i. and incubated for 24 hours in vitro; Atg7wt/wt (n = 8) and Atg7fl/fl (n = 9). (C) Transcriptome of microglia sorted from Atg7wt/wt (n = 4) and Atg7fl/fl (n = 4) mice 35 days p.i. analyzed by RNA sequencing and compared with microglia gene sets associated with disease and tissue homeostasis. (D) Representative images of immunofluorescence of CLEC7A and IBA1-expressing microglia at days 32 to 37 p.i. in spinal cord from Atg7wt/wt and Atg7fl/fl mice. DAPI defines nuclei. Scale bars correspond to 500 μm and 50 μm in the left and right panels, respectively. (E and F) Flow cytometry analysis of CLEC7A-expressing subpopulations in Atg7wt/wt (n = 6) and Atg7fl/fl (n = 6) microglia at day 35 p.i. (G) Flow cytometry analysis of the density of surface MSR1 staining in microglia from Atg7wt/wt (n = 6) and Atg7fl/fl (n = 6) mice in different subpopulations defined by levels of CLEC7A expression as shown in (E and F). (H) Uptake of labeled myelin by sorted Atg7wt/wt [control (Ctrl), n = 9; block, n = 8] and Atg7fl/fl (control, n = 6; block, n = 6) microglia after blocking of MSR1. (I) Flow cytometry quantification of intracellular MSR1 from naïve Atg7wt/wt and Atg7fl/fl microglia exposed to myelin 12 hours in vitro with or without 6-hour bafilomycin A1 treatment. All conditions, n = 6. (J) Surface and intracellular MSR1 detection by flow cytometry in Atg7wt/wt (surface, n = 7; intracellular, n = 4) and Atg7fl/fl (surface, n = 7; intracellular, n = 5) microglia day 5 p.i. (K) Expression of Apoe in Atg7wt/wt (n = 4), Atg7fl/fl (n = 5), and bafilomycin A1–treated Atg7wt/wt (n = 5) microglia exposed to myelin for 7 days in vitro. (L) ELISA of TGF-β1 secretion from microglia from naïve and day 35 p.i. mice (Atg7wt/wt, n = 5; Ulk1fl/fl, n = 5; and Atg7fl/fl, n = 5) after 24-hour in vitro culture. (M) Intracellular myelin debris (dMBP) assessed in microglia from Atg7wt/wt (n = 12) and Atg7fl/fl (n = 12) mice 35 days p.i. by flow cytometry. (N) Representative images of immunofluorescence of tissue deposits of myelin debris (dMBP) and density of Mac3+ macrophages at 37 p.i. in spinal cord from Atg7wt/wt and Atg7fl/fl mice. DAPI defines nuclei. Statistics: (A, B, F to H, J, L, and M) Mann-Whitney U test and (I and K) Kruskal-Wallis test followed by Dunn’s post hoc test (***P < 0.001, **P < 0.01, and *P < 0.05). Error bars indicate SEM. Experiment (N) is representative of three independent experiments. Experiments (B and H to M) were performed twice and (A, F, and G) three times.

In patients with MS, an increased inflammatory activity during the first 2 years after diagnosis has been associated with the risk of clinical progression during early disease (38). Long-term disability, nevertheless, better correlates with brain atrophy, reflecting neuro-axonal loss (39, 40). Atg7 deletion resulted in a microglial phenotype that associated with neuronal function and neurodegenerative diseases (41), suggesting sustained transcriptional changes that are critical for disease progression (Fig. 2, A and C, purple). A late-stage transcriptional profile of Atg7fl/fl Cx3cr1CreERT2 microglia demonstrated an enrichment in microglial genes associated with neurodegenerative diseases (41, 42) and MS-associated microglia identified in recent single-cell RNA sequencing studies (Fig. 3C, fig. S5, and table S3) (12, 43). We confirmed differences in protein levels of CLEC7A and CD11c/ITGAX (fig. S6A). In contrast, the homeostatic and tolerogenic state genes (4144) were depleted in Atg7fl/fl Cx3cr1CreERT2 microglia (Fig. 3C and fig. S5).

Microglia deficient in Atg7 have impaired signs of debris uptake and degradation

The Clec7a gene encodes the C-type lectin dectin-1, which is known to induce LC3-associated phagocytosis and recognizes ligands released upon CNS damage (4547). Clec7a expression characterizes a microglia population associated with neurodegenerative conditions (41). Atg7fl/fl Cx3cr1CreERT2 showed markedly increased number of CLEC7A+ IBA1+ microglia compared with control mice (Fig. 3D and fig. S6B).

We observed that Atg7 deletion did not affect the frequency of the CLEC7AINT population, but it did lead to a near loss of CLEC7ALOW and a robust increase in CLEC7AHIGH cells (Fig. 3, E and F). CLEC7AHIGH, and particularly CLEC7AINT, microglial populations displayed markedly lower surface expression of the scavenging receptors macrophage scavenger receptor 1 (MSR1, SR-A, CD204), (Fig. 3G), CD36, and other receptors (e.g., CD200R and IA/IE) that are implicated in microglial function during inflammation (fig. S6C) (4850). We further demonstrated that myelin phagocytosis relies, at least in part, on uptake mediated by MSR1, because the uptake of pHrodo-labeled myelin by microglia in vitro could be blocked using an MSR1 antibody (Fig. 3H), and myelin induced robust surface expression of MSR1 (Fig. 3I). Whereas RNA sequencing implicated an increased Msr1 expression in Atg7fl/fl Cx3cr1CreERT2 microglia (fig. S6D), flow cytometry analysis revealed reduced surface receptor quantity (fig. S6E). A similar pattern was observed for CD36 (fig. S6, D and F). However, a pool of intracellular MSR1 was detected upon permeabilization in Atg7fl/fl Cx3cr1CreERT2 microglia (Fig. 3J), suggesting a blocking of retrograde trafficking of phagosomal receptors that has been described in the context of disrupted noncanonical autophagy (16). To confirm this, we blocked the fusion of phagosomes and lysosomes in microglia ex vivo using bafilomycin A1, which resulted in an increase in intracellular MSR1 and CD36 levels (Fig. 3I and fig. S6G). These data indicate that stalling myelin-loaded autophagosomes can lead to MSR1 retention that could further sustain microglial failure to clear myelin debris, as observed after repeated myelin pulsing (Fig. 1J). This pattern of intracellular accumulation in relation to myelin exposure, and genotype was not detected upon targeting other potential myelin scavenger receptors such as CR3, CD16, CD64, or MARCO (fig. S6H).

Transcriptomic alterations of late-stage microglia further supported the notion of impaired phagosomal degradation in Atg7-deficient microglia. A marked up-regulation of apolipoprotein E (Apoe) was observed in late-stage Atg7fl/fl Cx3cr1CreERT2 microglia by RNA sequencing together with other lipoproteins (Apoc1 and Apoc4) and lipoprotein lipase (Lpl) (fig. S6I). Up-regulation of APOE occurs in response to the binding of myelin-derived cholesterol to the endogenous liver X receptor (LXR), which is transcriptionally activated in Atg7fl/fl Cx3cr1CreERT2 microglia (Fig. 2B) and is crucial for the export of cholesterol from microglia to the extracellular space (37). Moreover, we observed increased expression of Apoe when phagosome-lysosome fusion was impaired using bafilomycin A treatment (Fig. 3K), again suggesting that Apoe up-regulation in Atg7fl/fl Cx3cr1CreERT2 microglia constitutes a response to intracellular myelin accumulation. We also observed that Triggering receptor expressed on myeloid cells 2 (TREM2)—which can bind APOE and other lipids (51), is an inducer of the APOE pathway (41), and plays a major role in recovery from EAE (52)—was up-regulated during late EAE stages (fig. S6J) and that the pathway of TREM2 was highly activated (fig. S5). Last, levels of transforming growth factor–β1 (TGF-β1), a marker of homeostatic microglia as opposed to LXR-APOE-TREM2 disease-associated microglia (41, 53), was markedly decreased in Atg7fl/fl Cx3cr1CreERT2 microglia (Fig. 3L). Atg7fl/fl Cx3cr1CreERT2 microglia thus appear to cope with increase myelin load by engaging the LXR-APOE-TREM2 pathway and by up-regulating the expression of the MSR1 and other scavenging receptors to compensate for reduced retrograde transport and failed myelin degradation.

Accumulation of myelin debris is a strong inhibitor of oligodendrocyte differentiation and remyelination, is suggested to hinder recovery from inflammatory insults during MS and EAE, and leads to neuro-axonal loss (54, 55). Staining of Atg7fl/fl Cx3cr1CreERT2 microglia ex vivo with an antibody against degraded myelin basic protein (dMBP) revealed an accumulation of this protein in late-stage disease compared with microglia from control animals (Fig. 3M). Immunostaining of spinal cords further demonstrated the accumulation of dMBP in tissue, which overlapped with areas of activated microglia/infiltrating BMDMs (MAC3 bright) (Fig. 3N and fig. S7A). This was accompanied by unresolved inflammation evident as increased representation of immune cells during late-stage EAE (fig. S7, B and C). Moreover, pulsing late-stage Atg7fl/fl Cx3cr1CreERT2 microglia with myelin ex vivo led to lower uptake as compared with microglia from control animals (fig. S7D). We addressed whether the increased pool of tissue-infiltrating BMDMs (fig. S7C) could compensate for the reduced phagocytic capacity in Atg7fl/fl Cx3cr1CreERT2 mice, but BMDM cells underperformed microglia in phagocytosis of myelin as assessed by both ex vivo myelin load in cells isolated from late disease stage and by an in vitro myelin clearance assay (fig. S7, D and E). Last, flow cytometry assessment of oligodendrocytes isolated from the spinal cord at this time point confirmed a lower frequency of CD45GALC+MOG+ myelinating cells in Atg7fl/fl Cx3cr1CreERT2 animals as compared with controls (fig. S7F). Accordingly, we recorded reduced myelination in the Atg7fl/fl Cx3cr1CreERT late-stage EAE spinal cords (fig. S7G). Atg7-dependent impairment of microglia needed to clear myelin upon an inflammatory demyelinating insult could not be compensated for by infiltrating phagocytes and was associated with reduced CNS myelination during late-stage disease.

Aged microglia recapitulate the phenotype of young Atg7-deficient microglia

Because age is the strongest risk factor for progressive MS (5, 56) and many age-associated neurodegenerative pathologies are characterized by impaired autophagy (68, 41), we compared the impact of Atg7 deficiency in the context of aging. We observed a similarity between the transcriptomes of Atg7fl/fl Cx3cr1CreERT2 microglia and microglia from aged mice (>80 weeks) compared with control young microglia (Fig. 4A and table S4). Both late disease stage microglia from Atg7fl/fl Cx3cr1CreERT2 mice and microglia from aged (>80 weeks) wild-type mice revealed enrichment of genes associated with microglia during neurodegenerative diseases (41, 42), as well as MS-associated microglia (fig. S5) (12, 43). Accordingly, aged mice developed aggravated EAE with a clinical course similar to that of Atg7fl/fl Cx3cr1CreERT2 mice (Fig. 4B) and exhibited signs of accumulated myelin load (Fig. 4C). Similar to Atg7fl/fl Cx3cr1CreERT2, in vitro time-lapse imaging demonstrated that microglia from aged mice had reduced lysosomal loading of myelin as assessed by colocalization with low pH-sensing dye (pHrodo) and the pH-indifferent dye (PKH26), the latter detecting accumulation of myelin-containing phagosomes (movie S2).

Fig. 4 Trehalose treatment boosts myelin clearance and ameliorates EAE in aged mice.

(A) Transcriptome heat map of microglia sorted from naïve aged (>80 weeks) wild-type (Wt) mice (n = 3; average score, 2.5) and Atg7wt/wt (n = 4; average EAE score, 1.5) and Atg7fl/fl (n = 4; average EAE score, 2.5) mice 35 days p.i. analyzed by RNA sequencing (table S4). The scale represents Z score–transformed expression values (with red and blue indicating up-regulated and down-regulated genes, respectively, compared with the mean value of a gene from all samples). (B) Clinical course of EAE and full recovery rate in young (12 to 22 weeks old) Atg7wt/wt (control, n = 11; treated, n = 11), Ulk1fl/fl (control, n = 7; treated, n = 9), and Atg7fl/fl (control, n = 9; treated, n = 9) mice (left) and aged Atg7wt/wt (control, n = 4; treated, n = 6), Ulk1fl/fl (control, n = 4; treated, n = 5), and Atg7fl/fl (control, n = 5; treated, n = 5) mice (right). Mice were fed with trehalose in water or water as control. Clinical course was compared using one-way ANOVA with Tukey’s post hoc test on area under curve. (C) Flow cytometry quantification of intracellular myelin debris in microglia from young (n = 3), aged (n = 4), and trehalose-treated aged (n = 4) mice at day 21 p.i. (D) Immunofluorescence image and quantification showing TFEB translocation from cytosol to nucleus upon 48-hour in vitro trehalose–treated (n = 14) and untreated (n = 8) microglia from aged mice at day 5 p.i. LAMP1 detects lysosomal structures. Data pooled from two experiments. (E) Expression of selected key autophagosome, lysosome, and phagosome vesicle biogenesis genes in microglia of aged mice (n = 9) after 48 hours of ex vivo trehalose treatment as fold change of untreated control. (F) In vitro myelin clearance assay of microglia sorted 5 days p.i. from young (n = 3), aged (n = 13), and trehalose-treated young (n = 7) and aged (n = 5) mice pulsed four times with PKH26-labeled myelin. (G) In vitro pulsing of microglia 5 days p.i. with CellVue- and pHrodo-stained myelin, assessed by flow cytometry (all conditions, n = 4; except aged 3 hours, n = 5). (H) CLEC7A and APOE detected by Flow cytometry on ex vivo day 21 EAE microglia from aged (n = 4) with or without trehalose treatment (n = 4). (I) ELISA of cytokine production by CD11b+ cells isolated from the CNS of young (n = 5), aged (n = 13), and trehalose-treated aged (n = 10) mice at day 21 p.i. cultured ex vivo for 24 hours. Statistics: (D and F to H) Mann-Whitney U test and (C and I) Kruskal-Wallis test followed by Dunn’s post hoc test (***P < 0.001, **P < 0.01, and *P < 0.05). Error bars indicate SEM. Experiments (B to D and F to I) were performed twice.

Trehalose boosts autophagy in aged microglia and promotes recovery from EAE

Although increased understanding of the immune dysfunction during early relapsing-remitting phases of MS has led to the recent development of effective disease-modulatory treatments, progressive stages still largely lack treatment options. In accordance with our data, boosting microglial autophagy in neurodegenerative diseases has been proposed to have considerable therapeutic potential (5759). However, whether this could have beneficial effects in MS, and whether canonical or noncanonical autophagy is involved in disease progression, is currently unknown. To target phagocytosed myelin through autophagy, we administered trehalose, a disaccharide known to induce autophagy and ameliorate age-associated diseases (6065). Daily trehalose administration starting 1 week before EAE induction led to a reduction in the clinical severity of EAE, and more than 40% of aged control mice recovered as well as aged Ulk1fl/fl Cx3cr1CreERT2 mice that have compromised canonical autophagy (Figs. 1D and 4B and fig. S8A). We did not observe any trehalose effect on clinical EAE in younger mice, regardless of the genotype, nor in aged Atg7fl/fl Cx3cr1CreERT2 mice (Fig. 4B and fig. S8A), suggesting that trehalose acts by boosting noncanonical autophagy. We observed that trehalose increased nuclear density of transcription factor EB (TFEB), a key transcription factor for autophagy and lysosome-associated genes (60), and an increased formation of lysosomes defined by lysosomal-associated membrane protein 1 (LAMP1), a major lysosome membrane component (Fig. 4D). We also detected increased expression of multiple autophagy and lysosome genes upon trehalose treatment of aged microglia (Fig. 4E).

After trehalose treatment, we detected a reduction of intracellular microglial myelin load in aged EAE mice ex vivo (Fig. 4C) and increased myelin clearance and degradation through lysosomes in vitro (Fig. 4, F and G). Trehalose treatment also reduced the frequency of disease-associated microglia defined by expression of CLEC7A and APOE (Fig. 4H) and the infiltration of BMDMs (fig. S8B). Furthermore, microglia from trehalose-treated aged mice displayed reduced secretion of the disease-associated cytokines IFN-γ, TNF, and IL-1β and increased TGF-β1 secretion that reflected a normalization of the profile evident in microglia from young mice (Fig. 4I). Together, we validated an age effect on autophagy-associated vesicular biogenesis, with a decline in lysosome loading of phagocytosed myelin debris, similar to that in Atg7fl/fl Cx3cr1CreERT2 microglia. In aged mice, this impairment was successfully mitigated with trehalose treatment, with consequent effects on transcription, lysosome biogenesis, cytokine secretion, and clinical disease outcome.


We herein demonstrate that a noncanonical form of autophagy in microglia is responsible for myelin degradation and clearance and that impairment of this pathway, which occurs during aging, contributes to the progression of MS-like disease. We show that we can modulate this process therapeutically, with implications in other age-related neuroinflammatory disease.

The cellular events underlying inflammatory bouts typical of MS and EAE are well characterized, while the events promoting resolution of inflammation and limiting progression are much less understood (66). However, accumulation of myelin and inflammatory debris in the target tissue are known factors with inhibitory effects on remyelination (54, 55). We now demonstrate that the impaired myelin clearance capacity of microglia leads to increased tissue deposits of myelin debris accompanied by reduced myelination and oligodendrocyte differentiation. The deletion of Atg7 in microglia caused persistent neuroinflammation and, by comparing Atg7fl/fl Cx3cr1CreERT2 mice with Ulk1fl/fl Cx3cr1CreERT2, we demonstrated that the phenotype was largely independent from canonical autophagy. In several in vivo and in vitro experimental settings, we observed that the lack of Atg7 drives microglial dysfunction in clearance and processing of myelin debris and apoptotic CNS cells. The reduced clearance capacity of the microglia is most likely a consequence of internalization of scavenger receptors due to impaired Atg7-dependent lysosomal degradation.

Our model presented an opportunity to study impaired degradation of phagocytosed components as a regulator of microglial phenotype, which is relevant for a broad range of myeloid cell–associated pathologies. The elevated infiltration of peripheral immune cells 21 days after EAE induction was associated with an altered Atg7fl/fl Cx3cr1CreERT2 microglial cytokine profile, likely in synergy with increased myelin tissue deposits and local CNS expansion of immune cell populations. The infiltrating macrophage population did not compensate for the reduced myelin clearance in Atg7fl/fl Cx3cr1CreERT2 mice. The capacity for myelin clearance of this population at a late disease stage was markedly lower than that of microglia, which corroborates previous findings (13) and supports the established idea of microglia being promoters of homeostasis, in contrast to the monocyte-derived macrophages, which exhibit a more inflammatory phenotype (67).

Although the day 21 EAE microglia of Atg7fl/fl Cx3cr1CreERT2 mice reflect an acute inflammatory state, the day 35 microglia represent a more unique phenotype relevant for evaluating the challenges of chronic inflammation and tissue degeneration. The microglial transcriptome from persistent EAE at day 35 in Atg7fl/fl Cx3cr1CreERT2 mice was similar to other reported disease-associated microglial transcriptomes, indicating shared microglial pathology (41, 42). Potentially pathogenic microglia are dependent on the TREM2-APOE axis (41, 68). In our Atg7fl/fl Cx3cr1CreERT2 day 35 EAE microglia, we observed an enriched TREM2 pathway accompanied by elevated Apoe expression ex vivo and in vitro as a consequence of increased intracellular myelin load. Intracellular lipids are sensed by LXR, which acts as a transcription factor inducing Apoe expression. We propose an LXR-mediated pathogenic feed-forward mechanism through APOE in which intracellular myelin is not functionally degraded. The enriched LXR pathway in Atg7fl/fl Cx3cr1CreERT2 microglia resembled the macrophage phenotype characteristic of atherosclerosis, a disease state in which there is pathogenic accumulation of intracellular lipid compounds (63).

Phagocytosis of tissue debris has been reported to be essential for the maintenance of tissue homeostasis (69), and we demonstrate that upon inflammation the re-establishment of an anti-inflammatory response and subsequent tissue recovery are curbed when phagocytosis is decoupled from downstream cargo degradation through noncanonical autophagy. Expression of molecules associated with disease states (e.g., CLEC7A and CD11c) was increased in Atg7fl/fl Cx3cr1CreERT2 microglia, whereas the expression of genes that characterize homeostatic microglia was reduced (e.g., P2Ry12, CSF1R, CD200R, MSR1, and TGF-β1) (53). Among these, CLEC7A is of great interest because it is a strong inducer of LC3-associated phagocytosis (46, 47). Elevated expression of CLEC7A reported here for EAE and previously associated with pathology-associated microglia in other models (41) is therefore directly linked to an important functional outcome.

Increased age leads to a decline in autophagy and is a risk factor for progressive MS and neurodegeneration (56, 70). Dysfunctional aged myeloid cells have been reported in other settings to be a consequence of inefficient autophagy (71). We thus explored the potential of induced autophagy in ameliorating age-associated aggravated EAE disease. Treatment with the autophagy-inducing disaccharide trehalose led to a robust recovery rate and decline in clinical symptoms in aged mice, reminiscent of the recovery characteristic of untreated young wild-type mice (64, 72). However, trehalose treatment did not affect recovery of young wild-type mice or Atg7fl/fl Cx3cr1CreERT2 mice, suggesting that noncanonical autophagy in microglia from young mice is already at sufficient capacity and that the effect of trehalose treatment in aged mice is dependent on microglial ATG7, regardless of whether trehalose exerts its effect upstream or downstream of ATG7. The trehalose treatment induced vesicle biogenesis through transition of transcription factor TFEB to the nucleus, increased lysosome density, and myelin clearance and degradation. Trehalose treatment in vivo also reduced the density of CNS infiltrating BMDMs during EAE and a reduced pro-inflammatory cytokine profile evident in microglia from aged mice.

The tools for defining and studying microglia have developed vastly over the past years, allowing for more accurate ex vivo and in vitro experiments, which, however, have an impact on the microglial phenotype, especially regarding contextual and dynamic processes such as autophagy. Another challenge for future work is the characterization of autophagy-associated phagocytosis in humans, especially in disease context as in progressive human neuroinflammation. Additional work on molecules of the microglial autophagy–associated phagocytosis including RUBICON would also add support to our findings and could unveil interesting pharmacological targets. Last, the phenotypes observed on autophagy-deficient microglia could partially stem from secondary effects such as increased phagosome load, which also warrants further exploration.

Together, our findings demonstrate that degradation of inflammatory myelin debris by microglia is dependent on the noncanonical arm of autophagy, a function necessary for cell and CNS tissue homeostasis. We thus provide a functional link between age, autophagy, and myeloid dysfunction. We associate our phenotype with the newly described microglial transcriptomes primarily described in neurodegenerative diseases, suggesting a shared pathology and providing a functional characterization. Last, we propose this pathway to be as a promising treatment target for age-associated CNS pathology.


Study design

Previous work by our laboratory established a link between Atg7 and disease severity in EAE, an animal model for MS (23). In this study, we set out to understand whether this effect was intrinsic to the immune system by broadly targeting the deletion of Atg7 in mice to either T cells or myeloid cells by crossing Atg7-floxed mice to LckCre or Lyz2Cre expressing strains, respectively. Because the effect of the deletion was observed during the recovery phase of EAE, which pointed to effects within the CNS, we further restricted Atg7 deletion to microglia and CNS-resident macrophage populations by crossing Atg7-floxed mice to a Cx3cr1CreERT2 expressing strain (see the “Experimental subjects” section). Experiments were performed with littermate controls, using both males and females. Animals were randomized, and the majority of analysis was done in a blinded fashion. Sample sizes varied depending on the goal of each experiments (i.e., dissection at one or multiple time points, in vitro cultures, etc.) and expected effect sizes, and numbers of animals and statistical analysis methods are thus given in each figure for every experimental setup. No animals were excluded from analyses apart from two samples in RNA sequencing that did not have correct genotype. Catalog numbers and the description of different primers, antibodies, and kits used throughout the study can be found as an additional technical sheet (table S5).

Ethics statement

Animal experiments were approved and performed in accordance with the guidelines from the Swedish National Board for Laboratory Animals and the European Community Council Directive (86/609/EEC) under the ethical permits N284/07, N332/06, N338/09, N138/14, and 9328-2019, which were approved by the North Stockholm Animal Ethics Committee (Stockholms Norra djurförsöksetiska nämnd). Mice were tested according to a health-monitoring program at the National Veterinary Institute (Statens Veterinärmedicinska Anstalt, SVA) in Uppsala, Sweden.

Experimental subjects

Gene-deleted mice on the C57BL/6 background were generated by cross-breeding of Atg7fl/fl or Ulk1fl/fl to LckCre, Lyz2Cre, or Cx3cr1CreERT2 transgenic mice. All strains were purchased from the Jackson laboratory except Atg7fl/fl that was a gift from K. Blomgren. All experimental Cre mice had a hemizygote genotype. Cx3cr1CreERT2 mice were treated with 4 mg of TAM (Sigma-Aldrich), dissolved in corn oil, and administrated subcutaneously three times at 48-hour intervals. Experiments were initiated at earliest 4 weeks after the first TAM administration to allow for repopulation of peripheral bone marrow–derived Cx3cr1CreERT2 expressing cells such as monocytes, while the gene-deletion effect is preserved in the self-renewing CNS-resident microglial population. Although border associated macrophages in meninges and perivascular spaces also express Cx3cr1, self-renew, and are targeted by TAM administration, they have been shown not to contribute to T cell activation and CNS damage during EAE and are absent in the parenchyma where the processes of myelin uptake and degradation described here occur (73, 74). No marked influence from the Atg7fl/fl allele or Cre+/− toxicity was observed. In EAE experiments, 10- to 22-week-old littermate mice were used. Aged mice were >80 weeks old.

Induction and clinical evaluation of EAE

Recombinant mouse myelin oligodendrocyte glycoprotein (rmMOG) amino acids 1 to 125 from the N terminus was expressed in Escherichia coli and purified to homogeneity using chelate chromatography, as previously described (75, 76). The purified protein, dissolved in 6 M urea, was dialyzed against phosphate-buffered saline (PBS). For EAE induction, mice were immunized with a single subcutaneous injection at the dorsal tail base with 100 μl of inoculum containing rmMOG, 18 to 30 μg per mouse in saline solution emulsified in a 1:1 ratio with a complete Freund’s adjuvant (Chondrex) (100 μg of Mycobacterium tuberculosis per mouse), all under isoflurane (Baxter) anesthesia. In addition, all experimental animals received an intraperitoneal injection of 200 ng per mouse pertussis toxin (Calbiochem) at days 0 and 2 p.i.

The clinical score was graded as follows: 0, no clinical signs of EAE; 1, tail weakness or tail paralysis; 2, hind leg paraparesis or hemiparesis; 3, hind leg paralysis or hemiparalysis; 4, tetraplegia or moribund; and 5, death. EAE remission was calculated as number of mice with full recovery (score 0) divided by total EAE incidence within the subgroup.

Single-cell suspensions from CNS

CNS cells were extracted using the Neural Tissue Dissociation Kit T (Miltenyi Biotec). Mice were anesthetized with isoflurane and transcardially perfused with ice-cold PBS. Brains and spinal cords were mechanically minced and resuspended in enzyme mix according to the manufacturer’s protocol. The CNS homogenates were then passed through a 40-μm cell strainer and washed with PBS containing 5 mM EDTA. The pellet was resuspended in a 38% Percoll (Sigma-Aldrich) solution and centrifuged at 800g for 15 min (no brake). The myelin gradient layer was extracted, and cells were resuspended in PBS.

Cell cultures

Cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Sigma-Aldrich) conditioned with fetal bovine serum (FBS) 10% (v/v) (Sigma-Aldrich) and penicillin/streptomycin 1% (v/v) (Sigma-Aldrich) and M-CSF (20 ng/ml; R&D Systems). For the starvation-induced autophagy experiment, cells were kept in Earle’s balanced salt solution (Sigma-Aldrich) for 5 hours before analysis. Bafilomycin A1 (Sigma-Aldrich) was used in concentration of 1 μM.

Flow cytometry

CNS cells were analyzed at several time points from naïve to day 35 p.i. EAE. Single-cell suspensions were plated and stained with conjugated antibodies and LIVE/DEAD Fixable Near-IR Dead Cell Stain (Invitrogen, L34976). Intracellular/intranuclear staining was performed after permeabilization using a fixation/permeabilization kit (BD Biosciences/eBioscience). LC3 was detected using a digitonin kit causing mild permeabilization, leaving mainly membrane-bound LC3 in the cell for analysis. Mitochondrial membrane potential was quantified using the MitoTracker deep red probe. Cells were acquired using a Gallios flow cytometer (Beckman Coulter) and analyzed using Kaluza software (Beckman Coulter). All antibodies and reagents are specified in the enclosed technical data file.

Cell sorting

Cells from mouse CNS were sorted using a BD Influx cell sorter. Microglia were sorted as live CD11b+ CD45Intermediate(INT) Ly6G and/or enhanced yellow fluorescent protein positive (eYFP+) (fig. S9). BMDMs were sorted as live CD11b+ CD45High (Hi) Ly6G and/or eYFP. Neutrophils were sorted as live CD11b+Ly6G+ and/or eYFP. Cells for some in vitro experiments were sorted from naïve or day 5 p.i. mouse CNS using CD11b magnetic beads and columns [magnetic cell sorter (MACS); Miltenyi Biotec]. Purity was determined using flow cytometry to be >90% YFP+ cells. The purpose of sorting cells from day 5 p.i. was to acquire activated microglia without accumulated intracellular myelin phagosomes. Blood monocytes were sorted using Ly6C magnetic beads (MACS, Miltenyi Biotec).

Mouse cell RNA, cDNA preparation, and expression analysis

Cell pellets from sorted mouse CNS cells were lysed in RLT buffer and RNA extracted using an RNeasy mini kit (QIAGEN). Reverse transcription of total RNA was performed using random hexamer primers (Invitrogen) and SuperScript Reverse Transcriptase (Invitrogen). Complementary DNA (cDNA) was stored at −20°C until use. Quantitative polymerase chain reaction (qPCR) was performed in triplicates using a CFX38 Real-Time PCR Detection Systems with SYBR green as fluorophore (Bio-Rad). C(t) values with interduplicate differences more than one cycle were excluded. Target expression was calculated using the Bio-Rad CFX Manager V1.6. software. Hprt, B-actin, Gapdh, or the geometrical mean of two of those was used as housekeeping gene reference.

Next-generation sequencing

Sorted microglia were pooled 1:1 female and male. RNA was prepared using the RNeasy Mini Kit (QIAGEN) followed by quality control assessed with a Bioanalyzer 2100 (Agilent). All samples included had high-quality RNA (RNA integrity numbers 8.5 to 10). RNA was amplified with the SMARTer Stranded Total RNA-Seq Kit–Pico Input Mammalian (Clontech). Next-generation sequencing and generation of bioinformatic data were performed by the National Genomics Infrastructure (NGI) at the Science for Life Laboratory using a HiSeq 2500 System with a HiSeq Rapid SBS Kit v2 (Illumina). Data normalization and analysis of differential gene expression were performed using the DESeq2 R package with a negative binomial test (77). The false discovery rate–adjusted P value was estimated using the Benjamini-Hochberg correction (78). Data were further analyzed using IPA (QIAGEN) and gene set enrichment analysis (GSEA; Broad Institute). GSEA analysis performed with standard settings—classic scoring scheme for the enrichment score signal-to-noise metrics for the ranked gene list. Heat maps show in different figures show the range of expression values in red and blue (denoting high and low) calculated from normalized counts.

Myelin isolation and staining

Pure myelin was obtained using a protocol adapted from Norton and Poduslo (79). Briefly, myelin was isolated through mechanical homogenization of perfused brains in homogenization buffer with PBS containing 0.32 M sucrose. After two washes in homogenization buffer, an 0.85 M sucrose solution in PBS underlay was added to the CNS homogenate. The CNS gradient was centrifuged at 4500g for 50 min. The interphase containing myelin was then washed twice in water. The purified myelin was then incubated with pHrodo dye and/or CellVue Plum and/or Fluoromyelin (all from Thermo Fischer Scientific) and/or PKH26 (Sigma-Aldrich) in PBS/Hepes according to the manufacturer’s instructions, followed by washing.


CNS cells were sorted and plated into poly-l-lysine–coated plates and incubated for 36 hours before adding purified myelin for an additional 12 hours. After washing, cells were fixed with 4% paraformaldehyde and permeabilized with 0.2% Tween-20. Nonspecific binding was blocked by adding 10% bovine serum albumin (BSA) and serum from secondary antibody-producing species. Cells were then incubated overnight with primary antibodies diluted in PBS containing 1% BSA and 0.2% Tween-20. After washing, secondary antibodies diluted in host serum were added and incubated at 37°C for 1 hour. Last, DAPI solution (4′,6-diamidino-2-phenylindole, dihydrochloride; 0.2 μg/ml, BD Biosciences) was added to the wells for 3 min before final washing. Samples were analyzed using a Leica Confocal microscope and Leica LAS-X software. Cellprofiler (Broad Institute) software was used for quantitative analysis (analysis pipeline enclosed). Images were acquired using the same settings for all samples.

Phagocytosis assays

Phagocytic uptake and loading to low-pH lysosomes were quantified by flow cytometry analysis of ex vivo CNS cells or sorted in vitro myeloid populations. For apoptotic cell phagocytosis panels CNS cells were sorted using MACS kits for either CD171, O4, or GLAST (Miltenyi Biotec). Apoptosis was induced by radiation using a Precision X-rad (1 gray/min at 320 kV, 12.5 mA) twice with 8-hour incubation pause (80). pHrodo-labeled cells were pulsed with stained myelin or apoptotic cells for 30 min followed by washing and flow cytometry.

Incucyte time-lapse imaging

Sorted microglia were pulsed with purified myelin stained with pHrodo Green and PKH26. Incubation plates were immediately transferred to an incubator connected to an Incucyte ZOOM instrument in which wells were analyzed using 20× ocular magnification every hour for 16 to 28 hours, generating time-lapse movies. Green: excitation wavelength, 460 nm and emission wavelength, 524 nm. Red: excitation wavelength, 585 nm and emission wavelength, 635 nm.

LC3B-II detection and digitonin permeabilization protocol

Membrane-bound lipidated LC3 (II) was detected after mild digitonin permeabilization extracting cytosolic LC3 according to a previously described protocol (81). Cells seeded in 96-well plates were treated with the nonionic detergent digitonin (Sigma-Aldrich) at a concentration of 50 μg/ml in PBS for 5 min at room temperature. This permeabilizes the membrane for extracellular diffusion of nonbound LC3, whereas membrane-bound LC3 remains in the autophagosomal membrane. Cells were then fixed in 4% (w/v) paraformaldehyde/PBS for 10 min at room temperature. After two washes in PBS, cells were incubated with an anti-LC3B antibody for detecting membrane-bound LC3B.

Acid wash stripping of surface molecules

Analysis of receptor internalization was achieved using a protocol for acid-wash stripping of surface receptors. Cultured cells were detached with EDTA and incubated with PBS containing glycine (100 mM) and NaCl (150 mM) (pH 2.5) for 5 min on ice. Cells were then stained and analyzed by flow cytometry after fixation with or without permeabilization.

Myelin clearance assay

Sorted ex vivo microglia from immunized mice (day 5 or 21 p.i.) were seeded at 5 × 104 cells per well in 96-well plates and incubated for 12 hours before PKH26-conjugated myelin was added. At the indicated time points, supernatant was removed and analyzed using a SpectraMax 384 microplate reader for fluorescence at 560 nm. The remaining myelin concentration was determined in relation to a standard dilution series.

Microglia-CD4 T cell coculture

Sorted ex vivo microglia from immunized mice (day 5 p.i.) were seeded at 2 × 104 cells per well in 96-well plates coated with poly-l-lysine. After 24 hours, 2 × 104 MACS-sorted CD4+ T cells (Miltenyi Biotec) from EAE mice (day 21 p.i.) were added per well. After 36 hours, cells were analyzed by flow cytometry after intranuclear Ki67 labeling as a marker of proliferation.

Enzyme-linked immunosorbent assay

Sorted ex vivo microglia from naïve mice and mice during EAE were incubated in DMEM (Sigma-Aldrich), conditioned with FBS 10% (v/v) (Sigma-Aldrich) and penicillin/streptomycin 1% (v/v) for 24 hours in 96-well plates. Supernatants were collected and cytokine production was quantified using ready-set-go enzyme-linked immunosorbent assay (ELISA) kits (eBioscience, Invitrogen) and a SpectraMax 384 microplate reader plate reader according to the manufacturer’s instructions.

Trehalose treatment of EAE

For studies of the clinical effects of trehalose, mice were treated with water supplemented with either 5% (w/v) d-(+)-trehalose dihydrate or 5% (w/v) sucrose (both from Sigma-Aldrich) starting at the day of immunization. Sucrose was used as control given its similarity to trehalose since both are disaccharides, thus excluding elevated calorie availability as a determinant of EAE recovery. Intraperitoneal injections of 20% (w/v) trehalose, 20% (w/v) sucrose, 20% (w/v), or PBS supplemented the treatment every third day starting from EAE onset until end of experiment. During severe EAE, mice were fed with trehalose, sucrose, and water at a final concentration of 20% (w/v). The intraperitoneal doses were equal to previously reported clinical experiments, whereas the drinking water was enriched to 5% (w/v) trehalose [compared with 3% (w/v)] (61). This was a result of a titration experiment using 0, 3, and 5% (w/v) with a stronger impact evident with the higher dose.

Trehalose and TFEB in vitro assays

Ex vivo microglia were cultured in medium with or without d-(+)-trehalose dihydrate at a final concentration of 3% (w/v). Immunocytochemistry (ICC) experiments were performed as described above using antibodies against TFEB (diluted 1:1000; rabbit, Bethyl) and LAMP1 (diluted 1:1000; rat, Sigma-Aldrich), which labels lysosomes, and secondary antibodies Alexa Fluor 647 goat anti-rat (Thermo Fisher Scientific) and goat-anti rabbit Alexa fluor 546 (Thermo Fisher Scientific), respectively. TFEB and LAMP1+ lysosomes were quantified in relation to DAPI-defined nuclei (0.2 μg/ml) using CellProfiler software.

Histopathology and immunofluorescence

Histopathological and immunofluorescence analyses were performed on 3- to 5-μm-thick paraffin-embedded spinal cord cross sections. Luxol fast blue (Kluever; Sigma-Aldrich) was used to assess tissue demyelination. Quantitative evaluation of demyelination presented as the demyelination score was performed on an average of seven complete cross sections of the spinal cord per mouse, as previously described by Storch et al. (82). All images were captured using a Leica Polyvar 2 microscope.

For immunofluorescence analyses, the paraffin-embedded spinal cord cross sections were treated as previously described (83). After deparaffinization in xylol, sections were transferred to 90% (v/v) ethanol. Endogenous peroxidase was blocked by incubation in methanol with 0.02% H2O2 for 30 min at room temperature and rehydration to distilled water followed via a 90% (v/v), 70% (v/v), and 50% (v/v) ethanol series. Antigen retrieval was performed with Dako target retrieval solution (Dako) for 1 hour in a steamer device at 98°C. Sections were subsequently incubated in 10% fetal calf serum in PBS for 30 min at room temperature before incubation with the primary antibody on 4°C overnight. Primary antibodies used in costainings were Mac3 (diluted 1:200; rat, BD Biosciences), dMBP (diluted 1:200; rabbit, Millipore), CLEC7A (diluted 1:500; rabbit, Abcam), and YFP (diluted 1:200; chicken, Abcam), CD45 (diluted 1:200; rat, BD Biosciences), and CLEC7A (diluted 1:200; rabbit, Abcam). After washing in PBS, sections were incubated with a secondary antibody for 1 hour at room temperature. Secondary antibodies were used in the following combinations: Alexa Fluor 555 donkey anti-rat immunoglobulin G (IgG; Abcam), Alexa Fluor 488 donkey anti-rabbit IgG (Abcam), or rabbit anti-chicken IgY fluorescein isothiocyanate (Thermo Fisher Scientific), respectively. DAPI (0.2 μg/ml) was included in the last washing step to visualize the nuclei. All images were acquired using a Zeiss LSM700 confocal microscope and the ZEN 2009 software. Representative images shown are maximum intensity projections of 3-μm-thick z-stacks. Quantifications of the specific immunoreactivity was performed on five whole spinal cord cross sections per mouse using ImageJ64 based on the number of pixels above an estimated threshold.

Statistical analysis

GraphPad Prism 8 ( was used for all the statistical analysis. In graphs with several comparisons, a dotted line separates the datasets that were compared. All figure legends include information regarding statistical tests used and sample size.


Fig. S1. Recovery from EAE requires functional autophagy in the myeloid but not in the T cell compartment.

Fig. S2. Atg7 and Ulk1 deficiency in microglia affect noncanonical and canonical autophagy, respectively.

Fig. S3. Atg7 deficiency induces alterations in microglial transcriptome during EAE.

Fig. S4. Atg7 deficiency in microglia increases T cell proliferation and polarization to an inflammatory phenotype.

Fig. S5. Gene set enrichment analysis.

Fig. S6. Atg7-deficient microglia have impaired scavenger receptor recirculation associated with increased inflammation and a reduced myelinating oligodendrocyte population in EAE.

Fig. S7. Late-stage EAE is characterized by extensive tissue destruction and signs of increase in inflammation in mice with Atg7-deficient microglia.

Fig. S8. Trehalose boosts EAE recovery and decreases immune infiltration in aged mice.

Fig. S9. Gating strategy for defining cell populations by flow cytometry.

Table S1. RNA-sequencing data.

Table S2. IPA, ORA, and REViGO analysis.

Table S3. Genes shared among homeostatic or pathogenic gene sets.

Table S4. RNA sequencing data.

Table S5. Technical data file.

Table S6. Raw data.

Movie S1. Accumulation of phagocytosed myelin in Atg7fl/fl Cx3cr1CreERT2 microglia.

Movie S2. Accumulation of myelin-containing phagosomes in microglia of aged mice.


Acknowledgments: We thank M. Komatsu at the Tokyo Metropolitan Institute of Medical Science for providing the Atg7fl/fl mice. We acknowledge K. Blomgren at the University of Gothenburg for providing breeding couples of Atg7fl/fl mice. We thank the staff at the animal facility at the Karolinska University Hospital and particularly H. Kungsmark for animal care taking. We also thank A. van Vollenhoven for assisting with flow cytometry sorting. We acknowledge support from the Science for Life Laboratory, NGI. Funding: This work was supported by grants from the Swedish Research Council, the Swedish Brain Foundation, the Swedish Association for Persons with Neurological Disabilities, the Stockholm County Council (ALF project), AstraZeneca (AstraZeneca-Science for Life Laboratory collaboration), European Union Horizon 2020/European Research Council Consolidator Grant (Epi4MS), the Knut and Alice Wallenbergs Foundation, Margeretha af Ugglas Foundation, Alltid Litt Sterkere, Foundation of Swedish MS research, NEURO Sweden, and Karolinska Institutet. Author contributions: The study was conceived by R.B., A.O.G.-C., M.T.-H., T.O., and M.J. Most experiments were conducted by R.B. with assistance from A.O.G.-C., H.L., and E.N. M.T.-H. and R.P. assisted in EAE characterization. Histopathology and immunofluorescence of CNS were assessed by M.Z., E.N., S.A., and M.Z.A. RNA extraction, purification, and quality control were performed by R.B. and S.R. RNA sequencing data preparation and analysis was done by E.E. and functional annotation analysis by M.J. R.B., A.O.G.-C., and M.J. wrote the manuscript with input from other authors. Statistical analysis were done by R.B. and E.E. The project was supervised by A.O.G.-C., R.A.H., T.O., and M.J. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The RNA sequencing data has been deposited in the Gene Expression Omnibus (GEO) and the accession number is GSE154920. The mouse strains can be purchased at RikenBRC (B6.Cg-Atg7<tm1Tchi) and the Jackson laboratory [B6.129-Ulk1tm1Thsn/J and B6.129P2(Cg)-Cx3cr1tm1Litt/J, B6.129P2-Lyz2tm1(cre)Ifo/J, and B6.Cg-Tg(Lck-cre)548Jxm/J]. All reagents are listed as technical data with company details and order number (table S5).

Stay Connected to Science Immunology

Navigate This Article