Impaired macroautophagy in oligodendrocyte precursor cells suppresses neuronal plasticity via a senescence-associated signaling

Abstract

Aging drives cognitive decline in the adult brain with unclear mechanisms. Previously, oligodendrocyte precursor cells (OPCs), the source cells of myelin-forming cells in the central nervous system, have been linked to brain aging by their compromised differentiation and regeneration capability. Whether a myelination-independent function of OPCs is involved in brain aging remains unknown. In this study, we herein report a myelination-independent role of OPCs in exaggerating cognitive decline in the aging brain via suppressing neuronal plasticity. Our results demonstrate that macroautophagic flux declines in aged OPCs. Inactivation of autophagy promotes the senescence of OPCs, which activates C-C motif chemokine ligand 3 (CCL3)/CCL5–C-C motif chemokine receptor 5 signaling. Through this, autophagy-defective OPCs impair glutamatergic transmission, neuronal excitability, and long-term potentiation, exaggerating the cognitive decline in the aging brain. Our study demonstrates a myelination-independent role of OPCs in brain aging and identifies that a declined autophagy in OPCs is a pivotal factor in driving aging-associated cognitive decline.

Senescent OPCs, caused by a declined autophagy, suppress neuronal firing and excitatory transmission via CCL3/5-CCR5 signaling.

INTRODUCTION

Aging contributes to cognitive decline in the adult brain with unclear mechanisms. Cellular senescence is characterized by an irreversible cell cycle arrest and featured by the senescence-associated secretory phenotype (SASP). The latter contains pro-inflammatory cytokines, chemokines, growth factors, and proteases (1), through which senescent cells affect themselves and the neighboring cells via autocrine and paracrine mechanisms. Distinct types of neural cells such as neurons, astrocytes, microglia, and oligodendrocyte precursor cells (OPCs) have been shown to express senescent markers, and these senescent cells accumulate in the brains with aging and neurodegenerative diseases such as Alzheimer’s disease (AD) and Parkinson’s disease (PD). The selective elimination of these senescent cells via a genetic or pharmacological approach attenuates cognitive deficits in naturally aged mice and reduces the accumulation of hyperphosphorylation of tau and amyloid-β in AD transgenic mice (2–7). However, despite these beneficial effects, which types of cells are predominant players in driving brain aging and how these senescent cells contribute to aging-related cognitive decline remain unknown.

OPCs, also termed NG2 glia due to their expression of the proteoglycan NG2 (chondroitin sulfate proteoglycan 4), evenly distributed throughout the adult brain, are the primary proliferative cells in the adult central nervous system (CNS) (8, 9). One of the pivotal roles of OPCs is their capability to generate oligodendrocytes (OLs), which produce myelin, ensuring the fast and reliable conduction of action potentials (APs) and providing metabolic support to axons (10, 11). Myelin remodeling through oligodendrogenesis is an adaptive response to neuronal activity, termed adaptive myelination. This process is required to consolidate and retrieve distinct forms of memory (12–14). The compromised differentiation of OPCs with aging contributes to cognitive deficits (15–17). Apart from being the cellular source of myelin, recent studies point out a myelination-independent function of OPCs in maintaining adult brain networks. OPCs regulate cognitive behaviors via secreting soluble factors (18–23), phagocytotic remodeling synapses and axons (24, 25). It is worth noting that OPCs form bona fide synapses with glutamatergic and GABAergic neurons (26, 27). However, it remains unknown how OPCs regulate neuronal plasticity and whether the myelination-independent functions of OPCs are involved in aging-associated cognitive decline.

Macroautophagy (referred to as autophagy hereafter) is a conserved process where cargoes such as the misfolded proteins and the damaged organelles are engulfed by autophagosomes and degraded when autophagosomes are fused with lysosomes. Autophagy is closely linked to the aging of distinct organs and life span (28). Autophagic flux has also been detected in OPCs and OLs (29–32), where it regulates the survival and differentiation of fetal OPCs (32) and controls the number of OLs in the developing brains and the myelin turnover in the aged brain (29, 30). Atg7 is a pivotal gene of conjugation systems, which is necessary for autophagosome formation. A deficiency of Atg7 impairs the formation of autophagosomes and autophagic flux (33). Herein, we observe a declined autophagic flux in aged OPCs. We thus have used the NG2-Cre mouse line, NG2-CreERT2 mouse line, and the platelet-derived growth factor receptor-α (PDGFRα)–CreERT2 mouse line to recombine to inactivate autophagy in OPCs. We show that the inactivation of autophagy in adult OPCs exaggerates aging-associated cognitive decline. Inactivation of autophagy in adult OPCs impairs neuronal excitability and glutamatergic transmission via promoting senescence of OPCs, which enhances C-C motif chemokine ligand 3 (CCL3)/CCL5-CCR5 (C-C motif chemokine receptor 5) signaling. Therefore, the present study points out that the accumulation of senescent OPCs, caused by a declined autophagy, is a pivotal factor exaggerating aging-associated cognitive decline via suppressing neuronal plasticity.

RESULTS

Autophagic flux declines in OPCs in aged brain

Autophagy is closely linked to aging. Autophagy exhibits declined activity in distinct types of cells, such as neurons, astrocytes, and microglia, in aged brains (28, 34). To address whether autophagy declines in aged OPCs, we immunostained the brain slices for P62 and PDGFRα (a marker of OPCs) in 26-month-old C57BL/6 mice. P62, an autophagy receptor, is degraded by autophagy, the accumulation of which usually indicates an impaired autophagic flux (35). The intensity of P62 in OPCs was increased in the hippocampus and cortex of aged mice compared to that in young adult (referred to as 2 to 6 months old) mice (Fig. 1, A and B), indicating autophagic flux declines in OPCs with aging. Immunostaining analysis also observed an increased intensity of P62 in CC1⁺ OLs in the brain slices from aged mice (fig. S1), indicating that autophagic flux declines in OLs as well with aging.

Fig. 1. Autophagy deficiency results in a precocious aging-like morphology of OPCs.

(A) The coronal brain sections of the dorsal hippocampus and cortex of 3-month-old and 26-month-old C57BL/6 mice were immunostained for P62 and PDGFRα. Scale bars, 20 μm in images with lower magnification; 5 μm in images with higher magnification. (B) Intensity of P62⁺ immunoreactivity in OPCs. n = 150 to 158 cells. (C) The coronal brain sections of the hippocampus CA1 of 12-month-old cKO^NGCE and 3-, 12-, and 26-month-old control mice were immunostained for PDGFRα. Morphologies of PDGFRα⁺ OPCs. Scale bars, 10 μm. (D) Sholl analysis of OPCs. n = 108 (Ctrl-3 M), 127 (Ctrl-12 M), 122 (Ctrl-26 M), AND 70 (cKO^NGCE-12 M) cells. Means ± SEM. ***P < 0.001. Student’s t test (B). Two-way ANOVA (D).

Inactivation of autophagy causes OPCs to resemble aged-like morphology

Next, we performed Sholl analysis to assess the morphological changes of OPCs in the hippocampal CA1 during aging. In 12-month-old mice, OPCs exhibit slightly fewer branches and processes than those in 3-month-old mice. However, OPCs displayed remarkably increased branches and processes in 26-month-old mice compared to those in 3- and 12-month-old mice (Fig. 1, C and D). These results indicate that OPCs change their morphology with aging to harbor more branches and processes.

As autophagic flux declines in OPCs with aging, we further examined the effects of inactivation of autophagy on the morphology of OPCs during aging. We used tamoxifen (TAM)–inducible NG2-CreERT2 mouse line (NG2-CreERT2^+/wt: Atg7^fl/fl, cKO^NGCE) (11, 36, 37) to specifically inactivate autophagy in adult OPCs via homozygous deletion of Atg7. Middle-aged (referred to as 10 to 17 months old) cKO^NGCE mice, which had received TAM for five consecutive days at 3 months old (fig. S2A), which exhibited 90% recombination efficacy in OPCs (fig. S4, A to C), lost ATG7 expression in OPCs (fig. S2, B and C). They largely decreased ATG7 levels in tissue homogenates of white matter and gray matter (fig. S2, F and G), confirming an efficient deletion of Atg7 in OPCs. Consistent with these observations, OPCs in middle-aged cKO^NGCE mice exhibited inactive autophagy as evidenced by accumulation of P62 in these cells (fig. S2, D and E) and in tissues from the gray and white matter (fig. S2, F and H). Intriguingly, in 12-month-old cKO^NGCE mice, OPCs already displayed remarkably increased branches and processes, which is similar to OPCs in 26-month-old control mice (NG2-CreERT2^wt: Atg7^fl/fl), indicating that autophagy deficiency causes OPCs to resemble a precocious aging-like morphology (Fig. 1, C and D). These results suggest that the inactivation of autophagy may accelerate the aging of OPCs.

Inactivation of autophagy in OPCs impairs neuronal plasticity

The suppression of neuronal plasticity is closely linked to aging-related cognitive decline (38). We thus examined whether the inactivation of autophagy exaggerates these two processes in the aging brain. We used an NG2 promoter-driven Cre recombinase-expressing mouse line to specifically inactivate autophagy in fetal OPCs via homozygous deletion of Atg7 (NG2-Cre^+/wt: Atg7^fl/fl, cKO^NGC), where Atg7 was deleted from OPCs as early as these cells were born (fig. S2, I to O). Young adult cKO^NGC mice exhibited loss of ATG7 expression and inactivation of autophagy in OPCs (fig. S2, I to O). Because the hippocampus is a critical region for cognition, we assessed whether the inactivation of autophagy in OPCs affects the basal glutamatergic and GABAergic neurotransmission in this brain region. We recorded spontaneous excitatory postsynaptic currents (sEPSCs) and spontaneous inhibitory postsynaptic currents (sIPSCs) from hippocampal CA1 pyramidal neurons of middle-aged cKO^NGC mice (Fig. 2A). Compared to control mice (NG2-Cre^wt: Atg7^fl/fl), the frequency of sEPSCs was notably reduced in middle-aged cKO^NGC mice. On the other hand, there were no notable changes in the amplitude (Fig. 2, B to D). Neither the frequency nor the amplitude of sIPSCs showed a difference in cKO^NGC versus control mice (Fig. 2, E to G). These results indicate that inactivation of autophagy in fetal OPCs impairs glutamatergic transmission while preserving GABAergic neurotransmission. Moreover, the frequency of spontaneous action potentials (sAPs, Fig. 2, H to J), rather than evoked action potentials (eAPs, Fig. 2, K to M), was decreased in the hippocampal neurons of middle-aged cKO^NGC mice, indicating a decreased neuronal excitability.

Fig. 2. Inactivation of autophagy in OPCs impairs neuronal plasticity.

(A) Graphical description of electrophysiological recording protocol. Created in BioRender. Hong, C. (2025); https://BioRender.com/nrj8njb. (B to D) sEPSCs were recorded from the hippocampal CA1 pyramidal neurons. Example traces (B). Frequency (C) and amplitude (D). n = 28 to 35 brain slices from five to seven mice per genotype (C and D). (E to G) sIPSCs were recorded from the hippocampal CA1 pyramidal neurons. Example traces (E). Frequency (F) and amplitude (G). n = 13 to 17 brain slices from three to four mice per genotype (F and G). (H to J) sAP was recorded from the hippocampal CA1 pyramidal neurons. Example spikes (H). Frequency (I) and amplitude (J). n = 16 to 23 brain slices from four to five mice per genotype (I and J). (K to M) eAP was recorded from the hippocampal CA1 pyramidal neurons. Example spikes (K). Number (L) and amplitude (M). n = 17 to 30 brain slices from six mice per genotype (L and M). (N) Schematic description of the timeline of experimental procedures. (O to S) mEPSCs were recorded from the hippocampal CA1 pyramidal neurons. Example traces (O). Frequency (P) and amplitude (R). Cumulative probability of frequency (Q) and amplitude (S). n = 20 to 22 brain slices from three to four mice per genotype (P and R). (T to Y) LTP was recorded from the hippocampal CA1 pyramidal neurons. Time course of normalized fEPSP slope before and after induction of LTP by four trains of high-frequency stimulation (HFS). Representative traces (T). Mean normalized fEPSP slope (U). Peak amplitude 60 min post-HFS (V). I/O curve (W). PPR (X and Y). n = 6 to 9 brain slices (U and V) and 4 to 6 brain slices (Y) from three mice per genotype. Means ± SEM. *P < 0.05; **P < 0.01. n.s., nonsignificance. Student’s t test.

We further examined whether the inactivation of autophagy in adult OPCs affects glutamatergic transmission using cKO^NGCE mice. Compared to age- and sex-matched control mice (NG2-CreERT2^wt: Atg7^fl/fl), middle-aged cKO^NGCE mice exhibited a similar but more robust electrophysiological phenotype, reducing the frequency and the amplitude of sEPSCs (Fig. 3, A to C). NG2 is also expressed in perivascular pericytes. Quantitative analysis revealed that 6.4% of hippocampal pericytes underwent Cre-loxP–mediated genetic recombination in aged NG2-CreERT2^+/wt Ribotag-HA^fl/fl mice (fig. S4, A to C). To exclude the possibility that pericyte is involved in the electrophysiological phenotypes observed in cKO^NGCE mice, we confirmed these results by using the PDGFRα-CreERT2 mouse line (PDGFRα-CreERT2^+/wt: Atg7^fl/fl, cKO^PαCE), where few pericytes could be recombined in brain (39). The frequency and amplitude of sEPSCs (Fig. 4, I to K) were reduced in the middle-aged cKO^PαCE mice (fig. S2, P to T), consistent with the observation in cKO^NGCE mice. Moreover, the frequency of sAP (Fig. 4, L and M) was also reduced in the middle-aged cKO^PαCE mice, which is consistent with the observation in cKO^NGC mice. To further elucidate whether a pre- or postsynaptic mechanism is involved in the decreased glutamatergic transmission in cKO^NGCE mice, we examined the miniature EPSCs (mEPSCs). The results showed that compared to control mice, the frequency, but not the amplitude, of mEPSCs was decreased in middle-aged cKO^NGCE mice (Fig. 2, N to S), indicating that the inactivation of autophagy in adult OPCs impairs glutamatergic transmission in a presynaptic mechanism.

Fig. 3. Senolytic treatment rescues impaired neuronal plasticity and cognitive deficits.

(A to C) sEPSCs recorded from hippocampal CA1 pyramidal neurons. (A) Example traces. Frequency (B) and amplitude (C). n = 11 to 15 brain slices from three to four mice per genotype (B and C). (D to I) LTP was recorded from hippocampal CA1 pyramidal neurons. Time course of normalized fEPSP slope before and after LTP induction by four HFS trains. (D) Representative fEPSP traces. (E) Normalized fEPSP slope. (F) Peak amplitude 60 min post-HFS. (G) I/O curve. (H and I) PPR. n = 6 to 11 brain slices (E and F) and 5 brain slices (I) from four mice per genotype. (J and K) Analysis of intensity of synaptophysin. (J) The coronal sections of hippocampus CA3 were stained for SYN. Scale bars, 100 μm. (K) SYN⁺ mean fluorescence intensity (MFI). n = 3 mice per genotype (K). (L to N) TEM analysis of the density of synapses in the hippocampal CA1. (L) Representative images of synapses. Scale bars, 500 nm. (M) Density of synapses. (N) Numbers of docked vesicles per presynapse. n = 3 mice per genotype (M and N). (O to R) Preference index (O and Q) and difference index (P and R) in novel object recognition (O and P) and object location task (Q and R). n = 12 to 21 mice (O and P) and 11 to 21 mice (Q and R) per genotype. (S to V) MWM. Escape latency (S and U) and swimming velocity (T and V) in the training trials (S and T) and reversal learning trials (U and V). n = 9 to 19 mice (S and T) and 6 to 11 mice (U and V) per genotype. Means ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001. n.s., nonsignificance. One-way ANOVA (B, C, E, F, I, K, and M to R). Two-way ANOVA (S to V).

Fig. 4. Neutralization of up-regulated CCL3/5 rescues impaired neuronal plasticity.

(A to C) Cytokine array analysis of the potential SASP in corpus callosum. Heatmap of changed proteins across groups of mice (A). Relative protein levels of CCL3/CCL4 (B) and CCL5 (C). n = 4 from three mice per genotype (B and C). (D) Relative levels of Ccl3 mRNA in P16⁻ and P16⁺ OPCs of hippocampus by reanalyzing scRNA-seq data (7). (E to G) RNAscope ISH analysis of Ccl3 and Ccl5 mRNA levels in PDGFRα⁺ cell in hippocampus and corpus callosum. Representative images. Scale bars, 5 μm (E). Number of Ccl3⁺ (F) and Ccl5⁺ (G) puncta per PDGFRα⁺ cell. n = 3 mice per genotype (F and G). (H) Experimental paradigm for neutralization with anti-CCL3 and anti-CCL5 antibodies. The acutely isolated brain slices were incubated with anti-CCL3 and anti-CCL5 antibodies–containing ACSF for 20 min. (I to K) sEPSCs recorded from hippocampal CA1 pyramidal neurons. Example traces (I). Frequency (J) and amplitude (K). n = 14 to 18 brain slices from three to six mice per genotype (J and K). (L to N) sAP recorded from hippocampal CA1 pyramidal neurons. Example spikes (L). Firing frequency (M) and amplitude (N). n = 13 to 18 brain slices from four to seven mice per genotype (M and N). (O to Q) LTP was recorded from hippocampal CA1 pyramidal neurons. Time course of normalized fEPSP slope before and after LTP induction by four HFS trains. Representative traces (O). Normalized slope (P). Peak amplitude 60 min post-HFS (Q). n = 5 to 10 brain slices from three to four mice per genotype (P and Q). Means ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001. n.s., nonsignificance. Student’s t test (F and G). One-way ANOVA (B, C, and I to Q). Unpaired t test (M).

In the electrophysiological recording of hippocampal neurons, middle-aged cKO^NGCE mice showed suppressed long-term potentiation (LTP) in hippocampal CA1 stratum radiatum, as noted in the reduced field excitatory postsynaptic potential (fEPSP) slope and amplitude with a standard input-output (I/O) curve (Fig. 2, T to W). To investigate the impact of the inactivation of autophagy in OPCs on presynaptic properties, paired-pulse facilitation of synaptic currents was examined by calculating the ratio of the slope of the second excitatory postsynaptic potential (EPSP) to that of the first EPSP. The results demonstrated a notable decrease in paired-pulse facilitation in middle-aged cKO^NGCE mice, with an interstimulus interval (ISI) of 150 ms (Fig. 2, X and Y). This result indicates that inactivation of autophagy in adult OPCs decreases Ca²⁺-dependent release probability at presynaptic terminals. Middle-aged cKO^PαCE mice also exhibited a suppressed LTP (Figs. 4, O to Q, and 5, L to N). Altogether, these results indicate that the inactivation of autophagy in adult OPCs impairs neuronal plasticity, which could be linked to a presynaptic mechanism.

Fig. 5. Inhibition of CCR5 rescues impaired neuronal plasticity and cognitive deficits.

(A and B) Western blotting analysis of CCR5 levels in hippocampus and corpus callosum. n = 7 to 8 mice per genotype (B). (C and D) The coronal hippocampal CA1 was immunostained for CaMKII and CCR5. Scale bars, 20 μm (C). Intensity of CCR5. n = 3 mice per genotype (D). (E) Experimental paradigm for analysis of MVC effects on neuronal plasticity. (F to H) sEPSCs recorded from hippocampal CA1 pyramidal neurons. Representative traces (F). Frequency (G) and amplitude (H). n = 14 to 19 brain slices from four to five mice per genotype (G and H). (I to K) sAP recorded from hippocampal CA1 pyramidal neurons. Representative spikes (I). Frequency (J) and amplitude (K). n = 8 to 12 brain slices from three to four mice per genotype (J and K). (L to N) LTP was recorded from hippocampal CA1 pyramidal neurons. Representative traces (L). Normalized slope (M). Peak fEPSP amplitude 60 min post-HFS (N). n = 5 to 8 brain slices from three to five mice per genotype (M and N). (O) Experimental paradigm for analysis of MVC effects in vivo. (P and Q) The coronal hippocampus CA3 was stained for SYN. Scale bars, 100 μm (P). SYN⁺ MFI. n = 3 to 4 mice per genotype (Q). (R and S) Preference index (R) and difference index (S) in novel object recognition. (T and U) Preference index (T) and difference index (U) in object location task. (V and W) Escape latency (V) and swimming speed (W) in MWM. n = 7 to 10 mice (R and S), 6 mice (T and U), and 5 to 9 mice (V and W) per genotype. Means ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001. n.s., nonsignificance. Student’s t test (B). One-way ANOVA (D, F to N, and Q to U). Unpaired t test (J). Two-way ANOVA (V and W).

Inactivation of autophagy results in demyelination and neurodegeneration

One of the predominant functions of OPCs is to differentiate into OLs, which form the myelin sheath to wrap axons to ensure the conduction of APs. We then wondered whether autophagy deficiency in adult OPCs affects myelination. The g ratio in the hippocampal CA1 and corpus callosum of middle-aged cKO^NGCE mice was increased by about 2.6 and 1.5%, respectively, versus control mice (Fig. 6, A to C). To exclude the influence of axonal diameter on the g ratio, we further calculated the g ratio of the axons with distinct sizes in the hippocampal CA1 and corpus callosum (Fig. 6, E and F). The results demonstrated that the g ratio of the larger axons in mutant mice was increased by 5 and 7% in the hippocampal CA1 and corpus callosum, respectively (Fig. 6, G and H). Despite the low rate of myelinated fibers in the hippocampal CA1, respectively, which shows 2.7 and 4.8% in cKO^NGCE and control mice, respectively, the proportion of myelinated fibers in the hippocampal CA1 and corpus callosum of middle-aged cKO^NGCE mice was decreased by about 43.1 and 17.6%, respectively, versus control mice (Fig. 6, A and D). These data indicate that the inactivation of autophagy in adult OPCs results in the loss of myelin while having a minor effect on the thickness of the myelin sheath. Consistent with demyelination, inactivation of autophagy in adult OPCs exacerbates neurodegeneration, as evidenced by middle-aged cKO^NGCE mice exhibiting more degenerated axons, which are characterized by the dark axoplasm jammed with the mitochondria, lysosomes, or neurofilaments and the impaired or empty myelin sheath (40) in the hippocampal CA1 and corpus callosum under transmission electron microscopy (TEM) analysis (Fig. 6, I and J), loss of synapses (Fig. 3, L and M), and decreased synaptophysin (SYN) levels (Fig. 6, K and L), a marker of synapses.

Fig. 6. Inactivation of autophagy results in demyelination, neurodegeneration, and impaired oligodendrogenesis.

(A to H) TEM analysis of myelinated axons in the hippocampal CA1 and corpus callosum (CC) of 11.5-month-old cKO^NGCE mice and control mice. (A) Representative images of myelinated axons. Scale bars, 2 μm. (B, C, and E to H) Axonal diameters or g ratios of total fibers (B and C) and fibers with distinct axonal size in the CA1 (E and G) and CC (F and H). (D) Percentage of myelinated axons. The proportion of alteration among groups of mice was shown in the graph, where the data in control mice were normalized with 100% (C, D, G, and H). n = 127 to 849 (B and C), 5 to 223 (E to H) fibers, and 17 to 32 microscope fields (D) from three mice per genotype. (I and J) TEM analysis of the degenerated axons in the CA1 and CC of 15-month-old cKO^NGCE and control mice. Representative images. Scale bars, 2 μm (I). Densities of degenerated axons (J). n = 3 mice per genotype (J). (K and L) The CA3 of 17-month-old cKO^NGCE mice and control mice were immunostained for SYN. Representative images. Scale bars, 100 μm (K). SYN⁺ mean fluorescence intensity (MFI, L). n = 3 mice per genotype (L). (M) Experimental paradigm for EdU pulse-chase analysis; 9.5- to 11.5-month-old or 17-month-old cKO^NGCE and their control mice, administrated with TAM at 3 to 4 months of age or 1 month old, were injected intraperitoneally with EdU for six consecutive days and subjected to coimmunostaining for EdU and ASPA 7 or 4 weeks later. (N) Representative images of EdU⁺ASPA⁺ cells. Scale bars, 100 μm. (O) Densities of EdU⁺ASPA⁺ cells. n = 3 mice per genotype (O). Means ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001. n.s., nonsignificance. Student’s t test.

The proliferation of OPCs was decreased in the hippocampus of middle-aged cKO^NGCE mice (fig. S3, F and H), but this was not enough to cause a change in the total number of OPCs in this area (fig. S3, A and B). The number of OPCs decreased slightly in the corpus callosum of middle-aged cKO^NGCE mice, where OPCs are more enriched than gray matter (fig. S3, A and B). Decreased oligodendrogenesis within 4 to 7 weeks of 5-Ethynyl-2'-deoxyuridine (EdU) pulse-chase analysis was observed in middle-aged, but not in aged cKO^NGCE mice (Fig. 6, M to O). Despite the decreased oligodendrogenesis, middle-aged cKO^NGCE mice exhibited comparable ASPA⁺ myelinating OL numbers (fig. S3, C and D), suggesting that the decreased oligodendrogenesis is not sufficient to alter the number of myelinating OLs. During learning and memory formation, oligodendrogenesis is induced to produce new myelin. We thus used a learning paradigm in the Morris water maze (MWM) test to examine learning-induced oligodendrogenesis (fig. S3E) (14). Middle-aged cKO^NGCE mice displayed low and comparable learning-induced oligodendrogenesis to control mice during MWM training (fig. S3, G and I). Therefore, although the inactivation of autophagy in adult OPCs decreases the generation of OLs in a longer timescale, it is insufficient to affect the number of myelinating OLs. We then surmised that the loss of myelin might be attributed to the OLs differentiated from the recombined autophagy defective OPCs, as autophagy in OLs is required for myelin turnover (30).

Autophagy deficiency in adult OLs causes demyelination and neuronal hyperexcitability

We noticed that about 30% of OLs in the aged brains were differentiated from the recombined OPCs, where Cre-mediated recombination was induced in young adulthood (fig. S4, A to C). To further examine whether the suppressed neuronal plasticity is ascribed to the impaired autophagy in OLs, which is differentiated from the recombined OPCs in cKO^NGCE mice, PLP-CreERT2^+/wt: Atg7^fl/fl (cKO^PLCE) mice were injected with TAM at 3 months old and subjected to electrophysiology recording and TEM analysis at 12 to 15 months of age (fig. S4D). Middle-aged cKO^PLCE mice exhibited less myelinated axons in the gray and white matter (fig. S4, E and H). In contrast, their myelin thickness was comparable to that of the control mice (fig. S4, E to G). Of note, despite a similar reduction in the proportion of myelinated axons to that of cKO^NGCE mice (fig. S4, E and H; Fig. 6, A and D), middle-aged cKO^PLCE mice displayed an increased frequency of sAP (fig. S4, O to Q), while comparable sEPSCs and sIPSCs compared to control mice (fig. S4, I to N). These results indicate that the suppressed neuronal plasticity observed in cKO^NGCE and cKO^PαCE mice is unlikely ascribed to the inactivation of autophagy in recombined OPCs-generated OLs, which causes fewer myelinated axons.

Inactivation of autophagy in adult OPCs exacerbates aging-associated cognitive decline

We performed behavioral tests to assess the impact of impaired OPC-specific autophagy on cognition (Fig. 7A). In comparison to age- and sex-matched control mice, cKO^NGCE mice developed impaired learning and memory, as shown by the reduced difference index in both the novel object recognition test (Fig. 7C) and the object location test (Fig. 7E) at 12 months of age, and which was not yet observed at 3 months of age, i.e., 2 months after induction of gene deletion. The preference index of cKO^NGCE mice was comparable to that of control mice (Fig. 7, B and D) in both tests, excluding the above results ascribed to the different preferences of mice. In the MWM test, cKO^NGCE mice exhibited longer escape latency and longer swimming distances to locate the platform in middle age (Fig. 7, I and J) rather than in young adulthood (Fig. 7, F and G). At the same time, the swimming speed of cKO^NGCE mice was comparable to that of age-matched control mice (Fig. 7, H and K). These results indicate that the inactivation of autophagy in adult OPCs results in cognitive deficits until middle age. Middle-aged cKO^PαCE mice exhibited similar behavioral phenotypes to cKO^NGCE mice (Fig. 7, L to T). In summary, these results indicate that the inactivation of autophagy in adult OPCs exacerbates aging-associated cognitive decline.

Fig. 7. Autophagy deficiency in OPCs results in aging-dependent cognitive decline.

(A) Schematic description of the timeline of experimental procedures. TAM was administrated into either cKO^NGCE (B to K) or cKO^PαCE (L to T) and their control mice at P30 for five consecutive days. The mice were then subjected to behavioral tests at 3 to 5 months old and 12 to 16 months old. (B to K) Behavioral tests in cKO^NGCE and control mice. (B and C) Novel object recognition. Preference index (B) and difference index (C). n = 9 to 12 mice per genotype (B and C). (D and E) Object location task. Preference index (D) and difference index (E). n = 6 to 7 mice per genotype (D and E). (F to K) MWM. Escape latency (F and I), swimming distance (G and J), and swimming speed (H and K) in training trials. n = 8 to 9 mice (F to H) and 8 to 10 mice (I to K) per genotype. (L to T) Behavioral tests in 12- to 16-month-old cKO^PαCE and control mice. (L and M) Preference index (L) and difference index (M) in novel object recognition. n = 9 to 13 mice per genotype (L and M). (N and O) Preference index (N) and difference index (O) in object location task. n = 5 mice per genotype (N and O). (P to T) MWM. Escape latency (P), swimming distance (Q), and swimming speed (R) in training trials. Numbers of mice that swam across platform (S) and time that mice spent in the targeted place where the platform originally was (T) in probe trials. n = 8 to 10 mice per genotype (P to T). Means ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001. n.s., nonsignificance. Student’s t test (B to E, L to O, and S and T). Two-way ANOVA (F to K and P to R).

Inactivation of autophagy promotes the senescence of OPCs

Because the suppressed neuronal plasticity caused by the inactivation of autophagy in adult OPCs is not ascribed to the OLs generated from the recombined autophagy-defective OPCs, which cause fewer axons myelinated, we then wondered whether a myelination-independent mechanism is involved in the suppressed neuronal plasticity induced by autophagy-defective OPCs. Cellular senescence is a hallmark of aging (1, 7). OPCs exhibit the highest proportion of expressing senescent marker P16^INK4A in aged mouse brains (7). Given the close link between autophagy and cellular senescence in cancer cells (41–43), we examined whether the inactivation of autophagy promotes senescence of OPCs. We further tested whether the inactivation of autophagy could promote senescence of OPCs autonomously using the rat OLN-93 cell line (44). Knocking down of ATG7 with shRNA impaired autophagic flux in OLN-93 cells as evidenced by the up-regulation of P62, an autophagy receptor under autophagic degradation, and the down-regulation of LC3-II, the lipidated form of LC3-I, which is activated upon autophagy induction (Fig. 8, A to E). Knocking down of ATG7 increased the expression of P21 (a marker of senescent cells) in OLN-93 cells, which was further exaggerated by hydrogen peroxide (H₂O₂), an efficient inducer of cellular senescence (45), which induced the senescence of OLN-93 cells as well (Fig. 8, F to H, and fig. S7, F to H). The up-regulation of the P21 protein was prevented by cotreatment with senolytic drugs dasatinib (D) and quercetin (Q), which are capable of eliminating senescent cells by initiating the apoptosis of senescent cells (2, 7) (Fig. 8, F and G, and fig. S7, M and N). These results indicate that impairing autophagic flux promotes the senescence of OLN-93 cells in a cell-autonomous manner. Coimmunostaining of OPCs with several markers of senescent cells such as P21, P16^INK4A, and senescence-associated β-galactosidase (SA-β-gal) in brain sections revealed notable accumulation of senescent OPCs in both the hippocampus and corpus callosum of middle-aged cKO^NGCE and aged cKO^PαCE mice (Fig. 8, I, J, L, and M, and fig. S5, A to D). In contrast, the proportion of senescent microglia in the hippocampus and corpus callosum, as well as senescent OLs in the hippocampus of middle-aged cKO^NGCE mice, remained comparable to controls (fig. S5, E to H). A modest increase in senescent OLs was observed exclusively in the corpus callosum of cKO^NGCE mice (fig. S5, E and G), suggesting that inactivation of autophagy in adult OPCs does not notably exacerbate senescence of other glial populations, including microglia and OLs. Collectively, these results indicate that the autophagy inactivation drives cellular senescence of OPCs in aging brains.

Fig. 8. Inactivation of autophagy accelerates the senescence of OPCs.

(A to E) OLN-93 cells were transfected with ATG7 shRNA. (A to D) Western blotting analysis of levels of ATG7, P62, and LC3 protein. (E) qPCR analysis of atg7 mRNA levels. n = 3 (B and C), 4 (D), and 7 (E) independent biological replicates. (F to H) OLN-93 cells were transfected with ATG7 shRNA and treated with either H₂O₂ or H₂O₂ combined with D + Q. Western blotting analysis of levels of P21 (F and G). qPCR analysis of p21 mRNA levels (H). n = 5 to 6 (G) and 3 to 4 (H) independent biological replicates. (I and J) The coronal sections of hippocampus and corpus callosum of 17-month-old cKO^NGCE mice and control mice were immunostained for PDGFRα and P21. Representative images. Scale bars, 50 μm (I). Percentage of P21⁺PDGFRα⁺ cells. The proportion of alteration among groups of mice was shown in the graph, where the control mice was normalized with 100%. n = 3 mice per genotype (J). (K) Experimental paradigm for senolytic treatment. cKO^NGCE mice, injected intraperitoneally with TAM for five consecutive days at 3 months old, were administrated orally with D + Q for 3 months starting at 8.5 to 13.5 months old and then subjected to the following experiments. (L and M) The coronal sections of hippocampus and corpus callosum of 14-month-old cKO^NGCE mice, treated orally with either D + Q or vehicle, were immunostained for P16^INK4A and PDGFRα. Representative images. Scale bars, 50 μm (L). Percentage of P16⁺PDGFRα⁺ cells. The proportion of alteration among groups of mice was shown in the graph, where the control mice was normalized with 100%. n = 3 mice per genotype (M). Means ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001. n.s., nonsignificance. Student’s t test (B to E and J). One-way ANOVA (G, H, and M).

Senolytic treatment rescues impaired neuronal plasticity, neurodegeneration, and cognitive deficits

To analyze whether the inactivation of autophagy in adult OPCs suppresses synaptic plasticity via promoting senescence of OPCs, we orally administered 8.5- to 13.5-month-old cKO^NGCE mice with D + Q for 3 months following a previously described protocol (2, 7, 46) (Fig. 8K). The senolytic treatment attenuated the accumulation of senescent OPCs in middle-aged cKO^NGCE mice (Fig. 8, L, and M). In contrast, the senolytic treatment attenuated the impaired glutamatergic transmission (Fig. 3, A to C) and rescued the suppressed LTP (Fig. 3, D to G) and the decreased paired-pulse ratio (PPR) (Fig. 3, H and I) of middle-aged cKO^NGCE mice. Neurodegeneration, as reflected by the reduced intensity of SYN (Fig. 3, J and K) and the density of synapses and synaptic vesicles (Fig. 3, L to N) in middle-aged cKO^NGCE mice, was also rescued by senolytic administration (Fig. 3, J to N). These results indicate that the suppressed neuronal plasticity caused by the inactivation of autophagy in adult OPCs is ascribed to the accumulated senescent OPCs in the brains. Consistent with these observations, senolytic treatment rescued the cognitive deficits of middle-aged cKO^NGCE mice, as shown in the novel object recognition (Fig. 3, O and P), object location test (Fig. 3, Q and R), and MWM test (Fig. 3, S to V). Senolytic treatment failed to affect the number of OPCs (fig. S3, A and B), ASPA⁺ OLs (fig. S3, C and D), and the proliferation of OPCs (fig. S3, F and H) in middle-aged cKO^NGCE mice. These results indicate that the beneficial effects of senolytic treatment on cognition are linked to the rescue of neuronal plasticity.

Inactivation of autophagy impairs neuronal plasticity via up-regulating CCL3/5

One feature of senescent cells is a distinct secretion, SASP, which includes cytokines, growth factors, and enzymes. To address how senescent OPCs suppress neuronal plasticity in aged brains, we performed a cytokine array with lysates from the hippocampus and the corpus callosum to screen the SASP of autophagy-defective OPCs. The cytokines and growth factors, which were up-regulated in the brains of middle-aged cKO^NGCE mice, were selected as candidates for a senolytic rescue treatment (Fig. 4A and fig. S6, A to C). We screened the candidates possibly involved in neuronal plasticity and focused on CCL3, CCL4, and CCL5, which displayed consistent alteration in both the corpus callosum and the hippocampus across groups of mice (Fig. 4, A to C, and fig. S6, A to C).

To determine whether CCL3, CCL4, and CCL5 are components of the SASP of senescent OPCs, we set up a cellular model of senescent OPCs using OLN-93 cells treated transiently with H₂O₂, a well-characterized inducer of senescence that does not trigger cell death (45, 47, 48). Transient H₂O₂ exposure impaired autophagy in both OLN-93 cells and primary OPCs, as indicated by increased P62 levels (fig. S7, A, C and E) and decreased LC3-II expression (fig. S7, A, D and E). Senescence induction was confirmed by increased p21 mRNA (fig. S7H) and P21 protein levels (fig. S7, F and G), which was prevented by cotreatment with D + Q (fig. S7, M and N). Transient H₂O₂ treatment did not compromise cell viability (fig. S7I), ruling out an acute cytotoxicity. Concomitant with senescence induction, H₂O₂ treatment up-regulated Ccl3 and Ccl5 transcription (fig. S7, J and K), an effect that was dose-dependently suppressed by D + Q (fig. S7, O and P). In contrast, although Ccl4 transcription was also enhanced by H₂O₂ (fig. S7L), it remained unaffected by D + Q (fig. S7Q), suggesting senescence-independent regulatory mechanisms for CCL4. These findings demonstrate that CCL3 and CCL5, but not CCL4, are SASP components of senescent OLN-93 cells.

To further validate the link between autophagy dysfunction and SASP secretion, we analyzed ATG7-knockdown OLN-93 cells. As previously shown (Fig. 8, F to H), ATG7 knockdown promoted cellular senescence and notably increased Ccl3 and Ccl5 (but not Ccl4) transcription, with further amplification under H₂O₂ exposure (fig. S7, R to T). Notably, D + Q treatment reversed autophagy deficiency-driven Ccl3 and Ccl5 up-regulation under both basal and H₂O₂-treated conditions (fig. S7, R to T), confirming that autophagy impairment elevates CCL3 and CCL5 expression via senescence-dependent mechanisms. Pharmacological inhibition of autophagy using bafilomycin A1 (Baf-A1), which inhibits the fusion of autophagosomes to lysosomes (49, 50) or 3-methyladenine (3-MA), an autophagosome formation inhibitor (51, 52), recapitulated the senescence phenotype, marked by increased p21 (fig. S7U) and Ccl3 (fig. S7V) and Ccl5 (fig. S7W) expression. These results phenocopied the effects of ATG7 knockdown (Fig. 8, F to H, and fig. S7, R and S). Conversely, rapamycin, an autophagy inducer via suppressing mTOR (53), attenuated senescence and reduced Ccl3 and Ccl5 levels in WT OLN-93 cells, but failed to do so in ATG7 knockdown cells (fig. S7, X to Z), underscoring autophagy dependence of this mechanism. Collectively, these results indicate that ATG7 deficiency drives cellular senescence through an autophagy-dependent mechanism.

Consistent with these observations, the mRNA levels of Ccl3 and Ccl5 were increased in the brains of aged cKO^PαCE mice (fig. S6D), and the protein levels of CCL3 and CCL5 were increased in the brains of middle-aged cKO^NGCE mice, which was rescued by senolytic treatment (Fig. 4, A to C, and fig. S6, A to C). Reanalysis of the single-cell RNA sequencing (scRNA-seq) data obtained from aged mouse brains (7) indicates an up-regulated Ccl3 transcription in senescent OPCs (Fig. 4D). RNAscope in situ hybridization (ISH) analysis further confirmed that the mRNA levels of Ccl3 and Ccl5 were increased in autophagy-defective OPCs in the gray and white matter of middle-aged cKO^PαCE mice (Fig. 4, E to G). scRNA-seq of the hippocampus in aged cKO^PαCE mice revealed comparable Ccl3 and Ccl5 mRNA levels in microglia, astrocytes, and OLs relative to control mice (fig. S8). These findings suggest that other glial populations do not notably contribute to the elevated CCL3 and CCL5 levels in cKO^NGCE mice and cKO^PαCE mice. These results indicate that CCL3 and CCL5 are critical components of the SASP of senescent OPCs.

To further investigate whether the up-regulation of CCL3 and CCL5 impairs neuronal plasticity, we treated the brain slices isolated acutely from middle-aged cKO^PαCE mice with anti-CCL3 and anti-CCL5 antibodies for 20 min and performed electrophysiological recording (Fig. 4H). The short-term incubation allowed us to exclude the involvement of neurodegeneration and demyelination because it is impossible to induce sufficient neuroregeneration and remyelination in cKO^PαCE mice with such a short-term timescale of incubation of anti-CCL3 and anti-CCL5 antibodies. Thus, this strategy allowed us to examine the direct effects of anti-CCL3 and anti-CCL5 antibodies on neurotransmission and neuronal firing without the involvement of their possible effects on neuroregeneration and remyelination. The neutralization efficacy of anti-CCL3 and anti-CCL5 antibodies was validated by analyzing the activation of cAMP response element binding protein (CREB) and the phosphoinositide-3-kinase/protein kinase B (PI3K/AKT) pathway, the downstream signaling of these cytokines (54–56) (fig. S6, E to H). Compared to control mice, middle-aged cKO^PαCE mice exhibited a reduction in the frequency and amplitude of sEPSCs (Fig. 4, I to K), the firing frequency of sAP (Fig. 4, L and M), and LTP (Fig. 4, O to Q) in hippocampal slices. Neutralizing CCL3 and CCL5 with their corresponding antibodies rescued almost all the electrophysiological phenotypes mentioned above in middle-aged cKO^PαCE mice (Fig. 4, I to Q). These results indicate that the inactivation of autophagy in adult OPCs impairs glutamatergic neurotransmission, neuronal excitability, and LTP, which is linked to an up-regulation of CCL3 and CCL5.

Inhibition of CCR5 rescues impaired neuronal plasticity and cognitive deficits

CCL3 and CCL5 share a common receptor, CCR5, a pivotal suppressor of neuronal plasticity and closely associated with cognitive activity (56, 57). Thus, we tested whether activation of CCR5 underlies the suppressed neuronal plasticity caused by the inactivation of autophagy in OPCs. The levels of CCR5 were increased in the brains of middle-aged cKO^NGCE mice (Fig. 5, A and B). Immunofluorescence analysis indicates that CCR5 is predominantly expressed in glutamatergic neurons but not in GABAergic neurons (fig. S9C). Neither OPCs, OLs, nor OLN-93 cells express CCR5 (fig. S9, A, B, and D). CCR5 levels were increased in the glutamatergic neurons of middle-aged cKO^NGCE mice, which were prevented by senolytic treatment (Fig. 5, C and D, and fig. S9, E and F), indicating that the up-regulation of CCR5 in glutamatergic neurons is induced by cellular senescence.

We then test the role of CCR5 in neuronal plasticity by using maraviroc (MVC), a selective antagonist of CCR5 approved by the Food and Drug Administration for clinical therapy of HIV (58), to treat the brain slices acutely isolated from middle-aged cKO^PαCE mice for 20 min before electrophysiological recording (Fig. 5E). MVC treatment within such a short-term scale rescued all the electrophysiological phenotypes of middle-aged cKO^PαCE mice, including the impaired glutamatergic transmission (Fig. 5, F to H), the reduced neuronal excitability (Fig. 5, I to K), and the suppressed LTP (Fig. 5, L to N), indicating that the suppressed neuronal plasticity caused by the autophagy defective OPCs is ascribed to the up-regulation of CCR5.

To examine the long-term effect of MVC, middle-aged cKO^NGCE mice were injected intraperitoneally with MVC for 28 consecutive days and then subjected to immunohistochemistry analysis and behavioral tests (Fig. 5O). The results showed that chronic administration of MVC rescued the loss of synapses (Fig. 5, P and Q), the deficits in short-term memory as revealed in the novel object recognition test (Fig. 5, R and S) and the object location test (Fig. 5, T and U), and the defective spatial memory as shown in the MWM test (Fig. 5, V and W) in middle-aged cKO^NGCE mice. These results indicate that the inactivation of autophagy in adult OPCs impairs neuronal plasticity and cognitive function via the overactivation of CCR5.

Senescence OPCs-derived CCL3/5 suppresses neuronal plasticity in young mice

In addition to the senescent OPCs, CCL3 and CCL5 are expressed by other glial populations such as microglia and astrocytes (55, 59–61). To directly test whether senescent OPCs-derived CCL3 and CCL5 impair neuronal plasticity, we cultured primary OPCs with high purity (96.2% Olig2⁺; Fig. 9, B and C), minimally contaminated with microglia or astrocytes (fig. S10). These OPCs were treated transiently with H₂O₂ to induce senescence (Fig. 9A), which increased p21 mRNA levels (Fig. 9D), without compromising the cell viability (Fig. 9E). Enzyme-linked immunosorbent assay (ELISA) analysis confirmed elevated secretion of CCL3 and CCL5 into the culture medium by H₂O₂-treated senescent OPCs (Fig. 9, F and G). To assess functional consequences, acute brain slices from a 3-month-old C57BL/6 mouse were exposed to conditioned media from senescent primary OPCs (H₂O₂-CM) for 20 min (Fig. 9A), and H₂O₂-CM selectively reduced the frequency, but not the amplitude, of sEPSCs (Fig. 9, H to J) and sAP (Fig. 9, K to M), recapitulating the suppression of excitatory neuronal transmission and neuronal firing observed in vivo. Critically, preincubation of H₂O₂-CM with neutralizing anti-CCL3/5 antibodies or the CCR5 antagonist MVC for 20 min abolished these inhibitory effects (Fig. 9, H, I, K, and L). Therefore, these results indicate that CCL3 and CCL5 derived from senescent OPCs impair neuronal plasticity via CCR5. In summary, these data show that inactivation of autophagy in OPCs impairs neuronal plasticity and cognitive function via the senescence CCL3/5-CCR5 axis.

Fig. 9. Senescence OPC-derived CCL3/5 suppresses neuronal plasticity via CCR5.

(A) Experimental paradigm. The primary rat OPCs were transiently treated with H₂O₂ to induce senescence. The conditioned culture medium of H₂O₂-induced senescent OPCs (H₂O₂-CM) was collected and treated, and the brain slices were acutely isolated from 3-month-old C57BL/6 mice for 20 min. Anti-CCL3 (0.5 μg/ml) and anti-CCL5 (1.0 μg/ml) antibodies or MVC (15 nM) were added to the conditioned medium for a rescue analysis. The conditioned medium derived from the young OPCs treated with saline served as controls (Saline-CM). Created in BioRender. Hong, C. (2025); https://BioRender.com/8n4h1qw. (B and C) Primary rat OPCs were immunostained for Olig2. Scale bars, 50 μm (B). Percentage of Olig2⁺ cells among DAPI⁺ cells (C). n = 4 images (C). (D, F, and G) Primary rat OPCs were transiently treated with H₂O₂. qPCR analysis of p21 mRNA levels (D). ELISA analysis of CCL3 (F) levels and CCL5 (G). n = 4 to 5 independent biological replicates (D, F, and G). (E) Primary OPCs were treated with either H₂O₂ or 5% DMSO. CCK8 assay of cell viability. n = 3 independent biological replicates (E). (H to J) sEPSCs recorded from hippocampal CA1 pyramidal neurons. Representative traces (H). Frequency (I) and amplitude (J). n = 12 to 16 brain slices from three to four mice per genotype (I and J). (K to M) sAP recorded from hippocampal CA1 pyramidal neurons. Representative spikes (K). Firing frequency (L) and amplitude (M). n = 5 to 6 brain slices from three to four mice per genotype (L and M). Means ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001. n.s., nonsignificance. Student’s t test (D, F, and G). One-way ANOVA (E, I, J, L, and M).

DISCUSSION

We herein identify a mechanism underlying aging-associated cognitive decline. The declined autophagy in aged OPCs exaggerates aging-associated cognitive decline via suppressing neuronal plasticity. Autophagy declines with aging, which promotes the senescence of OPCs. The latter increases the secretion of CCL3 and CCL5, which activate CCR5 in glutamatergic neurons. In this way, the autophagy-defective OPCs impair glutamatergic transmission, neuronal excitability, and LTP, leading to cognitive deficits (Fig. 10). Thus, we demonstrate a myelination-independent role of OPCs in driving aging-associated cognitive decline.

Fig. 10. Inactivation of autophagy in OPCs suppresses neuronal plasticity via CCL3/5-CCR5 signaling.

Autophagy influx declines with aging, which promotes the senescence of OPCs. The latter increases the secretion of CCL3 and CCL5, which results in the overactivation of CCR5 in glutamatergic neurons. In this way, the accumulation of senescence OPCs impairs glutamatergic transmission, neuronal excitability, and LTP, leading to cognitive deficit. Created in BioRender. Hong, C. (2025); https://BioRender.com/8jutk8q.

Myelin breakdown, degeneration, and reduced myelin turnover have been observed in aged brains (5, 17, 62). OPCs, as the cellular source of myelin sheath-forming OLs, are linked to brain aging by their compromised capability in differentiation and regeneration (15, 16). In this study, we observed fewer myelination fibers and a slightly thinner myelin sheath in the brains of middle-aged cKO^NGCE mice, which may be ascribed to decreased oligodendrogenesis and the inactivation of autophagy in OLs, as about 30% of OLs were generated from the recombined OPCs in cKO^NGCE mice, and the less myelinated axons were observed in cKO^PLCE mice. A recent study indicates that OLs and myelination limit neuronal plasticity in the visual cortex (63). However, several lines of evidence suggest that it is unlikely that the suppressed neuronal plasticity is caused by the autophagy-deficient OPC-induced loss of myelin. First, CCR5, which, upon inhibition, rescued the impaired neuronal plasticity of middle-aged cKO^NGCE mice, is not expressed by OL lineage cells. Second, suppressing CCL3/5-CCR5 signaling within a short timescale, which is impossible to allow sufficient remyelination and regeneration, rescued the suppressed neuronal plasticity in the brain slices of middle-aged cKO^NGCE mice. Third, middle-aged cKO^PLCE mice, which exhibit less myelinated axons, display neuronal hyperexcitability, indicating that the suppressed neuronal plasticity is not ascribed to the autophagy defective OLs. Fourth and most important, the conditioned culture medium derived from senescent OPCs, rather than that from young OPCs, suppress neuronal plasticity via secreting CCL3/5. Hence, this study presents a myelination-independent role of OPCs in brain aging: aged OPCs impair neuronal plasticity via senescence-associated signaling. This is a complementary finding to understand how OPCs contribute to brain aging. In addition to compromising myelination, aged OPCs exaggerate cognitive declines in aging brains by suppressing neuronal plasticity as well. As the rejuvenation of OPCs to facilitate remyelination is taken as one of the current prime strategies in aging-related neurological disorders (15, 17, 64), the present study indicates that the intervention of senescent OPCs is an additional promising therapeutic strategy for aging-related cognitive deficits.

OPCs form bona fide synapses with glutamatergic and GABAergic neurons (26, 27). While prior studies have focused on how OPCs receive neuronal signals via these “neuron-OPC” synapses to regulate proliferation and differentiation (65–70), the mechanisms by which OPCs regulate neuronal plasticity remain poorly understood. Only a few studies have shed light on this context. OPCs regulate neuronal plasticity via shedding a cleaved fragment of NG2 protein (18), releasing GABA (22) and FGF2 (19) and up-regulating WIF1 (23). However, it remains unclear how OPCs selectively regulate the activity of distinct subtypes of neurons. Here, we demonstrate that the senescent OPCs suppress neuronal firing and excitatory neurotransmission via CCL3/5-CCR5 signaling. CCR5, a known suppressor of neuronal plasticity, inhibits neuronal excitability, glutamatergic neurotransmission, and LTP by inhibiting activation of CREB and MAPK, both of which play central roles in neuronal plasticity (56). CCR5 displays a delayed increased expression after the formation of a contextual memory, which determines the temporal window’s duration for associating or linking that memory with subsequent memories (71). We herein describe that CCR5 is selectively expressed in glutamatergic neurons, with negligible levels in GABAergic interneurons. Autophagy inactivation in OPCs elevates neuronal CCR5 expression in a senescence-dependent mechanism. CCL3 and CCL5, after binding to CCR5, impair autophagic flux via promoting mTOR signaling, which, in turn, causes CCR5 accumulation as the latter is a substrate of autophagy (55). We herein propose a similar mechanism underlying the autophagy-defective OPC-induced neuronal accumulation of CCR5. The CCL3 and CCL5, released from senescent OPCs, impair autophagic flux in neurons by activating mTOR signaling, thus causing the neuronal CCR5 accumulation. Consistent with this idea, neurons in cKO^NGCE mice exhibited impaired autophagic flux, evidenced by elevated P62 (fig. S9, G and H) and CCR5 (Fig. 5, C and D, and fig. S9, E and F) levels, a phenotype reversed by senolytic treatment. In contrast, OLs (fig. S9, I and J) and microglia (fig. S9, K and L) showed unaltered autophagic flux, highlighting the neuron specificity of this mechanism. Collectively, these data indicate that CCL3 and CCL5, up-regulated and secreted by senescent OPCs, promote CCR5 accumulation in glutamatergic neurons, thereby impairing their excitability and neurotransmission.

A discrepancy has been observed in the electrophysiology recording data obtained from distinct cre-mediated ATG7 mutant mice in this context. In contrast to cKO^NGCE and cKO^PαCE mice, where both the amplitude and frequency of sEPSCs were decreased compared to control mice, the frequency, but not the amplitude, of sEPSCs was decreased in middle-aged cKO^NGC mice. Several factors may contribute to this discrepancy, such as the recombination time. In contrast to the cKO^PαCE and cKO^NGCE mice, where autophagy is inactivated in adult OPCs, autophagy is inactivated in OPCs from the time of their generation in cKO^NGC mice. The inactivation of autophagy in early developmental OPCs may result in complementary alterations, which might cause cKO^NGC mice to show comparable sEPSCs amplitude to control mice. In addition, the NG2-Cre mouse line was reported to exhibit Cre activity in astrocytes (72), which may also contribute to the discrepancy observed in distinct mutant mice. Notably, treatment of the young adult brain slices with the conditioned medium derived from senescent OPCs partially, but not completely, phenocopies the electrophysiological properties of the Cre-mediated ATG7 mutant mice. The conditioned medium derived from senescent OPCs reduced the frequency, rather than the amplitude, of sEPSCs in the brain slices isolated acutely from young adult brains, possibly because neuronal CCR5 is not up-regulated in young adult brains. Despite these discrepancies, the present study presents that senescent OPCs, caused by a declined autophagy, suppress neuronal firing and glutamatergic transmission via CCL3/5-CCR5 signaling.

Another novelty of this study is our elucidation of a mechanism underlying the function of cellular senescence in brain aging. Although cellular senescence is one of the hallmarks of aging, and selective elimination of senescent cells extends health span while mitigating aging-related cognitive decline and neurodegeneration in naturally aged mice and neurodegenerative disease models (e.g., AD and PD) (2, 3, 5, 7, 73), the precise contribution of cellular senescence to brain aging and aging-related neurodegenerative disorders remains unclear. Senescent cells are heterogeneous and dynamic, as is their associated secretory phenotype. There is thus an urgent need to elucidate which cell types undergo senescence and the nature of the SASP in aging brains (1). Among neural cell populations, OPCs are particularly susceptible to senescence, exhibiting the highest proportion of cells expressing senescent marker P16^INK4A in aged mouse brains (7). OPCs serve as the source cells of myelinating OLs, and age-related declines in OPC differentiation capability impairs myelin renewal and contributes to brain aging (16). Here, we identify a myelination-independent role for senescence OPCs in driving aging-related cognitive decline. Specifically, senescent OPCs suppress neuronal plasticity via secretion of CCL3 and CCL5. We demonstrate that these cytokines, as components of SASP, are secreted by autophagy-defective OPCs. While CCL3 and CCL5 are also expressed by microglia and astrocytes (55, 59–61), our scRNA-seq analysis reveals that other glial populations, including microglia, astrocytes, and OLs, do not notably contribute to the elevated CCL3 and CCL5 levels in the brains of mutant mice with autophagy-defective OPCs. This aligns with the finding that autophagy-defective OPCs do not exacerbate senescence or autophagic dysfunction in neighboring glial cells, such as microglia and OLs. Together, these results suggest that CCL3 and CCL5 derived from autophagy-defective OPCs is sufficient to impair neuronal plasticity. This idea is further supported by coculture experiments: The conditioned medium from senescent OPCs rapidly reduces neuronal firing and synaptic transmission in acute brain slices from young adult mice within 20 min. Crucially, this effect is abolished by preincubating the medium with anti-CCL3 and -CCL5 neutralizing antibodies. To model cellular senescence, we used transient H₂O₂ treatment to induce cellular senescence of OPCs, a well-established approach in aging research that induces senescence without triggering cell death (47, 48, 74, 75). However, we acknowledge that transient H₂O₂ treatment–induced senescent OPCs cannot fully recapitulate the in vivo complexity of aged OPCs. While genetic manipulation of OPCs (e.g., viral knockdown of CCL3/CCL5 or selective senescent OPC deletion) would strengthen causal inferences, such approaches face technical hurdles in aged mice, such as low viral transduction efficiency (76). Future studies using genetic OPC manipulation via crossbreeding mutant mice will help address these limitations. Nevertheless, our multimodal evidence converges to demonstrate that senescent OPC-derived CCL3/CCL5 is sufficient to impair neuronal plasticity.

The rapid restoration of neuronal plasticity following anti-CCL3/CCL5 neutralizing antibody treatment strongly supports a direct, demyelination- and neurodegeneration-independent role of OPC-secreted CCL3/CCL5 in impairing neuronal plasticity, because the short timescale of rescue (20 min) is incompatible with neuroregeneration or remyelination processes. Nevertheless, neurodegeneration and demyelination observed in mutant mice with autophagy-defective OPCs raise critical unanswered questions. Specifically, it remains to be determined whether autophagy inactivation in OPCs drives neurodegeneration and demyelination via mechanisms overlapping with neuronal plasticity suppression (e.g., sustained CCL3/CCL5 signaling) or through distinct pathways, potentially involving cross-talk with other glial populations such as microglia.

Autophagy has been reported as a critical regulator of OPC survival, maturation, and proper myelination in the developing brain. Inactivation of autophagy in OPCs by using PDGFRα-CreERT2 to recombine and delete a loxp-flanked allele of Atg5 at P5 to P9 impairs OPC survival, oligodendrogenesis, and myelination, resulting in a prolonged tremor by P12 and an earlier lethal at P15 in mice (32). Consistent with these observations, we herein conclude that inactivation of autophagy in adult OPCs results in fewer myelinated axons, which may be ascribed to both the impaired oligodendrogenesis and the loss of myelin caused by the autophagy-deficient OLs produced by the recombined OPCs. Inactivation of autophagy in adult OPCs impairs oligodendrogenesis in middle-aged brains. In contrast, it fails to affect oligodendrogenesis in aged brains, possibly due to the ceiling effect caused by the declined autophagy in the aged OPCs. Of note, the reduced oligodendrogenesis was not robust enough to affect the number of ASPA⁺ myelinating OLs in cKO^NGCE mice, suggesting that the loss of myelin in these mutant mice may be caused by the autophagy-deficient OLs produced by the recombined OPCs. This idea is supported by the fact that autophagy is required for myelin turnover (77) and that middle-aged cKO^PLCE mice exhibited loss of myelin. In this context, it is worth noting that a previous study indicates that inactivation of autophagy in OLs with the CNPCre mouse line caused increased myelin thickness (30). The discrepancy may be caused by the distinct recombination time and efficacy or the distinct analysis time point between CNPCre- and PLPCreERT-mediated ATG7 mutant mice, as the myelin component protein, which was accumulated by autophagy deficiency, fails to be incorporated into the myelin sheath, eventually leading to myelin breakdown (77).

Autophagy is closely linked to cellular senescence with conflicting evidence. Autophagy is generally thought to be an antisenescence mechanism. Inhibition of autophagy causes senescence under basal conditions (78). However, autophagy also acts as a prosenescence mechanism, which is activated upon senescence induction, and its impairment paradoxically delays senescence. Such conflicting evidence indicates that autophagy regulates cellular senescence in a cell context–dependent manner, which may be attributed to the fact that distinct substrates are degraded by autophagy under different conditions (41–43). We herein observe an antisenescence mechanism of autophagy in aged OPCs. Autophagy regulates senescence by impairing protein turnover, mitochondrial homeostasis, oxidative stress, and maintaining SASP (41, 43, 79). Further investigation is needed to determine whether autophagy in OPCs shares the same mechanism to promote senescence as that in other types of cells. Considering autophagy declines with aging, our study brings a mechanism in aging-related cognitive decline. The declined autophagy causes senescence of OPCs, which impairs neuronal plasticity and exacerbates neurodegeneration via CCL3/5-CCR5 signaling.

MATERIALS AND METHODS

Mice

Atg7^floxp/floxp mice (Atg7^fl/fl, RIKEN, RBRC02759, RRID: IMSR_RBRC02759) (37) were crossed to the NG2-Cre line (Jax laboratory, 008533, RRID: IMSR_JAX:008533) to generate NG2-Cre^+/wt: Atg7^fl/fl (cKO^NGC) mice. Atg7^fl/fl mice were crossed to the NG2-CreERT2 line (36) to generate NG2-CreERT2^+/wt: Atg7^fl/fl (cKO^NGCE) mice. Atg7^fl/fl mice were crossed to the PDGFRα-CreERT2 line (Jax, 018280, RRID: IMSR_JAX:018280) (39) to generate PDGFRα-CreERT2^+/wt: Atg7^fl/fl (cKO^PαCE) mice. Atg7^fl/fl mice were crossed to the PLP-CreERT2 line (Jax, 005975, RRID: IMSR_JAX:005975) (80) to generate PLP-CreERT2^+/wt: Atg7^fl/fl (cKO^PLCE) mice. Ribotag-HA^fl/fl mice (Jax, 029977, RRID: IMSR_JAX:029977) (81) were crossed to the NG2-CreERT2 line to generate NG2-CreERT2^+/wt: Ribotag-HA^fl/fl mice. All mouse lines were maintained through crosses of Cre^wt/CreERT2^wt: Atg7^fl/fl and Cre^+/wt/CreERT2^+/wt: Atg7^fl/fl breeders, generating Cre^wt/CreERT2^wt: Atg7^fl/fl (control mice) and Cre^+/wt/CreERT2^+/wt: Atg7^fl/fl littermates in the C57BL/6 background. For inactivation of autophagy in adult OPCs or OLs, Cre-mediated recombination was induced via daily intraperitoneal injection of TAM (Sigma-Aldrich, T5648) dissolved in caprylic/capric glycerides to 1- or 3-month-old cKO^NGCE, cKO^PαCE, and cKO^PLCE mice for five consecutive days at 100 mg/kg. All mice were group housed with three to five same-sex cage mates in standard mouse cages in a pathogen-free barrier facility on a 12-hour light-dark cycle with lights on at 0700 and a controlled temperature range of 22° to 25°C. Food and water were provided ad libitum. Behavioral tests and electrophysiological recordings were performed during the light phase. All experimental procedures were preapproved by the Ethics Committee of Soochow University and conformed to the Institutional Animal Care and Use Committee Guidelines of Soochow University (reference number: 202207A0249 and SUDA20241121A07).

Antibodies

Rabbit anti-ATG7 (Sigma-Aldrich, A2856, RRID: AB_1078239), rabbit anti-P62/SQSTM1 (Sigma-Aldrich, P0067, RRID: AB_1841064), goat anti-PDGFRα (R&D Systems, AF1062, RRID: AB_2236897), mouse anti-APC (Merck Millipore, OP80, RRID: AB_2057371), rabbit anti-ASPA/Nur7 (Merck Millipore, ABN1698, RRID: AB_2827931), goat anti-Olig2 (R&D Systems, AF2418, RRID: AB_2157554), rat anti-PDGFRβ (Thermo Fisher Scientific, 14-1402-81, RRID: AB_467492), goat anti-IBa1 (Abcam, ab5076, RRID: AB_2224402), mouse anti-GFAP (CST, 3670S, RRID: AB_561049), rabbit anti-LC3 (Novus Biologicals, NB100-2220, RRID: AB_578334), rabbit anti-SYN (Abcam, ab32127, RRID: AB_2286949), rabbit anti-P16^INK4A (Abcam, ab211542, RRID: AB_2891084), rabbit anti-P21 (Abcam, ab188224, RRID: AB_2734729), rabbit anti-P21 (Abcam, ab109199, RRID: AB_10861551), mouse anti-NeuN (CST, 94403S, RRID: AB_2904530), mouse anti-CaMKII (CST, 50049S, RRID: AB_2721906), rabbit anti-GABA (Sigma-Aldrich, A2052, RRID: AB_477652), rabbit anti-CCR5 (Invitrogen, PA5-114965, RRID: AB_2899601), rabbit anti-CREB (CST, 9197S, RRID: AB_331277), rabbit anti-p-CREB (Ser133) (CST, 9198S, RRID: AB_2561044), rabbit anti-AKT (CST, 4691S, RRID: AB_915783), rabbit anti-p-AKT (Ser473) (CST, 4060S, RRID: AB_2315049), goat anti–CCL3/MIP-1 alpha (R&D Systems, AF-450-NA, RRID: AB_354492), goat anti-CCL5/RANTES (R&D Systems, AF478, RRID: AB_355385), normal goat immunoglobulin G (IgG) control (R&D Systems, AB-108-C, RRID: AB_354267), mouse anti-GAPDH (Proteintech, 60004-1-Ig, RRID: AB_2107436), mouse anti–γ-tubulin (Sigma-Aldrich, T6557, RRID: AB_477584), and corresponding secondary antibodies conjugated with horseradish peroxide (HRP; Sigma-Aldrich) or Alexa fluorophore 488, 555, or 647 (Invitrogen) were used.

Drug treatment

Dasatinib (LC Laboratories, LC-D-3307) and quercetin (Sigma-Aldrich, Q4951) were dissolved in a dissolving solution containing 30% polyethylene glycol, molecular weight 400 (PEG-400; Sigma-Aldrich, 8074851000), 10% anhydrous ethanol, and 2.5% dimethyl sulfoxide (DMSO) to a storage concentration of 1 mg/ml (D) and 10 mg/ml (Q). cKO^NGCE mice were injected intraperitoneally with TAM for five consecutive days at 3 months old and were administrated with D + Q at 8.5 to 13.5 months old. The mice were treated every 2 weeks with D + Q by oral gavage for three consecutive days in a dose of 5 mg/kg (D) and 50 mg/kg (Q) for a total period of 10 or 12 weeks (2, 7, 46). Age- and sex-matched cKO^NGCE and their control mice (NG2-CreERT2^wt: Atg7^fl/fl) were treated with the dissolving solution as vehicle control mice.

MVC (Selleck, S2003) was dissolved with DMSO into a 100 mg/ml concentration as the stock solution. The stock solution was diluted with the dissolving solution containing 40% PEG-300 (MCE, HY-Y0873), 2% Tween-80 (MCE, HY-Y1891), and saline to a 5 mg/ml concentration. To analyze the effect of MVC in vivo, 9- to 11.5-month-old cKO^NGCE mice were injected intraperitoneally with MVC in a dose of 50 mg/kg daily for 28 days as previously described (55). Sex- and age-matched 9- to 11.5-month-old cKO^NGCE mice and their control mice (NG2-CreERT2^wt: Atg7^fl/fl) were injected daily with the dissolving solution as vehicle control mice.

For electrophysiological recording, the brain slices of 13- to 16-month-old cKO^PαCE mice were incubated with 15 nM MVC or anti-CCL3 (0.5 μg/ml) and CCL5 (1.0 μg/ml) antibodies in artificial cerebrospinal fluid (ACSF) for 20 min. The brain slices of 13- to 16-month-old cKO^PαCE mice and their control mice were incubated with DMSO or standard IgG control in ACSF as vehicle control. After incubation, brain slices were immediately transferred to the recording chamber with ACSF, and then the sEPSC, sAP, and LTP were recorded.

Cell culture

Rat OLN-93 cells (44) (Biobw, Bio-105984, NCBI_Iran catalog no. C617, RRID: CVCL_5850) were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Gibco, 11960044) with high glucose and supplemented with 10% fetal bovine serum (FBS; Gibco, 12484028) and 1% penicillin-streptomycin (Gibco, 15070063) under a humid atmosphere with 5% CO₂ at 37°C. Lentiviruses encoding ATG7-shRNA were produced by GENECHEM (Shanghai, China). The sequences of ATG7-shRNA were as follows: CAGCCTGGCATTTGATAAA. OLN-93 cells were infected with lentiviruses encoding ATG7-shRNA for 4 or 5 days before being subjected to Western blotting and quantitative real-time polymerase chain reaction (qPCR) analysis. OLN-93 cells were treated with 500 μM hydrogen peroxide for 1 hour to induce senescence. OLN-93 cells were treated with 0.1 μM D and 20 μM Q for 48 hours to clear the senescent cells. OLN-93 cells were treated with 400 nM, 1 μM, or 5 μM 3-MA (Selleck, S2767) for 24 hours or 50, 100, or 200 nM Baf-A1 (Selleck, S1413) for 6 hours to inhibit autophagy, and treated with 200 nM rapamycin (Sigma-Aldrich, V900930) for 48 hours to induce autophagy before being subjected to qPCR analysis.

Primary culture of rat OPCs

OPCs were cultured as described previously (82). Neonatal male SD rats (P0) were euthanized by decapitation. The brains were dissected under a stereomicroscope, and the cerebral cortices were collected after the meninges were removed in ice-cold Hanks’ balanced salt solution (Sigma-Aldrich, H6648). The tissues were cut into small pieces, digested with 0.125% Trypsin (Gibco, C25200072), supplemented with 0.1 mg/ml DNase I (Sigma-Aldrich, D5025), and incubated for 15 min at 37°C. The digestion was stopped with DMEM/F12 (Gibco, 11320033) containing 10% FBS (Vivacell, C04001). Then, the tissue was blown into a single-cell suspension with a pipette, and the cells were collected by centrifuging at 1000 rpm for 5 min. The cells were then plated into a 75-cm² tissue culture flask precoated with poly-d-lysine (PDL; 0.1 mg/ml; Sigma-Aldrich, P6407). Cells were maintained in DMEM/F12 containing 10% FBS and incubated at 37°C with 5% CO₂. Half of the culture medium was changed every 3 days. On day 8, the flask was shaken on a shaker at 200 rpm for 2 hours at 37°C, and the culture medium was centrifuged at 1000 rpm for 5 min to collect microglia. Then, the fresh medium was added to the flask. For OPC purification, the flask was shaken at 250 rpm for 16 to 18 hours, and the culture medium was centrifuged at 1000 rpm for 5 min, after which the OPCs were then collected. OPCs were seeded onto a PDL-coated plate and cultured in DMEM/F12 containing 10% FBS. After 3 to 4 hours of culturing, the medium of OPCs was changed to DMEM/F12 supplemented with 2% B27 Supplement (Gibco, 17504044), 1% N2 Supplement (Gibco, 17502001), PDGF-AA (20 ng/ml; PeproTech, 100-13A-10) and basic fibroblast growth factor (bFGF; 10 ng/ml; PeproTech, 100-18B-10). The purification of cultured OPCs was confirmed by immunostaining.

Primary rat OPCs were treated transiently with 500 μM hydrogen peroxide (H₂O₂) for 1 hour to induce senescence. After 3 days, the culture medium from senescent OPCs (H₂O₂-CM) or young OPCs (Saline-CM) was collected.

For electrophysiological recording, the brain slices of 3-month-old C57BL/6 mice were incubated with Saline-CM, H₂O₂-CM, and H₂O₂-CM supplemented with anti-CCL3 (0.5 μg/ml) and CCL5 (1.0 μg/ml) antibodies or H₂O₂-CM supplemented with 15 nM MVC for 20 min. After incubation, brain slices were immediately transferred to the recording chamber with ACSF, and then the sEPSC and sAP were recorded.

Cell viability assay

OLN-93 cells and primary OPCs were treated with 500 μM H₂O₂ for 1 hour to induce senescence. Primary OPCs were additionally cotreated with 500 μM H₂O₂ for 1 hour with a senolytic cocktail (0.1 μM D and 20 μM Q) for 48 hours to eliminate the senescent cells. A positive control for cell death induction was included using 5% DMSO in both OLN-93 cells and primary OPCs. Cell viability was assessed by the cell counting kit (CCK8; Yeasen, 40203ES60) according to the manufacturer’s instructions. Absorbance at 450 nm was measured with a Microplate Reader (INFINITE M PLEX, TECAN, Switzerland).

Transmission electron microscopy

TEM was performed in conjunction with the Electron Microscope Facility at Soochow University. Briefly, mice were anesthetized with isoflurane and transcardially perfused with ice-cold phosphate-buffered saline (PBS) followed by 2% paraformaldehyde (PFA) and 3% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4) (fixing solution). Brains were then immersed in the fixing solution at 4°C overnight. Corpus callosum, hippocampus CA1, and CA3 were dissected into small blocks (1 mm³) and postfixed in the fixing solution at 4°C overnight. Tissues were washed twice with 0.1 M phosphoric acid for 15 min and then postfixed in 1% OsO₄ in a cacodylate buffer for 1 hour at room temperature (RT). After dehydration with serials of diluted acetone, tissues were embedded in Lx-112 (Ladd Research Industries). After polymerization, ultrathin sections 40 to 60 nm thick were prepared and stained with 3% uranyl acetate and lead citrate. Image acquisition was performed with a transmission electron microscope (Tecnai G2 Spirit BioTwin, USA). All image analyses were conducted blinded to the experiment condition. g ratios were quantified from 97 to 856 myelinated fibers from three mice per group using ImageJ software (NIH, Bethesda, MA, USA). They were calculated as axonal diameter/total diameter of myelinated fiber (the axon plus the myelin sheath). Images were taken randomly under lower magnification to measure the percent myelinated axons in the hippocampus and corpus callosum. The percentage of myelinated fibers in one microscope field was calculated as one sample. Degenerated axons were identified by their lack of myelin sheath and distortion or swelling of the axoplasm.

RNAscope ISH combined with immunofluorescence

Briefly, mice were transcardially perfused with PBS, followed by 4% PFA. Brains were removed, frozen with dry ice and isopentane, and stored at −80°C overnight. Brains were embedded in the optimal cutting temperature compound (OCT) medium. Cryosections 20 μm thick were prepared in a microtome cryostat (Leica, CM1950) and mounted on Superfrost Plus slides (CITOGLAS, 188105W). Slides were postfixed with 4% PFA at 4°C for 15 min and then dehydrated in 50, 70, and 100% ethanol sequentially. RNAscope ISH combined with immunofluorescence staining was performed using the RNAscope Multiplex Fluorescent Reagent Kit V2 (ACD, 323100) and RNA-Protein Co-Detection Ancillary Kit (ACD, 323180) as instructed by the manufacturer. Briefly, slides were incubated with hydrogen peroxide at RT for 10 min and then washed with PBST (PBS containing 0.1% Triton X-100). Slides were incubated with Co-Detection Target Retrieval Reagent at 99°C for 5 min and then quickly transferred to distilled water. Slides were incubated with anti-PDGFRα antibody (diluted in Co-Detection Antibody Diluent, 1:50) at 4°C overnight and then washed with PBST. Slides were then incubated with 4% PFA at RT for 30 min, followed by a PBST wash. Ccl3 (RNAscope Probe-Mm-Ccl3-C2, ACD, 319471-C2) and Ccl5 probe (RNAscope Probe-Mm-Ccl5-C1, ACD, 469601) were hybridized in situ at 40°C for 2 hours and then washed with washing buffer. Slides were incubated with RNAscope Multiplex FL v2 AMP1 (40°C, 30 min), AMP2 (40°C, 30 min), and AMP3 (40°C, 15 min) sequentially. Slides were incubated with RNAscope Multiplex FL v2 HRP-C1 (40°C, 15 min), TSA Vivid Fluorophore 570 (ACD, 323272, 1:750) (40°C, 30 min) and RNAscope Multiplex FL v2 HRP blocker (40°C, 15 min) sequentially, and then followed by HRP-C2 (40°C, 15 min), TSA Vivid Fluorophore 520 (ACD, 323271, 1:750) (40°C, 30 min), and HRP blocker (40°C, 15 min). After being incubated with the final HRP blocker, the slides were incubated with a secondary antibody (diluted in Co-Detection Antibody Diluent, 1:200) for 1 hour at RT and then washed with PBST. The slides were mounted with mounting medium with 4′,6-diamidino-2-phenylindole (DAPI; SouthernBiotech, 0100-20). Images were captured under a ZEISS confocal microscope LSM700 or LSM900 and analyzed with ImageJ software. Ccl3 and Ccl5 mRNA intensity was analyzed by calculating the number of Ccl3 and Ccl5 mRNA puncta per OPC as described previously (83).

Analysis of proliferation and differentiation of OPCs with EdU pulse-chase assay

To analyze the differentiation of OPCs in spontaneous conditions, 9.5- to 11.5- and 17-month-old cKO^NGCE and their control mice were injected intraperitoneally with EdU (Invitrogen, A10044) once daily in a 100 mg/kg dose for six consecutive days. Mice were euthanized 7 or 4 weeks later and subjected to coimmunostaining for EdU and ASPA. To access adaptive adult oligodendrogenesis and OPC proliferation, mice were injected intraperitoneally with EdU once daily in a 100 mg/kg dose for six consecutive days during a learning paradigm in WMW. Mice were euthanized 24 hours later and subjected to coimmunostaining for EdU and either PDGFRα or ASPA. EdU was detected using the Click-iT EdU imaging kit (Invitrogen, C10086) following the manufacturer’s instructions. Images were captured under a ZEISS confocal microscope LSM700 or LSM900. Densities of EdU⁺PDGFRα⁺ and EdU⁺ASPA⁺ cells in the hippocampus were analyzed with ImageJ software and expressed as EdU⁺PDGFRα⁺ and EdU⁺ASPA⁺ cells per mm² by dividing the numbers of marker⁺ cells by the areas of the hippocampus.

SA-β-gal staining

Brain sections were stained for SA-β-gal using the CellEvent Senescence Green Detection Kit (Invitrogen, C10851) following the manufacturer’s instructions. The sections were subsequently subjected to immunofluorescence staining for PDGFRα. Images were captured under a ZEISS confocal microscope LSM700 or LSM900.

Immunofluorescence staining and image analysis

Detailed information related to immunofluorescence staining was provided in the Supplementary Materials. Images were captured under a ZEISS confocal microscope LSM700 or LSM900 with identical imaging parameters. Numbers of OPCs, OLs, senescent OPCs, and EdU-labeled OPCs and OLs in the corresponding regions were counted, and the areas of corresponding regions were measured by ImageJ software as described previously (84). The density of marker⁺ cells per square millimeter was calculated by dividing the cell numbers by the area of the corresponding region. To analyze the distribution of CCR5 or HA, numbers of cell-type marker⁺CCR5⁺ cells or marker⁺HA⁺ cells and cell-type marker⁺ cells in the random microscope field were counted. The percentage of cell-type marker⁺-CCR5⁺ cells or marker⁺HA⁺ cells among cell-type marker⁺ cells was quantified. For analysis of P62 or SYN intensity, the brain slices were immunostained under the same experimental conditions, with identical dilution ratios and incubation conditions for both primary and secondary antibodies. All comparison images were acquired with identical laser intensity, exposure time, and filter. The brightness/contrast adjustments were applied equally to all comparison images using a ZEISS confocal microscope LSM900 (P62: oil immersion 40× objective; SYN: 20× objective). The intensity of P62 was quantified using ImageJ software in the identical parameters as described previously (85). Briefly, images were converted to 8-bit, and the image was changed from color to grayscale. Then, the thresholds of all images were adjusted and set correctly to ensure that the background signal was not detected, while the signal was not lost. Notably, the thresholds in all images were set to similar levels (P62: 30, 230; SYN: 30, 245). Then, we can obtain the intensity data per image. The investigators were blinded to experimental conditions during image acquisition and quantification.

Sholl analysis and image acquisition

Detailed information related to immunofluorescence staining was provided in the Supplementary Materials. Z-stack images (7-μm depth, 1-μm steps, ×40 magnification) were captured under a ZEISS confocal microscope LSM700 with identical imaging parameters. Sholl analysis was performed by ImageJ software as described previously (86). Briefly, images were converted to 8-bit and binary images sequentially with the background removed, and individual OPCs were analyzed. Regions of interest were determined as the soma of an OPC, and concentric circles were drawn around the soma in 1-μm steps. The number of intersections per radium was plotted to obtain Sholl analysis plots for each OPC.

Representative images of OPCs were built by IMARIS software (Version 9.91, Oxford Instruments, UK) as described previously (87). The OPC surface was reconstructed with the following parameters: surface detail, 0.700 μm (smooth); thresholding background subtraction (local contrast); and diameter of the largest sphere, 2.00. The surface reconstruction was then used as a template for filament reconstruction with the following custom settings: detect new starting points: largest diameter, 7.00 μm, seed points, 0.300 μm; remove seed points around starting points: diameter of sphere regions, 15 μm. Image processing, three-dimensional reconstruction, and data analysis regarding the experimental conditions were performed in a blind manner.

Cytokine expression analysis

For the cytokine array, the Proteome Profiler Mouse XL Cytokine Array and Mouse XL Cytokine Array kit (R&D Systems, ARY028) were used as instructed by the manufacturer. The hippocampus and corpus callosum were dissected and homogenized in PBS containing protease inhibitor cocktail (Roche, 11697498001) and 1% Triton X-100. The samples were frozen below −70°C and then centrifuged at 10,000g for 5 min to remove cellular debris after thawing. Protein concentrations were measured by the Pierce BCA Protein Assay kits (Thermo Fisher, 23227). Protein lysis (200 μg) was used for each membrane. The intensity of each spot on the membrane was quantified using ImageJ software (NIH, Bethesda, MA, USA), corrected for background intensity, and normalized to the membrane’s positive control.

Single-cell RNA sequencing

scRNA-seq was performed at Genergy BIO-Technology (Shanghai, China). The workflow included sample preparation, library construction, sequencing, and data analysis.

Sample preparation

Hippocampal tissues were dissected from 25-month-old cKO^PαCE mice and age-matched control mice (n = 5). Tissues from each group were pooled into a single sample. Single-cell suspensions were generated using enzymatic dissociation reagents (e.g., tissue dissociation reagents), and cell viability was confirmed to exceed 95% before processing.

Library construction

Libraries were prepared using a 10× Chromium device.

Sequencing

Libraries were sequenced on an Illumina Nova Seq6000 system with PE150 mode, with sequencing runs over 2 days.

Data analysis

Raw sequencing data were processed using Cell Ranger to remove low-quality reads and contaminating sequences. Cells with mitochondrial gene content >10% and unique feature counts >7000 or <200 were excluded. Downstream data analysis was performed using the Seurat software package, including normalization, dimensionality reduction, and clustering. Differential expression analysis of Ccl3 and Ccl5 in microglia, astrocytes, and OLs was conducted using the Seurat software package. Violin plots were generated with the normalized data and plotted on a log scale.

Behavioral tests

cKO^NGC, cKO^NGCE, and cKO^PαCE mice and their corresponding control littermates matched in sex and age were evaluated for each behavioral test. The experimenter was blinded to the genotype during testing. All behavioral tests were performed in a dimly lit room without noise interference. All tests were recorded using Any-maze 7.2 (Stoelting, CO, USA). The experimental apparatus used in every test was cleaned with 30% alcohol after each tested subject.

Novel object recognition test

Mice were allowed to habituate in an apparatus (40 cm by 40 cm by 40 cm) for 10 min per day for three consecutive days. On the fourth day, the mice were exposed to two identical objects placed 10 cm from the side walls in two opposite corners of the apparatus for 10 min. Ninety minutes after the training session, the mice were allowed to explore the open field for 10 min in the presence of one familiar and novel object. The time that mice spent sniffing or chewing the object was recorded as exploring activities. The preference index is defined as time exploring one of the identical objects/time exploring the identical object pairs. The difference index is defined as time exploring the novel object minus time exploring the familiar object/time exploring the novel object and time exploring the familiar object.

Object location test

Mice were allowed to explore freely for 5 min to enable habituation in an apparatus (40 cm by 40 cm by 40 cm) with intramural cues on the walls. Following 15 to 30 min, mice were returned to the device and exposed to two identical objects placed equidistant from the side walls for 10 min (training). Twenty-four hours after training, mice were allowed to explore the apparatus for 5 min, where one of the objects was moved to a novel position. The preference index was defined as time exploring one of the objects/time exploring both objects in a training session. The difference index was defined as time exploring the object in a novel position minus time exploring the original position/time exploring both objects in a novel and original positions.

MWM test

Mice were subjected to MWM with a circular pool of 160 cm in diameter. Mice had received four trials per day. A different starting position was used for each trial. The duration of each trial was 90 s. Escape latency (time spent swimming from the start point to the platform), swimming distance (the distance from the start point to the platform), and swimming speed were recorded on five to seven consecutive days. The platform was removed for the probe trials on days 6 to 8, and the mice were allowed to probe. The number of mice that swam across the spot where the platform was initially located and the time that mice swam across the place where the platform was originally located were recorded. The reversal test was performed 2 weeks after the probe trials. In the reversal test, the platform was removed to the opposite quadrant. Mice received four trials per day for 2 or 3 days. Escape latency, swimming distance, and swimming speed were collected.

Electrophysiology and recording

The experimental procedures were conducted as described previously (88). Briefly, mice were deeply anesthetized with inhalational isoflurane. Brains were extracted and sectioned into 300-μm-thick slices using a vibratome (Leica VT1200S, Leica Biosystems, Germany). The slices were then transferred to a holding chamber filled with oxygenated (95% O₂/5% CO₂) ACSF containing the following components: 125 mM NaCl, 2.5 mM KCl, 25 mM NaHCO₃, 1.25 mM NaH₂PO₄, 1.25 mM MgCl₂, 2.5 mM CaCl₂, 10 mM glucose, and 1 mM ascorbic acid (pH 7.4 to 7.45, 305 to 315 mosmol). The hippocampal slices were allowed to equilibrate at RT in ACSF for at least 1 hour before being transferred to a submersion-recording chamber.

Electrophysiological recordings were performed using a glass pipette filled with 3 M NaCl solution (resistance 2 to 5 megohms) inserted into the stratum radiatum of the CA1 region. Stimulation was delivered via bipolar nichrome electrodes placed in the CA3 region. Baseline values were recorded at a frequency of 0.033 Hz. LTP was induced using high-frequency stimulation (four 100-Hz trains, each lasting 1 s, delivered 20 s apart). I/O curves were generated by single-pulse stimulation of the Schaffer collateral region, allowing evaluation of synaptic efficacy by adjusting the stimulus intensity in steps of 0.05 to 1.0 mA. Stimulus pulses were delivered at 0.033 Hz, and five responses at each current intensity were averaged. PPR was measured using a paradigm with ISIs ranging from 25, 50, 75, 100, 125, 150, and 200 ms. Facilitation was quantified as the ratio of the second pulse-evoked EPSP slope to the average of the first pulse-evoked EPSP, with five responses per pulse pair.

sAPs were recorded in current-clamp mode using a gap-free recording lasting at least 1 min. The protocol for eAPs included a series of 200-ms current pulses with intensities ranging from 20 to 200 pA, incremented by 20 pA. The pipette solution for recording APs contained the following components: 140 mM KCl, 10 mM Hepes, 10 mM EGTA, 0.3 mM Na₃-GTP, and 4 mM Mg-ATP (pH 7.4, 290 mosmol).

sEPSCs were recorded at a holding potential of −80 mV, while sIPSCs were recorded at 10 mV. The pipette solution for recording sEPSCs and sIPSCs contained the following components: 120 mM CsCl, 2 mM MgCl₂, 1 mM CaCl₂, 30 mM Hepes, 11 mM EGTA, 2 mM Mg-ATP, and 0.3 mM Na₃-GTP (pH 7.4, 290 mosmol). Only data with stable (<20% variation), low series resistances (<20 megohms), and normal launch were included in the analysis. sEPSC and sIPSC were detected automatically using template search methods. For mEPSCs, 1 μM tetrodotoxin was bath applied for >5 min before recording spontaneous events. Electrophysiological data were acquired using an Axon multiclamp 700 B amplifier, filtered at 0.1 to 5 kHz, and digitized at 10 kHz.

Immunofluorescence staining

Mice were anesthetized with isoflurane and then transcardially perfused with PBS and 2% PFA. Brains were removed and postfixed in 2% PFA for 4 hours at 4°C, followed by dehydration with 15 and 30% sucrose. Brains were embedded in an OCT embedding medium and sectioned into 20-μm-thick slices with Microtome Cryostat (Leica, CM1950). Brain slices were washed with PBS and then blocked with 10% donkey serum containing 0.5% Triton X-100 at RT for 1 hour. Slices were incubated with the primary antibody overnight at 4°C and the corresponding secondary antibody at RT for 1 hour following a standard protocol. Slides were mounted with a mounting medium containing DAPI (SouthernBiotech, 0100-20).

Western blotting analysis

Brain homogenates were extracted with brain lysis buffer (10 mM of tris-HCl, pH 9.0, 150 mM of NaCl, 1% sodium deoxycholate, 0.5% Triton X-100, 0.5% SDS, and 2 mM EDTA) supplemented with protease inhibitor cocktail (Roche, 11697498001). For cultured cells, the cells were lysed with radioimmunoprecipitation assay lysis buffer (Beyotime, P0013D) containing protease inhibitor cocktail for 30 min on ice. Samples were centrifuged at 14,000g for 15 min at 4°C, and supernatants were collected. Protein extracts were subjected to SDS–polyacrylamide gel electrophoresis and transferred to the polyvinylidene fluoride (PVDF) membrane. Following a standard protocol, the PVDF membrane (Merck Millipore, IPVH00010) was incubated with a primary antibody and the corresponding HRP-conjugated secondary antibody. The protein signals were detected by ECL (Merck Millipore, WBKLS). The images were captured with an E-Gel Imager (E-Blot, China). The intensity of each protein signal on the membrane was measured as grayscale values using ImageJ software (NIH, Bethesda, MA, USA), corrected for background intensity, and normalized to the internal control.

Quantitative real-time polymerase chain reaction

According to the manufacturer’s instruction, total RNA was isolated from frozen dissected brains and cultured cells using TRIzol Reagent (Ambion, 15596026). Following isolation, RNA to be used in qPCR was quantified by NanoDrop (Thermo Fisher Scientific, USA) and converted to cDNA using a reverse transcriptase kit (Yeasen, 1141ES10) as instructed by the manufacturer. Real-time PCR was performed in a 7500 Fast Real-Time PCR System (Applied Biosystems, UK) using SYBR qPCR Master Mix (Vazyme, Q711-02). qPCR samples were analyzed using the standard ΔΔCT method. mRNA levels of tested genes were normalized to averages of gapdh. All primers were synthesized at GENEWIZ (Suzhou, China). Primer sequences were described as the following: Rat Ccl3 forward: TGCTGCTTCTCCTATGGACG, reverse: TCTTGGTCAGGAAAATGACACC; Rat Ccl4 forward: CAAACCTCTCCCCGAGCAAC, reverse: GGAGGGTCAGAGCCTATTGGTG; Rat Ccl5 forward: GCTGCTTTGCCTACCTCTCC, reverse: TCCTTCGAGTGACAAAGACGAC; Rat p21 forward: GATCCACAGCGATATCGAGA, reverse: AATCTGTTAGGCTGGTCTGC; Rat Atg7 forward: CCTGTCAGCCTGGCATTTGAT, reverse: GTCACTCATGTCCCAGATCTCAGC; Rat Gapdh forward: GGTGCTGAGTATGTCGTGGA, reverse: GCCATGCCAGTGAGCTTCCC; Mouse Ccl3 forward: TTCTCTGTACCATGACACTCTGC, reverse: CGTGGAATCTTCCGGCTGTAG; Mouse Ccl4 forward: CTTCTGTGCTCCAGGGTTCTC, reverse: CTGCCTCTTTTGGTCAGGAATACCA; Mouse Ccl5 forward: ATATGGCTCGGACACCACTC, reverse: TTCGAGTGACAAACACGACTG; Mouse Gapdh forward: GAAGGTCGGTGTGAACGGAT, reverse: AATCTCCACTTTGCCACTGC.

Statistical analysis

All data analyses were performed using GraphPad Prism 8.0 to produce graph values. All graph values are presented as mean ± SEM. All statistical analyses were performed using SPSS 26.0 or GraphPad Prism 8.0. The number of mice included in each experiment was based on standards established in the literature rather than being predetermined by statistical methods. Tests for normality and equal variances were used to determine the appropriate statistical test to use. Data were analyzed using Student’s t test (a comparison of the difference between the two groups), one-way analysis of variance (ANOVA; a comparison of the difference between multiple groups) followed by LSD or Dunnett T3 post hoc tests, and two-way ANOVA (a comparison of the difference of escape latency, swimming distance, and swimming speed during the MWM) followed by LSD post hoc tests and unpaired t test. Statistical details of experiments such as statistical tests, statistical values, and the information related to n can be found in the figure legends. Significance in differences was accepted at P < 0.05 (*P < 0.05; **P < 0.01; ***P < 0.001).

Supplementary Materials

The PDF file includes:

Figs. S1 to S12

Legends for data S1 and S2

Other Supplementary Material for this manuscript includes the following:

Data S1 and S2