Abstract

   Monoclonal antibody immunotherapy targeting the clearance of amyloid-β
   (Aβ) has shown promise in Alzheimer’s disease (AD). However, current
   antibody treatments trigger Fc receptors and induce proinflammatory
   responses, in turn exacerbating neuronal damage. Here, we report a
   synthetic efferocytic receptor (SER) integrating Aβ-targeting scFv,
   efferocytosis receptor backbone based on TIM4 and downstream signal for
   microglia (MG) reprogramming, which enabled selective elimination of Aβ
   without inducing an inflammatory response. Specifically, our
   in-house–customized MG-editing mRNA lipid nanoparticles (MERLINs)
   efficiently introduced SER mRNA into MG to generate Aβ-specific SER-MG
   in situ. SER-MG exhibited robust Aβ-specific phagocytosis and
   stimulated anti-inflammatory efferocytosis typical signaling in vitro.
   In a mouse model of AD, SER expression in the MG markedly increased the
   clearance of Aβ and dampened inflammation, resulting in improved
   behavioral outcomes along with substantially reduced synapse
   elimination. Our findings establish that AD-associated aberrant MG can
   be in situ reprogrammed with SER for Aβ clearance in an
   anti-inflammatory manner, with broad application in other
   inflammation-related diseases.
     __________________________________________________________________

   Microglia with a synthetic receptor show enhanced Aβ clearance without
   inflammation, opening pathways for AD treatment.

INTRODUCTION

   Neuroinflammatory responses mediated by abnormal amyloid deposition and
   dysregulated immune cell activation are critical causative factors in
   neurodegenerative diseases, including Alzheimer’s disease (AD) ([70]1,
   [71]2). Characterized by the accumulation of amyloid-β (Aβ) plaques and
   neurofibrillary tangles of hyperphosphorylated tau proteins, AD
   manifests mainly as progressive cognitive decline and memory
   impairment, and currently affects millions of individuals without an
   available cure ([72]3–[73]6). Existing treatments, such as the
   monoclonal antibody aducanumab, are shown to selectively bind Aβ
   aggregates and substantially reduce Aβ burden in AD patients’ brains
   ([74]7–[75]9). Despite promoting amyloid clearance, these
   antibody-based treatments only offer modest symptomatic relief in
   patients at the early and middle stages of AD, accompanied by several
   side effects, including brain edema and microhemorrhage associated with
   Fc receptor activation and complement cascade reactions
   ([76]10–[77]12). Furthermore, antibody-induced inflammatory responses
   within the central nervous system (CNS) could result in increased
   neuronal damage and inhibited neurogenesis, potentially exacerbating
   cognitive impairment ([78]7, [79]12–[80]14). Therapies that can
   effectively eliminate Aβ and inhibit the inflammatory response are
   urgently needed, but such approaches remain unexplored.

   Microglia (MG), the resident immune cells of the brain, play a crucial
   role in maintaining CNS homeostasis by clearing protein aggregates and
   cellular debris ([81]15–[82]17). This function is achieved mainly
   through efferocytosis, a process by which MG eliminate dying cells
   while triggering signals that inhibit the inflammatory response. As a
   classical efferocytic receptor, T cell immunoglobulin mucin protein 4
   (TIM4) is predominant in maintaining immune homeostasis and preventing
   autoimmune responses ([83]18, [84]19). TIM4-derived chimeric receptor
   for efferocytosis (CHEF) generated by fusing the signaling domain of
   cytoplasmic adapter protein ELMO1 to the cytoplasmic tail of TIM4
   displays a marked increase in efferocytosis and dampening of
   inflammation in multiple models of tissue injury ([85]18). In the AD
   microenvironment, MG exhibit impaired phagocytosis and efferocytosis
   due to a disturbed local microenvironment, leading to the accumulation
   of toxic debris and fostering a proinflammatory environment that
   exacerbates neurodegeneration ([86]20–[87]22). Thus, reprogramming
   AD-associated aberrant MG in situ through constructing synthetic
   efferocytic receptor (SER) based on CHEF could restore MG phagocytosis
   capacity while simultaneously attenuating multiple inflammatory
   insults, thus eliminating Aβ and mitigating AD progression.

   Here, we introduce a therapeutic strategy in which MG-editing mRNA
   lipid nanoparticles (MERLINs) are used to generate SER-modified MG
   (SER-MG) in situ. MERLINs encapsulating modified SER-mRNA are
   formulated with custom ionizable
   lipids, 1,2-dimyristoyl-rac-glycero-3-methoxypolyethylene glycol
   (DMG-PEG)–Mannose for MG in situ targeting and
   lysophosphatidylcholine-oleate (LPC-oleate) as a helper lipid to
   achieve high-efficiency meningeal and ventricular penetration after
   intrathecal injection. The SER, a variant of CHEF, consists of a
   single-chain variable fragment (scFv) derived from a monoclonal
   antibody targeting Aβ and the backbone of the TIM4 receptor, combining
   the effective signaling domain of the cytoplasmic adapter protein ELMO1
   to trigger the downstream efferocytic signaling pathway ([88]18). The
   engineered MG are designed to specifically recognize and engulf Aβ
   aggregates, alleviate neurotoxicity, and thereby mitigate AD. Compared
   with aducanumab, the generated SER-MG manifests strong efferocytosis,
   promotes the clearance of Aβ without inflammation, and substantially
   reduces neuronal death and synaptic loss, thus leading to improved
   behavioral outcomes. Collectively, our work provides a reversible and
   broadly applicable treatment strategy for AD and potentially for other
   neurodegenerative diseases.

   This work highlights the critical need for synthetic biological
   approaches that could efficiently target and clear toxic aggregates
   without triggering a concomitant inflammatory response for AD
   treatment. By leveraging mRNA-based therapeutics, we offer a compelling
   strategy that could transform the landscape of AD therapy and provide
   synthetic receptor design strategies to treat inflammation-related
   diseases.

RESULTS

The SER mediates noninflammatory efferocytosis of Aβ

   TIM4 is a PtdSer-binding receptor that mediates efferocytosis and is
   expressed exclusively by phagocytes. During efferocytosis, phagocytes
   take up debris while processing the ingested materials in an
   anti-inflammatory manner, thereby maintaining homeostasis of the
   microenvironment ([89]23). Here, we describe a SER that combines an
   scFv targeting Aβ, the TIM4 efferocytic receptor backbone, and the
   effective structural domain of downstream ELMO1, which is expressed on
   the surfaces of MG through proper design, modification, and
   transfection of mRNA ([90]Fig. 1, A and B). On the basis of the mRNA
   sequence, we first predicted the spatial structure of the SER and
   modeled the molecular docking of Aβ with the anti-Aβ scFv via
   AlphaFold2 (fig. S1) ([91]24). The analysis revealed that the SER has a
   TIM4-like backbone along with other functional structural domains,
   including a green fluorescent protein (GFP) tag ([92]Fig. 1C and fig.
   S2). Then, we applied lipid nanoparticle (LNP) technology to deliver
   mRNA encoding the SER to BV2 cells and found that transient expression
   of the SER could be achieved ([93]Fig. 1D). To validate the structure
   of the SER, further analysis demonstrated that the transfection
   efficiency of the anti-Aβ scFv with the G4S linker reached
   approximately 38.27% and 32.99% in BV2 cells and primary MGs,
   respectively ([94]Fig. 1E and fig. S3).

Fig. 1. SER-MG mediate the phagocytosis of Aβ without inducing
proinflammatory responses.

   [95]Fig. 1.
   [96]Open in a new tab

   (A) Schematic of SER mRNA design. (B) Schematic illustrating structural
   design of SER and its variants. (C) Structural simulation of TIM4 and
   SER with AlphaFold2. (D) Immunofluorescence images of BV2 cells
   expressing the SER with a GFP tag. Blue, DAPI; green, SER. Scale bar,
   10 μm. (E) Flow cytometry profiles of SER expression in BV2 cells and
   primary MG with a G4S linker of anti-Aβ scFv. (F) Immunofluorescence
   results for SER-MG incubated with Aβ for 1 hour (left) and
   corresponding linear scan of fluorescence intensity along white dotted
   line (right). Blue, DAPI; green, SER; red, dock180. Scale bar, 2 μm.
   (G) Immunofluorescence results (left) and enlarged images (right) for
   SER-MG incubated with Aβ for 1 hour. Blue, DAPI; green, phalloidin;
   red, Aβ. Scale bars, 2 μm (left) and 0.5 μm (right). (H and I)
   Immunofluorescence results (H) and calculated phagocytic index (I) for
   BV2 cells in different treatment groups after phagocyting Aβ plaques
   for 2 hours. Blue, DAPI; red, Aβ; green, WGA. Scale bar, 1 μm (n = 10).
   (J) Fluorescence images of BV2 cells after incubation with Aβ for 6
   hours (left) and corresponding linear scan of fluorescence intensity
   along white dotted line (right). Blue, DAPI; green, LysoTracker; red,
   Aβ. Scale bar, 5 μm. (K) Quantitative analysis of secreted TNF-α, IL-6,
   and IL-1β in Aβ-coincubated MG with different treatments (n = 9). (L)
   Quantitative analysis of secreted IL-10 and TGF-β in Aβ-coincubated MG
   with different treatments (n = 6). The data are presented as the
   means ± SD. The statistical comparisons in (I), (K), and (L) were
   performed via one-way ANOVA (ns = not significant, *P < 0.05,
   **P < 0.01, ***P < 0.001, ****P < 0.0001). (D), (F), (G), (H), and (J)
   show representative images of the corresponding independent biological
   samples.

   Furthermore, two SER variants, SER [anti–fluorescein isothiocyanate
   (FITC)] and SER (6M), were designed for subsequent functionality
   validation and mechanism exploration. As shown in [97]Fig. 1B, the
   anti-Aβ scFv was replaced with anti-FITC scFv in SER (anti-FITC), and
   six active sites of ELMO1 C terminus in SER were mutated for
   constructing SER (6M) ([98]18). In the presence of Aβ,
   immunofluorescence staining showed tangible colocalization of SER and
   dock180, while not observed in SER(6M)-MG, followed by cytoskeletal
   deformation of SER-MG for Aβ endocytosis ([99]Fig. 1, F and G, and
   figs. S4 to S6) ([100]25). This finding indicated that ELMO1 domain
   endowed SER with the ability to mimic the Rac activation and actin
   rearrangement activity of the natural ELMO/Dock180 complex and promote
   cytoskeleton reorganization in efferocytosis. Coincubation of the
   SER-MG with Aβ revealed that SER-MG had more robust phagocytotic
   effects than those of the monoclonal antibody targeting Aβ
   (aducanumab), yet SER (anti-FITC)–MG and SER (6M)–MG showed the
   deficiency in Aβ phagocytic capacity, indicating that both anti-Aβ scFv
   and ELMO1 domain were indispensable for Aβ-targeting phagocytosis
   induced by SER ([101]Fig. 1, H and I). Moreover, the phagocytosed Aβ
   was degraded within the lysosomes of BV2 cells ([102]Fig. 1J). Further
   analysis revealed that aducanumab substantially increased the secretion
   of inflammatory factors, whereas SER-MG demonstrated reduced
   inflammatory factor secretion as well as increased anti-inflammatory
   factor production after the phagocytosis of Aβ ([103]Fig. 1, K and L).
   The lower expression of proinflammatory factors [tumor necrosis
   factor–α (TNF-α), interleukin-6 (IL-6), and IL-1β] and the upper
   expression of anti-inflammatory factors [IL-10 and transforming growth
   factor–β (TGF-β)] in the SER-treated group than that in the
   aducanumab-treated group also indicated a robust anti-inflammatory
   effect of SER-MG (fig. S7). Additionally, the impact of SER on native
   phagocytic pathways of MG was also investigated. When coincubated with
   apoptotic cells, SER-MG exhibited comparable phagocytosis efficiency to
   untransfected MG, demonstrating a low risk of interfering with MG’s
   normal physiological function (fig. S8). Collectively, our designed and
   synthesized SER-MG have a specific structure and demonstrate a superior
   capacity to enhance the phagocytosis of Aβ while concurrently
   mitigating inflammatory responses, establishing the basis for
   subsequent cognitive improvement in vivo.

Synthesis and characterization of LNPs targeting Mfsd2a

   Given that major facilitator superfamily domain-containing protein 2a
   (Mfsd2a) is selectively expressed on the cerebral vascular endothelium,
   meninges, and ventricles ([104]Fig. 2A), where it transports common
   LPCs carrying long-chain fatty acids, including LPC-oleate and
   LPC–docosahexaenoic acid (DHA) ([105]26–[106]28), we propose that
   substituting helper lipids in conventional LNP components with
   LPC-oleate may achieve meningeal and ventricular penetration to deliver
   RNA drugs effectively following intrathecal injection. First, to screen
   and obtain suitable carbon chains with desirable transfection
   efficiency, we modified LPC-oleate with saturated fatty acids,
   including 10-carbon, 12-carbon, 14-carbon, 16-carbon, and 18-carbon
   fatty acids, and their structures were confirmed by ^1H nuclear
   magnetic resonance ([107]Fig. 2B and fig. S9). Next, our
   house-synthesized thymine ionizable lipid T12 was applied to
   encapsulate the mRNA, and then a microfluidic device was used to
   formulate different LNPs ([108]Fig. 2C and fig. S10). As shown in
   [109]Fig. 2 (D and E), the longer the carbon chain of LPC-oleate was,
   the larger the resulting particle size. However, there was no
   discernible effect on the surface charge of the LNPs ([110]Fig. 2F). We
   next tested the mRNA delivery efficacy of LNPs after the replacement of
   helper lipids with LPC-oleate, and flow cytometry analysis revealed
   that the desirable transfection efficiency of LNPs was achieved
   ([111]Fig. 2, G and H). LNP-14 loaded with mRNA encoding GFP exhibited
   outstanding meninges-penetrating efficiency (approximately 65.4%) and
   achieved the transfection of target cells in a transwell model (fig.
   S11). The in vivo luciferase expression distribution following
   intrathecal injection ([112]Fig. 2J) also confirmed that LNP-14 had
   better cerebral delivery efficiency, which were 8.91, 2.94, 2.04, and
   1.43 times greater than those of LNP-10, LNP-12, LNP-16, and LNP-18,
   respectively ([113]Fig. 2, I and K). Furthermore, we designed a
   C14-based formulation optimization strategy and found that the cells
   were transfected with the highest efficiency when the lipid ratios used
   in regimen 2 were used; therefore, this regimen was adopted for
   subsequent in vivo and in vitro studies (table S1 and fig. S12). In
   summary, these results indicate that LNPs synthesized with LPC-oleate
   display superior performance and have the potential for RNA delivery to
   generate SER-MG in situ.

Fig. 2. Construction and characterization of brain-penetrating LNPs.

   [114]Fig. 2.
   [115]Open in a new tab

   (A) Expression of Mfsd2a in the meninges and ventricles according to
   immunofluorescence staining. The white dashed line indicates the
   surface of the cerebral parenchyma; blue, DAPI; green, Mfsd2a. Scale
   bar, 25 μm. (B) LPC-oleate library. (C) Schematic of LNP preparation.
   (D) Particle diameter distribution of LNPs using LPC-oleate equipped
   with saturated fatty acid chain ranging from C10 to C18 as helper
   lipids. (E) Average particle diameter of LNP-10, LNP-12, LNP-14,
   LNP-16, and LNP-18 (n = 3). (F) Zeta potentials of LNP-10, LNP-12,
   LNP-14, LNP-16, and LNP-18 (n = 3). (G) Flow cytometry results after
   the transfection of GFP-mRNA with LNPs based on different LPC-oleates.
   (H) Quantitative fluorescence assay of cells transfected with different
   LNPs (n = 3). (I to K) Scheme for intrathecal injection in mice (J),
   visualization of luciferase expression distribution (I), and
   quantitative analysis of brain bioluminescence intensity (BLI) (K)
   after the injection of different luciferase mRNA-laden LNPs into nude
   mice (n = 3). The data are presented as the means ± SD. The statistical
   comparisons in (H) and (K) were performed via one-way ANOVA (*P < 0.05,
   ****P < 0.0001).

Targeting MG for the cellular delivery of SER-mRNAs

   CD206, also known as the macrophage mannose receptor, is a membrane
   protein expressed on the surfaces of macrophages that is intimately
   implicated in the phagocytosis and immune phenotypes of MG.
   CD206-expressing MG generally exhibit a phenotype associated with
   anti-inflammatory responses and tissue repair ([116]29, [117]30).
   Before confirming the targeting of mouse MG, we found that CD206 was
   expressed on the surface of MG after immunofluorescence staining of AD
   mouse brain sections (fig. S13). To endow LNPs with the ability to
   target MG, mannose moieties capable of strongly binding to the CD206
   receptor were covalently modified to the end of the DMG-PEG. LNP-14, in
   which DMG-PEG2000 was replaced with DMG-PEG-Mannose, was designated as
   MERLIN ([118]Fig. 3A) The formulated MERLIN solution showed a typical
   Tyndall effect after exposure to a laser at 532 nm ([119]Fig. 3B). The
   images of MERLINs captured by transmission electron microscopy (TEM)
   revealed a spherical structure with an approximate size of 150 nm,
   consistent with the dynamic light scattering (DLS) measurements
   ([120]Fig. 2C). Further analysis of the stability after 15 days of
   storage at 4°C, as shown in fig. S14, revealed that no obvious mRNA
   leakage was detected in MERLINs for up to 15 days, with negligible
   changes in size and zeta potential.

Fig. 3. MERLINs selectively target MG and achieve intracellular mRNA release.

   [121]Fig. 3.
   [122]Open in a new tab

   (A) Schematic diagram of the process by which MGs internalize MERLINs
   and express the SER on the surface of the cell membrane for the
   clearance of Aβ. (B) Transmission of LNPs detected by a 532-nm laser.
   (C) TEM image showing the morphology of the constructed LNPs. (D and E)
   Fluorescence images (D) and corresponding intensity (E) of BV2 cells
   after treatment with PBS, LNP-Cy7, and MERLIN-Cy7 (n = 3). Scale bar,
   20 μm. (F and G) Flow cytometry profiles (F) and corresponding
   quantitative analysis (G) of GFP fluorescence intensity after
   administration of GFP-mRNA, LNP, and MERLINs encapsulating GFP-mRNA to
   BV2, C8-D1A, and bEnd3 cells, respectively (n = 3). (H) Representative
   immunofluorescence images of BV2 cells after 2, 4, and 6 hours of
   incubation with MERLINs. Blue, DAPI; green, LysoTracker; red, Cy5-mRNA.
   Scale bars, 1 μm (left) and 2 μm (right). (I) Immunofluorescence
   showing the distribution of MERLIN-Cy7 and LNP-Cy7 within the cerebral
   parenchyma (white arrows, colocalization between MG and MERLIN-Cy7).
   Blue, DAPI; green, IBA1; red, MERLINs/LNPs. Scale bars, 100 μm (top)
   and 10 μm (bottom). (J) Quantitative analysis of MG internalizing
   MERLINs or LNPs (n = 10). The data are presented as the means ± SD. The
   statistical comparisons in (E) and (G) were performed via one-way ANOVA
   and two-way ANOVA, respectively. The statistical comparison in (J) was
   calculated using Student’s t test (**P < 0.01, ****P < 0.0001). (D),
   (H), and (I) show representative images of the corresponding
   independent biological samples.

   We next used flow cytometry and confocal laser scanning microscopy
   (CLSM) for MG-targeting delivery assessment of MERLIN and found that
   LNP modified with mannose (MERLIN) substantially increased the uptake
   efficiency of MG ([123]Fig. 3, D and E). Compared with that of mouse
   astrocyte C8-D1A cells and cerebral endothelial bEnd3 cells, the
   efficiency of GFP-mRNA transfection of MERLINs in BV2 cells was notably
   higher, suggesting that MERLINs are capable of selectively delivering
   mRNA to MG in the CNS ([124]Fig. 3, F and G). SER expression was then
   monitored after MERLIN transfection, and it peaked at 48 hours and
   almost returned to baseline at 96 hours (fig. S15). In addition, we
   examined the lysosomal escape ability of MERLINs via CLSM and revealed
   that MERLIN@CY5-mRNA colocalized mainly with LysoTracker Green–stained
   organelles after 2 hours of incubation. The separation of the green and
   red fluorescence spots was more notable after 6 hours, suggesting that
   MERLIN@CY5-mRNA efficiently escaped from the endosomes or early
   lysosomes to the cytoplasm ([125]Fig. 3H and fig. S16).
   Immunofluorescence staining of whole-brain sections after drug
   administration revealed substantial infiltration of MERLINs within the
   brain parenchyma, with predominant colocalization with MG, but LNPs
   permeated the brain parenchyma nonspecifically ([126]Fig. 3, I and J).
   Together, these results demonstrate that the constructed MERLINs could
   effectively penetrate the cerebral parenchyma, target MG, and achieve
   intracellular mRNA release and expression, demonstrating great
   feasibility for MG reengineering in vivo.

SER-MG facilitates Aβ clearance and attenuates AD in vivo

   To evaluate the capability of SER-MG to eliminate Aβ and mitigate
   cognitive impairment in vivo, 9-month-old APP/PS1 mice were selected
   and divided into three groups ([127]31). Wild-type mice were also
   included. The mice were intrathecally injected with phosphate-buffered
   saline (PBS), aducanumab, or MERLINs every 3 days for a total of 10
   injections ([128]Fig. 4A). Aβ deposition, as demonstrated by thioflavin
   T staining of the brain tissues of mice, was reduced at the end of
   treatment, with a reduction of approximately 57.2% in the
   MERLIN-treated group and approximately 40.9% in the aducanumab-treated
   group compared with the control group. A sustainable Aβ reduction was
   observed in the MERLIN-treated group, indicating that SER-MG could
   effectively phagocyte and degrade Aβ, and ultimately alleviate the Aβ
   burden in the brains of AD mice ([129]Fig. 4, B and C). Moreover,
   cerebral Aβ deposition was monitored for 4 weeks following MERLIN
   treatment, and Aβ levels remained relatively low for 3 weeks but
   rebounded in the fourth week (fig. S17). Subsequently, two variants of
   SER were also used for further mechanism exploration of SER-induced Aβ
   clearance. Compared with the aducanumab-treated group, MERLIN
   encapsulating SER mRNA inhibited the abnormal deposition of Aβ and
   reduced the damage to neurons manifested by LAMP1 expression ([130]Fig.
   4, D and E), which may be attributed to the construction of SER-MG in
   situ, mimicking the traditional efferocytosis signaling pathway and
   mediating Aβ phagocytosis in an anti-inflammatory manner ([131]Fig. 4,
   F and G, and fig. S18). In contrast, neither Aβ endocytosis nor reduced
   neuron damage was observed in two variant groups, indicating that the
   anti-Aβ scFv and ELMO1 motif were the key elements contributing to Aβ
   clearance and neuronal repair in SER. Mice were then tested for novel
   object/location recognition, and the discrimination index of mice
   treated with MERLINs was substantially improved, indicating that SER-MG
   could improve the cognition and memory of AD model mice ([132]Fig. 4, H
   to M, and fig. S19). In additional Morris water maze tests, mice in
   MERLIN treatment groups exhibited similar learning function to
   wild-type mice and determined the location of the platform more
   precisely than that in the control group after training for 5 days,
   demonstrating the therapeutic impact of SER on learning, cognition, and
   memory (fig. S20). Thus, we speculate that, compared with aducanumab
   treatment, the SER-MG generated by synthesized MERLINs can notably
   increase the phagocytosis and clearance of Aβ by MG in an
   anti-inflammatory manner, inhibit neuronal damage, and thus improve
   cognitive conditions in AD model mice.

Fig. 4. SER-MG mediates Aβ clearance and behavior rescue in APP/PS1 mice.

   [133]Fig. 4.
   [134]Open in a new tab

   (A) Schematic illustration of the experimental design. WT, wild type.
   (B and C) Thioflavin T staining (B) and statistical analysis of
   fluorescence intensity (C) of brain sections from normal and APP/PS1
   mice after various treatments. Scale bar, 200 μm (n = 3). (D and E)
   Representative immunofluorescence images (D) and statistical analysis
   (E) of Aβ burden and neuron damage in wild-type or APP/PS1 mice from
   different groups. Blue, DAPI; green, LAMP1; red, Aβ. Scale bars, 100 μm
   (left) and 30 μm (right) (n = 3). (F and G) Immunofluorescence staining
   of wild-type and AD mouse brain sections (F) from different treatment
   groups to observe Aβ phagocytosis by MG and quantitative analysis of
   fluorescence intensity (G). Scale bars, 50 μm (left) and 10 μm (right)
   (n = 20). (H to J) Graphical illustration of the novel object location
   (NOL) test (H) and representative pictures of the locomotor
   trajectories of mice in different treatment groups (I) with statistical
   analysis (J) (n = 6). (K to M) Graphical illustration of the novel
   object recognition (NOR) test (K) and representative pictures of the
   movement trajectories of mice in different treatment groups (L) with
   statistical analysis (M) (n = 6). The data are presented as the
   means ± SD. The statistical comparisons in (C), (E), (G), (J), and (M)
   were performed via one-way ANOVA (ns = not significant, *P < 0.05,
   **P < 0.01, ***P < 0.001, ****P < 0.0001). (B), (D), (F), (I), and (L)
   show representative images of the corresponding independent biological
   samples.

Single-cell RNA sequencing reveals the neuronal improvement effects of SER-MG

   To further elucidate the mechanism of the therapeutic strategy of
   generating SER-MG via MERLINs in vivo, we performed single-cell
   sequencing of the hippocampi from AD mice treated with PBS, aducanumab,
   or MERLINs via the same procedure described above ([135]Fig. 4A). The
   data revealed that editing MG with MERLINs increased the number of MG
   in the hippocampal region, which is critical for the elimination of Aβ
   ([136]Fig. 5, A and B, and fig. S21). The number of Cajal-Retzius cells
   is higher in the brains of MERLIN-treated AD mice than in those of
   aducanumab-treated mice, which contributes to neuronal remodeling in
   the brain and underlies the improvement in cognition in AD mice
   ([137]32, [138]33).

Fig. 5. Single-cell sequencing analysis reveals the mechanism of SER-MG
treatment in APP/PS1 mice.

   [139]Fig. 5.
   [140]Open in a new tab

   (A) Single-cell sequencing analyses of the brain hippocampal region of
   AD mice in the PBS, aducanumab, and MERLIN treatment groups and the
   corresponding UMAP plots of cells depicting separation into nine
   clusters. (B) Cluster fractions of cells from the hippocampus of each
   treatment group. (C and D) GSEA comparing MERLIN and aducanumab
   treatment. Gene set terms are indicated as follows: FDR, false
   discovery rate; NeS, normalized enrichment score. (E) KEGG enrichment
   analysis of signaling pathway enrichment of neurons in the MERLIN
   versus aducanumab treatment groups. (F to H) GO enrichment analysis of
   neurons based on sequencing results from different treatment groups.

   MG from different treatment groups were further segregated for
   differential expression gene (DEG) analysis. Consistent with the in
   vitro data, Fc-mediated endocytosis was down-regulated in
   MERLIN-treated mice, and the activation of inflammation-related
   signaling pathways, such as the mammalian target of rapamycin (mTOR)
   pathway, was inhibited according to gene set enrichment analysis (GSEA)
   comparing MERLIN and aducanumab treatment, confirming that SER-MG
   activate efferocytosis rather than FcR-mediated Aβ phagocytosis
   ([141]Fig. 5, C and D). Further analysis of the scRNA-seq data explored
   the molecular mechanisms by which SER-MG induced an anti-inflammatory
   phenotype. We identified 235 genes up-regulated and 174 genes
   down-regulated in the MERLIN group when compared to the aducanumab
   group (fig. S22A). Quantitative heat map analysis of the
   inflammation-related genes highlighted in the volcano plot showed
   substantial up-regulation of anti-inflammatory genes and
   down-regulation of inflammatory genes including Tnf and Il6, which was
   consistent with the enzyme-linked immunosorbent assay (ELISA) results
   (fig. S22B). Gene ontology (GO) annotation of differentially
   overexpressed genes in MG revealed that differentiational genes are
   enriched in pathways associated with mitosis and cytoskeletal
   remodeling, indicating that the construction of SER facilitates
   microglial proliferation and promotes cytoskeletal reorganization,
   thereby enhancing their capacity to phagocytose amyloid proteins (fig.
   S22, C and D).

   To elucidate the influence of the construction of SER-MG on CNS,
   neurons from different treatment groups were segregated for DEG
   analysis similarly. The pathway enrichment analysis based on the Kyoto
   Encyclopedia of Genes and Genomes (KEGG) was then used to further
   explore the mechanism of SER’s improving effect on neuronal repair and
   cognition elevation ([142]34). Compared with those in the
   aducanumab-treated group, expression of genes related to axon guidance
   (Slit1, Slit2, Nrp1, Epha6, and Unc5c) and synapse formation (Prkca,
   Cacna1c, Kcnq5, and Grik4) was up-regulated in AD mice treated with
   MERLINs. Additionally, genes with up-regulated expression were enriched
   in the oxytocin signaling pathway, which is thought to be associated
   with neuroplasticity and the modulation of social behavior ([143]35,
   [144]36) ([145]Fig. 5E and fig. S23A). Further analysis by GO
   annotation revealed notable enrichment of genes related to axon
   generation or guidance in biological process (BP), synaptic formation
   in cellular component (CC), and signal transduction activities in
   molecular function (MF). Overall, these results collectively elucidate
   the mechanism by which MERLIN treatment improved cognitive function in
   AD mice compared with aducanumab treatment ([146]Fig. 5, F to H, and
   fig. S23, B to D), suggesting its potential for application in other
   neurodegenerative diseases.

   Collectively, we developed a reengineering strategy by using MERLINs to
   selectively deliver SER-mRNAs into MG and generate Aβ-specific SER-MG
   in situ, which effectively eliminate Aβ and suppress the intracranial
   inflammatory response more effectively than aducanumab treatment does,
   thus promoting neural repair and improving behavioral outcomes in AD
   model mice ([147]Fig. 6).

Fig. 6. Intracranial generation of SER-MG for anti-inflammatory clearance of
Aβ.

   [148]Fig. 6.
   [149]Open in a new tab

   (A) Schematic of MERLIN-enabled in situ generation of SER-MG for
   alleviating AD via Aβ clearance and inflammation resolution. (B)
   Transport of MERLINs penetrating the meninges or ventricles. (C) MG
   selectively internalize MERLINs via CD206 and express the SER on their
   surface, subsequently activating downstream signals to reorganize the
   cytoskeleton and mediate anti-inflammatory responses.

DISCUSSION

   Aberrant deposition of Aβ plaques in the brain extracellular space is a
   key hallmark of AD, and brain Aβ accumulation is commonly denoted as
   the initiating event in AD’s pathogenesis, resulting in subsequent
   neurofibrillary tangle formation, synapse loss, and neuronal cell death
   ([150]37–[151]39). Current treatments for AD involve Aβ clearance
   through multiple means, including the use of monoclonal antibodies.
   Aducanumab, the first antibody therapy approved by the U.S. Food and
   Drug Administration for AD treatment, has effectively reduced Aβ
   burdens within the brains of AD patients. However, owing to side
   effects such as Fc receptor activation and complement cascade
   reactions, such therapies may activate local immune cells while
   inducing neuroinflammation, further exacerbating cognitive and memory
   deficits in AD patients ([152]7, [153]9). Eliminating Aβ while
   restricting the inflammatory response is therefore of utmost importance
   in AD treatment ([154]40). The SER-MG described in this study could
   facilitate Aβ elimination and simultaneously suppress neuroinflammatory
   responses ([155]Figs. 1 and [156]4). Further single-cell RNA sequencing
   of the hippocampus elucidated the neural improvement effects of SER-MG
   in vivo, revealing up-regulated neuronal axon guidance and synapse
   restoration signaling pathways, indicating the potential for mitigating
   AD in patients ([157]Fig. 5).

   MG, the predominant type of immune cell in the brain, maintain
   homeostasis in the CNS through the phagocytic clearance of protein
   aggregates and cellular debris. However, MG in the AD brain display
   signs of impaired phagocytosis, and dysfunction in MG’s efferocytosis
   could lead to the accumulation of such debris and promote a
   proinflammatory and anti-repair microenvironment that may result in
   further brain tissue dysbiosis and neurodegeneration ([158]41). Thus,
   restoring the phagocytic functions of aberrant MG represents a
   promising therapeutic strategy. Here, we first reported a SER that
   enables MG to specifically recognize and phagocytose Aβ in a
   noninflammatory manner through the fusion of an scFv targeting Aβ, the
   TIM4 efferocytic receptor backbone, and the effective fragment of
   downstream ELMO1. SER-MG were demonstrated to mimic the TIM4 signaling
   pathway, facilitating Aβ clearance via cytoskeletal rearrangement, and
   exhibited more robust elimination effects than those of aducanumab
   without eliciting local neuroinflammatory responses ([159]Fig. 1). This
   therapeutic approach provides a strategy for the noninflammatory
   clearance of abnormal protein deposits, warranting further exploration
   in other inflammation-associated diseases.

   To achieve precise reprogramming of MG, we propose a MERLIN mRNA
   delivery system modified with mannose to target CD206, which is
   selectively expressed on MG, and observed desirable specific cellular
   internalization. Additionally, we demonstrated that substituting helper
   lipids in conventional LNP components with LPC-oleate achieved
   meningeal and ventricular penetration to deliver RNA drugs effectively
   after intrathecal injection. This optimized MERLIN delivery system
   could serve as an effective tool for delivering nucleic acid drugs into
   MG, providing therapeutic potential for the treatment of AD and other
   cerebral diseases ([160]Figs. 2 and [161]3).

   However, there are several limitations to our current study. Although
   MERLIN-mediated SER-MG in situ editing therapy has demonstrated notable
   efficacy in AD treatment, the limitation must be noted that its impact
   on nervous system still needs intuitive and rigorous evaluations.
   Additionally, further studies should continue to explore the long-term
   safety and clinical prospects of this system and evaluate its potential
   application in other brain diseases. In particular, assessing the
   immunogenic responses that may arise from repeated administrations is
   essential for future translation into the clinic.

   In summary, our findings establish that MG in the AD microenvironment
   can be reengineered to express Aβ-specific SER via MERLINs, thereby
   enabling them to eliminate Aβ plaques without inducing
   neuroinflammation to prevent neural and synaptic damage. Therefore, our
   work may provide a more efficient immunotherapeutic strategy for
   patients suffering from AD or other inflammation-related diseases.

METHODS

Animals

   APP/PS1 mice (female, 9 months old) were purchased from GemPharmatech
   Co. Ltd. (Nanjing, China). All mouse experiments were performed
   according to the Research Ethics Committee of Shandong University and
   the Ethics Committee of Qilu Hospital (Shandong, China), in compliance
   with all relevant ethical regulations. The mice were housed in a
   barrier environment under 12-hour light/dark cycle with a humidity
   between 40% and 60% and a constant temperature of 24°C.

Materials

   All solvents (analytical grade) and reagents were used as supplied by
   commercial sources unless otherwise indicated. Cell culture flasks were
   purchased from NEST Biotechnology (Wuxi, China). mRNA encoding
   luciferase was purchased from APExBIO (Beijing, China). DMG-PEG-Cy7 was
   obtained from Chongqing Yusi Pharmaceutical Technology Co. Ltd.
   (Chongqing, China). Cholesterol,
   1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE), and DMG-PEG were
   obtained from AVT Pharmaceutical Tech Co. Ltd. (Shanghai, China). Aβ
   was purchased from Scierbio (Hefei, China). Information of antibodies
   used was listed as follows: Mfsd2a (Cell Signaling Technology, clone
   name: E8U2O), IBA1 (Abcam, clone name: [162]EPR16589), CD206-APC
   (BioLegend, clone name: C068C2), G4S Linker-phycoerythrin (PE) (Cell
   Signaling Technology, clone name: E7O2V), DOCK180 (Cell Signaling
   Technology, clone name: C4C12), β-amyloid (Cell Signaling Technology),
   and LAMP1-FITC (BioLegend, clone name: 1D4B). Phalloidin-iFluor 488
   (Abcam), 4′,6-diamidino-2-phenylindole (DAPI), and Alexa Fluor
   647–labeled goat anti-rabbit immunoglobulin G (IgG) were purchased from
   Beyotime Biotech Inc. (Shanghai, China). DyLight 549–labeled goat
   anti-rabbit IgG was purchased from Abbkine Scientific Co. Ltd. (Wuhan,
   China). Thioflavin T was purchased from Sigma (Shanghai, China), and
   wheat germ agglutinin (WGA)–Alexa Fluor 488, LysoTracker, fetal bovine
   serum (FBS), Dulbecco’s modified Eagle’s medium (DMEM), penicillin, and
   trypsin-EDTA were obtained from Thermo Fisher Scientific Inc.
   (Shanghai, China). Serum-free cell freezing medium was purchased from
   New Cell & Molecular Biotech (Suzhou, China). ELISA kits for TNF-α,
   IL-6, IL-1β, IL-10, and TGF-β were purchased from PeproTech Inc.
   (Cranbury, NJ, USA) and Beijing Solarbio Science & Technology Co., Ltd.
   (Beijing, China).

Extraction of primary cells

   Meningeal tissues are dissected from anesthetized mice by carefully
   removing the meninges from the brain. The tissues are then minced into
   small pieces and enzymatically digested using trypsin at 37°C. After
   digestion, the cell suspension is filtered through a cell strainer (70
   μm) to remove debris, followed by centrifugation to collect the cells.
   The isolated cells are resuspended in a culture medium (DMEM
   supplemented with 10% FBS and 1% penicillin-streptomycin solutions) and
   then plated for cultivation at 37°C in a 5% CO[2] incubator. Initially,
   the culture medium was replaced every other day. Subculturing was
   performed every 2 to 3 days when the cell density exceeded 90%. Cells
   were purified following subculturing by using the differential adhesion
   method after trypsinization.

   For the extraction of primary MG, culture surfaces were aseptically
   coated with poly-d-lysine solution (0.1 mg/ml) for 30 min. The brains
   were dissected, placed in a dish containing cold DMEM medium, and
   dissociated by repeated pipetting followed by filtration through a
   70-μm cell strainer and centrifuged at 400g for 5 min at room
   temperature. The supernatant was discarded, and cells were resuspended
   in culture system (DMEM supplemented with 10% FBS and 1%
   penicillin-streptomycin solutions) and then plated for cultivation at
   37°C in a 5% CO[2] incubator. The culture medium was completely
   replaced the following day, and thereafter, half of the medium was
   replaced every 3 days. After 2 weeks, MG were harvested from the medium
   by shaking the flasks at 180 rpm for 2 hours at 37°C. The MG were then
   plated at a density of 5 × 10^4 for 24 hours before further
   experimentation.

Preparation of ionizable lipid

   N-tert-butoxycarbonyl-1,2-ethylenediamine (10 mmol), 1-bromododecane
   (22 mmol), potassium carbonate (20 mmol), and acetonitrile (60 ml) were
   added to a 250-ml round-bottom flask and refluxed at 80°C for 72 hours.
   The solids were removed by filtration, and the residue was purified
   through column chromatography, yielding tert-butyl
   (2-(didodecylamino)ethyl)carbamate. The intermediate (5 mmol) was then
   dissolved in 1,4-dioxane (10 ml), followed by the addition of 4 M
   hydrochloric acid 1,4-dioxane (12.5 ml). The mixture was stirred at
   room temperature for 2 hours to yield N,N-didodecylethane-1,2-diamine,
   which was subsequently reacted with thymine-1-acetic acid (5 mmol) at
   room temperature for 6 hours in the presence of
   1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDCI; 5 mmol) and
   N-hydroxysuccinimide (NHS; 5 mmol). The resulting product was then
   purified by column chromatography using a dichloromethane/methanol
   (50:1) eluent, yielding the final compound, T12.

Preparation of LPC-oleate

   Choline glycerophosphate and dibutyltin oxide were dissolved in
   isopropanol and stirred at 100°C for 6 hours. Triethylamine was then
   added at 0°C, followed by the successive addition of fatty acid
   chlorides, including decanoyl chloride, lauryl chloride, palmitoyl
   chloride, and stearoyl chloride. The reaction mixture was stirred at
   room temperature for 16 hours. The solvents were subsequently removed
   under reduced pressure, and the residues were purified by column
   chromatography to yield the intermediates LPC-C10, LPC-C12, LPC-C14,
   LPC-C16, and LPC-C18. Next, oleic acid, 1-methylimidazole (NMI), and
   2-methyl-6-nitrobenzoic anhydride (MNBA) were added to the
   dichloromethane solution. After stirring at room temperature for 20
   min, the LPC intermediates were added, and the mixtures were stirred at
   room temperature for 48 hours. The products were concentrated and
   purified by silica gel column chromatography to obtain LPC-oleate with
   different carbon chain lengths.

Preparation and characterization of the LNP

   To synthesize LNP, a microfluidic device was used to mix an aqueous
   phase containing mRNA with an ethanol phase containing lipids and
   cholesterol. Specifically, the aqueous phase was prepared using 50 mM
   citric acid buffer (pH 4.5) and mRNA, while the ethanol phase consisted
   of T12, helper lipid, cholesterol, and DMG-PEG at molar ratios
   described in table S1, respectively. The two phases were combined in a
   3:1 ratio within the microfluidic device. Following this mixing
   process, the LNP was dialyzed against PBS for 0.5 hours. Additionally,
   MERLIN was prepared replacing DMG-PEG with DMG-PEG-Mannose according to
   optimal formula.

   The average particle size, polydispersity index (PDI), and zeta
   potential of the mRNA-LNP were determined with Zetasizer Nano ZS90
   (Malvern Instruments, Malvern, UK). The data represent the average of
   three measurements from different samples. The morphology of MERLIN was
   examined with a transmission electron microscope (HT7800, Hitachi,
   Japan). Samples were dropped onto a carbon-formvar copper grid and
   negatively stained with a 2% phosphotungstic acid solution for this
   analysis. A gel electrophoresis retardation assay was performed to
   evaluate the encapsulating ability of MERLIN. Agarose gels (2%) were
   prepared using 1× Tris Acetate-EDTA buffer (TAE) buffer and heated in a
   microwave oven until the agarose was completely dissolved. The samples
   were then loaded onto the gels and subjected to electrophoresis at 120
   V for 25 min. The resulting bands were visualized under an ultraviolet
   (UV) lamp. To study MERLIN storage stability at 4°C, the average
   particle size and zeta potential were monitored for a week.

mRNA delivery property characterization of MERLIN

   To evaluate the MG-targeting property of MERLIN, 1 × 10^5 of BV2,
   C8-N1A, and bEnd3 cells were embedded on six-well plates. Then, cells
   were cultured with MERLIN for transfection. Particle phagocytosis was
   analyzed by confocal microscope imaging at 1 hour and the intracellular
   behavior of internalized MERLINs was recorded by confocal microscope
   imaging at 2, 4, and 6 hours.

In vitro transwell model

   An in vitro transwell model was established using a six-well transwell
   cell culture system. BV2 cells were seeded in the lower chamber at a
   density of 1 × 10^5 cells per well. Similarly, the upper chamber was
   seeded with mouse primary meningeal cells at the same density. Cellular
   integrity was assessed by measuring the
   transepithelial/transendothelial electrical resistance (TEER) values
   using a Millicell-ERS voltmeter (Millipore). A TEER value greater than
   200 Ω/cm^2 indicated the successful establishment of the transwell
   model. LNP-Cy7 encapsulating GFP-mRNA with different LPC-oleate
   compositions was added to the upper chamber, and supernatants of both
   chambers were collected to evaluate LNP concentration at 1 hour. After
   24 hours of incubation, fluorescence microscopy was used to visualize
   the BV2 cells in the lower chamber and quantify the fluorescence
   intensity of GFP.

In vivo distribution study

   Five groups of mice (three mice per group) were treated with luciferase
   mRNA-laden LNPs containing different LPC-oleate by intrathecal
   injection. After 12 hours of injection, mice were anesthetized with
   isoflurane and biofluorescence imaging (BLI) was performed using an in
   vivo imaging system (IVIS) (Spectrum, PerkinElmer, MA, USA) to monitor
   the distribution of expressed luciferase.

SER expression kinetics

   To determine the duration of SER expression, the cells were fixed at 0,
   12, 24, 48, 72, and 96 hours after transfection. The fixed cell samples
   were stained with DAPI and recorded by Leica TCS Sp8 (Wetzlar,
   Germany). The fluorescence intensity of GFP fused to the C terminus of
   SER was quantified using ImageJ.

Apoptotic cell phagocytosis of SER-MG

   The Jurkat cells were stained with
   2',7'-bis-(2-carboxyethyl)-5-(and-6)-carboxyfluorescein,acetoxymethyl
   ester (BCECF-AM) and irradiated under 254-nm UV light for 15 min to
   induce apoptosis and then added to BV2 cells pretreated with PBS or SER
   mRNA-laden LNP for 2 hours of coincubation. Subsequently, free Jurkat
   cells were removed, and the BV2 cells were fixed and stained with DAPI
   for detection with Leica TCS Sp8 (Wetzlar, Germany).

Aβ oligomerization

   For the oligomerization of Aβ, 1 mg of Aβ was dissolved in 400 μl of
   HFIP (1,1,1,3,3,3-hexafluoro-2-propanol; Sigma, 105228) at room
   temperature for 2 hours. The Aβ solution was then lyophilized using a
   freeze dryer to completely remove HFIP and subsequently resuspended in
   5% dimethyl sulfoxide (DMSO) in PBS to achieve a final concentration of
   500 μM, followed by 30 min of sonication. The dissolved Aβ was
   incubated at 4°C for 24 hours to induce oligomerization and stored at
   −20°C for further analysis.

Flow cytometry analysis

   After 48 hours of transfection by LNP with different LPC-oleate or
   MERLIN encapsulation GFP or SER-mRNA, cells were collected and
   suspended in PBS, filtered through a 70-μm nylon cell filter, and
   counted for flow cytometry analysis.

Cytokine measurement

   Cytokine secretion was detected using ELISA kits using a double
   antibody sandwich ELISA technique. The wells of the ELISA plates were
   coated with capture antibodies and incubated overnight at room
   temperature. This was followed by a 1-hour incubation with a blocking
   buffer. To observe inflammatory responses under our culture conditions,
   as to in vitro experiments, BV2 cells were pretreated with
   lipopolysaccharide (LPS) at a concentration of 200 ng/ml and then
   incubated with Aβ (5 μM) in the presence of PBS, aducanumab, or MERLIN.
   Standards and supernatants were detected under the instruction. As to
   in vivo evaluation, the harvested tissue samples were ground into a
   homogenate on ice, followed by centrifugation to obtain supernatant for
   detection. Color development was measured using a microplate reader
   (Bio-Rad, CA, USA) at 450 nm, with a wavelength correction set at 620
   nm.

   Measurements of mRNA levels of cytokines in cells were performed by
   quantitative polymerase chain reaction (qPCR). BV2 cells were
   stimulated with LPS (200 ng/ml) before the drug treatment, and then
   they were incubated with Aβ and treated with PBS, aducanumab, and
   MERLIN. mRNA was isolated after drug treatment. mRNA (100 to 500 ng)
   was reverse-transcribed to generate complementary DNA.

Phagocytosis assay

   For the phagocytosis of Aβ in vitro, BV2 cells were incubated with Aβ
   (5 μM) in the presence of PBS and aducanumab (5 μg/ml). The experiment
   group was treated with MERLIN for 24 hours previously. Then, samples
   were fixed with 4% paraformaldehyde for 15 min and their membranes were
   penetrated with 0.3% Triton. In vivo, the mice were divided into three
   groups (six mice per group) and treated via intrathecal injection 10
   times a month with 5 μl of PBS, aducanumab (250 μg/ml), or MERLIN
   (containing mRNA at 250 μg/ml). After the treatment, behavioral tests
   were conducted on them blindly. Subsequently, brain tissues were
   collected and prepared into frozen sections for immunofluorescence
   analysis. Hippocampal tissues were then harvested for scRNA-seq
   experiments. Additionally, MERLIN-treated mice (three mice per time
   point) were sacrificed at 1, 2, 3, and 4 weeks following treatment, and
   the brain tissues were collected for immunofluorescence analysis.

Immunofluorescence staining

   In immunofluorescence analysis, all frozen sections were incubated with
   corresponding primary antibodies for 12 hours at 4°C according to
   different experiments. They were washed with PBST (0.1% Tween 20 in
   1× PBS) three times before secondary antibodies were added for 1 hour
   at room temperature. DAPI or WGA (Invitrogen, 1:1000 dilution) was used
   for the staining of cell membranes or nuclei on demand. For thioflavin
   T staining, sections were sequentially washed in 70% and 80% ethanol
   for 1 min each. They were then stained with a 1% thioflavin T solution
   in 80% ethanol for 15 min at room temperature. Following staining,
   sections underwent washes in 80% and 70% ethanol for 1 min each and two
   washes in distilled water. After a 10-min dehydration process, they
   were mounted using a mounting medium and imaged with a confocal
   microscope. All confocal images of brain sections were captured using a
   Leica TCS Sp8 (Wetzlar, Germany).

NOR and NOL tests

   Both novel object recognition (NOR) and novel object location (NOL)
   tests were conducted in a custom-made acrylic square box (30 cm by 30
   cm by 28 cm, width by depth by height), with one side marked by a
   stripe for location information. Each task comprised three phases:
   habituation, training, and testing.

   For the NOL test, mice were first habituated to the testing environment
   by allowing free exploration of the open-field arena for 10 min.
   Twenty-four hours later, mice were introduced to the box containing two
   identical objects for a 10-min session. During the testing phase, one
   of the objects was relocated within the box, and the mice were observed
   for another 10 min. In the NOR test, the process was identical to the
   NOL test except that relocation was replaced with novel object
   recognition.

Morris water maze test

   Morris water maze test was conducted in a large circular pool filled
   with opaque water, where a liftable escape platform was in a fixed
   location. The experiment included a training phase over five
   consecutive days and a test phase on the sixth day. For the training
   period, mice were placed in the water facing the wall of the water
   maze, and the time and path to reach the escape platform beneath were
   recorded. In the testing phase, the platform was removed, and the time
   and path reaching the platform’s location were recorded.

Bioinformatic analysis and SER structure prediction

   For scRNA-seq experiments of drug-treated mice, hippocampal tissues
   were isolated into single cells, counted, processed, and sequenced
   following the manufacturer’s instructions. Downstream analyses were
   performed in R using the Seurat package (v.4.2.0). Clustering and
   Uniform manifold approximation and projection (UMAP) analysis were
   performed based on the statistically significant principal components.
   Data visualizations were performed using Seurat functions DimPlot,
   DotPlot, and FeaturePlot. Other data visualizations were performed
   using ggplot2. KEGG, GO, and GSEA analysis were conducted using
   clusterProfiler and AnnotationHub packages. SER structure was predicted
   under the construction of AlphaFold2 ([163]24).

Statistics and reproducibility

   All data were analyzed using GraphPad Prism 10 software, with results
   presented as the mean ± SD. The statistical significance of differences
   between groups was analyzed by Student’s t test, one-way analysis of
   variance (ANOVA), or two-way ANOVA. A value of P < 0.05 was considered
   to indicate statistical significance.

Acknowledgments