Clinical cohort study

Participants. This study was conducted from January 2017 to October 2022 at Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School. All participants were right-handed and possessed > 6 years of education. T2DM diagnosis was based on the American Diabetes Association criteria [26] and menopause was diagnosed as 12 months of amenorrhea without a pathological cause. Exclusion criteria were: (1) hyperprolactinemia or pregnancy or parturition, (2) hormone or steroid treatment or drugs that modulate ovarian steroid secretion in 6 months prior to the study, or oral contraceptive use in 6 months prior to the study, (3) premenopausal and perimenopausal women, (4) history of neurological and psychiatric disorders, (5) dysfunction of organs including heart, liver or kidney, (6) thyroid diseases, (7) Acute complications such as diabetic ketoacidosis, hyperglycemic hyperosmolar state, or hypoglycemic coma within 6 months. Written informed consent was obtained from all participants.

Clinical measurements

Resting blood pressure, height, weight of patients were collected. Plasma glucose, insulin, and C-peptide concentrations were detected at fasting. Serum fasting total cholesterol (TC), triglyceride (TG), high density lipoprotein cholesterol (HDL-C), low density lipoprotein cholesterol (LDL-C), and glycated hemoglobin A1c (HbA1c) concentrations were measured. Insulin resistance was estimated by the HOMA2 Calculator (HOMA2 v2.2.3; Diabetes Trials Unit, University of Oxford). Serum pituitary hormones included prolactin (PRL), follicle stimulating hormone (FSH), luteinizing hormone (LH), thyroid stimulating hormone (TSH), adrenocorticotropic hormone (ACTH) and growth hormone (GH) were measured by an automated chemiluminescent immunoassay (Siemens Immulite 2000, Germany) at 8:00 a.m. All patients fasted for 10 h and stayed quietly for 10 min before the measurement of serum pituitary hormones.

Cognitive assessments

General cognition was evaluated by Mini-Mental State Examination (MMSE) [27] and Montreal cognitive assessment (MoCA, Beijing Version) [28]. MCI was diagnosed according to the established criteria [29]: (1) Cognitive concerns from patients, informants or skilled clinicians. (2) Objective evidence of impairment in cognitive domains, which was demonstrated as subjects whose education-adjusted MoCA scores were less than 26. (3) Preservation of independence in functional abilities. Independence in daily functioning was evaluated by Activities of Daily Living. (4) Lack of evidence for dementia. Multiple cognitive subdomains, including immediate memory, visuospatial constructional, language, attention, and delay memory, were evaluated by the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) [30]. Processing speed was evaluated by the Trail Making Test (parts A and B) [31], and executive function was assessed by the completion time of Stroop Color-Word Test (parts I, II, and III) [32].

MRI data acquisition and analysis

Image data were obtained using a 3.0T MR scanner (Achieva TX; Philips Medical Systems, Eindhoven, Netherlands) with an 8-channel head coil. Structural data were acquired with high-resolution T1-weighted three-dimensional fast field echo structural scans (repetition time 9.7 ms, echo time 4.6 ms, field of view 256 mm×256 mm×192 mm, flip angle 8°, and voxel size 1 mm×1 mm×1 mm). Segmentation of hippocampal subregions was accomplished by FreeSurfer software 7.2 (Athinoula A. Martinos Center for Biomedical Imaging, USA). A 3D image of the right hippocampus was established using 3D Slicer software 5.2.1 (National Inst. Of Health, USA).

Animal studies

The PRL knockout (PRL KO) mice were generated by using CRISPR/Cas 9 technology to modify the PRL gene. The brief process was as follows: sgRNA was transcribed in vitro, Cas9 and sgRNA were microinjected into fertilized eggs of C57BL/6 mice to obtain F0 generation mice. The correct F0 generation positive mice were mated with C57BL/6 mice to obtain a stably heritable F1 generation positive mouse model as verified by PCR-sequencing. The offspring were genotyped by PCR with the following primers: wild type (WT) alleles: GPS00003371-Prl-KO-tF1 5′-AACAGTATGTGCAAGACCGTG-3′ andGPS00003371-Prl-KO-tR1 5′-TGGACAGTTTATGGCTCAGCTAC-3′; PRL KO alleles: GPS00003371-Prl-KO-tF1 5′-CAGCTAGTGAGCTATTTAACAGTGCTG-3′ and GPS00003371-Prl-KO-tR1 5′-TCAGCCAAACTCATGGATAGAGG-3′.The sequence of sgRNA and experiments to verify that PRL is not expressed are presented in Supplementary Fig. S5. Cx3cr1CreERT2 mice were crossed with PRLRfl/fl mice (purchased from the Model Animal Research Center of Nanjing University). Microglial PRLR conditional knockout (cKO) mice were generated after we treated Cx3cr1CreERT2: PRLRfl/fl mice with tamoxifen (TAM) at different stages indicated. After TAM induction, mice were randomly divided into two groups and fed either a normal chow diet (5010, LabDiet, USA) or an HFD diet (D12492, Research Diets, USA) for 12 weeks. All mice were housed in a specific-pathogen-free facility under a 12-h light-dark cycle with food and water available ad libitum.

PLX3397 and minocycline treatment

A total of 38 mice were treated with PLX3397 or its vehicle. PLX3397 was dissolved in 5% DMSO (HY-Y0320, MCE, USA) and 40% PEG300 (HY-Y0873, MCE, USA) in ddH2O. WT and PRL KO mice were treated with PLX3397 (1029044-16-3, MCE, USA) by oral gavage using a gavage needle. Treatment with PLX3397 was administered at a dose of 50 mg/kg per day for 4 weeks. A total of 39 mice were treated with minocycline or saline. Minocycline was purchased from Sigma Aldrich (M9511, Canada). WT and PRL KO mice were intraperitoneally injected with minocycline or saline for one month at a dose of 50 mg/kg/day.

Tamoxifen injection

Tamoxifen injections were performed in 35 Cx3cr1CreERT2: PRLRfl/fl mice and 34 PRLRfl/fl mice. Tamoxifen (HY-13757A, MCE, USA) was dissolved in 10% ethanol and 90% corn oil. Mice were injected with tamoxifen intraperitoneally once every 24 h for one week, with a dose of 75 mg tamoxifen/kg body weight (20 µg/µl).

Viral injection

A total of 20 C57BL/6 mice accepted the virus injection. Adeno-associated virus (AAV) adenoviruses expressing short-hairpin RNA (shRNA) specific for neuronal PRLR or scrambled shRNA were obtained from brainVTA (Wuhan, China). AAV-hSyn-EGFP-shRNA (PRLR) or scramble were infused into the bilateral hippocampus (anteroposterior=-2.8 mm, mediolateral = ± 3.0 mm, dorsoventral=-2.85 mm). The shRNA sequence for mouse PRLR was 5′-GCCACCTACCATAACTGATGT-3′. The virus (titer: 5e + 12vg/ml) was infused at a rate of 200nl/min (1.0 µl/side) followed by a 10-min of rest.

Intracerebroventricular injection of PRL

A total of 20 chow diet-fed mice and 20 high fat diet-fed mice accepted the intracerebroventricular injection of PRL or its vehicle. Osmotic pumps (1001 W, RWD, China) were loaded with recombinant murine PRL (321-10, PrimeGene, China) supplemented with 0.1% BSA or vehicle solution (PBS, containing 0.1% BSA), attached to the infusion cannula. The cannula was implanted in right lateral ventricle (anteroposterior = + 0.3 mm, mediolateral = + 1.0 mm, dorsoventral=-2.5 mm) and fixed to the skull for intracerebroventricular infusion. Mice implanted with pumps received 0.8 µg of PRL per day over a 4-weekes period at a flow rate of 0.25 µl/h.

Behavioral analysisOpen field (OF)

The open filed test was performed as previously described [33]. Test was performed in a 48 cm×48 cm×36 cm open field box. Mice were placed in the center of the open field box and allowed to explore freely. The box was cleaned with 75% ethanol between the trials. We quantified the total distance the mice traveled within 5 min and the mean speed with ANY-maze software 7.0 (Stoelting Co., IL, USA).

Novel object recognition test (NORT)

The NORT was carried out as the method described previously [34]. The test was conducted in a grey open field box (48 cm width×48 cm depth×36 cm height). Mice were positioned in the open field box with two identical objects and allowed to behave freely for 10 min. Replacing one of the old objects with a new object with the same size but a different shape. Each mouse was allowed to explore the open field for 5 min. The time mice spent on exploring the novel and the familiar objects was recorded by a video tracking system. The box was cleaned with 75% ethanol after testing of each animal to diminish olfactory cues. The discrimination index was calculated as ([novel object exploration time/total object exploration time] ×100%).

Morris water maze (MWM)

Learning and memory performance was assessed by MWM according to the previously described method [35]. A 120-cm circular pool was divided into four equal quadrants (northeast, southeast, southwest, and northwest). The procedures consisted of 1 day of the visible platform test, four 60s trials per day for 5 consecutive days of the hidden platform test and a probe trial. During 5 days of hidden platform training, the platform was placed in the middle of the northeast quadrant at 1 cm underneath the water surface. The starting points for releasing mice into water were equally distributed among the four maze quadrants. Time taken by the mouse to find and climb onto the platform was recorded as latency. If the mice were unable to find the hidden platform within 60s, the mice were guided onto the platform manually and allowed a rest on the platform for 15s. In the probe trial, each mouse was placed into the water in the opposite quadrant, and the percentage of total time in target quadrant was measured. Tracking of animal movement was recorded with a DigBehv-MM tracker system (MobileDatum Co. Ltd, China).

Glucose tolerance test and insulin tolerance test

For glucose tolerance tests, basal blood glucose levels of mice were measured, followed by intraperitoneal injection with 2g/kg weight glucose after a 12-h fast. Blood glucose levels were recorded after 15, 30, 60 and 120 min using blood obtained via the tail vein. For insulin tolerance tests, basal blood glucose levels of mice were measured, followed by intraperitoneal injection with insulin (0.5 unit/kg body weight) after a 4-h fast. Blood glucose levels were recorded after 15, 30, 60 and 120 min using blood obtained via the tail vein.

Cell cultures and treatmentsPrimary microglial culture

Primary microglia were isolated from mixed glia harvested from newborn WT mice and PRL KO mice. After 12–14 days in culture, the microglia on the mixed primary glia layer were isolated by shaking the flasks. Cells were seeded into 24-well plates with DMEM/F12 complete medium plus 10% fetal bovine serum and 1% penicillin/streptomycin. Primary microglia were treated with recombinant PRL protein (321-10, PrimeGene, China) (100ng/ml) or minocycline (20µM) for 24 h before engulfment test.

Microglial engulfment test in vitro

Equal amounts of pHrodo-green conjugated synaptosomes were added to cultured WT or PRL KO microglia. After 1.5 h of incubation with synaptosomes, microglia were washed in PBS and fixed in 4% paraformaldehyde (PFA) for 10 min. After fixation, microglia were washed again in PBS and blocked with 5% bovine serum albumin (BSA) with 0.3% triton-X 100 in PBS for 1 h at room temperature. Microglia were then incubated with primary antibody overnight at 4 °C, followed by secondary fluorescent antibody staining for 1 h at room temperature. Microglia were stained with DAPI for 10 min and rinsed twice with PBS before imaging with a Leica TCS SP8-MaiTai MP confocal microscope (Leica, Germany). The whole experiment was performed in the dark.

Adult microglia isolation

Mice were anaesthetized and perfused intracardially with ice-cold Dulbecco’s phosphate-buffered saline to obtain brains. The Adult Brain Dissociation Kit (130-107-677, Miltenyi Biotec, Germany) was used to prepare a single-cell suspension after tissue dissociation, debris removal and erythrocyte removal. Microglia were further isolated from the single cell suspension using MACS Separation Columns (MS) (130-042-201, Miltenyi Biotec, Germany) and magnetic CD11b microbeads (130-093-634, Miltenyi Biotec, Germany). The isolated microglia were treated with or without a Rap1 inhibitor (GGTI298, MCE, USA) before engulfment test.

RNA sequencing and analysis

Total RNA was extracted from microglia followed by library preparation according to Illumina standard instruction (VAHTS Universal V6 RNA-seq Library Prep Kit for Illumina®, Vazyme, China). Agilent 4200 bioanalyzer was employed to evaluate the concentration and size distribution of cDNA library before sequencing with an novaseq6000 (Illumina, USA). The protocol of high-throughput sequencing was fully according to the manufacturer’s instructions (Illumina). The raw reads were filtered by Seqtk before mapping to genome using Hisat2 (version:2.0.4). The fragments of genes were counted using stringtie (v1.3.3b) followed by TMM (trimmed mean of M values) normalization. Standard DESeq2 v1.42.0 workflow was then applied to the count matrix to identify DEGs. P-values were obtained with the Wald test and corrected by the BH method. Differentially expressed genes (DEGs) were defined by adjusted p-value < 0.05 and |log2FoldChange|>0.26 (equivalent to 1.2-fold change).

Quantitative real-time PCR

In brief, total RNA was isolated by TRIzol reagent (15596018, Invitrogen, USA), and then cDNAs were synthesized using the Evo M-MLV PT Premix for PCR (AG11706, AG, China). cDNAs were amplified by quantitative real-time RT–PCR using a SYBR Green kit (AG11701, AG, China). Data were collected by a real-time PCR machine (Roche LightCycler 480II, Switzerland) and its software. Levels of gene expression were normalized relative to glyceraldehyde-3-phosphate dehydrogenase (GAPDH). The primer sequences of genes were taken from literature [36] or Primer Bank. Primer sequences were listed in Table S1.

Western blotting

Mouse hippocampus tissues were washed with ice-cold PBS and lysed in lysis solution including RIPA buffer (89900, Thermo, USA) and phenylmethyl-sulphonyl fluoride (PI0011, LEAG ENE, China) on ice for 0.5 h, and centrifuged for 0.5 h at 13,000 rpm. The supernatants were transferred into new 1.5 ml-tubes and boiled in SDS loading buffer. After SDS–PAGE, proteins were separated on 10% SDS-polyacrylamide gels and transferred to polyvinylidene difluoride (PVDF) membranes. The membranes were incubated with the following primary antibodies at 4 °C overnight: rabbit anti-Prolactin (1:1000, Abcam, ab188229), rabbit anti-Prolactin receptor (1:1000, Abcam, ab170935) and rabbit anti-β-actin (1:1000, Bioworld, AP0060). After washing with 0.1% Tris-buffered saline with Tween (TBST), the membranes were incubated with secondary antibodies (1:5000, Biosharp, BL023A) at room temperature for 1 h. The color detection was carried out with an ECL Western Blotting Detection kit (180-5001, Tanon, China), and quantified by Fiji-ImageJ software 1.0 (National Inst. Of Health, USA).

Enzyme-linked Immunosorbent Assay (ELISA)

To determine the concentrations of PRL in mouse serum, ELISA assay was performed according to the manufacturer’s instructions by using Mouse Prolactin ELISA Kit (FMS-ELM117, FMS, China). To detect the production of inflammatory factors in hippocampus, ELISA assays were performed by commercially available ELISA kits (BioLegend, China).

Immunofluorescence and image analysis

Briefly, mouse brains were fixed in 4% PFA for 24 h, then soaked in 15% and 30% sucrose solution for dehydration. We use a Leica cryostat (Leica CM1950) to cut coronal sections of 25 μm. Brain sections were incubated in 5% BSA with 0.3% triton-X 100 in PBS for 1 h at room temperature. Then sections were incubated with primary antibodies at 4 °C overnight. After washing with PBS for three times, the sections were incubated with secondary antibodies for 1 h at room temperature. Tissue sections were stained with DAPI and rinsed three times with PBS before image capture. For primary cultured cell immunofluorescence, cells were fixed with 4% PFA for 10 min, and the following procedure was the same as brain tissues. Information of antibodies used were summarized in Table S2.

Synaptic density in vivo

In brief, 25 μm sections stained with synaptic markers (Vglut1/NeuN, Vglut1/PSD95) were captured by Leica TCS SP8-MaiTai MP confocal microscope (Leica, Germany). Captured images were used to quantify the number of colocalized pre- and post-synaptic puncta. Single-channel images were used to quantify single synaptic marker density by Fiji-ImageJ software 1.0 (National Inst. Of Health, USA). Synaptic density was determined as puncta given area (276 μm×276 μm).

Fluorescent intensity

Images were converted into 8-bit to obtain total gray value in given area (562 μm×562 μm) by Fiji-ImageJ software 1.0 (National Inst. Of Health, USA). The fluorescent intensity was calculated as follows: gray value (CD68)/(Iba-1).

Cell number

For hippocampal neuron counting, the number of neurons was quantified by NeuN+ cell number in a square field (750 μm×750 μm). Immunofluorescence–stained images of NeuN+ cells were captured by a Leica TCS SP8-MaiTai MP microscope (Leica, Germany). Images were converted to 8-bit, and cells were cropped and thresholded to produce a binary (black and white) image. The number of cells were manually counted by using the cell counter plugin for the Fiji-ImageJ software 1.0 (National Inst. Of Health, USA). Five brain sections per mouse were used for neuronal counting. The number of neurons in the experimental groups was normalized to the number of neurons in the control group (cell count in experimental group/cell count in control group*100%). For hippocampal microglia counting, the number of Iba-1+ cells was divided by the total area (3072 μm×2048 μm) of the acquired field to represent cell density. The z-stack images of Iba-1 stained hippocampal slices were acquired by a Leica TCS SP8-MaiTai MP microscope (Leica, Germany). After being performed z-stack projection, the images were converted to 8-bit. Cells were cropped and thresholded to produce a binary (black and white) image, and then automatically counted using the particle analyzer plugin for the Fiji-ImageJ software 1.0 (National Inst. Of Health, USA). Three brain sections per mouse were used for microglial counting and microglia number displayed as cells per square millimeter. The threshold for NeuN and Iba-1 kept constant during image analyses.

3D microglial engulfment quantification

Z-stack images (at 0.3 μm intervals) were captured by Leica TCS SP8-MaiTai MP confocal microscope (Leica, Germany). Images were captured by choosing microglia with Iba-1 positive channel randomly. Images were processed by subtracting background using ImageJ software. Afterwards, 3D volume surface renderings of each z-stack were created using Imaris 9.0 software (Bitplane, Switzerland). Surface rendered images were used to calculate the volume of the microglia and synaptic puncta. Engulfment percentage was calculated as volume of internalized synaptic volume of microglia.

2D microglial engulfment assay in vitro

Engulfment was analyzed by calculating the fraction of the Iba-1 area overlapped by the pHrodo area. Cells were imaged using Leica TCS SP8-MaiTai MP confocal microscope (Leica, Germany).

Synaptosome isolation and pHrodo labeling

Mouse forebrains were quickly removed and homogenized in ice-cold gradient buffer (320.0 mM sucrose, 5.0 mM HEPES, 0.1 mM EDTA, pH 7.5) (HEPES, 2185833, Thermo scientific, USA; EDTA, AM9260G, Thermo scientific, USA). The homogenate was centrifuged at 1000×g for 20 min to collect supernatant, and supernatant was centrifuged at 1200×g for 10 min. The supernatant was centrifuged at 10,000×g for 10 min to obtain pellet. The pellet was resuspended in gradient buffer and loaded onto a sucrose gradient (0.8 M:1.2 M = 1:1). The layer between 0.8 M sucrose and 1.2 M sucrose was collected carefully after centrifugation at 100,000×g for 1 h. Equal volume of ultrapure water was added to dilute the collected layer before centrifuged at 100,000×g again to acquire the purified synaptosome pellet. For pHrodo labeling, synaptosomes were incubated with pHrodo Green STP ester (P36013, Thermo scientific, USA) in sodium carbonate buffer pH 9.0 for 2 h at 4 °C in the dark. After incubation unconjugated pHrodo was washed by PBS.

Statistical analysis

SPSS software 26.0 (IBM, IL, USA) and GraphPad Prism software9.0 (GraphPad, USA) were used for statistical analysis. Data were analyzed using unpaired two-tailed t test, Mann-Whitney U test, one-way ANOVA analysis and post hoc tests (Dunnett’s multiple comparisons test), two-way ANOVA analysis (followed by Sidak’s multiple comparisons test) and Kruskal-Wallis analysis. Normality of data was tested by Shapiro-Wilk test (for a sample size < 50) or Kolmogorov-Smirnov test (for a sample size ≥ 50). Heterogeneity of data was tested by F-test. Relative weight (RW) analysis was performed as the reported method [37]. All tests used for statistical analyses were mentioned in the figure legends. P values of 0.05 or less were supposed to be statistically significant.