All animal experiments were approved by the Institutional Animal Care and Use Committee of the Institute of Laboratory Animal Science of Peking Union Medical College (ILAS-PL-2014-003). Animals were provided by the Institute of Experimental Animals of the Chinese Academy of Medical Science and provided with care according to the guidelines published in the National Institutes of Health Guide for Care and Use of Laboratory Animals.

Eight-month-old female APP/PS1 Tg mice and wild-type (WT) littermates were used in this study. All Tg mice expressed the Swedish (K670N/M671L) mutation of human APP together with PS1 deleted in exon nine on a C57BL/6J background, confirmed by PCR genotyping of mouse tail tissue. Twelve Tg mice were employed to receive hNSC transplantation (NSC group), 12 Tg mice were subject to PBS transplantation (PBS group) and 12 WT mice without any treatment were used as controls (WT group). All animals had free access to food and water and were housed in cages in an environmentally controlled room with a 12-h light/dark cycle. All procedures involving animals were performed according to guidelines published in the National Institutes of Health Guide for Care and Use of Laboratory Animals.

Human NSCs were obtained from a NSC line derived from human fetal brain (Angecon Biotech, Shanghai, China), transduced with the lentivirus-mediated gene encoding green fluorescent protein (GFP). Using the procedures as we reported previously (Zuo et al., 2015), primary single cell suspensions were isolated from human cortex tissues of legally terminated embryos under the supervision of National Health and Family Planning Commission of the People’s Republic of China. NSCs were cultured in the serum-free NSC medium (Angecon Biotech, Shanghai, China) at 37°C, 5% CO₂, and 80% relative humidity, and amplified with a 7–10 day interval to the third generation. Neurosphere cultures were digested with 0.05% trypsin-EDTA (Invitrogen, Carlsbad, CA, USA) followed by trypsin inhibitor (Roche, Mannheim, Baden-Württemberg, Germany). Before transplantation, NSCs were stained for Nestin to confirm the nature of NSCs and then adjusted to a concentration of 10⁵/μl. All procedures were approved by Human Experimentation and Ethics Committee.

Tg mice were anesthetized with intraperitoneal injection of 5% chloral hydrate (0.1 ml/10 g) and then fixed on a stereotaxic apparatus. The skin was shaved and sterilized with iodine complex. The scalp was dissected in the middle line and retracted bilaterally to expose the sagittal and coronal sutures. Based on Mouse Brain Stereotaxic Atlas, bilateral hippocampal CA1 regions were selected as sites for transplantation. Two vertical bone holes were drilled symmetrically at 1.8 mm besides sagittal suture and 2.0 mm posterior to anterior fontanel with anterior fontanel as the center. A Hamilton microsyringe fixed on the stereotaxic apparatus was inserted 2.5 mm under the dura, and a 4 μl NSC suspension (at 1 × 10⁵/μl) or PBS was injected into the brain gently (>10 min). The needle was maintained in place for 5 min and then was withdrawn at a slow pace (>3 min). After repeating washes with normal saline, the wound was sewed up and covered with sterile dressings.

Behavioral assessment was performed as previously described (Zhang L. et al., 2014). The anxiety of mice was assessed using the open field test (OFT). In the test, each mouse was allowed to freely move in the open-field square apparatus (50 cm × 50 cm) divided into the peripheral and central zones. The movements of mice were monitored using a computer-driven movement tracing system (Ethovision, Noldus Information Technology, Wageningen, Netherlands). The time of mouse staying in both zones, as well as the frequency of mouse’s immobility, high mobility or immobility, were recorded. The high mobility and immobility were defined as mobility rates >60% and <20%, whereas the intermediate mobility was between 60% and 20%. After the OFT, novel object recognition (NOR) task was conducted to assess the recognition ability. In the task, two identical objects were positioned in the apparatus and the mice were allowed to explore in the field 5 min each day for three consecutive days. Twenty four hours later, one object was displaced with a novel object of different shape and color, and again mice explored the field. Each score was recorded for the mouse when its nose touched the object or its head was oriented towards the object within a 1-cm distance. The exploration time for the familiar (T_F) or novel object (T_N) during the test was collected and the recognition ability was calculated using the discrimination index (DI; Yao et al., 2013; Zhang L. et al., 2014) as the following formula: DI = (T_N − T_F)/(T_N + T_F) × 100%.

The spatial memory learning and memory was assessed using Morris water maze (MWM) test. The test was conducted in a white circular pool (100-cm diameter) divided into four equal quadrants. The pool was filled with water made opaque with non-fat milk at a temperature of 22 ± 1°C. First, mice were allowed to swim freely for 60 s in the pool with a visible platform. Then, an escape platform (6-cm diameter) was placed in the center of quadrant II, submerged 1 cm below the water surface. Each mouse was given three trials each lasting 60 s for five successive days, and in each trial the mice were allowed to swim freely to escape on the platform. The time taken to the platform (escape latency) was recorded and the values of three trials were averaged. At 24 h after the last set of trials, each mouse was given a probe test in which the platform was removed from the pool and the mouse was allowed to swim freely for 1 min. The crossings of target platform were recorded for each mouse. After completing the WMM, mice were kept in a plastic holding cage placed on an electric heater.

¹H-MRS was conducted on a 7.0T MRI scanner (Bruker PharmaScan 70/16, Bruker, Germany) coupled with a 12-cm-diameter self-shielded gradient system and a 23-mm surface coil to receive signals. Three groups of mice all underwent MRI examination at 6 weeks after transplantation. Animals were anesthetized with 5% isoflurane for induction and 2% isoflurane for maintenance. Animals were fixed on the experimental bed in prone position with the frontal tooth hooked with a tooth bar. During examination, the body temperature was kept at 36–37°C and respiratory frequency was monitored using an animal vital sign guarding system. T2-weighted images were acquired using rapid acquisition relaxation enhancement (RARE) as following parameters: TR = 4000 ms, TE = 24 ms, slice thickness = 0.8 mm, field of view (FOV) = 20 mm × 20 mm, matrix = 256 × 256, and RARE factor = 4. The obtained images were used for guiding subsequent locations of MRS examination.

Two regions of interest (ROI) sizing 20 mm³ × 20 mm³ × 18 mm³ were located on the right hippocampus and adjacent cortical region (Figure 1). The ROI location was placed on the coronal, axial and sagittal planes of mice brains. To control the consistency of ROI placements two experienced investigators simultaneously checked the location. MRS data was collected using point-resolved water suppression pulse sequence (PRESS) as following parameters: TR = 2500 s, TE = 20 ms, EF = 1000, T = 40 min. The obtained data was processed with the Bruker processing software 2D WIN-NMR (Bruker-Franzen Analytic, Germany). After baseline correction and adjustment, area under the peak of each metabolite was calculated automatically using the NUTS-NMR Utility Transform Software (AcornNMR, Livermore, CA, USA). The evaluated metabolites included NAA, Cho, Glu, mI and Cr. Because Cr is consistent in various diseases, it was used as the internal standard to calculate NAA/Cr, Cho/Cr, Glu/Cr and mI/Cr in this study.

For histological examination, mice were re-anesthetized and subject to cardiac perfusion with 10% formalin (PHs7.4, 4°C). The brain was totally removed, segmented into 5-mm-thick coronal slices and then post-fixed for at least 24 h. In particular, the slice involving the middle part of dorsal hippocampus was collected from bregma 0 ventrally to −5.0. The selected slices were dehydrated by gradient ethanol, embedded in paraffin and cut into coronal sections (6 μm). Two series of brain sections were collected from −1.0 to bregma −1.4 and from −2.0 to bregma −2.4 for staining.

Nissl staining was used to evaluate neuronal loss. Based on earlier reports (Fu et al., 2016), sections were hydrated and incubated in 1% cresyl violet acetate at 60°C for 40 min. After three washes with PBS, sections were decolorized with 95% ethanol till appearing clear. Finally, sections were dehydrated, dried and cover-slipped. To reveal Aβ depositions, immunohistochemistry (IHC) was performed based on manufacturer’s instructions. Briefly, hydrated sections were blocked with 10% goat serum, and then incubated with Aβ_1–16 antibody (6E10, 1:500, Covance/Signet Laboratories, Dedham, MA, USA) for at 4°C overnight. After three washes with PBS, sections were stained with secondary antibody (HRP-labeled anti-rabbit/mouse IgG), visualized with DAB (ZSJQ, Beijing, China), and counterstained with hematoxylin. The stained sections were captured using a Leica digital camera (TCS SP5, Leica, Wetzlar, Germany). Immunofluorescent labeling was conducted based on standard protocols as we described previously. Primary antibodies included glial fibrillary acidic protein (GFAP, 1:400, Millipore, Billerica, MA, USA), Neuronal Nuclei (NeuN, 1:500, Millipore, Billerica, MA, USA), synaptophysin (SYP, 1:200, Sigma, St. Louis, MO, USA), and microtubule associated protein 2 (MAP2, 1:400, Ruiying Biological, Suzhou, China). Primary antibodies were incubated at 4°C overnight and visualized with appropriate Alexa Fluor conjugated secondary antibodies (Invitrogen, Carlsbad, CA, USA). All sections were counterstained with 4′,6-diamidino-2-phenylindole (DAPI, ZSJQ, Beijing, China). Immunofluorescent sections were visualized using a laser scanning confocal microscope (TCS SP5, Leica Microsystems, Heidelberg, Germany).

For quantification analysis, sections were coded and the captured images were analyzed using the ImageJ software package (National Institutes of Health, Bethesda, MD, USA) by two blinded investigators. On Nissl-stained sections, five microscopic fields from the hippocampal CA1, CA3, DG regions and cortical pyramidal (PCL) and multiform cellular layers (MCL) were selected to calculate neuronal counts in each view at 40× magnifications. Whereas Aβ plaque area and intensity was calculated on 6E10-stained sections of the frontal cortex and hippocampus, the positive-staining area was calculated on SYP- and MAP2-stained sections. Large images were scanned on positive sections using the confocal microscope to observe the distribution of engrafted NSCs and calculate the cell counts in multiple brain regions.

The Aβ peptides (including soluble and insoluble Aβ1–40 and 1–42) in the cortex and hippocampus were extracted according to previous studies. Briefly, frozen brain tissue were weighed, added with freshly-prepared Trisbuffered saline (TBS; 20 mM Tris-HCl, 150 mM NaCl, PH 7.4) at 4:1 (TBS volume/brain wet weight), and homogenized on a mechanical Fluka homogenizer. The homogenate was centrifuged at 20,000 g for 30 min at 4°C. The supernatant containing soluble Aβ peptides was collected and stored at −80°C. The sediment containing insoluble Aβ peptide was rehomogenized with an equal volume of TBS plus 5 M guanidine HCl (pH 8.0) and incubated at room temperature. After re-centrifugation, the supernatant was collected and stored at −80°C. Aβ1–40 and Aβ1–42 levels were quantified by ELISA kit according to manufacturer’s instructions (Invitrogen, Carlsbad, CA, USA).

Transmission electron microscopy (TEM) was performed to observe the ultrastructural changes. Tissue samples were prepared as we described previously (Li et al., 2010). In brief, mice were re-anesthetized and perfused cardically with 0.9% sodium chlorine (4°C) followed by 2.5% glutaraldehyde in 0.01M PBS (pH 7.4, 4°C). The brains were cut into 1-mm coronal slices, the hippocampus and frontal lobe were trimmed into blocks of approximately 2 mm³ × 1 mm³ × 0.5 mm³ (length × width × thickness) and then post-fixed at 4°C until processing. The tissues were rinsed in double-distilled water, dehydrated in gradient ethanol and propylene oxide. Finally, the blocks were embedded in epoxy resin, cut into 80-nm sections, and stained with uranyl acetate and bismuth subnitrite to increase contrast. The ultrathin sections were examined under TEM (JEM-1400plus; JEOL Ltd, Japan).

All results were expressed as means ± SEM. Statistical analyses were performed using GraphPad Prism 6.0 (GraphPad Software, Inc., La Jolla, CA, USA). All continuous variables were tested to confirm that they fit a normal distribution before further analysis. Difference comparisons among three groups were analyzed with one-way ANOVA, followed by post hoc Tukey’s test (α = 0.05) to assess the difference between any two groups. A repeated measures ANOVA was used to analyze the MWM learning curve. Difference between two groups was compared using unpaired t-test. Difference was considered as significant when p < 0.05.

The GFP-labeled hNSCs were used in this study to identify the location of engrafted cells in various brain regions of APP/PS1 mice at 6 weeks after transplantation. The GFP-positive cells were widely distributed from transplantation sites (hippocampal CA1 region) to ambient regions including the corpus callosum, adjacent cortex, and even to distant regions like thalamus and external capsule (Figures 1A–G). These cells were well co-localized with DAPI staining, confirming the nature of transplanted NSCs (t-test; t_(1,78) = 1.05, p = 0.38; Figures 1H,I). To disclose the phenotype of engrafted hNSCs, glial cells and neurons were labeled by the astrocyte (GFAP) or neuronal marker (NeuN). Whereas a minority of GFP-labeled hNSCs expressed GFAP in the corpus callosum, external capsule and thalamus (Figures 1J–M), some others were co-localized with NeuN mainly in the hippocampus and cortex (Figures 1N–Q). Quantification analysis showed that most of GFP-labeled cells co-localized well with DAPI staining (94.7 ± 5.6%), wherein some cells differentiated into GFAP-astrocytes (9.8 ± 8.7%) or mature neurons (2.4 ± 4.4%; Figure 1R). These results indicate that hNSCs survived, engrafted, migrated, differentiated and matured into neural cell types of central nervous system (CNS) after transplantation in AD mice brains.

The behavioral changes were assessed in all mice before histological examination, the results of which are summarized in Figure 2. The OFT was used to assess the anxiety of AD mice. In the test, both the PBS and NSC group mice exhibited higher frequencies of immobile, mobile and highly mobile states and spent more time in the central zone and lesser time in the periphery zone than WT mice (ANOVA with post hoc Tukey’s test, all p < 0.05; Figure 2A), suggesting that AD mice were more hyperactive and anxious. However, no significant difference was observed between two groups of AD mice in any of the five parameters including mobile frequencies and time spent in each zone (all p > 0.05; Figure 2A). Thus, hNSCs did not significantly alter anxiety level in AD mice. In the NOR task, NSC group mice showed notably higher DI values than PBS group, but there were no significant difference in DI between NSC group and WT mice (ANOVA with post hoc Tukey’s test, p = 0.001 and p = 0.60 respectively; Figure 2B), suggesting that hNSCs restored the impaired recognition in AD mice.

The MWM test was used to evaluate the learning and memory of AD mice after transplantation. In the whole test, NSC group mice showed similar performance with WT mice. In training tests, WT mice showed significantly decreased escape latency from day 2 compared with PBS group (repeated measures ANOVA with post hoc Tukey’s test, p < 0.05 for day 2; p < 0.01 for day 3–5). The NSC group mice showed significantly decreased escape latency from day 3 compared with PBS group (p < 0.01 for day 3–5; Figures 2C,D). In the probe test, % of distance traveled in target quadrant, mean speed, as well as crossings of target platform was significantly reduced in the PBS group mice as compared with WT mice (ANOVA with post hoc Tukey’s test, all p < 0.05; Figures 2D,E). These parameters were significantly increased in the NSC group mice (all p < 0.05 vs. the PBS group) and there was no difference between the NSC and WT group (all p > 0.05), suggesting that the learning and memory deficits were significantly alleviated in AD mice after hNSC transplantation. These results denote that hNSCs transplanted in the hippocampus rescued learning and memory deficits in AD mice.

To assess effects of hNSCs on Aβ pathology in AD brain, the Aβ level and plaque burden were both assessed in the NSC, PBS and WT group mice. Because insoluble Aβs were intermediates from soluble Aβs to plaques, two Aβ forms were tested separately, and the results are summarized in Figure 3A. Compared with WT mice, soluble and insoluble Aβ1–40 and Aβ1–42 peptides were all significantly increased in the frontal cortex and hippocampus of PBS group mice (ANOVA with post hoc Tukey’s test, all p < 0.05). Compared with PBS group, soluble Aβ1–40 and Aβ1–42 were significantly reduced in the frontal cortex and hippocampus of NSC group mice (all p < 0.05), to the equivalent levels as exhibited by WT mice. However, insoluble Aβ1–40 and Aβ1–42 were still significantly higher in the cortex and hippocampus of NSC group mice than that of WT mice (all p < 0.05), and there was no significant difference between the PBS and NSC group mice (all p > 0.05). These results suggest that hNSCs significantly reduced soluble but not insoluble Aβ1–40 and Aβ1–42 levels in the cortex and hippocampus of AD mice.

The Aβ plaque burden was evaluated on 6E10-stained sections (Figure 3B). No Aβ-positive staining was observed in the frontal cortex and hippocampus of WT mice. Extensive Aβ plaques were observed in the brains of PBS group mice, and Aβ plaques were alleviated in the NSC group mice compared with PBS group. Quantification analysis showed that the Aβ plaque number, area and intensity was both reduced in the cortex and hippocampus of NSC group mice compared that of PBS group, but the reduced plaque number and area did not reach significance (t-test, all p > 0.05; Figure 3C). These results denote that hNSCs only slightly alleviated the Aβ plaque burden in the brains of AD mice.

Nissl staining was conducted to assess effects of hNSCs on neuronal loss in AD brain; representative images are shown in Figure 4. Compared with WT mice, PBS group mice showed typical Alzheimer’s pathology, including neuron loss, nucleus shrinkage or disappearance of Nissl bodies in the frontal cortex and hippocampus (Figure 4A). Compared with PBS group, the neuronal number was significantly increased and the cell organization was notably improved in the hippocampal CA1, CA3, DG regions and cortical PCL and MCL areas of NSC group mice (ANOVA with post hoc Tukey’s test, all p < 0.05; Figure 4B). However, no notable difference was observed in neuronal counts between NSC and WT group mice (all p > 0.05). These results imply that hNSCs improved the number and structure of neurons in AD mice brains.

Synaptic injury is another feature of AD. To assess how synapses changed after hNSC transplantation, we examined the synaptophysin immunoreactivity. As shown in Figure 5A, synaptic density was significantly reduced in the frontal cortex and hippocampus of PBS group mice compared with WT mice (ANOVA with post hoc Tukey’s test, p = 0.002 and p = 0.001 respectively). HNSCs significantly increased synaptic density in AD mice compared with PBS treatment (p = 0.034 and p = 0.013 respectively), to levels equal to WT mice (p = 0.331 and p = 0.130 respectively). Ultrastructural analysis showed that synapses were sparsely distributed and structurally loose in the frontal cortex and hippocampus of PBS group mice, whereas most synapses were densely spaced and maintained a well-structured shape in NSC group mice (Figure 5B).

Synaptic injury is often associated with damage of neuronal dendrites in AD brain. Thus, brain sections were labeled with a MAP2 antibody to assess synapto-dendritic damage (Figure 6). Results showed that NSC and WT group mice exhibited comparable levels of MAP2 immunoreactivity in the frontal cortex and hippocampus (ANOVA with post hoc Tukey’s test, p = 0.316 and p = 0.321 respectively; Figure 6A). Compared with PBS group, NSC group mice exhibited a notable increase in % area of the neuropil covered by MAP2-staining dendrites (p = 0.019; p = 0.010), and there was no significant difference in MAP2 immunoreactivity between NSC group and WT mice (p = 0.316 and p = 0.321 respectively). Ultrastructure analysis showed that unmyelinated nerves appeared enlarged in appearance and mussy in organization in the frontal cortex and hippocampus of PBS group mice, whereas the damage was notably repaired in NSC group mice. Likewise, compared with PBS group, myelinated nerves were relatively intact in myelin sheath in both NSC group and WT mice (Figure 6B).

¹H-MRS was used to assess brain metabolic changes after transplantation in AD mice, the results of which are summarized in Figure 7. Compared with WT mice, NAA and Glu peaks decreased while Cho and mI peaks increased in the frontal cortex and hippocampus of PBS group mice (Figure 7A). Compared with PBS group, NAA and Glu peaks were elevated while Cho and mI was lowered in brains of NSC group mice, to levels exhibited by WT mice. Quantification analysis showed that NAA/Cr and Glu/Cr were significantly higher in the frontal cortex and hippocampus of NSC group mice than that of PBS group (ANOVA with post hoc Tukey’s test, all p < 0.05), and there was no significant difference between NSC group and WT mice (all p > 0.05; Figure 7B). Likewise, mI/Cr was significantly higher in the cortex and hippocampus of PBS group mice than that of WT mice (p = 0.043 and p = 0.023 respectively). But, no significant difference was found in mI/Cr between NSC and PBS group mice (p = 0.829 and p = 0.776 respectively), indicating that increased mI was not significantly reduced by hNSCs. By contrast, there was no notable distinction in Cho/Cr among three groups of mice (ANOVA, F_(2,33) = 0.395, p = 0.677; F_(2,33) = 1.01, p = 0.374, respectively). These results indicated that hNSCs significantly elevated the NAA and Glu level but hardly affects the mI and Cho level in AD mice brains, suggesting that hNSC transplantation improved neuronal metabolic activity in AD brains.