Abstract
Diabetes-associated cognitive dysfunction (DACD) is increasingly
recognized as a critical complication of diabetes. The complex
pathology of DACD remains unknown. Here, we performed single-nucleus
RNA sequencing (snRNA-seq) to demonstrate unique cellular and molecular
patterns of the hippocampus from a mouse model of diabetes. More
in-depth analysis of oligodendrocytes (OLs) distinguished five
subclusters, indicating different functional states of OLs and
transcriptional changes in each subcluster. Based on the results of
snRNA-seq and experiments in vivo, we observed demyelination and
disharmony of oligodendroglial lineage cell composition in male
diabetic mice. Serum/glucocorticoid regulated kinase 1 (SGK1)
expression was significantly increased in the hippocampus OLs of male
diabetic mice, and SGK1 knockdown in hippocampus reversed demyelination
and DACD via N-myc downstream-regulated gene 1 (NDRG1)-mediated
pathway. The findings illustrated a transcriptional landscape of
hippocampal OLs and substantiated impaired myelination in DACD. Our
results provided direct evidence that inhibition of SGK1 or the
promotion of myelination might be a potential therapeutic strategy for
DACD.
Subject terms: Diabetes complications, Genetics of the nervous system,
Dementia
__________________________________________________________________
Diabetes-associated cognitive dysfunction (DACD) is increasingly
recognized as a critical complication of diabetes. Here, the authors
show that SGK1 drives hippocampal demyelination and DACD via regulating
NDRG1 phosphorylation.
Introduction
Diabetes mellitus (DM) has become a global public health problem on
account of the rising prevalence and complex complications^[58]1.
Diabetes-associated cognitive dysfunction (DACD) was common but easily
overlooked among the variety of complications of DM^[59]2,[60]3.
Epidemiological studies have established that participants with DM had
a higher risk (about 20% to 60%) for developing all-cause cognitive
impairments compared to nondiabetic individuals^[61]2,[62]4. The
cognitive dysfunction of DM individuals, especially type 2 diabetes
mellitus (T2DM), is mainly manifested in memory, executive function,
processing speed, attention and other cognitive domains. The etiology
of DACD is multi-factorial, such as brain insulin resistance,
disordered glucolipid metabolism, brain microvascular injury,
mitochondrial dysfunction, and neuroinflammation^[63]5,[64]6. However,
a randomized clinical trial has disclosed that a standard therapeutic
strategy to control blood glucose has limited beneficial effects on
cognitive decline induced by T2DM^[65]7. Therefore, clarifying the
pathological mechanism and exploring effective prevention and treatment
strategies for DACD was urgent and necessitated further
action^[66]8,[67]9.
Loss of myelinated nerve fibers induced by hyperglycemia, insulin
deficiency and obesity has been considered the driving force underlying
the occurrence and development of diabetic neuropathy, especially in
the peripheral nerves^[68]10,[69]11. Emerging evidence during the last
decade suggested that demyelination also participated in the central
nervous system (CNS) of clinical and experimental diabetic
research^[70]12,[71]13. In addition, it has been widely acknowledged
that myelin sheaths and oligodendroglial lineage cells have a critical
impact on regulating cognitive dysfunction and neurodegenerative
diseases, such as Alzheimer’s disease (AD)^[72]14. Nonetheless, direct
evidence presenting the involvement of demyelination and
oligodendrocyte (OL) status in DACD has been uncovered, and the
potential mechanism remains unknown.
Analysis of molecular signatures responses in the hippocampus, a region
key for learning, memory, and cognition ability, is an appropriate
access to elucidate the pathological process of DACD^[73]15.
Single-nucleus RNA sequencing (snRNA-seq) is a novel practical approach
that allows transcriptomic profiling of each single nucleus, identifies
cell subclusters and defines their distinguishing
characteristics^[74]16. In the present study, we performed snRNA-seq to
comprehensively characterize the composition and transcriptional
changes in the hippocampus tissue of the diabetic mice model. Based on
integrated data, we identified diverse cell types in the hippocampus
and filtrated differentially expressed genes (DEGs) in diverse cell
types. After revealing the genic heterogeneity and diversity of OLs, we
confirmed that demyelination could be a critical pathophysiological
driver of DACD progression, including oligodendrocyte precursor cell
(OPC) differentiation and OL apoptosis. Intriguingly, the myelin
deficits of DACD were associated with serum/glucocorticoid regulated
kinase 1 (SGK1) and N-myc downstream-regulated gene 1 (NDRG1) signal
pathway in OLs.
Results
snRNA-seq analysis reveals the cell composition of the hippocampus in db/db
mice
Firstly, we evaluated the blood glucose and cognitive performance of
two diabetic models, db/db mice and high-fat diet (HFD)-fed mice, to
study DACD. Body weight and fasting blood glucose of control (db/m) and
diabetic (db/db) groups were recorded at 20 weeks of age (Supplementary
Fig. [75]1a, b). Then we performed the nest building test, novel object
recognition (NOR) test, and morris water-maze (MWM) test to evaluate
spontaneous behaviors, learning and memory performance. We found that
db/db mice presented significant cognitive impairment, including lower
nest score, lower discrimination index, longer escape latency, shorter
time in the target quadrant, and lesser passing time for crossing the
platform location (Supplementary Fig. [76]1c–i). Consistent with
previous studies^[77]17,[78]18, the swimming speed of db/db mice was
slower than that of db/m mice, possibly affecting the outcomes of the
MWM test. We further demonstrated the motor activity (walking speed) of
db/db mice was not damaged compared to the db/m mice (Supplementary
Figs. [79]1j, k). Similar results were also detected in the HFD-fed
mice (Supplementary Fig. [80]2).
We collected fresh hippocampus from db/m and db/db mice for snRNA-seq
analysis (Fig. [81]1a), and a total of 23,016 single nuclei were
sequenced. After filtering low-quality nuclei, 20,690 high-quality
nuclei were obtained for subsequent biological analysis (12,104 from
the control group and 8586 from the diabetic group, respectively).
Reducing single nuclei populations in db/db mice might be associated
with brain atrophy induced by diabetes. Consistently with other
research^[82]19–[83]21, we observed the distinct hippocampus atrophy in
db/db mice by evaluating the hippocampus thickness and weights
(Supplementary Figs. [84]1l, m).
Fig. 1. snRNA-seq analysis revealed the cell composition of the hippocampus
in db/db mice.
[85]Fig. 1
[86]Open in a new tab
a Schematic depicting our study design. b t-SNE plot presenting the
different cell types of the hippocampus. c Heat map showing
well-established gene marker expression to identify cell clusters. d
t-SNE plot and e bar chart presenting the relative abundances of
different cell types from the db/m and db/db mice. f Bar chart
presenting the relative abundances of different cell types from the
diabetic and control mice by flow cytometry, n = 4 mice per group. g
Representative FACS plots of OLs and OPCs in hippocampus from db/m and
db/db mice. h Quantification of FACS analysis from (g), n = 4 mice per
group. i Representative FACS plots of OLs and OPCs in hippocampus from
HFD or Chow-fed mice. j Quantification of FACS analysis from (i), n = 4
mice per group. Data are presented as mean ± SEM, and analyzed by
unpaired two-tailed Student’s t-tests. Source data are provided as a
Source data file. HFD, high-fat diet; OLs, oligodendrocytes; OPCs,
oligodendrocyte precursor cells.
To validate the nucleus sequencing data, we checked the Malat1
expression, a common nuclear-located RNA, and found high Malat1 levels
of each cluster in the major seven cell types of our nucleus sequencing
data (Supplementary Fig. [87]3a). Based on the acknowledged cell
type-specific markers, 27 clusters were classified into 7 major cell
types, neurons (Gad1, Gad2, Atp1b1, and Rtn1), OLs (Mal, Mbp, Mobp, and
Mog), OPCs (Pdgfra), endothelial cells (Bsg and Flt1), microglia (C1qa,
C1qb, and C1qc), fibroblasts (Col3a1, Dcn, and Fstl1), and astrocytes
(Aqp4, Slc1a2, and Slc1a3) (Figs. [88]1b, c and Supplementary
Fig. [89]3b–h). snRNA-seq results uncovered the dramatic shifts of
cellular composition in the two groups. Compared to the control group,
the relative abundance of OLs was remarkably decreased, while microglia
and astrocytes were increased in the diabetic group (Fig. [90]1d, e).
To verify the fidelity of the snRNA-seq results, we subsequently
conducted flow cytometry of hippocampus from two diabetic mouse models
(Supplementary Fig. [91]4a). We found that the results of the two
methods were consistent in the numbers and variations of multiple cell
types, including neurons, OLs, OPCs, microglia, and astrocytes
(Figs. [92]1f–j and Supplementary Fig. [93]4b, c).
Diabetes altered gene expression in diverse cell types of the
hippocampus in db/db mice. Based on the snRNA-seq data, we analyzed the
specific genes in diverse cell types in order to investigate the
pathomechanism of DACD. Overall, 5,817 genes were up-regulated, and
2699 genes were down-regulated in two groups (Supplementary
Fig. [94]5a). Top 50 differentially expressed genes (DEGs, 25 up- and
25 down-regulated) of diverse cell types between control and diabetic
hippocampal cells were ascertained and shown using heat maps
(Supplementary Figs. [95]5b–h). Kyoto Encyclopedia of Genes and Genomes
(KEGG) pathway enrichment analysis predicted pathways of diverse cell
types involved in diabetes based on identified DEGs. All kinds of cells
of hippocampus are primarily associated with oxidative phosphorylation,
Parkinson’s disease (PD), Huntington’s disease (HD), AD, thermogenesis,
retrograde endocannabinoid signaling, and ribosome (Supplementary
Fig. [96]6a). We detailed demonstrated the TOP 10 enrichment pathways
in different cell types (Supplementary Fig. [97]6b–h).
We also conducted the cell-cell communication analysis to diagram
cellular interaction networks. Neurons, OLs and OPCs were the most
active cell types in db/m and db/db mice (Supplementary Fig. [98]7a).
Moreover, the numbers of ligand-receptor pairs in db/db mice were
increased entirely compared with those in db/m mice. The top
ligand-receptor pairs in two groups were exhibited, such as
Nrxn3-Nlgn1, Nrxn1-Nlgn1, Ncam1-Ncam2, and Cntn1-Nrcam (Supplementary
Figs. [99]7b, c).
Interestingly, we noticed a dramatic change in OLs, known as the
central myelinating cells of the CNS, indicating myelination might be
closely related to the development of DACD. Therefore, we further
analyzed the expression levels of myelin-related genes in different
cell types. Mbp (a common marker of myelin) and Sox10 (effective
regulator of myelination) RNA expression levels were significantly
decreased in OLs of db/db mice, compared with db/m mice.
Transcriptional levels of other myelin-related genes, including Myrf
and Dlg4, were exhibited in Supplementary Fig. [100]8.
Subcluster analysis indicates the difference of OLs of the hippocampus in
db/db mice
Considering the possible important role of demyelination on DCAD
progression, we further performed a particular subcluster analysis of
OLs to reveal heterogeneity within OLs. According to the published
database^[101]22,[102]23, the above-mentioned OPCs and OLs were
identified into 5 subclusters to depict the differentiation and mature
process of OLs. The 5 subclusters included (1) OPCs (Cspg4 and Pdgfra),
(2) differentiation-committed oligodendrocyte precursors (COP) and
newly formed oligodendrocytes (NFOL) (Bmp4, Casr, and Tcf7l2), (3)
myelin-forming oligodendrocytes (MFOL) (Ablims and Tmen141), (4) mature
oligodendrocytes 1 (MOL1) (Ptgds, Opalin, and Il33), (5) MOL2 (Klk6,
Apod, and Cd59a). We showed the heat map of marker expression in
diverse OL subclusters and performed the significance test in MOL1 and
MOL2. There was a highly significant difference in cell markers among
the MOL1 and MOL2 groups (all p < 0.01; Fig. [103]2a, b; Supplementary
Fig. [104]9a, b). Fluorescence in situ hybridization (FISH) was
performed to assess the colocalization of the OL marker (Mbp) and MOL
markers (Ptgds and Klk6). The results showed that MOL1 populations
(Ptgds^+) were distinct from MOL2 (Klk6^+) (Supplementary
Fig. [105]9c). Obviously, the populations of different oligodendroglial
lineage cell subclusters in db/db mice were decreased sharply compared
to the db/m mice. Compared to the db/m mice, the relative abundance of
COP + NFOL, MFOL and MOL2 was significantly decreased in the db/db
mice. On the other hand, the relative abundance of OPC and MOL1 was
significantly increased in the db/db mice (Figs. [106]2c, d and
Supplementary Fig. [107]9d). Pseudo-temporal analysis was performed to
identify the dynamics of OL lineage cells in DACD progression, showing
the lineage transitions of OPC, COP and NFOL to MOLs in db/db mice
hippocampus (Fig. [108]2e).
Fig. 2. Subcluster analysis indicated the difference in OLs of the
hippocampus in db/db mice.
[109]Fig. 2
[110]Open in a new tab
a Heat map showing well-established gene marker expression to identify
oligodendrocyte subclusters. b t-SNE plot presenting the
oligodendrocyte subclusters of the hippocampus. c Bar chart and (d)
t-SNE plot presenting relative abundances of oligodendrocyte
subclusters from the db/m and db/db mice. e Differentiation
trajectories of these subclusters were analyzed using Slingshot. Violin
plot showing the expression of myelin-related genes (f) Mbp, (g) Sox10,
(h) Myrf and (i) Dlg4 in the oligodendrocyte subclusters. j Heat map
showing presenting the representative KEGG enrichment pathways in
oligodendrocyte subclusters. Data are analyzed by unpaired two-tailed
Student’s t-tests for (f–i) and hypergeometric test for (j). Source
data are provided as a Source data file. COP, differentiation-committed
oligodendrocyte precursors; KEGG, Kyoto Encyclopedia of Genes and
Genomes; MFOL, myelin-forming oligodendrocytes; MOL, mature
oligodendrocytes; NFOL, newly formed oligodendrocytes; OPC,
oligodendrocyte precursor cell.
The integrated data of subcluster analysis presented unique
transcriptional profiles and cellular heterogeneity of oligodendroglial
lineage cells as expected (Supplementary Fig. [111]10). Top 30 DEGs (15
up- and 15 down-regulated) of different oligodendroglial lineage cell
subclusters between db/m and db/db hippocampus were determined and
demonstrated using heat maps. Similar with the myelin-related genes
alterations in OLs, Mbp gene was down-regulated in MOL1 and MOL2 of
db/db mice, compared to db/m mice (Fig. [112]2f). Sox10 expression
levels were reduced in COP + MFOL, MOL1 and MOL2, Myrf expression
levels were reduced in MOL1 and MOL2, Dlg4 expression levels were
reduced in OPC, COP + MFOL and MOL1 of db/db mice, compared to db/m
mice (Fig. [113]2g–i). KEGG analysis of oligodendroglial lineage cell
subclusters revealed abundant pathways involved in diabetes, such as
ribosome, PD, oxidative phosphorylation, HD, thermogenesis, and AD
(Fig. [114]2j). We also exhibited the TOP 10 enrichment pathways in
each subcluster (Supplementary Fig. [115]11).
Demyelination is involved in DACD of diabetic mice
The snRNA-seq results demonstrated that demyelination was involved in
DACD, therefore, we detected myelination-related indicators in diabetic
mice. Firstly, Luxol fast blue (LFB) staining was performed to verify
reduced myelination in the corpus callosum, hippocampus and cortex of
db/db mice (Fig. [116]3a). In addition, we used immunofluorescence
analysis and western blot analysis to evaluate the MBP in the brain of
db/db mice. The results showed that MBP levels were significantly
decreased both in the hippocampus and cortex of the db/db group
compared to the db/m group (Fig. [117]3b–e). Finally, we measured the
SOX10 (effective myelination regulator) and PSD95 (vital synaptic
protein) levels using western blot analysis. Analogously, both SOX10
and PSD95 expressions were reduced in the hippocampus and cortex of the
diabetic group compared to the control group (Fig. [118]3f–i).
Fig. 3. Myelination was reduced in the hippocampus of db/db mice.
[119]Fig. 3
[120]Open in a new tab
a Representative images of LFB staining in db/m and db/db mice. Scale
bar: 200μm (corpus callosum, hippocampus) and 50μm (cortex). b
Representative immunofluorescence images of MBP expression in db/m and
db/db mice. Scale bar: 200μm and 50μm (hippocampus), and 50μm (cortex).
c Quantification of immunofluorescence images from (b), n = 4 mice per
group. d Western blot analysis of the expression of MBP in db/m and
db/db mice. e Quantification of protein bands from (d), n = 5 mice per
group. f Western blot analysis of the expression of SOX10 in db/m and
db/db mice. g Quantification of protein bands from (f), n = 5 mice per
group. h Western blot analysis of the expression of PSD95 in db/m and
db/db mice. i Quantification of protein bands from (h), n = 5 mice per
group. Data are presented as mean ± SEM, and analyzed by unpaired
two-tailed Student’s t-tests. Source data are provided as a Source data
file. LFB, Luxol fast blue.
Analogous demyelinating features were captured in the HFD-fed mice
(Fig. [121]4a–i). Transmission electron microscopy (TEM) visualized the
ultra-structural myelination status of HFD-fed mice; the myelin sheath
in the diabetic group was thin, loose and irregularly shaped compared
to those in the control group (Fig. [122]4j, k). The g-ratio (ratio of
the inner to the outer boundary of a myelinated axon) in HFD-fed mice
were higher than in Chow-fed mice, suggesting diabetes contributed to
thinner myelin sheath (Fig. [123]4l). Moreover, we implemented TEM to
observe the synapse status in the hippocampus at the ultra-structural
level. The HFD-fed mice were found synaptic deficits, such as fewer
asymmetric synapses per field and thinner postsynaptic density,
compared to the control mice (Fig. [124]4m–o). These results confirmed
that demyelination might participate in the pathological process of
DACD.
Fig. 4. Myelination was reduced in the hippocampus of HFD-fed mice.
[125]Fig. 4
[126]Open in a new tab
a Representative images of LFB staining in Chow or HFD-fed mice. Scale
bar: 200μm (corpus callosum, hippocampus) and 50μm (cortex). b
Representative immunofluorescence images of MBP expression in Chow or
HFD-fed mice. Scale bar: 200μm and 50μm (hippocampus), and 50μm
(cortex). c Quantification of immunofluorescence images from (b),
(n = 4 mice per group). d Western blot analysis of the expression of
MBP in Chow or HFD-fed mice. e Quantification of protein bands from
(d), n = 5 mice per group. f Western blot analysis of the expression of
SOX10 in Chow or HFD-fed mice. g Quantification of protein bands from
(f), n = 5 mice per group. h Western blot analysis of the expression of
PSD95 in Chow or HFD-fed mice. i Quantification of protein bands from
(h), n = 5 mice per group. j Representative electron micrographs of
myelinated axons of the hippocampus from different groups. Scale bar:
500 nm. k Bar chart showed sheath thickness from different groups. l
Scatter plot showed g-ratio from different groups, n = 5 images from 5
mice per group. m Representative electron micrographs of synapses of
the hippocampus from different groups. Scale bar: 500 nm and 200 nm. n
Bar chart showed the numbers of asymmetric synapses from different
groups, n = 5 mice per group. o Bar chart showed PSD thickness from
different groups, n = 5 images from 5 mice per group. Data are
presented as mean ± SEM, and analyzed by unpaired two-tailed Student’s
t-tests. Source data are provided as a Source data file. HFD, high-fat
diet; LFB, Luxol fast blue.
To further investigate the alteration of oligodendroglial lineage cells
in diabetic brains, we measured the population marker by OLIG2, OPCs
marked by PDGF receptor α (PDGFRα), and mature OLs marked by CC1. The
number of PDGFRα^+Olig2^+ OPC remained unchanged in the corpus
callosum, hippocampus and cortex between the db/db mice and db/m mice
(Supplementary Fig. [127]12). The number of CC1^+OLIG2^+ OLs was
reduced in the corpus callosum, hippocampus and cortex of db/db mice
(Fig. [128]5a, b). To probe the differentiation and/or apoptosis in
OLs, we administered 5-ethynyl-2’deoxyuridine (EdU) to trace the new
OLs and performed a TUNEL assay of OLs, respectively. The density of
EdU^+CC1^+ cells significantly decreased in the corpus callosum,
hippocampus and cortex of db/db mice (Fig. [129]5c, d). Furthermore,
the proportion of apoptotic OLs was evidently increased in db/db mice
(Fig. [130]5e, f). Consistent findings were observed in HFD-fed mice
(Supplementary Fig. [131]13a–h). All the results collectively indicated
the disturbances of OPC differentiation and OL survival in diabetic
brains.
Fig. 5. Differentiation and apoptosis of OLs were disturbed in the
hippocampus of diabetic mice.
[132]Fig. 5
[133]Open in a new tab
a Representative immunofluorescence images of CC1 and OLIG2-labeling
OLs in db/m and db/db mice. Scale bar: 50μm (corpus callosum), 200μm
and 50μm (hippocampus), 50μm (cortex). b Quantification of
immunofluorescence images from (a), n = 4 mice per group. c
Representative immunofluorescence images of CC1 and EdU-labeling new
OLs in db/m and db/db mice. Scale bar: 50μm and 25μm (corpus callosum),
200μm and 25μm (hippocampus), 50μm and 25μm (cortex). d Quantification
of immunofluorescence images from (c), n = 4 mice per group. e
Representative immunofluorescence images of CC1 and TUNEL-labeling
apoptotic OLs in db/m and db/db mice. Scale bar: 50μm and 25μm (corpus
callosum), 200μm and 25μm (hippocampus), 50μm and 25μm (cortex). f
Quantification of immunofluorescence images from (e), n = 4 mice per
group. g Representative immunofluorescence images of mGFP-positive
myelin in HFD-fed NG2-CreERT; Tau-mGFP mice. Scale bar: 50μm (corpus
callosum), 200μm (hippocampus), 50μm (cortex). h Representative
immunofluorescence images of mGFP-positive myelin in HFD-fed
PLP-CreERT; mT/mG mice. Scale bar: 50μm (corpus callosum, hippocampus,
cortex). Data are presented as mean ± SEM, and analyzed by unpaired
two-tailed Student’s t-tests. Source data are provided as a Source data
file. CA, cornu ammonis; DG, dentate gyrus; EdU,
5-ethynyl-2’deoxyuridine; HFD, high-fat diet; OLs, oligodendrocytes;
OPCs, oligodendrocyte precursor cells; TMF, Tamoxifen.
Recent research has shown that myelin changes dynamically in the adult
brain^[134]24. OPCs differentiate into newly generated OLs and then
form myelin sheaths, which might accompany the degeneration of
pre-existing myelin sheaths. Therefore, we labelled and traced newly
generated myelin and pre-existing myelin separately to understand the
myelin dynamics in diabetic mice. The NG2-CreERT; Tau-mGFP mice line
was employed to label the newly formed myelin sheaths^[135]25.
Tamoxifen (TMF) can induce Tau-mGFP alleles recombination in NG2^+
OPCs, and the mGFP expression is controlled by the Tau promoter.
Considering Tau is highly expressed in OLs while rarely low in OPCs,
membrane-bound GFP (mGFP, green) can only be visualized in newly
generated OLs and myelin sheaths. 7-week-old NG2-CreERT; Tau-mGFP mice
were administered intragastrically with TMF and then fed with HFD for
12 consecutive weeks to induce DACD. Compared to control mice, the
mGFP^+ myelin sheaths were significantly decreased in the corpus
callosum, cortex, and hippocampus of HFD-fed mice (Fig. [136]5g and
Supplementary Fig. [137]13i). In addition, we obtained PLP-CreERT;
mT/mG mice line to label and trace the pre-existing myelin
sheaths^[138]26. As a vital myelin-related protein, PLP is only
expressed in mature OLs but not newly-formed OLs. After recombination
induced by TMF, mGFP can express in pre-existing OLs and myelin
sheaths. Consistently, the area of mGFP^+ myelin was reduced in the
corpus callosum, cortex, and hippocampus of HFD-fed mice (Fig. [139]5h
and Supplementary Fig. [140]13j). Together, these results demonstrated
that new myelin generation was inhibited in combination with
accelerating degradation of pre-existing myelin in diabetic brains,
ultimately leading to myelin dysfunction and dyshomeostasis in DACD
mouse models.
Increased SGK1 expression is associated with demyelination in diabetic mice
We concentrated on the DEGs data of snRNA-seq analysis to identify
particular genes that modulate demyelination in diabetic mice. As the
top DEG in OLs, Sgk1 was found to be present predominantly in OLs of
db/db mice and significantly up-regulated in OLs, OPCs, and microglia
of db/db mice (Fig. [141]6a, b). We next analyzed the distribution of
Sgk1 across different OLs subclusters in which finding Sgk1 was
overexpressed in all kinds of OLs subclusters of db/db mice
(Fig. [142]6c, d), suggesting the crucial role in the development of
OLs and maintenance of myelin myelination. We also performed western
blot analysis and immunofluorescence analysis to assess the protein
level of SGK1. Consistent with the snRNA-seq analysis results, we
verified that SGK1 was observably increased in both the hippocampus and
cortex of db/db mice compared to those of db/m mice (Fig. [143]6e, f).
Costaining for SGK1 and PDGFRα, we found the SGK1 fluorescence
intensity within PDGFRα^+ cells increased in the hippocampus and cortex
of db/db mice compared to the db/m mice (Fig. [144]6g–h). As expected,
the SGK1 intensity within CC1^+ cells also increased in the db/db mice
(Fig. [145]6i, j). The expression of SGK1 in OPCs and OLs was
noticeably higher in HFD-fed mice than in the chow-fed group assessed
by western blot and immunofluorescence experiments (Supplementary
Fig. [146]14a–f ). The expression of SGK1 in neurons did not been
detected a significant up-regulation in db/db mice and HFD-fed mice
(Supplementary Fig. [147]14g–j). Based on the results, we decided to
focus on SGK1 for the following exploratory experiments. We
hypothesized that up-regulated SGK1 expression in OLs might serve as a
target to attenuate cognitive dysfunction in diabetes.
Fig. 6. SGK1 expression was upregulated in the hippocampus in db/db mice.
[148]Fig. 6
[149]Open in a new tab
a t-SNE plot and (b) violin plot presenting the expression of Sgk1 in
the different cell types from the db/m and db/db mice. c t-SNE plot and
(d) violin plot presenting the expression of Sgk1 in the
oligodendrocyte subclusters from the db/m and db/db mice. e Western
blot analysis of the expression of SGK1 in control and diabetic mice. f
Quantification of protein bands from (e), n = 5 mice per group. g
Representative immunofluorescence images of SGK1 and PDGFRα expression
in db/m and db/db mice. Scale bar: 200μm and 50μm (hippocampus), 50μm
(cortex). h Quantification of immunofluorescence images from (g), n = 4
mice per group. i Representative immunofluorescence images of SGK1 and
CC1 expression in db/m and db/db mice. Scale bar: 200μm and 50μm
(hippocampus), 50μm (cortex). j Quantification of immunofluorescence
images from (i), n = 4 mice per group. Data are presented as
mean ± SEM, and analyzed by unpaired two-tailed Student’s t-tests.
Source data are provided as a Source data file.
Specific SGK1 knockdown in OLs alleviates DACD in db/db mice
To clarify the effects of SGK1 in OLs on the anti-demyelination of
DACD, we administered intrahippocampal recombinant adeno-associated
virus (AAV) injection, which specifically inhibited SGK1 expression in
OLs of db/db mice (Fig. [150]7a, b). Meanwhile, we implemented
remyelination treatment (clemastine, 10 mg/kg, 14 d) for db/db mice to
verify demyelination was the pivotal etiology of DACD. Western blot
analysis results showed that SGK1 expression of the hippocampus was
markedly decreased in the db/db+shSgk1 group compared to the db/db+shNC
group, whereas clemastine treatment did not regulate the SGK1
expression in the db/db+Cle group (Fig. [151]7c, d). We detected the
cell types expressing EGFP by staining with cell-specific markers for
OLs (CC1), OPCs (PDGFRα), neurons (NeuN), microglia (IBA1), and
astrocytes (GFAP). As shown in Fig. [152]7e, EGFP^+ cells were mainly
colocalized with CC1-labeled OLs rather than OPCs, neurons, microglia,
or astrocytes. The expression rate of EGFP in OLs ranged from 59.77% to
72.93% in the hippocampus. With the high specificity and efficiency, we
confirmed the AAV exhibited satisfactory efficacy in transduction of
OLs (Supplementary Fig. [153]15a, b). Then, we evaluated whether AAV
regulated expression levels of SGK1 in diverse cells. The SGK1
intensity in CC1^+ cells was distinctly down-regulated in the
db/db+shSgk1 group compared to the db/db+shNC group (Fig. [154]7f and
Supplementary Fig. [155]15c). Besides, immunofluorescence results
showed that little change in SGK1 expression of OPCs, neurons,
microglia, and astrocytes between the two groups (Supplementary
Figs. [156]15d–g).
Fig. 7. Specific SGK1 knockdown in OLs alleviated DACD in db/db mice.
[157]Fig. 7
[158]Open in a new tab
a Schematic depicting the design of the AAV vector and the experimental
timeline. b Fluorescence image of the section that expressed AAV in the
hippocampus. Scale bar, 200μm. c Western blot analysis of the
expression of SGK1 in different groups. d Quantification of protein
bands from (c), n = 3 mice per group. e Representative
immunofluorescence images of EGFP and cell markers (CC1, PDGFRα, NeuN,
IBA1, GFAP) in db/db+shSgk1 groups. Scale bar: 20μm and 10μm. f
Representative immunofluorescence images of SGK1 and CC1 in db/db+shNC
and db/db+shSgk1 groups. Scale bar: 20μm and 10μm. MWM test parameters,
(g) escape latency, (h) representative track plots in the learning
phase, (i) time spent in the target quadrant, (j) times of mice passed
through the platform location, (k) representative track plots in probe
phase, n = 10 mice per group. Data are presented as mean ± SEM, and
analyzed by one way ANOVA followed by the Bonferroni post hoc test.
*p < 0.05, **p < 0.01, between db/m vs. db/db; ^#p < 0.05, between
db/db+shNC vs. db/db+shSgk1; ^&p < 0.05, between db/db vs. db/db+Cle,
in (g). Source data are provided as a Source data file. AAV,
adeno-associated virus; Cle, clemastine; DACD, Diabetes-associated
cognitive dysfunction; MWM test, Morris water-maze test; NOR test,
novel object recognition test; OLs, oligodendrocytes.
After 4 weeks of AAV injection, metabolic parameters and a battery of
behavior tests were performed, which confirmed the neuroprotective
function of knocking down SGK1 and clemastine treatment in cognitive
deficit. SGK1 knockdown and clemastine therapy did not influence the
body weights and fasting blood glucose of db/db mice (Supplementary
Figs. [159]16a, b). The nest building test results indicated that both
SGK1 knockdown and clemastine therapy improved spontaneous behaviors
and motor function in db/db mice (Supplementary Fig. [160]16c). Both
SGK1 knockdown and clemastine therapy enhanced the nonspatial memory
assessed by NOR test (Supplementary Fig. [161]16d). In the MWM test,
mice in db/db+shSgk1 group were found more excellent cognitive
performance compared to the db/db+shNC group, including shorter escape
latency in the study phase, more extended time in the target quadrant
and more times for crossing the platform location in probe phase
(Fig. [162]7g–k). The results also presented that the clemastine
therapy notably improved the cognitive dysfunction in the db/db+Cle
group as expected. There were no significant differences in swim speeds
and walking speeds in the db/db mice (Supplementary Figs. [163]16e, f).
Specific SGK1 knockdown in OLs promotes myelination via inhibiting NDRG1
phosphorylation in db/db mice
We observed demyelination of the hippocampus during the pathogenesis of
DACD. We next evaluated the impact of SGK1 or clemastine therapy on
myelination in diabetic mice. The MBP immunofluorescence showed the
fluorescence intensity of MBP in the hippocampus was increased in the
db/db+shSgk1 group compared to the db/db+shNC group. On the other hand,
clemastine therapy also improved the MBP expression in the hippocampus,
compared to the db/db group (Fig. [164]8a, b). We received consistent
results of MBP protein expression levels with western blotting
(Fig. [165]8c, d). Similar results were found in the western blotting
analysis of PSD95 expression (Fig. [166]8e, f).
Fig. 8. Specific SGK1 knockdown in OLs reversed demyelination in db/db mice.
[167]Fig. 8
[168]Open in a new tab
a Representative immunofluorescence images of MBP expression in
different groups. Scale bar: 200μm and 50μm. b Quantification of
immunofluorescence images from (a), n = 4 mice per group. c Western
blot analysis of the expression of MBP in different groups. d
Quantification of protein bands from (c), n = 3 mice per group. e
Western blot analysis of the expression of PSD95 in different groups. f
Quantification of protein bands from (e), n = 3 mice per group. g
Representative electron micrographs of myelinated axons of the
hippocampus from different groups. Scale bar: 500 nm. h Bar chart
showed sheath thickness from different groups. i Scatter plot showed
g-ratio from different groups. n = 5 images from 5 mice per group. j
Representative electron micrographs of synapses of the hippocampus from
different groups. Scale bar: 500 nm and 200 nm. k Bar chart showed the
numbers of asymmetric synapses from different groups, n = 5 mice per
group. l Bar chart showed PSD thickness from different groups, n = 5
images from 5 mice per group. Data are presented as mean ± SEM, and
analyzed by one way ANOVA followed by the Bonferroni post hoc test.
Source data are provided as a Source data file. Cle, clemastine; OLs,
oligodendrocytes.
We also performed the TEM to assess the ultra-structural myelination
status of db/db mice. Similarly with the HFD-fed mice, the myelin
sheath was thin and aberrant and the g-ratio was higher in the db/db
group in comparison to those in db/m mice (Fig. [169]8g–i). Myelin
damage in the hippocampus was extensively restored in db/db+shSgk1
group and db/db+Cle group, compared with db/db+shNC group and db/db
group, respectively, as implied by lower g-ratio (Fig. [170]8g–i).
Moreover, the db/db mice were found synaptic deficits, including fewer
asymmetric synapses per field and thinner postsynaptic density,
compared to the db/m mice (Fig. [171]8j–l). The specific SGK1 knockdown
in OLs and clemastine treatment could reverse the synaptic damage as
expected (Fig. [172]8j–l). These findings showed SGK1 knockdown and
clemastine therapy could reverse the demyelination and synaptic
deficits induced by diabetes.
The reverse of hypermyelination phenotype in the SGK1 knockdown db/db
mice might relate to transformations in the OLs pool. Therefore,
immunofluorescence was implemented to quantify populations of OPCs and
OLs. The numbers of OPCs did not change between the db/db+shNC group
and db/db+shSgk1 group, while the quantity of CC1-labelled OLs was
markedly increased in db/db+shSgk1 group (Fig. [173]9a, b). The number
of new OLs (Ki67^+CC1^+ cells) was increased in the hippocampus after
the SGK1 knockdown. Meanwhile, SGK1 knockdown restrained the apoptotic
OLs (TUNEL^+CC1^+ cells) induced by diabetes and diminished the cleaved
caspase-3 (CC3, an apoptosis executioner) expression in OLs
(Fig. [174]9c, d). These results collectively indicated that SGK1
played an important role in regulating OL differentiation and
maintenance.
Fig. 9. Specific SGK1 knockdown in OLs inhibited NDRG1 phosphorylation to
promote myelination in db/db mice.
[175]Fig. 9
[176]Open in a new tab
a Representative immunofluorescence images of PDGFRα/OLIG2-labeling
OPCs and CC1/OLIG2-labeling OLs in db/db+shNC and db/db+shSgk1 groups.
Scale bar: 200μm and 50μm. b Quantification of immunofluorescence
images from (a), n = 4 mice per group. c Representative
immunofluorescence images of CC1/Ki67-labeling new OLs,
CC1/TUNEL-labeling apoptotic OLs, and CC1/CC3-labeling apoptotic OLs in
db/db+shNC and db/db+shSgk1 groups. Scale bar: 200μm and 25μm. d
Quantification of immunofluorescence images from (c), n = 4 mice per
group. e Western blot analysis of the expression of p-NDRG1 and NDRG1
in different groups. f Quantification of protein bands from (e), n = 3
mice per group. g Representative immunofluorescence images of p-NDRG1
in CC1-labeling OLs in db/db+shNC and db/db+shSgk1 groups. Scale bar:
20μm and 10μm. h Quantification of protein bands from (g), n = 6 mice
per group. i Western blot analysis of the expression of BDNF in
different groups. j Quantification of protein bands from (i), n = 3
mice per group. k Schematic illustration elucidated the molecular
mechanism underlying SGK1-related anti-demyelination efficacy in DACD.
Data are presented as mean ± SEM, and analyzed by one way ANOVA
followed by the Bonferroni post hoc test. Source data are provided as a
Source data file. Cle, clemastine; DACD, diabetes-associated cognitive
dysfunction; HFD, high-fat diet; MWM test, Morris water-maze test; NOR
test, novel object recognition test; OLs, oligodendrocytes.
We further investigated the molecular mechanism underlying SGK1-related
anti-demyelination efficacy in DACD. Studies have shown that NDRG1 is
enriched in OLs and participates in myelination, plasticity, and axonal
support. Considering that NDRG1 was one of the SGK1-specific targets
and phosphorylation of NDRG1 was regulated by SGK1^[177]27, we also
assessed NDRG1 and phospho-NDRG1 expression in different groups.
P-NDRG1 was significantly elevated in the db/db group compared to the
db/m group, and p-NDRG1 was reduced in the db/db+shSgk1 group compared
to the db/db + NC group (Fig. [178]9e, f). According to the
immunofluorescent staining results, the down-regulation of p-NDRG1
occurred in CC1-labeled OLs but not in OPCs, neurons, microglia, and
astrocytes of db/db+shSgk1 group (Fig. [179]9g, h and Supplementary
Fig. [180]17). There was no difference of p-NDRG1 between db/db group
and db/db+Cle group. BDNF, one of the most common neurotrophins,
contributed to remyelination, plasticity synapses, neuronal
development, and differentiation^[181]28. A growing body of evidence
suggested that BDNF acts as an intermediary between gene or drug
therapy and improved neuroplasticity. Hence, we measured the BDNF
expression of hippocampus in the 5 groups using the western blotting
analysis. Compared to the control group, BDNF level was significantly
decreased in the db/db group, whereas BDNF level was significantly
increased in the db/db+shSgk1 group and db/db+Cle group (Fig. [182]9i,
j). We also verified that the SGK1/NDRG1/BDNF pathway was activated in
the HFD-fed mice, another DACD mouse model (Supplementary
Fig. [183]18). In summary, based on snRNA-seq and validation, this
study demonstrated increased SGK1 of OLs inhibited BDNF expression via
SGK1/NDRG1 pathway and aggravated demyelination in the hippocampus,
eventually leading to DACD in diabetic mice (Fig. [184]9k).
Discussion
In this study, we used snRNA-seq technology to systematically delineate
a high-resolution transcriptional landscape of hippocampus tissue at
male diabetic mice status. DEGs analysis between the control and
diabetic groups demonstrated gene transcription in each cell cluster
underwent widespread diabetes-induced variation. Due to the unique
shift of OLs, we further distinguished oligodendroglial lineage cell
subtypes and characterized unique expression patterns of each OL
subtype. We observed the abnormal dynamic changes of oligodendroglial
lineage cells and myelin deposition in the hippocampus of male db/db
and HFD-fed mice. Moreover, we found that SGK1, a key molecule involved
in demyelination, was significantly highly expressed in OLs of the
hippocampus in male diabetic mice. Combined with the experiments in
vivo, we demonstrated increased SGK1 expression induced demyelination
and DACD in a diabetic mouse model. Specific SGK1 knockdown in OLs
rescued myelin impairment and DACD via inhibiting NDRG1 phosphorylation
in male db/db mice. Therefore, these findings emphasized the
relationship between demyelination and DACD progression. We also
provided insights about the feasibility of targeting SGK1 and
facilitating remyelination as a promising therapeutic approach of DACD.
The flourishing development of snRNA-seq transcription approaches has
brought unprecedented resolution to examining of tissue samples;
however, the application of this technique in hippocampus samples of
diabetes animal models was still limited. Our team executed single-cell
RNA sequencing (scRNA-seq) to explore the heterogeneity of hippocampal
tissue in db/db mice and identified a subpopulation of pro-inflammatory
disease-associated microglia in 2022^[185]29. Xiang et al. also
depicted transcriptional profile of hippocampal tissue in db/db mice
using scRNA-seq in 2024^[186]30, whereas the relevant report was brief
and did not provide a detailed or comprehensive exhibition of scRNA-seq
analysis data. Compared to scRNA-seq, the snRNA-seq technique bypassed
the cell dissociation step and released the cell nucleus from intact
cells using special detergents to capture and profile brain tissue
cells comprehensively. For tissues that cannot easily dissociate into
single-cell suspensions, such as brain tissue, snRNA-seq may be a
better choice as it can minimize possible gene expression changes
caused by dissociation. In addition, snRNA-seq analysis has higher
sensitivity and generates more cell type
classifications^[187]16,[188]31. Hence, we performed the snRNA-seq to
characterize the composition and transcriptional changes in the
hippocampus tissue of the diabetic mice model in order to fill the gap
in this aspect.
Consistent with the previous research about the hippocampus using
snRNA-seq, 7 cell types were identified in the present
study^[189]32,[190]33. We conducted the DEGs analysis between the
control group and the diabetic group on the basis of different cell
types. There were significant differences in the transcriptional
characteristics of various cells, with neurons and astrocytes being the
most prominent. KEGG analysis suggested that except for the commonly
enriched pathways, the degree of enrichment in different pathways
varied among different cell types. Our findings revealed specific cell
types of corresponding KEGG pathways and provided more evidence for the
function of individual cell types. It was worth noting that substantial
experiments in vitro and in vivo were needed to verify further the
diversity of the enrichment pathways in the future. Intriguingly, we
noticed fewer cells in the hippocampus of the db/db mice than of the
control group, which could be associated with diabetes-induced brain
atrophy^[191]19,[192]20.
Myelin sheath that enfolds larger axons was formed by OLs in the CNS
and Schwann cells in the peripheral nervous system^[193]34. An OL
formed a myelin sheath at the internodes of several adjacent axons to
maintain information exchange and support the cellular function of the
nervous system^[194]35. Previous studies have shown that myelin
degradation occurs in aging and neurodegenerative disorders (such as
AD, PD, and amyotrophic lateral sclerosis). For example, myelin
dysfunction accelerated Aβ plaque formation, which interfered with its
maintenance and support effects on nervous system cells in the AD
model^[195]14,[196]36,[197]37. CNS myelin sulfatide deficiency induced
disease-associated microglia and astrocytes, AD risk genes expression,
and AD-like cognitive deficit in a mouse model of adult-onset myelin
sulfatide deficiency^[198]38. In summary, demyelinating injuries and
myelin dysfunction were known to contribute to hallmark pathologies of
AD and cognitive dysfunction.
Previous studies reported myelin damage induced by hyperglycemia and
dyslipidemia played a critical role in the pathogenic mechanisms of
diabetic peripheral neuropathy (such as sciatic nerve^[199]39, sural
nerve^[200]40) and diabetic optic neuropathy^[201]41. A few studies
measured myelin damage in the CNS at diabetic status. The latest
neuroimaging study found that insulin resistance and insulin level were
associated with altered myelin content, assessed by myelin water
fraction in cognitively unimpaired adults^[202]12. Another study showed
myelin is critically damaged in the cerebral cortex of the
streptozotocin-induced diabetic rat model through the detailed
lipidomic analysis of myelin^[203]13. Given the interaction between
peripheral nerve demyelination and diabetes, it was reasonable to
surmise that demyelination also occurred in the central nervous system
and finally resulted in neurodegeneration and cognitive impairment
under hyperglycemic conditions.
In the current study, we focalized the dynamic change of OLs at the
transcriptional level, highlighting the myelin-associated etiology of
DACD by analyzing OLs subclusters. Consistent with the previous reports
about the oligodendroglial lineage cells of brain tissue using
snRNA-seq, we identified 5 distinct populations throughout OPC
differentiation and OL maturation in the current
study^[204]22,[205]23,[206]36. Due to the brain tissue being collected
from adult mice at the age of 20 weeks old in our study, populations of
OPCs, COPs, NFOLs and MFOLs were considerably lower compared with those
in juvenile mice, while mature OLs (MOL1 and MOL2) were higher.
Diabetes distinctly altered OLs heterogeneity in the hippocampus. Our
results elucidated molecular and cellular changes of OL subtypes in
DACD via single-nucleus expression profiling analysis. We also
identified abnormal expression genes in OL subtypes of DACD and
characterized individual gene functions. Most of the myelination of the
human brain is accomplished by the end of the second year of life,
after which myelin is gradually lost with aging^[207]42,[208]43. In the
adult brain, myelin generation still occurs to a lesser extent, OPCs
maintain a dynamic balance in quantities and effectively respond to
myelin loss by regulating proliferation and
differentiation^[209]24,[210]44. After myelination, mature OLs require
more energy to maintain the lipid-rich myelin membrane. However, a
changing metabolic microenvironment, such as elevated lipid and iron
levels, may make mature OLs vulnerable to damage and even
apoptosis^[211]45,[212]46. Mature OLs death and accompanying myelin
loss lead to neuronal damage and functional deficits by suppressing the
axonal neuroprotective function. We observed impaired OPC
differentiation and aggravating OL apoptosis as characteristic
pathological changes in diabetes mouse brains. Our results indicated
that dysfunction of OLs and insufficient myelin regeneration were
closely related to the onset and progression of DACD.
Based on the DEGs analysis between the db/m and db/db groups, Sgk1
caught our attention due to its unique expression pattern and
clustering characteristics. As the top DEG in hippocampal OLs of db/db
mice, Sgk1 was present predominantly in OLs and significantly
up-regulated in hippocampal OLs of db/db mice. Previous studies
confirmed that SGK1 was expressed at elevated levels in
diabetes^[213]47 and neurodegenerative diseases such as AD^[214]48 and
PD^[215]49, and was associated with myelin formation^[216]50.
Therefore, we verified the unique expression feature of SGK1 in the
hippocampus of diabetic mice and demonstrated the increased SGK1
aggravated the demyelination and cognitive impairment in the mice.
SGK1, a member of the AGC protein kinase family, had deleterious
effects induced by glucocorticoids and aldosterone in the circulation.
SGK1 expression was stimulated by glucocorticoids and
mineralocorticoids, insulin, insulin-like growth factor 1, interleukin
6, transforming growth factor-β, follicle-stimulating hormone, and
hepatic growth factor^[217]51. SGK1 was well acknowledged to play a
pivotal role in maintaining cellular physiological activities, and the
elevated SGK1 expression was associated with a variety of diseases,
such as diabetes^[218]47, hypertension^[219]52, neurodegenerative
diseases^[220]48 and cancers^[221]53. SGK1 modified tau pathology via
formation of SGK1-GSK-3β-tau complex, resulting in neurodegeneration
and poor cognitive performance in HFD-treated tauopathy model
mice^[222]54. Increased SGK1 induced pro-inflammatory properties of
astrocytes and microglia, dopamine neuron degeneration via the multiple
inflammatory pathways (NF-κB and NLRP3 inflammasome) in PD animal
models^[223]49. SGK1 was overexpressed in the adipose tissue,
intestine, vascular smooth muscle and myocardial tissue of diabetes
models. Abnormally increased SGK1 inhibited insulin secretion via
upregulating voltage-sensitive Kv1.5 channel and activating Na^+/K^+
ATPase during plasma membrane repolarization^[224]55,[225]56. SGK1
aggravated fluid retention and hypertension, induced deposition of
matrix proteins resulting in diabetic complications. For example, SGK1
expression stimulated by hyperglycemia promoted renal Na^+ transporters
(such as epithelial sodium channel, sodium chloride cotransporter and
Na^+/K^+-ATPase), which led to secondary hypertension and renal
fibrosis^[226]57. Increased SGK1 facilitated Na^+ entry via the
Na^+/H^+ exchanger 1 pathway, disrupting ion homeostasis and resulting
in myocardial fibrosis and cardiac insufficiency^[227]58,[228]59.
Consistent with previous clinical and experimental research, our
findings showed SGK1 was a critical modulator of diabetes and diabetic
complications.
The current study was meaningful in demonstrating myelin damage of the
hippocampus and elucidated the mechanism underlying SGK1 and
demyelination in DCAD using a diabetic mice model. The majority of Sgk1
was co-localized with Mbp at mRNA level in the corpus callosum of adult
male rats exposed to acute stress^[229]60. Another study reported SGK1
regulated pro-myelination indicators (BDNF, MBP, and Krox20) and was
involved in remyelination in Schwann cells during peripheral nerve
injury^[230]50. As a SGK1-specific target, NDRG1 was enriched in
myelinating OLs, modulated myelin damage and maintain myelin
homeostasis^[231]61,[232]62. We found that abnormal SGK1 overexpression
regulated OL differentiation, maturation and apoptosis, and then
exacerbated demyelination via the NDRG1-meidated pathway, which
provided powerful evidence of the modulation of SGK1 on myelination. We
highlighted remyelination, and SGK1 may be potential therapeutic
targets in nerve regeneration and DACD.
However, there were some limitations in our current study. First, the
current study only described the transcriptional features in the mice
at 20 weeks of age. Hippocampal snRNA-seq of db/db mice at different
time points were needed to supplement to completely understanding of
pseudotime trajectory toward the OLs in DACD progression. Second, our
results were drawn from observations of male diabetic mice, possibly
leading to gender bias in our conclusion. Moreover, considering the
differences between animal and human hippocampus, further research was
necessary to verify this information in other animal models and
patients with DACD.
In summary, the current study provided a more in-depth insight into the
transcriptional signatures and diversity of neuronal, glial cell
populations in db/db mice of DACD progression. We emphasized that
oligodendrogenesis and myelination played pivotal roles in DACD. In
addition, our findings indicated that regulating the SGK1/NDRG1 pathway
in OLs could reverse demyelination and DACD. Our work illuminated the
molecular pathomechanism between demyelination and DACD and identified
SGK1 as a potential therapeutic target for future antidementia
medication development.
Methods
Animals
All mouse experiments were performed according to the Guidelines for
the Care and Use of Laboratory Animals from the National Institutes of
Health and were approved by the Animal Experimentation Ethics Committee
of Shandong Provincial Hospital Affiliated to Shandong First Medical
University (approval number: 2021-120). BKS.Cg-Dock7^m + /+Lepr^db/J
mice were purchased from Changzhou Cavens Laboratory Animal Co., Ltd
(Jiangsu, China). Wild-type C57BL/6 J mice were purchased from Beijing
HFK Bioscience Co., Ltd. The NG2-CreERT; Tau-mGFP and PLP-CreERT; mT/mG
mice lines were gifted from Professor Feng Mei of the Third Military
Medical University.
8-week-old male heterozygous Lepr^db/m (db/m) mice were identified as
controls and homozygous Lepr^db/db (db/db) mice were diabetic.
8-week-old male wild-type C57BL/6 J mice, NG2-CreERT; Tau-mGFP and
PLP-CreERT; mT/mG mice were fed with a standard diet (10% Fat; Research
Diets, D12450B) or a HFD (60% Fat; Research Diets, D12492) for 12 weeks
to establish the DACD mouse model. Body weight and fasting blood
glucose were recorded. After undergoing behavioural experiments, all
mice were sacrificed for subsequent biochemical tests at 20 weeks of
age. All mice were housed in a standard facility with an SPF
environment under the 12 h light/12 h dark cycle at a room temperature
of 22 °C ± 2 °C, humidity of 55% ± 5%, with ad libitum access to
irradiated food and sterilized water.
Behavioral testing
Nest building test
Nest building test was performed to detect the spontaneous mouse
behaviors^[233]63. Each mouse was individually housed overnight (12 h)
in a new standard cage with 6 pieces of nest material (5 cm × 5 cm).
The nest was evaluated according to the scoring standard: 4,
ball-shaped or cup-shaped nest with a cover; 3, cup-shaped or
bowl-shaped nest with sides; 2, platform-shaped nest in the cage; 1, no
visible nest.
NOR test
The NOR test apparatus consisted of a square open field
(45 cm × 45 cm × 45 cm). After the adaptation, each mouse was placed in
the apparatus with two plastic object of the same color, size and shape
on the bottom for 5 min. After 2 h, one of the objects was replaced by
a novel object with different color, size, and shape and then, the
mouse was placed in the apparatus for 5 min again. The discrimination
index was calculated as a ratio of the time spent exploring the novel
object to the total time spent exploring the novel and familiar objects
using Video Tracker software (Anymaze, Varese, Italy)^[234]17.
MWM test
MWM test was known as a popular method to detect spatial learning and
memory of mice^[235]3. The MWM system consisted of a circular tank
(1.2 m diameter, 0.5 m height) filled with opaque white-colored water
(22.0 ± 1.0 °C) and a submerged platform (10 cm diameter, 1 cm under
the water surface). During the learning phase, each mouse was released
into the pool from pseudo-randomly selected starting positions to find
the submerged platform within 60 s. If the mouse failed to locate the
submerged platform, it can be guided to the platform and stand in 10 s.
Learning trials were performed four times daily, 1 h apart for 5
consecutive days. During the probe phase, the submerged platform was
removed, and each mouse was released into the pool to swim for 60 s.
All tests were executed at fixed times (12:00 pm to 06:00 pm). Testing
parameters were recorded and calculated using Video Tracker software
(Anymaze, Varese, Italy).
snRNA-seq and analysis
Nucleus isolation and snRNA-seq
Mice were deeply anesthetized with isoflurane, and hippocampus tissues
were rapidly dissected for snRNA-seq. Each group consisted of a mixture
of the hippocampus of 6 mice. Collected fresh tissues were quickly cut
to pieces, snap-frozen by liquid nitrogen and stored at −80 °C.
Genedenovo Biotechnology Co., Ltd (Guangzhou, China) administered
nucleus isolation, sequencing and bioinformatic analysis under the
guidance of 10X Genomics. Briefly, the samples were homogenized with
ice-cold homogenization buffer (0.25 m sucrose, 5 × 10^−3 m CaCl[2],
3 × 10^−3 m MgAc[2], 10 × 10^−3 m Tris-HCl (pH 8.0), 0.1 × 10^−3 m
EDTA, 1X protease inhibitor, and 1U μL^−1 RiboLock RNase inhibitor),
the homogenates were filtered through a 70 × 10^−6 m cell strainer and
the cell nuclei were collected from a graded iodixanol solution
series^[236]21. The cell nuclei were resuspended in resuspension buffer
(0.04% bovine serum albumin, 0.2 U μL^−1 RiboLock RNase inhibitor,
500 × 10^−3 m mannitol and 0.1 × 10^−1 m phenylmethanesulfonyl fluoride
in PBS). After the quality control (the cell debris and large clumps
were removed after the cell nuclei were filtered through a 40 × 10^−6 m
cell strainer), the concentration was adjusted to 700-1200 nuclei μL^−1
after manually assessed using trypan blue counterstaining and a
hemocytometer. After gel beads-in-emulsion generation, libraries were
generated and sequenced using the Chromium Next GEM Single Cell 3′
Reagent Kit V3.1 on the 10X Genomics GemCode Single-cell instrument
according to the instruction from the manufacturer^[237]31,[238]64.
Quality control and snRNA-seq data preprocessing
Raw sequencing data were converted and preprocessed by 10X Genomics
Cell Ranger software V3.1.0 with the default parameters and aligned to
the pre-mRNA reference (Ensemble_release 100, Mus musculus). Droplets
with low-quality barcodes and unique molecular identifiers (UMI) were
excluded. We also filtered cells with abnormal parameters using the
DoubletFinder V2.0.3, such as gene counts out of the normal range
(500-4000 per cell), UMIs ≥ 8000, or mitochondrial gene percent ≥ 10%.
Each cell gene expression by total expression was normalized by the
global-scaling normalization method to produce normalized counts. The
expression matrix was scaled and integrated.
Cell clustering
Graph-based clustering of snRNA-seq data was performed using Seurat
V3.1.1. Cells were clustered using the Louvain method to maximize
modularity, and these data were finally visualized using t-distributed
stochastic neighbor embedding (t-SNE), which is a powerful tool for
dimensionality reduction that is well-suited to visualizing
high-dimensional data. According to the cell-type marker genes, 27
clusters were corresponded to 7 cell types.
Cell-cell communication analysis
To explore the difference in cell-cell communication in hippocampus
cells of db/db mice, expression abundance of ligand-receptor analysis,
number of ligand-receptor analysis, and number of significantly
enriched ligand-receptor analysis were executed using cellphoneDB V2.0.
We constructed a cell-cell communication network based on the number of
significantly enriched ligand-receptor pairs^[239]65.
DEGs analysis
DEGs were identified based on the following criteria:
[MATH: log2fold
change>0.36 :MATH]
, gene expression in >25% of cells in the target cluster, and p
value < 0.05. Relevant heat maps and violin maps were drawn using R
package V.4.2.2. The DEGs were also analyzed for the KEGG statistically
enriched pathways.
Cell lineage and pseudotime inference
Accurate lineage inference is a uniquely robust and flexible tool in
identifying dynamic temporal gene expression. We ordered the OLs
subclusters in pseudotime utilized Slingshot, a novel method for
inferring cell lineages, and pseudotime was used from single-cell gene
expression data^[240]66.
TMF administration
To induce Cre recombination, TMF (Selleck, S1238) was dissolved in
sterile corn oil (Beyotime, ST2308). Then, 7-week-old male NG2-CreERT;
Tau-mGFP and PLP-CreERT; mT/mG mice received TMF solution at a
100 mg/kg/day dose for successive 4 days through oral gavage^[241]26.
After 3 days of adaptation, these mice were fed with HFD to induce
diabetes.
Clemastine administration
To investigate the demyelination in the hippocampus of diabetes, mice
were administrated clemastine, a drug identified as promising for
remyelination^[242]67. 18-week-old db/db mice were treated with
10 mg/kg clemastine through intraperitoneal injection for 14
consecutive days (Selleck, S1847). Based on previous studies, the dose
of clemastine was confirmed to promote myelination effectively^[243]68.
The mice of the clemastine treatment group received behavioral tests
and tissue collection at 20 weeks of age, which was consistent with
other groups.
Stereotaxic virus injection
Recombinant adeno-associated virus (AAV) was constructed for specific
knockdown of SGK1 in mice OLs (OBIO Technology, Shanghai, China).
pAAV-MBP-EGFP-miR30shRNA(Sgk1)-WPRE or
pAAV-MBP-EGFP-miR30shRNA(NC)-WPRE were constructed for follow-up
experiments. To construct the AAV vectors, the following sequence
targeting the Sgk1 mouse gene was used: 5′-CGGCTGAGATGTACGACAATA-3′.
16-week-old mice were deeply anesthetized with isoflurane and fixed on
the stereotaxic head frame (RWD Life Science, Shenzhen, China). Virus
(2 μl, 10^9 vg) were injected bilaterally into the hippocampal region
(at ± 1.5 mm mediolateral, −2 mm anteroposterior, −2 mm dorsoventral
from the bregma) using a laboratory microsyringe at a rate of
0.1 μL/min. After injection, the microsyringe needle was left for 5 min
and then carefully and slowly removed. Behavioral tests and tissue
collection were conducted 4 weeks after the viral injection.
Flow cytometry
20-week-old db/db and HFD-fed mice were anesthetized and sacrificed,
and hippocampus tissues were rapidly collected and kept on ice.
According to the instructions of the High Activity Adult Mouse and Rat
Brain Enzymatic Digestion Kit (RWD, DHABE-5003), hippocampus tissues
were digested by the Single Cell Suspension Dissociator (RWD, DSC-410)
to maximize cell viability. We then resuspended the single cells in the
ice-cold HBSS for further flow cytometric staining.
Single cells were incubated with mouse FcR blocking reagent (STARTER,
S0B0599) for 15 min at room temperature. After being treated with
LIVE/DEAD^TM Fixbable Near IR (780) Viability Kit (ThermoFisher,
[244]L34992) for 1 h at room temperature, cells were stained with
fluorochrome-conjugated antibodies directed at cell surface antigens
for 1 h at room temperature in the dark. For intracellular staining,
cells were fixed and permeabilized for 30 min at 4 °C in the dark by
using the eBioscience™ Foxp3/Transcription Factor Staining Buffer Set
(ThermoFisher, 00-5523-00). Then cells were washed and blocked prior to
being incubated with relevant antibodies for 1 h at room temperature in
the dark. Cell solutions were centrifuged at 450 g for 5 min at 4 °C
and finally resuspended in fresh FACS buffer. Stained cell suspensions
were detected utilizing Cytek Aurora 5-laser Spectra Analyzer (Cytek
Biosciences, Shanghai, China), and the data were analyzed by Flowjo
V10.8.1 software.
Immunofluorescence
Anesthetized mice were transcardially perfused with 0.01 M PBS followed
by 4% paraformaldehyde. The whole brain was rapidly dissected and
soaked in paraformaldehyde for 24 h and dehydrated in 30% sucrose. Then
the tissue was embedded in optimal cutting temperature compound (OCT,
SAKURA, 4583) and cut into 20 µm coronal brain sections by the cryostat
microtome (Thermo Fisher Scientific, CryoStar NX50). The sections were
antigen retrieval and blocked with 5% bovine serum albumin (Servicebio,
[245]GC305010). Sections were incubated with primary antibodies
overnight at 4 °C. The sections were then probed with corresponding
second antibodies at room temperature for 2 h. The sections were
counterstained with 4',6-diamidino-2-phenylindole (DAPI, Beyotime,
P0131) for identification of cell nuclei. Finally, the sections were
examined and photographed by a fluorescence microscope (Nikon, Eclipse
C1) with constant acquisition parameters, in order to compare different
samples in the same experiment. The relative density of MBP
immunostaining was measured and quantified from a constant field of
view (FOV) utilizing by ImageJ software^[246]69. All antibodies used in
the present study were listed in Supplemental Table [247]1.
LFB
The whole brain was rapidly dissected and soaked in paraformaldehyde
for 24 h. Then the fixed tissue was embedded in paraffin and cut into
4 µm coronal brain sections by microtome (Leica, CM1860). The sections
were stained with LFB solution (Servicebio, G1030) at 60 °C for 4 h.
After washing with PBS, differentiation with lithium carbonate
solution, rehydration with a graded ethanol series, sections were
immersed in xylene for 10 min and mounted with neutral glue. Finally,
the sections were examined and photographed by an optional microscope
(Nikon, E100).
FISH
Mice were deeply anesthetized with isoflurane, and fresh whole brains
were collected, fixated in paraformaldehyde, dehydrated in 30% sucrose
solution and embedded in an OCT compound. Brain tissues were cut into
10 µm slices. According to the manufacturer’s instructions, a multiplex
RNAscope was performed to detect gene expressions^[248]22,[249]60. The
fluorescent photographs were captured by an optional microscope (Nikon,
Eclipse ci).
EdU labelling and detection
To assess differentiation, 18-week-old diabetic and control mice
received intraperitoneal injections with EdU (50 mg/kg, Beyotime,
ST067) for 5 consecutive days. Mice were deeply anesthetized with
isoflurane and sacrificed at 20 weeks of age. According to the
manufacturer’s recommendations, brain slices with cellular EdU were
visualized using a BeyoClick™ EdU Cell Proliferation Assay (Beyotime,
C0075S). The fluorescent photographs were captured by an optional
microscope (Nikon, Eclipse ci).
TUNEL assays
To assess the apoptosis, TUNEL assays were executed using the TUNEL
BrightRed Apoptosis Detection Kit (Vazyme, A113) according to the
manufacturer’s direction. In brief, mice were deeply anesthetized with
isoflurane, and fresh whole brains were collected, frozen sections were
permeabilized and blocked with 5% bovine serum albumin for 1 h at room
temperature. After equilibration, brain slices were treated with TdT
working solution for 1 h at 37 °C. Finally, the slices were incubated
with relative primary antibodies for further immunofluorescence
experiments.
TEM
Mice were deeply anesthetized with isoflurane, and fresh hippocampus
tissue was immediately collected and soaked in a pre-cooling fixative
solution (2.5% glutaraldehyde) at 4 °C for 18 h. Then the tissue was
post-fixed in 1% osmium tetroxide at room temperature for 2 h in the
dark. After rehydration with a graded ethanol and acetone series, they
were embedded in an acetone-resin mixture. The resin blocks were
sectioned ultrathin at 60 to 80 nm on an ultrathin microtome; ultrathin
sections were collected on 150-slot copper grids and stained with 2%
uranyl acetate and 2.6% lead citrate solution. Finally, the sections
were examined and photographed under an electron microscope (Hitachi,
HT7800). The myelinated fiber g-ratio is the ratio of the inner to the
outer diameter of the axon myelin sheath.
Western blotting
Mice were anesthetized, and fresh hippocampus and cortex tissue was
rapidly dissected for Western blotting. The brain tissue was
homogenized in Protein Extraction Reagent (Beyotime, P0033) to exact
protein. The protein concentrations were measured by bicinchoninic acid
assay (Beyotime, P0009). 50 ug of total protein was resolved by 10%-15%
sodium dodecyl sulfate-polyacrylamide gel electrophoresis and
transferred to sodium dodecyl sulfate-polyacrylamide gel
electrophoresis membranes. The membranes were incubated with
appropriate antibodies, and the immunoreactive bands were detected and
visualized by the chemiluminescence system (Bio-Rad, chemidoc MP
imaging system). All antibodies used in the present study were listed
in Supplemental Table [250]1.
Statistical analysis
Sample size was chosen based on experience with the used experimental
models. No statistical method was used to determine animal’s sample
size. All results were presented as mean ± SEM. A normality test was
carried out whenever possible to choose appropriate statistical
analysis. Data were analyzed by GraphPad Prism V8.0 software. A
two-tailed p value less than 0.05 was considered statistically
significant. The details for each experiment, the number of replicates
and statistical specifications are indicated in the figure legends,
results, and the Source Data file.
Reporting summary
Further information on research design is available in the [251]Nature
Portfolio Reporting Summary linked to this article.
Supplementary information
[252]SUPPLEMENTARY INFORMATION^ (9.8MB, pdf)
[253]Reporting Summary^ (3.2MB, pdf)
[254]Transparent Peer Review file^ (180.7KB, pdf)
Source data
[255]Source Data^ (16.1MB, xlsx)
Acknowledgements