ADSCs were isolated as described previously [11]. In brief, white adipose tissue from the inguinal region of rats was placed in a digestion solution containing 0.1% collagenase type I and subjected to continuous agitation at 37 °C for 1 h. The reaction was stopped by the addition of culture medium, and the cell suspension was centrifuged at 2000 rpm for 10 min. The acquired cells were resuspended and filtered through 100-µm filters. After being washed twice, the cellular precipitate was resuspended in cell culture medium and cultured in 25 mm${}^2$ cell flasks at 37 °C in a 5% CO${}_2$ incubator. The cells were cultured in DMEM containing 1 g/L glucose (#11885084, Gibco, Carlsbad, CA, USA), 10% fetal bovine serum (#10099141C, Gibco, Carlsbad, CA, USA) and 1% penicillin/streptomycin at 37 °C in 5% CO${}_2$, and the medium was changed twice per week. The cells were observed daily under an inverted phase contrast microscope. The cells were detached with 0.25% trypsin/0.01% EDTA and replated for passaging. We observed that third-generation ADSCs adipogenic and osteogenic differentiation potential, and the immunophenotype of surface markers was examined by flow cytometry. ADSCs of this generation and the following two generations were used in the experiments.

All studies were performed in accordance with the National Institute of Health guidelines and were granted formal approval by the Research Ethics Committee of Shandong Institute of Endocrine & Metabolic Diseases. Male 8-week-old SD rats were exposed to 12 h of light/darkness under constant temperature and humidity and had free access to water and regular food. All of the rats had similar body weights and growth statuses before streptozotocin (STZ) administration. Twenty rats were fasted for 16 h (with free access to water) and were then intraperitoneally injected with 60 mg/kg STZ (#S8050, Solarbio, Beijing, China). After the injection, the rats were allowed to roam free for three days to stabilize the results. Fasting blood glucose levels were taken from the tail tip and measured daily using a glucometer until both blood glucose measurements were above 16.7 mmol/L. At this time, the rats were diagnosed with STZ-induced diabetes, which was set as day 0 of the experiment. Fourteen days later, the rats were randomized again. One group was the ADSC treatment group (ADSC group), and these rats received 2.0 $\times$ 10${}^6$ ADSCs through the tail vein, while the other group was the diabetes group (DM group) and received the same dose of PBS. Body weight and blood glucose levels in the two groups were dynamically monitored weekly. After 6 weeks of ADSC treatment or control treatment, the rats were anesthetized after being fasted; blood was collected from the jugular vein, and tissue samples were collected prior to sacrifice.

A TRIzol kit (#15596026, Thermo Fisher Scientific, Waltham, MA, USA) was used to extract total RNA from liver tissue according to the manufacturer’s instructions. The total RNA concentration was determined by a NanoDrop microvolume spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). The ratio of OD260 to OD280 was considered an indicator of RNA purity. In addition, we used agarose gel electrophoresis to ensure RNA integrity and exclude gDNA contamination.

RNA-seq analysis was performed by Health Biotechnology (Shanghai, China). Briefly, the NEBNextTM Ultra II Directional RNA Library Prep Kit was used to construct sequencing libraries for validation and quantification. The Illumina HiSeq X platform was then used for 150 bp peer sequencing. The expression level of each gene was estimated by calculating sequences mapped to reference sequences (reads), of which the most commonly used method is read per million transcripts per kilobyte (RPKM). In the DEGseq v1.20.0 software package, the differentially expressed genes (DEGs) were analyzed using the random sampling (MARS) model. Differences in gene expression among samples were considered significant when all of the following conditions were met: fold change (FC) $>$2 or 0.5, false discovery rate (FDR) q value 0.05 and RPKM $>$2 for at least one sample.

The human hepatoma cell line HepG2 was provided without Mycoplasma contamination and authenticated using short tandem repeat (STR) profiling analysis by Procell Life Science & Technology (Wuhan, China) under the catalog number CL-0103. The cells were maintained in DMEM supplemented with 10% fetal bovine serum and 50 U/mL penicillin/streptomycin in an incubator under a humidified atmosphere of 5% CO${}_2$ and 95% air at 37 °C. We used H${}_2$O${}_2$ treatment to simulate oxidative stress in vitro and observed the effects of ADSC culture supernatant and hepatocyte growth factor (HGF), which is a cytokine secreted by ADSCs, on inflammation and oxidative stress.

HepG2 cells were plated in 6-well plates at 1 $\times$ 10${}^6$ cells/well and grown in medium for 24 h. Then, the cells were pretreated with or without ADSC supernatant, HGF (10 ng/mL), or the extracellular regulated protein kinase (ERK)1/2 inhibitor U0126 (2 µM) for 2 h and were cultured with H${}_2$O${}_2$ (600 µM) for another 6 h to determine the effect of ADSCs on ROS-induced inflammation.

To obtain whole protein extracts, rat livers were treated with lysis buffer containing protease and phosphorylation inhibitors. The stained membranes were scanned using the Odyssey CLx imaging system (LI-COR, Lincoln, NE, USA) to measure the fluorescence at 800 nm and 700 nm. The fluorescence intensity of the bands was also quantified with the Odyssey CLx imaging system, and the results are expressed as the fold change relative to the control after normalization to the respective internal control. Antibodies against P-ERK1/2 (#4370), ERK1/2 (#4695), P-P38 (#4511), P38 (#8690), P-c-jun N-terminal kinase (JNK) (#9255), JNK (#9252), P-STAT3 (#9145), STAT3 (#9139), GAPDH (#5174) and $\beta$-tubulin (#86298) were obtained from Cell Signaling Technology (CST, Danvers, MA, USA) and used for western blotting.

Using the QuantiTect Reverse Transcription Kit (QIAGEN, Hilden, Germany), total RNA in each sample was reverse transcribed into cDNA. Using the Light Cycler system, the resultant cDNA was used as a template for real-time quantitative PCR. Nonspecific amplification was used to perform melt curve analysis. The data were normalized to $\beta$-actin, which was used as an internal control for normalization. The oligonucleotide primers used in this study are shown in Table 1.

The gene expression data were input into Ingenuity Pathways Analysis (IPA) software (Ingenuity Systems; https://ingenuity.com/) to establish a causal network. Upstream regulator analysis, downstream effect analysis and mechanistic network analysis were performed to determine likely causal mechanisms, and the p value was network-bias-corrected.

The analyses were performed using Microsoft Excel 2021 and GraphPad Prism 9.0 software (San Diego, CA, USA). The results are reported as the mean $\pm$ SD. Comparisons of between groups were performed using unpaired t tests. p 0.05 was considered significant for the qRT‒PCR and western blot data.

Fig. 1a–d shows that ADSCs had adipogenic and osteogenic differentiation potential and were positive for CD29 and CD90 (90.75% and 95.38%) and negative for the leukocyte antigen CD45 (1.49%). These cells met the criteria for ADSCs. The weekly blood glucose levels and the oral glucose tolerance test (OGTT) results before the rats were sacrificed are shown in Fig. 1e,f. There was no significant change in the blood glucose levels between the two groups; however, the area under the curve (AUC) of the OGTT was reduced in the ADSC group compared to the DM group. Moreover, ADSC treatment alleviated hepatic injury in our previous study [11].

Fig. 1.

Characteristics of ADSCs and physiological differences between the two groups. (a) Morphology of ADSCs at Passage 3 ($\times$40). (b,c) Differentiation potential of ADSCs into adipocytes and osteoblasts, as shown by Oil Red O and Alizarin Red staining ($\times$100). (d) Surface markers of ADSCs, as shown by flow cytometry. (e) Blood glucose levels in rats were measured weekly. The day on which STZ-induced diabetes was established in rats was referred to as day 0 of the experiment. ADSCs were injected at the end of the second week. (f) OGTT results before the rats were sacrificed and the AUC. DM, Diabetes mellitus; ADSCs, adipose-derived stem cells; STZ, streptozotocin; OGTT, Oral glucose tolerance test; AUC, area under the curve.

To examine the effect of stem cell therapy on the transcriptional landscape, we performed full transcriptome RNA-seq analysis on the liver tissues of those STZ-induced hyperglycemic rats. Fig. 2a shows liver transcriptome differences in diabetic rats with or without treatment. With p 0.05 and $|$log FC$|$ $>$0.5 as the cutoff criteria, a total of 1770 upregulated transcripts and 362 downregulated transcripts were screened (Fig. 2b). We selected certain highly altered upregulated (Mbp, Ly6al, Acpp) and downregulated (OAas1f, Amy2a3, Onecut1) genes from the RNA-seq data and verified them by qRT‒PCR (Fig. 2c). The results were consistent with the RNA sequencing results.

Fig. 2.

DEGs in the two groups. (a) Heatmap showing DEG levels (2.0 FC, FDR p 0.05) in the liver tissues of eight rats based on the RNA-seq data. Different columns indicate different samples, and different rows indicate different transcripts. The color scale represents the expression level of each gene in a single sample. If not otherwise noted, the terms upregulated or downregulated are used to describe the changes in stem cell-treated animals compared to untreated animals. (b) Volcano plot showing the RNA-seq data from the two groups, with p 0.05 and $|$log FC$|$ $>$0.5 as the cutoff criteria. The blue dots indicate downregulated genes, the red dots indicate upregulated genes, and the gray dots indicate genes without significant differences in expression. (c) The expression levels of selected genes in the liver tissues of eight untreated and stem cell-treated diabetic rats were measured by qRT‒PCR. The data are presented as the mean $\pm$ S.E. values. (d) Significantly affected (2.0 Z score –2.0) canonical pathways were automatically generated by IPA. The horizontal bars show the different pathways calculated based on the Z score. Red represents activation, while blue represents suppression. DM, STZ-induced DM rats; ADSC, ADSC-treated DM rats; DEGs, Differentially expressed genes; FC, fold change; FDR, false discovery rate; IPA, Ingenuity Pathways Analysis.

Examining the upregulated gene transcripts into IPA and performing canonical pathway analysis showed that many pathways were activated or inhibited during this process, including the hepatic fibrosis signaling pathway, IL-6 signaling pathway, and insulin receptor signaling pathway, which were significantly enriched (Fig. 2d) and showed the ameliorative effect of stem cell treatment on liver damage, inflammation, and insulin secretion.

IPA of downstream effectors can be used to predict comprehensive increases or decreases in downstream biological activity and gene-related function based on the causal relationship between measured transcriptome level changes and these activities and functions. Supplementary Fig. 1a shows a high level of middle and downstream activity and function in the liver tissue of treated and untreated diabetic rats. Fig. 3 shows multiple functional categories that were enriched, such as liver pathological changes and metabolic disorders, which included liver tissue alterations, cell growth and proliferation disorders, and suggested that stem cells could improve liver function and liver injury in DM. Furthermore, functional categories associated with immune system diseases and endocrine system diseases showed significant alterations. Notably, functional categories associated with hepatotoxicity showed significant upregulation, and there were much greater increases than those associated with nephrotoxicity and cardiotoxicity (Fig. 3b and Supplementary Fig. 1b). Liver toxicity was mainly characterized by liver damage, hepatitis, liver steatosis, liver fibrosis (LF) and other factors (Fig. 3c). Taken together, our data revealed significant decreases in susceptibility to liver damage and DM after stem cell therapy.

Fig. 3.

Downstream effector analysis of the two groups. (a) The affected diseases and biological function categories based on the tree map in Supplementary Fig. 1a. (b) The top ten most significantly enriched pathways. (c) The affected hepatotoxicity-related functional categories based on the tree map in Supplementary Fig. 1b. The major colored rectangles indicate a family of associated biological functions or diseases; blue indicates a decrease and orange indicates an increase. The sizes (using Fisher’s exact test p value) of the rectangles indicate whether the category is predicted by IPA to increase or decrease significantly between groups; and higher absolute Z scores are represented by color intensities. NAFLD, nonalcoholic fatty liver disease; NASH, nonalcoholic steatohepatitis; ASH, alcoholic steatohepatitis.

Multiple liver-related functional categories were enriched in stem cell-treated diabetic rats compared to untreated diabetic rats. Among the differentially enriched categories, liver damage and liver steatosis were the most prominent (Fig. 3c). Fig. 4a shows a heatmap of the log${}_2$ expression values of liver damage-related genes (Ccl4, Abcb1a, Cblb, Cnr2, Ddit3, Hmgb, Igf1, Il18, Irf3, Mapk14, Nr1i3, Sdc4, Sirt1, Gsk3b, Tgfbr2, Stk24, Klf2, Mboat7 and Cps1) in rat liver tissue. Our previous study showed a strong correlation between TGFB expression and the progression of liver damage [11], which was consistent with the results of this analysis. Overall, the expression of Ccl4, Abcb1a, Cblb, Cnr2, Ddit3, Hmgb, Igf1, Il18, Irf3, Mapk14, Nr1i3, Sdc4 and Sirt1 in hepatocytes promoted the progression of liver damage in mice, and Gsk3b, Tgfbr2, Stk24, Klf2, Mboat7 and Cps1 expression alleviated liver damage in mice. Interestingly, MAPK14 has previously been shown to alleviate liver damage [18]. This may explain the bidirectional alterations in MAPK14 isoform expression in our data.

Fig. 4.

Heatmap showing the amelioration of liver damage and liver steatosis in diabetic rats. Heatmap showing the expression of several upregulated and downregulated genes related to liver damage (a) and liver steatosis (b) in eight diabetic rats treated with or without ADSCs. The data are presented as the log${}_2$ TPM (transcript per million) expression values. The expression values are shown according to the color scale.

Moreover, some genes associated with liver steatosis were downregulated in stem cell-treated diabetic rats compared to untreated diabetic rats (Acbd5, Bhlhe40, Casp2, Cnr2, Ddit3, Dnajc7, Egfr, Helz, Ifit2, Il18, Irf3, Isg15, Isg15, Lime1, Nr1i3, Ptpn2, Sirt1, Srsf2, Tp53inp1, Thra, Egfr, Epas1, Foxo1, Gldc, Gna11, Gpam, Insig2, Insr, Klf2, Lrp6, Mapk14, Mboat7, Mgll, Acox1, Adipor2, Atf6, Stk24, Pdpk1 and Plin2; Fig. 4b). Interestingly, our data suggest that the progression of liver damage and hepatic lipid accumulation may be ameliorated by the inhibition of Cnr2, Ddit3, Il18, Irf3, Mapk14, Nr1i3 and Sirt1 following stem cell therapy.

Upstream regulator analysis can predict upstream molecules. Thus, we identified relevant factors by which stem cell therapy inhibits processes related to liver damage. IPA was used to predict possible factors, resulting in 30 factors with a $|$Z score$|$ greater than 3 (Table 2). These 30 factors were examined by IPA, and the nine factors shown in Fig. 5a were indirectly associated with liver damage. Among them, DUSP1 plays an important role in the cytosol and mitochondria. Changes in these factors identified by RNA-seq (Fig. 5b,c) were analyzed and showed that IPA predicted the importance of DUSP1 in the effects of stem cell treatment on diabetes.

Fig. 5.

Upstream regulator analysis and mechanistic network analysis predict multiple affected signaling networks. (a) Upstream regulators that are indirectly associated with liver damage. (b) The possible upstream location of DUSP1 was jointly predicted based on changes in factors associated with DUSP1. (c) DUSP1 mechanism network diagram showing its associated mechanism network and possible changes in downstream factors.

Upstream regulator analysis of DEGs explained the mechanistic networks underlying the observed changes in gene expression during stem cell therapy for DM (Supplementary Fig. 1), including the STAT4 ($|$Z score$|$ = 4.618), PTGER2 ($|$Z score$|$ = 3.916), LEF1 ($|$Z score$|$ = 3.915), INHBA ($|$Z score$|$ = 3.893), AHR ($|$Z score$|$ = 3.543), DUSP1 ($|$Z score$|$ = 3.435), CCND1 V ($|$Z score$|$ = 3.200), and CREB1 ($|$Z score$|$ = 3.095) networks (Fig. 5c, Supplementary Fig. 2). DUSP1 was predicted with relatively high confidence to directly activate the IL-12 complex, IL-1B and tumor necrosis factor (TNF). Furthermore, the activated IL-12 complex was predicted to directly promote the expression of STAT3 and STAT4; moreover, the upregulation of TNF was predicted to activate the vitamin D receptor (VDR) and coactivate SP1 through the upregulation of IL-1B. At the terminus of the mechanistic network, CREBBP, SMAD3, and EP300 were activated, and CDKN2A was inactivated. In summary, DUSP1 was highly enriched among the genes in liver damage-related functional categories with relatively high confidence and numerous predicted relationships.

By analyzing the factors that were downstream of DUSP1, the functional networks identified by IPA helped us to identify a mechanistic network in which DUSP1 affects multiple functions through multiple factors. The network is shown as genes and biological relationships between the nodes and lines, and nodes with specific shapes indicate the regulatory effect determined by IPA. The first network we identified consisted of the DEGs, their upstream and downstream relationships, and the functional changes that might result (32; as indicated in the legend); in short, this network probably moderates sensory system development (Fig. 6a). The other interesting enriched network moderated the antiviral response (Fig. 6b). Interestingly, we identified DUSP1 as a key hub gene in liver damage and other liver disease networks (Fig. 6c). Furthermore, with the Grow and Path Explorer IPA functions, we confirmed the indirect relationship between DUSP1 expression and numerous liver diseases (Fig. 6d).

Fig. 6.

Numerous disease-associated enrichment networks in diabetic rats. (a) Schematic of sensory system development and (b) antiviral response functional networks revealed by IPA. In this figure, red indicates activation, while blue indicates suppression. (c,d) DUSP1 is closely related to a variety of liver diseases and lipid metabolism disorders.

However, we also sought to explore the mechanism by which DUSP1 causes liver damage. The protein encoded by DUSP1 can dephosphorylate mitogen-activated protein kinase (MAPK), thus playing an important role in multiple pathological processes. We examined the phosphorylation status of the ERK1/2, P38 and JNK pathways, which are three pathways that play key roles in glucose metabolism disorders, in stem cell-treated diabetic rats compared to untreated diabetic rats. The results showed that the phosphorylation of ERK1/2 was significantly downregulated in ADSC-treated DM rats (Fig. 7a,c). This finding was further confirmed in vitro. P-ERK1/2 was increased when HepG2 cells were cultured with H${}_2$O${}_2$. However, it was restored to some extent when the cells were pretreated with HGF or ADSC supernatant, and there was a similar change in response to the ERK inhibitor U0126 (Fig. 7b,d). Moreover, the change in P-STAT3 was similar to that in P-ERK1/2. P-STAT3 was also reduced when the cells were pretreated with an ERK inhibitor. This result suggested that STAT3 phosphorylation was mediated by the ERK signaling pathway. These results suggested that stem cell therapy could act through DUSP1-mediated regulation of the ERK/STAT3 pathway.

Fig. 7.

Phosphorylation status of the ERK1/2, P38, JNK and STAT3 pathways. (a,c) The phosphorylation of ERK2, P38 and JNK in stem cell-treated diabetic rats and untreated diabetic rats. An unpaired t test was used to compare the two groups. (b,d) The phosphorylation of ERK1/2 and STAT3 in HGF- or stem cell-treated HepG2 cells induced by H${}_2$O${}_2$. An unpaired t test was used to compare the H${}_2$O${}_2$ group with the control (Con) group and the HGF, ADSC or U0126 groups with the H${}_2$O${}_2$ group. ERK, extracellular regulated protein kinase; JNK, c-jun N-terminal kinase; STAT3, signal transducer and activator of transcription 3; HGF, hepatocyte growth factor.