Islet-Like Structures Generated In Vitro from Adult Human Liver Stem Cells Revert Hyperglycemia in Diabetic SCID Mice

Abstract

A potential therapeutic strategy for diabetes is the transplantation of induced-insulin secreting cells. Based on the common embryonic origin of liver and pancreas, we studied the potential of adult human liver stem-like cells (HLSC) to generate in vitro insulin-producing 3D spheroid structures (HLSC-ILS). HLSC-ILS were generated by a one-step protocol based on charge dependent aggregation of HLSC induced by protamine. 3D aggregation promoted the spontaneous differentiation into cells expressing insulin and several key markers of pancreatic β cells. HLSC-ILS showed endocrine granules similar to those seen in human β cells. In static and dynamic in vitro conditions, such structures produced C-peptide after stimulation with high glucose. HLSC-ILS significantly reduced hyperglycemia and restored a normo-glycemic profile when implanted in streptozotocin-diabetic SCID mice. Diabetic mice expressed human C-peptide and very low or undetectable levels of murine C-peptide. Hyperglycemia and a diabetic profile were restored after HLSC-ISL explant. The gene expression profile of in vitro generated HLSC-ILS showed a differentiation from HLSC profile and an endocrine commitment with the enhanced expression of several markers of β cell differentiation. The comparative analysis of gene expression profiles after 2 and 4 weeks of in vivo implantation showed a further β-cell differentiation, with a genetic profile still immature but closer to that of human islets. In conclusion, protamine-induced spheroid aggregation of HLSC triggers a spontaneous differentiation to an endocrine phenotype. Although the in vitro differentiated HLSC-ILS were immature, they responded to high glucose with insulin secretion and in vivo reversed hyperglycemia in diabetic SCID mice.

Keywords: 3D culture; Diabetes; Insulin-producing stem cells; Liver stem cells; Pancreatic islets; Pancreatic β cells.

Fig. 1

In vitro differentiation of HLSC into islet-like structures (HLSC-ILS). a Representative micrographs of 10 experiments showing 1.3E + 4/cm² non-stimulated cells (HLSC) and cells stimulated with protamine 10 μg/ml (HLSC+P) or with 10 IU/ml heparin –protamine (10 μg/ml) (HLSC+p + Hep). Scale bars = 200 μm. b Representative micrographs of scanning electron microscopy (left; scale bar = 100 μm) and of dithizone staining (right; Scale bars = 200 μm) of HLSC-ILS showing their characteristic phenotype. c Dependence of HLSC-ILS generation on protamine concentrations (n = 6). d Growth curve of HLSC-ILS (n = 6). e Dependence of HLSC-ILS generation on cell density (cells/cm²) (n = 6). f Contribution of glucose and FBS to the generation of HLSC-ILS. RPMI basal medium (R), fetal bovine serum (F), glucose (G) and protamine (P) (n = 6). g Size frequency distribution of the mean profile diameter of HLSC-ILS expressed as relative frequency (Scale bars = 300 μm). The insert shows an example of the diameter measurement. Data are expressed as media ± SE of three independent experiments. ANOVA with Dunnet’s multi-comparison test was performed; *p < 0.01 versus 0 (C), 1.3E + 4 (E) or versus F and FG (F)

Fig. 2

Representative transmission electron microscopy of cadaveric human islets (a) and HLSC-ILS structures formed after 4 (b), 7 (c) and 14 (d) day stimulation with 10 μg/ml protamine. Heterogeneity of granules in HLSC-ILS at 14 days at higher magnification (e and f)

Fig. 3

Representative immunofluorescence for β-cell transcription factors and pancreatic hormones in cryostat sections of HLSC-ILS. The appropriate irrelevant isotype control antibodies used as controls were all negative. Original magnification ×400 (scale bars = 50 μm). Images are representative of 10 experiments performed with similar results

Fig. 4

Flow cytometry analysis of stem cells markers (of CD73, CD90, CD105, CD31), albumin and islets specific markers (PDX1, Ngn3, synaptophysin, C-peptide and glucagon) in HLSC and HLSC-ILS. Three independent experiments were performed with similar results and are expressed as media ± SE, ANOVA with Newman–Keuls multi-comparison test was performed; *p < 0.05 versus HLSC

Fig. 5

Phenotypic characterization of non-differentiated HLSC (HLSC) and differentiated HLCS-ILS at different times of culture (7, 14 and 21 days). a The expression of human Ngn3, PDX1, C-peptide and glucagon were analyzed by Western Blot. b Representative micrographs of immunofluorescence and percentage of different cell subpopulations expressing insulin alone, insulin and glucagon and negative for both hormones

Fig. 6

In vitro hC-peptide secretion analysis in static and dynamic conditions. a In vitro human hC-peptide and glucagon basal secretion by HLSC-ILS (4, 7, 14 and 21 days) was evaluated in static condition and compared with human islets. A total of 400 HLSC-ILS or cadaveric human islets were stimulated with 2.8 mM of glucose. Dunnet’s multi-comparison test was performed; data are presented as mean ± SD of 3 independent experiments. *p < 0.001 versus HLSC-ILS 4D and **p < 0.001 versus HLSC-ILS. b Static in vitro hC-peptide secretion by HLSC-ILS (4, 7, 14 and 21 days) after stimulation with two different concentrations of glucose (2.8 and 28 mM) and 50 mM of KCl. (c) HG/LG ratio secretion of hC-peptide of HLSC-ILS (4, 7, 14 and 21 days). Dunnet’s multi-comparison test was performed; data are presented as mean ± SD of 3 independent experiments *p < 0.001 versus 2.8 mM. C) HG/LG ratio of secretion by HLSC-ILS (4, 7, 14 and 21 days) and (d) Dynamic secretion of hC-Peptide by HLSC-ILS at 14 days of differentiation stimulated by 2.8 mM of glucose (LG), 28 mM of glucose (HG) and high potassium (KCl). The concentration of hC-peptide was assessed by ELISA and normalized to total protein. Data are expressed as mean ± SD of 4 independent experiments

Fig. 7

qRT-PCR array gene expression analysis of non-differentiated HLSC and differentiated HLCS-ILS at different times of culture. a Hierarchical clustering analysis of HLSC and HLSC-ILS at 4 (ILS 4d), 7 (ILS 7d), and 14 (ILS 14d) days of culture. The analyzed genes are listed below the heat map and grouped by gene function. The expression levels are reported as row Z score of RQ (2^^-∆∆Ct) values (scale color: blue low expression, orange high expression) of three independent experiments run in triplicate. Average linkage clustering method and Euclidean distance measurement methods were used. The heat map was generated with heatmapper online software (http://www.heatmapper.ca/expression/). b Real Time PCR analysis of transcription factors involved in β cells maturation and pancreatic hormones. Values are reported as mean ± SD of RQ of three independent experiments run in triplicate. Expression is normalized for HLSC (RQ = 1, not shown). * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001

Fig. 8

In vivo reversal of diabetes in SZT-treated SCID mice by implantation of 800 HLSC-ILS (28,000 IEQ/kg) under the right renal capsule. a Blood glucose levels in sham non-diabetic (non-DM, n. 12), diabetic (DM, n. 8), DM mice implanted with HLSC-ILS (DM + HLSC-ILS, n. 12) and DM + HLSC-ILS before (Pre-expl) and after (Post-expl) explantation (n. 4). ANOVA with Newman–Keuls multi-comparison tests was performed. * p < 0.001 versus DM; ^#p < 0.05 versus Pre-expl. Data are presented as mean ± SD. b Correlation between the serum levels of mouse and human C-peptide (mC-peptide and hC-peptide, respectively) and glycemia in non-DM, DM and DM-HLSC-ILS before (Pre-expl) and after (Post-expl) explantation. ANOVA with Newman–Keuls multi-comparison tests was performed. * p < 0.001 versus DM; # p < 0.05 glycemia DM + HLSC-ILS Pre-expl versus Post-expl; ^ p < 0.001 DM + HLSC-ILS Post-expl versus Non-DM Post-expl. Data are presented as mean ± SD. hC-petide was detectable only in DM + HLSC-ILS Pre-expl. mCpeptide and hC-peptide were undetectable in DM + HLSC-ILS Post-expl. C) IPGTT curves in non-DM (n.10), DM (n. 3) and DM + HLSC-ILS Pre-expl (n. 7) and Post-expl (n. 4). ANOVA with Newman–Keuls multi-comparison tests was performed. * ⁺p < 0.001 versus DM; ** p < 0.001 versus Pre-expl. Data are presented as mean ± SD

Fig. 9

Histological characterization of HLSC-ILS explants. a Representative H&E micrographs showing the presence of spheroid structures under the mice renal capsule (ILS). Scale bars in micrographs are 200 μm. b Human insulin expression in human pancreatic tissue (left, positive control) and HLSC-ILS explants (right). Arrow is showing a human pancreatic islet while positive reactivity is showed in brown. Scale bars in micrographs are 200 μm. c Insulin FISH reaction (red) in human pancreatic tissue (positive control) and HLSC-ILS explants. Renal and pancreatic mouse tissue were used as negative controls. Scale bars in micrographs are 10 μm. Three experiments were done with similar results

Fig. 10

qRT-PCR array gene expression analysis of non-differentiated HLSC, HLCS-ILS before and after implantation in diabetic mice, and human islets. a Hierarchical clustering analysis of HLSC and HLSC-ILS at 14 days of culture in vitro (ILS 14d), HLSC-ILS explanted after 2 (EXP2w) or 4 (EXP4w) weeks of implantation in diabetic mice and human islets (Human). The analyzed genes are listed below the heat map and grouped by gene function. The expression levels are reported as row Z score of Ln(RQ) values (scale color: red low expression, green high expression) of three independent experiments run in triplicate. Average linkage clustering method and Euclidean distance measurement methods were used. The heat map was generated with heatmapper online software (http://www.heatmapper.ca/expression/). b Real Time PCR analysis of transcription factors involved in β cells maturation and pancreatic hormones. Values are reported as mean ± SD of Ln(RQ) of three independent experiments run in triplicate. Expression is normalized for HLSC (Ln(RQ) = 0, not shown). * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001