Enrichment of human embryonic stem cell-derived NKX6.1-expressing pancreatic progenitor cells accelerates the maturation of insulin-secreting cells in vivo


  • Author contributions: A.R.: conception and design, collection and/or assembly of data, data analysis and interpretation; final approval of manuscript; J.E.B.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript; J.X.: conception and design, collection and/or assembly of data, and data analysis and interpretation; K.N., J.F., and J.J.O.: collection and/or assembly of data; T.J.K.: conception and design, data analysis and interpretation, and final approval of manuscript. A.R., J.E.B., and J.X. contributed equally to this article.

Correspondence: Timothy Kieffer, Ph.D., Room 5308-2350 Health Sciences Mall, University of British Columbia, Vancouver, British Columbia V6T 1Z3, Canada. Telephone: 604-822-2156; Fax: 604-822-2316; e-mail: tim.kieffer@ubc.ca


Human embryonic stem cells (hESCs) are considered a potential alternative to cadaveric islets as a source of transplantable cells for treating patients with diabetes. We previously described a differentiation protocol to generate pancreatic progenitor cells from hESCs, composed of mainly pancreatic endoderm (PDX1/NKX6.1-positive), endocrine precursors (NKX2.2/synaptophysin-positive, hormone/NKX6.1-negative), and polyhormonal cells (insulin/glucagon-positive, NKX6.1-negative). However, the relative contributions of NKX6.1-negative versus NKX6.1-positive cell fractions to the maturation of functional β-cells remained unclear. To address this question, we generated two distinct pancreatic progenitor cell populations using modified differentiation protocols. Prior to transplant, both populations contained a high proportion of PDX1-expressing cells (∼85%–90%) but were distinguished by their relatively high (∼80%) or low (∼25%) expression of NKX6.1. NKX6.1-high and NKX6.1-low progenitor populations were transplanted subcutaneously within macroencapsulation devices into diabetic mice. Mice transplanted with NKX6.1-low cells remained hyperglycemic throughout the 5-month post-transplant period whereas diabetes was reversed in NKX6.1-high recipients within 3 months. Fasting human C-peptide levels were similar between groups throughout the study, but only NKX6.1-high grafts displayed robust meal-, glucose- and arginine-responsive insulin secretion as early as 3 months post-transplant. NKX6.1-low recipients displayed elevated fasting glucagon levels. Theracyte devices from both groups contained almost exclusively pancreatic endocrine tissue, but NKX6.1-high grafts contained a greater proportion of insulin-positive and somatostatin-positive cells, whereas NKX6.1-low grafts contained mainly glucagon-expressing cells. Insulin-positive cells in NKX6.1-high, but not NKX6.1-low grafts expressed nuclear MAFA. Collectively, this study demonstrates that a pancreatic endoderm-enriched population can mature into highly functional β-cells with only a minor contribution from the endocrine subpopulation. Stem Cells 2013;31:2432–2442


Diabetes is characterized by chronic high blood glucose levels resulting from deficient insulin-producing pancreatic β-cells. There are numerous experimental strategies aimed to restore endogenous regulated insulin production in patients with type 1 diabetes [1]. To date, islet cell transplantation is the most effective clinical therapy [2, 3], but widespread application is severely limited by the shortage of cadaveric donor islets [4]. Human embryonic stem cells (hESCs) may be a potential renewable source of insulin-producing cells to bridge the gap between cell supply and clinical demand [5]. Step-wise differentiation protocols designed to mimic pancreatic development have successfully generated insulin-producing cells from hESCs in vitro, but these cells resemble fetal endocrine cells rather than adult pancreatic β-cells [6-17]. Mature, glucose-responsive insulin-secreting cells have only been produced from hESCs following transplantation of immature pancreatic progenitor cells; this was first demonstrated in healthy mice [12] and subsequently by our group in mice with diabetes [17, 18]. We demonstrated that graft-derived human insulin levels gradually increased over time and after a 3-month maturation period, previously insulin-dependent diabetic mice achieved normal fasting blood glucose levels without exogenous insulin treatment [17]. After a lengthy maturation period of approximately 30 weeks in vivo, human insulin secretion was regulated by various secretagogues, including meal, glucose, and arginine challenges, and a corresponding improvement in glucose tolerance was observed along with the appearance of insulin/V-maf musculoaponeurotic fibrosarcoma oncogene homolog A (MAFA) co-positive β-cells within kidney capsule grafts [17].

The pancreatic progenitor cells generated with our previous differentiation protocol were composed of mainly immature polyhormonal cells (insulin/glucagon-positive, NKX6.1-negative), endocrine precursor cells (NKX2.2/synaptophysin-positive, hormone/NKX6.1-negative), and pancreatic endoderm-like cells (NKX6.1/PDX1-positive, insulin-negative) [17]. The expression of NKX6.1 distinguishes these populations and is thought to mark the progenitor cells that arise during the second transition of pancreas development, as opposed to the polyhormonal cells that arise during the first transition [19]. Results from several recent studies suggest that polyhormonal cells may develop into pancreatic α-cells following transplantation [6, 16, 20]. In contrast, when hESC-derived pancreatic endoderm cells were purified by fluorescence-activated cell sorting (FACS) and transplanted into mice, functional islet-like clusters developed in vivo, although less efficiently than an unsorted progenitor cell population [6]. Thus, it remains unclear whether the NKX6.1-negative fraction contributes to maturation of the NKX6.1-positive cell fraction. Here, we examined the relative contributions of each pancreatic progenitor subpopulation to the development of mature insulin-secreting cells in vivo by transplanting cell populations that were enriched, but not FACS-purified, for either pancreatic endoderm or pan-endocrine cells.

Materials and Methods

In Vitro Differentiation of hESCs

The H1 hESC line was obtained from WiCell Research Institute, Inc. (Madison, WI, http://www.wicell.org). All experiments at the University of British Columbia (UBC) with H1 cells were approved by the Canadian Stem Cell Oversight Committee and UBC Clinical Research Ethics Board. H1 cells were cultured on 1:30 diluted Matrigel (BD BioSciences, San Diego, http://www.bdbiosciences.com; Cat#356231) in mTeSR-1 (Stem Cell Technologies, Vancouver, BC, Canada, http://www.stemcell.com; Cat#05850). At approximately 70%–80% confluence, cultures were rinsed with 1× Dulbecco's phosphate-buffered saline (DPBS) without Mg2+ and Ca2+ (Invitrogen, Carlsbad, CA, http://www.invitrogen.com; Cat#14190) followed by incubation with Versene (EDTA), 0.02% (LONZA, http://www.lonza.com, Cat#17–711E) for 12 minutes at room temperature. Released single cells were rinsed with mTeSR-1 and spun at 1,000 rpm for 5 minutes. The resulting cell pellet was resuspended in mTeSR-1 medium supplemented with Y-27632 (10 µM; Sigma-Aldrich; St. Louis, MO, http://www.sigmaaldrich.com; Cat#Y0503) and the single cell suspension was seeded at approximately 1.3 × 105 cells per centimeter square. Cultures were fed every day and differentiation was initiated 48 hours following seeding, resulting in approximately 90% starting confluence. The following differentiation protocol variations were used to generate either “NKX6.1-high” or “NKX6.1-low” cell populations (summarized in Fig. 1A):

Figure 1.

Characterization of NKX6.1-high and NKX6.1-low populations prior to transplantation in macroencapsulation devices. (A): Schematic summary of the 15-day in vitro differentiation protocols used to generate NKX6.1-low (green) and NKX6.1-high (brown) cells. (B): FACS quantification of differentiated hESC-derived populations at stage 4, day 4 prior to transplant. NKX6.1-low (n = 2) and NKX6.1-high (n = 3) populations were defined primarily based on their respective NKX6.1-expressing cell fractions. NKX6.1-low cells contained a relatively high proportion of endocrine cells (synaptophysin-positive and NKX2.2-positive) compared to NKX6.1-high cells. Representative FACS plots from both groups are provided in panel (C). (D): qPCR for expression levels of various pancreatic endocrine genes (hormones and transcription factors) in NKX6.1-high cells (n = 6) relative to NKX6.1-low cells (n = 4). Dashed red line indicates equal expression in both groups. (E–G): NKX6.1-low and NKX6.1-high cells were loaded into Theracyte devices pre-transplant and immunostained for insulin, glucagon and NKX6.1 (panel (E); scale bars = 100 µm), NKX2.2 (panel (F); scale bars = 50 µm) and DAPI (panel (G); scale bars = 50 µm). Abbreviations: DMEM-HG, Dulbecco's modified Eagle's medium-high glucose; DAPI, 4′,6-diamidino-2-phenylindole; FACS, fluorescence activated cell sorting; FBS, fetal bovine serum; FGF, fibroblast growth factor; hESC, human embryonic stem cells; LDN, LDN-193189; qPCR, quantitative polymerase chain reaction; TPB, (2S,5S)-(E,E)-8-(5-(4-(trifluoromethyl)phenyl)-2,4-pentadienoylamino)benzolactam.

Stage 1: Definitive Endoderm (3 Days)

Undifferentiated H1 cells plated on 1:30 Matrigel-coated surfaces were exposed to RPMI 1640 medium (Invitrogen; Cat#22400) supplemented with 1.2 g/L sodium bicarbonate (Sigma, MO; Cat# S6297), 0.2% fetal bovine serum (FBS; Hyclone, Logan, UT, http://www.hyclone.com; Cat# SH30071.02), 100 ng/mL activin-A (AA; Pepro-tech, Rocky Hill, NJ, http://www.peprotech.com), and 20 ng/mL of Wnt3A (R&D Systems, Minneapolis, http://www.rndsystems.com) for day 1 only. For the next 2 days, cells were cultured in RPMI with 0.5% FBS, 1.2 g/L sodium bicarbonate, and 100 ng/mL AA.

Stage 2: Primitive Gut Tube (3 Days)

Stage 1 cells were exposed to Dulbecco's modified Eagle's medium (DMEM)-F12 medium (Invitrogen) supplemented with 2 g/L sodium bicarbonate, 2% FBS, and 50 ng/mL of fibroblast growth factor 7 (Pepro-tech) for 3 days.

Stage 3: Posterior Foregut (4 Days)

Cultures were continued for 4 days in DMEM-HG (high-glucose) medium (Invitrogen) supplemented with 0.25 µM SANT-1 (Sigma-Aldrich), 2 µM retinoic acid (RA; Sigma-Aldrich), 100 ng/mL of Noggin (R&D Systems), and 1% (v/v) B27 (Invitrogen). At this stage, NKX6.1-low cells received 1 µM ALK5 inhibitor II (ALK5i; Axxora, San Diego, CA, http://www.axxora.com). ALK5i was not added to the NKX6.1-high cultures during stage 3.

Stage 4: Pancreatic Endoderm/Endocrine Precursors (5 Days)

Stage 3 cells were then cultured for 4 days in DMEM-HG medium supplemented with 100 ng/mL Noggin and 1% B27. For NKX6.1-high cells, 500 nM TPB (PKC activator; (2S,5S)-(E,E)-8-(5-(4-(trifluoromethyl)phenyl)-2,4-pentadienoylamino)benzolactam, EMD, http://www.emdmillipore.com/, Chemicals Inc, Gibbstown, NJ) was added to the stage 4 media. For NKX6.1-low cells, 1 µM ALK5i was added.

For the last day of culture, all cells were treated with 5 mg/mL dispase for 5 minutes at 37°C, pipetted gently to break into cell clumps (<100 µm) and transferred into Spinner Flasks (Corning, Acton, MA, http://www.corning.com). Cell clusters were spun at 80–100 rpm overnight in suspension with DMEM-HG supplemented with 0.2 µM ALK5i, 100 nM LDN-193189 (LDN; bone morphogenetic protein (BMP) receptor inhibitor, Stemgent, CA, https://www.stemgent.com/, Cat# 04–0074) and 1% B27.

Flow Cytometry

Stage 4 hESC-derived cells were assessed before transplant to ensure that the starting population met the appropriate quality control standards. Cell clusters were released into a single-cell suspension and either stained directly for surface markers or fixed and stained for various intracellular markers, as described previously [17]. Refer to Supporting Information Table 1 for antibody details.

Quantitative Reverse Transcriptase Polymerase Chain Reaction

Gene expression was assessed in NKX6.1-high and NKX6.1-low progenitor cells using custom Taqman Arrays (Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com), as previously described [17]. Data were analyzed using Sequence Detection Software (Applied Biosystems) and normalized to undifferentiated H1 cells using the ΔΔCt method.

Animal Studies

All experiments were approved by the UBC Animal Care Committee. Male 8–10-week-old SCID-beige mice (C.B-Igh−1b/GbmsTac-Prkdcscid-LystbgN7) were obtained from Taconic (Hudson, NY) and maintained on a 12-hour light/dark cycle with ad libitum access to a standard irradiated diet (Harlan Laboratories, Teklad Diet #2918; Madison, WI). Mice were rendered diabetic by a multiple low-dose streptozotocin (STZ) regimen of daily intraperitoneal STZ injections for 5 consecutive days (50 mg/kg per day).

Mice were anesthetized with inhalable isoflurane and received approximately 5 million hESC-derived stage 4 cells (NKX6.1-high or NKX6.1-low) transplanted subcutaneously on the right flank within a single 20 µL Theracyte planar macroencapsulation device (TheraCyte Inc., Laguna Hills, CA; n = 9 mice transplanted per group). These devices support neovascularization via woven outer membranes, while containing the engrafted cells within cell impermeable inner membranes [18, 21]. Although in our previous studies progenitor cells were transplanted under the kidney capsule [16, 17], we found that macroencapsulated progenitor cells develop into human C-peptide secreting cells with similar efficiency to cells implanted under the kidney capsule (Supporting Information Fig. 1; [18]). All mice received a single s.c. injection of enrofloxacin, Baytril, at the time of transplantation (10 mg/kg; Bayer Animal Health; Shawnee Mission, KS).

All metabolic analyses were performed in conscious, restrained mice on the indicated days. Blood glucose and body weight were monitored one to two times weekly following a 4 hour morning fast. For monthly meal challenges, blood glucose was measured and blood was collected after an overnight fast (16 hours) and then again after a 45 minute feeding period with normal chow. For the glucose tolerance test (GTT), mice received an oral bolus of glucose (2 g/kg, 30% solution; Vétoquinol, Lavaltrie, QC) and the arginine tolerance test (ArgTT), mice received i.p. arginine (2 g/kg, 40% solution; Sigma-Aldrich) following a 4 hour morning fast. For all tests, blood glucose was tested via saphenous vein using a handheld glucometer (Lifescan, Burnaby, BC), and saphenous blood samples were collected at the indicated time points using heparinized microhematocrit tubes. Plasma was stored at −30°C and later assayed using a human C-Peptide ELISA (Alpco Diagnostics, Salem, NH) or MSD Multi-Spot Assay for human glucagon-like peptide-1 (GLP-1), insulin, and glucagon (K15160C-2; Meso Scale Discovery, Gaithersburg, MD), according to manufacturer's instructions.


hESC-derived cells (pre-transplant and post-transplant) within Theracyte devices were fixed overnight in 4% PFA and stored in 70% ethanol prior to paraffin-embedding and sectioning (5 µm thickness; Wax-it Histology Services, Vancouver, Canada). Hematoxylin and eosin (H&E) and Masson's Trichrome staining were performed using standard protocols (Wax-it Histology Services), and all grafts were examined by an independent pathologist. H&E/Trichrome slides were scanned using the ScanScope CS system (Aperio, Vista, CA). Immunofluorescent staining was performed as previously described [16]; primary antibodies are detailed in Supporting Information Table 2. Images were captured using the ImageXpressMICRO Imaging System and analyzed using MetaXpress Software (Molecular Devices Corporation, Sunnyvale, CA, http://www.moleculardevices.com).

Endocrine cells were quantified within Theracyte devices harvested at 17–21 weeks post-transplant (n = 3 per group). Engrafted cells were coimmunostained for insulin, glucagon, and somatostatin, and whole slide fluorescence scanning was performed as described above. Individual images were stitched together to recreate the full length of each Theracyte device and the entire region between the inner membranes was quantified using MetaXpress software. The endocrine fraction was determined as the total number of insulin-, glucagon-, and/or somatostatin-positive cells relative to the total number of 4′,6-diamidino-2-phenylindole (DAPI)-positive cells within the inner membranes. The endocrine population was further subdivided into the number of insulin-positive, glucagon-positive, or somatostatin-positive cells relative to the total number of endocrine cells (insulin, glucagon, and/or somatostatin-positive), regardless of the number of hormones present per cell. Finally, the number of single hormonal (insulin-only, glucagon-only, or somatostatin-only) and polyhormonal (insulin/glucagon, insulin/somatostatin, glucagon/somatostatin, or insulin/glucagon/somatostatin) cells were quantified as a proportion of the total number of endocrine cells (i.e., insulin, glucagon, and/or somatostatin-positive).

Statistical Analysis

All statistics were performed using GraphPad Prism software (GraphPad Software Inc., LA Jolla, CA). Specific statistical tests for each experiment are described in the figure legends. Student-Newman-Keuls post hoc test was used for comparisons between groups with all one-way ANOVAs. For all analyses, p < .05 was considered statistically significant


Characterization of NKX6.1-High and NKX6.1-Low Cells Pre-transplant

Our revised in vitro differentiation protocols were designed to generate two different mixed populations of pancreatic progenitor cells (see Materials and Methods section for details and Fig. 1A for summary schematic). These two cell populations were thoroughly characterized by FACS (Fig. 1B, 1C), quantitative polymerase chain reaction (Fig. 1D), and immunofluorescent staining (Fig. 1E–1G) prior to transplant. The pancreatic endoderm-enriched population was designated as NKX6.1-high and the endocrine-enriched population as NKX6.1-low, since they contained approximately 80% versus 25% NKX6.1-positive cells, respectively (Fig. 1B, 1C). NKX6.1 expression was inversely proportional to the endocrine compartment; NKX6.1-high (pancreatic endoderm-enriched) and NKX6.1-low (endocrine-enriched) populations contained approximately 11% and 60% synaptophysin-positive cells, respectively (Fig. 1B, 1C). With the exception of NKX6.1 and PDX1, pancreatic endocrine transcription factors were generally upregulated in the NKX6.1-low cell population compared to the NKX6.1-high population (Fig. 1B, 1D). Both populations contained approximately 90% PDX1-positive cells (Fig. 1B) and PDX1 gene expression was similar (Fig. 1D). Immunofluorescent staining confirmed the profound difference in NKX6.1 expression between the two progenitor populations (Fig. 1E) and the enrichment of polyhormonal endocrine cells in the NKX6.1-low population (Fig. 1E–1G). Insulin/glucagon copositive cells were NKX6.1-negative (Fig. 1E) and NKX2.2-positive (Fig. 1F) prior to transplant.

Figure 2.

In vivo metabolic tracking of NKX6.1-low and NKX6.1-high cell function following transplantation in macroencapsulation devices. (A): Fasting blood glucose levels (mM) in diabetic mice (low-dose STZ model) transplanted with macroencapsulated NKX6.1-low or NKX6.1-high cells. Blood glucose levels were significantly lower in mice with NKX6.1-high cells compared to NKX6.1-low cells after 21 days post-transplant (*, p < .05, t test NKX6.1-high vs. NKX6.1-low). (B): NKX6.1-low and NKX6.1-high cells secreted similar levels of human C-peptide after an overnight fast at all ages, but at 3 and 4 months post-transplant, NKX6.1-high cells secreted significantly higher levels of human C-peptide following a 45-minute meal challenge (*, p < .05, paired t test, fast vs. fed). (C–E): Response to an ipArgTT (2 g/kg) at 17 weeks post-transplant. (C): Blood glucose (*, p < .05, t test; NKX6.1-high vs. NKX6.1-low), (D) human insulin secretion (*, p < .05, paired t test; 0 vs. 15 minutes), and (E) glucagon secretion (*, p < .05, one-way ANOVA; fasting glucagon NKX6.1-high vs. NKX6.1-low, sham). Sham mice were nontransplanted, nondiabetic SCID-beige controls. (F): Blood glucose levels and (G) human C-peptide secretion in response to an oral GTT (2 g/kg) at 20 weeks post-transplant. AUC values are shown to the right (*, p < .05, t test; NKX6.1-high vs. NKX6.1-low). Abbreviations: AUC, area under the curve; ArgTT, arginine tolerance test; GTT, glucose tolerance test; ipArgTT, intraperitoneal arginine tolerance test; STZ, streptozotocin.

Metabolic Profile of NKX6.1-High and NKX6.1-Low Cells Following Transplant

Macroencapsulated NKX6.1-high and NKX6.1-low cells were transplanted into mice with STZ-induced diabetes (average fasting blood glucose of 17.6 ± 0.94 mM at the time of transplant). Mice that received NKX6.1-high cells exhibited significantly lower fasting blood glucose levels compared to mice that received NKX6.1-low cells throughout the study beginning at 3 weeks post-transplant (Fig. 2A) and also during ArgTT and GTT at 17 and 20 weeks, respectively (Figs. 2C, 2F). By approximately 90 days post-transplant, NKX6.1-high cells had completely reversed STZ-induced hyperglycemia, whereas mice with NKX6.1-low cells remained moderately hyperglycemic until the endpoint of the study at 135 days (Fig. 2B). Blood glucose tracking data for individual animals are provided in Supporting Information Figure 2A and individual fasting glucose values pre-transplant (day 2) and post-transplant (day 113) are provided in Supporting Information Table 3.

Overnight fasted human C-peptide levels were similar between groups at all ages, whereas only recipients of NKX6.1-high cells displayed meal-stimulated human C-peptide secretion (1.7- and 2.4-fold increase after feeding at 3 and 4 months, respectively; Fig. 2B). Refer to Supporting Information Figure 2B and Supporting Information Table 3 for individual human C-peptide tracking data during monthly meal challenges. NKX6.1-high cells secreted significantly higher levels of human insulin/C-peptide at 15–30 minutes following arginine and glucose challenges, unlike NKX6.1-low cells, which were not responsive to either secretagogue in vivo (Fig. 2D, 2G). Fasting glucagon levels were significantly higher in NKX6.1-low versus NKX6.1-high mice; there was no difference in arginine-stimulated glucagon secretion between groups (Fig. 2E).

Characterization of Grafts Derived from NKX6.1-High and NKX6.1-Low Cells

At 5 months post-transplant, there were no obvious morphological differences in graft histology between NKX6.1-high and NKX6.1-low cells by H&E or Trichrome staining (Fig. 3; n = 3 per group). An independent pathology analysis of H&E-stained Theracyte devices indicated that grafts from NKX6.1-low and NKX6.1-high mice were largely composed of pancreatic endocrine cells (confirmed by quantification of % insulin, glucagon, and/or somatostatin-positive cells in Fig. 4B; NKX6.1-low: 67.3% ± 1.3%, NKX6.1-high: 73.5% ± 2.3%), as well as some loose mesenchymal-like cells (Fig. 3A, 3B, 3E, 3F). Based on Masson's trichrome staining, these regions of “loose meschenchymal tissue” (generally located adjacent to the inner membranes) were identified as collagen fibers (Fig. 3C, 3D, 3G, 3H). Notably, none of the grafts analyzed in this study contained mature bone or cartilage tissue, unlike grafts generated with our previous differentiation protocol [17].

Figure 3.

Morphology of engrafted NKX6.1-high and NKX6.1-low cells within Theracyte devices at 5 months post-transplant. H&E staining (panels (A), (B), (E), (F)) and Masson's Trichrome staining (panels (C), (D), (G), (H)) of harvested Theracyte devices at 5 months post-transplant derived from: (A–D) NKX6.1-low cells (enriched for polyhormonal endocrine cells) or (E–H) NKX6.1-high cells (enriched for pancreatic endoderm cells). For H&E panels, low magnification images through the full length of devices are shown on top and magnified images below; corresponding scale bar dimensions are indicated on each image. Pancreatic endocrine and ductal cells are indicated on images. For trichrome staining, two magnified regions are shown to the right, illustrating varying degrees of collagen staining (blue) within the inner membranes.

Figure 4.

Endocrine profiles of macroencapsulated NKX6.1-low and NKX6.1-high cells at 5 months post-transplant. (A): Harvested Theracyte devices were stained for insulin (red), glucagon (blue), somatostatin (green), and DAPI (white). Individual channels are shown for each magnified region in the bottom right of all panels. Two different animals are shown from each treatment group. Scale bars = 100 µm. (B): Quantification of the % endocrine cells (insulin-, glucagon-, and/or somatostatin-positive) relative to the total number of DAPI-positive nuclei within devices (n = 3/group). Grafts from both groups contained a similar proportion of endocrine cells. (C): Quantification of individual endocrine subpopulations, expressed as a proportion of the total insulin-/glucagon-/somatostatin-positive population (n = 3 per group). Grafts derived from NKX6.1-low cells contained significantly more glucagon-positive cells overall, as well as more glucagon-only and glucagon/somatostatin copositive cells compared to NKX6.1-high cells. In contrast, NKX6.1-high grafts contained significantly more insulin-positive and insulin-only cells compared to NKX6.1-low grafts. There was also a trend toward more polyhormonal insulin-expressing cells in NKX6.1-high cells compared to NKX6.1-low cells. *, p < .05, t test. Abbreviation: DAPI, 4′-6-diamidino-2-phenylindole.

Immunofluorescent staining of macroencapsulated cells revealed dramatic differences in the pancreatic endocrine subpopulations between groups (Fig. 4). NKX6.1-low cells developed into mainly glucagon-positive cells (96.5% ± 0.4%), whereas NKX6.1-high cells became mainly insulin-positive cells (88.7% ± 5.7%); both groups contained a similar number of somatostatin-positive cells (28.6% ± 4.6% and 34.4% ± 5.4% in NKX6.1-low and NKX6.1-high grafts, respectively) (Fig. 4). Quantification of the insulin-only, glucagon-only, and somatostatin-only populations, as well as the various polyhormonal populations are provided in Figure 4B and 4C. Paradoxically, devices from NKX6.1-high mice, which secreted human insulin in response to various secretagogues (Fig. 2B, 2D, 2G), also contained approximately twice as many insulin-positive cells that coexpressed at least one other hormone (Fig. 4). Both groups contained a small ghrelin-positive population and rare cells expressing pancreatic polypeptide (Fig. 5).

Figure 5.

Ghrelin and pancreatic polypeptide expression in Theracyte devices at 5 months post-transplant. Macroencapsulated human embryonic stem cell-derived pancreatic endocrine cells contained a small ghrelin-positive population and rare cells that expressed pancreatic polypeptide (white arrows). Cells derived from NKX6.1-low cells are shown on the left and NKX6.1-high cells on the right. Scale bars = 100 µm. Abbreviation: DAPI, 4′-6-diamidino-2-phenylindole.

We next examined the transcription factor profiles of endocrine cells derived from NKX6.1-high and NKX6.1-low cells (Fig. 6). Insulin-positive cells in both groups expressed nuclear NKX6.1 (Fig. 6A, 6B) and PDX1 in the nucleus and cytoplasm (Fig. 6C). Insulin-positive cells that were negative for PDX and NKX6.1 generally coexpressed either glucagon or somatostatin (Fig. 6A–6C, double arrows). Notably, MAFA expression was distinctly nuclear in insulin-expressing cells from NKX6.1-high grafts versus strictly cytoplasmic in insulin-expressing cells from NKX6.1-low grafts (Fig. 6D). Otherwise, endocrine cells were similar between groups, with somatostatin-positive cells generally expressing nuclear Hhex1 (Fig. 6E) and glucagon-positive cells expressing nuclear ARX (Fig. 6F), regardless of insulin coexpression.

Figure 6.

Transcription factor expression in endocrine cells from macroencapsulated NKX6.1-low and NKX6.1-high cells at 5 months post-transplant. (A, B): NKX6.1 was expressed in insulin-positive cells from both groups (single arrowheads) but was generally not expressed in cells coexpressing either insulin/glucagon (A) or insulin/somatostatin (B), as indicated by double arrowheads. (C): Insulin-positive cells in both groups expressed PDX1 in the cytoplasmic and nuclear compartments (single arrowheads); insulin/glucagon co-positive cells did not express PDX1 (double arrowheads). (D): Insulin-positive cells in grafts derived from NKX6.1-low cells expressed MAFA in the cytoplasm, but not the nucleus, whereas NKX6.1-high grafts contained insulin-positive cells with nuclear MAFA expression (single arrowheads). MAFA was not expressed in polyhormonal insulin-positive cells (double arrowheads). (E, F): In both groups somatostatin-positive cells were co-positive for Hhex1 (E) and glucagon-positive cells for ARX (F), even when insulin was also expressed in the same cell (double arrowheads). Unihormonal insulin-positive cells were negative for both Hhex1 (E) and ARX (F), as indicated by single arrowheads. For all panels, magnified regions illustrate three channels on top, blue/green/DAPI channels in the middle and red/green/DAPI channels on the bottom. Scale bars = 50 µm. Abbreviation: DAPI, 4′-6-diamidino-2-phenylindole.

Finally, we examined other markers found in mature pancreatic β-cells to determine if there were additional differences between insulin-positive cells derived from NKX6.1-high versus NKX6.1-low populations. In both groups, insulin-positive cells uniformly expressed glucose transporter 1 (GLUT1) and sporadically expressed amylin (Fig. 7A, 7B, respectively). Prohormone convertase (PC) 1/3 was robustly expressed in nearly all insulin-positive cells from NKX6.1-high grafts but only in a subset of insulin-positive cells from NKX6.1-low mice (Fig. 7C). Surprisingly, neither group appeared to express PC2 in the insulin-positive cell population, although this enzyme was highly expressed in glucagon-positive cells (Fig. 7D). The KATP channel sulfonylurea receptor, SUR1, was uniformly expressed in insulin- and glucagon-positive cells from both groups, and if anything, at a higher level in insulin-positive cells from NKX6.1-low mice (Fig. 7E). SerpinB10 was recently reported to be exclusively expressed in mature β-cells within adult human pancreas [22]. We observed robust expression of serpinB10 in the majority of insulin-positive cells from both NKX6.1-high and NKX6.1-low grafts and never in glucagon-positive cells (Fig. 7F).

Figure 7.

Characteristics of insulin-expressing cells in NKX6.1-low and NKX6.1-high cells post-transplant. NKX6.1-low and NKX6.1-high grafts at 5 months post-transplant were immunostained for insulin, glucagon, DAPI, and various markers of mature pancreatic β-cells, including: (A) GLUT1, (B) amylin, (C) PC 1/3, (D) PC2, (E) SUR1, and (F) serpin peptidase inhibitor (serpin) B10. Insulin-positive cells from both groups expressed similar levels of GLUT1 (A), amylin (B), SUR1 (E) and serpinB10 (F) but were generally negative for PC2 (D). PC1/3 was more uniformly expressed in insulin-positive cells from NKX-high grafts than in NKX6.1-low grafts, where many PC1/3-negative insulin-expressing cells were observed (C). Scale bars = 100 µm. Abbreviations: DAPI, 4′-6-diamidino-2-phenylindole; GLUT1, glucose transporter 1; PC, prohormone convertase; SUR1, sulfonylurea receptor.


This study demonstrates that hESC-derived progenitor populations containing a relatively high proportion of NKX6.1-expressing precursor cells may be more suitable for transplantation into patients with diabetes than a population enriched for immature endocrine cells. Progenitor cells enriched for NKX6.1-positive cells matured more quickly in vivo into glucose-responsive insulin-secreting cells compared to an endocrine-enriched progenitor population. Moreover, the insulin secretion kinetics displayed by NKX6.1-high cells within only 3 months of transplantation are attractive from a clinical cell therapy perspective. Not only did appropriate secretagogues, including glucose, stimulate rapid secretion of human insulin but also circulating C-peptide levels returned to baseline within an hour of a glucose challenge, thus reducing the risk of hypoglycemia.

Based on previous work by Kelly et al. [6], we predicted that the hESC-derived PDX1/NKX6.1 copositive subpopulation of pancreatic progenitor cells may have the developmental potential to become glucose-responsive β-cells following transplantation. We also suspected that hESC-derived polyhormonal cells likely contribute important developmental cues for endocrine cell maturation and thus should not be eliminated by FACS purification. Therefore, we generated two mixed pancreatic progenitor populations using modified in vitro differentiation protocols (summarized in Fig. 1A): (a) NKX6.1-high cells were enriched for pancreatic endoderm cells and contained a relatively minor endocrine component, whereas (b) NKX6.1-low cells were enriched for polyhormonal endocrine cells and expressed high levels of pancreatic endocrine transcription factors (except for NKX6.1). It is also possible that the presence of nonpancreatic cells may differ between progenitor populations and potentially influence the development of pancreatic endocrine cells post-transplant. Following transplant, NKX6.1-high and NKX6.1-low populations both developed into mainly pancreatic endocrine cells (∼70% insulin, glucagon, and/or somatostatin-positive), although the endocrine subpopulations differed dramatically between groups. Moreover, NKX6.1-high grafts not only contained a higher proportion of insulin-positive cells post-transplant but these cells also exhibited an advanced maturation state compared to insulin-positive cells from NKX6.1-low grafts. For instance, both groups secreted similar basal levels of human C-peptide, but NKX6.1-high cells secreted insulin in response to meal, glucose, and arginine challenges, and showed a remarkable capacity to treat diabetes. Notably, we observed meal responsiveness much earlier with NKX6.1-high cells than in our previous studies (at 12 vs. 30 weeks [17]) and the level of stimulation was considerably more robust (2.4-fold vs. 1.3-fold meal stimulus [17]). NKX6.1-high cells also displayed excellent insulin secretion kinetics, including rapid secretion within 15 minutes of a glucose challenge and return to baseline levels by 60 minutes. This is preferable over the lagging insulin secretion kinetics reported previously from hESC-derived cells, in which glucose-induced insulin secretion was observed by 60 minutes, but return to baseline insulin levels was not reported [6, 12, 17]. These findings begin to address important concerns about potential insulin overproduction by a surrogate β-cell lacking a dynamic “off switch” in response to falling glycemia [23]. Indeed, mice in the Kroon et al. [12] studies displayed significant hypoglycemia relative to controls (nonfasting levels of 55 mg/dL compared to 139 mg/dL). Hypoglycemia was never observed in mice engrafted with NKX6.1-high cells, either during a glucose challenge or weekly blood glucose tracking, likely due to the ability of engrafted cells to appropriately reduce insulin secretion.

We next sought to understand what distinguishes insulin-expressing cells derived from NKX6.1-high versus NKX6.1-low progenitor populations, since the increased number of insulin-secreting cells observed post-transplant in NKX6.1-high grafts could not explain the differences in insulin secretion kinetics. GLUT1 (the predominant glucose transporter in human β-cells [24, 25]) and SUR1 (regulatory subunit of the KATP channel) were each similarly expressed in insulin-positive cells from both groups and did not correspond with differences in insulin secretion kinetics. This supports recent findings that gene expression for glucose transporter and ATP-sensitive K+ channel subunits was similar between immature neonatal and glucose-responsive adult rat β-cells [26]. We also examined amylin and PC1/3 expression, as our previous study demonstrated that these β-cell specific proteins were absent from immature hESC-derived insulin-positive cells in the early post-transplant period [17]. Here, we observed a similar pattern of expression for amylin between groups, excluding this as an important maturation factor in our system. Interestingly, PC1/3 expression was more uniform in insulin-positive cells from NKX6.1-high grafts compared to NKX6.1-low grafts, whereas PC2 expression was relatively low in β-cells of both groups. Loss of PC1/3 expression has been reported to cause more severe defects in proinsulin processing than PC2 deficiency [27]. Considering that PC1/3 expression is required and sufficient for processing of proinsulin [27], the lack of PC1/3 expression in many insulin-positive cells from NKX6.1-low grafts may lead to proinsulin processing defects but still would not explain the immature insulin secretion kinetics in these mice. Finally, we examined a relatively new putative marker of mature β-cells, SerpinB10 [22], and confirmed that it was specifically localized to insulin-expressing cells but was not differentially expressed between NKX6.1-low and NKX6.1-high grafts.

We believe that impaired MAFA nuclear translocation may be a key defect in insulin-secreting cells derived from NKX6.1-low progenitor cells. MAFA is a known regulator of glucose-stimulated insulin secretion and key gene required for mature β-cell function [28-30]. Indeed, immature neonatal rat β-cells lacking glucose-stimulated insulin secretion not only express low levels of MAFA relative to adult β-cells but also expression is almost exclusively restricted to the cytoplasm of β-cells until postnatal day 15 [30], thus resembling the insulin-positive cells in NKX6.1-low grafts. Furthermore, overexpression of nuclear MAFA in neonatal (postnatal day 2) rodent β-cells promoted glucose-induced insulin secretion, suggesting that MAFA nuclear localization is a key regulator of β-cell maturity during normal pancreas development [30]. Altered MAFA localization may also account for a loss of β-cell function during the pathogenesis of diabetes. Cytoplasmic localization of MAFA was observed in β-cells from hyperglycemic db/db mice [31] and humans with type 2 diabetes [32]. Moreover, overexpression of an endogenous antioxidant enzyme was able to preserve intranuclear MAFA expression as well as ameliorate hyperglycemia in the db/db mouse model [31]. Although the physiological mechanisms that control MAFA localization are still unknown, the abundant nuclear MAFA expression could be a key factor driving the improved insulin secretion kinetics in β-cells derived from NKX6.1-high pancreatic progenitor cells.

Previous evidence in mice suggested that the timing of NGN3 induction may control the competency for pancreatic progenitors to generate specific endocrine cell types [33]. Therefore, we predicted that the timing of NGN3 induction during hESC differentiation would also influence the production of insulin-producing cells in vivo. Indeed, our findings suggest that the manner and timing of NGN3 induction in developing hESCs does control the relative proportions of pancreatic endocrine lineages. NGN3 timing was manipulated in our studies by adjusting exposure to ALK5i and PKC activator. We had previously reported that combined inhibition of transforming growth factor beta (TGF)β/BMP signaling with ALK5i/Noggin during stage 4 caused dramatic induction of pancreatic endocrine markers, including NGN3 [16, 17], whereas addition of a PKC activator reduced NGN3 while increasing NKX6.1 expression [17]. Here, the addition of ALK5i during stages 3–4 and absence of the PKC activator during stage 4 in the NKX6.1-low protocol not only reduced NKX6.1 expression at stage 4 but also caused a dramatic induction of NGN3 during stages 3–4 (∼20-fold induction compared to NKX6.1-high cells). In contrast, induction of NGN3 was delayed in NKX6.1-high cells due to the absence of ALK5i during stage 3; PKC activation also reduced NGN3 and increased NKX6.1 expression. Similar to mice in which early induction of NGN3 favored α-cell development [33], early induction of NGN3 during stage 3 in the NKX6.1-low population resulted in the development of mainly mature α-cells in vivo, as indicated by both circulating glucagon levels and the overwhelming number of glucagon/ARX co-positive cells within the harvested devices. Our observations also support previous studies [6, 16, 20], which collectively indicate that polyhormonal cells may be lineage committed toward an α-cell fate and challenge a long-accepted lineage tracing study, which concluded that fetal insulin/glucagon copositive cells did not contribute to adult α- or β-cell populations [34]. We also observed an unexpected population of cells expressing either insulin/glucagon or insulin/somatostatin derived from NKX6.1-high cells at 5 months post-transplant, particularly around the edges of encapsulation devices. The insulin/somatostatin copositive population is unusual, as this cell type is not typically observed during human fetal pancreas development [35] and appeared to resemble mature δ-cells rather than β-cells (hex-positive, NKX6.1-negative). The potential contribution of these polyhormonal cells to the function of surrounding unihormonal insulin-secreting cells remains to be determined.


Taken together, these studies indicate that when enriched (but not purified), hESC-derived pancreatic endoderm cells may be a suitable cell therapy product for treating patients with diabetes. Within a relatively short time frame (3 months), the NKX6.1-enriched population developed into insulin-secreting cells that responded to various physiological secretagogues and were capable of reducing insulin secretion when necessary. Moreover, these cells expressed key markers of mature pancreatic β-cells, including robust nuclear MAFA expression. In contrast, the hESC-derived population enriched for polyhormonal endocrine cells developed into mostly mature α-cells following transplant and produced immature pancreatic β-cells, that did not secrete insulin in a glucose-responsive manner or express nuclear MAFA. These data support the concept that hESCs may be a feasible alternative to cadaveric islets for transplantation within macroencapsulation devices into patients with type 1 diabetes.


T.J.K. was supported by a senior scholarship from the Michael Smith Foundation for Health Research. J.E.B. was funded by a JDRF postdoctoral fellowship, CIHR postdoctoral fellowship, and the CIHR Transplantation Training Program. J.E.B. also received a L'Oréal Canada for Women in Science Research Excellence Fellowship. We thank Mr. Ali Asadi for his technical expertise with immunofluorescent staining and Stem Cell Technologies for their financial support; Dr. Clifford Bogue from the Yale University School of Medicine for kindly providing the hex antibody. This work was funded by the Canadian Institutes of Health Research (CIHR) Regenerative Medicine and Nanomedicine Initiative, the Stem Cell Network (SCN) and the Juvenile Diabetes Research Foundation (JDRF).

Disclosure ofPotentialConflicts ofInterest

A.R., J.X., K.N., and J.J.O. are employees of Janssen R&D, LLC; T.J.K. received financial support from Janssen R&D, LLC, for the research described in this article. No other potential conflicts of interest relevant to this article were reported.