Reversal of diabetes with insulin-producing cells derived in vitro from human pluripotent stem cells.

A RT I C L E S

Reversal of diabetes with insulin-producing cells derived in vitro from human pluripotent stem cells

npg

© 2015 Nature America, Inc. All rights reserved.

Alireza Rezania1, Jennifer E Bruin2, Payal Arora1, Allison Rubin1, Irina Batushansky1, Ali Asadi2, Shannon O’Dwyer2, Nina Quiskamp2, Majid Mojibian2, Tobias Albrecht2, Yu Hsuan Carol Yang2, James D Johnson2,3 & Timothy J Kieffer2,3 Transplantation of pancreatic progenitors or insulin-secreting cells derived from human embryonic stem cells (hESCs) has been proposed as a therapy for diabetes. We describe a seven-stage protocol that efficiently converts hESCs into insulin-producing cells. Stage (S) 7 cells expressed key markers of mature pancreatic beta cells, including MAFA, and displayed glucose-stimulated insulin secretion similar to that of human islets during static incubations in vitro. Additional characterization using single-cell imaging and dynamic glucose stimulation assays revealed similarities but also notable differences between S7 insulin-secreting cells and primary human beta cells. Nevertheless, S7 cells rapidly reversed diabetes in mice within 40 days, roughly four times faster than pancreatic progenitors. Therefore, although S7 cells are not fully equivalent to mature beta cells, their capacity for glucose-responsive insulin secretion and rapid reversal of diabetes in vivo makes them a promising alternative to pancreatic progenitor cells or cadaveric islets for the treatment of diabetes.

Type 1 diabetes is one of the most common endocrine disorders in children, characterized by chronic hyperglycemia due to autoimmune destruction of insulin-producing pancreatic islet beta cells. Patients require exogenous insulin delivery, but the challenges of managing insulin dosing may lead to poor overall glycemic control. Islet cell transplantation can achieve superior glucose homeostasis compared with insulin therapy1,2 but is limited by its reliance on organ donations, challenges with isolation of islets from the pancreas and life-long use of immunesuppressive drugs. Pluripotent stem cells have tremendous potential to address the shortage of donor islets. We3–7 and others8–15 have developed multistep protocols based on developmental paradigms to differentiate hESCs into pancreatic progenitor cells capable of maturation in vivo. Pancreatic progenitor cells can produce glucose-responsive, insulin-secreting cells and prevent13 or reverse4–6 diabetes in mice several months after transplantation. These findings suggest that pancreatic progenitor cells are on course to become mature beta cells; however, current in vitro differentiation protocols have been unable to efficiently direct these progenitor cells further down the beta-cell development pathway, often leading instead to polyhormonal (insulin+/glucagon+/somatostatin+) cells3,12,14. We expect that mature endocrine cells generated in vitro would reverse diabetes more rapidly than pancreatic progenitor cells after transplantation. Moreover, cultures of mature beta cells may be useful for drug screening, regenerative medicine development and as an experimental model to understand the pathogenesis of diabetes. It has proven challenging to make authentic, glucose-responsive, insulin-secreting beta cells in vitro from human pluripotent stem cells16–19.

Mature beta cells secrete insulin within minutes of a glucose stimulus and subsequently shut off production as needed to prevent hypoglycemia. Therefore, in this study, functional characterization of S7 cells included assessment of the kinetics of glucose-induced insulin secretion. Furthermore, key markers of beta-cell development were assessed in vitro at the mRNA and protein levels. PDX1 (a pancreatic homeodomain transcription factor) and NKX6.1 (a homeobox transcription factor) are co-expressed in multipotent pancreatic progenitor cells, which give rise to all adult pancreatic endoderm cells. PDX1 expression precedes NKX6.1 expression in pancreas development, and co-expression of both transcription factors becomes restricted to beta cells20–22. NGN3 (a basic helix-loop-helix transcription factor) marks endocrine progenitor cells that give rise to hormone-expressing cells23. NGN3 is transiently expressed in the developing pancreas, whereas downstream targets of NGN3, such as NEUROD1 and NKX2.2, persist in most pancreatic endocrine cells, including beta cells. MAFA (a basic leucine zipper transcription factor) is expressed in adult beta cells and is absent in developing beta cells and other pancreatic cells. MAFA is thought to be critical in establishing betacell function in adult islets24–27. We used these transcription factors as key markers in our differentiation protocol to distinguish differentiation from a pancreatic progenitor state (PDX1+/NKX6.1+ cells) to an endocrine progenitor state biased towards the beta-cell lineage (PDX1+/NKX6.1+ cells that are also NEUROD1+), to immature beta cells (PDX1+/NKX6.1+/ NEUROD1+ cells that are insulin+/glucagon–/somatostatin–), and finally to maturing beta cells that gain expression of MAFA (PDX1+/NKX6.1+/ NEUROD1+/MAFA+ cells that are insulin+/glucagon–/somatostatin–).

1BetaLogics

Venture, Janssen R&D LLC, Raritan, New Jersey, USA. 2Department of Cellular and Physiological Sciences, Life Sciences Institute, University of British Columbia, Vancouver, British Columbia, Canada. 3Department of Surgery, University of British Columbia, Vancouver, British Columbia, Canada. Correspondence should be addressed to T.J.K. ([email protected]) or A.R. ([email protected]). Received 4 July; accepted 5 September; published online 11 September 2014; corrected online 16 September 2014; doi:10.1038/nbt.3033

NATURE BIOTECHNOLOGY VOLUME 32 NUMBER 11 NOVEMBER 2014

1121

A RT I C L E S a

Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 Stage 6 Stage 7 Figure 1 Overview of Definitive Primitive Posterior Pancreatic Pancreatic Immature Maturing differentiation protocol endoderm gut tube foregut endoderm endocrine precursors beta cells beta cells and characterization of 3d 2d 2d 3d 3d 7–15 d 7–15 d S4–S6. (a) Summary 10 mM glucose; planar culture 20 mM glucose; air-liquid interface of seven-stage GDF8 FGF7 FGF7 FGF7 SANT ALK5 inh II ALK5 inh II differentiation protocol, GSK3β inh VitC VitC VitC 50 nM RA T3 T3 including the important 1 μM RA 100 nM RA ALK5 inh II LDN N-Cys growth factors and SANT T3 GS inh XX AXL inh SANT small molecules that TPB TPB LDN were added at each LDN LDN stage. Key markers of the differentiating pancreatic endocrine LEGEND: FOXA2+ PDX1+ PDX1+/NKX6.1+ PDX1+/NKX6.1+/NEUROD1+ PDX1+/ NKX6.1 +/NEUROD1+/ MAFA+ INS + /GCG–/SST– cells are illustrated below (INS, insulin; GCG, glucagon; SST, GCG SST INS CHGA NGN3 1,500 6 × 10 6 × 10 4 × 10 4 × 10 S4 somatostatin). (b) Gene *b S5 *d expression profile of 3 × 10 3 × 10 1,000 4 × 10 4 × 10 *c S6 D7 hESC-derived cells 2 × 10 2 × 10 at S4, S5, S6 day S6 D14 2 × 10 500 * d 2 × 10 (D) 7 and S6 D14 *b Human islets * c 1 × 10 1 × 10 *b c * c of differentiation, a/c a/c a b a a a/b a a/b a a 0 0 0 0 compared with adult 0 human islet preparations NEUROD1 NKX6.1 PDX1 NKX2.2 SOX9 MAFB 1.5 × 10 1,800 2.0 × 10 150 * 8,000 a 1.5 × 10 (n = 6 biological b *c * * b * c b replicates for all 1,500 *c b b 1.5 × 10 6,000 a/b *b genes, except n = 4 for 1.0 × 10 1,200 100 1.0 × 10 b b b SOX9; n = 2 technical 1.0 × 10 b a/b 900 4,000 a replicates per point). 5.0 × 10 5.0 × 10 a 600 50 a * 5.0 × 10 * a 2,000 a Individual biological c c 300 a replicates are shown 0.0 0.0 0.0 0 0 0 on box and whisker plots. Different letters G6PC2 PCSK1 ABCC8 SLC30A8 MAFA PCSK2 3 × 10 8,000 3,000 2,000 1,000 1 × 10 represent significant * c * c 8 × 10 800 differences between 6,000 1,500 b * 2 × 10 2,000 hESC-derived cells 6 × 10 b 600 b * 4,000 1,000 throughout S4–S6 b b 4 × 10 400 1 × 10 1,000 (one-way ANOVA b b * 2,000 500 2 × 10 b b 200 a * with Tukey test for a a b c a a a a a a a a 0 multiple comparisons). 0 0 0 0 0 *P < 0.05 for each stage of differentiation S4 D3 100 80 versus human islets S5 D3 (one-way ANOVA 80 S6 D7 with Dunnett test for 60 multiple comparisons). S6 D14 60 (c) FACS quantification 40 40 of cells throughout differentiation (S4 D3, 20 20 S5 D3, S6 D7 and S6 D14) for the percentage 0 0 CHGA PDX1 NKX2.2 NKX6.1 PAX6 ISL1 PDX1+ INS+ CHGA+ INS+ INS+ INS– of cells expressing Ki67+ NKX6.1+ NKX6.1+ GCGGCG+ GCG+ various transcription factors and pancreatic Stage 4, day 3 Stage 5, day 3 Stage 6, day 7 Stage 6, day 14 hormones (S4, n = 1–6 0.5 105 26.7 42.3 105 29.3 51.5 105 4.1 78.7 biological replicates; 105 61.1 S5, n = 4 biological replicates; S6, n = 1–7 104 104 104 104 biological replicates). Data are presented as 103 103 103 103 individual biological replicates with bars 102 102 102 102 representing the mean. 0 3.9 0 3.3 0 0.3 0 36.6 1.8 27.0 16.0 16.9 (d) Representative FACS 2 3 4 5 2 3 4 5 2 3 4 5 2 3 4 plots illustrating protein 10 10 10 10 10 10 10 10 10 10 10 105 0 10 0 10 0 10 0 10 expression of NKX6.1 Chromogranin A and chromogranin A in populations of hESCderived S4, S5 and S6 cells. Red text indicates the % co-expression of NKX6.1 and chromogranin A.

b

4

6

6

4

6

6

5

4

6

6

5

4

6

4

4

4

4

3

3

4

4

4

Fold change relative to H1 cells


5

3

6

4

3

6

3

3

6

3

% of total population

d

NKX6.1

npg

c

1122

VOLUME 32 NUMBER 11 NOVEMBER 2014 NATURE BIOTECHNOLOGY

A RT I C L E S

npg


We describe a seven-stage in vitro differentiation protocol (Fig. 1a) that builds upon protocols previously used to specify pancreatic progenitors3–14. Introduction of vitamin C at early stages of differentiation results in production of PDX1+/NKX6.1+ pancreatic progenitors with low expression of NGN3 and its downstream targets (S4). Further differentiation of pancreatic progenitors using a particular combination of reagents, including an ALK5 inhibitor, BMP receptor inhibitor and thyroid hormone (T3), results in upregulation of NGN3 and cell populations of which a substantial fraction co-expresses PDX1, NKX6.1, NEUROD1 and NKX2.2 (S5). Continued exposure to ALK5 inhibitor, BMP receptor inhibitor and T3 with addition of a notch inhibitor results in the generation of cell populations in which a substantial percentage of PDX1+/NKX6.1+/NEUROD1+ cells express insulin but not glucagon or somatostatin (S6). Finally, screening of a number of additional reagents identified R428, an inhibitor of AXL, which, in combination with ALK5 inhibitor and T3, potently induces MAFA expression in PDX1+/NKX6.1+/NEUROD1+ cells that are insulin+/glucagon–/somatostatin– (S7). The resulting highly differentiated cells display certain key characteristics of mature beta cells, including glucose-induced insulin secretion, and rapidly reverse diabetes after transplantation in mice. RESULTS Optimization of S1–4 to generate PDX1+/NKX6.1+ cells We have previously reported the efficient generation of S4 pancreatic progenitor cells containing two distinct populations: polyhormonal (insulin+/glucagon+/somatostatin+) cells and PDX1+/NKX6.1+ cells3–6. The protocol presented here uses our previously published refinements3–6, including: (i) serum-free conditions; (ii) use of GDF8 (a TGFb family member) and a GSK3b inhibitor in place of the typical activin A/WNT3A combination for induction of definitive endoderm during S1; and (iii) FGF7, TPB (PKC activator) and LDN, a small-molecule inhibitor of the BMP receptor, in place of noggin to enhance the endoderm progenitor population yield at the end of S4. In the present work, we improved the previous protocol by addition of vitamin C during S2–4. This increased total cell numbers and confluency at S2,3 (data not shown), which was predicted given the known role of vitamin C in extracellular matrix production28. Moreover, addition of vitamin C at S2–4 reduced mRNA expression of NGN3, a master regulator of pancreatic endocrine cell differentiation23, and its downstream targets, including NEUROD1 and NKX2.2, at S3,4 while not affecting PDX1 expression (Supplementary Fig. 1). Suppression of NGN3 during early stages of differentiation is thought to be important as previous work suggested that premature induction of NGN3 in pancreatic endoderm cells primes the cells toward populations enriched with polyhormonal cells expressing glucagon and other hormones29. Inhibition of NGN3 at S3,4 while maintaining high coexpression of NKX6.1 and PDX1 distinguishes the present S4 cells from those generated by previous protocols4,5,12,13,30, although our new protocol still produces a small fraction of polyhormonal cells. Generation of S5 NKX6.1+/NEUROD1+ cells Our goal for S5 was to begin inducing the pancreatic endocrine program, marked by expression of NGN3 and NEUROD1. We first investigated the effect of moving from planar culture throughout S1–4 to an air-liquid interface during later stages, in which S4 cells are spotted onto a filter insert to create ~1- to 2-mm diameter cell clusters (Supplementary Fig. 2a). An air-liquid interface culture environment allows for basal and apical polarity of cells31,32 and exposure to atmospheric oxygen levels33, factors that may modulate beta-cell differentiation from endocrine progenitors34. Air-liquid interface culture during S5 resulted in upregulation of NGN3 transcript, as well as the pancreatic hormones insulin (INS)

and glucagon (GCG), compared with planar culture (Supplementary Fig. 2b). We next tested a panel of ALK5 (TGFb receptor) inhibitors during S5, as inhibition of TGFb signaling was previously reported to induce endocrine development in pancreatic progenitors generated from pluripotent stem cells5,7,14. Consistent with our previous work7, the addition of ALK5iII during S5 induced mRNA expression of NGN3 and strongly induced the pancreatic hormones INS, GCG and somatostatin (SST) transcripts compared with all other ALK5 inhibitors tested (Supplementary Fig. 3). We suspect that effective inhibition of ALK4 (closely related to ALK5 and a member of the TGFβ receptor I family35) coupled with minimal inhibition of other kinases (Supplementary Table 1) made ALK5iII the most effective inhibitor. We also examined the effect of thyroid hormone (T3) given recent evidence that it promotes pancreatic beta-cell maturation in rats36. Notably, exposure to T3 during S5 did not affect the total percentage of cells expressing NKX6.1 protein but enhanced co-expression of NKX6.1 and insulin (Supplementary Fig. 4). With the optimized S5 differentiation protocol, hESC-derived cells maintained robust co-expression of NKX6.1 and PDX1, similar to S4 cells from both the present differentiation protocol and our previously published four-stage protocol4 (Supplementary Fig. 5). In addition, a subset of the PDX1+/NKX6.1+ S5 population expressed key endocrine precursor markers, including NGN3, NKX2.2 and NEUROD1 (Supplementary Fig. 6). A fraction of NKX6.1+ cells expressed SOX9 during S5 (Supplementary Fig. 6a), although we noted a continuous drop in SOX9 mRNA expression from S4–6 (Fig. 1b), consistent with previous results showing loss of SOX9 expression during maturation of beta-cell precursors37. NGN3 mRNA expression was transiently induced in the S5 population along with its downstream targets, including NEUROD1, NKX2.2 and NKX6.1 (Fig. 1b). Moreover, the proliferative capacity of PDX1+/NKX6.1+ cells began to decline in S5 and continued through S6; we also began to observe co-expression of chromogranin A and insulin with NKX6.1 in S5 (Fig. 1c,d). Generation of S6 NKX6.1+/insulin+ cells Our goal for S6 was to differentiate the PDX1+/ NKX6.1+/NEUROD1+ cells generated in S5 into cells that express insulin but not other pancreatic hormones, such as glucagon and somatostatin. We first observed that continued exposure of S5 cells to ALK5iII resulted in downregulation of NGN3 (in contrast to induction during S5) and strong upregulation of INS, GCG, and SST transcript levels at S6 compared to other TGFβ inhibitors (Supplementary Fig. 3). Addition of T3 (along with ALK5iII) during S6 at a dose of 1 mM induced expression of INS and mature beta-cell markers (Supplementary Fig. 7), but the combination of ALK5iII, T3 and LDN resulted in the most effective mRNA induction of INS and GCG, along with NKX6.1 and two markers of more mature beta cells, MAFA and ABCC8 (Supplementary Fig. 4a). We next tested the effect of adding a Notch pathway inhibitor, gamma secretase inhibitor XX (GSiXX), during S6 based on evidence that misexpression of activated Notch in PDX1+ progenitor cells prevents differentiation of pancreas lineages38, and the reported role of Notch signaling in the pancreatic endocrine/exocrine fate choice39. GSiXX treatment induced mRNA expression of genes involved in beta-cell maturation and inhibited expression of PTF1A, a marker of the pancreatic exocrine lineage (Supplementary Fig. 4b). Combining GSiXX and T3 increased the fraction of NKX6.1+/insulin+/glucagon– cells from ~25% to ~50% relative to cultures without GSiXX (Supplementary Fig. 4c). We also noted that similar populations of NKX6.1+/insulin+/glucagon– cells could be generated by transferring S4 cells made using our previous protocol4 into suspension culture, rather than air-liquid interface, and culturing for 7 d


1123

A RT I C L E S

0 10

102 0 59.1 12.1 2

Insulin

10

3

10

10

4

10 3.2 104

Glucagon

42.4

0 10


10

5

c

1.1

103

44.9 2

Insulin

10

3

10

4

10

5

102 0

43.1

0 10

35.2 2

10

3

10

4

10

50

*

40

Insulin PDX1 DAPI

0.2

S6 cells Human islets

10 3.3 16.7 3.3 16.7 Glucose (mM)

Insulin NKX2.2 DAPI

60 30

d

150

*

100 50 0

600

Batch 1 Batch 2

400

200

0 15 G

KCI

Arg

Treatment

5

f

*

90

0

0.0

20

Synaptophysin CK19 Trypsin DAPI Stage 4 Stage 6

Insulin NKX6.1 DAPI

0.4

30

0

120

*

0.6

Insulin Glucagon Somatostatin DAPI Stage 4 Stage 6

Insulin MAFA DAPI

Insulin PCNA DAPI Insulin NEUROD1 DAPI

Stage 6

npg

Stage 4

g

10

4

104

103

e

10

3

1 0

2.7 2

9.5 105 20.6

5

102 0

0 10

5

2

0.8

Proinsulin (ng/ml)/ DNA (μg/ml)

102 0 8.0

*

3

Glucagon (ng/ml)/ DNA (μg/ml)

103

NKX6.1

103

4

4

Human insulin (μU/ml)

10

4

31.3

Insulin (μg/ml)/ DNA (μg/ml)

10

b

Human islets

44.0 105 7.0

C-peptide (μg/ml)/ DNA (μg/ml)

Stage 6, day 14 105 36.0

Human C-peptide (ng/ml)

a

Figure 2 S6 insulin+ cells co-express key transcription factors, such as NKX6.1, unlike S4 pancreatic progenitor cells. (a) Representative FACS plots for S6 cells (day 14) compared with adult human islet preparations for expression of insulin with either NKX6.1 or glucagon. Red text highlights the % of NKX6.1+/insulin+ cells or insulin+/glucagon– cells. Additional human islet FACS plots are provided in Supplementary Figure 11. (b) Hormone content relative to DNA content for different batches of S6 cells and human islet preparations (insulin/proinsulin: n = 8 for human islets (including three biological and two or three technical replicates), n = 6 for S6 cells (including three biological and two technical replicates); glucagon: n = 9 for human islets (three biological and two to four technical replicates), n = 6 for S6 cells (two biological and two to four technical replicates); C-peptide: n = 6 for human islets (including three biological and three technical replicates) and n = 4 for S6 cells (including two biological and two technical replicates)). Data are presented as box and whisker plots showing individual replicates. *P < 0.05, unpaired two-tailed t-test. (c) Human C-peptide secretion from S6D7 cells and human islet preparations in response to low and high glucose concentrations under static conditions (S6D7 cells: n = 5 biological replicates, human islets: n = 5 biological replicates). Data are presented as box and whisker plots showing individual replicates. *P < 0.05, paired one-tailed t-test (3.3 versus 16.7 mM). (d) Human insulin secretion in response to various secretagogues (G, glucose; KCl; Arg, arginine) in a perifusion assay (n = 3 technical replicates per batch). Data presented as mean ± s.e.m. (e–g) Representative immunofluorescent staining of S4 cells (generated according to our previously published four-stage protocol5) and S6 cells (generated with the current six-stage protocol) for: synaptophysin (endocrine marker, red), CK19 (ductal marker, green) and trypsin (exocrine marker, blue) (e); insulin (red), glucagon (green) and somatostatin (blue) (f); and co-immunostaining of insulin (red) with various transcription factors and proliferation marker PCNA (green) (g). Nuclear DAPI staining is shown in gray. Scale bars, in low magnification images, 500 mm (e,f), high magnification insets,125 mm (f); 100 mm (g).

1124


npg


A RT I C L E S with the S5 and/or S6 factors ALK5iII, T3, GSiXX, LDN and heparin (Supplementary Fig. 8). Thus, our protocol may be amenable to largerscale suspension cultures to generate clinically relevant cell numbers. With our optimized S6 protocol, we consistently generated populations with similar or higher transcript levels of several key transcription factors (NEUROD1, PDX1, NKX6.1, NKX2.2, MAFB) compared with adult human islets, but lower levels of INS, GCG and SST, and mature beta-cell markers (e.g., MAFA and G6PC2) (Fig. 1b). Notably, analysis of protein expression in single cells revealed that S6 populations derived from hESCs contained ~50% NKX6.1+/insulin+ cells (Figs. 1c and 2 and Supplementary Fig. 9). NKX6.1+/insulin+ cells were also generated from a human induced pluripotent stem cell (iPSC) line, although less efficiently (Supplementary Fig. 10). The fraction of NKX6.1+/ insulin+ cells and the fraction of insulin+/glucagon– cells were higher in hESC-derived S6 cells than in adult human islet preparations (Fig. 2a; fluorescence-activated cell sorting (FACS) plots for human islets are representative of typical preparations, with a second batch of islets shown in Supplementary Fig. 11). This is likely because of contamination of human islet preparations with acinar and ductal tissues (Supplementary Fig. 12)40,41. Islet impurity as well as the cell damage and stress associated with islet isolation contribute to substantial variability between islet preparations and serve as caveats to considering human islet preparations as the gold-standard positive control. Approximately 50% of S6 cells expressed insulin but not glucagon (Figs. 1c and 2a). Many of the insulin+/glucagon– cells expressed key beta-cell transcription factors such as NKX6.1, PDX1, NKX2.2 and NEUROD1 (Fig. 2a,g and Supplementary Fig. 9b,d). Moreover, the S6 population displayed a sharp reduction in CK19+ cells (Fig. 2e) and CK19+/PDX1+/NKX6.1+ cells (Supplementary Fig. 13a) compared with S4 progenitor cells generated with our previous four-stage differentiation protocol4. This is consistent with the known expression of CK19 in human pancreatic progenitors and its subsequent loss following NGN3 induction and beta-cell differentiation42. In sharp contrast to the S6 population, S4 cells generated with the previous or current protocols contained ~60% PDX1+/NKX6.1+ cells (Supplementary Fig. 5 and Supplementary Fig. 13a), but 0.05) from pre-STZ). In contrast, S6 cells reversed STZ-induced hyperglycemia in ~8–12 weeks in three independent cohorts of diabetic mice (Fig. 3e, P > 0.05 for day 83 versus pre-STZ; Fig. 3g, P > 0.05 for day 62 versus pre-STZ and Supplementary Fig. 14; P > 0.05 for day 60 versus pre-STZ), and survival nephrectomy confirmed that glycemic control was attributed to the engrafted insulin-secreting cells (Fig. 3e and Supplementary Fig. 15a). Diabetic recipients of S6 cells had significantly improved glucose excursions following oral meal (Supplementary Fig. 15b, P < 0.0001) and intraperitoneal (i.p.) glucose challenges (Fig. 3h, P = 0.0008) compared with S4 recipients at 16 and 20 weeks post-transplant, respectively, presumably as a result of higher human C-peptide levels produced by S6 cells (Supplementary Fig. 15c and Fig. 3a,b,i). Moreover, at 20 weeks S6 cells showed statistically significant glucose-induced human C-peptide secretion, whereas S4 cells did not (Fig. 3i). Human C-peptide levels peaked at 30 min and returned to basal by 60 min with S6 cells (Fig. 3i), a pattern confirmed in a second cohort of diabetic mice at just 10 weeks post-transplant (Supplementary Fig. 14b). Consistent with our previous studies5, mature S4 kidney capsule grafts contained a small trypsin+ population, whereas trypsin immunoreactivity was rare in S6 grafts (Fig. 3j and Supplementary Fig. 16). The presence of trypsin+ cells is not unexpected given that S4 cells are pancreatic progenitors that can give rise to all pancreatic lineages. However, dilated ducts were observed in about half of the S6 grafts, but were only rarely detected in the S4-derived tissues (Supplementary Fig. 16), despite S4 cells containing a much higher proportion of CK19+ cells than S6 cells before transplant (Fig. 2e). Notably, the CK19+ ductal structures in both S4 and S6 grafts maintained expression of PDX1, but S6 grafts showed reduced NKX6.1 expression in the ductal compartment compared with S4 grafts (Supplementary Fig. 13b). In the endocrine compartment, the distribution of pancreatic hormones was similar between S4 and S6 grafts (Fig. 3k), and insulin/C-peptide+ cells derived from both S4 and S6 grafts expressed PDX1, NKX6.1, NKX2.2 and MAFA (Fig. 3l). Moreover, co-localization of PDX1/NKX6.1, PDX1/MAFA and NKX6.1/NKX2.2 was observed in endocrine regions of the grafts (Supplementary Fig. 17). PCNA immunoreactivity was rarely observed in insulin+ cells from either S4 or S6 grafts (Fig. 3l). S7 cells express MAFA in NKX6.1+/insulin+ cells Our first goal for S7 was to identify compounds that induce expression of MAFA (Fig. 4), given the key role of MAFA in regulating genes important for beta-cell maturation, including those involved in glucoseregulated insulin secretion24–27,43. We first screened a library of >40 small molecules and growth factors and identified R428, a selective small-molecule inhibitor of the tyrosine kinase receptor AXL, as an inducer of MAFA (Fig. 4a). Notably, growth arrest specific protein 6 (GAS6), an agonist of the AXL receptor tyrosine kinase subfamily, has been proposed to play a role in beta-cell maturation through downregulation of Mafa expression in rodents44. Indeed, R428 dose-dependently upregulated MAFA mRNA levels during S7 culture to approximately half that of adult human islets (Fig. 4b). We also investigated the effect of N-acetyl cysteine (N-Cys) during S7, as previous work suggested that antioxidants may promote maintenance of nuclear MAFA protein in rodents45–47. Although N-Cys treatment did not affect MAFA tran-


1125

A RT I C L E S a

c

j

Blood glucose (mM )

h



Blood glucose (mM)

f





e

k

Stage 4

l

Stage 6


g

Blood glucose (mM)

d

npg

b

Stage 4 cells Stage 6 cells Human islets Figure 3 S6 insulin+ cells develop 5 14 5 * faster in vivo than S4 pancreatic 12 * 4 4 * * progenitor cells. (a–c) S4 cells * 10 5 3 * 3 (generated as previously described , 8 * * * * * * ~5 × 106 cells/animal; n = 7/cohort), 2 6 2 * 4 S6 cells (~2.5 × 106 cells/animal; 1 * 1 2 n = 7/cohort) and human islets 0 0 0 1 4 8 (2,000 islet equivalents/animal; 2 4 8 12 16 2 4 8 12 16 2 4 8 12 16 2 4 8 121620 2 4 8 121620 2 4 8 1216 20 Time post-tx Time post-tx (weeks) Time post-tx (weeks) n = 10) were transplanted under the (weeks) Cohort 1 Cohort 2 Cohort 3 kidney capsule of nondiabetic male Cohort 1 Cohort 2 Cohort 3 NSG mice and random fed human * * 14 4 4 4 40 hESC tx 4 C-peptide levels were measured at * * Fasted Post-glucose 12 Nephrectomy the indicated times up to 20 weeks * * 3 3 3 30 3 10 post-transplant (tx). Human C-peptide STZ 8 * 2 2 2 * 20 2 levels from individual mice are shown 6 on box and whisker plots. *P < 0.05, 4 1 1 10 1 1 one-way repeated measures ANOVA 2 0 0 0 0 0 (Dunnett test for multiple comparisons 0 –15 0 15 30 45 60 75 90 105 2 4 8 12 16 Human 16 weeks 8 weeks 12 weeks versus 1 or 2 weeks post-transplant). Day Time post-tx islets Stage 6 cells (weeks) (d) Human C-peptide levels were hESC measured after an overnight fast 35 6 25 i *b * S4 cells tx b *b 30 and 60 min following an i.p. glucose b S6 cells STZ *b *b 20 25 bolus at 8 (n = 7), 12 (n = 7) and 4 a 20 15 b 16 (n = 10) weeks post-transplant of a 15 S6 cells and in mice engrafted with 10 a 2 a b 10 a 2,000 human islet equivalents (n a a a 5 a 5 a = 6). *P < 0.05, paired one-tailed 0 0 0 0 30 60 90 120 0 15 30 45 60 –30 0 30 60 90 120 150 180 t-test. (e,f) Diabetic cohort 1: S6 Time (min) Time (min) Day 6 cells (~2.5 × 10 cells/animal) were Synaptophysin CK19 Trypsin DAPI Insulin Glucagon Somatostatin DAPI transplanted into male NSG mice Stage 4 Stage 6 Stage 4 Stage 6 with streptozotocin (STZ)-induced diabetes (n = 4). (e) Blood glucose was assessed weekly throughout the study and a survival nephrectomy was performed on all mice on day 93 to remove the engrafted kidney (indicated by red line). Data are presented as mean ± s.e.m. (f) Random fed human C-peptide levels were measured at the indicated times up to 16 weeks post-transplant. C-peptide levels from individual mice are shown on box and whisker plots. *P < 0.05 (one-way repeated C-Peptide NKX6.1 DAPI Insulin PDX1 DAPI C-Peptide NKX2.2 DAPI C-Peptide MAFA DAPI Insulin PCNA DAPI measures ANOVA with Dunnett test for multiple comparisons versus 2 weeks post-transplant). (g–i) Diabetic cohort 2: S4 cells (generated as previously described5, ~5 × 106 cells/animal; n = 10) and S6 cells (~2 × 106 cells/animal; n = 8) were transplanted into male SCID-beige mice with STZ-induced diabetes. All data are presented as mean ± s.e.m. (g) Fasting blood glucose levels were measured throughout the study duration in a subset of mice (S4, n = 4; S6, n = 5). Red and blue arrows indicate when a mouse was euthanized. (h,i) Blood glucose and human C-peptide levels were measured during an i.p. glucose tolerance test at 20 weeks post-transplant (S4, n = 4; S6, n = 7; *P < 0.05, two-way repeated measures ANOVA with a Bonferroni test for multiple comparisons between S4 versus S6 at each time point; different letters indicate significant differences between time points within each group (two-way repeated measures ANOVA with a Dunnett test for multiple comparisons versus t = 0)). (j–l) Representative immunofluorescent staining of S4 and S6 cells engrafted under the kidney capsule and harvested at 115 d post-transplant. Grafts were stained for synaptophysin (endocrine marker, red), CK19 (ductal marker, green) and trypsin (exocrine marker, blue) (j); insulin (red), glucagon (green) and somatostatin (blue) (k); co-expression of insulin or C-peptide (red) with transcription factors and proliferation marker PCNA (green) (l). Nuclear DAPI staining is shown in gray. Scale bars, in low magnification images, 500 mm, high magnification insets, 125 mm (j,k); 100 mm (l).

1126


A RT I C L E S script levels at any dose during S7 (Fig. 4c), 1,000–2,000 mM N-Cys increased nuclear MAFA protein (Fig. 4g). NKX6.1 immunoreactivity was unaffected by exposure to any dose of N-Cys (Fig. 4g). Addition of vitamin E, another antioxidant, did not increase MAFA protein levels (data not shown). ALK5iII was also tested during S6 and S7 and found to be a potent inducer of MAFA mRNA, particularly at a dose of 10 mM during S7 (Fig. 4d). Overall, the complete S7 medium

400

200

200

0

0

100 μM N-Cys

1,000 μM N-Cys

Human islets

S7 D14–21

Glucose (mM) Basal Complete S7 media S7 media

S6 D7

ALK5iII (µM) Stage 6 Stage 7

0

S5 D3

20

5

0

5

0

1,000

S7 D7–8

N-Cys ( µM)

5,000 Human islets

1,000

2,000

0

100 500

5.0

Human islets

2.0

1.0

0

0

500

2,000 μM N-Cys

Insulin NKX6.1

0 μM N-Cys

2,000

100

100

0

3,000

1,000

200

200

200

f 4,000

20

400

e 1,500

400 300

300

R428 (µM)

g

400

1 No data 5 10

600

d

500

1 5 10

c

800

Human islets PDGF AA 10 ng/ml PDGF AA 50 ng/ml PDGF BB 10 ng/ml PDGF BB 50 ng/ml Serotonin HCl 50 ng/ml GDF11 10 ng/ml GDF11 20 ng/ml R428 2 µM Chemerin 100 ng/ml Panobinostat 1 µM SAH A 1 µM SBH A 1 µM Scriptaid 1 µM Trichostatin A 1 µM Tubacin 1 µM NSC625987 1 µM ML141 1 µM LDN211904 1 µM CPI203 1 µM P7C3 1 µM AdipoRon 1 µM LY4 11575 1 µM DMSO CI-994 1 µM ZLN005 1 µM WDR5-C47 1 µM MS-275 1 µM M344 1 µM CS2100 1 µM ZCL278 1 µM OC 1 ng/ml OC 10 ng/ml OC 100 ng/ml WNT5A 50 ng/ml WNT5A 100 ng/ml CTGF 100 ng/ml

400

Human islets Activin A 10 ng/ml Activin A 20 ng/ml TGF 10 ng/ml TGF 20 ng/ml Wnt4A 10 ng/ml Wnt4A 50 ng/ml Untreated Muscimol 10 µM Muscimol 50 µM Antioxidant 1:500 Antioxidant 1:1000 DMSO SKP2-C25 1 µM SKP2-C25 2 µM DMSO DMSO GDF8 10 ng/ml GDF8 20 ng/ml BMP2 10 ng/ml BMP2 20 ng/ml BMP5 10 ng/ml BMP5 20 ng/ml DMSO JIB04 1 µM JIB04 2 µM Nicotinamide 1 mM Nicotinamide 10 mM DKK1 10 ng/ml DKK1 50 ng/ml IWP2 1 µM IWP2 2 µM XAV939 1 µM XAV939 2 µM

600

0.1

MAFA (fold change relative to H1 cells)

b

600

Insulin MAFA

npg


a

formulation (including the AXL inhibitor, N-Cys, ALK5iII and T3) produced a ~16-fold induction of MAFA transcript levels under both low and high glucose conditions (Fig. 4e). Moreover, exposure of differentiating cells to the complete S7 medium resulted in increasing MAFA transcript levels over time, such that on days 14–21 of extended S7 culture, they were approximately double those of human islets (Fig. 4f).

Figure 4 Optimization of S7 differentiation protocol for induction of MAFA expression. (a) Results of small-molecule screen for induction of MAFA gene expression (green circles); MAFA levels in human islets are shown as a reference (gray circles, n = 3; mean ± s.e.m.). R428 (AXL inhibitor; red square) was identified as a hit compound. For all conditions, S7 basal media was used (BLAR with ITS, BSA, heparin and zinc-sulfate) plus ALK5iII and T3. (b) Addition of R428 during S7 caused dose-dependent induction of MAFA gene expression at S7 D7 (n = 4, including two biological and two technical replicates). (c) Addition of N-Cys during S7 did not affect MAFA gene expression at S7 D8 (n = 2 technical replicates); ALK5iII, T3, heparin and vitamin E were included in S7 culture media for all conditions. (d) ALK5iII dose-dependently increased mRNA expression of MAFA during both S6 and S7 (n = 4, including two technical and two biological replicates per condition). (e) Addition of the complete S7 media formulation (including ALK5iII, N-Cys, T3 and R428) resulted in a strong induction of MAFA under both 5 mM and 20 mM glucose conditions (n = 4 per condition, including two technical and two biological replicates). (f) MAFA gene expression increased over time throughout extended culture of S7 cells (n = 2 biological replicates), resulting in higher MAFA levels compared with human islets (n = 10 biological replicates). Data in a–f are presented as individual biological replicates on box and whisker plots. (g) Immunofluorescent staining for insulin (green) with MAFA (red) or NKX6.1 (red) on S7 cells exposed to different doses of N-Cys. Scale bars, low magnification images, 100 mm, and high magnification insets, 10 mm.


1127

A RT I C L E S

GCG

*

3 × 104

2 × 106 1 × 107 1 × 106

3,000

* 2,000

*

PDX1

*

2 × 105

3 × 104

6,000

100

4,000

0

0

8,000

NKX6.1 3 × 104

NKX2.2 3,000

PAX4 3 × 104

*

S7 cells

*

Human islets

2,000

50

*

CHGB

3,000

1,000

0

1,000

*

0

PAX6

1,500

*

MAFA

600

2 × 104

2,000

2 × 104

1,000

400

1 × 104

1,000

1 × 104

500

200

0

GCK

15

4,000

10

2,000

5

40 20

3,000

0

0

105 14.5

40.0

*

104

104

103

103

102 0 10.0 35.5 0102 103 104 105 NKX6.1

MAFB

*

3,000

0

43.5 0102 103 104 105 Insulin

6.

2 × 10

5

*

0

SLC30A8

*

9,000

UCN3

*

6,000

1 × 105 3,000

0

d

5 4 3 2 1 0

0

0

0.6 0.4 0.2 0.0

400 300 200 100 0

N

H

KX

2,000

PCSK2

1,000

11.0

102 0 41.7

5,000 4,000

1,000

0

105 3.7

0

0

PCSK1

2,000

2,000

FO

X 6. A2 + 1+ /IN 1 S+ + N KX /C 6. -Pe 1+ p+ /C H IN GA + S+ /G C IN S+ GPD /GC X1 G + + /K i6 7+

0

*

6,000

Glucagon

60

8,000

KCNK3

4,000

c

80

0

0

KCNK1

*

0

KX

5,000

*

2,000

150

0

2,000

0

G6PC2

100

N

CHGA

4,000 2 × 104

1 × 104

0

3,000

*

1 × 105

2 × 104 1,000

NGN3

5

6,000

4 × 104

200

c an ells is le ts

0

ABCC8 4 × 104

6 × 104

250

um

0

ISL1

*

Proinsulin (ng/ml)/ DNA (μg/ml)

2,000

1 × 10

0

IAPP

S7

4,000

2 × 104

1 × 104

Insulin

Fold change relative to H1 cells

4 × 10

% of total population


b

6,000

4

2 × 105

3 × 105

*

H

*

6 × 104

2 × 104

PPY

c an ells is le ts

NEUROD1 8,000

3 × 105

um

8 × 104

4 × 105

*

H

0

SST

C-peptide (μg/ml)/ DNA (μg/ml)

0

4 × 104

S7

3 × 106

Insulin (μg/ml)/ DNA (μg/ml)

INS

2 × 107

S um 7 c an ells is le ts

a

npg

e

Synaptophysin CK19 Trypsin DAPI Stage 7 Human islets

f

Insulin Glucagon Somatostatin DAPI Stage 7 Human islets

Insulin PCNA DAPI InsulinNEUROD1 DAPI

Human islets

Stage 7

g Insulin NKX6.1 DAPI Insulin PDX1 DAPI Insulin NKX2.2 DAPI Insulin MAFA DAPI

Figure 5 Stage 7 cells have a similar gene expression profile to human islets, and insulin+ cells co-express key beta-cell maturation markers, including MAFA. (a) Gene expression profile of S7 day (D) 13 cells relative to human islets (n = 3 biological replicates per group, except n = 2 for GCK). Data are expressed as the fold change relative to pluripotent hESCs (H1) and raw Ct values are provided in Supplementary Table 2. *P < 0.05; two-tailed t-test. (b,c) FACS quantification of the S7 population (n = 4 biological replicates) (b) and representative FACS plots showing expression of insulin with either NKX6.1 or glucagon (c), during S7 culture. Red text highlights the % of NKX6.1+/insulin+ cells or insulin+/glucagon– cells. (d) Hormone content relative to DNA content for S7 cells and human islets (insulin, proinsulin and human C-peptide: n = 4, including two biological and two technical replicates). (e–g) Representative immunofluorescent staining of S7 cells and a typical preparation of human islets for synaptophysin (endocrine marker, red), CK19 (ductal marker, green) and trypsin (exocrine marker, blue) (e); insulin (red), glucagon (green) and somatostatin (blue) (f); and co-expression of insulin (red) with various transcription factors and proliferation marker PCNA (green) (g). Nuclear DAPI staining is shown in gray. Scale bars, in low magnification images, 500 μm, and high magnification insets, 100 μm (e,f); 100 μm (g). For all data, individual biological replicates are shown on box and whisker plots.

1128


S7 cells have functional similarities to human islets To determine whether our protocol generated insulin-secreting cells capable of responding acutely to glucose, we assessed the ability of hESC-derived S7 cells to rapidly increase cytosolic Ca2+ concentrations in response to glucose and subsequently return to baseline. We first confirmed that S7 cells and mature human islets maintained robust nuclear expression of MAFA in insulin+ cells following dispersal and seeding onto coverslips (Supplementary Fig. 20). Depending on the batch, 5–10% of S7 cells exhibited a significant Ca2+ response to 20 mM glucose above the stable baseline, whereas robust glucose-induced signaling was observed in virtually every human beta

Stage 7 cells

0.4 ~5–10% 0.1 0.7

10 min

20 mM Glucose

KCl

0.5 0.3

~30 % Ex4

0.1

20 mM Glucose

KCl

0.7 0.5 0.3

~20 % Ex4

0.1

b

KCl

Stage 7 cells

175 150 125 100 75 50 25

Ex4

0 15 min

20 mM Glucose

Cytosolic Ca2+ (340/380) Cytosolic Ca2+ (340/380) Cytosolic Ca2+ (340/380)

0.7


Cytosolic Ca2+ (340/380) Cytosolic Ca2+ (340/380) Cytosolic Ca2+ (340/380)

a


We next compared expression of key pancreatic beta-cell markers in S7 cells and human islets (Fig. 5). S7 day 13 cells were more similar than S6 cells to human islets. For example, INS, MAFA and G6PC2 transcript levels were indistinguishable between S7 cells and human islet preparations (Fig. 5a; all Ct values provided in Supplementary Table 2), whereas these genes had been expressed at relatively low levels in S6 cells (Fig. 1b). Although most pancreatic betacell transcription factors and maturation markers were expressed at levels similar or higher to those of adult human islet preparations, we did note several genes whose expression remained significantly (P < 0.05) lower at S7 (IAPP, CHGB, KCNK1, KCNK3, UCN3; Fig. 5a). The proportion of insulin+ cells co-expressing NKX6.1+ remained stable during S7 (Fig. 5b,c) compared with S6 (Fig. 1c, 2a), and the proportion of endocrine cells was consistently greater than in typical human islet preparations (Fig. 5e,f and Supplementary Fig. 12). However, S7 cells had moderately higher proinsulin content compared with human islets, suggesting a possible deficiency in insulin processing (Fig. 5d). Immunofluorescence staining revealed that the vast majority of S7 cells were endocrine (synaptophysin+; Fig. 5e), insulin+ (Fig. 5f) and expressed key transcription factors, such as NKX6.1, PDX1, NKX2.2 and NEUROD1 (Fig. 5g). Furthermore, we consistently achieved robust nuclear MAFA expression in the S7 insulin+ population (Fig. 5g), which was only sporadically observed during S6. Overall, we estimate that our protocol yields one NKX6.1+/insulin+ cell at S7 from every two hESCs. Our seven-stage protocol also induced the pancreatic endocrine program in an iPSC line and generated NKX6.1+/insulin+ and insulin+/MAFA+ cells during S7 (Supplementary Figs. 18 and 19a–c), although not as efficiently as with the H1 hESC line used in our studies (Supplementary Fig. 19c,d).

KCI

4

*

* 3

Human beta cells 0.7 0.5 0.3

>70%

0.1

20 mM Glucose

KCl

0.7 0.5 0.3

>70%

0.1

Ex4 20 mM Glucose

KCl

0.7 0.5 0.3

>50% KCl

Ex4

0.1

Human islets

175 150 125 100 75 50 25

Ex4

0

20 mM Glucose

d

c C-peptide (fold over basal)

npg


A RT I C L E S

Stage 7 cells

KCI

Human beta cells

Human islets S7 cells

2

1

0

3.3 16.7

3.3 16.7

Glucose (mM)

Figure 6 Stage 7 cells have certain functional similarities to human beta cells, but cannot be considered mature human beta cells. (a) Representative examples of calcium signaling traces from S7 day (D) 8 cells (green, left) and human beta cells (black, right). Three different experiments were performed: 20 mM glucose alone, 20 mM glucose ± exendin-4 (Ex4; 10 nM) and exendin-4 alone (10 nM). KCl (30 mM) was included at the end of all three experiments and baseline glucose was always 3 mM. (b) Secretion of human insulin by S7D17 cells (green, left; n = 6, including two biological replicates and three technical replicates per batch) and human islets (black, right; n = 3 technical replicates) in response to 20 mM glucose ± 10 nM exendin-4 and 30 mM KCl within a perifusion system. Data are presented as mean ± s.e.m. (c) Static glucose-stimulated human C-peptide secretion by S7 cells (n = 17, including five biological replicates and two to six technical replicates per batch) compared with adult human islets (n = 4 biological replicates with two technical replicates per batch). Data are expressed as the fold change relative to basal C-peptide levels and individual biological replicates are shown on a box and whisker plot. Raw C-peptide values are provided for individual experiments in Supplementary Figure 22. *P < 0.05, paired one-tailed t-test. (d) Transmission electron microscopy images of different fields of view (top and bottom) of S7 cells (left) and human islet beta cells (right) illustrating the ultrastructure of endocrine granules. Insulin granules in human beta cells could be categorized into three main types: pale, diffuse gray core (top inset); dense round core (middle inset); dense rod-shaped core (bottom inset). Examples of each type of insulin-like endocrine granule were observed in hESC-derived S7 cells. Scale bars, in low magnification images, 500 nm, and high magnification insets, 125 nm.


1129

A RT I C L E S

a

b

c

e

Blood glucose (mM)

f

j

i

C-Peptide NKX6.1 NKX2.2 DAPI

C-Peptide NKX2.2 DAPI

C-Peptide NKX6.1 DAPI

C-Peptide MAFA DAPI

h


Blood glucose (mM)




g


Human islets


d

npg

Stage 7 cells

Figure 7 S7 cells rapidly 8 8 10 * * Fasted 25 S7 tx * reverse diabetes and Post-glucose STZ 8 20 6 6 develop glucose-responsive * 6 Nephrectomy 15 insulin secretion in vivo. 4 4 4 (a,b) S7 cells (~1.25 × 10 2 2 106 cells/animal; n = 7 2 5 per cohort) or human islets 0 0 0 0 2 12 16 2 4 8 2 4 8 -10 0 10 20 30 40 50 60 70 80 6 weeks (4,000 islet equivalents/ Time post-tx (weeks) Time post-tx Time post-tx Day animal; n = 10) were (weeks) Cohort 1 Cohort 2 transplanted under the kidney capsule of * 20 20 20 20 20 20 15 nondiabetic male NSG Nondiabetic, no tx Fasted Fasted * Post-glucose Post-insulin S7 cell recipients 16 mice. (a) Human C-peptide 15 15 15 15 15 10 levels were measured under 12 * 10 10 10 10 10 random fed conditions at * * 8 * 5 the indicated time points 5 5 5 5 5 4 2–16 weeks post-transplant * 0 0 0 0 0 0 0 (tx). C-peptide levels 0 20 40 60 80 100 120 10 weeks 4 weeks 6 weeks 2 weeks PostTime (min) nephrectomy from individual mice are Time post-transplant shown on box and whisker plots. *P < 0.05, one-way Insulin Glucagon Hematoxylin Synaptophysin CK19 repeated measures ANOVA Somatostatin DAPI & eosin Trypsin DAPI (Dunnett test for multiple comparisons versus 2 weeks post-transplant). (b) Human C-peptide levels following an overnight fast and 60 min after an i.p. glucose bolus at 6 weeks. *P < 0.05, paired one-tailed t-test. (c,d) S7 cells were transplanted into SCID-beige mice with streptozotocin (STZ)induced diabetes (n = 6). (c) Fasting blood glucose levels showed reversal of diabetes by ~40 d post-transplant and rapid return to hyperglycemia following removal of the engrafted kidney by survival nephrectomy on day 78 (red arrow); data are presented as mean ± s.e.m. (d) Human C-peptide secretion after an overnight fast and 60 min following an i.p. glucose bolus at 2, 4, 6 and 10 weeks post-transplant. Human C-peptide was also measured postnephrectomy, after a 4-h fast and 60 min after an i.p. glucose bolus. C-peptide levels from individual mice are shown on box and whisker plots. *P < 0.05, one-tailed paired t-test. (e,f) An insulin tolerance test (ITT) was performed at 8 weeks posttransplant after a 4-h fast (S7 cells, n = 5; control mice, n = 3). Control mice were nondiabetic C57/Blk6 mice without cell transplants. (e) Blood glucose was measured during the ITT and data are presented as mean ± s.e.m. *P < 0.05, two-way repeated measures ANOVA with Dunnett’s test for multiple comparisons (each time point versus t = 0 within both groups). (f) Plasma was collected at 0 and 60 min post-insulin for measurement of human C-peptide. *P < 0.05, paired t-test. C-peptide levels from individual mice are shown on box and whisker plots and also in Supplementary Figure 23. (g–j) Representative histology and immunofluorescent staining of S7 cells engrafted under the kidney capsule and harvested at 10 weeks post-transplant. (g) Hematoxylin & Eosin (H&E) image of a whole S7 graft (high magnification inset provided below), illustrating minimal expansion of the hESC-derived population under the kidney capsule. See Supplementary Figure 24 for images of the entire kidney, plus two other examples of whole S7 graft histology. Immunofluorescent staining of whole grafts are also provided for synaptophysin (endocrine marker, red), CK19 (ductal marker, green) and trypsin (exocrine marker, blue) (h); insulin (red), glucagon (green) and somatostatin (blue) (i). (j) Co-immunostaining of C-peptide (red) with various transcription factors (green) and triple immunofluorescent staining of C-peptide (blue), NKX6.1 (red) and NKX2.2 (green). Nuclear DAPI staining is shown in gray. Scale bars, in low magnification images, 500 mm, and high magnification insets, 100 mm (g–i); 100 mm (j).

1130


npg


A RT I C L E S cell (Fig. 6a). However, compared with mature human beta cells, Ca2+ signals in hESC-derived cells exhibited reduced amplitude and a slower time to peak. Moreover, Ca2+ transients in many of these cells continued after high glucose stimulation was terminated. Notably, the response to direct depolarization with 30 mM KCl was rapid and transient in both S7 cells and mature human beta cells, indicating the presence of voltagegated Ca2+ channels and normal Ca2+ efflux pumps. Stimulation with the GLP-1 receptor (GLP-1R) agonist exendin-4 induced Ca2+ signals in ~30% of S7 cells, and also augmented Ca2+ signals in mature human beta cells. Both cell types responded to 10 nM of exendin-4 at basal glucose concentrations. Notably, GLP-1R gene expression increased throughout differentiation and peaked in S7 day 7 cells at levels ~30-fold higher than those in adult human islet preparations (Supplementary Fig. 21). After establishing that hESC-derived S7 cells exhibit glucose sensitivity, although in a manner that is quantitatively and qualitatively different from that of human beta cells, we assessed the dynamics of glucose-stimulated insulin release by perifusing cells in a temperatureand gas-controlled closed system with automated fraction collection. In keeping with the Ca2+ imaging data, S7 cells exhibited a very small and gradual response to the step-wise increase from 3 mM to 20 mM glucose (Fig. 6b). Moreover, insulin secretion did not return to the pre-treatment baseline, and the insulin secretory response to KCl was somewhat blunted. In contrast, adult human islets responded maximally to glucose within 5 min and returned completely to the pretreatment baseline before responding sharply to direct depolarization with KCl. Although our data showed that hESC-derived S7 cells did not rapidly secrete insulin in response to high glucose, we observed a statistically significant (P < 0.0001) accumulation of human C-peptide from S7 cells exposed to high glucose over the course of a 1-h static incubation (Fig. 6c; raw values from individual batches provided in Supplementary Fig. 22). Analysis of ultrastructure by transmission electron microscopy (TEM) revealed that S7 cells contained a high density of endocrine granules (Fig. 6d). In mature human beta cells, three types of insulin granules were generally observed by TEM: (i) light gray, diffuse core, (ii) dense, round core, and (iii) dense, rod-shaped core with a crystalline appearance (Fig. 6d; examples provided in insets). Notably, we observed examples of each type of insulin granule in hESC-derived S7 cells (Fig. 6d). Together, these data suggest that S7 cells are capable of producing appropriately packaged insulin granules and possess exocytosis machinery. Moreover, a subpopulation of S7 cells can respond to glucose, although with delayed kinetics, which indicates an improved maturation state relative to S6 cells but functional immaturity relative to adult human beta cells. S7 cells rapidly reverse diabetes in vivo We evaluated the function of S7 cells in nondiabetic and diabetic mice (Fig. 7). Following transplantation of 1.25 million S7 cells into nondiabetic mice, human C-peptide levels reached >1 ng/ml by just 2 weeks and within 4 weeks were equivalent to those produced by ~4,000 engrafted human islets (estimated to contain ~1.4–2.0 million beta cells; Fig. 7a). Moreover, at 16 d post-transplant into STZ-diabetic mice, blood glucose levels were reduced to levels that were not significantly different (P > 0.05) from pre-STZ levels, and normal fasting blood glucose levels were reached by 40 d post transplant (Fig. 7c). By 60 d post-transplant, blood glucose levels were significantly (P = 0.0235) lower than pre-STZ levels (Fig. 7c). Return to hyperglycemia was observed within 48 h of graft removal (Fig. 7c), which corresponded to loss of human C-peptide in the circulation post-nephrectomy (Fig. 7d). Notably, statistically significant glucose-stimulated human C-peptide secretion was observed beginning at 6 weeks post-transplant in both nondiabetic (Fig. 7b, P = 0.012) and diabetic mice (Fig. 7d, P = 0.032). A trend toward glucose-

responsiveness was observed as early as 2 weeks post-transplant but this did not reach statistical significance (P = 0.063). To assess the response of transplanted S7 cells to a hypoglycemia-inducing stimulus, S7 cell recipients and healthy control mice (nondiabetic, no transplants) were injected with a bolus of fast-acting insulin. Despite a lower blood glucose starting point, S7 cell recipients remained stable throughout the experiment and did not require intervention for life-threatening hypoglycemia (Fig. 7e). The degree of blood glucose lowering by exogenous insulin was similar between transplant recipients and controls (Supplementary Fig. 23). Notably, in response to the insulin bolus, the engrafted hESCderived cells significantly (P = 0.04) reduced human C-peptide secretion in five of five animals tested (Fig. 7f and Supplementary Fig. 23). At 10 weeks post-transplant, excised S7 cell grafts were highly compact and homogenous, and in contrast to the S6 grafts (Supplementary Fig. 16), did not have regions of expanded ducts (Fig. 7g,h; whole grafts from three representative mice are shown in Supplementary Fig. 24). S7 grafts were composed of mainly endocrine cells (synaptophysin+; Fig. 7h and Supplementary Fig. 24), with the vast majority being insulin+ (Fig. 7i; individual images for pancreatic hormones provided in Supplementary Fig. 25) and showing robust expression of key transcription factors, including MAFA, NKX6.1 and NKX2.2 (Fig. 7j). DISCUSSION Although the challenges of translating a stem cell–based transplantation therapy for diabetes are substantial, the clinical path for an islet cell therapy has been proven1,2. Unlike most other potential stem cell therapies, a diabetes cell therapy does not require functional integration of transplanted cells into damaged tissue as transplantation into ectopic sites can provide adequate insulin replacement to control glycemia48. Thus, it is reasonable to anticipate successful treatment of diabetes with hESC-derived cells provided that methods can be developed to produce safe and effective cells for transplantation while protecting the cells from auto-immunity and allogeneic rejection. Our seven-stage protocol builds on a substantial body of research on the complex developmental cues that guide the step-wise formation of mature islet cells in the pancreas20,49–51 and on protocols for converting hESCs to definitive endoderm, primitive gut tube, posterior foregut, pancreatic endoderm and endocrine precursors3–6,8–14,52. Here we focused on extending the differentiation of hESCs in vitro to mature beta cells. The cells we obtained at S7 were ~50% insulin+, and the vast majority of insulin+ cells expressed key beta-cell transcription factors, such as PDX1 and NKX6.1. In contrast, previous protocols yielded only a small fraction of insulin+ cells, the majority of which were polyhormonal and resembled fetal endocrine cells3,5–7,14,15,53. Notably, our protocol generated endocrine cells with an insulin content similar to that of human islet cells and capable of glucose-stimulated insulin secretion in vitro and rapid reversal of diabetes in vivo. Our protocol addresses some of the major shortcomings of previous studies. Although the PDX1+/NKX6.1+/insulin– cells generated with previous four-stage protocols developed into glucose-responsive insulin-secreting cells following transplant6,15, they did not typically acquire post-NGN3 markers such as NEUROD1 or NKX2.2 at later stages in vitro. This may have been due to the lack of NGN3 expression in hESC-derived NKX6.1+ cells, as previous studies have shown that a significant fraction of NKX6.1+ cells express NGN3 at embryonic day (e) 10.5 and NKX2.2 at e15 in mice21,22. Through numerous empirical iterations, we identified a protocol for S5 that resulted in formation of >50% NKX6.1+/NEUROD1+ cells. We believe that these cells are competent to form NKX6.1+/insulin+ cells, a process we induced in part by the addition of thyroid hormone and a gamma secretase inhibitor. However, the NKX6.1+/insulin+ cells generated at S6 were unable to


1131

npg


A RT I C L E S secrete insulin in response to glucose in vitro. We predicted that the absence of MAFA was a key deficiency in S6 cells, based on the known role of this transcription factor in regulating expression of genes involved in beta-cell maturation25–27,43. In the next step, we identified a cocktail of factors that, when applied directly to H1 hESCs, robustly induced expression of MAFA. Applying this cocktail at S7 induced MAFA and enabled development of glucose-regulated insulin release under static conditions in vitro. Insulin+/MAFA+ S7 cells were also generated with an iPSC line, but with reduced efficiency, suggesting that this protocol will require optimization when applied to other pluripotent cell lines. To our knowledge, no previous report has described the induction of MAFA, at levels similar to that of human islets, during pancreatic endocrine differentiation from human pluripotent cells in vitro. Mature adult pancreatic beta cells are functionally defined by their rapid response to elevated glucose. Brisk release of insulin minimizes glucose excursions during meals, and attenuation of insulin release as glucose levels fall is the first defense against hypoglycemia. Within a 1-h time frame in vitro, we observed consistent glucose-induced insulin release from multiple batches of S7 cells, which were similar in magnitude to that of adult human islets during static incubations. However, a more rigorous analysis revealed key differences between our hESC-derived cells and adult human islet cells. By perifusion analysis, glucose-stimulated release of insulin was delayed, gradual, and minor in magnitude with S7 cells, whereas human islets exhibited the stereotypical rapid and robust release of insulin that was highly synchronized with changes in glucose concentrations. By single-cell imaging, we found that only a small fraction of our S7 cells produced characteristic54,55 changes in intracellular Ca2+ oscillations when exposed to elevated glucose concentrations, although the majority had a rapid and transient Ca2+ response to depolarization with KCl. As much as half of postprandial insulin release is due to meal-induced release of gut hormones, or incretins, that stimulate insulin secretion by activation of specific G protein–coupled receptors56. The expression of GLP-1 receptors was strongly upregulated during hESC differentiation, and we found that a GLP-1 analog, exendin-4, induced Ca2+ responses in ~1/3 of the S7 cells. We conclude that our S7 cells possess intact incretin signaling pathways, functional voltage-gated Ca2+ channels and generally normal Ca2+ handling systems. The slower Ca2+ response to glucose, but not direct depolarization, implies a deficiency in glucose metabolism and/ or the ATP-sensitive potassium channel triggering of electrical activity. The blunted insulin response to glucose and KCl is also consistent with a reduction in the size of the readily releasable pool of insulin and/or impairments of vesicle traffic to the membrane and exocytosis. Together, these data suggest that a subpopulation of S7 cells can respond to glucose, although with delayed kinetics, and that this minor subpopulation is overshadowed in the perifusion assay by less mature cells. Overall, our data imply that S7 cells are functionally immature relative to human beta cells and highlight the need for careful assessment of rapid insulin secretory responses in putative beta cells, beyond static release assays. Despite these differences from adult human beta cells, S7 cells rapidly reversed diabetes following transplantation in immunodeficient mice using one-quarter the cell dose compared with S4 progenitor cells4–6. Moreover, maturation of pancreatic progenitor cells and resolution of diabetes required ~23 weeks with S4 cells, whereas normal fasting glucose levels were achieved within 6 weeks with S7 cells. Reversal of diabetes was accelerated with both S6 and S7 cells compared with S4 cells, but human C-peptide levels achieved by 1.25 million S7 cells at 4 weeks were two to four times higher than those achieved by 2.5 million S6 cells and similar to those of ~4,000 engrafted human islets (estimated to contain ~1.4–2.0 million beta cells, the approximate number needed to consistently reverse diabetes in mice). Notably, following treatment 1132

with hESC-derived S6 and S7 cells, fasting glucose levels were lower than in nondiabetic rodents, but similar to healthy humans, which is consistent with results following transplantation of human islets57 and human fetal pancreas tissues58,59. Transplant recipients were stable and healthy throughout the study, including during an insulin tolerance test in which the engrafted S7 cells repressed human C-peptide production in response to an exogenous insulin bolus. Thus, like adult human islets, hESC-derived cells are capable of secreting insulin in response to glucose and of reducing insulin release to prevent dangerous hypoglycemia. If cells similar to S7 cells could be manufactured in a reliable and scalable manner, they may provide a more consistent cell product and more predictable outcome following transplantation compared with cadaveric human islets or S4 pancreatic progenitor cells. METHODS Methods and any associated references are available in the online version of the paper. Note: Any Supplementary Information and Source Data files are available in the online version of the paper. ACKNOWLEDGMENTS This work was supported in part by funding from the JDRF, the Canadian Institutes of Health Research (CIHR) Regenerative Medicine and Nanomedicine Initiative, and the Stem Cell Network (SCN). J.E.B. was funded by a JDRF postdoctoral fellowship and L’Oreal Canada for Women in Science Research Excellence Fellowship. Y.H.C.Y. was the recipient of an NSERC Post Graduate Scholarship and M.M. was the recipient of a Mitacs Elevate Postdoctoral Fellowship. The CIHR Transplantation Training Program provided funding for J.E.B., N.Q. and M.M. We also thank Stem Cell Technologies for their financial support, T. Webber, S. Karuna and B. Hu for their technical assistance, and G. Warnock and Z. Ao from the Irving K. Barber Human Islet Isolation Laboratory (Vancouver, BC) for providing human islets. AUTHOR CONTRIBUTIONS A. Rezania, J.E.B., J.D.J. and T.J.K. designed experiments, analyzed and interpreted results. J.E.B., P.A., A. Rubin, I.B., A.A., S.O., N.Q., M.M., T.A. and Y.H.C.Y. performed experiments. A. Rezania and T.J.K. conceived the project and supervised its participants. A. Rezania, J.E.B., J.D.J. and T.J.K. wrote the manuscript. A. Rezania, J.E.B., A.A., J.D.J., T.J.K., N.Q., M.M., T.A. and Y.H.C.Y. contributed to manuscript revisions. COMPETING FINANCIAL INTERESTS The authors declare competing financial interests: details are available in the online version of the paper. Reprints and permissions information is available online at http://www.nature.com/ reprints/index.html. 1. Barton, F.B. et al. Improvement in outcomes of clinical islet transplantation: 1999– 2010. Diabetes Care 35, 1436–1445 (2012). 2. Warnock, G.L. et al. A multi-year analysis of islet transplantation compared with intensive medical therapy on progression of complications in type 1 diabetes. Transplantation 86, 1762–1766 (2008). 3. Bruin, J.E. et al. Characterization of polyhormonal insulin-producing cells derived in vitro from human embryonic stem cells. Stem Cell Res. 12, 194–208 (2014). 4. Bruin, J.E. et al. Maturation and function of human embryonic stem cell-derived pancreatic progenitors in macroencapsulation devices following transplant into mice. Diabetologia 56, 1987–1998 (2013). 5. Rezania, A. et al. Maturation of human embryonic stem cell-derived pancreatic progenitors into functional islets capable of treating pre-existing diabetes in mice. Diabetes 61, 2016–2029 (2012). 6. Rezania, A. et al. Enrichment of human embryonic stem cell-derived NKX6.1expressing pancreatic progenitor cells accelerates the maturation of insulin-secreting cells in vivo. Stem Cells 31, 2432–2442 (2013). 7. Rezania, A. et al. Production of functional glucagon-secreting alpha-cells from human embryonic stem cells. Diabetes 60, 239–247 (2011). 8. Chen, S. et al. A small molecule that directs differentiation of human ESCs into the pancreatic lineage. Nat. Chem. Biol. 5, 258–265 (2009). 9. Xu, X., Browning, V.L. & Odorico, J.S. Activin, BMP and FGF pathways cooperate to promote endoderm and pancreatic lineage cell differentiation from human embryonic stem cells. Mech. Dev. 128, 412–427 (2011). 10. Shim, J.H. et al. Directed differentiation of human embryonic stem cells towards a pancreatic cell fate. Diabetologia 50, 1228–1238 (2007).


npg


A RT I C L E S 11. Cho, C.H. et al. Inhibition of activin/nodal signalling is necessary for pancreatic differentiation of human pluripotent stem cells. Diabetologia 55, 3284–3295 (2012). 12. D’Amour, K.A. et al. Production of pancreatic hormone-expressing endocrine cells from human embryonic stem cells. Nat. Biotechnol. 24, 1392–1401 (2006). 13. Kroon, E. et al. Pancreatic endoderm derived from human embryonic stem cells generates glucose-responsive insulin-secreting cells in vivo. Nat. Biotechnol. 26, 443–452 (2008). 14. Nostro, M.C. et al. Stage-specific signaling through TGFbeta family members and WNT regulates patterning and pancreatic specification of human pluripotent stem cells. Development 138, 861–871 (2011). 15. Kelly, O.G. et al. Cell-surface markers for the isolation of pancreatic cell types derived from human embryonic stem cells. Nat. Biotechnol. 29, 750–756 (2011). 16. Schiesser, J.V. & Wells, J.M. Generation of beta cells from human pluripotent stem cells: are we there yet? Ann. NY Acad. Sci. 1311, 124–137 (2014). 17. Nostro, M.C. & Keller, G. Generation of beta cells from human pluripotent stem cells: Potential for regenerative medicine. Semin. Cell Dev. Biol. 23, 701–710 (2012). 18. Halban, P.A., Kahn, S.E., Lernmark, A. & Rhodes, C.J. Gene and cell-replacement therapy in the treatment of type 1 diabetes: how high must the standards be set? Diabetes 50, 2181–2191 (2001). 19. Aguayo-Mazzucato, C. & Bonner-Weir, S. Stem cell therapy for type 1 diabetes mellitus. Nat. Rev. Endocrinol. 6, 139–148 (2010). 20. Wells, J.M. & Melton, D.A. Vertebrate endoderm development. Annu. Rev. Cell Dev. Biol. 15, 393–410 (1999). 21. Nelson, S.B., Schaffer, A.E. & Sander, M. The transcription factors Nkx6.1 and Nkx6.2 possess equivalent activities in promoting beta-cell fate specification in Pdx1+ pancreatic progenitor cells. Development 134, 2491–2500 (2007). 22. Sander, M. et al. Homeobox gene Nkx6.1 lies downstream of Nkx2.2 in the major pathway of beta-cell formation in the pancreas. Development 127, 5533–5540 (2000). 23. Rukstalis, J.M. & Habener, J.F. Neurogenin3: a master regulator of pancreatic islet differentiation and regeneration. Islets 1, 177–184 (2009). 24. Matsuoka, T.A. et al. The MafA transcription factor appears to be responsible for tissuespecific expression of insulin. Proc. Natl. Acad. Sci. USA 101, 2930–2933 (2004). 25. Wang, H., Brun, T., Kataoka, K., Sharma, A.J. & Wollheim, C.B. MAFA controls genes implicated in insulin biosynthesis and secretion. Diabetologia 50, 348–358 (2007). 26. Zhang, C. et al. MafA is a key regulator of glucose-stimulated insulin secretion. Mol. Cell. Biol. 25, 4969–4976 (2005). 27. Hang, Y. et al. The MafA transcription factor becomes essential to islet beta-cells soon after birth. Diabetes 63, 1994–2005 (2014). 28. Choi, K.M. et al. Effect of ascorbic acid on bone marrow-derived mesenchymal stem cell proliferation and differentiation. J. Biosci. Bioeng. 105, 586–594 (2008). 29. Johansson, K.A. et al. Temporal control of neurogenin3 activity in pancreas progenitors reveals competence windows for the generation of different endocrine cell types. Dev. Cell 12, 457–465 (2007). 30. Schulz, T.C. et al. A scalable system for production of functional pancreatic progenitors from human embryonic stem cells. PLoS ONE 7, e37004 (2012). 31. Pruniéras, M., Regnier, M. & Woodley, D. Methods for cultivation of keratinocytes with an air-liquid interface. J. Invest. Dermatol. 81, 28s–33s (1983). 32. Pezzulo, A.A. et al. The air-liquid interface and use of primary cell cultures are important to recapitulate the transcriptional profile of in vivo airway epithelia. Am. J. Physiol. Lung Cell. Mol. Physiol. 300, L25–L31 (2011). 33. Heinis, M. et al. Oxygen tension regulates pancreatic beta-cell differentiation through hypoxia-inducible factor 1alpha. Diabetes 59, 662–669 (2010). 34. Cortijo, C., Gouzi, M., Tissir, F. & Grapin-Botton, A. Planar cell polarity controls pancreatic beta cell differentiation and glucose homeostasis. Cell Reports 2, 1593–1606 (2012). 35. Shi, Y. & Massagué, J. Mechanisms of TGF-beta signaling from cell membrane to the nucleus. Cell 113, 685–700 (2003).

36. Aguayo-Mazzucato, C. et al. Thyroid hormone promotes postnatal rat pancreatic betacell development and glucose-responsive insulin secretion through MAFA. Diabetes 62, 1569–1580 (2013). 37. Seymour, P.A. et al. SOX9 is required for maintenance of the pancreatic progenitor cell pool. Proc. Natl. Acad. Sci. USA 104, 1865–1870 (2007). 38. Murtaugh, L.C., Stanger, B.Z., Kwan, K.M. & Melton, D.A. Notch signaling controls multiple steps of pancreatic differentiation. Proc. Natl. Acad. Sci. USA 100, 14920–14925 (2003). 39. Apelqvist, A. et al. Notch signalling controls pancreatic cell differentiation. Nature 400, 877–881 (1999). 40. Pisania, A. et al. Quantitative analysis of cell composition and purity of human pancreatic islet preparations. Lab. Invest. 90, 1661–1675 (2010). 41. Street, C.N. et al. Islet graft assessment in the Edmonton Protocol: implications for predicting long-term clinical outcome. Diabetes 53, 3107–3114 (2004). 42. Piper, K. et al. Beta cell differentiation during early human pancreas development. J. Endocrinol. 181, 11–23 (2004). 43. Aguayo-Mazzucato, C. et al. Mafa expression enhances glucose-responsive insulin secretion in neonatal rat beta cells. Diabetologia 54, 583–593 (2011). 44. Haase, T.N. et al. Growth arrest specific protein (GAS) 6: a role in the regulation of proliferation and functional capacity of the perinatal rat beta cell. Diabetologia 56, 763–773 (2013). 45. Harmon, J.S. et al. beta-Cell-specific overexpression of glutathione peroxidase preserves intranuclear MafA and reverses diabetes in db/db mice. Endocrinology 150, 4855–4862 (2009). 46. Harmon, J.S., Stein, R. & Robertson, R.P. Oxidative stress-mediated, post-translational loss of MafA protein as a contributing mechanism to loss of insulin gene expression in glucotoxic beta cells. J. Biol. Chem. 280, 11107–11113 (2005). 47. Mahadevan, J. et al. Ebselen treatment prevents islet apoptosis, maintains intranuclear Pdx-1 and MafA levels, and preserves beta-cell mass and function in ZDF rats. Diabetes 62, 3582–3588 (2013). 48. Merani, S., Toso, C., Emamaullee, J. & Shapiro, A.M. Optimal implantation site for pancreatic islet transplantation. Br. J. Surg. 95, 1449–1461 (2008). 49. Kumar, M. & Melton, D. Pancreas specification: a budding question. Curr. Opin. Genet. Dev. 13, 401–407 (2003). 50. Jørgensen, M.C. et al. An illustrated review of early pancreas development in the mouse. Endocr. Rev. 28, 685–705 (2007). 51. Kinkel, M.D. & Prince, V.E. On the diabetic menu: zebrafish as a model for pancreas development and function. BioEssays 31, 139–152 (2009). 52. D’Amour, K.A. et al. Efficient differentiation of human embryonic stem cells to definitive endoderm. Nat. Biotechnol. 23, 1534–1541 (2005). 53. Basford, C.L. et al. The functional and molecular characterisation of human embryonic stem cell-derived insulin-positive cells compared with adult pancreatic beta cells. Diabetologia 55, 358–371 (2012). 54. Luciani, D.S. & Johnson, J.D. Acute effects of insulin on beta-cells from transplantable human islets. Mol. Cell. Endocrinol. 241, 88–98 (2005). 55. Misler, S., Barnett, D.W., Gillis, K.D. & Pressel, D.M. Electrophysiology of stimulussecretion coupling in human beta-cells. Diabetes 41, 1221–1228 (1992). 56. Nauck, M.A. et al. Incretin effects of increasing glucose loads in man calculated from venous insulin and C-peptide responses. J. Clin. Endocrinol. Metab. 63, 492–498 (1986). 57. Davalli, A.M. et al. A selective decrease in the beta cell mass of human islets transplanted into diabetic nude mice. Transplantation 59, 817–820 (1995). 58. Tuch, B.E. Reversal of diabetes by human fetal pancreas. Optimization of requirements in the hyperglycemic nude mouse. Transplantation 51, 557–562 (1991). 59. Tuch, B.E. & Monk, R.S. Regulation of blood glucose to human levels by human fetal pancreatic xenografts. Transplantation 51, 1156–1160 (1991).


1133

ONLINE METHODS Cell sources. Human islets were obtained from the Irving K. Barber Human Islet Isolation Laboratory (Vancouver, BC) and Prodo Laboratories, Inc. (Irvine, CA) with informed consent. The H1 hESC line was obtained from WiCell Research Institute, Inc. (Madison, WI). An episomal reprogrammed iPSC line was purchased from Life (Gibco Human Episomal iPSC, Cat # A18945, Life Technologies, CA). Both cell lines have been authenticated by Cell Line Genetics (Madison, WI) and confirmed to be mycoplasma-free by using the MycoSEQ Mycoplasma Detection Kit (Life, Cat#4399363).

npg


In vitro differentiation of human pluripotent stem cells. H1 and iPSCs were cultured on 1:30 diluted Matrigel (BD BioSciences, CA, Cat#356231) in Essential 8 (E8) medium (Life Technologies, Cat# A1517001). At ~70–80% confluency, cultures were rinsed with 1× DPBS without Mg2+ and Ca2+ (Invitrogen, Cat#14190) followed by incubation with TrypLE Express Enzyme (1×) (Life, Cat# 12604021) for 3–5 min at 37 °C. Released single cells were rinsed with E8, and spun at 1,000 r.p.m. for 5 min. The resulting cell pellet was resuspended in E8 medium supplemented with Y-27632 (10 mM; Sigma-Aldrich; MO, Cat#Y0503) and the single cell suspension was seeded at ~1.3–1.5 × 105 cells/cm2 on Matrigel-coated surfaces. Cultures were fed every day with E8 medium and differentiation was initiated 48 h following seeding, resulting in ~90% starting confluency. S1: definitive endoderm (3 d).Undifferentiated pluripotent stem cells plated on 1:30 Matrigel-coated surfaces were first rinsed with 1× DPBS without Mg2+ and Ca2+ and then exposed to MCDB 131 medium (Life, Cat# 10372-019) further supplemented with 1.5 g/l sodium bicarbonate (Sigma, MO, Cat# S6297), 1× Glutamax (Life, Cat#35050-079), 10 mM final glucose (Sigma, Cat# G8769) concentration, 0.5% BSA (fatty acid free BSA, Proliant, IA, Cat#68700), 100 ng/ ml GDF8 (Pepro-Tech; Rocky Hill, NJ, Cat#120-00), and 1 mM of MCX-928 (GSK3b inhibitor3, internal compound to Janssen) for day 1 only. For day 2, cells were cultured in MCDB with 0.5% BSA, 1.5 g/l sodium bicarbonate, 1× Glutamax, 10 mM glucose, 100 ng/ml GDF8 and 0.1 mM of MCX-928. On day three, cells were cultured in MCDB with 0.5% BSA, 1.5 g/l sodium bicarbonate, 1× Glutamax, 10 mM glucose and 100 ng/ml GDF8. CHIR-99021 (GSK3b inhibitor, SelleckChem, Cat#S2924) also proved effective in efficient induction of S1 cells (data not shown). For the iPSC line, 1.5 mM of MCX-928 was used on day 1 of S1. This protocol routinely resulted in 98–99% FOXA2/CXCR4 expression (data not shown). S2: primitive gut tube (2 d). S1 cells were rinsed with 1X DPBS (without Mg2+ and Ca2+) and then exposed to MCDB 131 medium further supplemented with 1.5 g/l sodium bicarbonate, 1× Glutamax, 10 mM final glucose concentration, 0.5% BSA, 0.25 mM ascorbic acid (Sigma, Cat# A4544) and 50 ng/ml of FGF7 (R & D Systems, Cat#251-KG) for 2 d. S3: posterior foregut (2 d). Cultures were continued for 2 d in BLAR medium (Custom media formulation; see Supplementary Table 3 for detailed formulation) further supplemented with 2.5 g/l sodium bicarbonate, 1× Glutamax, 10 mM final glucose concentration, 2% BSA, 0.25 mM ascorbic acid, 50 ng/ml of FGF7, 0.25 mM SANT-1 (Sigma, Cat# S4572), 1 mM retinoic acid (RA; Sigma, Cat#R2625), 100 nM LDN193189 (LDN; BMP receptor inhibitor, Stemgent, CA, Cat#04-0019), 1:200 ITS-X (Life, Cat#51500056), and 200 nM TPB (PKC activator, custom synthesis, ChemPartner, China). MCDB 131 could be substituted for BLAR media. S4: pancreatic endoderm, PDX1+/NKX6.1+ cells (3 d). S3 cells were exposed to BLAR medium further supplemented with 2.5 g/l sodium bicarbonate, 1× Glutamax, 10 mM final glucose concentration, 2% BSA, 0.25 mM ascorbic acid, 2 ng/ml of FGF7, 0.25 mM SANT-1, 0.1 mM retinoic acid, 200 nM LDN193189, 1:200 ITS-X, and 100 nM TPB for 3 d. After 3 d of culture, the S4 cells were treated for 4 h with 10 mM Y-27632. Cells were then rinsed with 1× DPBS without Mg2+ and Ca2+ and then exposed to TrypLE (1×) for 3–5 min at room temperature. The released cells were washed with basal BLAR medium and spun at 1,000 r.p.m. for 3 min. The resulting cell pellet was resuspended as single cells at a density of ~0.5 × 105 cells/ml on filter inserts (BD, Cat#35-3493 or Corning Cat#3420); 5–10 ml per spot for a total of 0.25–0.5 × 106 cells/spot) at an air-liquid interface. Each spotted area measured ~1–2 mm in diameter depending on the volume of cells added. For 6-well filter inserts (BD),

doi:10.1038/nbt.3033

1.5 ml/well was added to the bottom of each insert whereas 8 ml was added for 10-cm filter inserts (Corning). Typically 10–15 spots were used per well of a 6-well insert and 80–90 spots were used for 10-cm inserts (Supplementary Fig. 2a). S5: pancreatic endocrine precursors, PDX1+/NKX6.1+/NEUROD1+ (3 d). S4 cells prepared as described above were exposed to BLAR medium further supplemented with 1.5 g/l sodium bicarbonate, 1× Glutamax, 20 mM final glucose concentration, 2% BSA, 0.25 mM SANT-1, 0.05 mM retinoic acid, 100 nM LDN193189, 1:200 ITS-X, 1 mM T3 (3,3′,5-Triiodo-l-thyronine sodium salt, Sigma, T6397), 10 mM ALK5 inhibitor II (Enzo Life Sciences, NY, Cat# ALX-270-445), 10 mM zinc sulfate (Sigma, Z0251) and 10 mg/ml of heparin (Sigma, H3149) for 3 d. Addition of heparin resulted in improved viability of differentiated cells clusters. S6: NKX6.1+/insulin+ cells (7–15 d). S5 cells were exposed to BLAR medium further supplemented with 1.5 g/l sodium bicarbonate, 1× Glutamax, 20 mM final glucose concentration, 2% BSA, 100 nM LDN193189, 1:200 ITS-X, 1 mM T3, 10 mM ALK5 inhibitor II, 10 mM zinc sulfate, 100 nM gamma secretase inhibitor XX for the first 7 d only (EMD MilliPore, MA, Cat# 565789) and 10 mg/ml of heparin for 7–15 d. S7: NKX6.1+/insulin+/MAFA+cells (7-15 d). S6 cells were exposed to BLAR medium further supplemented with 1.5 g/l sodium bicarbonate, 1× Glutamax, 20 mM final glucose concentration, 2% BSA, 1:200 ITS-X, 1 mM T3, 10 mM ALK5 inhibitor II, 10 mM zinc sulfate, 1 mM N-acetyl cysteine (N-Cys, Sigma, Cat# A9165), 10 mM Trolox (Vitamin E analogue, EMD, Cat#648471), 2 mM R428 (AXL inhibitor, SelleckChem, Cat# S2841) and 10 mg/ml of heparin for 7–15 d. Unless otherwise specified, for all stages, the cultures were fed every day. At the end of S6 and S7, each spotted area contained 0.12–0.25 × 106 cells depending on the initial volume of cells added per spot. Overall, when 22 × 106 hESCs/T-150 flask were seeded, ~46 × 106 cells/T-150 flasks were retrieved at S4. From each S4 flask, we generated ~180 clusters (~0.25 × 106 cells/cluster) at air-liquid interface and by S6 each cluster contained ~0.125 × 106 cells. Therefore, we estimated that for every ES cell seeded, we generate ~1.2 S6/S7 cells. Assuming ~50% NKX6.1+/insulin+ cells generated at S6/S7, two hESCs result in one NKX6.1+/insulin+ cell. Differentiation of S4 cells in suspension aggregate cultures. For this study, H1 cells were differentiated according to our previously published 14-day, four-stage protocol4. On the last day of culture, S4 cells were treated with 5 mg/ml dispase (Corning, Cat# 354235) for 5 min at 37 °C, followed by gentle pipetting to break into cell clumps (100–200 mm). Next, clusters were transferred into ultra-low attachment 6-well plates (Corning, Cat# 3471) in suspension with DMEM-HG supplemented with 10 µM ALK5iII, 100 nM LDN193189, 1 mM T3, 100 nM gamma secretase inhibitor XX, 10 μg/ml of heparin, and 1% B27 for 7 d. Cell clusters were harvested for FACS analysis (Supplementary Fig. 8). In vitro screening studies. Eight different ALK5 inhibitors were added to differentiation cultures at a concentration of 1 mM during S5 and 6 to determine their effects on expression of islet genes. The following inhibitors were tested (all purchased from EMD Millipore, Billerica, MA, USA). Transforming Growth Factor-b (TGF-b) receptor inhibitor V (#616456), TGF-b receptor inhibitor I (#616451), TGF-b receptor inhibitor IV (#616454), TGF-b receptor inhibitor VII (#616458), TGF-b receptor inhibitor VIII (#616459), TGF-b receptor inhibitor II (#616452; ALK5 inhibitor II), TGF-b receptor inhibitor VI (#616464), TGF-b receptor inhibitor III (#616453). The efficiency of TGF-b receptor inhibitors for targeting different kinases was assessed by SelectScreen Kinase Profiling Services using 100 mM ATP concentration (Life Technologies, Carlsbad, CA, USA). Libraries of small molecules and growth factors were tested to identify an inducer of MAFA gene expression during S7. Small molecules/growth factors were added at various concentrations (indicated in Fig. 4a) for 7 d during S7, and MAFA gene expression was assessed on day 7 of S7 (details of all compounds are provided in Supplementary Table 4). Unless otherwise stated, MCDB media was used for S1 and S2 and BLAR for all other stages. Quantitative RT-PCR. Gene expression was assessed in differentiated cells using custom Taqman Arrays (Applied Biosystems; Foster City CA), as previously described5. Data were analyzed using Sequence Detection Software (Applied

NATURE BIOTECHNOLOGY

Biosystems) and normalized to undifferentiated H1 cells using the DDCt method. Refer to Supplementary Table 5 for primer details.

npg


Flow cytometry. Differentiated cells were released into a single-cell suspension, fixed, permeabilized, and stained for various intracellular markers, as described previously5. Dead cells were excluded during FACS analysis and gating was determined using isotype antibodies. Refer to Supplementary Table 6 for antibody details. Immunohistochemistry. For paraffin sections, hESC-derived cells were fixed overnight in 4% paraformaldehyde (PFA) at 4 °C and then embedded in 1% agarose before paraffin-embedding and sectioning (5 μm thickness; Wax-it Histology Services; Vancouver, Canada). Immunofluorescent staining was performed as previously described7; primary antibodies are detailed in Supplementary Table 7. Images were captured using the ImageXpressMICRO Imaging System and analyzed using MetaXpress Software (Molecular Devices Corporation, Sunnyvale, CA, USA). For cryosectioning, S5–S7 clusters were rinsed with PBS followed by overnight fixation in 4% PFA at 4 °C. Following fixation, PFA was removed and cells were rinsed three times with PBS and incubated overnight at 4 °C in 30% sucrose solution. The samples were next overlaid with OCT solution and flash frozen using liquid Nitrogen and stored at –80 °C. A microtome was used to cut 5 μm sections and placed on Superfrost plus slides. The sections were next rinsed with PBS and permeabilized with 0.5% Tween for 20 min, rinsed again with PBS and then blocked with appropriate serum for 30 min at room temperature. Primary antibodies were added at appropriate dilutions overnight at 4 °C. Secondary antibodies were added for 30 min at room temperature followed by rinsing with PBS and adding Vectastain mounting reagent with DAPI. The sections were visualized using a Nikon Ti-S fluorescence microscope. Transmission electron microscopy. S7 hESC-derived cells and human islets were stored in 2% glutaraldehyde and processed as previously described5 by the Electron Microscopy Facility at McMaster University, Faculty of Health Sciences (Hamilton, Ontario, Canada). Grids were examined with a Tecnai G2 Spirit electron microscope (FEI Co., Eindhoven, The Netherlands) and representative images were captured with a 4Kx4K FEI Eagle HS CCD camera. Hormone content. S6,or S7 cells and human islets were suspended in Tris-EDTA (pH 7.4) on ice followed by a brief sonication until cell membranes were dispersed. Cell debris were cleared by brief centrifugation and hormone levels were measured in an aliquot of the lysed cell suspension using the following ELISA kits: insulin (#10-1113-01; Mercodia), proinsulin (#10-1118-01; Mercodia), glucagon (#48-GLUHU-E01; Alpco Diagnostics, Salem, NH) and human C-peptide (#101141-01; Mercodia). For normalization, DNA content was determined using the Quant-iT PicoGreen dsDNA kit (#P7589; Invitrogen). In vitro analysis of cell function. Perifusion studies were conducted according to our previously established protocol54,60, and our n = 15 historical control data60 can be compared to the perifusion data shown in this manuscript. Groups of 150 hand-picked, size-matched human islets, five clusters of S6 hESC-derived cells (~2.50 × 105 cells/cluster), or five clusters of S7 hESC-derived cells (~1.25 × 105 cells/cluster), were suspended with Cytodex microcarrier beads (Sigma-Aldrich, St Louis, MO) in the 300 µl plastic chambers of an Acusyst-S perifusion apparatus (Endotronics, Minneapolis, MN, USA). Under temperature- and CO2-controlled conditions, the cells were perifused at 0.5 ml min-1 with a Krebs-Ringer buffer. This standard buffer contained (in mM): 129 NaCl, 5 NaHCO3, 4.8 KCl, 2.5 CaCl2, 1.2 MgSO4, 1.2 KH2PO4, 10 HEPES, 3 glucose and 0.5% radioimmunoassay-grade BSA (Sigma). Prior to sample collection, cells were equilibrated under basal (3 mM glucose) conditions for 1 h. During perifusion cells were exposed to 10 nM Exendin-4 (Sigma) and 30 mM KCl. Insulin secretion was measured by radioimmunoassay (Human Insulin RIA, HI-14K; Millipore, Billerica, MA). For static incubation glucose-stimulated insulin secretion assays, S6/7 cells (four to five clusters, equivalent to ~0.5–1.0 x 106 cells) and human islets (~20–50 islets) were rinsed twice with Krebs buffer (129 mM NaCl, 4.8 mM KCl, 2.5 mM CaCl2, 1.2 mM MgSO2, 1 mM Na2HPO4, 1.2 mM KH2PO4, 5 mM NaHCO3, 10 mM HEPES, 0.1% BSA, in deionized water and then sterile filtered) and then pre-incubated in Krebs buffer for 40 min. Cells were then incubated in Krebs buffer spiked with 3.3 mM glucose for 60 min. The cells were then transferred to


another plate containing Krebs buffer spiked with 16.7 mM glucose and incubated for additional 60 min. Supernatant samples were collected after each incubation period and frozen at –70 °C for human C-peptide ELISA (#10-1141-01; Mercodia) measurement. To validate our protocol and confirm the authenticity of our static glucosestimulated insulin secretion results in S7 cells, we performed several control experiments with human islets (Supplementary Fig. 26). Transferring islets from 3.3 mM into 16.7 mM glucose caused a significant accumulation of insulin, whereas transferring from 3.3 mM into 3.3 mM did not induce insulin secretion, confirming that the act of simply transferring cells did not induce insulin release. Second, we confirmed that preincubation in 0 mM glucose was not required to induce glucose-stimulated insulin secretion and finally, that parallel incubations (as opposed to sequential) in low and high glucose produced a similar foldresponse by human islets. Ratiometric Ca2+ imaging was performed on human islet cells and S7 hESCs using standard Fura-2 acetoxymethyl ester (Fura-2 AM) imaging cells as detailed elsewhere54. The S7 hESCs, compared with dispersed human islet cells, were smaller and did not take up and/or retain the Fura dye as efficiently. Therefore, 340/380 ratios should not be directly compared between the cell types. After loading and before recordings, cells were allowed to equilibrate for another 30 min in the baseline experimental solution, which was Ringer’s buffer containing (in mM): NaCl 144, KCl 5.5, MgCl2 1, CaCl2 2, HEPES 20 (adjusted to pH 7.35 by NaOH). Where indicated, KCl and glucose concentrations were increased by iso-osmotic substitution for NaCl. All imaging solutions contained 3 mM glucose unless otherwise is stated. Exendin-4 (Sigma) was provided at a 10 nM concentration. Imaging was conducted on continuously-perifused coverslips held at 37 °C on a Zeiss Axiovert 200M inverted microscope equipped with a FLUAR 20× objective and a Lambda DG-4 lightsource switcher (system custom built by Intelligent Imaging Innovation, Denver, CO). Fura-2 was excited at 340 nm and 380 nm and the emitted fluorescence was monitored through a D510/80m filter. Changes in [Ca2+] were expressed as the ratio of the fluorescence emission intensities (F340/F380). Animal studies. Animals (Diabetic and Non-Diabetic Cohorts). All experiments were approved by the UBC Animal Care Committee and Janssen IACUC committee. All mice were males and were maintained on a 12h light/dark cycle with ad libitum access to a standard irradiated diet (Harlan Laboratories, Teklad Diet #2918; Madison, WI). Mice were randomly assigned to treatment groups, such that body weight and blood glucose levels were matched before treatment. Each animal cohort is described in detail below and summarized in Supplementary Table 8. For nondiabetic cohorts in Figures 3a–d and 7a,b, 8- to 10-week-old NSG mice (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJStrain Code: 5557) were obtained from Jackson Laboratories, Bar Harbor, ME). Mice were anaesthetized with inhalable isoflurane and received one of the following cell transplants under the kidney capsule: S4 cells (generated as previously described5; ~5 × 106/animal), S6 cells (~2.5 × 106 cells/animal), S7 cells (~1.25 × 106 cells/animal) or human islets (2,000 or 4,000 islet equivalents/animal, as indicated). Since S7 cells were expected to produce more human C-peptide than S6 cells, we compared S7 cohorts to 4,000 engrafted human islet equivalents and S6 cohorts to 2,000 engrafted human islet equivalents as a positive control. Based upon the estimate of ~1,500 beta cells per human islet40 and reported average of between 23%41 to 32%40 beta cells in a typical preparation of isolated cadaveric human islets, we calculate that the recipients of 4,000 human islet equivalents were transplanted with roughly 1.4–2.0 × 106 human beta cells. For S6 Diabetic Cohort 1 (Fig. 3e,f), NSG mice (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJStrain Code: 5557) were obtained from Jackson Laboratories. Mice were rendered diabetic with multiple low-dose STZ (45 mg/kg/d for 5 d) and then transplanted with S6 cells (~2.5 × 106 cells/animal) under the kidney capsule. For S4/S6 Diabetic Cohort 2 (Fig. 3g-i and Supplementary Fig. 15), SCIDbeige mice (C.B-Igh-1b/GbmsTac-Prkdcscid-LystbgN7) were obtained from Taconic (Hudson, NY). Mice were rendered diabetic by a multiple low-dose streptozotocin (STZ) regimen of daily i.p. STZ injections for 5 consecutive days (57 mg/kg/d). Diabetic mice then received either S4 cells (differentiated as previously described5; ~5 × 106 cells/animal) or S6 cells (~2 × 106 cells/animal) under the kidney capsule at 9 or 10 weeks of age. For S6 Diabetic Cohort 3 (Supplementary Fig. 14), NOD SCID (NOD.CB17Prkdcscid/J) were originally obtained from Jackson Laboratory and subsequently bred at UBC. Mice were rendered diabetic with multiple low-dose STZ (35 mg/

doi:10.1038/nbt.3033

Metabolic analysis. All metabolic analyses were performed in conscious, restrained mice on the indicated days. 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. Blood glucose and body weight were monitored 1 or 2 times weekly following a 4 h morning fast. Glucose-stimulated human C-peptide secretion was assessed by collecting blood samples after an overnight fast (16 h) and 60 min following an i.p. glucose bolus (2 g/kg; 30% solution; Vétoquinol, Lavaltrie, QC). Intraperitoneal glucose tolerance tests (i.p.GTTs) were performed after an overnight fast and injection of glucose (2 g/kg); blood glucose and blood samples were collected at the indicated time points. For oral meal challenges, blood glucose was measured and blood collected after an overnight fast (16 h) and then after a 45 min feeding period with normal chow. The insulin tolerance test (ITT) was performed following a 4-h morning fast and administration of human synthetic insulin (0.7 IU/kg body weight; Novolin ge Toronto, Novo Nordisk, Mississauga, Canada). Plasma was collected at 0 and 60 min following the insulin injection. Plasma was stored at -30 °C and later assayed using a human C-Peptide ELISA (Alpco Diagnostics; Salem, NH).

sion throughout differentiation (S4–6) was performed using a one-way ANOVA with a Tukey test for multiple comparisons between different stages of differentiation and a Dunnett test for comparing each stage of differentiation with human islet preparations (Fig. 1b). Hormone content was compared between S6 cells and human islet preparations using an unpaired, two-tailed t-test (Fig. 2b). qPCR data for S7 cells and human islets were compared by an unpaired two-tailed t-test (Fig. 5a). For glucose, meal, or insulin-stimulated hormone secretion experiments, paired one-tailed t-tests were used to compare either low glucose versus high glucose (in vitro assays; Figs. 2c and 6c) or fasted versus postglucose/meal/insulin (in vivo assays; Figs. 3d, 7d and 7f and Supplementary Fig. 15b). To assess differences over the time course of a glucose, meal, or insulin tolerance test in vivo, a two-way repeated-measures ANOVA was used (Figs. 3h, 3i, 7e and Supplementary Fig. 15b,c). Multiple comparisons were performed between groups at each time point using a Bonferroni test at each time point (Fig. 3h,i) and a Dunnett test was used to compare each time point with t = 0 within a group (Figs. 3h,i and 7e). To track the development of hESC-derived cells or human islets over time following transplantation, random fed human C-peptide levels were compared by one-way repeated measures ANOVA with a Dunnett test for comparing each time point with the first measurement (1 or 2 weeks; Figs. 3a,b,c,f and 7a). For blood glucose tracking in these studies, values at selected days (indicated in results section; Figs. 3e,g and 7c and Supplementary Fig. 14a) were compared to pre-STZ values using a paired two-tailed t-test. For all analyses without multiple comparison adjustment, P < 0.05 was considered statistically significant. For all analyses with multiple comparison adjustment, adjusted significant levels were calculated using 0.05 as the unadjusted significance level. Investigators were not blinded to the treatment groups in our studies and no statistical test was used to determine sample size.

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. Briefly, analysis of gene expres-

60. Johnson, J.D. et al. Different effects of FK506, rapamycin, and mycophenolate mofetil on glucose-stimulated insulin release and apoptosis in human islets. Cell Transplant. 18, 833–845 (2009).

npg


kg/d for 5 d) and then transplanted with S6 cells (~2 × 106 cells/animal) under the kidney capsule at 24 weeks of age. Nondiabetic control mice with no cell transplant were used as a reference for normoglycemia. For the S7 Diabetic Cohort (Fig. 7c,d), 8- to 10-week-old male SCID-beige mice were obtained from Taconic. Mice were rendered diabetic by a single high-dose ip injection of STZ (190 mg/kg) and then transplanted with S7 cells (~1.25 × 106 cells/animal). Nondiabetic C57/Blk6 mice (obtained from the University of British Columbia Centre for Disease Modeling) with no cell transplant were used as controls for the insulin tolerance test experiment.

doi:10.1038/nbt.3033


E R R ATA

Erratum: Reversal of diabetes with insulin-producing cells derived in vitro from human pluripotent stem cells Alireza Rezania, Jennifer E Bruin, Payal Arora, Allison Rubin, Irina Batushansky, Ali Asadi, Shannon O’Dwyer, Nina Quiskamp, Majid Mojibian, Tobias Albrecht, Yu Hsuan, Carol Yang, James D Johnson & Timothy J Kieffer Nat. Biotechnol.; doi:10.1038/nbt.3033; corrected online 16 September 2014

npg


In the version of this article initially published online, there were several errors in the PDF and/or the HTML versions. PDF and HTML versions: In the abstract, the last sentence should have read “in vivo” rather than “in vitro.” In the Methods, under the heading “Differentiation of S4 cells in suspension aggregate cultures,” the units of the cell clump size should have been micrometers rather than millimeters, as follows: “On the last day of culture, S4 cells were treated with 5 mg/ml dispase (Corning, Cat# 354235) for 5 min at 37 °C, followed by gentle pipetting to break into cell clumps (100–200 μm).” In the Methods, under the heading “S4: pancreatic endoderm, PDX1+/NKX6.1+ cells (3 d),” milliliter should have been microliter, as follows: “The resulting cell pellet was resuspended as single cells at a density of ~0.5 × 105 cells/μl on filter inserts (BD, Cat#35-3493 or Corning Cat#3420); 5–10 μl per spot for a total of 0.25–0.5 × 106 cells/spot) at an air-liquid interface.” HTML version: In the legends to Figures 2, 3, 4 and 7, the units of the scale bars should have been micrometers rather than millimeters. The errors have been corrected for the print, PDF and HTML versions of this article.


Functional Enterospheres Derived In Vitro from Human Pluripotent Stem Cells.

Reversal of cellular phenotypes in neural cells derived from Huntington's disease monkey-induced pluripotent stem cells.

Functional properties of in vitro excitatory cortical neurons derived from human pluripotent stem cells.

In vitro organogenesis from pluripotent stem cells.

Fibroblasts derived from human pluripotent stem cells activate angiogenic responses in vitro and in vivo.

Radiation response of mesenchymal stem cells derived from bone marrow and human pluripotent stem cells.

Production of neural stem cells from human pluripotent stem cells.

Human intestinal tissue with adult stem cell properties derived from pluripotent stem cells.

The in vitro generation of lung and airway progenitor cells from human pluripotent stem cells.

Smooth Muscle Precursor Cells Derived from Human Pluripotent Stem Cells for Treatment of Stress Urinary Incontinence.

Neurotrophic activity of jiadifenolide on neuronal precursor cells derived from human induced pluripotent stem cells.

Osteogenesis from human induced pluripotent stem cells: an in vitro and in vivo comparison with mesenchymal stem cells.

Methods for Derivation of Multipotent Neural Crest Cells Derived from Human Pluripotent Stem Cells.

Genome modification leads to phenotype reversal in human myotonic dystrophy type 1 induced pluripotent stem cell-derived neural stem cells.

Engineered Human Muscle Tissue from Skeletal Muscle Derived Stem Cells and Induced Pluripotent Stem Cell Derived Cardiac Cells.

Defining differences among perivascular cells derived from human pluripotent stem cells.

Generation and characterization of functional cardiomyocytes derived from human T cell-derived induced pluripotent stem cells.

Interaction of Salmonella enterica Serovar Typhimurium with Intestinal Organoids Derived from Human Induced Pluripotent Stem Cells.

Germ cells from pluripotent stem cells: mouse versus human.

Studying pancreas development and diabetes using human pluripotent stem cells.

Mesenchymal stem cell population derived from human pluripotent stem cells displays potent immunomodulatory and therapeutic properties.

Modeling ALS with motor neurons derived from human induced pluripotent stem cells.

Three-dimensional Quantification of Dendritic Spines from Pyramidal Neurons Derived from Human Induced Pluripotent Stem Cells.

3D In vitro model of a functional epidermal permeability barrier from human embryonic stem cells and induced pluripotent stem cells.