Exogenous expression of the transcription factors Pdx1, Neurog3, and MafA in human fibroblasts

We first sought to examine whether the transcription factors Neurog3, Pdx1, and MafA could induce expression of the INSULIN (INS) gene in human fibroblasts as readout of the capacity of these cells to be transformed toward a -cell fate. To deliver these factors we employed a polycistronic adenoviral vector carrying the three transgenes (Ad-NPM hereafter), which had been previously used to promote -cell reprogramming from pancreatic acinar cells19. After optimization of adenoviral transduction in fibroblasts (see Methods), abundant (>80%) cells positive for Cherry, which is also encoded by Ad-NPM, were easily observable three and seven days after viral infection (Fig.1a). Likewise, high levels of transcripts encoding the NPM factors were expressed at both time points (Fig.1b). Three days after addition of Ad-NPM we detected marginal levels of INS mRNA that were increased >10-fold by day 7 (Fig.1c). As cell culture formulations can have a major impact on gene expression events and cellular reprogramming, we tested different conditions after Ad-NPM infection. We observed that moving to RPMI-1640 and, to a lesser extent, CMRL-1066 medium and lowering the fetal calf serum concentration to 6% dramatically boosted INS gene activation, reaching values that were 0.12% those of human islets (Fig.1d and Supplementary Figure1). Under the same culture conditions, only a very marginal induction of the INS gene occurred when the N+P+M factors were delivered simultaneously via distinct adenoviruses to human fibroblasts (Supplementary Fig.2). In addition to INS, we discovered that the NPM factors also activated the hormone genes GLUCAGON (GCG) and SOMATOSTATIN (SST), albeit at lower levels than INS as indicated by decreased relative expression values (compared to the housekeeping gene TBP) (Fig.1e). The NPM factors also induced expression of genes encoding islet differentiation transcription factors including NEUROD1, INSM1, PAX4, NKX2-2, and ARX (Fig.1f).

Human fibroblasts (HFF1) were infected with a polycistronic recombinant adenovirus encoding the transcription factors Neurog3, Pdx1, MafA, and the reporter protein Cherry (Ad-NPM). Untreated parental fibroblasts were used as controls (indicated as C in graphs). a Bright field images and Cherry immunofluorescence of control fibroblasts and fibroblasts infected with Ad-NPM at day 3 and 7 post-infection. Scale bar, 100 m. b qPCR of transgenes at day 3 (n=11) and 7 (n=6) after infection with Ad-NPM. c qPCR of human INS at day 3 (n=7) and 7 (n=13) after infection with Ad-NPM. d qPCR of human INS in fibroblasts maintained in the indicated culture media (DM=DMEM; CM=CMRL; RP=RPMI) during 7 days after infection with Ad-NPM or with an adenovirus expressing B-galactosidase (B-gal) (n=310). In yellow, INS mRNA levels in isolated human islets (n=10). e, f qPCR of islet hormone genes (GCG, SST) and islet differentiation transcription factors (NEUROD1, INSM1, PAX4, NKX2-2, ARX) at day 7 post-NPM (n=815). g qPCR of the indicated fibroblast markers at day 3 (n=511) and day 7 post-NPM (n=56). In bf, expression levels are expressed relative to TBP. In g, expression is expressed relative to control fibroblasts, given the value of 1 (dotted line). Data are presented as the meanSEM for the number of samples indicated in parentheses. *P<0.05; **P<0.01; ***P<0.001, between indicated conditions using unpaired t-test (bf), or relative to control fibroblasts using one sample t-test (g).

To further establish if the NPM factors promoted cell fate conversion and not simply activated their target genes in fibroblasts, we surveyed expression of genes associated with the fibroblastic signature, including several factors involved in maintenance of the fibroblastic transcriptional network such as TWIST2, PRRX1, and LHX920. We found that these genes were downregulated as early as three days after NPM introduction. Other fibroblast markers exhibited a more delayed response but, by day 7 post-NPM, all tested genes exhibited significant down-regulation (Fig.1g). Together, these experiments validate that islet cell fate can be induced in human fibroblasts using a defined set of transcription factors.

The observed induction of the islet hormone genes GCG and SST implied that the NPM factors might not specifically endorse -cell fate in fibroblasts. Furthermore, we found that these factors did not induce NKX6-1, which encodes a -cell specific factor required for the formation of pancreatic cells during development21 and key for optimal maturation of stem cell cells in vivo22,23 (Fig.2a). These findings indicated that the NPM factors sub-optimally promoted a -cell state in human fibroblasts. In order to enhance -cell fate over other islet cell identities, we opted to add new transcription factors to the reprogramming cocktail.

Human fibroblasts (HFF1) were infected with Ad-NPM alone or sequentially with Ad-NPM and adenoviruses encoding the transcription factors Pax4 and Nkx2-2. Ad-Pax4 and Ad-Nkx2-2 were added three days after NPM in the two-virus conditions. In the three-virus condition, Pax4 was added three days and Nkx2-2 and six days after NPM (condition called 5TF). All cells were collected ten days after infection with Ad-NPM. a qPCR of the indicated transgenes and endogenous genes. Expression levels are calculated relative to TBP. Values represent the meanSEM (n=412). b Representative immunofluorescence images showing insulin staining (in red) using two different antibodies, one against C-PEP, in untreated fibroblasts and in fibroblasts infected with Ad-NPM alone or with 5TF. Nuclei were stained with Hoechst (in blue). Scale bar, 25 m. *P<0.05; ****P<0.0001 relative to NPM in (b) using one-way ANOVA and Tukeys multiple comparison test.

Pax4 is activated downstream of Neurog3 during development24 and has been shown to favor - over -cell specification25,26, and to contribute to maintenance of the expression of Nkx6.1 in differentiating cells27. Despite that the NPM factors induced endogenous PAX4 mRNA, the expression levels attained might not be sufficient to endorse - over -cell fate. Hence, we treated fibroblasts with an adenovirus encoding Pax4 three days after NPM (Fig.2a). This resulted in the significant enhancement of INS expression as compared to NPM alone but, unexpectedly, GCG expression was also increased (Fig.2a), indicating that ectopic Pax4 improved islet hormone gene expression without apparent impact on - versus -cell fate conversion in human fibroblasts.

As the NKX6-1 gene remained silent in response to NPM+Pax4 (Fig.2a), we tried directly adding Nkx6-1 to the NPM reprogramming cocktail. However, exogenous Nkx6-1 resulted in considerable cell death irrespective of level of expression or timing of introduction. As an alternate approach, we added exogenous Nkx2-2, which also regulates early -cell differentiation and is an upstream activator of Nkx6-1 during mouse islet development21. Treatment with an adenovirus encoding Nkx2-2 three days after NPM led to endogenous activation of NKX6-1 expression with no compromise of fibroblast viability (Fig.2a). Nkx2-2 also induced PAX6, a pan-endocrine gene required to achieve high levels of islet hormone gene expression during mouse pancreas development28,29. Remarkably, ectopic Nkx2-2 reduced NPM-induced GCG gene activation without affecting INS gene expression (Fig.2a).

During development, Pax4 and Nkx2-2 are found in -cell precursors at around the same time, and their parallel activities are thought to enable the -cell differentiation program27. Hence, we tested the effects of including both transcription factors in the reprogramming cocktail. To ensure optimal expression of each transcription factor, we treated cells with Ad-Pax4 and Ad-Nkx2.2 sequentially, at day 3 and day 6 post-NPM, respectively. Following this protocol, the blockade of GCG gene activation and the induction of the NKX6.1 and PAX6 genes seen with NPM+Nkx2.2, and the higher INS expression elicited by NPM+Pax4 relative to NPM alone were all maintained (Fig.2a). Neither Pax4 nor Nkx2-2, added alone or together, had any impact on the minimal INS gene induction shown when the N+P+M factors were delivered via separate adenoviruses to human fibroblasts (Supplementary Fig.2).

Consistent with the gene expression data, staining for insulin protein was more robust in cells reprogrammed with NPM+Pax4+Nkx2.2 than in cells reprogrammed with NPM as assessed using two different antibodies, one against human insulin and another against human C-PEP to exclude possible insulin uptake from the media (Fig.2b). We quantified the immunofluorescence images and found that 67.96.2% of cells in the culture were INS+at day 10.

From here on, we used the sequential introduction of the five transcription factors (5TF protocol, Fig.3a) to generate insulin-producing cells from human fibroblasts (reprogrammed cells will be referred as 5TF cells). At day 10, 5TF cells displayed an epithelial morphology (Fig.3b) and hadnt grown as much as untreated fibroblasts (day 10; 5TF: 441033103 cells/well; control: 23810318103 cells/well, n=18). This decreased cell number was likely due to diminished proliferation, which was evident as soon as one day following Ad-NPM infection (Fig.3c). The capacity of cells to reduce the MTT compound, in contrast, was comparable to that of fibroblasts, indicating that viability was not compromised (Fig.3d).

a Scheme of the reprogramming protocol 5TF (NPM+Pax4+Nkx2.2) showing the sequence of addition of adenoviruses encoding the indicated transcription factor/s. Duration of incubation with each adenovirus is represented with a line. Cells were studied at days 10-11 after initial addition of Ad-NPM. b Representative bright field image of parental fibroblasts and 5TF reprogrammed fibroblasts at day 10. Scale bar, 75 m. c Cell proliferation measured by BrdU incorporation and d cell viability measured by MTT assay for n=3 independent reprogramming experiments. Bars represent values relative to control fibroblasts (given the value of 1, represented by a dotted line). Note that day 4 values are before Pax4 introduction. e, f qPCR of islet/-cell transcription factor and -cell function genes in untreated control fibroblasts (C, n=522), in fibroblasts infected with Ad-NPM alone (n=322) or with 5TF (n=522). Expression levels were calculated relative to TBP. qPCR of the indicated endogenous genes (g) and transgenes (h) at day 21 after initiation of reprogramming (n=913, from 7 reprogramming experiments). Transcript levels are expressed relative to levels in cells at day 10 of the reprogramming protocol (given the value of 1, shown with a dotted line). Data are meanSEM for the number of n indicated in parentheses. *P<0.05, **P<0.01, ***P<0.001 compared to control fibroblasts (c, d), or between indicated bars using unpaired t-test (e, f), or compared to day 10 5TF cells (g, h) using one-sample t-test.

Next we studied expression of selected differentiation transcription factor genes at days 10-11 of the protocol. All genes tested, except PAX4, were more expressed in 5TF relative to NPM (NEUROD1, INSM1, HNF1B, MAFB, PDX1, NEUROG3, NKX2.2) (Fig.3e). Likewise, several genes (PCSK1, KCNJ11, GLP1R, NCAM1) that are linked to -cell function were increased in 5TF cells as compared to NPM cells (Fig.3f). Remarkably, some genes were induced de novo by 5TF (ABCC8, GIPR) (Fig.3f). In line with a loss of GCG activation, the pro-convertase gene PCSK2, which is expressed at higher levels in than in cells30, was reduced by 5TF as compared to NPM (Fig.3f). These results support that sequential introduction of Pax4 and Nkx2-2 after NPM endorses the -cell differentiation program in human fibroblasts. -cell gene activation was sustained for at least twenty-one days after initiation of the protocol despite reduced expression of the reprogramming factor transgenes (Fig.3g, h). Furthermore, expression of several of the tested genes increased with time in culture including NKX6-1, PCSK1, KCNJ11, ABCC8 and CHGB among others (Fig.3g), suggestive of permanent cell lineage conversion.

Glucose-induced insulin secretion by cells is mediated by cellular glucose metabolism, closure of ATP-dependent potassium channels, membrane depolarization and opening of voltage-dependent calcium channels, resulting in an increase in cytosolic Ca2+ that triggers insulin exocytosis. We investigated whether 5TF cells increased intracellular Ca2+ in response to glucose and membrane depolarization elicited by high potassium. We found that 65% of the cells exhibited a response to glucose, high potassium, or both, whilst 35% of cells were unresponsive to either stimulus (Fig.4a and Supplementary Video1). Parental fibroblasts not engineered for 5TF expression were unresponsive to these stimuli (Fig.4b and Supplementary Video2). Among responsive cells, approximately half responded to both glucose and high potassium and half responded only to potassium (Fig.4a). We observed heterogeneity in the amplitude and kinetics of responses among individual cells (Fig.4c). Next, we performed static incubation assays to study GSIS and found that 5TF cells released similar amounts of human insulin at low (2mM) and high (20mM) glucose concentrations (Fig.4d). Thus, even though 5TF cells increased their intracellular calcium in response to glucose and membrane depolarization, they secreted insulin in a constitutive manner.

5TF cells were loaded with the calcium indicator Fluo-4-AM at day 10 of the reprogramming protocol. Single-cell imaging to detect cytosolic calcium was performed in the following sequence: low glucose (2mM, G2), high glucose (20mM, G20) and membrane depolarization with KCl (30mM). a Quantification of the frequency of cells (n=200, from six independent reprogramming experiments) that responded to glucose, membrane depolarization elicited by high potassium or both. Representative measurements of dynamic Fluo-4 fluorescence for (b) six fibroblasts and (c) four 5TF cells. d In vitro insulin secretion by 5TF cells. ELISA determination of secreted human insulin by control fibroblasts (n=413) and 5TF cells (n=16) under non-stimulatory conditions (glucose 2mM) and under stimulatory conditions (glucose 20mM). Data are meanSEM and correspond to six independent reprogramming experiments, 24 biological replicates per experiment.

The differentiation and functionality of many cell types vary dramatically between three-dimensional (3D) and two-dimensional (2D) monolayer cultures, the former being closer to the natural 3D microenvironment of cells in a living organism. Thus, we generated spheroids of 5TF cells (1200-1800 cells/spheroid; average diameter of 12827m) one day after the introduction of Nkx2-2 and maintained them in culture for three additional days (Fig.5a). At the time of collection, insulin-positive staining was easily identified but glucagon and somatostatin staining was undetectable (Fig.5b and Supplementary Fig.3). While INS transcript levels were nearly 2-fold higher in 5TF cell spheroids compared to 5TF cells kept in monolayer, other -cell marker genes, such as the prohormone convertase PCSK1 and the ATP-sensitive potassium channel subunits KCNJ11 and ABCC8, showed a higher response (4 to 5-fold) to 3D culture (Fig.5c). Thus, cell aggregation during the last stage of reprogramming (note that total length of the protocol was not changed) conferred improved activation of genes associated to -cell function. Despite increased gene activation, -cell gene expression in 5TF cell spheroids remained lower than in human islets, with differences ranging widely among examined genes (Fig.5c).

a Schematic representation of the modified 5TF protocol (5TF-3D): cells were moved from 2D to 3D culture during the last three days (days 710) of the protocol. Representative bright field image of 5TF cell spheroids. Scale bar, 100 m. b Representative immunofluorescence image showing insulin staining in red and nuclei in blue (marked with Hoechst) of a 5TF cell spheroid at the end of the reprogramming protocol. Scale bar is 50 m. c qPCR of the indicated genes in 5TF cell spheroids. Transcript levels are expressed as fold relative to levels in 5TF cells maintained in 2D culture throughout the 10-day protocol (given the value of 1, dotted line). Data are meanSEM for n=412. *P<0.05, **P<0.01, ***P<0.001 relative to 2D culture using one-sample t-test. Fold-change differences in expression levels between human islets and 3D-5TF reprogrammed cells are shown in the upper yellow box. d Heat map of differentially expressed genes between parental fibroblasts (C) and 5TF cell spheroids (n=3 reprogramming experiments). e GSEA plots on indicated gene sets and pathways. f Dot plots showing the enrichment analysis on Gene Ontology (GO) and KEGG categories of differentially expressed genes (gained in red, lost in blue) between fibroblasts (C) and 5TF cells. The X-axis represents the adjusted p value, the size of the dot represents the number of enriched genes (count) and the color intensity of the dots represents the percentage of hits in each category. g GSEA plot on -cell disallowed genes. h Relative expression levels of -cell disallowed genes repressed in 5TF cell spheroids as compared to fibroblasts (given the value of 1) based on RNA-seq data normalized expression values. Data are meanSEM (n=3). Insets show mRNA expression of the indicated genes in untreated control fibroblasts (n=5), 5TF cell spheroids (n=6) and human islets (n=5) as assessed by qPCR. Expression levels were calculated relative to TBP. Data are meanSEM. *P<0.05, **P<0.01 relative to control fibroblasts using unpaired t-test.

To obtain a more comprehensive understanding of thecell identity switch induced by the 5TF-3D reprogramming protocol, we performed RNA-sequencing of 5TF cell spheroids and parental fibroblasts. A total of 2806 genes (1186 upregulated, 1620 downregulated) were differentially expressed between both cell populations (adjusted p-value <0.05 and fold-change (FC)>2) (Fig.5d and Supplementary Data1). Gene set enrichment analysis (GSEA) showed that pancreas/-cell and peptide hormone metabolism gene sets were enriched in 5TF cells (Fig.5e). Biological functions associated with gained genes included epithelium development, synaptic signaling, ion transport, calcium sensing and secretion (Fig.5f). Among the upregulated genes related to stimulus-secretion coupling, there were synaptotagmins (SYT1,2,3,6,13,17), syntaxins (SYN2 SYN3), calcium sensors (SCG2) and SNARE protein complexes (VAMP1). Correlating with our previous results, cell cycle and mitotic function genes were enriched among repressed genes (Fig.5e, f). Additionally, GSEA demonstrated that 5TF cells had a lower expression of the gene set associated with the epithelial-mesenchymal transition (Fig.5e). In agreement, functions including cytoskeleton organization and cellular migration were overrepresented among lost genes (Fig.5f). Interestingly, GSEA also revealed that the -cell disallowed gene set, which includes genes that are selectively suppressed in cells and believed to be detrimental for cell function31,32,33, was reduced in 5TF cells (Fig.5g). A total of 23 previously recognized -cell disallowed were significantly downregulated in 5TF cells (Fig.5h). By using qPCR, we confirmed the repression of three of these genes -OAT, LDHA, and SMAD3- which are regarded as part of the core disallowed unit33. Of note, the levels of these genes in 5TF cells matched those of human islets (Fig.5h). Collectively, these results show that 5TF-3D reprogramming promotes a change in the fibroblast transcriptome, including selective gene activation along with specific gene repression events, enabling a change in cell identity from fibroblast towards a -cell fate.

Consistent with gene activation events identified in prior gene expression analyses, immunofluorescence staining showed the presence of the mature -cell markers PCSK1, NCAM1, and KCNJ11 (Kir6.2) in many insulin-positive 5TF cells. PTPRN (IA2) was also expressed albeit more sporadically in insulin-positive 5TF cells (Fig.6a). Using conventional electron microscopy, we looked for the existence of secretory granules and discovered that most cells contained multiple spherical electron-dense prototypical secretory vesicles (Fig.6b). These vesicles showed a high degree of morphological heterogeneity, presumably as consequence of their degree of maturation and/or loading. Although they did not have the appearance of typical insulin-containing granules from primary cells, which are characterized by a clear halo surrounding a dark polygonal dense core34, some of the vesicles exhibited a gray or less electron dense halo and looked like the granules described in immature insulin-positive cells generated in early stem cell differentiation protocols35,36.

a Representative confocal images of 5TF cell spheroids immunostained with the indicated antibodies. Scale bar, 10 m. b Conventional transmission electron microscopy showing a representative image of a 5TF cell spheroid. Prototypical electron dense secretory vesicles (asterisks) are observed dispersed in the cytoplasm. Well-preserved mitochondria (mit), endoplasmic reticulum (ER), Golgi membranes (G) and lipid droplets (LD) are also observed. Inset shows a detail of a secretory vesicle with an average diameter of 450nm. N, nucleus. Scale bars are 200nm (inset) and 500nm. c In vitro glucose-induced insulin secretion by 5TF cell spheroids (n=14, from 8 reprogramming experiments). Secretion by control spheroids composed of parental fibroblasts (n=5) is also shown. d Glucose stimulation Index (ratio between insulin secreted at 20mM glucose vs. 2mM glucose) of 5TF cells maintained in 2D or in 3D (spheroid) cultures (n=1618, from 8 to 10 reprogramming experiments). e Glucose dose curve of insulin secretion by 5TF cell spheroids (n=412, 5 reprogramming experiments). Data are presented as the meanSEM for the number of n indicated in parentheses. *P<0.05; ***P<0.001 between the indicated conditions using unpaired t-test (c), one sample t-test (d) or one-way ANOVA (e).

We next performed static incubation GSIS assays. 5TF cell spheroids exhibited significant insulin secretory response to glucose (fold 20mM/2mM: 2.020.18) as compared to 2D cultures (fold 20mM/2mM: 1.080.15) (Fig.6c, d). To establish the glucose threshold for stimulation of insulin secretion, 5TF cell spheroids were subjected to either 2,5,11 or 20mM glucose. Between 2mM and 11mM/20mM glucose, 5TF spheroids showed a 2.3-fold increase in insulin production on average (Fig.6e). In contrast, although there was some variability, they did not show a statistically significant increase in insulin secretion between 2mM and 5mM glucose (Fig.6e). These observations indicate that 5TF cell spheroids are stimulated at higher glucose threshold; it is interesting to note that human islets have a glucose threshold at 3mM and a maximal response at 15mM37.

The 5TF-3D protocol was repeated on an additional HFF line and produced results that were comparable (Supplementary Fig.4) proving the reproducibility of the reprogramming protocol.

Finally, we studied the stability of reprogramming in vivo. With this aim, we transplanted 300 5TF cell spheroids (10001200 cells/spheroid) into the anterior chamber of the eye (ACE) of non-diabetic immune-deficient NOD scid gamma (NSG) mice (Fig.7a). The ACE allows fast engraftment38 and in vivo imaging39. Ten days following transplantation, we used two-photon microscopy to evaluate in vivo graft re-vascularization and confirmed the presence of functioning vessels in the grafts (Fig.7b). Additionally, by observing the long-term tracer CFDAs fluorescence, we confirmed that the transplanted cells were alive (Fig.7b). To assess the maintenance of insulin expression in vivo, we harvested the eye grafts at day 10 for RNA extraction and immunostaining. Human INS mRNA was readily detectable and levels, calculated relative to human TBP, were comparable to those in 5TF cell clusters prior to transplantation (Fig.7c). In agreement, abundant HLA+(human cell marker) cells that stained for insulin were detected in the eye grafts by immunofluorescence staining (43.52.8% INS+HLA+/totalHLA+, n=5) (Fig.7d, e and Supplementary Fig.5). We observed positive staining for the reprogramming transcription factors in 2030% of the INS+cells (Supplementary Figure6). Although we were unable to discriminate between the two, high transgene expression found by qPCR analysis in eye grafts (Supplementary Fig.6) indicated that the staining represented virally encoded exogenous protein rather than endogenous protein. Since adenoviral vectors do not normally integrate into the host DNA, we speculate that the cessation of cell division induced by reprogramming may explain persistent transgene expression in 5TF cells. In fact, similar findings were reported in reprogrammed human duct-derived insulin-producing cells9. We were able to identify INS+cells in 4 (of 5) grafts harvested one month after transplantation even though their number was reduced relative to day 10 grafts (Supplementary Fig.7). The proportion of INS+HLA+ cells in 30-day grafts was more heterogeneous than in 10-day grafts, and in 3 (of 5) grafts, it was comparable or even higher than that of 10-day grafts, demonstrating the maintenance of reprogramming (Supplementary Fig.7).

a Schematic illustration and image showing 5TF cell spheroids transplanted into the anterior chamber of the eye (ACE) of a normoglycemic NSG mouse. b Vascularization of 5TF cell grafts ten days following transplantation into the ACE. Representative in vivo image depicting functional vessels (RITC-dextran, red) and viable 5TF cells (CFDA, green). Scale bar, 100 m. c qPCR of INS and TBP transcripts in eyes of non-transplanted mice (nt, n=3) and mice transplanted with either control fibroblast spheroids (C, n=3) or 5TF cell spheroids (n=5) collected ten days post-transplantation. INS gene expression in 5TF cell spheroids prior to transplantation is depicted in the blue bar (n=6). INS gene expression is calculated relative to TBP. Expression of TBP relative to mouse Tbp is shown to prove the presence of human cells in eyes receiving control and 5TF spheroids. Data are presented as meanSEM. d Representative immunofluorescence images showing HLA staining in red and insulin staining in green in 5TF cell grafts ten days post-transplantation. Scale bar, 25 m. e Percentage of cells doubly positive for insulin and HLA (relative to total HLA+cells) in 5TF cell grafts at day 10 following transplantation. Each dot corresponds to one eye graft (n=5). f ELISA determination of human insulin in the aqueous humor in un-transplanted mice (n=7), in mice transplanted with either 300 fibroblast spheroids (n=14) or 300 5TF cell spheroids (n=17) at day 10 post-transplantation and in mice transplanted with 150200 human islets (n=4) at day 1215 post-transplantation. Data are presented as meanSEM for the number of n indicated in parentheses. ***P<0.001; ***P<0.0001 between indicated samples using unpaired t-test.

To study if 5TF cells secreted insulin in vivo, we first measured the presence of human insulin by ELISA in the aqueous humor of the transplanted eyes. Human insulin was readily detectable in eyes carrying 5TF cell grafts (17 of 17, ranging from 76 to 1103 pmol/L) whilst no insulin was detected in eyes transplanted with parental fibroblast clusters or in non-transplanted mice (Fig.7f). For comparison, eyes containing 300 5TF spheroids showed on average approximately 20-fold lower levels of human insulin than eyes containing 150200 human islets (Fig.7f). Due to space limitations in the ACE, we transplanted a larger number of spheroids (35005000) into the omentum of normoglycemic NSG mice in order to detect circulating human insulin in host animals. We measured low amounts of human insulin in the plasma of most transplanted mice, and these levels increased in 6 (of 10) mice after receiving an intraperitoneal glucose injection on day 30 post-transplantation (3.60.9 vs 13.93.7pmol/L, p=0.014) (Supplementary Fig.8). Transplants were repeated in other locations yielding similar results (Supplementary Table2). As observed in the ACE grafts, a low number of INS+cells were identified in omentum grafts harvested at 30 days post-transplantation (Supplementary Fig.8). These findings show that, despite restricted survival, reprogramming is maintained and 5TF cells maintain the capacity to release insulin in an in vivo setting.

View original post here:
Direct reprogramming of human fibroblasts into insulin-producing ... - Nature.com