Diabetes treatment by conversion of gut epithelial cells to insulin‐producing cells

Abstract

Insulin‐deficient (type 1) diabetes is treated by providing insulin to maintain euglycemia. The current standard of care is a quasi‐closed loop integrating automated insulin delivery with a continuous glucose monitoring sensor. Cell replacement technologies are advancing as an alternative treatment and have been tested as surrogates to cadaveric islets in transplants. In addition, immunomodulatory treatments to delay the onset of type 1 diabetes in high‐risk (stage 2) individuals have gained regulatory approval. We have pioneered a cell conversion approach to restore insulin production through pharmacological conversion of intestinal epithelial cells into insulin‐producing cells. We have advanced this approach along a translational trajectory through the discovery of small molecule forkhead box protein O1 inhibitors. When administered to different rodent models of insulin‐deficient diabetes, these inhibitors have resulted in robust glucose‐lowering responses and generation of insulin‐producing cells in the gut epithelium. We review past work and delineate a path to human clinical trials.

Converting gut cells into insulin‐producing cells has emerged as an attractive therapeutic option to treat diabetes. Here, we summarize this field and review its clinical potential.

INTRODUCTION

Type 1 diabetes is caused by the loss of insulin‐producing cells through an autoimmune process that involves T‐cell‐mediated recognition of autoantigens, including certain peptides of the insulin molecule ¹ . Patients require lifelong insulin administration to live. Before the discovery of insulin, the disease was invariably fatal, and gruesomely so. Now, individuals with type 1 diabetes have a normal life expectancy, and women with type 1 diabetes can routinely bear children, something that well into the 1980s was marred by awful complications affecting both the fetus's and the mother's health ² . Nevertheless, life on an insulin replacement regimen is far from simple. Insulin is generally administered as multiple daily injections, or by way of an infusion device. Automated infusion devices are now standard of care in countries with advanced economies, much less so elsewhere. The key point is that, unlike type 2 diabetes patients, type 1 diabetes patients require around‐the‐clock insulin coverage to prevent ketoacidotic hyperglycemia, and are exquisitely sensitive to insulin‐induced hypoglycemia. Thus, germane to insulin administration is the need to closely monitor plasma glucose levels. Continuous glucose monitoring sensors, long in the making, have been gaining acceptance by providers and patients alike, but have limited availability in most countries ³ . With the development of sensors and infusion devices, insulin treatment has become much more akin to a physiological replacement, but its complexity and cost are barriers to a normal life for patients and their families. The constant need for monitoring, the perennial anxiety over possible life‐threatening episodes of low glucose and the sheer pain of finger pricks multiple times a day make for a less‐than‐ideal lifestyle that affects not only patients, but their families and coworkers too. Hence, the unmet need for a cure.

What is a “cure”? As a young patient once put it, it is “something that I don't have to think about every day”. Ideally, it is a permanent or at least long‐lasting restoration of physiological insulin production, so that glucose monitoring becomes unnecessary. Short of an intervention to achieve this goal, an oral treatment that does not require constant glucose monitoring would also be desirable.

PROPOSED SOLUTIONS: AN EVOLVING LANDSCAPE

We will not discuss ongoing efforts to prevent type 1 diabetes. Suffice it to say that individuals at risk can be identified using genetic and autoimmune markers; that is, anti‐islet or islet‐component antibodies. A landmark achievement in this area was the approval of Teplizumab – a humanized anti‐CD3 antibody initially investigated for its ability to prevent allograft rejection in kidney transplant ⁴ – as a preventative treatment that can delay disease onset by an average of 2 years in individuals with stage 2 type 1 diabetes (defined as multiple autoantibodies and dysglycemia, but no overt hyperglycemia) ⁵ . Several other immune modulators show promise, including rituximab, abatacept, thymoglobulin (leading to T‐cell depletion), alefacept ⁶ , golimumab ⁷ , as well as Janus kinase inhibitors ⁸ . The reader interested in this topic is referred to an excellent recent review ⁹ . Although the approval of teplizumab should be viewed as a milestone, this approach is not a cure. The disease is postponed, but not avoided altogether. Furthermore, identifying individuals at risk in the general population remains a tall order, thus limiting the usefulness of prevention‐based approaches. Type 1 diabetes surveys in population‐based cohorts show the magnitude of the effort required. For example, the yield of a screening campaign to detect islet autoantibodies in 90,000 Bavarian children netted 310 individuals with multiple positive antibodies. Of these, 17 were classified as stage 2 and, hence, would have qualified for teplizumab treatment ¹⁰ (note that replizumab is not yet approved for use in children aged <8 years, so this is simply to illustrate our point).

Advances in the treatment of established disease (Figure 1) can be subdivided into two main areas. One is based on automated insulin delivery: a tool consisting of an interstitial glucose monitoring sensor and a tethered or tubeless insulin infusion device connected to a smartphone app ³ . This is standard of care in many diabetes centers – including ours – operating in countries with advanced economies. Substantial improvements in predictive algorithms have partly addressed the issue of reduced sensitivity and delayed kinetics of glucose measurements in the interstitial fluid, improving quality of life for patients and their families through superior detection of hypoglycemia and correction of hyperglycemia, as well as meal coverage. Physicians and family members can also access the data remotely, simplifying consultation and management. This is a mature technology that should ideally be available to most patients with type 1 diabetes, but insurance coverage and socioeconomic barriers still limit its widespread adoption.

Figure 1.

Advances in type 1 diabetes treatment – a timeline. A simplified chronicle of alternative treatments for type 1 diabetes. IS, immune suppression.

A second area of research, still in the investigational stage, utilizes surrogate insulin‐producing β‐like cells to replace the body's own missing cells. Long held to be a visionary dream, β‐like cells derived from embryonic or induced pluripotent stem (iPS) cells are now a commodity ¹¹ . Clinical trials of β‐like cells transplanted through intra‐portal infusion in patients with hypoglycemia unawareness and repeated hypoglycemic episodes are underway ¹² , ¹³ . Patients receiving “naked” cells require subsequent immune suppression. Consistent with the preclinical data in animal models, it takes several weeks before the transplanted cells begin to produce insulin, probably reflecting the elusive concept of “in vivo maturation” of iPS‐derived β‐like cells ¹¹ . The trial of VX‐880, a combination of stem cell‐derived β‐like cells and immunosuppression, is currently on hold due to two patient deaths unrelated to treatment. An additional trial is evaluating the delivery of cells through an encapsulation device, thus circumventing the need for immune suppression. Delivering cells in a barrier device offsets the risk of “misbehaving” cells, but requires initial surgical placement and periodic replacement. For this treatment to go mainstream, the device must be easy to implant and remove, be of limited size, and the cells' performance must be long‐lasting. These factors are presently unknown.

A TRANSFORMATIVE IDEA: GUT AS THE NEW PANCREAS

Along with the aforementioned approaches, the idea of converting other body cell types into β‐like cells has been entertained for many years. However, bizarre claims and poor reproducibility have marred this field, which, as a result, has fallen out of favor just as the other approaches were gaining momentum ¹⁴ . There is, however, a glimmer of hope in the conversion of intestinal cell types into glucose‐sensing, β‐like cells. The Kieffer laboratory established the feasibility of this approach by showing that insulin expression can be conferred to intestinal K cells by a gastric inhibitory peptide promoter‐driven insulin transgene ¹⁵ . The transgenically‐encoded insulin confers protection from diabetes in non‐obese diabetic (NOD) mice, a model of autoimmune insulin‐deficient diabetes, providing proof‐of‐principle that insulin is functional, and that this approach can circumvent the autoimmunity ¹⁶ . The intestine contains the body's largest assembly of endocrine cells. However, unlike endocrine glands, such as the pancreas, thyroid and adrenals, enteroendocrine cells are not clustered, but scattered throughout the length of gut epithelium. Enteroendocrine progenitor cells share much in common with pancreatic endocrine progenitor cells, arise from neurogenin‐3‐expressing cells and, not surprisingly, give rise to similar cell types in both organs, such as somatostatin or ghrelin cells ¹⁷ (Figure 2). In other instances, progenitors in the two organs give rise to cells that make peptides derived from alternative splicing of the pre‐proglucagon gene ¹⁸ . Finally, there are types of endocrine cells that are present only in one, but not the other, organ. Insulin‐producing cells are an example of pancreas‐specific cells, and cholecystokinin‐producing cells are an example of gut‐specific cells.

Figure 2.

Derivation of pancreatic and enteric endocrine cells. In both the intestine and pancreas, endocrine progenitor cells are marked by the transient expression of transcription factor neurogenin‐3 (Neurog3). However, further instructive signals direct the development of organ‐specific cells, of which pancreatic β‐cells are a prime example. 5‐HT, 5‐hydroxytryptamine; FOXO1, forkhead box protein O1.

FORKHEAD BOX PROTEIN O1 AND ENTEROENDOCRINE CELL DIFFERENTIATION

Forkhead box protein O1 (FOXO1) is one of three highly homologous genes encoding forkhead box (a deoxyribonucleic acid‐binding domain)‐containing (hence the name) transcription factors that convey terminal differentiation and metabolic signals in various cell types ¹⁹ . It is not a “pioneering” transcription factor that drives early development of organs and tissue types. However, in studies of terminal cellular differentiation, FOXO1 has been shown to control the maturation of adipocytes ²⁰ , myocytes ²¹ and pancreatic endocrine progenitors ²² . In human, nonhuman primate and rodent intestine, it becomes gradually restricted to the enterochromaffin cell ²³ , ²⁴ . When FOXO1 is ablated by gene knockout or inhibited by small interfering ribonucleic acid (RNA) ²⁴ , a subset of enterochromaffin cells convert to β‐like cells (Figure 3) that not only make insulin, but also secrete it in a remarkably physiological fashion ²³ . The newly arisen intestinal insulin‐producing cells have all the markings of pancreatic β‐cells, and can take over the function of pancreatic β‐cells after the latter have been destroyed by the toxin, streptozotocin, effectively “curing” diabetes in rodents ²³ . Unlike iPS‐derived β‐like cells, enteroendocrine‐derived β‐like cells show glucose‐dose‐dependent insulin release as soon as they arise and do not require maturation ²³ , ²⁵ .

Figure 3.

Gut insulin‐immunoreactive cells. Intestinal insulin immunohistochemistry (red). Ablation of forkhead box protein O1 or forced expression of MafA, Pdx1 and neurogenin‐3 can convert a subpopulation of gastrointestinal epithelial cells into insulin‐immunoreactive cells. Experiments in rodent models of insulin‐deficient diabetes, as well as human and mouse organoids, show that these cells release insulin in a glucose‐dose‐dependent fashion, and can be inhibited by the adenosine triphosphate‐sensitive potassium channel activator, diazoxide.

These experiments found independent confirmation in work carried out in the Stanger ²⁶ and Zhou laboratories ²⁷ . These investigators asked which tissues or organs harbor cells with the potential to be converted into insulin‐producing cells by providing three genes – neurogenin‐3, Pdx1 and MafA – that are sequentially required during embryonic development to generate β‐cells, and can together convert exocrine pancreatic cells into β‐cells ²⁸ . They found that a multicistronic transgene encoding the three transcription factors is sufficient to convert intestinal epithelial cells to functional β‐like cells ²⁶ , ²⁷ . Interestingly, the same three genes are activated in response to the FOXO1 knockout ²³ , ²⁴ . Thus, either FOXO1 knockout or activation of neurogenin‐3, Pdx1 and MafA seem to impinge on the same differentiation pathway. Hence, there is a reasonable consensus that this pathway can be leveraged to convert intestinal cells into insulin‐producing cells.

HUMAN RELEVANCE

Although remarkable, the initial data were obtained in laboratory animals, and might or might not have been applicable to humans with type 1 diabetes. To show the feasibility of this approach in humans, we used both iPS‐derived and primary human gut organoids to test the conversion of gut cells into insulin‐producing cells, and showed that treating human gut organoid cultures with adenovirus encoding a dominant‐negative FOXO1 mutant or short hairpin RNA directed against FOXO1 yields cells with β‐like cell features that make insulin and release it on glucose challenge ²⁴ . Thus, the ability of gut endocrine cells to become insulin‐producing cells is not limited to rodents, but it is a shared property of human cells.

The present study also allowed us to further pinpoint a cell type that undergoes conversion into β‐like cells in human gut organoid cultures as the enterochromaffin cell that synthesizes serotonin (5‐hydroxytryptamine [5‐HT]). Interestingly, pancreatic insulin‐producing β‐cells also make 5‐HT ²⁹ , ³⁰ . In fact, 5‐HT‐producing enterochromaffin cells are the closest relative to an insulin‐producing cell in the pancreas. We surmise that enterochromaffin cells are the evolutionary remnant of hitherto unknown fetal gut insulin‐producing cells, and that by removing FOXO1 we restored this differentiation stage. Several lines of evidence support this statement. First, using cell lineage tracing in transgenic mice and gut organoids, we were able to show that converted β‐like cells arise from enterochromaffin cells (defined by the 5‐HT biosynthetic pathway master regulator, Tph1) ³¹ . Second, evidence from human fetuses indicates that insulin‐immunoreactive cells are present in the intestine, at least until gestational week 16 ³² . Interestingly, these cells appear to be FOXO1‐negative, and are comprised of at least three subpopulations: one containing both insulin protein and messenger RNA, one containing only protein, and one containing only messenger RNA. There are several different explanations for this finding, one of which is that the insulin‐immunoreactive cells are not replicating and represent a transient cell type ³¹ . Third, it has been shown that enterochromaffin cells are a substantial byproduct of differentiation of stem cells to β‐like cells, consistent with the possibility that enterochromaffin and β‐cells are developmentally related ³³ . Based on these converging lines of evidence, a potential explanation of our findings is that FOXO1 ablation arrests enteroendocrine progenitor cell differentiation at a fetal‐like stage that includes the insulin‐immunoreactive cells. To illustrate this point, we carried out single‐cell messenger RNA analysis of gut‐derived β‐like cells, and found extensive similarities with pancreatic β‐cells ³¹ .

PHARMACOLOGICAL FOXO1 INHIBITION

To move forward toward clinical applications, we resolved to develop small molecule inhibitors of FOXO1. Over the past decade, we have worked with colleagues in the pharmaceutical industry to discover, characterize and test these compounds. We carried out a high‐throughput screening of a small molecule library using a FOXO1‐specific reporter assay as a readout, followed by a specificity assay to rule out FOXO3 cross‐reactivity and a counter‐screen against FOXA2. In an initial collaborative publication with AstraZeneca, we reported structures of 13 compounds that were further refined for cytotoxicity and potency, and then tested in primary hepatocytes for their effects on glucose production, a classic effect of FOXO1 ³⁴ . This work established the concept of selective FOXO1 modulators, in that it led to the discovery of compounds with selective activity against genes that are activated or inhibited by FOXO1. These compounds also served as a backbone for further independent work carried out at Forkhead Biotherapeutics (New York City, NY, USA), which led to the discovery of ~400 additional compounds with improved potency and specificity ³⁵ .

There are precedents for the synthesis of FOXO1 inhibitors. However, when we tested the commercially available AS1842856 compound, we found that its actions were indistinguishable in FOXO1 knockout animals compared with wildtype, suggesting that they are not selective for FOXO1, possibly affecting FOXO3 or other elements of the Akt signaling pathway ³⁶ . The demonstration that FOXO1 can be modulated by small molecule inhibitors should allay concerns about the tractability of this biologically important target ³⁸ . The drug development industry is understandably cautious about targeting FOXO1, despite its clear biological relevance to several disease processes, for two main reasons: druggability and potential adverse effects of FOXO1 inhibition. The work described in this review should address the issue of druggability. Whether small molecules are the best approach remains to be seen, but the success of this work at least shows feasibility.

Regarding potential adverse effects of FOXO1 inhibition, cancer and autoimmunity are often quoted based on the mouse knockout literature. In this regard, it is important to note that an orally available FOXO1 inhibitor can be modified to be gut‐restricted and have minimal systemic exposure, thus allaying such concerns to an extent. In addition, with regard to cancer, it should be noted that evidence in the literature is mixed as to whether FOXO1 loss‐of‐function predisposes to or protects from cancer. Limited neoplastic predisposition is observed in mice genetically lacking all FOXO isoforms (1, 3a and 4) ³⁹ , and should therefore not affect FOXO1‐specific inhibitors, such as the inhibitors we discovered. On the contrary, there is evidence of activating somatic mutations of FOXO1 in non‐Hodgkin's lymphoma ⁴⁰ and hepatocellular carcinoma ⁴¹ . Indeed, a FOXO1 inhibitor has been shown to exert a strong anti‐neoplastic effect in mouse lymphoma ⁴² . Thus, there are also potential indications for FOXO1 inhibitors in cancer treatment.

Concerns about autoimmunity in the wake of FOXO1 inhibition arise from genetic knockout of FOXO1 in regulatory T cell (Treg) lymphocytes, giving rise to a congeries of potentially lethal autoimmune manifestations, including pancreatitis, hind limb paralysis, multiorgan lymphocyte infiltration, anti‐nuclear antibodies and expanded germinal centers ⁴³ . A key role of FOXO1 in human Tregs also emerged from a CRISPR screen, although in this case it appears to be dependent on specific cues; for example, interleukin‐12 activation ⁴⁴ . Whether the mouse knockout is predictive of human phenotypes remains to be seen. In the aforementioned CRISPR study, the redundancy of transcriptional regulation of Treg development appears to provide substantial backup mechanisms that should intervene in the absence of FOXO1 ⁴⁴ , as we also showed in a recent study of the effects of FOXO1 ablation in the liver ⁴⁵ . Be that as it may, close monitoring of Treg function will be an essential early safety readout in clinical trials of FOXO1 inhibitors.

FOXO1 INHIBITORS LOWER GLYCEMIA IN INSULIN‐DEFICIENT DIABETES

We tested various small molecule FOXO1 inhibitors for their ability to generate insulin‐producing cells in vivo (Table 1). We have shown that genetic FOXO1 ablation expands the enteroendocrine progenitor cell pool ²³ . Using the small molecule FOXO1 inhibitor, FBT10, we reproduced this finding using primary mouse organoid cultures and observed increased numbers of enteroendocrine progenitors, as well as terminally differentiated enteroendocrine cells. When administered orally to Akita mice, a model of insulin‐deficient diabetes due to impaired processing of the insulin molecule, FBT10 was well tolerated and lowered glucose by ~200 mg/dL while generating insulin‐positive cells in the gut at a rate of approximately one cell per intestinal villus. We also detected robust insulin secretion in response to glucose in primary organoids treated with this compound ²⁵ .

Table 1.

Summary of the anti‐diabetic effects of forkhead box protein O1 inhibitor treatment

Metabolic parameter in mouse model	Glucose‐lowering Akita	Glucose‐lowering STZ	Glucose‐lowering NOD	Gut insulin cells (% per villus)	GTT
Treatment
Single agent FBT374	200 mg/dL	200 mg/dL	NT	0.4	NGT
Dual agent FBT374 or FBT432 + Notch‐i	140 mg/dL	300 mg/dL	NT	0.8	NGT
Triple agent FBT10 + Notch‐i + Tgfβ‐i	NT	NT	400 mg/dL	1	NGT

These data summarize the effects of different small molecule forkhead box protein O1 inhibitors used as monotherapy or combination therapy in different murine models of insulin‐deficient diabetes. GTT, glucose tolerance test; i, inhibitor; NGT, normal glucose tolerance; NT, not tested; Tgfβ, transforming growth factor‐β.

FBT374 and FBT432 are orally bioavailable, potent, and selective FOXO1 inhibitors developed by Forkhead Biotherapeutics ³⁷ . In animals treated with streptozotocin to induce hyperglycemia, administration of either compound resulted in a significant glucose‐lowering effect (100–400 mg/dL) that persisted until the end of the experiment, with lower fasting glucose, significant improvement of intraperitoneal glucose tolerance tests and reversal of ketonuria. In this treatment, we observed insulin‐immunoreactive cells in the duodenum, and an increased total number of enteroendocrine as well as enterochromaffin cells, similar to data in human gut organoids ²⁴ . We did not observe increased plasma insulin levels, and our explanation is that the insulin is not released in bulk, as it is by islets, but in a drip‐like fashion over the entire length of the gut, and is avidly taken up by the liver, making the assessment of plasma levels problematic. However, we were able to detect restoration plasma insulin levels in NOD mice, as shown below.

Based on prior research showing a synergistic interaction of FOXO1 and Notch signaling ²¹ , ⁴⁶ , and based on the finding that Notch inhibition increases the enteroendocrine cell pool, we tested combination therapy with FOXO1 and Notch inhibition ⁴⁵ . We showed that combination treatment with the first‐generation FOXO1 inhibitor, FBT10, and the gamma‐secretase inhibitor, PF‐03084014, has additive effects on gut β‐like cell generation and glucose levels in Akita mice ²⁵ . We also showed a synergistic glucose‐lowering effect by combining another FOXO1 inhibitor, FBT432, with PF‐03084014 in streptozotocin‐treated mice ⁴⁷ . The effects of treatment on fasting glycemia, intraperitoneal glucose tolerance, glucosuria and ketonuria were remarkable. Combination treatment with PF‐03084014 and FBT432 caused a synergistic increase of insulin‐positive cells, mostly localized to the base of crypts in the proximal gut. Collectively, these results are consistent with the hypothesis that FOXO1 inhibition drives enteroendocrine differentiation into β‐like cells, and that this effect is significantly enhanced by Notch inhibition.

Endocrine progenitor cells are marked by expression of transcription factor neurogenin‐3. From this pool of progenitor cells also arise limited subsets of goblet and Paneth cells ¹⁷ . This observation raised the question of whether these sublineages can also be converted to insulin‐producing β‐like cells. To address this, we implemented a two‐pronged approach in which gut organoids carrying a reporter insulin 2 knock‐in allele were tested in a medium‐throughput screen to detect pharmacological combinations that give rise to converted cells, which were detected by quantitative flow cytometry based on activation of the reporter insulin gene. We found that a combination of FOXO1, Notch and transforming growth factor‐β inhibition resulted in a strong induction of insulin gene expression. We tested this hypothesis in vivo in mice and found that this combination was highly effective at inducing insulin‐immunoreactive cells, and lowered glycemia not only in streptozotocin‐diabetic mice, but also in an autoimmune model of diabetes, NOD mice. We also detected a significant increase in plasma insulin levels in this model ³¹ . To investigate the nature of cells undergoing conversion, we used lineage tracing approaches, and found that a subset of goblet and Paneth cells can also undergo conversion to insulin‐immunoreactive cells ³¹ . In summary, experiments in multiple laboratories by independent investigators using different compounds alone and in various combinations in an array of murine diabetic models all point to the same conclusion, that FOXO1 inhibition can convert gut cells into insulin‐producing cells (Table 1).

ADVANTAGES OF THE APPROACH

Conversion of gut cells into β‐like cells might leverage properties intrinsic to a developmental stage of certain cell types to reactivate the insulin production program. It is neither cell replacement nor transplant, but rather a semi‐physiological intervention. Assuming that it can be realized in vivo in humans, what would the advantages of this approach be?

1. Gut cells are readily accessible to orally delivered drugs, and they can be targeted by compounds with minimal systemic exposure. This would eliminate the need for injections or the wearing of infusion devices, and lower concerns related to systemic FOXO1 inhibition.

2. Converted cells show glucose dose‐dependent insulin release, indicating that they respond to physiological changes in plasma glucose levels. This would obviate the need for continuous glucose monitoring, as well as the risk of hypoglycemic episodes, a bane of the lives of individuals with type 1 diabetes.

3. Enteroendocrine cells are plentiful and, unlike pancreatic endocrine cells, they regenerate constantly throughout life. In fact, unlike pancreatic endocrine cells, which turnover very slowly, gut epithelial cells are replenished every 3–5 days. Thus, treatment might potentially be given weekly or even less frequently.

4. Enteroendocrine cells are already poised to make and release hormones, because they possess all the machinery required to sense nutrients, and synthesize, process and release peptides. They do not need to be reprogrammed from the ground up. The process of converting them into insulin‐producing cells does not require fundamental changes in cell lineage.

5. Once converted to insulin‐producing cells, these cells might be less exposed to the autoimmunity that killed the patient's own pancreatic β‐cells, for two reasons. First, the gut has a different immune profile from the pancreas (for one thing, as it is constantly exposed to foreign agents in the form of food, the immune system in the gut is more tolerant to antigens than in the rest of the body); second, as cell turnover is rapid, they are more likely to escape or outlast the autoimmune attack. We aim to focus initial human clinical trials on safety and efficacy in generating gut β‐like cells; once this hurdle has been cleared, a separate proof of concept will be required to assess survival of converted cells in an autoimmune setting, similar to trials of iPS‐derived β‐like cells. It should be pointed out that every approach to replace insulin faces this hurdle.

6. Obvious as it might sound, it is also worth remembering that, in addition to a best‐case scenario in which this treatment can do away with insulin altogether, there are intermediate scenarios in which this treatment can restore sufficient basal insulin production as to free patients from the need to be tethered for life to an insulin infusion device, while being constantly watchful of hypoglycemia. This can also be considered a treatment breakthrough.

In summary, multiple small molecule FOXO1 inhibitors, alone or in combination with Notch and transforming growth factor‐β inhibitors, result in the generation of insulin‐immunoreactive gut cells that secrete insulin and ameliorate glycemia in three different mouse models of insulin‐deficient diabetes: NOD, Akita and streptozotocin. In all instances, the glucose‐lowering effect is comparable to that of exogenously administered insulin. Long‐term safety and toxicity studies, as well as testing in non‐human primates, are underway. These data provide support for the vision of an oral treatment to replace insulin by converting gut epithelial cells.

DISCLOSURE

DA is an Editorial Board member of Journal of Diabetes Investigation and a co‐author of this article. To minimize bias, he was excluded from all editorial decision‐making related to the acceptance of this article for publication. The other authors declare no conflict of interest.

Approval of the research protocol: N/A.

Informed consent: N/A.

Approval date of registry and the registration no. of the study/trial: N/A.

Animal studies: All animal studies have been reviewed and approved by the respective institutional animal care and utilization review boards at Columbia University and Forkhead Biotherapeutics.