Harmine and exendin-4 combination therapy safely expands human β cell mass in vivo in a mouse xenograft system

Abstract

Five hundred thirty-seven million people globally suffer from diabetes. Insulin-producing β cells are reduced in number in most people with diabetes, but most individuals still have some residual β cells. However, none of the many diabetes drugs in common use increases human β cell numbers. Recently, small molecules that inhibit dual tyrosine-regulated kinase 1A (DYRK1A) have been shown to induce immunohistochemical markers of human β cell replication, and this is enhanced by drugs that stimulate the glucagon-like peptide 1 (GLP1) receptor (GLP1R) on β cells. However, it remains to be demonstrated whether these immunohistochemical findings translate into an actual increase in human β cell numbers in vivo. It is also unknown whether DYRK1A inhibitors together with GLP1R agonists (GLP1RAs) affect human β cell survival. Here, using an optimized immunolabeling-enabled three-dimensional imaging of solvent-cleared organs (iDISCO⁺) protocol in mouse kidneys bearing human islet grafts, we demonstrate that combination of a DYRK1A inhibitor with exendin-4 increases actual human β cell mass in vivo by a mean of four- to sevenfold in diabetic and nondiabetic mice over 3 months and reverses diabetes, without alteration in human α cell mass. The augmentation in human β cell mass occurred through mechanisms that included enhanced human β cell proliferation, function, and survival. The increase in human β cell survival was mediated, in part, by the islet prohormone VGF. Together, these findings demonstrate the therapeutic potential and favorable preclinical safety profile of the DYRK1A inhibitor–GLP1RA combination for diabetes treatment.

INTRODUCTION

Type 1 diabetes (T1D) results from autoimmune destruction of pancreatic β cells, but most, and perhaps all, people with T1D have residual β cells at autopsy and can secrete small amounts of insulin (1, 2). β cell mass is also reduced in type 2 diabetes (T2D) by ~40 to 60%, attributable to genetic predisposition, metabolic stress, death, and de-differentiation (3, 4). Despite the clear need for β cell regenerative therapies, none of the current diabetes drugs addresses the fundamental problem of β cell deficiency. Even in first-world countries, 70% of people with diabetes fail to achieve targets for blood glucose control (5–7). These considerations have led to alternate approaches, such as automated closed-loop insulin delivery systems, whole pancreas transplant, pancreatic islet transplant, and transplant of human embryonic stem cells differentiated into β cells (8–12). Although each of these approaches represents a major advance in diabetes care, none is scalable to the hundreds of millions of people with diabetes. Thus, an urgent need for alternate therapies remains.

Over the past decade, several groups have demonstrated that small-molecule dual tyrosine-regulated kinase 1A (DYRK1A) inhibitors, exemplified by harmine and others, alone or in combination with glucagon-like peptide 1 (GLP1) receptor agonists (GLP1RAs) are effective at activating human β cell proliferation (13–20). DYRK1A inhibitors have an additional benefit: They can enhance human β cell differentiation. For example, harmine increases expression of RNAs encoding β cell specification factors and differentiation markers such as PDX1, MAFA, MAFB, NKX6.1, PCSK1, GLP1R, and SLC2A2, accompanied by increased abundance of their cognate proteins and by enhanced glucose-stimulated insulin secretion (GSIS) from human islets derived both from healthy donors and from those with T2D (13–15). However, there is some doubt whether this will translate into clinically relevant increases in human β cell mass.

In contrast with their effects on human β cell proliferation, DYRK1A inhibitor potential for enhancing human β cell survival is understudied. For example, although GLP1RAs have been reported to enhance β cell survival in vitro and in vivo (21), whether DYRK1A kinase antagonists alone or in combination with GLP1RAs might have prosurvival effects is unknown. Similarly, there is no information regarding the biology of the DYRK family in islet vascularization.

The granin-like hormone VGF is synthesized by pancreatic islet and other cells (22, 23), and its posttranslational processing generates multiple smaller peptides, some of which are secreted upon stimulation (24). VGF peptides have biological roles including regulation of insulin secretion, food intake, and energy balance (25–27). In β cells, VGF is induced by NKX6.1 and cAMP (adenosine 3′,5′-monophosphate), and the VGF-derived peptide TLQP-21 enhances β cell survival and function (25–27). VGF-deficient mice display alterations in circulating insulin concentration and blood glucose (23). However, whether VGF family peptides contribute to the beneficial effects of the DYRK1A inhibitor–GLP1RA combination in human islets in vitro and in vivo is unknown.

Thus, despite progress in DYRK1A inhibitor–GLP1RA approaches to human β cell regeneration and function, critical unmet challenges remain. These include a demonstration that the regenerative effects of DYRK1A inhibitors, alone or in combination with GLP1RAs, translate into an increase in human β cell mass in diabetic models in vivo and that this treatment is safe. Additionally, it is unclear whether the DYRK1A inhibitor–GLP1RA combination modulates β cell survival and islet vascularization that might contribute to β cell mass expansion. We address these challenges and the underlying mechanisms in this report.

RESULTS

iDISCO⁺ imaging accurately quantifies human β cell mass in vivo

To assess human β cell mass in vivo, we transplanted human pancreatic islets into both right and left kidney capsules of immunodeficient Rag1^−/− mice, with subcutaneous osmotic minipump infusion of vehicle (water), harmine, the GLP1RA exendin-4, or harmine plus exendin-4 (hereafter referred to as “H+E”) for up to 3 months, followed by perfusion and tissue fixation (Fig. 1A). One kidney was harvested for histology and immunohistochemistry, and the other was harvested for iDISCO⁺ tissue clearing, immunolabeling for insulin and glucagon, light sheet microscopy, and three-dimensional (3D) image analysis to assess islet mass. Before studying harmine efficacy, we validated the accuracy and reproducibility of iDISCO⁺ (Fig. 1, B to F, and movie S1) and Imaris quantitative assessments of human β and α cell mass (Fig. 1, G to J, and movies S2 to S9). In these control experiments, islets from healthy donors [100, 300, 500, and 1000 islet equivalents (IEQs)] were transplanted into the renal capsular space of untreated Rag1^−/− mice. The kidneys were harvested either immediately or 14 days after transplant for assessment of β and α cell mass. This allowed comparison of actual measured initial transplanted islet numbers and IEQs versus the iDISCO⁺- measured islet mass (β and α cell) in in situ islet grafts harvested after transplant. The iDISCO⁺ β (Fig. 1, G and H, and movies S2 to S5) and α cell mass (Fig. 1, I and J, and movies S6 to S9) measurements were linearly consistent and reproducible, showing that β and α cell volumes quantified by iDISCO⁺ increased in an islet dose–related manner, correlating with the predicted β, α, and total cell volumes. iDISCO⁺-inferred volume of islets harvested immediately (red line) loosely followed the theoretical black line but was ~35% lower, reflecting the fact that not all cells in whole islets (black line) are β cells (red line) (Fig. 1H). By 2 weeks (blue line), β cell volume in all islet preparations had declined by ~30%, likely reflecting additional β cell death in the 2-week interval between transplant and harvesting (Fig. 1H). Additionally, α cells comprised a smaller volume of total islet volume as expected, and the sum of β cell volumes (Fig. 1H) and α cell volumes (Fig. 1J) was similar to the total theoretical islet cell volume (black lines), reflecting that the large majority of islet cells were either α or β cells.

Fig. 1. Quantification and validation of iDISCO⁺ α and β cell imaging.

(A) Schematic of the study design. Human islets were transplanted under both kidney capsules of immunodeficient mice implanted with continuous infusion minipumps containing vehicle (water), harmine, exendin-4, or H+E. After 1 week, 1 month, or 3 months of treatment, mice were perfusion-fixed, and kidneys were harvested and clarified by iDISCO⁺, then immunolabeled with insulin or glucagon antibodies, and examined by light sheet microscopy, with 3D images analyzed using Imaris for calculation of islet cell volume. Generated with BioRender. (B) A 3D image of β cells (green) in one human islet graft (smooth muscle actin, red) harvested from an untreated mouse 2 weeks after transplantation. Scale bar, 2000 μm. See movie S1. (C to F) Images of the human islet graft (insulin, green; smooth muscle actin, red) in (B) at different magnifications [scale bars, 2000 μm (C), 1000 μm (D), and 500 μm (E), respectively] in the light sheet microscope and (F) in a confocal microscope (scale bar, 100 μm). Insets indicated by a box. (G) Example images of islets transplanted in progressively larger IEQs under the kidney capsule in untreated immunodeficient mice and harvested immediately (insulin, green). Scale bar, 500 μm. Inset contains an image of the created surface used for volume quantification. See movies S2 to S5. (H) Mathematical validation of iDISCO⁺ quantification of human β cell mass. Black line represents the calculated theoretical volume of 100, 300, 500, and 1000 human IEQs (1 IEQ = 125-μm-diameter islet), including all islet cell types, of which ~30 to 70% are β cells or α cells. Red line represents iDISCO⁺-measured β cell volume in 100, 300, 500, and 1000 human IEQs transplanted as in (G), after which kidneys were removed immediately and processed as in (A) (n = 3 different human islet preparations). The blue line is the same as the red line, except that the islet grafts were left in place for 2 weeks before harvesting and processing (n = 4 to 6 different human islet preparations). Mice transplanted with human islets for these experiments were untreated. (I and J) Similar to (G) and (H) but for validation of α cell volume. (I) Example images of α cells in grafts immunolabeled for glucagon in 100, 300, 500, and 1000 human IEQs. Scale bar, 500 μm. Inset contains an image of the created surface used for volume quantification. See movies S6 to S9. (J) Assessment of human α cell volumes in human islet grafts. Black line, calculated theoretical volume; red line, iDISCO-measured α cell volume of islet grafts harvested immediately after transplantation; blue line, iDISCO-measured α cell volume of islet grafts harvested 2 weeks after transplantation.

DYRK1A inhibitors together with exendin-4 increase human β cell mass in nondiabetic mice

We next assessed the effects of vehicle, harmine, exendin-4, or the H+E combination on human β cell volume, proliferation, and death in islet grafts (300 IEQs) transplanted into mice treated for 4 or 12 weeks. After 4 weeks of treatment, β cell mass was comparable in the vehicle, exendin-4 alone, and harmine alone groups (0.1 to 0.2 mm³) but significantly greater (0.325 mm³, P = 0.020) in the H+E group (fig. S1, A and B). Compared with the vehicle and exendin-4 groups, human β cell proliferation (Ki67) was threefold higher in the H+E group, whereas β cell death [terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) assay] was not different among the groups (fig. S1, C and D). In contrast, α cell proliferation was only marginally detectable, and α cell mass did not change (fig. S1, E and F).

On the basis of these results, we extended these studies to 3 months in a second cohort of mice using seven or eight different human islet preparations. After 3 months of treatment, average β cell volume was ~0.1 mm³ in the vehicle and exendin-4 groups, 0.36 mm³ in the harmine group, and 0.75 mm³ (four- and sevenfold) in the H+E combination group (Fig. 2, A to C, and movies S10 and S11). Human β cell proliferation was two- to threefold higher in the harmine alone and H+E groups than in the vehicle and exendin-4 groups (Fig. 2D), whereas β cell death was essentially absent in all groups (Fig. 2E). β cell size was comparable among all four groups (Fig. 2F). Steady-state circulating exendin-4 and harmine concentrations in these mice were 4 ng/ml (~1 nM) and 4 to 5 ng/ml (~16 to 20 nM), respectively (Fig. 2, G and H).

Fig. 2. Human β cell mass after 3 months of harmine/compound 2.2c with or without exendin-4 in euglycemic mice.

(A) Study protocol. Mice were transplanted with 300 human IEQs per kidney, and 4-week minipumps were implanted and replaced monthly to provide 3 months of continuous infusion of vehicle (V) or H+E. At month 3, both kidneys were harvested and blood was obtained. (B) Static images of kidney grafts treated for 3 months. Inset contains an image of the created surface used for volume quantification. See movies S10 and S11. (C to H) Measurements in the treated mice described in (A) at 3 months. Each dot represents an individual islet graft from one human islet donor transplanted in a mouse. (C) Human β cell volume, (D) Ki67 labeling index, (E) TUNEL labeling quantification, (F) β cell size, (G) plasma harmine, and (H) plasma exendin-4 concentrations per group. (I) Three-month study pro-tocol with compound 2.2c. (J) Human β cell volume in the four groups of mice at 3 months. (K) Ki67 labeling index in human β cells in islet grafts from the four groups. (L) The 3-month treatment plus 1-month drug withdrawal study protocol. Four-week minipumps were implanted and replaced every month to provide 3 months of continuous infusion. After that, no more pumps were implanted, and at month 4, both kidneys were harvested for analysis. (M and N) Human β cell volume (M) and Ki67 labeling index (N) in the four groups of mice 1 month after drug withdrawal. (O and P) Plasma harmine (O) and exendin-4 (P) concentrations after 1 month of drug withdrawal. In all panels, bars indicate mean ± SEM. *P < 0.05 versus V; **P < 0.05 versus V and E; ^P < 0.05 versus V and H, using Dunn post hoc test [(C) to (E), (G), (H), (N), and (P)] and Tukey post hoc HSD [(F), (J), (K), (M), and (O)]. Scale bars, 500 μm.

These results were reproduced and confirmed using a second more potent and selective DYRK1A inhibitor, compound 2.2c (28). Compound 2.2c, at a dose threefold lower than harmine (1 mg/kg per day), combined with exendin-4 induced four- to fivefold increases in human β cell mass, results similar to those observed with H+E treatment (Fig. 2, I and J). Human β cell proliferation was three- to fourfold higher in the compound 2.2c alone and compound 2.2c–exendin-4 groups compared with the vehicle and exendin-4 groups (Fig. 2K).

To assess duration of these beneficial effects after drug withdrawal, we studied a separate cohort of human islet–transplanted mice that were treated for 3 months (as above) and then followed for an additional month. One month after drug termination, the four- and sevenfold increases in human β cell volume induced by harmine alone and H+E treatment, respectively, were maintained (Fig. 2, L and M). In contrast, human β cell proliferation declined to baseline values (Fig. 2N), and circulating harmine and exendin-4 concentrations in these mice were negligible or undetectable (Fig. 2, O and P).

We next assessed possible effects of human donor characteristics. We found no correlation between human islet donor age, body mass index (BMI), hemoglobin A1c (HbA1c), islet purity, or days in culture after islet isolation and human β cell mass induced by 3-month treatment of harmine/2.2c ± exendin-4 (fold increase over vehicle treatment) (fig. S2, A to E). However, there was a significant correlation (r = 0.597, P = 0.015) between the effects of harmine and 2.2c alone versus harmine/2.2c + exendin-4, indicating that islet preparations that responded to harmine/2.2c responded further to H+E (fig. S2F).

Together, these studies demonstrate that DYRK1A inhibitors together with GLP1RAs markedly increase human β cell mass in euglycemic mice transplanted with human islets. Of note, this increase was maintained for at least 1 month after treatment withdrawal.

Enhanced glucose tolerance and insulin secretion with harmine–exendin-4 treatment in euglycemic mouse recipients

We next assessed additional parameters among the four groups in Fig. 2A after 3 months of treatment. Body weight gain was not different among groups (Fig. 3A), but nonfasting blood glucose was lower in all three treatment groups (Fig. 3B). Plasma human insulin, but not plasma glucagon, was elevated in the H+E group (Fig. 3, C and D). Glycemic responses to intraperitoneal glucose administration were superior in the exendin-4 group and in the H+E group (Fig. 3E). Immunohistochemical evaluation of key β cell transcription factors PDX1, NKX6.1, and MAFA in the human islet grafts indicated that these were maintained and comparable in all four groups (Fig. 3F). GSIS was enhanced in human islet grafts from mice treated with H+E (Fig. 3, G and H). As additional evidence of normal differentiation and function, no animals in Fig. 2 or fig. S1 developed hypoglycemia despite their increase in human β cell mass. Insulin tolerance test (ITT) responses were similar in vehicle- and H+E-treated mice (Fig. 3, I to K), and C-peptide in plasma decreased similarly in both groups (Fig. 3L), suggesting that insulin secretion was down-regulated in response to hypoglycemia. Collectively, these findings demonstrate that it is possible to markedly increase human β cell mass while also retaining or improving function over the long term in vivo.

Fig. 3. Human β cell function after 3 months of treatment in euglycemic mice.

(A) Body weight gain, (B) random blood glucose, (C) plasma human insulin, and (D) plasma glucagon in vehicle (V), exendin-4 (E), harmine (H), and H+E-treated mice after 3 months of treatment. (E) Intraperitoneal glucose testing in the four groups. Inset shows area under the curve (AUC) per group. (F) Immunolabeling of PDX1, NKX6.1, and MAFA in islet grafts from the four groups. (G) GSIS measured by perifusion of islet grafts from mice transplanted with human islets (n = 4 human islet preparations) and treated with vehicle or H+E for 3 months. (H) AUCs for different steps of the perifusion experiments. (I) ITT in mice transplanted with human islets (n = 4 human islet preparations) and treated with vehicle or H+E for 3 months. (J) ITT displayed as percentage of blood glucose at time 0. (K) AUC of the ITT results shown in (J). (L) Human C-peptide in the mice in (I) at time 0 and 15 min after insulin injection. In all panels, bars indicate mean ± SEM. *P < 0.05 versus vehicle using one-way ANOVA with Holm multiple comparisons test for simultaneous comparisons with vehicle treatment (A to C), Dunn post hoc test (D), Tukey post hoc HSD (E), two-tailed Mann-Whitney test (H), and two-tailed Student’s t test (K and L). Scale bars, 10 μm.

Harmine and exendin-4 effects on the human α cell in vivo

Harmine and other DYRK1A inhibitors can activate rodent and human α cell proliferation in vitro (13–17). Therefore, we assessed human α cell volume and proliferation in the islet grafts from the mice transplanted with human islets shown in Fig. 2. Despite the increased human β cell mass in the renal islet grafts, there were no noticeable changes in human α cell mass in the same grafts with any treatment at any time point (fig. S3, A to D, and movies S12 to S17). Thus, β cell mass expansion was dissociated from α cell mass changes. Harmine and H+E treatment caused a small (0.2 to 0.3%) but significant (P = 0.045) increase in α cell proliferation at 3 months (fig. S3E) but not at 3 months with compound 2.2c–exendin-4 (fig. S3F) or 1 month after harmine–exendin-4 withdrawal (fig. S3G). α cell death was negligible and was not different among groups (fig. S3H).

Safety profile of 3-month harmine–exendin-4 treatment in vivo

Because DYRK1A is ubiquitous and GLP1R expression is not exclusive to the β cell (21, 29), we explored cell proliferation and tissue histology in the same human islet–transplanted Rag1^−/− mice treated for 3 months with harmine, exendin-4, or H+E. Tissues examined included liver, spleen, intestine, adipose tissue, lung, heart, kidney, exocrine and endocrine pancreas, and brain (figs. S4 to S6). Whereas proliferation increased as expected in endogenous mouse β cells in H+E-treated mice (fig. S4E), no evidence of abnormal proliferation or tissue histology was observed in any other organ with any treatment except with exendin-4, which induced a significant increase in lung cell proliferation (figs. S4 and S5). We did not observe abnormal posturing, feeding, or grooming behaviors. Equal numbers of mice displayed innate immune cell abnormalities in the lung, presumably reflecting their immunocompromised state, but there were no histopathologic differences in any tissue among the groups (fig. S6).

Extended harmine–exendin-4 treatment increases human β cell mass and normalizes glucose homeostasis in diabetes

Studies thus far were performed on euglycemic mice. To assess the effect of continuous H+E treatment on human β cell mass and glucose homeostasis in a conventional model of diabetes, we performed longer-term, 3-month experiments in streptozotocin (STZ)–induced diabetic immunodeficient Rag1^−/− mice transplanted with a marginal or therapeutically inadequate number of human islets (Fig. 4A). Diabetic mice transplanted with human islets and treated with H+E combination displayed a rapid, considerable, and sustained normalization of blood glucose concentrations that was maintained for the entire 3 months of treatment and that was superior to the vehicle and single-treatment groups (Fig. 4, B and C). This associated with a significant (P = 0.012) increase in plasma human insulin in the H+E group (Fig. 4D). Glucose tolerance at 3 months was also improved in the H+E-treated mice compared with the vehicle-treated group (Fig. 4, E and F). Notably, during repeated blood glucose monitoring, no animal developed hypoglycemia at any time point. This is relevant because transplanting human pancreatic islets into diabetic mice has been reported to transfer the glycemic set point of the donor species islets to the recipient mouse, independently of the islet mass in the graft (30). Moreover, human islet grafts sense glucose concentrations and adjust their insulin secretion accordingly, an effect attributed to adjacent α cells (31). Together, these data indicate that H+E treatment provides sustained improvement in glucose homeostasis in a diabetic model of insulin insufficiency.

Fig. 4. Effects of 3 months of harmine, exendin-4, and combined treatment on human β cell mass, function, and glucose homeostasis in STZ-induced diabetic mice.

(A) Three-month study protocol in diabetic mice. Mice were rendered diabetic by STZ injection, and then 200 to 250 human IEQs per kidney were transplanted under the kidney capsule. Four-week minipumps were implanted and replaced monthly to provide 3 months of continuous infusion. At month 3, kidneys were harvested and blood was obtained. (B) Blood glucose at the time points indicated, before and after islet transplantation and during the 3 months of treatment, and (C) the AUC for the blood glucose for the four groups. Each dot represents an individual islet graft from one human islet donor transplanted in a mouse. (D) Plasma human insulin concentrations per group at the end of 3 months. (E) Intraperitoneal glucose testing in the four groups and (F) AUC for each group. (G and H) Human β cell (G) and α cell (H) volume per group at 3 months. In all panels, bars indicate mean ± SEM. *P < 0.05 versus V, **P < 0.05 versus V and E, using Dunn post hoc test (C and G) and Tukey post hoc HSD [(D), (F), and (H)]. Scale bars, 10 μm.

We next assessed human β cell mass in the β cell–deficient STZ model. Human β cell volume was significantly higher (~700%, 0.08 mm³ versus 0.65 mm³, P < 0.05) in the H+E group than in the other groups (Fig. 4G) and comparable to the increase observed in the euglycemic group described earlier. Additionally, human α cell volume did not change in the STZ-diabetic model over 3 months, despite the marked increase in β cell volume (Fig. 4, G and H), similar to our previous results in the euglycemic model. Together, these data indicate that H+E therapy expands human β cell mass and normalizes blood glucose in diabetic mice without altering α cell mass.

Harmine–exendin-4 combination treatment increases human β cell mass and survival at early stages after transplantation

β cell transplantation in mouse and humans is associated with substantial β cell death in the first days after transplant (32). We wondered whether, and to what extent, harmine alone or the H+E combination group might contribute to β cell survival in the peri-transplant period and whether this might affect the overall increase in β cell mass at 3 months. To address this question, we transplanted human islets in four groups of animals, as in previous studies, but delayed drug treatment for 4 weeks, to allow equal human islet engraftment and vascularization in all groups, so that they would begin treatment with the same initial β cell mass (Fig. 5A). We then repeated the 3-month experiments and analyzed β cell volume in human islet grafts. After 3 months of treatment, average β cell volume was ~0.1 mm³ in the vehicle and exendin-4 groups, 0.21 mm³ in the harmine group, and 0.39 mm³ in the harmine–exendin-4 group (Fig. 5, B and C). These results support two important interpretations. First, an early human β cell prosurvival effect and enhanced β cell mass induced by the combination likely play an important role in the ultimate increase in β cell mass observed at 3 months of treatment, because the same treatment initiated 1 month after transplant leads to a more moderate β cell mass expansion (0.75 versus 0.39 mm³); second, the three- to fourfold increase in β cell mass in this context (Fig. 5, B and C) correlated with the three- to fourfold increase in human β cell proliferation observed with H+E compared with vehicle-treated mice. These observations suggest that both β cell proliferation and enhanced survival may contribute to the overall increase in β cell mass observed with H+E.

Fig. 5. Effects of harmine, exendin-4, and the combination on human β cells when treatment is started 4 weeks after transplant or when treatment is only for 1 week.

(A) Protocol for a 4-week delay after transplant with 300 human IEQs per kidney. Islets were transplanted, and 4 weeks later, minipumps were implanted and replaced monthly to provide 3 months of continuous infusion, after which kidneys were harvested for iDISCO⁺. (B and C) Human β cell volume per group at 3 months (B) and summary of these changes (C). Values at time 0 and 28 days are from Fig. 1H and fig. S1B (vehicle-treated mice), respectively, and serve as a reference point for the changes induced by 3 months of treatment. (D) The 1-week study protocol. Islets were transplanted, minipumps were implanted, and 1 week later both kidneys were harvested for immunohistochemistry and iDISCO⁺. (E) Human β cell volume per group at 1 week of treatment. Each dot represents an individual islet graft from one human islet donor transplanted in a mouse. (F) Examples of Ki67 immunolabeling in V and H + E groups. Arrows indicate Ki67⁺ insulin⁺ cells. Ki67 labeling indices in human β cells in islet grafts are also shown. (G) Examples of labeling for TUNEL in human β cells from grafts in mice treated with V or H+E. Arrows indicate TU-NEL⁺ insulin⁺ cells. Quantification of TUNEL labeling in β cells of islet grafts from the four groups. (H) Plasma harmine and (I) exendin-4 concentrations per group at 1 week of treatment. (J) Changes in human β cell volume at all time points in the four treatment groups. In all panels, bars indicate mean ± SEM. *P < 0.05 versus V; **P < 0.05 versus V and E; and ^P < 0.05 versus V and H using Dunn post hoc test (B and F) and Tukey post hoc HSD [(E) and (G) to (I)]. Scale bars, 10 μm.

To directly query whether harmine or the H+E combination might enhance human β cell survival in the early posttransplant period, we transplanted human islets into four groups of animals as previously, but treated the mice only for 1 week (Fig. 5D). At 1 week of treatment, human β cell volume was 50% higher in the H+E group than in the vehicle group (Fig. 5E), associated with an increase in β cell proliferation (Ki67) and a corresponding reduction in β cell labeling for TUNEL, a marker of cell death (Fig. 5, F and G). Steady-state circulating harmine and exendin-4 concentrations in these mice were approximately 3 to 4 ng/ml (~12 to 16 nM) and 4 ng/ml (~1 nM), respectively (Fig. 5, H and I), similar to those in the 3-month study. Together, these observations suggest that the long-term increase in β cell volume induced by H+E reflects components of both enhanced β cell proliferation and survival.

The collective results over time are summarized in Fig. 5J, which illustrates several points. First, β cell mass declined in the first week after transplant in all groups, as expected because of inadequate vascularization with resultant β cell ischemia (32). However, this initial decline was less in the harmine–exendin-4 group, reflecting increased β cell survival and proliferation at this early time point. β cell mass then progressively increased over the ensuing 3 months in the harmine and H+E groups, at which point β cell mass was four- and sevenfold greater than at the time of initial transplant, respectively, as compared with the vehicle and exendin-4 groups. This provides evidence that it is possible to markedly increase human β cell mass in vivo.

Harmine–exendin-4 combination enhances human β cell survival in part via VGF up-regulation

To identify potential mechanisms of the enhanced human β cell survival induced by harmine–exendin-4, we performed RNA sequencing (RNA-seq) analysis on human islet grafts after 1 week of treatment (Fig. 6A). Differential gene expression analysis revealed that 20 human genes were increased by at least 1.5-fold, and nine were decreased by at least 0.67-fold in the islet grafts of the H+E-treated mice (Fig. 6B). Gene set enrichment analysis identified “cell secretion” and “cell adhesion” pathways as the top biological processes (Fig. 6C). Increased expression of VGF, PECAM1, and CD93 in human islet grafts of the H+E-treated mice was confirmed using quantitative polymerase chain reaction (qPCR) (Fig. 6D). Among these, VGF, a well-known regulator of β cell function and survival (25–27), was the most abundant mRNA and was further up-regulated by H+E (fig. S7A). Treatment of human islets in vitro with the individual or combination treatments confirmed that H+E treatment increased VGF mRNA expression and induced secretion of both the VGF prohormone and VGF-derived peptide TLQP-21 (Fig. 6, E to G).

Fig. 6. VGF mediates the human β cell protection induced by harmine–exendin-4 in vitro.

(A) The 1-week study protocol for posttransplantation human islet transcriptomics analysis. Mice were transplanted with 500 human IEQs per kidney, minipumps were implanted, and 1 week later the human islet grafts in both kidneys were harvested for RNA isolation and sequencing. (B) Volcano plot with human genes that were significantly (P < 0.05) changed (>1.5- and <0.67-fold) in H+E-treated versus V-treated mice marked in red. (C) Gene ontology analysis shows changes in cell secretion and cell adhesion pathways. P values are shown next to bars. (D) qPCR of representative human genes from (B) significantly up-regulated in human islet grafts from H+E-treated mice. Data are means ± SEM; n = 3 mice per treatment. (E) Human VGF expression in human islets treated in culture with vehicle, 10 nM exendin-4, 10 μM harmine, or H+E for 24 hours. Bars are means ± SEM; n = 3 different human islet preparations. (F) VGF and (G) TLQP-21 concentrations in the conditioned medium of the human islets from (E). (H) Human VGF expression in human islets treated in culture with vehicle or H+E for 24 hours as in (E). Before treatment, islet cells were dispersed and infected with adenoviruses carrying scrambled shRNA (control) or human VGF shRNA for 24 hours. Bars are means ± SEM; n = 3 different human islet preparations. (I) Representative images of human cells stained with insulin and TUNEL. Human islet cells were dispersed, transduced with the indicated adenovirus, and cultured with 500 nM thapsigargin with or without H+E for 24 hours. Arrows indicate TUNEL⁺ insulin⁺ cells. (J) Quantification of TUNEL labeling in β cells of human islets treated as in (I). Bars indicate means ± SEM; n = 3 or 4 different human islet preparations. *P < 0.05 versus V or as indicated in the figure, **P < 0.05 versus V and E using Tukey post hoc HSD [(D), (E), (H), and (J)] or two-tailed Student’s t test (F and G). Scale bars, 10 μm.

Transplanted islet grafts experience hypoxia for several days after transplantation before engraftment and vascularization (32). Hypoxia causes endoplasmic reticulum (ER) stress in β cells, and human islet grafts transplanted into mice exhibit signs of ER stress and eventual β cell apoptosis because of induction of the proapoptotic unfolded protein response (33–35). Therefore, we next explored the effects of harmine, exendin-4, and H+E in the ER stress–induced human β cell death using thapsigargin. The H+E combination, but neither agent alone, reduced human β cell death (fig. S7B). Silencing VGF expression with adenovirus-VGF-shRNA (short hairpin RNA) (Fig. 6H) completely blocked the protective effect of H+E in human β cells in vitro (Fig. 6, I and J) and in vivo in human islet grafts 1 week after transplantation (Fig. 7, A and B). VGF expression downregulation also blunted the human β cell mass increase induced by H+E at 1 week (Fig. 7C), without altering human β cell proliferation (Fig. 7D). Together, these studies suggest that VGF appears to be one of several potential mediators of the protective effects of the H+E combination on human β cells.

Fig. 7. VGF helps mediate the protection and expansion of human β cells induced by harmine–exendin-4 in vivo.

(A) The 1-week study protocol for adenovirus-transduced human pseudoislets. Human islet cells are dispersed and transduced with adenovirus scrambled shRNA or adenovirus VGF shRNA, and pseudoislets were generated. Pseudoislets (300 human IEQs per kidney) were transplanted, minipumps with vehicle or harmine–exendin-4 were implanted, and 1 week later the human pseudoislet grafts in both kidneys were harvested for immunohistochemical and iDISCO⁺ analysis. (B) Quantification of TUNEL labeling in β cells of human pseudoislet grafts. (C) Human β cell volume in n = 5 or 6 mice per group at 1 week. (D) Ki67 labeling index in β cells of human pseudoislet grafts. Each dot represents an individual islet graft from one human islet donor transplanted in a mouse. In all panels, bars indicate mean ± SEM. *P < 0.05 versus their corresponding control shRNA; **P < 0.05 versus control shRNA and ^P < 0.05 versus control shRNA + H+E, using Tukey post hoc HSD (B and C) and two-tailed Mann-Whitney test (D).

Harmine–exendin-4 combination enhances islet graft vascularization

RNA-seq analysis indicated that endothelial genes PECAM1, ESM1, and VWF were up-regulated in islet grafts in mice treated with the drug combination for 1 week (Fig. 6B). This suggests that human islet graft vascularization might be increased after H+E treatment. To address this question, we performed PECAM1/CD31 immunostaining in human islet grafts from the treated mice. PE-CAM1/CD31 staining showed an enhanced vascular network in human islet grafts from mice treated for 1 month with H+E compared with vehicle- and single drug–treated mice (Fig. 8A). Vessel network analyses of human islet grafts showed that total graft vascularization volume, number of branch points, and vessel volume per islet area, but not segment length and vessel diameter (36, 37), were increased in H+E combination–treated mice compared with vehicle-treated mice at 1 month after transplantation (Fig. 8, B to F). Similar results were observed in H+E-treated mice at 1 week after transplantation, in this case with increases in total graft vascularization volume, segment length, and vessel volume per islet area (fig. S8).

Fig. 8. Effects of 1 month of harmine, exendin-4, and combined treatment on human islet graft vascularization.

(A) Representative images of human islet grafts from mice treated for 1 month with V, H, E, or H+E and stained for PECAM1/CD31 (red), insulin (green), and DAPI (4′,6-diamidino-2-phenylindole) (blue). Vessel network analysis of human islet grafts in these mice showing (B) total vascular graft volume, (C) mean segment vessel length in islets, (D) normalized number of branch points per islet volume, (E) vessel diameter, and (F) islet vessel volume. Each dot represents an individual islet graft from one human islet donor transplanted in one mouse. In all panels, bars indicate mean ± SEM. *P < 0.05 versus V, ^#P < 0.05 versus V, E, and H using Tukey post hoc HSD. Scale bars, 20 μm.

DISCUSSION

We demonstrate that DYRK1A inhibitors in combination with a canonical and widely used GLP1RA increased human β cell mass, function, and glycemic control in vivo. These findings apply to human β cells transplanted into both euglycemic mice and mice with severe diabetes. These observations may enable pharmacologic approaches that are scalable to millions of people with diabetes. These studies raise several important questions. For example, how much β cell regeneration and mass expansion are enough to treat T1D and T2D? Answering this can only be addressed through human trials in T1D and T2D evaluating C-peptide secretion, insulin reserve, and glycemic control. However, a four- to sevenfold increase in human β cell mass over 3 months, with normalization of glycemic control and in the absence of hypoglycemia as shown here in diabetic conditions, is promising. Specifically, to reverse the 50 to 60% reduction in β cell mass in people with T2D (3, 4), a four- to sevenfold increase in 3 months would seem more than sufficient, and the accompanying improvement in β cell survival and function should be particularly beneficial. T1D provides a greater challenge because baseline β cell mass is lower in established T1D than in T2D (1, 2), but a four- to sevenfold increase in 3 months should improve glycemic control and reduce “fragility,” and, of course, longer-term treatment may be feasible if needed. β cell autoimmunity will remain a challenge in T1D, but recent advances in the control of autoimmunity in T1D provide reason for optimism (38–40).

Because DYRK1A is ubiquitously expressed, one might worry about undesired off-target effects in tissues outside β cells. Safety and efficacy data at 1 week (15) and at 3 months (described herein) are encouraging. Safety is further supported by the low human β cell proliferation rates in vivo (~0.3 to 0.7% Ki67) compared with those in vitro (2 to 8%) (13–20) and by the observation that in vivo β cell mass increases are sustained but proliferation declines to baseline when treatment is withdrawn. Of course, safety and duration of effect will need to be assessed over even longer treatment periods and in large animal preclinical studies before human safety and efficacy trials can be performed. Of note, many GLP1RAs and dipeptidyl peptidase IV (DPP-IV) inhibitors that elevate endogenous GLP1 are currently in use in millions of people with diabetes (21, 41); adding an orally administered DYRK1A inhibitor for a matter of months in this setting to restore β cell mass could potentially be a simple, inexpensive, and highly scalable approach to diabetes treatment. The breadth of GLP1RAs and DPP-IV inhibitors available (21, 41) and the flexibility in DYRK1A inhibitor dosing (15, 28) are promising. Some would argue that β cell–specific targeting may be required for DYRK1A inhibitors, but the existing 1-week (15) and 3-month safety data described here suggest that such targeting may be unnecessary.

The synergistic effects with GLP1RAs allow the use of lower doses of harmine (15). Here, we selected a dose of harmine that increases β cell proliferation in vivo but does not increase α cell proliferation, with the goal of expanding human β cells without altering α cell mass. Cotreatment with low doses of harmine and exendin-4 led to a substantial increase in human β cell proliferation and mass expansion without α cell proliferation at 1 or 4 weeks of treatment and only modest α cell proliferation at 3 months, resulting in no alteration in α cell mass or circulating glucagon concentrations during 3 months of treatment. We attribute this β cell selectivity to the low dose of harmine used and the low abundance or absence of GLP1Rs on α cells (29). Notably, the beneficial effects of harmine were reproduced by a second more selective and potent DYRK1A inhibitor, compound 2.2c (28), which at a threefold lower dose than harmine produced four- to fivefold increases in human β cell mass, again without changes in α cell proliferation or mass. Of note, the harmine-induced increase in human β cell mass persisted for at least a month after drug withdrawal, when β cell proliferation and circulating concentrations of the drugs were decreased to basal values.

The mean four- to sevenfold increase in human β cell mass over 3 months is unexpected in the face of β cell proliferation rates or labeling indices that, although consistently higher than in controls, are nonetheless low, below 1.0%. Although this bodes well from a safety standpoint, it raises a question regarding the mechanism(s) responsible for the robust increase in human β cell mass. Notably, in addition to β cell proliferation, we found that β cell survival in the immediate posttransplant period was enhanced by combination treatment, as evidenced by the lower rates of TUNEL labeling in the drug-treated groups, the greater β cell mass at 1 week in the harmine–exendin-4 group, and the smaller (three- to fourfold) increase in β cell mass when transplanted islets were allowed to engraft and vascularize for 1 month before initiation of 3 months’ treatment. In addition, the H+E combination provided enhanced protection to human β cells in vitro against the ER stress inducer thapsigargin. The in vivo and in vitro protective effect of H+E appears to be mediated, at least in part, by VGF, an islet prohormone that potentiates GSIS and provides protection against β cell cytotoxicity (26, 27). Furthermore, the VGF-derived peptide TLQP-21 has been shown to preserve islet mass and to slow diabetes onset in prediabetic ZDF rats (26). Our results also highlight the importance of VGF in β cell survival in the hypoxic and nutrient-deprived environment of the early posttransplant period. This observation places VGF and its derived peptides as potential agents to enhance islet engraftment and pancreatic β cell survival in diabetes. Three cell surface binding targets have been proposed for TLQP-21: C3aR1, gC1qR, and HSPA8 (42–44). Future studies are needed to define whether these receptors or others mediate TLQP-21 effects in β cells. Last, in addition to the enhanced survival induced by H+E, we observed an up-regulation of endothelial genes (PECAM1, VWF, and ESM1) in human islet grafts at the early 1-week time point after transplantation, hinting at enhanced islet graft vascularization. Immunostaining for PECAM1/CD31 revealed enhanced islet graft vascularization parameters. It is thus likely that the increase in β cell survival observed in the immediate posttransplant period may, in part, reflect enhanced/accelerated islet vascularization induced by this treatment combination. Whether VGF plays a role in enhanced islet graft vascularization warrants further study.

Also of note, accurate quantification of human β cell mass in animal transplant models has been tedious and poorly reproducible. Thus, the marriage of iDISCO⁺ tissue clearing with 3D imaging—widely used in neuroscience (45–47)—to equally widely used models of human islet transplant could allow accurate in vivo β cell mass quantification to be widely performed.

This study has limitations. First, there is a large variation in the β cell mass expansion response of human islet grafts to H+E. This could reflect the inherent variable response of human islets observed previously with drug treatments in vitro (13–16). It is important to note that those human islet preparations with a larger response to harmine or 2.2c further responded to H/2.2c + E, suggesting perhaps variations in DYRK1A levels in different human islet preparations. In addition, we observe that human β cell volumes in islet grafts from untreated mice 2 weeks after transplantation are variable,which could reflect changes in the survival of different human islet preparations in the islet transplant setting. This variability in graft survival could provide different starting points for the enhancement in β cell mass induced by H+E. Furthermore, the large variation could also reflect the complexity of the tissue clarification process, 3D staining, imaging, and image analysis that might result in variability in the evaluation of human α and β cell mass in the grafts. Technical improvements as well as additional studies analyzing the effect of different DYRK1A inhibitors are warranted. Second, the marked expansion of β cell mass in the face of low β cell proliferation raises the possibility that harmine or harmine + exendin-4 may encourage human α-to-β cell transdifferentiation, as has been reported in rodent islets (48–50). Documenting this possibility will require genetic lineage tracing studies. Until such tools are available, a potential role for α-to-β transdifferentiation must remain speculative. Third, it remains uncertain whether β cell–specific targeting will be necessary and, if so, what the ideal targeting molecule might be. Our preclinical histopathology and proliferation studies suggest that targeting may not be necessary, but longer-term studies in higher preclinical models will be required to clarify this issue. Fourth, the cellular mechanisms through which the many VGF family peptides potentiate harmine-driven human β cell mass expansion (25) remain to be defined. Fifth, development of potent and more selective DYRK1A inhibitors, exemplified by compound 2.2c (28, 51), should enhance their therapeutic efficacy. Last, whether the substantial increase in β cell mass induced by H+E in human islets transplanted in immunodeficient mice will occur in humans treated with the combination is unknown at this point. Differences in the human islet microenvironment in the islet graft compared with the intact human pancreas, the absence of a complete immune system in immunodeficient mice, and the obvious difference in species could affect the response to H+E. Future clinical trials will define the potential of this combination therapy for β cell regeneration in humans.

In summary, we demonstrate that the combination of a DYRK1A inhibitor with a GLP1RA markedly enhances human β cell mass in vivo over a period of 1 to 3 months, which appears to be sustained after drug withdrawal for at least 1 month. This therapy appears to be somewhat β cell specific, which could potentially avoid undesirable adverse effects on non–β cell tissues and cell types. This therapy also maintained β cell differentiation, survival, and function in vivo, both in a euglycemic model and in the conventional β cell depletion model of diabetes. The prosurvival effects of this treatment combination may be mediated at least in part by the islet prohormone VGF, a well-known regulator of β cell function and survival. Further refinements in dosing, duration of treatment, and deeper exploration of safety and potential toxicities are required for human translation. Last, iDISCO⁺/3D imaging technology may be a potentially effective tool for assessing human β cell mass.

MATERIALS AND METHODS

Study design

Having demonstrated that DYRK1A inhibitors together with the GLP1RAs induce human β cell proliferation ex vivo (15), we aimed at testing whether this increase in human β cell proliferation translated into an increase in human β cell mass in vivo using human islets transplanted into immunodeficient mice. We first optimized the iDISCO protocol for mouse kidneys bearing human islet grafts to analyze human islet cell volume in three dimensions. For this purpose, we transplanted human islets (0 to 1000 IEQs) under the kidney capsules of immunodeficient Rag1^−/− mice and analyzed α and β cell volumes at 0 and 15 min after transplantation in human islet grafts using iDISCO⁺ tissue clearing, immunolabeling for insulin and glucagon, light sheet microscopy, and 3D image analysis to assess islet mass. During our development of this quantification approach, we initially performed these experiments in n = 3 to 6 human islet preparations (one per mouse) per human islet “dose.” These studies revealed strong correlations between measured β cell and α cell volumes in the grafts and in the transplanted islets, assuring accuracy of the method for subsequent treatment studies.

We then transplanted human pancreatic islets into both right and left kidney capsules of immunodeficient Rag1^−/− mice, followed by subcutaneous implantation of continuously infusing osmotic minipumps that delivered vehicle, a DYRK1A inhibitor (harmine or compound 2.2c), exendin-4, or the H+E combination, for 7, 28, or 84 days, changing the pump every 28 days for the latter time point, followed by in vivo perfusion and tissue fixation. One kidney was harvested for human α and β volume analysis and the other for immunohistochemical analysis of islet graft cellular parameters such as proliferation, survival, size, and vascularization. These experiments were done in a minimum of n = 4 human islet preparations (mice) per treatment per time point. All outliers were included and are visualized in scatterplot versions of the figures. Analysis of human islet grafts, qPCR, mouse histopathology, and other assays were performed in a double-blinded fashion. Investigators who assessed, measured, or quantified the results were blinded to the intervention. All mice were maintained in a climate-controlled vivarium on a 12:12-hour light/dark cycle. Four mice were housed per cage with ad libitum access to food and water. All procedures were performed with the approval of and in accordance with guidelines established by the Icahn School of Medicine at Mount Sinai Institutional Animal Care and Use Committee (IACUC #2015–0107). Detailed methods are available in the Supplementary Materials.

Statistical analysis

Data presented as linear graphs, bar graphs, and scatterplots show means ± SEM. Statistical significance was set at P < 0.05. Normally distributed data by Shapiro-Wilk log-normality test were evaluated using a two-tailed unpaired Student’s t test or by two-tailed probability values of Pearson’s correlation coefficients for comparison between two groups (GraphPad and https://www.danielsoper.com/statcalc/default.aspx). Data with more than two groups were evaluated by one-way analysis of variance (ANOVA) with Tukey’s or Holm’s post hoc highly significant difference (HSD) (GraphPad and http://astatsa.com). Nonparametric tests (two-tailed Mann-Whitney test between two groups or Kruskal-Wallis with Dunn’s post hoc test for more than two groups) were used for the data that did not pass the normality test. Figure legends include the individual statistical test used, number of samples, and P values.

Data and materials availability:

All data associated with this study are present in the paper or the Supplementary Materials. Data from figures are available in data file S1. All raw RNA-seq reads for the RNA-seq data generated from engrafted human islet tissues are compiled and available from NCBI SRA PRJNA1035045 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE249310) under accession GSE249310.