Nutrient Excess Stimulates β-Cell Neogenesis in Zebrafish

Adult zebrafish were raised in Aquatic Habitats systems on a 14-/10-h light/dark cycle. Embryos were collected and raised in 0.3× Danieau solution at 28.5°C in an incubator with lights on a 14-/10-h light/dark cycle. The transgenic lines used for this study were Tg(−1.2ins:H2BmCherry), Tg(−1.2ins:tagRFP), Tg(−1.2ins:EGFP) (23), Tg(−5.1mnx1:tagRFP), Tg(−3.0hb9:GFP) (23), Tg(sst2:memRFP) (23), Tg(nkx2.2:mEGFP)^VU17 (23), Tg(gcga:EGFP), and Tg(ptf1a:EGFP) (23). We also generated a transgenic line with three elements separated by a 2A peptide, a membrane targeted TagRFP, nuclear mCherry, and nitroreductase and referred to as (−1.2ins:mR-nC-NTR).

All feeding experiments used 0.3× Danieau solution as the medium. For Hatchfry encapsulon (Argent Laboratories) feeding, a small amount was placed on the surface of the solution. For egg yolk feeding, chicken eggs were obtained from local grocery stores, the yolk separated, and diluted to 5% by volume. The solution was shaken vigorously to ensure suspension. For purified amino acid treatment, L-amino acids were obtained (Sigma-Aldrich), dissolved in 1 M HCl at 50 mg/mL, diluted to appropriate concentration based on data from the U.S. Department of Agriculture nutrient database, and the pH adjusted to 7.4. Intralipid (Sigma-Aldrich) was diluted to 1.33% by volume. Rapamycin (Sigma-Aldrich) was dissolved in DMSO at 10 mmol/L and used at a working concentration of 2.5 µmol/L. To ablate β-cells, transgenic larvae were incubated in 10 mmol/L metronidazole (Sigma-Aldrich) for 24 h. After the appropriate exposure interval, larvae were killed in 3-amino-benzoic acid ethyl ester (Sigma-Aldrich) and fixed in 4% paraformaldehyde overnight at 4°C.

After fixation, larvae were washed with 1× PBS plus 0.1% Tween-20 (PBST) and flat mounted in Aqua-Mount (Richard-Allan Scientific) with larvae oriented with their right side facing the coverslip. The larvae were flattened sufficiently to disrupt the islet slightly but allow better resolution of individual nuclei. All slides were coded to avoid bias. The β-cell number was determined by manually counting positive nuclei using a Zeiss AxioVert 100 under a 20× DIC lens or using a Zeiss AxioImager under a 40× lens (Carl Zeiss).

Where necessary, after fixation, larvae were dehydrated in methanol and stored at −20°C. Larvae were rehydrated, permeabilized in acetone for 30 min at −20°C, and washed in PBST. Nonspecific binding was blocked using PBST plus 1.0% DMSO, 1% BSA, and 5% normal goat serum. Primary and secondary antibodies were diluted in PBST plus 1.0% DMSO, 1% BSA, and 2% normal goat serum. Antibodies and dilutions used were insulin (A0564; DakoCytomation), 1:1000, green fluorescent protein (GFP; ab6556; Abcam), 1:500, and appropriate Alexa Fluor secondary antibodies (Invitrogen), 1:2500. For proliferation analysis using the Click-iT EdU Alexa Fluor 488 Imaging Kit (C10337; Invitrogen), 2 nL of 100 μmol/L 5-ethynyl-2-deoxyuridine (EdU) was injected into the pericardial region of larvae. After recovery, larvae were treated for 8 h with appropriate stimuli, fixed in 4% paraformaldehyde, and staining performed (according to Ref. 23). All images were collected using a Zeiss LSM510 or Zeiss LSM710 (Carl Zeiss).

After fixation, samples were washed in PBS and stained with filtered 0.3% Oil Red O in 60% 2-propanol for 2 h. Larvae were washed in PBS before imaging.

The Click-iT TUNEL imaging assay kit (Invitrogen) was used with the following modifications. Larvae were permeabilized with 20 μg/mL proteinase K in PBST for 45 min followed by refixation in 4% paraformaldehyde with 0.2% Tween-20. Larvae were then treated with 1× PBS with 1.0% DMSO and 0.5% Triton X-100. Manufacturer instructions were followed for the terminal deoxynucleotidyltransferase-labeling reaction and detection.

After 5 dpf, zebrafish have nearly exhausted the maternally provided nutrient store and begin feeding. At this time period, the second wave of endocrine differentiation from the ventral pancreas bud has begun (22,28,30). To facilitate quantification of β-cell number, we developed a transgenic line using a 1.2-kb fragment of the insulin promoter/5′−untranslated region to drive expression of a nuclear mCherry protein in the β-cells (Tg:-1.2ins:H2B-mCherry). We confirmed that the mCherry expression in the transgenic line labels all of the insulin-expressing cells (Fig. 3C and D). At 5 dpf, we found 32.7 ± 5.4 (mean ± SD) β-cells in the principal islet of unfed larvae (Fig. 1A), consistent with other reports (22). Larvae were fed ad libitum hatchfry encapsulon, a zebrafish larval diet, which consists of 50% protein and 12% lipid, or incubated in lipid-rich 5% chicken egg yolk solution, which consists of 26.5% lipid and 15.8% protein. When fed for 8 h, the number of β-cells increased to 36.7 ± 6.3 with the larval diet (Fig. 1C) and 43.1 ± 7.1 (Fig. 1B and C) with egg yolk. This was specific to the β-cells because the number of α-cells was similar in unfed larvae (17.9 ± 4.6 cells; n = 54) and egg yolk–fed larvae (18.7 ± 3.9 cells; n = 56) as determined using the Tg(gcga:EGFP) transgenic line (Fig. 4A and B). We also observed that the size of the β-cells is significantly larger in the egg yolk–fed larvae using the (Tg:-1.2ins:tagRFP) transgenic line, in which tagRFP expression is throughout the cell (data not shown). However, there is not a difference in the overall growth of unfed larvae or those fed for 8 h with the larval diet or egg yolk solution. Fish at 21 dpf had a similar β-cell response to 8-h incubation in egg yolk (Fig. 1E). A detailed time-course analysis of the response to egg yolk (Fig. 1F) revealed that the increase in β-cell number was evident after 4 h of feeding (ANOVA, P < 0.001, Tukey honest significant difference [HSD]), and the full effect was seen after 6 h of feeding. After incubation, larvae were found to have substantial egg yolk in the digestive tract (Fig. 3B and F), increased lipid in the liver (Fig. 3G), and chylomicrons in blood vessels (Fig. 3H). To determine whether the addition of β-cells is induced simply by the introduction of nutrients, larvae were treated with three rounds of 1-h incubation in the egg yolk solution followed by 2 h in yolk-free media. At the end of 9 h, the number of β-cells was not different from unfed larvae (Fig. 1D), suggesting that persistent feeding during the 8-h exposure period is required for the response. These data support that β-cell expansion in zebrafish larvae is induced by overfeeding.

We next investigated whether the β-cell expansion observed in larvae fed with egg yolk was due to a particular macronutrient type or growth factors and other components that may be present in the egg yolk. There are three major macronutrient classes in an egg yolk: amino acids, lipids, and glucose at a low concentration (1.5 mmol/L). To mimic these nutrient classes, we used defined solutions of purified amino acids, intralipid, and glucose at equivalent concentrations to that found in egg yolk based on U.S. Department of Agriculture data. Larvae were incubated for 8 h either with each individual nutrient class or in combinations (Fig. 1G). There was no change in the number of β-cells with the individual types of nutrients. The combination of amino acids and glucose together also had no significant effect. Intralipid in combination with amino acids or glucose significantly increased the number of β-cells over that of controls (ANOVA, P < 0.001, Tukey HSD), albeit not to the extent induced by egg yolk. However, all three nutrient classes together increased the number of β-cells to a similar degree as the egg yolk. The data strongly suggest that the β-cell addition induced by chicken egg yolk is due to a combination of the three macronutrient classes. The data further demonstrate that β-cell expansion is due to nutrient load rather than stimulation by growth factors in the chicken egg yolk.

The central role of pancreatic β-cells is to regulate blood glucose levels, and both adult and larval zebrafish have been shown to be glucose-responsive (26,31). To determine if glucose stimulates an increase in β-cells, larvae were incubated in a solution of 20 mmol/L glucose. After 8 h of exposure, the number of β-cells increased to 38.4 ± 5.7 (ANOVA, P < 0.001, Tukey HSD) (Fig. 1C). Although a significant increase over the unfed control, the degree of increase is not as large as in the lipid-rich egg yolk. The glucose-treated larvae did not exhibit abnormalities in morphology or behavior, nor did they have an increase in apoptosis (data not shown). Overt glucotoxicity has also not been observed in rats when the blood glucose was kept >20 mmol/L for 4 days (32). Incubation of larvae in 20 mmol/L mannitol for 8 h did not change the number of β-cells (Fig. 1C), suggesting the induction of β-cells was not a result of changes in osmolarity. Similar to the results with the egg yolk, intermittent incubation in 20 mmol/L glucose was insufficient to increase the β-cell number (Fig. 1D). Taken together, continuous presence of nutrients is necessary to induce additional β-cells.

To discern the role of mTOR, a major nutrient sensor, in the overnutrition-induced β-cell expansion, we used rapamycin to inhibit TORC1. Zebrafish larvae are very amenable to this type of pharmacological approach. Using rapamycin in conjunction with egg yolk or 20 mmol/L glucose for 8 h, we found that mTOR activity was only necessary for glucose stimulated β-cell expansion (Fig. 2A). In larvae cultured in 20 mmol/L glucose, rapamycin treatment abolished the β-cell expansion (t test, P < 0.001). In contrast, rapamycin had no effect on the β-cell expansion induced by egg yolk (Fig. 2A). We did not observe a difference in apoptosis between the treatment groups (data not shown). These data suggest that different signaling pathways are activated to induce addition of β-cells depending on the nutrient context.

Autocrine/paracrine insulin/IGF-1 signaling has been implicated in postnatal β-cell expansion and high-fat diet-induced compensatory hyperplasia in mice (19). To determine if insulin/IGF-1 signaling in β-cells has a role in regulating compensatory β-cell expansion, we made use of a dominant-negative IGF-1R (dnIGF-1R) (33,34). In mice, dnIGF-1R forms hybrid receptors between insulin receptor and IGF-1R and interferes with signaling (35). We developed a transgenic zebrafish that express the dnIGF-1R in β-cells under the control of the insulin promoter. We confirmed that dnIGF-1R was expressed exclusively in the pancreatic β-cells (Fig. 2B and C). In unfed larvae, we did not observe any abnormalities in the number and morphology of the β-cells. However, when larvae are overfed egg yolk for 8 h, β-cell number was significantly reduced compared with controls (Fig. 2D) (t test, P < 0.001), but the response to 20 mmol/L glucose was not statistically different than control larvae. This suggests that, similar to the mouse, autocrine/paracrine insulin/IGF-1 signaling is necessary in regulating the compensatory response to high-fat diet.

We sought to determine the origin of the new β-cells induced by overnutrition. New β-cells could form through three distinct mechanisms: β-cell self-replication, transdifferentiation from other pancreatic cells, or neogenesis from precursor/progenitor cells. To determine whether the new β-cells result from β-cell replication, EdU was used to label replicating cells during the β-cell expansion period. In unfed larvae (n = 34), we observed many proliferating cells that are readily labeled with EdU. However, we rarely detected insulin-positive, EdU-positive β-cells (Fig. 3M). Strikingly, in larvae with overnutrition (n = 40), we observed no additional EdU-positive β-cells (Fig. 3N and O). EdU-positive cells are occasionally observed within the islet, but these cells do not colocalize with insulin. Similar results were obtained with 20 mmol/L glucose (Fig. 3N) or egg yolk (Fig. 3O) as the stimulus. The lack of EdU-positive β-cells suggests the new β-cells are not from self-replication but arise potentially through transdifferentiation from non–β-cells or neogenesis from postmitotic precursor cells.

We also did not observe an increase of apoptosis in the pancreas of larvae with overnutrition (Fig. 3L). Transferase-mediated dUTP nick-end labeling (TUNEL)-positive cells were rare in the pancreas of 6 dpf unfed larvae (Fig. 3K) and overfed larvae, although TUNEL-positive cells in the intestine could be identified in both. This suggests that the β-cell increase was not triggered by apoptosis in the pancreas or islet. Apoptotic cells throughout the larvae were more prevalent in unfed larvae (Fig. 3I) than larvae with overnutrition (Fig. 3J).

To determine the cell of origin of the new β-cells, we used an array of transgenic lines that highlight cell types implicated as a potential source of β-cells. Given the addition of β-cells is a rapid process, and fluorescent proteins are fairly stable in this time frame, we reasoned that new β-cells should retain the fluorescent marker present in the previous state. We first examined transdifferentiation from α-cells using Tg(gcga:EGFP) transgenic larvae, which labeled all α-cells (data not shown). Although we did observe endocrine cells that coexpress the markers for glucagon and insulin, there is no change in the number of these cells between unfed (2.3 ± 1.4; n = 18) (Fig. 4A) and larvae that were incubated for 8 h in egg yolk (1.6 ± 1.3; n = 22) (Fig. 4B). We did not observe transdifferentiation of acinar cells into β-cells using Tg(ptf1a:EGFP) (Fig. 4C and D), as no cells were found to coexpress the two markers. We also excluded transdifferentiation from δ-cells using Tg(sst2:memRFP) (Fig. 4E and F), as no cells that coexpress the two markers were identified in either unfed larvae or those overfed with egg yolk. The number of δ-cells was difficult to quantify due to morphology and the membrane localization of RFP. These data suggest that new β-cells are not derived from transdifferentiation of differentiated pancreatic endocrine or exocrine cells.

The newly differentiated β-cells arose via neogenesis from postmitotic endocrine precursor cells. Using Tg(−5mnx1:tagRFP), we found that overnutrition increased the number of TagRFP-positive β-cells from 17.1 ± 2.4 cells in the unfed larvae (n = 16) (Fig. 5A) to 23.3 ± 5.3 cells in larvae overfed egg yolk for 8 h (Fig. 5B) (n = 30) (t test, P < 0.001). Although highly expressed during early development (data not shown) (36,37), TagRFP was expressed only in a subset of the β-cells in the larval islet (Fig. 5A and B). The increase in TagRFP-positive cells likely highlights newly differentiated, immature β-cells. This was suggested by the observations that mnx1 (hb9) promoter–directed marker expression preceded insulin promoter–directed marker expression during the time course of egg yolk feeding and that mnx1 promoter–directed marker expression was absent in the secondary islets of adult fish (Supplementary Fig. 1). Furthermore, after nitroreductase-mediated β-cell ablation in Tg(−3.0hb9:GFP);Tg(−1.2ins:mR-nC-NTR) larvae, GFP-positive mCherry-negative cells appeared in the regenerating islets (Fig. 5D), whereas all GFP-expressing cells in untreated larvae are mCherry positive (Fig. 5C). These data strongly suggest that the increase of cells with mnx1 promoter activity was due to newly differentiated β-cells rather than reactivation of the promoter in mature β-cells. However, the increase in TagRFP-positive cells, an average of six cells, did not fully account for the 10 new β-cells induced by overnutrition. We also observed an increase in enhanced GFP (EGFP) expression both in and surrounding the islet in overfed Tg(nkx2.2:mEGFP)^VU17; Tg(ins:TagRFP) larvae (Fig. 5F) compared with unfed larvae (Fig. 5E). There were significantly more cells that coexpress both EGFP (nkx2.2) and tagRFP (insulin) in egg yolk–overfed (8.06 ± 2.93 cells; n = 16) larvae compared with unfed larvae (4.14 ± 1.51 cells; n = 14) (t test; P < 0.001). However, like the increase of TagRFP cells in Tg(−5.1mnx1:TagRFP), the four new EGFP-positive cells in Tg(nkx2.2:mEGFP)^VU17 do not account for all 10 of the new β-cells induced by overnutrition. This may indicate two separate pathways of β-cell differentiation. In mice, two parallel pathways of β-cell differentiation have been suggested: one Pax6-dependent and the other Pax4-Mnx1(Hb9)-dependent (38,39). We have found that the β-cells with TagRFP expression and those with expression of EGFP are distinct populations in Tg(−5mnx1:tagRFP);Tg(nkx2.2:mEGFP)^VU17 larvae in both unfed (Fig. 5G) and overnutrition (Fig. 5H) conditions. Because the nkx2.2 promoter is active in both ductal cells and endocrine precursor cells (40), these new EGFP-positive β-cells could be from transdifferentiation of differentiated pancreatic ductal epithelial cells, differentiation of endocrine precursor cells, or both. More focused lineage tracing will be required to distinguish these possibilities. Taken together, these data support that a large fraction of the new β-cells are likely from differentiation of postmitotic endocrine precursor cells.

The maintenance of an appropriate number of β-cells and adequate production of insulin is essential for glucose homeostasis. Although zebrafish have been used to study the development of the pancreas and β-cells (20,22,23) and regeneration of β-cells in adult animals (21), how β-cells respond to an increased metabolic demand has not been determined. We have shown in this study that a rapid increase in β-cells occurs upon overfeeding and suggests that the zebrafish β-cells are indeed able to sense and respond to an increase of metabolic demand by 5 dpf. We did not observe an increase with meal-style feeding, suggesting that the response requires a persistently elevated metabolic demand and is not simply a normal response to nutrients. This is consistent with the compensatory hyperplasia of β-cells that occurs in the face of nutrient excess and obesity. In humans and mice, this is an essential mechanism to prevent, or at least delay, the development of diabetes (9–11). Because childhood overnutrition and obesity predicts type 2 diabetes later in life, our model may potentially be used to understand the underlying mechanism (41). Given that zebrafish are an excellent platform for genetic and chemical screens (41,42), this system can be adapted to identify genes and pathways that are involved in sensing and responding to nutrient excess.

Type 2 diabetes results when β-cells can no longer supply sufficient insulin to maintain glucose homeostasis, which may be a consequence of β-cell apoptosis, β-cell failure, or defective β-cell hyperplasia (1,43). A chronic excess in nutrients leads to obesity and is a major contributor to β-cell failure (44). Although we have not observed β-cell failure after 4 days of overnutrition (data not shown), a longer duration of overnutrition may be required. If β-cell failure does result with a longer period of overnutrition, the zebrafish system may also be used to identify genes and pathways leading to β-cell failure in conditions of nutrient excess.

We have observed differences in signaling pathways that are involved in the β-cell differentiation induced by glucose and egg yolk. Although the glucose effect requires mTOR but not autocrine/paracrine insulin/IGF signaling, the opposite is true for the egg yolk effect. Although we cannot pinpoint if mTOR activity in β-cells is necessary, the data from genetic suppression of insulin/IGF1 signaling indicate that differentiated β-cells are an important component in the compensatory response. The requirement of insulin signaling for β-cell expansion has also been observed in adult mice, in which β-cell replication seems to be the main mechanism (18,19), although in our model, β-cell replication does not occur. It is possible that in all systems, a common factor or factors is produced by β-cells under conditions of overnutrition. Whether this factor(s) results in β-cell proliferation or differentiation of precursor cells may depend on the replication competence of β-cells or the availability of precursor cells.

Because we do not observe β-cell proliferation in this model of nutrient excess, the population of cells that are recruited to become β-cells is of great interest. It is unlikely that the increase of β-cells is simply a phenomenon of increased insulin promoter activity in pre-existing β-cells because all of the insulin-positive cells are marked by our transgenic approach and the β-cell numbers observed in unfed larvae are consistent with other reports (22). It is conceivable that a postmitotic precursor cell pool resides within the islet and is poised to differentiate with appropriate cues. Identification of the precursor population may be challenging, however, because many of the genes used to highlight potential precursor populations in zebrafish, such as pax6 and neurod, are also expressed in differentiated endocrine cell types. In rodent models, NGN3 expression is used as a marker of a precursor cells (45). However, ngn3 expression in the zebrafish pancreas has not been observed (46). The mnx1 promoter-directed reporter gene expression in the Tg(−5.0mnx1:tagRFP) and Tg(−3.0hb9:GFP)-transgenic lines likely mark a subpopulation of differentiating precursor cells and immature β-cells. Interestingly, the activity of the mnx1 promoter in the transgenic zebrafish contrasts to that of the endogenous Mnx1 in mice in which it has been shown to be active in both developing and adult β-cells (47). Whether our observation is a consequence of the lack of additional regulatory elements in the promoter used or a fundamental difference between zebrafish and mammals remains to be resolved. Overall, our findings support overnutrition can induce new β-cells in zebrafish through induction of precursor differentiation into β-cells. Through a combination of genetics, pharmacology, and transgenesis, the zebrafish larvae is well poised to identify the mechanisms, signals, and cells involved in this response, which can lend insights into β-cell responses involved in conditions of nutrient excess, insulin resistance, and the development of diabetes.

This work was supported by the Diabetes Research and Training Centers and Grant DK-088686 (to W.C.). We used the core(s) of the Vanderbilt Diabetes Research and Training Center funded by Grant DK-02593 from the National Institute of Diabetes and Digestive and Kidney Diseases, and confocal imaging was performed in the Vanderbilt University Medical Center Cell Imaging Shared Resource (supported by National Institutes of Health Grants CA-68485, DK-20593, DK-58404, HD-15052, DK-59637, and EY-08126).

No potential conflicts of interest relevant to this article were reported.

L.A.M. and W.C. designed the experiments. L.A.M. researched data. L.A.M. and W.C. wrote the manuscript. W.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank Kaitlin Joest (Vanderbilt University) for excellent technical help. Several of the transgenic zebrafish lines were provided by Drs. Michael Parsons (Johns Hopkins University), Francesco Argenton (University of Padova, Italy), Bernard Peers (University of Liege, Belgium), Li-En Jao and Susan Wente (Vanderbilt University), and Bruce Appel (University of Colorado). The authors also thank Drs. Patrick Page-McCaw (Vanderbilt University) and Maureen Gannon (Vanderbilt University) for critical reading of the manuscript.