Compounds that increase β-cell number can serve as β-cell replacement therapies in diabetes. In vitro studies have identified several agents that can activate DNA synthesis in primary β-cells but only in small percentages of cells and without demonstration of increases in cell number. We used whole well multiparameter imaging to first screen a library of 1,280 compounds for their ability to recruit adult rat β-cells into DNA synthesis and then assessed influences of stimulatory agents on the number of living cells. The four compounds with highest β-cell recruitment were glucocorticoid (GC) receptor ligands. The GC effect occurred in glucose-activated β-cells and was associated with increased glucose utilization and oxidation. Hydrocortisone and methylprednisolone almost doubled the number of β-cells in 2 wk. The expanded cell population provided an increased functional β-cell mass for transplantation in diabetic animals. These effects are age dependent; they did not occur in neonatal rat β-cells, where GC exposure suppressed basal replication and was cytotoxic. We concluded that GCs can induce the replication of adult rat β-cells through a direct action, with intercellular differences in responsiveness that have been related to differences in glucose activation and in age. These influences can explain variability in GC-induced activation of DNA synthesis in rat and human β-cells. Our study also demonstrated that β-cells can be expanded in vitro to increase the size of metabolically adequate grafts.
- drug screening
a reduced β-cell mass is a major defect in type 1 and a subgroup of type 2 diabetes patients. Strategies for its restoration are directed toward implanting an exogenous source for insulin-producing cells or at regenerating endogenous β-cells. Both approaches can benefit from compounds that induce the replication of β-cells and thus lead to an increased functional β-cell mass. Several studies have been undertaken to identify such agents by assessing their ability to increase the percentage of β-cells with DNA synthetic activity, as previously reviewed (12, 14, 46 ,47). These mostly were in vitro studies, using rodent islet cells and short exposure times, i.e., <72 h (46, 47). The reported effects mostly involved only small percentages of cells, with a few that shown promising in vivo effects (1, 2, 12, 40, 42, 48). They were not accompanied by data showing the efficacy of the compounds to increase β-cell number, a key index for therapeutic significance, together with evidence that the newly formed β-cells exhibit normal functional properties. Recent articles have provided a comprehensive overview of these agents and their underlying mechanisms but pointed to the absence, or scarcity, of similar effects on human β-cells (5, 19, 43).
We screened a library of pharmacologically active compounds (LOPAC) for compounds that increase the number of functional β-cells. We first examined which agents recruited cultured rat β-cells into DNA synthesis. The stimulatory agents were then assessed for their effect on the number of living β-cells. An in vitro expanded β-cell preparation was then evaluated for its ability to correct diabetes in mice. The strongest recruitment was observed for glucocorticoids (GCs). GCs did, however, not activate all cells. We investigated the basis for this heterogeneity as this might be informative on conditions for in vivo responses or failures as well as on reasons why a proportion of rat and human β-cells remained unresponsive in vitro.
MATERIALS AND METHODS
Cell culture media and supplements were purchased from Invitrogen (Life Technologies, Paisley, UK). The LOPAC (1,280 compounds) was purchased from Sigma-Aldrich (St. Louis, MO). The 804G cell line (provided by Dr T. Otonkoski, University of Helsinki, Helsinki, Finland) was used to prepare a matrix (6). Cells were seeded in imaging-grade black 384-well plates (Greiner Bio-One, Frickenhausen, Germany) or in six-well plates (BD Biosciences, San Jose, CA).
Purification, culture, and analysis of rat β-cell proliferation.
The present study was approved by our institutional ethical committee and conducted in accordance with European Community Council Directive 86/609/EEC. β-Cells were FACS purified from neonatal (1–3 days old) or adult male Wistar rats (8–10 wk; Janvier Bioservices) on the basis of cellular light scatter and their higher FAD-related autofluorescence in 2.8 mM glucose (32). Glucose high-responsive β-cells were separated from low-responsive cells based on their higher NAD(P)H autofluorescence after a 15-min exposure to 7.5 mM glucose, as previously described (18). Sorted cells were seeded in 804G matrix-coated 384-well plates for serum-free culture in Ham's F10 medium with 10 mM glucose and analyzed for number, viability, and proliferation activity as previously described (3). Whole well counts were performed after 1 day of culture to determine the number of viable cells at the start; this was performed by whole well image quantification (BD Pathway Bioimager 855 system, BD Biosciences) after live-dead staining (Hoechst 33342 for all cells and propidium iodide for dead cells). The number of living β-cells was obtained by subtracting propidium iodide-positive cell numbers from Hoechst 33342-positive cell numbers. The numbers counted at later time points were expressed as a percentage of the numbers on day 1.
Purification and culture of human β-cells.
Human pancreases were obtained from organ donors through the Eurotransplant Foundation according to Eurotransplant criteria for their use in clinical transplantation and associated projects with approval by our institutional ethical committee (CME 2005/118 2010/193). Islet cells were prepared and cultured as previously described (24). After 1–3 wk of serum-free culture, preparations were enzymatically dispersed and then incubated with the zinc-binding fluorochrome 6-methoxy-(8-p-toluenesulfonamido)quinoline (Invitrogen) before FACS purification of the fluorescently labeled cells, of which 60–70% were β-cells. After an overnight reaggregation in suspension, cell aggregates (<100 μm in diameter) were seeded in collagen type IV (Sigma-Aldrich)-coated 384-well plates and cultured in 7.5 mM glucose, which is known to maintain human β-cell survival and functions (24). The percentage of β-cells in proliferative activity was determined as previously described for rat β-cells (3) in the absence or presence of GCs and of a GC receptor (GR) antagonist.
Screening for compounds that recruit β-cells into DNA synthesis.
Freshly purified rat β-cells were preincubated in control medium for 24 h before the addition of a LOPAC compound (final concentration: 1 μM) or vehicle (0.1% DMSO) for culture until day 6. The thymidine analog 5-ethynyl-2′-deoxyuridine (EdU; 10 μM, Invitrogen) was added for the last 72 h of culture to label DNA synthesis; the percentage of EdU-positive cells was determined after whole well imaging and quantification using the BD Pathway Bioimager (3). The EdU data were normalized using the Z-score, which adjusts compound effects to within-plate variation. The Z-score of any compound is obtained by subtracting the average of percent EdU of all compounds in the plate from that induced by the compound and dividing the difference by the SD of measurements in the plate (25). A minimal Z-score level of 3 was set to select potential hits, i.e., compounds that increase the percent EdU-positive cells by at least three SDs above the plate-averaged percentage.
Properties of GC-amplified rat β-cell preparations.
GC-amplified rat β-cell preparations were examined in vitro and in vivo for characteristic β-cell properties. These preparations had been cultured in 6- or 24-well plates to provide sufficient cells for these experiments. In vitro experiments consisted of comparing their cellular glucose metabolism and cellular and medium insulin content with those in control preparations. Rates of glucose utilization and oxidation (22, 23) were measured after 6 days in culture in the absence or presence of hydrocortisone (HC): samples of 5 × 104 cells were incubated for 2 h with d-[5-3H]-glucose and d-[U-14C]-glucose (Amersham Biosciences, Buckinghamshire, UK), and the conversion to 3H2O and 14CO2 was measured.
They were also examined for their ability to correct diabetes in 8-wk-old male NOD/SCID mice (alloxan-Sigma-Aldrich, 50 mg/kg iv, stable 2-h fasting glycemia above 400 mg/dl). β-Cells cultured in six-well plates for 15 days were detached with accutase (Sigma-Aldrich) and reaggregated for 1 h before transplantation. Aggregates of 3–5 × 105 cells were embedded in a gelatin sponge (Gelfoam, Pfizer) and implanted in the abdominal fat pad of anesthetized animals. Blood glucose levels were measured (Glucocard X-Sensor, Menarini Diagnostics, Florence, Italy) in 2-h fasted mice or after an intraperitoneal glucose injection (3 g/kg body wt). Plasma rat C-peptide was measured with a Linco assay kit (St. Charles, MO). The insulin content of retrieved implants and the pancreas was determined by a radioimmunoassay (32).
Data are expressed as means ± SE of n independent experiments. Statistical significance of differences was assessed with a two-tailed unpaired Student's t-test or one-way ANOVA with Tukey's test for multiple comparisons using GraphPad Prism (GraphPad Software, San Diego, CA).
High-content screening assay identifies GCs as potent activators of DNA synthesis in β-cells.
A library of 1,280 compounds was screened using a recently described high-content screening assay (3) for compounds that recruit adult rat β-cells into proliferative activity during a 6-day culture period. Cells were preincubated for 1 day in control medium before a test compound at 1 μM was added for a 5-day culture and EdU (10 μM) was added for the last 3 days. The percentage of EdU-positive cells was normalized using the Z-score, and four of the compounds tested gave a Z-score higher than five: triamcinolone (5.9), beclomethasone (5.9), betamethasone (5.5), and HC (5.4), all of which are GR ligands. None of them increased the percentage of dead cells. We selected a natural GC, HC, and its synthetic analog, 6-methylprednisolone (MP), for further investigation.
GCs time dependently recruit β-cells into the cell cycle, leading to an increase in β-cell number, an effect not observed in other islet cells.
The effects of HC and MP were followed over a 2-wk exposure time. They did not increase the number of β-cells in DNA synthesis during the first 2 days (from days 1 to 3) but did significantly so during each of the subsequent 3-day periods (Fig. 1A). The effect of HC progressively increased with time, reaching the highest percentage of EdU-positive cells between days 12 and 15, whereas that of MP was already maximal between days 3 and 6 and maintained until day 12, followed by a decline. This decline in percentage might be explained by the increase in total cell number by that time (see further), whereas the number of recruited cells remained constant.
Sequential labeling with EdU and 5-bromo-2′-deoxyuridine (BrdU) was performed to examine whether cells that have entered DNA synthesis between days 3 and 6 are also positive during a subsequent 3-day labeling and are therefore counted twice. This experiment could also indicate whether β-cell replication occurs by a repetitively replicating subpopulation or whether it proceeds as a time-dependent recruitment of new cells into the cell cycle. It was conducted for HC, using EdU during the first period and BrdU during the second period. Only 3% of the BrdU-incorporating cells were also positive for EdU (data not shown); the majority thus corresponded to a newly recruited population. Thus, β-cells differ in their individual responsiveness to GC-induced activation of DNA synthesis and exhibit intercellular differences in the exposure time needed before starting DNA synthesis. Under the present culture conditions, this prior exposure time varied between 48 and 264 h (Fig. 1A).
Summation of the percentages measured over the successive 72-h labeling periods indicated that over 60% of the cells at the start became analog labeled during 14 days of culture in the presence of HC or MP, whereas this was only about 13% in their absence. This difference should be an indirect index for an increase in cell number if all labeled cells proceed to replication. This appears to be the case, as shown by the total cell counts at the end of culture (Fig. 1B). HC and MP almost doubled the initial number in 2 wk, whereas only about a 10% increase occurred in the control condition. This effect was time dependent, as was also the case for the recruitment into DNA synthesis. At the end of culture, the purity in insulin-positive cells was as high as at start (Fig. 1C), indicating that the increase was achieved by replicating β-cells that kept their hormone positivity.
In dispersed rat islet cell preparations, GC-induced DNA synthetic activity was only found in insulin-positive cells but not in other cell types (data not shown). This was also the case in adult human islet cell preparations, where 6 days of exposure to MP increased the percentage EdU-positive β-cells threefold with a wide variation in individual responses, but all suppressed by a GR antagonist (Table 1). In 10 of 14 tested cases, MP activated 1–8% of β-cells above the 0–2.5% in control medium; this MP-induced recruitment involved less cells than in rats and was too low to detect an impact on cell number. It was absent or negligible in the four other cases (<1% activated by MP). This variation could not be related to differences in age, sex, or body weight. No stimulation was seen in the INS1832/13 cell line when we used the same serum-free culture medium as in the present study (data not shown). These data suggest that GC-induced proliferation is mediated through signaling pathways occurring in primary adult β-cells.
The HC-induced increase in β-cell number was suppressed by mifepristone (RU-486), a GR antagonist (11), but not by spironolactone, a mineralocorticoid receptor (MR) antagonist (10) (Table 2). Aldosterone, a MR agonist (9), exerted no effect on β-cell number, further supporting that GC activation of DNA synthesis is mediated by GR. Mifepristone (RU-486) also suppressed GC-induced DNA synthesis in human β-cells (Table 1).
GCs induce the replication of glucose-responsive adult rat β-cells without causing toxicity; they are toxic for neonatal β-cells.
The observation that individual β-cells differed in GC responsiveness, with a fraction entering DNA synthesis after 2 days of stimulation and others after longer exposure times, led us to examine whether this correlated with differences in the glucose-responsive state of the cells. In prior work, we have shown that adult rat β-cells exhibit intercellular differences in the glucose-regulated metabolic redox state whereby the subpopulation with high glucose responsiveness is mainly responsible for the biosynthetic activity (18). This characteristic has been documented by functional studies on subpopulations that were separated according to differences in this metabolic state. We therefore compared the GC effect in β-cell subpopulations that were sorted on basis of their redox response to 7.5 mM glucose (18) and then cultured for 14 days in the absence or presence of HC. In HC-free medium, no difference was seen: for both subpopulations, the number of viable cells at the end of culture was similar to that at the start (Table 3). In the presence of HC, this number increased only in the subpopulation sorted on basis of the higher metabolic glucose response, indicating the dependency of the GC effect on the glucose-responsive state of the cells, a feature that can explain its β-cell specificity.
Culture in 10 mM glucose has previously been shown to maintain the glucose-responsive state of adult rat β-cells as well as their viability (23). This was also the case in the presence of HC, as judged by the dose-dependent glucose utilization and oxidation rates measured after 5 days of exposure (Fig. 2). In fact, the amplitude of this metabolic response was higher than in control cells. This HC-induced amplification might be related, at least causally, to its recruitment of cells into DNA synthesis.
To further assess this dependency of the glucose responsive state of the cells, we examined whether HC exerted the same effect during 2 wk of culture with low (5 mM) or high (20 mM) glucose concentrations, which are known to impair the metabolic state of adult rat β-cells (23). The addition of HC only marginally maintained the number of β-cells at 5 mM and had no substantial effect at 20 mM (Table 4). Along this rationale, we assessed its effect in neonatal rat β-cell preparations, which have not yet developed a glucose-responsive state (26) but exhibit a high replication rate under basal conditions (3). The addition of HC was cytotoxic at the three tested glucose concentrations, reducing initial β-cell number by >50% and suppressing the basal expansion of the cells with 5 and 10 mM glucose (Table 4).
GC-induced replication of β-cells amplifies their functional mass for transplantation.
The higher β-cell number at the end of 2 wk of culture with GC raises the possibility that this compound can be used to generate a larger β-cell mass for transplantation. However, the cells appeared heavily degranulated (74% lower cellular insulin content than control cells when cultured with HC; Table 5), which is considered to interfere with their role in metabolic correction. Exposure for 8 days (from days 1 to 9) resulted in lower values than in the no-GC control conditions, but the difference did not reach statistical significance, and no further decline occurred when GCs were removed for the subsequent 6-day period (Table 5). Interestingly, this discontinuation did not affect the HC-induced amplification as seen on day 15 under continued exposure. Similar observations were made during culture with MP, with an extension to the amount of insulin that was released in the medium, i.e., no decline during the first 8 days of exposure but a marked reduction during the subsequent 6 days in the presence of MP, not after its discontinuation (Table 5).
The functional capacity of the GC-amplified β-cell mass was then assessed by transplantation in diabetic mice. Equal numbers of β-cells were cultured at 10 mM glucose for 15 days in the absence (control) or presence of HC (from days 1 to 9). At day 15, cell yield after HC treatment was 54–110% higher than in control preparations (paired cultures from three independent experiments), thus allowing the preparation of more grafts per experiment; control and HC grafts contained the same number of β-cells and a similar total hormone content (between 10 and 15 μg insulin/graft). When implanted in the abdominal fat pad of alloxan-diabetic mice, both groups corrected hyperglycemia within 1 wk and remained normalized over the 6-wk followup period (Fig. 3A), with stable glucose-stimulated rat C-peptide levels (15 min after glucose load in recipients of control and HC grafts, respectively, at posttransplant week 1: 2.5 ± 0.1 and 2.2 ± 0.2 nM and at posttransplant week 5: 2.6 ± 0.3 and 2.6 ± 0.2 nM). Both groups exhibited an equally rapid normalization of glycemia after intraperitoneal glucose load, with similar kinetics as normal controls (Fig. 3B). Removal of implants was rapidly followed by reversal to hyperglycemia. For both groups, the insulin content of implants at posttransplant week 6 was ∼50% of that in the initial grafts and exceeded the corresponding pancreatic insulin content by almost 10-fold (data not shown). These data demonstrate that the HC-amplified β-cell mass generates β-cell grafts that, on a cell number basis, are equally potent as control preparations.
A recently developed image-based high-content assay was used to screen a LOPAC for agents that recruit adult rat β-cells into DNA synthesis. Four of the 1,280 tested compounds generated a Z-score above five; they all corresponded to GR ligands. Selected natural and synthetic GCs specifically increased the percentages of proliferating rat and human β-cells but not those of other islet cell types.
In vivo administration of GCs has been shown to increase the percentage of β-cells in proliferative activity (21, 35, 36, 45). This effect has been considered as indirect, being a consequence of the hyperglycemic and hyperinsulinemic state that is caused by the GC-induced insulin resistance (35, 36, 45). Its impact on β-cell number has remained unclear since total cell counts in the pancreas have not been performed; neither was the balance made with other processes that can affect the β-cell population in chronic hyperglycemia, such as an increase in β-cell size and β-cell toxicity (38). The present in vitro study on purified rat β-cells shows that GCs can exert direct effects on β-cells, which, depending on their age and functional state, can consist of an increased replication or cytotoxicity. The model allows assessing these effects at different glucose concentrations and exposure times; it was thus found that prolonged GC exposure, after a GC-induced replication, resulted in reduced insulin production and marked cellular degranulation. As with all in vitro studies with cells and tissue, data in these simplified models are not necessarily representative for the effects in humans but insights in their relationship with cell biologic variables should help interpret observations in (patho)physiological conditions. Identification of conditions under which GCs are, or are not, cytotoxic for β-cells do caution against the statement that their inclusion in immune suppressive protocols is cytotoxic for transplanted β-cells. The clinical success of the GC-free Edmonton protocol may have led to this assumption but was in fact not shown to be attributable to a lack of GC-induced β-cell destruction; it has several other candidate reasons (41).
Both HC and MP recruited adult rat β-cells into proliferative activity, increasing their numbers over a 2-wk culture period with 10 mM glucose. This effect is mediated through GRs without contribution from MRs. An exposure of minimally 2 days is needed before cells become progressively activated into DNA synthesis and involves up to ∼60% of the initial cell number over 12 days. Analysis of the replicative history of GC-treated cells using EdU-BrdU sequential labeling (3, 44) revealed that GCs increase cell numbers by progressive recruitment of resting cells into proliferative activity, not through repetitive proliferation of a small subpopulation of β-cells; this is in line with observations in other conditions (3, 7, 44). The data indicate a major intercellular difference in β-cell propensity to undergo GC-induced proliferation. A high glucose concentration (20 mM) was previously reported to also activate adult rat β-cells into replication, but this involved only 15% of the initial cell number (3). These intercellular differences in susceptibility and time to recruitment into DNA synthesis represent another manifestation of the functional heterogeneity that has been observed for other properties of adult rat and human β-cells (15, 18, 23, 31).
The GC-induced β-cell replication started after minimally 2 days of exposure, with intercellular differences in the time required before activation. This suggests that GCs do not act as mitogens but as replication competence-inducing factors that bring quiescent β-cells toward cell cycle entry. Progression factors are then needed to advance them through the cell cycle (16, 20, 33). Such permissive action has previously been described for GCs in growth factor-induced proliferation of other cell types (4) and is recognized in their anti-inflammatory effects (27, 28, 39). It might be mediated through GC amplification of glucose metabolism, which is compatible with the view that a high β-cell workload is a determinant factor for its replication (8) and consistent with induction by a glucokinase activator (3, 29, 34). In addition, we showed that GC-induced replication was achieved in the β-cell subpopulation that initially exhibited a high responsiveness to glucose. Thus, efficient glucose sensing and metabolic activity sustain the completion of the cell cycle, making glucose a likely progression factor. Insulin released by cultured β-cells may also serve as a cell cycle progression factor, as proposed from in vivo studies (37) and as reported for other cell types (16, 17); it is difficult to assess its possible role in cultured islet cell preparations in view of the high medium insulin levels that are inherent to this model concentration (>10−7 M in our study). Other progression factors could be generated by the signal transduction pathways activated by the laminin-5-rich 804G extracellular matrix on which the cells were seeded. Prior studies have demonstrated that this matrix improves in vitro β-cell function, survival, and proliferation (13, 30).
We examined whether the in vitro expanded β-cell population had retained the ability to correct diabetes in mice. This experiment was conducted with cells collected after 2 wk, where HC had been omitted during the last 6 days to avoid their degranulation. Grafts contained the same β-cell number as those prepared from the condition without HC. Since the HC treatment had almost doubled the initial β-cell number, it resulted in more grafts. Both types of grafts were equally effective in reducing hyperglycemia in alloxan-diabetic mice and establishing normal glucose tolerance curves. These data demonstrate that HC-expanded β-cells exhibit and maintain their homeostatic function. To our knowledge, they are also the first demonstration of an in vitro expansion of primary β-cells so that the size or the number of β-cell grafts can be increased while maintaining their functional properties.
A fraction of the initial β-cell number remained quiescent despite 2 wk of exposure to the two synergistic activators, glucose and HC. This was also the case when the glucose concentration was increased to 20 mM. It is unknown whether this unresponsive cell subpopulation requires a longer exposure time or different activators or whether its life cycle has reached a phenotype that is no longer recruitable to replication. The size of this subpopulation may well vary with the age of the cells and their host as well as with the species. In adult human β-cell preparations, we also noticed a GC-induced activation into DNA synthesis, with marked individual differences in the percentages of recruited β-cells but all suppressed by a GR antagonist. The proportion of human β-cells that were not activated by GC within 6 days was higher (>90%) than that in adult rat β-cells (<75%), but the reasons might be similar. Comparative analysis of corresponding phenotypes may identify cellular characteristics for this unresponsiveness as well as indicate ways to overcome it. Differences in donor characteristics, duration of intensive care, and cold preservation can be expected to cause differences in the phenotype of isolated human β-cells and thus in their responsiveness to GCs. The serum-free culture condition that we developed for maintaining survival and glucose-responsive functions of β-cells (22–24) may reduce these influences and their variability, but it generates more standardized preparations for in vitro analysis of the biological properties of human β-cells.
Whereas young adult rat β-cells required stimulatory glucose plus GC concentrations to double their cell number over 2 wk, neonatal rat β-cells exhibited this property at basal glucose (5 mM) in the absence of GCs, with a majority in DNA synthesis during the first 9 days of culture (data not shown). The addition of HC does not increase β-cell number beyond the amplification that occurred under basal conditions; on the contrary, it reduces the initial number by >50%. These data demonstrate a major age-dependent difference in HC effects on isolated rat β-cells. GC-induced cytotoxicity in neonatal β-cells markedly reduces their number and suppresses their normal basal growth process, which contrasts with the drugs' ability to induce replication in glucose-activated young adult β-cells. Although this in vitro observation on rodent cells is not necessarily indicative for a similar process in humans, it raises sufficient (patho)physiological interest to be further examined.
This work was supported by grants from the Agency for Innovation by Science and Technology in Flanders (IWT) (IWT645), the Industrial Research Fund (IOF742), the Research Foundation Flanders (G.0428.09N), and UZ-VUB-Wetenschappelijk Fonds Willy Gepts (WFWG-BOR16). Z. Assefa was recipient of an IWT Postdoctoral Innovation Mandate and S. Akbib and A. Lavens were recipients of an IWT PhD fellowship.
No conflicts of interest, financial or otherwise, are declared by the author(s).
Z.A., Z.L., K.H.H., and D.P. conception and design of research; Z.A., S.A., A.L., G.S., and K.H.H. performed experiments; Z.A., S.A., A.L., G.S., Z.L., K.H.H., and D.P. analyzed data; Z.A., G.S., Z.L., K.H.H., and D.P. interpreted results of experiments; Z.A. and G.S. prepared figures; Z.A. and D.P. drafted manuscript; Z.A., S.A., A.L., G.S., Z.L., K.H.H., and D.P. approved final version of manuscript; Z.L. and D.P. edited and revised manuscript.
The authors thank our collaborators for preparing quality-controlled rat and human beta-cells and performing immunochemical assays and characterizations.
- Copyright © 2016 the American Physiological Society