Glucose-stimulated insulin secretion is the hallmark of the pancreatic β-cell, a critical player in the regulation of blood glucose concentration. In 1974, the remarkable observation was made that an efflux of intracellular inorganic phosphate (Pi) accompanied the events of stimulated insulin secretion. The mechanism behind this “phosphate flush,” its association with insulin secretion, and its regulation have since then remained a mystery. We recapitulated the phosphate flush in the MIN6m9 β-cell line and pseudoislets. We demonstrated that knockdown of XPR1, a phosphate transporter present in MIN6m9 cells and pancreatic islets, prevented this flush. Concomitantly, XPR1 silencing led to intracellular Pi accumulation and a potential impact on Ca2+ signaling. XPR1 knockdown slightly blunted first-phase glucose-stimulated insulin secretion in MIN6m9 cells, but had no significant impact on pseudoislet secretion. In keeping with other cell types, basal Pi efflux was stimulated by inositol pyrophosphates, and basal intracellular Pi accumulated following knockdown of inositol hexakisphosphate kinases. However, the glucose-driven phosphate flush occurred despite inositol pyrophosphate depletion. Finally, while it is unlikely that XPR1 directly affects exocytosis, it may protect Ca2+ signaling. Thus, we have revealed XPR1 as the missing mediator of the phosphate flush, shedding light on a 45-year-old mystery.
Glucose-stimulated insulin secretion is the central function of the pancreatic β-cell for which loss or malfunction contributes to type 1 and type 2 diabetes, respectively. An observation made more than 40 years ago by Freinkel et al. (1) and confirmed in several further studies (2) was that this glucose-stimulated release of insulin was accompanied by an efflux of [32P]-labeled inorganic phosphate (Pi) in pancreatic islets prelabeled with [32P]Pi. This “phosphate flush” was shown to be specific to pancreatic β-cells (3) and did not occur in other islet cells or the exocrine pancreas (3). The efflux occurred simultaneously with first-phase insulin release (1) but was not driven by the Ca2+ influx that accompanies insulin secretion (1,2). It was a dose-dependent phenomenon (4) and specific for D- over L-glucose (5) but could be promoted by a number of other insulin secretagogues (2). The source of this Pi was identified as existing under the plasma membrane (6). Over the last 45 years, there have been a number of proposals advanced to explain the phosphate flush but the presumed Pi transporter mediating this effect has not been identified, and thus, there are many unresolved questions surrounding this phenomenon. Pi metabolism has specific importance in pancreatic β-cells, as Pi is required for glycolysis and thus the operation of the β-cell as a glucose sensor (7). Furthermore, plasma buildup of Pi in, for example, chronic kidney disease could be detrimental to β-cell viability (8). Thus, the control of Pi is of central importance to the β-cell’s ability to maintain blood glucose homeostasis.
Studies in plants have revealed that a number of different Pi transporters possess an SPX domain, named after proteins Syg1, Pho81, and XPR1 (9). In the mammalian genome, XPR1 is the only protein that possesses the SPX domain. XPR1 was originally characterized as a receptor for xenotropic and polytropic murine retroviruses (10). However, new studies have revealed XPR1 as the first Pi efflux transporter to be identified in metazoans (11–14). In addition, inositol pyrophosphates, which play a central role in β-cell regulation (15,16), are proposed to specifically bind and activate SPX domains, including that of XPR1 (17–19). XPR1 is therefore the first potential candidate mediating the phosphate flush. Using a pancreatic β-cell line and a primary β-cell model, we tested the hypothesis that XPR1 is responsible for the phosphate flush and studied its regulation by inositol pyrophosphates. We demonstrate that XPR1 is important for the regulation of intracellular Pi in the β-cells and is indeed the missing player that mediates the phosphate flush.
Research Design and Methods
All animal experiments were approved by the Animal Ethics Committee of Northern Stockholm, Sweden, and carried out in accordance with the National Institutes of Health’s Guide for the Care and Use of Laboratory Animals.
Cell culture reagents, Hanks’ balanced salt solution, M-PER Mammalian Protein Extraction Reagent, Pierce BCA Protein Assay, Lipofectamine 2000 and RNAiMAX, TaqMan gene expression assays, and siRNAs were obtained from Thermo Fisher Scientific (Stockholm, Sweden). Silencing was implemented by IP6K siRNAs (15,16) or XPR1 siRNAs (identification numbers s72989 and s201933) with silencer select negative control 1 (catalog number 4390843). Common chemicals, Accutase, and N2-(m-Trifluorobenzyl), N6-(p-nitrobenzyl) purine (TNP) were from Sigma-Aldrich (Stockholm, Sweden) or Merck KGaA (Darmstadt, Germany). Carrier-free [32P]orthophosphate and the insulin AlphaLISA assay were obtained from PerkinElmer (Stockholm, Sweden). PiColorLock malachite green–based Pi assay was obtained from Expedeon/Abcam (Cambridge, U.K.).
Cell Culture and RNA Silencing in MIN6m9 Cells
Preparation and RNA Silencing of Pseudoislets
Islets were isolated by collagenase digestion from the pancreata of male C57BL/6J mice and immediately extracted for RNA analysis or cultured in complete RPMI medium with 10% FBS (16). The following day, the islets were washed with Hanks’ balanced salt solution and dissociated into cells using 100 µL Accutase/tube containing ∼200 islets in a ThermoMixer shaker (Eppendorf, Göteborg, Sweden) at 500 rpm and 37°C for 10 min, followed by gentle pipetting. Then, fresh Accutase (200 µL) was added, and the step was repeated. Cells were transferred by centrifugation (180g for 8 min) to complete DMEM without antibiotics and counted. The siRNAs were premixed with Lipofectamine RNAiMAX in Opti-MEM for 5 min. This was added to the cell suspension (10-fold dilution). After 5 min, the cell suspension was distributed (200 µL/well) into PerkinElmer CellCarrier Spheroid ULA 96-well microplates (2,500 cells, 10 pmol siRNAs, and 0.6 µL RNAiMAX in each well). After plate centrifugation (180g for 6 min), cells were allowed to reaggregate into three-dimensional structures (∼1 pseudoislet/well) for ∼5 days in a cell-culture incubator.
RNA Extraction and Real-time RT-PCR
Total RNAs were extracted from MIN6m9 cells, islets, or pseudoislets (RNeasy Mini and Micro kits; Qiagen, Sollentuna, Sweden) and reverse transcribed as previously described (15,16). The relative expression of mRNA was measured by real-time RT-PCR TaqMan Gene Expression Assays: XPR1 (Mm00495501_m1), IP6K1 (Mm00501996_m1), and IP6K2 (Mm01232057_m1) using ABI 7300 instrument and QuantStudio 5 (Thermo Fisher Scientific). 18S rRNA (TaqMan rRNA Control Reagent) was used as endogenous control.
Dynamic Incubation Assay
MIN6m9 cells were incubated at 37°C in dynamic assays as described previously (16) in order to determine either Pi efflux or insulin secretion (see detailed protocols in Supplementary Fig. 1A). The cells were incubated for 90 min with 25–50 µCi carrier-free [32P]Pi (for Pi efflux measurements) or without this label (for measurements of insulin secretion) in modified Krebs buffer containing 5.5 mmol/L glucose. After additional 60 min in unlabeled Krebs buffer with 0.5 mmol/L glucose, the cell monolayers were subjected to the dynamic phase, exchanging buffer every minute. The radioactivity of the collected dynamic assay samples and the total cellular material was determined using a liquid scintillation counter. Insulin was determined by the AlphaLISA assay and normalized with total protein, extracted using M-PER Mammalian Protein Extraction Reagent, and determined by the BCA Protein Assay. For determining the intracellular Pi content, cells were extracted with 5% perchloric acid (PCA) at the time point indicated in Supplementary Fig. 1B.
Static Incubation Assay
For the determination of Pi efflux, pseudoislets were incubated at 37°C in a similar regimen as the MIN6m9 cells in the presence of [32P]Pi in modified Krebs buffer with 5.5 mmol/L glucose for 90 min. [32P]Pi was omitted in the experiments for measuring insulin secretion and intracellular Pi (see protocol in Supplementary Fig. 1C). After 50 min in 3 mmol/L glucose, the pseudoislets were subjected to the static incubation phase (i.e., basal incubation at 3 mmol/L glucose for 10 min followed by stimulation with 16.7 mmol/L glucose for 10 min). [32P]Pi and insulin were determined in the samples by scintillation counting and the AlphaLISA assay, respectively. PCA extraction of the pseudoislets was performed to obtain both total intracellular Pi and cellular protein.
Measurements of Intracellular Pi
Measurements of Intracellular Free Calcium Concentrations
Pseudoislets were loaded with 2 μmol/L Fura-2 LeakRes (AM) (TEFLabs, Inc., Austin, TX) in modified Krebs buffer containing 3 mmol/L glucose for 60 min. After loading, a single pseudoislet was transferred to an open perifusion chamber maintained at 37°C. The pseudoislet was perifused with Krebs buffer containing 3 mmol/L glucose and 16.7 mmol/L glucose was used for stimulation. Intracellular free calcium concentration ([Ca2+]i) was measured as the 340/380 nm fluorescence ratio using a Spex Fluorolog spectrophotometer coupled to a Zeiss Axiovert 35M microscope with a Zeiss Fluar 40×/1.30 oil immersion objective (Carl Zeiss, Göttingen, Germany).
The data, expressed as means ± SEM, were statistically analyzed with GraphPad Prism software version 5.0. Details of analyses are indicated in figure legends, text, and Supplementary Table 1.
Data and Resource Availability
The data sets generated during and/or analyzed during the current study are available from the corresponding authors on reasonable request.
Results and Discussion
In the original studies, it was shown that in pancreatic islets, only the β-cells mediated the phosphate flush (3). In order to investigate this phenomenon, we sought to use a currently established glucose-responsive β-cell line from mouse (MIN6m9), as it is a more experimentally amenable system. MIN6m9 cells are a good β-cell surrogate, both in terms of their response to glucose and the regulation of exocytosis (16,21). We first determined that the phosphate flush occurred in this cell model. Thus, we prelabeled MIN6m9 cells with carrier-free [32P]Pi using similar protocols to the original studies in islets (1). This was followed by a dynamic incubation assay (16,21) that allows us to assess temporal changes in Pi release and insulin secretion in response to glucose with a 1-min time resolution (Supplementary Fig. 1A). Figure 1A and B illustrate that glucose stimulation of these cells leads to the expected enhanced efflux of [32P]Pi, as well as to an increase in insulin secretion (Fig. 1C and D).
Intracellular Pi in the MIN6m9 cells cultured in complete DMEM was 41.3 ± 1.9 pmol Pi/µg protein (mean ± SEM; n = 6), which is comparable to other cells (18,19). However, it has been reported that intracellular Pi accumulates during the low-glucose condition used to achieve basal insulin secretion (22). Under these settings (Supplementary Fig. 1B), we also observed an accumulation of Pi (107.7 ± 13.0 pmol Pi/µg protein; n = 6). This accumulation was significantly reduced (50.9 ± 16.1%; n = 6; P < 0.05, one-sample t test) by maintaining high glucose (10 mmol/L) during the preincubation period. Thus, the phosphate flush is a possible mechanism normalizing intracellular Pi during glucose stimulation to avoid reaching very high levels associated with β-cell dysfunction (8).
We hypothesized that XPR1 plays a crucial role in Pi efflux in β-cells, mediating the phosphate flush (Fig. 1E). We detected XPR1 mRNA expression in both MIN6m9 cells and mouse pancreatic islets (Fig. 2A). We then used siRNA silencing to knockdown the expression of XPR1 in MIN6m9 cells to ∼30% of its normal level (Fig. 2B). This treatment did not affect either the cell number or the cell viability (Supplementary Fig. 2). We then used the dynamic assay to assess the effect of silencing XPR1 on the phosphate flush. In this case, we normalized the Pi efflux to the total phosphate in the cells in order to control for any disturbance to phosphate metabolism mediated by the silencing. Figure 2D and E show that knockdown of XPR1 dramatically reduced the glucose-stimulated Pi efflux to ∼10% of the control value. Normalizing Pi efflux to protein revealed that XPR1 also significantly impacted basal Pi efflux (Fig. 2, legend). We then used pseudoislets, a primary β-cell model in which we exploited the siRNA approach and reached a similar level of XPR1 knockdown as in MIN6m9 cells (Fig. 2C). A consequent reduction of the phosphate flush was observed (Fig. 2F). Consistent with the decrease in Pi efflux, intracellular Pi concentrations were significantly enhanced upon XPR1 silencing, both in MIN6m9 cells and pseudoislets (Fig. 2M). Thus, XPR1 is responsible for the β-cell glucose-stimulated phosphate flush.
An important question, unresolved in previous studies (2), is whether or not the phosphate flush is necessary for insulin secretion or is simply a parallel phenomenon that accompanies exocytosis. Indirect evidence suggested that the two processes could be disassociated. For example, removal of extracellular Ca2+ or blockade of its entry prevents insulin release but not the phosphate flush (1,2). Conversely, stimulation of secretion by sulphonylureas does not increase the phosphate flush (7). With the identification of XPR1 as the mediator of the flush, we could, for the first time, test directly the relationship between the phosphate flush and insulin exocytosis. Using identical conditions to those used to assess the phosphate flush, we measured the impact of XPR1 silencing on insulin release. Figure 2G and H show only a modest impact on first-phase insulin secretion under conditions in which the phosphate flush is almost ablated in MIN6m9 cells. XPR1 knockdown in pseudoislets did not significantly impact the five-fold increase in insulin secretion provoked by 10 min of glucose stimulation (Fig. 2I). Taken together, these data suggest that the XPR1-mediated phosphate flush is not essential for insulin secretion.
Freinkel (2) originally proposed that an interaction between Ca2+ ions and the phosphate flush could contribute to biphasic insulin secretion. Interestingly, several reports have highlighted the relationship of XPR1-mediated Pi efflux to cellular calcium metabolism (12–14), including calcification due to Pi buildup and calcium phosphate precipitation (13,14). We therefore examined the glucose-induced changes in [Ca2+]i in pseudoislets in which XPR1 was knocked down (Fig. 2J and K). If XPR1 was preventing calcium phosphate precipitation, we might anticipate a blunted [Ca2+]i response in its absence. There was no significant impact on the glucose-stimulated Ca2+ transient of XPR1 knocked down pseudoislets compared with controls (Fig. 2J–L). However, a subsequent stimulation by 25 mmol/L potassium, which drives a higher peak increase in [Ca2+]i, led to a significantly reduced increase in [Ca2+]i in the XPR1 knocked down pseudoislets (Δ Fura-2 ratio; control siRNA–treated, 0.716 ± 0.049; XPR1 siRNA–treated, 0.544 ± 0.022, means ± SEM; n = 6–7; P < 0.05, Student t test). These data suggest that under superphysiological conditions or prolonged stimulation, the XPR1-mediated phosphate flush could serve to prevent calcium phosphate precipitation also in pancreatic β-cells, thus protecting the integrity of [Ca2+]i signaling, a key driving force of insulin exocytosis (23,24).
The SPX domain found in XPR1 is regulatory and not necessary for Pi transport (9,17). Since inositol derivatives and particularly the inositol pyrophosphates (17) can specifically bind to SPX domains, it is possible that they regulate the phosphate flush through this domain (9,17). We therefore investigated whether inositol pyrophosphates have any role in the regulation of the phosphate flush in MIN6m9 cells, a good β-cell model in studies on inositol pyrophosphates (15,16). We used two complementary approaches to curtail the production of the critical inositol pyrophosphates, 5-diphosphoinositol pentakisphosphate (5-PPIP5/5-IP7) and 1,5-bisdiphosphoinositol tetrakisphosphate (IP8) (Fig. 3A).
Firstly, we used siRNA to knock down the two IP6Ks that exist in pancreatic β-cells (IP6K1 and IP6K2) (15,16). This prevents the production of 5-IP7 and thus also the formation of IP8 (Fig. 3A). Under these conditions, IP6K1 and IP6K2 are knocked down by ∼40% and 50%, respectively, which leads to a 75% reduction in IP7 (15) (Supplementary Fig. 3A and B). Again, we applied the dynamic protocol to study the Pi efflux. Reduction in the levels of IP6K1 and IP6K2 caused a significantly increased glucose-induced phosphate flush, without affecting basal Pi efflux (Fig. 3B and C). However, under these conditions, the expression of XPR1 mRNA was doubled (Supplementary Fig. 4), suggesting a compensatory mechanism for the loss of the stimulatory inositol pyrophosphates. In addition, there was an accumulation of intracellular Pi (Fig. 3F), as reported in other cells in which IP6Ks were knocked down (18).
We also used a short-term pharmacological approach applying the pan-specific IP6K inhibitor TNP to circumvent the longer-term adaptive responses seen above. This too blocks the formation of both 5-IP7 and IP8 (Fig. 3A). In contrast to the long-term knockdown of IP6Ks, TNP reduced Pi efflux under basal conditions and delayed the phosphate flush rise (Fig. 3D and legend). These data show that acute knockdown of 5-IP7 and IP8 production inhibits basal phosphate efflux via XPR1, consistent with two recent reports in other cells (18,19). However, TNP treatment does not reduce the glucose-stimulated phosphate flush (Fig. 3D and E). It is possible that the acute TNP treatment leads to a short-term compensatory adaption by which more existing XPR1 is translocated to the plasma membrane of β-cells, as observed for active KATP membrane channels upon glucose stimulation (25). Another factor that could alter Pi efflux kinetics is the buildup of intracellular Pi that we (Fig. 3F) and others (18) observe on IP6K knockdown. Furthermore, while our strategy reduces 5-IP7 and IP8, it will not impact a glucose-generated production of 1-IP7, which has recently been shown to be more potent that 5-IP7 in activating XPR1 (19). Thus, we would suggest that under basal glucose conditions Pi efflux via XPR1 is directly driven by inositol pyrophosphates. However, under the glucose stimulatory conditions of the phosphate flush, other so far unidentified factors linked to glucose metabolism contribute to a more complex regulation of the XPR1-dependent Pi efflux.
Overall, our study has solved an important aspect of the β-cell phosphate flush (1). We demonstrate that XPR1 is the mediator of glucose-stimulated Pi efflux and that this process is largely independent from inositol pyrophosphate regulation. Furthermore, we show that the phosphate flush is unlikely to contribute to glucose-stimulated insulin secretion, while providing a hint that the ultimate role of this mechanism may be to prevent calcium phosphate precipitation and thus protect pancreatic β-cell Ca2+ signaling.
C.J.B. and F.H.G.T. contributed equally to this study.
See accompanying article, p. 27.
This article contains supplementary material online at https://doi.org/10.2337/figshare.12821192.
Acknowledgments. The authors thank Professor S. Seino (Kobe University Graduate School of Medicine, Kobe, Japan) for the gift of MIN6m9 cells; Dr. J.H. Park (Soonchunhyang University College of Medicine, Cheonan, Korea), Dr. B. Leibiger and Dr. V.M. Lauschke (Karolinska Institutet, Stockholm, Sweden), Professor D. Piston (Washington University School of Medicine, St. Louis), and Dr. C. Reissaus (Novo Nordisk Research Center, Indianapolis) for advice about pseudoislets; and Professor A. Saiardi and Dr. M.S. Wilson (UCL, London, UK) for advice on Pi assays.
Funding. This study was supported by the Vetenskapsrådet, the Novo Nordisk Foundation, Karolinska Institutet, the Swedish Diabetes Association, The Family Knut and Alice Wallenberg Foundation, Diabetes Research and Wellness Foundation, Stiftelsen för Strategisk Forskning, Berth von Kantzow’s Foundation, The Skandia Insurance Company Ltd., Strategic Research Programme in Diabetes at Karolinska Institutet, ERC-2013-AdG (338936-BetaImage), ERC-2018-AdG (834860 EYELETS), the Stichting af Jochnick Foundation, the Family Erling-Persson Foundation, Fundação de Amparo à Pesquisa do Estado de São Paulo grants 2010/02272-0, 2014/05214-1, and 2017/11540-7, CNPq (PQ-1D) grant 301617/2016-3, and Swedish Foundation for International Cooperation in Research and Higher Education and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior grants.
Duality of Interest. P.-O.B. is Chief Executive Officer of the biotech company Biocrine AB, and C.J.B. and M.K. are consultants with this company. No other potential conflicts of interest relevant to this article were reported.
Author Contributions. C.J.B. and P.-O.B. conceived the project. C.J.B., F.H.G.T., S.d.S.F., R.S., T.S.A., M.K., J.O.M., E.D., and P.-O.B. designed experiments. C.J.B., F.H.G.T., S.d.S.F., R.S., T.S.A., M.K., S.S.R., and E.D. carried out experiments. C.J.B., F.H.G.T., S.d.S.F., R.S., T.S.A., M.K., and E.D. analyzed data. C.J.B. and P.-O.B. wrote the article with input from all authors. C.J.B. and P.-O.B. are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.