The magnesium transporter NIPAL1 is a pancreatic islet–expressed protein that conditionally impacts insulin secretion

Abstract

Type 2 diabetes is a chronic metabolic disease characterized by pancreatic β-cell dysfunction and peripheral insulin resistance. Among individuals with type 2 diabetes, ∼30% exhibit hypomagnesemia. Hypomagnesemia has been linked to insulin resistance through reduced tyrosine kinase activity of the insulin receptor; however, its impact on pancreatic β-cell function is unknown. In this study, through analysis of several single-cell RNA-sequencing data sets in tandem with quantitative PCR validation in both murine and human islets, we identified NIPAL1 (NIPA-like domain containing 1), encoding a magnesium influx transporter, as an islet-enriched gene. A series of immunofluorescence experiments confirmed NIPAL1's magnesium-dependent expression and that it specifically localizes to the Golgi in Min6-K8 cells, a pancreatic β-cell–like cell line (mouse insulinoma 6 clone K8). Under varying magnesium concentrations, NIPAL1 knockdown decreased both basal insulin secretion and total insulin content; in contrast, its overexpression increased total insulin content. Although the expression, distribution, and magnesium responsiveness of NIPAL1 in α-TC6 glucagonoma cells (a pancreatic α-cell line) were similar to the observations in Min6-K8 cells, no effect was observed on glucagon secretion in α-TC6 cells under the conditions studied. Overall, these results suggest that NIPAL1 expression is regulated by extracellular magnesium and that down-regulation of this transporter decreases glucose-stimulated insulin secretion and intracellular insulin content, particularly under conditions of hypomagnesemia.

Type 2 diabetes (T2D) is a chronic metabolic disease characterized by pancreatic β-cell dysfunction and peripheral insulin resistance (1). Insulin secretion from the β-cell associated with glucose levels is highly dependent on a coordinated pattern of ionic flux events, with the roles of important cations K⁺ and Ca²⁺ having been well-established in this process (2–4). However, the relevance of Mg²⁺ to islet secretion and diabetes remains relatively underexplored. This is due in part to the wide variety of roles played by magnesium: it has been implicated in membrane stability, cell growth, energy metabolism, and enzyme activity caused by its role as an ATP-carrying co-factor (4–6). A host of studies have shown that substantial fluxes of Mg²⁺ occur across the plasma membrane under appropriate hormonal or metabolic stimuli (7, 8). However, cellular buffering systems in several models have been found to maintain relatively constant intracellular magnesium levels despite major changes in total magnesium content (9). Exchange is regulated via several Na⁺-dependent and independent pathways (9–11); notable transporters include TRPM6/7 (12, 13), MAGT1 (14), the SLC41 (15) and CNNM (16) families, and Golgi-specific MMGT1/2 (17). Of these, TRPM7 and CNNM4 have been found to be the most highly expressed in pancreatic β-cells (18). However, despite transporter activity, ATP complexing remains the most important regulator of free Mg²⁺ (19). It is also worth noting that unlike free Ca²⁺, cytosolic Mg²⁺ is not likely to act as an intracellular secondary messenger (7).

Overall, the importance of free Mg²⁺ in islets remains unclear. An estimated 30% of patients with T2D are believed to suffer from hypomagnesemia (i.e. <0.7 mmol/liter Mg²⁺ in plasma) (20–23). This strong association makes hypomagnesemia a potential risk factor for the development and progression of T2D (18, 24, 25). This may be due in part to the importance of Mg²⁺ in regulating insulin receptor tyrosine kinase activity (22, 26–28), and it has been posited that Mg²⁺ levels are directly linked to insulin resistance and dyslipidemia (20). Two independent meta-analyses have suggested that correcting for hypomagnesemia with long-term oral Mg²⁺ supplementation improves both insulin sensitivity and fasting plasma glucose in T2D patients (29, 30). However, the apparent benefits of added Mg²⁺ may not translate well to islet function.

In vitro studies have demonstrated that pancreatic Mg²⁺ perfusion inhibits both insulin and glucagon secretion, because of magnesium's capacity to disrupt critical Ca²⁺ binding and Ca²⁺ channel activity; this, in turn, alters the sensitivity of these cells to glucose (31–34). The effect of magnesium becomes even more complicated in vivo. One study demonstrated that magnesium-deficient rats develop significantly elevated blood-glucose values versus pair-fed controls (35); however, another study showed that low dietary Mg²⁺ intake was beneficial in ameliorating high-fat diet–induced obesity in mice, marked by decreased fasting serum glucose levels and improved insulin sensitivity (36). The relevance of Mg²⁺ to islet function thus remains poorly understood.

In this study, several recent single-cell RNA-sequencing (scRNA-seq) and whole-islet RNA-seq studies were explored to discover novel islet-specific proteins. Through this process, magnesium transporter NIPAL1 (NIPA-like domain containing 1) was identified in islets and subsequently investigated in vitro to determine its relation to pancreatic β-cell function.

Results

Islet-specific candidate selection: NIPAL1

A sequential discovery process was designed to identify novel gene transcripts in pancreatic islets (i.e. α- and β-cells) and determine their influence in the secretion of insulin, glucagon, or both (Fig. S1). In the candidate selection phase, 915 candidates were independently investigated and ultimately filtered down into a final shortlist of 28 genes (Fig. 1). The specific publications, data sets, and preprocessing and selection strategies used can be found under “Experimental procedures.” The 28 shortlisted genes were investigated in both human and mouse tissues (i.e. liver, kidney, islets, and exocrine pancreas) through quantitative real-time PCR (qPCR) to identify those with significant islet enrichment. The transcripts that demonstrated statistically significant higher expression in mouse islets versus liver, kidney, and pancreas include Nipal1, Syt13, Fam159b, Vat1l, Syt4, Tm4sf4, and Tspan13 (Fig. 2A). In humans, candidates that demonstrated statistically significant higher expression in islets versus liver and pancreas include NIPAL4, IGSF11, NIPAL1, CYSTM1, and BEST3 (Fig. 2B). Of all the genes investigated, magnesium transporter NIPAL1 was the only candidate found to be statistically significant in islets versus all other tissues in both species and was thus selected for further characterization and functional investigation using both mouse pancreatic α and β-cell lines (summarized in Fig. 3A).

Figure 1.

Overview of candidate selection process to find novel candidates of interest in α- and β-cells. Candidates were taken from seven datasets highlighting the most enriched candidates: one highlighting differentially expressed candidates under T2D, one showing differential expression under glucose induction, and one highlighting islet-specific eQTL. All significance thresholds for log₂-transformed expression and FDR were selected by the studies and were provided in the form of prefiltered lists. A combined total of 915 unique candidates met the initial criteria across all sources. This initial pool was investigated for relative α/β-cell specificity, islet specificity, and novelty in islets. Through this process, the pool of 915 candidates was narrowed down to a final shortlist of 28 candidates, all of which progressed to the next phase of transcriptional validation via qPCR. MeSH, medical subject headings; diff. exp., differential expression.

Figure 2.

Shortlisted candidate gene expression was measured in mouse (A) and human (B) tissues via qPCR, with results split into low and high value graphs. In mice, expression was determined in islets, liver, kidney, and exocrine pancreas. In humans, expression was measured in islets, liver, and exocrine pancreas. Raw expression values were normalized to β-actin and then displayed as fold expression over pancreas. The results were then organized by decreasing fold islet expression, after controls PEPCK (liver), OAT3 (kidney), and KCNJ11 (islet). Candidates that were significantly enriched in islets versus all other tissues are indicated (*). NIPAL1 is significant in both species (indicated by arrows). Statistical significance between islets and other tissues was evaluated via one-way analysis of variance and Tukey's honest significant difference (not shown here) (n = 3; values displayed as means ± S.E.).

Figure 3.

A, workflow linking transcriptional validation phase to translational validation phase. B, analysis via qPCR identified substantial Nipal1 expression in Min6-K8 cells, comparable with expression of islet/Min6-enriched potassium channel Kcnj11 (n = 3). C, representative images of immunofluorescent staining of NIPAL1 in Min6-K8 cells, co-stained with 4′,6′-diamino-2-phenylindole (DAPI) and insulin. NIPAL1 staining was observed to have a punctate (likely vesicular) presence throughout much of the cytosol but is most concentrated in regions adjacent to the nuclei, as denoted by arrows. D, subcellular localization studies with GM-130 (Golgi) staining identified that this concentrated region was confined to the Golgi apparatus. However, NIPAL1 still shows distribution outside of the Golgi. E, PCC analysis was performed to quantify co-localization between NIPAL1 and other markers (n = 30–45 cells across three independent studies; values displayed as means ± S.E.). All IF experiments had n = 3.

NIPAL1 is concentrated in the Golgi of Min6-K8 cells

Following the identification of NIPAL1 in both murine and human islets, its expression was similarly assessed via qPCR in Min6-K8 cells (i.e. glucose-responsive murine pancreatic β-cell–like cell line). Nipal1 expression in Min6-K8 cells was comparable with the functionally important gene Kcnj11, encoding a K_ATP channel that is enriched in β-cells (Fig. 3B). In immunofluorescence (IF) studies, abundant NIPAL1 protein expression was also observed in Min6-K8, with a punctate cytosolic distribution suggesting a vesicular presence (Fig. 3C). Interestingly, NIPAL1 appeared to be concentrated in an area directly adjacent to the nucleus in all cells. Co-localization studies were thus performed using an antibody against GM-130 (Golgi matrix protein 130) (Fig. 3D). A strong overlap between the concentrated region of NIPAL1 and GM-130 indicated that NIPAL1 was indeed most abundant in the Golgi. The co-localization of NIPAL1 was quantified using Pearson's correlation co-efficient (PCC) analysis. Relative co-localization was found to be significantly higher with both insulin granules (0.38 ± 0.04) and the Golgi (0.44 ± 0.04) compared with the nucleus (Fig. 3E).

Mg²⁺ enhances NIPAL1 expression in Min6-K8 cells

Because NIPAL1 is a magnesium transporter, experiments were conducted to determine whether NIPAL1 expression in Min6-K8 cells is directly responsive to variations in extracellular magnesium content using qPCR (Fig. 4A). The results showed that Nipal1 gene expression is roughly halved under magnesium-free (0.0 mm) conditions relative to standard (0.8 mm) conditions, with no substantial change under high magnesium (5.0 mm) conditions. These findings thus suggest that Nipal1 expression depends on the presence of Mg²⁺. IF studies were performed on cells cultured under similar conditions (Fig. 4B). Qualitatively, there was a visible increase in staining intensity under high Mg²⁺ (5.0 mm) within a small area inside each cell (likely the Golgi). Mean NIPAL1 pixel intensity for each condition was thus quantified via ImageJ software analysis (Fig. 4C). However, the difference in mean intensity observed between 5.0 and 0.8 mm Mg²⁺ did not reach significance (p = 0.078).

Figure 4.

Investigation of effect of magnesium on NIPAL1 expression. A, following 24-h culture in standard medium, Min6-K8 cells were cultured in medium with varying Mg²⁺ concentrations (0.0, 0.8, and 5.0 mm), and Nipal1 gene expression was measured via qPCR. Expression was halved in magnesium-free conditions (0.0 mm) relative to standard conditions (0.8 mm), with no change under high conditions (5.0 mm) (n = 3; values displayed as means ± S.E.) *, p < 0.05. B, IF studies of NIPAL1 expression (×63) were performed under similar conditions. Laser intensities and detection thresholds were kept consistent. NIPAL1 intensity appeared stronger under high Mg²⁺ conditions (5.0 mm), especially adjacent to nuclei (previously identified as Golgi). C, average NIPAL1 pixel intensity analysis of defined cell regions was measured to quantify this. Although NIPAL1 intensity appeared to be elevated under 5.0 mm Mg²⁺ relative to 0.8 mm Mg²⁺, this did not reach significance (p = 0.078) (n = 30–45 cells across three independent studies; values displayed as means ± S.E.). *, p < 0.05.

NIPAL1 is important for Mg²⁺-dependent insulin secretion in Min6-K8 cells

To characterize the role of NIPAL1 in pancreatic β-cell function, glucose-stimulated insulin secretion (GSIS) assays were conducted in transfected Min6-K8 cells (i.e. NIPAL1 overexpression and knockdown) under both regular and hypo/hypermagnesemic conditions. NIPAL1 overexpression was confirmed via IF using a FLAG antibody (Fig. 5A), with mean pixel intensity of both NIPAL1 and FLAG being significantly elevated in cells transfected with NIPAL1 cDNA (Fig. 5B) relative to control. In the case of NIPAL1 knockdown using siRNA, qPCR was carried out to evaluate the efficacy of silencing. Results showed a consistent decrease ∼70% of expression (or <0.3-fold expression) relative to scramble control (Fig. 5C). Attempts were taken to further validate these observations using immune blots; however, the NIPAL1 antibody (used here) was found not suitable for this application (not shown).

Figure 5.

Validation of NIPAL1 overexpression and knockdown. A, IF studies of NIPAL1 in Min6-K8 with FLAG epitope antibody were used to confirm overexpression following 24-h transfection of NIPAL1 cDNA-FLAG and empty vectors via Lipofectamine (n = 3). B, mean pixel intensity quantification confirmed higher intensity for NIPAL1 and FLAG staining under overexpression conditions. C, Min6-K8 cells were transfected with either nontargeting scramble siRNA or NIPAL1 siRNA via Lipofectamine, and expression was measured via qPCR; NIPAL1 siRNA fold expression showed ∼70% decreased expression compared with scramble (n = 3).

In GSIS under standard culture conditions (0.8 mm Mg²⁺), neither overexpression nor knockdown of NIPAL1 in Min6-K8 cells showed significant alteration in overall insulin secretion (Fig. 6, A and B). However, it was hypothesized that any observable effect might be magnesium-dependent. In subsequent experiments, the role of NIPAL1 was evaluated under magnesium-induced stress using a range of magnesium concentrations (0.0, 0.4, 0.8, 5.0, and 10.0 mm Mg²⁺) with 2% FBS (to minimize additional magnesium content) as displayed in Fig. 6C, following confirmation that treatments did not impact cell viability (Fig. 6, I and J); GSIS was performed in both untransfected (Fig. 6D) and siRNA-transfected Min6-K8 cells (Fig. 6, E and F). In untransfected cells, a positive correlation between Mg²⁺ concentration and insulin secretion was readily apparent. (Fig. 6D). This suggests an importance for Mg²⁺ in insulin secretion in Min6-K8 cells. In the case of NIPAL1 knockdown, basal insulin secretion (0 mm glucose) was significantly decreased under almost every magnesium concentration (Fig. 6E). However, these differences were abolished under treatment with 16.7 mm glucose and KCl (Fig. 6F). In addition, knockdown of NIPAL1 significantly reduced total insulin content at lower Mg²⁺ concentrations (0.4 and 0.8 mm Mg²⁺) but not at higher concentrations (5.0 and 10.0 mm Mg²⁺) (Fig. 6G). In contrast, overexpression of NIPAL1 significantly increased total insulin content under almost every concentration (Fig. 6H). These observations thus indicate a role for NIPAL1 in insulin secretion and storage in a magnesium-dependent manner.

Figure 6.

GSIS experiments in Min6-K8. A, 24-h transfection with empty vector or NIPAL1 cDNA was performed via Lipofectamine. Following 1-h incubation in glucose-free KRB medium, GSIS was performed on cells transfected with empty vector or NIPAL1 cDNA using basal (0 mm), high-glucose (16.7 mm), and high-glucose + KCl conditions. Insulin samples were collected after 20-min incubations at 37 °C and measured via HTRF. The results were normalized to DNA. No significant difference was observed between treatment groups under any glucose condition. B, 24-h transfection with scramble or NIPAL1 siRNA was performed via Lipofectamine, followed by GSIS as described before, although no significant differences were observed. C, Min6-K8 cells were cultured under various Mg²⁺ concentrations. D, GSIS was performed. A clear trend was observed between increasing Mg²⁺ concentration and insulin secretion, with significant increases in basal and KCl-induced secretion under 5.0 and 10.0 mm Mg²⁺ treatments. E, cells cultured under the same conditions were transfected with nontargeting scramble siRNA or NIPAL1 siRNA. NIPAL1 knockdown was found to significantly decrease basal secretion under all Mg²⁺ concentrations except 5.0 mm. F, secretion under NIPAL1 knockdown appeared to be rescued with KCl induction. G, NIPAL1 knockdown was also found to decrease total insulin under 0.4 and 0.8 mm Mg²⁺ conditions, but this recovered under hypermagnesemic 5.0 and 10.0 mm Mg²⁺. H, total insulin was also investigated in cells cultured under the same conditions and transfected with either empty vector control or NIPAL1 cDNA. Total insulin was found to be significantly increased under 0.4, 0.8, 5.0, and 10.0 mm Mg²⁺ conditions. I and J, trypan blue assays confirmed that transfection treatments under varying magnesium conditions with 2% FBS did not impact cell viability or proliferation relative to standard (STD) conditions. §, standard conditions = 0.8 mm Mg²⁺ and 10% FBS. All experiments had n = 3 (means ± S.E.). *, p < 0.05.

NIPAL1 is present in the glucagon-producing α-TC cells

The experiments conducted in Min6-K8 cells were additionally carried out in α-TC6 cells (i.e. murine α-cell–like cell line, α-TC1 clone 6). Nipal1 gene expression was observed in α-TC6 as assessed via qPCR, although at notably lower levels than Kcnj11 (Fig. 7A). Follow-up IF experiments confirmed healthy expression at the protein level (Fig. 7B). As in Min6-K8, NIPAL1 protein distribution appeared cytosolic and punctate in α-TC cells, with concentration in the Golgi (Fig. 7C). PCC analysis of NIPAL1 with glucagon was notable (0.60 ± 0.09), potentially indicating a higher degree of co-localization (Fig. 7D). IF experiments of α-TC cells cultured under varying magnesium conditions (Fig. 7E) showed stronger NIPAL1 staining under high Mg²⁺ (5.0 mm), which was confirmed to be significantly increased under mean pixel intensity quantification relative to standard Mg²⁺ (0.8 mm) (Fig. 7F). In similar fashion to previous GSIS experiments, α-TC6 cells were transfected (i.e. NIPAL1 overexpression and knockdown), and both basal glucagon secretion (Fig. 7G) and total glucagon were measured (Fig. 7H). No significant differences were observed between any of the treatment groups with respect to glucagon secretion or content.

Figure 7.

A collection of NIPAL1 experiments in α-TC6 cells. A, qPCR analysis identified substantial Nipal1 expression in α-TC6 cells, although at lower levels than islet/α-TC6–enriched potassium channel KCNJ11. B, α-TC6 cells were co-stained with NIPAL1, 4′,6′-diamino-2-phenylindole (DAPI), and glucagon/insulin. As in Min6-K8, NIPAL1 was observed to have a punctate (likely vesicular) presence throughout much of the cytosol. C, the concentrated regions denoted by arrows were confirmed to be localized to the Golgi when co-stained with a GM-130 marker. D, PCC analysis was performed to quantify co-localization between NIPAL1 and other markers. PCC = +1 (perfect correlation); PCC = 0 (absence of relationship). E, IF studies of cells cultured under varying Mg²⁺ conditions with consistent imaging conditions. F, NIPAL1 intensity appeared stronger under high Mg²⁺ conditions (5.0 mm), as confirmed via the significantly higher pixel intensity analysis of NIPAL1 staining in defined cell regions (n = 30–45 cells across three independent studies). G, following 1-h incubation in glucose-free KRB medium, the cells were transfected with either empty vector control, NIPAL1 cDNA, scramble siRNA control, or NIPAL1 siRNA, and basal glucagon samples were taken following 20-min incubation. H, total glucagon samples were also collected. No significant differences were observed in glucagon secretion or content. All results are displayed as means ± S.E. *, p < 0.05 (n = 3).

NIPAL1's interactome

An interactome was generated based on NIPAL1's known interactions with other proteins. A total of 10 proteins are known to interact with NIPAL1, as follows: MMGT1, MAGT1, PPM1G, SLC35C1, SLC35C2, SLC35B3, TMEM144, FAM35A, NFXL1, and CNGA1 (Fig. 8). Many of these proteins are integrated into membranes (both of the cell and various organelles), with notable representation in the Golgi apparatus and endoplasmic reticulum. (Fig. S3). However, NIPAL1's location is not defined here. In this study, we have shown that NIPAL1 is concentrated in the Golgi. The major suggested biological function of NIPAL1 is transmembrane transport of ions and biomolecules which include Mg²⁺, carbohydrate derivatives, nucleotide sugars, inorganic cations, metals, and carbohydrates. In addition, NIPAL1 may play a role in glycosylation and in the regulation of the Notch signaling pathway (Fig. S3).

Figure 8.

An interactome of the known protein-protein interactions of NIPAL1, constructed with String 11.0 software on a Homo sapiens background. Clustering analysis was performed with protein-protein interaction enrichment P value < 0.05 and local clustering co-efficient > 0.85.

Discussion

A better understanding of T2D pathophysiology entailing the role of various contributing genes/proteins, especially in pancreatic islets, is essential for ultimately devising novel interventions. As such, omics-based technology such as scRNA-seq holds enormous potential to discover novel disease-related factors. In this study, NIPAL1 was identified as a novel islet-enriched protein via a manually curated data mining approach employing several T2D-related scRNA-seq data sets in combination with qPCR validation in both human and murine islets. To understand the functional roles of NIPAL1 in islets, we first examined its expression in α- and β-cell–like cell lines. The abundant expression of Nipal1 in both α- and β-cell–like cell lines is supported by observations of robust Nipal1 expression in murine α- and β-cells in scRNA-seq studies by Benner et al. (37) and DiGruccio et al. (38). The study of Benner et al. in particular places Nipal1 among its top 200 most enriched murine β-cell candidates. Although the NIPAL1 rankings in the human study by Segerstolpe et al. (39) are unremarkable (12,147th and 13,603rd for β- and α-cells, respectively), it is worth noting that the study by Varshney et al. (40) ranks NIPAL1 within the top decile of human islet specificity when compared with 14 other human tissue types; however, the use of whole-islet RNA-seq prevents the authors from pinpointing specific cell types.

NIPAL1 (NIPA-like domain containing 1) is a magnesium transporter gene, also known as NIPA3 (nonimprinted in Prader–Willi/Angelman syndrome 3) on account of its structural similarity to paralogs NIPA1 and NIPA2. Although both NIPA1 and NIPA2 have been implicated in neurological disorders (41, 42), a role for NIPAL1/NIPA3 as a disease-causing factor remains unestablished. The human Genotype-Tissue Expression database indicates that NIPAL1 has significant levels of expression in the skin, colon, bladder, and vagina, with more moderate expression in the small intestine and liver; whereas islet data are unavailable on the platform, whole pancreas expression is notably absent (RRID:SCR_013042, accessed June 16, 2019). In mice, the GeneAtlas MOE430 data set on the BioGPS database (ID 152519, accessed April 7, 2020) indicates high Min6 expression and notable expression in thymocytes, stomach, and intestine (islet data are unavailable, but whole pancreas expression is also absent). At present, there are only a handful of publications available that directly reference NIPAL1, none of which have studied it in the context of islets or diabetes.

In the present study, NIPAL1 protein was found to have substantial Golgi co-localization and distribution with both insulin and glucagon granules (Figs. 3E and 7D). It also demonstrated magnesium-dependent expression in Min6-K8 cells (Fig. 4A). Insulin secretion experiments revealed that magnesium effectively influenced insulin secretion (Fig. 6D). Importantly, knockdown of NIPAL1 expression under hypo- and hypermagnesemic conditions was found to decrease secretion (Fig. 6E), and total insulin content was reduced (Fig. 6G). In contrast, insulin content was increased with NIPAL1 overexpression (Fig. 6H). Although the impact of solely magnesium on total insulin was not measured directly, the recovery of total insulin content under hypermagnesemic (i.e. 5.0 and 10.0 mm) conditions to control levels despite NIPAL1 knockdown (Fig. 6G) strongly suggests a positive correlation between magnesium and total insulin in Min6-K8 cells. Although interesting observations were made regarding insulin, no effect was ultimately observed on glucagon secretion in α-TC6 cells (Fig. 7, G and H). Overall, NIPAL1 appears to be important in promoting cellular insulin content and stabilizing the insulin secretory response in the face of abnormal magnesium levels.

A previous study showed that Nipal1 expression was up-regulated in kidney tissue harvested from mice on a magnesium-restricted diet (43). The results suggested that this increased Nipal1 expression helps promote Mg²⁺ influx into cells, thereby regulating intracellular magnesium homeostasis when extracellular magnesium is scarce. In contrast, we found that NIPAL1 expression was increased with increasing magnesium levels in Min6-K8 and α-TC6. Furthermore, magnesium-dependent pancreatic β-cell function/sensitivity (Fig. 6D) was inhibited under basal conditions with NIPAL1 knockdown (Fig. 6E). As such, these islet-specific results indicate that the Mg²⁺ requirement of different tissues varies widely with their functions. Although higher NIPAL1 expression has the potential to allow for higher Mg²⁺ influx, hypermagnesemia, to the best of our knowledge, is not associated with T2D.

The link between Mg²⁺ sensitivity and NIPAL1 expression may potentially be explained by the predicted interaction of NIPAL1 with PPM1G, as displayed in the interactome (Fig. 8). PPM1G is a magnesium-dependent protein phosphatase that has been shown to form protein complexes involved in transcriptional regulation (44). When this is considered alongside its magnesium-dependent function, it is possible then that PPM1G acts as an intermediary sensor between magnesium levels and NIPAL1 expression.

The detection of NIPAL1 in the Golgi is also of interest. The Golgi is one of the few organelles known to possess its own local magnesium transporters (43), highlighting the importance of Mg²⁺ to Golgi function. Given this, NIPAL1 could play a role in facilitating magnesium entry into the Golgi in a manner that directly impacts the ionic flux governing the trafficking of insulin vesicles within the organelle. A potential driving force may be magnesium's role as a primary ATP-binding co-factor (Mg-ATP) involved in facilitating phosphate transfer reactions (5, 6) that are critical for fueling secretory granule movements (45, 46) and both glycolysis and the citric acid cycle (47, 48). Both Golgi and mitochondrial functions are highly dependent on the availability of ATP delivered by bound magnesium. It is thus possible that higher Mg²⁺ increases Mg-ATP complex formation, which subsequently increases mitochondrial output, insulin granule maturation, and exocytosis. Because NIPAL1 overexpression is specifically observed to increase total insulin content (and knockdown decreases it), greater delivery of magnesium through NIPAL1 may bolster the cell's insulin granule reserve pool, which in turn translates proportionally to greater secretion. Alternatively, increased Mg-ATP formation may influence the ATP:ADP ratio regulating the closure of K_ATP channels, whereas extracellular free Mg²⁺ has also been proposed to regulate K_ATP channels (49). Fluctuation in extracellular Mg²⁺ levels may also have a substantial effect on membrane potential and by extension the likelihood of triggering secretion.

In summary, this study has demonstrated the importance of magnesium in the preservation of β-cell function as regulated by the magnesium influx transporter NIPAL1, which has been newly identified in islets. Experiments demonstrated that NIPAL1 expression correlates with magnesium and that it is functionally important in permitting magnesium-dependent increases in insulin production and secretion from Min6-K8 cells. However, it is important to emphasize that the importance of NIPAL1 likely only emerges under abnormal conditions such as hypomagnesemia and not under healthy physiological conditions. It is also important in future to further validate the immunofluorescent study results shown here with Western blotting and, particularly in cells where NIPAL1 is ablated, to eliminate any doubt regarding the antibody's specificity. Furthermore, it would be worthwhile to determine whether NIPAL1 influences expression of insulin and other key β-cell genes. In one study, acute lowering of extracellular Mg²⁺ levels in mouse islets and rat INS-1 cells was found not to affect expression of key glucose metabolism genes (e.g. Gck, Abcc8); knockdown of magnesium transporter Trpm7 also had no effect on expression of these genes nor that of the Ins1 gene (50). The same case would thus be expected for NIPAL1. As a final point, further experiments are needed to determine whether magnesium and NIPAL1 are related to glucagon secretion. Overall, these results are pertinent to the emerging role of magnesium in the context of diabetes and dietary metabolism. Further studies investigating NIPAL1 in human pancreatic β-cells may serve to elucidate the role of magnesium in the context of islet health and secretion.

Experimental procedures

Data set selection and basic preprocessing for candidate selection

Several scRNA-seq and traditional RNA-seq studies provided the data sets used for candidate selection in this study (an unpublished scRNA-seq data set from the Wheeler laboratory was additionally included). All data used were acquired in supplementary data sets with results preprocessed or filtered to varying degrees. Specific details can be found in Table 1.

Table 1.

Selected data sets and associated preprocessing for candidate selection

Publication	Application	Islet species	Data set type	Specific data used from Publication	Preprocessing by author	Preprocessing in this study
Benner et al. (37)	scRNA-seq	Human and mouse	Most enriched genes	Additional file 12	Mean RPKM values of genes in mouse and human α/β cells	Sorted candidates by decreasing mean RPKM and took top 200 α/β genes for both species
Segerstolpe et al. (39)	scRNA-seq	Human	Most enriched genes + differentially expressed under T2D	Tables S1 and S2 (enriched genes)	All tables had mean/median log2RPKM ≥ 1, FDR adjusted p < 0.01	Top 200 α/β genes used from enriched tables; α/β tabs of Table S6 used as provided
				Table S6 (differential expression under T2D)
Xin et al. (55)	scRNA-seq	Human and mouse	Most enriched genes	Table S4	Mean RPKM values of genes in mouse and human α/β cells	Sorted candidates by decreasing mean RPKM and took top 200 α/β genes for both species
Li et al. (56)	scRNA-seq	Human	Most enriched genes	Data Set SEV6	Mean RPKM values of genes for islet and exocrine cells	Sorted candidates by decreasing mean RPKM and took top 200 α/β genes
DiGruccio et al. (38)	scRNA-seq	Mouse	Most enriched genes	File in Gene Expression Omnibus (GSE80673)	Mean log2RPKM values for FACS-sorted cell clusters provided	Cell clusters labeled (matched based on paper details) and sorted in decreasing order of mean expression for α/β (top 200 taken from each group)
Lawlor et al. (57)	scRNA-seq	Human	Most enriched genes	Table S6, NonT2D.Beta.versusAlpha tab	Mean log2CPM values of genes for α/β cells	Sorted candidates by decreasing mean log2CPM and took top 200 α/β genes
Ottosson-Laakso et al. (58)	Whole-islet RNA-seq	Human	Differentially expression under glucose exposure	Tables S4, S5, and S8	FDR-corrected p value [q] < 0.05, mean log2CPM provided	Only included initial candidates with log2-transformed fold change ≥0.5 or ≤−0.5
Varshney et al. (40)	Whole-islet RNA-seq	Human (islets + 14 other tissues)	Islet-specific eQTL	Interactive platform: RRID:SCR_018692	Only islet eQTL variants (3314); classified into deciles of islet specificity + epigenetic state	Top decile of islet specificity, default parameters for strong transcription, active TSS, ATAC-seq “footprint”

TSS, Transcription start site; ATAC-seq, Assay for Transposase-Accessible Chromatin using sequencing.

All 915 initially selected candidates were filtered into a final shortlist if they met the following specific criteria: (1) Strong α/β-cell specificity, where mean expression of a candidate was at least double that of other islet (or exocrine) cell types in at least one of the scRNA-seq data sets in which it was identified (where applicable); candidates identified in a whole-islet RNA-seq data set were cross-referenced in scRNA-seq data sets (wherever possible). (2) Strong islet specificity, assessed using the GeneAtlas U133A (gcrma) human tissue data set (79 tissues total) on the BioGPS database. The data set easily distinguishes candidates with mean expression in human islets that is at least triple the median of all tissues, and this was selected as the cut-off point for this criterion. If unavailable on BioGPS, the Genotype-Tissue Expression and ProteinAtlas databases were used to ensure that candidates were not ubiquitously expressed. (3) Novelty in islets was assessed via a thorough PubMed search employing a broad range of medical subject headings terms, e.g. “NIPAL1 AND (islet OR diabetes OR pancreas OR pancreatic OR insulin OR glucagon OR secretion).” Candidates with any detailed/focused studies in an islet/diabetes context were deemed not to be novel and excluded. Candidates with no publications were still considered.

Tissue sample acquisition and preparation

Healthy male CD1 mice (8–12 weeks old) were obtained from Charles River and fed a standard chow diet. The mice were housed in the Division of Comparative Medicine facility, University of Toronto. All animal procedures and maintenance were conducted in compliance with protocols approved by the Animal Care Committee at the University of Toronto and the guidelines of the Canadian Council of Animal Care. Liver, kidney, and pancreatic islet and exocrine samples were isolated as described previously (51). If not used immediately, samples were stored at −80 °C. Human samples from six nondiabetic donors (three male and three female; average age, 49.8 ± 16.4 years; average body mass index, 29.8 ± 4.5) were received from either the IsletCore laboratory at the Alberta Diabetes Institute (University of Alberta) or through the Islet Program organized by the Centre for Islet Research, University Health Network. Sample procurement was approved by either the University of Alberta's Human Research Ethics Board or the Research Ethics Boards of the University Health Network (University of Toronto). All human aspects of this study abide by the Declaration of Helsinki principles. Both islets and exocrine tissue were handpicked. The samples were cultured or stored in the same manner as mouse samples.

Quantitative PCR

Total RNA was isolated and spectrophotometrically analyzed, reverse transcribed into cDNA, and measured via qPCR analysis as described previously (52). The data were normalized to β-actin mRNA. Primers targeting exon–exon junctions were designed using the NCBI Primer-BLAST tool.

Cell culture and transfection

Min6-K8 cells (a gift from S. Seino [Kobe University, Kobe, Japan] and J. Miyazaki [Osaka University, Suita, Osaka, Japan]) were cultured in high-glucose DMEM (Millipore-Sigma, D6429) supplemented with 10% FBS, 100 units/ml P/S, and 1.75 μl of β-mercaptoethanol. α-TC6 cells were cultured in low-glucose DMEM (Millipore-Sigma, D5921) supplemented with 10% FBS, 100 units/ml P/S, 0.02% BSA, and 15 mm HEPES. The cell lines were maintained at 37 °C and 5% CO₂. For varying magnesium experiments, MgSO₄ was dissolved in custom magnesium-free high-glucose DMEM (Wisent) supplemented with 2% FBS (to minimize magnesium content) and 100 units/ml P/S; medium was subsequently filter-sterilized. Following seeding and 24-h culture in standard medium, the cells were washed thrice with magnesium-free PBS and replacement with special magnesium medium was done concurrently with transfection. Transfection was performed for 24 h with Lipofectamine 2000 (Invitrogen, 11668030) according to the manufacturer's instructions. For overexpression, the cells were transfected with empty vector or mouse pCMV6-Nipal1-Myc-DDK (Origene, MR224805). For knockdown, the cells were transfected with either siGENOME nontargeting siRNA pool 1 or siGENOME mouse Nipal1 (Dharmacon, M-054062-01-0005). Cell viability and count was measured using trypan blue and a Countess II automated cell counter (Thermo Fisher, AMQAX1000) according to the manufacturer's instructions.

Immunofluorescence (IF) studies

IF studies in cell lines were performed as described previously (52). The following primary antibodies were used where applicable: mouse anti-glucagon (1:200, Abcam, ab133195), guinea pig anti-insulin (1:100, Dako, A0564), rabbit anti-NIPAL1 (1:200, Abcam, ab17028), mouse anti-FLAG (1:1000, Origene, TA50011), mouse anti-GM-130 (1:1000, BD, 610823, courtesy of Dr. Sergio Grinstein). The secondary antibodies were goat anti-mouse Alexa Fluor^®555 (Abcam, ab150114), donkey anti-rabbit Cy5 (Jackson Immunoresearch, 711-175-152), and donkey anti-guinea pig Alexa Fluor^®488 (Jackson ImmunoResearch, 706-545-148). The images were acquired on a Zeiss LSM 880 Elyra confocal microscope, and the images were processed in ZEN 2.6 blue edition software. Co-localization analysis of images was performed using the Coloc2 feature of Fiji ImageJ (53) software with 100 Costes randomizations. Relative NIPAL1 intensity was determined by thresholding images to create regions of interest confined to stained cells and then measuring mean pixel intensity of NIPAL1 staining in cells across multiple images (30–50 cells/treatment group).

Glucose-stimulated secretion assay (GSIS) and homogenous time-resolved fluorescence (HTRF)

GSIS was performed on Min6-K8 cells as described previously (54). The insulin samples were stored at −20 °C until HTRF analysis. DNA and total insulin were extracted via acid ethanol. HTRF assays were performed according to the manufacturer's instructions (Cisbio, 62IN2PEG) and read on a PHERAstar plate reader (BMG Labtech). Insulin levels were normalized to DNA.

Glucagon secretion assay and glucagon ELISA

Glucagon samples were collected from α-TC6 cells in similar fashion to GSIS protocols, although without the addition of glucose or KCl; the samples were instead collected without anything added (i.e. basal). Collection media included mammalian protease inhibitor to preserve samples. Total glucagon was collected in the same manner as total insulin. Glucagon was measured via an ELISA kit (Sigma, RAB0202) according to the manufacturer's instructions, and the results were normalized to DNA.

Protein–protein interactome

The NIPAL1 interactome was constructed using String 11.0 software (RRID:SCR_005223) on a Homo sapiens background. To identify the interacting proteins, a clustering analysis was conducted by keeping protein–protein interaction enrichment p values < 0.05 and average local clustering co-efficient > 0.85. All other analyses such as biological functions (GO), molecular functions (GO), and cellular location (GO) were carried out by setting the false discovery rate (FDR) at <0.05.

Statistical analyses

For qPCR results, one-way analysis of variance and post hoc Tukey's honest significant difference tests were performed to determine statistical significance between islet expression and expression in other tissues. Student's t test was employed for all other statistical analyses. *, p < 0.05 was considered statistically significant.

Data availability

All data are available within the manuscript.