Glucose Regulates Microtubule Disassembly and the Dose of Insulin Secretion via Tau Phosphorylation

Pancreatic β-cells are pivotal to glucose homeostasis by secreting insulin. Insufficient insulin secretion causes diabetes, a disease characterized by prolonged high blood glucose and related morbidity. Too much insulin secretion results in hyperinsulinemic hypoglycemia that causes brain damage. Thus, tightly regulating insulin secretion is critical for normal physiology and health.

Each β-cell contains >10,000 insulin secretory vesicles that comprise two pools (1), the readily releasable pool (RRP), accounting for 1%–5% of the total vesicles, and the reserve pool (RP). RRP contains vesicles that are docked onto the plasma membrane and are primed for release, while the RP contains vesicles that need to be transported for docking and priming (2). Upon high glucose induction, some vesicles in both the RRP and RP are released within the first phase of secretion, while some in the RP are mobilized and released within the second prolonged phase (3,4). Intriguingly, each glucose stimulation only induces the release of 1%–5% of total insulin, a property that is necessary for sustainable secretion in response to repeated stimulation. Thus, a challenge for β-cells is to properly allocate the vesicles into the RRP and RP while timely immobilizing the reserved vesicles in response to stimulatory signals.

Regulation of the number of RRP vesicles involves their transportation and plasma-membrane docking (4,5). Vesicles originating from the trans-Golgi network are located in the β-cell interior and transported to locations of their temporary intracellular storage, release, or degradation (6). Insulin-vesicle movement is aided by the cytoskeleton (actin and microtubules) and motor proteins (myosin 5a, kinesins, and dynein) (7–10). Intriguingly, both actin and microtubules displayed dual functions in glucose-stimulated insulin secretion (GSIS). The actin cytoskeleton, F-actin, and myosin 5a facilitate insulin vesicle movement to the plasma membrane for stimulated release (8); yet, cortical F-actin disassembly enhances insulin secretion (10). Regarding the microtubules, some studies suggest that microtubules are required for insulin secretion (11,12). Paradoxically, microtubule-stabilizing agents inhibit insulin secretion, which underscores the complex roles of microtubules in β-cells (13,14).

We recently reported that the majority of microtubules in β-cells derive from the Golgi apparatus instead of the centrosome (14) and form a densely intertwined meshwork without obvious directionality (15,16). Consequently, long-distance directional insulin vesicle movement in β-cells is relatively rare (14,15,17). We further discovered that a key role of microtubules, especially for those close to the plasma membrane, is to move vesicles away from the plasma membrane. Thus, chemically disrupting the microtubule networks acutely increases GSIS in both isolated islets and in mice (14,18). Importantly, high glucose stimulates the dramatic rearrangement of microtubules, including a significant reduction of long-lived microtubules (14) and new microtubule growth mediated by cAMP/EPAC2 signaling (15,19). The mechanism of glucose-induced microtubule disassembly in β-cells is unknown.

Microtubule-associated protein tau is important in neuronal development and function. Upon association with microtubules, tau can regulate motor protein movement (20) and microtubule life span (21). The tau-microtubule association is inhibited by phosphorylation (22). Recent studies have linked dysregulated tau level to impaired insulin secretion and glycemic control. Puzzlingly, both tau overexpression in rat insulinoma cells (23) and tau knockout in mice compromise GSIS (24). These emerging yet contradicting findings indicate the importance of tau in regulating β-cell functions but raise questions on how tau influences GSIS.

In this study, we used real-time imaging to directly assay the microtubule disassembly in live mouse β-cells in islets. Our findings support a model where high glucose enhances microtubule disassembly in cell periphery to enhance GSIS via tau phosphorylation.

Mouse usage followed protocols approved by the Vanderbilt University Institutional Animal Care and Use Committee for G.G. and I.K. Mice were euthanized by isoflurane inhalation. Wild-type CD-1 (ICR) mice were from Charles River Laboratories (Wilmington, MA). Tau^−/− (B6.129X1-Mapt^tm1Hnd/J) and C57BL/6J mice were from The Jackson Laboratory (Bar Harbor, ME). Ins2^H2B.Apple mice have previously been described (25).

Islets were isolated from 8- to 16-week-old mice (14). Briefly, ∼2 mL of 0.8 mg/mL collagenase P (MilliporeSigma, St. Louis, MI) in Hanks’ balanced salt solution (Corning, Corning, NY) was injected into the pancreas through the common bile duct. The pancreas was digested at 37°C for 20 min. Islets were handpicked into RPMI 1640 media with 11 mmol/L glucose (Gibco, Dublin, Ireland) plus 10% heat-inactivated (HI) FBS (Atlanta Biologicals, Flowery Branch, GA) and cultured at 37°C with 5% CO₂. For MIN6 cells, DMEM with 25 mmol/L glucose, 0.071 mmol/L β-mercaptoethanol (MilliporeSigma), 10% HI FBS, 100 μU/mL penicillin, and 100 μg/mL streptomycin (Gibco) was used.

The tau-targeting shRNA (shTau no. 1, 5′-TTGTGATGGATGTTCCCTAAC-3′, and no. 2, 5′-AATCTTCGACTGGACTCTGTC-3′) plasmid, pLKO.1-shTau-puro, was from the Eli and Edythe L. Broad Institute of MIT and Harvard (Cambridge, MA). The control plasmid, pLKO.1-shCtrl-puro, was from Addgene (cat. no. 1864). pLKO.1-shRNA-Eos-tubulin-puro was constructed by first cloning the Eos-tubulin cDNA from ptdEos2-tubulin-C35 (a gift from Dr. V.I. Gelfand, Northwestern University) into BamHI/NheI sites of pCDH-CMV-MCS-EF1-Puro (System Biosciences, Palo Alto, CA). The Eos-Tubulin cassette was next subcloned into SpeI/BamHI sites of pLKO.1-shRNA-puro. pLKO.1-shRNA-copGFP was constructed by replacing the PGKpr-puro cassette of pLKO.1-shRNA-puro with copGFP from pCDH-CMV-MCS-EF1-copGFP (System Biosciences) using EcoRI/KpnI sites. Lentivirus production and infection followed standard methods (26). For quantitative RT-PCR, total RNA extracted from MIN6 cells was purified using RNA Clean & Concentrator (Zymo Research), converted to cDNA using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA), and quantitative PCR amplified using SsoAdvanced Universal SYBR Green Supermix (Bio-Rad, Hercules, CA) on a Bio-Rad CFX96 Thermal Cycler. Primers used are as follows: Ins1, 5′-GGGACCACAAAGATGCTGTT-3′ and 5′-CAGCAAGCAGGTCATTGTTT-3′; Ins2, 5′-TGAAGTGGAGGACCCACAAG-3′ and GTAGTGGTGGGTCTAGTTGC-3′; Mapt, 5′-CGCTGGGCATGTGACTCAA-3′ and 5′-TTTCTTCTCGTCATTTCCTGTCC-3′; and Gapdh, 5′-AACTTTGGCATTGTGGAAGG-3′ and 5′-GGATGCAGGGATGATGTTCT-3′.

For tau and tubulin staining, islets were extracted with methanol at −20°C for 5 min to remove microtubule-unbound tau and fixed with 4% paraformaldehyde for 1 h. For all other antigens, 4% paraformaldehyde plus 0.1% saponin (MilliporeSigma) was used. Antibodies used are as follows: anti–α-tubulin (cat. no. ab18251; Abcam, Cambridge, U.K.), anti-E-cadherin (610181; BD Biosciences, San Jose, CA), anti-tau (sc-32274; Santa Cruz Biotechnology, Dallas, TX), anti-insulin (A0564; Dako, Santa Clara, CA), anti-proinsulin (GS-9A8-C; Developmental Studies Hybridoma Bank, Iowa City, IA), anti–detyrosinated tubulin (AB3201; MilliporeSigma), anti-glucagon (G2654; MilliporeSigma), anti-GM130 (610823; BD Biosciences, Franklin Lake, NJ), and anti-TurboGFP (PA522688; Thermo Fisher Scientific, Waltham, MA). Secondary antibodies are from Thermo Fisher Scientific. Images were captured using a Nikon Eclipse A1R laser scanning confocal microscope equipped with a CFI Apochromat TIRF 100×/1.45 oil objective, except in Fig. 5B and C, where an EVOS FL microscope (equipped with an LPlan PH2 20×/0.4 objective) was used. Images are quantified using Nikon NIS-Elements. Mean intensity (average fluorescence intensity per pixel after background subtraction) of the epitope was measured from individual islet β-cells outlined by E-cadherin and insulin signals. Background subtraction used the average intensity of a region without cells.

Isolated mouse islets were incubated with 2.8 or 20 mmol/L glucose for 2 h and lysed in Tris-lysis buffer (10 mmol/L Tris-Cl, pH 7.5; 100 mmol/L NaCl; 1% Triton X-100; 10% glycerol; and cOmplete proteinase inhibitor cocktail [Roche, Basel, Switzerland]) on ice. For in vitro dephosphorylation assay, MIN6 cells were lysed with Lambda protein phosphatase reaction buffer (New England Biolabs, Ipswich, MA) plus proteinase inhibitor cocktail and incubated with or without 20 units/μL Lambda protein phosphatase (New England Biolabs) at 30°C for 30 min. Antibodies used are as follows: IRDye 800 goat anti-rabbit IgG (Rockland Immunochemicals, Pottstown, PA) and IRDye 700DX goat anti-mouse IgG (LI-COR, Lincoln, NE). Blots were imaged using the Odyssey CLx imager (LI-COR).

Lentivirus-transduced whole mouse islets were cultured for 5 days on glass-bottom 35-mm dishes (MatTek, Ashland, MA) coated with Matrigel (Corning) plus human extracellular matrix (ECM) (Corning). Eos-tubulin–labeled microtubules were photoconverted within a region of interest (diameter ∼2.1 μm) with a 405-nm laser. Images were taken with 7 z-steps (Δz = 0.6 μm) covering ∼4.5 μm in the z-axis, sufficient to capture most microtubules movement along the z-axis. For quantifying the photoconverted microtubules, the total fluorescence (red) in regions-of-interest of all z-stacks were measured and then subtracted by the cytoplasmic background, measured in areas outside the photoconverted regions in the same cells.

Isolated islets were pretreated with specific inhibitors or 0.05% DMSO (solvent for inhibitors) for 2 h, stimulated with 20 mmol/L glucose plus inhibitors/DMSO for 2 h, and fixed for immunofluorescence (IF) staining. Inhibitors used are AIP-II, MARK Inhibitor 39621, PKI 14-22, DMSO, 2-DG, GKA-50, and MG-132 (MilliporeSigma); AZ191, IC261, and SB216763 (Cayman, Ann Arbor, MI); BIRB796, CYC202, Gö6983 (LC Laboratories, Woburn, MA); CX-4945, JNK-IN-8, and SCH772984 (Selleck Chemicals, Houston, TX); GSK-3β inhibitor VI and XI (Santa Cruz Biotechnology); and Verapamil, 6-Bez-cAMP-AM, and 8-Br-cAMP-AM (Enzo Clinical Labs, Farmingdale, NY).

Isolated mouse islets were dissociated with Accumax (Innovative Cell Technologies, San Diego, CA), mixed with lentivirus in RPMI 1640 media, and loaded to the GravityPLUS hanging drop 96-well microplate (InSphero, Schlieren, Swiss) for 4 days to form pseudoislets, which were transferred to an ECM-coated 96-well plate and cultured for 1 day before analyses. Islet GSIS strictly follow that outlined in (14). ELISA kits were from ALPCO, Salem, NH.

The reagents generated during the current studies are available from the corresponding authors upon reasonable request.

To visualize microtubule disassembly in live β-cells, we labeled microtubules with an Eos-tubulin fusion protein, which emits green fluorescence, with photoconversion to red fluorescence after ultraviolet excitation (27). The method was evaluated in mouse insulinoma cells (MIN6) (Supplementary Fig. 1A–C).

Photoconversion efficiently labeled preexisting microtubules within a region of interest (∼1/20th β-cell volume), which accounts for a small fraction of the total cytosolic tubulin pool and contributes negligible amount of converted Eos-tubulin to newly polymerized microtubules. Free Eos-tubulin dimers that were photoconverted diffuse evenly throughout the cytoplasm within 10 s (Supplementary Fig. 1D) and are excluded after background subtraction. Photobleaching levels were negligible (∼0.5% of the red fluorescence after each imaging [Supplementary Fig. 1E]). The intensity of photoconverted microtubules was reduced after each event of microtubule depolymerization (Supplementary Fig. S1F–1M), indicating that this assay directly detects microtubule disassembly in the cell. Thus, this approach enabled measurement of microtubule disassembly over time by monitoring the decay of red fluorescence without interference from new microtubules polymerized from free tubulin.

We next examined microtubule disassembly in β-cells in mouse islets. Islets were isolated from adult Ins2^Apple mice, which express a histone H2B-Apple fusion protein exclusively in β-cells and were infected with lentivirus carrying Eos-tubulin (Fig. 1A). They were then incubated in 2.8 mmol/L glucose for an hour and switched to 2.8 or 20 mmol/L glucose. After 0, 5, 10, 15, or 25 min of stimulation, individual islet cells were imaged at 10 s and 145 s after photoconversion (Fig. 1B–E and Supplementary Movie 1). Note that the nuclear H2B-Apple fluorescence did not interfere with detection of cytoplasmic Eos-tubulin (Fig. 1C–E).

The ratio of red fluorescence intensity at 145 s to that at 10 s after photoconversion was calculated to evaluate microtubule disassembly (Fig. 1F–J). High glucose (20 mmol/L) induced significantly faster microtubule disassembly compared with that sustained in low glucose (2.8 mmol/L) at all time points after stimulation (Fig. 1J). The detected microtubule disassembly is consistent with a reduction in detyrosinated tubulin (14) (Supplementary Fig. 2A–D), an established marker of long-lived microtubules.

We have previously shown that microtubules near the cell periphery suppress insulin secretion by withdrawing vesicles from the plasma membrane (14). We therefore examined glucose-stimulated microtubule turnover in this region. In high glucose, more microtubules are disassembled within ∼1 µm from the cell boundary than in the cell interior (Fig. 1K–T). Taken together with previous reports, this indicates that high glucose triggers both microtubule growth near the trans-Golgi for new insulin vesicle formation (14,15,19) and peripheral microtubule disassembly to facilitate vesicle secretion.

Glucose stimulates β-cells to secret insulin via enhanced glucose metabolism, membrane depolarization, and subsequent Ca²⁺ influx. To explore the mechanism(s) by which glucose destabilizes peripheral microtubules, we first tested whether it requires enhanced glucose metabolism. Nonmetabolizable 2-deoxy-d-glucose does not induce microtubule turnover (Supplementary Fig. 2E), whereas the glucokinase activator GKA50 induces microtubule turnover in 5.6 mmol/L glucose (Supplementary Fig. 2F), indicating that the increased microtubule dynamics is due to enhanced glucose metabolism.

Because Ca²⁺ is known to destabilize microtubules (28), we tested whether high glucose induces microtubule destabilization via Ca²⁺. Blocking Ca²⁺ influx with verapamil shows no effect on microtubule turnover induced by high glucose, while potassium chloride–induced depolarization does not facilitate microtubule turnover in low glucose (Supplementary Fig. 2G). Similarly, introducing cAMP analogs did not influence microtubule dynamics (Supplementary Fig. 2H). These results suggest that glucose-induced microtubule disassembly is independent of β-cell depolarization and downstream Ca²⁺ signaling. We therefore started to explore the roles of other glucose metabolism–regulated factors in microtubule disassembly.

We focused on tau as a potential regulator based on its microtubule-modulating activities. To examine whether tau plays any role in β-cell microtubule dynamics, we knocked down tau (tau-KD) (>70% efficacy) (Fig. 2A–E) in adult islet cells. This approach avoids any potential β-cell adaptation during postnatal growth in Tau null mice. While microtubule disassembly was facilitated upon tau-KD at low glucose, high-glucose treatment failed to further enhance microtubule turnover (Fig. 2F–L). Consistent with these findings, tau-KD caused a significant reduction of detyrosinated tubulin in low glucose but did not affect its levels in high glucose compared with control cells (Fig. 2I). These results indicate that tau is an essential player in microtubule disassembly stimulated by high glucose.

The ability of tau to bind/modulate microtubules is largely regulated by its phosphorylation (22). We investigated whether this mechanism is conserved in β-cells. We used methanol extraction to efficiently eliminate soluble cytosolic proteins (Supplementary Fig. 3), which allow us to examine the level of microtubule-bound tau. After methanol extraction, high glucose–stimulated β-cells have significantly less tau colocalized with microtubules, suggesting that tau binding to microtubules in β-cells is coupled to glucose stimulation (Fig. 3A–G). Intriguingly, the increased tau-microtubule binding in low glucose is not observed in α-cells (Fig. 3H–K).

We next tested whether high glucose affects tau phosphorylation by gel-shift assay (29). High glucose–treated islet cells show an upshift of tau protein (Fig. 3L), which is eliminated by protein phosphatase treatment (Fig. 3M). Note that the total protein level of tau in high and low glucose is comparable. This result suggests that the reduction of tau seen in IF images after methanol extraction (Fig. 3A and D) is not due to degradation of tau in high glucose. Thus, the combined findings suggest that high glucose enhances tau phosphorylation in β-cells, which impacts tau-microtubule binding.

Several kinases have been shown to phosphorylate tau. High glucose activates six of these kinases, including GSK3 (30), CDK5 (31), p38MAPK (32), ERK (32), PKA (33), and PKC (34). We found that inhibitors of GSK3, CDK5, PKA, and PKC, but not p38MAPK, ERK, or other potential tau kinases, effectively suppress the high glucose–induced microtubule turnover (Fig. 3N and Supplementary Fig. 4A).

We have further focused on GSK3, a major tau kinase in neurons (35). Indeed, inhibition of GSK3 by lithium nearly abolished tau phosphorylation in β-cells (Fig. 3L). In addition, chemical inhibition of GSK3 suppressed glucose-induced microtubule turnover in dose- and tau-dependent manners (Supplementary Fig. 4B and C). Acute GSK3 inhibition suppresses GSIS in a tau-dependent manner as well (Supplementary Fig. 4D). These findings suggest that GSK3 facilitates microtubule dynamics and insulin secretion in high glucose through modulating the phosphorylation states of tau.

While high glucose induces microtubule disassembly preferentially in the cell periphery (Fig. 1L), we also found that in low glucose, microtubule-bound tau is more abundant in the cell periphery than in cell interior (Fig. 4A–E, P, and Q). This biased tau distribution is not observed in high glucose–treated cells (Fig. 4F–J and Q), suggesting that high glucose causes tau dissociation from peripheral microtubules. Bulk inhibition of GSK3 increases the overall tau-microtubule binding in the entire cell (Fig. 4R) but does not restore its spatial localization (Fig. 4K–O and Q), suggesting that regional activity of GSK3 in cells is likely important for proper tau distribution. Such glucose-regulated tau binding to peripheral microtubules may explain their sensitivity to high glucose–induced turnover.

Since our findings support a key role of tau in mediating glucose-induced microtubule disassembly, we determined the effect of tau-KD on insulin secretion. For efficient tau-KD in β-cells, islets were dissociated, infected with shRNA-carrying lentivirus, and allowed to reaggregate into pseudoislets. Tau-KD pseudoislets displayed significantly elevated insulin secretion at basal glucose but reduced secretion at high glucose (Fig. 5A–D), corresponding to reduced insulin content in tau-KD β-cells (Fig. 5E–I).

Impaired GSIS in tau-KD cells may be due to inhibited insulin transcription (23). Yet, MIN6 cells with tau-KD had mildly increased Ins1 and Ins2 mRNA (Supplementary Fig. 5A), which might be a compensatory effect due to reduced total insulin in tau-KD cells (Fig. 5I). Additionally, the level of proinsulin in tau-KD islet β-cells does not differ from that in controls (Supplementary Fig. 5B–K). Thus, tau-KD does not impair proinsulin production.

We further tested whether tau-KD influences the conversion of proinsulin to insulin. We followed the fate of proinsulin when both translation and ER-associated protein degradation were inhibited. In this condition, the reduction of proinsulin over time is mainly due to its processing into insulin. We found no difference in the turnover of proinsulin in Tau^−/− and control islets (Supplementary Fig. 5L). These results indicate that the defective GSIS in tau-KD pseudoislets is not due to reduced insulin biogenesis. Rather, enhanced basal insulin secretion and reduced total insulin are consistent with a model where tau-KD and the consequent uncoupling of microtubule dynamics from glucose stimulation can deplete the insulin storage, a condition we have found in immature β-cells (26).

Lastly, tau-KD islet β-cells have unevenly distributed insulin vesicles in the cytoplasm (Fig. 5G). Specifically, these cells have increased insulin in the cell periphery, whereas they have decreased insulin in the cell interior (Fig. 5J). Thus, we conclude that a function of tau is to reduce the number of insulin vesicles in the cell periphery, likely by modulating the microtubule-assisted vesicle transportation between the cell interior and periphery.

Upon entry into β-cells, glucose was metabolized to trigger Ca²⁺ influx, which is the primary inducer for vesicle–plasma membrane fusion, while promoting insulin biosynthesis and vesicle movement for sustainable insulin secretion (36). A major unknown is why β-cells only secrete a limited number (<5%) of insulin vesicles at a given stimulus. Although most studies agree that vesicle docking onto the plasma membrane near Ca²⁺ entry sites is necessary for stimulated secretion (4,5,37), additional regulatory mechanisms are apparently needed (26). We now show that high-glucose stimulation acutely promotes microtubule disassembly in β-cell periphery, which results in the enrichment of insulin vesicles near the plasma membrane and potentiates secretion.

We have recently reported that high glucose induces microtubule growth from the Golgi (in the cell interior) to facilitate new vesicle biogenesis and replenishment of the stored insulin pool (19). Thus, in the long-term, microtubules are necessary for GSIS. However, the role of microtubules in transporting insulin vesicles throughout the cytoplasm is more complex. Several studies indicate that microtubules and microtubule-dependent motors facilitate the random-yet-rapid movement of vesicles (7,14,17). According to our model, such nondirectional movement has dual functions in positioning insulin vesicles to the cell periphery. One function is the net transport of vesicles from their biogenesis sites, which leads to an even distribution of vesicles throughout the cell. This long-range net transport brings insulin vesicles within the reach of SNARE complex for docking. The second role is to prevent docking by short-range movement of microtubule-dependent vesicles away from the plasma membrane or by rapid movement of vesicles parallel to the membrane along peripheral microtubule bundles (14). This means that, in the short term, microtubules prevent accumulation of excessive vesicles at the cell periphery, thus negatively regulating insulin secretion. Therefore, their depolymerization causes an acute increase of vesicle docking and promotes GSIS.

Importantly, our data support a link among GSK3, the glucose-dependent regulation of microtubule dynamics, and insulin secretion. We show that acute inhibition of GSK3 suppresses both glucose-induced microtubule dynamics and GSIS by regulating tau phosphorylation. These acute effects differ from that of chronic β-cell–specific GSK3 inactivation in 12-week-old mice (38), suggesting that these previous findings likely reflect long-term impact of GSK3 loss on β-cell proliferation and survival (39,40).

Our findings that high glucose enhances GSK3-mediated tau phosphorylation are consistent with the reported activation of GSK3 in high glucose (30,41). Increased tau phosphorylation results in tau dissociation from microtubules and increased microtubule turnover. This occurs predominantly near the periphery of β-cells where tau-bound microtubules are enriched. These findings imply that β-cells have two subpopulations of microtubules: the peripheral microtubules, which are more sensitive to high glucose–induced disassembly, and the less sensitive interior microtubules. Therefore, the differential tau association with peripheral and interior microtubules may contribute to their distinct sensitivity to high glucose–induced destabilization. Furthermore, microtubule-bound tau in β-cells forms distinct puncta along microtubules. This is consistent with the recently reported tau self-aggregation into islands, which form selective barriers for kinesins but not dynein (42,43). In our model, the association of tau with peripheral microtubules likely inhibits kinesin-mediated vesicle transportation toward the plasma membrane but does not prevent dynein-dependent vesicle withdrawal. Logically, this subpopulation of microtubules that are sensitive to glucose enables fine-tuning of vesicle allocation in response to glucose stimulation.

Several unresolved issues exist. First, the roles of CDK5, PKA, and PKC in high glucose–induced microtubule disassembly are unclear. Because GSK3-mediated tau phosphorylation depends on priming by other kinases including PKA and CDK5 (44,45), future studies on CDK5- and PKA-mediated priming are important. Second, how exactly tau regulates microtubule dynamics is not clear. Conventional models suggest that tau stabilizes microtubules and promotes microtubule growth (46), whereas a recent study indicates that neuronal tau indirectly promotes the growth of the labile domain rather than stabilizing axonal microtubules (47). Furthermore, tau islands were shown to inhibit microtubule severing by katanin, indirectly extending microtubule life span (21,43). Our results that tau-KD significantly increases microtubule turnover in the β-cell periphery are consistent with the latter findings. Future studies are needed to test whether microtubule-severing proteins, such as katanin, are involved in the regulation of microtubule turnover and GSIS. Third, why tau preferentially binds to microtubules near the cell periphery is unknown. Subcellular and regional GSK3 localization and/or activation/inhibition, which has been described in several cell types (48,49), could account for this difference. Lastly, whether tau also regulates vesicle–plasma membrane fusion in β-cells is unknown. Since tau cross links microtubules and F-actin (50), it is possible that tau directly regulates glucose-induced disassembly of actin cortex, which facilitates the exocytosis of insulin vesicles.

In summary, we report a novel feed-forward mechanism by which glucose regulates both short-term insulin secretion and long-term β-cell function. In low glucose, tau suppresses the turnover of microtubules in the cell periphery to limit basal insulin secretion and augment the insulin pool. High-glucose stimulation promotes tau dissociation from microtubules in the cell periphery to facilitate insulin vesicle release. Meanwhile, high glucose enhances new microtubule formation from the Golgi and does not induce as strong an effect on interior microtubule disassembly. We envision that manipulating these processes may prevent β-cell dysfunction, a hallmark of type 2 diabetes.

Acknowledgments. The authors thank Drs. Chris Wright, Xiaodong Zhu, and Kathryn Trogden (Vanderbilt University) for constructive discussion and technical help.

Funding. This work was supported by Eli Lilly and Company Lilly Innovation Fellowship Award (LIFA) fellowship 0101420 (to K.-H.H.) and National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, grants R01-DK106228 (to G.G. and I.K.), DK065949 (to G.G.), and R35-GM127098 and R01-GM078373 (to I.K.).

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

Author Contributions. K.-H.H., I.K., G.G., O.C., and M.L.H. conceptualized the work and designed the study. K.-H.H. designed most of the experiments and did most of the work. A.B.O. and M.A.M. provided Ins2^H2B.Apple mice. X.Y. performed quantitative RT-PCR experiments and data analyses. All authors contributed to writing the manuscript. I.K. and G.G. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 79th Scientific Sessions of the American Diabetes Association, San Francisco, CA, 7–11 June 2019.