Expression of PIP4P2 regulates alpha cell function

Our previous patch-seq studies of alpha cells from human donors identified hundreds of transcripts correlated with alpha cell function [18]. However, potential causative roles for most of these remain to be elucidated. Mining this data, a lipid phosphatase, TMEM55A (encoded by PIP4P2), stood out, as it appeared to associate with single-cell function in alpha cells. However, when we analysed both PIP4P2 mRNA expression and TMEM55A protein abundance in donors with or without type 2 diabetes (using data available at www.humanislets.com [20]), no obvious differences were found, partly due to the large variation in expression levels in each group (Fig. 1a, b). Nevertheless, we did confirm the variable expression of TMEM55A using snap-frozen islets from both donors without and donors with type 2 diabetes (Fig. 1c). We further confirmed the expression of TMEM55A in human alpha and beta cells by immunostaining biopsies from donors with and without type 2 diabetes (Fig. 1d). On analysis of a large patch-seq dataset [20], we found that the expression of PIP4P2 was positively correlated with alpha cell size, exocytosis, Ca2+ currents and Na+ currents (Fig. 1e), while it did not correlate with most beta cell electrophysiological properties (Fig. 1f). We further compared the correlations in alpha cells from donors with or without type 2 diabetes at 1 mmol/l glucose and 5 mmol/l glucose and found that PIP4P2 demonstrated a loss of correlation (or even opposite correlation) with alpha cell electrophysiology parameters in type 2 diabetes (Fig. 1g, h), indicating that the function of TMEM55A might be dysregulated.

Fig. 1

TMEM55A and islet cell function. (a) PIP4P2 transcript expression in alpha and beta cells from single-cell RNA-seq data from donors with and without type 2 diabetes. (n=1617, 493, 565 and 121 ND and T2D alpha cells and ND and T2D beta cells, respectively). (b) TMEM55A protein abundance in donors with and without type 2 diabetes (n=17 and 117 donors, respectively). (c) Western blot of TMEM55A from snap-frozen islets from four donors without and four donors with type 2 diabetes; Mean blot intensities normalised to GAPDH levels are shown. (d) Representative immunofluorescence images confirm TMEM55A expression in alpha and beta cells from donors with and without type 2 diabetes. Scale bar, 50 µm (n=ten islets from three donors without diabetes and nine islets from three donors with type 2 diabetes). (e, f) Correlation of PIP4P2 transcript expression (encodes TMEM55A protein) with electrophysiological properties in human alpha cells (e; n=1114 cells from 45 donors) and beta cells (f; n=229 cells from 37 donors). (g, h) Correlation of PIP4P2 transcript expression with electrophysiological properties in human alpha cells at 1 mmol/l glucose (g; n=393 cells from 33 donors without diabetes and 75 cells from seven donors with type 2 diabetes) or 5 mmol/l glucose (h; n=524 cells from 43 donors without diabetes and 211 cells from 11 donors with type 2 diabetes). Data in a, b, e, f are from www.humanislets.com. *p<0.05, **p<0.01 and *** p<0.001. ND, no diabetes; T2D, type 2 diabetes

To better characterise the role of TMEM55A in islets, PIP4P2 was knocked down by RNA interference in islet cells isolated from humans. Knockdown of TMEM55A was confirmed at both mRNA (86% reduction) and protein (57% reduction) levels (Fig. 2a, b), and by single-cell immunofluorescence (Fig. 2c and ESM Fig. 1a). Interestingly, we did not find any differences in alpha cell size on knockdown of PIP4P2, although we cannot rule out that more chronic knockdown or developmental effects may be required to impact alpha cell size. Next, we functionally characterised the role of TMEM55A using the patch-clamp technique. Here, we still did not find any differences in alpha cell size or depolarisation-induced Ca2+ entry following knockdown of PIP4P2 but found that the alpha cell exocytotic response (at 1 mmol/l glucose) was dramatically decreased (Fig. 2d, e), consistent with the correlation analysis above. In beta cells, we found little effect following PIP4P2 knockdown (ESM Fig. 1b, c). As this single-cell exocytosis measurement using the capacitance as a readout can inform about mechanistic underpinnings of alpha cell function but does not measure secretion directly, we assessed glucagon secretion stimulated by low glucose (1 mmol/l) together with GIP (100 nmol/l) and the amino acid alanine (10 mmol/l) or KCl alone following PIP4P2 knockdown, as we find these to be robust stimuli of glucagon secretion. Consistently, we found that glucagon secretion from the si-PIP4P2 transfection group was significantly reduced on KCl or GIP and alanine stimulation compared with control (Fig. 2f). These data together suggest the involvement of TMEM55A in the regulation of glucagon secretion.

Fig. 2

Knockdown of PIP4P2 decreases alpha cell exocytosis. (a) qPCR analysis of PIP4P2 mRNA expression from control and PIP4P2-knockdown human islet cells (n=3 donors). (b) Representative western blot of TMEM55A from control and PIP4P2-knockdown human islet cells; mean blot intensities normalised to β-actin are shown (n=3 donors). (c) Representative immunofluorescence images showing TMEM55A in control and PIP4P2-knockdown human alpha cells. Scale bar, 5 µm. Positive glucagon staining was used to confirm alpha cell identity. Mean intensities (normalised to cell area) per cell and cell size indicated by cell diameter (n=15 and 14 cells, respectively, from three donors) are shown. (d) Representative capacitance and current traces induced by a train of ten depolarisations from −70 mV to 0 mV (grey trace) from control and PIP4P2-knockdown human alpha cells. Electrophysiological parameter measurement included cell size (initial membrane capacitance), early exocytosis (capacitance change following the first-time depolymerisation), late exocytosis (capacitance change after the first-time depolymerisation), total exocytosis (capacitance change for tenth depolymerisation) and Ca2+ integral (Ca2+ entry following the first depolymerisation). (e) Mean cell size (n=28 and 29), early exocytosis (n=28 and 29), late exocytosis (n=28 and 29), total exocytosis (n=28 and 29), Ca2+ integral (n=12 and 10) and exocytosis normalised to Ca2+ (n=12 and 10) for control and PIP4P2-knockdown human alpha cells, calculated as total exocytosis normalised to the Ca2+ integral, in (d) (n=5 donors). (f) Glucagon secretion from dispersed control and PIP4P2-knockdown human islets at basal (5 mmol/l glucose) and stimulated conditions. Total glucagon content is also shown (n=3 donors). (g) Representative current traces induced by depolymerisation from −70 mV to −10 mV (grey trace). Electrophysiological parameter measurement included Na+ currents (peak currents induced by the depolymerisation), early Ca2+ currents (initial currents after the closure of Na+ channels) and late Ca2+ currents (plateau currents after the closure of Na+ channels). (h) Mean Na+ currents (n=22 and 25), early Ca2+ currents (n=24 and 22) and late Ca2+ currents (n=24 and 22) for control and PIP4P2-knockdown human alpha cells in (g) (n=5 donors). (i) Representative current traces obtained using a voltage jump protocol (grey trace) from control and PIP4P2-knockdown human alpha cells; averaged and normalised I-V curves are shown (n=13 and 11 cells, respectively, from three donors). (j) Representative capacitance traces following infusion of 200 nmol/l free Ca2+ from control and PIP4P2-knockdown human alpha cells; averaged and normalised capacitance increase (ΔCm) after 200 s infusion is shown (n=10 and 9 cells, respectively, from three donors). Data are presented as mean ± SD. *p<0.05, **p<0.01 and ***p<0.001 by Student’s t test (a–c, e, h, j) or one-way ANOVA and Holm–Sidak post-test (f). Ala, alanine; Glu, glucose; si-scramble, control human islets; si-PIP4P2, PIP4P2-knockdown human islets

We also measured voltage-dependent Na+ and Ca2+ currents. The PIP4P2-knockdown alpha cells showed decreased Na+ currents. However, we did not detect any significant changes in Ca2+ currents (Fig. 2g, h). Consistent with the little effect on beta cell exocytotic response, no differences in the Na+ or Ca2+ currents were found on PIP4P2 knockdown in beta cells (ESM Fig. 1d, e). As intracellular Ca2+ inhibits voltage-dependent Ca2+ channels, we also used Ba2+ as a charge carrier and included the Na+ channel inhibitor tetrodotoxin (0.5 µmol/l) in the bath to more closely interrogate Ca2+ channel activity. We still found no significant effect of PIP4P2 knockdown on Ca2+ channel-mediated currents (Fig. 2i), indicating that reduced Ca2+ channel activity is not responsible for the decreased alpha cell exocytosis observed. Indeed, on infusion of 200 nmol/l free Ca2+ into the cell interior, exocytosis from the PIP4P2 siRNA-transfected alpha cells was still lower than from control cells (Fig. 2j).

A role for PIP2 and PI5P in the regulation of alpha cell exocytosis

TMEM55A was initially identified as a phosphatase that dephosphorylates PIP2 to PI5P [17]. As a signalling phospholipid, PIP2 plays critical roles in exocytosis [21, 22], while the physiological relevance of PI5P to exocytosis has never been explored. We therefore examined whether PIP2 or PI5P can regulate alpha cell exocytosis. The intracellular dialysis of 1 µmol/l diC8 PIP2 or PI5P, a water-soluble dioctanoyl analogue, decreased and increased human alpha cell exocytosis, respectively (Fig. 3a, b), directionally consistent with what we would expect based on our TMEM55A knockdown studies. Indeed, PI5P infusion rescued the reduced exocytosis in si-PIP4P2 transfected cells to levels comparable with those of controls (Fig. 3c). Similar results were obtained using mouse alpha cells (Fig. 3d, e and ESM Fig. 2). To better illustrate the role of PIP2 and PI5P on glucagon secretion, we treated mouse islets with 1 µmol/l PIP2 or PI5P (which are membrane permeable [23, 24]) overnight and found that PIP2 had no obvious effect, while PI5P-treated islets demonstrated increased glucagon secretion (Fig. 3f). This suggests that the decreased glucagon secretion caused by PIP4P2 knockdown might be due to decreased intracellular PI5P levels, rather than increased PIP2 levels. Of note, neither PIP2 nor PI5P affected glucagon content (Fig. 3g), which is consistent with our knockdown study (Fig. 2f).

Fig. 3

The effect of PIP2 and PI5P on glucagon secretion. (a) Mean total exocytosis with or without the infusion of 1 µmol/l PIP2 in human alpha cells (n=23 and 18 cells, respectively, from four donors). (b) Mean total exocytosis with or without the infusion of 1 µmol/l PI5P in human alpha cells (n=13 and 11 cells, respectively, from three donors). (c) Mean total exocytosis from human alpha cells treated with control scrambled RNA, PIP4P2 siRNA, and PIP4P2 siRNA with PI5P infusion (n=11, 13 and 11 cells, respectively, from three donors). (d) Mean total exocytosis from control and Pip4p2-knockdown mouse alpha cells (n=16 and 10 cells, respectively, from three mice). (e) Mean total exocytosis from control mouse alpha cells, Pip4p2-knockdown mouse alpha cells, and Pip4p2-knockdown mouse alpha cells with PI5P infusion (n=9, 10 and 9 cells, respectively, from three mice). (f) Glucagon secretion induced by 2.8 mmol/l glucose, GIP and alanine in control mouse islets and islets with PIP2 or PI5P treatments, together with respective AUCs (n=6 mice). (g) Total glucagon content from (f). Data are presented as mean ± SD. *p<0.05 and **p<0.01 by Student’s t test (a, b, d), two-way ANOVA test with mixed-effects analysis (f) or one-way ANOVA followed by Holm–Sidak post-test (c, e, g). Ala, alanine; Ctrl, control; si-scramble, control human or mouse alpha cells; si-PIP4P2, PIP4P2-knockdown human alpha cells; si-Pip4p2, Pip4p2-knockdown mouse alpha cells

Phosphatase activity of TMEM55A

It has been suggested that TMEM55A does not, in fact, have lipid phosphatase activity [25]. To investigate this, and to confirm whether TMEM55A lipid phosphatase activity is involved in the regulation of alpha cell function, we generated a WT GFP-tagged human TMEM55A (GFP-TMEM55A) and ‘phosphatase dead’ mutant TMEM55A (GFP-TMEM55A C107S). Interestingly, HEK 293 cells expressing GFP-TMEM55A demonstrated a more rounded shape than those expressing GFP or GFP-TMEM55A C107S (ESM Fig. 3), suggesting that some activity of the WT enzyme influences cell morphology. Moreover, we detected a greater hydrolysis of PIP2 by GFP-TMEM55A than with GFP or GFP-TMEM55A C107S, which was even more robust than that by the positive control, the phosphatase enzyme SH2 domain-containing inositol phosphatase 2 (SHIP2) (Fig. 4a). As PI5P is reported to be an oxidative stress-induced second messenger and is increased in response to hydrogen peroxide (H2O2) in diverse cell lines [26], we examined whether TMEM55A activity could be modulated by cellular redox in vitro. Here, we found that under reducing conditions with 1 mmol/l dithiothreitol (DTT), TMEM55A lipid phosphatase activity was inhibited while subsequent incubation with an oxidiser (1 mmol/l H2O2) activated the enzyme (Fig. 4b). Considering the reducing intracellular environment, we assumed that TMEM55A is sensitive to oxidation on exposure to air and was fully activated during our preparation.

Fig. 4

TMEM55A regulation of glucagon secretion requires its phosphatase domain. (a) Mean phosphatase activity of GFP, GFP-TMEM55A, GFP-TMEM55A C107S and SHIP2 (n=3). (b) Mean phosphatase activity of GFP-TMEM55A following treatment with DTT, H2O2 or DTT and H2O2 (n=4). (c) Representative images of alphaTC1–9 cells transfected with RFP-PH-PLC (red) and GFP (green), GFP-TMEM55A or GFP-TMEM55A C107S at the basal level and after treatment with H2O2 (1 mmol/l) for 15 min. Scale bar, 10 µm. Quantification of the dynamic change in PIP2 signal intensity in alphaTC1–9 cells on H2O2 stimulation is shown (n=8, 10 and 10 cells, respectively, from three independent experiments). (d) Static glucagon secretion stimulated by KCl (55 mmol/l) from alphaTC1–9 cells transfected with scrambled RNA + GFP, Pip4p2 siRNA + GFP, Pip4p2 siRNA + GFP-TMEM55A or Pip4p2 siRNA + GFP-TMEM55A C107S (n=3). (e) Static glucagon secretion stimulated by KCl (55 mmol/l) from alphaTC1–9 cells with or without PI5P treatment (n=4). Data are presented as mean ± SD. **p<0.01 and ***p<0.001 by two-way ANOVA test with mixed-effects analysis (c) or one-way ANOVA followed by Holm–Sidak post-test (a, b, d, e). Ctrl, control (no PIP5); Pi, inorganic phosphate; si-scramble, control cells; si-Pip4p2, Pip4p2-knockdown cells

Next, we tested the sensitivity of TMEM55A to H2O2 in situ in alphaTC1–9 cells. Because there is no reliable PI5P sensor, we overexpressed GFP-TMEM55A and the PIP2 probe, mCherry-PH-PLC [27], at the same time. Then, we monitored the dynamic PIP2 changes following treatment with 1 mmol/l H2O2 using live-cell imaging. As reported previously [28], external application of H2O2 induced PIP2 hydrolysis. Interestingly, while treatment with H2O2 activated PIP2 hydrolysis in cells expressing WT GFP-TMEM55A, this effect was lost in cells expressing GFP-TMEM55A C107S (Fig. 4c and ESM Videos 1–3), indicating that H2O2-induced PIP2 hydrolysis depends on the phosphatase activity of TMEM55A.

Because of the ease of genetic manipulation, we performed similar studies in alphaTC1–9 cell lines to those carried out in primary alpha cells. Knocking down TMEM55A decreased glucagon secretion stimulated by KCl from alphaTC1–9 cells (Fig. 4d). When we overexpressed GFP-TMEM55A and GFP-TMEM55A C107S, we found that only WT GFP-TMEM55A cells showed a partial recovery of glucagon secretion following TMEM55A knockdown. This indicates that control of glucagon secretion by TMEM55A depends on its phosphatase activity. Moreover, the product of TMEM55A, PI5P, significantly increased glucagon secretion stimulated by KCl in alphaTC1–9 cells (Fig. 4e). Taken together, these data suggest that the regulation of glucagon secretion by TMEM55A depends on its lipid phosphatase activity.

RhoA is the downstream signalling molecule of PI5P

PI5P is the least-characterised phosphoinositide and its functions remain elusive [29]. It was reported that PI5P might regulate actin dynamics in cell migration [30], and intracellular or extracellular application of PI5P in different cell lines causes F-actin stress fibre breakdown via activation of ras-related C3 botulinum toxin substrate 1 (RAC1) [24, 31]. To address the downstream signalling effect of PI5P in alpha cells, we monitored the levels of active RAC1 together with two other major GTPases, CDC42 and RhoA, in alphaTC1–9 cells following treatment with PI5P. We found that treatment with PI5P led to significant inactivation of RhoA, with no obvious effects on RAC1 or CDC42 (Fig. 5a). Although this contradicts observations in mouse embryonic fibroblast (MEF) cell lines [24], it is in line with the role of RhoA identified in alpha cells [15]. Similar results were found in human islets (Fig. 5b). Expression of GFP-TMEM55A in alphaTC1–9 cells also resulted in the inactivation of RhoA, while GFP or GFP-TMEM55A C107S transfection had no effect (Fig. 5c). These observations suggest that RhoA is a downstream effector of PI5P and TMEM55A in alpha cells.

Fig. 5

PI5P and TMEM55A inactivate RhoA and have no effects on RAC1 and CDC42. (a) Pull-down assays showing the levels of active forms of RAC1, CDC42 and RhoA in alphaTC1–9 cells treated with or without PI5P (10 µmol/l) for 1 h; mean active forms of the GTPases are shown (n=4, 3 and 3, respectively). (b) Pull-down assays showing the levels of active forms of RAC1, CDC42 and RhoA in human islets treated with or without PI5P (1 μmol/l) overnight; mean active forms of the GTPases are shown (n=3, 3 and 3, respectively). (c) Pull-down assays showing the levels of active forms of RAC1, CDC42 and RhoA for GFP-, GFP-TMEM55A- or GFP-TMEM55A C107S-transfected alphaTC1–9 cells treated with or without PI5P (10 µmol/l) for 1 h; mean active forms of the GTPases are shown (n=3, 3 and 3, respectively). Data are presented as mean ± SD. *p<0.05, **p<0.01 and ***p<0.001 by Student’s test (a, b) or one-way ANOVA followed by Holm–Sidak post-test (c). Ctrl, control (no PIP5)

PI5P regulates alpha cell F-actin remodelling

F-actin depolymerisation is mainly governed by Rho-family small GTPases [32]. After validation that RhoA is downstream of TMEM55A in alpha cells, we wanted to visualise whether PI5P could disrupt F-actin in alpha cells to enhance glucagon secretion. We first confirmed the relationship between F-actin and alpha cell exocytosis. Human alpha cells treated with 10 µmol/l latrunculin B or jasplakinolide for 1 h to disrupt and enhance actin polymerisation demonstrated increased and decreased exocytotic responses, respectively, directionally consistent with the effects of TMEM55A and PI5P (Fig. 6a). We then examined the distribution of F-actin by Phalloidin staining using confocal microscopy following PI5P treatment. Consistent with previous studies on MEF and HeLa cell lines [24, 31], PI5P significantly decreased F-actin intensity in dispersed human and mouse primary alpha cells (Fig. 6b, c), in whole islets (Fig. 6d, e) and in alphaTC1–9 cells (Fig. 6f). Using live-imaging of F-tractin-mCherry, a probe for monitoring global F-actin [33], we found that PI5P disrupted F-actin in a time-dependent manner (Fig. 6g and ESM Videos 4, 5). As TMEM55A increases intracellular PI5P, we overexpressed GFP-TMEM55A in alphaTC1–9 cells and found that these cells demonstrated greater F-actin disruption than cells expressing GFP-TMEM55A C107S or GFP alone (Fig. 6h). Conversely, when we knocked down PIP4P2 in human alpha cells, we detected a significant increase in F-actin intensity (Fig. 6i), which explains the decreased exocytosis in our functional studies (Fig. 2e). Taken together, these data strongly suggest that TMEM55A and PI5P are involved in F-actin depolymerisation.

Fig. 6

F-actin depolymerisation is involved in the regulation of glucagon secretion by TMEM55A. (a) Mean total exocytosis without treatment or after incubation with 10 µmol/l latrunculin B or jasplakinolide for 1 h in human alpha cells (n=12, 11 and 9, respectively, from three donors). (b, c, f) Representative immunofluorescence images of human alpha cells (n=96 and 85 cells, respectively, from five donors) (b), mouse alpha cells (n=32 and 32 cells, respectively, from three mice) (c) and alphaTC1–9 cells (n=34 and 34 cells from three independent experiments) (f) following no treatment or treatment with PI5P for 1 h. F-actin was visualised by staining with AlexaFluor 647-Phalloidin. Positive glucagon staining was used to confirm alpha cell identity in dispersed human or mouse islets. Scale bar, 5 µm. F-actin fluorescence intensity line profile analysis and mean F-actin intensity per cell are shown. (d, e) Representative immunofluorescence images of human (d) and mouse (e) whole islets following no treatment or treatment with PI5P overnight from three independent experiments. Yellow arrows indicate the disruption of F-actin. Scale bar, 50 µm. (g) Representative images of alphaTC1–9 cells transfected with F-tractin-mCherry at the basal level and after treatment with PI5P (10 µmol/l) for 10 min. Scale bar, 10 µm. Quantification of the dynamic change of the F-actin signal intensity in the alphaTC1–9 cells on PI5P treatment is shown (n=6 or 7 cells, respectively, from three independent experiments). (h) Representative immunofluorescence images of alphaTC1–9 cells transfected with GFP, GFP-TMEM55A or GFP-TMEM55A C107S. White asterisks indicate GFP-positive cells and yellow arrows indicate the disruption of F-actin. Scale bar, 10 µm. Mean percentage of cells showing F-actin disruption from GFP-positive cells, from three independent experiments including 40–59 cells, is shown. (i) Representative immunofluorescence images showing the F-actin level in control and PIP4P2-knockdown human alpha cells. F-actin was visualised by staining with AlexaFluor 647–Phalloidin. Positive glucagon staining was used to confirm alpha cell identity. Scale bar, 5 µm. F-actin fluorescence intensity line profile analysis and mean F-actin intensity per cell are shown (n=33 and 34 cells, respectively, from three donors). Data are presented as mean ± SD. *p<0.05, **p<0.01 and ***p<0.001 by Student’s t test (b, c, f, i), one-way ANOVA followed by Dunnett’s T3 multiple comparison test (a, h) or two-way ANOVA test with mixed-effects analysis (g). Ctrl, control