Previous work has established the ER-localized lipid-transport protein TMEM24 as an important regulator of insulin secretion from pancreatic β-cells (Lees et al., 2017; Pottekat et al., 2013). Transient reduction of TMEM24 expression or knockout of the TMEM24 gene have been found to impair insulin secretion from clonal rat and mouse β-cells. The underlying mechanism has been proposed to involve reduced Ca2+-independent recruitment of insulin granules to the PM (Pottekat et al., 2013) or disturbed regulation of voltage-dependent Ca2+ influx through alterations in the PM lipid composition (Lees et al., 2017). In this work, we further explored the mechanisms underlying TMEM24-dependent regulation of insulin secretion. In contrast to previous studies, we found that reduced expression of TMEM24 has little impact on voltage-dependent Ca2+ influx and only modestly impaired insulin secretion from β-cells. Instead, we identified TMEM24 as an important regulator of Ca2+ homeostasis at both the ER and the mitochondria, and show that this protein regulates mitochondrial ATP production.

TMEM24 is anchored to the ER through an N-terminal transmembrane domain and contains, in sequence, a lipid-binding SMP-domain, a C2-domain and a polybasic C-terminus. It is highly expressed in neuronal and endocrine tissue, including the pancreatic islets of Langerhans, where it is involved in tethering the ER to the PM (Lees et al., 2017; Sun et al., 2019). Its localization to the PM depends on electrostatic interactions between the C-terminus and negatively charged lipids in the PM, and neutralization of charged amino acids in TMEM24 by PKC-dependent phosphorylation results in TMEM24 dissociation. Using both intact and permeabilized cells, we found that the Ca2+-dependent dissociation of TMEM24 from the PM can be triggered even by nanomolar elevations of cytosolic Ca2+ through intracellular release or extracellular influx. Interestingly, we found that TMEM24 is already partially displaced from the PM under resting Ca2+ concentrations, perhaps due to constitutive PKC activity. This is consistent with findings from neuron-like cells, where endogenous TMEM24 has been found both at the PM and in the bulk ER under resting conditions (Sun et al., 2019). Such observations would support the hypothesis that TMEM24 acts at additional cellular locations, perhaps driven by interactions with lipids in the membranes of organelles known to form contacts with the ER (Cohen et al., 2018; Eden et al., 2010). By locally photobleaching a small region of the PM in cells expressing TMEM24–GFP, we could also estimate the mobility of TMEM24 at ER–PM junctions. We found that TMEM24 weakly interacts with the PM, and that a large fraction was highly dynamic under resting conditions, which is in sharp contrast to E-Syt2, another SMP-domain-containing protein constitutively localized to ER-PM contact sites (Xie et al., 2019). This observation is also consistent with TMEM24 executing functions at other cellular locations than the PM.

In contrast to previous studies (Lees et al., 2017; Pottekat et al., 2013), we did not find support for an absolute requirement of TMEM24 for normal insulin secretion. Both transient knockdown of TMEM24 using siRNA and CRISPR/Cas9-mediated knockout of TMEM24 had little effect on either glucose- or depolarization-induced Ca2+ influx and insulin secretion from MIN6 cells cultured as monolayers. If anything, cells with reduced expression performed slightly better than control cells. Interestingly, when we instead allowed MIN6 cells to aggregate into islet-like cell clusters (pseudo-islets) we found that the sustained glucose-stimulated insulin secretion was reduced following TMEM24 KO. One previous study, in which TMEM24 expression was stably reduced by shRNA, showed impaired glucose-stimulated insulin secretion from both clonal rat INS1 and mouse MIN6 β-cell pseudo-islets. The Ca2+ responses in these cells were normal, and secretion in response to direct depolarization was also unaffected by TMEM24 knockdown (Pottekat et al., 2013), which is similar to what we report here. CRISPR/Cas9-mediated knockout of TMEM24 in INS1 cells resulted in complete inhibition of both glucose-induced Ca2+ increases and insulin secretion, which was restored by re-expression of full-length TMEM24 (Lees et al., 2017). Although INS1 cells secrete insulin in response to glucose, the mechanism is likely different from that of primary β-cells in that it not only depends on KATP-channel closure (Herbst et al., 2002) and voltage-dependent Ca2+ influx (Dorff et al., 2002). Insulin secretion may instead be triggered by Ca2+ released from intracellular stores, since addition of the ER Ca2+ ATPase (SERCA) inhibitor thapsigargin causes Ca2+ oscillations in these cells (Herbst et al., 2002). Another possibility for the observed differences in insulin secretion may be that TMEM24 only regulate secretion under certain conditions. The observation here that TMEM24 KO impaired sustained insulin secretion from pseudo-islets despite having little effect on glucose-induced cytosolic Ca2+ changes indicate that it may be involved in the amplifying pathways of insulin, which operate in parallel with the Ca2+-dependent triggering pathway. The amplifying pathway requires mitochondrial metabolism, and is more prominent in pseudo-islets than cell monolayers (Chowdhury et al., 2013). Consistent with this, we observe impaired mitochondrial oxidative phosphorylation in TMEM24 KO cells, which is accompanied by depolarization of the inner mitochondrial membrane and in hyper-accumulation of Ca2+. The mitochondria were still able take up and release Ca2+ in response to changes in cytosolic Ca2+, arguing against defects in the major uptake and extrusion pathways. Regulation of mitochondrial Ca2+ is closely linked to ER Ca2+, and it is possible that the changes in mitochondrial function observed after TMEM24 KO are secondary to changes in ER Ca2+ handling. We found that TMEM24 KO cells have increased Ca2+ accumulation in the ER, observed by both measurements of cytosolic Ca2+ following ER-store depletion and by direct measurements of ER Ca2+. It is not clear how TMEM24 contributes to ER Ca2+ homeostasis. The major route of Ca2+ uptake into the ER of β-cells is the SERCA (Roe et al., 1994) but we did not find any evidence for increased SERCA activity in TMEM24 KO cells. If anything, the activity was slightly reduced, as indicated by the lack of an initial Ca2+ lowering effect of glucose in TMEM24 KO cells, although this might also reflect impaired ATP production or the steeper concentration gradient in these cells. There is a continuous leakage of Ca2+ from the ER, which results in rapid loss of Ca2+ upon SERCA inhibition, but the mechanism behind this leak is not clear. Numerous mediators of ER Ca2+ leak have been identified, including presenilin-1/2 and TMCO1 (Tu et al., 2006; Wang et al., 2016). Reduced ER Ca2+ leak could explain the increased accumulation of Ca2+ in the ER of TMEM24 KO cells. Interestingly, loss of both presenilin-1 and TMCO1, like TMEM24, results in ER Ca2+ overload and in impaired mitochondria function (Tu et al., 2006; Wang et al., 2016, 2019). Although TMEM24 unlikely functions as a Ca2+ channel, it may modulate other release mechanisms either through direct interactions or modulation of the lipid environment. Because of its dynamic nature, TMEM24 may provide means to acutely regulate ER Ca2+ permeability in response to increases in cytosolic Ca2+ or PM DAG concentrations. An alternative explanation for how TMEM24 regulate mitochondria function is by acting in trans at ER-mitochondria contact sites. Increases in the cytosolic Ca2+ concentration causes the dissociation of TMEM24 from the PM and is followed by accumulation of TMEM24 at mitochondria, as shown by the increased overlap between TMEM24-GFP and an ER-mitochondria proximity reporter. This is similar to other ER-localized lipid transport proteins, such as ORP5/8 and Vps13A/C, which have been shown to bind more than one organelle membrane (Galmes et al., 2016; Kumar et al., 2018). Although our observations are consistent with a role of TMEM24 at ER-mitochondria contacts, this would need to be confirmed using super-resolution microscopy or correlative light and electron microscopy. ER-mitochondria contacts are sites of Ca2+ and lipid exchange that control mitochondria Ca2+ levels, ATP production and morphology in β-cells (Rieusset, 2018). It is possible that TMEM24 controls mitochondria Ca2+ by modulating the Ca2+ transfer reaction at ER-mitochondria contact sites. These contact sites concentrate many of the key components of organellar Ca2+ homeostasis, including the MCU, the voltage-dependent anion channel (VDAC), SERCA, Presenilin-1 and IP3 receptors (Area-Gomez et al., 2009; Vecellio Reane et al., 2020). The MCU has low affinity for Ca2+ and is kept inactive under resting cytosolic Ca2+ concentrations (Marchi and Pinton, 2014). However, MCU is still important for maintaining basal energy production, likely via sensing Ca2+ microdomains formed by Ca2+ release from the ER at ER-mitochondria contacts (Rossi et al., 2019). It is possible that TMEM24 controls the transfer of Ca2+ between the two compartments either by directly modulating the function of one or several components of the contact sites, or indirectly through its effect on ER Ca2+ concentration and Ca2+ mobilization. Another intriguing possibility is that TMEM24 modulates the lipid composition of the mitochondria membranes, and that lack of this transport alters mitochondria function. TMEM24 has a strong preference for phosphatidylinositol, and the importance of this lipid for mitochondria function has been known since the 1960s (Vignais et al., 1963). More recently, it has been shown that that the outer mitochondrial membrane is indeed rich in phosphatidylinositol (Zewe et al., 2020), and that phosphorylated derivatives of this lipid are required for mitochondrial function (Nagashima et al., 2020; Rosivatz and Woscholski, 2011). However, it remains to be discovered how phosphatidylinositol is delivered to the mitochondria. One possibility is that TMEM24 contributes to this transport and couples it to changes in Ca2+ concentration and energy demand. Interestingly, one study found that depletion or masking of PI(4,5)P2 on the mitochondria surface caused mitochondrial fragmentation, which could be prevented by PKC activation and would also trigger TMEM24 dissociation from the PM to enable interactions with the mitochondria (Rosivatz and Woscholski, 2011). A Ca2+-dependent feedback system to control mitochondria function may be particularly important in the β-cells, where mitochondrial metabolism is tightly coupled to Ca2+ influx in order to adjust insulin secretion and maintain blood glucose homeostasis.