TWIK-related two-pore domain K+ channels (TREKs) contribute to the negative resting membrane potential as a partial ‘leak’ K+ conductance. TREKs are activated by various physicochemical conditions including membrane stretch, unsaturated fatty acid, temperature and acidic intracellular pH. For the proper activation of TREKs, tonic inhibited state is necessary. However, the determining factors and related molecular structure are not clearly understood yet. Especially, there has been controversy over the direction of regulation by PI(4,5)P2 that are dynamically generated by phosphatidylinositide (PI) kinases using ATP. TREK-1 and TREK-2 currents are commonly increased spontaneously by dialysis with ATP-free pipette solution for whole-cell recording (ITREK,w-c) or by membrane excision for inside-out patch clamp recording. Consistently, both ITREK-1,i-o and ITREK-2,i-o were inhibited by ATP to the cytoplasmic side. The ATP-dependent inhibition was prevented by PI kinase inhibitor. Also, direct application of PI(4,5)P2 strongly inhibited ITREK,i-o. The putative PI(4,5)P2-mediated inhibition of TREK-2 by intracellular ATP was also observed in the intrinsically expressed cells such as astrocytes and B cells. Treatment with polycationic poly-L-lysine (PLL) totally abolished ITREK-2,i-o, then the application of PI(4,5)P2 showed dose-dependent dual effects; activation and inhibition by relatively low and high ranges, respectively. When co-expressed with voltage-sensitive PI(4,5)P2 phosphatase (Dr-VSP), ITREK-2,w-c was initially activated and then inhibited by sustained activation of Dr-VSP, further indicating the dual effects of PI(4,5)P2 on TREK-2. To identify the residue(s) of pCt for the ATP-dependent TREK-2 inhibition, site-directed substitution of caionic residues with Ala was introduced to human TREK-2. Neutralization of Lys330 (K330A) showed higher basal activity and was not inhibited by ATP. In whole-cell recordings, K330A also showed higher basal activity and the further activation by 2APB, arachidonic acid or acidic pHe (6.9) was less significant. On the contrary, neutralization of the three consecutive Arg distal to Lys330 (R355-7A) showed very low basal activity even without ATP. Finally, I investigate the acidic pHi-sensing residues. Consistent with the previous study of TREK-1, neutralization of a Glu in the pCt of hTREK-2 (E332A) showed higher basal activity and insensitivity to acidic pHi. However, different from hTREK-1, an additional Glu335 of hTREK-2 was contributing to the acidic pHi-sensitivity; E335A showed partial attenuation of the pHi effect. Taken together, I suggest that the ATP-dependent intrinsic PI(4,5)P2 induces tonic inhibition of TREKs, which might be beneficial for the wide changes of ITREK modulation for physiological responses.