AKT Ser/Thr kinase increases V-ATPase – dependent lysosomal acidification in response to amino acid starvation in mammalian cells

The vacuolar H 1 -ATPase (V-ATPase) is an ATP-dependent proton pump that is essential for cellular homeostasis. V-ATPase activity is controlled by the regulated assembly of the enzyme from its component V 1 and V 0 domains. We previously reported that amino acid starvation rapidly increases V-ATPase assembly and activity in mammalian lysosomes, but the signaling pathways controlling this effect are unknown. In testing inhibitors of pathways important for controlling cellular metabolism, we found here that the cAMP-dependent protein kinase (PKA) inhibitor H89 increases lysosomal V-ATPase activity and blocks any further change upon starvation. The AMP-activated protein kinase (AMPK) inhibitor dorsomorphin decreased lysosomal V-ATPase activity and also blocked any increase upon starvation. However, CRISPR-mediated gene editing revealed that PKA and AMPK are not required for the starvation-de-pendent increase in lysosomal V-ATPase activity, indicating that H89 and dorsomorphin modify V-ATPase activity through other cellular targets. We next found that the AKT Ser/Thr kinase (AKT) inhibitor MK2206 blocks the starvation-dependent increase in lysosomal V-ATPase activity without altering basal activity. Expression of AKT1 or AKT3, but not AKT2, was required for increased lysosomal V-ATPase activity in response to amino acid

Amino acid homeostasis, which involves balancing the production and utilization of free amino acids, is essential to cellular and organismal viability. To achieve homeostasis, cells must continually sense amino acid availability and respond accordingly. A major way that cells replenish the available supply of amino acids is through lysosomal proteolysis, which depends upon the low lumenal pH generated by vacuolar H 1 -ATPase (V-ATPase). The V-ATPase is an ATP-dependent proton pump that functions to transport protons from the cytosol into various intracellular compartments or across the plasma membrane of specialized cells (1)(2)(3)(4). V-ATPase-dependent acidification is essential for a variety of cellular processes. Acidification of early endosomes promotes ligand/receptor dissociation, which is important for signal termination and receptor recycling (5). Similarly, acidification of late endosomes causes dissociation of lysosomal proteases from mannose 6-phosphate receptors as they are trafficked to the lysosome (6). The formation of endosomal carrier vesicles is also pH-dependent, as is autophagosome/lysosome fusion (7,8). Degradation of lysosomal contents depends on the activation of acid-dependent proteases. Finally, the proton gradient generated by the V-ATPase drives the coupled transport of small molecules, such as amino acids, out of the lysosome (9). Because these various processes occur within discrete pH ranges, the activity of the V-ATPase is highly regulated.
The V-ATPase is composed of two multisubunit domains: the peripheral V 1 domain that hydrolyzes ATP and the integral V 0 domain that transports protons (1,2). V-ATPase activity is controlled by reversible dissociation of the V 1 and V 0 domains, a process termed regulated assembly, where both the catalytic and transport functions of the enzyme are inhibited upon disassembly (1,4,10). Regulated assembly has been best characterized in yeast, where glucose deprivation rapidly lowers assembly in order to preserve energy (4,10,11). We showed this process is controlled by the Ras/cAMP/PKA pathway, where hyperactivation of the pathway causes the V-ATPase to remain assembled even upon glucose starvation (12). Regulated assembly also occurs in mammalian cells, but the stimuli and signaling pathways controlling it are not the same as those in yeast. In mammalian cells, elevated glucose causes increased V-ATPase assembly and activity via the phosphoinositide-3-kinase (PI3K)/AKT pathway (13). Moreover, we recently reported that acute glucose starvation also increases V-ATPase assembly and lysosomal activity in mammalian cells, and that inhibitors of PI3K/AKT and AMPK block this effect (14). This was the first demonstration of AMPK control of assembly in any system, while PI3K is known to be important for increasing assembly in cells infected with influenza virus and in dendritic cells upon maturation (15,16). V-ATPase assembly and activity also respond to changes in amino acid availability, with amino acid starvation causing increased assembly and lysosomal acidification (17). Interestingly, unlike the examples discussed above, inhibition of either PI3K or one of its major effectors, mTOR complex 1 (mTORC1), failed to block the increase in assembly and lysosomal acidification upon amino acid starvation (17).
In the present study, we sought to identify the pathways controlling V-ATPase-dependent lysosomal acidification in response to amino acid starvation. Using a combination of pharmacological and genetic approaches, we find that the PKA inhibitor H89 and the AMPK inhibitor dorsomorphin alter V-ATPase-dependent lysosomal acidification through off-target effects. In contrast, results using the AKT inhibitor MK2206 and genetic analysis demonstrate that AKT is important for controlling lysosomal V-ATPase activity upon amino acid starvation. AKT, also known as protein kinase B (PKB), is a serine/ threonine kinase with diverse substrates that functions to promote survival, growth, and other anabolic processes (18). We now report that AKT1 or AKT3, but not AKT2, is required for the increase in V-ATPase-dependent lysosomal acidification in response to amino acid withdrawal.

PKA and AMPK inhibitors modulate lysosomal V-ATPase activity in the presence and absence of amino acids
We previously reported that in mammalian cells, amino acid starvation increases V-ATPase assembly and lysosomal V-ATPase activity independently of PI3K or its major effector, mTORC1 (17). Because previous studies have suggested the involvement of PKA and AMPK in modulating V-ATPase activity (12,14,19,20), we wished to test the involvement of these important signaling pathways in controlling the starvation-dependent increase in lysosomal V-ATPase activity. To test the role of the PKA pathway, we first determined the minimum concentration of the PKA inhibitor H89 necessary to inhibit forskolin-induced, PKA-dependent phosphorylation of vasodilator-stimulated phosphoprotein (VASP) (21) in HEK293T cells. As shown in Fig. 1A, 50 mM H89 completely inhibits PKA activity. We next tested the effect of H89 on lysosomal V-ATPase activity in HEK293T cells maintained in amino acids or starved for 1 h. As shown in Fig. 1B, H89 significantly increases basal lysosomal V-ATPase activity and blocks any further changes to activity upon amino acid starvation. This suggested that PKA is important for controlling the steady-state level of lysosomal V-ATPase activity in mammalian cells and, possibly, the response to amino acid starvation.
To test the involvement of the AMPK pathway, we first determined the minimum concentration of the AMPK inhibitor dorsomorphin necessary to inhibit AMPK in HEK293T cells, as measured by phosphorylation of the AMPK substrate acetyl-CoA carboxylase (ACC) (22). As shown in Fig. 1C, dorsomorphin markedly reduces ACC phosphorylation, although, consistent with other reports, some AMPK activity appears to remain, even at high drug concentrations (23). Therefore, we selected 5 mM as the lowest dose at which maximal inhibition of ACC phosphorylation occurs. As shown in Fig. 1D, 5 mM dorsomorphin markedly reduces lysosomal V-ATPase activity and blocks the increase upon amino acid starvation, suggesting AMPK is important both for maintaining basal V-ATPase activity and for increasing activity during starvation.
PKA and AMPK are not required to increase lysosomal V-ATPase activity in response to amino acid starvation H89 and dorsomorphin were found to modulate lysosomal V-ATPase activity in the presence and absence of amino acids, suggesting PKA and AMPK regulate V-ATPase activity both under basal conditions and in response to starvation. However, because these compounds are known to have numerous off-target effects (23,24), we wished to confirm these results with genetic disruption of PKA and AMPK. We chose to Figure 1. PKA and AMPK inhibitors modulate lysosomal V-ATPase activity in the presence and absence of amino acids. A, HEK293T cells were incubated for 1 h with DMSO (vehicle) or 50 mM forskolin in the presence of increasing concentrations of the PKA inhibitor H89. Lysates were prepared and subjected to Western blotting with the indicated antibodies as described in Experimental procedures. Representative images are shown (n = 3). B, HEK293T cells were incubated with FITC-dextran to allow endocytic uptake into lysosomes and then treated for 1 h with DMSO (vehicle) or 50 mM H89 in the presence (1AA) or absence (2AA) of amino acids. The rate of V-ATPasedependent fluorescence quenching in lysosomes was determined as described in Experimental procedures. Values are expressed relative to the unstarved, vehicle-treated condition. *, p , 0.05; ns, p 0.05. Error bars represent S.E. C, HEK293T cells were treated with either DMSO (vehicle) or increasing concentrations of the AMPK inhibitor dorsomorphin. Lysates were subjected to Western blotting with the indicated antibodies. Representative images are shown (n = 2). D, The rate of V-ATPase-dependent fluorescence quenching in FITC-dextran-loaded lysosomes was determined for HEK293T cells treated for 1 h with either DMSO (vehicle) or 5 mM dorsomorphin in the presence (1AA) or absence (2AA) of amino acids. Values are expressed relative to the unstarved, vehicle-treated condition. *, p , 0.05; ns, p 0.05. Error bars represent S.E. functionally disrupt PKA and AMPK by targeting the genes encoding their respective catalytic subunits using CRISPRmediated genome editing. The PKA catalytic subunit has two main isoforms: PKA Ca, encoded by PRKACA, and PKA Cb, encoded by PRKACB (25). To disrupt PKA, we first transfected HEK293T cells with a plasmid containing the Cas9 nuclease, an enhanced GFP reporter, and guide sequences targeting PRKACA. Individual GFP-positive cells were sorted by FACS, and clonal populations were expanded. Multiple clones were derived and tested for knockout of PKA Ca by Western blotting. A successful PKA Ca knockout clone then was subjected to a second round of CRISPR to knock out PKA Cb. Successful deletion of both PKA Ca and Cb was confirmed in multiple clones by Western blotting (Fig. 2A). PKA function was then probed by stimulating cells with the PKA activator forskolin, and PKA activity was monitored by phosphorylation of VASP. As shown in Fig. 2B, PKA Ca/Cb double knockout cells have nearly undetectable levels of phospho-VASP, even after forskolin stimulation. We next tested the ability of PKA Ca/Cb double knockout cells to show increased V-ATPase activity upon amino acid starvation. Surprisingly, the PKA Ca/Cb double knockout clones are unaffected in their ability to show increased lysosomal V-ATPase activity upon amino acid starvation (Fig. 2C), suggesting that PKA is not required for this effect. Furthermore, PKA Ca/Cb double knockout clones do not exhibit a significantly different level of basal lysosomal V-ATPase activity compared with WT HEK293T cells (Fig. 2D). This suggests that PKA does not function to suppress lysosomal V-ATPase activity under basal conditions. Finally, H89 increased lysosomal V-ATPase activity in a PKA Ca/Cb double knockout clone, further indicating that H89 modulates V-ATPase activity independently of its effect as a PKA inhibitor (Fig. 2E).
We used a similar double knockout approach to disrupt AMPK. The AMPK catalytic subunit has two isoforms: AMPKa1, encoded by PRKAA1, and AMPKa2, encoded by PRKAA2 (26). We confirmed knockouts by Western blotting using isoform-specific antibodies and also confirmed that phosphorylation of ACC was nearly undetectable in AMPKa1/ a2 double knockout cells (Fig. 3A). Surprisingly, AMPKa1/a2 double knockouts are also unaffected in their ability to show increased lysosomal V-ATPase activity in response to starvation, suggesting that AMPK also is not required for this effect (Fig. 3B). Moreover, the AMPKa1/a2 double knockout clone did not have reduced basal lysosomal V-ATPase activity relative to that of WT cells (Fig. 3C). We also found that dorsomorphin reduced lysosomal V-ATPase activity in this AMPKa1/a2 double knockout clone, further indicating that dorsomorphin modulates V-ATPase activity independently of its effect as an AMPK inhibitor (Fig. 3D). Collectively, these results demonstrate that PKA and AMPK are not required for amino acid-dependent changes in lysosomal V-ATPase activity.
AKT1 and AKT3, but not AKT2, are involved in the increase in lysosomal V-ATPase activity upon amino acid starvation We next tested inhibitors of other signaling pathways that have been implicated in controlling V-ATPase activity. Inter-estingly, while the PI3K inhibitor LY294002 does not prevent the starvation-dependent increase in V-ATPase activity (17), the AKT inhibitor MK2206 was found to block the response to starvation (Fig. 4A). This result suggests that AKT is important in controlling the starvation-dependent increase in V-ATPase activity despite the fact that no starvation-dependent change in AKT activity (as measured by phospho-AKT Western blotting) is observed in HEK293T cells (Fig. 4B). We next sought to test the function of AKT in amino acid-dependent changes in V-ATPase activity genetically. AKT triple knockout cells do not proliferate and, therefore, are not suitable for cell culture (27). However, murine lung fibroblasts have been generated that lack endogenous AKT and instead express one of each of the three individual human AKT isoforms (28), as confirmed by Western blotting using isoform-specific AKT antibodies (Fig.  4C). We tested the ability of these AKT knockout fibroblasts to display increased lysosomal V-ATPase activity upon amino acid starvation. Interestingly, cells expressing only AKT1 or AKT3 exhibit the same 1.5-fold increase in V-ATPase activity observed with the WT HEK293T cells upon starvation (Fig.  4D). In contrast, fibroblasts expressing only AKT2 show no increase in V-ATPase activity in response to amino acid starvation. These results suggest that AKT is important for controlling lysosomal V-ATPase activity in response to changes in amino acid availability, and that this dependence is specific to particular isoforms of AKT.
We next sought to further confirm our finding that particular AKT isoforms are important for controlling lysosomal V-ATPase activity in response to amino acid starvation. Because the mouse fibroblasts tested overexpress human AKT isoforms, we wished to test human cells expressing endogenous levels of AKT. Interestingly, Western blot analysis indicates that HEK293T cells express only AKT1 and AKT2 but lack AKT3 (Fig. 4C). Since fibroblasts expressing only AKT2 show no starvation-dependent increase in V-ATPase activity, we hypothesized that knocking out AKT1 in HEK293T cells (leaving only AKT2 intact) would block this response, whereas knocking out AKT2 (leaving only AKT1 intact) would have no effect. We targeted the AKT1 and AKT2 genes separately by CRISPR and confirmed, by Western blotting, the absence of the corresponding isoform in multiple independently derived clones (Fig. 5A). As further shown in Fig. 5A, there does not appear to be any substantial compensation by overexpression of other isoforms (including AKT3) in any of the derived clones.
We next tested the effect of amino acid starvation on lysosomal V-ATPase activity in the AKT knockout clones. As shown in Fig. 5B, AKT2 knockout cells show a normal starvation-dependent increase in lysosomal V-ATPase activity relative to WT cells. In contrast, all AKT1 knockout clones tested have a greatly diminished response of lysosomal V-ATPase activity to starvation. Four clones showed no significant increase in V-ATPase activity upon starvation, while in four other clones the increase was significantly lower than that in WT cells. Thus, the mean V-ATPase activity upon starvation relative to the unstarved condition was 1.64 6 0.02 in WT cells, 1.64 6 0.09 in the negative-control clones, 1.14 6 0.07 in AKT1 knockout clones, and 1.75 6 0.09 in AKT2 knockout clones. . PKA is not required for increased lysosomal V-ATPase activity upon amino acid starvation. A, CRISPR-mediated genome editing was used to disrupt PKA by targeting the PRKACA and PRKACB genes as described in Experimental procedures. Western blotting was performed on lysates from untransfected WT cells (WT), clones targeted for disruption of PKA Ca followed by a nontargeting control (PRKACA 2/2 ), and clones targeted for disruption of PKA Ca followed by the disruption of PKA Cb (PRKACA 2/2 B 2/2 ). The antibody recognizes both Ca and Cb isoforms at the correct molecular weight of 40 kDa. The antibody also produces a nonspecific band at ;50 kDa. Successful knockout results in loss of the 40-kDa band. Vinculin was used as a loading control. Representative images are shown (n = 2). B, To assess PKA activity, WT, nontarget negative control (Neg), PKA Ca knockout (PRKACA 2/2 ), and PKA Ca/Cb double knockout (PRKACA 2/2 B 2/2 ) HEK293T cells were stimulated with DMSO or 50 mM forskolin for 1 h, followed by Western blotting to detect phosphorylation of VASP. Representative images are shown (n = 2). C, WT, nontargeted negative control (Neg), and PKA Ca/Cb double knockout (PRKACA 2/2 B 2/2 ) HEK293T cells were incubated with FITC-dextran and then either maintained in amino acids (1AA) or starved (2AA) for 1 h. The rate of V-ATPase-dependent fluorescence quenching in lysosomes was determined as described previously. Values are expressed relative to the unstarved condition for each clone tested. *, p , 0.05; ns, p 0.05. Error bars represent S.E. D, Lysosomal V-ATPase activity for unstarved PKA Ca/Cb double knockout clones (PRKACA 2/2 B 2/2 ) is shown relative to the unstarved WT cells. ns, p 0.05. Error bars represent S.E. E, The rate of V-ATPase-dependent fluorescence quenching in FITC-dextran-loaded lysosomes was determined for WT and PKA Ca/Cb double knockout HEK293T cells (PRKACA 2/2 B 2/2 ) treated for 1 h with either DMSO (vehicle) or 50 mM H89 in the presence of amino acids. Values are expressed relative to the unstarved, vehicle-treated condition. *, p , 0.05; ns, p 0.05. Error bars represent S.E.
Since it is possible that the lack of a starvation-dependent increase in activity is due to elevated basal activity in the AKT1 knockout cells, we also analyzed basal activity relative to WT cells. As shown in Fig. 5C, there is variability in basal activity among all groups, including the negative-control clones, although mean basal activities were quite similar to those of the WT. Thus, the mean basal lysosomal V-ATPase activity, relative to the WT, was 1.08 6 0.12 in the negative control clones, 1.08 6 0.08 in the AKT1 knockout clones, and 1.01 6 0.09 in the AKT2 knockout clones. Importantly, among the four AKT1 knockout clones that showed no significant increase in V-ATPase activity upon starvation, no clone had a basal activity higher than 0.98 relative to the WT. Thus, for these clones, the lack of a starvation-dependent increase in lysosomal V-ATPase activity was not due to elevated basal activity.
We next wished to test whether the reduced response of lysosomal V-ATPase activity observed in the AKT1 knockout clones was due to an attenuated increase in V-ATPase assembly. To measure V-ATPase assembly, Western blot-ting was performed on membrane and cytosolic fractions isolated from cells that had either been maintained in amino acids or starved for 1 h. Because the peripheral V 1 domain is only present in membranes in fully assembled V-ATPase complexes, the abundance of V 1 subunits in the membrane fraction reflects the level of V-ATPase assembly (14,17). As shown in Fig. 6, amino acid starvation led to significantly increased assembly in WT HEK293T cells and an AKT2 knockout clone, while there was no significant change in the AKT1 knockout clone.
Since it is possible that the lack of a starvation-dependent increase in assembly is due to elevated basal assembly in the AKT1 knockout cells, we also analyzed basal assembly relative to WT cells. As shown in Fig. 6C, the AKT1 knockout clone did have significantly increased basal assembly relative to the WT, while the AKT2 knockout clone did not. This suggests that increased basal assembly may contribute to the failure to show increased V-ATPase assembly upon amino acid starvation in this AKT1 knockout clone. . AMPK is not required for increased lysosomal V-ATPase activity upon amino acid starvation. A, CRISPR-mediated genome editing was used to disrupt AMPK by targeting the PRKAA1 and PRKAA2 genes as described in Experimental procedures. Western blotting was performed on lysates from untransfected WT cells, clones transfected with nontargeting control plasmids (Neg), clones targeted for disruption of AMPKa1 followed by nontargeting control (PRKAA1 2/2 ), and clones targeted for disruption of AMPKa1 followed by disruption of AMPKa2 (PRKAA1 2/2 2 2/2 ). Isoform-specific AMPKa antibodies were used to assess protein knockdown, and ACC phosphorylation was used to measure AMPK activity. Vinculin was used as a loading control. Representative images are shown (n = 2). B, WT and AMPKa1/2 double knockout (PRKAA1 2/2 2 2/2 ) HEK293T cells were loaded with FITC-dextran and then either maintained in amino acids (1AA) or starved (2AA) for 1 h. The rate of V-ATPase-dependent fluorescence quenching in lysosomes was determined as described previously. Values are expressed relative to the unstarved condition for each clone tested. *, p , 0.05; ns, p 0.05. Error bars represent S.E. C, Lysosomal V-ATPase activity for unstarved AMPKa1/2 double knockout cells (PRKAA1 2/2 2 2/2 ) is shown relative to unstarved WT cells. *, p , 0.05; ns, p 0.05. Error bars represent S.E. D, The rate of V-ATPase-dependent fluorescence quenching in FITC-dextran-loaded lysosomes was determined for WT and AMPKa1/2 double knockout HEK293T cells (PRKAA1 2/2 2 2/2 ) treated for 1 h with either DMSO (vehicle) or 5 mM dorsomorphin (Dorso) in the presence of amino acids. Values are expressed relative to the unstarved, vehicle-treated condition. *, p , 0.05; ns, p 0.05. Error bars represent S.E.
These results, taken collectively with data from the AKT knockout fibroblasts and the effect of the AKT inhibitor MK2206, demonstrate that AKT is involved in the starvationdependent increase in lysosomal V-ATPase activity and assembly, and that this involvement depends upon the isoform of AKT expressed.

Discussion
Amino acid sensing is essential to cellular homeostasis, and the V-ATPase has emerged as a central component of the amino acid sensing machinery in mammalian cells (29). The V-ATPase is required for amino acid-dependent association of mTORC1 with the lysosomal membrane, where it is activated (30,31). Furthermore, our laboratory previously reported that amino acids modulate V-ATPase assembly and activity on lysosomes, and that this effect is PI3K-and mTORC1-independent (17). We also showed that this starvation-dependent increase in lysosomal V-ATPase activity is not required for amino acid-dependent changes in mTORC1 activity (17). Rather, the increase in lysosomal acidification observed upon amino acid starvation likely facilitates a rapid increase in lysosomal protein degradation, thereby helping to maintain amino acid homeostasis.
In the present study, we sought to identify the signaling pathways controlling lysosomal V-ATPase activity in response to changes in amino acid availability. We first investigated the roles of PKA and AMPK, because these kinases appear to be important for controlling V-ATPase assembly in response to changes in glucose levels in yeast and mammalian cells, respectively (12,14). Treatment of cells with the PKA inhibitor H89 robustly increases lysosomal V-ATPase activity in the presence of amino acids and prevents any further increase upon amino acid starvation. It should be noted that with the concentration of H89 at which PKA is fully inhibited, V-ATPase activity is likely saturated, which may be why no further increase is observed upon starvation. This result would suggest that PKA functions to suppress lysosomal V-ATPase activity in mammalian cells. In contrast, in yeast, PKA functions to increase V-ATPase assembly and activity upon glucose starvation (12). Based upon this pharmacological result, we expected that PKA Ca/Cb double knockout cells would have elevated lysosomal V-ATPase activity, which would be insensitive to amino acid starvation. Surprisingly, we observed that PKA Ca/Cb double knockout cells have normal basal levels of lysosomal V-ATPase activity and display a normal increase in activity upon amino acid starvation. This suggests that PKA does not function to regulate lysosomal V-ATPase activity in mammalian cells in response to starvation, and that H89 modulates V-ATPase activity by altering the activity of other targets. H89 was originally described as a Rho-associated protein kinase (ROCK) inhibitor and is now known to potently inhibit numerous other kinases as well (24). H89 was reported to induce autophagy in a PKAindependent manner, and this effect was partially abrogated by the inhibition of AKT (32). While lysosomal acidification was not measured in that study, it is tempting to speculate that the increased lysosomal V-ATPase activity we observed upon H89 treatment is related to the increased autophagic flux that these authors observed. Further studies will be necessary to elucidate the mechanism by which H89 increases lysosomal V-ATPase activity.
In contrast to H89, which increases lysosomal V-ATPase activity, we found that the treatment of HEK293T cells with the AMPK inhibitor dorsomorphin significantly lowers lysosomal V-ATPase activity and also prevents any increase upon amino acid starvation. Dorsomorphin was originally discovered as an Figure 5. AKT1 knockout HEK293T cells have reduced ability to increase lysosomal V-ATPase activity in response to amino acid starvation. A, 3 different guide sequences were used to target either AKT1 or AKT2 for CRISPR-mediated disruption in HEK293T cells. Western blotting was performed on lysates from untransfected WT cells, clones transfected with nontargeting control plasmids (Neg), clones targeted for disruption of AKT1 (AKT1 2/2 ), and clones targeted for disruption of AKT2 (AKT2 2/2 ). Isoformspecific AKT antibodies were used to assess protein knockdown, and vinculin was used as a loading control. Representative images are shown (n = 3). B, WT, nontargeted negative control (Neg), AKT1 knockout (AKT1 2/2 ), and AKT2 knockout (AKT2 2/2 ) HEK293T cells were incubated with FITC-dextran and then either maintained in amino acids (1AA) or starved for 1 h (2AA). The rate of V-ATPase-dependent fluorescence quenching in lysosomes was determined as described previously. Values are expressed relative to the unstarved condition for each clone tested. AKT2 knockout cells displayed a statistically significant increase in V-ATPase activity following amino acid starvation, whereas AKT1 knockout cells did not. *, p , 0.05; ns, p 0.05. Error bars represent S.E. C, Lysosomal V-ATPase activity for unstarved CRISPR negative control (neg), AKT1 knockout (AKT1 2/2 ), and AKT2 knockout (AKT2 2/2 ) cells is shown relative to unstarved WT cells. *, p , 0.05; ns, p 0.05. Error bars represent S.E. Figure 6. AKT1 knockout HEK293T cells fail to show increased V-ATPase assembly in response to amino acid starvation. A, WT HEK293T cells, an AKT1 knockout clone (AKT1 2/2 ), and an AKT2 knockout clone (AKT2 2/2 ) were maintained in amino acids or starved for 1 h, followed by cell fractionation, as described in Experimental procedures. Western blotting was performed on the membrane (M) and cytosolic (C) fractions using the indicated antibodies. B2 is a V 1 subunit and, therefore, indicates the abundance of the V 1 domain in each fraction. d1 is a V 0 subunit and was used as a membrane loading control, while vinculin was used as a cytosolic loading control. The relative abundance of V 1 in the membrane fraction versus the cytosolic fraction is a measure of V-ATPase assembly. Representative images are shown (n = 3). B, Quantification of Western blots described in panel A. The intensity of B2 in the membrane fraction was normalized to d1. The intensity of B2 in the cytosolic fraction was normalized to vinculin. The ratio of the normalized intensities gives the relative assembly. Values are expressed relative to the unstarved condition for each clone tested. WT and AKT2 knockout (AKT2 2/2 ) cells displayed a statistically significant increase in V-ATPase assembly following amino acid starvation, whereas AKT1 knockout (AKT1 2/2 ) cells did not. *, p , 0.05; ns, p 0.05. Error bars represent S.E. C, V-ATPase assembly for an unstarved AKT1 knockout clone (AKT1 2/2 ) and an unstarved AKT2 knockout clone (AKT2 2/2 ) is shown relative to unstarved WT cells. *, p , 0.05; ns, p 0.05. Error bars represent S.E.
AKT regulates V-ATPase activity upon amino acid starvation AMPK inhibitor but is now known to inhibit numerous kinases more potently than AMPK, making interpretation of results obtained with this compound challenging (23,24). Using CRISPR-mediated gene disruption, we found that AMPKa1/ a2 double knockout cells display normal levels of lysosomal V-ATPase activity and also show increased activity upon amino acid starvation. This suggests that in mammalian cells, AMPK is not required to maintain lysosomal V-ATPase activity and does not control the response to amino acid starvation. Recent studies have reported that AMPK is activated at the lysosome in a V-ATPase-dependent manner during energy stress (33). Therefore, while AMPK may not be required for changes to lysosomal V-ATPase activity upon amino acid starvation, it may be important during other forms of nutrient depletion, such as glucose starvation (14).
We previously showed that the treatment of cells with the PI3K catalytic inhibitor LY294002 does not prevent starvationinduced increases in V-ATPase assembly and activity (17). In response to external stimuli, such as growth factors and insulin, PI3K converts plasma membrane PI(4,5)P 2 to PI(3,4,5)P 3 , which recruits AKT for activation. Therefore, it was surprising that cells treated with the AKT allosteric inhibitor MK2206 do not show a starvation-dependent increase in lysosomal V-ATPase activity. It was further surprising that AKT activation (measured by Western blotting for AKT phospho-S473) does not increase upon amino acid starvation. These findings suggest that the existing pool of active AKT is sufficient to control the response of V-ATPase to amino acid starvation. Thus, LY294002 prevents any further AKT activation, but active AKT molecules may still be able to regulate downstream substrates until they are deactivated by phosphatases. Conversely, MK2206 directly inhibits all cellular AKT, which could explain the different responses to the two drugs.
A major downstream effector of AKT is mTORC1, which is activated at the lysosome following amino acid-dependent recruitment by a complex containing the V-ATPase (31). We previously showed that the starvation-dependent increases in V-ATPase activity and assembly still occur in the presence of mTORC1 inhibition with rapamycin (17). The fact that mTORC1 is not required for the increased V-ATPase activity upon amino acid starvation raises mechanistic questions of how AKT controls V-ATPase activity. AKT has been reported to directly bind the V-ATPase to increase endosomal acidification during rotavirus infection (34). This presents the possibility that AKT binding to the V-ATPase is a conserved mechanism to control activity in mammalian cells. Alternatively, AKT may act on intermediate proteins, such as assembly factors, to regulate lysosomal V-ATPase activity. Whether acting on the V-ATPase directly or indirectly, AKT may translocate to lysosomes to regulate lysosomal V-ATPase activity and/or assembly. There is a growing body of evidence to support a role for the intracellular translocation of AKT. Starvation of cells was shown to lead to lysosomal recruitment of AKT by pleckstrin homology domain-containing family F member 2 (Phafin2) and the serine/threonine kinase VRK2 (35,36). Moreover, lysosomal recruitment of AKT appears to be important for lysosomal acidification and autophagic flux (35). Although those studies were done with a more prolonged and complete starvation than was performed in the present study, they nevertheless provide further evidence in support of a role for AKT in regulating lysosomal acidification during nutrient limitation.
We have begun to investigate the role of AKT isoforms in controlling lysosomal V-ATPase activity in response to amino acid availability. AKT exists as three highly homologous isoforms, encoded by the genes AKT1, AKT2, and AKT3. While largely redundant in their activation and downstream substrates, genetic studies have revealed AKT isoform-specific phenotypes (18). Because AKT triple knockout cells are not viable for long-term culture, we first took advantage of mouse fibroblast lines, which express only one of the three human AKT isoforms and in which the endogenous mouse AKT has been disrupted (28). We found that cells expressing only AKT1 or only AKT3 display increased lysosomal V-ATPase activity upon amino acid starvation, whereas cells expressing only AKT2 show no change upon starvation. This result provides additional evidence for a role of AKT in the regulation of lysosomal V-ATPase activity and suggests that particular AKT isoforms are important for this process. Since we find HEK293T cells have no detectible AKT3 expression, we hypothesized that knocking out AKT1 (leaving AKT2 intact) would abrogate the increased V-ATPase activity upon amino acid starvation. Indeed, CRISPR-mediated AKT1 knockouts display a significantly reduced response to amino acid starvation. Although some starvation-dependent increase in activity is still observed in some clones, it is possible that these cells still express low levels of AKT1, or that other mechanisms exist that can compensate for AKT1 loss under certain circumstances.
Assembly is a major form of V-ATPase regulation, and we previously showed that amino acid starvation leads to a significant increase in V-ATPase assembly in mammalian cells (17). Other V-ATPase-dependent mechanisms that can influence lysosomal acidification include changes to the coupling efficiency of the enzyme and changes in the density of V-ATPase complexes per lysosome (1,4). Therefore, it was important to test whether the reduced response to amino acid starvation in the AKT1 knockouts is due to a failure to increase V-ATPase assembly. We find that while WT and AKT2 knockout cells display a significant increase in V-ATPase assembly upon amino acid starvation, AKT1 knockout cells show no significant change. However, the failure to show increased assembly upon starvation may be because this AKT1 knockout clone displays significantly elevated basal V-ATPase assembly. This clone does not have elevated basal lysosomal V-ATPase activity, suggesting that the assembled lysosomal V-ATPases are not active. Alternatively, since assembly was measured in whole-cell lysates, whereas activity is measured in lysosomes, it is possible that the observed basal elevation is due to the assembly of V-ATPase complexes in other cell membranes. However, studies using several different types of mammalian cells have reported that the V 0 domain (and, thus, assembly-competent V-ATPase) is most abundant in lysosomal fractions (37)(38)(39). Furthermore, we have previously found that changes in lysosomal V-ATPase activity strongly correlate with changes in total cellular V-ATPase assembly in HEK293T cells (14,17). Therefore, it is likely that the lysosomal pool of V-ATPases contributes to a significant fraction of the observed change in total cellular V-ATPase assembly.
In summary, we have shown that AKT plays an important role in modulating V-ATPase-dependent lysosomal acidification in response to amino acid starvation. Future studies will investigate the mechanism by which AKT regulates V-ATPase assembly and activity in response to nutrient stress and test the hypothesis that this response is a primitive mechanism to rapidly increase protein degradation upon starvation to replenish the supply of free amino acids. Understanding how cells sense and respond to amino acid availability may enable the development of treatments for conditions in which nutrient homeostasis is dysregulated, such as cancer and diabetes. In addition, understanding how lysosomal acidification is controlled may lead to the development of therapies for lysosomal storage and neurodegenerative diseases, such as Alzheimer's disease, which are characterized by the failure to degrade toxic protein aggregates.

Amino acid starvation media
A 103 stock solution of amino acids was prepared by dissolving individual powdered amino acids in PBS (no. 10010; Gibco), filter sterilized, and stored as aliquots at 220°C. The final (13) amino acid concentrations were the following: 0.05 mM L-alanine, 0.7 mM L-arginine hydrochloride, 0.05 mM L-asparagine, 0.05 mM L-aspartic acid, 0.2 mM L-cysteine, 0.05 mM L-glutamic acid monosodium salt, 2.5 mM L-glutamine, 0.25 mM glycine, 0.15 mM L-histidine hydrochloride monohydrate, 0.42 mM Lisoleucine, 0.45 mM L-leucine, 0.5 mM L-lysine hydrochloride, 0.12 mM L-methionine, 0.21 mM L-phenylalanine, 0.15 mM Lproline, 0.25 mM L-serine, 0.45 mM L-threonine, 0.04 mM L-tryptophan, 0.21 mM L-tyrosine, and 0.45 mM L-valine. All amino acids were purchased from Sigma-Aldrich. DMEM/F12 without amino acids (no. D9811-01; US Biological) was prepared in ultrapure water, brought to pH 7.2, and filter sterilized. Prepared amino acid-free medium was discarded after 1 month. The 103 amino acid solution (or an equivalent volume of PBS) was added to amino acid-free medium immediately before each starvation experiment, for a final concentration of 13.
V-ATPase-dependent lysosomal proton pumping 2 3 10 6 cells were seeded onto 6-cm cell culture dishes that had been coated with 0.1 mg/ml poly-D-lysine (PDL; no. A-003-E; Millipore) and allowed to attach overnight. The following day, the growth medium was replaced with medium containing 2 mg/ml FITC-dextran (average molecular weight, 40 kDa; no. FD40; Sigma-Aldrich), and the cells were incubated overnight to allow the uptake of the dextran by fluid-phase endocytosis. The next day, the cell monolayers were rinsed with warm PBS and incubated with fresh medium without FITC-dextran for 1 h to chase the internalized FITC-dextran to lysosomes. Following the chase period, cell monolayers were rinsed with warm PBS and then incubated for 1 h with starvation medium containing 13 amino acids or PBS in the presence or absence of the indicated inhibitors. Following the 1 h treatment period, cell monolayers were rinsed with ice-cold PBS and scraped into 200 ml ice-cold activity assay buffer (125 mM KCl, 50 mM sucrose, 20 mM HEPES, pH 7.5, 1 mM EDTA, 1 mM PMSF, 5 mg/ ml leupeptin, 2 mg/ml aprotinin, 1 mg/ml pepstatin). The cell suspensions were centrifuged at 300 3 g for 5 min at 4°C, and the resulting pellets were resuspended in 400 ml fresh ice-cold activity assay buffer. Cells were lysed by passing through a 27gauge needle 20 times, and the crude lysates then were cleared of unbroken cells and nuclei by centrifugation at 1,000 3 g for 10 min at 4°C. The cleared supernatants then were centrifuged at 16,100 3 g for 15 min at 4°C to sediment FITC-dextrancontaining lysosomes. The resulting pellets were resuspended in 100 ml activity assay buffer.
To measure V-ATPase-dependent lysosomal proton pumping, ;10-20 ml of sample was added to a quartz cuvette (101-058-40-QS; Hellma Analytics) containing activity assay buffer prewarmed to 37°C. Samples were excited at 490 nm, and fluorescence emission was continuously collected at 520 nm using a Perkin-Elmer LS50B luminescence spectrophotometer. V-ATPase-dependent proton pumping was initiated by adding magnesium-ATP to a final concentration of 1 mM ATP and 2 mM Mg 21 . The rate of fluorescence quenching was determined by linear regression using GraphPad Prism 8.0, and the first 10 s of quenching were used for statistical analysis.
Cell fractionation 5 3 10 6 cells were seeded onto 10-cm, PDL-coated culture dishes and allowed to attach overnight. The next day, cell monolayers were rinsed with warm PBS and then incubated for 1 h with starvation medium containing 13 amino acids or PBS. Following the 1 h treatment period, cell monolayers were rinsed with ice-cold PBS and scraped into 500 ml ice-cold fractionation buffer (250 mM sucrose, 10 mM HEPES, pH 7.2, 1 mM EDTA, 1 mM PMSF, 5 mg/ml leupeptin, 2 mg/ml aprotinin, 1 mg/ml pepstatin, 2 mM b-glycerophosphate, 1 mM NaF). The cell suspensions were centrifuged at 300 3 g for 5 min at 4°C, and the resulting pellets were resuspended in 500 ml fresh icecold fractionation buffer. Cells were lysed by passing through a 27-gauge needle 20 times, and the crude lysates then were cleared of unbroken cells and nuclei by centrifugation at 1,000 3 g for 10 min at 4°C. The cleared supernatants then were centrifuged at 100,000 3 g for 30 min at 4°C using a Beckman Coulter L7-55 ultracentrifuge to sediment the membrane fractions. The resulting supernatants containing the cytosolic fractions were concentrated using Amicon Ultra 10K centrifugal filter units (UFC501096; Millipore) according to the manufacturer's instructions. The membrane pellets were solubilized in fractionation buffer supplemented with 1% SDS.
For fractionation experiments, band intensities were quantified using ImageJ 1.52p. The intensity of subunit B2 in the cytosolic fraction of each sample was normalized to the intensity of the corresponding cytosolic loading control, vinculin. The intensity of subunit B2 in the membrane fraction was normalized to the intensity of the corresponding membrane loading control, subunit d1. V-ATPase assembly was determined by taking the ratio of normalized intensities of B2 in the membrane and cytosolic fractions.

CRISPR-mediated disruption of AMPK
The strategy to disrupt AMPK was to target first the a1 isoform of the catalytic subunit, encoded by PRKAA1, and then the a2 isoform, encoded by PRKAA2. The first exon of PRKAA1 was searched using the online tool at crispr.mit.edu. The guide sequence GAAGATCGGCCACTACATTC was chosen because of its high specificity score and cloned into the plasmid pSpCas9(BB)-2A-Puro (PX459; no. 48139; Addgene) according to the method of Ran et al. (40). Plasmids containing either the guide sequence or empty backbone were transfected into HEK293T cells using Lipofectamine 2000 (no. 11668; Invitrogen) according to the manufacturer's instructions. Successfully transfected cells were selected for growth in 1 mg/ml puromycin and then seeded at 0.5 cells/well into 96-well plates by serial dilution to isolate individual clones. Clonal lines were expanded until there were enough cells to determine protein expression of AMPKa1 by Western blotting. A single AMPKa1 knockout clone was chosen for a second round of CRISPR to target PRKAA2, this time by cloning a published guide sequence, GAAGATCGGACACTACGTGC (41), into pSpCas9(BB)-2A-GFP (PX458; no. 48138; Addgene) according to the method of Ran et al. (40). Plasmids containing either the guide sequence or empty backbone were transfected into cells using Lipofectamine 3000 (no. L3000; Invitrogen) according to the manufacturer's instructions. The following day, cells were detached by trypsinization, and GFP-positive cells were sorted individually into the wells of 96-well plates by FACS. Clonal lines were subsequently expanded until there were enough cells to determine protein expression of AMPKa2 by Western blotting.

CRISPR-mediated disruption of PKA and AKT
The protein-coding regions of the PRKACA, PRKACB, AKT1, and AKT2 genes were searched using the online tool CHOPCHOP (42), which scores potential single guide RNAs according to multiple parameters, including specificity, cleavage efficiency, and the likelihood of causing a frameshift. The following guide sequences were chosen to target PRKACA: ACGAATCAAGACCCTCGGCA, TCGTTCAAACTGATC-CAAGT, and TGAACCTTCCGATCCGCCGT. The following guide sequences were chosen to target PRKACB: GCTGCATA-GAACCGTGCATG and GCTACAATAAGGCAGTGGAT. The following guide sequences were chosen to target AKT1: GAGCGACGTGGCTATTGTGA, TCACGTTGGTCCACA-TCCTG, and GACAACCGCCATCCAGACTG. The following guide sequences were chosen to target AKT2: TCTCG-TCTGGAGAATCCACG, GAGCCACACTTGTAGTCCAT, and GAGATAGTCGAAGTCATTCA. Plasmids containing either guide sequences or the empty backbone were transfected into HEK293T cells using Lipofectamine 3000. The following day, cells were detached by trypsinization, and GFP-positive cells were sorted individually into the wells of 96-well plates by FACS. Clonal lines were subsequently expanded until there were enough cells to determine protein expression by Western blotting. For PKA, a single PKA Ca knockout clone was selected for a second round of CRISPR to disrupt PKA Cb.
Statistics p values were calculated using a two-tailed Student's t test.

Data availability
All data are available upon request to the corresponding author, Dr. Michael Forgac.