Identification of intracellular cavin target proteins reveals cavin-PP1alpha interactions regulate apoptosis

Caveolae are specialized domains of the plasma membrane. Formation of these invaginations is dependent on the expression of Caveolin-1 or -3 and proteins of the cavin family. In response to stress, caveolae disassemble and cavins are released from caveolae, allowing cavins to potentially interact with intracellular targets. Here, we describe the intracellular (non-plasma membrane) cavin interactome using biotin affinity proteomics and mass spectrometry. We validate 47 potential cavin-interactor proteins using a cell-free expression system and protein-protein binding assays. These data, together with pathway analyses, reveal unknown roles for cavin proteins in metabolism and stress signaling. We validated the interaction between one candidate interactor protein, protein phosphatase 1 alpha (PP1α), and Cavin-1 and -3 and show that UV treatment causes release of Cavin3 from caveolae allowing interaction with, and inhibition of, PP1α. This interaction increases H2AX phosphorylation to stimulate apoptosis, identifying a pro-apoptotic signaling pathway from surface caveolae to the nucleus.


Supplementary Figure 7. PP1 translocation to the cytosol following hypo-osmotic treatment in MCF-7 cells.
MCF7 cells treated with hypo-osmotic medium were fixed for immunofluorescence of PP1a visualised under confocal microscopy. Images were inverted to grey scale and the nuclei were outlined in red. Scale bar, 10 µm.

Supplementary Figure 8. PLA controls.
(a-b). Fluorescence microscopy analysis of PLA signals generated in A431 cells either left untreated or treated with hypo-osmotic medium or UV treatment in the absence of primary antibodies, PLA probe controls or individual Cavin1, Cavin3 and PP1α antibody as negative controls for PLA signals with the existence of probes and Cavin3/PP1α or Cavin1/PP1α antibody pairs. Representative images are from at least two independent experiments as shown.
(c). PLA controls following knockdown of PP1a and Cavin3. Three independent experiments were performed. Scale bar, 10 μm. Confocal microscopy images represent the PLA signals in MDA-MB231 cells with and without UV treatment. PLA signals alone is presented as inverted images (left panel). The merged images include PLA signal (red), Alexa Fluor 488-Phalloidin (green) and DAPI (blue) channels. Phalloidin staining was used to identify the borders of cells. The nucleus is indicated by DAPI staining. Scale bar, 10 µm. Images are representative of two independent experiments performed. (a). Representative immunofluorescence images of Cavin3 (green) and PP1α (red) in untreated or UV treated A431 cells from three independent experiments. DNA was outlined by blue circle. Scale bar, 10 μm.
(b). GFP Trap assays of A431 cells transfected with GFP-Cavin3 or GFP with and without UV treatment western blotted with anti-PP1a antibodies. Transfection efficiency was confirmed with GFP antibodies. a-Tubulin was used as the loading control. Western blots are representative of three independent experiments.
(g). LDH release as a % of the total LDH following UV treatment was calculated where Ctrl vs GFP-

Supplementary Discussion: Potential functions for Cavin interacting proteins in cells Potential non-caveolar cavin interacting proteins and their localization in the cell
The proteins identified as potential non-caveolar cavin interacting proteins are distributed throughout the cell, with 38% of identified proteins localized predominantly in the cytosol, 33% of proteins localized predominantly in the nucleus, 6% of proteins localized predominantly in the endoplasmic reticulum, plasma membrane and mitochondria, 3% of proteins localized predominantly to the cytoskeletal and nucleolus, 2% of proteins localized predominantly in the Golgi apparatus, and less than 1% of proteins localized predominantly to lysosomes, caveolae and peroxisomes, respectively.
These data suggest that cavin proteins can interact with many other intracellular compartments when released from caveolae with a propensity for interaction with many nuclear and cytosolic proteins that freely shuttle between these two compartments in response to stress stimuli. For example, several proteomic studies have identified changes in the subcellular localization of Cavin1 and Cavin3 in response to cellular stressors including the oxidative stress induced nuclear accumulation of Cavin1 and Cavin3 in human fibroblasts (10,11). In addition, DDX21 and other DDX proteins (DDX1 and DDX5) as potential non-caveolar cavin proteins can translocate from the nucleolus to the nucleoplasm in response to stress stimuli (12). In the course of the review of this manuscript, a paper by Mendoza-Topaz et al. (13) applied BioID to identify proteins that interact with Cavin1. Candidate Cavin1-interacting proteins included a number of proteins located predominantly in the nucleus with functions related to the role of Cavin1 in regulating ribosomal RNA synthesis (14). Collectively, these findings suggest that the reach of the caveola function now extends from the plasma membrane to intracellular targets in compartments such as the cytoplasm, nucleus and nucleolus and specific cellular processes which could not have been envisaged before.

Metabolism and Protein Synthesis
The BioID/MS approach identified a number of key glycolytic enzymes as potential cavin-interacting proteins, including alpha enolase (ENO1), fructose bisphosphate aldolase A (ALDOA), pyruvate kinase (PKM) and phosphoglycerate kinase A (PGK1). Of particular interest is Pyruvate Kinase (PKM), the rate limiting enzyme in this process. These findings have potential importance for the role of cavins in cancer (15)(16)(17)(18)(19) as transformed cells predominantly metabolize glucose by glycolysis to produce energy in order to sustain their increased metabolic requirement, a process known as the Warburg effect (20).
Recent reports suggest that the in vitro consequences of the loss of Cavin3 by gene knockout studies include the induction of Warburg metabolism (aerobic glycolysis) and the in vivo consequence of Cavin3 loss in a mouse model system is increased lactate production (21) suggesting that Cavin3 may play an important role in this process.
The metabolic shift from oxidative phosphorylation to aerobic glycolysis is partially achieved by a switch in the splice isoforms of PKM from PKM1 to PKM2 that is demonstrated here as a potential Cavin3 interacting protein (22). Switching from PKM1 to PKM2 promotes aerobic glycolysis and thus provides a selective advantage for tumor formation. This switching mechanism that involves the generation of alternative splicing of two mutually exclusive exons for PKM is controlled by heterogenous nuclear ribonucleoprotein (hnRNP) family members along with the polypyrimidine tract binding protein (PTB; known also as hnRNPI) (23). Interestingly, BioID/MS analysis of potential eEF2-K phosphorylates eukaryotic Elongation Factor 2 (eEF2) on Threonine (Thr)-56, thereby inactivating this key elongation factor that blocks protein translation and inhibits protein synthesis.
Indeed, eEF2 was identified by the BioID/Mass Spectrometry analysis as a potential Cavin3 interacting protein. Additionally, eEF2-K confers tolerance to stress conditions in cancer cells (25). Collectively, these findings suggest involvement of Cavin3 in stress responsive pathways that influence a number of processes including metabolism and protein translation.