CIP2A deficiency activates glycolytic metabolism in NSCLC cells
Our previous work showed that CIP2A was elevated and had an important role in human NSCLC 19, 33, 34. During daily culture of NSCLC cell lines, we observed that the medium of CIP2A-knockdown cells turned orange much more rapidly than that of control cells, even at the same confluence (Fig. 1a). This observation indicated an increased production of acidic metabolites in CIP2A-knockdown cells, which prompted us to investigate the role of CIP2A in glucose metabolism. We applied XFe Seahorse energy metabolic stress assays to determine the impact of CIP2A on glucose metabolism, and the rate of glycolysis was assessed by ECAR. As anticipated, the glycolysis level, maximum glycolytic capacity, and glycolytic reserve were significantly increased in CIP2A-knockdown H1299 cells (Fig. 1b), whereas the reverse was observed in CIP2A-overexpressing cells (Fig. 1c). In CIP2A knockdown H1299 cells, a significant reduction in the basal mitochondrial respiration level, maximum rate of oxidative phosphorylation (OXPHOS) and reserve capacity of OXPHOS, as measured by OCR, were observed (Fig. 1d). In contract, a significant increase in the basal mitochondrial respiration level, maximum rate of OXPHOS and reserve capacity of OXPHOS, were seen in H1299 cells that were transfected with pCDH-CMV-CIP2A (Fig. 1e).
Identification of the glycolytic enzyme PKM2 as an endogenous CIP2A-associated protein
To gain mechanistic insights, endogenous CIP2A was immunoprecipitated from extracts of A549 and H1299 cells, the eluates were resolved by SDS/PAGE, and several major bands not seen in the control lane were silver-stained and subjected to MS sequencing to identify CIP2A-interacting proteins. Interestingly, PKM was discovered as a CIP2A-binding protein (Fig. 2a). To verify this interaction, co-immunoprecipitation (Co-IP) analyses were performed using cell lysates of A549 and H1299 cells, and we showed that CIP2A could pull down PKM2 and vice versa (Fig. 2b). Additionally, GST-CIP2A could pull down PKM2 from H1299 cell lysates (Fig. 2c), suggesting that CIP2A could interact directly with PKM2. We detected whether CIP2A interacts with PKM1, despite the much lower expression of PKM1 in NSCLC cells than in SCLC cells (Fig. 2d), which was consistent with a previous study 35. We found that PKM1 was unable to interact with CIP2A (Fig. 2e). To further confirm the interaction between CIP2A and PKM2, we detected the colocalization of CIP2A and PKM2 by IF staining in A549 and H1299 cells. The results indicated that CIP2A and PKM2 were primarily localized in the cytoplasm and that CIP2A colocalized with PKM2 (Fig. 2f). We next determined the interacting domain for the CIP2A-PKM2 interaction. As shown in Fig. 2g, a series of truncated mutants fused to the HA tag for PKM2 were generated to test the binding affinity with Flag-tagged CIP2A in 293T cells by Co-IP. We showed that deletion of the A2 domain (amino acids between 218 and 388) abrogated the interaction between PKM2 and CIP2A, whereas the other fragments of PKM2 retained the ability to bind CIP2A with various affinities (Fig. 2h). These results demonstrated that CIP2A could bind PKM2 at the A2 domain, and the functional consequence of this interaction warrants further investigation.
CIP2A stabilizes PKM2 oligomerization
To explore how CIP2A regulates PKM2, we first found that knocking down CIP2A had no effect on the total PKM2 protein expression (Fig. S1a). Then, we clarified whether CIP2A influences PKM2 oligomerization and pyruvate kinase activity. Size exclusion chromatography clearly indicated that the tetrameric formation of PKM2 was blocked in the shCIP2A group compared with the control group in A549 cells (Fig. 3a), and CIP2A overexpression skewed PKM2 towards high-molecular-weight forms in 293T cells (Fig. 3b). Further crosslinking experiments revealed that PKM2 migrated as a single band at ~60 kDa (monomer) in the absence of crosslinking reagent, indicating that SDS treatment resulted in a complete dissociation of potential PKM2 multimers (Fig. 3c). Following crosslinking conditions (0.01% glutaraldehyde at 37 °C for 9 minutes), tetrameric (240 kDa) PKM2 was increased after CIP2A overexpression in H1299 cells (Fig. 3c). Furthermore, the tetramer was significantly reduced and replaced by dimeric (120 kDa) and monomeric forms after CIP2A knockdown in H1299 cells (Fig. 3d). To validate whether CIP2A influences the interaction between the subunits of PKM2 in cells, we found that the amount of endogenous PKM2 that coprecipitated with HA-PKM2 was reduced in cells transfected with shCIP2A (Fig. 3e). These observations suggested that CIP2A facilitates PKM2 tetramer assembly.
CIP2A enhances pyruvate kinase activity and hinders the nuclear translocation of PKM2
We then assessed whether the PK activity of PKM2 is interfered with CIP2A. In vitro reactions showed that purified CIP2A protein from bacteria increased bacterial PK activity of PKM2 in a dose-dependent manner (Fig. 3f). On the contrary, there was a nearly 50% decrease in PKM2 activities in lysates of CIP2A-depleted H1299 and A549 cells (Fig. 3g). In addition, PK activity was decreased in H1299 cells treated with the PKM-specific inhibitor Compound 3K 36, and this response was attenuated by CIP2A overexpression (Fig. S1b). To further validate the cellular distribution of PKM2, we analyzed confocal microscopy images by comparing the PKM2 signaling intensity overlapping with high DAPI staining (nuclear region) by ZEN software 37. A higher-intensity PKM2 signal from the nucleus was visualized in H1299 cells after knockdown of CIP2A (Fig. 3h). Consistent with the IF findings, western blot analysis of subfractionated cellular compartments indicated that CIP2A deficiency substantially induced the nuclear accumulation of PKM2 in the lysates collected from H1299 and A549 cells (Fig. 3i). The mild reduction of PKM2 in the cytosolic fractions under CIP2A deletion may be responsible for the nuclear PKM2 is indeed a small portion compared with the quantity of cytoplasmic PKM2 (Fig. 3i). These results indicated that CIP2A increases PK activity and inhibits the nuclear translocation of PKM2.
Metabolic flux analysis was further carried out to characterize the involvement of PKM2 in metabolic reprogramming regulated by CIP2A. The PKM2-specific activator TEPP-46 38 entirely restored the PK activity reduced by CIP2A knockdown (Fig. 3j) and abolished the shCIP2A-mediated ECAR elevation (Fig. 3k). Consistent with the altered glycolysis phenotype, CIP2A deficiency-inducible mitochondrial respiration reduction, as measured by OCR, was blocked by TEPP-46 treatment (Fig. S1c). Additionally, TEPP-46 also reversed the acidic culture medium induced by CIP2A depletion in A549 cells (Fig S1d). These data suggested that CIP2A increases PK activity and impedes the nuclear translocation of PKM2, which supports the notion that CIP2A modulates metabolic reprogramming in NSCLC cells by intervening in multiple aspects of the PKM2 state and functions.
PP2A regulatory subunit B56α controls PKM2 activity and reprogrammed glycolysis
PP2A is a ubiquitously expressed heterotrimer containing the scaffolding A subunit, the catalytic C subunit, and the substrate specificity-determining regulatory B subunit 39, and CIP2A inhibits the phosphatase activity of PP2A 8. We found that siCIP2A resulted in reduced tetrameric and increased dimeric/monomeric PKM2, whereas knockdown of PP2A A subunit by siRNA increased tetrameric and reduced dimeric/monomeric PKM2, and co-transfection of siPP2A-Aα/β attenuated the effects induced by CIP2A depletion (Fig. S2a). The B’ family of PP2A regulatory subunits is composed of a large array of different members, and only a subset of the PP2A subunit directly interacts with and is inhibited by CIP2A 40. We validated which B subunit mediates the modulation of PKM2 activity by CIP2A. We identified that only B56α (PPP2R5A) interacts with ectopic expression of HA-PKM2 (Fig. 4a). Co-IP analysis of endogenous proteins indicated that B56α, but not B56γ (PPP2R5C) or B56ε (PPP2R5E), interacted with PKM2 (Fig. 4b). Further IF analysis demonstrated colocalization between B56α and PKM2 in A549 and H1299 cells (Fig. 4c). These results suggested that CIP2A, PKM2 and PP2A might form a ternary complex (Fig. 4d) to regulate cell metabolism.
A short linear motif, [LMFI]xx[ILV]xEx (where x is any amino acid), acts as the preferred docking site for the complementary pocket conserved in the B56 subunit, whereas substitution of the two best consensus residues at positions 1 and 4 to alanine would remarkably reduce the binding affinity 41, 42. We therefore analyzed the amino acid sequence of PKM2 and found two possible B56α binding motifs in the A2 domain (Fig. 4e). We constructed two mutated versions of the two potential binding motifs and found that F280A/I283A, but not L218A/V221A, completely disrupted the interaction between B56α and HA-PKM2, which could be the binding site (Fig. 4f).
We tested the effects of B56α on PKM2 PK activity and found that depletion of B56α notably upregulated the PKM2 PK activity reduced by CIP2A knockdown (Fig. 4g). Importantly, B56α deficiency substantially blocked CIP2A depletion-elicited PKM2 nuclear distribution and ECAR (Fig. 4h, i). In parallel, the increased OXPHOS level (Fig. S2b) and the yellow discoloration of the culture medium (Fig. S2c) due to CIP2A depletion were also reversed by co-depletion of B56α. These results indicate that the regulation of multiple functions of PKM2 and metabolic reprogramming are dependent on PP2A and specifically on the B56α regulatory subunit.
PKM2 serine 287 is a novel phosphorylation site
We next assessed the potential B56α-mediated posttranslational modifications of PKM2. Given that many known phosphorylation sites were found in the vicinity of the consensus sequence, especially the S/T residues at positions 2/7/8/9, we generated Ala substitution mutants of Ser harbored near the two motifs and transfected into 293T cells with these mutants individually. The mutated proteins were purified by IP followed by western blotting with an anti-phosphoserine antibody. We found that mutation of S287 significantly reduced PKM2 phosphorylation, while the other 3 mutants only slightly reduced the overall phosphorylation of PKM2 (Fig. 5a). Additionally, MS also suggested that S287 is the most promising candidate serine phosphorylation site (Fig. 5b), and a genomic analysis showed that S287 is highly conserved among different species through evolution (Fig. 5c), demonstrating that S287 is a major phosphorylation site under this condition.
To study whether the phosphorylation status at S287 affects CIP2A/PP2A-mediated dimer/tetramer conversion of PKM2, we fractionated 293T cell lysates by gel filtration analysis and found that the distribution of total WT PKM2 spread throughout multiple fractions and that coexpression of CIP2A further switched PKM2 from monomer and dimer to tetramer (Fig. 5d). Notably, S287-phosphorylated PKM2 was only detected in a low molecular-weight fraction, and ectopic expression of CIP2A failed to shift fractions of PKM2 to higher molecular weight complexes (Fig. 5d). These results demonstrated a strong effect of phosphorylation at S287 in promoting tetramerization of PKM2.
CIP2A modulates PKM2 phosphorylation at S287
To study whether S287 is phosphorylated in vivo, we generated an antibody specific to phospho-S287 and carried out a series of experiments to test its specificity. The dot plot assay showed that the PKM2 phospho-S287 antibody preferentially detected the phosphorylated peptide but not the unmodified peptide and the signal gradually improved with increasing peptide concentration (Fig. S3a). Western blotting and IHC assays indicated that the staining was absent when the antibody was preincubated and neutralized with the phosphorylated peptide compared with the unmodified peptide (Fig. S3b, c). In parallel, when PKM2 was knocked down by expression of a shRNA, the band disappeared, as detected by the phospho-S287 antibody (Fig. S3d). These results verified the specificity of the antibody.
We showed that CIP2A silencing remarkably downregulated phospho-S287 (Fig. 5e), while overexpression of CIP2A elevated endogenous S287-phosphorylated PKM2 and could be reversed by coexpression of B56α (Fig. 5f). We next carried out IHC staining analysis of 15 primary human lung adenocarcinoma tissue samples using a NSCLC tissue microarray (Table. S3) to examine the clinical relevance of PKM2 S287 phosphorylation. We observed that phospho-S287 predominantly localized to the cytoplasm and was hardly detected in the nucleus (Fig. 5g, left panel). Notably, the expression levels of PKM2 pS287 and CIP2A were correlated with each other (Fig. 5g, right panel). By using the Online Survival Analysis Software43 (http://kmplot.com/analysis/index.php?p=service&cancer=lung), we found that patients with CIP2Ahigh/PKM2high had worse prognosis than patients with CIP2Ahigh/PKM2low (Fig. 5h, left), and patients with PKM2high/CIP2Ahigh exhibited poorer clinical outcome than patients with PKM2high/CIP2Alow (Fig. 5h, right).
S287 is critical to PKM2 activity
To obtain insights into the function of S287 phosphorylation of PKM2 regulated by CIP2A, we first replaced endogenous PKM2 with either shRNA-resistant HA-PKM2, the phospho-mimetic mutant S287D HA-PKM2, or S287A HA-PKM2 in H1299 cells (Fig. 5i, j). Subcellular fractionation analysis indicated that S287D PKM2 accumulated predominantly in the cytoplasm and that resistance to CIP2A depletion-induced nuclear translocation (Fig. 5i). In contrast, compared to WT HA-PKM2-expressing cells, S287A HA-PKM2 presented lower PK activity (Fig. 5j), lower OXPHOS (Fig. 5k), and higher ECAR (Fig. S3e), and was incapably modulated by CIP2A ectopic expression (Fig. 5j, k, Fig. S3e). Additionally, H1299 cells expressing S287A HA-PKM2 maintained more acidic culture medium regardless of CIP2A expression (Fig. S3f). These results revealed that PKM2 phosphorylation at S287 is required for the change in oligomerization, nuclear translocation, PK activation and glucose flux regulated by CIP2A.
S287 facilitates S37 in modulating PKM2 nuclear localization
We further used the other two available phospho-specific antibodies against PKM2 to determine whether it is influenced by CIP2A. In H1299 and A549 cells, we found that CIP2A depletion had no effect on the phosphorylation level of PKM2 Y105 but significantly increased S37 phosphorylation and downregulated S287 phosphorylation (Fig. 5e). We tested whether these observations were associated with PP2A, and found that ectopically expressed CIP2A reduced phospho-S37 and increased phospho-S287 and was reversed by coexpression of B56α (Fig. 5f); conversely, depletion of the PP2A A subunit blocked CIP2A silencing-induced phospho-S37 upregulation and phospho-S287 downregulation (Fig. 6a). These results seemed inconsistent with the fact that CIP2A inhibits PP2A-mediated serine/threonine dephosphorylation and prompted us to explore whether there might be crosstalk between the phosphorylation of S37 and S287.
Previous studies indicated that the phosphorylation of PKM2 at S37 was ERK1/2 dependent and could promote PKM2 binding to importin α5, and translocated to the nucleus 32. Interestingly, crystal structure analysis (PDB code 3SRD) revealed that the ERK1/2 binding sites (Ile 429/Leu 431) and the pivotal residues Arg399/400 of the nuclear localization signal (NLS) sequence, which bind to importin α5, are buried in the prominent C-C interface (Fig. 6b), whereas CIP2A-induced tetramerization may obstruct accessibility to PKM2, and the dimer stage of PKM2 is likely exposed to recruit ERK1/2 and importin α5, which promote S37 phosphorylation and nuclear PKM2 translocation. We found that ectopic CIP2A expression caused a clear reduction of the binding of HA-PKM2 to endogenous ERK1/2 and importin α5, whereas the HA-PKM2 S287A mutant, which presented in a dimer state, reinforced this interaction and was largely unaffected by transfection with Flag-CIP2A (Fig. 6c). Consistent with previous described 32, S37A PKM2 failed to transport into the nucleus (Fig. 6d); surprisingly, S37A mutant also abrogated the PKM2 nuclear translocation induced by CIP2A knockdown (Fig. 6d). These results suggested that both S287 and S37 are required for nuclear translocation of PKM2 caused by CIP2A depletion.
Modification of PKM2 S287 controls cell proliferation
We then examined the role of PKM2 S287 phosphorylation on cell proliferation, and found that compared to cells expressing WT PKM2, cells expressing S287A PKM2 exhibited enhanced, while cells expressing phosphorylation mimetic S287D PKM2 showed reduced cell proliferation (Fig 7a). Consistent with these observations, we found by the three-dimensional cell culture assays seeding the same number of cells, that H1299 cells expressing S287A PKM2 presented the greatest number of tumor spheroids compared with cells expressing WT and S287D transcripts of PKM2 (Fig.7b). Notably, HA-PKM2 S287A mutant enhanced the expression of Myc and phosphorylation of STAT3 in H1299 cells (Fig. 7c).
Given that PKM2 S287 dephosphorylation displays a significant growth advantage and may counteract the tumor suppression effect of CIP2A inhibition, PKM2 activation and Warburg effect inhibition could potentiate the tumor-suppressive effect of CIP2A depletion. We tested this possibility and found that in H1299 cells expressing WT PKM2, CIP2A silencing significantly increased sensitivity of the cells to TEPP-46 and a glycolysis inhibitor 2-DG, but no such phenomena were observed in S287D cells (Fig. 7d), supporting that S287 dephosphorylation restricts the tumor suppressive effect of CIP2A deficiency.
Combinatory modifications of CIP2A and PKM2 inhibits tumor growth in vivo
As an alternative approach for depleting CIP2A, we induced CIP2A degradation using celastrol, a small natural compound previously discovered by our group that can trigger proteasomal degradation of CIP2A with the carboxyl terminus of Hsp70-interacting protein (CHIP) as the E3 ligase 19. We showed that in celastrol-treated cells, CIP2A and p-S287 PKM2 were downregulated while p-S37 PKM2 was upregulated, in a dose-dependent manner, although the PKM2 protein was not changed (Fig 7e). We next tested the combined effect of celastrol and TEPP-46 or 2-DG, and found that both TEPP-46 and 2-DG enhanced the inhibitory effect of celastrol on proliferation of NSCLC cells (Fig. 7f, left panel), and a Bliss synergism analysis verified the synergistic effects of the two combination strategies (Fig. 7f, lower right panel). In addition, the crystal violet staining experiment revealed that dual treatment of A549 and H1299 cells with TEPP-46 (30 μM)/celastrol or 2-DG (25 mM)/celastrol (1 μM) substantially suppressed the clonogenic growth and that the combinatory regimens showed much stronger effects than each single agent treatment (Fig. 7g).
We next examined the in vivo antitumor efficiency of this combination treatment in a xenograft model. We observed that monotherapy with either celastrol, TEPP-46 or 2-DG had a modest effect on tumor growth (Fig. 8a-c) and did not affect body weight of the mice (Fig. 8d). Monotherapy using these agents inhibited tumor cell proliferation, as indicated by Ki67, compared to the controls (Fig. 8e). Importantly, celastrol/TEPP-46 and celastrol/2-DG combinatory regimens exhibited substantially enhanced inhibition on tumor growth (Fig.8a-c) and suppression of cell proliferation reflected by the expression level of Ki67 (Fig. 8e), compared to each monotherapy. We found that in tumor tissues isolated from mice treated with celastrol alone or celastrol-based combination regimens, CIP2A and p-S287 PKM2 were substantially inhibited (Fig. 8e). The results indicated that PKM2 activation and glycolysis inhibition were able to synergize with CIP2A-targeted therapy in suppressing NSCLC cell proliferation in vivo.