INPP4B Is a PtdIns(3,4,5)P₃ Phosphatase That Can Act as a Tumor Suppressor

Thyroid cancer is the most common endocrine malignancy, and its frequency is increasing dramatically in both men and women (1, 2). Indeed, thyroid cancer is predicted to be the fourth leading cancer diagnosis by 2030 (3). The histopathology of thyroid cancers is diverse. About 80% of all malignant thyroid neoplasms are papillary thyroid carcinoma (PTC), which are usually not aggressive. In contrast, about 10% to 15% of thyroid neoplasms are follicular thyroid carcinoma (FTC), which can invade blood vessels and metastasize to lung or bone. Follicular variant of papillary thyroid carcinoma (FV-PTC) has a unique histopathology characterized by the growth pattern of an FTC but the nuclear features of a PTC (4). The mechanisms driving the formation of these thyroid cancer variants are largely unknown.

Phosphoinositide (PI) signaling is a lipid second messenger cascade involving the sequential phosphorylation of phosphatidylinositol (PtdIns) to generate first PtdIns monophosphate (PtdInsP) and then phosphatidylinositol bisphosphate (PtdInsP₂). Phosphorylation of PtdInsP₂ by PI3Ks then generates phosphatidylinositol 3,4,5-trisphosphate [PtdIns(3,4,5)P₃]. In mammalian cells, a total of eight PIs (three regioisomers each for PtdInsP and PtdInsP₂) can be interconverted, and intracellular levels of each class of these lipids are regulated by the activities of 19 kinases and 29 phosphatases (5). PIs function as direct regulators of a broad range of intracellular proteins with diverse functions (6). To date, the list of molecules that bind to PIs includes protein kinases, phospholipases, ion channel proteins, scaffold proteins, cytoskeletal proteins, and regulators of membrane trafficking. Hence, PIs control cellular responses such as proliferation, inhibition of apoptosis, motility, secretion, and endocytosis, all of which are altered in cancer cells. Given the myriad functions of PI target proteins, it is not surprising that genetic mutations or alterations to the protein expression of PI kinases and phosphatases cause a variety of disorders. For instance, the inherited disease Cowden syndrome/multiple hamartoma syndrome is caused by either germline heterozygous inactivating mutations in the PtdIns(3,4,5)P₃ phosphatase PTEN (OMIM #158350, CWS1), or germline activating mutations or amplifications of the PtdIns(3,4,5)P₃-synthesizing enzyme PIK3CA (OMIM #615108, CWS5; ref. 7). Patients with Cowden syndrome show a predisposition to developing breast, thyroid, and endometrial cancers, and about 70% of these individuals have benign thyroid abnormalities, such as multinodular goiter, adenomatous nodules, and follicular adenoma (8). These abnormalities are presumably due to the deleterious effects of excessive intracellular PtdIns(3,4,5)P₃, although formal experimental proof is lacking.

It is widely inferred that elevated PtdIns(3,4,5)P₃ triggers activation of its targets in precancerous cells, including the proto-oncogene product AKT that promotes cell survival, proliferation, oncogenesis, motility, and metastasis. Recently, another phosphoinositide phosphatase called inositol polyphosphate 4-phosphatase B (INPP4B; ref. 9), which dephosphorylates PtdIns(3,4)P₂, has emerged as a tumor suppressor. The INPP4B gene locus (4q31.1–3) shows loss of heterozygosity in breast and ovarian carcinomas (10, 11), and expression of INPP4B protein is decreased in prostate cancer (12). Currently, however, the mechanisms underlying the putative tumor-suppressive role of INPP4B are a mystery. Intriguingly, alterations in INPP4B expression are more frequent in cancer patients who bear PTEN mutations than in those retaining wild-type (WT) PTEN function (10, 11). This dual inactivation of PTEN and INPP4B in cancers implies functional collaboration between these tumor suppressors, although how such concomitant aberrations of two lipid phosphatases would promote tumorigenesis is unclear.

Studies of PTEN-deficient mice by our laboratory and other groups have unequivocally shown that PTEN is a highly effective tumor suppressor in a wide variety of tissues (5). Pten^+/− mice have a shortened lifespan, with 30% dying within 1 year of birth. Aged Pten^+/− survivors develop mammary or endometrial tumors or T-cell lymphomas, among other malignancies. Interestingly, thyroid cancers, a type of tumor associated with PTEN mutations in humans, have not been detected in Pten^+/− mice (13). Our mutants thus provide a unique system with which to analyze cooperation among gene mutations associated with thyroid carcinogenesis. In this study, we generated gene-targeted mice lacking the phosphatase domain of Inpp4B (Inpp4B^Δ/Δ). Although these mutants are viable and healthy, Inpp4B^Δ/Δ;Pten^+/− compound mutant mice develop thyroid carcinomas with complete penetrance. Moreover, our initial studies in human tissues suggest that reductions in INPP4B and PTEN co-occur in human thyroid and endometrial cancers. INPP4B may therefore be an important regulator of cancer progression, especially in the context of PTEN insufficiency.

To gain insight into the mechanism by which INPP4B exerts its tumor-suppressive function, we scrutinized the ability of INPP4B to use each PI as a substrate. We overexpressed INPP4B in 293T cells and recovered an immunoprecipitate that was used as an enzyme source in vitro. To our surprise, in addition to PtdIns(3,4)P₂, INPP4B dephosphorylated PtdIns(3,4,5)P₃ (Fig. 1A). This activity toward PtdIns(3,4,5)P₃ was intrinsic to INPP4B and not due to any proteins that might have bound to it in the immunoprecipitate because immunoprecipitates of the INPP4B C842S–mutant protein, which lacks hydrolase activity, did not dephosphorylate PtdIns(3,4,5)P₃ (Supplementary Fig. S1A). PtdIns(3,4,5)P₃ dephosphoryation catalyzed by PTEN reached a plateau at a substrate concentration of 1 mmol/L (Fig. 1B), whereas PtdIns(3,4,5)P₃ dephosphoryation by INPP4B did not. These results suggest that INPP4B requires a higher PtdIns(3,4,5)P₃ concentration than does PTEN to exert its phosphatase activity. Indeed, when our plots of reaction velocity versus substrate concentration were fitted to the sigmoidal Hill equation, nonlinear regression analysis revealed that the PtdIns(3,4,5)P₃ concentration needed to obtain a velocity half of maximal (K_half) for PTEN was 0.27 ± 0.04 mmol/L, whereas the K_half for INPP4B was 0.70 ± 0.09 mmol/L. We confirmed this activity using an in-house, purified preparation of PtdIns(3,4,5)P₃ (Supplementary Fig. S1B), excluding the possibility that our earlier results were due to INPP4B-mediated dephosphorylation of a contaminant [e.g., PtdIns(3,4)P₂] in our commercial source of PtdIns(3,4,5)P₃.

The above results raised the possibility that INPP4B might help to control PtdIns(3,4,5)P₃ levels, particularly when PTEN is deficient and intracellular levels of this lipid second messenger rise as a result. To investigate this hypothesis, we generated Inpp4b^Δ/Δ mice lacking Inpp4B exon 21, which encodes the active site motif in the phosphatase domain (Fig. 1C). Successful disruption of the Inpp4B gene was confirmed by Southern blotting and immunoblotting (Fig. 1D). Inpp4B^Δ/Δ mice were born at the expected Mendelian ratio, were healthy and fertile, had a normal lifespan, and showed no gross tissue abnormalities. These findings stand in sharp contrast to the early embryonic lethality of Pten^−/− mice and the tumorigenesis occurring in various tissues of aged Pten^+/− survivors (13). Thus, INPP4B is dispensable for PtdIns(3,4,5)P₃ metabolism, at least when PTEN is functional.

We then crossed Inpp4b^Δ/Δ mice with Pten^+/− mice (14) to generate Inpp4B^Δ/Δ;Pten^+/− compound mutants. Although Inpp4B^Δ/Δ;Pten^+/− mice were born at the expected ratio and appeared healthy at birth, they had significantly shorter lifespans than Pten^+/− mice, with half-lifetime values of only 26.2 weeks (Fig. 1E). Furthermore, all double mutants displayed hoarseness and signs of respiratory distress before they died. Gross examination revealed that their airways were compressed by abnormally large thyroid glands. These data suggest that the functions of INPP4B and PTEN overlap to some degree, especially in the thyroid gland.

As noted above, genetic mutations and decreased protein expression of PTEN have been implicated in human thyroid neoplasia. However, inactivation of PTEN alone is thought to be insufficient to induce thyroid cancers because these malignancies tend to affect older adults, and the increase in lifetime risk for developing thyroid cancer in Cowden syndrome patients is less than 10% (8). In mice, heterozygous loss of Pten results in thyroid follicular hyperplasia, but no tumors progress to aggressive malignancy (13, 15). Thus, genetic alteration(s) in addition to Pten deficiency are required for benign thyroid disorders to evolve into malignant cancers.

Our Inpp4B^Δ/Δ;Pten^+/− mice had larger thyroid glands than WT, Inpp4B^Δ/Δ, and Pten^+/− mice, prompting us to histologically examine thyroid cells of these animals at 15 weeks of age. After hematoxylin and eosin (H&E) staining, Inpp4B^Δ/Δ thyroid follicles were indistinguishable from WT (Fig. 2Aa,b). Consistent with previous studies (13, 15), Pten^+/− mice had a tendency to form larger follicles than WT mice, but the overall architecture and nuclear morphology of these structures were normal (Fig. 2Ac). In sharp contrast to Pten^+/− follicles, Inpp4B^Δ/Δ;Pten^+/− follicles were filled with irregularly arranged epithelial cells (Fig. 2Ad,e), and only a few glandular lumens displayed recognizable colloid (Fig. 2Af). Cells in Inpp4B^Δ/Δ;Pten^+/− follicles also displayed numerous nuclear abnormalities associated with PTC (Fig. 2Ba–e). Importantly, Elastica-Masson staining and thyroglobulin immunostaining revealed that Inpp4B^Δ/Δ;Pten^+/− thyroids exhibited thyroglobulin-positive satellite nodules surrounded by elastic fibers (Fig. 2Ca–c), demonstrating vascular invasion by carcinomas of follicular cell origin. Furthermore, double-mutant mice developed pulmonary metastases of thyroid follicular cells (Fig. 2Da–c), whereas no local lymph node metastases were detected. These findings suggested that the thyroid tumors in Inpp4B^Δ/Δ;Pten^+/− mice resembled FTC. Because of their PTC-like nuclear morphology but FTC-like histopathology, Inpp4B^Δ/Δ;Pten^+/− thyroid tumors were deemed to fall under the definition of FV-PTC. Similar thyroid histopathologies and pulmonary metastasis were observed in Inpp4B^+/Δ;Pten^+/− mice (Supplementary Fig. S2), indicating that Inpp4B is haploinsufficient for tumor suppression. Finally, except for thyroid cancer, there appeared to be no difference in the spectrum of tumors arising in Pten^+/− versus Inpp4b^Δ/ΔPten^+/− mice.

Decreased expression of INPP4B protein has been documented in human breast, ovarian, and prostate malignancies (10–12), but its status in thyroid cancers has yet to be reported. To extend our findings in mice to humans, we first mined the recently published papillary thyroid cancer data in The Cancer Genome Atlas (TCGA) dataset using cBioPortal (16, 17). Using OncoPrint analysis, we found that INPP4B expression was downregulated in 30.6% (15 of 49 cases) of human FV-PTCs (Supplementary Fig. S3). Next, we generated an anti-INPP4B monoclonal antibody (Ab) suitable for human immunohistochemistry and immunostained sections of thyroid tissues from FTC and PTC patients. Noncancerous thyroid glands from patients with goiter served as controls. High, medium, and low levels of Ab staining intensity were established such that the immunostaining of control glands fit into the medium group (Supplementary Fig. S4). Anti-INPP4B staining of thyroid follicular epithelial cells was reduced compared with noncancerous controls in seven of eight FTC samples and seven of 39 PTC samples (Fig. 3A). These results are the first demonstration that INPP4B expression is decreased in human thyroid cancers.

Double immunostaining of our human FTC samples with anti-INPP4B and anti-PTEN Abs detected concurrent reductions in INPP4B and PTEN in three FTC cases (Fig. 3A and Fig. 3Ba–c), as has been observed for breast and ovarian carcinomas (10, 11). To extend this finding to additional thyroid cancer samples plus another tumor type, we used cBioPortal to conduct mutual exclusivity and co-occurrence analyses of the 511 cases of thyroid carcinoma and 240 cases of endometrial carcinoma listed in the TCGA database (16, 17). We identified strong tendencies toward the co-occurrence of INPP4B and PTEN mRNA downregulation in thyroid carcinoma (Supplementary Fig. S5A), and alterations to the INPP4B and PTEN genes in endometrial carcinoma (Supplementary Fig. S5B). INPP4B mutations also showed a significant positive association with PIK3CA mutations in the endometrial dataset. These profiles suggest that loss of INPP4B function may be advantageous to cancer cells when they fail to degrade, or overproduce, PtdIns(3,4,5)P₃.

A previous study has shown that genetic alterations of PTEN in thyroid cancers are associated with AKT hyperactivation (15). In line with this observation, we found that AKT phosphorylation was modestly increased in a INPP4B^hiPTEN^lo human FTC sample compared with control thyroid gland (Fig. 3Ca,b). However, INPP4B^loPTEN^lo FTC samples showed dramatically increased AKT phosphorylation (Fig. 3Cc), suggesting that INPP4B is a potent negative regulator of AKT activation.

To investigate whether the reduced INPP4B in INPP4B^loPTEN^lo FTC actually caused the observed AKT hyperactivation, we compared the activation status of AKT and its downstream effectors in thyroid follicular cells of Inpp4B^Δ/Δ;Pten^+/− mice and controls. Immunoblot analyses of Inpp4B^Δ/Δ;Pten^+/− follicular cells revealed hyperphosphorylation of PDK1 (Fig. 3D), a direct target of PtdIns(3,4,5)P₃ that is recruited to the plasma membrane and phosphorylates AKT. Accordingly, double-mutant thyroids exhibited increased AKT phosphorylation, just as observed in INPP4B^loPTEN^lo human FTC. Consistent with their activated AKT, Inpp4B^Δ/Δ;Pten^+/− follicular cells showed enhanced phosphorylation of PRAS40, mTOR, and S6 (Fig. 3D). Histologic examination of thyroid serial sections revealed positive staining for phosphorylated (phospho)-AKT at the plasma membrane of double-mutant follicular cells (Fig. 3E) and for phospho-S6 in the cytosol (Fig. 3F). Low levels of phospho-AKT and phospho-S6 were present in Pten^+/− follicular cells, but staining intensities were weaker than in Inpp4B^Δ/Δ;Pten^+/− cells. It should be noted that no activation of AKT signaling was observed in Inpp4B^Δ/Δ thyroids, suggesting that INPP4B is dispensable for downregulating AKT activation in the presence of normal PTEN. These data demonstrate that INPP4B has a key role in preventing the hyperactivation of AKT and its downstream effectors that occurs in situations of PTEN deficiency.

Two AKT isozymes occur in thyroid follicular cells: AKT1 and AKT2. The enhanced activation of AKT signaling we observed in Inpp4B^Δ/Δ;Pten^+/− follicular cells prompted us to examine whether separate deletion of each AKT isozyme could ameliorate the thyroid tumorigenesis and metastasis in Inpp4B^Δ/Δ;Pten^+/− mice. We generated triple-mutant mice lacking either AKT1 or AKT2 in the Inpp4B^Δ/Δ;Pten^+/− background and found that aggressive thyroid cancers developed in Akt1^−/−;Inpp4B^Δ/Δ;Pten^+/− mice (Fig. 4Aa,b), just as observed in Inpp4B^Δ/Δ;Pten^+/− animals (Fig. 2D). In contrast, Akt2^−/−;Inpp4B^Δ/Δ;Pten^+/− mice had a much milder phenotype, with no cytologic evidence of malignancy or pulmonary metastasis (Fig. 4Ac,d). Consistent with these results, levels of AKT (Fig. 4B) and S6 (Fig. 4C) phosphorylation were significantly lower in mutant cells lacking AKT2 compared with those lacking AKT1. Over a 48-week observation period, 38% of Akt2^−/−;Inpp4B^Δ/Δ;Pten^+/− mice remained viable, whereas only 10% of Akt1^−/−;Inpp4B^Δ/Δ;Pten^+/− mice survived this long (Fig. 4D). These results suggest that AKT2 hyperactivation in thyroid follicular cells makes a greater contribution to the mortality of Inpp4B^Δ/Δ;Pten^+/− mice than AKT1 hyperactivation.

Although it has long been assumed that PTEN deficiency causes PtdIns(3,4,5)P₃ accumulation in tumor cells in vivo, quantitative data validating this assumption is lacking for both mouse and human cancers. To elucidate the molecular mechanism underlying the functional collaboration between Inpp4B and Pten in tumor suppression, we quantitated PtdIns(3,4,5)P₃ levels in thyroid tissues by modifying a nonradioactive method of PtdIns(3,4,5)P₃ measurement developed by Clark and colleagues (18). As shown in Fig. 4E, heterozygous deletion of Pten increased cellular PtdIns(3,4,5)P₃ levels in whole murine thyroid gland extracts. Consistent with the weak phosphatase activity of INPP4B toward PtdIns(3,4,5)P₃in vitro (Fig. 1A), disruption of Inpp4b alone caused only a small (but still statistically significant) increase in PtdIns(3,4,5)P₃. However, PtdIns(3,4,5)P₃ accumulation in the thyroid was enormous when Inpp4B was deleted in the Pten heterozygous background. These results clearly demonstrate that INPP4B is critical for dephosphorylating the PtdIns(3,4,5)P₃ that accumulates under conditions of PTEN deficiency. In contrast to PtdIns(3,4,5)P₃, there was no difference in PtdIns(3,4)P₂ levels between Pten^+/− and Inpp4B^Δ/Δ;Pten^+/− thyroid tissues (Supplementary Fig. S6). Taken together, our data suggest a model (Fig. 4F) in which INPP4B's dephosphorylation of PtdIns(3,4,5)P₃ comprises a “back-up” mechanism that prevents tumorigenic accumulation of this lipid messenger when PTEN activity is insufficient.

Gene mutation, heterozygous deletion, and decreased protein expression of INPP4B have been implicated in several types of human cancers (10–12, 19). However, until our study, it had not been definitively shown that INPP4B deregulation is not just a consequence of tumorigenesis but an active contributor. Here, we have used Inpp4B single, double, and triple mutant mice to establish a cause-and-effect relationship between loss of INPP4B function and tumorigenesis.

Our work is the first to demonstrate that a loss of Inpp4B predisposes mice to cancer development, providing reverse-genetics evidence for a tumor-suppressive function of INPP4B. Although mice with disruption of Inpp4B alone did not develop any tumors, all Inpp4B^Δ/Δ;Pten^+/− mice exhibited spontaneous and early-onset formation of metastatic thyroid cancers. These double mutants displayed a dramatically reduced lifespan compared with both Pten^+/− mice (this study) and thyrocyte-specific Pten-null mice (20). Thus, INPP4B is dispensable for thyroid tumor suppression in WT mice but is essential for preventing tumorigenesis in this tissue when PTEN activity is insufficient. Studies of the genomes of human thyroid cancer cells have revealed that these malignancies accumulate various genetic alterations that may promote thyroid cell dedifferentiation and drive thyroid cancer initiation and progression (1, 2). Our results are in line with this observation and indicate that loss of Inpp4B is one step of the many driving thyroid cancer progression.

Because Pten and Inpp4b are deleted in a systemic manner in our mouse model, a concurrent loss of these enzymes in non-thyroid cells may also contribute to the tumor-prone phenotype of the thyroid gland. Coincident decreases in expression of INPP4B and PTEN have been observed in human breast and ovarian carcinomas (10, 11). Our mutational profiling of human thyroid and endometrial cancers using the TCGA dataset also showed that genetic alterations in INPP4B and PTEN have a strong tendency to co-occur, as do mutations of INPP4B and PIK3CA. In addition, although not statistically significant, we detected coincident decreases in INPP4B and PTEN expression in 38% of human FTC specimens. Our biochemical data showing that INPP4B hydrolyzes PtdIns(3,4,5)P₃ with a K_half value lower than that of PTEN may explain such concurrent genetic alterations in PI-metabolizing enzymes. Furthermore, this finding suggests that loss of INPP4B function may be particularly advantageous to cancer cells when they fail to degrade PtdIns(3,4,5)P₃, or overproduce PtdIns(3,4,5)P₃. Such situations can arise due to either sporadic or hereditary mutations of PTEN or PIK3CA, which presumably cause deleterious increases in intracellular PtdIns(3,4,5)P₃. Future study should investigate the position of phosphate that INPP4B hydrolyzes.

Hyperplasia of the thyroid gland, mammary gland, and uterus are the most common manifestations of Cowden syndrome, and these patients have increased risks of developing cancers of these origins (8). For example, 66% of Cowden syndrome patients have benign thyroid abnormalities such as follicular adenomas, whereas 3% to 10% of Cowden syndrome patients develop thyroid malignancies. It is tempting to speculate that the presence or absence of INPP4B could determine whether the benign lesions in Cowden syndrome patients undergo malignant transformation. We have demonstrated in mice that concurrent PTEN and INPP4B deficiencies synergistically elevate PtdIns(3,4,5)P₃ in the thyroid gland in vivo (Fig. 4E). We therefore propose that there must be a threshold PtdIns(3,4,5)P₃ concentration above which benign hyperplastic thyroid cells acquire cancerous properties (Fig. 4F).

A provocative question arising from our work is whether INPP4A, a PI phosphatase highly related to INPP4B, also functions as a tumor suppressor. INPP4A and INPP4B share 37% amino acid sequence identity (5). Remarkably, the CX₅R active site motif in the phosphatase domains of these isozymes is identical (CKSAKDR). As expected, we found that INPP4A is also capable of dephosphorylating PtdIns(3,4,5)P₃ (Supplementary Fig. S1). We have previously reported that loss of the Inpp4a gene in mice leads to neurodegeneration in the central nervous system as a result of enhanced glutamate excitotoxicity (21). These mutants die within 1 month of birth, precluding an analysis of tumorigenesis. Nevertheless, we speculate that INPP4A may be a potential candidate for a novel tumor-suppressive phosphatase. Examination of genetic alterations in endometrial and lung cancers listed in the TCGA database has revealed several mutations in INPP4A, in particular a missense mutation that replaces the arginine residue in the CX₅R active site motif. Future experiments with single- and double-mutant mice bearing conditional Inpp4a deletions should answer this intriguing question.

In conclusion, our study has established a role for INPP4B in tumor suppression and demonstrated its critical function in PtdIns(3,4,5)P₃ metabolism. Our Inpp4B^Δ/Δ (and parental Inpp4B^flox/flox) mice provide useful models for studying disorders emanating from aberrant accumulation of PtdIns(3,4,5)P₃. Notably, downregulation of INPP4B has been observed in several aggressive human cancers, including in >80% of basal-like breast cancers (11), and so may be an important molecular signature of malignancy. Future studies in which the biologic properties of PtdIns(3,4,5)P₃-driven cancers are determined by reconstituting the underlying genetic alterations in mice may point toward novel clinical interventions. Because the INPP4B gene is not lost in human cancers but is only minimally transcribed, INPP4B is a plausible anticancer drug target. Our mutants may be useful for testing both PI3K inhibitors and chemical compounds capable of enhancing residual INPP4B activity in cancer cells. This approach may open up new therapeutic strategies for a wide range of hyperplasias and cancers that harbor mutations in PTEN and/or PIK3CA and accumulate deleterious levels of intracellular PtdIns(3,4,5)P₃.

The conditional targeting vector was constructed to delete a genomic fragment containing the 21st coding exon of the mouse Inpp4b gene by homologous recombination (Fig. 1C). One loxP site was introduced into intron 20 and two into intron 21. The LacZ-PGK-Neo^r cassette was inserted in antisense orientation to Inpp4b transcription between the two lox sites in intron 21. The linearized construct was electroporated into 1 × 10⁷ E14K mouse embryonic stem (ES) cells (22). ES cell colonies resistant to G418 (0.3 mg/mL; Life Technologies) were screened for homologous recombination by standard PCR. Recombinant clones were confirmed by standard Southern blotting of HindIII-digested genomic DNA fragments using a 590 bp probe. Targeted ES cells were injected into C57BL/6J blastocysts (CLEA Japan). Chimeric male mice were crossed with C57BL/6JJ females to achieve germline transmission. Deletion of the selection cassette to generate Inpp4b^+/flox mice was achieved by crossing to MeuCre40 transgenic mice (23). Likewise, Inpp4b^+/Δ mice were obtained by crossing Inpp4b^+/flox mice with MeuCre40 mice. Inpp4b^Δ/Δ mice were distinguished from Inpp4b^+/Δ and Inpp4b^+/+ mice by PCR. An oligo primer (5′-CCTGCCATGGGTAGATTTCT-3′) common to the Inpp4A⁺ and Inpp4A^Δ alleles, a primer specific for the Inpp4A⁺ allele (5′-GTTTACATTTGACAGGGTGGTTGG-3′), and a primer specific for the Inpp4A^Δ allele (5′-TGCTGTCGCCGAAGAAGTTA-3′), were combined in the same PCR reaction. Inpp4b^+/ΔPten^+/− mice were obtained by crossing Inpp4b^+/Δ mice with Pten^+/− mice (14). To obtain Inpp4b^Δ/ΔPten^+/− mice, Inpp4b^+/ΔPten^+/− mice were crossed with Inpp4b^+/Δ mice. Triple mutants were obtained by crossing Inpp4b^+/ΔPten^+/− mice with Akt1^−/− mice or Akt2^−/− mice, both of which were purchased from The Jackson Laboratory. All experimental protocols were reviewed and approved by the Akita University Institutional Committee for Animal Studies.

FLAG-tagged INPP4A, INPP4B, or PTEN was expressed in 293T cells and purified using anti-FLAG antibody (Sigma-Aldrich) as previously described (24). Recombinant FLAG-INPP4A or FLAG-INPP4B (0.4 μg) was incubated for 30 minutes at 37°C with 100 to 1,000 μmol/L 1,2-dioctanoyl-PtdIns(3,4,5)P₃ (Cayman Chemical) in 50 mmol/L Tris–HCl (pH7.4) plus 10 mmol/L EDTA and 200 mmol/L NaCl. Released phosphate was determined using the malachite green assay (25). V_max and K_half values were determined by nonlinear curve fitting analysis using the GraphPad Prism6.0 software tool (GraphPad Software, Inc.). The initial velocity (pmol/min/μg) versus substrate concentration data were fit to the sigmoidal Hill equation: v = (V_max [S]ⁿ)/(Kⁿ + [S]ⁿ), where v = reaction rate; V_max = maximum reaction rate; [S] = substrate concentration; K = the Hill constant, that is, the substrate concentration producing half-maximal velocity; and n = Hill coefficient (26).

Radiolabeled [D3-³²P]PI(3,4,5)P₃ was prepared enzymatically by incubating dipalmitoyl-phosphatidylinositol 4,5-bisphosphate [PI(4,5)P₂] (Cayman) with [γ-³²P]ATP (NEN) in the presence of an immunoprecipitate containing class IA PI3-kinases that was prepared by incubating mouse liver lysates with anti-p85 Ab (Cell Signaling Technology). The kinase reaction was quenched by the addition of chloroform/methanol/8% HClO₄ (5:10:4). After vigorous vortexing, chloroform/HClO₄ (1:1) was added to isolate the organic phase, which was washed three times in chloroform-saturated 1% HClO₄ before drying. Dried lipids were resolved in chloroform/methanol (95:5) and spotted onto a silica gel 60 plate (MERCK) that had been pretreated with 1.2% (w/v) potassium oxalate in methanol/water (2/3). PI(3,4,5)P₃ was fractionated by thin layer chromatography (TLC) in chloroform/methanol/acetone/acetic acid/water (7/5/2/2/2). The corresponding spot was scraped off of the silica gel plate and extracted using the Bligh–Dyer method. The resultant radioactive PI(3,4,5)P₃ (4,000 cpm) was mixed with nonradioactive dipalmitoyl-PI(3,4,5)P3 (3 nmol, Cayman) in 15-μL assay buffer [50 mmol/L Tris–HCl (pH 7.5), 0.2% Triton X-100, and 2 mmol/L dithiothreitol]. Purified recombinant FLAG-tagged phosphatase (5–10 μg/15 μL) was added to the substrate and incubated at 30°C for 20 hours. After reaction completion, lipids were extracted by the Bligh–Dyer method and isolated by TLC as described above. Radiolabeled PI(3,4,5)P₃ was quantitated by autoradiography using Typhoon FLA 9500 (GE Healthcare).

Anti-INPP4B antiserum used for immunoblotting was kindly provided by Dr. Jean Vacher (Clinical Research Institute of Montreal, Canada). INPP4B-specific rat monoclonal Ab (mAb) against a purified recombinant human INPP4B fragment (2–235 amino acids) was raised for immunohistochemistry. The mAbs against PTEN, pan-AKT, phospho-AKT (S473), phospho-AKT (T308), phospho-PDK1 (S241), phospho-PRAS40 (T246), phospho-mTOR (S2448), phospho-S6 (S235/236), and phospho-S6 (S240/244) were all from Cell Signaling Technology. Anti-thyroglobulin mAb was from DAKO. Anti–α-tubulin polyclonal Ab was from Medical & Biological Laboratories. Antibodies conjugated to Alexa Fluor 488 or 568 were from Invitrogen, and other secondary antibodies were from DAKO or Vector Laboratories.

Thyroid tissue was homogenized in lysis buffer [50 mmol/L Tris–HCl (pH 8.0), 100 mmol/L NaCl, 1 mmol/L EDTA, 1% Triton X-100, 1 mmol/L DTT, 1 mmol/L Na₃VO₄, 30 mmol/L sodium pyrophosphate, 50 mmol/L sodium fluoride, and protease inhibitor cocktail (Roche Applied Science)]. After centrifugation at 20,000 × g for 15 minutes, supernatants (10 μg total protein) were subjected to standard SDS-PAGE and immunoblotting as described previously (21).

Mouse thyroid glands were fixed in 10% formalin neutral buffer solution and embedded in paraffin. Sections (3–5 μm) were cut and stained with H&E or Elastica-Masson in accordance with standard procedures. For immunohistochemistry, deparaffinized sections were incubated in 10 mmol/L citrate buffer (pH 6.0) at 121°C for 10 minutes, treated with 3% H₂O₂/methanol for 30 minutes, blocked with 10% BSA, and incubated overnight at 4°C with primary Ab recognizing thyroglobulin, phospho-AKT S473, phospho-S6 S235/236, or PTEN (all at 1:100 dilution). Immunostained sections were incubated with peroxidase-conjugated secondary Abs that were detected using DAB solution (WAKO).

For immunofluorescent examination of INPP4B, PTEN, and phospho-AKT S473 in human thyroid tissues, paraffin-embedded human thyroid tissue arrays (thyroid cancer–normal) were purchased from SuperBioChips Laboratories. Deparaffinized sections were boiled in 10 mmol/L citrate buffer (pH 6.3) and incubated with primary mAb recognizing PTEN or phospho-AKT S473, or with anti-INPP4B. Binding was visualized using secondary Abs conjugated to Alexa Fluor 488 or 568. For counterstaining, DAPI was added at 1 μg/mL for 5 minutes. Fluorescent images were acquired on a LSM 780 microscope (Carl Zeiss) and processed with ZEN software (Carl Zeiss).

To score expression levels of INPP4B and PTEN in human thyroid tissue samples, the intensity of Ab staining (0, +1, +2) was multiplied by the percentage of immunoreactive cells to generate three groups (Low: <50; Medium: 50–150; High: >150). For each section, staining intensities were determined for 100 cells in at least three distinct areas of the slide. This method allowed classification of nine control (goiter/adenoma) samples in the “Medium” groups of INPP4B and PTEN expression. Samples were then categorized into nine (3 × 3) subgroups representing all combinations of low, medium, and high INPP4B and PTEN expression.

For PtdIns(3,4,5)P₃ measurement, acidic phospholipids were extracted from thyroid tissues by the Bligh–Dyer method as described previously (21). PtdIns(3,4,5)P₃ was measured by modifying a method developed by Clark and colleagues (ref. 18; Nakanishi and colleagues, manuscript in preparation). An UltiMate 3000 LC system (Thermo Fisher Scientific) equipped with HTC PAL autosampler (CTC Analytics) was used for column chromatography. TSQ-Vantage (Thermo Fisher Scientific) was used for electrospray ionization MS/MS analysis. For PtdIns(3,4)P₂ measurement, mice were intravenously injected with 2 mCi [³²P]-phosphorus (PerkinElmer) and thyroid glands were isolated 4 hours later. Total lipids were extracted, deacylated, and subjected to high-performance liquid chromatography (HPLC) analysis using a Partisphere SAX column (Whatman) as previously described (27).

cBioPortal, the open platform for exploring multidimensional cancer genomics data, was used to obtain summary statistics on mutual exclusivity and co-occurrence of genomic alterations in human INPP4B and PTEN (16, 17). The portal calculated odds ratios that indicated the likelihood that the mutations in each pair of genes were mutually exclusive or concurrent. The 511 cases of PTC and the 240 cases of uterine corpus endometrial carcinoma listed in the TCGA database were analyzed for mutual exclusivity of INPP4B and PTEN alterations. P values were determined by the Fisher exact test, with the null hypothesis that the frequency of occurrence of a pair of alterations in two genes was proportional to their uncorrelated occurrence in each gene (17).

Among the 511 cases of PTC, 96 samples were coded as “8340/3.” According to the International Classification of Disease for Oncology (third edition ICD-O-3), 8340/3 specifies “Papillary carcinoma, follicular variant” (FV-PTC). Sample IDs for all FV-PTC were obtained and the “OncoPrint” results were analyzed on the basis of the International Classification of Disease for Oncology histology code. The percentages of FV-PTC and non–FV-PTC samples with decreased INPP4B protein were calculated.

Statcel2 (OMS, Japan) was used to perform the Kaplan–Meier cumulative mouse survival analyses. For PtdIns(3,4,5)P₃ measurements, data were analyzed using the unpaired Student t test. P < 0.05 were considered significant.

No potential conflicts of interest were disclosed.

Conception and design: J. Sasaki, T. Sasaki

Development of methodology: H. Nakanishi, T. Sasaki

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Kofuji, H. Kimura, H. Nakanishi, H. Nanjo, S. Takasuga, S. Eguchi, R. Nakamura, N. Ueno, K. Asanuma, M. Huang, A. Koizumi, M. Yamazaki, J. Sasaki

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Kofuji, H. Kimura, H. Nakanishi, H. Nanjo, S. Takasuga, H. Liu, S. Eguchi, R. Nakamura, N. Ueno, K. Asanuma, M. Huang, A. Koizumi, M. Yamazaki, T. Sasaki

Writing, review, and/or revision of the manuscript: S. Kofuji, T. Sasaki

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Itoh, M. Huang, A. Suzuki, J. Sasaki, T. Sasaki

Study supervision: T. Habuchi, T. Sasaki

The authors thank K. Kofuji, At. Kato, Ak. Kato, S. Kumagai, Y. Kadowaki, Y. Sugihara, H. Takahashi, Y. Tsuya, T. Ayukawa, K. Asanuma, C. Horie, Y. Horie, Y. Moribayashi, S. Nishino, Y. Kuribayashi, S. Ooi, and K. Yanase for critical discussions and technical support, and H. Watanabe for tissue processing and histological analyses.

This work was supported in part by research grants from Japan Science and Technology Agency (JST); the Ministry of Education, Culture, Sports, and Technology of Japan; the Japan Society for the Promotion of Science (JSPS); Astellas Foundation for Research on Metabolic Disorders; Ono Medical Research Foundation; Uehara Memorial Foundation; and Takeda Science Foundation. T. Sasaki and J. Sasaki received funding from Core Research for Evolutional Science and Technology (CREST; JST) and Precursory Research for Embryonic Science and Technology (PRESTO; JST).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.