The primary glioblastoma cell line U-87 MG, the hormone-refractory prostate cancer cell line DU 145, and the large cell lung cancer cell line NCI-H1915 (hereinafter “H1915”) were purchased from the American Type Culture Collection (ATCC, Manassas, VA). DU 145 and H1915 cell lines were originally isolated from a brain metastasis. All cell lines were maintained according to the supplier’s instructions and routinely tested for mycoplasma contamination using MycoAlerTM (Lonza, Walkersville, MD, USA).

Perifosine was obtained from Aeterna Zentaris GmbH (Frankfurt, Germany). For the in vitro assays, perifosine was dissolved in dimethyl sulfoxide (DMSO), and the final concentration of DMSO was adjusted to 0.1%. For the in vivo studies, perifosine was dissolved in physiological saline (FUSO Pharmaceutical Industries, Ltd., Osaka, Japan). The antibodies used in this study are listed in Supplementary Table 1 . EnVision System HRP-Labelled Polymer Secondary antibody was purchased from Agilent Technologies (Santa Clara, CA, USA). All other reagents were obtained from Sigma-Aldrich (St. Louis, MO, USA) unless otherwise specified.

Dose response curves were generated for perifosine to determine the inhibitory concentration required to achieve 50% cell death (IC₅₀ values). Briefly, cells were grown overnight in 96-well plates and then left untreated or treated with perifosine at different concentrations. After 48 h, the extent of cell viability was assessed according to a WST-8 dye-based assay (Kishida Chemical, Osaka, Japan) as described previously (21). For each treatment condition, the mean values from triplicate wells were calculated.

Five-week-old male BALB/c nude mice were purchased from Japan SLC, Inc. (Shizuoka, Japan).

We designed a loading (180 mg/kg) and maintenance (45 mg/kg) dosing regimen for a series of experiments, with the aim of efficiently delivering perifosine into tumor sites. The chemotherapy experiments were performed using two different protocols as described below.

Brain or subcutaneous tumors were fixed in 10% formalin for approximately 72 h, embedded in paraffin, and sectioned at 4 µm thickness. Sections were stained with hematoxylin and eosin (H&E) for morphological observations, and with van Gieson for detection of fibrosis using a standard procedure. For immunostaining, the sections were preincubated with 3% H₂O₂ to block endogenous peroxidases. Sections were then incubated with antibodies against cytokeratin (clone AE1/AE3) and Ki-67 (clone SP6) overnight at 4°C and cleaved caspase-3 for 60 min at room temperature. After rinsing in Tris buffer, the secondary antibody was applied for 30 min at room temperature, and this was followed by the addition of 3,3’-diaminobenzidine (DAB) as a substrate. The sections were then counterstained with hematoxylin. Ki-67, cleaved caspase-3, and van Gieson-stained sections were graded by a pathologist (TS) as follows: -, absent; ±, minimal; 1+, mild; and 2+, moderate ( Supplementary Figure 1 ).

Tumor lysates were prepared from mice in each group on day 22 in Protocol Number 2 (subcutaneous ectopic tumor models). Western blot analysis of the phosphorylation/activation patterns of the relevant molecules was performed as previously described (21). All the western blots were normalized to β-actin, and protein intensity was quantified using the Image J software (NIH, Bethesda, MD).

Non-tumor-bearing mice were orally administered a single dose of perifosine at 45 or 180 mg/kg. At 4, 8, 24, 96, 168, and 360 h after the dose, blood was collected from the abdominal vena cava under isoflurane anesthesia, and the brain was resected after exsanguination. In a separate set, the DU 145 and H1915 subcutaneous ectopic tumor models were administered perifosine as either a single dose (45 or 180 mg/kg) or repeated doses (180 mg/kg loading dose followed by four consecutive daily doses of 45 mg/kg [180/45]). At 24, 48, 96, 144, and 168 h after the first dose, blood and tumor tissues were obtained in a manner similar to that described above.

The plasma was mixed with an equal volume of 1% formic acid and stored at -80°C. It (10 μL) was then mixed with an equal volume of acetonitrile containing 1% formic acid (FA/ACN) and 180 μL of internal standard (IS) solution (200 ng/mL of ethyl p-hydroxybenzoate [EHB] solution in FA/ACN). Next, the supernatant was collected after deproteinization (Sirocco, Waters Corp.) for LC-MS/MS assay analysis. The brain and tumor tissues were mixed with an equal volume (w/v) of 1% formic acid, homogenized, mixed with a 3-fold volume of FA/ACN, and centrifuged at 15,000 rpm (16,617 × g) for 3 min at 4°C to collect the supernatant. The supernatant (20 μL) was mixed with 80 μL of IS solution (200 ng/mL of EHB solution in “FA/ACN”/water [3:1] solution), and the new supernatant was obtained after deproteinization using the LC-MS/MS assay. The maximum plasma drug concentration (C_max), the area under the concentration-time curve up to the last sampling time-point (AUC_0-last), the mean residence time (MRT), and the terminal half-life (T_1/2) were calculated using the mean values at each point. The AUC_0-last was calculated using the trapezoidal rule. The MRT was calculated as follows: MRT   =   AUMC / AUC   ( AUMC :   area   under   moment   curve   = ∫ last ∞ tC d t ) .

Data were analyzed using SAS System Release 8.2 (SAS Preclinical Package, Version. 5.0; SAS Institute Japan Ltd., Tokyo, Japan). The values are expressed as mean ± standard deviation (SD). Differences in survival among experimental groups were analyzed according to Kaplan-Meier survival curves using the log-rank test. The Student’s t-test or Welch’s t-test (F-test; p < 0.05) was used to detect the statistical differences in brain weight, TV, tumor weight, RBW, and protein levels in tumor tissues between two groups. P values < 0.05 were considered to be statistically significant. 

To investigate whether perifosine is extensively distributed in brain tissues, we examined the plasma and brain tissue levels of perifosine in normal mice with intact BBBs. Perifosine was administered at doses of 45 or 180 mg/kg, as the maximum tolerated dose of perifosine was confirmed to be 45 mg/kg (once-daily regimen) and 250 mg/kg (once-weekly regimen). The plasma perifosine concentrations reached maximum levels at 24 or 96 h post-dosing and exhibited a gradual decline thereafter ( Figure 1A ). In contrast, brain concentrations increased gradually and remained at high levels for at least up to 360 h (Brain C_max: 6.20 μM [2.86 μg/g] for 45 mg/kg and 16.22 μM [7.49 μg/g] for 180 mg/kg, respectively) ( Figure 1B ). High and rapid distribution in brain tissues was observed in the 180 mg/kg group compared to that observed in the 45 mg/kg group ( Figure 1B ). When calculating the brain-to-plasma ratios for perifosine, a trend toward a higher retention in brain tissues was observed at 360 h post-dosing ( Figure 1C ); however, this was not observed at the earlier time-points. The in vitro perifosine IC₅₀ values for DU 145 and H1915 were 28.8 μM and 2.5 μM, respectively, at 48 h; these were approximately 1.8-fold higher and 6.5-fold lower, respectively, than the brain C_max values after a single oral dose of 180 mg/kg ( Figure 1D ).

Intracranial inoculations were performed using human primary glioblastoma U-87 MG cells and two human metastatic brain tumor cell lines. Of these cell lines, U-87 MG cells formed a unifocal, unilateral tumor in the cerebral parenchyma, which was consistent with previous reports (23, 33) ( Supplementary Figure 2 ). Injection with DU 145 cells caused scattered formation of small tumor nests in the cerebral parenchyma near the injection site. These multiple tumors were found separately ( Figures 2A, Ba, b ), unlike a solitary lesion induced by U-87 MG cells. Injection of H1915 cells resulted in multifocal tumors. Tumor cells were observed not only in the brain parenchyma but also in the subarachnoid spaces, ventricles, and vascular spaces (all four mice tested; Figures 2C, Da-g ). Our ability to observe tumor cells even in the left lateral ventricle ( Figure 2Dd ) suggests that this separation was highly unlikely to be caused due to technical reasons, but was instead a result of cell migration; this was also corroborated by the in vitro wound-healing assay ( Supplementary Figure 3 ). Thus, the H1915 model appeared to closely resemble the clinical features of leptomeningeal carcinomatosis (34–36). The formation of multiple lesions such as leptomeningeal meningitis makes surgical resection impossible, and such patients cannot typically be treated with surgery only (36).

Brain weights of H1915 tumor-bearing mice were increased significantly, with marked enlargement, compared to those of sham-operated mice on day 21 ( Figure 2E ), suggesting that clinical features, including elevated intracranial pressure caused by metastatic brain tumors, were reflected at least in the H1915 model (2, 3). Therefore, these models are likely to provide valuable evaluative tools for predicting the clinical benefits of drug candidates.

We used the intracranial DU 145 and H1915 models to evaluate the survival benefit of perifosine against metastatic brain tumors. In the DU 145 model, the MST significantly increased to 49.5 (%ILS = 30, p < 0.01) upon administration of perifosine from day 3, while the control group exhibited an MST of 38.0 ( Figure 3A ). In the H1915 model, the MST of the control group was 24.5, while eight out of 10 mice survived until the final day of observation (day 64) in the perifosine (D3) group (MST was not reached, p < 0.001). Likewise, treatment with perifosine from day 7 onwards significantly prolonged survival compared to that of the control group, where the MST was 59.0 (%ILS = 141, p < 0.01) ( Figure 3B ). Although the survival benefit was lower than that of the perifosine (D3) group, four out of 10 mice survived until day 64 in the perifosine (D7) group. Based on the observation that the control mice began to reach the endpoint as early as on day 19 and the time required for perifosine to reach values close to C_max in brain tissues was at least four days after administration ( Figure 1B ), it is noteworthy that perifosine resulted in significant survival benefits, even in the perifosine (D7) group.

Some agents targeting the PI3K/Akt signaling pathway are associated with hyperglycemia due to interaction with the insulin-glucose regulatory axis (37, 38). Therefore, hyperglycemia and polyuria are frequently reported adverse effects of PI3K/Akt inhibitors. Here, the bedding of survivors, all of which were mice in perifosine-treated groups, was visibly wet after day 54, suggesting the symptom of polyuria. In order to assess the possibility of hyperglycemia, mice surviving for 64 days were analyzed for the level of glucose in plasma. As a result, perifosine did not induce an increase in plasma glucose when compared to the historical control data. Similarly, there were no severe changes in T-CHO, T-BIL, BUN, CRE, and ALP on day 64 ( Supplementary Figure 4 and Supplementary Table 2 ).

In the control group of the H1915 model, the brain weight was clearly higher than that in the normal and sham-operated mice, which is suggestive of aggressive tumor progression and/or cerebral edema ( Figures 3C, D ). In contrast, the brain weights in both the perifosine groups were significantly lower than those in the control group, and these values were almost comparable to the values in the normal and sham-operated groups ( Figures 3C, D ).

Tumor cells were observed multifocally in the control group on day 14 ( Figures 3Ea-c ). However, no apparent viable tumor cells were observed in the perifosine (D3)-treated mice on day 41 in a separate experiment (n = 2), thus indicating complete responses ( Figures 3Ed-f ). In agreement with the results of brain weight, it is apparent that perifosine dramatically inhibited H1915 tumor growth.

It has been documented that high correlations between in vitro and in vivo sensitivity of drugs are not always observed (39–41). Although the exact reason for the discrepancies is unknown, this can be explained, at least in part, by the modification of sensitivity through multiple factors in the tumor microenvironment.

Here, we demonstrated that the growth patterns of DU 145 and H1915 were largely different in brain tissues ( Figures 2A–D ). Therefore, we speculated that the growth patterns of both cell lines in brain tissues may have influenced the in vivo sensitivity of perifosine.

To confirm this possibility, DU 145 ( Figures 4A–E ) and H1915 ( Figures 4F–K ) cells were subcutaneously inoculated and allowed to form a solitary tumor, and the antitumor effect of perifosine was examined. Perifosine, when used according to the same treatment regimen as was used in the brain orthotopic models, exerted a moderate antitumor effect with a TGI value of 34% in the DU 145 xenograft model, despite the lack of significance [p = 0.058 for TV on Day 22 ( Figure 4A ); p = 0.086 for tumor weight ( Figure 4C )]. In accordance with these results, perifosine down-regulated phosphorylation levels of Akt at both Ser473 and Thr308 by 61% (p = 0.074) and 44% (p < 0.05) respectively without diminishing total Akt protein level in DU 145 xenografts ( Figure 4E ). Meanwhile, perifosine also elicited much stronger antitumor responses against subcutaneous H1915 growth, which included regressions of the established tumor on and after day 7 (93% reduction in TV; Figures 4F, G ). These results indicated a similar tendency between both in vitro cytotoxicity and in vivo orthotopic findings, thus demonstrating that perifosine is inherently more potent against H1915 than against DU 145; this was independent of the different growth patterns in brain tissues. Furthermore, perifosine treatment significantly abrogated the Akt signaling pathway in H1915 xenografts on Day 22 ( Figure 4J ), almost consistent with other tumor models ( Supplementary Figure 5 ). Notably, both p-Akt and total Akt were reduced at this time point, which is suggestive of an increase in tumor cell death ( Figure 4J ). We also monitored the systemic toxicity of mice receiving this therapy. No adverse effects in general conditions or any treatment-related macroscopic changes in major organs including the brain, liver, and kidneys were induced by this therapeutically effective regimen. Moreover, as the RBW values were more than 0.8, this therapy was deemed tolerable ( Figures 4D, I ).

Histological examination revealed that the tumor cells were large, round-to-oval, and possessed distinct multiple nucleoli in the clear oval nucleus; all of these characteristics resemble the morphological features of large-cell lung carcinoma ( Figures 4Ka, b ). Although there was still no big difference in H1915-tumor volume between the vehicle and perifosine-treated tumors on day 7, we observed marked morphological changes in tumor cells in response to perifosine treatment, which is suggestive of a decrease in tumor cell proliferation and/or an increase in cell death not only on day 22, but also even at this early stage ( Figures 4Kc-f ).

To further investigate the mechanism of perifosine-mediated tumor growth inhibition, the levels of Ki-67 and cleaved caspase-3 were assessed in the DU 145 ( Figures 5A–J ) and H1915 ( Figures 5K–T ) tumors. H1915 tumor cells grew into a solitary solid mass, with central necrosis, in the control group on day 22 ( Figure 5K ). Perifosine exhibited a marked decrease in tumor cell proliferation, as detected by Ki-67 and increased apoptosis (cleaved caspase-3 staining), compared to that in tumor cells from the control group, and the majority of the residual tumor tissues treated with perifosine were composed largely of collagen fibers (van Gieson) and well-formed granulomas, which are thought to be formed by the effect of perifosine ( Table 3 and Figures 5K–T ). Consistently, an increase in the number of tumor cells with apoptotic morphology such as apoptotic body formation and nuclear condensation was observed in the H1915 tumor mass of the perifosine-treated animals ( Table 3 ).

Regarding the DU 145 xenografts, although an increase in the number of cleaved caspase-3 positive tumor cells was detected in response to perifosine treatment, there were no obvious differences in the morphological observations, staining patterns of Ki-67, and van Gieson staining between the control and perifosine groups ( Table 3 and Figures 5A–J ). Special and immunohistochemical stains revealed the changes underlying the antitumor responses to perifosine, thus indicating consistency with the above-mentioned findings.

It has been documented that drug accumulation differs among tumor xenografts (42). Herein, we hypothesized that the different sensitivities to perifosine depend on the differences in drug levels between DU 145 and H1915 xenografts, and characterized the pharmacokinetic properties of perifosine ( Figure 6 and Supplementary Table 3 ).

The plasma concentration of perifosine gradually increased, reached C_max at 24 or 48 h after a single dose of 45 or 180 mg/kg, respectively, and then decreased slowly ( Figures 6A, D ). There were no apparent differences in plasma C_max and AUC between the DU 145 and H1915 subcutaneous transplantation models. Although the tumor levels of perifosine remained almost stable at 48 h and beyond in both models ( Figures 6B, E ), some differences were observed. In the H1915 tumor xenografts, the AUC of perifosine was 10.7-fold (45 mg/kg) and 6.5-fold (180 mg/kg) higher than that in the DU 145 xenografts; this was corroborated by the data demonstrating that perifosine was favorably delivered into the H1915 tumor tissues, where the tumor-to-plasma ratios were much higher in the H1915 model ( Figures 6C, F , and Supplementary Table 3 ). These findings indicate H1915-preferential targeting of perifosine and are closely consistent with the above-mentioned results, which showed anti-proliferative activity in vitro and antitumor activity in orthotopic and ectopic tumor models were much stronger against H1915 than against DU 145 ( Figures 1D , 3 , and 4 ). Although tumor perifosine concentrations after repeated oral administrations in the 180/45 mg/kg group were higher than those after single doses in the 45 and 180 mg/kg groups, the difference was slight ( Figures 6B, E ).