Phosphoinositide 3-kinase δ inactivation prevents vitreous-induced activation of AKT/MDM2/p53 and migration of retinal pigment epithelial cells

Phosphoinositide 3-kinases (PI3Ks) are a family of lipid kinases that play a critical role in transmitting signals from cell-surface molecules to intracellular protein effectors. Key PI3Ks include PI3Kα, PI3Kβ, and PI3Kδ, which are regulated by receptors. The signaling pathway comprising the PI3Ks, along with a Ser/Thr kinase (AKT), a proto-oncogene product (mouse double minute (MDM)2), and a tumor suppressor protein (p53), plays an essential role in experimental proliferative vitreoretinopathy (PVR), which is a fibrotic blinding eye disorder. However, which PI3K isoforms are involved in PVR is unknown. A major characteristic of PVR is the formation of epi (or sub)-retinal membranes that consist of extracellular matrix and cells, including retinal pigment epithelium (RPE) cells, glial cells, and macrophages. RPE cells are considered key players in PVR pathogenesis. Using immunoblotting and immunofluorescence analyses, we herein provide the evidence that PI3Kδ is highly expressed in human RPEs when it is primarily expressed in leukocytes. We also found that PI3Kδ inactivation through two approaches, CRISPR/Cas9-mediated depletion and a PI3Kδ-specific inhibitor (idelalisib), not only blocks vitreous-induced activation of AKT and MDM2 but also abrogates a vitreous-stimulated decrease in p53. Furthermore, we demonstrate that PI3Kδ inactivation prevents vitreous-induced proliferation, migration, and contraction of human RPEs. These results suggest that PI3Kδ may represent a potential therapeutic target for RPE-related eye diseases, including PVR.

Phosphoinositide 3-kinases (PI3Ks) 2 are a family of lipid kinases, which phosphorylate the 3-hydroxyl of the inositol ring of inositol lipids to generate phosphatidylinositol 3-phosphate, phosphatidylinositol 3,4-bisphosphate, and phosphatidylinositol 3,4,5-trisphosphate at the inner leaflet of the plasma membrane (1,2). Receptor-regulated PI3Ks include heterodimeric PI3K␣, PI3K␤, and PI3K␦ (1,2). Their regulatory subunit p85 that contains Src homology 2 and 3 domains can bind to the phosphorylated tyrosine in the YXXM motif of receptor tyrosine kinases (e.g. platelet-derived growth factor receptor) so that their catalytic subunit (p110␣, p110␤, or p110␦) can phosphorylate phosphatidylinositol 4,5-bisphosphate to become phosphatidylinositol 3,4,5-trisphosphate, which in turn can be bound by pleckstrin homology domain-containing proteins such as AKT, an oncogene product also known as protein kinase B (1,(3)(4)(5)(6). This binding facilitates AKT to be phosphorylated by phosphoinositide-dependent kinase I at threonine 308 and by the mammalian target of rapamycin complex 2 at serine 473 (1,6,7). Activation of AKT can stimulate multiple cellular processes such as cell survival, proliferation, growth, as well as migration (6,8,9). So PI3Ks play an essential role in transmitting signals from cell-surface molecules into the intracellular enzymes as well as in stimulating cellular responses (1,2,10). However, the contribution of most of the PI3K isoforms to organismal function and disease still remains obscured.
Deregulation of PI3Ks may initiate multiple diseases such as cancer and proliferative vitreoretinopathy (PVR) (14 -16), a blinding eye disease with a major characteristic of epi (or sub)-retinal membranes (ERMs) that consists of extracellular matrix and cells, which include RPEs, glial cells, fibroblasts, and macrophages (17,18). In the pathogenesis of PVR, RPEs are exposed to vitreous during open global injury or retinal re-attachment surgery, and these cells along with other cell types (e.g. glial cells) migrate, proliferate, grow, and secrete extracellular matrix, resulting in the formation of ERMs whose contraction causes retinal detachment (17)(18)(19)(20)(21)(22). However, which PI3K isoform(s) contribute to PVR pathogenesis is still unknown.
In this study, we found that PI3K␦ is highly expressed in human RPEs, and inactivation of PI3K␦ with its specific inhibitor idelalisib or CRISPR/Cas9-mediated depletion suppresses vitreous-induced activation of AKT and MDM2 and degradation of p53, proliferation, migration, as well as contraction of human RPEs.

PI3K␦ is highly expressed in RPEs
Previous studies have demonstrated that PI3Ks play a central role in experimental PVR (14,16), but which PI3K isoform contributes to the pathogenesis of PVR is still a mystery. In addition, PI3K␦ was originally reported to be restricted to hematopoietic cells (2,(23)(24)(25); however, surprisingly, Western blot analysis of the cell lysates showed that p110␦, a catalytic isoform of PI3K␦, was highly expressed in cultured human RPEs, including ARPE-19, a spontaneously arising retinal pigment epithelial cell line, and primary human fetal RPE (PRPEs), whereas low levels of p110␦ expression were detected in other cell lines examined, including human corneal fibroblast (HCF), mouse cone cells (661W), a human retinoblastoma cell line (Y-79), rabbit conjunctival fibroblast (RCF), and porcine aortic vascular endothelial cells (PAECs) (Fig. 1A). In addition, p110␣ (Fig. 1, A and D) was expressed in all the examined cell lines at similar levels, and p110␤ was expressed approximately twice more and p110␥ less in RPEs than those in 661W cells ( Fig. 1, B, E, and F). Specifically, expression of p110␦ in RPEs was ϳ10 times higher than that in 661W cells and around 1-fold lower than that in a mouse macrophage RAW264.7 cell line, which served as a positive control (Fig. 1, C and G). Taken together, these results suggest that PI3K␦ may play an essential role in vitreous-induced signaling events (e.g. AKT activation) in RPEs.

Depletion of p110␦ with CRISPR/Cas9
The technology of clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein (Cas) can be harnessed for depleting gene expression by genome editing (26 -29). Next, we identified that one sgRNA (named PK2, AGAGCGGCTCATACTGGGCG) from exon 4 in the human genomic PIK3CD locus (Fig. S1A) was able to guide SpCas9 to cleave the locus at its expected site (Fig. S1B), resulting in ϳ90% depletion of p110␦ in ARPE-19 cells analyzed by Western blotting (Fig. S1, C and D); in addition, p110␦ was hardly detectable by immunofluorescence in most of the ARPE-19 cells expressing PK2-sgRNA/Cas9 compared with those expressing lacZ-sgRNA/Cas9 (Fig. S1F). Furthermore, Western blotting and immunofluorescence analysis showed that this CRISPR/Cas9mediated depletion of p110␦ did not have significant impact on expression of p110␣ ( Fig. S1, C, E, and G), suggesting that this PK2-sgRNA is specific to PIK3CD, and the p110␦ antibody we used herein is specific to p110␦. were subjected to Western blot analysis using the indicated antibodies. This is representative of three independent experiments. D-G, intensity of the p110␣, -␤, -␥, and -␦ bands in A-C was first normalized to that of the corresponding ␤-actin bands and then calculated to establish the ratio of the control in the 1st lane. The graphs are mean Ϯ S.D. of three independent experiments. Data were analyzed using one-way ANOVA followed by the Tukey post test. **, ***, and **** are Ͻ0.01, Ͻ0.001, and Ͻ0.0001, respectively. ns, not significant.

PI3K␦ inactivation blocks AKT/MDM2/p53 in RPEs Depletion of p110␦ attenuates vitreous-induced activation of AKT and MDM2 and decrease of p53
Previously, we found that PI3K is critical for vitreous-mediated activation of AKT and MDM2 and decrease of p53 (16), and blockade of the interaction of MDM2 with p53 suppresses experimental PVR (30). Therefore, we investigated whether depletion of p110␦ in ARPE-19 cells had any influence on the pathway of AKT/MDM2/p53. Compared with treatments with a single growth factor or cytokine or their combination, vitreous treatment is more relevant to an in vivo environment for PVR development (31)(32)(33)(34). In addition, rabbits are usually used to induce PVR (15,(34)(35)(36) because the smaller size of the rabbit lens as compared with that of the eyeball permits manipulations to be performed within the eye, without causing any damage to the lens and retina; however, other common laboratory animals such as rats and mice are difficult to work with because of the small volume of vitreous (20,37,38). Therefore, ARPE-19 cells were treated with normal vitreous from experimental rabbits (RV) for 2 h based on a previously established time course (16). As shown in Fig. 2, vitreous induced an increase (4.1 Ϯ 0.3-fold) in phosphorylated AKT (Ser-473) and (3.3 Ϯ 0.2-fold) in phosphorylated MDM2 (Ser-166), and an ϳ70% decrease in p53, but depletion of p110␦ in these cells blocked these biochemical events, suggesting that PI3K␦ is a primary PI3K isoform mediating the vitreous-induced pathway of AKT/MDM2/p53.

Re-expression of p110␦ restores vitreous-induced activation of AKT and MDM2 and decrease of p53
To exclude the off-target effects of CRISPR/Cas9, we re-expressed a mouse PIK3CD cDNA in APRE-19 cells whose endogenous one had been depleted by CRISPR/Cas9 because the mouse PIK3CD would not be recognized by the humanspecific sgRNA (PK2) of the CRISPR/Cas9. As shown in Fig. 3, transducing the mouse PIK3CD into the p110␦-depleted ARPE-19 cells raised their p110␦ levels slightly higher than their original ones; in addition, re-expression of mouse p110␦ in the p110␦-depleted ARPE-19 cells recovered RV-induced AKT and MDM2 activation and p53 degradation (Fig. 3). These results further illustrate that PI3K␦ plays an essential role in the vitreous-induced signaling pathway of PI3K/AKT/MDM2/p53 in RPE cells.

Pharmacological inhibition of PI3K␦ effectively prevents vitreous-induced activation of AKT and MDM2 and reduction of p53
Next, a pharmacological approach was used to assess whether PI3K␦ was required for vitreous-induced activation of AKT and MDM2 and degradation of p53. Idelalisib, a specific inhibitor of PI3K␦, is a Food and Drug Administration (FDA)approved drug for treating certain cancers. To establish the maximum-tolerated dose of idelalisib in cultured RPEs, we treated ARPE-19 cells with serially increasing concentrations of idelalisib: 5, 10, 20, 40, 60, and 80 M in serum-free media for 72 h (10). The results showed that 80 M idelalisib became obviously toxic to ARPE-19 cells by morphology (e.g. cells shrinking) (Fig. S2A), and in these cells the IC 50 of idelalisib was 61.5 M, which was determined by MTT assay (Fig. S2B). However, 10 M idelalisib did not show obvious toxicity to ARPE-19 cells except that cell number was less than those treated by its vehicle dimethyl sulfoxide (DMSO), which was assessed by morphology under a microscope and MTT assay (Fig. S2).
To examine the impact of idelalisib on the PI3K/AKT/ MDM2/p53 pathway, we still monitored phosphorylation of AKT at serine 473 and phosphorylation of ERK at tyrosine 204 as indirect measures of activation of these kinases (39). In addition, it has been reported that the clinical trial dose of idelalisib for drinking is 150 mg twice per day (23), and there are a number of publications using 10 M idelalisib to inactivate PI3K␦ in cultured cells (40 -42) even though its IC 50 to inhibit p110␦ activity is 19 nM, which was evaluated by an in vitro biochemical assay (10). So we treated ARPE-19 cells with idelalisib (0.1, 1, and 10 M) in addition to RV for 2 h. Western blot analysis of their lysates showed that idelalisib at 1 M completely abrogated vitreous-induced activation of AKT, in contrast with ERK activation, which was not affected even if the cells were treated with idelalisib at 10 M (Fig. 4, A and B). Importantly, idelalisib at 10 M failed to inhibit PDGF-BB-induced AKT activation in 661W cells (Fig. 4, E and F), in which p110␦ was expressed at very low levels compared with that in ARPE-19 cells (Fig. 1), indicative of idelalisib at 10 M being selective. Furthermore, BYL719, whose IC 50 to inhibit p110␣ activity is 4.0 -4.8 nM (43), did not inhibit vitreous-induced AKT activation at 400 nM (100 times of its IC 50 ) in ARPE-19 cells (Fig. 4, C and D), but at this dose BYL719 inhibited PDGF-BB-induced activation of AKT in 661W cells (Fig. 4, E and F). These results suggest that PI3K␦ is a predominantly functional receptor-regulated PI3K isoform for vitreous-induced activation of AKT in RPEs. Therefore, we next evaluated the impact of idelalisib on the AKT/ MDM2/p53 pathway. As shown in Fig. 4, A, B, G, and H, idelalisib at 1 M blocked vitreous-induced activation of AKT and MDM2 and suppressed vitreous-stimulated impact on p53 expression. This result is consistent with the depletion of p110␦ Their lysates were subjected to Western blot analysis using the indicated antibodies. This is representative of three independent experiments. B, intensity of the bands (e.g. in A) from three independent experiments was quantified in the graphs. **** indicates significant difference (p Ͻ 0.001) using one-way ANOVA followed by the Tukey post test. ns, not significant.

Inactivation of PI3K␦ prevents vitreous-stimulated proliferation of RPEs
Activation of the PI3K/AKT/MDM2/p53 pathway can stimulate cell proliferation (2,16), which is also a main characteristic for pathogenesis of PVR (16). Thus, we examined whether inactivation of PI3K␦ using the two approaches could block vitreous-induced cell proliferation. As shown in Fig. 5, A and B, although vitreous stimulated (1.8 Ϯ 0.2-fold) proliferation of ARPE-19 cells, either depletion of p110␦ with CRISPR/Cas9 or inhibition of PI3K␦ with idelalisib significantly suppressed vitreous-induced proliferation.

Inactivation of PI3K␦ blocks vitreous-induced migration of RPEs
Activation of the PI3K/AKT/MDM2/p53 signaling pathway also affects cell motility (2,16). During PVR pathogenesis, RPE cells proliferate and migrate from the RPE monolayer to form sheets of dedifferentiated cells, a major cellular constituent of ERMs from patients with PVR (44). We next examined the impact of PI3K␦ inactivation using the two approaches on vitreous-induced migration of ARPE-19 cells with a wound-healing assay, in which the wound was created by scratching the cells on culture plates with a 200-l pipette tip (37,45). As illustrated in Fig. 6, while vitreous stimulated cell migration into the wound areas, inactivation of PI3K␦ by either p110␦ depletion with CRISPR/Cas9 or a specific pharmacological inhibitor idelalisib at 10 M reversed vitreous-induced migration of ARPE-19 cells to levels that were even lower than in controls, indicating that both induced and constitutive PI3K␦ activity contribute to cell migration.

Inactivation of PI3K␦ abrogates vitreous-induced contraction of RPEs
A main characteristic of PVR is the formation of ERMs, whose contraction causes retinal detachment (17,34,46). To mimic the interaction of cells with extracellular matrix for contraction, an in vitro contraction assay was used even though this assay was far from the in vivo environment (37,46). To investigate the impact of PI3K␦ inactivation on the capability of the cell contraction, two approaches by either depletion of p110␦ with CRISPR/Cas9 or inhibiting its kinase activity with a specific inhibitor idelalisib were utilized. As shown in Fig. 7A, vitreous significantly induced contraction of cells expressing lacZ-sgRNA/Cas9, and the contraction capability was significantly attenuated in the RPEs expressing PK2-sgRNA/Cas9; in addition, idealisib also blocked the vitreous-induced contraction of ARPE-19 cells in this collagen contraction assay (Fig.  7B).

Discussion
In this paper, we report that p110␦ is highly expressed in RPEs, and inactivation of PI3K␦ using molecular and pharmacological approaches blocks vitreous-induced activation of AKT/MDM2/p53 as well as cellular responses intrinsic to the development of PVR. RPEs have several functions (light absorption, epithelial transport, spatial ion buffering, visual cycle, phagocytosis, secretion, and immune modulation) (47) and are involved in age-related macular degeneration (48), retinitis pigmentosa (49), diabetic retinopathy (50,51), and PVR (35). p110␦ has been found to be mostly expressed in hematopoietic systems, including myeloid cells, B cells, and T cells, and plays key roles in leukocyte signaling, proliferation, differentiation, activation, and chemotaxis (2, 13, 52, 53). Our findings reported herein expand the research scope of PI3K␦, which at present is A, ARPE-19 cells whose endogenous p110␦ had been depleted by expressing PK2-sgRNA/Cas9 were transduced with retroviruses that contained a mouse PIK3CD mRNA, and an empty retroviral vector pLNCX 3 was serum-deprived for 24 h and then treated with RV for 2 h. Their lysates were subjected to Western blot analysis using the indicated antibodies. This is representative of three independent experiments. B, intensity of the bands in A from three independent experiments was quantified in the graphs. **** indicates significant difference (p Ͻ 0.001) using one-way ANOVA followed by the Tukey post test. ns, not significant. (m)p110␦, mouse p110␦.

PI3K␦ inactivation blocks AKT/MDM2/p53 in RPEs
mostly focused on the hematopoietic system, indicating deregulation of PI3K␦ may involve in RPE-related pathology such as PVR.
Based on that, idelalisib inhibits vitreous-induced activation of AKT in RPEs but not PDGF-BB-stimulated AKT activation in 661W cells (Fig. 4). Because RPEs express much more p110␦ than do 661W cells (Fig. 1), we propose the following three hypotheses: 1) in RPE cells PI3K␦ is one of the predominant isoforms of PI3Ks; 2) vitreous induces a specific p85 regulatory subunit to activate p110␦; and 3) vitreous stimulates a specific AKT isoform via PI3K␦ for activating MDM2. These hypotheses are actively being examined in my laboratory.
Idelalisib is a potent, small-molecule inhibitor of PI3K␦ that is highly selective for the ␦ isoform of PI3K, as compared with the ␣, ␤, and ␥ isoforms of PI3Ks. In lymphoid cell lines and primary samples from patients, idelalisib blocks PI3K␦-AKT signaling (23,42). Similarly, inactivation of PI3K␦ abrogated vitreous-stimulated activation of AKT and degradation of p53 (Figs. 2-4). AAV-based CRISPR/Cas9 has been in clinical trial for treating human diseases especially eye disease (29); in addi- Then, their lysates were subjected to Western blot analysis using the indicated antibodies. This is representative of three independent experiments. E, serum-deprived 661W cells were treated with PDGF-BB in addition to idelalisib (1 and 10 M), BYL719 (0.1, 0.2, and 0.4 M), or its vehicle (0.1% DMSO, V) for 15 min. Then, their lysates were subjected to Western blot analysis using the indicated antibodies. This is representative of three independent experiments. B, D, F, and H, intensity of the p-AKT, p-MDM2, or p53 bands was first normalized to that of the corresponding AKT, MDM2, or ␤-actin bands and then calculated to establish the ratio of the control in the 1st lane. The graphs are mean Ϯ S.D. of three independent experiments. **** indicates a significant difference between the compared two groups at p Ͻ 0.0001, respectively, using one-way ANOVA followed by the Tukey post test. ns, not significant.

PI3K␦ inactivation blocks AKT/MDM2/p53 in RPEs
tion, idelalisib has been approved by the FDA for the treatment of certain hematological malignancies (2,53); therefore, these genome-editing and pharmacological (idelalisib) approaches to targeting PI3K␦ have great potential for therapeutic treatment of RPE-related diseases, including PVR.
To express SpGuides in the targeted cells, the top oligonucleotides 5Ј-CACCG-20-nt (target PI3KCD DNA sequence PK2 (AGAGCGGCTCATACTGGGCG) or the lacZ sgRNA sequence)-3Ј and the bottom oligonucleotides 5Ј-AAAC-20-nt (20-nt: complementary target PI3KCD DNA sequences or lacZ sgRNA sequence)-C-3Ј were annealed and cloned into the len-tiCRISPR v2 vector by BsmBI (New England Biolabs, Boston, MA), respectively. All clones were confirmed by DNA sequencing using primer 5Ј-GGACTATCATATGCTTACCG-3Ј from a sequence of the U6 promoter that drives expression of sgRNAs. DNA synthesis and sequencing were done by Massachusetts General Hospital DNA Core Facility (Cambridge, MA).

Production of lentivirus
Lentivirus was produced as described previously (72). Briefly, the lentiCRISPR v2 vector inserted with sgRNA, the packaging plasmid psPAX2 (Addgene, 12260), and the envelope plasmid VSV-G (Addgene, 8454) were mixed together and then added to a mixture of Lipofectamine 3000 (Thermo Fisher Scientific) with Opti-MEM (Thermo Fisher Scientific). This transfection mix was kept at room temperature for 30 min and then carefully added into HEK 293T cells in a 60-mm cell culture dish. After 18 h (37°C, 5% CO 2 ), the medium was replaced with growth medium supplemented with 20% FBS, and lentiviruses were harvested at 24 h after changing the medium and then daily for 2 days. The virus-containing media were pooled and centrifuged at 800 ϫ g for 5 min to remove the cell debris. The supernatant was used to infect APRE-19 cells supplemented with 8 g/ml Polybrene (Sigma). The infected cells were selected in media using puromycin (Sigma) (4 g/ml), and the resulting cells were examined by Western blotting (21,31,75).

Re-expression of p110␦
This experiment was performed as described previously (37). Briefly, the mouse PIK3CD cDNA in the vector pCMV-SPORT6 (MMM1013-202765002, GE Dharmacon, Waltham, MA) was subcloned into a retroviral vector of pLNCX 3 (LTR-CMV-SalI-HpaI-SacI-NotI-BgeII-ClaI-LTR) derived from pLNCX (GenBank TM accession no. M28247; National Institutes of Health, Bethesda) by SalI/NotI (37) and confirmed by Sanger DNA sequencing. The sequence-confirmed retroviral vector was transfected into 293GPG cells for producing retroviruses, which were used for infecting ARPE-19 cells whose endogenous p110␦ had been depleted by the CRISPR/Cas9. This mouse PIK3CD cDNA circumvented recognition by the human originated PK2-sgRNA of the CRISPR/Cas9, and its expression was confirmed by Western blotting.

Immunofluorescence
This experiment was performed as described previously (37). Briefly, cells were fixed in 3.7% formaldehyde/PBS for 10 min at PI3K␦ inactivation blocks AKT/MDM2/p53 in RPEs room temperature. Then, the samples were blocked with 5% normal goat serum in 0.3% Triton X-100/PBS for 30 min and incubated with a primary antibody from rabbit against p110␦ (Abcam) or p110␣ (1:100 dilution) in combination with mouse against pan-keratin or nonimmune mouse and rabbit IgG overnight at 4°C. After thorough washes with PBS to get rid of the nonspecific binding, the samples were incubated with fluorescently-labeled secondary antibodies Dylight 549 (anti-rabbit IgG) and Dylight 488 (anti-mouse IgG) (Vector Laboratories, Inc., Burlingame, CA) (1:300 dilution in a blocking buffer) for 1 h. After thorough washes with PBS, the slides were mounted with a mount medium with 4Ј,6-diamidino-2-phenylindole (DAPI) (Vector Laboratories). These slides were observed and photographed under a fluorescent microscope (30,37,69).

MTT assay
The toxicity of Idelalisib on cell viability was assessed using the MTT assay (Solarbio). ARPE-19 cells were seeded into 96-well plates in their respective medium supplemented with 10% FBS and incubated for 24 h before exposure to a range of concentrations (0, 12.5, 25, 50, and 100 M) of idelalisib dissolved in DMSO in 5% FBS-supplemented medium or serumfree medium for 24 -72 h. Subsequently, the medium was removed and replaced with 200 l of 0.5 mg/ml MTT in 5% FBS-containing medium or serum-free medium, and the cells were incubated in the CO 2 incubator at 37°C for 2 h. Supernatants were removed from the wells, and the MTT dye was solubilized in 150 l/well DMSO. Absorbance was measured at 570 nm on a plate reader. Each condition was tested with six replicates, and all assays were performed in triplicate (76,77).

Cell proliferation assay
Cells were plated in 24-well plates at a density of 3 ϫ 10 4 cells/well in DMEM/F-12 supplemented with 10% FBS. Following attachment, the cells were treated with DMEM/F-12 or RV (1:3 dilution in DMEM/F-12) plus idelalisib (10 M) or its vehicle DMSO. After treatment for 48 h, the cells were trypsindetached from the plates and counted in a hemocytometer under a light microscope. At least three independent experiments were performed as described previously (31,33,75).

Cell migration assay
After cells in 24-well plates were grown to confluence, they were serum-starved overnight. Then, a wound was created by scratching the monolayer with a 200-l pipette tip (37,45). The cells were washed with PBS and treated with DMEM/F-12 or RV (1:3 dilution in DMEM/F-12) plus idelalisib (10 M) or its vehicle DMSO. The scratched area was photographed to capture the initial width and photographed again 16 -20 h later. Data were analyzed using Adobe Photoshop CS4 software. At least three independent experiments were performed.

Collagen contraction assay
Cells were resuspended in 1.5 mg/ml neutralized collagen I (INAMED, Fremont, CA), pH 7.2, on ice at a density of 1 ϫ 10 6 /ml cells/ml (15,75). The cell and gel mixture was transferred into wells of 24-well plates, which had been preincubated overnight with 5 mg/ml BSA/PBS. When the gel became a solid jelly-like soft material after incubation at 37°C for 90 min, 0.5 ml of DMEM/F-12 or RV (1:3 dilution in DMEM/F-12) was added with or without idelalisib (10 M). On day 3, the gel diameter was measured, and the gel area was calculated using a formula 3.14 ϫ r 2 , where r is the radius of the gel. At least three independent experiments were performed as described in previous reports (37,75,78).

Statistics
The experimental data were analyzed using an unpaired t test between two groups and one-way ANOVA among more than two groups followed by the Tukey post test using GraphPad Prism Version 5c. A p value less than 0.05 is considered statistically significant.