Quantification and visualization of phosphoinositides by quantum dot-labeled specific binding-domain probes.

Phosphoinositides (PI) play important regulatory roles in cell physiology. Localization and quantitation of PIs within the cell is necessary to understand their precise function. Currently, ectopic expression of green fluorescent protein (GFP)-fused PI-binding domains is used to visualize PIs localized to the cell membrane. However, ectopically expressed PI-binding domains may compete with endogenous binding proteins, thus altering the physiological functions of the PIs. Here, we establish a novel method for quantification and visualization of PIs in cells and tissue samples using PI-binding domains labeled with quantum dots (Qdot) as specific probes. This method allowed us to simultaneously quantify three distinct PIs, phosphatidylinositol 3,4,5-triphosphatase [PtdIns(3,4,5)P3), PtdIns(3,4)P2, and PtdIns(4,5)P2, in crude acidic lipids extracted from insulin-stimulated cells. In addition, the method allowed the PIs to be visualized within fixed cells and tissues. Sequential and spatial changes in PI production and distribution were detected in platelet-derived growth factor (PDGF)-stimulated NRK49F cells. We also observed accumulation of PtdIns(3,4)P2 at the dorsal ruffle in PDGF-stimulated NIH3T3 cells. Finally, we found PtdIns(3,4,5)P3 was enriched in lung cancer tissues, which also showed high levels of phosphorylated Akt. Our new method to quantify and visualize PIs is expected to provide further insight into the role of lipid signaling in a wide range of cellular events.

Phosphoinositides (PI) play important regulatory roles in a variety of cellular events, including signal transduction, cytoskeletal regulation, and membrane traffi cking (1)(2)(3). The formation and breakdown of PIs are catalyzed by families of PI kinases and phosphatases. Recent studies have revealed that the improper functioning of these with PBS containing 0.05% Tween 20, and then scanned with a FLA-8000 fl uorescent image analyzer (Fujifi lm, Tokyo, Japan).

Phosphoinositide extraction from cells
Cells were seeded into 60 mm dishes and starved with serumfree medium before stimulation with insulin, as described below. Lipids were extracted from cell lysates as described previously ( 18 ). The lipids were spotted onto nitrocellulose membranes and quantifi ed with Qdot-labeled 6Cys-PH domains, as described above.
Cell culture, transfection, and fl uorescence microscopy NIH3T3 cells were cultured in DMEM (Sigma-Aldrich, St. Louis, MO) containing 10% calf serum. Chinese hamster ovary (CHO) cells stably expressing the wild-type human insulin receptor (CHO-IR) were cultured in Ham's F-12 medium (Invitrogen) containing 10% fetal calf serum (FCS). NRK49F cells were maintained in DMEM with 10% FCS. NIH3T3 and NRK49F cells were starved with DMEM for 8 h before stimulation with 20 ng/ml PDGF-BB. CHO-IR cells were starved with Ham's F-12 medium for 8 h before stimulation with 100 nM insulin.
Transient transfection of CHO-IR cells was performed using Lipofectamine 2000 reagents (Invitrogen). After 24 h incubation, cells were starved with Ham's F-12 medium and stimulated for varying times with 100 nM of insulin.

Immunohistochemistry
Arrays of frozen human cancer tissues were purchased from ISU ABXIS (Seoul, South Korea). Tissues were fi xed as described above and incubated with anti-pAkt Ser473 antibody (Cell Signaling Technology, Beverley, MA) and Qdot-6Cys-GRP1-PH domains. TO-PRO-3 iodide was used for nuclear staining. Fluorescence images were obtained using an FV-1000 confocal microscope.

Construction of PI-specifi c probes
To determine whether ectopically expressed PI-binding domains infl uence cell morphology, we transfected CHO-IR cells with various PH domain constructs. Cells overexpressing GFP were used as a control and developed membrane ruffl ing upon insulin stimulation (supplementary Fig. I). In contrast, cells overexpressing any of the individual PH domains, except PLC ␦ 1, showed poor membrane ruffl ing, consistent with previous observations ( 17 ). other processes through their intrinsic membrane-bending abilities ( 14 ).
The intracellular localization of each PI species has been studied by transfection of cultured cells with green fl uorescent protein (GFP)-fused PH domains ( 15 ). However, this method may not accurately detect intracellular PIs. Balla et al. reported that the GFP-PLC ␦ 1-PH domain was localized only to the plasma membrane despite some of its target phosphoinositide, PtdIns(4,5)P 2 , being present in the Golgi membrane and secretory vesicles as well ( 16 ). In addition, simultaneously visualizing distinct PI species is diffi cult by using this conventional method, and overexpression of PI-binding domains could interfere with the normal function of PIs by competing with their target proteins. Indeed, Varnai et al. demonstrated that overexpression of isolated PH domains could inhibit PtdIns(3,4,5) P 3 -regulated cellular pathways ( 17 ). Finally, the major drawback of the GFP-PH domain method is that it cannot be applied to tissues or clinical samples.
In this study, we developed a new method for quantifying and visualizing multiple PIs using quantum dot (Qdot)-labeled PI-binding domains as specifi c probes. With this method, we succeeded in simultaneously quantifying the levels of PtdIns(3,4,5)P 3 , PtdIns(4,5)P 2 , and PtdIns(3,4)P 2 in crude lipid samples. We also visualized each PI in fi xed cells or tissues, and we demonstrated their differential membrane distribution, both spatially and temporally.

Constructs
To construct the GST fusion 6Cys proteins, tgctgctgctgctgctgc was inserted into the BamHI site of the pGEX6P-1 vector (GE Healthcare, Little Chalfont, Buckinghamshire, UK). The sequences of the GRP1-PH, PLC ␦ 1-PH, and TAPP1-2×PH domains were inserted into the appropriate sites of the vector. Site-specifi c mutagenesis of GRP1-PH and TAPP1-2×PH was performed by PCR using mutated primers. GST fusion proteins were purifi ed from E. coli BL21 with glutathione-sepharose (GE Healthcare) according to the manufacturer's protocol, and the GST tag was removed by incubation with PreScission protease at 4°C overnight. Each purifi ed 6Cys-PH domain was labeled with the Qdot antibody conjugation kit (Invitrogen, Carlsbad, CA) according to the manufacturer's protocol.

Quantifi cation of PtdIns(3,4,5)P 3 , PtdIns(4,5)P 2 , and PtdIns(3,4)P 2 in cells
We next measured the amount of PtdIns(3,4,5)P 3 , PtdIns(4,5)P 2 , and PtdIns(3,4)P 2 in insulin-stimulated CHO-IR cells. Acidic lipids were extracted from cell lysates, spotted onto a nitrocellulose membrane, and the levels of all three PIs were quantifi ed simultaneously with the Qdot-labeled probes. PtdIns(3,4,5)P 3 was found to be elevated within 10 s of insulin stimulation and then decreased over the next 5 min ( Fig. 4 ). In contrast, PtdIns(3,4) P 2 levels increased more slowly and reached a maximum 60 s after stimulation. This result supports previous observations that PtdIns(3,4)P 2 is produced mainly from the To overcome the limitations associated with the GFP-PH domain constructs, we established a new method using PIbinding domains as specifi c probes. For this, GRP1-PH, PLC ␦ 1-PH, and TAPP1-2×PH were selected as specifi c probes for PtdIns(3,4,5)P 3 , PtdIns(4,5)P 2 , and PtdIns(3,4) P 2 , respectively. We used TAPP1-2×PH for the detection of PtdIns(3,4)P 2 because this probe possesses two PH domains in tandem, owing to which the binding affi nity of this probe is approximately 10-fold higher than that with a single PH domain ( 6 ). Initially, we labeled the PH domain by covalently modifying the lysine residues with Qdots. However, these PH domains showed very low affi nities for the PIs, possibly due to electrostatic repulsion between the positively charged residues in the PH domains and the negatively charged phosphates in the PIs (data not shown). Thus, six tandem cysteine residues were introduced at the amino-termini of the PH domains to allow labeling of thiols with amine-derivatized, polyethylene glycol (PEG)coated Qdot nanocrystals, using the thiol crosslinker succinimidyl-4-( N -maleimidomethyl)cyclohexane-1-carboxy-(6-amidocaproate) (SMCC) ( Fig. 1A ). Newly designed Qdot probes (hereafter referred to as Qdot-6Cys-PH) retained the lipid specifi cities and sensitivities of the parent PH  less effi cient. We also carried out the same staining using a single Qdot probe to exclude the possibility that the localizations of multiple probes were affected by each other. Data shown in supplementary Fig. IV clearly demonstrate that there is virtually no difference between single-and multiple-stained samples (compare supplementary Fig. IV  with Fig. 5 ), confi rming that Qdot probes can be used not only to detect single PI but also to simultaneously detect three distinct PIs in cultured cells. In addition, we stained cells that expressed GFP-lipid binding domain with Qdot-labeled probe. GFP-PLC ␦ 1 PH domain was localized throughout plasma membranes and, correspondingly, Qdot-labeled PLC ␦ 1 PH probe also stained plasma membranes (supplementary Fig. V).

Visualization of PIs in insulin-stimulated CHO-IR cells
Next, we determined if the Qdot probes could be used to visualize PIs in cultured cells. First, various concentrations of saponin for permeabilization were tested with a slight modifi cation as described previously ( 19 ). As shown in supplementary Fig. III, saponin concentration of 1.5 mg/ml was suffi cient for the Qdot probes to visualize plasma membrane, although lower concentration appeared PIs at discrete regions of the plasma membrane. We examined whether there was a difference in the localization of the PIs in PDGF-stimulated NIH3T3 fi broblasts. We found that PtdIns(3,4,5)P 3 , PtdIns(4,5)P 2 , and PtdIns(3,4) P 2 were all localized at peripheral membrane ruffl es in response to PDGF stimulation ( Fig. 7A ). Interestingly, PtdIns(3,4)P 2 was more concentrated at dorsal ruffl es, plasma membrane protrusions that rapidly form in response to activation of PI3K. The observed increase in PtdIns(4,5) P 2 could result from activation of type I phosphatidylinositol-4-phosphate-5-kinases (PIP5KI), which are known to generate PtdIns(4,5)P 2 and affect organization of the actin cytoskeleton ( 21 ). LY294002 blocked the insulin-stimulated increases in PtdIns(3,4,5)P 3 and PtdIns(3,4)P 2 , as well as lamellipodia formation (supplementary Fig. VI). Importantly, the insulin-stimulated increase in PtdIns(3,4,5)P 3 and PtdIns(3,4)P 2 was not detected in cells expressing mutated PH domains (supplementary Fig. VII).

PtdIns(3,4)P 2 enrichment in PDGF-induced dorsal ruffl es
The PI3K-driven signaling pathway induces rapid reorganization of the actin cytoskeleton, which gives rise to not only lamellipodia but also peripheral and dorsal membrane ruffl es ( 22,23 ). These transient and dynamic changes in the actin cytoskeleton are initiated by the production of For quantifi cation of PtdIns(4,5)P 2 , lipids were diluted 100-fold before spotting. A standard curve was used for quantifi cation. Data are mean ± SD of three independent experiments.  consistent with previous reports ( 24,25 ) ( Fig. 7B ). As expected, no increase in PtdIns(3,4,5)P 3 and PtdIns(3,4)P 2 was detected at peripheral or dorsal ruffl es using the PH domain mutants (supplementary Fig. VIII).

Localization of PIs in polarized cells
The morphological changes that occur during cell migration are associated with a polarized distribution of PIs. PtdIns(3,4,5)P 3 is known to mediate the effect of growth cultured cells and tissue samples. Our method enables sensitive and simultaneous quantifi cation of three PIs in small amounts of crude acidic lipids more easily than conventional methods. PI metabolism has been studied using radioisotope labeling combined with TLC or HPLC techniques, which are technically demanding and complex ( 29,30 ). In addition, variations in metabolic labeling efficiency cannot theoretically be excluded. Although a new nonradioisotope method has been developed to quantify PtdIns(3,4,5)P 3 ( 31 ) , specialized HPLC-mass spectrometry equipment is required. In contrast, our method permits quantifi cation and visualization of PIs more easily and in a more direct manner. Moreover, PtdIns(3,4,5)P 3 and PtdIns(3,4)P 2 , which are present in very low amounts in the cell, were readily detected by the Qdot-labeled PH domain probes.
There are currently two methods to visualize phosphoinositides in cells: one is to express lipid binding domains as GFP-fused proteins in intact cells, and the second is to use commercially available anti-phosphoinositide antibodies or lipid binding domains in fi xed cells. Although GFPfusion PH domain probes can be used for live cell imaging, their expression allows limited resolution and may alter cell function, either by sequestering lipid or by promoting and/or interfering with lipid-protein interactions. Indeed, Varnai et al. showed that overexpression of isolated PH factors on the formation of peripheral ruffl es implicated in cell movement ( 26 ). Consistently, we noted that all three PIs were localized at the leading edge of the cell membrane in migrating cells; however, only PtdIns(4,5)P 2 was also found at the opposing side of the plasma membrane ( Fig. 8A -C ). This fi nding confi rms that the PIs display polarized distribution to the front and back of the cell during movement.

DISCUSSION
In this study, we describe a new method that allows quantifi cation of cellular PIs and their noninvasive visualization in membrane structures. Triton, which is commonly used in immunocytochemical experiments, induces clustering of PtdIns(4,5)P 2 detectable by fl uorescence resonance energy transfer (FRET), with visible clustering observed at 0.005% Triton ( 40 ). Therefore, saponin was used for membrane permeabilization. By doing so, PtdIns(3,4,5) P 3 , PtdIns(4,5)P 2 , and PtdIns(4)P at the plasma membrane were visualized in fi xed cells. However, this method allows only one species of PI to be stained ( 19,34,39 ), and it is important to analyze a number of PIs simultaneously because cellular events are coordinately controlled by different PI species. Our method allows visualization of three different PIs in stimulated cells, and the spatial resolution is much improved compared with the method using GFPfusion PH domains.
The number of cellular processes known to be directly or indirectly controlled by PIs has expanded dramatically. Thus, the ability to detect and visualize multiple species of PIs will play a critical role in understanding the many functions of these signaling lipids.
domains could inhibit PtdIns(3,4,5)P 3 -regulated pathways ( 17 ). In addition, because the expressed PH domains were distributed throughout the plasma membrane, this approach lacks the ability to spatially resolve two or three PIs at the plasma membrane.
Electron microscopic techniques have been used to map PI distribution at high resolution, using lipid binding domains and antibodies. Chemical fi xation of lipids is an essential methodological consideration when measuring PIs. Downes et al. have shown that aldehyde-based fi xation methods more effi ciently immobilize membrane proteins than membrane lipids ( 32 ). Fixation of lipids requires the inclusion of glutaraldehyde, either alone or in combination with formaldehyde ( 33,34 ). By using this method, PtdIns(3,4,5)P 3 was shown to localize at both plasma and nuclear membranes of agonist-induced cells, whereas PtdIns(3,4)P 2 localized at the plasma membrane and intracellular organelles ( 35,36 ). More recently, Fujita et al. developed an electron microscopic method that used freeze-fracture replicas for the detection of PtdIns(4,5)P 2 without chemical fi xation ( 37,38 ). Their method is superior in resolution but requires more sophisticated techniques.
Light microscopy imaging techniques have also been used to detect PIs in fi xed cells, including detection by anti-phosphoinositide antibodies ( 19,34,39 ). For this method, glutaraldehyde was used to preserve plasma