Psoralen and Ultraviolet A Light Treatment Directly Affects Phosphatidylinositol 3-Kinase Signal Transduction by Altering Plasma Membrane Packing*

Psoralen and ultraviolet A light (PUVA) are used to kill pathogens in blood products and as a treatment of aberrant cell proliferation in dermatitis, cutaneous T-cell lymphoma, and graft-versus-host disease. DNA damage is well described, but the direct effects of PUVA on cell signal transduction are poorly understood. Because platelets are anucleate and contain archetypal signal transduction machinery, they are ideally suited to address this. Lipidomics on platelet membrane extracts showed that psoralen forms adducts with unsaturated carbon bonds of fatty acyls in all major phospholipid classes after PUVA. Such adducts increased lipid packing as measured by a blue shift of an environment-sensitive fluorescent probe in model liposomes. Furthermore, the interaction of these liposomes with lipid order-sensitive proteins like amphipathic lipid-packing sensor and α-synuclein was inhibited by PUVA. In platelets, PUVA caused poor membrane binding of Akt and Bruton's tyrosine kinase effectors following activation of the collagen glycoprotein VI and thrombin protease-activated receptor (PAR) 1. This resulted in defective Akt phosphorylation despite unaltered phosphatidylinositol 3,4,5-trisphosphate levels. Downstream integrin activation was furthermore affected similarly by PUVA following PAR1 (effective half-maximal concentration (EC50), 8.4 ± 1.1 versus 4.3 ± 1.1 μm) and glycoprotein VI (EC50, 1.61 ± 0.85 versus 0.26 ± 0.21 μg/ml) but not PAR4 (EC50, 50 ± 1 versus 58 ± 1 μm) signal transduction. Our findings were confirmed in T-cells from graft-versus-host disease patients treated with extracorporeal photopheresis, a form of systemic PUVA. In conclusion, PUVA increases the order of lipid phases by covalent modification of phospholipids, thereby inhibiting membrane recruitment of effector kinases.

Psoralen and ultraviolet A light (PUVA) are used to kill pathogens in blood products and as a treatment of aberrant cell proliferation in dermatitis, cutaneous T-cell lymphoma, and graftversus-host disease. DNA damage is well described, but the direct effects of PUVA on cell signal transduction are poorly understood. Because platelets are anucleate and contain archetypal signal transduction machinery, they are ideally suited to address this. Lipidomics on platelet membrane extracts showed that psoralen forms adducts with unsaturated carbon bonds of fatty acyls in all major phospholipid classes after PUVA. Such adducts increased lipid packing as measured by a blue shift of an environment-sensitive fluorescent probe in model liposomes. Furthermore, the interaction of these liposomes with lipid order-sensitive proteins like amphipathic lipid-packing sensor and ␣-synuclein was inhibited by PUVA. In platelets, PUVA caused poor membrane binding of Akt and Bruton's tyrosine kinase effectors following activation of the collagen glycoprotein VI and thrombin protease-activated receptor (PAR) 1. This resulted in defective Akt phosphorylation despite unaltered phosphatidylinositol 3,4,5-trisphosphate levels. Downstream integrin activation was furthermore affected similarly by PUVA following PAR1 (effective half-maximal concentration (EC 50  Psoralens are linear furanocoumarins. Their heterocyclic aromatic core reversibly intercalates in nucleic acid clefts of cell nuclei (1). Absorption of ultraviolet A (UVA) 2 light excites the psoralen, supplying energy for covalent bond formation. When in the vicinity of nucleic acids, the reaction is primarily with pyrimidine moieties (2). This chemical addition is irreversible and prevents normal replication of DNA (3), explaining why severe phytophotodermatitis occurs in sun-exposed skin that has been in contact with psoralen-producing plants. This biostatic photochemistry is utilized in a pathogen inactivation technology (Intercept Blood System, Cerus Corp., Concord, CA) for plasma and platelet concentrates in transfusion medicine (4) where it prevents replication of potentially infectious pathogens transmitted from the donor to the blood product. In dermatology, PUVA is used to treat psoriasis, certain dermatoses, and cutaneous T-cell lymphoma (5). In these cases, the psoralen is administered topically or systemically followed by local or whole body UVA illumination. Treatment modalities are diverse and may differ by psoralen derivative and/or by doses of UVA light (6). Furthermore, PUVA is under clinical investigation for the treatment of multiple other diseases mostly involving uncontrolled cell proliferation (7). However, although broadly applied in medicine, the biochemical consequences of PUVA are poorly understood.
Psoralens are not clinically efficacious without UVA illumination, and despite its well documented photoreaction with DNA, PUVA photochemistry is not specific for this particular biologic target alone. In theory, photoexcited psoralens can react with many other common biomolecules. Therefore, the net effect of PUVA in a cellular milieu is unlikely to be restricted to consequences of its chemistry with solely DNA as was acknowledged already in the late seventies (8). This random reaction with multiple biomolecules explains why the quest for understanding PUVA and optimizing therapies has been exceptionally challenging. For example, the systemic PUVA treatment modality ECP is used to treat cutaneous T-cell lymphoma, and although clinically successful (9), the mechanism of action has remained elusive ever since its first application. Researchers speculate that induction of apoptosis through DNA damage causes priming of the immune system. However, apoptosis induction alone is insufficient for the typical immunomodulation caused by PUVA (10,11). Despite this uncertainty, ECP is used "off-label" for (auto)immune diseases including GVHD, bronchiolitis after stem cell transplantation, and Crohn's disease (12,13). But without comprehension of definite biochemical mechanisms, both protocol uniformity and outcome variability will remain problematic.
Adduct formation of photoexcited psoralens with targets other than DNA (11,14) is primarily with membrane components (15,16). Psoralens efficiently partition into lipid bilayers (17,18) where chemical reactions with double carbon bonds in unsaturated acyl chains take place (19,20). As a consequence, rational hypotheses on the cell biological consequences of these modifications have been postulated (21,22), but mechanisms explaining how membrane modifications exactly impact downstream cell signal transduction are unknown.
We chose platelets as model cells to study this because they have no nucleus, are easily isolated from whole blood, and contain a complex signal transduction machinery exemplary of key signaling processes in most eukaryotic cells. Furthermore, recent work has shown that PUVA significantly affects platelet function in an in vitro model of hemostasis (23). Our data now show that PUVA can inhibit signal transduction in the phosphatidylinositol 3-kinase (PI3K) pathway by preventing efficient plasma membrane recruitment of crucial effector kinases such as Akt (alternative name, protein kinase B) and Bruton's tyrosine kinase (Btk). Our findings are not restricted to platelets because T-lymphocytes in patient samples treated with ECP displayed attenuation of Akt phosphorylation as well.

PUVA Modifies Phospholipids by Covalent Adduct
Formation-Adducts between nucleic acids and psoralens like amotosalen or 8-methoxypsoralen (8-MOP) are formed in nucleated cells, but chemistry with (phospho)lipids has been described as well (15,24). Mass spectrometry of platelet membrane extracts confirmed that PUVA causes amotosalen-lipid adduct formation, which was not found in control cells that were incubated with this psoralen in the dark (Fig. 1, a and b). All major phospholipid classes were targeted (Fig. 1c) provided they contained acyl chains with unsaturated carbon bonds (19). Fragmentation MS/MS spectra allowed identification of the main adducts that formed, and Fig. 1d depicts exemplary MS/MS data for PS(18:0/20:4), bisphosphorylated phosphatidylinositol (PIP 2 )(18:0/20:4), amotosalen, and their UVA-dependent adducts.
PUVA Increases Lipid Order, Causing Inhibition of Protein Binding-Because photochemical addition of psoralens is to unsaturated carbon bonds, we hypothesized that PUVA causes changes in the order of lipid phases because this is typically controlled by acyl chain (un)saturation. To assess this, the environment-sensitive fluorescent probe NR12S (25,26) was used in liposomes composed of PC(16:0/18:1). The emission peak of NR12S shifts blue in liquid-ordered liposomes, causing a rise in the generalized polarization (GP) value. In a validation experiment, this was demonstrated by addition of cholesterol, which is known to enhance lipid packing (Fig. 2a) (26). Ultraviolet A light treatment of amotosalen-containing liposomes caused apparent increases in lipid order similar to the addition of cholesterol, suggesting that phospholipid-amotosalen adducts tighten the acyl chain alignment.
PUVA Affects Akt and Btk Plasma Membrane Binding and Subsequent Signal Transduction-During platelet activation, essential interactions between membrane (phospho)lipids and cytoplasmic soluble proteins take place. These then relay signals to secondary messenger molecules, establishing successful primary hemostasis. Because PUVA affects protein interaction with model liposomes, we questioned whether this membrane binding of key cytoplasmic signaling proteins like Akt and Btk was intact in PUVA-treated platelets. To activate the initiating PI3K pathway in platelets, the collagen GPVI or the thrombin PAR1 was stimulated. Membrane binding of Akt and Btk was assessed by Western blotting of isolated membrane fractions. Significantly less Akt and Btk was found in PUVA versus paired untreated controls, pointing to defects in membrane binding in live cells as well (Fig. 3a). This defective binding moreover caused significant decreases in subsequent Akt phosphorylation, a major event in the PI3K signal transduction pathway ( Fig. 3b) (31).
Binding of Akt and Btk to the activated membrane requires at least increased levels PI(3,4,5)P 3 , which is produced by phosphorylation of PIP 2 by PI3K. Quantitative mass spectrometry on membrane extracts ruled out that PI(3,4,5)P 3 formation is abnormal in PUVA-treated resting and activated platelets (Fig.  4a). Furthermore, the relative amount of PIP 2 -amotosalen adducts to total PIP 2 was only 1.2%, confirming that direct effects on the PI3K substrate were unlikely to cause decreased Akt membrane recruitment and phosphorylation. Furthermore, signal transduction in the concurrent phospholipase C (PLC) pathway was not affected by PUVA because the level of pleckstrin phosphorylation (Fig. 4b) and calcium release (data not shown) were normal in response to GPVI and PAR1 stimulation. Of note, platelet activation by another thrombin ana-logue that specifically targets PAR4 instead of PAR1 was not affected by PUVA and led to normal Akt phosphorylation (Fig.  4c). This is important because it demonstrates that the intrinsic potential to phosphorylate Akt is still present in PUVA-treated cells but depends on the kind of activatory input. Together, these data show that, despite generic adduct formation of psoralen with acyl chains, consequences for signal transduction are relatively selective and specific.
Platelet Integrin Activation Is Decreased by PUVA-Downstream of the investigated signaling circuit, platelets respond by activation of integrin ␣ IIb ␤ 3 . Integrin activation of PUVAtreated platelets was investigated with flow cytometry using the conformation-sensitive monoclonal antibody PAC-1 (32 (Fig. 5, a and b). In line with the Akt phosphorylation data, activation through PAR4 was normal (EC 50 , 50 Ϯ 1 versus 58 Ϯ 1 M) (Fig. 5c). As a control, the plasma membrane expression levels of the receptors GPVI and PAR1 was determined in flow cytometry and found to be normal (Fig. 5d).
PUVA Decreases Phosphorylation of Akt in T-lymphocytes from GVHD Patients-As a model for PUVA treatment of nucleated human cells, leukocytes were isolated from healthy consenting volunteers. The T-lymphocytes were activated by anti-CD3/CD28 beads. The effect of PUVA on T-cell activation was determined for 8-MOP and amotosalen in a paired experiment. The data show that both psoralens dose-dependently inhibited phosphorylation of Akt ( Fig. 6, a and b), which was not caused by immediate cell death (data not shown). The data show that amotosalen was more efficient than 8-MOP, requiring lower concentrations to achieve half-maximal inhibition. To confirm whether Akt activation is affected in clinically relevant tissue, buffy coat cells from GVHD patients enrolled in a clinical efficacy study were collected before and after ECP. In these samples, Akt activation in T-lymphocytes was also significantly decreased (Fig. 6, c and d). Together, these data show that PI3K-dependent signaling is affected by PUVA in cells other than platelets and in the clinical setting of ECP treatment.

Discussion
Despite its widespread clinical use, many aspects of PUVA's mode of action have remained uncertain. Our data show for the first time that PUVA treatment specifically inhibits the PI3K-Akt pathway by interfering with efficient recruitment of effector kinases Akt and Btk to the activated plasma membrane in response to specific stimuli. Our model (Fig. 7) suggests that psoralen-phospholipid adduct formation in the plasma membrane prevents efficient interaction between effector proteins specifically interacting with microenvironments of decreased lipid packing harboring PI(3,4,5)P 3 .
Covalent binding of photoexcited psoralen to unsaturated fatty acids has been described before in cell-free test tubes and in cell extracts (19,20,24). Our mass spectra of PUVAtreated platelet extracts confirm covalent reaction with unsaturated acyl side chains. All investigated phospholipid classes formed adducts with the amotosalen, indicating no preference for the hydrophilic headgroup. Despite this seemingly unspecific photochemistry, signal transduction inhibition was remarkably specific for the PI3K pathway. Phosphorylation of Akt was decreased, but the concurrent PLC axis showed no defect even though both pathways utilize the same PI(4,5)P 2 substrate. Furthermore, PI3K activity was not directly affected by the presence of free amotosalen and its adducts with PI(4,5)P 2 because PI(3,4,5)P 3 levels were normal in response to platelet activation and because these particular adducts were not abundant. Consequently, specific modification of only inositides is not underlying the PI3K signal transduction defect.
Instead, PUVA inhibits normal plasma membrane recruitment of Akt and Btk in the presence of normal PI(3,4,5)P 3 levels, suggesting that the interaction is dependent on auxiliary facilitators. Because unsaturated fatty acids are the prime target of PUVA chemistry, such facilitators thus require normal (levels of) fatty acyl unsaturations to fulfill their role. Plasma membrane binding of small effector kinases is well studied, and even though it is assumed to be primarily between the PH domain sequence KX n (K/R)XR (33) and PI(3,4,5)P 3 , additional factors (in)directly support binding. For instance, the related PH domains of PLC␦1 and GRP1 are known to harbor amino acids that penetrate the inner hydrophobic core of the membrane (34,35). Such membrane insertion often requires the targeted membrane compartment to be less tightly packed (29,36). Now our data show that binding of packing-sensitive probes ALPS and ␣-synuclein to model bilayers is hampered by PUVA. This indicates that amino acid insertion into the membrane core facilitated by unsaturated phospholipids is inhibited by PUVA. We thus hypothesize that lipid packing is the additional factor that cooperatively strengthens Akt and Btk binding as suggested by the coincidence detection (37) hypothesis of PH domain interactions with the plasma membrane. Our data provide the first proof in live cells that the acyl side chain milieu of phospholipids co-determines binding success of major PH domain-containing effector kinases in the presence of normal PI(3,4,5)P 3 levels.
The extent of Akt and Btk inhibition by PUVA, however, varies with the stimulus. Integrin ␣ IIb ␤ 3 activation was most affected following GPVI stimulation, less by PAR1, and not at all by PAR4. This shows that the defect is not restricted to either G-protein-coupled or immune receptor activation. It also means that the integrin conformational machinery is not affected directly by PUVA and able to respond normally if signals come in correctly from the upstream cue. How this variability in response is established, however, remains uncertain. Maybe the separate ligand-receptor circuits reside in plasma membrane microdomains of specific phospholipid composition (38) that is variably susceptible to modifications by PUVA. This follows the biochemical dogma of PI3K-dependent signaling stating that PI(3,4,5)P 3 molecules are scaffolding molecules restricted in space and time for controlled signalosome construction (39). The experimental setting allows activation through a single receptor-ligand couple, resulting in a distinct identifiable signal. However, in a real life physiologic environment, multiple extracellular chemicals act in concert at varying concentrations in space and time. Therefore, the net effect of PUVA on the diverse tissues that are clinically treated is difficult to predict based on our findings. However, because many signaling pathways rely on membrane recruitment, the output of PUVA-treated tissue in response to physiologically relevant chemicals will be undoubtedly altered.
In platelets, this altered output presents as delayed thrombus formation on immobilized collagen in vitro (23). Experimental inhibition of PI(3,4,5)P 3 -dependent signaling in platelets indeed shows significant function defects for instance in mouse  (40), or PI3K isoform deficiency (41). In human B-cell malignancy, Btk inhibition by ibrutinib is clinically successful but increases the risk of bleeding by platelet inhibition (42). Platelets are currently treated by PUVA in certain blood banks in Europe to prevent transmission of pathogens during transfusion. This treatment efficiently reduces pathogen and white blood cell proliferation by damaging their DNA. Originally platelets were believed to be unsusceptible to PUVA because they do not contain a nucleus. Our findings have now revealed the mechanism of decreased function caused by side effects of PUVA and unrelated to DNA damage. The clinical consequences of transfusing psoralen-modified phospholipids are unknown but worthwhile investigating.
Our findings also have consequences for other fields that commonly apply PUVA. The exact biochemical drivers of PUVA's clinical success in skin pathologies or GVHD for instance remain poorly understood. We propose a substantial role for selective PI3K axis inhibition given the results of the buffy coat samples from our GVHD patient that clearly demonstrate reduced Akt phosphorylation. However, this single readout merely demonstrates attenuation of just one effector kinase (Akt), in one cell type, and to one ligand-receptor combination. The exact cellular composition of PUVA-treated tissue by type and number as well as their individual susceptibility to PUVA will eventually determine the response in a relevant physiological environment. Further study will thus require a systems approach (43) to unravel the dependence of relevant ligand-receptor couples to PUVA and combine the findings in an integrated view, ideally also taking into account effects on DNA replication and transcription where possible.
In conclusion, our research shows that PUVA selectively inhibits recruitment of kinases Akt and Btk to activated plasma membranes. This suggests that unsaturations in the plasma membrane co-determine kinase-membrane binding. We furthermore propose that the variable susceptibility of ligand-receptor couples to PUVA contributes to complex rewiring of cell  (n ϭ 4). b, whole lysates of activated platelets were analyzed by Western blotting using antibodies to phosphorylated tyrosines (P-Tyr). Loading and specificity were controlled by reprobing with anti-pleckstrin (PLK) antibodies. Molecular marker bands in the left and right lanes represent molecular masses of 50 and 35 kDa. Paired control samples for resting platelets were incubated with the respective vehicle buffers. c, platelets were activated via the thrombin PAR4, and whole lysates were analyzed by Western blotting using antibodies to P-Akt. Loading was controlled by reprobing with total Akt (Akt). Molecular marker bands in the left lanes represent molecular masses of 75 and 50 kDa. To control for PI3K pathway specificity, paired experiments in the presence of 0.5 M Wm were performed. The decrease in mean protein band density of the ratio of phosphorylated kinase to total kinase (mean data and error bars representing S.D., n ϭ 3) in PUVA-treated platelets relative to untreated paired control is shown in the bar graph. Null change is emphasized by the horizontal dashed line. and tissue fate, explaining the complexity in output and in clinical success.
PUVA Treatment-PUVA treatment of platelets, liposomes, and peripheral blood mononuclear cells (PBMCs) was with the Intercept Blood System pathogen inactivation technology using amotosalen or 8-MOP as the active photosensitizing psoralen. Ultraviolet A light in the wavelength range 320 -400 nm was delivered to the samples in a closed and calibrated illumination device.
Platelet concentrates for transfusion were prepared from human voluntary whole blood donations using the standardized buffy coat method (45), which is by pooling of six buffy coats and resuspension in additive solution (SSPϩ, Macopharma, Tourcoing, France) as described (23). To generate paired samples, two platelet concentrates were pooled, mixed, and split again in two equivalent volumes. One volume remained untreated (control), and the other was PUVA-treated in accordance with the instructions of the Intercept Blood System manufacturer (46,47). For this, the platelet concentrate was transferred to an illumination bag while 17.5 ml of amotosalen (final 150 M) solution was added followed by UVA illumination at a dose of 3.9 J/cm 2 for 260 s on average. Following overnight incubation on an orbital shaker at room temperature with a compound adsorption device to remove unreacted amotosalen and its photoproducts, platelet concentrates were trans- ferred to a storage bag. Storage was under standard blood bank conditions. Experiments on platelets were performed 16 -24 h after UVA illumination using low volume samples taken through sterile welded connections under a laminar flow hood.
Signal Transduction Analysis-Platelet samples were washed in Tyrode's buffer (5 mM HEPES, pH 7.4, with 136.5 mM NaCl, 2.7 mM KCl, 11.9 mM NaHCO 3 , 0.4 mM NaH 2 PO 4 , 1 mM MgCl 2 , and 0.1% (w/v) D-glucose) as described (48). Washed platelets were rested for maximally 30 min at 37°C prior to activation with the indicated agonists and incubation times. Activation through PARs for measurement of phosphorylation downstream of PI3K activation was in conjunction with 10 M MeSADP. Activated platelets were lysed in ice-cold buffer (50 mM Tris, pH 8.0, with 100 mM NaCl, 0.5% (v/v) octylphenoxypolyethoxyethanol (Igepal-CA630), 0.4% (w/v) sodium dodecyl sulfate (SDS), 10 mM activated Na 3 VO 4 , and Halt protease mixture) and mixed. Separation of platelet membrane fractions was performed by ultracentrifugation as described by Badolia et al. (49). In brief, following platelet activation, the reaction was stopped by the addition of Halt protease mixture (final 2ϫ). Four freeze/thaw cycles in liquid nitrogen were performed next. Lysates were centrifuged at 1500 ϫ g for 10 min at 4°C to precipitate poorly lysed cell remnants. The supernatant was centrifuged in an ultracentrifuge at 100,000 ϫ g for 30 min at 4°C. The supernatant contained the cytosolic fraction and was kept stored at Ϫ20°C. The pellet, which contained the membrane fraction and cytoskeleton, was rinsed with 1 ml of icecold Tris-buffered saline (TBS) (pH 7.4; 25 mM Tris with 150 mM NaCl and 2 mM KCl) containing 1% (v/v) Triton X-100 and 2ϫ (final concentration) Halt protease mixture before it was resuspended in 100 l of the same buffer. These suspensions were again centrifuged at 15,000 ϫ g for 10 min at 4°C to precipitate the cytoskeleton. The supernatant contained the membrane fraction. Sample protein concentration was determined using a bicinchoninic acid kit (Thermo Scientific, Rockford, IL).
Platelet samples (membrane fraction or whole platelet lysates) were loaded on 4 -15% polyacrylamide Tris-glycine TGX gels in reducing Laemmli sample buffer. Western blotting was onto precut nitrocellulose membranes. Primary and secondary antibodies to blotted targets were prepared in TBS with 0.1% (v/v) Tween 20 and 5% (w/v) skimmed milk. Reprobing of membranes was by incubation for 30 min at 60°C in Western reprobing buffer (50 mM Tris, pH 7.0, 2.0% (w/v) SDS, and 100 mM ␤-mercaptoethanol). Bound antibody was detected with enhanced chemiluminescence and an automated imaging system (ChemiDoc MP) equipped with a charge-coupled device camera and analysis software (ImageLab v4.0.1) for densitometry (all reagents, hardware, and software from Bio-Rad).
Lipid Extraction and Lipid Sample Preparation-The solvents used for the lipid extraction and LC-MS separation were HPLC grade chloroform, ultra-LC-MS grade water, and ultra-LC-MS grade methanol (Biosolve, Valkenswaard, The Netherlands). HPLC grade ammonium formate, HCl, and HPLC grade methyl tert-butyl ether were from Sigma-Aldrich. Phosphoinositides from platelet lipid extracts were derivatized and analyzed as described (50) with minor modifications. In brief, 3 volumes of chloroform/methanol (1:2, v/v) were added to 1 volume of platelet suspension for extraction of neutral lipids. The samples were vortex-mixed followed by centrifugation at 13,000 ϫ g for 2 min. The supernatant was transferred to an LC-MS vial and used as such for analysis of phospholipids and phospholipids in complex with amotosalen. Internal standards PI(4,5)P 2 (17:0/20:4) and PI(3,4,5)P 3 (17:0/20:4) were added at 5 g/ml to the remaining pellet. A volume of 1365 l of methyl tert-butyl ether, methanol, and 2 M HCl (200:60:13, v/v/v) was used to resuspend the fraction followed by incubation for 15 min at room temperature with intermittent mixing. Next, 250 l of 0.1 M HCl was added, vortex-mixed, and centrifuged at 6,500 ϫ g for 2 min. The upper organic phase was transferred to a new tube, and 500 l of prederivatization wash solution (lower phase of methyl tert-butyl ether, methanol, and 0.01 M HCl (100:30:25, v/v/v)) was added. The samples were again centrifuged at 6,500 ϫ g for 2 min, and the upper phase was transferred for derivatization. To this, 50 l of 2 M trimethylsilyldiazomethane in hexane was added and mixed for 10 min at room temperature followed by addition of 6 l of glacial acetic acid (all Sigma-Aldrich) to stop the methylation reaction. Next, 500 were added to isolated PBMCs from healthy donors and subsequently illuminated with UVA light or not as indicated on top of each panel. The condition without psoralen contained vehicle solution and was used as a negative control. c, PBMCs were isolated from clinical buffy coat samples before (pre) and after (post) ECP treatment for GVHD. T-lymphocyte activation in the presence of 0.5 M Wm was used to control for specificity of the PI3K axis. a-c, T-lymphocytes were activated by receptor ligation in the presence of IL2 followed by Western blotting of whole cell lysates. Antibodies to P-Akt were used, and membranes were reprobed for total Akt (Akt) to control for loading. Molecular marker bands in the left lanes represent molecular masses of 75 and 50 kDa. d, the mean decrease in the P-Akt over Akt signal ratio in samples post-ECP relative to pre-ECP is depicted (mean with error bars representing S.D., n ϭ 3 different ECP treatments). Null change is emphasized by the horizontal dashed line.

Effects of Psoralen and UVA Light on Signal Transduction
l of postderivatization wash solution (upper phase of chloroform/methanol/water (120:60:45, v/v/v)) was added, vortexmixed, and centrifuged for 5 min at 1,500 ϫ g. The upper phase was transferred to a high recovery glass vial and dried down in a centrifugal vacuum evaporator (MiVac duoconcentrator, Genevac, Ipswich, UK) at 37°C. The dried sample was reconstituted in 50 l of chloroform/methanol (1:2, v/v) before injection.
Analysis of (Phospho)lipids-Lipid analysis was based on high resolution liquid chromatography coupled to high resolution quadruple time-of-flight (TOF) mass spectrometry (51,52). For lipidomics analyses, a 1290 Infinity LC system (Agilent Technologies, Waldbronn, Germany) was used. Lipid extracts were analyzed on an Acquity UPLC BEH Shield RP18 column (2.1 ϫ 100 mm; 1.7 m; Waters, Milford, MA) placed in a Polaratherm 9000 series oven (Selerity Technologies, Salt Lake City, UT) at 80°C. Elution was carried out with a multistep gradient of mobile phase A (20 mM ammonium formate, pH 5.0) and mobile phase B (methanol) starting from 50% B to 70% B in 5 min followed by a slow gradient of 70 -90% B in 30 min. Mobile phase B subsequently reached 100% in 0.1 min where it was maintained for an additional 5 min. The flow rate was 0.5 ml/min, and the injection volume was 10 l for both positive and negative electron spray ionization (ESI) measurements. The whole system was allowed to re-equilibrate under starting conditions for 15 min. For measurement of derivatized phosphoinositides, elution was optimized using the same mobile phases at a flow rate of 1.0 ml/min, starting from 50% B to 85% B in 5 min followed by a very slow gradient of 85-90% B in 15 min. At the end of the analysis, mobile phase B reached 100% in 2 min where it was maintained for an additional 3 min. High resolution accurate mass measurements were obtained on an Agilent 6550 quadruple TOF mass spectrometer equipped with a dual Jetstream ESI source. The instrument was operated in both positive and negative ESI modes. Needle voltage was optimized to Ϯ3.5 kV; the drying and sheath gas temperatures were set to 290 and 400°C, respectively; and the drying and sheath gas flow rates were set to 13 and 12 liters/min, respectively. Data were collected in centroid mode from m/z 100 to 3,200 at an acquisition rate of two spectra/second in the extended dynamic range mode (2 GHz), offering an in-spectrum dynamic range of 10 5 and a resolution of Ϯ20,000 full-width half-maximum in the lipid m/z range. MS/MS experiments were performed in targeted MS/MS mode. The quadrupole was operated at medium resolution, and the collision energy was fixed at either 20 or 35 eV. Study samples were analyzed in randomized order. Samples were kept at 4°C in the autosampler tray while waiting for injection.
Abundances (area under the curve (AUC)) of lipid species were extracted by the Find by Ion extraction algorithm in the MassHunter Qualitative Analysis 7.0 software package (Agilent Technologies) using predefined mass (10 ppm) and retention time extraction windows (15 s). The Agile integrator was used with an absolute peak height cutoff of 1,000 counts.
Molecular formulas, based on accurate mass, isotopic abundance, and spacing both in positive and negative ionization modes, were complemented with accurate mass database searching in an in-house-built database (populated with LIPID Metabolites And Pathways Strategy (MAPS) and Human Metabolome Database (HMDB) entries and theoretical lipid structures) and MS/MS measurement in both modes. Lipid fragmentation mechanisms and spectra in both positive and negative ionization modes were equal to those reported in the literature (52). Each lipid class displays characteristic fragment ions in positive and/or negative ESI mode. Other parameters such as lipid elution behavior and adduct formation interpretation further assisted in the identification. Regarding phosphoinositides, LC-MS cannot discriminate between PIP 2 isomers differing in the position of phosphates on the inositol ring. For trisphosphorylated PI(3,4,5)P 3 , the only known isomer is PI(3,4,5)P 3 . Throughout the paper, the nomenclature of the International Lipid Classification and Nomenclature Committee, i.e. the "comprehensive classification system for lipids," has been used (53)(54)(55).
Flow Cytometry-Membrane expression of activated integrin ␣ IIb ␤ 3 was analyzed with an acoustic focusing Attune flow cytometer (Life Technologies) using fluorescein-labeled PAC-1 monoclonal antibody (BD Biosciences). Platelets were incubated with 5% (v/v) PAC-1 and the indicated agonists or the respective vehicle buffer for 10 min at room temperature in HEPES-buffered saline, pH 7.4, with 1.0 mM MgSO 4 and diluted a thousand-fold immediately before readout according to previously published work (56). Signals of the isotype antibody controls were used to set threshold gates including 0.5% of 10,000 negative events. A total of 10,000 events staining positive for the platelet marker CD61 (allophycocyanin-labeled anti-CD61, Life Technologies) were measured.
Lipid Packing Analysis in Liposomes-Liposomes were prepared by mixing the (phospho)lipids from their stock solutions in chloroform and drying under nitrogen gas. Next, these were rehydrated in HKac buffer (50 mM HEPES, pH 7.2, with 120 mM potassium acetate, 1 mM MgCl 2 , and 1 mM dithiothreitol) followed by five freeze/thaw cycles in liquid nitrogen and a 37°C water bath. The samples were sonicated for 15 min at room temperature in a bath at 80 kHz. Liposome preparations were kept under nitrogen gas to avoid oxidation and were prepared fresh for every experiment. UVA treatment was similar to that of platelet concentrates.
The order of lipid phases was analyzed using the environment-sensitive NR12S probe (25,26). Probe and lipid concentrations were optimized first in separate validation experiments. First, liposomes of PC(16:0/18:1) with or without 35 mol % cholesterol containing 150 M amotosalen were UVA-illuminated or not. Next, 50 M of liposomes by lipid content were mixed with 0.1 M NR12S and allowed to equilibrate for 7 min in the dark at room temperature. Fluorescence was measured in a fluorescence spectrophotometer (Tecan, Männedorf, Switzerland) at 580-and 620-nm emission wavelengths using 510-nm excitation filters. The emission peak of NR12S shifts blue when partitioned to lipid phases of increased order. The fluorescence intensities at each emission wavelength were used to calculate a GP value (57). This is a relative measure between ϩ1 (most ordered) and Ϫ1 (least ordered).
Binding experiments of the peptidic lipid-packing sensors ␣-synuclein and ALPS were based on the work by Vanni et al. (29) and Pranke et al. (30). First, the peptides were covalently labeled with NBD to their N termini. ␣-Synuclein at 1 mg/ml was labeled by incubation with 0.5 mM NBD chloride in citrate buffer, pH 7.0, with 1 mM EDTA for 24 h in the dark at room temperature. The neutral pH renders the reaction specific for the peptide N terminus, avoiding labeling auxiliary amines in the amino acid sequence (58). The reaction mixture was then re-equilibrated to HEPES-buffered saline, pH 7.4, with 5 mM KCl and 1 mM MgSO 4 by desalting column buffer exchange (Zeba Spin columns, Thermo Scientific). The ALPS peptide contained just one cysteine (at the N terminus) and was kept in a reducing environment (10-fold molar excess of tris(2carboxyethyl)phosphine). Labeling was to that free N-terminal thiol using IANBD at a molar ratio of IANBD of 10:1 in HK buffer (50 mM HEPES, pH 7.2, with 120 mM potassium acetate) for 2 h at room temperature in the dark. The reaction mixture was dialyzed (membrane molecular weight cutoff, 2,000) to 4 liters of HK buffer during 24 h at 4°C. The degree of labeling for both peptides was determined by optical absorbance at wavelengths for peptides and bound and free label.
T-cell Activation-PBMCs were prepared by Ficoll gradient centrifugation of buffy coats from healthy donors. Isolated PBMC preparations were washed three times in sterile phosphate-buffered saline, pH 7.4, by soft centrifugation to remove contaminating platelets. Finally, the cells were resuspended in phenol red-free RPMI 1640 medium with 2 mM L-glutamine, 10 mM HEPES, and 10% (v/v) AB human serum and rested for 2 h at room temperature. The rested cells were resuspended to 40,000 leukocytes/l in Tyrode's buffer and mixed with 8-MOP or amotosalen in a final amount of 60 ϫ 10 6 leukocytes. Following PUVA, T-cells were activated in Tyrode's buffer with 2.5 ϫ 10 6 CD3/CD28 human T-activator Dynabeads/10 6 leukocytes in the presence of 10 ng/ml IL2 for 30 min at 37°C on a rotator. Cells were lysed with ice-cold lysis buffer and centrifuged at 20,000 ϫ g for 3 min. Supernatant was analyzed by SDS-PAGE Western blotting in reducing sample buffer. Cell viability was unaltered by PUVA during the course of T-cell receptor activation as monitored by Sytox Green (Life Technologies) staining in flow cytometry.
Extracorporeal Photopheresis Samples-A patient enrolled in a clinical trial for ECP at the Hospital Network Antwerp (Belgium) for the treatment of chronic GVHD gave informed consent for sampling of cells from the buffy coat suspension before and after ECP. The study was approved by the hospital's ethics committee (approval number 4591). The ECP procedure involved a leukapheresis (Spectra Optia, Terumo BCT, Lakewood, CO) yielding a buffy coat cell suspension that was diluted with saline in an illumination bag to a final 300 g by mass. 8-MOP was added to a final concentration of 0.9 M followed by illumination in a Macogenic G2 UVA (peak wavelength, 365 nm) illumination device (Macopharma), which delivered 2 or 2.5 J/cm 2 depending on the suspension's hematocrit of Ͻ2% or Ն2%, respectively, for 10 -15 min. Before and after illumination, 2-ml samples were taken aseptically from the bag for this study, and PBMCs were prepared followed by activation of T-cells as described above. A total of three independent ECP procedures were performed over a period of 2 months. Three technical replicates per sample series were performed to control for variation in SDS-PAGE Western blotting and densitometry.
Author Contributions-B. V. A. designed research, executed experiments, interpreted data, and assisted in writing the paper. R. D. executed experiments and interpreted data. R. t'K. and K. S. were responsible for mass spectrometry experiments and their data analysis and figure composition. P. Z. provided patient samples and essential patient data and interpreted data. P. V. provided essential tools and facilitated research. V. C. designed research, interpreted data, assisted in writing the paper, and managed the research team. H. B. F. designed research, interpreted data, wrote the paper, and managed research. All authors critically reviewed the manuscript and amended the text where deemed appropriate.