Cellular and Physiological Roles for Phospholipase D1 in Cancer*

Phospholipase D enzymes have long been proposed to play multiple cell biological roles in cancer. With the generation of phospholipase D1 (PLD1)-deficient mice and the development of small molecule PLD-specific inhibitors, in vivo roles for PLD1 in cancer are now being defined, both in the tumor cells and in the tumor environment. We review here tools now used to explore in vivo roles for PLD1 in cancer and summarize recent findings regarding functions in angiogenesis and metastasis.

Phospholipase D enzymes have long been proposed to play multiple cell biological roles in cancer. With the generation of phospholipase D1 (PLD1)-deficient mice and the development of small molecule PLD-specific inhibitors, in vivo roles for PLD1 in cancer are now being defined, both in the tumor cells and in the tumor environment. We review here tools now used to explore in vivo roles for PLD1 in cancer and summarize recent findings regarding functions in angiogenesis and metastasis.
The phospholipase D (PLD) 2 enzyme superfamily is defined by a conserved catalytic site and a functional commonality of transphosphatidylation activity at phosphodiester bonds found in a wide range of substrates. PLDs are present in bacteria and viruses as well as in yeast, plants, and animals. Classical PLD enzymes hydrolyze the abundant membrane lipid phosphatidylcholine to yield the second messenger phosphatidic acid (PA) and choline (1). Most vertebrates have two classic isoforms, PLD1 (2) and PLD2 (3), which are 50% identical in protein sequence in mammals and arose from a gene duplication event early in the vertebrate lineage. PLD2 is particularly intriguing in that it has a compact gene structure (4,5) and resides within a 50-kb intron of another gene. Other animals, such as Drosophila melanogaster and Caenorhabditis elegans, have single PLD genes that are intermediate between mammalian PLD1 and PLD2 (1,6,7). PLD1 and PLD2 encode an identical series of protein regions, including Pleckstrin homology (PH) and Phox (PX) domains and a phosphatidylinositol 4,5bisphospate-interacting motif that regulate association with specific subcellular membranes during signaling events, in addition to the pair of split catalytic domains (1). PLD function has been studied using biochemical, cell biological, and now physiological approaches. Potential roles for PLD in general or for PLD1 specifically have been reported in numerous physiological settings including ones relevant to cancer such as survival signaling (8 -11), control of cell polarity (12,13), Ras activation (14 -19), and cell migration (13, 20 -26). Moreover, a PLD1 single nucleotide polymorphism (SNP) associates with the risk of non-small cell lung cancer and increased PLD1 expression and/or PLD activity have been reported in multiple types of cancer (27)(28)(29)(30)(31)(32)(33), although the mechanisms underlying this increase and the specific advantage this confers to the tumor cells are not known. As will be discussed as well, roles for PLD1 in the tumor microenvironment have also been uncovered, specifically in relationship to platelet activation (34 -36) and angiogenesis (22,26,37).
In this review, we discuss physiological roles, in particular in the context of cancer, that have been identified for PLD1 using PLD lipase activity inhibitors and genetically modified animal models.

Tools Used for Study of PLD Function
Cell biological roles for PLD enzymes have long been explored using a variety of types of inhibitors, the most popular of which has involved primary alcohols. Although classic PLDs use water as a nucleophile in the transphosphatidylation reaction to hydrolyze their target lipids and thus generate PA, they can also use short-chain primary alcohols to generate phosphatidyl (Ptd) alcohols (38,39). Ptd alcohols can be generated only through the action of PLD, whereas PA can be generated through three other enzymatic pathways (de novo synthesis, acetylation of lysoPA, and phosphorylation of diacylglycerol (40 -42)), and the Ptd alcohols are metabolized slowly, making them excellent reporters of PLD activity (38). Because primary alcohols such as ethanol and 1-butanol possess more than 1000-fold greater nucleophilicity than water (39), they are preferentially used by PLD when present and thus divert the PLD from generating PA. Because the Ptd alcohols were thought to be biologically inert, ethanol and 1-butanol have been used to inhibit PLD generation of PA for more than two decades. However, concerns were raised in 2002 that the concentrations of ethanol and 1-butanol commonly used for inhibition were not sufficient to block PA production and that the amounts of alcohol required to fully block PA production had many other effects on cellular lipids that complicated interpretation of the ensuing findings (43). In addition, Ptd ethanol was found to trigger cell biological responses (44), and a number of groups reported that the primary alcohols block receptor activation at the receptor level, including the insulin receptor (45,46), creating additional challenges for drawing definitive conclusions through use of alcohols to suppress PA formation. We then reported that whereas 1-butanol blocked glucose-stimulated insulin release but did not affect plasma membrane levels of PA, a new and highly effective and specific PLD inhibitor, 5-fluoro-2-indolyl des-chlorohalopemide (FIPI), blocked increases in the level of PA but did not block glucose-stimulated insulin release, indicating that the 1-butanol was blocking insulin release through a nonspecific mechanism (47). Finally, with the availability of primary cells prepared from mice lacking both PLD1 and PLD2, it has been conclusively shown that although primary alcohol blocks neutrophil degranulation in response to bacterial peptide stimulation, ablation of PLD1 and PLD2 has no effect on the process, revealing a nonspecific inhibition mediated by the primary alcohol in this event (48). Several indirect inhibitors of PLD activity have been reported including neomycin (49) and ceramide (50 -52), and some natural products also possess PLD inhibitory effects albeit through indirect mechanisms (53). However, none of these specifically inhibit PLD. Taken together, there is a need to reassess prior studies that made use of primary alcohol or indirect PLD inhibitors using modern, PLD-specific small molecule inhibitors, RNAi, or PLD-ablated cells to confirm that the phenotypes previously reported are reproduced with these other methods.
More recently and usefully, a large scale screen performed by Steed and colleagues (54) at Novartis for small molecule PLD inhibitors resulted in the identification of FIPI. Initially reported as a PLD2 inhibitor, we and others subsequently reported that FIPI is a potent inhibitor of both PLD1 and PLD2 (47,55), and Brown and colleagues (56) have pursued combinatorial chemistry approaches to generate subtype-selective PLD inhibitors. FIPI and the isoform-selective inhibitors have become widely used to explore roles for PLD in cells (11,13,21,34,(57)(58)(59)(60)(61)(62)(63)(64)(65). To date, the inhibitors have not been associated with off-target or nonspecific effects, excepting one of the PLD1-selective compounds (66). FIPI has also been employed in vivo to study cancer growth and metastasis and platelet activation (22,35,67), which was achievable given its 5.5 h half-life and acceptable bioavailability and solubility (54). The subtypespecific PLD inhibitors currently available are less amenable for in vivo studies with shorter half-lives (45 min to 1 h) and efficient first-pass clearance (56), but with continued development could become invaluable tools for study in animal models and as potential therapeutics.
As useful and popular as these inhibitors are becoming for the study of PLD function, it should be cautioned that PLD inhibition may not always yield the same outcome as PLD gene ablation or RNAi because both PLD1 and PLD2 have been reported to mediate non-enzymatic functions in specific settings (23,68). More generally, PLD1 has been reported to interact with 28 other proteins and PLD2 been reported to interact with 29, only 11 of which are shared in common (see www.ncbi.nlm.nih.gov/gene (accession numbers 5337 and 5338) for details), and in most instances these interactions should presumably not be dependent on PLD activity. Thus, there may be many changes in signaling pathways that occur when the proteins are absent that are not observed when the enzymatic function is simply suppressed.

Roles for PLD1 in Cancer Signaling from Studies of Cell Lines in Culture
PLD activity has been proposed to stimulate pro-survival tumor cell pathways (69) and to facilitate a switch in energy metabolism from normal oxidative phosphorylation glycolysis to the aerobic one characteristic of cancer (the Warburg effect) (70). These roles will be discussed in separate minireviews within this series. Here we will mainly focus on other roles proposed for PLD1 in cancer based on study of cell lines.
Combatting metastasis remains the key challenge in managing cancer (71), and PLD1 has been reported to affect a number of processes relevant to it. One of the earliest steps in many types of tumors involves persistent activation of the small GTPase Ras. Although regulation of Ras has been more broadly connected to PLD2 (18,19), PLD1 has also been shown to promote Ras and ERK activation and transformation in epithelial cells (14) via interaction with PEA-15 (72). Another early step involves loss of polarization for epithelial cells. PLD1 has been linked to epithelial cell polarization downstream of the tumor suppressor LKB1 via effects on the small GTPase Rap2A and the actin-binding protein Ezrin (12) and for motile cells such as neutrophils via control of recruitment of Rac1 to the leading edge of the cell and subsequent actin cytoskeletal reorganization (13). An important subsequent step involves increasing invasive capacity to facilitate movement into the vascular or lymphatic circulation. PLD1 has been linked by many groups to increased release of extracellular proteases such as MMP-2 and MMP-9 that facilitate invasion for multiple types of tumors (73)(74)(75)(76). Increased cell motility is similarly important in most types of cancer. PLD1 has been linked to cell migration in multiple settings for non-transformed cells (21)(22)(23)(24)(25)(26) and may have similar roles in some types of cancer cells. The most interesting finding related to PLD1 and motility of cancer cells, however, addresses a setting in which loss of PLD1 activity results in decreased motility that enhances the oncogenic process. Specifically, PLD1 facilitates endosomal recycling of the small GTPase Rap1 to the plasma membrane, leading to activation of LFA-1, an integrin that regulates lymphocyte entry into and exit from lymph nodes (77). A subset of chronic lymphocytic leukemia cases is characterized by defective activation of PLD1, resulting in loss of Rap1 translocation and blunted LFA-1 activation, causing the malignant cells to become arrested in lymph nodes where they are exposed to survival/proliferation signals in the tissue microenvironment, which correlates with shorter survival of the patients (20,78). Taken together, roles for PLD1 in tumor cells have been described that are generally but not always pro-oncogenic, suggesting that a PLD1 therapeutic would be potentially useful in many settings but might be contraindicated in others such as chronic lymphocytic leukemia.

Danio rerio (Zebrafish)
Two studies using D. rerio as a model organism are interesting in connection to PLD1 and oncogenesis. First, Zeng et al. (37) found that use of antisense morpholino oligonucleotides to inhibit PLD1 expression led to impaired intersegmental vessel development. The phenotype could be rescued in part by transplantation of non-targeted mesenchymal notochord tissue, suggesting that the defect arose from loss of a notochord morphogen that stimulated angiogenesis. Although intriguing, mice lacking PLD1 do not appear to exhibit this developmental phenotype, and the phenotype may not relate well to the role of PLD1 in neoangiogenesis in adult mice, which directly connects to the vascular endothelial cells (22,26), as will be discussed subsequently.
Another intriguing story was reported by Liu et al. (79) who discovered that PLD1 is up-regulated and hyperactivated in the fat-free (ffr) mutant, accompanied by decreased levels of phos-phatidylcholine, and showed that this phenotype could be ameliorated through suppression of PLD activity. This study suggests a potential role for PLD1 in cell metabolism that has not yet been explored in cancer pathways, although PLD1 has been linked to lipid droplet generation and storage in non-transformed cells (80).

Mouse Models
PLD1 and PLD2 are expressed in most tissues but vary in relative and absolute levels (81,82). The isoforms are both activated by G-protein-coupled and receptor tyrosine kinase pathways and use similar mechanisms to target subcellular membrane sites (83), but subtle differences in the protein sequences of the conserved domains and the only partially overlapping regulatory mechanisms and sets of binding-partners lead PLD1 and PLD2 to localize in general to different intracellular sites (84) and to exhibit isoform-specific functions (85), although redundancy to differing degrees has also been reported as discussed below.
Mice lacking PLD1 are viable and grossly normal (34), but have an interesting set of phenotypes. The first study on PLD1 Ϫ/Ϫ mice described defective platelet function, principally involving blunted activation of integrin ␣ IIb ␤ 3 (34). Activated ␣ IIb ␤ 3 binds fibrinogen, a key step in converting monolayers of platelets on damaged vessel walls into full-scale thrombi. Exploring the pathophysiological consequences of this defect led to the demonstration that PLD1 Ϫ/Ϫ mice were protected from major thrombotic events such as pulmonary emboli, strokes, and aortic occlusion after vessel wall damage, without affecting tail bleeding times. The phenotype could be reproduced using in vivo administration of the PLD1/2 small molecule inhibitor FIPI (35), both providing a potential therapeutic opportunity for PLD inhibitors and demonstrating that the defect ensued from loss of PA generation rather than physical loss of PLD1. Both PLD1 and PLD2 are expressed in platelets and activated by thrombin (86), albeit PLD1 is responsible for generation of PA significantly earlier than PLD2 (34). Ablation of PLD2 has no effect on platelet activation; however, the absence of both PLD1 and PLD2 strengthened the PLD1 phenotype, revealing an additional parameter of ␣-granule release (36). Depending on the strength of the inhibition desired, a PLD1-selective or a dual PLD1/2 inhibitor might be most useful for therapeutic purposes. Although not appreciated at the time, the PLD1 Ϫ/Ϫ platelet phenotype is relevant to metastasis as discussed below.
Our initial studies on inhibition of PLD activity in several types of tumor cells were unrewarding in that no changes in proliferation, survival, or migration on standard substrates were discernable. A very interesting finding was obtained, however, through implantation of wild-type melanoma and lung tumors into mice lacking PLD1. In brief, the xenografts grew slowly and exhibited virtually no tumor angiogenesis (22). This was traced to blunted signaling through the VEGF receptor, leading to minimal neoangiogenesis in response to VEGF signaling as the tumor grew hypoxic. Part of the aberrant response appears to involve interaction of the vascular endothelial cells with extracellular matrix because aortic endothelial cells lacking PLD1 had difficulty in attempting to generate microvessels in response to serum stimulation. In addition, PLD1 Ϫ/Ϫ lung endothelial cells proliferated equally well and appeared identical to wild-type lung endothelial cells when cultured in standard tissue culture dishes, but failed to reorganize into vascular cords when replated onto extracellular matrix substrates and adhered poorly to fibronectin, vitronectin, and collagen, signifying poor activation of integrins ␣ 5 ␤ 1 , ␣ v ␤ 3 , and ␣ 2 ␤ 1 , the extracellular receptors that interact with them. These findings were reminiscent of the blunted activation of ␣ IIb ␤ 3 for PLD1 Ϫ/Ϫ platelets (34), suggesting a common underlying mechanism. PLD1 has also been linked to hypoxia-induced, VEGF-stimulated pathological retinal angiogenesis via siRNA studies, although the signaling pathways involved may differ in detail (26).
Metastasis to the lung after introduction of tumors cell intravenously was also reduced in PLD1 Ϫ/Ϫ mice, and this finding was discovered to reflect aberrant interaction of the tumor cells with the PLD1 Ϫ/Ϫ platelets (22). Although part of the platelet deficiency could involve poor ␣ IIb ␤ 3 activation because this integrin is known to be important for tumor-platelet interactions and the efficiency of metastatic seeding (87), a second role for PLD1 was uncovered as well, which we hypothesize involves transmission of a signal from the platelets to the tumor cells to promote epithelial-mesenchymal transformation (EMT). Although EMT is classically thought to underlie the conversion from epithelial cells to mesenchymal cells that promotes emigration from the primary tumor, it has recently been shown that sustained contact between platelets and circulating tumor cells in the vasculature reinitiates EMT pathways in the tumor cells and facilitates their successful colonization at sites of metastasis (88). The platelet-promoted EMT process involves both TGF␤ release that stimulates a SMAD-dependent pathway and a physical interaction that activates NF-B signaling. Given the observations that PLD-deficient platelets have a secretion defect and that PLD1 Ϫ/Ϫ platelets have decreased physical affinity for tumor cells (22), PLD1 could be involved in either or both of these stimulatory mechanisms (Fig. 1). Finally, both the tumor angiogenesis and metastasis could be suppressed equally well in wild-type mice through administration of the PLD inhibitor FIPI, indicating the enzymatic role of PLD1 in this setting and the potential utility of a PLD1 inhibitor for multiple types of cancer. PLD1 has been linked to some additional interesting pathways. Neovascularization of tumors often results in morphologically and functionally abnormal blood vessels that are leaky, i.e. have increased permeability, and this is thought to prevent cytotoxic drugs from fully infusing into the core of malignant tumors. Normalization of tumor blood vessels has been proposed as a goal to improve nanomedicine delivery (89). PLD1 Ϫ/Ϫ but not PLD2 Ϫ/Ϫ mice exhibit decreased Evan's blue ear leakage following irritation with mineral oil (21), indicating that PLD1 Ϫ/Ϫ mice have reduced vascular permeability. VEGF receptor 2 (VEGFR2), which is also called vascular permeability factor, plays an important role in regulating vascular permeability (90). We have shown reduced ERK1/2, p38-MAPK, and Akt signaling downstream of VEGFR2 stimulation in PLD1 Ϫ/Ϫ primary endothelial cells (22). Taken together, these results suggest that VEGFR2 signaling impairment in the absence of PLD1, which leads to reduced vascular permeability, could potentially normalize tumor blood vessels and improve chemotherapeutic delivery to pre-established and vascularized tumors.
Autophagy is a widely studied topic in oncogenesis. The autophagic process recycles cellular organelles, facilitating cell survival in settings of nutrient starvation, radiotherapy, and certain cytotoxic drugs (91). PLD1 ablation decreases starvation-induced expansion of LC3-positive compartments, a marker for autophagy (92), although whether this is pro-or anti-oncogenic depends on the type of tumor and setting (93).
Another complicated topic involves the innate immune response to tumors. Tumor-associated macrophages can be either supportive or suppressive for tumor growth. PLD1 Ϫ/Ϫ macrophages exhibit several types of defects including blunted phagocytosis, spreading, and migration (21). Because local macrophages tend to be tumor-promoting and recruited ones tend to be tumor-suppressive, such migration defects in PLD1 Ϫ/Ϫ mice may favor a shift in the balance toward tumor promotion.
Aside from tumorigenesis, roles for PLD1 have been assessed in other mouse models that may provide insights for cancer research. TGF␤ is a pathogenic mediator of acute lung injury, of which the major consequence is human acute respiratory distress syndrome. TGF␤ increases endothelial and epithelial permeability and facilitates alveolar flooding, the primary symptom of acute respiratory distress syndrome (94). The epithelial sodium channel (ENaC) promotes clearance of water from the alveolar (95). Explorations in a mouse acute lung injury model have shown that TGF␤ promotes ␤ENaC internalization through a PLD-driven pathway, i.e. siRNA-mediated knockdown of PLD1 prevents ␤ENaC internalization in human lung epithelial A549 cells (96). PLD1 knockdown also blocks TGF␤-induced reactive oxygen species production, which depending on the setting could be either pro-oncogenic or antioncogenic. Ribosomal S6 kinase 2 (RSK2) functions as a prooncogenic manner in many pathways (97) and also regulates PLD1 via phosphorylation in the setting of neurite extension (98), but whether this also plays a role in cancer progression is unknown.
Finally, to emphasize some points made earlier, in some settings PLD1 and PLD2 function in relatively distinct manners with virtually no overlap in function. For example, PLD1 is required for ␣ IIb ␤ 3 activation on platelets, whereas PLD2 is not. In contrast, Fukui and colleagues (64) have shown that PDGFstimulated dorsal ruffling on mouse embryo fibroblasts proceeds with normal kinetics in the presence of either PLD1 or PLD2, but not when both isoforms are absent, presenting a cell biological process in which there is substantial overlap in function for the isoforms. Thus, roles in cancer-connected processes will have to be examined on a case-by-case basis to determine whether it would be most advantageous to inhibit one or both of the PLD isoforms for optimal therapeutic benefit.

Perspective
Over the past two decades, hundreds of studies have explored roles for PLD isoforms using cell culture approaches. Many of these need to be reassessed because they employed Not shown: evidence has been published to suggest that PLD or PLD1 specifically may affect the steps leading up to this stage through 1) stimulating Ras activation to increase proliferation, 2) generating survival signals that suppress apoptosis in settings of nutrient deprivation, hypoxia, and anoikis driven by detachment from the basal lamina extracellular matrix, and 3) regulating cell polarity to enable the detachment (8 -19). Shown: VEGF released by tumor cells stimulates neoangiogenic vascularization of the tumor. Middle left, PLD1 may also increase tumor cell invasive capacity via MMP release and facilitate migration and thus entry into circulation. Once in circulation, the tumor cells aggregate through interaction with platelets and fibrinogen, shielding the tumor cells from the immune system, physically protecting them from flow stress, promoting EMT, and (lower left) assisting them in seeding at distal metastatic sites. Right, in the absence of PLD1, vascular endothelial cells respond poorly to the VEGF released by the primary tumor, strongly decreasing neoangiogenesis. For those tumor cells that enter into circulation, decreased interaction with platelets leads to reduced efficiency of metastatic seeding, via any or multiple of the mechanisms listed above. primary alcohol as a PLD activity inhibitor. However, even for the more recent studies that employed RNAi or highly specific small molecule inhibitors, cell culture findings are not always predictive of organismal responses. The limited number of animal studies that have been performed, however, suggest that both PLD1 and PLD2 play roles in tumor progression and that there will be many interesting stories to come in the context of primary cancer cell migration, inflammatory responses, and metabolic reprogramming. New techniques such as advanced intravital imaging, which can now be employed down to the subcellular level to examine tumor-stroma interactions (99), genetically encoded PA sensors (16,62), or in vivo cell tracking using MRI and single-photon emission computed tomography (100) will advance our understanding of the role of PLD1 in the early steps of tumor development and metastasis. Therapy resistance is one of the major challenges in treating cancer. Both intrinsic genetic changes in cancer cells and changes in the extrinsic tumor microenvironment can contribute to the development of resistance (101). PLD may play any of several roles in this process. In addition, the role of PLD in CTLA4 processing (102) may offer the potential to predict immune therapy outcomes. Inhibition of PLD activity as a means of decreasing platelet activation might also provide benefit in management of the hypercoagulation that is a common complication of cancer and cancer therapeutics (22,103) and increases the risk of stroke (104).

Conclusion
Evidence for roles of PLD1 in cancer progression are becoming well established, setting the stage for crossing PLD1 Ϫ/Ϫ mice into genetic mouse models of cancer and conducting pharmacological studies on genetic models, potentially in combination with other chemotherapeutic agents. Mice lacking both PLD1 and PLD2 are viable, fertile, and overtly normal, suggesting that isoform-selective or dual PLD1/2 inhibitors should be well tolerated even for extended periods of time. Nonetheless, given the partial redundancy between PLD1 and PLD2 function and the several other means of generating PA in the cell as a form of induced compensation, critical roles for the PLD isoforms will need to be established empirically in each setting.