Aquaporin-1 facilitates angiogenic invasion in the pathological neovasculature that accompanies cirrhosis


  • Robert C. Huebert,

    1. Gastroenterology Research Unit, Mayo Clinic and Foundation, Rochester, MN
    2. Department of Internal Medicine, Mayo Clinic and Foundation, Rochester, MN
    Search for more papers by this author
  • Meher M. Vasdev,

    1. Gastroenterology Research Unit, Mayo Clinic and Foundation, Rochester, MN
    Search for more papers by this author
  • Uday Shergill,

    1. Gastroenterology Research Unit, Mayo Clinic and Foundation, Rochester, MN
    Search for more papers by this author
  • Amitava Das,

    1. Gastroenterology Research Unit, Mayo Clinic and Foundation, Rochester, MN
    Search for more papers by this author
  • Bing Q. Huang,

    1. Center for Basic Research in Digestive Diseases, Mayo Clinic and Foundation, Rochester, MN
    Search for more papers by this author
  • Michael R. Charlton,

    1. Center for Basic Research in Digestive Diseases, Mayo Clinic and Foundation, Rochester, MN
    2. Department of Internal Medicine, Mayo Clinic and Foundation, Rochester, MN
    Search for more papers by this author
  • Nicholas F. LaRusso,

    1. Center for Basic Research in Digestive Diseases, Mayo Clinic and Foundation, Rochester, MN
    2. Department of Internal Medicine, Mayo Clinic and Foundation, Rochester, MN
    3. Center for Cell Signaling, Mayo Clinic and Foundation, Rochester, MN
    Search for more papers by this author
  • Vijay H. Shah

    Corresponding author
    1. Gastroenterology Research Unit, Mayo Clinic and Foundation, Rochester, MN
    2. Department of Internal Medicine, Mayo Clinic and Foundation, Rochester, MN
    3. Center for Cell Signaling, Mayo Clinic and Foundation, Rochester, MN
    • Gastroenterology Research Unit, Mayo Clinic and Foundation, 200 First Street SW, Rochester, MN 55905
    Search for more papers by this author
    • fax: 507-255-6318

  • Potential conflict of interest: Nothing to report.


Increasing evidence suggests that hepatic fibrosis and pathological angiogenesis are interdependent processes that occur in parallel. Endothelial cell invasion is requisite for angiogenesis, and thus studies of the mechanisms governing liver endothelial cell (LEC) invasion during cirrhosis are of great importance. Emerging research implicates amoeboid-type motility and membrane blebbing as features that may facilitate invasion through matrix-rich microenvironments. Aquaporins (AQPs) are integral membrane water channels, recognized for their importance in epithelial secretion and absorption. However, recent studies also suggest links between water transport and cell motility or invasion. Therefore, the purpose of this study was to test the hypothesis that AQP-1 is involved in amoeboid motility and angiogenic invasion during cirrhosis. AQP-1 expression and localization was examined in normal and cirrhotic liver tissues derived from human and mouse. AQP-1 levels were modulated in LEC using retroviral overexpression or small interfering RNA (siRNA) knockdown and functional effects on invasion, membrane blebbing dynamics, and osmotic water permeability were assayed. Results demonstrate that AQP-1 is up-regulated in the small, angiogenic, neovasculature within the fibrotic septa of cirrhotic liver. AQP-1 overexpression promotes fibroblast growth factor (FGF)-induced dynamic membrane blebbing in LEC, which is sufficient to augment invasion through extracellular matrix. Additionally, AQP-1 localizes to plasma membrane blebs, where it increases osmotic water permeability and locally facilitates the rapid, trans-membrane flux of water. Conclusion: AQP-1 enhances osmotic water permeability and FGF-induced dynamic membrane blebbing in LEC and thereby drives invasion and pathological angiogenesis during cirrhosis. HEPATOLOGY 2010

Cirrhosis and its complications are associated with significant morbidity, mortality, and healthcare expenditures.1 Therefore, there is a need for expanded understanding of the mechanisms driving fibrosis. An increasing body of evidence suggests that hepatic fibrosis and pathological angiogenesis are interdependent processes that occur in tandem.2 Indeed, the fibrotic septa surrounding cirrhotic nodules contain a dense neovasculature.3, 4 The chronic inflammatory milieu of cirrhosis is thought to stimulate the expression and release of multiple angiogenic molecules such as fibroblast growth factor (FGF), vascular endothelial growth factor (VEGF), platelet-derived growth factor, and angiopoietins from stromal cells, and epithelium.2, 5 In turn, the neovasculature undergoes complex interactions with the cirrhotic microenvironment,6 provides nourishment to areas of active scarring and tissue remodeling, and serves as a source of inflammatory cytokines and chemokines, thereby driving chronic inflammation and disease progression.7 Further support for angiogenesis as a driver of liver fibrosis comes from studies in which anti-angiogenic therapy reduced fibrosis and portal pressure in cirrhotic animals.3 However, better understanding of the basic underlying mechanisms is required because not all angiogenic targets may be useful,8 and thus therapeutic approaches need to be refined toward biological targets most likely to have therapeutic benefits.9, 10

Although the role of VEGF has been widely studied in liver angiogenesis, FGF is another molecule known to be involved in fibrogenesis2, 11, 12 and liver angiogenesis,13 and it has prominent effects on endothelial cell motility and vascular integrity.14 The cellular source of increased FGF levels in fibrosis is not entirely clear, but it is presumed to be derived from activated hepatic stellate cells.11

Traditionally, the study of cell motility and invasion has been heavily focused on traditional, mesenchymal migration mechanisms based on actin dynamics in two-dimensional culture. However, recent advances in three-dimensional culture methods and in vivo imaging have revealed that many cells behave quite differently in extracellular matrix (ECM) in vivo, including mode-switching from mesenchymal motility to an invasive, amoeboid phenotype involving dynamic membrane blebbing.15, 16

Aquaporins (AQPs) are integral membrane water channels that allow for rapid, bidirectional flux of water in response to local osmotic gradients.17 Whereas the expression and function of AQPs have been extensively studied in secretion and absorption across epithelial barriers,18, 19 these proteins are also expressed in endothelia, where their role is less clearly understood. Endothelial motility and invasion are well recognized as prerequisites for angiogenesis,20 and we noted several features of AQPs suggesting that they may contribute to amoeboid invasion in liver angiogenesis and cirrhosis. First, recent studies show that AQPs may influence cell motility and angiogenesis in general.21, 22 Second, AQPs localize to areas of focal plasma membrane shape change and protrusions.23 Third, AQPs can directly interact with signaling molecules relevant to cell motility in addition to numerous solute/ion transporters.23, 24 Lastly, recent genetic studies in patients with chronic hepatitis C have identified an AQP single-nucleotide polymorphism as part of a genetic signature identifying patients at risk for progression to cirrhosis.25 However, direct mechanistic evidence for AQP regulation of liver endothelial cell (LEC) invasion in the context of cirrhosis is lacking.

Therefore, we sought to test the hypothesis that AQP-1 is involved in FGF-induced pathological angiogenesis during cirrhosis and to gain relevant mechanistic insights into this process. The experimental results from the current study provide several novel pieces of information regarding the mechanisms controlling LEC invasion through ECM. The work also begins to develop a foundation for plausible anti-angiogenic therapies targeting water channels in the treatment of cirrhosis and portal hypertension. Numerous AQP inhibitors in development make this direction ideal for future human translation.26


AQP, aquaporin; CCL4, carbon tetrachloride; ECM, extracellular matrix; FGF, fibroblast growth factor; HHSEC, human hepatic sinusoidal endothelial cells; IF, immunofluorescence; IHC, immunohistochemistry; LEC, liver endothelial cell; NAFLD, nonalcoholic fatty liver disease; RT-PCR, reverse transcription polymerase chain reaction; SE, standard error; SEM, scanning electron microscopy; siRNA, small interfering RNA; TSEC, transformed sinusoidal endothelial cells; VEGF, vascular endothelial growth factor; vWF, von Willebrand factor.

Materials and Methods

Additional experimental details and references can be found in the Supporting materials.

Human Tissue Specimens.

Tissue from normal and cirrhotic human liver were obtained from biopsy specimens or surgical waste with informed consent and Institutional Review Board approval. No donor organs were obtained from executed prisoners or other institutionalized persons.

Animal Models.

Cirrhosis was induced in C57BL/6J mice (Harlan) with chronic carbon tetrachloride injection (CCl4), using a well-established protocol with appropriate Institutional Animal Care and Use Committee approval.27, 28 Animals received humane care according to the criteria outlined in the Guide for the Care and Use of Laboratory Animals by the National Academy of Sciences.

Isolation of Mouse LEC.

LECs were isolated from whole mouse liver by mechanical disruption, enzymatic digestion, and immunomagnetic bead separation, as previously described, with modifications.29-31

Cell Culture.

Freshly isolated mouse LECs, human hepatic sinusoidal endothelial cells (HHSEC; ScienCell), or transformed sinusoidal endothelial cells (TSEC),32 an SV40-immortalized mouse cell line that largely recapitulates the phenotype of pathological vasculature (Robert Huebert; unpublished data), were grown in standard tissue culture conditions in Endothelial Cell Media (ScienCell).

Quantitative Reverse Transcription Polymerase Chain Reaction.

RNA was isolated using the RNeasy kit (Qiagen) according to the manufacturer's instructions. RNA was reverse transcribed using the SuperScript III System (Invitrogen), and TaqMan-based real-time reverse transcription polymerase chain reaction (RT-PCR) was performed according to the manufacturer's instructions (Applied Biosystems).


Western blotting was performed from liver lysates or endothelial cell lysates, as previously described.18


Immunohistochemistry (IHC) was performed from normal or cirrhotic paraffin-embedded human liver tissue, as previously described.18


Immunofluorescence (IF) was performed on normal or cirrhotic frozen liver tissues from mouse or human, as previously described.18

Viral Transduction.

The complementary DNA sequence of AQP-1 was subcloned into the pMMP retroviral vector and used to generate retroviral supernatant in 293T cells. TSEC were incubated with supernatant for 24 hours. AdhAQP1 was provided by Dr. Bruce Baum).

Chemotaxis Assays.

Chemotaxis in LECs, human hepatic sinusoidal endothelial cells, or TSEC was measured by using a modified Boyden chamber assay (Becton Dickinson) in response to FGF, serum, or vehicle.

Collagen Invasion Assays.

Invasion was measured in TSEC overexpressing LacZ or AQP-1 using a three-dimensional collagen assay33 in response to FGF or vehicle.

Small Interfering RNA Transfection.

Prevalidated small interfering RNA (siRNA) from Qiagen was transfected using RNAiFect (Qiagen) according to the manufacturer's instructions. The final concentration of siRNA during transfection was 100 nM. Negative control siRNA was used for all experiments. Protein knockdown was confirmed by western blot.

Measurement of Blebbing Dynamics.

Primary mouse LECs or TSEC overexpressing LacZ or AQP-1 were stimulated using FGF or VEGF. Cells were imaged and measured using time-lapse, phase-contrast microscopy, and volume, surface area, osmotic water permeability, and water flux were calculated.

Apoptosis Assays.

TSEC overexpressing LacZ or AQP-1 were treated with 25 ng/mL mouse FGF or 10 ng/mL tumor necrosis factor alpha and incubated for 18 hours. Caspase activation was assayed using the Apo One Homogeneous Caspase-3/7 Assay (Promega) according to the manufacturer's instructions.

Visualization of Osmotic Water Flux.

Calcein fluorescence quenching was used to assess areas of localized water influx as previously described.23

Electron Microscopy.

Immunogold labeling and scanning electron microscopy (SEM) were performed as previously described34 on TSEC overexpressing AQP-1 or LacZ and treated with FGF.

Statistical Analysis.

Data are presented as means ± standard error of the mean. Bar graphs, blots, and micrographs represent typical experiments reproduced at least three times. Statistical analyses were performed by two-tailed Student t tests.


AQP-1 Expression Is Increased in Cirrhotic Liver.

To begin testing our hypothesis proposing a pathophysiological role for AQP-1 in cirrhosis, we assessed the expression of AQP-1 messenger RNA and protein in normal and cirrhotic liver from humans. We noted a dramatic increase in AQP-1 messenger RNA in nonalcoholic fatty liver disease (NAFLD) that correlated with stage of cirrhosis using real-time quantitative RT-PCR (Fig. 1A). Western blotting confirmed overexpression of AQP-1 protein in cirrhosis attributable to both NAFLD (Fig. 1B) and chronic hepatitis C (Fig. 1C). We next sought to recapitulate these findings in the context of a mouse model of cirrhosis, carbon tetrachloride (CCl4)27 injection, and found that AQP-1 expression was increased in CCl4 as compared with vehicle-treated mice (Fig. 1D). This cross-species consistency provides a useful animal model in which to test further hypotheses.

Figure 1.

AQP-1 expression is increased in cirrhotic liver. (A) Complementary DNA from human liver tissue at various stages of NAFLD-induced cirrhosis amplified using quantitative RT-PCR (n = 3; mean ± standard error [SE]). (B-D) Representative immunoblots for AQP-1 or actin (50 μg/lane) in control or cirrhotic liver tissue from human or mouse and densitometric quantification (n = 4-9; mean ± SE). Panels represent: (B) multiple human NAFLD patients; (C) multiple human hepatitis C patients; and (D) CCL4 model. *P < 0.05.

AQP-1 Is Localized to Angiogenic Neovasculature Within Fibrotic Septa.

To confirm and extend our findings regarding AQP-1 expression during cirrhosis, we used IHC and IF to measure angiogenesis and to localize the source of increased AQP-1 in cirrhotic human liver. IHC for Von Willebrand factor (vWF), a marker of liver endothelia,35 showed significantly increased angiogenesis in cirrhotic human liver compared with normal controls (Fig. 2A). IF (Fig. 2B) and IHC (Supporting Fig. 1A) localized the increased AQP-1 signal to small, angiogenic vessels within fibrotic septa, consistent with reports linking fibrogenesis and angiogenesis, and suggesting a role for AQP-1 in these processes.8 Similar results were seen in the CCl4 mouse model (Fig. 2B). Western blotting confirmed AQP-1 overexpression in endothelial cells isolated from cirrhotic animals as compared with cells isolated from control animals (Supporting Fig. 1B). Costaining showed significant colocalization of the increased AQP-1 signal with additional endothelial markers endothelial nitric oxide synthase and platelet/endothelial cell adhesion molecule, but not with cytokeratin 19, a biliary marker, nor with alpha smooth muscle actin, a stellate cell marker (Fig. 2C). Together, these data demonstrate robust overexpression of AQP-1 in the angiogenic neovasculature of cirrhotic liver.

Figure 2.

AQP-1 is localized to endothelia within fibrotic septa. (A) Control or cirrhotic human liver tissue stained with IHC for vWF (brown), counterstained with hematoxylin (blue), and quantified (n = 5; mean ± SE). *P < 0.05. (B) Control or cirrhotic human and mouse liver tissue stained with IF for AQP-1 (red) and the dimeric cyanine nuclear stain, TOTO-3 iodide (TOTO-3) (blue). See Supporting Fig. 1 for AQP-1 IHC. (C) IF costaining on cirrhotic liver tissue for AQP-1 (red) and endothelial nitric oxide synthase, platelet/endothelial cell adhesion molecule, alpha-smooth muscle actin, or cytokeratin-19 (green).

AQP-1 Is Significantly Overexpressed After Viral Transduction.

To allow efficient studies of AQP-1 effects on LEC in vitro, we modulated AQP-1 expression levels in cultured cells. Interestingly, we observed that although endothelial cells in situ as well as freshly isolated LECs express AQP-1, cells in culture, including TSEC, quickly lose expression of AQP-1 when removed from the in vivo microenvironment. We therefore created a retroviral transduction system to stably overexpress AQP-1 in vitro (pMMP-AQP1). IF (Fig. 3A) and western blotting (Fig. 3B) demonstrated robust overexpression of AQP-1 after treatment with pMMP-AQP1 compared with pMMP-LacZ controls, providing a mechanistic in vitro model in which to study the biological effects of AQP-1.

Figure 3.

AQP-1 is overexpressed in TSEC after retroviral transduction. (A) TSEC transduced with retroviral AQP-1 or LacZ control and stained by IF for AQP-1 (red) and nuclear TOTO-3 (blue). (B) Representative immunoblots (50 μg/lane) for AQP-1 or actin from TSEC treated with pMMP-AQP1 or LacZ control.

AQP-1 Enhances FGF-Stimulated HSEC Invasion.

Based on our a priori hypothesis that AQP-1 promotes angiogenesis, we speculated that AQP-1 up-regulation would increase LEC motility. We therefore tested the effects of AQP-1 overexpression on LEC chemotaxis using modified Boyden chamber chemotaxis assays. However, contrary to our initial hypothesis, we found that after AQP-1 overexpression with pMMP-AQP1, traditional chemotaxis in LEC was actually reduced compared with LacZ controls, both in the basal state and in response to FGF (Fig. 4A). Similar results were observed using human hepatic sinusoidal endothelial cells, various chemotactic agents, and both adenoviral and retroviral overexpression (Supporting Fig. 2A). Using AQP-1-specific siRNA or scrambled siRNA, we found, again, that AQP-1 expression was inversely correlated with HSEC chemotaxis in primary cells (Supporting Fig. 2B). Attempts to modulate AQP-1 function with chemical inhibitors, such as mercuric chloride, resulted in endothelial cell toxicity and therefore were not pursued in greater depth (Supporting Fig. 3). Recent studies have revealed that, in the context of desmoplasia, cells frequently modify their migration pattern from a traditional, actin-based, mesenchymal mechanism, to a membrane deformation mechanism that has been referred to as ameboid motility.36 This invasive form of motility, although slower, is nonetheless, more adaptable in circumstances requiring cell shape deformation and dynamic membrane blebbing events to squeeze through confined areas. We hypothesized that the dense fibrotic ECM of the cirrhotic microenvironment requires invasion and that LECs undergo mode-switching to a more primitive form of amoeboid motility in this setting. We therefore tested the effects of AQP-1 overexpression on FGF-induced endothelial cell invasion capacity using three-dimensional collagen invasion assays. In striking contrast to our chemotaxis results, we observed that AQP-1 overexpression in TSEC significantly increased both basal and FGF-induced invasion (Fig. 4B), suggesting that bleb-based amoeboid motility occurs in this setting.

Figure 4.

AQP-1 enhances TSEC invasion capacity. (A) TSEC transduced with retroviral AQP-1 or LacZ control subjected to chemotaxis assays in the presence or absence of FGF. Migrated cells are stained with 4′,6-diamidino-2-phenylindole (DAPI (blue) and quantified (n = 3; mean ± SE). See Supporting Fig. 2 for additional chemotaxis data. (B) TSEC transduced with retroviral AQP-1 or LacZ controls were subjected to three-dimensional collagen invasion assays in the presence or absence of FGF. Invading cells are stained with 4′,6-diamidino-2-phenylindole (blue) and quantified (n = 3; mean ± SE). *P < 0.05 versus LacZ. #P < 0.05 versus LacZ+FGF.

AQP-1 Enhances FGF-Induced Dynamic Membrane Blebbing.

Because dynamic membrane blebbing is the hallmark of amoeboid motility,15 we used phase-contrast, time-lapse video microscopy to examine blebbing behavior in TSEC. We hypothesized that altered blebbing dynamics may contribute to amoeboid motility and explain the increased invasion capacity conferred by AQP-1. Phase contrast and SEM (Fig. 5A) demonstrated that treatment with FGF induced dynamic membrane blebbing in TSEC that was further enhanced by AQP-1 overexpression (Fig. 5A, Supporting Movies 1 and 2). In contrast, we did not observe dynamic membrane blebbing after treatment with VEGF, suggesting that amoeboid invasion may be FGF specific in these cells (Supporting Fig. 4). The time-course of bleb formation and retraction revealed bleb formation occurring rapidly over a period of seconds, with ensuing retraction occurring more slowly (Fig. 5B), consistent with amoeboid blebbing.15 Interrogation into the precise nature of the enhanced blebbing activity showed that AQP-1 overexpression in TSEC significantly increased maximum bleb size as assessed by both phase contrast and SEM (Fig. 5C). We quantified these changes and found that AQP-1 overexpression increased maximum bleb volume and surface area, and that this effect was reversible with AQP-1-specific siRNA (Fig. 5D). To confirm the enhanced membrane dynamics in primary cells, we repeated the analysis on freshly isolated LECs from normal or cirrhotic mice. Mice treated with CCL4 showed significantly increased blebbing dynamics compared with control mice, an effect that was abrogated with AQP-1-specific siRNA (Fig. 5E), thus confirming relevance to the in vivo cirrhotic milieu. The stimulatory effects of AQP-1 on blebbing dynamics provide a cell biological mechanism to correlate with our functional invasion data. Because membrane blebs in healthy cells can be indistinguishable from those associated with apoptosis, we performed caspase 3, 7 activation assays on cells overexpressing LacZ or AQP-1 in the presence and absence of FGF. We found that whereas tumor necrosis factor alpha, a potent inducer of apoptosis, caused intense activation of apoptotic pathways, the experimental conditions that induce membrane blebbing showed no such activation (Fig. 6A, B). Furthermore, on removal of the FGF stimulus, blebbing ceases, and TSEC revert to a traditional actin-based migration phenotype (Fig. 6C-F, Supporting Movie 3). Thus, we conclude that AQP-1 enhances nonapoptotic, FGF-induced, dynamic membrane blebbing.

Figure 5.

AQP-1 enhances dynamic membrane blebbing. (A) TSEC transduced with retroviral AQP-1 or LacZ control, treated with FGF or vehicle, and imaged using phase-contrast, time-lapse, video microscopy (left panels) or scanning electron microscopy (right panels). See Supporting Movies 1 and 2. (B) Time course of a dynamic membrane bleb in TSEC at high magnification (1 second/frame). (C) TSEC transduced with retroviral AQP-1 or LacZ control, treated with FGF, and imaged using phase-contrast microscopy (upper panels) or scanning electron microscopy (lower panels). (D) TSEC were transduced with retroviral AQP-1 or LacZ control or transfected with AQP-1-specific siRNA or scrambled siRNA, treated and imaged as above, and measured to calculate maximum bleb volume and surface area. (E) Freshly isolated LECs from mice treated with CCl4 or vehicle were treated, imaged, and measured, as above. *P < 0.05 versus control.

Figure 6.

Dynamic membrane blebbing in TSEC is nonapoptotic. (A) TSEC were transduced with retroviral LacZ, treated tumor necrosis factor alpha, or FGF, and subjected to caspase 3,7 activity assays (n = 6; mean ± SE). (B) TSEC transduced with retroviral AQP-1 were treated and assayed as above (n = 6; mean ± SE). *P < 0.05 versus control. (C-F) TSEC were imaged using phase-contrast, time-lapse, video microscopy during dynamic blebbing after withdrawal of FGF at times (C) 0 hours, (D) 1 hour, (E) 2 hours, and (F) 3 hours. See Supporting Movie 3.

AQP-1 Localizes to Membrane Blebs and Facilitates Localized Water Influx.

To further define the mechanism of AQP-1-enhanced membrane blebbing, we investigated the ultrastructural localization of AQP-1 in cells undergoing membrane blebbing. Immunogold labeling coupled with SEM showed clear localization of AQP-1 to the periphery of plasma membrane blebs in cells treated with pMMP-AQP-1, unlike cells treated with pMMP-LacZ (Fig. 7A). IF confirmed the subcellular localization of AQP-1 on plasma membrane blebs (Fig. 7B). In costaining experiments, AQP-1 decorated blebs with a myosin II-positive base, a common marker associated with blebbing.37 We next preloaded TSEC with a self-quenching fluorescent dye, Calcein-AM (the intensity of which increases on dilution) and induced blebbing to show that localized water influx is occurring across the bleb membrane (Fig. 7C). To quantify the effects of AQP-1 on localized water influx, we used time-lapse, phase-contrast microscopy to measure volume changes over time in response to an osmotic gradient and estimated both relative osmotic water permeability (Pf) and water flux (Jv). AQP-1 overexpression increased both Pf and Jv (1.77-fold and 3.29-fold, respectively), an effect that was inhibited with AQP-1 siRNA (Fig. 7D, E). Coupled with the changes seen in bleb dynamics and invasion, these results strongly support a role for AQP-1-mediated water transport as a biophysical component of the forces driving dynamic membrane blebs, thereby facilitating FGF-induced amoeboid invasion in LECs.

Figure 7.

AQP-1 localizes to membrane blebs and facilitates regional water flux. A TSEC overexpressing AQP-1 or LacZ control labeled with AQP-1-specific immunogold particles and imaged with scanning electron microscopy at 10,000× or 50,000×. (B) TSEC overexpressing AQP-1 treated with FGF and stained by IF for AQP-1 (green) or both AQP-1 and myosin II (red). (C) TSEC overexpressing AQP-1 were preloaded with calcein AM, treated with FGF, and imaged during blebbing using phase-contrast or confocal microscopy. Pseudocolor enhancement shows increased fluorescence within blebs (representing localized dilution). Inset shows high magnification of a bleb. (D, E) A 30-mOsm external osmotic gradient was applied to TSEC transduced with retroviral AQP-1 or LacZ, or transfected with AQP-1-specific siRNA or scrambled siRNA, and treated with FGF, imaged, and measured over time to calculate osmotic water permeability (D) and water flux (E).


An understanding of the precise mechanisms controlling endothelial cell invasion and angiogenesis in liver, especially in a pathophysiological context, is an important area of investigation given recent implications of anti-angiogenic therapies on the treatment of liver disease.3, 4, 8, 10 In this regard, the current study provides the following novel observations: (1) AQP-1 expression is increased in the neovasculature within cirrhotic liver in vivo; (2) FGF promotes mode-switching toward an invasive, amoeboid phenotype that is sufficient to drive endothelial cell invasion through ECM; (3) AQP-1 overexpression enhances both dynamic membrane blebbing and invasion capacity in LEC; and (4) AQP-1 localizes to plasma membrane blebs, where it allows for the rapid, trans-membrane flux of water. This data provide several conceptual advances across disciplines. First, we have elucidated a new mechanism for endothelial cell invasion in the cirrhotic liver involving FGF, an understanding of which might ultimately allow for more refined targeting of anti-angiogenic therapy in cirrhosis. Second, although amoeboid motility is increasingly recognized as an important form of invasion in the contexts of embryology, immunology, and malignancy, there are surprisingly few studies in endothelial cells.38, 39 Indeed, to our knowledge, this is the first study to demonstrate amoeboid motility in the context of angiogenic invasion. Third, our data implicate channel-mediated water transport across dynamic membrane blebs, a concept that could substantially alter our understanding of these structures and their role in liver relevant processes.

Considerable controversy exists in the literature regarding the causal relationship between hepatic fibrosis and pathological liver angiogenesis. There is evidence of the anti-angiogenic compound, Sunitinib, reducing hepatic fibrosis in experimental animals.3 In contrast, the integrin αvB3 inhibitor, Cilengitide, worsens fibrosis in similar models.8 What appears congruent is that angiogenesis and fibrosis occur together, are closely intertwined, and that there is considerable molecular and paracrine crosstalk in the signals driving each process. Less apparent are the precise mechanisms by which one process perpetuates the other and the therapeutic implications of inhibition of either pathway. It is anticipated that resolution of these issues will require further studies to provide a more detailed understanding of the molecular mechanisms driving endothelial neovascularization in the setting of enhanced ECM deposition, as we pursued here.

Whereas lower organisms such as Amoeba proteus and Dictyostelium discoidium rely entirely on bleb-based amoeboid motility,16, 36 there is evidence in diverse cellular circumstances that higher eukaryotic cells undergo a so-called “mesenchymal to amoeboid transition” in situations requiring rapid deformation of cellular shape.36, 37 Examples of this transition include diapedesis of leukocytes,38 metastatic invasion,39 and embryogenesis. This mode of motility may be favored in microenvironments containing dense three-dimensional ECM, such as that seen in cirrhosis. Our data suggest that mode-switching toward amoeboid invasion may be a previously unrecognized, yet important mechanism in the development of blood vessels in fibrotic liver.

The biophysics at play in the dynamic expansion and retraction of blebs is complex, involving expansion by cytoplasmic streaming (a hydraulic force caused by contraction of the cytoskeletal cortex), and mechanical retraction (a force caused by myosin II activation). We propose that a third force also may be at play, an osmotic force, driving water influx and efflux. Indeed, we see localization of water channels at the periphery of dynamic membrane blebs, similar to the dynamic protrusions in C. parvum infection of cholangiocytes.23 Our data support a role for channel-mediated, trans-membrane water flux in membrane blebs that is sufficient to enhance FGF-induced blebbing and to promote invasive angiogenesis (Fig. 8). This provocative idea suggests that angiogenesis in general could be driven, in part, by local osmotic gradients. Physiological interactions between AQPs and several ion/solute transporters, including the Na+/H+ exchanger,40, 41 the Cl/HCOmath image exchanger (AE2),24 the cystic fibrosis transmembrane regulator,24 and the Na+/glucose cotransporter 123 are well described in other cell types. However, the ion/solute transporters that create the osmotic gradients to help drive the expansion and retraction of endothelial blebs are currently unknown.

Figure 8.

Working model. Upper panel depicts FGF-stimulated, AQP-1-mediated water transport that enhances dynamic membrane blebbing and promotes amoeboid invasion through the cirrhotic microenvironment. Lower panel depicts a single bleb demonstrating a combination of hydraulic, mechanical, and osmotic forces driving dynamic membrane bleb expansion and retraction.

Numerous small molecule inhibitors of AQPs are currently known, including mercurial agents, gold compounds, dimethyl sulfoxide, quaternary ammonium compounds, carbonic anhydrase inhibitors, and plant flavonoids such as phloretin.26, 42 However, none are suitable for clinical applications because of toxicity and lack of specificity. As rapid screening techniques for water channel function continue to become available,42 large-scale testing of pharmaceutical compounds should accelerate the discovery of new AQP inhibitors.43 Currently, mechnistic in vivo studies will require the use of genetic AQP knockout models.

In summary, our findings identify a mechanism whereby LECs can adapt to the cirrhotic microenvironment and pursue invasion, despite the presence of fibrotic scar, thereby driving pathological angiogenesis and progression of fibrosis. We anticipate that these results may provide insights in other pathophysiological contexts such as hepatic malignancies and angiogenesis in other vascular beds. It is also conceivable that the findings could be generalizable to other dynamic cell membrane events, such as phagocytosis, apoptosis/autophagy, cytokinesis, and amoeboid motility in other cell types.


The authors thank Patrick Splinter and Helen Hendrickson for technical support and Theresa Johnson for secretarial support.