These authors contributed equally to this work
Article first published online: 7 JAN 2013
Copyright © 2012 American Association for the Study of Liver Diseases
Volume 57, Issue 1, pages 205–216, January 2013
How to Cite
Sun, X., Han, L., Seth, P., Bian, S., Li, L., Csizmadia, E., Junger, W. G., Schmelzle, M., Usheva, A., Tapper, E. B., Baffy, G., Sukhatme, V. P., Wu, Y. and Robson, S. C. (2013), Disordered purinergic signaling and abnormal cellular metabolism are associated with development of liver cancer in Cd39/Entpd1 null Mice. Hepatology, 57: 205–216. doi: 10.1002/hep.25989
Potential conflict of interest: Nothing to report.
Supported by National Institutes of Health (NHLBI P01-HL076540 and R01-HL094400).
- Issue published online: 7 JAN 2013
- Article first published online: 7 JAN 2013
- Accepted manuscript online: 1 AUG 2012 07:35AM EST
- Manuscript Accepted: 20 JUL 2012
- Manuscript Received: 6 MAR 2012
Liver cancer is associated with chronic inflammation, which is linked to immune dysregulation, disordered metabolism, and aberrant cell proliferation. Nucleoside triphosphate diphosphohydrolase-1; (CD39/ENTPD1) is an ectonucleotidase that regulates extracellular nucleotide/nucleoside concentrations by scavenging nucleotides to ultimately generate adenosine. These properties inhibit antitumor immune responses and promote angiogenesis, being permissive for the growth of transplanted tumors. Here we show that Cd39 deletion promotes development of both induced and spontaneous autochthonous liver cancer in mice. Loss of Cd39 results in higher concentrations of extracellular nucleotides, which stimulate proliferation of hepatocytes, abrogate autophagy, and disrupt glycolytic metabolism. Constitutive activation of Ras-mitogen-activated protein kinase (MAPK) and mammalian target of rapamycin (mTOR)-S6K1 pathways occurs in both quiescent Cd39 null hepatocytes in vitro and liver tissues in vivo. Exogenous adenosine 5′-triphosphate (ATP) boosts these signaling pathways, whereas rapamycin inhibits such aberrant responses in hepatocytes. Conclusion: Deletion of Cd39 and resulting changes in disordered purinergic signaling perturb hepatocellular metabolic/proliferative responses, paradoxically resulting in malignant transformation. These findings might impact adjunctive therapies for cancer. Our studies indicate that the biology of autochthonous and transplanted tumors is quite distinct. (HEPATOLOGY 2013)
Hepatocarcinogenesis is linked to chronic inflammation. Under these circumstances, disordered cellular proliferation, with decreases in autophagy and aberrant metabolism, might predispose to malignant transformation. The mammalian target of rapamycin (mTOR) has been shown to play a critical role in these processes.1 More specifically, in these settings, Ras and phosphatidylinositol-3-OH kinase (PI3K) pathways converge to activate mTOR in response to nutrients and to mitogens.2
The role of purinergic signaling3, 4 in hepatocarcinogenesis is unexplored. In hepatocytes, extracellular nucleotides (specifically adenosine 5′-triphosphate [ATP] and uridine 5′-triphosphate [UTP]) up-regulate Ca2+ signaling and activate mitogen-activated protein kinase (MAPK) cascades (namely, c-Jun NH2-terminal kinase [JNK] and extracellular signaling-related kinase [ERK]) as well as transcription factor nuclear factor kappa B (NF-κB) through the activation of type 2 purinergic (P2) receptors.4, 5 Although such molecular pathways are clearly associated with tumorigenesis, it is unknown whether such effects occur by way of the mTOR signaling pathway, given that Ras and PI3K are often components of P2 receptor signaling.6
CD39/ENTPD1 (nucleoside triphosphate diphosphohydrolase-1) is the dominant ectonucleotidase expressed by hepatic endothelium, Kupffer cells, and sinusoidal lymphocytes and catalyzes nucleotide phosphohydrolysis.3, 4 We have previously shown that CD39 expression on regulatory T cells (Treg) inhibits natural killer (NK) cell activity and is permissive for the growth of metastatic tumors in the liver.7 Further, vascular-expressed CD39 boosts angiogenesis8 and directly promotes tumor cell growth by scavenging cytotoxic extracellular ATP.9
We have further demonstrated that mice globally deficient in Cd39 exhibit metabolic disturbances such as glucose intolerance, increased hepatic glucose production, insulin resistance, and increased plasma levels of insulin and fatty acids, all associated with heightened inflammatory markers.10 Curiously, these mutant mice also exhibit disordered liver regeneration and increased liver injury with impediments to endothelial cell-dependent hepatocyte proliferation, which is then further compromised by failure of vascular reconstitution in these mutant mice.11 The end result is the observation of persistent hepatocyte proliferation at 21 days post 70% hepatectomy.11
We show here that Cd39 deletion paradoxically promotes development of both induced and spontaneous liver cancer, in stark contrast to the demonstrated effects on transplanted tumors to the liver.7 We suggest that altered purinergic responses modulate gene regulation in the Cd39 null mouse liver promoting malignant transformation. This develops in either nitrosamine-injured tissues or in the setting of accumulated genetic rearrangements with associated DNA point mutations.12 Our observations have implications for development of adjunctive therapies for liver cancer. These paradoxical responses to Cd39 deletion further emphasize that the biological characteristics of autochthonous versus transplanted tumors are quite distinct.13, 14
Materials and Methods
Mice and General Reagents/Antibodies.
We used 5 to 12-week-old male Cd39-null mice on the C57BL/6 background (interbred and backcrossed ×12).15 Age-, sex-, and strain-matched wildtype (WT) mice were purchased from Taconic (Hudson, NY). The animal experimentation protocol was reviewed and approved by the Institutional Animal Care and Use Committees (IACUC) of Beth Israel Deaconess Medical Center (BIDMC). More details are described in Supporting Materials and Methods.
Diethylnitrosamine (DEN)-Induced Hepatocellular Carcinogenesis.
The DEN-induced liver tumor model was described by Koen et al.16 Briefly, mice were given an intraperitoneal injection of 5.0 μg DENA/g body weight in 30 μL of saline at 15 days of age. Control mice were given an equivalent volume of saline. Mice were sacrificed at 1 year and examined for liver tumor formation.
Spontaneous Development of Hepatocellular Carcinoma (HCC).
Age-, sex-, and strain-matched Cd39-null and WT mice were maintained at the Animal Research Facility at Center for Life Science of BIDMC. Mice were aged over 18 months, sacrificed, and examined for spontaneous formation of liver tumors.
In vivo hepatic infusion of ATP and histology was as described10 with modifications detailed in the Supporting Materials and Methods.
Primary hepatocyte culture was as established in our laboratory11 and detailed in the Supporting Materials and Methods.
Cell Proliferation Assays.
Intracellular Nucleotide Measurements.
Extracellular Acidification Rate (ECAR) Analysis and Lactate Accumulation.
Western Blotting and Ras Activation Assay.
Statistical Analysis. Results are expressed as ± standard error of the mean (SEM).
For statistical analyses, the two-tailed Student's t-test was used. Significance was defined as P < 0.05.
Cd39 Deletion and HCC.
Two murine models of HCC were established and examined in this study. First, de novo HCC formation was induced in C57/BL6 WT and Cd39-null mice. Although all mice developed liver cancers, the tumor burden in Cd39-null mice was markedly increased when compared to WT controls (Fig. 1A-C). Incidence of tumors greater than 5 mm in diameter was 69% (9/13) in null livers and 6.7% (1/15) in WT livers (Supporting Table S1A, P < 0.001). Using a morphological classification of “mouse liver tumors” (MLT),22 30% of null tumors had high-grade malignancy (MLT type III and IV) (Fig. 1C, lower), whereas tumors arising in WT livers were of low-grade malignancy (chiefly MLT type I and II) (Fig. 1C, upper).
We next studied HCC arising spontaneously and in an autochthonous manner in mice. The incidence of liver neoplasia in aged (18-24 months old) Cd39-null mice (70%, 19/27) was significantly higher than that of WT mice (29%, 2/7) (Table S1B; P = 0.04; Fig. 1D,E). Accordingly, the mouse tumor biology database at Jackson Laboratories indicates that the frequency of spontaneous liver tumor varies from 8.8% to 30% in aged WT C57/BL6J mice (tumor.informatics.jax.org). Hematoxylin-eosin (H&E) staining of liver tumor sections from Cd39-null mice confirmed that tumors were of the hepatocellular type with curiously marked hypervascularity (Fig. 1E).
Hepatocyte Proliferation Is Increased in Cd39-Null Mice.
Cd39-null mice developed spontaneous hepatic necrosis concomitantly accompanied by adjacent hepatocellular dysplasia at a young age (as early as 5 weeks old) (Fig. 2A). These mutant mice, in comparison to WT controls, exhibited significantly increased liver-to-body weight ratios, (Fig. 2B; P < 0.0001). Next we used RT-PCR to characterize the complement of P2 receptors expressed by WT hepatocytes. Mouse hepatocytes expressed messenger RNA (mRNA) transcripts for P2X4, P2X5, P2X7 (weaker), P2Y1, P2Y2, and P2Y12-14 (Supporting Fig. S1).
We then examined the impact of Cd39 deletion on hepatocyte proliferation. First, Cd39-null cells exhibited a heightened baseline proliferation rate compared to WT cells (Fig. 2C). This occurred regardless of the culture conditions and appeared indicative of prior “set-points” in null cells (considered as possible “preprogramming”). Second, insofar as ATP has differential effects on cell proliferation at different concentrations, we evaluated the responses of WT cells to increasing concentrations of ATP. Figure 2D shows that at the 10 to 100 μM range, ATP stimulated cell proliferation. In contrast, high concentrations of ATP (>2,500 μM) inhibited proliferation (Fig. 2D). This is in keeping with our recent observations.9 Third, we observed that hepatocyte proliferation was enhanced by ATP, as determined by the classical 3H-TdR-incorporation method (Fig. 2E) and by the Cell Counting Kit-8 (CCK-8) that measures the activity of cellular dehydrogenases (Fig. S2A). ATP-stimulated hepatocyte proliferation was completely abolished by coincubation with the global P2 receptor antagonist suramin (Fig. 2F). Additionally, similar stimulatory effects were also noted with UTP (50 μM) (Fig. S2B).
Autophagy Is Suppressed in Cd39-null Hepatocytes.
Autophagy is a cellular degradation response to starvation/stress removing damaged/surplus proteins and organelles to thereby tightly control cell growth. Autophagy defects have been linked to various pathogenic conditions, particularly cancers.23 A sensitive marker for autophagy, light chain 3-II (LC3-II), was used here. Figure 3A shows that starvation-induced elevation of LC3-II levels was significantly inhibited by ATP and that apyrase (a soluble NTPDase) reversed this ATP-mediated suppression. LC3-II levels in WT cells were increased 12 hours after starvation and peaked at 24 hours (Fig. 3B). In contrast, levels of LC3-II were remarkably low in Cd39-null cells (Fig. 3B).
In parallel, mRNA expression of most autophagy-associated genes examined (Beclin-1, ATG-5, and ATG-7) were also significantly suppressed by ATP in WT cells (Fig. 3C). Similarly, ATP-induced inhibition of autophagy genes was observed in Cd39-null cells as well (Fig. S3). Finally, mRNA expression of major autophagy genes (Beclin-1, ATG-5, ATG-7, ATG-12, and Vps34) were significantly decreased in null cells post-serum/mitogen-deprivation (Fig. 3D).
Taken together, the data indicate that autophagy suppression in Cd39-null hepatocytes is, at least in part, mediated by way of disordered extracellular nucleotide-initiated purinergic responses.
Deletion of Cd39 Disrupts Glycolytic Pathways in Hepatocytes.
Autophagy is a basic cellular catabolic process that fuels oxidative phosphorylation by supplying essential molecules by way of the break down of nonfunctional intracellular components. As such, the inhibition of autophagy in Cd39-null hepatocytes (Fig. 3) suggests the dominance of anabolic pathways. Interestingly, proliferation assays assessing the activity of dehydrogenases using the CCK-8 kit (Fig. 2C) depict a higher proliferation rate of null cells compared to WT cells, indicating that Cd39-null cells are metabolically more active and proliferate more rapidly. We now provide evidence indicating Cd39-null hepatocytes are preferentially deviated towards aerobic glycolysis.
First, we examined pyruvate kinase M2 (PKM2) and lactate dehydrogenase A (LDH-A), enzymes that are the key metabolic control points for aerobic glycolysis. Decreases in PKM2 activity and increases in LDH-A expression promote pyruvate conversion to lactate and thereby drive glycolysis. Proliferating cells and cancer cells exclusively express PKM2. For example, FGFR1 (fibroblast growth factor receptor 1) directly phosphorylates Tyr105 of PKM2, thereby inhibiting the formation of its active tetramer, suggesting that tyrosine phosphorylation of PKM2 may serve as a critical glycolytic switch in cancer cells.24 As shown in Fig. 4A (left), ATP stimulated Tyr105 phosphorylation of PKM2 in both WT and Cd39-null hepatocytes in vitro, albeit with a more dramatic stimulatory effect on null cells. We then examined tyrosine phosphorylation of PKM2 in mouse livers after portal venous administration of ATP in vivo. ATP also promoted Tyr105 phosphorylation of PKM2 in WT livers and that Cd39-null livers had much higher tyrosine phosphorylation even before ATP infusion (Fig. 4A, right). Furthermore, total PKM2 proteins were increased by ATP in both WT and null cells (Fig. 4B). In parallel, expression of LDH-A mRNA was elevated in quiescent Cd39-null cells as compared to WT cells and was further enhanced by ATP in both cells (Fig. 4C, left). Moreover, similar patterns were also noted with LDH-A proteins (Fig. 4C, right).
We next evaluated the impact of Cd39 deletion on mitochondrial function in hepatocytes. As shown in Fig. 4D (top), ATP downregulates expression of cytochrome B mRNA and upregulates expression of mitochondrial uncoupling protein UCP2 mRNA in WT cells (see also bottom panel for comparison of WT versus null, Fig. 4D; Fig. S4A,B for other genes). Decreased expression of cytochrome B, Cox2 (cyclooxygenase 2), glucagon (a mitochondrial regulator), and increased levels of UCP2 were noted in quiescent Cd39-null livers in contrast to intact WT livers (Fig. S4C), associated with enhanced ribosome biogenesis (indicated by mRNA levels of 18S rRNA) (Fig. S4D). Importantly, decreases in intracellular ATP levels were noted in Cd39-null cells (Fig. 4E; Table S2). In addition, lactate formation, the endproduct of glycolysis, was significantly increased in null cells (Fig. 4F, P = 0.01), as measured in a real time fashion.
Collectively, these data show that ATP-initiated purinergic signaling is associated with altered bioenergetic metabolism of hepatocytes that promotes aerobic glycolysis by modifying glycolytic enzyme expression and/or activity as well as disrupting mitochondrial function to favor anabolic pathways and promote cell growth.
Ras, PI3K, and mTOR/S6K1 Signaling in Cd39-Null Hepatocytes and Livers.
We first studied the activation of Ras in response to ATP in vitro. Levels of activated Ras in WT cells were increased by ATP (Fig. 5A). Ras was also elevated in quiescent Cd39-null cells and increased following ATP stimulation, in contrast to WT cells (Fig. 5A).
We next examined the phosphorylation of downstream components of Ras, PI3K, and mTOR pathways by ATP in hepatocytes. In both WT and null cells, ATP enhanced the phosphorylation of components of many pathways, including ERK1/2, JNK/SAPK, NF-κB in Ras-MAPK signaling, AKT in PI3K signaling, and mTOR, S6K1, S6 in mTOR signaling (Fig. 5B). Boosting these signaling pathways was also noted in quiescent as well as in ATP-stimulated null cells, as compared to WT cells (Fig. 5B).
Finally, we assessed the phosphorylated levels of these signaling molecules in mouse livers posthepatic ATP infusion as described above. The phosphorylation of these signaling components in both untreated and ATP-stimulated Cd39-null livers nearly exactly recapitulates the same pattern observed in cells (Fig. 5C). These data indicate that Cd39 deletion results in persistent activation of oncogenic pathways with an increased incidence of autochthonous tumor formation in the liver.
Rapamycin Corrects Purinergic Signaling-Mediated Responses in Hepatocytes.
We further sought to delineate the role of mTOR by addition of rapamycin to these model systems. First, hepatocyte proliferation was significantly inhibited by rapamycin in both ATP-stimulated and nonstimulated cells (Fig. 6A). Furthermore, ATP-stimulated proliferative responses could be almost completely blocked by this approach (Fig. 6A). However, Cd39-null cells exhibited a higher proliferation rate than WT cells, regardless of the treatment strategy (Fig. 6A).
Second, we evaluated the effect of rapamycin on hepatocyte autophagy. This fully restored the previously noted ATP-induced inhibition of autophagy in WT cells at the level of LC3 degradation (Fig. 6B).
Third, we analyzed mitochondrial gene/regulator mRNA expression post-rapamycin treatment. The dysregulation of mitochondrial genes (LDH-A, cytochrome B, UCP2, Cox1, Cox2, NADHsub1, and NADHsub2) and regulators (PGC-1β, TFAM, NRF, and glucagon) in Cd39-null cells could be reversed by mTOR inhibition (Fig. 6C; Fig. S5). We also noted that rapamycin exhibited comparable effects on WT cell signaling responses, albeit with variable potency (Table S3).
Fourth, we examined the impacts of rapamycin on lactate production by hepatocytes. Accumulated lactate levels in both WT and null cells were significantly inhibited by this approach (Fig. S6). No significant differences between WT and null cells were noted.
Fifth, to further investigate the signaling pathways impacted by rapamycin, we studied the phosphorylated levels of mTOR-S6K1-S6 and downstream targets of Ras. We noted three key findings. In both rapamycin-treated WT and Cd39-null cells, mTOR phosphorylation was significantly decreased, whereas phosphorylation of the downstream S6K1 and S6 was completely abolished (Fig. 6D). AKT phosphorylation was increased after short-time exposure to rapamycin (Fig. 6E,F). Interestingly, phosphorylation events of downstream components of Ras signaling, e.g., MEK and JNK/SAPK were also enhanced by rapamycin (Fig. 6E), suggesting a possible negative-feedback loop on Ras signaling by mTOR-S6K1 in hepatocytes. However, phosphorylation of NF-κB was not affected by rapamycin (Fig. 6E).
Finally, we explored the effect of AKT-PI3K-mTOR pharmacologic inhibitors. As shown in Fig. 6F, ATP-stimulated S6K1 phosphorylation could be completely blocked by inhibitors to AKT (AKT-IV) or PI3K (LY294002, wortmannin, PI-103, PI3-Kα, PI3-Kγ), but not by inhibitors to PDK (PDK1) nor to PKC (RO-31-8220). In parallel, AKT phosphorylation was completely abrogated by inhibitors to PDK, PI3K, or PKC. These data indicate that PI3K-AKT is upstream of ATP-stimulated mTOR-S6K1 signaling, whereas PDK-AKT and PKC-AKT signaling are not involved in this process.
We infer that abnormal responses in Cd39-null hepatocytes resulting from disordered purinergic signaling are, at least in part, mediated by way of mTOR signaling (as illustrated in Fig. 7).
In this work, we show that CD39/ENTPD1 modulates crucial cell metabolic and proliferative elements in vitro and impacts cellular transformation that is linked with development of liver cancer in vivo (as illustrated in Fig. 7). We implicate disordered purinergic signaling in the evolution of both induced and spontaneous liver cancer.
The model by which disordered purinergic signaling promotes hepatocarcinogenesis may involve the following molecular mechanisms. First, heightened levels of extracellular ATP (eATP) provoke hepatocyte dysfunction, as observed in Cd39-null mice, in keeping with previous studies.4, 5, 25 eATP levels are determined by release from stressed, activated inflammatory cells or injured parenchymal elements, and by altered expression of ectonucleotidases. In turn, eATP-initiated responses are mediated by diverse hepatocellular P2 receptors. We show that these effects are blocked by a global P2 receptor antagonist suramin (Fig. 2F) and can further implicate P2Y2 in this process (Fig. S2B).
We also establish the possible role of mTOR in eATP-modulation of intracellular ATP (iATP) levels. Higher levels of eATP result in constitutive stimulation of the mTOR-S6K1 pathway in liver cells, which may incorrectly imply abundant iATP, despite these levels being decreased in these cells (Fig. 4E; Table S2).
Second, we link purinergic signaling responses to deviation of cellular metabolic pathways that support rapid cell proliferation. Direct tyrosine phosphorylation of PKM2 by FGFR1 has been recently identified as a key mechanism promoting glycolysis and tumor growth.24 Here we show that purinergic signaling modulates tyrosine phosphorylation as well as expression of PKM2 in proliferating hepatocytes (Fig. 4A,B). We have also previously demonstrated that inhibition of LDH-A, another essential glycolytic enzyme frequently overexpressed in cancer, limits tumor growth.20 We now show LDH-A expression to be upregulated by ATP-stimulated purinergic signaling in an mTOR-dependent manner (Figs. 4C, 6C).
Cunningham et al.26 have shown that mTOR controls mitochondrial oxidative function by formation of a transcriptional complex with YY1 (Ying Yang 1)-PGC-1α (peroxisome-proliferator activated receptor coactivator-1α). Inhibition of mTOR resulted in global suppression of mitochondrial gene expression (such as PGC-1α, PGC-1β, LDH-A, NRF-1, and UCP2). We found, however, that rapamycin decreases gene expression of LDH-A and UCP2 (Fig. 6C), both known to be expressed at high levels and associated with metabolic reprogramming in various cancers, while increasing the expression of all other mitochondrial genes/regulators (Fig. 6C; Fig. S5). That our results differ from these of the prior study might be explained by the very different nature of the cell phenotypes studied (which are skeletal myocytes and myoblasts versus primary hepatocytes).
Third, we further define the role of up- and downstream effectors of mTOR signaling in hepatocarcinogenesis. The mTOR pathway is often up-regulated in many human cancers inclusive of HCC.1 Rapamycin-related compounds demonstrate antitumor efficacy in a wide range of human malignancies.27 Recently, a novel mTORC1-MAPK feedback loop (mediated by way of the S6K1-PI3K-Ras-ERK pathway) has also been identified in both normal cells and cancer cells.28 Rapalog exposure unfortunately disrupts crucial negative feedback mechanisms and results in subsequent activation of PI3K-AKT and/or PI3K-MAPK pathways. Therefore, antitumor efficiency has been shown to be enhanced by inhibiting mTOR and PI3K pathways in parallel.29 Hence, targeting upstream purinergic signaling, as well as mTOR and PI3K, might have utility in treating cancer patients.
Fourth, autophagy is a fundamental catabolic process to maintain cellular homeostasis by sustaining protein and organelle quality control, the regulation of which has promising chemotherapeutic potential. This is an antitumor mechanism linked to various cancers, including HCC.23, 30 In this study, we provide evidence that mTOR-mediated suppression of autophagy is modulated through purinergic signaling pathways, in response to extracellular nucleotides and further regulated by CD39 in a tightly controlled manner (Figs. 3, 6B).
Recent studies have also noted that murine hepatocytes enter a senescence program triggered by excessive proliferative signaling, which has features (at least in part) of the cellular phenotype observed in these current studies in Cd39-null hepatocytes.31 Senescent liver cells are subject to surveillance and immune clearance impacting development of cancer. These aspects may also be abnormal in Cd39-null mice, as we have documented NK and NKT cell dysfunction.18, 32
CD39L4/ENTPD5, another ENTPD/ectonucleotidase family member, has been recently linked to induction of glycolytic metabolism and survival of transplanted tumors.33 However, mice null for this ectoenzyme exhibit heightened incidence of primary liver neoplasms, for unclear reasons.34 These comparable features might be associated with the aberrant metabolic effects following deletion of Cd39 that we describe here.
Further defining the components of purinergic signaling pathways that initiate and promote tumor formation will be critical for the development of effective prevention and intervention strategies. A better understanding of the complexity of mTOR signal transduction (inclusive of negative S6K1-feedback loops and impact of eATP) will be required to understand the biological consequences of inhibiting specific components of these cellular regulatory networks.
In summary, we describe a small animal model for autochthonous hepatocellular cancer in the noncirrhotic liver, with distinct outcomes as those seen in acute inflammatory responses to transplanted tumors. We have characterized novel purinergic mechanisms that regulate mTOR signaling in proliferating liver cells. These pathways are associated with malignant cell transformation resulting in development of liver cancer in mice. Further experimental dissection of the role of purinergic mediators in hepatocellular transformation might impact adjunctive therapies in this devastating disease.
- 12Metabolism, genomics, and DNA repair in the mouse aging liver. Curr Gerontol Geriatr Res 2011; 2011: 859415., , .
Additional Supporting Information may be found in the online version of this article.
|HEP_25989_sm_SuppFig1.tif||2193K||Supporting Information Figure 1.|
|HEP_25989_sm_SuppFig2.tif||1425K||Supporting Information Figure 2.|
|HEP_25989_sm_SuppFig3.tif||964K||Supporting Information Figure 3.|
|HEP_25989_sm_SuppFig4.tif||2118K||Supporting Information Figure 4.|
|HEP_25989_sm_SuppFig5.tif||2251K||Supporting Information Figure 5.|
|HEP_25989_sm_SuppFig6.tif||983K||Supporting Information Figure 6.|
Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.