Copper Transport Protein Antioxidant-1 Promotes Inflammatory Neovascularization via Chaperone and Transcription Factor Function

Copper (Cu), an essential micronutrient, plays a fundamental role in inflammation and angiogenesis; however, its precise mechanism remains undefined. Here we uncover a novel role of Cu transport protein Antioxidant-1 (Atox1), which is originally appreciated as a Cu chaperone and recently discovered as a Cu-dependent transcription factor, in inflammatory neovascularization. Atox1 expression is upregulated in patients and mice with critical limb ischemia. Atox1-deficient mice show impaired limb perfusion recovery with reduced arteriogenesis, angiogenesis, and recruitment of inflammatory cells. In vivo intravital microscopy, bone marrow reconstitution, and Atox1 gene transfer in Atox1−/− mice show that Atox1 in endothelial cells (ECs) is essential for neovascularization and recruitment of inflammatory cells which release VEGF and TNFα. Mechanistically, Atox1-depleted ECs demonstrate that Cu chaperone function of Atox1 mediated through Cu transporter ATP7A is required for VEGF-induced angiogenesis via activation of Cu enzyme lysyl oxidase. Moreover, Atox1 functions as a Cu-dependent transcription factor for NADPH oxidase organizer p47phox, thereby increasing ROS-NFκB-VCAM-1/ICAM-1 expression and monocyte adhesion in ECs inflamed with TNFα in an ATP7A-independent manner. These findings demonstrate a novel linkage between Atox1 and NADPH oxidase involved in inflammatory neovascularization and suggest Atox1 as a potential therapeutic target for treatment of ischemic disease.

Copper (Cu), an essential micronutrient and catalytic cofactor, plays an important role in physiological repair processes including wound healing and angiogenesis as well as various pathophysiologies including tumor growth and inflammatory diseases such as atherosclerosis [1][2][3][4][5] . Cu directly stimulates cell proliferation and migration in cultured ECs 3 . Furthermore, angiogenic lesions in cancer and atherosclerosis have higher Cu levels in cell nuclei than normal tissues 6 . In vivo, Cu induces neovascularization and its concentration is increased in angiogenic tissue 6 , while Cu chelators, which have been developed to treat Wilson disease (a disease of Cu toxicity), inhibit tumor growth and angiogenic responses [1][2][3][4] . Of importance, clinical trials for the treatment of solid tumors by Cu chelation show efficacy in disease stabilization 7 . Cu also plays an important role in inflammatory responses involved in innate and adaptive immunity 8,9 , which is in part via activating NFκ B 10,11 . Cu deficiency alters intravascular adhesion of leukocytes to the activated ECs and expression of adhesion molecules, such as ICAM-1/VCAM-1 12,5,13 . Despite a critical role of Cu in angiogenesis and inflammation, the molecular mechanisms that underlie these functions are entirely unknown.
Since excess Cu is toxic, bioavailable Cu level is tightly controlled by Cu transport proteins including cytosolic Cu chaperone Antioxidant-1 (Atox1) which obtains Cu via Cu importer CTR1 and then transfers Cu to the Cu transporter ATP7A that delivers Cu to the secretory Cu enzymes for their activity, or exclude Cu 14,15 . Atox1 contains the conserved Cu-binding motif (MTCXXC) and Atox1-deficienct mice show reduced secretory Cu enzyme activity 16 . We previously demonstrated a novel function of Atox1 as a Cu-dependent transcription factor to regulate cell growth 17 . Indeed, Atox1 has two conserved lysines residues (Lys56 and Lys60) which are nuclear localization signal (NLS), and it is shown to be localized in the nucleus in vitro 17,18 and in vivo 19 . We have shown that Atox1 binds to DNA at cis elements (GAAAGA; Atox1-responsive elements)(Atox1-RE) in the promoters and induces transactivation 17 . Cu chelator TTM, which blocks Atox1 function 20 , prevents tumor progression in clinical trials 1,2,4 .
We performed the present study to examine the role of Cu transport protein Atox1 in ischemia-induced neovascularization which is dependent on angiogenesis/arteriogenesis, inflammation, and reactive oxygen species (ROS) 21 . Here we demonstrate that Atox1 plays an essential role in inflammatory neovascularization via Cu chaperone function to increase angiogenesis via regulating extracellular matrix (ECM) modifying secretory Cu enzyme lysyl oxidase (LOX) activity as well as transcription factor function to promote ROS production and inflammatory responses via increasing NADPH oxidase organizer p47phox expression. We used Atox1 −/− mice with hindlimb ischemia model and sponge implant model, and various approaches including micro-CT, in vivo intravital microscopy, gene transfer, bone marrow transplantation (BMT), in situ O 2 − detection, bioluminescence imaging of NFκ B reporter mice, and in vitro reporter gene assay, and chromatin immunoprecipitation (ChIP) assay. These findings demonstrate a novel linkage between Cu transport protein Atox1 and NADPH oxidase involved in inflammatory angiogenesis and provide insight into Atox1 as a potential therapeutic target for inflammation-dependent ischemic cardiovascular diseases.

Atox1 in tissue resident cells is required for ischemia-induced neovascularization.
To gain insight into the role of Atox1 in ischemia-induced neovascularization, we first looked at Atox1 expression in mice and patients with CLI. Atox1 protein was significantly increased after hindlimb ischemia of mice (Fig. 1A) and it was expressed in both ECs and inflammatory cells ( Supplementary  Fig. 1). Consistently, ischemic limb muscle samples from CLI patients showed approximately 40-fold increase in Atox1 mRNA as compared to healthy subjects (p < 0.01) (Supplementary Fig. 2) (For detailed patient characteristics see Supplementary Table 1). To examine the role of endogenous Atox1 in post-ischemic neovascularization, wild-type (WT) and Atox1 −/− mice were subjected to unilateral hindlimb ischemia. All Atox1 −/− mice survived after ischemic surgery. Figure 1B showed that the blood flow recovery after ischemia was significantly reduced in Atox1 −/− mice as compared to WT mice. Of note, ischemic foot showed increased toe necrosis in Atox1 −/− mice ( Fig. 1B) with enhanced muscle damage and delayed repair at day 14 (data not shown). To examine the relative role of Atox1 in tissue resident cells and BM cells in post-ischemic neovascularization, we performed BMT. Lethally irradiated Atox1 −/− mice transplanted with WT-BM, but not irradiated WT mice reconstituted with Atox1 −/− BM, showed impaired limb perfusion recovery compared to control group (WT mice reconstituted with WT-BM) (Fig. 1C). In addition, direct injection of adenovirus expressing Atox1 (Ad. Atox1) into Atox1 −/− ischemic limbs significantly improved blood flow recovery compared to Atox1 −/− mice treated with Ad.LacZ (control) (Fig. 1D). These results suggest that Atox1 in tissue resident cells is required for ischemia-induced revascularization.
Atox1 is required for ischemia-induced arteriogenesis and angiogenesis. We next examined a role of Atox1 in arteriogenesis and angiogenesis, which contributes to ischemia-induced neovascularization. Micro-CT analysis at day 7 after ischemia showed significant decrease in the number of arterial density in Atox1 −/− ischemic limbs ( Fig. 2A). Moreover, immunohistochemical analysis showed that the numbers of CD31 + ECs (Fig. 2B) and α SMA + arterioles (Fig. 2C) in ischemic muscles were significantly decreased in Atox1 −/− mice.

Figure 1. Atox1 in tissue resident cells, but not BM cells, is required for ischemia-induced neovascularization.
(A) Atox1 protein expression in ischemic and non-ischemic muscles from wild type (WT) and Atox1 knockout (KO) mice at day 7 post-surgery. Actin is loading control (n = 5). (B) Blood flow recovery after hindlimb ischemia in WT and Atox1 KO mice, as determined by the ratio of foot perfusion between ischemic (left) and non-ischemic (right) legs. Bottom panels show representative laser Doppler images (white arrows shows ischemic foot) and toe necrosis characterized by edematous fingers and degenerative nail beds (black arrows) in Atox1 KO mice (n = 10). (C) Bone marrow transplantation (BMT) showing a role of Atox1 in tissue resident cells for ischemia-induced neovascularization. After 6 weeks of BMT, mice were subjected to hindlimb ischemia and limb blood flow was measured (n = 6). (D) Atox1 gene transfer in Atox1 KO mice rescues blood flow recovery after hindlimb ischemia. Purified adenoviruses (Ad. Atox1-WT, Ad. LacZ (control), 1 × 10 9 pfu) were injected into the adductor and gastrocnemius muscles in Atox1 KO mice at one day prior to ischemic injury, and limb blood flow was measured. Representative laser Doppler images (white arrows shows ischemic foot) and quantitative analysis are shown (C,D). **p < 0.01, ***p < 0.001.  assay was performed by implanting polyvinyl alcohol sponge containing VEGF subcutaneously into WT and Atox1 KO mice. Representative images for H&E staining and isolectin immunostaining for blood vessel formation in sponges harvested on day 21. Right panels show quantitative analysis of the number of red blood cells (RBC) and isolection+ ECs. Scale bars = 50μ m. (B) HUVECs were transfected with control, Atox1 or ATP7A siRNAs or treated with Cu chelator TTM (20 nM, 24 hrs) and seeded on Matrigel-coated plates in culture media containing VEGF for 6 h. Four random fields per well were imaged, and representative pictures are shown (left). Averaged numbers of capillary tube branches, branching points, and tube length per field are shown (Right). (C) HUVECs were treated with LOX inhibitor β -aminopropionitrile (BAPN, 100 μ M) for 24 hrs and capillary tube formation was measured (n = 3). (D) Activity of LOX was measured in ischemic gastrocnemius muscle of WT and Atox1 KO mice (left)(n = 4) or in culture medium from VEGF (20ng/ml)-stimulated HUVECs transfected with siControl or siAtox1 (right)(n = 4). *p < 0.05.
Scientific RepoRts | 5:14780 | DOi: 10.1038/srep14780 Atox1 promotes angiogenesis via activating Cu enzyme LOX in ECs in an ATP7A-dependent manner. To demonstrate further a role of Atox1 in angiogenesis, we used a mouse sponge implant model 22 . Figure 3A showed that Atox1 −/− mice exhibited marked decrease in new blood vessel formation assessed by number of red blood cells and isolectin positive capillaries 23 . To assess the mechanism by which Atox1 is involved in angiogenesis, we first examined a role of Atox1 in cell proliferation, because we previously showed that Atox1 functions as a Cu-dependent transcription factor for cyclin D1 to regulate cell growth 17 . However, either BrdU labeling or cyclin D1 expression was not altered in Atox1 −/− mice ( Supplementary Fig. 3). We then examined a role of Atox1 in VEGF-induced capillary-network formation, an in vitro model of angiogenesis. Figure 3B showed that depletion of Atox1 using siRNA in human umbilical vein endothelial cells (HUVECs) significantly reduced the number of capillary tube branches, branching sprouts as well as tube length. Similar inhibitory effects were observed by Cu chelator TTM or siRNAs for Cu importer CTR1 ( Supplementary Fig. 4) or Cu transporter ATP7A.
Cu chaperone function of Atox1 is required for activating secretory Cu enzymes such as ECM-modifying Cu enzyme LOX which is mediated through Cu delivery via ATP7A. Thus, we next examined a role of LOX in capillary tube formation. Figure 3C showed that a specific LOX activity inhibitor, aminopropionitrile (BAPN), significantly reduced the number of capillary tube branches, branching sprouts as well as tube length. Moreover, LOX activity was markedly increased in ischemic tissues as well as VEGF-stimulated ECs, which were inhibited by Atox1-/-mice and Atox1 siRNA-treated HUVECs, respectively (Fig. 3D). These findings suggest that Cu chaperone function of Atox1 is involved in VEGF-induced angiogenesis via activating LOX through the CTR1-Atox1-ATP7A pathway 14 in ECs.

Atox1 is required for inflammatory cell recruitment to the ischemic sites in vivo.
Since ischemia-induced neovascularization also depends on inflammation, we examine the role of Atox1 in inflammatory cells recruitment. Figure 4A showed that the numbers of Gr-1+ neutrophils and Mac-3+ macrophages infiltrated in the ischemic muscles were significantly decreased in Atox1 −/− mice. This was associated with reduced expression of inflammatory cytokines TNFα and IL-6 ( Fig. 4B) as well as VEGF protein (Fig. 4C) which were mainly secreted by the infiltrated macrophages 24 . By contrast, VEGF receptor2 expression was not changed in ischemic tissue of Atox1 −/− mice (data not shown). Atox1 gene transfer into Atox1 −/− mice increased macrophages infiltration to the ischemic sites ( Supplementary Fig.  5A) and reduced muscle necrosis (not shown), thereby restoring post-ischemic limb perfusion (Fig. 1D). Moreover, sponge implant model also showed that macrophage infiltration was significantly inhibited in Atox1 −/− mice ( Supplementary Fig. 5B). These results suggest that Atox1 in tissue resident cells plays an important role in inflammatory cell recruitment, and thus promoting neovascularization.
Since inflammatory cell accumulation into sites of injury is dependent on leukocytes adhesion to ECs, we examined a role of Atox1 for neutrophil adhesion to the inflamed ECs in vivo using real-time intravital microscopy with a mouse inflammation model 25 (Fig. 4D and Supplementary Video 1). Mouse neutrophils were monitored by infusion of the Alexa Fluor 647-labeled anti-Gr-1 antibody 26 . Neutrophil adhesion was increased in inflamed venule stimulated with TNFα , which was significantly inhibited in Atox1 −/− mice. By contrast, the neutrophils rolling influx over the inflamed venule wall was increased in Atox1 −/− mice compared to WT mice while there was no difference of the rolling velocity of neutrophils between WT and Atox1 −/− mice. Furthermore, the number of detaching neutrophils after firm adhesion over 30 sec increased to 3.8-fold in Atox1 −/− mice. These results suggest that Atox1 plays an important role in neutrophil adhesion to the inflamed endothelium, but not in neutrophil rolling in vivo.
Atox1 is required for monocyte adhesion to inflamed ECs in a Cu dependent, but ATP7A-independent manner. To determine the mechanism by which Atox1 in ECs promotes inflammatory cell recruitment, we examined a role of Atox1 in monocytes adhesion to inflamed ECs in vitro using THP-1 monocytes and cultured HUVECs. Figure 4E showed that adhesion of THP-1 cells to HUVEC activated by inflammatory cytokine TNFα was significantly inhibited in Atox1-depleted ECs with siRNA or ECs treated with Cu chelator BCS or Cu importer CTR1 siRNA, but not by Cu transporter ATP7A siRNA. These results suggest that Atox1 is involved in monocyte adhesion to the activated ECs in a Cu-CTR1-dependent manner but independent of Cu chaperone function of Atox1 which requires ATP7A.
Atox1 is required for adhesion molecules expression in a Cu-dependent manner via activating NFκB in inflamed ECs. Leukocyte/monocyte adhesion to the activated ECs is mediated through adhesion molecules ICAM-1 and VCAM-1 27,28 . Figure 5 showed that ischemia-induced expression of ICAM-1 and VCAM-1 mRNAs (Fig. 5B) and VCAM-1 protein in the endothelial layers of vessels (  Fig. 7A). These results suggest that Atox1 is involved in upregulation of ICAM-1/VCAM-1 expression in inflamed ECs, which may promote inflammatory cell recruitment to the ischemic tissue. Since transcription of ICAM-1/VCAM-1 is dependent on NFκ B 27,28 , we next examined a role of Atox1 in NFκ B activation. Atox1 siRNA did not affect TNFα -induced degradation of Iκ B, which is involved in activation of NFκ B  (C) HUVECs transfected with Atox1 or control siRNAs, or treated with BCS were stimulated with TNFα (10ng/ml), and ICAM-1 and VCAM-1 mRNAs were measured (n = 12). (D) HUVECs transfected with Atox1 or control siRNAs were incubated with TNFα for 18 hours, and VCAM-1 and ICAM-1 protein expression was measured. Actin was loading control (n = 8). (E) ChIP assay showing a role of Atox1 in TNFα -induced p65 NFkB binding to the ICAM-1/VCAM-1 promoter in vivo. HUVECs transfected with Atox1 or control siRNAs stimulated with TNFα were precipitated with the anti-p65NFkB antibody. The ICAM-1/VCAM-1 promoter region was amplified by PCR. Input of nuclear DNA was used as PCR control. (n = 4). (F) In vivo bioluminescence imaging of NFkB reporter mice (HLL mice) and HLL mice crossed with Atox1 KO mice before and after hindlimb ischemia at day 3. (n = 3). Representative images (left) and quantification of bioluminescence intensity (right). (G) Luciferase activity in homogenates of non-ischemic and ischemic gastrocunemious tissue from HLL and HLL/Atox1 KO mice at day 3 (n = 5). *p < 0.05, **p < 0.01, ***p < 0.001 (data not shown). By contrast, ChIP assays showed that siAtox1 blocked NFκ B subunit p65/RelA binding to the ICAM-1/VCAM-1 promoters in HUVECs stimulated by TNFα (Fig. 5E). Furthermore, in vivo Bioluminescence imaging of NFκ B reporter mice (HLL mice) demonstrated that NF-kB activation induced by ischemia was blocked in HLL mice crossed with Atox1 −/− mice compared to control HLL mice (Fig. 5F). In parallel, ex vivo luciferase assay of ischemic and non-ischemic muscles confirmed that ischemia-induced NFκ B activity was significantly inhibited in HLL-Atox1 −/− mice (Fig. 5G).

Atox1 is required for endothelial ROS production in ischemic tissues in vivo and TNFα-induced
ROS production in ECs in a Cu-dependent and ATP7A-independent manner. Since ICAM-1 and VCAM-1 are redox sensitive genes and ROS are involved in NFκ B activation 27,28 , we then examined a role of Atox1 in ROS production. Figure 6A showed that injection of O 2 •− specific dye, DHE 29 into the mice revealed that extent of decrease in DHE + /CD31 + cells (~90%) was much higher than that of CD31 + cells (~54%) in Atox1 −/− mice compared to WT mice, suggesting that Atox1 is involved in ischemia-induced endothelial ROS production. To confirm this finding in vitro, we next examined a role of Atox1 in H 2 O 2 production in cultured ECs using DCF-DA which mainly detects intracellular H 2 O 2 30 . Figure 6B showed that TNFα -induced increase in H 2 O 2 production was blunted by siRNAs for Atox1 or CTR1, or PEG-catalase or Cu chelator BCS, but not by ATP7A siRNA. These results suggest that Atox1 is involved in ROS production in inflamed ECs in a Cu-dependent manner, but independent of Cu chaperone function through ATP7A.

Atox1-mediated p47phox transcription is required for adhesion molecule expression in a Cudependent manner in ECs.
To determine the mechanism by which Atox1 is involved in ROS production in ECs, we examined a role of Atox1 in expression of NADPH oxidase, a major source of ROS involved in angiogenesis 31 . We found that TNFα -induced p47phox mRNA in HUVECs was almost completely inhibited by siRNAs for Atox1 or CTR1 or treatment with BCS ( Fig. 7A and Supplementary Fig. 6) while Atox1 siRNA did not reduce p22 phox mRNA (Fig. 7A) or Nox2 or Nox4 mRNA ( Supplementary  Fig. 8). We confirmed that TNFα -induced p47phox protein was inhibited in Atox1 si-treated HUVECs (Fig. 7B) or Atox1 −/− ECs (Fig. 7C), and that ischemia-induced p47phox protein was markedly reduced in Atox1 −/− mice ( Supplementary Fig. 7B). Of importance, inhibition of TNFα -induced VCAM-1 expression in Atox1 −/− ECs was rescued by re-expression of p47phox-EGFP (Fig. 7D), supporting the specificity of the Atox1-p47phox axis for increasing adhesion molecule expression.

Atox1 functions as a Cu dependent transcription factor for p47phox in ECs. Since Atox1
was involved in ROS production and monocyte adhesion to inflamed ECs in a Cu-dependent but Cu chaperone-independent manner (Figs 4 and 6), we examined whether Cu-dependent transcription factor function of Atox1 is involved in p47phox transcription in ECs. Immunofluorescence analysis revealed that Atox1 was localized in the cytosol in basal state and TNFα stimulation promoted Atox1 translocation into the nucleus (Fig. 8A), which was blocked by siCTR1 or BCS ( Fig. 8A and Supplementary Fig. 9). Since p47phox promoter contains several Atox1-REs (GAAAGA) 17,32 at − 305, and − 110 bp upstream of the transcription start site, we performed luciferase reporter assays using p47phox promoters. Figure 8B showed that p47phox promoter activity was significantly increased by overexpression of WT-Atox1, but not by Atox1 mutated at Cu-binding sites (C12, C15S) or nuclear localization signal motif (K56, 60E). A 5′ -deletion analysis of the p47phox promoter identified functional Atox1-REs from − 110 to − 105 bp at the promoter, which had no response to the Atox1 mutants (Fig. 8B). Furthermore, mutation or deletion of the GAAAGA sequence (− 110/− 105) in the p47phox promoter abolished TNFα -induced activation of p47phox promoter (Fig. 8C). To test whether endogenous Atox1 binds to the p47phox promoter in vivo, we performed ChIP assays using specific Atox1 antibody in HUVECs (Fig. 8D). Atox1 associated with the p47phox promoter including Atox1-RE at − 110 in a time-dependent manner in ECs stimulated by TNFα , which was abolished by siAtox1. These results suggest that Atox1 binds to the p47phox promoter at Atox1-RE at − 110 in response to TNFα , thereby promoting H 2 O 2 production that stabilizes p65/RelA binding to ICAM-1/VCAM-1 promoters, which in turn increases adhesion molecule expression and leukocytes/monocytes adhesion in ECs.

Discussion
Role of Cu in angiogenesis and inflammatory responses has been appreciated [1][2][3]8,9 ; however, its molecular mechanisms are entirely unknown. Here we demonstrate that Cu transport protein Atox1 plays an essential role in inflammatory neovascularization via Cu chaperone function through Cu transporter ATP7A to increase angiogenesis via regulating secretory Cu enzyme LOX mediated activity as well as Cu-dependent transcription factor function to promote ROS production and inflammatory responses via increasing p47phox expression in an ATP7A-independent manner. The following key observations support this conclusion. First, we observed that Atox1 expression is significantly upregulated in patients and mice with CLI. Second, mice lacking Atox1 exhibit impaired ischemia-induced perfusion recovery, arteriogenesis/angiogenesis and infiltration of inflammatory cells that produce angiogenic cytokines including VEGF and TNFα . Third, BMT, Atox1 gene transfer, in vivo intravital microscopy experiments reveal that Atox1 in ECs, but not BM cells, is required for post-ischemic revascularization and recruitment of inflammatory cells. Fourth, Atox1 functions as a Cu chaperone to promote VEGF-induced angiogenesis in ECs via increasing LOX activity in a Cu/ATP7A-dependent manner. Fifth, Atox1 translocates to nucleus in response to TNFα , resulting in binding to p47phox promoter for its transactivation, thereby promoting the ROS/NFkB-dependent ICAM-1/VCAM-1 expression in ECs in an ATP7A-independent manner, which facilitates recruitment of inflammatory cells. Thus, our study demonstrates a novel linkage  between Atox1 and NADPH oxidase as well as positive feed-back loops that promote inflammatory neovascularization organized by a Cu chaperone and transcription factor function of Atox1 (Fig. 8E).
In this study, we show that Atox1 expression is significantly upregulated in mice and patients with CLI, supporting the clinical significance of Atox1 in peripheral vascular disease. Ischemia hindlimb model with micro-CT analysis and sponge implant assay reveal that Atox1 is involved in ischemia/ inflammation-induced angiogenesis and arteriogenesis. We previously reported that Atox1 functions as a Cu-dependent transcription factor to stimulate cell growth via regulating cyclin D1 in fibroblast in an ATP7A-independent manner 17 . However, BrdU staining and cyclin D1 expression in ischemic tissues are not altered in Atox1 −/− mice. The discrepancy between our previous and current study may be attributable to differences in cell types or experimental models. We thus examined if Cu chaperone function of Atox1 which depends on Cu transporter ATP7A that delivers Cu to the secretory Cu enzymes 14,15 , is involved in angiogenesis. We found that activity of Cu enzyme LOX, which is involved in angiogenesis 33,34 , is markedly decreased in ischemic tissue of Atox1 −/− mice or Atox1-depleted HUVECs. VEGF-induced capillary tube formation in ECs is blocked by Cu chelator or siRNAs for CTR1 or Atox1 or ATP7A; or specific LOX inhibitor. Furthermore, we previously reported that Atox1 35 or CTR1-ATP7A-LOX axis 35,36 are involved in PDGF-induced vascular smooth muscle migration, which may contribute to arteriogenesis and collateral formation 37 . Thus, these results suggest that Atox1 plays an important role in VEGF-induced angiogenesis in ECs and post-ischemic angiogenesis/arteriogenesis at least via Cu chaperone function to increase LOX activity through the CTR1-Atox1-ATP7A pathway.
Cu deficiency is shown to inhibit vascular adhesion of leukocytes to the activated ECs 12 and pathological angiogenesis 1-4 but its mechanism remains unclear. Infiltrated inflammatory cells after ischemia are the major source of angiogenesis growth factors and cytokines, thereby promoting revascularization 24 . Here we show that ischemia-induced macrophage and neutrophil infiltration is inhibited in Atox1 −/− mice, which is associated with decreased expression of TNFα and VEGF. Experiments with BM transplantation and Atox1 gene transfer to Atox1 −/− mice of ischemic muscles reveal that Atox1 in tissue resident cells such as ECs is important for restoring limb perfusion and macrophage recruitment. This conclusion is further supported by the in vivo intravital microscopy study showing that exogenous labeled Atox1 + neutrophil adhesion to the TNFα -inflamed endothelium is inhibited in Atox1 −/− mice. Moreover, we previously reported that BM-derived macrophage migration is not inhibited in Atox1 −/− mice 35 . Thus, our data are consistent with the conclusion that Atox1 in ECs, but not BM cells, is required for recruitment of inflammatory cells that secrete TNFα and VEGF, thereby enhancing neovascularization and perfusion recovery in ischemic limbs.
Previous reports suggest that Cu plays an important role in activation of redox-sensitive transcription factor NFκ B and induction of its downstream genes including adhesion molecules ICAM-1/VCAM-1 which mediate inflammatory cell adhesion to ECs 10,38,39,5,13 . However, its underlying mechanism is entirely unknown. Here we provide the first evidence that Atox1 is involved in ROS production, NFκ B-DNA binding, VCAM-1/ICAM-1 expression and monocyte adhesion to ECs stimulated by TNFα in a Cu/ CTR1-dependent manner. Of note, these Atox1-mediated inflammatory responses are independent of ATP7A, suggesting that Cu chaperone function of Atox1 is not involved. Role of Atox1 in promoting ROS and NFκ B activity is further confirmed by in vivo studies demonstrating a significant decrease in DHE + /CD31 + ECs in ischemic tissues as well as NFκ B activity measured by bioluminescence imaging of NFκ B-reporter mice. Consistent with our result, we previously used EC-specific catalase transgenic mice to demonstrate that EC-derived H 2 O 2 plays an essential role in ischemia-induced arteriogenesis/angiogenesis, NFκ B-dependent adhesion molecule expression and inflammatory cells recruitment 40 . By contrast, Tirziu et al. 41 reported that EC-specific overexpression of inhibitor of NFκ B in mice show reduced limb perfusion and adhesion molecule expression and monocyte influx in ischemic tissues with excess branched arterial network and immature vessels. The discrepancy regarding arteriogenesis might be due to the difference between overexpression of IkBα mutant and endogenous ROS inhibition, or there are possible additional downstream targets of Atox1 than p47phox NAPDH oxidase, as discussed below.
We previously reported that Atox1 functions not only as a Cu chaperone but also as a Cu-dependent transcription factor by binding to DNA at Atox1-RE 17,32 . The present study demonstrates that Atox1 functions as a Cu-dependent transcription factor for p47phox, a cytosolic organizer of NADPH oxidase in ECs activated by TNFα , thereby increasing ROS production and redox-sensitive NFκ B-dependent ICAM-1/VCAM-1 expression, which promotes monocytes adhesions to ECs. It is shown that p47phox is involved in adhesion molecule expression and inflammatory cell adhesion in ECs 27,42-44 and that NADPH oxidase plays an important role in post-ischemic neovascularization in context-and disease state-specific manner 23,[45][46][47][48] . In this study using mmunofluorescence, luciferase, and ChIP assays, we found that Atox1 translocates to nucleus in response to TNFα , and then bind to the Atox1-RE 17,32 in the p47phox promoter, in a Cu/CTR1-dependent manner. Moreover, overexpression of Atox1, but not Atox1 mutants lacking the Cu binding or nuclear translocation sequence, increases p47phox promoter activity. p47phox protein expression is significantly decreased in Atox1 −/− ECs or Atox1 siRNA-treated HUVECs stimulated by TNFα as well as ischemic tissues of Atox1 −/− mice. Among NADPH oxidase components, p47phox seems to be the specific target of Atox1, since expression of other subunits such as p22phox and Nox2 as well as Nox4 are not affected by siAtox1 in TNFa-stimulated ECs, and the promoter region of p47phox, but not other components of NADPH oxidase such as p22phox or Nox2, contains Atox1-RE regions at − 305, and − 110 bp upstream of the transcription start site. Importantly, Scientific RepoRts | 5:14780 | DOi: 10.1038/srep14780 decrease in TNFα -induced VCAM-1 expression in Atox1 −/− ECs is rescued by re-expression of p47phox, supporting the specificity of the Atox1-p47phox-VCAM-1 axis.
It is shown that VEGF-induced ROS play an important role in angiogenesis in ECs 31 , and that VEGF stimulation increases the expression of ICAM-1/VCAM-1 49 . However, we found that Atox1 translocation to nucleus is not observed by VEGF or other growth factors such as FGF or HGF, and that Atox1siRNA has no effects on VEGF-induced ROS production or ICAM-1/VCAM-1 expression in ECs ( Supplementary Figs 10-12). As discussed above, Atox1 is involved in recruitment of inflammatory cells that produce VEGF in ischemic tissues (Fig. 4C). Thus, it is conceivable that Atox1 may be upstream of VEGF signaling linking to ROS-dependent induction of ICAM-1/VCAM-1 during ischemia-induced neovascularization, which may explain why expression of these adhesion molecules is completely suppressed in Atox1 −/− ischemic tissues (Fig. 8E). Moreover, in ECs stimulated with VEGF, Atox1 functions as a cytosolic Cu chaperone for LOX through ATP7A, instead of Cu-dependent transcription factor, to promote angiogenesis (Fig. 8E). Addressing molecular mechanism by which Atox1 translocates to the nucleus in response to TNFα specifically or other inflammatory cytokines is the subject of future study.
p47phox transcription is reported to be negatively regulated by binding of HBP1 to the p47phox promoter at − 1300bp 50 , while it is positively regulated by binding of Ets family protein PU.1 to the p47phox promoter at − 40bp 51 . Thus, these transcription factors may also participate in regulating Atox1-mediated p47phox transcription. In addition, we previously reported that Atox1 acts as a Cu-dependent transcription factor for secretory copper enzymes extracellular SOD (ecSOD) in vascular smooth muscle 19,32 , which plays a role in ischemia-induced reparative neovascularization by regulating extracellular ROS levels 30,52 . HIF1α is also Cu-dependent 53 . It is thus conceivable that Cu-dependent transcription factor Atox1 may have additional downstream targets to promote neovascularization. Atox1 functions as a Cu chaperone for ecSOD 32 to increase its activity via binding to ATP7A 14 . However, since ecSOD is not/little expressed in cultured ECs 54 , ecSOD may not be involved in Atox1-mediated angiogenesis and inflammatory responses in ECs. Moreover, Cu-dependent angiogenesis may be mediated through increased binding of growth factors to cell surface or secretion of proangiogenic peptides such as FGF1 and IL-1α , or stabilization of extracellular matrix such as fibronectin, which are shown to be regulated by Cu 1-3 .
In summary, our study will enrich our understanding about the molecular mechanisms of how Cu is involved in inflammation and angiogenesis-dependent diseases. We provide the novel evidence that Atox1 plays an essential role in inflammatory neovascularization via Cu chaperone function for LOX to increase angiogenesis as well as Cu transcription factor function to promote recruitment of inflammatory cells which secrete TNFα and VEGF via the p47phox/ROS-NFkB-ICAM-1/VCAM-1 pathway. Clinical and animal studies indicate that Cu deficiency therapies by either Cu deficient diet or Cu chelation can be effective for several pathological conditions such as tumor progression and inflammatory diseases 1 . Function of ROS is dependent on its concentration; optimal levels of ROS are required, but excess ROS produced under pathological states are inhibitory for neovascularization 21,31 . Thus, understanding the Cu chaperone/transcription factor function of Atox1 in physiological and disease conditions will lead to new therapeutic strategy for inflammation-dependent ischemic cardiovascular diseases.

Methods
Animals. Atox1 −/− mice (backcrossed eight times to C57Bl/6J (Jackson laboratory) were obtained from Mutant Mouse Regional Resource Centers. Atox1 −/− mice were originally reported to have various phenotypes from perinatal death (45% of pups) to the phenotype which could survive more than one month 16 . In our laboratory, they were further backcrossed to C57Bl/6J mice more than ten times, and thus C57Bl/6J mice were used as control. "Survivor" mice were intercrossed with more than 10 times and Atox1 −/− mice used in the present study survived more than six months (90%). For the experiments, Atox1 −/− mice were weaned at 4 weeks of age and maintained on regular chow for 2 to 3 months. Diet and water were provided ad libitum. All studies were carried out in accordance with the guidelines approved by the Animal Care and Use Committee of the University of Illinois-Chicago Mouse hindlimb ischemia model. Mouse hindlimb ischemia was induced as described previously 29,55 . Briefly, following anesthesia (87 mg/kg ketamine, 13 mg/kg xylazine), the left femoral artery was exposed under a dissection microscope. The proximal portion of the femoral artery and the distal portion of the saphenous artery were ligated. All branches between these 2 sites were ligated or cauterized, and arteriotomy was performed. We measured ischemic (left)/nonischemic (right) limb blood flow ratio using a laser Doppler blood flow (LDBF) analyzer (PeriScan PIM 3 System; Perimed) as we described 29,55 . Mice were anesthetized and placed on a heating plate at 37 °C for 10 minutes to minimize temperature variation. Before and after surgery, LDBF analysis was performed in the plantar sole. Blood flow was displayed as changes in the laser frequency, represented by different color pixels, and mean LDBF values were expressed as the ratio of ischemic to nonischemic LDBF. Tissue was extracted in radioimmunoprecipitation assay lysis buffer supplemented with protease inhibitor. The lysates were clarified by ultracentrifugation (13200 rpm, 10 minutes) at 4 °C and protein were measured by bicinchoninic acid assay. For adenovirus experiment, recombinant purified adenovirus (Ad.Atox1-WT, Ad. LacZ (control), 1 × 10 9 pfu) was injected into the adductor and gastrocnemius muscles in Atox1 KO mice at one day prior to hindlimb ischemia.
Bone marrow (BM) transplantation. BM transplantation was performed as we previously described 29 . BM cells were isolated by density gradient separation. Recipient mice were lethally irradiated with 9.5 Gy and received an intravenous injection of 3 million donor BMCs 24 hours after irradiation. To determine the transplantation efficiency, transplantation was performed between CD45.1 mice (Jackson Laboratories) and either Atox1 −/− or WT mice and peripheral white blood cells were stained with anti-CD45.1 (BD) and anti-CD45.2 (BD). Hindlimb ischemia was induced 6 to 8 weeks after bone marrow transplantation.
Histology and immunohistochemistry. Mice were sacrificed at 3-14 days after surgery, and muscles of the lower limbs or upper limbs were harvested, methanol fixed, and paraffin embedded. Frozen sections were prepared by overnight 4% PFA incubation followed by sucrose dehydration and OCT embedding. Capillary density in the ischemic muscles was determined in methanol fixed 5 μ m thick sections that were stained with anti-mouse CD31 antibody (BD Biosciences) followed by biotinylated anti-mouse IgG antibody (Vector Laboratories). Arterioles were stained with anti-α SMA antibody (Sigma). Adhesion molecule expression was stained by goat anti-VCAM-1 (Santa Cruz). Monocytes/macrophages were labeled with anti-Mac3 and neutrophils with anti-Gr1 antibody followed by biotinylated anti-rat IgG (Vector Laboratories). Rabbit polyclonal anti-Atox1 17 and rat anti-CD45 (BD Biosciences) staining was performed in frozen sections. For immunohistochemistry, we used R.T.U. Vectorstain Elite (Vector Laboratories) followed by DAB (Vector Laboratories). Images were captured by Axio scope microscope or confocal microscopy and processed by AxioVision 4.8 or Zeiss LSM510 software, respectively.
To quantify proliferating cells, mice were injected intraperitoneally with the thymidine analog 5-bromo-2′ -deoxyuridine (BrdU; 40 mg/kg body weight, 500 μ l, SigmaAldrich) at 12 hours and 1 hour before sacrifice asdescribed previously 52 . Mice were euthanized and perfused through the left ventricle with saline, and hindlimb muscles were embedded in OCT compound and snap-frozen in liquid nitrogen. Sections were incubated with a rat anti-BrdU antibody (Abcam) to detect proliferating cells. For nuclear staining, BrdU-labeled tissue sections were counterstained with 4′ , 6-diamidino-2-phenylindole (DAPI).
Volumetric Micro-CT. Mice were euthanized and vasculature was flushed with 0.9% normal saline containing, papaverine (4 mg/L), and adenosine (1 g/L) for 3 minutes. After vasodilation, 2% paraformaldehyde was perfused for 5 minutes at 100 mm Hg pressure to fix the vasculature and wash with saline. Contrast agent containing bismuth (0.5 mL) (Generously provided by Dr. Michael Simons, Section of Cardiovascular Medicine, Yale School of Medicine, New Haven, CT) 41 in 5% gelatin was injected over a period of 2 minutes with a syringe pump. The mice were then immediately chilled on ice and immersion-fixed in 2% paraformaldehyde overnight. The hindlimb vasculature was imaged with a high-resolution micro-computed tomography (micro-CT) imaging system (Scanco μ CT 50, Bassersdorf, Switzerland) in MicroCT/Histology Core, Rush University, Chicago, IL, USA. The samples were scanned at 70 kV voltage and 114 μ A current, with a500ms integration time and 16 μ m resolution. A region of interest was reconstructed using 1355 z-sections (~21.68 mm) of the hindlimb. During postprocessing of the image, a 530 gray-scale value was set as a threshold to eliminate bone with minimal sacrifice of vessel visualization. Quantification was performed by using Scanco software.
Intravital microscopy analysis. WT (C57BL/6J) or Atox1 −/− mice were prepared for intravital microscopy as described previously 25,26 . Mice were pre-anesthetized by intraperitoneal injection of ketamine (87 mg/kg, Bedford Laboratories, Bedford, OH) and xylazine (13 mg/kg, Akorn Inc., Decatur, IL). A tracheal tube was inserted and the mouse was placed on a thermo-controlled blanket (37 °C). To maintain anesthesia, the ketamine and xylazine solution, 50 μ l, was administered every 30 minute through a cannulus placed in the jugular vein. Mice were kept on a 37 °C thermo-controlled rodent blanket (Thermal Care, Niles, IL) throughout the experiment. For TNFα stimulation, mice were treated with intrascrotal injection of murine TNFα (0.5 μ g in 250 μ l saline) for 3 hr before exteriorization of the cremaster muscle. After the scrotum was incised, the cremaster muscle was exteriorized onto an intravital microscopy tray. The cremaster preparation was superfused with thermo-controlled (37 °C) and aerated (5% CO 2 , 95% N 2 ) bicarbonate buffered saline throughout the experiment.
Three hours after TNFα injection, the cremaster muscle was exposed as described previously 25,26 . Mouse neutrophils were monitored in an area of 0.02 mm 2 over 5 min in the inflamed cremaster muscle venules by infusion of an Alexa Fluor 647-labeled anti-Gr-1 antibody (0.04 μ g/g body weight, BioLegend). Five to seven different venules were monitored in a mouse. Rolling and adhesion of neutrophils in post-capillary venules was recorded using an Olympus BX61W microscope with a 60x objective. Digital images were acquired using a high-speed digital camera (Hamamatsu C9300) through an intensifier (Video Scope International, Dulles, VA). Fluorescence images were analyzed using Slidebook v5.0 (Intelligent Imaging Innovations, Denvor, CO). Fluorescence images were captured at exposure times of 100 millisec and bright field images were captured with exposure times of 20 millisec.
Wall shear rate (Wsr) in the mouse cremaster muscle venule averaged 500-700 s −1 and was calculated from the following equation as described previously 25  Dv were calculated from measurements derived from infusion of FluoSpheres yellow-green fluorescent microspheres (0.2 μ m, Molecular Probes) into the cremaster venule and analyzed using the SlideBook software. Wall shear stress (Wst) within the mouse cremaster venule was calculated using the following equation: Wst = Wsr × viscosity, where Wsr is in s −1 and viscosity is in poise [P]. A value of 0.035 [P] was used for the viscosity of blood.

O 2
•− Detection in Mice. Dihydroethidium (DHE) (Invitrogen, Grand Island, NY) was prepared as a 1 mg/ml solution in 1% Dimethyl sulfoxide (DMSO) and administered at 10 mg per kg of body weight by intravenous injection as described previously 29,55 . Mice were killed and immediately perfusion-fixed with 4% paraformaldehyde 1 hour after DHE injections. Frozen sections of ischemic muscles were prepared and observed by confocal microscope (Zeiss, Thornwood, NY) with excitation at 510-550 nm and emission > 585 dsx7Anm to detect oxidized ethidium.
In vivo measurement of luciferase gene expression by bioluminescence. HLL and Atox1 −/− HLL mice were anesthetized, and the hair was removed over the back side of mice before imaging as described previously 56 . Luciferin (150 mg/kg/mouse in 200μ l of isotonic saline) was administered by i.p. injection. After 12-16 min, mice were imaged with using Xenogen IVIS spectrum (Caliper Life Sciences, Hopkinton, MA, USA). For the duration of photon counting, mice were placed inside a light-tight box. Light emission from the mouse was detected as photon counts using the intensified charge-coupled device camera and expressed as photon counts. Images were obtained before and following hind-limb ischemic surgery. Luciferase activity was measured in ischemic tissue samples by adding 100 μ l of freshly reconstituted luciferase assay buffer to 20μ l of the homogenated ischemic tissue that was homogenized using reporter lysis buffer (Promega). Luciferase activity was expressed as relative light units normalized for protein content.
Sponge implant model. To prepare sponge for implantation 22 , polyvinyl alcohol sponge (sponge CF120, 5mm diameter ×2.75 mm thick, Merocel) was hydrated using saline overnight and sterilized hydrated sponges by autoclaving for 25 min at 121 °C. Sterilized sponges were implanted into WT and Atox1 KO mice in the presence of VEGF. Implants were excised at 3 weeks following implantation for histological analysis because inflammatory angiogenesis peaks during this period as showed previously 22 .
The sponge was harvested, fixed with paraformaldehyde and embedded in paraffin. The paraffin section (5 μ M) was used for immunostaining and histology.

Isolation of coronary vascular endothelial cells. Mouse coronary endothelial cells (MCECs) were
isolated as previously described 57 . Initially, beads coated with sheep anti-rat IgG were incubated with purified rat anti-mouse CD31 monoclonal antibody (1μ g/ml) at 4 °C overnight and then washed with PBS containing 0.1% BSA and 2 mM EDTA. Next day, dissected heart tissues were minced and incubated with M199 containing 1 mg/ml collagenase II for 1 h at 37 °C with mild agitation every 15 min. The digested material was filtered through sterile 40μ m nylon mesh and washed in 2% fetal calf serum in M199. Cell suspensions were incubated with previously prepared antibody/beads complex for 30 min at 4 °C and beads/endothelial cells complex were isolated using magnet, washed and plated with medium containing endothelial growth factors.
Quantitative real-time PCR. Total RNA of gastrocnemius and HUVECs was isolated by using phenol/chloroform and isolated using Tri Reagent (Molecular Research Center Inc.). Reverse transcription was carried out using high capacity cDNA reverse transcription kit (Applied biosystems) using 2 ug of total RNA. Quantitative PCR was performed with the ABI Prism 7000, the SYBR Green PCR kit (Qiagen) and the QuantiTect Primer Assay (Qiagen) for specific genes. Samples were all run in triplicates to reduce variability. Expression of genes was normalized and expressed as fold-changes relative to GAPDH. Immunofluorescence analysis. HUVECs on glass coverslips were rinsed quickly in ice-cold PBS, fixed in freshly prepared 4% paraformaldehyde in PBS for 10 min at room temperature, permeabilized in 0.05% Triton X-100 in PBS for 5 min, and rinsed sequentially in PBS, 50 μ mol/L NH 4 Cl and PBS for 10 min each. After incubation for 1 h in blocking buffer (PBS+ 3%BSA), cells were incubated with Atox1 antibody for 18 h at 4 °C, rinsed in PBS/BSA, and then incubated in Alexa Fluor 488-conjugated goat anti-rabbit IgG for 1 h at room temperature and cells rinsed with PBS. Cells on coverslips were mounted onto glass slides using Vectashield (Vector Laboratories) and observed using confocal microscopy.
For protein expression in ischemic muscle, mice were perfused with cold phosphate buffer saline. Muscle samples were harvested and frozen in liquid nitrogen. Muscle samples were crushed and lysed with RIPA lysis buffer (5 mM Tris-HCl (pH 7.6), 150 mM NaCl, 1% NP-40, 1%sodium deoxycholate, 0.1% SDS) with protease inhibitor followed by brief sonication as described previously 40 . Equal amount of protein was separated by SDS-PAGE. Following primary antibodies were used: anti-Atox1 (home-made), anti-p47phox (Santa Cruz), or anti-VEGF antibody (Abcam). Protein expression was visualized by ECL (Amersham). Band density was quantified by ImageJ.
Capillary network formation assay. HUVECs transfected with control siRNA or Atox1 siRNA were seeded on top of the thick growth factor-reduced Matrigel-coated wells (BD Biosciences) and incubated for 6 h at 37 °C. Images were taken with a Nikon digital camera, and eight random fields per well were analyzed.
Chromatin Immunoprecipitation (ChIP) Assay. HUVECs grown to confluence were treated with TNFα . ChIP assays were performed by following the EpiTect ChIP One-Day Kit protocol (SA Bioscience). Cells were treated with formaldehyde (final concentration of 1%) for 15 min at 37 °C to cross-link proteins to DNA before harvesting. Then cells were rinsed twice with ice-cold PBS containing protease inhibitor, scraped into conical tubes, and pelleted for 4 min at 2000 rpm at 4 °C. Cells were resuspended in 150μ l of lysis buffer (1% SDS, 10mM EDTA, 50mM Tris, pH 8.1, containing protease inhibitor) and incubated for 10 min on ice. Resuspended cells were sonicated 10 times for 10s each on ice and a second round of 5 times for 5s each. After centrifugation, the supernatant was diluted 1:10 with dilution buffer (0.01% SDS, 1.1% Triton X-100, 1.2mM EDTA, 16.7 mM Tris-HCl, pH 8.1, 167 mM NaCl). The cell lysate was precleared by incubation at 4 °C for 1 h with 45μ l of salmon sperm DNA/protein A-agarose beads. The cleared lysates were incubated with anti-p65/RelA (Santa Cruz), monoclonal Atox1 antibody (Z31112) as described above, histone (Santa Cruz), or normal mouse IgG overnight. The following primers were used to amplify ICAM-1: Forward 5′ -TGTTCCCAGGTGAGTCGGGGTG-3′ and reverse 5′ -TCGTGGAGGTATGCAGGGTCTGG-3′ . VCAM-1: Forward 5′ -AAATCAATTCACATGGCATA-3′ and reverse 5′ -AAGGGTCTTGTTGCAGAGG-3′ . -110 region of p47phox: Forward 5′ -GCGACGGG AAGGACAAGTGTAAA-3′ and reverse 5′ -GCTGCGCCTTAAATGCACTGGAAA-3′ . LOX activity assay. LOX activity in cultured medium and ischemic tissue lysate was measured by a high-sensitivity fluorescence assay as previously described 35 . Protein samples was incubated in the presence and absence of 500 μ mol/L BAPN at 37 °C for 30 min with final reaction mixture supplied by Amplite Fluorimetric Lysyl Oxidase Assay kit (AAT Bioquest) to the manufacturer's instruction. The reaction was stopped on ice, and differences in fluorescence intensity (540-nm excitation wavelength and 590-nm emission wavelength) between samples with and without BAPN were determined. Specific activity was determined by the ratio of activity to relative amount of protein.
Monoclonal Atox1 antibody production. Mouse monoclonal Atox1 antibody (Z3112) was developed by using baculovirus particles displaying the surface glycoprotein gp64-fusion protein as the immunizing agent. Briefly, the Atox1 cDNA was ligated into the gp64 gene to create a fusion protein that is expressed in the viral surface, as previously described 58 . Positive clones were selected using recombinant Atox1 protein or cells overexpressing Atox1 by adenovirus 17,32 . Statistical analysis. All the experiments were repeated at least three times, and all values were expressed as mean ± SE. Blood flow recovery in the ischemic hindlimb was compared between the two groups by two-way repeated measures ANOVA, followed by Bonferroni post hoc analysis. Comparison between groups was analyzed by unpaired Student 2-tailed t test (2 groups) or ANOVA for experiments with more than two subgroups followed by Bonferroni post hoc analysis. For clinical studies, the distributions of the data were scrutinised for normality using the Shapiro-Wilk test. Parametric and non-parametric pairwise comparisons were analysed with the Student's t-test and Mann-Witney U test, respectively. Differences in proportions were tested using the z-test. All tests performed were 2-sided with a p < 0.05 considered statistically significant. Statistical tests were performed using Prism v6 (GraphPad Software, San Diego, CA).