Midkine Controls Arteriogenesis by Regulating the Bioavailability of Vascular Endothelial Growth Factor A and the Expression of Nitric Oxide Synthase 1 and 3.

Midkine is a pleiotropic factor, which is involved in angiogenesis. However, its mode of action in this process is still ill defined. The function of midkine in arteriogenesis, the growth of natural bypasses from pre-existing collateral arteries, compensating for the loss of an occluded artery has never been investigated. Arteriogenesis is an inflammatory process, which relies on the proliferation of endothelial cells and smooth muscle cells. We show that midkine deficiency strikingly interferes with the proliferation of endothelial cells in arteriogenesis, thereby interfering with the process of collateral artery growth. We identified midkine to be responsible for increased plasma levels of vascular endothelial growth factor A (VEGFA), necessary and sufficient to promote endothelial cell proliferation in growing collaterals. Mechanistically, we demonstrate that leukocyte domiciled midkine mediates increased plasma levels of VEGFA relevant for upregulation of endothelial nitric oxide synthase 1 and 3, necessary for proper endothelial cell proliferation, and that non-leukocyte domiciled midkine additionally improves vasodilation. The data provided on the role of midkine in endothelial proliferation are likely to be relevant for both, the process of arteriogenesis and angiogenesis. Moreover, our data might help to estimate the therapeutic effect of clinically applied VEGFA in patients with vascular occlusive diseases.

Introduction

Vascular occlusive disease such as myocardial infarction, stroke or peripheral artery diseases are still a major cause of morbidity and mortality worldwide. Searching for new, non-invasive options to treat affected patients, much effort is made to understand the molecular mechanisms of arteriogenesis. Arteriogenesis is a tissue- and even lifesaving process and presents the growth of pre-existing collateral arterioles to natural biological bypasses. It is provoked by increased fluid shear stress and is mediated by sterile inflammation (Chillo et al., 2016).

Midkine (MK) is a retinoic acid inducible cytokine, which is highly expressed during embryogenesis (Kadomatsu et al., 1990). In the adulthood its expression is restricted to certain tissues but strongly induced during inflammatory processes (Badila et al., 2015) such as rheumatoid arthritis, inflammatory bowel disease, and multiple sclerosis (for an overview see (Weckbach et al., 2011)), which are associated with angiogenesis. MK has also been implicated in angiogenesis and hence endothelial cell (EC) proliferation under conditions of tissue ischemia (Weckbach et al., 2012) and tumor growth (Choudhuri et al., 1997, Muramaki et al., 2003) as well as in neointima formation (Hayashi et al., 2005, Horiba et al., 2000). The functional role of MK in angiogenesis is still not well defined, however, it was shown that MK induces chemotaxis of neutrophils and supports neutrophil adhesion during inflammation (Takada et al., 1997, Weckbach et al., 2014). Neutrophils themselves express high levels of MK, but do not release it (Weckbach et al., 2012, Weckbach et al., 2014, Narita et al., 2008). However, after stimulation with the chemokine (C-X-C) ligand 1 (CXCL1) neutrophils release angiogenic growth factor such as vascular endothelial growth factor A (VEGFA) (Scapini et al., 2004).

The functional role of MK in arteriogenesis has never been investigated. Moreover, the functional role of VEGFA in arteriogenesis is controversially discussed (Jazwa et al., 2016). In vitro data from Tzima et al., who identified the VEGF receptor 2 (VEGFR-2), to be part of a mechanosensory complex and to be relevant for phosphoinositide 3-kinase (PI3K) activation, even suggested that VEGFR-2 is ligand-independent activated in response to shear stress (Tzima et al., 2005). Here, we demonstrate that leukocyte domiciled midkine controls the plasma bioavailability of VEGFA, which in turn is relevant for the expression of neuronal nitric oxide synthase (Nos1) and endothelial Nos (Nos3) (but not inducible Nos (Nos2)) in endothelial cells of collaterals. For arteriogenesis all three isoform of NOS, namely NOS1, NOS2 and NOS3 are described to be involved, although the function of the individual NOS is not well defined (Troidl et al., 2010, Pagel et al., 2011). Exclusively for NOS3 it has been shown that it mediates vasodilation in growing collaterals (Troidl et al., 2010). Here we show that Nos1 or Nos3 expression, respectively, is essential for collateral endothelial cell proliferation, whereby both isoforms are likely to be able to substitute for each other.

Collectively, our data provide mechanistic insights into MK-mediated endothelial cell proliferation, which are likely to be relevant not only for the process of arteriogenesis, but also for other inflammatory processes as well as tumor growth being associated with angiogenesis and hence endothelial cell proliferation.

Materials and Methods

Animals and Treatments

All experiments were carried out according to the German animal legislation guidelines and were approved by the Bavarian Animal Care and Use Committee. Mice were kept under a 12-hour (h) day and night cycle with chow and water provided ad libitum.

Mice were treated with either 100 μl DETA NONOate (Cayman Chemical, Ann Arbor, MI) 0.12 mg/ml intraperitoneal once daily in 0.9% NaCl (saline) (Braun, Melsungen, Germany), with 50 μl recombinant murine midkine (rmMK) (PeproTech, Rocky Hill, NJ) 0.1 mg/ml subcutaneously twice daily in saline or with 100 μl recombinant murine VEGFA164 (rmVEGFA) (Sigma-Aldrich, St. Louis, MO) 0.01 mg/ml intraperitoneal once daily in saline. Control groups received saline only. Sodium nitrite (Sigma-Aldrich) was applied via drinking water at a concentration of 1 g/l (Kumar et al., 2008, Hannas et al., 2010). Control groups received normal drinking water. Treatment was started 3 days (d) before the surgical procedure and continued until the end of the experiments.

Femoral Artery Ligation, Laser Doppler Imaging and Bone Marrow Transplantation

Collateral artery growth was investigated in a murine hindlimb model (Chillo et al., 2016) using 8- to 10-week-old male Mdk−/− (carrying the Mdk allele, RRID: MGI:3579532) (Nakamura et al., 1998), Nos1−/− (B6.129S4-Nos1/J, RRID: IMSR_JAX:002986) (Huang et al., 1993) mice and appropriate WT controls (C57BL/6J mice, RRID: IMSR_JAX:000664; Charles River, Sulzfeld, Germany). Mice underwent ligation of the right femoral artery distally to the origin of the profunda femoris branch using a 7-0 silk braided suture (Peasalls Sutures, Somerset, Great Britain). During the surgical procedures mice were under general anesthesia with a combination of fentanyl (0.05 mg/kg body weight) (Janssen-Cilag, Neuss, Germany), midazolam (5.0 mg/kg body weight) (Ratiopharm, Ulm, Germany) and medetomidine (0.5 mg/kg body weight) (Pfizer, New York, NY), which was administered subcutaneously. The left leg underwent sham operation and was used as control. Perfusion of the paws was assessed using a LDI technique (Moor LDI 5061 and Moor Software Version 3.01, Moor Instruments, Remagen, Germany) under temperature-controlled conditions. The measurements were conducted before FAL (basal), immediately after FAL (aFAL), 3d and 7d after FAL. Color-coded images of the paws representing the flux value were used to calculate the ratio of the tissue perfusion of the occluded (Occ) to the Sham-operated (Sham) paw. For bone marrow transplantation see the Supplemental Experimental Procedures.

Tissue and Blood Sampling

Heparinized blood gained by cardiac puncture was centrifuged (2.300g at 4 °C for 20 min) and plasma was stored at − 80 °C. Prior to harvesting superficial collateral arteries for qRT-PCR (quantitative real-time PCR), both hind limbs were perfused with latex flexible compound (Chicago Latex, Chicago, IL) via a catheter in the abdominal aorta. For each mouse two superficial collateral arteries per side were isolated, snap frozen in dry ice and stored at − 80 °C until further investigations.

For histological analysis of collateral arteries, mice were perfused with 20 ml adenosine buffer (1% adenosine (Sigma-Aldrich), 5% bovine serum albumin (BSA, Sigma-Aldrich) dissolved in phosphate buffered saline (PBS, PAN Biotech, Aidenbach, Germany), pH 7.4), to assure maximum vasodilatation (Chillo et al., 2016), followed by 20 ml 4% paraformaldehyde (PFA, Merck, Darmstadt, Germany) (for paraffin embedding) or 20 ml 3% PFA (for cryopreservation) in PBS, pH 7.4.

For histological analysis adductor muscles were rinsed in 4%PFA for 24 h and after paraffin embedding cut in 4 μm cross-sections. For immunofluorescence staining the samples were placed in 15% sucrose (Sigma-Aldrich), dissolved in PBS, for 4 h, followed by 30% sucrose, dissolved in PBS, overnight at 4 °C. Afterwards the adductor muscle tissue was cryopreserved in tissue tek (Sakura, Alphen aan den Rijn, The Netherlands) and cut in 6 μm cross-sections.

Histology and Immuno-Histology

Paraffin embedded cross-sections were stained with GIEMSA according to standard procedures. The number of collaterals on the Occ and Sham side of WT and Mdk−/− mice was counted 7d after FAL with an Axioskop 40 microscope and Axiocam 105 color (Carl Zeiss AG, Feldbach, Switzerland). On the Occ side only collaterals with a luminal diameter larger than 25 μm were defined as growing collaterals. Collaterals with a smaller diameter were not analyzed in the following investigations, since their diameter was comparable to the diameter of collaterals on the Sham side and hence didn't grow after FAL (Fig. S1). Luminal diameter and vascular wall of collateral arteries were analyzed with AxioVs40 V4.8.2.0 software (Carl Zeiss AG). For each mouse 5 slices with 3 growing collaterals on each slice were measured.

Mice were treated daily with 100 μl bromodeoxyuridine (BrdU) (BD Biosciences, San Jose, CA) (12.5 mg/ml dissolved in PBS, intraperitoneal) starting at the day of the surgical procedure. A standardized BrdU In-Situ Detection Kit (BD Biosciences) was used to analyze the proliferation of ECs and SMCs in collateral arteries 7d after FAL. For immunofluorescence staining 24 h after FAL, cryofixed tissue sections were stained with combinations of the following antibodies: primary anti-MK antibody (Cat#: PA5-19640, RRID: AB_10985009, Thermo Fischer Scientific, Waltham, MA, dilution 1:100, for 12 h at 4 °C), or primary anti-VEGFA164 antibody (Cat#: AF-493-SP, RRID: AB_354506, Novus Biological, Littleton, CO, dilution 1:50, for 12 h at 4 °C), secondary donkey anti-rabbit antibody (Cat#: A10040, RRID: AB_2534016, Thermo Fischer Scientific, dilution 1:200, for 1 h at room temperature), anti-Ly-6G antibody (Cat#: ab25024, RRID: AB_470400, Abcam, Cambridge, MA, dilution 1:100, for 12 h at 4 °C) and DAPI (Cat#: D1306, RRID: AB_2629482, Thermo Fischer Scientific, dilution 1:1000, for 20 min at room temperature). Cross-sections were covered with Mowiol (Sigma-Aldrich) and analyzed with a confocal microscope (Leica SP5, Leica, Wetzlar Germany).

Protein Levels of VEGFA and MK

The VEGFA plasma level was measured with a standardized ELISA Kit (R&D Systems, Minneapolis, MN) following the manufacturer's protocol. The ready-SET-go Midkine ELISA-Kit (USCN Life Science Inc., Wuhan, China) was used to measure the MK plasma level following the manufacturer's protocol.

RNA Isolation, cDNA Synthesis and qRT-PCR

After isolating RNA from collateral arteries or primary endothelial cells with Trizol (Life technologies, Carlsbad, CA), RQ1 RNase-Free DNase (Promega, Madison, WI) digestion was used to remove residual genomic DNA. 250 ng purified RNA (using RNeasy MinElute columns (Qiagen, Hilden, Germany)) was reverse transcribed into cDNA with the QuantiTect Rev. Transcription Kit (Qiagen). For qRT-PCR analysis the SYBR Power Green Kit (Life Technologies) was used according to the manufacturer's protocol with 1 μl of the 1:5 diluted cDNA and 0.5 μM of each primer per reaction. For further information see the Supplemental Experimental Procedures.

Cell Culture

Mouse aortic ECs and SMCs were isolated as previously described (Kobayashi et al., 2005). In brief, isolated murine aortas were filled with collagenase type II (Biochrom AG, Berlin, Germany) solution (2 mg/ml, dissolved in serum-free Dulbecco's Modified Eagle's Medium (DMEM, Sigma-Aldrich)) for 45 min at 37 °C, and afterwards the ECs were flushed out of the aortas with DMEM containing 20% fetal calf serum (FCS, PAN Biotech). ECs were cultured with Complete Mouse Endothelial Cell Medium (Cellbiologics, Chicago, IL) on collagen A (Biochrom AG) coated flasks. The remaining aorta was cut into 2 mm pieces and placed on a 1% gelatin-coated dish (PAN Biotech). After 10 days the tissue pieces were removed and SMCs were cultured with Complete Mouse Smooth Muscle Cell Medium (Cellbiologics) on 1% gelatine coated flasks. Medium was changed every other day and passages 1 to 3 were used for experiments.

In Vitro BrdU Proliferation Assays

A 96-well plate with 1.2 × 104 ECs or SMCs per well and a standardized BrdU proliferation kit (Roche, Basel, Switzerland) was used to analyze the in vitro proliferation of isolated ECs or SMCs. For further information see the Supplemental Experimental Procedures.

Statistical Analysis

Experimenters were blind to outcome assessment for all experiments. All statistical analysis were performed with GraphPad software PRISM6 (GraphPad Software, La Jolla, CA) and are specified in the Figure legends. Data shown are means ± SD. Results were considered to be statistically significant at P < 0.05.

Results

Midkine Is Essential for EC Proliferation of Growing Collateral Arteries

To characterize the functional role of MK in arteriogenesis, we investigated a murine hindlimb model in which femoral artery ligation (FAL, Occ) results in collateral artery growth (Chillo et al., 2016). Our data showed that FAL results in a reduced increase in caliber size and tortuosity of superficial collateral arteries as well as a significantly reduced perfusion recovery in Mdk deficient (Mdk−/−) compared to wild-type (WT) mice (Fig. 1a–c). To investigate whether the reduced perfusion recovery observed in Mdk−/− mice was due to a reduced number of pre-existing arteriolar connections, we performed histological analyses. Our results revealed identical numbers of collaterals with increased luminal diameter (3 per slice per mouse) as well as total numbers of all collaterals (Occ = 5, Sham = 5, per slice per mouse) in WT and Mdk−/− mice (Fig. S1). Moreover, the luminal diameters of collateral arteries of the Sham side were comparable in WT and Mdk−/− mice (Figs. S1 and S2). However, compared to WT mice, Mdk−/− mice showed a significantly smaller increase in luminal diameter after FAL (Fig. 2a and b). Moreover, the ratio of the vessel wall thickness to the luminal diameter was significantly increased in Mdk−/− mice (Fig. 2c). BrdU staining revealed a significantly decreased percentage of proliferating ECs in Mdk−/− mice, whereas there was no significant difference in the number of proliferating smooth muscle cells (SMCs) (Fig. 2d and e).

Fig. 1

Mdk−/− mice show reduced perfusion recovery after FAL. (a) Representative images of superficial collateral arteries in the adductor muscles of WT (upper panels) and Mdk−/− (lower panels) mice 7d after FAL (Occ) or sham operation (Sham). Scale bars: 3 mm. (b) Representative LDI images of WT (upper panel) and Mdk−/− (lower panel) mice 3d and 7d after FAL. (c) Line graph displaying the perfusion ratio (Occ/Sham) of WT and Mdk−/− mice before FAL (basal), immediately after FAL (aFAL), 3d and 7d after FAL. n = 5. *P < 0.05 WT compared with Mdk−/−, two-way ANOVA with Bonferroni's multiple comparison test. Data shown are means ± SD.

Fig. 2

Collateral vessel morphology after FAL. (a) Representative pictures of Giemsa stained slices showing collaterals in the adductor muscle of WT (left) and Mdk−/− mice (right) 3d (upper panel) and 7d (lower panel) after FAL. Scale bars: 50 μm. Bar graphs representing (b) the luminal diameter and (c) the ratio of the vascular wall to the luminal diameter of WT and Mdk−/− collaterals 3d and 7d after FAL. 5 slices with 3 collaterals per slice were analyzed per mouse and time point. (d) Representative pictures of BrdU stained tissue sections with BrdU+ ECs (arrowhead) and BrdU+ SMCs (arrow) 7d after FAL. (e) Bar graph displaying the percentage of BrdU+ and hence proliferated ECs and SMCs in WT and Mdk−/− collaterals 7d after FAL. 3 slices with 3 collaterals each were analyzed per mouse. n = 5. *P < 0.05, one-way ANOVA with Bonferroni's multiple comparison test (b, c) or unpaired Student's t-test (e). Data are means ± SD.

Leukocyte Domiciled MK Promotes Collateral Artery Growth

MK plasma levels prior to any surgical procedure (basal), 12 h and 24 h after FAL were below the detection range. At 3d after FAL, MK plasma level increased to 0.268 ± 0.012 ng/ml and decreased again at 7d (Fig. 3a). The Mdk mRNA expression levels significantly increased in collaterals 12 h and 24 h after FAL (Fig. 3b). Immunofluorescence double staining against MK and Ly-6G performed on tissue samples isolated 24 h after FAL evidenced that MK is present in perivascular neutrophils (Fig. 3c). To investigate the relevance of leukocyte derived MK, bone marrow from WT mice was transplanted into Mdk−/− mice (Mdk−/− WT) and vice versa (WT ). WTWT and Mdk−/− transplanted mice were used as control groups. Bone-marrow transplantation itself didn't affect perfusion recovery, as quantified by LDI measurements of transplanted and non-transplanted mice (Fig. S3). In comparison to Mdk−/− mice, perfusion recovery was significantly increased in Mdk−/− WT as well as WT mice, and reached values of WTWT mice 7d after FAL (Fig. 3d). Histological analysis evidenced a significant increase of the luminal diameter of Mdk−/− WT mice compared to Mdk−/− mice, but not in WT mice (Fig. 3e) pointing to a relevance of bone marrow cell MK for collateral EC proliferation and hence arteriogenesis. Interestingly, MK was only detectable in plasma of WTWT and WT mice but not in Mdk−/− WT mice 7d after FAL (Fig. 3f).

Fig. 3

Bone marrow cell delivered MK is essential for collateral artery growth. (a) Bar graph displaying the plasma level of MK (ng/ml) in WT mice prior to any surgical procedure (basal), 12 h, 24 h, 3d and 7d after FAL. Dashed line shows the detection limit of the ELISA kit (0.128 ng/ml). n = 5 in triplicates. (b) Bar graph representing the expression level of Mdk mRNA in collaterals of WT mice prior to any surgical procedure (basal), 12 h, 24 h and 36 h after FAL as quantified by qRT-PCR. Data were normalized to the expression level of the 18S rRNA. n = 5 in triplicates. *P < 0.05, one-way ANOVA with Bonferroni's multiple comparison test. (c) Immunofluorescence staining for MK (red, lower panel, left), neutrophils (Ly-6G, green, lower panel, middle) and nuclei (DAPI, blue, lower panel, right) 24 h after FAL. Ly-6G+ cells in perivascular tissue stained positive for MK (upper panel). Blue dotted lines indicate the endothelial layer of the collateral artery. Scale bar: 25 μm. (d) Line graph displaying the perfusion ratio (Occ/Sham) as measured by LDI in bone marrow transplanted mice before FAL (basal), immediately (aFAL), 3d and 7d after FAL. For the bone marrow transplant studies WT mice received Mdk−/− donor bone marrow (WT) and vice versa (Mdk−/− WT). WTWT and Mdk−/− transplanted mice were used as control groups. n = 4 per group. §P < 0.05 (WTWT compared with Mdk−/− ), #P < 0.05 (Mdk−/− WT compared with Mdk−/− ), %P < 0.05 (WT compared with Mdk−/− ), two-way ANOVA with Bonferroni's multiple comparison test. (e) Bar graph representing the luminal diameter of collaterals in bone marrow transplanted mice 7d after FAL. n = 4. 5 slices with 3 collaterals each were analyzed per mouse. *P < 0.05, one-way ANOVA with Bonferroni's multiple comparison test. (f) Bar graph displaying the MK plasma level (ng/ml) in bone marrow transplanted mice 7d after FAL. The dashed line shows the detection limit of the ELISA kit (0.128 ng/ml). n = 4. Data shown are means ± SD.

Rescue of Arteriogenesis in Mdk−/− Mice

In order to elucidate the function of exogenous MK on the growth of collaterals, mice were treated with recombinant murine MK (rmMK). Treatment with rmMK normalized perfusion recovery, the luminal diameter, the ratio of the vascular wall to the luminal diameter and EC proliferation in Mdk−/− mice, but did not influence SMC proliferation (Fig. 4a and c–f). WT mice treated with rmMK showed significantly increased perfusion recovery 7d after FAL compared to saline treated WT mice, but the proliferation rate of vascular cells and the growth of collateral arteries were unchanged (Fig. 4a and c–f). Thus exogenous MK improved perfusion recovery in Mdk−/− mice via collateral growth, whereas increased perfusion recovery in WT mice was due to vasodilatation. This data suggested that MK not only promotes collateral growth, but also regulates the expression of Nos responsible for vasodilation. Accordingly, Mdk−/− and WT mice were treated with the NO-donor DETA NONOate. In Mdk−/− mice DETA NONOate treatment significantly increased perfusion recovery and normalized EC proliferation and collateral artery growth, whereas SMC proliferation was unaffected (Fig. 4b–f). In WT mice perfusion recovery, collateral growth as well as vascular cell proliferation was unchanged by DETA NONOate treatment (Fig. 4b–f). qRT-PCR analysis on collaterals isolated 12 h after FAL evidenced significantly decreased Nos1 and Nos3 expression levels in Mdk−/− mice compared to WT mice, while Nos2 expression levels showed no significant differences (Fig. 4g). Since it is well described that Nos3 deficiency is not associated with reduced collateral artery growth (Troidl et al., 2010, Mees et al., 2007), but that NOS3 is relevant for proper vasodilation during arteriogenesis (Troidl et al., 2010), we asked about the relevance of Nos1 for arteriogenesis. Nos1−/− mice did not show a reduced perfusion recovery after FAL. However, in contrast to WT and Mdk−/− mice, NO-donor treatment strikingly interfered with perfusion recovery in Nos1−/− mice (Fig. S4).

Fig. 4

Treatment with DETA NONOate as well as with rmMK rescues arteriogenesis in Mdk mice. Line graphs displaying the perfusion ratio (Occ/Sham) as measured by LDI in WT and Mdk mice after treatment with saline, rmMK (a), or DETA NONOate (b), respectively. n = 5. #P < 0.05 (compared with Mdk), $P < 0.05 (WT + rmMK compared with WT + saline and compared with Mdk), two-way ANOVA with Bonferroni's multiple comparison test. Bar graphs displaying the luminal diameter (c) and the ratio of the vascular wall to the luminal diameter (d) of WT and M Mdk−/− , collaterals 7d after FAL. 5 slices with 3 collaterals each were analyzed per mouse. Bar graphs displaying the percentage of BrdU+ ECs (e) and SMCs (f) in WT and Mdk collaterals 7d after FAL. 3 slices with 3 collaterals each were analyzed per mouse. n = 5. *P < 0.05, one-way ANOVA with Bonferroni's multiple comparison test. (g) Bar graph representing the expression level of Nos1, Nos2 and Nos3 in WT and Mdk collaterals 12 h after FAL as quantified by qRT-PCR. Data were normalized to the expression level of the 18S rRNA. n = 5 in triplicates. *P < 0.05, unpaired Student's t-test. Data are means ± SD.

Inhibition of NOS Blocks the Proliferative Effect of rmMK on ECs In Vitro

Primary ECs of Mdk−/− showed significantly reduced basal expression levels of Nos1 and Nos3 (Fig. 5a) as well as a significantly reduced proliferation rate (Fig. 5b) compared to primary WT ECs. Treatment with the non-specific NOS inhibitor L-NIO significantly reduced the proliferation of WT ECs, but not of Mdk−/− ECs. Treatment with DETA NONOate as well as with rmMK increased EC proliferation of WT and Mdk−/− ECs, however, compared to WT ECs, the proliferation rate of Mdk−/− ECs was still reduced. Treatment with L-NIO abolished the positive effect of rmMK on EC proliferation in both types of cells indicating that the impact of rmMK on EC proliferation is mediated by NOS. Similar to the in vivo experiments, SMC proliferation was neither influenced by MK loss, nor by treatment with rmMK or DETA NONOate (Fig. 5c).

Fig. 5

In vitro analyses of vascular cells. (a) Bar graph representing the mRNA expression level of Nos1 and Nos3 in primary WT and Mdk ECs. n = 5. *P < 0.05 unpaired Student's t-test. Data shown are means ± SD. (b, c) Proliferation rate of primary ECs and SMCs isolated from WT and Mdk aortas as quantified by BrdU uptake. Cells were cultured for 48 h with 2%FCS in DMEM medium alone (Control), or in combination with the non-specific NOS inhibitor L-NIO (0.1 μmol/ml), DETA NONOate (50 μmol/l), rmMK (100 ng/ml), or rmMK and L-NIO. Bar graphs represent the relative proliferation rate of ECs (b) and SMCs (c). Values for WT cells cultured with 2%FCS in DMEM alone (Control) were defined as 100%. n = 5 in triplicates. *P < 0.05 two-way ANOVA with Bonferroni's multiple comparison test. Data are means ± SD.

Relationship Between MK and VEGFA

Since NO has been described to induce the expression of VEGFA and vice versa (Kroll and Waltenberger, 1998, Kimura and Esumi, 2003), we investigated the plasma level of VEGFA of WT and Mdk−/− mice. Prior to any surgical procedure (basal), the plasma level of VEGFA was comparable in WT and Mdk−/− mice (Fig. 6a). 24 h after FAL the VEGFA plasma level increased significantly in WT mice, but not in Mdk−/− mice. Treatment with rmMK normalized the VEGFA plasma level of Mdk−/− mice, but showed no effect on the plasma level of WT mice. DETA NONOate treatment in contrast did not influence the plasma level of VEGFA either in WT or Mdk−/− mice. Immunohistological analysis identified neutrophils as potential source VEGFA (Fig. 6b).

Fig. 6

Influence of VEGFA. (a) Bar graph displaying ELISA results on VEGFA plasma levels (pg/ml) in WT and Mdk mice treated with saline, DETA NONOate or rmMK, respectively. Plasma was collected prior to any surgical procedure (basal) or 24 h after FAL. n = 5 in triplicates. *P < 0.05 one-way ANOVA with Bonferroni's multiple comparison test. (b) Triple immunofluorescence staining for VEGFA (right, lower panel), Ly-6G (right, middle panel) and DAPI (right, upper panel) 24 after FAL. Ly-6G+ cells in perivascular tissue stained positive for VEGFA (left panel). Blue dotted line indicates the endothelial layer of the collateral artery. Scale bar: 25 μm. (c) Line graph displaying the perfusion ratio (Occ/Sham) as measured by LDI in WT and Mdk mice treated with saline or rmVEGFA. Data are show for before FAL (basal), immediately (aFAL), 3d and 7d after FAL. n = 4. #P < 0.05 (Mdk + rmVEGFA compared with Mdk + saline), §P < 0.05 (WT + rmVEGFA compared with Mdk + saline), %P < 0.05 (WT + saline compared with Mdk + saline), two-way ANOVA with Bonferroni's multiple comparison test. (d) Bar graph displaying the luminal diameter of collaterals in WT and Mdk mice treated with saline or rmVEGFA 7d after FAL. n = 4. *P < 0.05, one-way ANOVA with Bonferroni's multiple comparison test. (e) Bar graph displaying the expression levels of Nos1, Nos2 and Nos3 as quantified by qRT-PCR in WT and Mdk collaterals 12 h after FAL. Data were normalized to the expression level of the 18S rRNA. n = 5 in triplicates. *P < 0.05, two-way ANOVA with Bonferroni's multiple comparison test. (f) Bar graph representing the relative in vitro proliferation rate of WT and Mdk ECs as quantified by BrdU uptake analysis. Cells were treated for 48 h with 2%FCS in DMEM medium alone (Control), or with the specific VEGFR-2 inhibitor SU1498 (100 ng/ml), with rmVEGFA (50 ng/ml), or a combination of SU1498 and rmVEGFA. Values for WT cells cultured with 2%FCS in DMEM alone (Control) were defined as 100%. n = 5. *P < 0.05 two-way ANOVA with Bonferroni's multiple comparison test. Data are means ± SD.

7d after FAL, treatment with rmVEGFA significantly increased perfusion recovery and luminal diameter of Mdk−/− mice, however, showed no effect on WT mice (Fig. 6c and d). Moreover, treatment with rmVEGFA normalized Nos1 and Nos3 expression levels in collaterals of Mdk−/− mice, whereas Nos2 expression levels were not affected either in WT or Mdk−/− mice (Fig. 6e).

Accompanying in vitro results evidenced that inhibition of VEGFR-2 by SU1498 significantly reduced the proliferation rate of primary WT ECs, but not of primary Mdk−/− ECs (Fig. 6f). Although having a broader specificity, SU1498 has been used in many experiments to block VEGFR-2 (e.g. (Greenberg and Jin, 2013)). Treatment with rmVEGFA in contrast significantly increased the proliferation rate of primary WT ECs and primary Mdk−/− ECs.

Discussion

In this study we identified MK as factor relevant for proper collateral EC proliferation during arteriogenesis. Our data indicate that MK regulates the bioavailability of VEGFA and modulates the differential expression of Nos1 and Nos3 in vascular ECs. A proposed model for the action of MK is shown in Fig. 7.

Fig. 7

Proposed model for the mechanistic function of midkine in arteriogenesis. After induction of arteriogenesis, CXCL-1, either released from platelets or EC, activates neutrophils thereby inducing MK mediated VEGFA expression and release. Upon binding to VEGFR-2, VEGFA induces the expression of Nos1 and Nos3, relevant for vascular cell proliferation. However, MK also activates either directly, or indirectly by binding to ALK, PI3K resulting in NOS activation and vasodilation.

Femoral artery ligation resulted in the growth of an identical number of collateral arteries in WT and Mdk−/− mice. However, the collaterals of Mdk−/− mice showed a reduced increase of the luminal diameter with an increase of the ratio of vascular wall to luminal diameter. Thus, perfusion recovery in Mdk−/− mice was impaired due to diminished outward remodeling, although limited vasodilatation might also have contributed (see below). This kind of hypertrophic outward remodeling occurs whenever EC proliferation (intima proliferation) is compromised, while the proliferation of the media and adventitia layer is not affected (Mulvany, 1999). Indeed, our BrdU staining revealed a reduced proliferation of collateral ECs in Mdk−/− mice, while the proliferation of SMC was not negatively affected. Together, these data identify MK as cytokine relevant for collateral EC proliferation and confirm the importance of EC proliferation for the process of arteriogenesis (Moraes et al., 2013).

The initial phase of collateral growth relies on extravasation of leukocytes, especially neutrophils (Chillo et al., 2016). Our immunohistochemical analyses together with our bone marrow transplantation experiments identified neutrophil-domiciled midkine as relevant for arteriogenesis while non-bone marrow cell derived MK influenced perfusion recovery by increased vasodilatation. Although our data evidenced that MK is not released from leukocytes, thereby confirming previous results (Weckbach et al., 2012), we found increased plasma levels of MK 3d and 7d after FAL. FAL results – due to increased shear stress - in arteriogenesis in the upper leg, and timely delayed – due to ischemia in angiogenesis in the lower leg (Ito et al., 1997). Hypoxic ECs were recently identified as source of soluble MK in the vascular system (Weckbach et al., 2012). Accordingly it is likely that the MK observed in plasma derived from hypoxic ECs from the lower leg.

Treatment with rmMK restored perfusion recovery in Mdk−/− mice by promoting EC proliferation, while it promoted vasodilation in WT mice. These results suggested that MK might be involved in regulating NO synthases. Indeed, we found that the expression level of Nos1 and Nos3 was significantly reduced in ECs (but not SMCs) of Mdk−/− mice, whereas the expression level of Nos2 was not affected. To investigate whether reduced expression of NO synthases was causative for reduced EC proliferation in Mdk−/− mice, mice were treated with the NO-donor DETA NONOate. Interestingly, DETA NONate treatment rescued arteriogenesis in Mdk−/− mice. In WT mice, however, DETA NONOate treatment did not improve arteriogenesis, and in contrast to rmMK treatment, DETA NONOate did not promote vasodilation. These data suggest that endogenous produced NO is more effective in vasodilation than an exogenous NO-donor. The receptors for MK are manifold (Weckbach et al., 2011), however, it has been shown in human umbilical vein endothelial cells (HUVECs) that stimulation with MK resulted in binding and phosphorylation of anaplastic lymphoma kinase (ALK, the receptor for PTN (pleiotrophin)) and activation of PI3K (Stoica et al., 2002). There is ample of evidence relating PI3K to Nos1 expression (Fujibayashi et al., 2015) as well as to NOS1 (Wu et al., 2016) and NOS3 (Fisslthaler et al., 2000) activation. Together, these data suggest that ALK might have mediated the vasodilatory function of rmMK in WT mice (see also Fig. 7). However, it might be interesting to mention that intracellular MK is also capable to activate pI3K (Khan et al., 2017).

NO, particularly derived from NOS3 has been described to promote EC proliferation in vitro and angiogenesis in vivo (Fukumura et al., 2001, Cai et al., 2006, Lee et al., 1999). But studies on Nos3−/− mice suggested that NOS3 is not essential for arteriogenesis, and reduced perfusion recovery was attributed to reduced vasodilation (Troidl et al., 2010, Mees et al., 2007). Nos1 has previously been described to be upregulated during arteriogenesis, however, its functional role remains to be elucidated (Pagel et al., 2011). Our current study showed that reperfusion recovery is not reduced in Nos1−/− mice. However, in contrast to WT or Mdk−/− mice, in which treatment with an NO-donor either did not affect (WT mice) or even promoted (Mdk−/− mice) arteriogenesis, it showed deleterious effects on perfusion recovery in Nos1−/− mice indicating that the loss of NOS1 was compensated in a manner not tolerating increased levels of NO.

Whereas NOS2 and NOS3 are mainly described to be involved in NO production, NOS1 has been more implicated in H2O2 generation (Capettini et al., 2010, Costa et al., 2016). However, also NOS3 can generate H2O2 when uncoupled (Li et al., 2002). Interestingly, NO (Cai et al., 2006) as well as H2O2 (Oshikawa et al., 2010, Li et al., 2015) have been shown to play a role in EC proliferation, and both NOS isoforms are described to be capable to substitute for each other (Son et al., 1996). For example it has been show in a rodent model of cerebral aneurisma (CA) formation that expression of Nos1 (but not Nos2) was upregulated in Nos3−/− mice (Aoki et al., 2011). Moreover, the incidence of CA formation was neither affected in Nos1 nor Nos3 deficient mice, however, increased in mice deficient for both genes (Aoki et al., 2011).

We have previously shown that VEGFA is not differentially expressed in the muscle tissue surrounding growing collaterals and is not at all expressed in collaterals (Deindl et al., 2001). Hence it was likely that the increased plasma levels of VEGFA detected 24 h after induction of arteriogenesis derived from a different source. Our immunohistological results evidenced that VEGFA is present in neutrophils. Interestingly, we have recently found that neutrophils are recruited to growing collaterals at the time point we now observed increased plasma levels of VEGFA (Chillo et al., 2016). A study on mesenchymal stem cells revealed that overexpression of MK resulted in increased mRNA expression levels of Vegfa, as well as increased amounts of VEGFA in cell supernatant (Zhao et al., 2014). Accordingly, our results suggest that expression of VEGFA in neutrophils was responsible for increased plasma levels of VEGFA in rmMK treated Mdk−/− mice, but also in WT mice showing MK expression in neutrophils.

Neutrophils are well described to play an important role in angiogenesis by supplying growth factors such as VEGFA (Gong and Koh, 2010, Scapini et al., 2004). Moreover, it has been demonstrated in a femoral artery ligation model that angiogenesis occurring in ischemic tissue is dependent on VEGFA released from neutrophils (Ohki et al., 2005). We have recently shown that angiogenesis is severely compromised in ischemic muscles of Mdk−/− mice after femoral artery ligation (Weckbach et al., 2012). Together, these data suggest that MK is relevant for the bioavailability of neutrophil-derived VEGFA not only in arteriogenesis, but also in angiogenesis.

It has been shown very elegantly by Scapini in vivo (Scapini et al., 2004) that neutrophil-derived VEGFA dependent angiogenesis is mediated by CXCL-1. Moreover, it has been demonstrated by several studies that blocking the CXCL-1 receptor CXCR2 interferes with angiogenesis (e.g. (Sukkar et al., 2008)). Platelets, of which we have shown to play an important role in arteriogenesis by activating neutrophils (Chandraratne et al., 2015, Chillo et al., 2016) are a rich source of CXCL-1. However, CXCL1, is also upregulated in ECs due to shear stress in vitro (Hagiwara et al., 1998) and in growing collaterals in vivo (Vries et al., 2015). Indeed it was recently demonstrated that blocking the CXCL-1 receptor CXCR2 interferes with arteriogenesis (Vries et al., 2015). Together these data suggest that CXCL-1 induced expression and release of VEGFA from neutrophils is mediated by MK. A proposed model for arteriogenesis is shown in Fig. 7.

Although not investigated in the present study, it is likely that at later time points of arteriogenesis (d3 and d7), where monocytes play a role (Troidl et al., 2013), VEGFA is supplied by this type of leukocyte, since also monocytes are a source of both, MK (Badila et al., 2015) and VEGFA (Ramanathan et al., 2003).

In our study, administration of rmVEGFA rescued arteriogenesis in Mdk mice substantiating the relationship between MK and VEGFA. Administration of rmVEGFA to WT mice in contrast showed no effect on arteriogenesis confirming previous results (Jazwa et al., 2016). Together these data suggest that MK mediated increased bioavailability of VEGFA is necessary and sufficient to promote arteriogenesis under normal physiological conditions by regulating the expression level of Nos1 and Nos3 via activation of VEGFR-2. Increasing the current number of mice per group (n = 5 or n = 4) should strengthen the data in future studies.

In summary, we provide evidence that NO synthase, in particular NOS1 and NOS3 isoforms are decisive for EC proliferation in arteriogenesis. Moreover, we show that the expression of these NO synthases is mediated by MK by increasing the bioavailability of VEGFA. Together, these results might have profound impact on the treatment of patients with vascular occlusive diseases.

Funding Sources

This work was supported by the Deutsche Forschungsgemeinschaft (SFB 914/A2 to B. Walzog). J.-I. Pagel and J. Borgolte were supported by the FöFoLe Program from the LMU Munich and T. Lautz by the Lehre@LMU Program from the LMU Munich.

Conflicts of Interest

All authors declare no conflicts of interest.

Authors Contributions

TL, ML, EK, J-IP and JB performed in vivo measurements. TL performed qRT-PCR, in vitro assays and histological analyses. KT performed immunohistology, AC-M helped with qRT-PCR and in vitro analyses, and TL with writing the manuscript. BW participated in scientific discussions. DE performed experimental design, data analysis, conducted scientific direction, and wrote the manuscript.