Cell Origin of Human Mesenchymal Stem Cells Determines a Different Healing Performance in Cardiac Regeneration

The possible different therapeutic efficacy of human mesenchymal stem cells (hMSC) derived from umbilical cord blood (CB), adipose tissue (AT) or bone marrow (BM) for the treatment of myocardial infarction (MI) remains unexplored. This study was to assess the regenerative potential of hMSC from different origins and to evaluate the role of CD105 in cardiac regeneration. Male SCID mice underwent LAD-ligation and received the respective cell type (400.000/per animal) intramyocardially. Six weeks post infarction, cardiac catheterization showed significant preservation of left ventricular functions in BM and CD105+-CB treated groups compared to CB and nontreated MI group (MI-C). Cell survival analyzed by quantitative real time PCR for human GAPDH and capillary density measured by immunostaining showed consistent results. Furthermore, cardiac remodeling can be significantly attenuated by BM-hMSC compared to MI-C. Under hypoxic conditions in vitro, remarkably increased extracellular acidification and apoptosis has been detected from CB-hMSC compared to BM and CD105 purified CB-derived hMSC. Our findings suggests that hMSC originating from different sources showed a different healing performance in cardiac regeneration and CD105+ hMSC exhibited a favorable survival pattern in infarcted hearts, which translates into a more robust preservation of cardiac function.


Introduction
Cell transplantation utilizing different cell types including skeletal myoblasts [1,2], cardiomyocytes [3,4], smooth muscle cells [5,6], bone marrow cells [7] and hematopoietic stem cells [8], has emerged as a promising therapeutic avenue for cardiac regeneration following myocardial infarction damage. Experimental and clinical evidences have demonstrated that transplanting different cell types in the myocardium is safe and could contribute to the angiogenesis in the infarcted area and the improvement of cardiac functions [9][10][11][12][13][14].
Wagner et al. evaluated hMSC from AT, UCB and BM in terms of immune phenotype with a panel of 22 surface antigen markers and analyzed their global gene expression profiles [33]. hMSC populations from different sources displayed similar phenotypic characteristics and a consistent and reproducible gene expression profile [33]. Kern et al. characterized the hMSC pertaining to their morphology, frequency of colonies, expansion characteristics, multilineage differentiation capacity, immunophenotype, and success rate of isolating the cells under identical in vitro conditions [34]. Although phenotypically similar, these culture-expanded hMSC exhibited cell source-related heterogeneity in colony frequency, proliferative potential and differentiation potential [34], suggesting these different sources might greatly affect the survivability and behavior of transplanted hMSC in the hostile environment of ischemia, inflammation, pro-apoptotic factors and scarring from myocardial infarction [35].
In this study we compared the therapeutic potential of hMSC dervied from different sources in cardiac regeneration in a SCID mouse left anterior descending (LAD) ligation model via intracardiac injection with respect to cell survival, infarct size, angiogenesis, cardiac remodelling and fuctional improvement. We found that CD105 + hMSC exhibited a favorable survival pattern in infarcted hearts, which translated into a more robust preservation of cardiac function.

Immunophenotypic analysis and functional differentiation
Cell isolation, expansion, characterization, and differentiation of hMSC have been established according to previous reports [23,33]. The morphology of hMSC from different sources displayed a homogenous spindle-shaped population and maintained a similar morphology during the subsequent passages. FACS analysis was employed to identify the surface marker expression at passage 3. The hMSC culture was shown to be devoid of CD45, which is one of the markers for hematopoietic cells. In contrast, a high expression of CD29, CD44, CD73 and CD105 markers were observed from three different sources ( Figure 1A). Compared to AT-and BM-hMSC, CD105 expression level in CB-hMSC was significantly lower ( Figure 1A and Table 1). The results from FACS analysis were confirmed by immunocytochemistry (data not shown).

CD105 + isolation
For characterization of isolated CD105 + cells from human CBderived hMSC compared to unseparated CB group, surface protein expression of 4 donors immediately after magnetic separation were examined by flow cytometry. A high percentage of cells express CD105 antigen with a mean value over 90% after magnetic isolation whereas a smaller population below 80% express CD105 antigen in unseparated CB group as previously described by Kern et al [34] ( Figure 1A). To observe the CD105 antigen presence in later passages of isolated CD105 + cells, we examined CD105 surface protein expression of 3 donors after passage 4 and 5. During the growth of the cells in culture the CD105 surface marker decreased over the culture time (Table 2). hMSC can differentiate into multiple lineages (such as bone, cartilage, and adipose tissue), and this ability is taken as a functional criterion defining hMSC precursor cells [36]. To verify whether the CD105 enrichment affected the differentiation capacity, the hMSC from different sources and CD105 MSC-CB underwent adipogenic, and chondrogenic differentiation using the methods previously described [37,38] at 24 hours post-enrichment. After 21 days of induction toward an adipogenic lineage, a characteristic morphological change with accumulation of lipid vacuoles was observed ( Figure 1B left). Immunostaining revealed the presence of FABP-4, which is a marker protein for adipocytes. Chondrogenesis was assessed by immunostaining for aggrecan after 4 weeks of culture under chondrogenic conditions. The chondrocyte-like cells showed positive staining for aggrecan protein ( Figure 1B right). Taken together, this evidence indicated that hMSC prior to in vivo experiment retain their multidifferentiation potential into adipogenic, and chondrogenic, lineages.

Cardiac functions
Hemodynamic measurement of the cardiac performance ( Figure 2A) demonstrates an improvement of functional param-eters in case of stem cell treatment both under baseline conditions as well as after stress induction. Figure 2A also shows improved endsystolic values and stroke volume for hearts with implanted human BM-and CD105-purified CB-hMSC (MI-CB105) compared to MI-CB. Functional parameters in MI-BM and MI-CD105 groups present significant improvements on ejection fraction (EF) and stroke work (SW), both under baseline and stress induction in comparison to the MI-CB ( Figure 2B). Significant values have also been found on stroke volume (SV), endsystolic volume (ESV), maximum pressure (Pmax) and cardiac output (CO; Figure 2B).

Infarct size
Ligation of the LAD consistently resulted in a transmural MI with its typical histologic changes including the thinning of the left ventricular free wall (Fast green) and extensive collagen deposition (Sirius red) 6 weeks after infarction. Representative heart sections 6 weeks after myocardial infarction following hMSC application or no injection of cells in the MI-C group are shown in Figure 3A. All hMSC treated animals show a markedly smaller infarct size, while the application of human BM-and CD105-purified CB-hMSC could significantly reduce (p,0.05) the myocardial damage compared to MI-C ( Figure 3B).

Capillary density
The capillary density was determined by CD31-staining 6 weeks after myocardial infarction. Examples of staining from the BZ (border zone) of the infarct area present a lower capillary density in MI-CB treated hearts and the MI-C group. ( Figure 4A) Both, at the BZ as well as the RA (remote area), hearts implanted with AT-, BM-and CD105-purified CB-hMSC show a significant higher capillary density compared to hearts which have been treated with cells derived from CB (MI-CB) and the MI-C group, respectively ( Figure 4B).

Cardiac remodeling
Postinfarct cardiac remodeling serves as an important compensatory mechanism of congestive heart failure, characterized by progressive ventricular chamber dilatation, hypertrophy, fibrosis and prolonged cardiomyocyte apoptosis. Fibrosis resulted in extensive collagen deposition (Sirius red) and increased distance between myocytes (Fast green) 6 weeks after infarction. Figure 5A shows representative staining images from the BZ indicating a higher portion of collagen deposition in the MI-CB and the MI-C group. Hearts implanted with CD105-purified CB-hMSC showed a significant decrease of collagen deposition compared to MI-CB and the MI-C group, respectively in RA and in the BZ ( Figure 5B). Figure 6A represents apoptotic nuclei of cardiomyocytes 6 weeks after myocardial infarction. A significantly reduced percentage of apoptotic cardiomyocytes could be found in the BZ of hearts implanted with CD105-purified CB-hMSC and BM-hMSC, respectively, compared to the MI-C group ( Figure 6B).

Engraftment and characterization of hMSC in infarcted murine hearts
We evaluated human GAPDH expression at 3 different parts from the infarction area and identified implanted cells in the mouse tissue 6 weeks after myocardial infarction following hMSC application with selective binding human nuclei antibody (HNA) ( Figure 7A). Double immunofluorescence staining with HNA and CD31 antibody revealed that at least some of the hMSC appeared to display endothelial cell-like phenotype ( Figure 7B). Six weeks after cell transplantation, we observed a very low number of hMSC colocalized with cardiac Troponin T (cTnT) ( Figure 7C). The frequency of cTnT-HNA double-positive cells from the engrafted stem cells was extremely low. There was no significant difference between different hMSC groups. It was not clear whether the transplanted cells had fused or differentiated into cardiomyocytes. Furthermore, higher human GAPDH expression was detected in the lower and the middle section of the infarcted hearts transplanted with BM-hMSC and CD105-purified CB derived hMSC in comparison to MI-AT and MI-CB ( Figure 7D). There were no significant differences in the upper heart section. A. FACS analysis showed that the cells were negative for CD45 expression and positive for CD29, CD44, CD73 and CD105, which are phenotypes currently known to be characteristic of hMSC. The gray line indicates the control of the CD marker isotypes. B. In vitro differentiation capacity of transplanted hMSC. hMSC from bone marrow were cultured in adipogenic and chondrogenic medium. Chondrogenic differentiation (left). Immunostaining for aggrecan (red). Nuclei were counterstained with DAPI (blue). Adipogenic differentiation (right). Immunostaining with fatty acid binding protein-4 (brown). doi:10.1371/journal.pone.0015652.g001 As the acidification is closely linked to the cellular energy metabolism, we measured the acidification rate of hMSC under normoxic as well as hypoxic conditions. While under normoxic conditions the cells showed no significant difference (data not shown), CB derived hMSC showed significantly higher metabolic activity than AT-, BM-and CD105-purified CB-hMSC under hypoxic conditions ( Figure 8).

Tube formation of hMSC in vitro
To observe the influence of CD105 in the acceleration of network formation we compared nature BM derived hMSC with CD105 low BM-hMSC (antisense) using antisense phosphothiate-ODN. Figure 9 shows the delivery efficiency of antisense into hMSC and antisense delivery decreased network formation of BM-hMSC compared to nature and scrambled group.

Cell proliferation of hMSC in vitro
The MTT assay is widely used to measure metabolic activity and cell proliferation. At different time points we detected metabolic activity of BM105-antisense and BM105-scrambled. After 24 h we observed a reduced viability for BM derived hMSC treated with BM105-antisense compared to scrambled group. The antisense blocking of CD105 seems to lose at later time points we measured ( Figure 9C).

Discussion
The present study for the first time systematically evaluated the cardiac regenerative capability of hMSC derived from different human origins in a SCID mouse left anterior descending (LAD) ligation model via intracardiac injection.
We showed that hMSC originating form different sources could induce significant morphological and functional differences in cardiac parameters. Infarcted hearts with hMSC-injection derived from BM displayed (I) a significant improvement in myocardial performance in comparison to those with hMSC-injection from AT and CB. Furthermore, BM-hMSC treated animals presented (II) a significantly reduced infarction area following diminished cardiac remodeling and (III) a better capillary density in the border zone of the MI. Significant higher localization of human cells could be seen in the middle and apex section of BM-hMSC treated hearts, which might be the result of a migration effect similar than that seen by cardiac stem cells [39]. BM-derived hMSC also showed (IV) the lowest metabolic activity in comparison to all other cells. In addition, the definitive application of a pure fraction of CD105 + -hMSC from CB revealed overall a better myocardial performance than the whole proportion of CB derived hMSC and was similar to that of the MI-BM group.
hMSC are found in many adult tissues and represent an attractive stem cell pool due to their self renewing ability, high proliferative capacity and mesodermal differentiation potential [15]. For the isolation of hMSC, the BM displays one of the main sources next to alternative sites such as CB and AT [26,28]. hMSC derived from different human sources are well characterized and share a consistent and reproducible gene expression profile [33], however they may behave differently with respect to morphology, expansion rate and differentiation potential under in vitro conditions [34,40,41]. It has been not determined if there might be therapeutic differences. In order to compare the differentiation potential of hMSC originating from CB, AT and BM, a MI model in the mouse was established as previously shown by our group [42]. Herein we are able to analyze structural, functional and molecular changes associated with acute MI.
Bieback and coworkers demonstrated that among different stem cell surface markers (CD29, 44, 73, 90, 105), CD105 was significantly lower expressed in CB-derived hMSC than in AT and BM [34]. Therefore, we have also analyzed a very pure fraction of CD105 + -hMSC derived from human CB, which has been purified by an immunomagnetic isolation technique. CD105 or endoglin is a type I membrane glycoprotein, which is located on the cell surface and is also part of the TGF-b receptor complex [43]. Besides cytoskeletal organization, endoglin is also associated with the development of the cardiovascular system and vascular remodeling [43,44]. Furthermore, it is a proliferation-associated and hypoxia-inducible protein which is efficiently expressed in endothelial cells during (tumor) angiogenesis [43,[45][46][47]. Hence, low or less CD105 might be a potential candidate for the overall worse performance of animals treated with the whole fraction of CB-derived hMSC. Additionally, the results of presented study show that after injection of the pure CD105 + -fraction nearly equivalent values could be obtained as seen in the MI-BM-group. This might be due to the fact that CD105 + -cells strongly activate the TGF-b1 receptor pathway which can then interact with downstream signaling to the Smad proteins, which seem to be involved in cardiac fibrosis and scar remodeling [48,49]. The stem cell surface marker CD105 additionally might be of importance during the regeneration process of the infarcted heart. Herein it could be shown that CD105 prevents hypoxia-induced apoptosis in endothelial cells [50] and that downregulation of CD105 mRNA and protein expression resulted in a reduced inhibitory effect of TGF-b1 on cell proliferation, migration and microvessel formation [51]. Further studies with genetic manipulation of stem cells by virus vector [52,53], nanoparticles [54,55] or geneactivated matrix [56,57] and genome-wide transcriptome profiling analysis [58] are necessary to elucidate the underlying mechanism by which CD105 + -cells improve the therapeutic efficacy of cell transplantation in treating myocardium.
Taken together, the presented study demonstrated that hMSC display different regenerative effects in the post-infarct period. Especially for CB-derived hMSC, this might be due the fact of low CD105 purity. These results underscore the importance of a detailed evaluation of the different sources of hMSC prior to their clinical application, in order to increase the patient benefit of stem cell therapy after MI. According to requirements of the START-MSC-Project we used hMSC from the three different human sources cord blood (CB), bone marrow (BM), and adipose tissue (AT) which were isolated and prepared as previously described in detail [34]. We cultured all the hMSCs at 37uC at a humified atmosphere containing 5% CO 2 and a mean cell density of 461610 3 /cm 2 in stem cell medium (MSCGM; Lonza, Walkersville, MD, USA) to 70-80% confluency. Cells were harvested at sub-confluency using trypsin. After the third passage, cells have been used for subsequent in vitro and in vivo experiments.

CD105 + separation
To isolate CD105 + cells from human CB derived hMSC we used magnetic separation with CD105 MicroBeads following the instructions of MACSH (Miltenyi Biotec, Germany) using 20 ml MicroBeads per 250.000 augmented hMSC. The positive fraction was used for subsequent experimentation. A number of 10 7 cells were resuspended in PBS with 2 mM EDTA and 0.5% bovine serum albumin and loaded into the separation columns after MicroBeads incubation. The positive fraction was used for in vivo experiments without further culture. In order to analyze the surface expression of CD105 by Flow Cytometry, magnetic separation was carried on at passage 3. Subsequently, CD105 + signal was monitored at passages 3, 4 and 5.

Immuno phenotypic analysis
Cell surface antigen phenotyping of CB, AT and BM derived hMSC was performed at passage 3. CD105 expression level in

Experimental design of the animal model
The federal animal care committee of LALLF Mecklenburg-Vorpommern (Germany) approved the study protocol (approval number LALLF M-V/TSD/7221.3-1.1-036/07). SCID mice (strain CB17/Icr-Prkdc-scid) were purchased from Charles River Laboratories (Sulzfeld, Germany). SCID mice (male, 2061 g, Charles River Laboratories) were randomly assigned to 5 groups: Sham operation (Sham, n = 10) and 4 MI groups with implanted hMSC of the respective source (MI-CB n = 10, MI-CB105 n = 10, MI-AT n = 10, MI-BM n = 10). Infarcted animals treated with BD Matrigel TM Matrix alone served as controls (MI-C n = 10). A subset of randomly selected mice (n = 7) were assessed for functional measurement, histological and real time polymerase chain reaction (real time-PCR) evaluation at 6 weeks after LAD-ligation.

Generation of MI in mice and stem cell implantation
Mice were anesthetized with tribrom ethanol (AvertinH 0,35 mg/kg, intraperitoneal). After thoracotomy and preparation, the left anterior descending coronary artery (LAD) was permanently ligated. Immediately after LAD-ligation, each mouse received an intramyocardial injection of 400.000 hMSC in BD Matrigel TM Matrix (BD Biosciences USA), or BD Matrigel TM Matrix alone for MI-C similar to a previous study [59]. Along the border of the blanched myocardium 465 ml injections were given. Sham operated mice underwent identical surgical procedures without LAD-ligation but followed by intramyocardial BD Matrigel TM Matrix injection without cells.  via hypertonic saline (5%) injection. Data were analyzed with IOX Version 1.8.3.20 software (emka Technologies). After P/V loop measurements, mice were euthanized. Hearts were arrested in diastole with potassium chloride. Each heart was removed, embedded in O.C.T. TM Compound (Tissue-TekH; Zoeterwoude, Niederlande) and snap-frozen in liquid nitrogen. For histological and biomolecular investigations the infarct area of heart tissue has been divided into 4 horizontal levels from top to bottom within each given amount of 5 mm sections were cut. The three interlayers between the mentioned levels have been collected separately for RNA isolation.

Infarction size and fibrosis analysis
Heart sections of 4 horizontal infarct levels (5 mm) were stained with Fast Green FCF (Sigma-Aldrich) and Sirius Red (Division Chroma, Mü nster, Germany). Two contiguous levels of the heart (n = 7 for each group) which represent the major infarct ratio were analyzed using computerized planimetry (Axio Vision LE Rel. 4.5 software; Zeiss, Jena, Germany). To evaluate fibrosis (n = 5 for each group), the sirius red positive regions of collagen deposition in the remote area (RA) near endocardial border were examined in 10 randomly chosen fields per section (one section per level; 6306) using computerized planimetry. Collagen density was expressed as the ratio of collagen deposition to myocardial tissue in percentage.

Late cardiomyocytes apoptosis
To analyse cardiomyocyte apoptosis, heart sections of two contiguous levels of the heart (n = 5 for each group) which represent the major infarct ratio underwent terminal deoxynucleotidyl transferase-mediated dUTP nick end-labelling assay (DeadEnd Colorimetric TUNEL System; Promega; Madison, WI, USA) according to the manufacturer's instructions. Subsequently, slides were stained with rabbit polyclonal anti-troponin I primary antibody (Santa Cruz; Santa Cruz, USA) and goat antirabbit Alexa-FluorH 568 (Molecular Probes TM ; Carlsbad, USA) conjugated secondary antibody, counterstained with TO-PROH-3 iodide (Molecular Probes TM ) for nuclei and examined with confocal microscope. The number of TUNEL positive cardiomyocytes was counted in 10 randomly chosen HPFs (6306) per section (one section per level) for both RA and border zone (BZ; n = 5 for each group). Results were expressed as the proportion of the TUNEL positive cardiomyocytes nuclei to the total number of cardiomyocytes in percentage.

Determination of capillary density
For immunohistological detection of capillaries, heart sections of two contiguous levels of the heart (n = 5 for each group) which represent the major infarct ratio were immunostained with polyclonal goat anti-CD31 (Santa Cruz) primary antibody followed by anti goat Alexa-FluorH 568 (Molecular Probes TM ) conjugated secondary antibody and counterstained with DAPI. The sections were analyzed within the BZ and RA of the heart. Capillary density was assessed by counting the number of capillaries in 5 RA and 5 BZ randomly-chosen fields (4006). Results were expressed as capillaries per high power field (HPF).

Human cell detection
For identification of implanted hMSC total RNA was isolated from the three separately collected interlayers of cryosectioned hearts (n = 7 for each group) following the instructions of the TRIZOLH Reagent (Invitrogen; Carlsbad, USA). For reverse transcription of total RNA amount (2 mg) and first-strand synthesis of cDNA, SuperScriptH III Reverse Transcriptase (Invitrogen) and oligo (dT) 15

Human cell differentiation potential
In order to investigate the differentiation capacity of hMSC after transplantation into the infarcted heart multiple antibodies staining was performed. Polyclonal goat anti-CD31 (Santa Cruz) primary antibody was initially applied to the section followed by anti goat Alexa-FluorH 488 (Molecular Probes TM ) secondary antibody incubation. Subsequently, human nuclei were stained following the protocol previously described and a rabbit polyclonal anti-troponin I primary antibody (Santa Cruz) was applied to the sections. A goat anti-rabbit Alexa-FluorH 568 (Molecular Probes TM ) was utilized during secondary antibody reaction. Counterstaining was achieved by TO-PROH-3 iodide (Molecular Probes TM ) nuclear staining. The samples were analyzed using a LSM 780 confocal microscopy (Carl Zeiss, Jena).

Real time acidification of viable hMSC in vitro
The silicon sensor chip technology allows the observation of cellular behaviour in cell cultures. Online monitoring was performed with the BionasH 2500 analyzing system (Bionas, Rostock, Germany) [60,61]. The cells were seeded in duplicate 24 h before measurement directly on the chip surface to assure highly specific signal detection. The cell concentration was adapted in such a way that the cells reach approximately 80% confluence on the sensor chip after 24 h. The measurement is noninvasive and label-free. The media flow over the cells is stopped periodically. Breakdown products (lactate, CO 2 ) and the oxygen consumption of cells result in a change of pH and oxygen content in the medium. These changes are measured in the stop phases of the pump cycle. In the following pump phase ''used'' medium is exchanged for fresh medium. The stop and go cycle of 8 min each is carried out over the whole experiment. Acidification rates are calculated as the slope of changes in every stop phase related to the basal signal in %. The changes of pH are measured on the sensor chip by Ion Sensitive Field Effect Transistors (ISFETs). For measurement in the BionasH 2500 analyzing system medium without bicarbonate buffer (running medium) with 1 mM HEPES, 0.1% FCS, 10.000 U penicillin and 10 mg streptomycin/ ml was used. The pH of the running medium was adjusted to 7.4 and the osmolarity to 290 mOsm/kg. For measurement under hypoxic conditions the analyzing system was operating in a nitrogen environment. At the end of the experiment the cells were killed by addition of 0.2% Triton X-100 to the running medium to get a basic signal without living cells on the sensor surfaces (negative control).

In Vitro Functional Differentiation Assay
To induce adipogenic differentiation, hMSC from different sources were seeded at a density of 3610 3 cells per cm 2 and cultured for up to 3 weeks in cell culture medium supplemented with 1028 M dexamethasone, 2.5 mg/ml insulin, and 100 mM indomethacin. To induce chondrogenic differentiation, 3610 5 hMSC were cultured in 1 ml of chondrogenic induction medium (cell culture medium supplemented with 0.1 mM dexamethasone, 1 mM sodium pyruvate, 0.17 mM l-ascorbic acid 2-phosphate, 0.35 mM l-proline, 6.25 mg/ml insulin, 6.25 mg/ml transferrin, 6.25 ng/ml selenite, 5.33 mg/ml linolic acid, 1.25 mg/ml bovine serum albumin, and 0.01 mg/ml transforming growth factor-b3) in the tip of a 15-ml conical tube to allow aggregation of the cells in suspension culture. The induction of chondrogenic differentiation was performed for 4 weeks. The differentiation capacity toward different cell lineages was verified by morphology changes and immunostaining for specific markers, that is, aggrecan for chondrocytes and, fatty acid binding protein (FABP-4) for adipocytes

Antisense-Oligodesoxynukleotide (ODN) blockade
BM-hMSC at passage 3 was seeded 24 h before transfection. The cell concentration was adapted in such a way that the cells reach approximately 60% confluence on the well after 24 h. To reduce the expression of CD105 in BM derived hMSC phosphothiate-ODN with the antisense-sequence 59-ATGCTGTCCACGTGGG-39 (Eurogentec, Germany) was transfected into the cells by Lipofectamine 2000 (Invitrogen) as the manufacturer described. A scrambled-ODN with the nonsense sequence 59-ACTCGTGC-TACGGTGG-39 (Eurogentec) was used as control.

Tube forming assay
To observe the network formation potential of BM derived hMSC with reduced expression compared to nature cells, BM- hMSC were seeded in 24-well plates and transfected as described. After 24 h 7610 4 cells were cultured in 4-well plates with 200 ml BD Matrigel TM Matrix (BD Biosciences) at 37uC and a humified atmosphere containing 5% CO 2 . After 30 min 200 ml endothelial cell medium (EGM-2; Lonza) was applied to each well, cultured continuously and daily changed.

Statistical analysis
Statistical analysis was performed using SigmaStat 3.0 (Chicago, USA). Results are expressed as mean 6 SEM. Overall comparisons of the treatment groups were performed by using the one-way analysis of variance (ANOVA) method that applies post-hoc multiple Holm-Sidak tests, and by using the nonpara- metric Kruskal-Wallis (failing normality) or post-hoc multiple Dunn tests. P values ,0.05 were considered as statistically significant.