Neuronal hypoxia in vitro: Investigation of therapeutic principles of HUCB-MNC and CD133+ stem cells

Background The therapeutic capacity of human umbilical cord blood mononuclear cells (HUCB-MNC) and stem cells derived thereof is documented in animal models of focal cerebral ischemia, while mechanisms behind the reduction of lesion size and the observed improvement of behavioral skills still remain poorly understood. Methods A human in vitro model of neuronal hypoxia was used to address the impact of total HUCB-MNC (tMNC), a stem cell enriched fraction (CD133+, 97.38% CD133-positive cells) and a stem cell depleted fraction (CD133-, 0.06% CD133-positive cells) of HUCB-MNC by either direct or indirect co-cultivation with post-hypoxic neuronal cells (differentiated SH-SY5Y). Over three days, development of apoptosis and necrosis of neuronal cells, chemotaxis of MNC and production of chemokines (CCL2, CCL3, CCL5, CXCL8, CXCL9) and growth factors (G-CSF, GM-CSF, VEGF, bFGF) were analyzed using fluorescence microscopy, FACS and cytometric bead array. Results tMNC, CD133+ and surprisingly CD133- reduced neuronal apoptosis in direct co-cultivations significantly to levels in the range of normoxic controls (7% ± 3%). Untreated post-hypoxic control cultures showed apoptosis rates of 85% ± 11%. tMNC actively migrated towards injured neuronal cells. Both co-cultivation types using tMNC or CD133- reduced apoptosis comparably. CD133- produced high concentrations of CCL3 and neuroprotective G-CSF within indirect co-cultures. Soluble factors produced by CD133+ cells were not detectable in direct co-cultures. Conclusion Our data show that heterogeneous tMNC and even CD133-depleted fractions have the capability not only to reduce apoptosis in neuronal cells but also to trigger the retaining of neuronal phenotypes.


Background
Transplantation of adult stem cells has been shown to be an auspicious and effective treatment for degenerative and traumatic neurological diseases [1]. Among degenerative neurological disorders acute ischemic stroke is the leading cause of disability and death in industrial nations [2][3][4].
Acute stroke leads to an increased release of hematopoietic stem and progenitor cells from bone marrow into peripheral blood [5]. It is assumed that these cells take part in self-healing processes occurring after neuronal injury. They are supposed to promote the survival of the injured brain tissue by producing neurotrophic factors [6], to enhance endogenous angiogenesis [7] and neurogenesis [8] or even to transdifferentiate into neuronal cells [9]. However, the stroke induced endogenous release of hematopoietic stem and progenitor cells seems not to be sufficient to compensate massive loss of brain tissue after extended ischemic stroke. Therefore, external application of hematopoietic stem and progenitor cells is expected to complement current treatment of acute stroke based on thrombolytic therapy. An appropriate source of hematopoietic stem cells is the mononuclear cell (MNC) fraction of human umbilical cord blood (HUCB) [10][11][12]. Transplantation of HUCB-MNC as well as enriched HUCB hematopoietic stem cells into animals which were subjected to focal stroke caused by middle cerebral artery occlusion (MCAO) ameliorated the animals' functional outcome and reduced the lesion size [13]. However, there are still manifold unanswered questions addressing the beneficial influence of such grafts on injured neuronal cells.
It has been documented that there is no neuronal transdifferentiation of hematopoietic stem cells in vitro [14][15][16]. Though so far there is no convincing proof that locally administered hematopoietic stem cells transdifferentiate into functionally neuronal cells forming the basis of the animals' behavioral progression [17].
It has recently been shown that there is no need for MNC to enter the brain for neuroprotection. Soluble factors like GDNF, NGF, BDNF or G-CSF are known to promote neuroprotection over long-distances [18,19]. This raises many questions about the cellular mechanisms causing the functional improvement after grafting [20]. Prevention of neurons from apoptotic cell death [21] is considered to be supported by the transplantation and could be directly connected to improved tissue conservation, lesion size reduction and superior functional outcome [22].
Cell culture models of neuronal hypoxia complement the exploration of particular interactions between grafts and neuronal tissue. Our study is based on a well established post-hypoxic neuronal cell culture model (SH-SY5Y). This model was used to address (i) the neuroprotective potential of stem cell enriched and -depleted HUCB derived cell fractions, (ii) the impact of these cells especially on apoptotic status of oxygen-deprived neurons, and (iii) the mediation of cell-derived survival signals (soluble or cellattached).

Direct co-cultivation with each fraction of HUCB-MNC reduced apoptosis in post-hypoxic neuronal cells
Hypoxic cultivation (48 hours) of fully matured neuronal SH-SY5Y cells resulted in an initial rate of apoptosis of 26% ± 13%. Within the following three days rate of apoptosis increased to 85% ± 11%. By contrast, normoxic control cultures showed a stable amount of apoptotic cells (7% ± 3%) over the whole observation time (data not shown). Direct co-cultivation with tMNC and CD133showed pronounced reduction of neuronal apoptosis. Similar results were obtained after application of CD133 + . By application of 4.5 × 10 3 CD133 + they were given in equal amounts as they exist in tMNC. Though the whole cell amount of CD133 + was 100 times less then tMNC administered (Fig. 1A).
Levels of necrosis in post-hypoxic control cultures remained nearly stable (approximately 25%) over three days. tMNC and CD133cell application also induced a significant reduction of necrosis. CD133 + cells did not influence the level of necrosis (Fig. 1B).

In indirect co-cultures tMNC and CD133were also sufficient to decrease apoptosis of post-hypoxic neuronal cells
Over the entire observation period, direct as well as indirect co-cultivation with tMNC or CD133exhibited a significant reduction of apoptosis. In all co-culture set-ups percentage of annexin-V positive cells was significantly lower (p ≤ 0.001) than in post-hypoxic control cultures ( Fig. 2A, B). Direct and indirect co-cultivation of CD133resulted in similar rates of apoptosis continuously below 5% of annexin-V positive cells (Fig. 2B). However, still generating strong neuroprotective effects, the number of apoptotic neuronal cells in indirect tMNC co-cultures was significantly higher than in direct co-cultures at Day 2 and Day 3 (p ≤ 0.001) as shown in figure 2A. Direct co-cultivation with tMNC resulted in a stable level of 6% ± 1% neuronal apoptosis and was therefore significantly lower than in post-hypoxic control cultures (Day 2: 46% ± 20%; Day 3: 85% ± 11%).
When tMNC were indirectly co-cultured neuroprotection was as pronounced as in direct co-cultures on the first day after hypoxia. Two and three days after application there was still a significant, but compared to control cultures reduced, protective effect in the indirect co-cultures while protection in direct co-cultures was as distinct as on Day 1 (7% ± 8%).
The comparison of both application types using tMNC and CD133showed that soluble factors seem to have strong therapeutic potential ( Fig. 2A).
The positive influence of indirect co-cultivation on the amount of apoptotic cells, revealed by annexin-V detection, was also confirmed by typical patterns in the cleavage of PARP, a late marker of apoptosis (Fig. 3). In indirect co-cultures with tMNC and CD133quantities of cleaved PARP were nearly at the same level on Day 1 posthypoxia. On Day 2 and Day 3 neuroprotection by CD133resulted in concentrations of cleaved PARP ranging in the level of the normoxic control (Fig. 3).
Only direct co-cultures of tMNC and CD133displayed improved protection from necrosis/late apoptosis as revealed by Propidium Iodide labeling of post-hypoxic neuronal cells (Fig. 2C, D). Indirect co-cultivation did not reduce the percentage of necrotic/late apoptotic neuronal cells.

tMNC localized in close proximity to post-hypoxic neuronal cells
In direct co-cultures many tMNC were found in close spatial relation with post-hypoxic neuronal cells already at Day 1 (Fig. 4B). This became even more evident at later time points (Fig. 4D, F). At Day 3, the vast majority of tMNC was found adjacent to neuronal somata and processes (Fig. 4F). Interestingly, co-cultivation with tMNC seemed to have strong positive effects on the preservation of typical neuronal cell morphology as the formation of branched processes.

Cytokine secretion patterns induced by direct cocultivation with tMNC or CD133were similar; cytokines produced by CD133 + were not detectable
In supernatants of direct co-cultures and corresponding mono-cultures we assessed concentrations of CCL2, CCL5, CXCL8, VEGF and bFGF (Fig. 5).
Post-hypoxic neuronal cells secreted the chemokines CXCL8 and CCL2 and the growth factor VEGF. Low amounts of CCL5 and bFGF were detectable, as well.
Mono-cultures of tMNC and CD133displayed similar growth factor and chemokine secretion patterns: CXCL8 and CCL2 were produced in considerable amounts while CCL5 and bFGF were secreted at low levels. VEGF was not present. None of the investigated soluble factors was detected in supernatants of purified CD133 + mono-cultures.

Up-regulation of soluble factors in indirect CD133cocultures was associated with long term neuroprotection
In a second set of experiments we investigated different effects of direct and indirect co-culturing on cytokine secretion. Therefore, concentrations of CXCL9, CCL3, VEGF, G-CSF and GM-CSF were measured in normoxic and post-hypoxic mono-cultures, and co-cultures with tMNC and CD133 -.
In co-cultures with tMNC direct co-cultivation generated more pronounced effects than indirect co-cultivation, as seen for the relative up-regulation of CXCL9, GM-CSF (both, p ≤ 0.05) and CCL3 and the relative down-regulation of VEGF.
Only direct CD133co-cultivation had an impact on regulation of CCL3, VEGF and GM-CSF. Secretion of all three cytokines was markedly decreased compared to controls whereas indirect co-cultivation with CD133had no effect.
In contrast, the secretion of CXCL9 and G-CSF (p ≤ 0.01) was markedly induced by indirect co-cultivation with CD133 -. For these cytokines direct co-cultivation showed no effect (G-CSF) or resulted in a remarkably reduced production (CXCL9) as compared to indirect co-cultivation with CD133 - (Fig. 6).
The experiment also revealed the influence of different HUCB-MNC-derived cell preparations on the cytokine secretion.
Most prominent, direct co-cultivation with tMNC significantly increased expression of CCL3 (p ≤ 0.05) whereas direct co-cultivation with CD133resulted in a relative reduction of this chemokine (p ≤ 0.01).

Discussion
It has been previously shown that HUCB derived MNC as well as nearly pure stem cell populations obtained from MNC are able to improve the clinical outcome of animals after MCAO [23]. In an in vitro model of neuronal hypoxia we discern cell populations within MNC being able to improve neuronal survival and disclose potential neuroprotective mechanisms.
First we studied the effects of tMNC application on the apoptotic status of post-hypoxic neurons. We found that direct application of tMNC results in preservation of neuronal morphology based on a constant protection from apoptosis ( Fig. 2A).
For in vivo experiments it has been described that none or only a minority of systemically administrated cells were detected in the brain while large quantities were found in the spleen, in the lungs and in the blood of the animals. Nevertheless, some studies showed that cell-treatment improved behavioral deficits [18,23,24]. This indicates the importance of soluble factors for neuroprotection. Our data obtained from indirect co-cultures strongly support this hypothesis. tMNC application was highly protective, although the anti-apoptotic effect was slightly weakened after two days ( Fig. 2A). Probably, soluble factors act as "first aid" messengers while longer protection seems to demand close proximity between tMNC and neuronal cells. The comparison of direct application of tMNC to normoxic and post-hypoxic neuronal cells revealed that only post-hypoxic neuronal cells attracted tMNC (Fig. 4). The clustering of tMNC around neuronal cells could explain the enhanced protection from apoptosis in direct co-cultures with tMNC at Day 2 and Day 3 (Fig. 4, Fig. 2A). Spatial contiguity is possibly associated with higher concentrations of neuroprotective mediators in the micro-environment of injured neuronal cells. Post-

FR Q WU R O G LU HF W FR FX OW X UH LQ G LU H F W F R F X OW X UH
*0 &6) &&/
To focus the neuroprotective effects of stem cells we enriched or depleted CD133 + cells from HUCB-MNC. Purities of 97.38% for the stem cell preparation and a reduction of CD133 + up to 0.06% for the stem cell depleted fraction were achieved. In direct co-cultures CD133 + were applied to neuronal cells in quantities of 4.5 × 10 3 resembling 1% of the applied tMNC, since CD133 + account for about 1% of total cell number in tMNC preparations. Neuroprotective capacity of nearly pure CD133 + cells was comparable to tMNC (Fig. 1A). According to our analyses of cytokine production CD133 + cells did not secrete measureable cytokines neither in mono-cultures nor in co-cultures with injured neuronal cells (Fig. 5). Additional experiments using even tenfold elevated numbers of stem cells (4.5 × 10 4 ) also revealed no detectable cytokine concentrations and did not result in an increased neuroprotective capacity (data not shown). Neuroprotection by the absence of measurable soluble mediators argues for a stem cell specific therapeutic mechanism through contiguity. Due to this assumption we did not include CD133 + in the investigation of indirect co-cultures in this study.
Surprisingly, CD133cell fractions were also highly sufficient in protection from apoptosis. Therefore, the observed anti-apoptotic neuroprotective effects of tMNC in our experiments do not only relay on hematopoietic stem cell-specific mechanisms. This assumption is supported by the missing significance in the effect of CD133 + on necrotic/late apoptotic loss of post-hypoxic neuronal cells (Fig. 1B). tMNC and CD133significantly reduced the percentage of PI-positive neuronal cells, whereas CD133 + did not. Since CD133in indirect co-cultures were superior to tMNC in protection from apoptosis and because of very low amounts of CD133 + in this preparation, CD133seem to mediate additional neuroprotective effects. Possibly, the separation process influenced the functionality of CD133cells. FACS analyses of the activation markers CD25, CD38, CD71 and HLA-DR on tMNC and on the separated CD133fraction did not reveal an increased population of activated cells in our preparations (data not shown). However, the expression of other activation molecules cannot be excluded, since this cell population displayed an enhanced secretion of cytokines like G-CSF, GM-CSF and CCL-3 (Fig. 6). We also cannot rule out that remaining CD34 + cells could account for the neuroprotective activity of the CD133fraction, since depletion did only reduce the number of CD34 + cells to 77% (Tab. 1).
The investigation of soluble mediators exhibited that hypoxia induced a significant increase of VEGF in neuronal cells (Fig. 6, [28]). VEGF is documented to inhibit proapoptotic signaling by Bad (BCL2 antagonist of cell death), and cleavage of caspase-3, and caspase-9 [29] and therefore can be claimed as an autocrine self-protection mechanism of damaged neuronal cells. The neuroprotective impact of tMNC and CD133cells in direct applications was accompanied by prevention of VEGF production which is typically induced in post-hypoxic neuronal cells (Fig. 5).
Neuroprotective effects of cell application could be mediated by G-CSF that was found only in mono-cultures of CD133but not in mono-cultures of post-hypoxic SH-SY5Y cells. Schneider et al., 2005 [30] pointed out that human SH-SY5Y neuroblastoma cells express the G-CSF receptor and that activation by the neurotrophic G-CSF reduced NO-induced poly-ADP ribose polymerase (PARP) and caspase-3 cleavage. Our data support these observations. At Day 3 in indirect co-cultures supply of G-CSF by CD133was associated with cleaved PARP levels in the range of normoxic cultures (Fig. 3, Fig. 6). G-CSF levels in mono-cultures of tMNC were only slightly above detection limit and could not provide this anti-apoptotic effect (Fig. 6).
The decrease of cleaved PARP observed in post-hypoxic neuronal mono-cultures at Day 3 (Fig. 3) was probably induced by rising lack of energy due to increased rate of apoptosis (Fig. 1).
Noticeably, there are application-specific differences in the regulation of cytokine secretion in co-cultures with CD133fractions. Concentrations of CCL3 and G-CSF were significantly higher in indirect co-cultures than in direct co-cultures. Possibly these enhanced concentrations are responsible for a protection from apoptosis in indirect co-cultures similar to that in direct co-cultures at Day 3. Indirect co-cultivation with tMNC did not induce enhanced cytokine levels (Fig. 6) and at the same time did not exert the same neuroprotective effect on post-hypoxic SH-SY5Y neurons. This could be explained by spatial effects: paracrine released cytokines could be more effective than action of cytokines over a longer distance.

Conclusion
In this study we investigated human umbilical cord blood derived cell populations (tMNC, CD133 + , and CD133 -) according to their ability to protect post-hypoxic neuronal cells.
For different reasons, as the missing systemic effects and the disregard of brain cell interactions this in vitro system does only simplified reflect the action of MNC after hypoxic brain lesions in vivo. But taken this into account, our study delivers useful indications for the in vivo application of such cells: So, since purified CD133 + fractions are not superior to total HUCB-MNC in mediating neuroprotective antiapoptotic effects, expensive and time consuming stem cell separations are not necessarily needed to yield neuroprotective cell populations. Furthermore, our study underlines the importance of MNC derived soluble factors for the mediation of neuroprotective effects visible as prevention of neuronal cells from apoptosis. Therefore, future therapeutic approaches should focus on the sufficient supply of soluble anti-apoptotic mediators, to reduce post-hypoxic brain damage.

Preparation of HUCB samples and isolation of CD133 + cells
HUCB samples of healthy full-term neonates were obtained in accordance with ethical prescripts immediately after delivery. Samples were processed and analyzed as described previously [25]. The total MNC (tMNC) fraction gained from Ficoll density gradient (Tab. 1) was stored in the gaseous phase of liquid nitrogen. Cellular sub-fractions of tMNC were characterized using CD3-Phycoerythrin (

Direct and indirect co-cultivation of post-hypoxic neuronal cells with tMNC, CD133 + and CD133cell fractions
Subsequent to hypoxia, direct and indirect co-cultivation with tMNC, CD133 + or CD133was carried out under normoxic conditions over a period of three days. Added tMNC, CD133 -(4.5 × 10 5 cells, both) and CD133 + (4.5 × 10 3 cells) were dissolved in 500 μl co-culture medium (DMEM-Ham's F12, 5 ng/ml BDNF and 0.1% HSA) and added to differentiated post-hypoxic neuronal SH-SY5Y cells cultivated in adequate volume of post-hypoxic medium. Cell ratio of post-hypoxic neuronal cells to MNC was 1:15 and to CD133 + 1:0.15. For indirect co-cultivation tMNC or CD133were added in cell impassable cell culture inserts with a pore size of 0.4 μm (Greiner Bio-One GmbH, Frickenhausen, Germany).
Prior to co-cultivation with post-hypoxic neuronal cells, tMNC as well as CD133 + and CD133were labeled with CFSE.

Cell viability assay of neuronal cells
Within direct and indirect co-cultures and control cultures the influence of tMNC, CD133 + and CD133on neuronal viability was detected via i) Propidium Iodide (PI, Invitrogen, Karlsruhe, Germany) assay for necrosis and late apoptosis and ii) annexin-V assay (Becton-Dickinson, Heidelberg, Germany) for apoptosis. The PI-and annexin-V assays were performed in cell culture plates as described previously [25]. tMNC, CD133 + and CD133were distinguished from annexin-V-PE or PI positive neuronal cells by the green CFSE staining.
Cytometric Bead Array for human apoptosis (CBA, Becton Dickinson, Erembodegem, Belgium) was used to quantify the apoptosis specific parameter cleaved Poly-ADP-Ribose-Polymerase I (PARP) in lysates of post-hypoxic neuronal cells after indirect co-cultivation with tMNC and CD133 -. For cell lyses adherent neuronal cells were rinsed with PBS and incubated on ice in the supplied buffer for 20 minutes. For analyses pooled samples obtained from three independent experiments were used.

Cytokine profiling
For cytokine profiling supernatants of direct and indirect co-cultures were analyzed on Day 3. Supernatants of tMNC, CD133 + and CD133mono-cultures and those of post-hypoxic neuronal cells were also investigated on Day 3. Cytokines were simultaneously measured using CBA for human soluble proteins (Becton Dickinson, Erembodegem, Belgium). Supernatants were screened for the fol-lowing cytokines: CCL2, CCL3, CCL5, CXCL8, CXCL9 and for the growth factors basic Fibroblast Growth Factor (bFGF), Granulocyte Colony-Stimulating Factor (G-CSF), Granulocyte-Macrophage Colony-Stimulating Factor (GM-CSF) and Vascular Endothelial Growth Factor (VEGF). The detection limit was 20 pg/ml, except for CCL2 and VEGF (40 pg/ml).

Statistical analyses of data
Except for apoptosis and necrosis rates all results have been reported as mean ± SD. Statistical differences were analyzed by Student's t-test or Mann-Whitney rank sumtest. P values of ≤ 0.05 were considered statistically significant (* p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001). Apoptosis and necrosis rates were logit-transformed to obtain normally distributed quantities. The effects of time, experimental setting (post-hypoxia), experimental run and the investigated well were determined univariately, and, finally multivariately using a mixed-model approach with time and experimental setting as fixed effects and well and experimental run as random effects.
Cytokine concentrations of indirect and direct co-cultures with tMNC, CD133and CD133 + were compared with the sum of the concentrations obtained after post-hypoxia of neuronal cells and corresponding tMNC, CD133 + and CD133mono-cultures using a bootstrapping algorithm. Therefore, we added concentrations which were resampled from e.g. the experiment post-hypoxia and the tMNC mono-culture and compared the results with concentrations obtained from e.g. the experiment of indirect co-culture with tMNC. Results were compared with Student's ttest or Mann-Whitney rank sum test. P-values reported are based on 10,000 bootstrapping simulations.
Box plots (if applicable) and univariate analyses were determined using the software package SPSS (SPSS Inc., Chicago IL, USA). Mixed Model analyses were performed using PROC MIXED of the statistical software package SAS 9.1 (SAS Institute Inc., Cary, NC, USA). Bootstrapping analysis was performed using the statistical software package "R" [31].