Neuroprotection of the hypoxic-ischemic mouse brain by human CD117+CD90+CD105+ amniotic fluid stem cells

Human amniotic fluid contains two morphologically-distinct sub-populations of stem cells with regenerative potential, spindle-shaped (SS-hAFSCs) and round-shaped human amniotic fluid stem cells (RS-hAFSCs). However, it is unclear whether morphological differences correlate with functionality, and this lack of knowledge limits their translational applications. Here, we show that SS-hAFSCs and RS-hAFSCs differ in their neuro-protective ability, demonstrating that a single contralateral injection of SS-hAFSCs into hypoxic-ischemic P7 mice conferred a 47% reduction in hippocampal tissue loss and 43–45% reduction in TUNEL-positive cells in the hippocampus and striatum 48 hours after the insult, decreased microglial activation and TGFβ1 levels, and prevented demyelination. On the other hand, RS-hAFSCs failed to show such neuro-protective effects. It is possible that SS-hAFSCs exert their neuroprotection via endoglin-dependent inhibition of TGFβ1 signaling in target cells. These findings identify a sub-population of CD117+CD90+CD105+ stem cells as a promising source for the neuro-protection of the developing brain.

is a major healthcare burden, being the fourth leading cause of childhood mortality and resulting in one million neuro-disabled children each year 9,10 . Whilst moderate hypothermia is now standard care, it is only partially effective and cannot be applied to pre-term babies 11 . With currently no effective cure addressing the underlying loss of cerebral tissue, there is an urgent need to develop a simple, safe and effective treatment to directly modulate pathophysiological processes and protect the developing brain of HIE-affected babies. Stem cell therapy has the potential to lessen brain injury either by replacing lost cells, promoting the differentiation of host progenitors, and/or modulating the host immune system 12,13 .
Here we show that, contrary to the results obtained with RS-hAFSCs, a single contralateral injection of SS-hAFSCs into the hypoxia-ischemia (HI) mouse brain contributed to lessening the size of the brain lesion, reduced the number of TUNEL + cells, decreased microglial activation, prevented demyelination and reduced TGFβ1 levels. We are the first to demonstrate that SS-hAFSCs, which can easily be isolated from heterogeneous populations of hAFSCs based on their cell surface co-expression of CD105 and CD90, have potential for the protection of the developing brain.

Results
Differential cell surface molecule expression in SS-hAFSCs and RS-hAFSCs. The human mid-trimester amniotic fluid contains a heterogeneous population of plastic-adherent cells presenting either a polymorphic round-shaped appearance (RS-hAFSCs) or an elongated spindle-shaped cytoplasm (SS-hAFSCs) (Fig. 1a,b). The cells studied here were karyotypically normal, isolated from the same gestational age (2 samples16 weeks and 3 days of gestation and 2 samples 16 weeks of gestation) and pooled from four different donors undergoing prenatal diagnostic (healthy pregnancy) to reduce inter-individual variability. Both sub-populations were characterised by the positive expression of the stem cell marker CD117 in the cytoplasm (which is expressed on the cell surface of freshly isolated samples but gets internalised during ex vivo expansion) (Fig. 1b), the positive expression of the MSC marker CD73 and the negative expression of the hematopoietic markers CD34 and CD45 (Fig. 1c). However, contrary to SS-hAFSCs, RS-hAFSCs do not express the cell surface markers CD90 (Thy-1) and CD105 (endoglin) (results obtained from 7 samples of SS-hAFSCs and 6 samples of RS-hAFSCs), indicating that these surface markers can also be used for purification of heterogeneous cell adherent CD117 + (C-Kit + ) human amniotic fluid stem cell populations.

Injection of SS-hAFSCs decreases microglial activation.
To analyse the mechanistic role of donor stem cells, we assessed microglia activation by the expression of CD11b (αM) (Fig. 3a) and cell morphology (Fig. 3b), (uninjured control group in Supplementary Figure 1c). HI insult led 48 hours later to strong microglia activation. We showed that, in comparison to PBS-injected brains, injection of SS-hAFSCs significantly decreased CD11b expression in the hippocampus (3.  (Fig. 3a). However, SS-hAFSCs injection had no effect on CD11b expression in the cortex (1.2 ± 0.1, n = 22 vs. 1.0 ± 0.2, n = 16; P = 0.06) and external capsule (2.9 ± 0.2 vs. 2.5 ± 0.3, n = 16;  Injection of SS-hAFSCs prevents demyelination in the striatum and external capsule. Neuronal myelination was assessed by measuring the intensity of myelin basic protein (MBP) immunostaining in the striatum and external capsule. Results revealed that in PBS-injected brains, the HI insult significantly decreased MBP luminosity in the external capsule (43.5 ± 2.3, n = 10 vs. 28.7 ± 1.5, n = 18, P = 0.0007, two-way ANOVA followed by Bonferroni's multiple comparison test), but was not significant in the striatum (38.3 ± 2.1 mean ± SEM, n = 11 vs. 27.7 ± 1.5, n = 18; P = 0.21) (Fig. 5a). However, injection of SS-hAFSCs tended to prevent demyelination in both regions (35.8 ± 6.1, n = 14; P = 0.18 in the striatum and 35.2 ± 2.5, n = 14; P = 0.21 in the external capsule). Notably, the injection of RS-hAFSCs did not show any positive effect on myelination in either brain regions (25.6 ± 3,4, n = 15; P = 0.99 in the striatum and 26.1 ± 3.4, n = 15; P = 0.99 in the external capsule) (Fig. 5a). Immunohistochemistry for MBP confirmed the trend towards prevention of demyelination in the HI brains injected with SS-hAFSCs (Fig. 5b). Injection of SS-hAFSCs led to a 65% decrease in TGFβ1 + cells. The number of hippocampal TGFβ1 + cells was significantly increased in the brains of mice that received an HI insult and were not injected (HI) and those who received an HI insult but were injected with PBS alone (PBS), compared to those that did not receive any HI insult (naïve) (100.1 ± 10.3, n = 9 vs. 96.9 ± 8.5, n = 17 vs. 40.2 ± 5.7, n = 6; P = 0.003, one-way ANOVA followed by Bonferroni's multiple comparison test). Injection of SS-hAFSCs, but not RS-hAFSCs, led to a 65% decrease in TGFβ1 + cells (58.6 ± 8.6, n = 16; P = 0.005 for SS-hAFSCs and 101.6 ± 7.4, n = 15, P > 0.99 for RS-hAFSCs) (Fig. 6a). Co-immunostaining for TGFβ1 and either CD11b (microglia) or GFAP (astroglia) on sections of PBS-injected brains confirmed reduced levels of TGFβ1, CD11b and GFAP levels in SS-hAFSC-transplanted mice, as shown in Fig. 6b. Neuroprotection of the HI mouse brain by SS-AFSCs is likely to be via endoglin-dependent inhibition of TGFβ1. We injected 100,000 cells in 2 μl PBS in the contralateral side of the brain and assessed the presence of donor cells in the ipsilateral side 48 hours later using an anti-human Nuclear Antigen antibody, which reacts with human but not mouse tissue. We found that donor SS-hAFSCs or RS-hAFSCs were absent in the ipsilateral side of the brain, suggesting that the beneficial effects observed by donor cells are not mediated via direct differentiation but through the release of soluble factors.

Discussion
Human amniotic fluid, which can be retrieved during caesarean section or at birth, is a valuable source of fetal stem cells that can be used without ethical concerns for allogeneic and autologous applications in regenerative medicine. These cells present a higher ability to differentiate and exhibit higher growth kinetics than adult stem cells, whilst showing no sign of tumorigenicity 3 . Despite having been characterised by their plastic-adherence and expression of the stem cell surface marker CD117, human amniotic fluid stem cells (hAFSCs) are heterogeneous with regards to their morphology. We, and others, have in particular distinguished between cells presenting a spindle-shaped (SS-hAFSCs) and round-shaped (RS-hAFSCs) morphology, which show differences with regards to their proteomic profile, growth characteristics and differentiation potential 7,17 . In the present study, we isolated and pooled clonal populations of SS-hAFSCs and RS-hAFSCs and showed that both cell types express CD73 but, unlike RS-hAFSCs, SS-hAFSCs uniquely co-express CD90 (Thy-1) and CD105 (endoglin). Thy-1 has several cellular functions, being involved in cell-cell interactions, regulating apoptotic signalling, modulating proliferation, survival and cytokine/growth factor response, and promoting T cell activation 18 . Endoglin is part of the TGFβ receptor complex, thus modulating the response to the binding of TGFβ, activin A and BMPs. It plays an important role in angiogenesis 19 . However, there is no evidence that phenotypical differences translate into in vivo functional differences. Here, we show that a single injection of SS-hAFSCs into the HI mouse brain immediately after the HI insult resulted 48 hours later in a significant reduction of brain lesion size and TUNEL-positive cells. Notably, injection of RS-hAFSCs failed to show such neuro-protective effects. Donor cells were absent from the site of injury 48 hours after intracranial injection, suggesting they mediate their neuro-repair effects through paracrine/endocrine mechanisms. In this study, we pooled together several donor samples, to reduce inter-individual differences. Further analysis will be required to develop in vitro screening platforms to compare the therapeutic efficacy of individual samples. The HI mouse model used in this study, which is a modification of the Vannucci model of unilateral ligation/occlusion of the common carotid artery at postnatal day 7 followed by a period of hypoxia, presents some phenotypic similarities to periventricular white matter damage in humans. These similarities include prevention of demyelination, ventricular enlargement, loss of neurons, damage to axons and dendrites, oligodendrocyte cell death and alterations of oligodendrocyte development, which are the hallmarks of brain injury in preterm/term infants 20 . We are the first to report the heterogeneity of the hAFSC population in relation to its neuro-regenerative properties, and to identify a sub-population of CD117+ spindle-shaped cells that can be easily isolated and purified based on their spindle-shape morphology and co-expression of CD90 and CD105.
In the specific disease model used here, we demonstrated that understanding the mechanistic properties of the cells is important to further improve stem cell-based therapy. Specifically, our data show that both cell types decreased reactive astrogliosis, indicating that the neuro-protective effects of SS-hAFSCs are not attributable to a reduction in astroglial activation and indeed astrogliosis failed to influence on the extent of HI brain injury in neonatal mice 20 . However, SS-hAFSCs, but not RS-hAFSCs, lowered the levels of astroglial TGFβ1, decreased microglia activation and prevented demyelination, suggesting that donor SS-hAFSCs modulated the host immune system by decreasing resident TGFβ1 levels, thereby lowering the microglial response to neuronal death and restoring the oligodendrocyte/astrocyte differentiation balance. Not surprisingly, further analysis of   the phenotypical differences between SS-hAFSCs and RS-hAFSCs showed that only SS-hAFSCs release endoglin. These results indicate that inhibition of the TGFβ1 signaling pathway in target cells may be mediated by the release of endoglin.
In conclusion, this study provides the first evidence for the neuro-protective effects of a subset of spindle-shaped CD73 + CD90 + CD105 + fetal stem cells that can be easily and ethically isolated from mid-trimester amniotic fluid and will provide the basis for the development of innovative stem-cell based therapeutics for the protection of the developing brain. Cell culture. Human mid-trimester amniotic fluid was collected at mid-gestation from normal pregnancy and immediately spun before the cell pellet was resuspended into a single cell suspension in StemMACS MSC expansion media under xeno-and serum-free conditions at low density (10 2 cells/cm 2 ) onto a plastic culture dish and let to expand until clones of >50 cells formed. The clones had a homogeneous morphology, with some clones presenting a round-shaped cytoplasm, whilst others presented a spindle-shaped one. 5-10 clones presenting similar morphology were mechanically detached, pooled, spun, resuspended into single cells and plated (10 4 cells/cm 2 ) on plastic culture dishes without feeders in Dulbecco's modified Eagle's medium (DMEM-HG) (Invitrogen) supplemented with 10% fetal bovine serum (Biosera), 2 mM L-glutamine, 50 IU/ml penicillin and 50 mg/ml streptomycin (Gibco-BRL) (D10), at 37 °C in a 5% CO 2 incubator.

Enzyme-linked immunosorbent assays (ELISAs).
Cells were plated at 2 × 10 4 cells/cm 2 in four wells of a 6-well plate. 72 hours after seeding, 8 ml conditioned medium was removed and concentrated using an Amicon ultra-4 15 ml 3 K NMWL filter by centrifugation at 4000 g for 40 minutes. 300 µl of D10 was then added to the concentrated medium and 100 µl of this was added per well of the ELISA plate. The human TGFβ1and endoglin ELISA kit (Abcam) were performed according to the manufacturer's instructions. Hypoxic-ischemic insult. The animals used were from the C57/BL6 strain (Charles River, Kent, UK). The surgical procedures to induce HI insult were performed as previously described 21 . Briefly, seven-day-old mice were anaesthetised with isofluorane (5% induction followed by 1.5% maintenance) before undergoing permanent left common carotid artery occlusion with 8/0 polypropylene suture. The mice were left to recover at 36 °C and returned to the dam for 2 hours before being placed in a hypoxia chamber and exposed to humidified 8% oxygen /92% nitrogen for 1 hour (severe insult). Mice were randomized to receive either 100,000 cells (SS-hAFSCs or RS-hAFSCs) in 2 μl PBS or with PBS only (with no cells) via freehand intracranial injection through the skull in the brain hemisphere contralateral to the arterial occlusion. The mice were then returned to the dam and sacrificed 48 hours later by intraperitoneal injection of pentobarbitone and perfused with 30 ml 4% paraformaldehyde in PBS (intracardiac injection). Mice from at least five different litters were used in each experimental group, and mice from each litter were randomly assigned to all experimental groups, taking gender into account in a way that both genders were equally distributed among groups.  21 . The brains were then frozen on dry ice, cut on a cryostat into sequential 40 μm sections and stored at −80 °C until processed for analysis by an experimenter blinded to the group analysed.
Detection of donor cells in the ipsilateral side of the brain was performed 48 hours after injection in the contralateral side using anti-human Nuclear Antigen antibody (Abcam).
Scoring and measurement. The brain infarct volume was measured on Nissl-stained sections, as previously reported 20 . Briefly, sections scans were imported into the Optimas 6.5 image analysis software (Bothell, WA, USA) and areas of intact staining were outlined and measured bilaterally in cortex, hippocampus, striatum, pyriform cortex, thalamus and external capsule. The percentage of tissue loss was the calculated by converting the measured areas into square millimetres and converting the volume by multiplying by 400 μm. The sum of these volumes was used to calculate the percentage of intact brain tissue as injured/intact volume x 100.
For TUNEL immunoreactivity, the number of positive cells was counted bilaterally in three different optical fields at 20× magnification in cortex, hippocampus, striatum, pyriform cortex, thalamus and external capsule, and the counts were averaged per animal and per group.
The intensity of GFAP staining was quantified using optical luminosity values. Images for both ipsilateral and contralateral sides of the brain were captured with a Sony AVT-Horn 3CCD colour video camera (24 bit RGB, 760 × 570 pixel resolution) and three different optical fields in the cortex, hippocampus, striatum, pyriform cortex, thalamus and external capsule, as well as the surrounding glass at x20 magnification. We used Optimas 6.5 software to obtain the mean and standard deviation (SD) for optical luminosity values (OVL). SD was subtracted from the values obtained for the surrounding glass, as previously described 22 . Statistical analysis. Statistical analysis was performed using GraphPad 7.0b. Two-way ANOVAs, followed by Bonferroni's multiple comparison post-hoc test, were used to test if there was an interaction between variables and obtain p values. One-way ANOVA were used to determine if there was any statistical difference between the means of more than two independent groups. Unpaired T-test was used to compare the means of two independent groups. P < 0.05 was considered significant.