Retrieval of germinal zone neural stem cells from the cerebrospinal fluid of premature infants with intraventricular hemorrhage

Abstract Intraventricular hemorrhage is a common cause of morbidity and mortality in premature infants. The rupture of the germinal zone into the ventricles entails loss of neural stem cells and disturbs the normal cytoarchitecture of the region, compromising late neurogliogenesis. Here we demonstrate that neural stem cells can be easily and robustly isolated from the hemorrhagic cerebrospinal fluid obtained during therapeutic neuroendoscopic lavage in preterm infants with severe intraventricular hemorrhage. Our analyses demonstrate that these neural stem cells, although similar to human fetal cell lines, display distinctive hallmarks related to their regional and developmental origin in the germinal zone of the ventral forebrain, the ganglionic eminences that give rise to interneurons and oligodendrocytes. These cells can be expanded, cryopreserved, and differentiated in vitro and in vivo in the brain of nude mice and show no sign of tumoral transformation 6 months after transplantation. This novel class of neural stem cells poses no ethical concerns, as the fluid is usually discarded, and could be useful for the development of an autologous therapy for preterm infants, aiming to restore late neurogliogenesis and attenuate neurocognitive deficits. Furthermore, these cells represent a valuable tool for the study of the final stages of human brain development and germinal zone biology.


| INTRODUCTION
Intraventricular hemorrhage (IVH) is a common complication of premature infants, occurring in 15% to 40% of preterm infants weighing less than 1500 g at birth and being particularly common in extremely low birthweight neonates. 1-3 IVH is classified into four grades according to the extent of hemorrhage, development of subsequent ventricular dilatation, and parenchymal involvement: grade I-a hemorrhage restricted to subependymal region; grade II-a hemorrhage bleeding into de ventricles without dilation; grade III-an IVH with ventricular dilatation; and grade IV-an IVH with associated adjacent brain parenchyma infarction. 4 Some authors classify grade IV IVH separately because the presence of periventricular hemorrhagic infarction or other parenchymal lesions is generally not caused simply by extension of IVH into brain parenchyma and should thus be considered as a different pathological condition. 5 Around 10% to 15% of preterm infants develop severe (grade III-IV) IVH and those infants are at high risk to develop posthemorrhagic hydrocephalus, an expansion of the ventricles due to cerebrospinal fluid (CSF) accumulation (reviewed in Reference 6), and to present long-term neurological deficits with cognitive and motor disabilities. 6,7 IVH initiates in the periventricular germinal zone (Gz), also known as germinal matrix, a highly proliferative, highly vascularized region around the lateral ventricles with a dense and fragile, endothelial-lined, vessel network. 8 From 24 to 32 weeks of gestation the Gz is most prominent in the caudo-thalamic groove, forming the ganglionic eminences of the ventricular zone (VZ), where late migrating cortical and thalamic neurons and oligodendrocyte precursors are born. 9 The ultimate cause of Gz bleeding remains unclear, but it is commonly accepted that it results from the combination of Gz vasculature vulnerability and blood pressure fluctuations associated with prematurity. 8 In the hemorrhage phase, there is a rupture of the Gz, occurring most often at the level of the medial ganglionic eminence, that entails loss of neural stem cells (NSCs) and disturbs the cytoarchitecture of the zone leading to abnormal neuronal, ependymal, and glio-genesis. 10,11 Current treatments for IVH are intended to decrease the intracranial pressure that can cause periventricular white matter compression and damage, impairment of brain development, and even death. 12,13 There is a standardized protocol neither for the type nor for the timing of the intervention, [14][15][16] but it has recently been shown that early removal of hemorrhagic CSF by neuroendoscopic lavage is a safe procedure that may mitigate the adverse effects of the accumulation of blood products, decrease the need for permanent shunt placement, and potentially reduce neurological disability. [17][18][19] NSCs are the self-renewing, multipotent cells that generate neuronal and glial cell populations during development. During brain development, primary NSCs located in the VZ have radial processes (apical radial glial cells), divide symmetrically, are highly polarized, and express prominin-1 (CD133). 20 Radial glia are more multipotent than the intermediate progenitors of the subventricular zone (SVZ) that include basal radial glia, transient amplifying, and neural progenitors. 21,22 Regional differences in the transcriptional profile of radial glia dictate the fate of their postmitotic progeny. In addition to their differentiation potential, NSCs produce neurotrophic and neuroprotective molecules, making them attractive for regenerative approaches. 23,24 In this regard, clinical-grade human NSC lines are usually obtained from the fetal central nervous system (CNS).
Human NSCs can also be isolated from the adult CNS in patients undergoing surgical procedures, 25 or be derived from pluripotent stem cells (PSC) and from somatic cells through reprogramming protocols. 24,26 Advanced therapy medicinal products based on clinical-grade human allogenic fetal NSCs are being tested in clinical trials for various neurological disorders and, although efficacy has yet to be ascertained in a clinical setting, their safety profile has been repeatedly confirmed (reviewed in Reference 24).

Significance statement
Intraventricular hemorrhage (IVH), occurring in 15% to 40% of preterm births, is frequently associated with long-term neurological deficits. The rupture of the proliferative germinal zone in IVH disturbs late neuronal, ependymal, and glio-genesis. Using a minimally invasive neuroendoscopic procedure, neural stem cells can be retrieved from the cerebrospinal fluid, which can be expanded, cryopreserved, and differentiated in vitro and in vivo, and are not tumorigenic.
These cells display distinct hallmarks related to their origin in the germinal zone of the ventral forebrain and could be useful for the development of an autologous cell therapy aiming to attenuate neurocognitive sequelae.
Here we demonstrate that a novel class of NSCs can be robustly isolated from the hemorrhagic CSF of preterm neonates during neuroendoscopic lavage. These NSCs that we named Gz-NCS display distinctive features corresponding to their origin in the ventral forebrain. Gz-NSCs could be useful to develop an autologous cell therapy aiming to reduce neurological disability in preterm infants and to further our understanding of human Gz biology.  Scientific #A12856-01) before cell seeding. Cells were expanded for 3 (early) and 7-10 (late) passages for characterization. Passage 7, which corresponds to 13 ± 1 accumulated population doublings, was considered "late passage" given that it will not be possible to extensively expand the cells in a clinical setting.

| Immunofluorescence
Cells grown over Matrigel-coated coverslips were fixed with 4% para-    Table S1. For Ki-67 detection, we first performed an antigen retrieval step in which cells were heated for 10 seconds in a microwave with citrate buffer pH 6.0 (Sigma-Aldrich #C9999) letting cells cool down 20 minutes. Acquisition of fluorescence images was performed in a Leica TCS-SP5 or a Nikon Eclipse Ti fluorescence microscope. Images were processed using the Adobe Photoshop CS5 or ImageJ software. 28 Positive cells were counted using the ImageJ software from at least three random fields per preparation. Purification efficiency was determined by flow cytometry analysis staining with CD34-phycoerythrin (PE; Miltenyi Biotec #130-120-520) antibody. Further analyses were performed using the Transcriptome Analysis Console (TAC, Affymetrix) v4.0 software and R version 3.5.0. 33 Functional enrichment analysis was performed using the bioinformatics tool EnrichR (http://amp.pharm.mssm.edu/Enrichr/). 34,35 Neuroanatomical references were obtained from the Allen Atlas of the developing human brain (www.brainspan.org). 36

| Statistics
Data are presented as mean ± SEM. Significance was determined using two-tailed Student's t test for comparisons between two samples. P < .05 was considered significant. Paired t-tests or repeated measures (R) ANOVA were used to compare samples from the same individual at different stages. All statistical analyses were performed using the GraphPad Prism 8.01 software. Bioinformatic analyses were performed using the Affymetrix and R software, using t, ANOVA, and R-ANOVA tests and selected thresholds, as indicated in the text and figure legends. A false discovery rate (FDR) <5% was established for significance.

| Hemorrhagic CSF of preterm IVH patients contains NSCs
Eight consecutive cases with a clinical and radiological diagnosis of grade IV IVH (Table 1) Figure 2D). Likewise, expression of the stem cell marker prominin-1 (CD133) was maintained through passages (58.99 ± 8.7 vs 57.93 ± 10.82; Figure 2E). An exception was the 42-weeks-old sample ( Figure 2E, pink symbols) in which the percentage of CD133 + cells dropped drastically upon passaging. This case was excluded from further analyses given that IVH in full-term neonates most often originates in the choroid plexus. 8 We next analyzed whether the cell population obtained from hemorrhagic CSF had a similar expression pattern of CD surface antigens than that described for fetal NSCs. 40,41 Like fetal NSCs, most cells in CSF samples were positive for CD133 and all were negative for CD45, displaying a variable expression of CD24 ( Figure 2E and Table S3). Intriguingly, some samples contained a substantial percentage of CD34 positive cells ( Figure 2E and  Figure 2F and Table S3).
We attempted to recover cells from subsequent CSF samples, col- 3.2 | NSCs isolated from the hemorrhagic cerebrospinal fluid present distinctive regional hallmarks We next performed a transcriptomic analysis to study the differences and similarities between NSCs isolated from hemorrhagic CSF, fetal forebrain NSCs and NSCs derived from iPSCs. Given that CSF samples contained mostly blood cells we also included in the analysis F I G U R E 3 NSC cells isolated from the CSF display a ventral forebrain gene-expression profile. A, PCA analysis of global gene-expression profiles. B, Venn diagrams showing the number of differentially expressed genes (DEG), 2-fold change, FDR P < .05. C, Volcano plot of DEG in NSCs from fetal brain (red) and CSF (blue) sources. Highlighted are markers that identify regional populations including genes that have been previously associated with germinal zones and forebrain regionalization (see also schematic in D) and putative candidates for prospective identification of germinal zone NSCs. Expression levels of NSC and regional forebrain markers (D) and candidate DEG genes that could identify this NSC population (E). F, Semi-quantitative RT-PCR of NSC markers and candidate DEG genes. G, Enrichment network analysis of upregulated genes relative to fetal NSCs, profiled across brain regions according to the Allen brain atlas, and schematic neuroanatomical representation on coronal brain  Figure 3A). Pairwise comparisons showed a significant overlap in the expression profiles of the three types of NSCs ( Figure 3B).
Notwithstanding, there were 1073 differentially expressed genes (DEG) between CSF-derived NSCs and fetal NSCs, using a false discovery rate (FDR) P value < .05 and ±2-fold change ( Figure 3C). Consistent with an NSC identity, expression of radial glia and neural progenitor markers, such as SOX2, FABP7, FOXG1, and NES was similar in CSF-derived and fetal NSCs ( Figure 3D). GFAP was highly expressed in both types, but significantly higher in the NSCs from CSF. GFAP expression is restricted to the VZ during primate brain development. 36 Figure 3C). There were also remarkable differences in the expression of forebrain regional transcription factors ( Figure 3D) with high expression of ventral and posterior forebrain markers, OTX2 and NKX2.1, while dorsal ones such as PAX6 and GSX2 were lower than in fetal NSCs. In addition, we identified several markers that could provide a distinctive signature for this NSC population ( Figure 3C,E). Among those, there was a remarkable upregulation of genes related to antigen presentation and immune response, in particular pertaining to the major histocompatibility complex II (MHCII) ( Figure 3E), which according to the developmental human brain atlas are highly expressed in germinal zones during midgestational stages (www.brainspan.org). 36 Differential expression of selected transcripts was validated by PCR ( Figure 3F) Interestingly, enrichment analysis showed that the genes upregulated in CSF-derived NSCs relative to fetal NSCs mapped to the ventral forebrain structures, including the periventricular nuclei-basal ganglia, thalamic, and septal nuclei ( Figure 3G). This regional topography corresponds to the anatomical structures surrounding the ganglionic eminences, most often affected by IVH in preterm infants. This is consistent with an origin of the cells isolated from the CSF in the germinal zone of the ventral forebrain and therefore we have named them Gz-NSCs.
Next, we used immunofluorescence to evaluate the expression at the protein level of typical radial glia markers such as SOX2, nestin, and brain lipid binding protein (BLBP, FABP7). Quantification confirmed expression of these proteins by most cells (SOX2: 96.57 ± 1.56; NESTIN: 84.40 ± 4.27; BLBP: 80.80 ± 10.83; Figure 4A). In addition, the majority of the cells expressed regional transcription factors corresponding to ventral and posterior forebrain,  Figure 4B). Other differentially expressed transcripts that we selected based on a putative membrane localization were also expressed at the protein level ( Figure 4C). However, one of these, PLPP4, a poorly characterized phospholipid phosphatase expressed in the brain (www. proteinatlas.org) 44 showed a clear nucleolar localization pattern.
To establish the differentiation potential of Gz-NSCs, we cultured the cells in FBS without mitogens for 2 weeks. Cells showed in vitro trilineage potential, upregulating neuronal, astrocyte, and oligodendrocyte markers in various proportions as shown in the graph ( Figure 4D), and downregulating Ki67 (P < .01 compared with undifferentiated cells).
Because OLIG2 is also expressed in neuronal progenitors, we examined the coexpression of OLIG2 and doublecortin (DCX). About 10% (9.9 ± 7) of the OLIG2 positive cells coexpressed DCX in these in vitro conditions, representing 1.5% of the total cells ( Figure 4E). To further confirm differentiation into glial lineages, we also examined the expression of PDGFR, another oligodendrocyte marker, and S100β that was consistently coexpressed with GFAP in Gz-NSCs differentiated in 2% B27 for 2 weeks ( Figure S4).

| CD133 + purified cells maintain Gz-NSC features
Cells initially isolated from hemorrhagic CSF samples are a heterogeneous mixture of cellular types at different developmental and maturation stages, in particular taking into account that all these cases had parenchymal involvement (grade IV). Therefore, in order to better define putative Gz-NSC-specific features and obtain a more homogenous population for future in vivo applications, we selected CD133 + cells by magnetic activated cell sorting (MACS).
CD133 has been used for the isolation of NSCs from normal brain tissues and CD133 + cells differentiate in vitro and in vivo into the three neuroectodermal lineages. 40,[45][46][47] Following MACS purification, we could expand and cryopreserve the cells, which maintained their typical morphology and the expression of CD133 ( Figure 5A).
The percentages of CD34 + and CD24 + double positive cells were variable and did not significantly change with sorting, although we observed that CD34 + cells tended to remain in the negative fraction ( Figure 5B). We examined transcriptomic changes related to in vitro propagation and CD133 purification comparing early, late, and CD133 + -sorted Gz-NSCs (R-ANOVA, FDR F < 0.05) with no significant DEG between early and late passages ( Figure 5C). Nevertheless, although there were no DEG at FDR < 0.05 between early and late passages, they appeared to be segregated along PCA2 (y-axis, Figure 5C), therefore we further analyzed the data using a less stringent cutoff ( Figure S5).
Those analyses showed that, upon passaging, markers corresponding to more mature phenotypes tended to decrease, without weakening the NSC ventral identity ( Figure S5).
Upon purification, there were significant changes relative to both early and late passage cells. Enrichment analysis of the common genes (early and late vs CD133 + ) showed that the expression of genes implicated in neural and synaptic specific pathways was decreased ( Figure 5D). On the other hand, CD133 + -sorted cells showed a relative enrichment in genes expressed at less differentiated stages and in less specific pathways ( Figure 5D). Transcriptomic analysis confirmed that CD133 + cells maintained the expression of radial glia and NSC markers as well as the pattern of regional transcription factors. Most of the putative membrane markers that we had selected, including genes related to antigen presentation, were also expressed ( Figure 5E). In contrast, (TREK2) expression was significantly decreased, but was still significantly higher than in fetal NSCs.
We validated the expression of our candidate genes at the protein level using flow cytometry, before and after MACS enrichment ( Figure 5F). The selected markers were expressed at the protein level by the majority of the cells (>90.7 ± 9%) and the expression was in general maintained or increased after CD133 purification with rare exceptions. MHCII was more variable, although in this case we cannot discard internalization. In contrast, TREK2 expression was maintained in most cells (97.5%-95.8%) despite the decrease at the RNA level.

| Transplantation study
To evaluate the safety of Gz-NSCs, we first verified their trophic factor dependence ( Figure S6A). Cells were expanded until passage 12-which corresponds to 22.25 ± 4.54 accumulated population doublings-and maintenance of normal karyotype was verified for the cell lines at this stage ( Figure S6B). Recovery of purified and unpurified cells upon thawing and other cell growth characteristics relevant to scale-up manufacturing were also analyzed and are shown in Figure S6.
Next, we transplantedCD133 + purified cells in the striatum of nude mice ( Figure 6A). None of the transplanted animals presented weight loss, neurological focal signs, or any adverse reactions for the duration of the study. Using a human-specific antibody (huN), Gz-NSCs were identified in the striatum of all animals at 3 weeks (n = 3) and 6 months (n = 7) after transplantation, albeit at low numbers ( Figure 6B). Transplanted cells formed small grafts that did not cause anatomical distortion and a few migrated over the striatum and adjacent white matter tracts, occasionally reaching the VZ ( Figure 6C). Transplanted cells showed low mitotic activity measured by Ki67 expression (4.59 ± 0.76 and 5.97 ± 2.22; Figure 6D).
Phenotypic analyses showed the presence of human cells coexpressing GFAP, βIIItubulin or OLIG2 with huN ( Figure 6D). At both time points, the cells appeared to predominantly adopt an oligodendrocyte fate.

| DISCUSSION
We report here the isolation of a distinct class of NSCs, the Gz-NSCs, from hemorrhagic CSF samples of premature neonates diagnosed with IVH grade IV. It is well established that NSCs from a given anatomical location give rise to corresponding regional cellular subtypes and, thus, display transcriptional and phenotypical differences with NSCs from other locations. [48][49][50] We describe here features of Gz-NSCs related to their regional and developmental origin in the ventral ganglionic eminences of the forebrain, which set them apart from other available human fetal forebrain NSC lines. Importantly, human interneuron neurogenesis continues into the third trimester of gestation, largely at the medial ganglionic eminence, which is also a source of oligodendrocyte precursors. Previous studies have shown that late neurogenesis is suppressed in premature births and that IVH arrests proliferation in the Gz at the level of the ganglionic eminences, 9,11 which could result in decreased GABA interneuron production and decreased myelination due to loss of oligodendrocyte precursors. These perturbations may contribute to persistent impairments in neurocognitive function in these children.
F I G U R E 5 Gz-NSC signature is maintained after CD133 sorting. A, Cell morphology and flow cytometry of CD133 after MACS purification.
Results are the mean of 6 independent biological replicates. Scale bar: 100 μm. B, Flow cytometry analysis of CD24 and CD34 before and after purification for CD133. Percentages of CD24 + and CD34 + before sorting (blue) and in both the CD133 negative (white) and CD133 positive (green) fractions showed no significant differences. Data are shown as mean ± SEM of 6 independent biological samples. C, PCA analysis of early, late, and sorted Gz-NSC populations. D, Venn diagrams representing the transcriptional changes related to cell propagation (early vs late) and CD133 sorting (2-fold change, R-ANOVA FDR F < 0.05). There were no DEG between early and late passages. Genes downregulated in the CD133 sorted cells with respect to early and late passages corresponded to GO pathways related to neuronal and synaptic activity while those upregulated in the sorted cells were indicative of a less differentiated stage. E, Expression of NSCs, regional and Gz-NSC markers at the RNA level in early, late, and CD133 + purified cells. In addition to conspicuous differences in the expression of regional transcription factors, we report the expression of novel Gz-NSC markers that differentiate these cells from fetal (dorsal) forebrain NSCs and were maintained after sorting for CD133 + cells, such as PODXL, IL1RAP, HLA-DR, DLK1, and FZD5. These markers could serve to isolate and identify human Gz-NSCs. However, more cases are required to validate these given that they can be developmentally regulated and our samples show some heterogeneity. PODXL is an interesting glycoprotein involved in apical polarity, which belongs to the CD34 family of syalomucins, whose absence has been reported to cause ventricular enlargement in mice. 51  Here we describe a novel class of NSCs, the Gz-NSCs that can be easily and robustly isolated from the CSF of preterm neonates with grade IV IVH undergoing neuroendoscopic lavage. We found that these cells, while being similar to fetal forebrain NSCs, have several distinctive hallmarks related to their regional and developmental origin in the ventral forebrain. Gz-NSCs can be expanded and cryopreserved showing in vitro and in vivo differentiation potential, and pose no ethical concerns as the fluid is usually discarded. Thus, Gz-NSCs could represent an optimal source for the development of an autologous cell therapy for infants with IVH, as well as a useful tool for studying the late stages of human neural development.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available upon reasonable request to the corresponding author.