Microglia and Stem-Cell Mediated Neuroprotection after Neonatal Hypoxia-Ischemia

Neonatal hypoxia-ischemia encephalopathy (HIE) refers to a brain injury in term infants that can lead to death or lifelong neurological deficits such as cerebral palsy (CP). The pathogenesis of this disease involves multiple cellular and molecular events, notably a neuroinflammatory response driven partly by microglia, the brain resident macrophages. Treatment options are currently very limited, but stem cell (SC) therapy holds promise, as beneficial outcomes are reported in animal studies and to a lesser degree in human trials. Among putative mechanisms of action, immunomodulation is considered a major contributor to SC associated benefits. The goal of this review is to examine whether microglia is a cellular target of SC-mediated immunomodulation and whether the recruitment of microglia is linked to brain repair. We will first provide an overview on microglial activation in the rodent model of neonatal HI, and highlight its sensitivity to developmental age. Two complementary questions are then addressed: (i) do immune-related treatments impact microglia and provide neuroprotection, (ii) does stem cell treatment modulates microglia? Finally, the immune-related findings in patients enrolled in SC based clinical trials are discussed. Our review points to an impact of SCs on the microglial phenotype, but heterogeneity in experimental designs and methodological limitations hamper our understanding of a potential contribution of microglia to SC associated benefits. Thorough analyses of the microglial phenotype are warranted to better address the relevance of the neuroimmune crosstalk in brain repair and improve or advance the development of SC protocols in humans. Graphical abstract


Introduction
Perinatal hypoxic-ischemic (HI) brain injury is the result of a decreased blood and/or oxygen supply to the brain, namely asphyxia. It can occur in preterm and term infants (see Box 1 on definitions), in which latter case the condition is termed neonatal HI encephalopathy (HIE). The risk factors of hypoxia-ischemia are multiple, and include fetal, maternal, or placental conditions [1]. The incidence of HIE is estimated to be around 1.5 per 1000 live births in developed countries [2], and it is an important cause of mortality in neonates [3]. Surviving infants, depending on the severity of HI and maturational state of the brain, may suffer lifelong neurological sequelae, such as hearing and visual impairments, sensorimotor or learning disorders, seizures, and cerebral palsy (CP) [4].
At present, therapeutic hypothermia (TH) is the only approved intervention for term neonates diagnosed with moderate or severe HIE; it requires to be initiated within six hours after birth, and consists of either whole-body cooling or selective head cooling for up till 72 h to reach a body core temperature of 34 ± 0.5°C. Randomized controlled trials of TH demonstrate a reduction in mortality without an increase in major disabilities in surviving infants at 18-22 months [5], but the narrow time window of application and incomplete efficacy of TH call for the development and evaluation of complementary and/or alternative interventions.
Stem cell (SC) treatment represents such a promising therapeutic option, as it allows for a wider therapeutic window in comparison to TH [6], and therefore is more clinically relevant. Preclinical studies indicate that SC based therapy can ameliorate or modify diverse aspects of HI-induced cerebral damages such as infarct size, apoptosis levels, axonal sprouting, neurite outgrowth, microglial activation, and provide some degree of functional improvements (e.g. motor, sensorimotor and cognitive behaviors) [7][8][9][10]. The results of SC based clinical trials for CP are also encouraging and show a modest yet significant effectiveness in improving gross motor function [11]. While the underlying biological mechanisms still remain elusive, the modulation of the immune system, in the central nervous system (CNS) and/or in the periphery is often put forward as a major mechanism responsible for this therapeutic benefit. Neonatal HI indeed elicits a sterile immune response that plays a crucial role in the progression of HI-induced injury. The innate immune system is activated from the very early phases until the tertiary phase of HI. In the brain, this process involves principally microglia-the resident myeloid cells-and astrocytes, which are also immune competent cells. Mast cells have also been shown to contribute to the neuroinflammatory response [12,13]. A role for various peripheral innate immune cells into HI cerebral damage has also been demonstrated, in particular for monocytes that can infiltrate the brain parenchyma [14,15], as well as for neutrophils [16][17][18] and natural killer cells [17,19]. An adaptive immune response implicating T cells also seems activated after neonatal HI [17,20,21], but it is less characterized than that of the innate immune arm. Purinergic signaling (e.g. ATP/ inflammasome axis [22]), specific cytokines, chemokines and their downstream signaling pathways also play a role in sterile inflammation and in the evolution of HI-induced cerebral injury [17,[23][24][25][26][27][28][29][30][31][32]. The complexity and multiplicity of all these inflammatory components has been the subject of recent extensive and comprehensive reviews [33][34][35][36]. Thus, to avoid redundancy, a particular emphasis is placed on discussing the microglial phenotype after neonatal HI in light of recent advances in the field, and the impact of developmental age at time of injury on the microglial response. The link between microglia and brain repair is then assessed by reviewing whether and how immune related treatments and stem cell therapy provide neuroprotection and impact microglia. Finally, the clinical evidence for immune modulation by SC and association with neuroprotection and neuroregeneration is discussed. The focus of this review being neuroinflammation/ microglia and SC therapy in the context of HIE, only the findings from the corresponding postnatal days (P) 7-10 rodent model will be discussed.

Box 1 Definitions
Cerebral palsy (CP) is a neurodevelopmental disorder. Its most recent definition is as follows [37]: "CP describes a group of permanent disorders of the development of movement and posture, causing activity limitation, that are attributed to nonprogressive disturbances that occurred in the developing fetal or infant brain. The motor disorders of cerebral palsy are often accompanied by disturbances of sensation, perception, cognition, communication, and behaviour, by epilepsy, and by secondary musculoskeletal problems." Prematurity refers to any live birth before 37 weeks of gestation, and is further divided in moderate or late preterm (32 to <37 weeks), very preterm (28 to <32 weeks), and extremely preterm (<28 weeks) [38]. Term pregnancy for singleton until recently referred to 37-42 completed weeks of gestation. Due to heterogeneity in neonatal outcomes depending on gestational age within this five weeks' time [39], the American College of Obstetricians and Gynecologists Committee has now refined the notion of term and suggests to use "full term" for deliveries between 39 through 40 weeks of gestation [40]. This new definition of term pregnancy is not yet reflected in the clinical definition of neonatal HIE, which very often refers to term pregnancy as 36 weeks of gestation or more.
Human neonate (synonym for newborn): the neonatal period corresponds to the first 28 days of life after birth. extracellular accumulation of excitatory amino acids, and death of neurons, mainly through necrosis but also through apoptosis. Depending on the severity and duration of the HI episode, a latent phase may follow during which partial recovery of neuronal metabolism can occur; although its duration and exact timing is unknown, it is often considered that this phase is an appropriate time window to begin potential neuroprotective treatments, such as TH or SCs injection, in order to limit the ongoing toxic processes and prevent the progression towards the secondary phase. In the absence of intervention or if the primary phase is prolonged and severe, then the latent phase is short and a secondary delayed energy failure phase (six to 72 h after the insult) ensues, during which further deterioration and cell death will occur, characterized by excitotoxicity, mitochondrial failure, acute inflammation, oxidative stress, and increased seizure activity. Finally, there is evidence that some injury processes may persist over months or perhaps years, resulting in so called tertiary damages [45]; the suggested mechanisms underlying such damages include astrogliosis, persistent inflammation and epigenetic changes, all of which may lead to exacerbated cell loss and sensitization to potential second hits.

Modeling HIE in Rodents: The Rice-Vannucci Model
The most widely used experimental model to study HIE is the Rice-Vannucci model [46]. Initially developed in the immature Sprague-Dawley rat, it was later adapted to the mouse [47,48], thus allowing to study the contribution of individual or combination of genes to the HI-induced pathological processes using genetically modified mice. This model combines unilateral ligation of the common carotid artery (CCA) and subsequent exposure to 8-10% oxygen/balance nitrogen (hypoxia) for a duration of between 40 min up to 3.5 h. The time interval between permanent CCA ligation and subsequent exposure to hypoxia influences the development of brain injury in the rodent neonate. If initiated three to four hours after CCA ligation, hypoxia induces a discernible neurologic lesion histologically and a drop in ATP levels in the ipsilateral hemisphere. Nevertheless, the brain injury is almost absent if hypoxia is initiated 24 h post-ligation [49,50]. This is most likely because collateral compensatory blood flow occurs within 24 h, as demonstrated in the adult rat in which one to two days and up to six weeks after CCA ligation, a significant enlargement of the anastomosis of the ipsilateral posterior communicating artery was measured [51,52]. Hypoxia alone also does not cause obvious neuronal death, but rodent neonates exposed to global hypoxia between P7-P10 show a heightened susceptibility to provoked seizures, and develop spontaneous seizures shortly after hypoxia exposure, and during adulthood [53,54]. Thus hypoxia-only models are valuable to study short and long-term consequences of neonatal epilepsy (reviewed in [54]). The brain injury induced by HI is observed in the ipsilateral hemisphere, i.e. on the same side of carotid ligation, and generally leaves the contralateral hemisphere intact. Duration of hypoxia obviously influences the severity of lesion [55], but even when using the same exposure time to hypoxia within an experiment, animals still display variable brain damage, ranging from none, mild, moderate to almost fully infarcted ipsilateral hemisphere [56,57]. The lesions are commonly observed in both grey and white matter, and affected structures include the cortex, striatum, thalamus and hippocampus. This neuropathology is accompanied by short-term sensorimotor deficits [58] and long-lasting motor, learning and memory impairments [59][60][61]. Of note, the HI-induced sensorimotor and motor deficits are lateralized, and observed on the contralateral limbs [62].
The HI surgery in rodents is typically performed at P7 in rats, and between P7-10 days in mice. While not explicitly stated in the original publications, the choice of these days was based on the work by Dobbing and Sands [63], who reported that the peak of the brain growth spurt occurred at P7 in the rat brain and around birth in humans (see Box 2 on the brain growth spurt). There is no definitive answer as to whether the degree of maturation of the rat brain at P7 day corresponds to that of the human brain at birth, because it mainly depends on the developmental indices considered. For instance, in terms of functional cortical development (measured by the number of synapses, electrical activity and activity of some neurotransmission system), the brain of a human full-term neonate resembles that of P12-13 rat [64]. When white matter maturation is considered, the human brain between 30 and 36 weeks of gestation then roughly corresponds to that of a P7 rodent brain [65]. Based on these criteria and others, some authors posit that the P7 rat brain may still be premature in comparison to that of a human term infant, and argue that the P10-11 stage may be more appropriate to study HIE and potential therapies [56,66]. Notwithstandingly, the overwhelming majority of investigators-including ourselveshas and continues to use the classical P7 rat model and thus the prevalence of this model facilitates inter-studies comparisons (see all Tables). As models translating the timing of neurodevelopment events from humans to rodents are now available (http://translatingtime.org) [67][68][69], an informed choice can be made concerning the postnatal age at which HI is induced, keeping in mind that a three to four day shift may impact neuropathological and behavioral outcomes.
HI is also performed in rodent neonates younger than P7. As P1-P5, in terms of brain maturation, roughly corresponds to a human fetus between 24 to 32 weeks of gestation, HI induction at this postnatal age attempts to model brain lesions observed in the very preterm infant, in particular white matter damage and its most severe form periventricular leukomalacia (PVL). There is currently no single model of "encephalopathy of prematurity " [70], as the day of HI induction, the severity of hypoxia (typically 5% O 2 because 8% O 2 fails to induce neuronal damage [71]) and its duration, vary greatly between studies [72][73][74][75][76][77]. In addition, HI is sometimes combined with an inflammation trigger (e.g. LPS) [78,79], since inflammation is thought to be a major cause of brain injury in preterm infants [80,81]. The pattern of brain lesion overlaps to a certain extent with that of the P7 model, for instance alteration in cortical development and myelination can be observed, albeit with subtle differences in neuronal populations affected, and short as well as long-term sensorimotor, motor and cognitive defects are also observed [82][83][84][85]. Nevertheless, contrary to the HI-P7 model, the hippocampus is typically spared in neonates exposed to HI before P7 [71].

Box 2 Age Matters/the Brain Growth Spurt
Age is admittedly a critical factor in the field of neurosciences [86], but even more so in neurodevelopment. Why do we model a brain injury occurring at around birth in humans in seven-day-old rodents? The answer is that the relation of birth to the degree of maturation of the brain differs between species. While humans, just as rodents, are considered an altricial species, their brain maturation at birth is actually more advanced than that of rodents.

The Brain Growth Spurt
The growth of the brain, in terms of weight, is not uniform across development. If age is plotted against brain weight (expressed as percent of adult value), it yields a sigmoid curve, whichever species considered. When derivated into a velocity curve, then a peak becomes visible: it corresponds to a phase of extremely fast growth, referred to as the brain growth spurt. A major difference between species is the timing of this growth spurt in relation to birth. In humans, there is a major brain growth spurt that begins at around mid-gestation, peaks at birth, and ends at around 3 to 4 years [87]; between birth and the age of 3-4 years, the weight of the human brain quadruples, reaching almost 80% of the adult brain weight. In the rat, the brain growth spurt is postnatal only: it starts at birth, peaks at postnatal day 7, and ends at around postnatal day 25. The timing of the peaks, birth in humans, and P7 in rats, is the basis for the P7 rat model of neonatal HI. It is still frequent to read in the literature that the developmental stage of the rat brain at P7 roughly corresponds to that of humans at term birth. Such a statement is yet a shortcut, as other developmental aspects beyond the growth spurt can also be considered, as discussed in the main text. Thus, the decision-making concerning the time point at which an injury is induced in rodent neonates to mimic a human condition will depend on the research question and should be based on the available neurodevelopmental data.
The Impact of Neonatal HI on Microglia/Macrophages

Microglial Phenotype after Neonatal HI
In the healthy developing brain, microglia are implicated in various ongoing neural processes, such as synaptic pruning [88,89], maintenance of normal brain structure and olfaction [90], phagocytosis of dead cells and debris, but also of viable neural precursor cells in the subventricular zone (SVZ) [91], axon guidance [92], as well as neurogenesis and oligodendrogenesis [93] . Experimental data in rodents indicate that neonatal HI induces early (two hours post-HI) and persisting (up until 17 days) phenotypical changes in microglia in the ipsilateral hemisphere, including an increase in cell count, a spectrum of morphological changes from a ramified to an amoeboid state, and an elevated expression of markers of microglia activation (OX-6, major histocompatibility complex II; OX-18, major histocompatibility complex I; CD68) [94,95] (see Table 1 for an overview on microglia and other cerebral immune-related findings in the rodent model of neonatal HI). The time course of these changes is region-dependent, with the hippocampal microglia showing earlier activation than their cortical, striatal and white matter counterparts [32,96]. There is a commonly held view that in the earlier phases of HI injury, microglia adopt a M1 "classically activated" pro-inflammatory phenotype, while later on switch to a M2 "alternatively activated" anti-inflammatory phenotype (see Boxes 3 and 4 for a historical perspective and a summary on the M1/M2 concept). Thus, a study reported that M1 microglia predominate in the ipsilateral hemisphere three hours after neonatal HI, while at 24 h, both M1 and M2 microglia can be detected, suggesting a switch in the inflammation state of the brain [31]. Nevertheless, a limitation in this study is that a single marker for each state was used (iNOS for M1, and CD206 for M2). A more recent study revealed a slightly more complex picture, with higher and concomitant expression of both anti-inflammatory and pro-inflammatory genes in the ipsilateral than in the contralateral hemisphere [97]. In addition, FACS analysis of the CD11b + population (microglia/ macrophages) for CD86 (classical) and CD206 (alternative) cell markers revealed an expanded and predominant population of CD11b + CD86 + CD206 − cells in the injured hemisphere over several days post injury, a relative suppression of CD11b + CD86 + /-CD206 + cells, and a small "non-polarized" cell population expressing neither CD86 and CD206, but expressing galectin-3, an immunomodulatory mediator. Our own investigations of the microglial phenotype in the SVZ -three and 13 days after neonatal HI in rat-also converge towards the same point since CD11b + microglia isolated from the ipsilateral subventricular zone (SVZ) of HI-exposed rat neonates upregulate both pro-and anti-inflammatory genes until 13 days post-injury, and do not differentially express markers of acute inflammation such as IL-1β, IL-6, TNF-α, and IFN-γ when compared to microglia isolated from the SVZ from sham animals [98]. This underlies the specificity of the SVZ microglial response versus that observed in regions nearby the ischemic core. Further comparative analyses of our microglial gene datasets with  Hemispheric myeloperoxidase (MPO) activity MPO activity in both hemispheres of HI-exposed animals is higher than in control animals at all times of sacrifice, with a peak between 4 and 8 h post-HI. There is no difference in MPO activity between Il and Cl from HI-injured rats. Neutrophil staining with ANS (anti-neutrophil serum) in the brain The count of neutrophils in HI-Il and HI-Cl hemispheres does not differ significantly from that in normal brain tissue. Overall there are significantly more neutrophils in the Il than in the Cl hemisphere. Staining indicates that the majority of neutrophils are located within the leptomeningeal blood vessels on the surface of the brain or within the choroid plexus, thereby indicating that the neutrophils do not extravasate from blood to the brain parenchyma.
Neutrophils may play a role in the pathogenesis of neonatal HI, via yet unclear mechanisms.  -HI stimulates MCP-1 expression in the brain even in the absence of IL-1β activity. -The activity of ICE participates to the regulation of the acute inflammatory response to HI brain injury in neonatal mice.  -After neonatal HI, the first immune activation is that of CD11b + cells in the brain and correspond to microglia.  positive T-cells, not seen in shams, is present in the HI hemisphere at 3 mo but disappears at 7 mo.
-The activation of microglia may prepare for the activation of adaptive immune response. -It is hypothesized that the infiltrating T-cells 1 w after HI remain in the brain, and then become activated. HI in the liver differs from that observed in the brain: a downregulation of pro-inflammation mediators is observed 3 h post HI; 24 h post-HI, the hepatic cytokine/chemokine expression levels are back to these measured in control animals. -Overall, the authors suggest that the cerebral inflammation induced by HI is responsible for the downregulation of the hepatic pro-inflammatory cytokine response. The data illustrate that the traditional concept of "classical" versus "alternative" activation of microglia/macrophages cannot be applied to the complex phenotype observed in vivo after neonatal HI. -It is the unpolarized population of CD11b + cells that express the most galectin-3, an immunomodulatory factor. These cells may play a role in the recruitment of cells in the brain and in the immune response after neonatal HI. published ones revealed that, three days after neonatal HI, the transcriptome of the SVZ microglia from HI-exposed animals does not fit a M1 or M2 state, but rather present similarities with that of primed microglia isolated from mouse models for aging and neurodegenerative diseases, characterized by an altered immune profile associated with phagocytic clearance and possibly neurotrophic features [99]. Thus the concept of "disease-associated microglia" may apply beyond the neurodegenerative environment [100,101]. Altogether, these data uncover a previously unrecognized microglial diversity/ complexity after neonatal HI that surpasses the traditional M1/M2 frame of thought [102,103]. This reinforces the idea that the M1/M2 concept, as currently applied to microglia should be questioned, if not abandoned [104,105]. With the advent of RNA-sequencing (RNA-Seq) analyses of microglia, an extraordinary complexity and unique microglial signature during physiological development has been revealed. Thus, mouse microglia harvested between embryonic day E14.5 and P9 harbor a transcriptomic signature distinct from earlier embryonic stages or adult microglia, and these so called "pre-microglia" express of a common cluster of genes involved in neural migration, neurogenesis and cytokine secretion [106]. Further refining these observations, single-cell RNA-sequencing (scRNA-Seq) of murine microglia at E14.5 and P5 reveals a greater diversity of microglial subpopulations at these ages than at juvenile, adult or old stages [107]. Collectively these data highlight the peculiarity of the microglial phenotype during development, the regional and temporal specificities, and agrees with the reported heterogenous functions in the developing brain.

Does Age at Time of Injury Influence the Ischemic-Induced Microglial Response? If So, how?
Six studies have directly compared the cerebral immune response after induction of HI at different postnatal ages, out of which three were conducted by the same laboratory (see Table 2). In these reports, the intensity of the microglial and cytokine response after neonatal HI (P9 mice or P12 rats) was compared to that after HI in newborn rats (P1), juvenile mice (P21 or P30) or adult mice (three months of age). While HI-induced microglia proliferation (five weeks post-HI) and brain expression of IL-18 and MCP-1 (three days post-HI) appeared more pronounced in the ipsilateral hippocampus of mice subjected to HI at P21 than in mice subjected to HI at P9, opposite results were documented in studies comparing neonatal HI to older juvenile or adult HI. In particular HI-associated elevation in hippocampal microglial population (CD11b +high /CD45 +low ) and in pro-inflammatory cytokines (TNF-α, IL-1β) two days after lesioning was significantly more robust in P9 than in P30 mice. Morphological analyses of microglia at the same time point post-HI corroborated these findings, with microglia displaying a more activated phenotype in P9 than in P30

Reference
Species and Model of HI Immune-related findings in the ipsilateral (Il) brain in comparison to the contralateral (Cl) brain

Conclusion(s)
-The proportion of proliferating microglia (Iba1 + BrdU + /Iba1 + ) is augmented in the Il SVZ only at P10. -Morphologically, the proportion of amoeboid microglia in the Il SVZ rises significantly at P10, but not thereafter. Nevertheless, the number of ball-and-chain buds detected in Iba1 + microglia is higher at P40 in the Il SVZ than in the sham SVZ. Microarray analyses of Cd11b + microglia micro-dissected from the SVZ -After neonatal HI, both pro-and anti-inflammatory related genes as well as neurotrophic genes (Igf-1) are upregulated at P10 and P20. -KEGG pathway analysis of differentially expressed genes of HI and sham SVZ microglia reveals an enrichment of pathways associated with neurodegenerative diseases in P10-HI SVZ microglia.
-Overall there is a SVZ-specific response of microglia to neonatal HI, associated with a complex immunomodulatory and potential neurotrophic gene expression pattern.
All sacrifice time-points are given in post-HI Immunohistochemistry in the hippocampus only (dentate gyrus=DG and Cornu Ammonis=CA) -In the brain of sham-exposed animals, the total number of BrdU-labeled cells is greater in the immature than in the juvenile mice. Nevertheless, 5 w after HI, a greater increase in the number of BrdU + cells is measured in the brain from juvenile mice than from immature mice. -In the CA: the number of BrdU + /Iba1 + cells decreases significantly during normal brain development, does not increase after HI at P9, but increases 113-fold after HI at P21. -In the DG: the number of BrdU + /Iba1 + cells does not change significantly during normal brain development or after HI at P9, but increases 8-fold after HI at P21. -The number of MCP-1 and IL- 18 immunopositive cells in the CA and DG is twice higher in the Il hemisphere of mice subjected to HI at P21 than in that of mice subjected to HI at P9.
The post-ischemic hippocampal inflammation evaluated in terms of microglia proliferation and expression of IL-18 and MCP-1 is more pronounced in the hippocampus from mice subjected to HI at the juvenile stage than from mice subjected to HI at the neonatal stage. -In the neonatal P9 brain, the microglial response in the hippocampus occurs earlier (peak on d 2 post-HI) than in the striatum and cortex (peak on d 9 post-HI). -The microglial proliferation and pro-inflammatory cytokine responses in P9 brains after HI are more robust in comparison to P30 brains. -Such differential inflammatory response induced by HI at the two postnatal stages may have implications for tissue repair.    post-HI is 3 times higher in the ipsilateral -First report that demonstrates age-dependent differences in the activation of resident microglia and infiltrating blood-derived macrophages in the hippocampus after HI brain injury. -The inflammatory response appears more pronounced in the immature brain. -Overall, the resident microglia, rather than infiltrating blood-derived macrophages, proliferate and are activated earlier in the immature than in the adult hippocampus, but remain increased longer in the adult brain. much more moderate decrease in branching morphology 2 d post-HI, and overall retain a highly ramified morphology. Gene expression in CD11b + microglia isolated from hippocampus -In mice subjected to HI at P9, a significant reduction in TGF-β receptor expression, but an increase in Serpine1 (part of the TGF-β receptor signaling pathway) expression are observed. Also, there is a significant reduction in MerTK (Mer Tyrosine Kinase, phagocytic receptor) and SOCS3 (suppressor of cytokine signaling, component of the MerTK pathway) expression. -In mice subjected to HI at P30, the expression of TGF-β receptor is also significantly reduced; Serpine1 expression is not changed 2 d after HI, but its expression remains dramatically elevated compared to P9. As in P9 mice after HI, there is a reduction in MerTK and SOCS3 expression; but the expression of both is significantly higher in the Il P30 compared to Il P9 hippocampus.
-During physiological development (between P2 and P60), microglia in the hippocampus undergo significant changes in morphology, reaching a mature phenotype by P30. These morphological changes are accompanied by increases in the expression of members of the TGF-β receptor and MerTK signaling pathways. -Based on morphological evaluation, 2 d after HI, there is an intense activation of microglia in the Il hippocampus of P9 mice, but not in the P30 mice. -The age-related differences in the microglial response to neonatal HI may be linked in part to the age-associated changes in the TGF-β receptor and MerTK signaling pathways.
All sacrifice time-points are given in post-HI mice. The acute neuroinflammatory response (48 h) after HI was also much stronger in P12 rats than in rats subjected to HI at P1 [79]. Using Cx3cr1 GFP/+ Ccr2 RFP/+ double transgenic mice to allow differential labeling of resident microglia (GFP + ) and blood derived macrophages (RFP + ), Umekawa et al. [14] showed that the increase in the total number of hippocampal microglia after HI was approximately the same in P9 and adult mice, but (i) it occurred later in adult mice and (ii) the proportion of activated microglia (GFP and galectin-3 double positive) was much higher in the immature than in the adult brain. In addition, one day after HI, the concentration of the pro-inflammatory chemokine CCL2 was around three times higher in the ipsilateral hippocampus of the P9 neonates than in the adult mice. Altogether the majority of the studies indicate that the intensity of the microglial response after HI is stronger in the immature rodent brain equivalent to a term-like human brain than in a less (P1 or preterm-like human brain) or more developed juvenile or adult brain. While the time course of HI-induced microglial activation is similar in neonatal and P30 brain, it seems delayed in the adult brain in comparison to the neonatal brain. These age-related differences may also impact response to treatment, as suggested by report of Cikla et al. [108] and detailed in next section. Overall, it is very likely that the microglial heterogeneity at the transcriptional level described during development [103] contributes to this differential response to HI and treatment, but exactly which signaling pathway(s) is (are) involved remain to be investigated.
Box 3 the Fundamental Breakthroughs in Immunology that Led to the M1/M2 Concept Mackaness, in his fundamental work between 1962-1970 [109][110][111][112] (for an in-depth explanation of his work, see Van Epps [113]) introduced first the term "activated macrophage". Using a mouse model of infection with the intracellular bacteria Listeria monocytogenes, he demonstrated that the mechanism of "acquired cellular resistance"-the fact that after a primary infection, mice become temporally resistant to a subsequent reinfection with the same or unrelated intracellular pathogen-was dependent upon a change into the host macrophages that became "activated". He further showed that macrophage activation depended on a soluble factor released by lymphocytes-now recognized as a specific subset of T helper CD4 + cells-exposed to an antigen [111]. He defined the activated macrophage as "larger and much more complex morphologically; it has a marked propensity to spread on glass, a property which is related to its enhanced capacity for phagocytosis. The content of acid hydrolases, the digestive capacity, the respiratory rate and the mitotic rate of activated macrophages are all conspicuously raised." [112].
The major cytokine responsible for the observed macrophage activation (now referred to as "classical" activation) was thereafter identified as IFN-γ [114]. Further studies on the effect of different cytokines on macrophage activation showed that IL-4 [115] and IL-13 [116] caused a macrophage phenotype distinct from that induced by IFN-γ. This "alternative" activation was characterized, among others, by a downregulation of IFN-γ mediated bactericide signals such as NO production, a reduction in the release of inflammatory cytokines, and an increased expression of an endocytic receptor, namely mannose receptor (or CD206). Amidst the work on macrophage activation, a major step in the understanding of immune regulation was brought about by the T helper type 1 (TH1)-TH2 hypothesis, proposed by Coffman & colleagues in 1986 [117,118]. Further insight into the model and its historical context can be found in [119,120]. In essence, it provided an invaluable framework to describe how an organism responds to different pathogens (e.g. intracellular vs extracellular). The TH1 and TH2 clones were distinguished by the pattern of cytokine release, with TH1 clones secreting mainly IFN-γ, and TH2 clones IL-4, IL-5, and IL-13. In general, TH1 and TH2 responses are associated with cellular and humoral immunity, respectively.
Box 4 the M1/M2 Concept of Macrophage/Microglia Activation Mills in 2000 applied the TH1/ TH2 nomenclature to define macrophage activation, hence proposing the concept of M1/M2 macrophage activation [121]. The M1/M2 model stemmed from the main following in vitro data: (i) macrophages from C57BL/6 and Balb/c mice, described to be prototypical Th1 and Th2 strains respectively, respond from a metabolic point of view, i.e. in terms of NO and ornithine production, differently to the same Th1-like stimuli, namely IFN-γ, LPS, or IFN-γ + LPS. In particular, while stimulated macrophages from C57BL/6 produce predominantly NO (iNOS or kill pathway), macrophages from Balb/c produce predominantly ornithine/urea (arginase or Heal pathway). (ii) The same observation was made for macrophages isolated from SCID or NUDE C57BL/6 or Balb/c mice, in which lymphocytes are reduced or absent. This led Mills to hypothesize that macrophages have an inherent capability to display a classical M1 or alternative M2 phenotype, and do not necessarily need "instruction" from lymphocytes. He further proposed that macrophages are in fact the first line of defense cells that can then direct the nature of the adaptive immune response of the T lymphocytes, i.e. TH1, TH2 or other [122][123][124]. This latter view remains nevertheless contested by immunologists who consider the lymphocytes and the cytokines they release as the orchestrators of macrophage activation [125]. A notable difference between the Th1/Th2 and the M1/M2 hypotheses is that M1/M2 polarized macrophages do not constitute separate clones, like Th1/Th2 cells, but rather represent the extremes of a spectrum of different phenotypes depending on stimulus/environment.
The M1/M2 nomenclature was then enlarged, with the introduction of M2a, M2b, M2c [126] and M2d subgroups [127,128], defined by partially overlapping or distinct transcriptional profiles/signaling cascades induced by specific stimuli. Immunologists now suggest to use a more precise polarization nomenclature that includes the source of macrophage, the stimulus used and corresponding set of activation markers [125]. Thus, while the value of the M1/M2 framework is acknowledged, it is also recognized that it does not fully capture the complexity of macrophage responses especially in in vivo settings, as now revealed by the numerous analyses of macrophage transcriptomes via high-throughput RNA sequencing-bulk, single cell or single nucleus-methodologies in diverse pathological conditions (for reviews, see [129,130]).
Even though it may only seem natural that the M1/M2 concept was also applied to microglia, the brain resident macrophages, neuro/immunologists are reconsidering the use and the validity of such concept [104,105,131]. As for macrophages, microglial gene expression profiles typically cannot be fitted into M1-or M2-like subphenotypes, as observed in rat models of adult ischemic stroke [132] or neonatal HI [98], in rodent models of neurodegenerative disorders [100,107,133,134] and in human microglia isolated from Alzheimer's patients [135,136]. There again, the complexity of microglia cannot be reflected into the M1/M2 concept.
Modulation of the Immune Response and Neuroprotection-Correlative or Causal Crosstalk?
Treatments known to confer some degree of neuroprotection can modulate aspects of HI-induced inflammation. For instance, hypothermia can reduce infarct size, improve behavioral outcome and these effects are concomitant with a decrease in cytokine expression in the injured hemisphere (IL-18, IL-6, TNF-α), and a reduction in the number of amoeboid microglia in cortex and corpus callosum [137][138][139]. But whether the direct modulation of the immune response associates with neuroprotection is a challenging question, because anti-inflammatory agents/treatments can have biological actions unrelated to inflammation, for instance on angiogenesis, cell proliferation and apoptosis. Such broad effects are documented for erythropoietin [140], cyclooxygenase-2 (COX-2) inhibitors [141], doxycycline and minocycline [142].

General considerations.
Twelve immune-related therapies have been tested in the rodent model of neonatal HI (see Table 3). Among them, only EPO has been used in clinical trials for neonates with HIE (reviewed in [143]); seven are in clinical use for other human conditions (recombinant human interleukin-1 receptor antagonist, platelet-activating factor antagonists, minocycline, doxycycline, cromolyn, etanercept and fingolimod); one is going out of clinic due to cardiac side effects (COX-2 inhibitor); and three are for preclinical testing only (anti-rat neutrophil serum (ANS), IKK/NF-κB inhibitor and JNK inhibitor). Regardless of clinical use, testing in animal models remains invaluable because it provides fundamental knowledge on the molecular or cellular inflammatory components contributing to the pathogenesis of neonatal HI, and on the phase of injury post-HI these components may be involved in. Eleven out of these twelve treatments conferred neuroprotection, but with highly variable degrees (see Table 3). Putative factors contributing to such variability for different or sometimes same treatments are listed in Fig. 1A. Studies in which minocycline was evaluated illustrate this point. In a rat model of neonatal HI, a single intraperitoneal dose of minocycline (45 mg/kg body weight BW) injected before or immediately after hypoxia induced a near complete neuroprotection at seven days post-HI. Nevertheless, if administered three hours post-HI, neuroprotection was almost lost. A lower dose (22.5 mg/kg BW) given before or immediately after hypoxia did not provide significant neuroprotection [144]. In a mouse model of neonatal HI, two intraperitoneal injections of minocycline (40 mg/kg BW) at two and 24 h after HI improved neurological injury two and nine days later, but the cerebral atrophy, measured with T2-weighed MRI nine and 60 days post-HI was similar to that of non-treated HI-exposed mice. Assessment of cognitive performance 60 days post-HI revealed no significant difference between treated and untreated HI mice. In that study, authors report that the same minocycline treatment protocol applied to juvenile mice subjected to HI at P30 ameliorated cognitive performance and cerebral atrophy measured 60 days after HI [108]. These two studies demonstrate how several parameters may ultimately influence response to treatment, namely (i) species, (ii) timing of administration, (iii) dosage, (iv) age at surgery, (v) time and methodology for evaluation of neuroprotection. Differing degrees of neuroprotection were also reported for recombinant human interleukin-1 receptor antagonist (rhIL-1ra, anakinra), namely >50% in Martin et al. [145] and only~24% in Hagberg et al. [24]. Here the contributing factors to variability include the route of administration (subcutaneous-s.c. versus intracerebral-i.c.), the treatment protocol (multiple versus single injection), and dosage (multiple s.c. doses of~1.5 g per rat in Hagberg's report, versus a single intracerebral dose of 3.3 or 5 μg per rat in Martin's study).
The role of dosage was documented for NS398, a COX-2 inhibitor for which a higher dose was better than a lower dose [146], while the platelet-activating factor antagonist BN 52021 exerted beneficial impact in a dose-independent manner [147].
A common feature between these studies is that all the immune-related therapies were either administered before hypoxia or within few hours after HI. Four out of five studies in which impact of treatment was tested prophylactically (before the onset of hypoxia, anakinra, PAF antagonist, minocycline, ANS) reported superior neuroprotective outcome in comparison to that with post-HI treatment, which could be of interest for at risk pregnancies. Of note, only one study reported worsening of infarct severity upon treatment (i.e., with Fingolimod, [21]). This potential bias towards publishing only beneficial outcomes may be counterproductive, as any "detrimental" treatment actually reveals the key involvement of the targeted pathway in mitigating the injury, a fact that could be harnessed further in therapies. Overall, based on these studies, the following cells and molecular pathways have been identified to play a role in the pathogenesis of neonatal HI in the acute phase: neutrophils, mast cells, platelet-activating factor, IKK/NF-κB signaling, and cyclooxygenase-2 since inhibiting them confer neuroprotection. No or a limited role in neuroprotection post-HI was reported for TNF-α and JNK pathway, while a beneficial role was documented for peripheral lymphocytes. The role of IL-1  Degree of neuroprotection 5 d post-HI, assessed on cresyl violet stained brain slices: -Incidence of infarction 90% in vehicle-treated rats vs 33% in BN52021-pretreated rats (**) and 30% in BN52021-post-treated rats (*). -Neuroprotection per region • Pretreatment: 82% in cortex, 74% in striatum, 90% in hippocampus. • Post-treatment: 72% in cortex, 46% in striatum, 73% in hippocampus. 5 d post-HI, assessed by interhemispheric weight differences for the post-treatment dose-effect study: 18.9% (NS), 44.8% (*) and 32.5% (*) neuroprotection observed in rats treated with 12.5, 25 or 50 mg/kg/dose compared with controls; differences in neuroprotection among the three doses not statistically different. Cerebral inflammation post-treatment Not examined Conclusion PAF appears to be one of the acute inflammatory factor mediating perinatal HI brain injury; PAF antagonists may represent a pharmacologic approach to attenuate the severity of HI-brain injury. Anti-rat neutrophil serum (ANS) compared to allopurinol Note: rabbit polyclonal anti-rat neutrophil serum (ANS) is used to deplete neutrophils; Degree of neuroprotection 42 h post-HI, assessed by quantifying brain swelling (wet vs dry weight of posterolateral half of each cerebral hemisphere):  Anti-rat neutrophil serum (ANS) Route: s.c. Treatment protocol: Prophylactic, or in the acute phase post-HI -before HI group: immediately after carotid ligation -after HI group: 4-8 h after hypoxia -2 corresponding control groups (HI animals exposed to normal rabbit serum) In clinical use for other conditions: no, experimental only Degree of neuroprotection 42 h post-HI, measurement of brain swelling: -In the before HI group, neutropenia reduces brain swelling by 75% (***). -In the after HI group, no reduction in brain swelling. -14 d post-HI, measurement of brain atrophy: -In the before HI group, neutropenia reduces IL hemisphere atrophy by 61% (*). -In the after HI group, neutropenia is not protective.    Degree of neuroprotection 1, 2 and 4 w post-HI, scoring of Fluoro-Jade B stainings, and measurement of IL/CL hemisphere ratio on H&E stained brain slices: -61-75% (*) improvement in score of brain damage at the time-points examined. -Cavitation is absent in all cromolyn treated rats, whereas it is present in 30% and 50% of vehicle-treated animals at 2 and 4 w post-HI, respectively. -IL/CL ratio restored to normal in cromolyn treated rats (*). Cerebral inflammation post-treatment Astrocytic and microglial activation, scored with GFAP and OX-42 antibodies respectively, are significantly reduced at the time points examined.

Conclusion
Mast cells appear to contribute to the inflammatory response to neonatal HI and could be the initiators of the immune response.  -NS-30 group: 79% at 2 w (*) and 75% (*) at 6 w. 6 w post-HI, behavior analysis: improvement in, or restoration of many HI-induced neurobehavioral deficits in NS-10 and NS-30 treated animals. Cerebral inflammation post-treatment -COX-2 immunopositive signals (western blotting and IHC staining of ipsilateral brain) are significantly reduced in NS398 treated HI animals in comparison to HI-vehicle animals. -At 72 h post-HI, IL-6 staining and concentration in ipsilateral brain is increased in HI-vehicle exposed rats, and significantly reduced by NS-30. Qualitative evaluation of single immunofluorescent staining for Iba1, CD68 and MPO in ipsilateral brain indicates that NS30 attenuates significantly HI-induced increase in these markers. Conclusion COX-2 inhibition is neuroprotective; effects may be mediated by anti-inflammatory actions, through IL-6 reduction, decreased activation of microglia and of infiltration of macrophages and neutrophils. Degree of neuroprotection 2 and 9 d post-HI, scoring of damage on MAP2 stained brain slices: -P9 neonatal brain: 73% (*) and 87% (*) protection at 2 and 9 d post-HI. -P30 juvenile brain: no protection at both days; of note, HI-induced brain damage less severe at P30 than at P9. Brain atrophy at 9 and 60 d post-HI by T2-weighted MRI: -P9 neonatal brain: no significant impact on volume loss at both days. -P30 juvenile brain: no significant impact on volume loss at 9 d, but a slight impact at 60 d. Hippocampal learning by Morris water maze test: -P9 neonatal brain: no amelioration in HI-induced learning deficits. -P30 juvenile brain: significant improvement. Cerebral inflammation post-treatment signaling pathway remains currently unclear and requires further investigation.
Is the microglial phenotype modulated by immune rel a t e d t r e a t m e n t s a n d d o e s i t c o r r e l a t e w i t h neuroprotection?
Among the 14 studies listed in Table 3, five examined the phenotype of microglia, but the depth of examination varied widely between studies. The data described below present results observed in the ipsilateral hemisphere from treated versus untreated HI-exposed animals. The injection of Impact of minocycline on total counts of microglia (CD11b + /CD45 + ) -Significantly suppresses HI-induced increase in microglia counts in the IL hippocampus from P9 and P30 brains 2 and 9 d post-HI; this is also observed in the cortex and striatum from P9 brains, but less in the same regions from P30 brains. Impact of minocycline on percentage of activated microglia (CD45 +med /total microglia) -2 d post-HI, in the IL hippocampus, induces~70% reduction in the proportion of activated microglia in both neonatal and juvenile brains. -9 d post-HI in the IL cortex and striatum -P9 neonatal brain: significant decrease in the percentage of activated microglia -P30 juvenile brain: no impact of minocycline on HI-induced late microglia activation Conclusion Minocycline has different neurological outcome depending on the age at which HI-injury is induced. Early improvements in neurologic injury do not necessarily predict long-term improvements in neurologic function. Degree of neuroprotection 1 w post-HI, scoring of injury in cresyl violet stained brain slices: -Significant worsening of neuropathology in FTY720-treated animals versus saline-exposed mice. Cerebral inflammation post-treatment -FTY720 treatment induces a slight but not significant increase in Iba1 protein levels (assessed by western blot). -Despite FTY720 induced depletion of circulating T cells, the total cerebral leukocyte infiltration assessed qualitatively by IHC is unaffected by FTY720. This is explained by an increased infiltration of innate immune cells, mainly neutrophils and inflammatory macrophages. Conclusion Pharmacological mediated T cell depletion increases HI-induced brain injury; this contrasts with the neuroprotective effect of FTY720 observed in adult models of stroke. Thus, neonatal T cells may promote endogenous neuroprotection.
All sacrifice time-points are given in post-HI H&E: hematoxylin-eosin BBB: Blood brain barrier † Percent neuroprotection was either reported by study authors, either calculated by reviewers using the formula 100*(1-%damage in treated animals/% damage in vehicle treated animals), as in [146]. The significance is that reported by authors, and unless indicated, represents the significance of HI drugtreated animals versus HI vehicle-treated animals. Significance values * P < 0.05; ** P < 0.01; *** P < 0.001; **** P < 0.0001 § Various rhEPO treatment protocols have been tested in the P7 rodent model of neonatal HI (for a review, see [143]). This study is included in the table because it is the only one that reports on CD68 immunostaining in the ipsilateral hemisphere from HI-exposed animals. The study authors attribute CD68 staining to infiltrating leukocytes, but it could also be microglia/macrophages doxycycline before or within 3 h of HI reduced neuronal loss by 70% as measured with NeuN immunoreactivity in the hippocampus at seven days post-HI, and concomitantly the number of ED1 (CD68) positive cells declined in the ipsilateral hippocampus, thalamus and cortex [148]. Administration of cromolyn, a mast cell stabilizer curtailed significantly brain damage score at one, two and four weeks post-HI (by roughly 70% as measured with Fluoro-Jade B) and a significant reduction in the number of OX-42 (CD11b) positive cells was noted in the ipsilateral thalamus at the same time points [13]. Administration of the COX-2 inhibitor NS398 diminished significantly brain atrophy (assessed via ipsi:contralateral brain weight ratio, degree of neuroprotection ranging between 57 and 79% (two and six weeks post-HI), restored some HI-induced neurobehavioral defects and those effects were associated with a qualitative reduction in single Iba1 and CD68 immunoreactivities in the ipsilateral cerebral cortex at three days post-HI [146]. The exacerbation of neonatal HI-induced brain injury reported after treatment with Fingolimod did not affect significantly Iba1 protein levels one-week post-HI [21]. Cikla et al. [108] carried out the most detailed examination of the microglial phenotype after minocycline treatment using a combination of flow cytometric and immunohistochemical analyses. Mice, subjected to HI at either P9 (mouse model of neonatal HI) or P30 (mouse model of juvenile HI), were injected twice with minocycline at two and 24 h after surgery. In the HI-exposed neonatal mice, the total microglial count (CD11b/CD45 double positive cells) and the percent of activated microglia (CD45 med+ /total) increased in the ipsilateral hippocampus two days post-HI, and then in the ipsilateral cortex and striatum nine days post-HI. The same spatio-temporal rise in microglia was observed in HI-exposed juvenile mice, although not as strong as that observed in HI-exposed neonatal mice. In the mouse model of neonatal HI, minocycline reduced microglial counts (total and percent activated) at both days post-surgery in all regions considered; in the mouse model of juvenile HI, minocycline also impacted microglial counts in the hippocampus two days post-HI, but not at nine days post-HI in cortex and striatum. Surprisingly, long-term neuroprotection was only observed in HI-exposed juvenile mice, but not in HI-exposed neonates. Authors hypothesized that this differential response to treatment was linked to the persistent microglial activation observed in the striatum and cortex of treated HI juvenile mice, thereby suggesting a positive correlation between microglial activation and beneficial outcome. Thus, while the four aforementioned studies suggest a correlation between treatment-associated neuroprotection and a decline in the number of microglia in the ipsilateral hemisphere from treated animals in comparison to that from untreated animals, the latter study actually proposes a link between neuroprotection and persisting microglial activation. The role of microglia in the early phase after neonatal HI was recently addressed in murine models of neonatal HI using genetic and pharmacological based tools to deplete microglia [149,150]. In [149], microglia depleted male mice (generated by conditional  Table 4 deletion of Cx3cr1 induced by tamoxifen) subjected to HI at P10 displayed an aggravated neuropathology and astrocyte reaction three days after injury, and overall lower levels of TGF-β and IL-10 in the ipsilateral hemispheres. These data suggest that microglia and specific cytokines play a neuroprotective role in the early phase post-injury. In contrast, in [150], PLX3397 (Plexxikon) was administered orally from P4 to P11, and mouse neonates were subjected to HI at P9. Two days after HI, a significant amelioration in neuropathology was observed, accompanied by a reduction in brain infiltration of circulating immune cells and an improved sensorimotor function, overall suggesting that in the early phase post HI, CSF1R mediated pathways exert a detrimental role in the pathogenesis of the disorder. Reasons underlying such opposing results are unclear but may be linked to differences in HI model (P10 mice, 1 h hypoxia in [149] versus P9 CD1 mice, 30 min hypoxia in [150]) and to the fact that CSF1R inhibition has unspecific effects, as reported in [151]. Another recently reported issue associated with the use of PLX3397 is the actual silencing of typical microglial genes such as Iba1 and Cx3cr1, rather than a depletion of or a reduction in microglial density [152]. Additional microglia targeting agents to be tested include gadolinium chloride (GdCl 3 ), a drug that reportedly can deplete selectively pro-inflammatory microglia [153], as well as others potential depletion tools, such as lipid nanocapsules, or polyamidoamine dendrimers as reported in [154].

Therapeutic Benefit of Stem Cell Transplantation in the Rodent Model of Neonatal HIE: Is Microglia Involved?
Immune modulation is often put forward as a major mechanism of SC mediated therapeutic benefit. We could find 19 studies-from 12 distinct research laboratories-that reported on SC transplantation in the rodent model of neonatal HI and examined changes in neuroinflammation or in peripheral inflammation post-treatment (see Table 4). A quick glimpse at the table indicates that each study used its own experimental protocol, thus rendering comparison of results between studies difficult. The only common point between these studies is that a single injection of stem cells was used. Variability is otherwise detected at all experimental steps, starting from the rodent model of neonatal HI, through the treatment (type/source, dose and route of SC transplantation), and to the evaluation of neuroinflammation and neuroprotection, as depicted in Figs. 1, 2 and 3. Even when the same treatment protocol is used, variation in other parameters eventually can lead to contrasting outcomes, as illustrated in the two following studies by the same group. In the study of McDonald et al. [155] duration of hypoxia was 180 min and readout endpoint was seven days post-HI, while in the study of Penny [156] these parameters were 90 min and 43 days post-HI. Perhaps mainly due to longer exposure to hypoxia in the first study, deficits in the negative geotaxis test were observed seven days post-HI in the untreated HI-injured neonates, but not in the second study at 43 days post-HI. Stem cell treatment (one million human UCB mononuclear cells transplanted i.p. 24 h after HI) almost fully protected animals from brain damage in the first study, and this was accompanied by an improvement in negative geotaxis test and a significant reduction in the number of Iba1+ cells in the frontal motor cortex. In the second report, stem cell treatment did not limit HI-induced brain damage and an increase in the number of Iba1 activated microglia was observed in the somatosensory cortex, but not in the motor cortex. Thus, experimental variability eventually complicates interpretation. Does stem cell therapy change the microglial phenotype after neonatal HI? If so, how? Is it correlated with neuroprotection and functional outcome?
Six studies reported a reduction in Iba1-related immunohistochemical parameters-such as number of activated cells, immunointensity and proliferation-in the cortex, the frontal motor cortex, the penumbra of the cortex, the hippocampus, the thalamus at 1-, 5-, 6-, 10-, 14-, 18-and 21 days post-transplantation. This observation was not correlated with a particular cell type, dose, route or time of transplantation (rat/ mouse/human bone marrow-BM derived mesenchymal stem cells-MSCs, human amniotic fluid stem cells, mouse neural stem cells, human umbilical cord blood-UCB derived mononuclear cells, T regulatory cells-Tregs, endothelial progenitor cells-EPCs and monocytes, injected i.p., i.v., i.c. or i.n., doses of 13′333 or 66′667 cells per gBW and 333′333-666′667-1′ 666′667 cells per g brain weight-BrW, injected either 24 h, three or ten days post-HI). Behavioral outcome was evaluated in five studies, and significant improvements in sensorimotor function were reported (assessed in the cylinder rearing test or negative geotaxis test), with some nuances: for instance, amelioration in sensorimotor function was found transient in one report, being detected at three weeks post-HI but not at ten weeks post-HI [157]. Also, the individual cell subtypes of human UCB had differential ability to affect HI-induced deficits in the negative geotaxis test, with UCB derived monocytes being ineffective, in contrast to UCB derived EPCs [155]. The degrees of neuroprotection (assessed with cresyl violet, H&E, MAP2 stainings, ipsi/contra area ratios) were variable, ranging from none to 42%, 58%, 70% to almost full neuroprotection.
Six studies reported a reduction in the CD68 microglia/ macrophage activation marker (levels in whole ipsilateral hemisphere, number of positive cells or immunointensity) in the cortex, temporal cortex, hippocampus and dentate gyrus, penumbra area at 1-, 7-and 35 days post-HI. This was again independent of a particular cell type, dose, route or time of transplantation (human UCB derived mononuclear cells and MSCs, or rat dedifferentiated fat cells, injected i.p., i.v. or i.c.     with Iba1 include single or double stainings (e.g. Iba1 with CD68 or with iNOS). (C) maximum image projection from a P30 rat exposed to HI and WJ-MSC treatment as an example of Iba1 and CD68 double immunostaining. CC: corpus callosum, LV: lateral ventricle, SVZ: subventricular zone, STR: striatum, dotted line marks SVZ area negative geotaxis reflex) in only one study, but this effect was transient, being observed four days post-HI, but not seven days post-HI [158]. In two studies, either sensorimotor function was not improved [159], or only trended towards improvement [160]. In two studies, HI-untreated animals did not show any deficits in motor function (gait analysis) [161], or learning (active avoidance test and novel object recognition test), thus rendering the assessment of therapeutic benefit of SC unattainable [160,161]. Absence of obvious deficits among HI untreated animals versus their sham counterparts was also anecdotally reported in [156], suggesting that some tests may be insensitive or that HI animals could be less affected. Neuroprotection was either detected and significant [158,159], although in one report restricted to the striatum [158], or absent [160,161]. Conversely, four studies reported a significant elevation in Iba1-related immunohistochemical measurements (immunoreactivity/overall cell count of cells or of activated cells) in the striatum, periventricular striatum, somatosensory cortex at 7-, 29-, 43-and 49 days post-transplantation. In two independent studies, this elevation was reported to be transient [162], and region-specific [156]. Variability in sources, (human/ mouse NSCs or human UCB derived mononuclear cells) routes, i.c. (two studies), i.v. (one study) and i.p. (one study), timing of transplantation, 24 h (three studies) or seven days (one study) post-HI, and doses (66′667 or 666′667 cells/gBW for i.p. and i.v. and 300′000 or 3′200'000 cells/gBrW for i.c. route) excludes a relation between microglial phenotype after SC treatment and any of these factors. Concomitantly, a significant improvement in sensorimotor functions was clearly observed in three studies, the report of Penny [156] being peculiar because HI-untreated animals did not show deficits in individual tests but the calculation of an overall behavioral burden score did reveal an improvement in the HI SC treated group. A common finding between these four studies is that neuroprotection, based on H&E or cresyl violet brain stainings, was not observed. In [156], an exacerbation of neuronal damage was even observed in the sensorimotor cortex after SC treatment. Nevertheless, the number of studies is too low to conclude on a relationship between an increase in Iba1-related measurements and SC associated amelioration in sensorimotor function.
Finally, three studies reported absence of modulation of Iba1-associated measures, such as count [155,163], or CD68 immunointensity [159] after stem cell transplantation performed at either six or 24 h post-HI. In [163], rat adipose derived stem cells injected i.v. 24 h post-HI did not impact Iba1 microglial count or activation status in the cortex, hippocampus, and basal ganglia or serum chemokine/cytokine levels. The treatment also did not affect the number of apoptotic cells in the hippocampus, and was actually associated with an increased mortality, thus demonstrating a detrimental effect of these particular cells. In [155], i.p. transplantation of human UCB derived mononuclear cells or individual cell subtypes (at 24 h post-HI) show that transplantation of monocytes did not impact the number of activated Iba1 cells in the cortex, and neither neuroprotection nor behavioral improvement were detected. I.c. transplantation of human UCB derived MSCs at 6 h post-HI was not associated with changes in CD68 immunointensity in the HI penumbra area, but neuroprotection was observed and HI-induced deficits in sensorimotor function remained. Nevertheless, combination of this treatment with hypothermia led to a reduction in CD68 immunointensity, together with a better neuroprotection and significant behavioral improvement [159].
Altogether, twelve studies reported a reduction in the activation of microglia upon SC administration, accompanied or not by a concomitant reduction in other cerebral or serum inflammation parameters. In some reports, but not all, this observation was accompanied by some degree of neuroprotection and/or improvement in behavioral outcome. In contrast, four studies reported an increase in these parameters typically accompanied by an improvement in behavioral outcome but not associated with significant neuroprotection. Absence of microglia modulation was also associated with absence of therapeutic benefit. Thus, overall the phenotype of microglia is modulated by SC transplantation, but bothlessened or augmented activation-phenotypes can be associated with a therapeutic benefit. Nonetheless, a major caveat in these studies is that the evaluation of the microglial phenotype relied mostly on single immunohistochemical stainings, which remains insufficient to fully capture the complexity of microglial response after treatment. Future studies could address it more precisely (using a combination of double immunostainings, and/or additional techniques i.e. flow cytometric, genomic and proteomic approaches) taking into account different developmental times and specific regions. The absence of a standardized methodology to assess changes in the microglial phenotype together with the multiplicity of experimental designs prevent for now any definitive conclusion.
Transplanted SCs, microglia and neuroprotection/regeneration: mechanisms of the crosstalk Benefits of SC transplantation are currently mainly attributed to paracrine (over a short distance, i.e. within the damaged brain) and endocrine-like (over a long distance, i.e. from outside the injured brain) signaling, since evidence for cell differentiation and replacement remains scarce. SCs can migrate and home to the damaged brain (Table 4), as they express receptors for chemotactic cues such as stromal cell-derived factor-1 (SDF-1), but homing in the injured site may not be required to elicit benefits. SCs released factors, soluble or present in extracellular vesicles (EVs, e.g. exosomes, microvesicles) are thought to have pleiotropic activities in the host, including immunomodulatory, antiapoptotic, angiogenic, and neurotrophic effects. A particular interest in the role of EVs shed by exogenous SCs on endogenous microglia has recently emerged, as microglia appear crucial "EV recipient cells" in comparison to other cell types in the brain [164][165][166]. Until now three studies have investigated the therapeutic potential of bone marrow MSCs derived EVs (administration routes were either i.p., intranasal or intracardiac) in the murine model of neonatal HI, and all reported a reduction in microglial activation, with two studies also reporting functional benefits [167][168][169], similar to findings of studies using actual MSCs (Table 4). Thus, overall, a complex crosstalk between exogenous SCs, endogenous microglia and neuronal cells mediated by EVs may explain in part their therapeutic potential. A question remains though whether the EVs released by SCs after transplantation are of the same nature of those collected from in vitro cultured and expanded cells, as the diseased environment of the host may impact SCs secretion profile.

Is Inflammation in HIE and CP Modulated by Stem Cell Treatment? Knowledge from Clinical Studies
The tools allowing to evaluate the immune-system associated changes in living humans after cell-therapy are mainly the measurement of cytokine levels in plasma/serum or cerebrospinal fluid (CSF). Modern imaging techniques listed hereafter may also be useful, although with limitations. Positron emission tomography-computed tomography with the radiotracer 18 F-fluorodeoxyglucose ( 18 FDG PET-CT), which measures glucose metabolism, is a valuable tool, as inflammatory cells are highly glycolytic and foci of increased uptake of 18 FDG in the brain may be interpreted as sites of inflammation. Brain magnetic resonance imaging with diffusion tensor imaging sequences (MRI-DTI) provide relevant information on brain connectivity, indicative of white matter pathologies. Single-photon emission computed tomography (SPECT) can answer questions on inflammation, and on cerebral blood flow and blood brain barrier injury. These three imaging techniques are nevertheless not specific for immune-related changes, and therefore caution in data interpretation is warranted. In the past decade, clinical trials have investigated mainly safety, feasibility, and efficacy of cell-based therapies in HIE and CP (Table 5). Up to now, among 25 trials, we found three studies reporting on cytokine levels after SC transplantation, two in plasma [170,171] of CP patients, and one in the CSF of intracerebral hemorrhage (IVH) patients [172]. 18 FDG PET-CT imaging was also performed in [170,173] and three additional studies [174][175][176]; brain MRI-DTI data were reported in [173,174,177,178] and SPECT data in [177]. To the best of our knowledge, studies using PET radioligands targeting the 18 kDa translocator protein (TSPO), a surrogate marker of neuroinflammation (e.g. activated microglia and astrocytes), have not been performed in infants diagnosed with CP or an early HI brain injury. Promising studies in rodent models of adult stroke and in human stroke patients have been reported [179][180][181]; nevertheless, questions on cellular specificity persist, and insufficient binding affinity of these radioligands is documented in patients harboring polymorphisms in TSPO [182].

General Considerations on SCs Used in Clinical Trials
In clinical trials, umbilical cord blood (UCB) derived cells (i.e. total nucleated cells-TNCs, mononuclear cells-MNCs and MSCs) are the most frequently used cells (Fig. 4). SCs from neonatal tissue may display improved features in comparison to those isolated from adults, for instance MSCs derived from neonatal tissue display improved capacity for proliferation, expansion and engraftment in comparison to adult MSCs [183]. Cells from birth related tissue (UC, placenta) also raise no ethical issues; practically, they can be easily harvested and banked and are therefore rapidly available for both autologous or allogenic use in human patients. Alternative strategies for endogenous brain repair also include the mobilization of endogenous BM derived SCs to the bloodstream; this is typically achieved with granulocyte colony stimulating factor (G-CSF) but additional methods are being tested, for instance co-injection of NOx-12, a compound that acts in a similar fashion to G-CSF, with SDF-1 to promote recruitment of the cells and favor paracrine effects at site of injury, as reported in a model of retinal degeneration [184].

Immune related findings in clinical trials
Allogenic infusion of UCB-TNCs in CP patients resulted in elevated levels of plasma pentraxin 3 (PTX3), interleukin-8 (IL-8) and in an increased number of blood cells expressing Toll-like receptor 4 (TLR4) during a consecutive short period after infusion. The elevation in PTX3, IL-8 and TLR4 levels correlated with a better Gross Motor Function Measure (GMPM) outcome up to six months after treatment [170]. The role of the cytokine PTX3 is not fully understood but several studies suggest an anti-inflammatory, protective effect in various conditions [185][186][187]. IL-8 is a chemokine that promotes chemotaxis towards the injured area and was shown to have an angiogenic effect [188]. A study further investigating the role of PTX and TLR4 after allogenic UCB infusion in CP is currently pending (NCT03130816). In that same study, two weeks after treatment, a decreased activity in 18 FDG PET-CT in the white matter of the occipital and temporal areas was detected. The authors interpreted this finding as an anti-inflammatory effect of UCB-TNC, as these areas are -Levels of CSF-IL-6 significantly decline after MSC transplantation.
-Baseline CSF-IL-6 is higher in infants who underwent shunt surgery than in infants without shunt placement.
-Correlation between baseline levels of CSF-IL-6 and TNF-α and baseline ventricular index.       typically inflamed in PVL, a significant cause of CP. Increased activity was also observed in cortical areas, which the authors interpreted as a possible correlate of improved motor function. In a previous study, the same group had investigated the effect of allogenic UCB infusion in combination with EPO [174]. They found significant fractional anisotropy (FA) value increments in all measured loci and especially a correlation of the changes in the posterior limb of the internal capsule (PLIC) with the GMPM in the first six months [174]. An additional effect of repeated administration of allogenic UCB-TNC was found in a large open label study including 80 patients [189]. Also, Sun et al. found a dosage dependent effect after autologous UCB reinfusion with an improved brain connectivity analyzed via MRI-DTI, supporting the notion that paracrine signaling underlies the UCB effect leading to GMFM scores above those predicted by age and severity in CP patients [178]. Main limitation of the aforementioned studies with allogeneic and autologous UCB-TNC transplantations is the administration of the immunosuppressants cyclosporin and/or methylprednisolone which could explain findings as no cyclosporin/methylprednisolone-only controls are included. Regarding HIE, autologous UCB infusion without prior cryopreservation was safe in a Phase 1 study [190], and feasibility was also demonstrated in cases where cord blood collections were insufficient for banking criteria [191]. The impact of administration of autologous G-CSF mobilized peripheral blood MNCs (mPB-MNCs) on plasma cytokines was also reported in a randomized double-blind crossover study including 16 CP patients [171]. Their neuroregenerative effect was rather marginal, and motor-functional improvements did not differ after placebo or mPB-MNC reinfusion. Interestingly after G-CSF treatment the improvements were significantly higher than during the period of reinfusion of cells. Overall, no significant difference in plasma levels of the following factors, i.e. G-CSF, brain-derived neurotrophic factor (BDNF), vascular endothelial growth factor (VEGF), insulin-like growth factor IGF-1, IL-6, IL-8, IL-10 was observed between the intervention arms. However, in clinical responders, plasma levels of IL-6 and G-CSF were higher one month after G-CSF infusion, and levels of IGF-1 and BDNF were lower, suggesting that these cytokines may be used as prognostic factors in G-CSF trials and that they may be associated with the G-CSF driven neurological improvements. MRI-DTI showed significantly higher FA-values in patients who received cell therapy one month after G-CSF treatment in comparison to those who received it seven months later, suggesting a possible synergetic effect if mPB-MNC are infused shortly after G-CSF, although 18 FDG-PET did not show the expected differences [171,173,192]. In another study enrolling 18 CP patients, mPB-MNC were considered safe for combined intrathecal/ intravenous administration with possible benefit on neurological function at six months after infusion. In that study the collection and reinfusion of cells were performed on the same day, four days after G-CSF mobilization [193]. Since G-CSF is an immune system stimulating factor, its contribution to positive effects cannot be ruled out; a clinical trial including a study arm with G-CSF alone is registered (NCT02866331, last updated 2016).
In a clinical study assessing the effect of intraventricular administration of allogenic UBC derived MSC in nine preterm infants with severe intracerebral hemorrhage (IVH), the CSF levels of IL-1β, IL-6, TNF-α, VEGF, TGF-β1/2, BDNF and fibroblast growth factor (FGF) were investigated before and after MSC transplantation [172]. Results showed that the concentration of IL-6 significantly declined after transplantation,  Table 5 while that of other cytokines or growth factors also tended to decline, but not significantly. Of note, the baseline levels of IL-6 were significantly higher in infants with shunt surgery than in infants without, which may confound the results observed after MSC transplantation. A correlation between baseline CSF levels of IL-6 and TNF-α and ventricular index was observed. Authors suggest that these cytokines may be used as markers of early neuronal injury, but also acknowledge the small sample size and the need for further studies. In a trial using autologous BM derived MSC in 40 CP patients, 18 FDG PET-CT changes were observed correlating to clinical improvement after intrathecal administration [175]. A safety study comparing an intraparenchymal with intrathecal administration of autologous BM-MSCs did not find a superior effect of intraparenchymal versus intrathecal administration, supporting again the hypothesis of paracrine signaling for SC-mediated positive outcome. Of note, the invasive intraparenchymal access was also considered safe [194]. In contrast, one study also testing autologous BM derived MSC in 94 CP patients showed no functional short-term improvements [195]. Allogenic MSC derived from UCB or UC Wharton's Jelly were also tested in two Chinese studies and showed motor-functional improvements during twelve [176] and 24 months [196] of follow up, respectively. Improvements in cerebral structure could not be observed in MRI-DTI, but 18 FDG-PET revealed increased glucose uptake in some patients of the treatment group which could support functional improvements, since lowered glucose metabolism can be found in CP patients [197].
Overall, the efficacy of stem cells via the modulation of the immune system at this stage remains hypothetical given the limited number of studies. Immunosuppressive or stimulating growth factors like cyclosporine and G-CSF may play a role in the inflammatory status of the brain or other organs which eventually could confound the results.

Conclusion
Stem cell-based therapies hold promise as alternative or complementary neuroprotective/regenerative therapeutic strategies for infants diagnosed with neonatal HIE or CP. In a context where these therapies are being advertised by uncontrolled institutes, and understandably, families are eager to offer treatment to their children to ameliorate the persisting neurological disabilities, it is crucial to improve our basic understanding of their mechanism of action to eventually optimize and advance the clinical development of SC protocols.
Our review of the preclinical literature indicates that stem cell treatments can modulate the microglial phenotype after neonatal HI, but a causal link between this immune-related modulation and neuroprotection is difficult to establish, due partly to a limited number of studies, caveats in methodology and highly heterogenous experimental designs. The concept that microglia may be a therapeutic target is appealing [152], but in its current state, requires further research. Clinical data on immunomodulation after SC transplantation in HIE and CP patients remain anecdotical and hold considerable uncertainties that also call for further investigations. Collaborative efforts between researchers and medical practitioners from different fields (neurodevelopment, neuroimmunology, neurobehavior, pediatry, neurosurgery) as well as general guidelines as to how to characterize microglia -similar to those established for macrophages-and various immune related parameters, would advance our understanding of the link between the cerebral immune system, microglia and neuroprotection.
Funding Open Access funding provided by Universität Basel (Universitätsbibliothek Basel). This research was supported by the Neurosurgery department and the Department of Biomedicine of the University Hospital Basel.

Availability of Data and Material Not applicable.
Code Availability Not applicable.

Declarations
Conflicts of Interest/Competing Interests Not applicable.
Ethics Approval Not applicable.
Consent to Participate Not applicable.

Consent for Publication Not applicable.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.