Therapeutic hypothermia is now standard care for near-term to term infants with moderate to severe HIE. Clinical guidelines recommend therapeutic hypothermia should be started as early as possible within 6 h of birth and continued for a period of 72 h, with a target brain temperature of 33.5 ± 0.5 °C ^[1][77]. Therapeutic hypothermia is associated with immunosuppressive changes ^[2][155], including inhibition of microglial activation, chemotaxis, production of pro-inflammatory cytokines, and nuclear translocation nuclear factor kappa-B ^[3][4][5][156,157,158]. Confirming extensive preclinical evidence ^[6][159], a meta-analysis of 11 randomized controlled trials of systemic and head cooling in infants with HIE (≥ 35-week gestation) showed that hypothermia reduces brain damage on imaging after rewarming ^[7][160], and improves survival without disability at 18 to 24 months of age ^[8][129]. Long-term follow-up data support improved outcomes, including reduced risk of death, or CP, or an IQ score < 55 in 6–7-year-old children ^[9][10][130,161]. Moreover, hypothermia has had an excellent safety record, with benign physiological effects such as sinus bradycardia, mild thrombocytopenia, and scalp edema during cerebral cooling ^[8][129]. Thus, therapeutic hypothermia is a remarkable example of a successful bench to cotside translation.

There are limited clinical evidence for therapeutic hypothermia in preterm (≤ 35 weeks’ gestation) infants. Thus, therapeutic hypothermia is not the standard of care for neonates born very preterm. Preclinical evidence, however, strongly supports a similar pharmacodynamic profile in preterm animals as at term equivalent. In preterm-equivalent fetal sheep, cerebral cooling for 72 h (with extradural temperature titrated to 29.5 ± 2.6 °C) started 90 min after severe asphyxia was associated with basal ganglia and hippocampal neuroprotection, protection of immature OLs in periventricular and parasagittal white matter, and reduced overall microgliosis and apoptosis ^[11][12][74,133]. In association with this histological improvement, higher EEG frequencies recovered faster, cephalic blood flow was restored, and the stereographic seizure amplitude reduced ^[11][74]. Importantly, anatomical brain structures that are frequently injured in term and preterm infants with acute HIE were protected.

Historically, systemic-mild hypothermia was associated with reduced survival in extremely low birth weight (≤1000 g) neonates ^[13][162]. The well-known systemic effects of hypothermia have raised concerns that induced cooling might promote postnatal hypotension, intracranial bleeding, or respiratory compromise in the extremely preterm infants ^[14][163]. In a cohort of neonates cooled outside the standard hypothermia criteria (n = 36, e.g., infants at 34–35 weeks’ gestation, or with postnatal collapse, or cardiac disease), compared with infants cooled to the protocol (n = 129), complication rates and neurologic outcomes at 18–20 months were similar, except that five newborns who developed intracranial hemorrhage had very poor outcomes ^[15][164]. In a retrospective cohort of 31 preterm infants born at 34–35 weeks’ gestation, hypothermia appeared to be associated with increased mortality compared to 32 term infants (12.9% vs. 0%, p = 0.04) and a greater prevalence of white matter damage on modern imaging (66.7% vs. 25.0%, p = 0.001) ^[16][132]. By contrast, a retrospective cohort study of preterm neonates (gestational age range 33–35 weeks) with HIE found death or moderate-severe NDI in 11/22 (50%) cases, similar to rates of adverse outcomes in treated term infants in the large controlled hypothermia trials ^[8][129]. Thus, hypothermia is clearly feasible in preterm infants, and appropriately powered controlled clinical trials are essential. One multi-center randomized controlled trial is currently in progress testing safety, feasibility, and efficacy of therapeutic hypothermia started within 6 h of birth in preterm infants at 33–35 weeks’ gestation with moderate to severe HIE (ClinicalTrials: NCT01793129).

One factor that might affect response to therapeutic hypothermia is that preterm newborns have high rates of prenatal and postnatal infection/inflammation ^[17][165]. There is some evidence that suggests a possible deleterious interaction between infection and induced hypothermia. For example, a randomized clinical trial in adult patients diagnosed with bacterial meningitis reported that cooling (32–34 °C, for 48 h) was associated with excess mortality (51% vs. 31%), compared to normothermic patients ^[18][166]. In neonatal p7 rats subjected to HI, hemispheric and hippocampal protection with systemic cooling was lost after pre-insult sensitization with gram-negative LPS ^[19][134] but not gram-positive PAM3CSK4 ^[20][135], suggesting a pathogen-dependent effect. The potential effect of infection in encephalopathic infants treated with hypothermia is unclear, as known infection was an exclusion criterion in the hypothermia trials ^[8][129].

Cell-based interventions as a therapeutic strategy for injury to the (neonatal) brain have attracted much attention in the past decade ^{[21][22][23][24][25][26][27][28][29]}[136,167,168,169,170,171,172,173,174]. Many different types of stem cells, derived from fetal, placental, and adult tissues, are currently under investigation ^{[30][31][32][33][34][35]}[175,176,177,178,179,180]. Stromal cells, including mesenchymal stem cells and multipotent adult progenitor cells (MAPC) are a subset of progenitors that have been shown to differentiate into multiple lineages (i.e., osteoblasts, adipocytes, and chondrocytes) ^{[30][31][32][33]}[175,176,177,178]. They have been a particular focus of research as they are easily obtainable (e.g., from cord blood, Wharton’s jelly and bone marrow) and do not have the ethical and safety concerns of embryonic stem cells ^[27][29][172,174]. Moreover, bone marrow-derived adherent stromal cells have low immunogenicity due to a lack of expression of MHC class II antigens, allowing their use for allogenic therapy ^[29][174]. The therapeutic potential of mesenchymal stem cells (MSCs) has mainly been attributed to their immune-modulatory and regenerative potential ^{[27][28][32][36]}[172,173,177,181]. MSCs modulate innate and adaptive immune responses from a pro-inflammatory status towards an anti-inflammatory status, thereby reducing tissue injury and creating an environment supporting tissue repair and regeneration ^[32][33][37][177,178,182].

Besides immune modulation in favor of regeneration, MSCs directly affect the injured CNS through secretion of neurotrophic factors that stimulate and maintain neurogenesis of the endogenous neural stem cell population and its subsequent differentiation into neuronal and oligodendroglial lineages ^{[28][32][36][38][39][40]}[173,177,181,183,184,185].

The neuroprotective potential of systemically administered MSCs and Multipotent Adult Progenitor Cells (MAPC) was shown in a preclinical preterm ovine model of global hypoxia-HI injury ^[41][42][65,67]. MSC administration during the latent phase following global HI resulted in functional improvement and prevented hypomyelination of the subcortical white matter at seven days post HI. These protective effects were attributed to anti-inflammatory effects since neuroinflammation, and peripheral immune activation were significantly reduced ^[41][42][65,67]. Specifically, intravenous MSC treatment reduced the proliferative capacity of splenic T lymphocytes with concomitant reduced cerebral T cell infiltration. In addition, systemic administration of MSCs ameliorated splenic involution caused by global HI, implicating a key role of the spleen in the protective mechanisms of stem cell therapy. The results from this study were supported and extended by studies from Walker et al. ^[43][44][186,187] and Depaul et al. ^[45][188], demonstrating a crucial role for the spleen in (1) the pathophysiology of CNS injury and (2) providing mechanistic insight for the beneficial effects of stem cell therapy. Moreover, there is increasing evidence for the effectiveness of stem cell administration via the intranasal route, which is a feasible route for the pharmacological treatment of neonates. This concept of intranasal administration is based on the passage of stem cells over the cribiform plate along the olfactory nerve, allowing for rapid dispersion throughout the brain, where local effects can be exerted ^{[46][47][48][49][50][51]}[73,189,190,191,192,193]. In a study by van Velthoven et al. delayed (10 days post HI injury, induced at P9), nasal administration of MSC therapy improved neurological outcomes 28 days post HI. Interestingly, improved outcomes were attributed to the regenerative potential of MSC rather than inhibition of injurious processes or prevention of injury since cells were administered when injury was readily established ^[50][192]. The delayed intravenous MSC administration in fetal sheep five days after global HI was less effective than acute administration 12 h after reperfusion ^[52][72]. These results were attributed to systemic and neuro-immunomodulatory effects ^[52][72]. Collectively, the administration route in relation to the timing of administration appears to be crucial for the action of MSCs (immune modulation and regeneration) to establish optimal neuroprotection. As such, repeated dosing by different administration routes throughout different phases (latent, secondary, and even tertiary) of disease progression might ultimately lead to the greatest benefit.

Therapeutic hypothermia has improved intact survival and neurodevelopmental outcome in infants with moderate-severe neonatal encephalopathy ^[8][9][53][129,130,131]. However, neuroprotection is partial, and a significant proportion of asphyxiated infants still die or suffer life-long consequences, including CP, cognitive deficits, and epilepsy ^[8][9][53][129,130,131]. Encouragingly, preclinical and clinical studies are now investigating therapeutic hypothermia as a potential treatment for extending over several days after the insult infants with mild encephalopathy ^[54][194]; however, hypothermia is counter-indicated for preterm infants as discussed above, and there is currently no neurotherapeutic treatment for these patients. Critically, additional therapeutic strategies that can augment the neuroprotective effect of therapeutic hypothermia in infants with moderate-severe encephalopathy, and reduce the neurological burden in (extreme) preterm infants are under active investigation ^[22][167]. 

More recent clinical trials have also demonstrated that stem cells are safe, with promising protective results in multiple neonatal diseases, including HIE ^[55][197], bronchopulmonary dysplasia (BPD) ^[56][57][58][198,199,200], IVH ^[59][60][137,138], and established CP ^[61][201].

Within this context, recent findings indicated that MSCs may interact either synergistically ^[62][202] or antagonistically ^[63][203] with therapeutic hypothermia in experimental models of neonatal HIE, depending on the timing of administration. These conflicting findings indicate that the translational gap between preclinical research and the clinical application still needs to be addressed before stem cells could be safely translated into clinical practice for neonatal diseases. Several clinical trials testing the protective potential of stem cells are still recruiting or pending, including for NCT04255147 for BPD, NCT03356821 for perinatal stroke, and NCT02612155 for HIE.

Although a large body of experimental animal studies have demonstrated the beneficial effects of cell-based therapies for preterm brain damage, (pre)clinical studies confirming these data are limited. In part, this mismatch can be attributed to different methodological approaches between animal models and clinical practice in terms of the use of single-hit animal models whilst clinical etiology is multifactorial, use of inadequately characterized and heterogeneous stem cell populations, route of administration, and dosing strategies. At the same time, clinical trials often use top down approaches in which stem cells are administered at trivial time-points and the clinical outcomes are measured. Therefore, future studies testing stem cells should closely align with the underlying pathophysiology and stage of injury to address the multi-factorial nature of preterm brain injury.

In addition, the stem cell secretome, which is modulated by micro-environmental cues during the different phases leading to preterm brain injury, defines efficacy of administered stem cells. Thus, a detailed analysis of the biodistribution of stem cells over time combined with detailed secretome analysis might (1) unravel novel potential pathways involved in the pathophysiology preterm brain injury, (2) and enable adjunctive cell-free therapies comprising specific trophic and immunomodulatory factors, and other regulators (such as miRNA), and most importantly (3) provide insight into timing, which appears to be a crucial determinant for optimal therapeutic efficacy.

Despite that immunomodulatory and regenerative effects have been shown, the underlying mechanisms of action of stem cell therapies remain largely unknown. It was initially thought that the therapeutic action of stem cells relied on direct replacement of dead and injured cells. However, since the number of cells that reach the site of injury is minimal, with marginal engraftment and short cell survival, this theory has been largely discredited ^[45][64][139,188]. 

Remarkably, these therapeutic effects could not be explained by the known anti-inflammatory effects of the MSC-EVs, as observed after MSC treatment ^[65][66]. This prompted us to focus on alternative explanations for the pharmacologic effects of MSC-EVs, in particular, restoration of the injured BBB after global HI. There is accumulating evidence that the BBB becomes functional during the second trimester ^[66][67][85,207]. Nevertheless, a global HI insult would result in the release of reactive oxygen species and excitotoxic molecules into the extracellular environment (i.e., the direct effects of HI), with increased cytokine release by the peripheral and local innate immune system (secondary inflammatory component), which leads to BBB dysfunction ^[68][208]. In turn, increased BBB permeability allows intracranial infiltration of peripheral immune cells (e.g., macrophages, leukocytes, and T-cells) that aggravate white matter injury via the release of pro-inflammatory mediators. Thus, strengthening or restoring BBB integrity by enforcing endothelial cells which would attenuate the degree of white matter injury.

ANXA1 (37 kDa), formerly known as macrocortin, renocortin, lipomodulin, or lipocortin-1, was first described in the 1980s, and initially known for its anti-inflammatory effects as a downstream mediator of glucocorticoids ^[69][209]. It is a calcium-dependent phospholipid-binding protein that has received more research attention in recent years due to its multimodal function that extends beyond suppressing inflammation ^[70][210].

Increased BBB permeability in ANXA1 knock-out (KO) mice and progressive loss of endogenous ANXA1 in the cerebrovasculature and plasma of patients with multiple sclerosis led to better understanding of the function of ANXA1 ^[71][211]. ANXA1 strengthens BBB integrity by 1) binding of extracellular ANXA1 to the FPR2 (formyl peptide receptor 2), inhibiting Rho A kinase, and thereby stabilizing tight junctions; 2) direct interaction of intracellular ANXA1 with actin molecules, and thereby promoting cytoskeleton stability and tight junction formation between endothelial cells ^[71][72][211,212]. The intravenous administration of MSC-derived extracellular vesicles containing ANXA1 prevented the previously described HI-induced depletion of endogenous ANXA1 in the cerebrovasculature, thereby preventing BBB leakage ^[73][81]. Given the time-dependent drop of ANXA1 in the course of antenatal infection, the ANXA1 administration in this infectious context might be a promising therapeutic strategy. 

Clinical studies and preclinical animal models have reported acute increases in systemic levels of cytokines such as tumor necrosis factor (TNF)-α, IL-1β, and IL-6 after HI ^{[74][75][76][77][78]}[142,143,144,213,214] and intra-amniotic exposure to inflammation ^[79][80][54,56]. High CSF/serum ratios indicate that local production of cytokines within the CNS also contributes to increased cytokine levels ^[77][213]. Thus, decreasing the levels of pro-inflammatory cytokines represents a promising strategy to suppress neuroinflammation and ensure neuroprotection. In preclinical fetal ovine models, neutralizing antibodies against IL-1β and IL-6 or pharmacological antagonism of TNF-α have already been tested and shown promising short-term effects ^{[74][75][76][81]}[142,143,144,145]. One day after HI, the systemic infusion of anti–IL-6 mAb attenuated BBB dysfunction and decreased cerebral IL-6 levels. Similarly, IL-1β neutralizing antibodies improved BBB integrity, lowered IL-1β levels in the brain, and reduced IL-1β transport across the BBB ^[76][82][144,146]. Further, histological studies have shown that infusion of anti-IL1β decreased short-term ischemic reperfusion-related parenchymal brain injury ^[81][145].

The abovementioned studies have shown promising results, but mainly focused on the short-term effects after HI. Determining the long-term effects is crucial to reduce the detrimental processes that occur during the tertiary phase of the brain injury after HI. As IL-1β and IL-6 expression changes throughout gestation, and they fulfill physiologic functions that are not detrimental by definition, caution has to be paid to the right dosing and timing ^[78][214].

Recombinant human erythropoietin (rEpo) is a common treatment for anemia in preterm infants and pediatric patients with chronic renal disease. In addition to stimulating erythropoiesis, rEpo has shown anti-apoptotic, anti-inflammation, and anti-excitotoxic effects after HI and infection-induced perinatal brain injury ^[83][84][215,216], suggesting rEpo could be a promising neuroprotectant for encephalopathic infants. In the long-term, it could promote OL and neuronal maturation and replacement ^[85][86][147,148], and so might promote neurorestoration after injury. Compelling preclinical evidence has shown that early administration of rEpo is neuroprotective over a broad dose range, from 1000 to 30,000 IU/kg, and that continued exposure to high-dose rEpo is more effective, but that optimal treatment regimens are likely to be paradigm-specific.

For example, after maternal LPS injection in rats at 18–19 days gestation, peripheral rEpo (5000 UI/kg) injection at birth was associated with reduced IL6, IL1, and TNF-α concentrations, and apoptosis and demyelination at p7 ^[87][217]. Critically, after HI in P7 rats, repeated rEpo injections (5000 IU/kg, at days 1, 2, and 3) provided greater protection than either a single (5000 IU/kg) dose or three injections with 2500 or 30,000 IU/kg rEpo ^[88][149], but brain protection was largely lost when the treatment was delayed until 1–3 h after HI ^[89][150]. By contrast, P5 mice treated with ibotenic acid (an excitotoxin) had fewer white matter lesions after a single rEpo injection (5000 IU/kg, at 1 h), but additional injections did not augment protection ^[90][151]. For post-HI treatment, these data are consistent with the experience with therapeutic hypothermia where optimal protection was achieved when brain cooling was started as soon as possible during the latent phase and continued until the delayed secondary events, such as overt seizures had resolved, after ~72 h, with loss of protection if the treatment is delayed more than ~6 h after global ischemia. Supporting this concept, in preterm-equivalent fetal sheep, prolonged rEpo (5000 IU/kg, from 30 min to 72 h) infusion after asphyxia was associated with partial subcortical white matter and neuroprotection, and improved electrophysiological EEG recovery, in association with reduced apoptosis and inflammation ^[91][62]. Similarly, intravenous rEpo (5000 IU/kg) boluses administered once daily for three days to preterm fetal sheep with endotoxin-induced brain damage, reduced axonal damage, microglial, and astrocytic responses in white matter, and improved myelination ^[92][152].

There is strong clinical evidence that rEpo treatment is safe in neonatal cohorts and shows indicative evidence of possible benefit. A meta-analyses in preterm infants suggested that rEpo/Darbepoetin started within eight days from birth reduced rates of IVH, PVL, and necrotizing enterocolitis, and improved neurologic deficits, without adverse effects, at 18–22 months’ corrected age (four randomized controlled trials, 1130 cases, relative risk (RR) 0.62, 95% Confidence interval (CI) 0.48 to 0.80; risk difference −0.08, 95% CI −0.12 to −0.04) ^[93][218]. However, disappointingly, a more recent large multi-center, randomized, trial that assessed high-dose rEpo treatment for perinatal brain damage in extremely preterm infants (741 patients, 24 to 27 weeks and six days) found that repeated rEpo (1000 IU/kg, i.v.) doses at 48 h intervals for a total of six doses, followed with maintenance doses (400 IU/kg, s.c.) three times per week through 32 completed weeks of postmenstrual age, did not reduce death or severe neurodevelopmental impairment, compared to placebo (26% vs. 26%, RR 1.03, 95% CI 0.81–1.32, p = 0.80) at 22–26 months of age ^[94][154]. There were also no significant differences between the treatment groups in the rates of intracranial hemorrhage, sepsis, NEC, death, or frequency of serious complications.

These negative results are disappointing; however, there were several limitations that should be kept in mind. First, the optimal rEpo regimen for preterm brain damage is still unknown. Treatment was started from 24 h after birth, which might be too late for benefit. Second, the treatment duration covered the period when PVL is prominent (i.e., 24–32 weeks’ gestation), but it is possible that rEpo treatment should have been continued for longer. There is extensive evidence that chronic microgliosis and extracellular matrix disturbance continue to contribute to disrupted myelination later in gestation, as discussed above ^[95][219]. Third, cognitive testing for mild impairments is more sensitive later in life ^[96][220], and so potentially long-term follow-up might still show small but beneficial effects. Thus, further studies are required to conclusively determine whether timely and well-targeted treatment with rEpo could alleviate perinatal brain damage in preterm infants.