Whether Erythropoietin can be a Neuroprotective Agent against Premature Brain Injury: Cellular Mechanisms and Clinical Efficacy

Abstract

Preterm infants are at high risk of brain injury. With more understanding of the preterm brain injury's pathogenesis, neuroscientists are looking for more effective methods to prevent and treat it, among which erythropoietin (Epo) is considered as a prime candidate. This review tries to clarify the possible mechanisms of Epo in preterm neuroprotection and summarize updated evidence considering Epo as a pharmacological neuroprotective strategy in animal models and clinical trials. To date, various animal models have validated that Epo is an anti-apoptotic, anti-inflammatory, anti-oxidant, anti-excitotoxic, neurogenetic, erythropoietic, angiogenetic, and neurotrophic agent, thus preventing preterm brain injury. However, although the scientific rationale and preclinical data for Epo's neuroprotective effect are promising, when translated to bedside, the results vary in different studies, especially in its long-term efficacy. Based on existing evidence, it is still too early to recommend Epo as the standard treatment for preterm brain injury.

1. INTRODUCTION

Preterm birth accounts for 11% of all live births worldwide [1]. Although the survival rate of preterm infants has significantly increased over the last three decades, unfortunately, many survivals, especially those born extremely preterm (<28 weeks' gestation), are accompanied by a broad spectrum of neurodevelopment impairment, including cerebral palsy and cognitive, visual, social-behavioral, attention and learning disabilities [2-4].

To date, neuroprotection in preterm infants remains limited to supportive measures rather than targeted therapies. With more understanding of preterm brain injury's pathogenesis, neuroscientists and physicians have embarked on what they hope will be revolutions in treatment to prevent it, among which erythropoietin (Epo) is considered as a prime candidate. Epo, named after its role in hematopoiesis, recently caught attention because of its neuroprotective potential. Preclinical models have demonstrated that Epo signaling is essential for cerebral development [5, 6] and can prevent or ameliorate premature brain injury [7, 8]. Some clinical trials also support its efficacy in preterm neuroprotection [9, 10], while a large randomized trial completed recently (ClinicalTrials.gov, NCT: 01378273) reported that the postnatal Epo treatment failed to improve neurodevelopmental prognosis [11].

Since it remains controversial to use Epo as a pharmacological intervention for preterm brain injury, this review tries to clarify the possible mechanisms of Epo in preterm neuroprotection and summarize updated evidence discussing the safety and neuroprotective efficacy of Epo in preterm infants.

2. MECHANISMS OF EPO IN PRETERM NEUROPROTECTION

2.1. Pathogenesis of Premature Brain Injury

Premature birth occurs during a critical period when the brain experiences several major sequential developmental events. The unique cerebrovascular anatomy and physiology of the premature brain underlie its vulnerability to insults such as hypoxia, ischemia, and inflammation. These insults can coexist and amplify their effects, damaging the developing brain through two principal downstream mechanisms: excitotoxicity and free radical attack [12]. Following the initial brain injury, a series of dysmaturational disturbances in both white and gray matter lead to an amalgam of damage termed encephalopathy of prematurity [13, 14]. During recent years, improvements in perinatal care have reduced the incidence of cystic white matter injury (WMI), and diffuse WMI becomes the encephalopathy most frequently seen in preterm infants [4].

A notable characteristic of preterm WMI is hypomyelination. Normal ensheathment is critical for axonal differentiation, which further plays an essential role in cerebral cortical development [14]. Pre-myelinating oligodendrocytes (OLs) are extremely vulnerable to various insults following preterm birth and are considered a principal cellular target in WMI [15, 16]. Once injured, pre-OLs can be replenished in the subacute period but fail to differentiate to later phases, leading to the subsequent hypomyelination. Moreover, the failure in pre-OLs' differentiation is accompanied by vigorous and persistent astrogliosis and microgliosis, contributing to subsequent dysmaturational events [17-20].

Oxidative stress is among the main causes of diffuse WMI [4]. During the perinatal period, the concentration of free radicals can drastically increase because of inflammation, hyperoxia, hypoxia, ischemia-reperfusion, neutrophil and macrophage activation, etc. These excessive free radicals may disturb the normal redox state of cells, thereby increasing oxidative stress and causing toxic effects. The neonatal brains, especially the preterm brains, are susceptible to oxidative stress due to their immature anti-oxidant defense systems and reduced anti-oxidant capacity [21-25]. Therefore, the brains of preterm infants experience a greater accumulation of reactive oxygen species after insults [26, 27]. As mentioned earlier, pre-OLs are selectively vulnerable to oxidative stress [28]. Multiple genes regulating the maturation of OLs can be activated by oxidative stress. By promoting global histone acetylation, oxidative stress can block OL differentiation [4, 29, 30]. By activating the caspase systems, oxidative stress can lead to the apoptosis of pre-Ols [25]. Besides, axons also display maturation-dependent vulnerability to oxidative stress [31].

In recent years, accumulating evidence indicates that normal neuroinflammation is pivotal for neurodevelopment. Early brain development is a dynamic process that involves rapid rates of neural cell proliferation, differentiation, and death, in which numerous newly generated brain cells are pruned back. This dynamic process generates large amounts of DNA damage, cellular debris, and byproducts of cellular stress, all of which can activate immune signaling. Deficits in the immune surveillance and response to these insults can lead to neurodevelopmental impairment [32]. Recent studies found that some key inflammasome components such as AIM2 are highly expressed during the fetal period [33], which may work through recognizing and removing genetically compromised cells [34].

Microglia, the resident macrophages in the central nervous system, play an important role in neuroinflammation. They are essential for axonal development, synaptogenesis, synaptic pruning, neural circuit formation, OLs proliferation and differentiation, and vascularization [35-38]. In humans, microglia become prominent in the forebrain during 16~22 weeks' gestation and migrate progressively through the white matter from 20 to 35 weeks, then to the cerebral cortex [39, 40]. Preterm birth happens during the critical period when microglia are heavily populated in cerebral white matter. They can be activated by a variety of inflammatory insults accompanied with preterm birth and transform into macrophage-like phenotypes, which have the ability for phagocytosis, proliferation, and migration into the areas of injury [41]. These cells can generate free radicals, secrete proinflammatory cytokines, and enhance excitotoxicity, thereby exacerbating the disturbances in the premature brain [42]. Previously, microglia have been generally characterized as proinflammatory and anti-inflammatory phenotypes. A recent study identified at least nine distinct microglial subpopulations with unique molecular signatures that changed throughout the course of development [43]; therefore, their functions must be much more complex than we currently realize.

The etiology of neurodevelopmental impairment after preterm birth is multifactorial, with inflammatory processes being a crucial driving force. In animals subjected to insults such as infection, inflammation, hypoxia, and hypoxic-ischemia during the perinatal period, significant myelination deficits are seen in association with neuroinflammation [44]. In humans, sustained elevations of acute inflammation-related proteins during the first postnatal month of life in extremely preterm neonates were associated with an increased risk of adverse neurodevelopment at ten years old [45]. Clinical observational data revealed that elevated levels of inflammatory proteins in the cerebrospinal fluid of infants were associated with a higher risk of neurocognitive and behavioral impairment [46-48]. It is widely accepted that normal labor involves physiological inflammatory processes [49]. Preterm birth exposed the premature brain to labor-related inflammation ahead of schedule, which might be harmful. Moreover, preterm birth is usually precipitated by pathological proinflammatory events (“first inflammatory hit”) and is accompanied by postnatal exposure to sustained inflammation caused by hypoxia, artificial devices, drugs, etc (“second inflammatory hit”) [50]. Together, these factors disturb the normal developmental trajectory of the brain and lead to preterm brain injury, as shown in Fig. (1).

Fig. (1).

Inflammatory processes as a driving force in preterm brain injury. Normal labor involves physiological inflammatory processes, while preterm birth exposed the premature brain to such labor-related inflammation ahead of schedule. Moreover, preterm birth is usually precipitated by pathological proinflammatory events, including maternal systemic or intrauterine infection/inflammation (“first inflammatory hit”). After birth, preterm infants are inevitably exposed to hypoxia, artificial devices, drugs, and so on, which may cause sustained inflammation (“second inflammatory hit”). These inflammation processes involve the expression of proinflammatory cytokines, activation of immune cells, increased production of reactive oxygen and nitrogen species, excitotoxicity, and mitochondrial impairment. Together, these inflammatory processes lead to the maturation arrest or death of oligodendrocytes and axonal loss or death of neurons. Also, they turn the microglia and astrocytes into an activated state, which can further aggravate neuroinflammation and cause insults to oligodendrocytes and neurons (dashed arrow). Eventually, these factors disturb the normal developmental trajectory of the preterm brain and lead to preterm brain injury, resulting in neurodevelopment impairment, including cerebral palsy and cognitive, visual, social-behavioral, attention, and learning disabilities.

Since the disturbances affecting brain development occur over a relatively long period, there is a chance for clinicians to provide interventions. Based on the pathogenesis of premature brain injury, neuroprotective interventions should prevent initial cell death and regulate secondary dysmaturational events and neuroinflammation. To date, many experimental agents seem to have neuroprotective potential, among which erythropoietin (Epo) is considered a prime candidate [51].

2.2. Structure, Synthesis, and Regulation of Epo

Human Epo is a 34kDa glycoprotein cytokine composed of 4 helical bundles. Its gene is situated at chromosome 7q11-22 and encodes a 193-amino acid polypeptide chain [52, 53]. The mature circulating Epo is a heavily glycosylated protein, nearly 40% of which consists of carbohydrate moieties, including 3 N-linked sugars at positions 24, 38, and 83 and 1 O-linked sugar at position 126 [54], which are necessary for the biological activity of Epo. Besides glycosylation, the sialylation process is also essential for the normal function of Epo since it adds stability to the Epo molecule and prolongs Epo's circulating time [55].

In humans, the liver is the primary site of Epo production during fetal and neonatal life, while renal Epo mRNA expression increases significantly after 30 weeks of gestation, indicating that the Epo-production site switches from the liver to kidneys during this period [56]. Besides the liver and kidneys, Epo is also produced by cells from various tissues, including the heart, spleen, lung, testis, ovaries, retina, and the brain [57], where it has non-erythropoietic functions. Among all these Epo-producing sites, the identification of Epo in the brain has sparked the greatest enthusiasm. Epo-producing cells within the nervous system, including neurons, astrocytes, oligodendrocytes, microglia, and endothelial cells, release Epo as a paracrine and autocrine signal in a regulated manner [58, 59]. Additionally, a recent study identified brain pericytes as major sites for Epo production under hypoxic conditions [60]. Interestingly, previous lineage-tracing studies have suggested that both brain and kidney pericytes are neural crest-derived, which might be a possible explanation for their similar function in different sites [61].

Since Epo is a large glycosylated protein (molecular weight >30,000 Daltons), the normal blood-brain barrier (BBB) is relatively impermeable to Epo [62]. To some extent, brain-derived Epo is separated from the systemic circulation, so detecting Epo in the brains of both fetuses and adults lets people guess that it might have a continuous role in brain development and function maintenance.

The production and secretion of Epo in the nervous system are induced by hypoxia, mechanical damage, infection, metabolic stress, etc. [63-66]. Among these challenges, hypoxia has been thoroughly studied and has proven to be the most influential factor for Epo production. Under hypoxic conditions, transcription factor complex hypoxia-inducible factor-1 (HIF-1) is at the center of the signaling cascade of Epo production [53, 67]. Notably, although the expression of Epo seems to be stimulated by prolonged hypoxia (on the order of hours) in all organs, the temporal pattern of regulation is tissue-specific. For example, the expression of Epo in kidneys and liver is transient, peaking during 6~24 hours in vivo (depending on the severity of hypoxia) and at 6 hours in vitro [68], while the mRNA level of Epo in the brain remains elevated for the whole duration of the hypoxic stimulus and even mild hypoxia keeps Epo mRNA level up for more than 12 hours [69].

Besides Epo, some artificial variants of Epo have been developed to modify its pharmacological properties. Darbepoetin, which has extended circulating half-life because of the additional oligosaccharide chains, is one of the few variants which have been clinically tested in preterm infants (NCT01471015 [70], NCT01207778 [71, 72]). The carbamylated derivative of Epo (cEpo) has been reported to be a non-erythropoietic form of Epo, preventing neuronal cells' apoptosis without inducing erythroid cell growth [73, 74]. However, it has not been assessed in clinical trials till now.

2.3. Mechanism of Epo in Neuroprotection

In the presence of Epo, signal transduction is induced by a conformational change of the extracellular region of Epo receptor (EpoR), which consists of two domains that provide two discrete binding sites for a single Epo. These two sites are bound sequentially, forming a homodimeric receptor complex. Then the transmembrane domain transfers this conformational change across the cell membrane to the intracellular region [75, 76].

As the prime regulator of erythrocyte production, Epo's hematopoietic mechanism has been thoroughly studied. By inhibiting programmed cell death (apoptosis) of erythroid precursors, Epo allows erythrocytes' maturation and promotes erythroid differentiation. Such anti-apoptotic effects also work in neurons and protect brain tissues [77]. Although EpoR has no endogenous tyrosine kinase activity, the conformational change of the EpoR dimer can activate Janus kinase 2 (JAK2), thereby activating multiple signaling cascades including signal transducer and activator of transcription 5 (STAT5), mitogen-activated protein kinase (MAPK), phosphatidylinositol 3-kinase (PI3K), serine/threonine kinase (Akt), and nuclear factor kappa B (NF-κB) [78-85]. Then NF-κB and STAT5 move into the nucleus and act as transcription factors for anti-apoptotic genes such as B-cell leukemia/lymphoma (Bcl)-xL, eventually upregulating anti-apoptotic proteins [86], as shown in Fig. (2).

Fig. (2).

When an Epo molecule binds to 2 adjacent receptors on the cell surface, the conformational change of the EpoR dimer activates JAK2, thereby activating downstream signaling messengers including STAT5, PI3K/Akt, MAPK, and NF-κB. NF-κB and STAT5 can move into the nucleus and act as transcription factors for anti-apoptotic genes of the BcL family, such as BcL-xL, thus resulting in the upregulation of anti-apoptotic proteins. The activity of glycogen synthase kinase 3β (GSK3β) was also inhibited, thus inhibiting the mitochondria-related proapoptotic process. Some of the pathways act directly, while others act indirectly by inhibiting the activity of a group of caspases, which in turn inhibits DNA fragmentation in the nucleus and cellular phosphatidylserine membrane exposure. The upregulation of hypoxia-inducible factor-1 (HIF-1) results in increased Epo expression, therefore enhancing the effect.

Besides the anti-apoptotic effect, in vitro and in vivo studies reported that Epo also acts as an anti-inflammatory [87, 88], anti-excitotoxic, anti-oxidant [89], angiogenetic, and neurotrophic agent [90]. Evidence has shown that Epo also decreases neuron and oligodendrocyte death and promotes neurogenesis [91] and oligodendrogenesis [92-95]. Moreover, its erythropoietic effect, in return, increases iron utilization, thus decreasing circulating, potentially unbound iron, which may produce harmful free radicals [96]. Epo also shows additional neurotrophic effects and promotes the differentiation of neurons from stem cells [6, 97]. During in vitro studies, Epo promotes axonal sprouting [98] and neurite outgrowth [99], which enhances neuron proliferation [100]. One possible mechanism of this neurotrophic effect is increasing the level of brain-derived neurotrophic factors [101], which is vital for the survival and growth of neurons [102]. Notably, the beneficial effects of Epo were usually accompanied by the reduction in microglial activation [103, 104], and further investigation found that Epo modulated the polarization of microglia towards the protective M2 phenotype via the JAK2-STAT3 pathway, thus alleviating neuroinflammation [105]. The most likely mechanism of Epo’s anti-oxidative effect is that it can promote the extracellular signal-regulated kinase (ERK)/nuclear erythroid 2-related factor 2 (Nrf2) and anti-oxidant response element (ARE) pathway [89, 106, 107]. The neuroprotective effect of Epo via the Nrf2-ARE pathway has been reported in rodent models of subarachnoid hemorrhage (SAH) [108], traumatic spinal cord injury [109], and cerebral ischemia [110]. Nrf2 is a global promoter of the anti-oxidant and anti-inflammatory response. Epo induces nuclear translocation of Nrf2, which is an important step of Nrf2 activation [111]. The activation of Nrf2 upregulates its target genes in neural cells, such as peroxiredoxin (PRX), glutathione peroxidase (GPX), and heme oxygenase-1 (HO-1), thus modulating neural oxidative stress by inducing anti-oxidant and detoxifying enzymes [111-113]. Briefly, current studies indicate that Epo's neuroprotective effect may work through multiple functions (Fig. 3).

Fig. (3).

Epo's neuroprotective effect may work through anti-apoptotic, anti-inflammatory, anti-oxidant, anti-excitotoxic, neurotrophic, angiogenetic, erythropoietic, and neurogenetic functions.

Additionally, several new pathways have been discovered recently. Sims reported that Epo positively regulates the expression and activity of system Xc−, which is the transporter responsible for the uptake of cystine for the production of glutathione, thus exerting its neuroprotective effect [114]. Jantzie's group found out that excess calpain activity contributed to the pathogenesis of encephalopathy of prematurity, and Epo could modulate cerebral and serum degradation products from excess calpain activation [115]. Garzón et al. reported the neuroprotective effect of a low sialic form of Epo. They found it protected cortical neurons from glutamate-induced apoptosis by increasing Bcl-2 and inhibiting glutamate-induced activation of caspase-3 [116]. NRF1 and REST, which are important for neural development and maintenance, were also found to be overexpressed after Epo administration, suggesting they might also be the mediators of Epo signaling [117].

3. EFFICACY AND SAFETY OF EPO IN NEUROPROTECTION

In 1985, with the identification of the corresponding nucleotide sequences, two groups of investigators independently cloned the human Epo gene [118], paving the way for industrial manufacturing of recombinant human Epo (rhEpo). In preterm infants, the low plasma level of Epo provides a rationale for the exogenous administration [119]. Prophylactic rhEpo treatment has been recommended to reduce red blood cell transfusion in preterm infants, and the recommended dose is 750 IU/kg per week for six weeks [120]. As for neuroprotection, the optimal dose and optimal time to start therapy are still unknown, but previous animal studies indicate that neuroprotection requires higher doses [121]. Since the molecular weight of rhEpo is larger than the cut-off for lipid-mediated free diffusion across BBB [62], it is difficult for the rhEpo molecule to cross normal BBB. However, under hypoxic or other pathological conditions, the endothelial barrier may become leaky. Besides, the BBB of preterm infants has not fully developed and has higher permeability. The finding that the EpoR is expressed at the brain capillaries indicates that systemically administered rhEpo may initiate endocytosis and be translocated across the BBB. A full understanding of how EPO mediates its effects across the BBB has not been elucidated yet, but current experimental evidence suggests that when administered at high doses, rhEpo administered systematically can cross BBB to the degree sufficient for neuroprotection [9, 122].

3.1. Animal Models

3.1.1. Efficacy

Various animal models have reported the neuroprotective potential of Epo. Mazur et al. established a prenatal systemic hypoxia-ischemia (TSHI) rat model to mimic human early-third trimester placental insufficiency, and their results showed that prenatal TSHI significantly upregulated EpoRs on oligodendrocytes and neurons. Postnatal rhEpo administration after TSHI minimized histological damage and rescued oligodendrocytes and gamma-aminobutyric acidergic interneurons [123]. Chung et al. reported consistent results in a pediatric mouse model that rhEpo can rescue the decrease in hippocampal neurogenesis and production of pre-OLs caused by chronic hypoxia during postnatal day P21 to P35 [124]. Similarly, another group created a hyperoxia-induced brain injury rat model and revealed that high-dose rhEpo (5000 IU/kg per day, from P3~P6) significantly reduced myelination deficits and improved behavioral outcomes in adolescent rats [125]. Recently, in a germinal matrix-intraventricular hemorrhage model created by administering intraventricular collagenase to P7 mice, rhEpo significantly reduced neuroinflammation, limited brain atrophy and ventricle enlargement, restored neuronal density, and ameliorated dendritic spine loss, thus preserving the learning and memory abilities [126].

Genetic approaches were also used to study the neuroprotective function of Epo. EpoR knockout in mice led to severe neuronal apoptosis, while culturing neurons with a functional receptor significantly increased cell viability compared to knockouts [127]. Similarly, mice without either Epo or EpoR gene demonstrated severe embryonic neurogenesis defects, suggesting the essential role of Epo in embryonic neurogenesis. Furthermore, they created a post-stroke model with these EpoR knockdown mice and demonstrated that deletion of EpoR resulted in a significant reduction in cell proliferation in the subventricular zone after stroke [6].

In addition to rodents, preterm brain injury models were also created with larger mammals. In an ovine model, lipopolysaccharide (LPS) was given at 107±1 days of gestational age (term is 147 days) through a catheter implanted into a fetal femoral vein to induce fetal inflammation. A single dose of rhEpo (5000 IU/kg) or saline was injected into the fetuses. The results showed that Epo was beneficial to myelination and reduced brain injury [128]. Later, the study by Barton et al. provided a more detailed description of the differential region-specific effects of early rhEpo (5000 IU/kg) on white matter in preterm lambs after mechanical ventilation. In their study, lambs were delivered at 126 days of gestation and received high tidal volume ventilation to create a ventilation-induced lung injury model. According to their study, compared to the Ventilation group, the rhEpo+Ventilation group had: 1) a higher level of IL-1β and IL-6 mRNA in periventricular white matter (WM), a higher level of IL-8 mRNA in subcortical WM, 2) a lower density of microglia or astrocytes in subcortical WM, 3) a higher level of occludin and claudin-1 mRNA in periventricular WM, and 4) a lower rate of vascular leakage [129].

3.1.2. Underlying Risks

As a pleiotropic cytokine, Epo has been clinically used for its erythropoietic effect. When used for neuroprotection, it might have a cross-talk with other systems and lead to unwanted outcomes [130]. The possible impact of Epo on different systems is discussed as follows.

In the brain, by comparing the infarct volumes and number of immunoreactive cells in stroke mice between Epo and control group, Wiessner et al. demonstrated that Epo administration increased hematocrit and blood viscosity, thus counteracting its beneficial effects and eventually amplifying brain injury [131]. Additionally, pharmacological doses of rhEpo may also stimulate platelet production and increase platelet reactivity towards thrombin in dogs, which may be responsible for the potentiated thrombogenicity [132, 133]. These adverse effects have only been studied in adult animals, and whether neonates may have similar responses remains unknown.

Since Epo has exhibited similar angiogenic potential as the vascular endothelial growth factor (VEGF) [134], the relationship between Epo and retinopathy of prematurity (ROP) is worth debating. Previous observational studies suggested Epo treatment might lead to pathologic intravitreal neovascularization, thus increasing the risk of ROP [135]. However, from a developmental perspective, this statement is too general. After preterm birth, the sudden postnatal increase of tissue oxygen tension suppresses the VEGF mRNA expression, leading to delayed physiologic retinal vascular development (phase I of ROP), which further creates hypoxia and stimulates pathological vasoproliferation (phase II of ROP) [136]. Based on the two phases of ROP, researchers hypothesized that Epo might contribute to both physiologic and pathologic angiogenesis, depending on the administration time. Bretz et al. placed knock-in mice with reduced EpoR signaling and wild-type mice in 75% oxygen from P7 to P12 to induce retinopathy and compared their retinal vascularization at P12 and P17. The results revealed a significant increase in the amount of avascular area and a significant decrease in intravitreal neovascularization in knock-in mice, suggesting that EpoR signaling contributed to pathologic retinal angiogenesis after hyperoxia [137]. With the same ROP model, they recently found that reduced EpoR signaling resulted in reduced retinal function after injury evaluated at 8 weeks [138, 139], which supports Epo's role in restoring retinal function. Unlike previous observational studies, recent random-controlled trials reported that Epo has no significant influence on ROP development [140, 141]. As a result, it may be appropriate to acknowledge that prior studies have been refuted, and Epo does not significantly alter the risk of ROP in humans. Future studies should assess the influence of Epo on ROP based on the developmental stages.

Some studies have mentioned that an early high dose of Epo might exacerbate lung injury. In LaRosa's study, rhEpo was administered via the umbilical vein to preterm lambs who received invasive ventilation immediately after birth. Their data revealed that early low-doses of rhEpo neither provided any short-term respiratory benefit nor exacerbated ventilation-induced lung inflammation and injury, while high doses (³3000 IU/kg) exacerbated lung inflammation and injury in ventilated preterm lambs [142]. Such dose-dependent effect of Epo was confirmed in their later research [143]. Similarly, Polglase et al. reported that administration of rhEpo at a relatively high dose (5000 IU/kg) led to more severe lung injury in ventilated preterm lambs [144]. Therefore, the dosage of Epo should be carefully considered when administered to ventilated babies.

These underlying risks temper enthusiasm for rhEpo as a neuroprotective agent in preterm infants. It seems that molecules retaining the neuroprotective effect without interfering with other systems would be preferred when administered systematically.

3.2. Clinical Trials

After searching systematically on PubMed, Embase, and Cochrane Library, and tracking clinical trials registered on clinicaltrials.gov [9], we summarized and evaluated currently available clinical trials focused on the neuroprotective effect of Epo in preterm infants. The literature search was following search formula (taking PubMed as an example): ((Epo) OR (erythropoietin)) AND ((preterm) OR (premature) OR (low birth weight)) AND ((brain injury) OR (neuroprotective) OR (neurodevelopment) OR (Bayley) OR (IQ) OR (DQ) OR (intelligence) OR (cerebral palsy)). To date, 11 clinical trials involving 8 teams from 3 countries have been registered on clinicaltrials.gov to investigate the neuroprotective effect of Epo in preterm infants. All registered clinical trials are listed in Table 1, and more details are shown on www.clinicaltrials.gov.

Table 1.

Summary of clinical trials.

Investigator and Site (NTC Number)	Study Design (Phase)	Number of Patients (Gestational Age)	Intervention	Current Status	Main Results
Hans Ulrich Bucher, Swiss Neonatal Network, Switzerland (NCT00413946)	RCT (Phase II)	448 (26~32 weeks)	rhEpo, 3000IU/kg i.v. at 3hrs, 12~18hrs, and 36~42hrs after birth, a total of 3 doses	Completed	Pharmacokinetics: significantly higher renal Epo excretion in more immature infants (<29 weeks GA). Epo did not accumulate after multiple doses. Safety: no excess in major adverse events in the Epo group. Short-term efficacy: significantly higher FA value in Epo group at the genu and splenium of the corpus callosum, the anterior and posterior limbs of the internal capsule, and the corticospinal tract bilaterally at 40 weeks PMA. Long-term efficacy: no significant differences in BSID II score at 24 months corrected age.
Hans Ulrich Bucher, Swiss Neonatal Network, Switzerland (NCT02076373)	RCT (Phase III)	120 (23~31 weeks)	rhEpo, 2000IU/kg i.v., first dose at d5 (±2 days) of life, 2nd and 3rd dose at 24 hrs and 48 hrs later, maintaining dose at 10 and 17 days after the first dose	Recruiting	Safety (not reported yet): incidence of adverse events. Short-term efficacy (not reported yet): MRI at 40 weeks PMA. Long-term efficacy (not reported yet): Intelligence Quotient at 5 years of age.
Huiqing Sun, Zhengzhou Children's Hospital, China (NCT02036073)	RCT (Phase IV)	800 (£32 weeks)	rhEpo, 500IU/kg i.v., beginning within 24 hrs after birth, every other day for 14 days	Completed	Safety: no pronounced side effect in Epo group. Short-term efficacy: significantly lower incidence of PVL and grade III-IV ICH, NEC, sepsis in EPO group, similar incidence of ROP and BPD between groups. Long-term efficacy: significantly lower incidence of moderate or severe neurological disability assessed by BSID II at 18 months corrected age in Epo group, significantly lower incidence of deafness in the rhEpo group, similar incidence of cerebral palsy, blindness between groups.
Huiqing Sun, Zhengzhou Children's Hospital, China (NCT02745990)	RCT (Phase IV)	440 (£28 weeks)	rhEpo, 500IU/kg i.v., beginning within 72 hrs after birth, every other day up to 32 weeks PMA	Recruiting	Short-term efficacy (not reported yet): early blood biomarkers at 4 weeks after birth, MRI at 40 weeks PMA. Long-term efficacy (not reported yet): Bayley II score, mortality, the incidence of cerebral palsy, blindness, and deafness at 24 months corrected age
Changlian Zhu, Third Affiliated Hospital of Zhengzhou University, China (NCT03914690)	RCT (Phase II)	308 (£32 weeks)	rhEpo, 500IU/kg i.v., beginning within 72hrs after birth, every other day for 2 weeks	Partially completed	Short-term efficacy: significantly higher FA values at the posterior limb of the internal capsule, the splenium of the corpus callosum, frontal white matter, and occipital white matter in the Epo group. Long-term efficacy (not reported yet): Bayley IIscore, mortality, the incidence of cerebral palsy, blindness, deafness at 18 months corrected age
Xihui Zhou, First Affiliated Hospital of Xian Jiaotong University, China (NCT03110341)	RCT (Phase III)	400 (£32 weeks)	rhEpo, 750 IU/kg i.v., starting within 72hrs after birth, every other day for 2 weeks	Unknown	Short-term efficacy (not reported yet): DTI at 40 weeks PMA. Long-term efficacy (not reported yet): Bayley IIscore at 18 months corrected age.
Wenhao Zhou, Children's Hospital of Fudan University, China (NCT02550054)	RCT (Phase II)	312 (£32 weeks)	rhEpo, 1000IU/kg i.v., beginning within 48 hrs after birth, every 48hrs for 6 doses after birth, then 400IU/kg s.c. 3 doses per week to 34 weeks PMA	Recruiting	Safety (not reported yet): incidence of adverse events. Short-term efficacy (not reported yet): serum biomarkers for oxidative stress, inflammation, and immune response at 34 PMA, incidence of IVH and PVL during the neonatal period. Long-term efficacy (not reported yet): MRI at 9 months and 18 months corrected age, Bayley III score at 18 months corrected age.
Sandra Juul, University of Washington, USA (IND 12656)	prospective, dose-escalation, open-label (Phase I/II)	60 (£28 weeks)	rhEpo, 500, 1000, or 2500 IU/kg i.v., beginning within 24 hrs after birth,3 injections at 24 hrs interval	Completed	Pharmacokinetics: steady-state of plasma Epo concentration was achieved within 24 to 48 hours after administration, the doses of 1000 IU/kg and 2500 IU/kg could produce neuroprotective serum concentrations, while lower dosage could not. Safety: no significant difference in adverse events between groups, including hypertension, ROP, RDS, BPD, NEC, sepsis, and death during hospitalization.
Sandra Juul, University of Washington, USA (NCT01378273)	RCT (Phase III)	941 (24~28 weeks)	rhEpo, 1000IU/kg i.v., beginning within 24 hrs after birth, every 48hrs for 6 doses after birth, then 400IU/kg s.c. injection 3 doses per week to 33 weeks PMA	Completed	Safety: no significant differences between groups in the incidence of ROP, ICH, BPD, NEC, sepsis or death, or other severe adverse events. Long-term efficacy: no significant differences between groups in the incidence of severe neurodevelopmental impairment assessed by Bayley III at 24~26 months corrected age.
Robin K Ohls, University of New Mexico, USA (NCT00334737)	RCT (Phase II)	102 (£32 weeks)	rhEpo, 400 IU/kg s.c., 3times a week; darbepoetin, 10 mics/kg sc. once a week; both from the first week of life through 35 weeks PMA or ×10 weeks	Completed	Safety: no differences between groups in the incidence of ROP, IVH, NEC, BPD, PDA, or other adverse events. Long-term efficacy: significant higher Bayley III cognitive score at 18~22 months corrected age in ESA (Darbe and rhEpo) recipients. significantly lower incidence of cerebral palsy in ESA recipients, no significant difference in the incidence of visual or hearing impairment.
Robin K Ohls, University of New Mexico, USA (NCT01207778)	observational prospective (not applicable)	105 (£32 weeks)	rhEpo, 400 IU/kg s.c., 3times a week; darbepoetin, 10 mics/kg s.c. once a week; both from the first week of life through 35 weeks PMA or ×10 weeks	Completed	Long-term efficacy: significantly better neurocognition assessed by Full Scale Intelligence Quotient (FSIQ) and performance Intelligence Quotient (IQ) in the ESA group at 42~48 months and 66~72 months old, better performance on executive function tasks in the ESA group, significantly lower incidence of neurodevelopmental impairment in the ESA group.
Bai-Horng Su, China Medical University Hospital in Taiwan, China (NCT00910234)	RCT (Phase I/II)	100 (£32 weeks)	darbepoetin, 10 mics/kg s.c. once a week;	Unknown	Safety: incidence of IVH, PVL, ROP, sepsis, NEC, PDA, and BPD

RCT: randomized controlled trial, GA:gestational age, PMA: postmenstrual age, DTI: diffusion tensor imaging, FA: fractional anisotropy, BSID: Bayley Scales of Infant and Toddler Development, PVL: periventricular leukomalacia, IVH: intraventricular hemorrhage, ICH: intracerebral hemorrhage, NEC: necrotizing enterocolitis, ROP: retinopathy of prematurity, BPD: bronchopulmonary dysplasia, RDS: respiratory distress syndrome, PDA: Patent ductus arteriosus, i.v.: intravenous injection, s.c.: subcutaneous injection.

3.2.1. Pharmacokinetics and Safety

Low-dose Epo has been used to prevent or treat anemia in preterm infants for many years, and side effects are rare [119]. However, since the doses for neuroprotection are much higher (1000~5000 U/kg/dose) compared with the doses to treat anemia (100~500 U/kg/dose) [145], the safety of Epo must be reevaluated before being used as a neuroprotective agent.

In 2006, a phase I/II clinical trial by Juul's team, registered with the FDA (IND# 12656), was carried out to obtain pharmacokinetic and preliminary safety data of rhEpo in preterm neuroprotection. Extremely low birth weight infants were enrolled in the study and randomly received 3 times of intravenous rhEpo at one of the 3 doses (500, 1000, or 2500 IU/kg) at 24-hour intervals. Their result showed such doses were well tolerated by extremely low birth weight infants and caused no adverse events. The steady-state of plasma Epo concentration was achieved within 24 to 48 hours, and doses of 1000 and 2500 IU/kg produced peak serum Epo concentrations comparable to neuroprotection previously seen in experimental rats, while doses of 500 IU/kg did not [146]. Further, this team studied the pharmacokinetics of Epo in 47 neonates who were ³36 weeks gestational age and diagnosed with moderate to severe hypoxic-ischemic encephalopathy (HIE). Notably, they stated that Epo exposure with the dose of 1000 IU/kg every 48 hours was slightly lower than the target for neuroprotection (AUC_48h=140,000 mU*h/ml). Shortening the dosing interval from every 48 hours to 24 hours during the first two days could ensure adequate exposure [145, 147]. However, as shown in Table 1, except for Bucher's team, others still choose the dosing interval of 48 hours. A possible reason for this may be that the BBB of preterm infants has higher permeability, so adequate Epo concentrations in the brain can already be reached with this dosing strategy.

In the same year (2006), Bucher's team started a randomized, placebo-controlled, multicenter trial (NCT00413946). They enrolled preterm infants whose gestational ages are between 26~32 weeks and gave repetitive short-term infusions after birth (3 × 3000 IU/kg within 42 hours).The Epo concentrations were measured in the first two consecutive urine specimens after each rhEpo infusion. The results demonstrated that the renal Epo excretion was significantly higher in preterm infants with gestational age less than 29 weeks than in more mature infants, which might be attributed to a higher glomerular filtration rate of the more immature kidneys. Additionally, their data paralleled Juul's results by stating that the urinary Epo concentrations after the first, second, and third high-dose rhEpo infusions were close, indicating that rhEpo did not accumulate after multiple doses in very preterm infants [148]. According to their report, no significant adverse effects of early high‐dose rhEpo treatment were identified in all enrolled preterm infants during hospitalization. No significant differences between groups were found in the incidence of mortality, ROP, intraventricular hemorrhage (IVH), sepsis, necrotizing enterocolitis (NEC), and bronchopulmonary dysplasia (BPD) [149-151].

3.2.2. Short-term Efficacy

Advanced magnetic resonance imaging (MRI) technique such as diffusion tensor imaging (DTI) and magnetic resonance spectroscopy (MRS) has provided a new perspective for understanding premature brain injury and assessing brain development. With these new methods, it becomes possible to monitor the brain's structural or functional changes in vivo.

After evaluating the pharmacokinetics and safety of Epo in very preterm infants, the Swiss team (NCT00413946) followed up the neurodevelopment of included infants. High-resolution DTI was obtained at the term equivalent age. Preterm infants treated with rhEpo demonstrated significantly higher mean fractional anisotropy (FA) value in the genu and splenium of the corpus callosum, the anterior and posterior limbs of the internal capsule, and the corticospinal tract bilaterally [152]. Besides, this team established a scoring system to semiquantitatively assess brain injury, with which they found that Epo treatment was associated with a significantly lower incidence of white matter injury, white matter signal intensity, periventricular white matter loss, and gray matter injury [153, 154]. Notably, using graph theory analysis, this team's latest report revealed that Epo treatment led to stronger connections in a structural network mainly consisting of frontal and temporal lobe bilaterally, especially the peripheral and feeder connections of the core structural connectivity network, thus resulting in a globally increased clustering coefficient, higher global and average local efficiency [155].

Another research team from China (NCT03914690) reported a similar improvement in FA value in the rhEpo group at the corrected gestational age of 35~37 weeks. They enrolled preterm infants whose gestational ages were less than 32 weeks and randomly assigned them to the rhEpo group or control group. Significantly higher FA values were found in the posterior limb of the internal capsule, the splenium of the corpus callosum, frontal white matter, and occipital white matter [156].

3.2.3. Long-term Efficacy

Although short-term efficacy assessed by advanced MRI all support Epo's role in promoting WM development in preterm infants, long-term neurodevelopmental tests demonstrated divergent results.

The trial in Switzerland mentioned above (NCT00413946) collected neurodevelopmental outcome data using Bayley Scale of Infant and Toddler Development, third edition (BSID III) at a mean age of 23.6 months, and no statistically significant differences were found between the rhEpo treatment group and the saline group [157].

In 2020, Juul's team completed a multicenter, randomized, double-blind trial (NCT01378273) and yielded similar results. In their study, 941 extremely preterm infants born at 24⁺⁰ weeks to 27⁺⁶ weeks of gestation were randomly assigned to receive Epo or placebo. The initiating dose of rhEpo is 1000 IU/kg given intravenously every 48 hours. A total of 6 doses are given, followed by a maintenance dose of 400 IU/kg, given by subcutaneous injection 3 times per week till 32 weeks of postmenstrual age. Regarding the primary outcomes, there were no significant differences between the Epo group and the placebo group in the incidence of death or severe neurodevelopmental impairment at 2 years of age. To date, this is the largest clinical trial aiming at assessing the neuroprotective efficacy and safety of Epo in extremely preterm infants. With a reasonable design and enough enrolled patients, this study's results were relatively reliable [11].

However, some studies have still found evidence to support Epo's long-term neuroprotective effect. Ohls's team reported that rhEpo or darbepoetin treatment improved preterm infants' neurodevelopmental outcomes assessed at 2 and 4 years of age with the BSID III (NCT01207778 and NCT00334737). In their study, enrolled patients were randomly given one of the following treatments: darbepoetin (10 mg/kg, once a week, subcutaneously), rhEpo (400 IU/kg, 3 times a week, subcutaneously), or placebo (sham dosing, 3 times a week, subcutaneously) [71, 158]. Further, they evaluated the interaction between Epo and socioeconomic conditions and reported that the beneficial effect of Epo was maximal in children with lower socioeconomic composite scores, indicating that socioeconomic conditions should also be taken into consideration when giving Epo treatment to different individuals [72]. However, when they used advanced MRI to get more objective insights into the change of brain after Epo administration, the result was different. After analyzing the DTI and MRS, they reported that the use of rhEpo or darbepoetin in preterm infants during hospitalization failed to increase FA, cortical thickness, or cortical surface area at 3.5~4 years old compared to the control group [159]. The brain metabolite levels were not statistically significant between groups either [160, 161]. The conclusions from neurobehavioral tests and neuroimages are divergent but explicable because functional changes are not always in line with the structural changes of the brain.

A phase IV clinical trial in China (NCT02036073) enrolled preterm infants whose gestational ages were less than 32 weeks. They were then administered the first dose of rhEpo at 500 IU/kg within 72 hours after birth, followed by the same dosage every other day for 2 weeks. The neurodevelopmental assessment was carried out at 18 months of age by BSID II, and the results demonstrated that the rate of

death or moderate/severe neurological disability in the rhEPO group was significantly lower than the placebo group [162]. Meanwhile, considering the possible gastrointestinal trophic effect of Epo, this research group compared the incidence of NEC between groups (NCT03919500) and found that Epo treatment significantly decreased the incidence of NEC (stage I, II, and III), especially confirmed NEC stage II and III [163].

Longer follow-up data was obtained from an observational study by Neubauer's team. They observed the neurodevelopmental outcome of extremely low birth weight infants treated with either 750 or 1500IU/kg dosages of Epo per week from day 3 of life to 37 weeks corrected age. At 10 to 13 years old, the neurodevelopmental and school outcome of enrolled infants was compared with that of untreated children, showing that the rhEpo group scored significantly better than untreated children in the overall developmental assessment as well as in the psychological examination [164].

Taken together, the long-term outcomes of preterm infants treated with Epo vary from center to center. It is easy to notice that different centers enrolled patients of different gestational age range, applied different treatment strategies, and assessed long-term neurodevelopment at different ages with various measures. Previous preclinical models and clinical trials have proved that a therapeutic concentration for neuroprotection might be much higher than in erythropoiesis [146]. Contrary to this assumption, studies showed that early high-dose Epo failed to improve the neurodevelopment, while lower dosage succeeded. We guess that the difference in gestational age range might be one reason for the divergence because preterm brains develop rapidly and might respond differently to Epo treatment.

4. SUMMARY

As an erythropoietic cytokine, EPO is essential for red blood cell survival and differentiation. Moreover, it is also a pleiotropic agent having the potential to protect pre-OLs from injury and alleviate WMI in preterm infants. Till now, many preclinical models have identified EpoR-mediated neuroprotective effects. However, when results generated from these basic science studies are translated to bedside, Epo's neuroprotective effect varies in different clinical trials.

The previous meta-analysis has merged the results from existing researches and evaluated the safety and efficacy of exogenous Epo in preterm neuroprotection. In 2017, a meta-analysis by Fischer et al. included 4 RCTs comprising 1133 infants. It demonstrated that prophylactic rhEpo administration improved the cognitive development of very preterm infants assessed by BSID at the corrected age of 18~24 months [10]. In the same year, Ohlsson's systematic review found that neurodevelopmental outcomes at about 2 years old varied significantly in different studies [119]. However, obvious heterogeneity exists in the studies included in their review, and the quality of the evidence is low.

The unwanted cross-talk with other systems is one of the major challenges for developing therapeutic Epo. The discovery of other types of EpoR brings hope for solving this problem. In addition to the classic Epo receptor (EpoR), the existence of other types of EpoR in the nervous system has been confirmed by many studies [165-169], and these variants also show neuroprotective potential [170, 171]. Ostrowski reviewed the 3 major isoforms generated from alternative splicing, including the full-length protein as part of the functional receptor, a soluble protein that lacks the transmembrane and intracellular domains, and a truncated protein that lacks large parts of intracellular domains [166]. Besides these 3 isoforms generated from alternative splicing, several other types of EpoR are also expressed in the nervous system. They may serve in tissue protection instead of hematopoiesis, including a heteroreceptor consisting of EpoR and common beta receptor (βcR), the Ephrin (Eph)-B4 receptor, and the human orphan cytokine receptor-like factor 3 (CRLF3) [165]. After combining with different types of EpoR, Epo and Epo-like ligands may activate different signal transduction cascades, thereby avoiding the unwanted hematopoietic effects. Further investigations are needed to map the concise distribution of EpoR variants in the brain and understand the similarities and differences of their molecular pathways.

Another challenge for developing therapeutic Epo is the difficulty for Epo to cross the BBB. Current experimental evidence suggests that neuroprotection requires high doses of systemic rhEpo administration, which may increase the risk of other side effects. Hence, enhancing the BBB permeability of Epo might increase its efficacy for neuroprotection and alleviate other adverse effects. To overcome this challenge, numerous new drug delivery strategies have been developed and tested in rodents. Cell-penetrating peptide (CPP) modification of a therapeutic agent is one of the non-invasive drug delivery approaches for the central nervous system [172-174]. Molecular Trojan horse technology that harnesses endogenous receptor-mediated transcytosis systems such as the BBB transferrin receptor (TfR) has also been reported to have the potential to transport Epo across the BBB [175, 176]. Future work will focus on translating these results to other species, including humans.

Besides Epo, several drugs with promising neuroprotective efficacy have been assessed, such as magnesium sulfate [177], melatonin [178], and vitamin D supplementation [179, 180]. As for preterm infants with a history of asphyxia, therapeutic hypothermia has been considered, but concerns remain around its safety [181]. Additionally, increasing preclinical evidence suggests that mesenchymal stem cells (MSCs) therapy may be a viable strategy for preterm brain injury, which can modulate the inflammatory response and promote proliferation, growth, and differentiation of neurons and Ols [182, 183].

Another attractive strategy is preconditioning, and its effect has been validated in many aging-related neural disorders. The underpinning of the preconditioning strategy is hormesis, which refers to a dose-response phenomenon characterized by low-dose stimulation and a high-dose inhibition [184]. Therefore, using a weak stimulation to activate the endogenous cellular defense system can prepare the body for a stronger one. For example, transient exposure of neurons to a low level of superoxide can protect the neurons against a subsequent lethal level of exposure [185]. Some olive oil polyphenols may prevent neurological symptoms associated with aging through hormesis [186, 187]. Polyphenols also show their potential to protect the neonatal mouse brain against a hypoxia-ischemia insult [188, 189]. Preconditioning can also be used to promote the local production of endogenous Epo. It has been reported that tissue damage triggers a protection and repair system mediated by locally-produced hyposialated Epo [190]. Also, functional hypoxia can trigger neuronal Epo/EpoR expression, thus increasing neuroplasticity and neurogenesis [91]. In addition to hypoxia, preconditioning with inflammation has also been investigated. Chronic intra-amniotic preexposure to Ureaplasma can prevent subsequent LPS-induced brain injury [191]. Preconditioning can also be harnessed as improvement strategies for other treatments. For example, preconditioning of MSCs may enhance their therapeutic efficacy. Although the extraordinary efficacy of several preconditioning strategies has been reported in many animal models, the underlying mechanisms remain obscure, and translation to the clinics has not been completely successful yet. One of the obstacles for translation to bedside is to “get the dose right.”

CONCLUSION

Although there have been many attempts till now, neuroprotective strategies with strong evidence are still very limited. Only refinements of clinical care of preterm infants are regarded as a consensus, and antenatal steroids and magnesium sulfate are recommended in some cases [192, 193]. Since preterm brain injury involves a complex mechanism, strategies that target multiple mechanisms seem to be the most promising. To conclude, based on the evidence we currently know, when the principle of modern neonatal care advocates “less is more,” it is still too early to recommend Epo as the standard treatment for premature brain injury. Considering the safety, Epo might have a cross-talk with other systems and lead to unexpected adverse events. As for the efficacy, the variations in different trials must be better understood before recommending it. Since Epo's effect is dose-dependent, the optimal treatment strategy must be established. Moreover, drug metabolism may vary widely in preterm infants of different gestational ages, so each factor must be individually evaluated in different populations. Besides, different types of injury may benefit differently from Epo treatment and require different treatment strategies. For example, brain injury resulting from preterm birth tends to be a chronic phase, while the injury resulting from asphyxia during the perinatal period tends to be acute. As a result, the latter group may require a shorter duration of therapy. Future studies may focus on the following points: 1) further exploration of the mechanisms of premature brain injury and searching for better therapeutic targets, 2) developing Epo variants retaining the neuroprotective effect without interfering with other systems and with higher permeability across BBB, 3) optimizing Epo treatment strategy for different groups of preterm infants.

CONSENT FOR PUBLICATION

Not applicable.

FUNDING

This study was supported by the National key research and development plan key special project of digital diagnosis and treating equipment (2018YFC0114405).

CONFLICT OF INTEREST

The authors declare no conflict of interest, financial or otherwise.