Several classes of drugs, including NMDA antagonists [1], GABA_A agonists and alcohol [2, 3], which has both NMDA antagonist and GABA agonist properties, trigger widespread neuroapoptosis throughout the developing brain of several animal species [3–11], including non-human primates [12–19]. Included among drugs that have apoptogenic potential are several that are sometimes abused by pregnant women (alcohol, PCP, ketamine, benzodiazepines, barbiturates) and many that are used frequently in pediatric and obstetric medicine (sedatives, anesthetics, anticonvulsants)(reviewed in Creeley and Olney [20]). Evidence documenting neuroapoptosis induced by alcohol and anesthetic drugs is particularly concerning in that millions of human fetuses or infants, every year throughout the world, are exposed to one or more of these agents. The period of peak sensitivity to this neurotoxic mechanism coincides with the brain growth spurt [1, 2], during which neurons throughout the brain are undergoing rapid synaptogenesis. In rodents, the brain growth spurt is confined primarily to the first several weeks after birth, but in humans it spans from mid gestation to several years after birth [21]. During this sensitive period, a single drug exposure, at clinically relevant doses, is sufficient to trigger apoptosis of neurons throughout many regions of the brain [3–7, 10, 11], and cause long-term neurobehavioral disturbances in both rodents [6, 20–22] and NHPs [16].

The human relevance of the above findings remains to be fully elucidated. It is well recognized that alcohol can damage the developing human brain and cause a wide variety of neurodevelopmental disabilities referred to as Fetal Alcohol Spectrum Disorders (FASD) [23–25]. Initially, doubts were raised whether anesthetic drugs can cause similar outcomes [26, 27], but recently reported evidence from several independent research groups suggests that brief exposure of human infants to anesthesia is associated with increased risk for neurobehavioral disturbances, including learning disabilities and attention deficit/hyperactivity disorder [28–34]. While the original findings pertaining to the apoptogenic action of alcohol and anesthetic drugs on the developing brain were described exclusively in terms of an impact on various neuronal populations, we recently reported that exposure of the neonatal NHP brain to isoflurane [35] or propofol [14] anesthesia triggers a robust apoptotic cell death reaction affecting both neurons and oligodendroglia [OL]. Guerri et al. [36], based primarily on in vitro findings, have championed the hypothesis that alcohol might have a toxic effect on astroglia, but OLs have not been suspected of being a target of alcohol’s toxic action. We have reported that exposure of the fetal NHP brain to alcohol on a single occasion can cause widespread death of neurons [15]. In the course of that study, we made the preliminary unreported observation that there was immunohistochemical evidence of a cell death reaction distributed diffusely throughout white matter (WM) regions of these brains. The present study was undertaken to clarify the nature and extent of this toxic reaction to alcohol in WM regions of these alcohol-exposed NHP fetal brains.

Following up on our preliminary observation that sections from alcohol-exposed fetal NHP brains have abundant AC3-positive profiles distributed widely throughout WM regions [Figure  1a], we stained sections adjacent to the AC3-stained sections with the DeOlmos cupric silver method (marker for dying cells), and observed that approximately the same number of cells in the same WM locations were silver positive [Figure  1b], which provides confirmatory evidence that these WM cells are undergoing cell death. There is evidence that cells in the OL lineage can be stained with antibodies to fractin when they are undergoing apoptotic degeneration [37, 38]. Therefore, we stained sections adjacent to those we had stained for AC3 or silver, with antibodies to fractin, and found that fractin staining readily detected a similar number of cells in the same WM locations [Figure  1c], and the appearance of these degenerating WM cells when stained with fractin was very similar to their appearance when stained with AC3 or silver [compare Figure  1a,b,c].

The majority of cells in the WM are either astrocytes or OLs. Therefore, next we performed an immunofluorescent double staining experiment in which AC3-stained sections were also stained with antibodies against GFAP (marker for astrocytes), and found that the AC3-positive profiles did not co-label for GFAP, therefore were not astrocytes [Figure  1d]. In other double staining experiments, consistent with our previously published evidence [35] pertaining to isoflurane-induced apoptosis of OLs, we found that AC3-positive profiles in the WM did not co-label for neuN, PDGFRα, or Iba1, signifying that the vulnerable WM cell is not a neuron, or OL progenitor or a microglial cell.

Collectively, the above findings support the hypothesis that the vulnerable WM cells are OLs that are beyond the progenitor stage of maturation, i.e., OLs that are engaged, or are preparing to engage in myelination. To test this hypothesis we used IHC staining with antibodies to myelin basic protein (MBP) as our primary tool, because MBP IHC provides a means of studying both the myelination process and the role of OLs in mediating that process. We performed MBP staining at several rostrocaudal levels and will present the findings at a rostral level where little or no myelination was occurring, a mid-rostrocaudal level where there was abundant evidence for early myelination, and a caudal level where WM pathways were densely myelinated.

Findings at a rostral level

In sections from the G120 brain cut through the frontal cortex at or rostral to the level of the septum, MBP staining revealed no evidence for myelination [Figure  2a], but at this level there were numerous MBP-positive immature OLs which, judged by their MBP content, were in a state of readiness to begin the myelination process. All of the MBP-positive WM cells at this rostral level had a similar morphology characterized by a dense arbor of radially oriented processes, each of which displayed web-like protoplasmic extensions [Figure  2b]. In alcohol-exposed brains, some of these MBP-positive OLs appeared to be undergoing degeneration, as was evidenced by a halo of particulate debris surrounding the cell body in starburst configuration [Figure  2c], resembling the appearance of degenerating WM profiles stained with silver or fractin [compare Figures  1b and c with 2c]. In double staining experiments it was readily determined that AC3-positive profiles were typically also MBP-positive (Figures  2d,e,f), and DAPI counterstaining revealed an abnormal nuclear chromatin pattern indicative of apoptotic cell death [Figure  2f inset). These findings document that OLs are vulnerable, and onset of vulnerability coincides with the stage when OLs are beginning to synthesize MBP in preparation for myelinating axons. In the literature, OLs at this stage of maturation have been given various designations [37, 39, 40], depending on the model studied and immunological reagents used. In this writing we will use the nomenclature recommended by Butt and Berry [40] which recognizes a progenitor stage in which the OL cell body stains positive for PDGFRα and negative for MBP. The next stage is the promyelination stage in which the OL cell body becomes positive for MBP and negative for PDGFRα. In this stage the OL is accumulating MBP and other constituents of myelin in its cytoplasm in preparation to begin myelinating axons. When the promyelinating OL begins ensheathing axons and transferring MBP to the sheaths, it is referred to as a myelinating OL, which stains positive for MBP, as do the newly formed myelin sheaths.

Findings at a mid-rostrocaudal level

In sections from G120 brains at a mid-rostrocaudal level, MBP staining provided evidence for active myelination in several brain regions, including the centrum semi-ovale, corona radiata, internal capsule and globus pallidus [Figure  3a]. At this level, MBP-positive OLs were plentiful, and many were engaged in myelinating axons [Figure  3b & c]. Typically, the early myelination scene featured many small patches of axons undergoing myelination, each patch being attended by a single OL that extended its processes to multiple axonal segments that became visible by virtue of the MBP that the OL was incorporating into their newly formed sheaths [Figure  3b,c]. OLs that were actively myelinating axons, displayed an arbor of multiple slender processes, each of which contacted and appeared to be myelinating several axons. Some of these profiles that were engaged in myelination appeared to be degenerating [Figure  3c]. Interestingly, degenerating OLs in MBP-stained sections presented in two different patterns, depending on the stage of myelination the OL was in when it committed to cell death. If commitment occurred before the OL had effectively transferred its MBP content to axonal processes, the pattern consisted of MBP-positive debris in starburst configuration surrounding a condensed cell body [Figure  2c]. The second pattern consisted of MBP-positive debris distributed in a linear pattern conforming to the layout of an axonal patch [Figure  3c], signifying that the dying OL had committed to cell death after having already myelinated a group of axons. Double staining experiments documented that MBP-positive profiles that were actively engaged in myelination frequently colabeled for AC3 [Figure  3d,e,f], signifying that they were dying by apoptosis, and DAPI counterstaining of these profiles revealed an abnormal nuclear chromatin pattern confirming the apoptosis diagnosis [Figure  3f inset].

Findings at a caudal level

At a caudal cerebellar/brain stem level, ascending WM tracts were heavily myelinated and MBP staining provided an excellent means of visualizing these myelinated pathways, but MBP did not stain OL cell bodies in these densely myelinated pathways. However, in alcohol-exposed brains, if these densely myelinated pathways were double stained with MBP and AC3, it was possible to detect many cell bodies that were negative for MBP and positive for AC3 amid a dense bed of MBP-positive myelinated axons. In Figure  4a this phenomenon is illustrated in the cerebellar peduncle, a very densely myelinated pathway. Staining of adjacent sections with the silver method (Figure  4b) confirmed that large numbers of cells with OL morphology were undergoing apoptotic cell death in the cerebellar peduncle following alcohol exposure. The observation that two cell death markers identify many dying cells with typical OL morphology in a densely myelinated pathway that has no MBP-positive OL cell bodies, suggests that at this late stage of myelination, the OL cell bodies have ceased generating new MBP and have transferred all of their MBP content to myelin sheaths, which do stain positive for MBP.

The illustrations we have chosen to document the effects of alcohol on the fetal NHP brain are from brain sections of animals at an age early in the third trimester, because this documents that the developing primate brain is vulnerable at the earliest age evaluated in this study, which is an age when human fetuses are frequently exposed to alcohol. We have previously demonstrated [35] that isoflurane anesthesia induces a robust oligoapoptosis response in the primate brain on postnatal day 6. In the present study, the response we observed to alcohol in the G155 fetal brain (1 week before term) was similarly robust (see quantitative evaluation) and showed the same specificity for OL lineage cells at or beyond the maturational stage when they achieve myelination potential. Collectively, our present findings, together with previous findings [35], document vulnerability of the developing primate brain to this toxic mechanism throughout the third trimester of gestation. Thus, exposure of the human fetus to alcohol at any time in the third trimester entails risk of inducing an oligoapoptosis response in the fetal brain.

Quantitative observations

Quantitative counts performed on WM regions of AC3-stained brain sections revealed that the mean (± SEM) density of apoptotic profiles in the 5 ethanol-exposed brains was 840.3 ± 29.11 profiles/mm3 compared to 66.37 ± 17.97 profiles/mm3 for the 4 control brains (Figure  5). This amounts to a mean 12.7-fold increase in glial apoptotic profiles in the alcohol-exposed compared to the control brains. The difference between the means was 773.9 (95% confidence interval 678.9 to 868.9, p < 0.0001). While the increase in profile density caused by ethanol exposure, when averaged across the three ages, was 12.7 fold, the fold-increase was19.44, 15.27 and 8.45 for the early, mid and late third trimester ages respectively. Thus, the greatest difference between experimental and control density counts was at the earlier ages. Cell death counts based on AC3 staining reveals that the dying OLs in all WM regions at all stages of maturation have a remarkably similar morphology. While the density of degenerating OLs (profiles/mm3) decreased from the early to late gestation period, the toxic impact (in terms of total OLs deleted) remained approximately the same across the different ages because brain mass increases two-fold between the beginning and end of the third trimester. Therefore, if the number of deleted OLs per unit mass of tissue becomes half as many as the tissue mass becomes two times larger, the total number of deleted OLs remains the same.

We have previously reported that a single exposure of the third trimester fetal cynomolgus macaque to alcohol causes widespread apoptotic degeneration of neurons throughout gray matter regions of the developing macaque brain [15]. Here we show in the same group of NHP subjects that alcohol also causes widespread apoptotic degeneration of glial cells throughout WM regions, and the type of glial cell affected is of the oligodendrocyte (OL) lineage. During the NHP third trimester, cells in the OL lineage undergo maturational changes in which OL progenitors differentiate into promyelinating, and then myelinating OLs. During the promyelinating stage they accumulate MBP and other constituents of myelin in their cytoplasm in preparation for myelinating axons. Our evidence indicates that it is early in the promyelination stage that they become vulnerable to alcohol-induced apoptosis, and they continue to be vulnerable throughout both early and late stages of myelination.

An interesting feature of alcohol-induced apoptosis in both the mouse [7] and monkey brain is that the cell death reaction affecting both neurons and OLs is bilaterally symmetrical. Cell death in any region of gray or white matter affecting one hemisphere of an alcohol-exposed brain, is precisely mirrored by the same number and distribution of dying cells among homologous cell populations in the contralateral hemisphere. This feature of alcohol neurotoxicity may significantly limit the capacity for functional recovery, in that the extent of recovery may depend, in part, on the availability of intact contralateral cell populations with similar functional properties to fill in for the cell loss.

It is not clear whether alcohol causes oligoapoptosis by the same mechanism(s) as it causes neuroapoptosis. There is substantial evidence that the neuroapoptotic action of alcohol is triggered by a dual mechanism - blockade of NMDA glutamate receptors and hyperactivation of GABA_A receptors [1–3]. Growing evidence for intimate glutamatergic and GABAergic signaling between neurons and OLs via synaptic contacts [41, 42] argues in favor of similar cell surface receptor mechanisms triggering these two toxic phenomena. It has been demonstrated that anesthetic drugs with NMDA antagonist or GABA_A agonist properties also trigger neuroapoptosis in the developing rodent [1, 5, 6, 43, 44] or NHP [12, 14, 17–19] brain, and that anesthetic drugs also trigger oligoapoptosis in the developing NHP brain [14, 35]. Regarding intracellular signaling pathways that may be involved, it has been shown in the in vivo mouse brain that this apoptosis process involves Bax-mediated release of cytochrome c from mitochondria which results in activation of caspases 9 and 3 [45–47]. An important upstream event that occurs within minutes after administration of alcohol or anesthetic drugs (prior to Bax protein activation), is the suppression of ERK (extracellular signal-regulated kinase) phosphorylation. Within minutes after administration, both alcohol and anesthetic drugs trigger suppression of pERK, and lithium counteracts this pERK suppressant action, and prevents alcohol or anesthetic drugs from triggering neuroapoptosis [48, 49]. It remains to be determined whether these apoptogenic agents suppress ERK phosphorylation in oligodendrocytes, and whether lithium protects against oligoapoptosis induced by these agents.

The neurodevelopmental disabilities associated with FASD are heterogeneous; in childhood, they vary from mild attention deficit/hyperactivity disorder (AD/HD) and related learning disturbances to frank mental retardation [24, 25]. Later in life, individuals with a FASD diagnosis have a high incidence of adult-onset neuropsychiatric disturbances, including a 40% incidence of psychosis and 44% incidence of major depressive disorder [23]. The observation that alcohol deletes glia, as well as neurons, from the developing primate brain, and that the glial cell type affected is of the OL lineage, strengthens the proposal that the apoptogenic action of alcohol can explain a wide range of FASD neuropsychiatric disturbances. Deletion of OLs at a stage when they are just beginning to myelinate axons that interconnect neurons throughout the developing brain, could add to both the severity and complexity of long term neurobehavioral consequences of alcohol exposure during fetal life.

While the present findings pertain only to alcohol, and have primary significance in relation to FASD, the potential public health significance is extended by mounting evidence that many drugs (sedatives, anesthetics, anti-convulsants) used in obstetric and pediatric medicine mimic the apoptogenic action of alcohol in the developing brains of several animal species, including NHPs [20]. In addition, a single exposure to these agents can cause long-term neurobehavioral disturbances in several animal species [6, 16, 43, 44, 50, 51] similar to those caused by alcohol in rodents [22] or in humans [25]. Of potentially greater import is recent evidence reported by several independent research groups [28–34] suggesting that brief exposure of human infants to anesthesia is associated with increased risk for FASD-like neurobehavioral disturbances, including learning disabilities and AD/HD. It was found in several of these studies that increased risk for neurobehavioral disturbances was positively correlated with an increased number (or total duration) of exposures to anesthesia. These observations suggest an explanation why alcohol causes more conspicuous or readily detected neuropsychiatric disturbances than other apoptogenic agents; i.e., it is more common for the developing human brain to be exposed on multiple occasions during critical stages of development to alcohol than to other apoptogenic drugs. Also relevant to the multiple exposure hypothesis, we have found [3] that valproate is one of the most potent anticonvulsants in triggering neuroapoptosis in the infant rat brain, and there is evidence from recent multicenter studies [52, 53] that daily exposure of human fetuses to valproate during the third trimester of pregnancy is associated with a 9 point deficit in IQ, and that multiple third trimester exposures to other anticonvulsant drugs ((carbamazepine, lamotrigine, phenytoin) is associated with impaired verbal abilities at age 4.5 years.

Our findings document that OLs become vulnerable to alcohol-induced apoptosis at a maturational stage when they are just beginning to generate MBP in preparation to myelinate axons. At each developmental age evaluated (early, mid, late third trimester) there were abundant OLs at or beyond this maturational stage. This signifies that the developing NHP brain is vulnerable to alcohol-induced oligoapoptosis throughout the entire third trimester of gestation. Early research focusing on alcohol-induced apoptosis of neurons established that the window of vulnerability for that phenomenon coincided with the brain growth spurt period when billions of recently differentiated neurons are expanding their dendritic surfaces to accommodate incoming synaptic contacts. Thus, vulnerability of neurons appeared to be linked to the period of rapid synaptogenesis. Our present data suggest that vulnerability of OLs is linked to a different functional parameter - myelinogenesis. Therefore, it is possible, indeed likely, that the window of vulnerability for alcohol-induced oligoapoptosis will be found to have a different time schedule - one that corresponds to the progression of myelination events and to the maturational status of OLs that are responsible for the generation and maintenance of myelin. Our present findings indicate that onset of vulnerability to the apoptogenic action of alcohol corresponds not to the time when myelin sheaths are formed, but rather to the earlier time when promyelinating OLs are beginning to generate constituents of myelin, which in the human developing brain may occur 2–3 months earlier than myelination of axons is scheduled to begin [54].

The finding that alcohol and anesthetic drugs cause degeneration of OLs in the WM of the developing fetal primate brain raises a question how this toxic reaction might relate to another form of WM injury, periventricular leukomalacia (PVL), which occurs frequently in the brains of prematurely born human infants and results in cerebral palsy. The mechanism(s) underlying PVL are poorly understood, but hypoxia is a leading candidate, and the pathological reaction is thought to involve death of both neurons and OLs [55]. However, in animal research pertaining to hypoxic brain injury, the evidence suggests that death of both neurons and OLs is triggered by an excitotoxic (not apoptotic) mechanism, and OLs that are preferentially killed have been identified as OL progenitors that have not yet begun to generate constituents of myelin in their cytoplasm [37, 56, 57]. Another distinction is that OL degeneration in PVL is focally concentrated in periventricular regions, whereas OL apoptosis induced by alcohol or anesthetic drugs is diffusely distributed throughout WM pathways. Therefore, these are apparently two distinctly different forms of WM injury. However, if the third trimester brain were exposed to alcohol or anesthetic drugs and hypoxic conditions at the same time, the two forms of pathology, each involving death of both neurons and OLs, would undoubtedly act in concert to increase the severity of brain injury and long-term neurodevelopmental disability. Unfortunately, in pediatric medicine there are settings that are particularly conducive to this type of compound neurotoxicity. For example, in neonatal intensive care units throughout the world, premature infants are commonly exposed intermittently or continuously for days or weeks to anesthetic drugs for purposes of procedural sedation, and these same infants have a weak respiratory reflex, making them prone to spells of apnea (hypoxia).

An important limitation of our current findings is that they do not shed light on how alcohol exerts toxic effects on the brain during very early stages of development. However, Sulik and colleagues [58, 59] have demonstrated that in utero exposure of mice to alcohol at early stages of embryogenesis markedly increases the rate of apoptotic cell death among specific embryonic precursor cell populations that are destined to make important contributions to the glial and neuronal makeup of the brain. Collectively, our findings and those of Sulik et al., support a unifying and research-guiding hypothesis that, although apoptotic cell death is a natural phenomenon during development, pathological augmentation of apoptotic cell death is unnatural, and represents an occult mechanism by which alcohol, or other drugs that mimic alcohol’s apoptogenic action, can disrupt CNS development at any point from early embryogenesis to several years after birth.

In utero exposure of the fetal NHP brain to alcohol on a single occasion triggers widespread acute apoptotic death of neurons (previous study) and of OLs (present study) throughout WM regions of the developing brain. The rate of OL apoptosis in alcohol-exposed brains was 12.7 times higher than the natural OL apoptosis rate. OLs become sensitive to the apoptogenic action of alcohol when they are just beginning to generate constituents of myelin in their cytoplasm, and they remain vulnerable throughout later stages of myelination. There is growing evidence for a similar apoptotic response of both neurons and OLs following exposure of the developing brain to anesthetic and anticonvulsant drugs. Collectively, this body of evidence raises important questions regarding the role that neuro and oligo apoptosis may play in the human condition known as fetal alcohol spectrum disorder (FASD), and also poses a question whether other apoptogenic drugs, although long considered safe for pediatric/obstetric use, may have the potential to cause iatrogenic FASD-like developmental disability syndromes.