We studied the brains of fetuses 24 h after they had been subjected to 20 min of IUI, and compared them to control brains from sham operated E19 fetuses not subjected to IUI. Sur1 expression was examined using immunohistochemistry and immunoblot. Sur1 protein expression was prominent 24 h after IUI, with upregulation found in the choroid plexus, the ependymal lining of the lateral ventricles, and in periventricular tissues, including the internal capsule, the subventricular zone (SVZ), the hippocampus and corpus callosum (Figure 1A,B,D,E). By comparison, Sur1 expression was negligible in control tissues (Figure 1C). Upregulation following IUI was especially prominent in elongated structures consistent with microvessels (Figure 1B,E), as previously reported for models of focal ischemia [46]. Upregulation of Sur1 protein 24 h after a 20 min episode of IUI was confirmed with immunoblots of periventricular tissues; minimal Sur1 protein was detected in tissues from controls not subjected to IUI, compared to a strong single band 24 h after IUI (Figure 1F). Overall, the finding of Sur1 upregulation 24 h after 20 min of global ischemia was consistent with previous observations of prominent upregulation of Sur1 in models of focal ischemia, with both permanent and temporary occlusion [43].

In our previous report describing the tandem insult model [41], we showed that an IP injection of 50% glycerol (13 μL/g) was associated with a rise in intrathoracic pressure. We hypothesized that the rise in intrathoracic pressure would be associated with a rise in venous pressure, which would be transmitted to periventricular veins weakened by IUI. Here, we examined this hypothesis directly by measuring jugular venous pressure. An IP injection of 50% glycerol (13 μL/g IP) was associated with a significant increase in jugular venous pressure, which rose from −5 to +25 mm Hg over the course of 15 min (Figure 2A).

The concentration of glycerol used for the IP injection had a pronounced effect on the severity of the hemorrhages observed in pups that had been subjected to 20 min of IUI. Keeping the volume of injection constant (13 μL/g, IP), we compared the effect of 25%, 50% and 75% glycerol. In pups subjected to IUI, the use of 75% glycerol resulted in extensive hemorrhages that typically were fatal, whereas 50% glycerol yielded modest periventricular hemorrhages and an acceptable mortality rate (Figure 2B; details on mortality below). In the absence of IUI, 50% glycerol produced no hemorrhages [41], but with IUI, 50% glycerol produced hemorrhages in periventricular regions including choroid plexus, hippocampus, SVZ and corpus callosum (Figure 3A–E). All subsequent experiments were performed with the tandem insults of 20 min IUI and post-natal IP injection of 50% glycerol, 13 μL/g IP.

Blockade of Sur1 has been found to be protective after focal cerebral ischemia, both in humans and in animal models [42,43,45]. However, the effect of Sur1 blockade has not been examined in any model of global intrauterine ischemia. In previous experiments on focal ischemia in adult rats, in which drug was delivered directly to the affected organism, we found that an infusion of 200 ng/h of glibenclamide (in addition to a loading dose of 10 μg/kg) was associated with significant protection [43]. Here, we planned to administer glibenclamide to the pregnant dam, necessitating transplacental passage to the affected organism. In the rat, glibenclamide crosses the placental barrier, with the ratio of drug in fetal tissue to that in maternal blood being 0.535 ± 0.068 [47]. We compensated for the indirect delivery to the fetus by administering 400 ng/h (in addition to a loading dose of 10 μg/kg) to the dam. Infusions of this dose of glibenclamide had no significant effect on levels of serum glucose in either the dams at the time of giving birth, or on the pups at birth (Figure 4).

We compared mortality at 30 h in 3 groups: pups that had not been subjected to any insult or exposed to any treatment (naïve), tandem insult pups from untreated or vehicle-treated mothers (TI-CTR), and tandem insult pups from mothers that received glibenclamide after IUI (TI-GLIB). Mortality was nil in the naïve group versus 17% in pups from the TI-CTR group [41]. By comparison, mortality was 6% (5/79) in pups from the TI-GLIB group (Figure 5A).

We compared brain hemorrhages 24 h after the second insult in pups from the TI-CTR group versus the TI-GLIB group. H & E-stained coronal sections were scored by 3 blinded investigators to assess the severity and distribution of brain hemorrhages. Data were compiled from 16 and 14 rats in the TI-CTR and TI-GLIB groups, respectively. Compared to controls, glibenclamide treatment of the dam was associated with a dramatically reduced severity and incidence of brain hemorrhages 24 h after the second insult (Figure 5B–D). Although there were small hemorrhages near the lateral ventricles in pups from the TI-GLIB group (Figure 5E,F), the majority of the hippocampus and developing cortex typically was spared in these pups.

In adjacent sections of the same tissues used to evaluate hemorrhages, we also examined ED-1 (rat homologue of CD68), which identifies activated microglia, macrophages and monocytes. ED-1-positive cells in periventricular tissues and the internal capsule were examined. Visual inspection of representative sections showed that ED-1-positive cells were scarcer in tissues from pups in the TI-GLIB group (Figure 6A–D). This observation was confirmed by counts of ED-1-positive cells in both regions, which showed significant elevations in tissues from TI-CTR pups compared to naïve, and significantly fewer cells in tissues from TI-GLIB compared to TI-CTR (Figure 6E,F).

For the neurological functional tests (beam walk, accelerating Rotarod and the Morris water maze (MWM) paradigms), we compared the performances in three groups of rats: naïve, TI-CTR and TI-GLIB.

Rats were weighed at periodic intervals during development. When measured as % change, rats in the TI-CTR group lagged behind those in the naïve group (Figure 8A). By contrast, those in the TI-GLIB group exhibited better growth, which was indistinguishable from that in the naïve group.

At the time of sacrifice (P52), body and brain weights were significantly lower in rats from the TI-CTR group, compared to naïve, consistent with marked developmental delay associated with the tandem insults [41]. However, pups from TI-GLIB groups were similar to those in the naïve group; there was no significant difference in body weight between the 2 groups, and only a small, albeit significant, difference in brain weight (Figure 8B). These findings suggested that glibenclamide had largely preserved normal somatic and brain development.

Examination of H & E-stained sections of these brains revealed no hydrocephalus or other apparent abnormality in the TI-CTR group, compared to the naïve or TI-GLIB groups. More detailed immunohistochemical analyses will be needed to determine the underlying neuropathological correlate(s) of the functional abnormalities identified in the TI-CTR group.

The three principal findings of the present study are: (1) Sur1 is upregulated in periventricular tissues of rats after brief IUI, similar to our findings in premature infants [44]; (2) a low-dose glibenclamide infusion administered to the mother at the end of pregnancy protects pups subjected to the tandem insults of intrauterine ischemia and post-natal elevated venous pressure; (3) the dose of glibenclamide required for protection does not significantly lower serum glucose in either the mothers or the pups. The benefit of glibenclamide was realized early on, in the form of reduced brain hemorrhages secondary to elevated venous pressure and, because of this early protection, the benefit of treatment also was realized later in life, when as young adults, the rats were tested for performance on complex vestibulomotor and complex spatial learning tasks. Arguably, the most notable finding of the present study was that glibenclamide treatment of the mother significantly attenuated developmental delay of the pups and preserved their brain and body mass. Other antenatal treatments with significant efficacy in animal models of periventricular hemorrhage include the selective cyclooxygenase (Cox) 2 inhibitor, celecoxib [29] and glucocorticoids [28,53,54], although antenatal glucocorticoids may have adverse effects on brain function and fetal growth [55,56].

The effect of glibenclamide previously was examined in a rat pup model of neonatal ischemia/hypoxia induced by unilateral carotid ligation plus 2- or 2.5-h exposure (“moderate” and “severe” insults, respectively) to 8% O₂ in 10-day old pups (modified Rice-Vannucci model) [57]. Glibenclamide (10 μg/kg IP) was administered first shortly after the insult and a second time at 24 h. Brain edema, infarct volume and brain tissue loss were not changed by drug. However, in the moderate injury group, glibenclamide was associated with improvements of motor function in the foot-fault test and in postural reflex test at 3 weeks after ischemic/hypoxic brain injury. The smaller benefit of glibenclamide reported in the Zhou et al. [57] study, compared to the present study, may have been due to a more severe ischemic/hypoxic injury than what we studied here. Alternatively, the smaller benefit of glibenclamide may have been due to the amount and manner of drug administered, i.e., to the pharmacokinetics and selectivity of glibenclamide in the rat. The experience in our laboratory [43] and that of others [58] is that a continuous low-dose infusion is important for maintaining constant occupancy of Sur1, for avoiding the off-target Sur2 effect of vasoconstriction [59,60], and for obtaining the most beneficial effect, whereas episodic injections yield positive, albeit less favorable results [61,62]. Here, beginning immediately after IUI, fetuses were exposed continuously to a low dose of glibenclamide while in utero, at a concentration half of that in the mother’s serum (ratio of drug in fetal rat tissue to that in maternal blood, 0.535 ± 0.068 [47]). Although the mother continued to receive drug while nursing, it is doubtful that the newborn pups continued to receive glibenclamide, as the transfer of glibenclamide via breast milk is negligible [63,64].

Several animal models of periventricular hemorrhage in neonates have been reported [65]. Direct infusion of blood [16] or of collagenase to induce bleeding [66,67,68] reaffirms the toxic effects of blood products on immature periventricular cells [17]. Models intended to shed light on biological mechanisms responsible for hemorrhage generally incorporate a vascular insult of some sort that is superimposed on a predisposing condition such as prematurity [69,70,71] or IUI [41]. The most frequently utilized vascular insult is an IP injection of glycerol. Glycerol is an osmotic agent long said to cause hemorrhage due to a reduction in intracranial pressure (ICP) [70,72], even though the physiological grounds linking a reduction in ICP to the induction of hemorrhage are dubious. Recently, we showed that IP glycerol also causes an increase in intrathoracic pressure [41], and here we show that it causes an increase in jugular venous pressure (Figure 2), which accounts for our previous observation that hemorrhages originate from post-capillary venules [41]. Together, these findings provide a more plausible explanation for the induction of hemorrhage by IP glycerol. Also, these findings support the hypothesis that in humans, mechanical ventilation that may trigger hemorrhage [33,34] is due to elevated venous pressure. Intraperitoneal glycerol has been used to induce hemorrhages in premature rabbits at E29 (full term, 32 days) [70,71,73] and in rats at term that were subjected to IUI on E19 (full term, 21 days) [41]. The rabbit model with IP glycerol has several advantages [71], but suffers from the paucity of validated vestibulomotor and cognitive tests available for this species. Conversely, numerous well validated vestibulomotor and cognitive tests are available for rats, making the rat model with IP glycerol particularly attractive for the study of longer term functional outcomes.

A combination of mechanisms likely accounts for the fragility of periventricular veins during the several days before term that renders periventricular tissues susceptible to hemorrhage. As detailed in the excellent review by Ballabh et al. [20], selective vulnerability has been attributed to an immature basal lamina, incomplete glial support, and poor matrix support, which have been identified in both animals [21,22,23] and humans [24,25,26,27,28]. Immature vessels have prominent ongoing angiogenic activity, and two angiogenic inhibitors, the Cox-2 inhibitor, celecoxib, and the vascular endothelial growth factor receptor 2 inhibitor, ZD6474, reduce this vulnerability [29]. Immature periventricular vessels also are preferentially susceptible to injury by ischemia/hypoxia [30]. As we show here, ischemia/hypoxia results in upregulation of Sur1 in periventricular tissues, and selective blockade of Sur1 using glibenclamide reduces the vulnerability induced by ischemia/hypoxia. In focal ischemia involving both animal models and humans, glibenclamide similarly has been found to protect from ischemia-associated hemorrhage [42,74]. Whether Sur1 is involved here via the Sur1-Trpm4 channel [75] or via some other molecular mechanism has yet to be determined. Notably, exposure of brain endothelium to hypoxic conditions results in upregulation of the transcription factor, specificity protein (Sp) 1 via hypoxia inducible factor 1 [46], and Sp1 upregulation is known to drive the transcriptional expression of both Cox-2 [76] and Sur1 [46,77] in brain endothelium. This shared transcriptional program for Cox-2 and Sur1 in brain endothelium provides an interesting molecular link between the findings of Ballabh et al. [29] regarding Cox-2 inhibition and those reported here regarding Sur1 inhibition.

Glibenclamide (United States (US) adopted name, glyburide) is a safe drug approved by government health regulatory agencies worldwide that has been used for more than 3 decades to treat adult onset diabetes mellitus. In the last decade, there has been a growing interest in using glibenclamide to treat gestational diabetes mellitus (GDM) [78,79]. Since 2000, several studies have reported an 80%–85% success rate with the use of glibenclamide in the management of GDM. The use of glibenclamide during pregnancy has given rise to 2 concerns: potential teratogenicity, and fetal hypoglycemia. Some authors have cautioned against possible increases in the risk of preeclampsia, macrosomia, admission to a neonatal intensive care unit and the need for phototherapy with glibenclamide [80,81]. However, animal and human studies assessing the teratogenic effects of oral antidiabetic agents including glibenclamide have yielded equivocal or even negative findings [82], reflecting the fact that the risk of major malformations in infants of mothers with GDM is primarily related to maternal glycemic control, not the specific antidiabetic therapy that is used [83,84]. Some authors have been quite sanguine about the use of glibenclamide for the treatment of GDM, expressing the opinion that “unless future studies refute current data regarding the efficacy and safety of glyburide, we believe that, owing to its ease of administration, convenience and low cost, glyburide will become the first line of medical treatment in patients with GDM within the next few years” [78].

The second issue, that of fetal hypoglycemia, is of concern when treating GDM, especially if a large dose of drug is required. At equivalent doses, glibenclamide plasma concentrations are approximately 50% lower in pregnant women than in nonpregnant subjects, potentially necessitating dose escalation [85]. Notably, at the time of delivery, the average umbilical cord/maternal plasma glibenclamide concentration ratio is 0.7 ± 0.4 [85], consistent with some degree of fetal protection from excessive doses. However, the findings presented here as well as in models of focal ischemia [43] indicate that a low, non-hypoglycemogenic dose suffices for protection. Thus, when considering the use of glibenclamide as prophylaxis against the ravages of hemorrhagic EP, it seems unlikely that fetal hypoglycemia would pose a serious safety concern.