Subjects were male Wistar rats born to time mated dams (Charles River Laboratories, Wilmington, MA, USA). Dams were shipped to the University of Connecticut on embryonic day 4 (E4), and were housed in tubs in an approved animal facility. Upon birth (P1), litters were culled to 10 pups (8 males and 2 females per litter). Only male pups were used in the current study, based on established sex differences between males and females following HI injury (with males showing a more robust pattern of deficits than females) [18]. Two female pups were retained in each liter to maintain normal maternal behavior. Subjects were weaned on P21 and pair housed with like-treated littermates until P55, when they were single housed for the duration of the study. All subjects were housed in a 12-h light/dark cycle with food and water available ad libitum.

On P7, each pup was randomly assigned one of four treatment groups: HI vehicle (sterile saline), HI caffeine, sham vehicle, or sham caffeine. All treatment conditions were balanced across litters. For the HI procedure, pups were anesthetized with isoflurane (2.5%) and a midline incision was made longitudinally on the neck on the ventral side. The right common carotid artery was located and separated from surrounding tissue, cauterized, and the incision was sutured. Pups were marked with a footpad injection to indicate treatment group at weaning. Pups were then allowed to recover from the anesthesia under a warming lamp, and were returned to their dam for two hours to feed. Animals who were assigned to the sham condition were placed under anesthesia, but only had a midline incision with no manipulation of the carotid artery. Sham treated pups were allowed to recover and were also returned to the dam. Two hours after all pups had received the surgery, HI subjects were placed in an airtight oxygen container flowing 8% humidified oxygen (balanced with nitrogen) for a period of 120 min. Sham animals were placed in an open container with room air exposure for the same duration.

Immediately following the 120 min chamber exposure, animals received an intraperitoneal (i.p.) injection of either caffeine (10 mg/kg) or vehicle (sterile saline in comparable volume), and all pups were then returned to their dam. On P21, subjects were weaned and paired housed with like-treated littermates for later behavioral testing

Upon the completion of behavioral testing, animals were anesthetized with Ketamine (100 mg/kg) and Xylazine (15 mg/kg) and transcardially perfused with 0.9% saline and 10% buffered formalin phosphate. Brains were removed and placed in the 10% buffered formalin phosphate. Prior to slicing, brains were placed in a 30% sucrose solution for cryoprotection for 24 h, and were then sliced in a coronal plane at 60 μm on a cryostat. Every third section was mounted on a slide and stained using cresyl violet. With the use of a Zeiss Imagine A.2 microscope and Micro Bright Field (MBF) Stereo Investigator computer software (Willinston, VT, USA) cortical volume was measured and calculated using Cavalier’s estimator. Due to complications with the slicing and/or staining procedures, some tissue sections were lost, so that there were not enough sections to create an accurate representation of total cortical volume in nine rat brains (5 HI saline and 4 HI caffeine). All measurements were performed blind to treatment group.

For all measures we employed planned paired comparisons among HI saline, HI caffeine, sham saline, and sham caffeine groups. Specific comparisons based on a priori hypotheses were made between: HI saline vs. sham (to confirm replication of HI effects); HI saline vs. HI caffeine (to assess any beneficial effects of caffeine); and HI caffeine vs. sham (to determine whether caffeine-treated subjects were equivalent to shams). In addition, we compared caffeine versus saline treated shams to ascertain any effects of caffeine on normal subjects, and/or to pool shams.

Repeated Measure Multiple Analysis of Variance (ANOVA) was used to analyze MWM scores. Unless stated otherwise, all analyses were two-tailed. In comparing the HI saline to sham groups on behavioral and volumetric measurement outcomes, our lab has previously shown significant deleterious effects, and thus one-tailed analyses were used for these specific comparisons [19,20,21]. Variables analyzed and presented in the results section for the MWM task include Treatment (3 levels; sham [pooled sham saline and sham caffeine], HI saline, and HI caffeine) and Day (5 levels). For the MWM task, total latency (in seconds) to reach the platform for the four trials each day was the dependant variable. A “Learning Index (LI)” score was calculated to assess possible learning differences between groups (i.e., an index of slope for each animal). The equation LI = (Day1 latency - Day2) + (Day 2 - Day 3) + (Day 3 - Day 4) + (Day 4 - Day 5) was applied to each animal, and scores were analyzed using a one-way ANOVA. Finally, a non-parametric Kruskal-Wallis ANOVA was used to analyze right cortical volumes (i.e., the side of injury) by Treatment, using the same paired comparisons delineated above. SPSS 15.0 with a criterion of alpha 0.05 was used for all analyses.

A Univariate ANOVA revealed no significant effect of Treatment groups [F(2,35) = 0.87, p > 0.05] on water escape latencies indicating comparability in each group’s ability to see the platform, and to swim to it.

An initial 5 (Day) x 3 (Treatment) repeated measures ANOVA revealed a significant between groups effect of Treatment for the three groups: HI saline, HI caffeine and sham [F(2,33) = 4.697, p < 0.05]. There was also a significant within subjects effect of Day [F(4,144) = 6.445, p < 0.01], reflecting decreasing latencies with learning. Additional ANOVAs were used to compare specific paired groups. We did not see a Treatment x Day interaction.

A 5 (Day) x 2 (Treatment) repeated measures ANOVA revealed a significant Treatment difference between HI saline and sham treated animals [F(1,21) = 3.274, p < 0.05, one-tailed], with the HI saline group taking longer to reach the platform than shams (Figure 1a). A further 5 (Day) x 2 (Treatment) repeated measures ANOVA also revealed a significant Treatment difference between HI saline and HI caffeine treated animals [F(1,28) = 8.477, p < 0.01], with the HI saline group taking longer to reach the platform compared to the HI caffeine group (Figure 1b). Finally a 5 (Day) x 2 (Treatment) repeated measures ANOVA revealed no significant differences between HI caffeine and sham animals [F(1,21) = 0.363, p > 0.05] (Figure 1c). The within variable Day showed a trend for the HI saline and sham comparison [F(4,84) = 1.599, p = 0.09, one tailed], but was significant for the HI saline vs. HI caffeine comparison [F(4,112) = 2.7, p < 0.05] as well as for the HI caffeine and sham comparison [F(4,84) = 2.824, p < 0.05]. As above, these Day effects reflect learning.

Finally, to assess putative learning effects, a “Learning Index” was calculated for each animal. The results of a one-way ANOVA revealed no significant difference between groups on this measure [F(2,35) = 0.083, p > 0.05], however, scores were in the expected direction (shams > HI saline).

In addition we performed a paired samples t-test between Day 1 versus Day 5 within each Treatment group. For the HI saline group this effect was non-significant [t(14) = 1.22, p > 0.05, one tailed], indicating there was no learning effect over Days. In the HI caffeine group, a paired samples t-test revealed a significant difference between Day 1 and Day 5 [t(14) = 1.88, p < 0.05, one-tailed]. Additionally, in the sham group, a paired samples t-test revealed a significant difference for learning as well, [t(7)=1.513, p < 0.05, one tailed].

The cognitive deficits as well as neuropathological changes evident following neonatal HI injury are thought to be mediated in part by elevated adenosine levels following HI. Typically in neonatal rodents, interstitial adenosine levels (basal) are around 50 nM, but are believed to increase to upwards of 1000 nM during ischemic events [24,34]. Additionally, adenosine deaminase (ADA), a key enzyme to deaminate adenosine and thus maintain levels of adenosine, is also increased following neonatal HI in the cortex [35] and hippocampus [36]. This dramatic increase in adenosine may promote accelerated expression of apoptotic activity, leading to a rise in free radical and caspase formation [24]. During typical development in rodents, adenosine A1R increases in density between P9 and P15. Moreover, although receptors are found throughout the brain, highest levels are specifically seen in the cortex and hippocampus by adulthood [23]. Studies have shown that rodents treated with an A1 receptor agonist from P9 to P14 exhibit increased ventricular volume, reduced periventricular white matter volume, and a reduced number of neurons in the cortex and hippocampus [37] as well as decreased expression of myelin [24]. Theses changes are thought to reflect the fact that excess adenosine binding to the A1 receptor leads to the activation of hypoxia-inducing factor (HIF)-α, which in turn promotes the apoptotic cascade leading to cell death [24,38]. Additionally, adenosine accumulation following hypoxia is thought to be responsible for electrical suppression in the brain. For example, the release of adenosine was directly related to the depression of excitatory synaptic transmissions in hippocampal slices of P10–24 rats [39,40] and increased adenosine levels following hypoxia as measured in vivo in adult animals led to suppressed electrically evoked synaptic transmission [41]. Again these latter deleterious effects are thought to be mediated via activation of A1Rs. Consistent with the negative effects of A1 receptor activation in the neonatal brain, blockade of the A1 receptor has been found to exert protective effects in HI injured mice [42]. For example, knock-out mice deficient in the A1 receptor and reared in a hypoxic chamber from P3 to P14 showed decreased ventricular volume and increased white matter density as compared to wild-type HI mice that did express the A1 receptor [39].

Although the effects of adenosine expression and resulting brain damage following early HI injury appears to be primarily mediated through the A1 receptor, activation of the A2A receptor has also been implicated. For example, knock-out mice for the A2A (A2A -/-) receptor that had HI injury induced on P7 were evaluated three weeks to three months later, and brain injury was found to be exacerbated in the A2A -/- animals as compared to wild types. Furthermore, A2A -/- animals also showed impairments on rota-rod and beam walking tasks [43]. Since A2A receptors are located on the capillaries, an over-activation resulting from adenosine increases after HI injury could contribute to intraventricular hemorrhage via rupture of these vessels (although this has not been empirically demonstrated) [24]. In sum, A2A receptor activation is not well understood in development but may play a role in HI pathology.

In the adult brain, the effects of caffeine are primarily mediated through antagonism of A1Rs and A2ARs [25,44], where antagonism of A1 affects transmitter release and neuronal firing, and antagonism of A2As seems to affect dopaminergic transmission [44]. Additionally, caffeine has been shown to attenuate long-term potentiation in hippocampal slices [45]. The neuroprotective effects of caffeine following neonatal HI, however, are primarily via antagonism of the A1 receptor [34], consistent with evidence that AR antagonism appears to be mediated differently in the adult brain as compared to the embryonic fetal and neonatal brain [29]. Importantly, in the neonatal brain, A1R activation offers no protection in ischemia-induced damage but instead is associated with brain injury [23]. Studies have shown deleterious effects of caffeine (not seen in adults) during normal fetal brain development, including a wide array of behavior, neurochemical, and electrophysiologic changes [46,47,48,49,50]. However, serum caffeine levels in caffeine treated premature infants were found to be correlated with a change in pro-inflammatory and anti-inflammatory cytokine levels in the peripheral blood, with a greater concentration of anti-inflammatory molecules correlated with higher levels of caffeine (between 10 and 20 μg/mL). Additionally, reductions in interleukin (IL)-6 and tumor necrosis factor (TNF)-α (both pro-inflammatory) and increases in IL-10 (anti-inflammatory) were seen in these subjects [51]. Finally, activation of A2ARs and A2BRs (located on capillaries [25]) could induce capillary leakage, which might contribute to hemorrhages and HI. Thus antagonism of A2ARs and A2BRs may explain in part why caffeine treated infants show lower incidence of intra-ventricular hemorrhage (IVH) [24]. Caffeine treatment also down-regulates TNF-α and the release of cytokines in response to lipopolysaccharides (an endotoxin that elicits a large inflammatory response from the immune system; LPS) in term injured infants [52]. There are likely to be additional cellular factors involved in cognitive protection based on the work of others. For example, in cultured developing cortical neurons, caffeine enhanced CREB dependant gene expression and mediated the activity-dependant BDNF and TrkB expression [53]. Neonatal mice reared in hypoxia treated with caffeine also showed more normally arranged axon orientation and increased the proportion of immature oligodendrocytes [33]. Premature baboons treated with caffeine also showed improved myelination [34]. Similarly, a study investigating the effects of induced early life convulsions in P7 rodents, injured animals showed deficits in selective memory in adulthood and a loss of presynaptic glutamate terminals. Caffeine prevented these memory deficits and prevented the loss of nerve terminal markers in the hippocampus [54]. Although loss of gross brain volume is a significant complication following neonatal HI injury, it could be that the protective effects of caffeine allow in part for cellular reorganization and plasticity effects contributing to the rescue of behaviors. Additional studies are needed to confirm this hypothesis and to characterize how different cell types might reorganize themselves in order to prevent brain injury and resulting behavioral deficits following early HI injury.

As stated above, caffeine was originally used in premature infants to improve respiratory drive and treat apnea of prematurity by decreasing the threshold for sensitivity of hypercapnia, and increasing contractility of the diaphragm [55]. Caffeine has also been shown to lead to a decreased rate of bronchopulmonary dysplasia in treated compared to non-treated premature children [27] as well as improved cognition at 18 months. Due to improved and more consistent adequate blood oxygenation in these treated infants, it is not possible to conclude definitely that the neuroprotective effects of caffeine in this population are mediated exclusively by central brain related changes versus the elimination of intermittent hypoxia. Animal models expressing the A1 receptor show suppression of respiratory drive with adenosine, and stimulation of respiratory drive with caffeine, while in A1R knock-out mice for A1Rs, adenosine showed no suppressive effects on respiratory drive, and no stimulatory effects with caffeine [24]. Blockade of A2A receptors are also implicated in the neuroprotective role of caffeine. Specifically, in the medulla of the brainstem (an area implicated in cardiac function and respiratory drive), about half of the GABAergic neurons (measured by in situ hybridization and retrograde tracing) express mRNA for A2A receptors in the developing mouse brain. This suggests that A2A receptors might play an important role in respiration early in life as well [56]. While brain injury in premature infants are typically associated with repetitive respiratory-induced mild to moderate drops in brain oxygenation over a period of weeks, hypoxia brain injury in term infant are generally perinatal acute events, much like the acute injury used in our rodent model. Thus we hypothesize the neuroprotective effects of caffeine in our model are due to the direct effect of caffeine via the interaction with the adenosine receptor and not due to changes in breathing and long term improved oxygenation.