Acute Anemia Elicits Cognitive Dysfunction and Evidence of Cerebral Cellular Hypoxia in Older Rats with Systemic Hypertension

• ❖ Postoperative cognitive dysfunction occurs more commonly in the aged. Anemia has been associated with postoperative cognitive impairment.

• ❖ In a laboratory study of young versus  aged spontaneously hypertensive rats, acute isovolemic anemia produced age-dependent impairment of learning and memory as well as molecular signs of cellular hypoxia.

• ❖ These animal data suggest a possible link among anemia, cellular hypoxia, and postoperative cognitive dysfunction with aging.

POSTOPERATIVE cognitive dysfunction (POCD) primarily affects the working memory domains of attention, cognitive speed, and executive function, as well as hippocampal based memory acquisition or learning.1POCD may occur in 40-60% of patients after cardiac surgery and somewhat less frequently after major vascular and orthopedic surgery.1,2Aging3appears to be the most important risk factor for its occurrence. Patients presenting for cardiac and noncardiac surgery often suffer from longstanding hypertension and preexisting hypertension-related cerebrovascular disease, also an important risk factor for cognitive impairment.4 

Mechanisms potentially contributing to POCD remain poorly understood. Research has focused largely on processes related to the conduct of cardiopulmonary bypass in cardiac surgery. The “Off-Pump” approach for coronary artery bypass grafting, designed to avoid the presumed cerebral side effects associated with extracorporeal circulation, has failed to diminish the incidence of POCD.5In light of this, and the parallel occurrence of POCD after major noncardiac surgery, it is plausible that POCD after both major cardiac and noncardiac surgery shares common origins, at least in part unrelated to anesthetic exposure. Finally, although POCD occurs with high frequency postoperatively, it resolves in most cases over several weeks to months, and there is little evidence of permanent disability.6Is there then a contributing process that shares such a time course?

We propose that a role for anemic hypoxia be considered in the genesis of POCD. Acute anemia has been found by Weiskopf et al.  7to elicit cognitive impairment even in healthy nonsurgical subjects. Major surgery in all arenas frequently results in acute and severe anemia. Severe anemia during cardiopulmonary bypass has been shown to lead to developmental delay in infants after cardiac surgery.8Kulier et al.  9and Karkouti et al.  10have demonstrated a correlation between adverse cerebral outcomes and lower preoperative hemoglobin levels. Mathew et al.  11randomized 108 elderly subjects to hematocrits of more than 27% or less than 18% during cardiopulmonary bypass, with subjects undergoing more profound hemodilution experiencing a greater degree of cognitive impairment. Finally, anemia is recoverable over a time course similar to recovery from POCD.12 

Mild to moderate hypoxia associated with high-altitude exposure and intermittent hypoxia associated with sleep apnea have also been linked to working memory and learning impairment.13–15It is not surprising then that mild to moderate hypoxia has also been found to impair neuronal protein synthesis and synaptic plasticity,16key processes requisite for learning. Exposure to hypoxia triggers a cascade of events focused upon cell survival. The hypoxia-inducible factor (HIF) transcription system17is the “master” regulator of this hypoxic response.18The prolyl hydroxylases family of proteins (prolyl hydroxylases-1, -2, and -3), along with factor-inhibiting HIF, responsible for degrading ubiquitously present HIF during normoxia, are themselves inhibited during hypoxia, allowing levels of HIF to rise.19This results in the transcription of hundreds of hypoxia-responsive genes20involved in regulating a host of activities focused on cell survival, including glucose transport (glucose transporter-1, GLUT-1), glycolysis (phosphoglycerate kinase-1, PGK-1), oxygen transport (erythropoiesis-erythropoeitin), and angiogenesis (vascular endothelial growth factor, VGEF).21Effective hypoxia sensing and adaptation to hypoxia are critical for cellular function in all organs, including the brain.22HIF and related gene products are exquisitely sensitive biomarkers of cellular hypoxia. Critical to the management of our aging population, aging cells appear to have a diminished, putatively cytoprotective, hypoxic response.23 

In this study, we test the hypothesis that acute isovolemic anemia leads to cognitive impairment with aging. We also test the hypothesis that aging with hypertension predisposes to cerebral tissue and cellular hypoxia with anemia. These experiments were conducted in a rodent model of chronic hypertension and cerebrovascular disease, the spontaneously hypertensive rat (SHR). We chose the SHR rat model in which to test these hypotheses because of the many similarities between this strain and humans with hypertensive cerebrovascular disease who present for cardiovascular, vascular, and major noncardiovascular surgery.

Methodologies are described below for the three main experiments conducted: 1) memory testing with anemia, 2) cerebral oxygenation with anemia, and 3) molecular markers of hypoxia with anemia.

Approval from the Institutional Animal Care and Use Committee at the University of Pennsylvania, Philadelphia, Pennsylvania, was obtained before the conduct of this study.

Experiments were conducted in a rodent model of chronic hypertension and cerebrovascular disease, the spontaneously hypertensive rat. In this model, hypertension starts developing by 4 months of age with the appearance of hypertension-related changes in the brain microvasculature by 6 months of age.24Mortality in the SHR is approximately 50% at 18 months of age.25SHR were obtained from Harlan Sprague-Dawley Inc. (Indianapolis, IN). A more detailed description of cerebrovascular changes in this model can be found in Supplemental Digital Content 1, which provides a more detailed description of methods for all experiments than can be provided in the main manuscript, https://links.lww.com/ALN/A613. Three age groups were studied: young = 3 months (Y), mature = 9-12 months (M), and aged = 13-18 months (A). Two treatment subgroups were included. The anemic (An) subgroups, YAn, MAn, and AAn, underwent the true hemodilution protocol, and a sham (Sh) protocol was conducted in the YSh, MSh, and ASh subgroups.

To conduct the hemodilution and sham hemodilution protocols for the memory testing and molecular markers of hypoxia experiments, rats were anesthetized in a plastic box containing 4% isoflurane in 100% oxygen. A nose cone with spontaneous ventilation was used for the maintenance of anesthesia utilizing 1.5-2.0% isoflurane in 21% oxygen. Animals were recovered for subsequent testing.

For the cerebral oxygenation studies, rats were initially anesthetized in a plastic box containing 4% isoflurane and 100% oxygen, tracheotomized with a 14-gauge catheter, and ventilation controlled. Anesthesia was maintained with 1.5-2% isoflurane in 21% oxygen and α-chloralose (80 mg/kg) injected into the axillary fat pad before craniotomy. Isoflurane was discontinued after craniotomy, and a period of 45 min was allowed to pass to allow for the elimination of isoflurane and for stabilization of the brain tissue before data collection. α-Chloralose was utilized in this experiment because it has minimal effects on cerebrovascular reactivity as compared with its volatile anesthetic counterparts. α-Chloralose has been used extensively in research for brain functional activation studies because responses appear to mimic the awake state.26,27Anesthetic management did not include muscle relaxants.

Under sterile conditions, an incision over the right groin was made. The right femoral vein was cannulated with polyethylene-10 tubing that was fixed in place with suture for the hemodilution process and for blood sampling for hemoglobin and hematocrit. The isovolemic hemodilution process involved three weight-based withdrawals of 2-5 ml (7 ml/kg) venous blood. For the YAn, MAn, and AAn groups, the blood removed was immediately replaced with warmed normal saline at 37°C, in a ratio of 1: 3 (blood:normal saline). For the sham groups (YSh, MSh, and ASh), the withdrawn blood was immediately reinjected. The hemodilution protocol was designed to decrease the hemoglobin concentration in a step-wise fashion from a baseline of ≈12 g/dl to a final level of ≈5 g/dl after the third hemodilution. Hematocrit and hemoglobin were measured after each hemodilution utilizing a Stat Profile pHOx system (Nova Biomedical, Waltham, MA). The sham hemodilution groups allowed us to control for anesthetic exposure, hypotension related to experimental hemorrhage, surgical stress, and wound-related disability.

For animals in the memory testing and molecular markers of hypoxia experiments, after hemodilution was completed, the groin incision was closed with absorbable suture and the anesthetic discontinued. Animals were recovered under warming lights in cages in the laboratory. When animals demonstrated normal ambulation they were returned to the animal facility.

Postoperative cognitive dysfunction typically affects working memory and learning, while sparing long-term or reference memory. Therefore, the impact of aging and isovolemic anemia upon visuospatial reference memory, as well as working memory and learning, was tested, using Morris water maze paradigms designed to differentiate among these specific memory categories. In the Morris water maze, animals are challenged to swim in a circular tank, partially filled with opaque water, using primarily visual cues located in the testing room, to find a submerged platform upon which they can escape. Final group sizes then were: 1) YSh, n = 12; 2) YAn, n = 13; 3) ASh, n = 14; and 4) AAn group, n = 15.

Prehemodilution and posthemodilution visuospatial reference memory was tested through the assessment of performance on a fixed position, submerged platform paradigm, after a period of prehemodilution training on this same platform location. Posthemodilution visuospatial working memory and learning were tested using a criteria-driven, moving platform paradigm described by Chen et al.  28Success in this latter paradigm demands utilization of attention, conceptual, and procedural working memory to achieve learning upon acutely presented and changing spatial information.

Prehemodilution flagged (cued) training, using a fixed platform location, occurred during days 1-5 and was followed by noncued fixed platform training on days 8-11. Prehemodilution reference memory testing then occurred on day 12 and included measurement of escape latency, swim speed, and distance traveled while searching for a fixed, submerged platform. Reference memory was further tested using a single “probe” trial and was also performed on day 12, where time, percentage of time spent in each quadrant, and platform crossings were recorded while the animal searched for a submerged platform that had been removed from its previously located and trained upon quadrant. Surgery and hemodilution occurred on day 15. Posthemodilution reference memory performance was tested on day 17, using the same performance measures listed above and using the identical platform location used for prehemodilution training and testing. Posthemodilution visuospatial working memory and learning were then tested in a longitudinal fashion on days 18-26, using the criteria-driven, working memory and learning-sensitive moving platform paradigm (see fig. 1).

1. Escape latency(s): time between leaving the starting point and climbing onto the escape platform.

2. Probe testing.

   • Distance (cm): distance traveled from starting point to platform escape.

   • Platform crossings (#): number of crossings over the escape platform.

   • Quadrant time percent (%): the percentage of time spent in the quadrant of the escape platform.

Trials-to-criteria (#): number of trials required to reach working memory success criteria.

Anemia has been reported to impair cerebral tissue oxygenation and could contribute to cognitive impairment through this mechanism. We therefore measured the effect of hemodilution upon brain tissue oxygenation and blood flow. The parietal association cortex was studied because of its documented role in visuospatial memory acquisition29and ease of access via  craniotomy. Experimental groups were as follows: YAn, n = 14; AAn, n = 16. A 450-μm-diameter Optronix multiparameter probe (Oxford Optronix, Oxford, United Kingdom) was placed 2-4 mm into the parietal association cerebral cortex, allowing continuous measurement of laser Doppler flow and brain tissue partial pressure of oxygen (P^Bro²). Mean arterial pressure (MAP) and rectal temperature were also measured continuously. Arterial blood gas and hemoglobin measurements were made after each hemodilution step and ventilation adjusted to maintain arterial partial pressure of carbon dioxide (P^artco²), 38-42 mmHg.

N = 10 animals in each of three age groups (Y, M, A) underwent isovolemic (An) or sham (Sh) hemodilution protocols. At 24 or 48 h after hemodilution, the animals were sacrificed and their brains retrieved for RNA extraction. Primers against 18S RNA (Applied Biosystems, Inc., Foster City, CA) were used as endogenous controls for ΔC^Tcalculations. Primer sets were generated against rat HIF-1α, HIF-2α, carbonic anhydrase-9, erythropoietin, erythropoietin receptor, GLUT-1, PGK-1, VEGF, endothelial nitric-oxide synthase (NOS), inducible NOS, and neuronal NOS.

N = 3-4 animals in each of three age groups (Y, M, A) underwent isovolemic (An) or sham (Sh) hemodilution protocols. Animals were sacrificed at 24 or 48 h and cortical and hippocampal tissue obtained. Samples from the 3-4 animals in each group were pooled before immunoblot analysis for HIF-1α, VEGF, GLUT-1, and neuronal NOS. Samples for Western blotting were repeated several times to ensure consistency; the graphs presented are representative of those results.

For memory testing, a repeated measures analysis of variance (ANOVA) approach was used to test for intragroup differences over time. Intergroup comparisons utilized the Tukey-Kramer HSD (Tukey-HSD) methodology for multiple comparisons. For the analysis of the role of group (aged vs.  young) and treatment subgroup (anemia vs.  sham), a linear mixed-effects model was implemented.

For cerebral oxygenation and blood flow data, intragroup means were compared via  repeated-measures ANOVA. Comparison of intergroup means was conducted using the paired Student t  test.

For molecular markers of hypoxia, intragroup means were tested with a two-tailed Student t  test. ANOVA regression analysis was used to test the dependence of each gene level upon: 1) An versus  Sh treatment, 2) age group, 3) the interaction between the treatment and the age group (An × Age Grp), 4) the sampling time (24 vs.  48 h), and 5) the site of sampling (hippocampus vs.  cortex).

For all analyses, significance is cited for P < 0.05. All analyses were performed with the JMP Statistical Discovery Software (SAS, Inc., Cary, NC). Means ± SD are reported in the results section.

Additional details on all experimental methods can be found in Supplemental Digital Content 1, https://links.lww.com/ALN/A613.

The sham hemodilution group hemoglobin levels averaged 12.7 ± 0.6 g/dl, and the experimental hemodilution group hemoglobin levels averaged 4.7 ± 0.4 g/dl after treatment. Prehemodilution, posthemodilution, and final recovery hemodilution hemoglobin levels by group can be found in tabular format under Supplemental Digital Content 2, https://links.lww.com/ALN/A614. Swim speeds can be found in tabular format under Supplemental Digital Content 3, https://links.lww.com/ALN/A615.

The mean number of trials-to-criteria by subgroup and platform location are graphed in figure 2and summarized in tabular format in Supplemental Digital Content 4, https://links.lww.com/ALN/A616. The figure in Supplemental Digital Content 5 demonstrates percentage failures per day for each subgroup, https://links.lww.com/ALN/A617. The mean number of trials-to-criteria for platform locations 2-10, figure 3, were as follows: ASh (5.8 ± 2.3), AAn (8.0 ± 4.0), YSh (4.6 ± 1.7), and YAn (4.9 ± 1.7). We tested the dependence of trials-to-criteria upon age group (young vs.  aged) and treatment subgroup (An vs.  Sh). Working memory and learning were found to be impaired by aging (age group, P < 0.001), by hemodilution (An vs.  Sh, P < 0.001), and by the age group × treatment interaction (AAn vs.  YAn, P < 0.01). Further statistical details describing group differences are summarized in tabular format under Supplemental Digital Content 6 and 7, https://links.lww.com/ALN/A618and https://links.lww.com/ALN/A619, respectively.

Following swim speed can account for the impact of anemia on fatigue and lethargy, independent of potential cognitive effects. Average posthemodilution swim speed (cm/s) slowed markedly for ASh and AAn, from an average of 36.4 ± 7.4 cm/s to 29.0 ± 7.6 cm/s (P < 0.001), now slower for the aged animals than for the younger groups, 32.9 cm/s (P < 0.001), the opposite of what had been seen in the prehemodilution phase (Supplemental Digital Content 3, https://links.lww.com/ALN/A615). Although anemia may have resulted in greater fatigue in older animals, there remained no important difference in swim speed between the ASh and AAn subgroups (P = 0.75), while there was a small difference between the YSh and YAn subgroups (P < 0.001) after the sham or anemic treatments. Escape latency, reported as the average of four trials to the hidden platform for P-1 on the first day of testing posthemodilution (Day 16), is listed in Supplemental Digital Content 8, where reference memory variables are summarized in tabular format for prehemodilution and posthemodilution phases, https://links.lww.com/ALN/A620. All groups demonstrated roughly parallel increases in latency from prehemodilution to posthemodilution phases. Within-group analysis (YSh vs.  YAn, or ASh vs.  AAn) of latency measures demonstrated no significant differences (Supplemental Digital Content 8, https://links.lww.com/ALN/A620). Distance traveled increased in all subgroups, and remained increased in the aged animals relative to younger cohorts. Posthemodilution probe testing revealed no remarkable decrement in platform crossings, quadrant time, or quadrant time percent from the prehemodilution phase.

Chronic hypertension and atheromatous changes in the vasculature accelerate with aging, which may render tissue especially vulnerable to anemic hypoxia. We therefore tested the impact of isovolemic anemia upon local brain partial pressure of oxygen and laser Doppler flow in young (YAn) and aged (AAn) male SHR.

A summary of the results of bivariate analyses for all physiologic measurements is shown in table 1. Surprisingly, steady-state tissue brain partial pressures of oxygen values did not change appreciably from baseline values (P = 0.49).

We did not attempt to control MAP during this experiment, letting it “float” under the conditions of the experiment. MAP was only significantly different between groups, however, after the third hemodilution, falling by 17 ± 26% in the AAn group and 41 ± 13% in the YAn animals (P = 0.005). A positive correlation was apparent between the change in brain partial pressure of oxygen that occurred from baseline to completion of the hemodilution protocol and percentage change in MAP for the AAn animals (P = 0.017), although none appeared to exist for the younger animals (P = 0.57); see figures in Supplemental Digital Content 9, https://links.lww.com/ALN/A621. The slope of the regression line climbed more steeply for the AAn animals by ≈37% relative to YAn animals. This difference persisted (28%) even after the young outlier (red circle) was withdrawn from the analysis.

Laser Doppler flow measurements increased from baseline by 31 ± 41% for the AAn animals and increased to a somewhat lesser degree, 17 ± 28%, for YAn animals, from baseline to final hemodilution, (P = 0.028). Stepwise details can be found in table 1.

To test the cellular sensing of and responses to anemic hypoxia, we measured levels of the HIF-1α protein and examined the gene expression profiles for downstream HIF-regulated genes, including HIF-1α, HIF-2α, carbonic anhydrase-9, erythropoietin, erythropoietin receptor, GLUT-1, PGK-1, VEGF, endothelial NOS, and inducible NOS after sham or anemic exposures. The hypoxia-sensitive neuronal NOS gene was also measured. Finally, we also measured translation of HIF-1α, VEGF, GLUT-1, and neuronal NOS. Hemoglobin levels for animal groups used in messenger ribonucleic acid (mRNA) and protein analyses are summarized in tabular format in Supplemental Digital Content 10, https://links.lww.com/ALN/A622.

HIF-1α protein levels in the cortex and hippocampus were increased in response to anemia in all age groups, but also appeared to be increased in relatively greater magnitude in aged anemic animals relative to younger cohorts, especially at 48 h in both the hippocampus and cortex; see figure 4a. HIF-1α levels were increased roughly 5-fold in the hippocampus and doubled in the cortex at 48 h. This compared with a 3-fold elevation in the hippocampus, and a decreased level in the cortex, in younger cohorts.

A subset of HIF-regulated (VEGF and GLUT-1) or hypoxia-regulated proteins (neuronal NOS) were also examined; these results can be found in figure 4, b–d. VEGF and GLUT-1 protein expressions appeared to be increased, and neuronal NOS was depressed in the ASh groups relative to YSh and MSh. Anemia resulted in the increased expression of all tested proteins (neuronal NOS, VEGF, and GLUT-1). In the AAn group, the translation of GLUT-1, a key protein in switching from aerobic metabolism to glycolysis during oxygen deprivation, appeared to be blunted relative to younger cohorts.

HIF- and hypoxia-regulated downstream mRNA levels were measured in the sham and anemic subgroups within each age group. In figure 5, mRNA levels for all groups are compared with normalized YSh mRNA levels. Multiple downstream genes known to be regulated by the HIF-1α and HIF-2α transcription factors, or by hypoxic exposure, were consistently increased within both the cortex and hippocampus in response to anemia, indicating the sensing of hypoxia in all age groups in response to the anemic challenge.

Table 2summarizes the ANOVA regression analysis testing the dependence of each gene level upon: 1) An versus  Sh treatment, 2) age group, 3) the interaction between the treatment and the age group (An vs.  Sh × Age Grp), 4) the sampling time (24 vs.  48 h), and 5) the site of sampling (hippocampus vs.  cortex).

In table 2, the column labeled An versus  Sh shows that the coefficients for HIF-1α and HIF-2α and all other genes, except for inducible NOS and carbonic anhydrase-9, were positive and significant, indicating that the anemic exposure resulted in the cellular sensing of hypoxia in all age groups.

In the table 2column labeled Age Group, significance of the coefficients tests the effect of age group upon the tested genes in the sham groups. Advanced age in the SHR, independent of any effect of anemic exposure, was associated with increased levels of HIF-1α and HIF-2α mRNA, or an increased sensing of hypoxia, as evidenced by the 31% increase in the HIF-1α coefficient for the ASh group relative to YSh. Erythropoietin receptor and PGK-1 mRNA levels were markedly increased in the mature group (M) relative to the young (Y) animals.

Aging impacted the transcriptional response to the anemic exposure. In table 2, in the column labeled An × Age Grp, we tested the interaction (×) of the effect of anemic exposure within the three age groups. The AAn group experienced a greater elevation of HIF-1α and HIF-2α genes relative to both MAn and YAn groups, particularly at 24 h, offering further support for the conclusion that anemia results in more cellular hypoxia with aging in the SHR. Anemic exposures in the aged (AAn) animals also resulted in elevation of several HIF-regulated or hypoxia-sensitive downstream gene products, including VEGF, endothelial NOS, and neuronal NOS. In contrast, in response to the anemic exposures, levels of HIF-1α and HIF-2α mRNA were not significantly different between the YAn and MAn groups.

In response to exposure to acute isovolemic anemia, we have demonstrated the following: 1) evidence of age-dependent working memory and learning dysfunction which spared reference memory, 2) an unimpaired local cerebral tissue blood flow response and maintenance of tissue oxygenation with aging, 3) molecular evidence of cellular sensing of hypoxia in spite of normal systemic (arterial) and tissue oxygen levels, and 4) an increased sensing of cellular hypoxia with aging.

Working memory, defined as the ability to temporarily store, manipulate, and process new information,30appears requisite for hippocampal based learning.31,32Working memory involves multiple brain regions, including the prefrontal lobe, parahippocamal structures, parietal association cortex, and subcortical network (thalamic nuclei, limbic system).33,34Although the hippocampus and mesial temporal lobe are considered critical for learning and memory, they also appear to have an important role in working memory.35Postoperative cognitive dysfunction primarily affects the working memory domains of attention, cognitive speed, and executive function, as well as hippocampal based memory acquisition or learning.1The impairment of visuospatial working memory and learning but not visuospatial reference memory in our translational model is consistent with the human experience.1,36 

Acute isovolemic anemia does not impair cerebral tissue oxygenation, as measured with an invasive cortical tissue oxygen probe, even with aging and chronic hypertension. The absence of an aging impact upon local cerebral oxygenation with hemodilution was unexpected, especially when using the SHR model. Although hemodilution resulted in stable cerebral oxygenation in both young and aged SHR groups, it should be noted that this was so in the younger group in spite of a 41% decrease in MAP. Stable P^Bro²levels in the aged SHR group were accompanied by a much smaller 17% decrease in MAP. The figures provided in Supplemental Digital Content 9, https://links.lww.com/ALN/A621, with graphs of P^Bro²versus  MAP for both YAn and AAn animals, demonstrate that there may indeed exist a steeper MAP:P^Bro²relationship with anemia in the setting of chronic hypertension.

The demonstration of a seemingly adequate increase in brain blood flow and stable tissue oxygen levels equates only with adequate local oxygen delivery, however. Invasive tissue oxygen probes, such as we employed and such as are used clinically, measure a milieu with contributions from the arteriole, the capillary, the cell, and the venule and may therefore poorly reflect the status of the cell. Further, we were able to sample only one site in the brain. The age- and hypertension-dependent up-regulation of molecular markers of cellular hypoxia, when oxygen delivery is apparently stable and tissue perfusion luxuriant relative to younger cohorts, may suggest that brain oxygen extraction, under conditions of acute isovolemic anemia, may be flow-dependent in the chronically hypertensive brain. An impairment in microvascular compliance and distribution of blood flow during isovolemic anemia with chronic hypertension could also be hypothesized. Finally, these measurements were conducted under the anesthetic influence of α-chloralose, after the discontinuation of isoflurane. α-Chloralose has been used extensively in animal research, including in experiments on functional magnetic resonance imaging and flow-activation coupling, because this anesthetic appears to cause minimal perturbations in cerebral oxygen metabolism and cerebrovascular reactivity. We recognize that this approach may not, and cannot exactly, reflect the truly awake state.

The prolyhydroxylase family of proteins interacts closely with mitochondria to sense cellular hypoxia at the level of the cell.19Hare37and McLaren38first documented, and we now confirm, that clinically relevant levels of acute isovolemic anemia initiate cerebral cellular molecular changes in hypoxia-related gene synthesis and protein translation, in the absence of systemic (arterial) or tissue hypoxia as measured with an invasive cortical tissue oxygen probe. The up-regulation of HIF and related molecular biomarkers of hypoxia may be triggered at oxygen tensions higher than previously thought to be associated with cellular hypoxia. Lopez-Lazaro39has coined the term dysoxia-inducible factor as an alternative to HIF. HIF may be up-regulated or stabilized by subtle changes in basal cellular oxygen metabolism, which in turn may be controlled by subtle changes in the perceived availability of oxygen supply.40Thus, the cell appears to initiate life-saving strategies in response to even mild, sublethal hypoxia, and we must consider the potential impact of this strategy upon cognitive processes.

Aging with chronic hypertension was associated with a markedly increased cellular sensing of hypoxia relative to younger cohorts, as evidenced by the pronounced upregulation of HIF-1α protein, and HIF-1α and HIF-2α genes. Genes and proteins downstream of HIF-1α and HIF-2α, or responsive to hypoxic stimuli (neuronal NOS), however, demonstrated a mixed response. VEGF, endothelial NOS, and neuronal NOS genes were markedly increased in the AAn group relative to MAn and YAn, but most others, including erythropoietin, erythropoietin receptor, GLUT-1, and PGK-1, did not follow suit. Overall, this may indicate a posttranscriptional or translational failure to synthesize key proteins with aging in response to hypoxia. Previous work in nonhypertensive animals has demonstrated an impaired HIF-regulated and hypoxia-sensitive gene and protein expression upon hypoxic exposure with aging.23,41–44We have studied a model where the impact of hypertension upon oxygen delivery and extraction is superimposed upon this aging background. Under these circumstances, therefore, one could hypothesize that anemia might cause a greater degree of hypoxia and a diminished ability to manage the threat successfully.

Consistent with reports of only temporary cognitive impairment after surgery, with fairly consistent reports of recovery at 6 weeks to 3 months, hypoxia need not result in cell death to impact cognition. Working memory and learning (memory consolidation) may be impaired by a temporary bottleneck in protein synthesis.45The dependence of working memory46and neuronal plasticity47upon new protein synthesis, to include probably the most important protein in this process, brain-derived neurotrophic factor, is well documented. The down-regulation of brain-derived neurotrophic factor during exposure to hypoxia,48chronic anemia,49and chronic hypertension50has also been documented, as has been the dependence of brain-derived neurotrophic factor recovery upon erythropoietin synthesis,51and an age-related impairment in erythropoietin response to hypoxia.52Finally, the tight dependence of erythropoietin synthesis upon HIF-2α in the brain53more clearly defines just why mild hypoxia, or anemic hypoxia, might impact working memory and learning in a temporary fashion and why an age-dependent impairment in the ability to synthesize erythropoietin and brain-derived neurotrophic factor might further impact working memory and learning.

In summary, in a translational animal model of aging upon a background of chronic hypertension, acute anemia resulted in age-dependent working memory and learning impairment, similar to the type of cognitive deficit that is seen clinically in humans with similar comorbidities. This was matched by an age-dependent increase in molecular markers of cellular hypoxia. Finally, molecular changes consistent with the sensing of hypoxia at the level of the cell may occur in the absence of systemic and cerebral hypoxia when measured via  invasive tissue oxygen probes. A potential role for anemia in POCD and the possible mechanism of cellular hypoxia is highly relevant in cardiac and noncardiac surgery arenas, especially in the aged, chronically hypertensive patient. The discordance of the tissue oxygen probe data with cellular biomarkers of hypoxia and with the cognitive impairment are particularly relevant in light of the growing utilization of comparable brain tissue oxygen monitors in neurointensive care and the use of surface cerebral oximetry probes in the operating room for the prevention of brain injury.

The authors acknowledge the assistance of Michelle Labricciosa, Research Assistant, Virginia Tech University, Blacksburg, Virginia; Erik Johansson, Research Assistant, La Salle University, Philadelphia, Pennsylvania; and Marcus Pochettino, B.S., Research Assistant, University of Pennsylvania, Philadelphia, Pennsylvania, in the completion of this work.