INTRODUCTION

Top Abstract Introduction Methods Results Discussion References

BOTH EXPERIMENTAL (1, 6, 26, 36) and clinical (2, 14, 15, 30, 40) anemia result in an increase in cerebral blood flow (CBF) that is inversely related to arterial O₂ content (Ca_O2). Because the increase in CBF does not require dilation of large cerebral arteries and pial arterioles (16, 19, 27, 28, 41), the passive decrease in viscosity is presumed to be sufficient for adequately decreasing cerebrovascular resistance. Interestingly, the increase in CBF during anemia is approximately the same as the increase in CBF during hypoxic hypoxia at equivalent reductions in Ca_O2 (20). With hypoxic hypoxia, the increase in CBF compensates for the decrease in Ca_O2, thereby maintaining cerebral O₂ transport (CBF × Ca_O2) to the cerebral microcirculation (20, 22, 42). With anemia, cerebral O₂ transport generally is well preserved (1, 16, 20), although small decreases have also been reported (6, 37). Whether preservation of O₂ transport during anemia is the result of active microcirculatory regulation by an O₂-sensitive mechanism or simply a fortuitous passive effect of decreased hematocrit is unclear.

Increases in CBF with methemoglobinemia (17, 18) or carbon monoxide exposure (23, 24, 30) imply that decreases in O₂-carrying capacity of blood at normal hematocrit and arterial partial pressure of O₂ (Pa_O2) are capable of producing cerebral vasodilation. Another way to separate effects of hematocrit and O₂-carrying capacity is to replace red blood cell-based hemoglobin with plasma-based hemoglobin. In our previous work (38) using an isovolumic exchange transfusion of cell-free, tetrameric bovine hemoglobin stabilized with a fumaryl cross-linker, we have shown that a reduction in hematocrit with albumin transfusion results in a greater increase in CBF than an equivalent reduction in hematocrit and viscosity with cross-linked hemoglobin transfusion. Because cerebral O₂ transport and cerebral metabolic rate of O₂ consumption (CMR_O2) were unchanged after either albumin or cross-linked hemoglobin transfusion, differences in CBF were attributed to differences in Ca_O2. These results support the hypothesis that O₂ transport to cerebrum is actively regulated and not simply related passively to changes in hematocrit and viscosity.

To test further the efficacy of O₂ transport regulation in the present study, we superimposed graded reductions in oxyhemoglobin saturation under conditions of normal hematocrit, decreased hematocrit and O₂-carrying capacity, and decreased hematocrit without a proportional decrease in O₂-carrying capacity. To achieve the latter condition, we performed an exchange transfusion using cell-free human hemoglobin with covalent sebacyl cross-links between the two -82 lysine residues and between the two -99 lysine residues (3, 4). This tetrameric hemoglobin has an O₂ affinity and cooperativity similar to cat blood. We tested the hypotheses 1) that CBF is determined solely by Ca_O2 without separate, independent effects of arterial hematocrit, hemoglobin concentration, and O₂ saturation, and 2) that resultant changes in CBF maintain cerebral O₂ transport, CMR_O2, and fractional O₂ extraction when hematocrit, hemoglobin concentration, and O₂ saturation are independently manipulated over the physiological range. Use of radiolabeled microspheres to measure CBF also permitted blood flow determinations to other organs, such as heart, that are highly sensitive to O₂ delivery.

	METHODS

Top Abstract Introduction Methods Results Discussion References

Hemoglobin preparation. Cross-linked hemoglobin solutions were prepared according to the method of Bucci et al. (4). Briefly, outdated human, banked blood was hemolyzed in the presence of 0.05 M phosphate buffer at pH 7.0 in the cold overnight. The stroma and lipids were removed by centrifugation and dialysis. After deoxygenation and under continuous stream of nitrogen, a 6% solution of hemoglobin in 0.1 M Tris and 0.05 M borate buffer (pH = 9) was incubated for 3 h at 37°C with 4 mg/ml of the cross-linking reagent bis-(3,5-dibromosalicyl)sebacate. The reaction was stopped by the addition of glycylglycine at a concentration of 0.1 M for 30 min at 37°C, followed by dialysis against 0.1 M glycine at 4°C. Hemoglobin solution was concentrated to ~15 g/dl by laminar-flow filtration using a membrane with a 30,000-Da nominal molecular-mass cutoff and then pasteurized at 70-75°C for 3 h. The solution underwent isovolumic dialysis with an isotonic solution containing 125 mM NaCl, 25 mM NaHCO₃, and 4 mM KCl. The final solution was centrifuged and sterile-filtered through a 0.45-nm membrane, collected into a sterile container, and stirred overnight with Detoxygel (Pierce Chemical, Rockford, IL), using 0.2 ml of gel/100 ml of solution to reduce endotoxin content. On the next day, the solution was centrifuged, sterile-filtered again, diluted to a concentration of 7 g/dl, and used within a few hours. Sterile endotoxin-free water was used for the preparation of all of the buffers and other solutions.

Under the conditions of synthesis, one-half of the final product is intramolecularly cross-linked between the 82 lysines of the two -subunits, and the remaining one-half is cross-linked between both the two -82 lysines and the two -99 lysines (4). The compound has been crystallized, and the folding of the sebacyl cross-linker has been modeled in the oxy- and deoxygenated states (3). The PO₂ at 50% saturation (P₅₀) is 34 mmHg at 37°C and a pH of 7.4. The O₂-binding cooperativity gives a Hill parameter of n = 2.2. All hemoglobin preparations contained <5% non-cross-linked material and 7-24% methemoglobin.

Surgical preparation. All procedures were approved by the institutional animal care and use committee. Twenty-five mixed breed male cats weighing 2.5-3.5 kg were anesthetized with pentobarbital sodium (40 mg/kg ip bolus and 6 mg · kg¹ · h¹ iv infusion). Muscle paralysis was produced by the periodic administration of pancuronium bromide (0.1 mg/kg iv) to facilitate monitoring and electrocauterization. Additional pentobarbital was administered with arterial pressure increase during surgery. Previous experience with pentobarbital anesthesia in cats, measuring evoked potentials and cerebral utilization of O₂, has indicated adequate anesthetic depth. Although pentobarbital anesthesia reduces CBF, this occurs in conjunction with reduced O₂ utilization such that the ratio of blood flow to metabolism is maintained. Inhalational anesthetics increase CBF disproportionately to O₂ utilization and may attenuate hypoxic reactivity. The basal level of O₂ utilization in this study (3.4 ml O₂ · min¹ · 100 g¹) is in the range anticipated for general anesthesia without an isoelectric electroencephalogram.

Mechanical ventilation was performed after oral endotracheal intubation to maintain the arterial partial pressure of carbon dioxide (Pa_CO2) at 35-40 mmHg. Supplemental O₂ was administered to maintain Pa_O2 at >80-120 mmHg during surgery. Arterial pH was maintained between 7.35 and 7.45 with intravenous sodium bicarbonate (1 meq/ml).

Femoral veins were catheterized for the administration of fluids and drugs, and femoral arteries were catheterized to measure mean arterial blood pressure (MABP) and for the withdrawal of blood during exchange transfusion and microsphere blood flow measurements. Through a left thoracotomy, a catheter was placed into the left atrium to measure left atrial pressure (P_LA) and for the injection of radiolabeled microspheres. The chest wall was apposed and the animals were turned prone. The head was secured in a frame such that the external auditory meatus was 5 cm above the level of the heart. The scalp was retracted and the sagittal sinus was cannulated at the level of coronal suture to measure sagittal sinus pressure (P_ss) and for sampling cerebral venous blood. A thermistor was placed in the epidural space underlying the calvarium over the right hemisphere. Temperature was maintained with a heating lamp and warming blanket.

Physiological measurements. Blood gases and pH were measured with Radiometer ABL3 electrodes and analyzer (Copenhagen, Denmark). Whole blood hemoglobin concentration and O₂ saturation were determined with a hemoximeter (model OSM3; Radiometer) using the algorithm for cat blood. Ca_O2 and sagittal sinus blood O₂ content (Cv_O2) was calculated assuming an O₂-carrying capacity of 1.36 ml O₂/g hemoglobin. In separate experiments, calculated O₂ content before and after hemoglobin transfusion closely agreed with measurements obtained on a Lex-O₂-Con fuel cell.

Regional blood flow was measured by the radiolabeled microsphere technique (22). Briefly, a well-mixed aliquot of 1.5 × 10⁶ spheres (15 ± 0.5 µm diameter) was injected into the left atrium and flushed with 5 ml normal saline over a period of 20 s. The isotopes (¹⁵³Gd, ^114mIn, ¹¹³Sn, ¹⁰³Ru, ⁹⁵Nb, and ⁴⁶Sc) were injected in random sequence, and the number of the microspheres injected was selected to allow at least 400 spheres to be delivered to the smallest tissue sample taken. The reference sample was withdrawn from the femoral artery catheter at a rate of 1.94 ml/min, beginning just before the microsphere injection and continuing 90 s after completing the infusion of the flush solution. After each experiment the anesthetized cats were killed with intravenous KCl. The brain was removed and placed into 10% Formalin for 1-2 days before sectioning into regions. Tissue and blood samples were analyzed in an autogamma scintillation spectrometer (Packard Instruments, Miniaxi model 5530, Downers Grove, IL). Blood flow was calculated as the product of the arterial withdrawal rate (1.94 ml/min) times the counts in the tissue (corrected for isotope overlap) divided by the counts in the arterial reference sample.

CMR_O2 was calculated as the product of the arterial-sagittal sinus O₂ content difference and blood flow to the cerebrum. Cerebral vascular resistance (CVR) was calculated as (MABP  P_ss)/CBF. Cerebral O₂ transport was calculated as Ca_O2 × CBF. Left heart ventricular resistance was calculated as (MABP  P_LA)/left ventricular blood flow.

Exchange transfusion. The free hemoglobin was diluted in a sterile salt solution containing (in mM) 140 Na⁺, 120 Cl, 25 HCO₃, 3 K⁺, and 1.5 Ca²⁺. A 7 g/dl hemoglobin solution was infused intravenously at a rate of 1.7 ml/min until the hematocrit was reduced to 20-22%. Depending on body weight and starting hematocrit, this required the infusion of 3.2-5.0 g of hemoglobin (~80 ml over 40-50 min). Arterial blood was withdrawn concurrently at a rate of 1.7 ml/min into heparinized syringes with a reciprocal infusion-withdrawal pump. Subsequent blood volume loss due to blood sampling was replaced with blood that was drawn during the exchange transfusion. In another group, cats underwent an exchange transfusion with 2.7-4.8 g of bovine albumin (6 g/dl concentration). The concentration of albumin was less than that of hemoglobin because the negatively charged albumin was presumed to have a Donnan effect and therefore to exert a greater oncotic pressure in vivo. The amount of infused albumin (3.2 ± 0.1 g; mean ± SE) was less than the amount of infused hemoglobin (4.0 ± 0.2 g). Because loss of albumin and hemoglobin into the interstitial space and lymph over time would cause a reduction in plasma volume and an increase in hematocrit, additional amounts of the albumin (1.5 ± 0.2 g) or hemoglobin (1.7 ± 0.2 g) solution were infused continuously after the initial exchange transfusion to keep hematocrit constant.

Experimental protocol. Cats were divided into three groups: anesthesia time control with no transfusion (n = 7), albumin exchange transfusion (n = 8), and hemoglobin exchange transfusion (n = 10). All groups received intravenous hydration with lactated Ringer solution at 8 ml · kg¹ · h¹ during the surgical preparation and 4 ml · kg¹ · h¹ thereafter. Baseline measurements were made 60 min after completion of surgery. In the albumin and hemoglobin groups, the exchange transfusion was then performed over a 40- to 50-min period. One hour after the end of transfusion, repeat physiological measurements were made at this new posttransfusion baseline. Three successively greater degrees of hypoxic hypoxia were then induced by adding increasing amounts of nitrogen to the ventilator air supply. Target levels of hypoxia were based on hemoglobin saturation of O₂ measurements of 80, 65, and 50% using the hemoximeter. Regional blood flow and arteriovenous O₂ content measurements were made at baseline, 60 min after exchange transfusion, and 10 min after each change in inspired O₂ concentration.

Statistical analysis. All variables were analyzed by two-way analysis of variance where transfusion treatment was a between-subject factor and repeated measures over time was a within-subject factor. Post hoc analyses were performed using the Newman-Keuls multiple-range test. Significance was assumed when P < 0.05. Values are expressed as means ± SE.

	RESULTS

Top Abstract Introduction Methods Results Discussion References

Hematocrit was well matched between albumin and hemoglobin groups and was maintained at ~30% less than in the time control group (Table 1). In the albumin group, total hemoglobin concentration and Ca_O2 were reduced in proportion to the reduction in hematocrit. In the hemoglobin group, whole blood hemoglobin concentration and Ca_O2 were reduced proportionately less than the reduction in hematocrit. Whole blood hemoglobin concentration and Ca_O2 in the hemoglobin group were at an intermediate level between those in the control and albumin groups after transfusion and during subsequent reduction of O₂ saturation. Approximately 2.5% of whole blood hemoglobin in vivo was methemoglobin after transfusion with the cross-linked hemoglobin solution. There were no differences in arterial pH or Pa_CO2 from baseline values among groups either after exchange transfusion or during hypoxia (Table 2). With each level of graded hypoxia, arterial O₂ saturation (Table 1) and Pa_O2 (Table 2) were reduced to similar levels among the three groups. Arterial glucose concentration increased during hypoxia in the hemoglobin-transfused group.

Mean arterial pressure increased after hemoglobin transfusion and during hypoxia (Table 3). P_ss increased slightly during hypoxia in all groups. Consequently, cerebral perfusion pressure was increased after hemoglobin transfusion and during hypoxia. As an indirect measure of left ventricular preload, P_LA also increased after hemoglobin transfusion and during hypoxia. Epidural brain temperature was well maintained at normothermic levels.

After exchange transfusion with the albumin solution, CBF increased by 23 ± 5 ml · min¹ · 100 g¹ (Table 4) without a change in arterial O₂ saturation (Table 1) or sagittal sinus PO₂ (Table 2). During hypoxia, CBF increased and was greater than that in the control group for comparable reductions in arterial O₂ saturation and sagittal sinus PO₂ (Fig. 1). After exchange transfusion with the hemoglobin solution, CBF increased by 11 ± 3 ml · min¹ · 100 g¹ (Table 4) in association with a small decrease in arterial O₂ saturation (Table 1) and sagittal sinus PO₂ (Table 2). The relationships of CBF to arterial O₂ saturation and to sagittal sinus PO₂ were more closely matched between the control and hemoglobin groups than between the control and albumin groups (Fig. 1).

View larger version (19K): [in this window] [in a new window]   Fig. 1.   Cerebral blood flow (CBF) vs. arterial O2 saturation and sagittal sinus PO2 in time control (n = 7), albumin-transfused (n = 8), and hemoglobin-transfused cats (n = 10) before and during graded hypoxia. Pretransfusion values are omitted for clarity. Values are means ± SE.

When the CBF data were plotted against Ca_O2 (including pretransfusion values), the relationships were indistinguishable among the three groups (Fig. 2). Thus the increase in CBF seen after hemoglobin transfusion (67 ± 4 ml · min¹ · 100 g¹) at normoxic O₂ saturation was similar to that seen in the control group (68 ± 7 ml · min¹ · 100 g¹) during mild hypoxia at comparable levels of Ca_O2 (11.9 ± 0.3 vs. 11.7 ± 0.6 ml/dl, respectively). Likewise, the increase in CBF seen after albumin transfusion (79 ± 10 ml · min¹ · 100 g¹) at normoxic O₂ saturation was similar to that seen in the hemoglobin group (76 ± 5 ml · min¹ · 100 g¹) during mild hypoxemia at comparable levels of Ca_O2 (10.1 ± 0.5 vs. 9.6 ± 0.4 ml/dl, respectively). With more severe hypoxemia, the increase in CBF in the albumin and hemoglobin groups was similar to that of the control group at equivalent levels of Ca_O2.

View larger version (16K): [in this window] [in a new window]   Fig. 2.   CBF and cerebrovascular resistance (CVR) vs. arterial O2 content (CaO2) in time control (n = 7), albumin-transfused (n = 8), and hemoglobin-transfused cats (n = 10) before and after transfusion and during graded hypoxia. Values are means ± SE.

After albumin exchange transfusion, CVR decreased from 2.15 ± 0.14 to 1.58 ± 0.14 mmHg · ml¹ · min · 100 g (Fig. 2). After hemoglobin transfusion, CVR was unchanged as a result of the smaller increase in CBF and the increase in cerebral perfusion pressure. During hypoxemia CVR decreased in all groups, but the relationship of CVR to Ca_O2 was not as tight among groups as the relationship of CBF to Ca_O2 because of increases in mean arterial pressure. In the control, albumin, and hemoglobin groups, the decrease in CVR during hypoxemia became significant starting at the 50, 80, and 65% levels of arterial O₂ saturation, respectively.

Cerebral O₂ transport was not significantly changed by albumin or hemoglobin exchange transfusion or by subsequent hypoxemia (Fig. 3). Over a wide range of Ca_O2 values generated by differences in hematocrit, hemoglobin concentration, and hypoxia among groups, cerebral O₂ transport, CMR_O2, and cerebral fractional extraction of O₂ were similar among groups (Fig. 3).

View larger version (20K): [in this window] [in a new window]   Fig. 3.   Cerebral O2 transport (CBF × CaO2), cerebral metabolic rate of O2 consumption (CMRO2), and cerebral fractional O2 extraction with varying CaO2 in time control (n = 7), albumin-transfused (n = 8), and hemoglobin-transfused cats (n = 10) before and after graded hypoxia. Values are means ± SE.

Analysis of regional brain blood flow measurements revealed a pattern similar to that seen in the entire cerebrum (Figs. 4 and 5). Increases in blood flow in the posterior fossa structures (cerebellum and brain stem), hippocampus, and white and deep gray (caudate nuclei) matter corresponded to Ca_O2 values in the control, albumin, and hemoglobin groups. Hemoglobin transfusion caused a decrease in blood flow to the neurohypophysis and choroid plexus (Table 4). These two regions have high basal flow and are devoid of a blood-brain barrier. Hypoxia increased neurohypophysial blood flow, but blood flow in the hemoglobin group remained less than that in the control and albumin groups. In choroid plexus, hypoxia produced a decrease in blood flow in the control and hemoglobin groups, whereas blood flow remained unchanged in the albumin group.

View larger version (20K): [in this window] [in a new window]   Fig. 4.   Cerebellum and brain stem blood flow vs. CaO2 in time control (n = 7), albumin-transfused (n = 8), and hemoglobin-transfused cats (n = 10) before and after transfusion and during graded hypoxia. Values are means ± SE.

View larger version (22K): [in this window] [in a new window]   Fig. 5.   Regional CBF to gray matter (caudate nuclei), white matter, and hippocampus vs. CaO2 in time control (n = 7), albumin-transfused (n = 8), and hemoglobin-transfused cats (n = 10) before and after transfusion and during graded hypoxia. Values are means ± SE.

Reduction in hematocrit with albumin exchange transfusion increased blood flow to small intestine, kidney, and forelimb skeletal muscle (Table 4). In contrast, hemoglobin transfusion resulted in unchanged intestinal and muscle blood flow and in decreased renal blood flow. Hypoxemia decreased blood flow to intestines and kidney in the hemoglobin group to values less than those in the control and albumin groups. Increased skeletal muscle blood flow during hypoxemia in the control and albumin groups was attenuated in the hemoglobin group.

Blood flow to the free wall of the left ventricle increased with the reduction in hematocrit and Ca_O2 after albumin and hemoglobin exchange transfusion and increased further with subsequent hypoxia (Fig. 6). The relationships of coronary blood flow and vascular resistance to Ca_O2 were similar in the control and albumin groups. In the hemoglobin group, the increase in coronary flow was attenuated during severe hypoxemia. Coronary vascular resistance was maintained near the baseline levels after hemoglobin exchange transfusion in association with the increase in MABP. During hypoxemia, coronary resistance in the hemoglobin group decreased in parallel with the other groups.

View larger version (22K): [in this window] [in a new window]   Fig. 6.   Mean arterial blood pressure (MABP) and left ventricular blood flow and vascular resistance vs. CaO2 in time control (n = 7), albumin-transfused (n = 8), and hemoglobin-transfused cats (n = 10) before and after transfusion and during graded hypoxia. Values are means ± SE.

	DISCUSSION

Top Abstract Introduction Methods Results Discussion References

The major findings of this study are that 1) CBF during anemia remains higher than that in the control group during arterial O₂ desaturation, 2) increasing O₂-carrying capacity at low hematocrit with plasma-based hemoglobin results in less of an increase in CBF at O₂ desaturation levels equivalent to those in the anemic groups, and 3) CBF is a unitary function of Ca_O2 when hematocrit, O₂-carrying capacity, and oxyhemoglobin saturation are independently varied. Consequently, bulk O₂ transport is maintained at equivalent levels during 1) hypoxemia, 2) anemia, 3) hypoxemia plus anemia, and 4) hypoxemia plus reduced hematocrit without a proportional decrease in O₂-carrying capacity.

Exchange transfusion with cell-free hemoglobin produces a moderate decrease in Ca_O2 at normoxic Pa_O2 because the oncotic pressure of tetrameric hemoglobin results in less hemoconcentration than can occur inside red blood cells. In addition, the hemoglobin solution contains some methemoglobin. Nevertheless, whole blood O₂-carrying capacity is augmented compared with exchange transfusion with an albumin solution. The present results with exchange transfusion of tetrameric human hemoglobin stabilized with a sebacyl cross-linker confirm our previous work on basal CBF after exchange transfusion of tetrameric bovine hemoglobin stabilized with a fumaryl cross-linker (38). In both cases, the increase in CBF normally seen at reduced hematocrit was attenuated. This attenuation was not attributed to nitric oxide scavenging by plasma-based hemoglobin in brain with tight endothelial junctions because differences in CBF between the albumin and fumaryl cross-linked hemoglobin-transfused groups persisted after N-nitro-L-arginine inhibition of nitric oxide synthase (38) and because endothelium-dependent dilation remains intact (Y. Asano, R. C. Koehler, J. A. Ulatowski, R. J. Traystman, and E. Bucci, unpublished observations). Rather, differences in CBF at normoxic Pa_O2 are more likely attributed to differences in Ca_O2. The present results indicate that the increases in CBF after albumin and hemoglobin exchange transfusion were appropriate for the resultant changes in Ca_O2 compared with a control group in which Ca_O2 was reduced by graded hypoxic hypoxia.

We have previously observed an ~25% decrease in whole blood viscosity at a hematocrit of 18% measured by cone viscometry for both albumin and hemoglobin transfusion (38). The increase in CBF after albumin exchange transfusion does not require pial arteriolar dilation (16, 19, 28). Pial arteriolar pressure remains as a constant fraction of aortic pressure after hemodilution (19). These results imply that the decrease in CVR after hemodilution is attributable to a decrease in viscosity in vivo rather than a decrease in vascular hindrance (resistance/viscosity) and that the segmental distribution of hindrance between small and large arterioles remains approximately constant. The lack of a change in cerebral venous PO₂ after hemodilution at normoxic Pa_O2 and arterial O₂ saturation indicates that any changes in arteriolar diameter act to minimize changes in the blood-to-tissue PO₂ gradient. Moreover, fractional O₂ extraction, which is the ratio of CMR_O2 to cerebral O₂ transport, remains constant after hemodilution. A constant fractional O₂ extraction at normoxic Pa_O2 and arterial O₂ saturation (and constant P₅₀) is also consistent with a constant venous PO₂. Results with hemoglobin transfusion suggest that constant venous PO₂ and fractional O₂ extraction are not simply a fortuitous effect of decreased viscosity at low hematocrit. With augmented Ca_O2 at reduced hematocrit, fractional O₂ extraction remained unchanged and cerebral venous PO₂ decreased by only 2 mmHg. This small decrease in venous PO₂ may be the result of a small decrease in arterial O₂ saturation (97 to 94%) and is in contrast to the 9- to 13-mmHg decrease previously observed after fumaryl cross-linked bovine hemoglobin (38). We attribute this difference in cerebral venous PO₂ to the low P₅₀ of the fumaryl cross-linked bovine hemoglobin (17 mmHg). The currently used sebacyl cross-linked human hemoglobin has a P₅₀ of 34 mmHg, which is close to that of native cat blood (36-37 mmHg). Thus, with a well-matched P₅₀ between red blood cell- and plasma-based hemoglobin, cerebral venous PO₂ is well maintained when Ca_O2 is increased at low hematocrit.

These results imply that active vasoconstriction was required after hemoglobin exchange transfusion to keep CBF from increasing to the same extent as occurred with albumin exchange transfusion. The resulting increase in vascular hindrance should partially offset the decrease in blood viscosity and lead to a small decrease in CVR to account for the small increase in CBF. However, MABP increased after hemoglobin transfusion, presumably because of nitric oxide scavenging by hemoglobin in the peripheral vasculature (39), and additional vasoconstriction was required to keep CBF at a level appropriate for the Ca_O2. Thus CVR did not decrease after hemoglobin transfusion, and CVR was greater than that after albumin transfusion.

In addition to showing that CBF after albumin and hemoglobin transfusion at normoxic Pa_O2 is appropriate for the level of Ca_O2, the present results show that CBF remains at a level appropriate for preserving cerebral O₂ transport when hypoxic hypoxia is superimposed at low hematocrit. Similar results were obtained in newborn lambs when Pa_O2 was varied at different hematocrits (20). The present results indicate that hypoxic vasodilation remains intact at low hematocrit even when baseline arteriolar tone is augmented by increases in O₂-carrying capacity with plasma-based hemoglobin. The decrease in viscosity or the increase in velocity after hemodilution would act to decrease or increase endothelial wall shear stress, respectively, and changes in vivo cannot easily be predicted. If changes in wall shear stress occur with hemodilution or with hemoglobin transfusion, our results indicate that these changes do not interfere with hypoxic vasodilation.

Hypoxic vasodilation presumably depends on release of mediators from neurons and glia in response to decreased tissue PO₂. Because isolated cerebral arteries can relax at low PO₂ (31, 32), intravascular PO₂ sensors may also be involved. Comparison of the albumin and hemoglobin groups at equivalent hematocrit implies a greater role for tissue PO₂ than vascular PO₂. Neither arterial PO₂ nor cerebral venous PO₂ was higher in the hemoglobin group at each level of hypoxia, yet CBF in the hemoglobin group was lower than that in the albumin group over a wide range of cerebral venous PO₂ (Fig. 1). Thus it is unlikely that vascular PO₂ at the arteriolar level was substantially higher in the hemoglobin group or that the lower CBF in the hemoglobin group was due to a higher arteriolar PO₂. Rather, the greater O₂ content at each PO₂ with plasma-based hemoglobin presumably permits greater O₂ diffusion into tissue, a greater tissue PO₂, and an attenuated CBF response for a particular venous PO₂. Consequently, CBF is more strongly related to Ca_O2 during hy- poxic hypoxia than to arterial or venous PO₂. The mechanism and the cellular site of regulation of blood flow to changes in O₂ content remains unknown.

One situation in which CBF is not related solely to Ca_O2 is when P₅₀ changes. For example, increasing P₅₀ lowers CBF at a particular Ca_O2 during normoxia and hypoxia (22), whereas decreasing P₅₀, such as with carbon monoxide, increases CBF at a particular Ca_O2 compared with either hypoxic hypoxia (23, 24) or anemia (30). Changes in P₅₀ have been modeled to account for changes in CBF through changes in tissue PO₂ (35). Therefore, the CBF response to hypoxemia at constant CMR_O2 can be predicted by two variables: P₅₀ and Ca_O2. The present study indicates that Ca_O2 completely accounts for effects of O₂ saturation and O₂-carrying capacity without additional effects of hematocrit and associated changes in viscosity.

The combination of 50% arterial O₂ desaturation at hematocrits of 22% did not critically limit CMR_O2 after either albumin or hemoglobin transfusion. Thus cerebral O₂ transport was adequate for preserving CMR_O2 without large increases in fractional O₂ extraction. Moreover, changes in CMR_O2 did not produce a confounding influence in the interpretation of the CBF results. Because most general anesthetics lower CMR_O2, it is possible that CMR_O2 would have become crit- ically limited in the unanesthetized state during combined anemia and hypoxic hypoxia. We chose pentobarbital as the general anesthetic agent because both basal CBF and the response to hypoxia are reduced in proportion to CMR_O2 (5).

In addition to cerebrum, single relationships of blood flow to Ca_O2 were obtained in most other brain regions. However, two regions in which blood flow was not uniformly related to Ca_O2 were neurohypophysis and choroid plexus. Increases in neurohypophysial blood flow during hypoxic hypoxia appear to be related to peripheral chemoreceptor activation and vasopressin secretion (12). The lack of an increase in flow during hemodilution in the present study is consistent with the lack of an increase in flow during carbon monoxide hypoxia (13). The decrease in neurohypophysial blood flow after hemoglobin transfusion may be due to scavenging of nitric oxide in this region devoid of a blood-brain barrier. We have previously observed (38) that N-nitro-L-arginine administration reduced neurohypophysial blood flow in control and albumin-transfused cats by the same amount as hemoglobin transfusion alone, whereas administration of the nitric oxide synthase inhibitor produced no further reduction in flow after hemoglobin transfusion. The increase in neurohypophysial blood flow during hypoxic hypoxia in the hemoglobin group suggests that any scavenging of nitric oxide by hemoglobin within the neurohypophysis does not interfere with the hypoxic response in this region.

Others have reported no change in blood flow to choroid plexus during hypoxic hypoxia (7, 21, 42). The reduction in blood flow that we observed during hypoxia may be related to vasopressin- and ANG II-mediated constriction (7, 8, 25). The decrease in blood flow after hemoglobin transfusion and the lack of a change with albumin transfusion suggest that hemoglobin may scavenge nitric oxide and cause constriction in choroid plexus.

We also observed that hemoglobin transfusion potentiated the intestinal and renal vasoconstrictor response to hypoxic hypoxia. Because N-nitro-L-arginine inhibits the constrictor response to hemoglobin transfusion in these vascular beds during normoxia (39), this potentiation during hypoxia may be related to the loss of nitric oxide counteraction of sympathetic vasoconstriction (29 33). In addition, endothelin contributes to the hypertensive response to cross-linked hemoglobin transfusion in the rat during normoxia (10, 11, 34) and may also potentiate peripheral vasoconstriction during hypoxia. Furthermore, arterial glucose concentrations increased after hemoglobin transfusion and during hypoxia in this group. It is possible that a sympathoadrenal response unopposed by nitric oxide could augment hyperglycemia or that hemoglobin perfusion of the liver augments gluconeogenesis.

In skeletal and cardiac muscle, blood flow increased after hemodilution and increased further during hypoxia. The increase in flow during hypoxia was attenuated in the hemoglobin group at equivalent O₂ desaturation. As in brain, Ca_O2 is a critical variable in these tissues. For example, blood flow to muscle increases nonlinearly with O₂ desaturation and O₂ demand (9). The relatively large but variable increase in skeletal muscle flow (24.5 ± 11.6 ml · min ¹ · 100 g¹) at 50% O₂ desaturation (Table 4) in the anemic group probably reflects the nonlinear response to Ca_O2 and the interanimal variability in Ca_O2 at this severe level of combined hypoxic and anemic hypoxia. The left ventricular blood flow response to hypoxia was nearly superimposable in the control and albumin groups for a particular level of Ca_O2 (Fig. 6). The coronary blood flow response to mild and moderate hypoxia also appeared to be appropriate for the level of Ca_O2 in the hemoglobin-transfused group, although interpretation cannot be made definitively without myocardial O₂ consumption measurements. The increase in MABP after hemoglobin transfusion and during mild hypoxia may have increased O₂ demand on the myocardium more than in the control and albumin groups. Moreover, the increase in coronary flow was attenuated at the most severe reduction of Ca_O2 in the hemoglobin group. Whether this represents improved tissue oxygenation or impaired hypoxic vasodilation is unclear. It is important to consider that plasma-based hemoglobin theoretically may improve tissue oxygenation 1) by improving capillary O₂ flux homogeneity by overcoming the effect of red blood cell flux heterogeneity, 2) by increasing the effective capillary surface area for O₂ diffusion that is normally limited by the particulate nature of red blood cell capillary transit, and 3) by facilitating O₂ transport between red blood cell-based hemoglobin and myoglobin. Thus a lower coronary blood flow during severe hypoxia may reflect less vasodilation because of improved tissue oxygenation by plasma-based hemoglobin. Alternatively, leakage of hemoglobin into the myocardial interstitium could scavenge nitric oxide and limit maximal vasodilation.

In summary, the present study shows that reducing hematocrit without a proportionate decrease in Ca_O2 by exchanging red blood cell-based hemoglobin with plasma-based cross-linked hemoglobin results in a lower CBF than reducing hematocrit and Ca_O2 proportionately with albumin transfusion. The differences in blood flow were attributable to differences in Ca_O2. During hypoxic hypoxia CBF remained related to Ca_O2 despite differences in hematocrit and O₂-carrying capacity. Thus cerebral O₂ transport appears well regulated when Ca_O2 is manipulated by independently changing hematocrit, hemoglobin concentration, and arterial O₂ saturation.