Journal of Applied Physiology

Chronic hypoxia and the cerebral circulation

Kui Xu, Joseph C. LaManna


Exposure to mild hypoxia elicits a characteristic cerebrovascular response in mammals, including humans. Initially, cerebral blood flow (CBF) increases as much as twofold. The blood flow increase is blunted somewhat by a decreasing arterial Pco2 as a result of the hypoxia-induced hyperventilatory response. After a few days, CBF begins to fall back toward baseline levels as the blood oxygen-carrying capacity is increasing due to increasing hemoglobin concentration and packed red cell volume as a result of erythropoietin upregulation. By the end of 2 wk of hypoxic exposure, brain capillary density has increased with resultant decreased intercapillary distances. The relative time courses of these changes suggest that they are adjusted by different control signals and mechanisms. The CBF response appears linked to the blood oxygen-carrying capacity, whereas the hypoxia-induced brain angiogenesis appears to be in response to tissue hypoxia.

  • altitude adaptation
  • angiogenesis
  • cerebral blood flow
  • cerebral blood volume


the relationship between brain function and blood flow has been studied since the classical paper of Roy and Sherrington (61) at the end of the 19th century. Despite the absence of quantitative methods, early investigators were able to establish some of the basic principles of the cerebral circulation; among these was a recognition of the dominating influence of carbon dioxide and the reciprocal effect of oxygen (72). Interestingly, even the idea that local cerebral blood flow (CBF) was transiently augmented with increased neuronal activation, alluded to by Roy and Sherrington, had received additional qualitative experimental support. A listing of the qualitative methods in use at the time, including methods such as venous outflow, radium emanation, and thermocouple, can be found in the 1936 Wolff review (72).

Nevertheless, despite more than 100 years of effort, there are still major areas concerning the control of the cerebral circulation that remain unknown. The observations that decreasing oxygen led to increased blood flow and that the increase was blunted perhaps by the accompanying hypoxia-induced hyperventilation and resultant decreased carbon dioxide remained wholly qualitative until a useful quantitative method, based on inert gases, was developed for measuring CBF and oxygen metabolism by Kety and Schmidt (33, 35). They recorded the blood flow increase in seven young men from the resting level of 54 ± 4 ml·100 g−1·min−1 (mean ± SE) to 73 ± 5 ml·100 g−1·min−1 while breathing 10% oxygen, with an average increase of 137 ± 9% (34). These investigators chose to report the blood flow measurements per 100 g tissue presumably to provide convenient units with whole numbers rather than decimals (35). This tradition has been followed ever since, although many investigators are beginning to switch to the more scientific international friendly units (ml·g−1·min−1).

Severinghaus and his colleagues (64) used the Kety and Schmidt method to measure CBF in seven volunteers breathing atmospheric air at sea level and at the Barcroft Laboratory at White Mountain, California, which is at an altitude of 3,810 m. This was the first quantitative study of the time course of the hypoxic CBF response. They reported a mean sea-level CBF of 42 ± 2 ml·100 g−1·min−1, increasing to 51 ± 4 ml·100 g−1·min−1 after 6–12 h at altitude, and falling back toward the sea-level CBF after 3–5 days acclimation (47 ± 3 ml·100 g−1·min−1). They realized that as the Pco2 had fallen to 30 Torr, the return of CBF implied a “resetting” of the central medullary chemoreceptor. The data from this study have been analyzed in great detail recently (70, 71).

Current models of chronic hypoxia.

This review concerns itself with the cerebrovascular adaptations that occur in response to chronic hypoxia. The term “hypoxia” is used to cover a wide variety of situations, and usually it needs to be defined through the use of specific delimiters. In general, hypoxia simply describes a lower than normal oxygen pressure. For this paper, we will consider hypoxia to refer to lower than normal atmospheric oxygen. This condition might be produced by decreasing the fractional component of oxygen in the inspired gas mixture, producing “hypoxic hypoxia.” Several systems are currently available to control the composition of gases in normobaric chambers, allowing the study of intermittent hypoxic exposures on an infinite variety of duty cycles. Hypoxia can also be produced by decreasing the overall barometric pressure, resulting in “hypobaric hypoxia.” Hypobaric hypoxia can be produced either naturally by ascent to altitude or in “Wright chambers,” which are chambers for housing small animals fitted for withstanding a partial vacuum (73). Most of the literature studies on hypoxic adaptation have been conducted at altitude, in research stations around the world or in laboratory simulations of climbs (for an exceptional review, see Ref. 69). There do not appear to be important differences in the physiological adaptive response to the various methods of producing continuous hypoxic exposure.

The degree of hypoxia also has had many descriptors, however, most investigators refer to three degrees (mild, moderate, and severe), without complete agreement on where the boundaries are drawn for each level. For the most part, hypoxia is considered mild if the arterial oxygen tension (PaO2) does not fall below ∼50 Torr (assuming sufficient red cell volume). This degree is considered mild because there is almost complete compensation and function is relatively undisturbed. The equivalent of 10% normobaric oxygen, or an altitude of ∼5,000 m, is the upper limit for mild hypoxia. Moderate hypoxia, PaO2 between ∼35 and 50 Torr, will manifest as a deterioration of thought processes, and below ∼35 Torr unconsciousness results. Although the results are not completely clear, mild hypoxia is not associated with permanent neuronal damage, but moderate and severe degrees of hypoxia lead to neuronal loss in a time of exposure/degree of hypoxia-dependent manner.

The term “chronic” is used loosely here to indicate an interval of time for a physiological response to occur. In the case of hypoxia, acute exposure occurs during an interval from a few minutes up to hours or even days, where chronic exposure occurs over weeks to months. Longer term adaptation occurs over generations and can be observed and studied in animal (e.g., Ref. 62) and human (e.g., Ref. 1), populations in the Himalayas, Andes, and Ethiopia.

This short review will concentrate on the physiological adaptations of the cerebral circulation to prolonged exposure to mild hypoxic environments. It will rely mainly on data from the rodent model because most of the quantitative animal studies have been done with the rat, but it will include sections on species comparisons and human studies.


If it is considered that decreased oxygen delivery is the environmental stimulus for triggering the adaptive responses, then it remains to be determined what are the significant variables contributing to the control mechanisms. These are well known from basic physiology. From the point of view of the systemic circulation, the first variable, obviously, is the PaO2, which is the partial pressure of oxygen as a gas dissolved in the arterial blood. Second is the concentration of the oxygen-carrier hemoglobin, which is located in the red blood cells and is given as milligrams per deciliter or as the packed red cell fraction, often referred to as the hematocrit. The saturation curve for hemoglobin is a third factor, and this curve is altered by temperature, pH, arterial Pco2 (PaCO2), and circulating 2,3-diphospho-d-glycerate. Within the central nervous system, the important considerations are the blood flow rate and the capillary density (hence intercapillary distance).


As noted above, CBF increases with exposure to hypoxia (15). Mild hypoxia results in an increase in CBF up to but not quite double despite a fall in PaCO2 (2, 5, 10, 36). The exact mechanism responsible for the increased CBF is unknown; however, it appears to have a major neurogenic component originating in the brain stem (54).

Local signals will also have an influence [e.g., more of the vasodilator nitric oxide (NO) will be present as the amount of oxygenated hemoglobin decreases], but environmental hypoxia is a systemic challenge and many mechanisms contribute to the adaptation. Local tissue factors may have more to do with intracerebral circulatory distribution rather than whole organ blood flow. Although local chemical and metabolic factors such as K+, adenosine, and NO, among other things, certainly become increasingly important with severity of hypoxia (58), the mechanism for at least half of the acute increase in brain flow in response to mild hypoxia is most likely via neuronal pathways that originate or pass through the brain stem (18, 19, 68) and closely related to the blood oxygen content (6, 28, 31). It may be part of a more generalized stress response (7). Hypoxia-induced hyperventilation reduces PaCO2 (53), which opposes cerebral vasodilation, thus limiting the increase in blood flow.

When hypoxic exposure is prolonged for more than a day or so, the CBF increase becomes attenuated (40, 74) so that by 3 wk of exposure the CBF has returned to prehypoxia baseline.

Packed red cell volume.

One possible reason for the renormalization of CBF is the well-known observation that the packed red cell volume increases with exposure to hypoxia (42). This means that the oxygen content of the circulating blood has been restored by balancing the loss of oxygen with increased carrier, so that the oxygen delivery in (ml O2·100 g−1·min−1) can be restored at the prehypoxic blood flow rate. In the rat, the packed red cell volume increases from the normoxic ∼45% to ∼70% (22–24, 40, 74). In fact, oxygen delivery remains within the normal range during the entire hypoxic exposure (74).

Angiogenesis and cerebral blood volume.

Although the oxygen delivery to the brain is relatively compensated during chronic hypoxic exposure, it could be presumed that the delivery of oxygen to the mitochondria within the parenchyma will be diminished because the driving force for diffusion from capillary to tissue is the Po2 difference between capillary and tissue. Because the tissue Po2 is close to zero, the driving force is essentially the PaO2, which remains low during continued hypoxic exposure. (There is a small, but significant, increase in PaO2 with chronic hypoxia to the extent of the hyperventilatory drop in PaCO2.) This decrease in the driving force may underlie the finding that there is an increase in capillary density through angiogenesis that is completed by 3 wk of hypoxic exposure (4, 11, 20, 40, 51, 56, 57, 60).

Angiogenesis occurs through the same hypoxia-inducible factor-1 (HIF-1) transcription factor mechanism that leads to upregulation of erythropoietin and increased packed red cell volume, but it works through upregulation of vascular endothelial growth factor. There is also a HIF-1-independent, cyclooxygenase-2/angiopoietin-2 dependent component of cerebral angiogenesis (59). Hypoxia-induced cerebral angiogenesis has been recently reviewed (39). The cartoon in Fig. 1 summarizes the key components of hypoxia-induced angiogenesis.

Fig. 1.

Diagram depicts the components of hypoxia-induced angiogenesis. In the normoxic state, angiopoietin-1 (Ang-1) acting through the Tie-2 receptor on cerebral endothelial cells maintains structural integrity. During hypoxia, cyclooxygenase-2 (COX-2) is activated and arachidonic acid (AA) is converted to prostaglandin E2 (PGE2). PGE2 induces release of angiopoietin-2 (Ang-2) from endothelial cells, which occupies the Tie-2 receptor resulting in structural instability and falling away of the pericyte. This allows access of hypoxia-inducible factor-1 (HIF-1) upregulated VEGF to its receptors, stimulating the hypertrophy and hyperplasia of capillary angiogenesis.

The increased capillary density should be detectable as an increase in cerebral blood volume (CBV). Acute hypoxic exposure has been shown to result in an increase in CBV (32, 65). The acute exposure result is most likely due to vasodilation. The effect of prolonged hypoxic exposure on CBV has also recently been reported (14) and related directly to capillary density.

Tissue oxygen tension.

Even under normoxic conditions, the mean tissue oxygen tension is low and the distribution is heterogeneous, with a left-shifted histogram (46, 66). Even so, regardless of the starting level, acute hypoxia results in a decrease of tissue oxygen (38, 43, 49, 66).

Assuming the hypoxic adaptations are successful, the end result should be a restoration of tissue oxygen tension (41). This indeed appears to be the case. After 3 wk of hypoxia, tissue oxygen tension as measured by an in vivo electron paramagnetic resonance technique show levels equivalent to the prehypoxic baseline (13). An important question for understanding the control mechanism, not yet answered, is at what time interval during the adaptation phase the oxygen tension is restored.

The relative time courses of the responses of CBF, packed red cell volume, capillary density, and HIF-1 are shown in Fig. 2. CBF initially responds rapidly then falls off by the fourth or fifth day (74). Packed red cell volume has begun to increase by day 3, reaching ∼80% of maximum by 7 days (74). Angiopoietin-2 is upregulated for the 2 wk during which angiogenesis is occurring, and then it falls to baseline by 3 wk (60). HIF-1, indicative of tissue hypoxia, is initially elevated, falling to about half after 4 days and then back to baseline by 3 wk (8). These data imply that the restoration of brain tissue oxygen tension does not occur until after 2 wk and before 3 wk of hypoxic exposure. This interpretation implies that CBF is not a function of tissue oxygen tension to any great extent but that the process of angiogenesis does directly respond to tissue signals.

Fig. 2.

Relative time courses of cerebral blood flow, packed red cell volume [hematocrit (Hct)], HIF-1α (Hif), capillary density, and angiopoietin-2 (Ang 2) in response to chronic mild hypoxic exposure. Data from sources cited in text.

Mean transit time.

Even though CBF has returned to prehypoxic baseline levels, it does not mean that the cerebral circulation has not undergone significant changes. Blood flow and blood volume are related to each other by consideration of the mean transit time through the Stewart-Hamilton principle (48). Thus, in its simplest form, the mean transit time can be calculated by dividing the blood volume by the blood flow. This variable refers to the average time it takes any unit of blood to traverse from the arterial end of the capillary bed to the venous end. For the rat, the average mean transit time is ∼1.5 s (3 ml/100 g divided by 130 ml·100 g−1·min−1 times 60 s/min). If CBV doubles after hypoxic adaptation, the mean transit time will become much longer, also doubling to ∼3 s. This means that there is much more unloading time for oxygen and, probably as significantly, for glucose. The effect on glucose delivery of the increased transit can be recognized in the large increase in glucose influx across the blood-to-brain barrier after adaptation to chronic hypoxia (22). Not only is there an increase in the capillary density, but also there is an increase in the number of GLUT1 transporter molecules per milligram of microvessel. Together with the increased residence time in the capillary, the glucose influx increases by three times compared with normoxic controls, reaching an extraction fraction of 0.7 or more.


Up to now, we have concentrated mostly on rat studies. Nevertheless, there are many commonalities with studies of other species. The general mammalian pattern holds for the most, although there are some interesting differences. With respect to CBF, hypoxia induces an acute increase. This includes sheep (29, 31, 37, 75) and cows (47). Hypoxic-adapted animals have a blood flow not too different from the normoxic. There are data in conflict with this, for example (37), where CBF remains elevated for 4 days in sheep. It appears that the discrepancy is related to the fact that these sheep develop acute mountain sickness (AMS). In sheep that did not develop AMS, the CBF tended to fall during the 4 days to about one-half of the hypoxic maximum, a result not too different from the rat. This is not to suggest that the increased CBF is a causative factor in AMS because this is unlikely (30).

Other species differences occur due to various niche issues that are beyond the scope of this review but include species that normally reside at higher altitudes, such as the llama (44), those that are diving mammals such as the seal (25), and those that hibernate such as the ground squirrel (12). In addition to these mammalian variations, there are an even greater number when vertebrates are considered such as birds (3, 16), the turtle, and other anoxia-tolerant vertebrates (27, 55). Major differences also exist in comparison between the fetus and the adult (17, 45).


To compare species properly, it is important to distinguish the differences that are due to size. For this we use the principles of allometry to understand the pertinent scaling variables (63). In general, blood flow and metabolic rate are proportional. Thus it is no surprise that blood flow is a scaled variable. CBF, expressed per unit weight, scales with about the −0.25 power of body weight; the smaller the mammal, the higher the blood flow. CBV and packed red cell volume do not scale, but they remain constant across mammals. Thus the ratio of these variables, the mean transit time, must also scale, and larger brains will have longer transit times. The longer transit times in the larger brains might be the reason for the unexplained observation that the human red blood cell has a significant amount of the GLUT1 transporter, whereas small mammals like the rat and mouse do not have GLUT-1 transporters on their red blood cells (21).

Human studies.

There is a long and rich literature concerning cerebrovascular physiology and adaptation to altitude (42, 69). In general, the findings appear to echo the same mammalian pattern that has been discussed. In many ways the human experiments have preceded and surpassed those of the animal studies. There are far more quantitative studies of CBF, cerebral metabolic rate of oxygen (CMRO2), and cerebral metabolic rate of glucose in humans than in animals. The basic observation of a hypoxia-induced, acute increase in CBF followed by adaptation to prehypoxic levels have been confirmed many times (9, 50, 52, 67). Resting CBF measured in populations living at altitude (highlanders) is quite similar to that of the average lowlander. In a recent study of nine healthy adults in whom CBF was measured after 3 wk at Chalcaltaya in Bolivia (5,260 m) and then some 8 mo or so after return to sea level, CBF was found to be unchanged at 53 ml·100 g−1·min−1 in both conditions (52). These investigators failed to detect the decrease in CMRO2 at altitude that had been predicted by Hochachka’s proposal of decreased metabolic rate as a defense against hypoxia (26). CMRO2 was also not altered by 3.5 days at 3,810 m, ∼3.0 ml O2·100 g−1·min−1 (64).


Exposure to mild hypoxia rapidly induces an acute increase in CBF that becomes attenuated over 5 days as the oxygen carrying capacity of the blood increases. A later growth factor driven angiogenesis leads to increased capillary density over a 2 wk interval, restoring tissue oxygen levels. The increased ventilation rate is driven by the Po2 and stays elevated as long as Po2 stays low. CBF, probably controlled by brain stem sensors and circuits, responds to the oxygen carrying capacity of the blood. The microvasculature responds to tissue hypoxia with transient angiogenesis. The delay between the fall in blood flow and the opening of new capillaries implies a period of ∼1 wk or more during which the tissue is in uncompensated hypoxia. It is unknown whether during this period normal neural function is affected or if there is a limit to the maximal work rate of tissue until full tissue oxygen is restored. Blood oxygen level-dependent functional magnetic resonance imaging studies in hypoxia-adapting subjects might prove revealing in this regard.

It remains to be determined whether or not these conclusions apply to all species, including humans. For example, it would be very useful to know whether there is a significant increase in blood volume that persists in humans adapting or adapted to altitude. More definitive studies are needed in all species as to whether decreased metabolic rate contributes to hypoxic adaptation.


This work has been supported by National Institutes of Health Grants NS-38632 and GM-066309.


View Abstract