In this review, the regional heterogeneity of the oxygen supply-consumption ratio within the heart is discussed. This is an important functional parameter because it determines whether regions within the heart are normoxic or dysoxic. Although the heterogeneity of the supply side of oxygen has been primarily described by flow heterogeneity, the diffusional component of oxygen supply should not be ignored, especially at high resolution (tissue regions ≪ 1 g). Such oxygen diffusion does not seem to take place from arterioles or venules within the heart, but seems to occur between capillaries, in contrast to data recently obtained from other tissues. Oxygen diffusion may even become the primary determinant of oxygen supply during obstructed flow conditions. Studies aimed at modelling regional blood flow and oxygen consumption have demonstrated marked regional heterogeneity of oxygen consumption matched by flow heterogeneity. Direct, non-invasive indicators of the balance between oxygen supply and consumption include NADH videofluorimetry (mitochondrial energy state) and microvascular PO2 measurement by the Pd-porphyrin phosphorescence technique. These indicators have shown a relatively homogeneous distribution during physiological conditions supporting the notion of regional matching of oxygen supply with oxygen consumption. NADH videofluorimetry, however, has demonstrated large increases in functional heterogeneity of this ratio in compromised hearts (ischemia, hypoxia, hypertrophy and endotoxemia) with specific areas, referred to as microcirculatory weak units, predisposed to showing the first signs of dysoxia. It has been suggested that these weak units show the largest relative reduction in flow (independent of absolute flow levels) during compromising conditions, with dysoxia initially developing at the venous end of the capillary.
Regional blood flow
Time for primary review 36 days.
The balance between oxygen supply and oxygen demand is of paramount importance for the heart since it determines whether the tissue is healthy or dysoxic (where oxygen need exceeds oxygen supply). Yet the assessment of this balance at a regional level has been exceedingly difficult to study. This is mainly due to the fact that both the regional oxygen supply to the myocyte, which is provided by diffusion, and the regional oxygen consumption are each difficult to assess separately, let alone simultaneously.
Oxygen supply is to a large part determined by convective blood flow and many excellent studies have been carried out on the heterogeneity of cardiac blood flow, demonstrating marked regional variations in flow within the heart [1,2]. However, because oxygen is delivered by convection and diffusion, the heterogeneity of oxygen supply is expected to differ from that of blood flow, especially when spatial resolution improves below 1-g tissue pieces. Such a discrepancy between flow supply and oxygen supply is supported by the observation of increased homogeneity of tissue oxygenation when diffusion between capillaries is incorporated in a mathematical simulation of the capillary circulation of the myocardium .
Regional oxygen consumption of the heart has primarily been determined by combining quick-freeze techniques with microspectrometric determinations of arterial and venous haemoglobin saturation . Promising new developments such as 15O-oxygen PET  and 13C-NMR  obtain estimates of regional cardiac oxygen consumption together with regional measurements of flow. However, these techniques are very elaborate, expensive and difficult to execute. Therefore, techniques which can give direct measures of the balance between oxygen supply and consumption are valuable tools, since such techniques will provide direct information about whether tissue cells are well oxygenated or dysoxic. Regional NADH videofluorimetry [7,8] and microvascular oxygen pressures measured by Pd-porphyrin phosphorescence  are two such direct indicators of the oxygen supply-consumption ratio. Since these indicators have been predominantly applied to the epicardium, heterogeneity in this review will be discussed in terms of regional heterogeneity of the epicardium, without discussing chamber or layer (endocardium versus epicardium) heterogeneity.
The aim of this paper is to review the current knowledge concerning both the regional supply side of oxygen as well as the consumption side of oxygen within the heart and compare this information with that obtained by direct indicators of the oxygen supply-consumption ratio. The heterogeneity of this ratio will be examined during normoxic and oxygen compromised conditions.
2 Flow heterogeneity
Micro-heterogeneity of flow has been established to a large extent by the use of microspheres since the early studies of Yipintsoi et al. , and confirmed by others (for review see Ref. ). An important feature of flow heterogeneity is that it increases as tissue piece size decreases . This can be accounted for by the tree-like coronary anatomy of vessels, from the large supplying arteries up to the terminal arterioles . Such fractal nature may explain the dependency of flow heterogeneity on piece size [12,13]. Being fractal, i.e. showing self similarity upon scaling, suggests increased flow heterogeneity as resolution is increased due to just one, similar, recurrent determinant of heterogeneity (such as branching patterns) for all piece sizes. However, the non-tree-like topology of capillaries  results in a breakdown of the fractal nature of flow heterogeneity and consequently in a different dependency of heterogeneity on piece size . This may explain why flow heterogeneity measured at a resolution <1 mm2 is much smaller than predicted from the relation between heterogeneity and piece size for resolutions much larger than 1 mm2. Although coronary anatomy is a primary determinant of flow heterogeneity, vascular tone also influences blood flow distribution . In this respect, it is important to note that flow heterogeneity (determined by deposition of molecular markers or indicator-dilution experiments) decreases with reductions in arterial O2 tensions at constant flow, probably as a result of decreased vascular tone [15,17]. In contrast, most studies directed at the effects of ischemia on heterogeneity have observed an increase in heterogeneity (determined by microspheres) with decreases in flow [18–21]. However, effects of hypoxia and ischemia on flow heterogeneity are critically dependent on sample size  which may be one explanation for the conflicting results concerning effects of hypoxia and ischemia on flow heterogeneity: the hypoxia studies have used sample sizes <0.01 g, whereas in the ischemia studies sample sizes were >0.01 g. Thus, regional flow heterogeneity is strongly dependent on sample size and condition of the heart (normal, hypoxia, ischemia). The important question remains, however, as to how this convective flow heterogeneity translates into the distribution of oxygen supply by diffusion.
3 Oxygen diffusion
3.1 Direct measurements of oxygen diffusion from the vasculature
In the first half of the 20th century, it was thought that diffusion of oxygen from the vasculature system occurred mainly from the capillaries . More recently, evidence has shown that oxygen diffuses to and from most elements of the vasculature system (thus not only from capillaries), and accounts for a significant portion of oxygen transport to the tissue cells . When going from large to small vessels in the arteriolar network, a longitudinal decrease in PO2 has been observed in the cheek pouch , cremaster muscle , mesenteric microvascular network , ileum  and dorsal skin preparation . In addition, vascular PO2 was also found to increase with increased vessel size in the venular network, with capillaries taking up oxygen from overlying venules [28,29].
All the aforementioned studies, however, have been conducted in tissue beds other than the heart, and very few studies can be found that have directly examined oxygen diffusion from large vessels within the heart. Weiss and Sinha  have observed no correlation between vessel size and arterial oxygen content, suggesting negligible oxygen diffusion from these vessels. Honig and Gayeski  found no change in haemoglobin saturation as a function of vessel size between 20 and 200 μm for arterioles and venules in frozen tissue pieces of rat and dog heart. It was concluded that due to high red blood cell velocity and short arteriolar length not enough time is available to allow a significant amount of oxygen to diffuse across the vessel wall. Their point is well taken: assuming the length of the cardiac arteriolar bed to be approximately 2 mm, while the diameter decreases from 200 to 20 μm , a decrease of only 6% in Hb saturation or 6 mm Hg in vascular PO2 can be calculated when extrapolating the measured amount of oxygen that diffuses from arterioles in the dorsal skin preparation . This anticipated small decrease would even be less when the higher flow rates in the arterial compartment of the heart, as compared to that in the dorsal skin preparations, are considered. Data shows, thus, that oxygen diffusion from large vessels is minimal for the heart due to the short flow path lengths and high flow rates.
3.2 Diffusive shunting of oxygen within the heart
Insight into the properties of O2 diffusion within the heart has been limited, due to the lack of appropriate techniques. The major available information concerning the nature and extent of diffusion within the heart has been obtained from experiments that have focused on a consequence of diffusion, i.e. diffusive shunting of compounds within the heart.
Diffusive shunting is defined as the bypassing of a part of the vascular system by diffusion, e.g. arteriovenous (direct diffusion from the arteriole to the venule) or inter-endcapillary (release and uptake of oxygen by different capillaries). This type of shunting can be envisaged by comparing the emergence function (for peak height or time of arrival) of compounds with different diffusibilities. For example, diffusive shunting is indicated when a diffusible compound has a larger peak height or earlier time of arrival in the venous effluent than a non-diffusible compound, which remains intravascular.
One of the first studies to examine diffusional shunting, in an isolated blood-perfused dog heart, found that the ratio of the venous dilution peak of the compound with the highest diffusibility (tracer water) to the compound of lower diffusibility (antipyrine) increased with lower flow-rates (<1 ml min−1 g−1) . This data demonstrated the presence of a diffusional shunt, even though the shunting fraction of water amounted to less than 3%. Roth & Feigl  examined diffusional shunting in closed-chest dogs with controlled coronary blood flow (0.21 to 1.17 ml min−1 g−1) and compared venous appearance times of hydrogen with intravascular indicators. At low flow (0.2–0.8 ml min−1 g−1), less than 0.1% of injected hydrogen preceded the intravascular, non-diffusible indicators. Because the diffusion constant of hydrogen is approximately twice that of oxygen, this data indicated, indeed, a very small degree of oxygen shunting. This finding has been corroborated by the observation that the transport of oxygen always lags behind the transport of an intravascular, non-diffusible compound during normal or high flow conditions in rabbit and dog hearts [35,36]. In a subsequent study by Bassingthwaighte et al.  it was concluded that, while the shunting of heat in isolated blood-perfused dog heart is great (diffusion constant of heat is two orders of magnitude higher than oxygen), the shunting of oxygen will not be large. It can be concluded from the above studies that arteriovenous or end-capillary shunting of oxygen within the heart is negligible at normal, physiological flows (∼ 1 ml min−1 g−1), and increases only marginally at low, pathological flows (<0.5 ml min−1 g−1). These findings may also explain the earlier findings by Steenbergen et al.  that regional anoxic zones in the isolated perfused rat heart appear at higher effluent PO2 during ischemia (low flow, normal PO2 perfusate) as compared to hypoxia (normal flow, low PO2 perfusate). Although this may be explained by differences in pH, there is also the possibility that due to the lower flow during ischemia, there is more time during ischemia than during hypoxia for diffusional shunting to occur. This ‘shunted’ oxygen is not used by the myocyte (anoxic zones) and increases the venous effluent PO2.
Analysis of emergence functions in whole heart studies, however, does not have the sensitivity to exclude the presence of shunting that bypasses only a small part of the vasculature system, i.e. between capillaries at less extreme ends of their length. That such diffusion probably takes place has been suggested by Wolpers et al. , where it was found that a model which allowed bypassing of a small part (<300 μm) of the capillary functional length (500 μm) best described the outflow curves of inert gases in closed-chest dogs. In comparison, models that completely bypassed the capillary bed (thus arteriovenous shunting) generated poor descriptions of the outflow curves.
In this context it is interesting to note that model analysis of the transport of water or oxygen through the heart, that does not take diffusional shunting into account between capillaries, resulted in low permeability estimates for water and oxygen, which decreased as flow was decreased [40,41]. These low permeabilities were explained by the suggestion that the capillary and the sarcolemmal membranes constitute a significant barrier to water and oxygen [36,42], in contrast to previous findings which suggest water transport to be mainly flow-limited . However, when diffusional interactions between capillaries is allowed in the models, by either using high diffusion constants in the model  or actually constructing interactions between capillaries , normal permeabilities were sufficient to describe the experimental data.
In conclusion, most data indicate that diffusive shunting of oxygen within the heart primarily takes place between capillaries and not between larger vessels. That diffusion between large vessels does not seem to take place within the heart, is in agreement with the limited data which shows no detectable oxygen loss from these vessels . The diffusion of oxygen between capillaries causes oxygen supply at the capillary level to be more homogeneous than oxygen supplied by convective flow alone, and may have important functional consequences during restricted oxygen supply conditions.
4 Discrepancy between blood flow and oxygen supply
Blood flow can be considered as the convective transport of blood through the vascular system, and perfusion as the transport of solutes (such as oxygen) by blood flow and diffusion. As such, regional blood flow has been conventionally estimated by infusion and counting of the number and distribution of particles such as microspheres that plug a minor part of the vasculature. Only small biases exist toward higher flow in high flow regions, and lower flows in low flow regions, especially when sample region size decreases . During normal physiological flow conditions, diffusible tracers also give approximately similar perfusion heterogeneities as determined by microspheres, although there is a tendency for higher heterogeneity with microspheres as compared with diffusible tracers . Under conditions of restricted flow, however, perfusion as determined by diffusible tracers is larger than flow determined by microspheres [45,46]. These findings are corroborated by studies examining the transition zone between the normoxic zone with normal blood flow (measured by fluorescein angiography) and the anoxic zone (marked by NADH fluorescence) with no flow . This transition zone was still normoxic despite the lack of blood flow and had a length of about 300 μm. Thus through diffusion alone, oxygen was supplied over a distance of 300 μm. Together, this data indicates that oxygen supply is probably more homogeneous and larger than blood flow in areas of restricted blood flow, due to the diffusional aspect of oxygen supply.
Wieringa et al. developed a network model of the myocardial microcirculation to evaluate microheterogeneity of tissue or capillary PO2. An assumption in the model was the presence of homogeneous oxygen consumption, so that effects of blood flow and different types of oxygen diffusion could be studied. This model demonstrated that intercapillary diffusion (an aspect not incorporated in the classical Krogh's cylindrical tissue model) resulted in more homogeneous capillary and tissue oxygenation. These simulations showed larger heterogeneity for flow than for oxygen and indicated that oxygen supply at the level of the capillaries cannot only be determined by convective blood flow.
Summarising, the diffusion component of perfusion results in oxygen supply being more homogeneous than blood flow at the high spatial resolution of capillaries during normal conditions. During restricted flow conditions, the impact of diffusion is probably extended to large tissue areas such that these areas may still be perfused although minimal blood flow is present. This illustrates that blood flow is not always a good indicator of oxygen supply, and shows the difficulties in determining the supply side of oxygen.
5 Regional myocardial oxygen consumption
One of the first studies to provide quantitative information on transmural myocardial oxygen consumption determined haemoglobin saturation of arterial and venular vessels (20–200 μm) in combination with regional blood flow determination by microspheres in frozen tissue samples of the dog heart . Subepicardial oxygen consumption was found to be approximately 20% lower than subendocardial oxygen consumption. These techniques have however not been used to examine microheterogeneity of oxygen supply-demand. It should also be noted that this technique is not without criticism: tissue sample freezing takes at least several seconds and is probably heterogeneous which may cause regional changes in venous oxygen content as a result of irregular beating tissue where metabolism continued and venous blood might redistribute .
Only recently has the direct examination of regional myocardial oxygen consumption been pursued, the interruption probably related to the difficulties of developing reliable techniques. Regional oxygen consumption has been obtained for the in situ dog heart, combining 15O-positron emission tomography with an axial distributed blood–tissue exchange model . Following intravenous injection of 15O-water for blood flow determination, the model estimates regional oxygen consumption from the regional time–activity curve of inhaled 15O-oxygen for regions as small as 0.5 g. Correlations of r=0.5–0.7 were observed among regional blood flow and oxygen consumption. Other studies have also demonstrated significant correlations between regional blood flow and metabolism: between regional flow (microspheres) and regional oxygen consumption calculated from the tricarboxylic acid cycle rate using 13C-NMR and mathematical modelling in frozen tissue samples of isolated rabbit hearts ; between regional blood flow (microspheres) and regional glucose uptake measured by 3H-deoxyglucose deposition in frozen tissue samples of in situ dog hearts ); between regional flow (microspheres) and regional O2 consumption estimated from H218O residue counting in frozen tissue samples of isolated rabbit hearts .
Both the 15O-PET and 13C-NMR techniques are state-of-the-art, highly developed, analysis tools to obtain quantitative data on regional myocardial oxygen consumption. However, most of the new techniques are very elaborate, time-consuming methods, that are also either expensive (PET) or destructive (13C-NMR, deoxyglucose- and H218O-accumulation), thus not allowing continuous measurement of regional oxygen consumption. It is also difficult with these techniques to obtain direct information concerning the functional heterogeneity (whether tissue regions are ischemic). In addition, no complete determination of the oxygen supply-consumption ratio is feasible with these techniques, because in most cases (studies using microspheres) only the convective supply side of oxygen is measured. The omission of the diffusive component will become increasingly important during restricted oxygen supply conditions and when piece sizes smaller than 1 g are examined: e.g. assuming that oxygen can travel a distance of 300 μm by diffusion, this oxygen can affect 11% of the volume of a 0.5 g tissue piece. Nevertheless, these techniques have collectively shown that oxygen consumption is heterogeneously distributed throughout the heart, which is, to a large part, matched to flow heterogeneity: low flow regions having low oxygen consumption, high flow regions having high oxygen consumption.
6 The oxygen supply and consumption ratio
6.1 Direct measurements of the oxygen supply-consumption ratio
When supply becomes limiting compared to consumption, and no downregulation of metabolism occurs such as seen during hibernation, the myocyte becomes dysoxic and the PO2 in the capillary/venous compartment stabilises at a very low or even zero level. In that case, parameters that reflect the transition to the dysoxic state of the myocyte are of functional importance. The term dysoxia, which is defined as O2-limited cytochrome turnover, is preferred because the prefix ‘dys’ describes both oxygen supply and consumption . This in contrast to terms such as ischemia (literally meaning ‘restrained flow’) or hypoxia (meaning low oxygen), which have been shown to cause confusion through the definitions .
A parameter that is viewed as the gold standard for the condition of dysoxia within cells is the mitochondrial NADH/NAD+ redox state [7,8]. NADH is visible due to its fluorescence upon excitation with ultraviolet light, whereas NAD+ is not [52,53]. The redox state can be measured in vitro by videofluorimetry and after suitable correction for the optical properties of blood in vivo by dual wavelength videofluorimetry . That NADH reflects the balance between supply and consumption has been shown by Steenbergen et al. : NADH fluorescence increased in isolated perfused rat hearts with tachycardia during constant low flow. It should be noted, however, that NADH levels have a low sensitivity to changes in oxygen supply or demand during normoxic conditions, along with a strong dependency on substrate usage [55,56]. NADH fluorimetry is, however, a sensitive indicator for the O2 balance during compromised O2 supply conditions.
The recent development of an optical, non-invasive technique now makes the determination of vascular oxygen tension possible [9,57]. This technique is based on oxygen-dependent quenching of the phosphorescence of Pd-porphyrin, a compound intravascularly administered to the animal. Following excitation by a flash of green light, the measured half-life of the phosphorescence signal can be quantitatively related to the oxygen tension, using proper calibration of the probe . Binding Pd-porphyrin to albumin has the advantages of confining the probe initially to the vascular compartment and increasing the half-life of phosphorescence such that physiological oxygen tensions can be measured. We have shown that when this phosphorescence PO2 is measured with optic fibres in the microcirculation of rat intestine, it primarily mirrors capillary and venular PO2 values . Therefore, the PO2 determined with the fibre phosphorimeter has been called microvascular PO2, μPO2. Most importantly, we directly demonstrated here that the μPO2 is an out measure of the balance between oxygen supply and consumption in Langendorff-perfused pig hearts (Fig. 1). An increase of heart rate (↑O2 consumption) at constant flow (constant O2 supply) decreased the μPO2, whereas a decrease in flow (↓O2 supply) at constant mechanical performance (constant O2 consumption) also decreased the μPO2 (Fig. 1). Although other techniques that determine myocardial oxygenation may also partly reflect the balance between O2 supply and consumption, such as surface  and tissue penetrating oxygen electrodes, the undefined origin (vascular, cellular, interstitial) of these electrode PO2's hampers a clear interpretation of these signals and, thus, will not be further discussed here.
Dependency of epicardial microvascular PO2 (μPO2) determined by the phosphorescence technique on oxygen consumption and supply. The μPO2 is determined on the surface of the left ventricle of an isolated Langendorff-perfused pig heart . The heart was perfused at constant flow with a blood–Tyrode's mixture (Hct±20%) containing 220 μM Pd-porphyrin, gassed with room air/5% CO2 (arterial PO2=110 mm Hg). The coronary vasculature was maximally vasodilated with nitroprusside. (A) The μPO2 as a function of O2 consumption at constant O2 supply (flow). O2 consumption was changed by varying the pacing frequency from 100 to 180 beats per min (bpm) as indicated in the figure. (B) The μPO2 as a function of O2 supply at constant O2 consumption. O2 supply was changed by varying the retrograde flow using a roller pump. Flow was varied between 3.4 and 2.1 ml min−1 g−1 as indicated in the figure.
A limitation of using these non-invasive surface illumination techniques is the restricted penetration depth of the used wavelengths (NADH: ∼0.36 mm ; Pd-porphyrin: ∼0.5 mm ). However, studies have shown that the heterogeneous epicardial NADH pattern is characteristic of the whole ventricle, for both small hearts  and large hearts [64,65]. In addition, these techniques may ideally be suited for the smaller heart size if most of the heart tissue needs to be investigated, such as the mouse heart, and will thus be an important analysis tool in the realm of molecular physiology.
6.2 NADH fluorescence imaging during dysoxia
The first NADH fluorescence measurements demonstrated the possibility of visualising dysoxic areas in perfused rat hearts . In a classical study  it was subsequently shown that these high fluorescence, dysoxic areas, were heterogeneously distributed over the surface of the heart during high flow graded hypoxia, graded ischemia and respiratory acidosis. In addition, the location and size of the dysoxic areas was similar during high flow hypoxia and ischemia (but not during acidosis) and did not vary over a period of 15 min under constant flow conditions. Repetitive periods of identical ischemia generate identical distribution patterns for a given heart . It was further demonstrated that the location of these patchy dysoxic areas was also identical during recovery from either total global ischemia or high flow anoxia or tachycardia for a given heart . This heterogeneity in dysoxic areas has also been shown during inhibition of NO synthesis in endotoxemic rat Langendorff-hearts , in hypertrophied rat Langendorff-hearts  and during haemorrhagic shock in in situ pig hearts .
Although NADH measurements have mostly been used to evaluate dysoxia in compromised hearts, most studies have also reported that normoxic perfusion shows a relatively homogeneous tissue fluorescence [38,66,67,70]. These results should be cautiously interpreted due to the low sensitivity of NADH to the O2 balance during normoxic conditions, but they do indicate matching of oxygen supply to oxygen consumption, in agreement with recent measurements of regional oxygen consumption with flow [5,6,48,49].
6.3 Size and origin of the microcirculatory (weak) unit
The benefit of visualisation of indicators of the oxygen supply-consumption ratio is that the structure and size of the dysoxic areas during the progress of dysoxia can be observed continuously, and may thus provide insight into the main determinant of dysoxia at each level of dysoxia (capillary ↔ arterioles, convection ↔ diffusion). An interesting observation is that the high NADH areas are always in the same location, independent of whether regional dysoxia was induced by hypoxia, ischemia or increased atrial pacing [38,66,67]. This strongly suggests that certain areas within the heart are predisposed to becoming the first dysoxic when oxygen supply is limited. These areas have been called microcirculatory weak units . Using surface NADH fluorescence photography, in the isolated working rat heart, it has been observed that anoxic zones 100–300 μm wide developed during the progression of both hypoxia and ischemia . The authors suggested from the size of the anoxic regions (larger than intercapillary distances), that oxygen supply be regulated at the level of the arterioles. However, Ince et al.  concluded from the use of microspheres of different sizes that the regulatory unit determining the patchy pattern during compromised oxygen supply was located at the level of the capillary (Fig. 2A and 2B). In an elaborate study using an in situ rat heart preparation in combination with quick freezing techniques and determination of NADH fluorescence in 5-μm thick sections, Vetterlein et al. showed that the size of the anoxic zones ranged from a few myocytes to several hundred microns . The size of the anoxic zones increased with further restriction of blood flow through the left anterior coronary artery. It was found that the dysoxic regions had more venous capillary segments than arterial capillary segments, as compared to non-dysoxic regions. This data would indicate that dysoxia mainly develops along the venous end of capillaries. In addition, it has been suggested by these authors  that the large areas of organ surface NADH fluorescence that have been observed  may be due to the interference from the fluorescence arising from layers beneath the focused plane of observation, which would obscure the visualisation of small areas of NADH fluorescence. In summary, data suggests that hypoxia/ischemia result in dysoxia, which starts at the level of the capillary [67,70], and which occurs primarily at the venous end of the capillary .
Heterogeneous NADH fluorescence images of left ventricle area of isolated Langendorff-perfused rat hearts, with apex of the heart always at the bottom of the figure: (A) NADH fluorescence patterns elicited during transition from anoxia (95% N2) to normoxia (95% O2) are, as shown in (B), the same as those elicited by embolization with 5.9-μm diameter microspheres . To aid comparison of the two images, corresponding areas in the two images are numbered. (C) Low homogeneous NADH fluorescence during pH 7.5 perfusion, whereas large NADH fluorescence areas (indicated by arrows) are visible during pH 7.0 perfusion (D). (E) and (F) show examples of hypertrophied hearts demonstrating the large areas (indicated by arrows) of high NADH fluorescence within 5 min after the start of perfusion . (Figs. 3A and 3B are reprinted with permission from Am J Physiol, Figs. 3C–F are reprinted with permission from Biochim Biophys Acta).
Hypertrophy [69,72] and pH [38,72] have been shown to generate a different distribution and larger areas of high fluorescence than interventions using decreased oxygen supply (Fig. 2C–F). Interestingly, superoxide dismutase was able to reduce these large areas of high fluorescence , suggesting that such large areas are determined by a vasoregulation mechanism at the level of the arterioles. This is in accordance with the observation that the larger areas of fluorescence occur during embolization with microspheres >10 μm which occlude arterioles . Thus, acidosis and hypertrophy are associated with dysoxia at the arteriole level, generating dysoxia in the relatively large area of the capillary bed of the arteriole.
Anterior view of three open-chest pig hearts showing the heterogeneity of μPO2 on the epicardial surface, as determined by the phosphorescence technique. The μPO2 was measured in nine different regions (∼1 cm2) for each heart. The animals were ventilated with 33% O2, resulting in an arterial PO2 of ±180 mm Hg. Preparation as in van Iterson et al. [74,75]. Abbreviations used: RA=right auricle; LA=left auricle; LAD=left anterior descending artery.
6.4 Microvascular PO2 measurements
In the first report of applying the phosphorescence technique to the heart, lifetime images of the phosphorescence signal were obtained through a microscope from the epicardial surface of new-born piglets . The microvascular PO2 of 0.3-mg regions was 16.8±4.2 mm Hg, at an arterial PO2 of 106 mm Hg. The coefficient of variation (CV=SD/mean), which is an index of μPO2 heterogeneity, was 25%, much smaller than the CV obtained for flow heterogeneity in such small tissue samples . In a study directed at resuscitation [74,75] in a model of shock in the pig, the μPO2 was determined with the phosphorescence technique by a fibre phosphorimeter [58,62] in nine different areas (∼1 cm2 or 0.04–0.08 g of tissue) on the surface of the ventricles (Fig. 3). The CV for these μPO2 values was between 11–12%, which again is considerably smaller than the CV of flow heterogeneity (30%) for equally sized tissue samples . Thus, the smaller CV's indicate a matching of O2 supply with consumption within the heart. Very few studies have used the phosphorescence technique to evaluate myocardial μPO2 and its heterogeneity during conditions of restricted O2 supply. It has been shown, however, that during the transition from high-flow anoxia to normoxia, a very heterogeneous μPO2 distribution occurs on the surface of the isolated rat heart . The μPO2 patterns coincided with the NADH patterns that were also recorded, supporting the notion that both NADH and μPO2 reflect a similar phenomenon, i.e. the O2 supply-consumption ratio.
Various studies have shown that oxygen diffuses to cells from arterioles, capillaries and venules. Within the heart, both direct, but limited, data on oxygen diffusion from large vessels and indirect data on diffusional shunting suggest that the release or uptake of oxygen by large vessels is insignificant. However, more research examining the PO2 along a considerable length of artery or arteriole within the heart is needed to answer this question conclusively. Furthermore, it also remains uncertain whether the regions that show the first signs of dysoxia have any relation with the normal local level of flow, flow reserve, O2 diffusion or O2 consumption. Two recent studies have shown that the resting normal, low and high blood flow regions are equally vulnerable to dysoxia during partial coronary stenosis [18,76], and that the relative flow reduction during stenosis was independent of low or high blood flow region . This data is corroborated by the observation that no correlation exists between the intensity of NADH fluorescence and the distance to the next perfused capillary , thus there seems to be no dependence of dysoxia on the absolute level of convective flow. Assuming matching between flow and O2 consumption, this would also indicate equal vulnerability for regions of low and high O2 consumption. Dysoxia was mainly predicted by the relative flow reduction upon stenosis , implying a vasoregulatory (at the level of the arterioles or higher) mechanism as a primary determinant of dysoxia. However, other studies have shown that these dysoxic areas can be best generated by obstructing capillaries, but not arterioles, with microspheres . It remains to be examined whether the obstructed capillaries mainly occur in the regions with the largest relative reductions in flow, indicating dysoxia develops at the capillary level of those areas.
The studies reviewed in this paper emphasise that it remains unknown whether a dependency may exist between dysoxia and oxygen supply by diffusion. Assuming diffusive shunting of oxygen within the capillary bed, for which evidence is present, it may be hypothesised that capillary regions where oxygen is shunted away from their microcirculation become the first dysoxic. The observation that the NADH fluorescent zones appear at higher venous PO2 during ischaemia as compared to high-flow hypoxia  is in support of this hypothesis. Why certain capillary regions are more prone to diffusional shunting is unknown, however, and may be related with capillary geometry (capillary length or number of cross-connections?).
The evidence in support of regional matching between oxygen supply and oxygen consumption within the heart is rather strong. The question that arises is how such matching occurs. One interesting explanation may be related to the recently observed adaptation of oxygen consumption to extracellular oxygen tensions in hepatocytes and embryonic cardiomyocytes [77,78]. Lowering the oxygen tension in suspensions of isolated cardiomyocytes decreased oxygen consumption within 5 min in these embryonic cells, offering a mechanism for regional matching of oxygen supply-consumption within the heart. This hypothesis may be tested in experiments directed at hibernation: the relation between regional oxygen supply and consumption is compared at different time points of the reduced flow condition, to study whether at some time point the heterogeneity of O2 supply-consumption ratio equals the heterogeneity observed during the control condition.
We gratefully acknowledge the skillful assistance of Charles N.W. Belterman and Joris de Groot of the Department of Experimental Cardiology of the University of Amsterdam with the isolated blood-perfused pig heart experiment.