Dynamics of blood oxygenation gives better insight into tissue hypoxia than averaged values

normal cellular function requires a continuous supply of oxygen. In higher animals a cardiovascular system is used to provide for a convective delivery of oxygen to the smallest branches of the network of blood vessels—the microcirculation—from which oxygen then passes to the cells through passive diffusion. The oxygen is transported through the cardiovascular system by hemoglobin (Hb), an iron-containing protein in the red blood cells.

The Hb oxygen saturation consists of a steady level and fluctuations around it, much like blood flow (5, 7). The characteristics of the steady contribution to oxygen transport have been studied extensively in vitro since the pioneering work of August Krogh ninety years ago (10, 11). He measured the diffusion coefficient of oxygen through tissues and the capillary density of these tissues. These data were then used in a mathematical model of oxygen transport that became known as the Krogh cylinder model. With the assumption of oxygen mass balance and a cylindrical vessel, the model provides a relationship between oxygen flow, saturation, and pressure. Because of its conceptual simplicity and ease of application, this model has remained a mainstay in the field up to the present day, even though some of its assumptions can be criticized as unrealistic (15).

In higher animals, oxygen deprivation rapidly leads to disability and death, so their cardiovascular systems are highly regulated, delivering oxygen as needed to meet the tissue requirements. Our understanding of the regulatory mechanisms has been limited by uncertainties in the oxygen concentration dependence of cellular metabolism and in the intracellular oxygen concentration. Several methods have been used in an effort to measure these parameters, including oxygen electrodes, oxygen binding to myoglobin, oxygen-dependent quenching of fluorescence, changes in the electron spin resonance of spin labels, and chemiluminescence from a bacterium (23).

The methods that have been used to study oxygen transport, especially recently developed methods for noninvasive imaging of tissue hypoxia (25), are beginning to be applied to pathophysiological conditions. A comparison with the knowledge that has been gained under normal physiological situations shows that there are significant alterations in oxygen transport under these circumstances.

The physics of oxygen transport is particularly important for studies of cancer. Differences in oxygen delivery in tumors and/or cancer cells manifest as low intercellular oxygen tensions (7). Because this characteristic enhances both the survival and aggressiveness of cancer cells (24), it can adversely impact on some forms of treatment and it has potential for the detection and monitoring of cancers. In addition to tumors, the deficiency in oxygen supply and blood flow regulation accompany many other physiological and pathological states. Areas that have received attention include aging (16), chronic heart failure (4), diabetes (6, 22), hemorrhagic shock (20), hypertension (9), malaria (8), obesity (6), postacute myocardial infarction (1), sickle-cell disease (2), sepsis (3, 21), and wound healing (13). The application of modern microcirculatory approaches to the study of oxygen transport in these important disease states is only beginning, and much remains to be learned.

A promising way forward can be seen by studying the fluctuational component of blood oxygenation and its dynamical properties. This is what Thorn et al. (19) now describe. They set out to establish whether or not mean blood oxygen saturation (S_mbO₂) is a reliable indicator of tissue oxygenation. They approached the problem by applying optical reflectance spectroscopy and near infrared spectroscopy, enabling them to monitor S_mbO₂ and the tissue oxygenation index. They observed large, spontaneous, oscillatory variations in S_mbO₂, to which they applied Fourier analysis to establish the characteristic frequencies. The physiological origins of the different frequencies were already known from earlier work on blood flow (12). One of several interesting conclusions drawn by Thorn et al. (19) is that there are two distinctly different processes giving rise to spontaneous decreases in S_mbO₂. The first of these apparently results from changes in arterial blood volume and could be linked to the effects of respiration, myogenic, sympathetic, and endothelial activity, but without any corresponding change in the deoxy-Hb level. The second process could be linked to endothelial and sympathetic activity only and was marked by a fall in oxy-Hb and a rise in deoxy-Hb. Hence, the measurement of a given fall in S_mbO₂ might correspond either to a change in the oxygen delivery from blood to tissue (the latter process) or to no change (the former), thus answering the original question: S_mbO₂ is not, in fact, a reliable indicator of tissue oxygenation.

However, the understanding of the role of the microcirculation in the supply of oxygen to tissues is still far from complete. One way forward is through computational modeling. Yet, without a deeper understanding of the underlying mechanisms of regulation, the models remain far from realistic. One important issue is the connection between nitric oxide (NO) and oxygen transport. Unlike oxygen, which has one primary source, the red blood cells, and one primary sink, the mitochondria, NO has multiple sources and sinks, most of which have yet to be uncovered.

The multitude of interactions results in a continuous fluctuation around the steady values. The article by Thorn et al. (19) is groundbreaking in that it is one of the first to exploit the information contained in these oscillatory fluctuations and to reach important conclusions on this basis.

By addressing the fluctuations in both oxy-Hb and deoxy-Hb measured in healthy humans in vivo, the authors have not only pioneered this application of optical reflectance spectroscopy but are also paving the way to physiological insight into the dynamical properties of oxygen transport in the unperturbed microcirculation. This is a very big step, as our current understanding comes mainly from in vitro studies and deals with the steady-average values only. The task of decoding the fluctuations is difficult, however, and faces a number of potential obstacles.

First, one has to have a measurement technique that senses the physiological process in question selectively. Noise is often inevitable, but any instrumental noise should be clearly distinguished from the natural fluctuations of the underlying physiological process.

Second, an appropriate measurement technique should have a small enough time constant that the natural dynamics are captured and not inadvertently averaged out.

Third, the sampling frequency should be high enough, and the time of observation long enough, for appropriate reconstruction of the dynamics. As we are necessarily restricted to a finite time of observation, we should be aware of the boundaries within which the dynamics can be reconstructed.

Fourth, the selection of the mathematical method by which the dynamical properties are reconstructed is crucial. One can apply modern complexity methods to quantify the amount of order or disorder in the recorded signals. However, most of these methods are very sensitive to the amount of information in the recorded data sets, i.e., on their lengths. Because real data sets are often short, such methods can easily end up predominantly quantifying noise. Of course, time domain analysis is in principle the easiest and most straightforward to interpret. But very few processes consist of single-frequency periodic oscillations. More often the signals on longer time scales just look like noise. Frequency analysis is a golden approach for gaining an initial insight into practically any signal, and the Fourier transform has long been a standard method. It is easy to apply, straightforward to interpret, and mathematically well elaborated. Thorn et al. (19) apply the Fourier transform to their recorded data. They then use the attributions of physiological origin of the frequency intervals obtained earlier by wavelet transform (17, 18), thereby opening up a whole new page in our understanding of the processes associated with oxygen transport. As the frequency intervals to date have been associated with neurogenic, myogenic, NO-dependent, and NO-independent endothelial activity, the study clearly shows potential for a multitude of future investigations. The coherence between blood flow and oxygen transport, the effect of temperature, and other vasodilation and vasoconstriction conditions should now be revisited with the oscillations in mind.

Fifth, we should bear in mind that some of the characteristic frequencies differ between species, so that we should proceed with caution, such as, in attempting to extrapolate results from rats to humans.

We can expect that the work by Thorn et al. (19) will pave the way to investigations of numerous consequences of rhythmical flow motion for nutrition and the oxygen supply to tissue, which are currently largely unknown. Interestingly, the subject has been recently addressed by another community; namely, functional MRI and in particular blood oxygen level-dependent (BOLD) MRI studies are repeatedly reporting oscillations other than that of cardiac or respiratory origin in the BOLD signal (14). Apart from the study by Li at al. (13), where the periodic variations of Hb oxygenation in compressed and uncompressed skin were evaluated with a reflection spectrometer using an in vivo Sprague-Dawley rat model, the study by Thorn et al. (19) is the first investigation of this kind to be performed on humans. It will undoubtedly be followed by many more as the nascent subject of nonlinear vascular physiology comes fully into its own.

GRANTS

The publication was supported in part by the Wellcome Trust, UK, and by the Slovenian Research Agency.