Oxygen binding by single red blood cells from the red-eared turtle Trachemys scripta

Transcription

1 J Appl Physiol 90: , Oxygen binding by single red blood cells from the red-eared turtle Trachemys scripta SEBASTIAN FRISCHE, 1 STEFANO BRUNO, 2 ANGELA FAGO, 1 ROY E. WEBER, 1 AND ANDREA MOZZARELLI 2 1 Danish Centre for Respiratory Adaptation, Department of Zoophysiology, University of Aarhus, 8000 Aarhus C, Denmark; and 2 Institute of Biochemical Sciences, University of Parma, Area Parco delle Scienze, Parma, Italy Received 8 August 2000; accepted in final form 2 December 2000 Frische, Sebastian, Stefano Bruno, Angela Fago, Roy E. Weber, and Andrea Mozzarelli. Oxygen binding by single red blood cells from the red-eared turtle Trachemys scripta. J Appl Physiol 90: , Oxygen-binding properties of single red blood cells from the red-eared turtle Trachemys scripta were measured by microspectrophotometry to describe the variation in oxygen affinity of red blood cells and to gain insight into the distribution of functionally different hemoglobins among red blood cells. Methodologically, this study represents the first report on the cell-to-cell variation in oxygen-binding properties based on oxygen-binding curves of single vertebrate red blood cells. The cells differed significantly with respect to oxygen affinity. Mean oxygen pressure at half saturation of the cells in a blood sample was found to be (SD) Torr. The distribution of oxygen affinities among red blood cells is unimodal, indicating that the two hemoglobins found in turtle blood are not segregated in distinct cells. Therefore, the functional interaction shown by these hemoglobins in vitro is likely to take place in vivo. The considerable variation in oxygen affinity between individual red blood cells calls for its incorporation in models of tissue oxygenation. oxygen transport; microspectrophotometry; hemoglobin ALTHOUGH THE DESCRIPTION OF vertebrate red blood cells as bags of hemoglobin is often a useful generalization, these very specialized and simplified cells may show significant variability in their biological function as oxygen carriers. This variability allows red blood cells to continuously adjust their properties in response to changes in factors like oxygen tension (PO 2 ) (27), carbon dioxide tension (PCO 2 ), osmolality, temperature, season (21), and concentration of hormones (14, 31). In contrast to that of humans, the blood of most vertebrate species contains several types of hemoglobin even in the adult state, thereby adding the distribution of hemoglobins among red blood cells to the list of factors that contribute to red blood cell variability. Moreover, the majority of vertebrates (e.g., fish, amphibians, reptiles, and birds) have nucleated red blood cells, which show high aerobic metabolic rates (22) and are capable of protein synthesis during circulation (30), making these cells even more complex than mammalian red blood cells. Two functionally different hemoglobins can be separated from the blood of the adult red-eared freshwater turtle Trachemys scripta (11). The hemoglobins differ in oxygen affinity, expressed as oxygen pressure at half saturation (P 50 ): 5.6 and 4.0 Torr (25 C, 100 mm HEPES, ph 7.3, and 100 mm KCl); moreover, there is a marked functional interaction in vitro. A 1:1 mixture of the two hemoglobins at similar conditions has a P 50 value of 10.0 Torr. Cooperativity [measured by the slope of the Hill plot (n)] changes were from 2.2 and 2.4 in the isolated hemoglobins to 2.0 in the 1:1 mixture (Frische, unpublished observations). For algebraic reasons, a mixture of noninteracting hemoglobins with different P 50 values will result in a reduced n (12). However, an increase in P 50 values is unexpected and indicates functional interaction between the hemoglobins. A prerequisite for this functional interaction to take place in vivo is the presence of both hemoglobins in the same cells. Because the hemoglobins show different affinities for oxygen, the distribution of hemoglobins among red blood cells will be reflected in the corresponding distribution of oxygen affinities. Apart from hemoglobin composition, the age profile of the circulating red blood cells affects cell-to-cell differences in oxygen affinity. Red blood cells are continuously formed and removed from the blood; the mean red blood cell life span in mammals is generally dependent on body weight, with the larger species having cells with longer life spans (33). Therefore, the cell-to-cell differences within a red blood cell population are often related to the activity of the different erythropoietic tissues, which may modify both the age profile of the circulating cells and their biochemical characteristics, e.g., hemoglobin composition (20). The red blood cells of turtles live for up to 11 mo (15) and may exhibit large age-related diversity. Moreover, turtle red blood cells have been proposed as a model for the evolutionary transition state between red blood Address for reprint requests and other correspondence: S. Frische, Danish Centre for Respiratory Adaptation, Dept. of Zoophysiology, Univ. of Aarhus, Universitetsparken Bldg. 131, 8000 Aarhus C, Denmark ( frische@biology.au.dk). The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact /01 $5.00 Copyright 2001 the American Physiological Society 1679

2 1680 OXYGEN BINDING BY SINGLE RED BLOOD CELLS cells relying on aerobic metabolism and the anaerobically metabolizing mammalian red blood cells (18), a transition that is homologous to that occurring in maturing mammalian red blood cells. To map the distribution of hemoglobins and to describe the variation of oxygen affinity in these longliving red blood cells, we measured oxygen-binding properties by microspectrophotometry of individual turtle red blood cells. All available methods for measuring blood oxygen dissociation curves measure mean values for a large number of cells (11). The possibility to obtain oxygen-binding curves on single cells by microspectrophotometry has been acknowledged for more than a decade (4, 17), but, to our knowledge, the technique has not been applied to vertebrate red blood cells. By determining oxygen affinity and cooperativity for individual cells, we can provide descriptions of the variation of these main functional properties in a population of nucleate red blood cells. MATERIALS AND METHODS Blood samples were taken with heparinized syringes from the coccygeal vein (26) of adult Trachemys scripta kept at 25 C. The blood was directly transferred to two gas-tight vials preequilibrated with 95% air and 5% CO 2. One vial was centrifuged for 5 min at 1,500 rpm to separate blood cells from plasma. The plasma was removed and centrifuged at 10,000 rpm for 5 min to remove all other cellular and particulate matter. Less than 1 l of blood from the second vial was diluted with 200 l of plasma. A drop of this dilute cell suspension was loaded in a Dvorak-Stotler flow cell (9), which was covered by a gas-permeable silicon copolymer membrane. The red blood cells settled one by one on the glass bottom and remained stationary for hours. The flow cell was mounted on the thermostated stage of a Zeiss MPM03 microspectrophotometer (24). The light transmitted through the red blood cell, and the surrounding solution was recorded using a circular measuring spot of 8 m in diameter. The spot was smaller than the average size of a single ellipsoid turtle red cell (19 11 m) (25). Spectra were recorded in the wavelength range nm at 1-nm intervals. Gas mixtures consisting of 5% CO 2, % O 2, and % He were prepared as follows. O 2 and He were mixed in discrete steps with a gas-mixture generator (Environics, series 200). The output gas was then mixed with He and CO 2 by a gas mixing pump (Digamix M301/a) to obtain 5% CO 2 in the final gas mixture. The final mixture was humidified by bubbling in a 0.9% NaCl solution and passed through the flow cell. The equilibration of the system to a new PO 2 required 20 min. All the experiments were carried out at 25 C and within 6.5 h after blood sampling. Absorbance spectra of fully oxygenated and deoxygenated red blood cells were recorded after equilibration with 5% CO 2-95% O 2 and 5% CO 2-95% He, respectively. The fractional oxygen saturation of hemoglobin was estimated by fitting a linear combination of these reference spectra and a baseline correction consisting of a horizontal offset and a variable slope to the observed absorbance spectra at intermediate PO 2 levels (23). The standard deviation of the fitted fractional saturation was estimated from the variances calculated in the fitting procedure and subsequent error propagation calculations (29). Data are given as means SD unless otherwise stated. Hematocrit was measured using glass capillary tubes centrifuged for 5 min at 12,000 rpm. The concentration of ATP in the blood samples was measured using the test kit from Sigma Diagnostics (no. 366A). In a control experiment, the concentration of ATP was monitored in a sample of blood in a vial connected to the outlet of the flow cell, during the oxygen-binding experiment. The erythrocytic ATP concentration was 2.32 mm initially, 2.27 mm after 3.5 h, and 2.08 mm after 6.5 h. RESULTS Oxygen saturation of a single cell. Four spectra of a single red blood cell at different oxygen pressures are presented in Fig. 1A. Two spectra were obtained at the same oxygen pressure (35.1 Torr), at the beginning and the end of the experiment, respectively. The two spectra gave nearly identical fractional saturation values of and , illustrating the reversibility of the oxygenation and the stability of the cell preparation during the oxygen-binding experiment, which typically lasted 5 h. The absorbance spec- Fig. 1. A: absorbance spectra of a single red blood cell exposed to PO 2 values of 35.1, 25.2, 15.4, and back to 35.1 Torr. B: comparison between the observed absorbance spectrum at 19.7 Torr O 2 and the fitted spectrum obtained from a linear combination of the oxy- and deoxy-reference spectra and baseline correction. Fractional oxygen saturation is estimated to be and root mean square The difference in absorbance ( ) between observed and calculated spectrum (dashed line) refers to the right y-axis.

3 OXYGEN BINDING BY SINGLE RED BLOOD CELLS 1681 Fig. 2. Distributions of fractional oxygen saturations at fixed PO 2 (25.2 Torr), PCO 2 (36.6 Torr), and temperature (25 C). Open bars: a population of 38 red blood cells. Gray bars: 11 measurements of a single cell. Black bars: 9 measurements of another single cell. Curves represent Gaussian distributions of the data presented in the histograms. tra are not identical because the position of the measuring spot is not the same, which reflects the variable thickness of the red blood cells. Figure 1B shows the observed spectrum at 19.7 Torr O 2 along with the spectrum calculated by linear least squares fitting, giving a mean fractional saturation of The SD of the mean fractional saturation was small [ (mean SD of 110 saturation measurements)], and the predicted 95% confidence interval of fractional saturation was To verify the accuracy of these intervals, the fractional saturation of two cells were measured repeatedly (9 and 11 times) at constant PO 2 (see Fig. 2). These cells had fractional saturations of and , resulting in 95% confidence intervals of and 0.027, respectively. These intervals are narrower than those reported from another microspectrophotometric system (10). This difference can probably be ascribed to our use of 71 absorbance values from the entire absorbance spectrum from 380 to 450 nm instead of the difference in optical densities at only two wavelengths. This fitting procedure significantly increases the overall precision of the measurement. Oxygen saturation in a red blood cell population. The variability of red blood cell oxygen affinity is evident from the distribution of fractional saturations of 38 cells at a fixed oxygen pressure of 25.2 Torr (Fig. 2). The mean fractional saturation was , and the corresponding 95% confidence limits were and The SD of the population data was about four times that of the repeated measurements of individual cells (see above). Thus it cannot be ascribed to measuring error but must be considered a reliable estimate of the intercellular variation. On the basis of an n value of 1.71 (6), these data indicate that the mean P 50 value for this cell population was Torr and the values of the two cells subjected to repeated measurements were and Torr. The distribution of saturations of the 38 cells is qualitatively unimodal. Assuming a normal distribution, we obtained a 2 value of 3.316, which is far from rejecting the null hypothesis of normality (0.7 P 0.95) (34). Oxygen binding of single cells. Hill plots of seven individual cells from two different turtles are presented in Fig. 3, and the corresponding seven sets of P 50 and n values are shown in Table 1. Table 2 presents the details of an analysis of covariance (ANCOVA) of the data from each sample. The ANCOVAs show the elevations of the regression lines within each sample to be significantly different (P in both samples), whereas the slopes of the plots (n) are not (P 0.75 in both samples) (35). This implies that the oxygen affinity differs significantly between the red blood cells of a given blood sample. Mean P 50 values of the two samples of cells are and Torr, and the mean n values of the two samples are and DISCUSSION This study represents the first report on cell-to-cell variation in oxygen affinity and binding cooperativity of vertebrate red blood cells. Literature values for P 50 and Hill s n are averages of many cells and may only be compared with the mean values of our single cell measurements. The average P 50 values found in this study, and Torr, are close to the value of 24.5 Torr reported for whole blood studies of Trachemys scripta at similar temperature and PCO 2 values (6). Similarly, the mean n values, and 2.13 Fig. 3. Hill plots showing oxygen-binding properties of 4 and 3 individual red blood cells from 2 turtles, one in A and the other in B. Symbol shapes indicate individual cells. Coefficients of determination [( y 2 resss)/ y 2 r 2 ] are as follows: ( ), (E), (ƒ), and ({) for A and ( ), (E), and (ƒ) for B. See text and Tables 1 and 2 for further statistical details.

4 1682 OXYGEN BINDING BY SINGLE RED BLOOD CELLS Table 1. P 50 and Hill s n calculated from the Hill plots in Fig. 3 Turtle a Turtle b { ƒ E Mean SD E ƒ Mean SD P n Symbols are from Fig. 3. P 50, O 2 pressure at half saturation. 0.09, are within the range of earlier reported values for Trachemys scripta and other chelonians (6). No directly comparable data exist for the variance of P 50 between cells or the distribution of fractional saturations at fixed PO 2 reported in this study. From a technical point of view, the only directly comparable study concerns the blood cells of the annelid worm Glycera dibranchiata (17), in which case the P 50 values at ph 7.4 and 20 C were and in the absence and presence of 20 mm KCN, respectively. Microspectrophotometry has previously been used to monitor fractional oxygen saturation of single red blood cells from trout to describe the distribution of hemoglobin with a root effect among red blood cells (5). The span of fractional saturations of individual cells at fixed ph and PO 2 was shown to be 0.12 in trout (from Fig. 1 in Ref. 5), but no statistics were given. In comparison, the range of saturations in Fig. 2 is Microspectrophotometric techniques have also been applied to monitor the kinetics of reversible binding of CO and O 2 to hemoglobin in single red blood cells (1, 2) and in a study of the diffusion of gases into red blood cells (8). In contrast to all prior microspectrophotometric studies, in which washed cells were resuspended in buffer to control ph, the cells used in this study were kept in natural plasma buffered with CO 2. The resultant data are thus directly comparable to traditional whole blood measurements. Table 2. Details of two ANCOVAs comparing the Hill plots presented in Fig. 3 Although a bimodal distribution cannot be completely excluded, the distribution of saturations (Fig. 2) appears unimodal and normal (0.7 P 0.95), indicating that the two functionally different hemoglobins (11) are not separated in distinct cells. Minor differences in the proportion of the two hemoglobins may influence the variance of the distribution of P 50 values among cells. In this respect, the results confirm the expression pattern of globin genes described for other adult vertebrates (5). Because differences in hemoglobin constitution of the red blood cells are unlikely to be involved, we must look for other factors to explain the significant differences in oxygen-binding properties of red blood cells found in this study. Cell age is the major generator of variance in red blood cell population parameters and is therefore a prime candidate for causing the observed cell-to-cell variance in oxygen affinity. Human red blood cells undergo gradual biochemical changes during their 120-day life span (7), such as loss of K, decreased enzyme activity, decreased 2,3-diphosphoglycerate (DPG) concentration, and increased hemoglobin concentration. These changes are accompanied by reduced cell deformability and increased density. The latter feature can be exploited in the traditional densitygradient centrifugation for separating red blood cells in age fractions (7). Similar age-dependent changes are shown in red blood cells from other mammals (33) and x 2 xy y 2 No. of Data Points b resss resdf Turtle a { ƒ E Regression Pooled Common Total Turtle b E ƒ Regression Pooled Common Total b, Slope of regression line; resss, residual sum of squares; resdf, residual degrees of freedom. Symbols are from Fig. 3.

5 OXYGEN BINDING BY SINGLE RED BLOOD CELLS 1683 in lower vertebrates, despite differences due to the presence of aerobic metabolism and hemoglobin synthesis during circulation (22, 30). In humans, the oxygen-binding properties of old red blood cells separated by centrifugation are characterized by a higher affinity for oxygen compared with young cells. The difference is mainly associated with decreasing erythrocytic DPG concentration with cell age (28). Other factors like glycosylation of hemoglobin and formation of methemoglobin in the cells may also contribute to the increased oxygen affinity of old red blood cells (28). We suggest that the differences in oxygen-binding properties between red blood cells documented in this study are paralleled by differences in the concentration of ATP, which is the dominant organic phosphate compound in Trachemys red blood cells (3) and which modifies the oxygen affinity of both hemoglobins (11). Turtle red blood cells have a much longer life span than their human counterparts; therefore, the magnitude of the age-dependent intercell variation in oxygen affinity is probably larger than in humans, although no conclusions can be drawn on the basis of the available data. Blood generally behaves as a homogeneous liquid with respect to oxygen transport, but, in the capillaries, oxygen diffuses to the surrounding tissue from cells that move in single file. Variation in oxygen affinity affects the amount of oxygen released from each red blood cell for a given decrease in PO 2. The cell-to-cell variation in oxygen-binding properties may therefore play a physiological role by influencing tissue PO 2 gradients and, at least locally, tissue oxygenation levels. Modeling has been an important constituent in the study of tissue oxygenation since the beginning of the last century. An extensive literature has grown from the original Krogh model (16), generally expanding the Krogh cylinder to deal with the effects of the particulate nature of blood, the shape of erythrocytes, variations in hematocrit, uneven temporal distribution of red blood cells, and anisotropic distribution of capillaries (13, 32, 34). In all models, P 50 is assumed to be constant for all red blood cells; however, from our results, it is clear that considerable variation exists between individual red blood cells with respect to oxygen affinity. This study therefore calls for modeling studies that include the consequences of variations in red blood cell oxygen affinity with respect to tissue oxygenation. Oxygen equilibria of single red blood cells may also prove useful in other areas of research. In diseases in which the erythrocytes are differentially affected, e.g., malaria, single cell measurements would be able to quantify the effects of oxygen binding on the cellular level. In the case of malaria, the oxygen binding of blood cells containing different stages of the parasite could be quantified. Because only a few microliters of blood is necessary, single cell oxygen binding could also be used to study variations in blood oxygen binding within natural populations of small animals. Another field in which the determination of oxygenation curves of single red blood cells may be useful is the testing of drugs affecting oxygen affinity. Agents that increase the oxygen affinity are of therapeutic benefit in sickle cell anemia, which is caused by the polymerization of deoxygenated sickle hemoglobin. The polymerization induces changes in the shape of the red blood cells and shortens their circulatory survival, resulting in severe anemia. The increase of oxygen affinity reduces the amount of the insoluble deoxygenated sickle hemoglobin in venous capillaries, thus preventing red blood cells from sickling. Previous microspectrosphotometric studies monitored the kinetics of sickle hemoglobin polymerization within single red blood cells and determined the distribution of sickling cells as a function of oxygen saturation (19). Microspectophotometry could be used to monitor both affinity changes and morphological alteration of single cells in response to drugs. We thank the Danish Natural Science Research Council and the Italian National Research Council for financial support and Academia di Danimarca, Roma, for providing housing for turtles. REFERENCES 1. Antonini E, Brunori M, Giardina B, Benedetti PA, Bianchini G, and Grassi S. Kinetics of the reaction with CO of human erythrocytes. FEBS Lett 86: , Antonini E, Brunori M, Giardina B, Benedetti PA, and Pintus N. Kinetics of oxygen reactions within single erythrocytes. In: Hemoglobin and Oxygen Binding, edited by Ho C et al. Elsevier, 1982, p Bartlett GR. Phosphate compounds in reptilian and avian blood cells; developmental changes. Comp Biochem Physiol A Physiol 61: , Brunori M, Coletta M, Bellelli A, Evangelista V, Benedetti PA, and Brumen M. Microspectroscopy of red blood cells. Mol Cell Biol 7: 89 93, Brunori M, Giardina B, Antonini E, Benedetti PA, and Bianchini G. Distribution of the haemoglobin components of trout blood among the erythrocytes: observations by single-cell spectroscopy. J Mol Biol 86: , Burggren W, Hahn CEW, and Foëx P. Properties of blood oxygen transport in the turtle Pseudemys scripta and the tortoise Testudo graeca: effects of temperature, CO 2 and ph. Respir Physiol 31: 39 50, Clark MR. Senescence of red blood cells: progress and problems. Physiol Rev 68: , Coletta M, Giardina B, Amiconi G, Gualtieri P, Benedetti PA, and Brunori M. Kinetics of the reaction of intraerythrocytic haemoglobin by single cell microspectroscopy: effect of shape and osmolarity. FEBS Lett 190: , Dvorak JA and Stotler WF. A controlled-environment culture system for high resolution light microscopy. Exp Cell Res 68: , Ellis CG, Ellsworth ML, and Pittman RN. Determination of red blood cell oxygenation in vivo by dual video densitometric image analysis. Am J Physiol Heart Circ Physiol 258: H1216 H1223, Frische S, Fago A, and Weber RE. Functional interaction of hemoglobins from the red-eared turtle. In: XXth ESCPB Conference: Molecular, Physiological and Behavioural Adaptations to Environmental Factors, edited by Weber RE, Jensen FB, and Larsen EH. Aarhus, Denmark: ESCPB, 1999, p Holland RAB, Rimes AF, Comis A, and Tyndale-Biscoe CH. Oxygen carriage and carbonic anhydrase activity in the blood of a marsupial, the Tammar wallaby (Macropus eugenii), during early development. Respir Physiol 73: 69 86, Hoofd L and Bos C. Extraction pressures calculated for rat heart and dog skeletal muscle and application in models of tissue oxygenation. Adv Exp Med Biol 428: , 1997.

6 1684 OXYGEN BINDING BY SINGLE RED BLOOD CELLS 14. Kaloyianni M, Cotoglou C, and Giagtzoglou N. Adrenergic regulation of Rana balcanica erythrocyte pyruvate kinase. J Comp Physiol [B] 170: 85 90, Kirkland CB and Altland PD. Red cell survival in the turtle. Am J Physiol 183: 91 94, Krogh A. The number and distribution of capillaries in muscle with calculations of the oxygen pressure head necessary for supplying the tissue. J Physiol (Lond) 52: , Mangum CP, Colacino JM, and Vandergon TL. Oxygen binding of single red blood cells of the annelid bloodworm Glycera dibranchiata. J Exp Zool 249: , Mauro N and Isaacks RE. Examination of reptilian erythrocytes as models of the progenitor of mammalian red blood cells. Comp Biochem Physiol A Physiol 116: , Mozzarelli A, Hofrichter J, and Eaton WA. Delay time of hemoglobin S polymerisation prevents most cells from sickling in vivo. Science 237: , Nikinmaa M. Vertebrate Red Blood Cells. Adaptations of Functions to Respiratory Requirements. Berlin: Springer-Verlag, Pagés T, Peinando VI, and Viscor G. Seasonal changes in hematology and blood chemistry of the freshwater turtle, Mauremys caspica leprosa. Comp Biochem Physiol A Physiol 103: , Phillips MCL, Moyes CD, and Tufts BL. The effects of cell ageing on metabolism in rainbow trout (Oncorhynchus mykiss) red blood cells. J Exp Biol 203: , Rivetti C, Mozzarelli A, Rossi GL, Henry ER, and Eaton WA. Oxygen binding by single crystals of hemoglobin. Biochemistry 32: , Rossi GL, Mozzarelli A, Peracchi A, and Rivetti C. Time course of chemical and structural events in protein crystals measured by microspectrophotometry. Philos Trans R Soc Lond A Math Phys Sci 340: , Saint Girons M-C. Morphology of the circulating blood cells. In: Biology of the Reptilia, Morphology C (1st ed.), edited by Gans C and Parsons TS. London: Academic, 1970, p Samour HJ, Risley D, March T, Savage B, Nieva O, and Jones DM. Blood sampling techniques in reptiles. Vet Rec 114: , Schindler JF and De Vries U. Scanning cytophotometry of carp, Cyprinus carpio L., erythrocyte populations: the influence of short-term hypoxic environment and recovery period following severe bleeding. J Fish Biol 28: , Schmidt W, Böning D, and Braumann KM. Red cell age effects on metabolism and oxygen affinity in humans. Respir Physiol 68: , Skoog DA and West DM. Fundamentals of Analytical Chemistry. New York: Holt-Saunders International, Speckner W, Schindler JF, and Albers C. Age-dependent changes in volume and haemoglobin content of erythrocytes in the carp (Cyprinus carpio L.). J Exp Biol 141: , Tetens V and Christensen NJ. Beta-adrenergic control of blood oxygen affinity in acutely hypoxia exposed rainbow trout. J Comp Physiol [B] 157: , Tsai AG and Intaglietta M. Local tissue oxygenation by statistically distributed sources. Microvasc Res 44: , Vácha J. Red cell life span. In: Red Blood Cells of Domestic Mammals, edited by Agar NS and Board PG. Amsterdam: Elsevier, 1983, p Wang C and Popel A. Effect of red blood cell shape on oxygen transport in capillaries. Math Biosci 116: , Zar JH. Biostatistical Analysis. London: Prentice-Hall, 1984.