Red blood cell thickness is evolutionarily constrained by slow, hemoglobin-restricted diffusion in cytoplasm

During capillary transit, red blood cells (RBCs) must exchange large quantities of CO2 and O2 in typically less than one second, but the degree to which this is rate-limited by diffusion through cytoplasm is not known. Gas diffusivity is intuitively assumed to be fast and this would imply that the intracellular path-length, defined by RBC shape, is not a factor that could meaningfully compromise physiology. Here, we evaluated CO2 diffusivity (DCO2) in RBCs and related our results to cell shape. DCO2 inside RBCs was determined by fluorescence imaging of [H+] dynamics in cells under superfusion. This method is based on the principle that H+ diffusion is facilitated by CO2/HCO3− buffer and thus provides a read-out of DCO2. By imaging the spread of H+ ions from a photochemically-activated source (6-nitroveratraldehyde), DCO2 in human RBCs was calculated to be only 5% of the rate in water. Measurements on RBCs containing different hemoglobin concentrations demonstrated a halving of DCO2 with every 75 g/L increase in mean corpuscular hemoglobin concentration (MCHC). Thus, to compensate for highly-restricted cytoplasmic diffusion, RBC thickness must be reduced as appropriate for its MCHC. This can explain the inverse relationship between MCHC and RBC thickness determined from >250 animal species.


Results
Using measurements of spatio-temporal [H + ] dynamics as a read-out of CO 2 diffusion. CO 2 cannot be imaged dynamically inside a single intact RBC, but its acidic chemistry allows pH-sensitive fluorescent dyes, such as cSNARF1 20 , to monitor its diffusion. H + ions are highly buffered in cells, therefore their apparent cytoplasmic diffusivity (D H app ) is determined exclusively by the mobility of H + -carrying buffers (Fig. 1A,B) 21,22 . Based on this biophysical principle, it is possible to describe the diffusive properties of cytoplasmic buffers from measurements of H + dynamics in intact cells 18,25 . To determine D H app , a local source of H + ions was produced in cytoplasm by photolytic uncaging from the membrane-permeant H + -donor 6-nitroveratraldehyde (NVA) 23 once every 0.13 s. This generates an acidic microdomain which dissipates at a rate determined by the buffers' concentrations, reaction kinetics with H + ions, and diffusion coefficients. To follow the progress of buffer-facilitated H + diffusion, pH-sensitive cSNARF1 fluorescence was imaged confocally across the RBC's horizontal plane at intervals between H + uncaging (e.g. Fig. 1C,D) 20 . Provided that membrane transport of H + and HCO 3 − ions is slow or inactivated, best-fitting these spatio-temporal [H + ] data with a diffusion equation (derived previously 18,23,26 ) gives a robust measure of D H app . During experiments, continuous superfusion of cells controls for temperature and CO 2 /HCO 3 − concentration 20 . In RBCs superfused with CO 2 /HCO 3 − -free solution, H + diffusion is facilitated solely by hemoglobin (Fig. 1A). Under these conditions, D H app reports H + diffusion facilitated by the translation and rotation of hemoglobin molecules 27 . When RBCs are superfused with CO 2 /HCO 3 − -containing solution, cytoplasmic diffusion of H + ions is additionally facilitated by CO 2 /HCO 3 − , thus D H app reports the combined effects of hemoglobin and CO 2 /HCO 3 − (Fig. 1B). Since these two buffer-shuttles are additive, the effect on D H app of introducing CO 2 / HCO 3 − into cytoplasm is a read-out of the turnover of the CO 2 /HCO 3 − buffer-shuttle. High carbonic anhydrase (CA) activity in RBCs ensures that the CO 2 /HCO 3 − buffer-shuttle is not meaningfully rate-limited by chemical reactions; instead, CO 2 /HCO 3 − -facilitated H + diffusion is strongly dependent on the cytoplasmic diffusion coefficients of CO 2 gas and HCO 3 − ions (D CO2 , D HCO3 ). Thus, the first step in quantifying D CO2 and D HCO3 is to probe the CO 2 /HCO 3 − -dependent and independent components of D H app . CO 2 /HCO 3 − weakly facilitates cytoplasmic H + diffusion in the cytoplasm of human RBCs. CO  -free buffer (at pH 7.8). H + ions diffuse slowly from uncaging site, as shown by maps of intracellular pH (pH i ) before and after 2 s of uncaging and pH i time courses in regions of interest (ROIs) at increasing distance from the uncaging site. The width of each ROI was 1/10 th of RBC's major diameter; data for ROIs 1 (uncaging site), 2, 3, 4, 5, 6 and 8 shown. Grey curves are best fit using a diffusion equation. (D) Experiment on RBCs superfused with 5% CO 2 /HCO 3 − (at pH 7.8). The diffusion equation is solved using the finite element method which takes fully into account differences in cell outline, as observed between the cells shown in panels C and D. (E) Apparent H + diffusion coefficients (D H app ; mean ± SEM of 10-35 cells), showing the small effect of the CO 2 /HCO 3 − buffer-shuttle. Where indicated, DIDS was added to block AE1. Symbols α and β denote significance (P < 0.05; P < 0.02) compared to measurements in the absence of CO 2 /HCO 3 − . that encompasses physiological pH c , experiments were performed in superfusates at pH 7.4 or 7.8. Membrane transport of HCO 3 − by AE1 is largely inactivated in the absence of CO 2 /HCO 3 − , therefore uncaged H + ions are retained in cytoplasm. Figure 1C shows the slow spread of H + ions through the cytoplasm of a human RBC in the absence of CO 2 /HCO 3 − . Measured D H app (Fig. 1E) was three orders of magnitude slower than in water (1.2 × 10 4 μ m 2 /s) 28 . Immature RBCs, or reticulocytes, have been proposed to contain small histidine derivatives, such as carnosine 29 , which could facilitate H + diffusion alongside hemoglobin. However, D H app in cells identified positively by Thiazole Orange staining as reticulocytes was no different to D H app in Thiazole Orange-negative (mature) RBCs (Fig. S1). Thus, the concentration of small-molecule histidine derivatives in reticulocytes is not sufficient to meaningfully influence H + diffusivity.
Next, experiments were performed on RBCs superfused with CO 2 /HCO 3 − -buffered NT (BNT) containing 5% CO 2 and either 22 mM HCO 3 − (for pH 7.4) or 55 mM HCO 3 − (for pH 7.8; Fig. 1D; solution osmolality was corrected to 296 mOsm/kg by balancing [NaCl]). If D CO2 were rapid, the CO 2 /HCO 3 − buffer-shuttle would accelerate D H app substantially and collapse pH gradients; however, introducing CO 2 /HCO 3 − into cytoplasm increased D H app by only 2-3 μ m 2 /s (Fig. 1E), indicating that CO 2 /HCO 3 − has a limited capacity to facilitate H + diffusion, comparable to that of hemoglobin. Activation of AE1 in CO 2 /HCO 3 − -containing superfusates may, in principle, compromise the accuracy of D H app measurements. However, D H app estimates were unaffected by blocking AE1 with 12.5 μ M or 163 μ M 4,4′ -diisothiocyano-2,2′ -stilbenedisulfonic acid (DIDS; Fig. 1E), indicating that the magnitude of membrane HCO 3 − transport is not sufficient to short-circuit the cytoplasmic CO 2 /HCO 3 − buffer-shuttle (NB: even the lower dose of DIDS was sufficient to inhibit AE1; Fig. S2). Cytoplasmic CA activity was not inactivated by DIDS, NVA or its photolytic derivative (Fig. S3), therefore low D H app was not an erroneous result of inhibited CO 2 /HCO 3 − reaction kinetics.

Solute diffusion in RBC cytoplasm is restricted by the high concentration of hemoglobin.
To investigate if slow cytoplasmic diffusivity is unique to CO 2 /HCO 3 − , the mobility of calcein (a fluorescent marker) was measured in intact human RBCs. Applying a high-intensity laser beam every 0.13 s to one end of the cell bleaches calcein locally, and drives a diffusive redistribution of fluorescence signal, which was imaged throughout the cell at intervals between bleaching events. Time delays in calcein fluorescence measured at different distances from the bleaching region provide a readout of cytoplasmic diffusivity (D calc ), measured to be 47.8 ± 5.95 μ m 2 /s, i.e. 13-fold lower than in water (~600 μ m 2 /s) 30 (Fig. 2A). This result indicates that RBC cytoplasm is a highly tortuous environment for diffusion.
Hemoglobin, which occupies a quarter of human RBC volume 20 , is likely to impose a substantial tortuosity to the movement of solutes in cytoplasm. Loosening hemoglobin density is expected to accelerate small-molecule diffusion, and this was tested in osmotically-swollen cells (Fig. 2B). RBCs were first pre-equilibrated in, and then superfused with 155 or 220 mOsm/kg solutions (prepared by reducing [NaCl]). In the absence of CO 2 /HCO 3 − , cytoplasmic dilution increased D H app (Fig. 2C), presumably because of less restricted translational and rotational movements of H + -carrying hemoglobin molecules 27 . The relationship between D H app and osmolality was steeper after introducing a constant concentration of CO 2 /HCO 3 − into cytoplasm (Fig. 2C). This finding indicates that at lower hemoglobin density, CO 2 /HCO 3 − is more effective in facilitating H + diffusion. This occurs despite CA activity dilution, adding further evidence that H + diffusion is not reaction-limited (i.e. even after 2-fold dilution, CA activity remains very high).
To explore the effect of naturally occurring differences in hemoglobin concentration on D H app , measurements were performed on RBCs from chicken, alpaca and Xenopus ( Fig. 3A-C). Previous studies have detected carnosine in nucleated erythrocytes 31 , which may augment D H app in chicken and Xenopus RBCs by acting as a mobile buffer alongside hemoglobin and CO 2 /HCO 3 − . However, using a modification of Pauly's assay, levels of small-molecule histidine derivatives were in the sub-millimolar range, with the highest levels detected in chicken RBCs (0.5 mM; Fig. S4). Over this low concentration range, histidine-containing small molecules cannot meaningfully increase D H app , therefore hemoglobin and CO 2 /HCO 3 − remain the principal H + -carriers. To measure D H app , alpaca and chicken RBCs were superfused in solution at 37 °C and pH 7.8, whereas Xenopus cells were superfused at room temperature and pH 7.5 to match their normal physiology. Nuclear regions in Xenopus and chicken RBCs were excluded in the analysis of pH (Hoechst 33342-positive nuclear areas also had higher intensity of cSNARF1 fluorescence, likely reflecting the ambient physicochemical environment of nucleoplasm; Fig. S5). Compared to human RBCs, chicken and Xenopus RBCs are larger and have a lower MCHC, whereas alpaca RBCs are smaller but with higher MCHC (Table S1). High CA activity was detected in hemolysates from all species studied, although it was lower in alpaca and Xenopus compared to humans (Fig. S6). Introducing CO 2 / HCO 3 − into cytoplasm increased D H app in RBCs from chicken and Xenopus, but not alpaca (Fig. 3D). Hypotonic swelling (i.e. loosening hemoglobin density) increased D H app further in chicken and alpaca RBCs (NB: Xenopus RBC are too fragile for this experiment). Figure 3E summarizes the relationship between D H app and MCHC; with the exception of cold-blooded Xenopus, all data-points followed an exponentially-declining relationship. In summary, the facilitatory effect of CO 2 /HCO 3 − on D H app was attenuated at higher MCHC, consistent with hemoglobin-imposed tortuosity.

Calculated CO 2 diffusivity in cytoplasm is slow and dependent on hemoglobin concentration.
Cytoplasmic H + diffusivity is related mathematically 21,22 to the concentration, mobility and protonation/deprotonation kinetics of hemoglobin and CO 2 /HCO 3 − . In this system, all variables except for CO 2 /HCO 3 − mobility (D CO2 , D HCO3 ), are known, therefore an algorithm can be designed to calculate D CO2 and D HCO3 from D H app measurements. Since CO 2 diffuses 46% faster than HCO 3 − (a difference attributable to molecular size) 2 , D CO2 and D HCO3 are numerically constrained to one another, reducing the system of equations to one unknown. This algorithm, described in more detail in the Supplement, performs simulations for a range of different test-values of D CO2 (and D HCO3 ) and for each generates a predicted D H app . Least-squares best-fitting this output to the experimentally-determined D H app derives D CO2 (and D HCO3 ). The 'known' variables required to run these simulations were obtained as follows. Firstly, the concentration of hemoglobin was obtained from MCHC measurements 20,32 , and the cytoplasmic concentration of CO 2 /HCO 3 − was calculated using the Henderson-Hasselbalch equation from extracellular [CO 2 ] and [HCO 3 − ] and the measured transmembrane pH gradient. Secondly, hemoglobin reaction kinetics were assumed to be instantaneous, whereas the reactions of CO 2 /HCO 3 − were modelled kinetically using data for CA activity, hydration and dehydration rate constants. Finally, hemoglobin-facilitated H + diffusivity is equal to D H app measured in CO 2 /HCO 3 − -free media. Cytoplasmic D CO2 and D HCO3 were found to be substantially lower than the rates measured in water (Fig. 4A) and were inversely correlated with MCHC (Figs 4B and S7). For example, CO 2 diffusivity in human RBC cytoplasm was 5% of the rate in water. The extent to which hemoglobin restricts small-molecule diffusion is substantially greater than previous estimates (inset to Fig. 4B; e.g. for human RBCs, the difference is one order of magnitude). Thus, highly restricted diffusivity of CO 2 gas and HCO 3 − anions in RBC cytoplasm is a hitherto unrecognized rate-limiting step in the process of gas exchange.

Discussion
Chemical reactions and membrane transport are recognized to be critically important in determining the rate of gas exchange, and their experimental characterization has informed our current model of RBC physiology. The results of this study indicate that diffusion inside RBCs is a strongly rate-limiting factor for gas turnover. We find that RBC cytoplasm is a highly tortuous environment that substantially restricts the movement of solutes, consistent with the uniquely high microviscosity inside RBCs 33 . In terms of resistances to overall gas transport 34 , the component attributable to RBC cytoplasm has hitherto been underestimated, and so it is plausible that the other resistances in series (imposed by membranes and extracellular unstirred layers) have been exaggerated in earlier analyses 3 .
The extent to which the cytoplasm of intact human RBCs restricts gas diffusion is greater than estimated previously from cell-free solutions 12,15 . This may relate to the significantly slower rotational diffusion of hemoglobin inside RBCs compared to solution 33 , which argues that encapsulation in a membrane produces a more rigid lattice of macromolecules. It is also possible that diffusivity measurements in cell-free solutions may have been over-estimated as a result of convective currents of the medium, which are much less likely to occur inside cells. Of note is that our measured effect of hemoglobin on CO 2 diffusivity is quantitatively similar to the effect of an unrelated macromolecule, Ficoll70, on rhodamine green diffusion 35,36 . Thus, hemolysates and hemoglobin-solutions may have only a limited ability to fully recapitulate the biophysical properties of intact RBC cytoplasm. The cytoplasm of human RBCs reduces CO 2 diffusivity by a factor of 23 and calcein diffusivity by 13-fold. The difference in tortuosity imposed on CO 2 and calcein may relate to the solute's reactivity with hemoglobin. CO 2 binds weakly and reversibly to hemoglobin 2 (forming carbamino-hemoglobin), which will retard CO 2 translational diffusion. In contrast, the highly negatively charged calcein will not interact with hemoglobin in this way. Also, because the CO 2 molecule (radius 2 Å) is 50-fold smaller than the calcein molecule (radius 7.4 Å), CO 2 may be able interact more intimately with the surface of the hemoglobin (radius 27.5 Å). Consequently, CO 2 may experience a more tortuous path around hemoglobin, compared to calcein.
Since diffusivity, path-length and cell shape are inter-related, our measurements are an important addition to our understanding of RBC form and function. In light of highly restricted diffusivity, the advantage of flattening an RBC can be explained in terms of minimizing critical time delays, which are proportional to the square of distance. To explore this proposal, a mathematical model (described more fully in the Supplement) simulated the time course of CO 2 penetration into human RBCs entering a hypercapnic (8% CO 2 ) environment (Fig. 5Ai). Predictions for normal human RBCs (half-thickness 0.9 μ m; diameter 8 μ m) were compared to those for a hypothetical spherical variant (diameter 8 μ m). Using D CO2 data derived previously from cell-free solutions (40% of diffusivity in water, labelled here as 'fast'), the time-to-reach 90% of the equilibrium state (τ 90 ) for CO 2 partial pressure (pCO 2 ) was adequately rapid in both the flat (0.03 s) and spherical (0.09 s) geometry (Fig. 5Aii). Simulations using the presently determined value of D CO2 (5% of diffusivity in water, i.e. 'slow') showed a dramatically prolonged τ 90 for pCO 2 equilibration: 0.09 s for a flat RBC and 0.38 s for its spherical variant (Fig. 5Aiii). The later delay is incompatible with efficient gas exchange with faster blood flows (e.g. in exercise). Thus, the critical importance of RBC thickness for efficient gas exchange becomes apparent only when slow cytoplasmic diffusivity is considered. The cytoplasmic restrictions imposed on CO 2 movement may also apply to O 2 diffusion, therefore RBC thickness could also be important in determining the rate of O 2 exchange.
Another consequence of profoundly restricted CO 2 and HCO 3 − diffusion is very slow H + ion mobility in RBC cytoplasm. Remarkably, H + diffusivity inside RBCs is the slowest among all ions studied in cells; this is somewhat surprising given that RBCs experience the highest acid-base turnover in the body 26 . H + ions are a powerful signal that regulates O 2 binding to hemoglobin through the Bohr 24 and Root Effects 37,38 , but this response is only effective if the RBC's mean cytoplasmic pH (pH RBC ) is able to track changes in ambient pH (e.g. during the transit through acidic microvasculature of contracting muscles). Slow cytoplasmic H + diffusion thus places an additional constraint on RBC thickness, which was analyzed, using the model, in terms of the pH RBC response to an ambient metabolic acidosis (Fig. 5Bi). Assuming the higher D CO2 value, a flattened RBC will acidify faster than a spherical cell, but in both cases, τ 90 was under 1 s (0.48 s for flat; 0.75 s for spherical; Fig. 5Bii). Using the lower D CO2 value measured herein, τ 90 would increase to 0.54 s and 0.95 s for flat and spherical cells, respectively (Fig. 5Biii). The delay associated with spherical geometry would result in only a partial manifestation of the Bohr Effect, and consequently suboptimal distribution of O 2 to tissues. Thus, in order for O 2 release from RBCs to respond adequately to changes in ambient pH, the cell must acquire a flattened form.
Circulating human RBCs undergo a dramatic velocity-dependent deformation (e.g. parachute-like shape) as they transit through the microvasculature [39][40][41] . As a result of cytoplasmic re-distribution, the intracellular diffusion distances may modestly increase towards the leading edge of the cell and decrease at rear. Nonetheless, the parachute RBC retains an overall flattened form that is compatible with efficient gas exchange. This ability of flattened human RBCs to attain a parachute shape in capillaries may be critical for maintaining minimal cytoplasmic path-lengths for efficient gas exchange.
Our work also quantified the effect of MCHC on gas diffusion. There is considerable inter-species variation in MCHC (Table S2) which, at least in part, is driven by demand for O 2 -carrying capacity, but nonetheless appears to be capped at ~550 g/L. High MCHC is associated with raised microviscosity and resistance to blood flow 6,42 , which could underlie the biological limit of MCHC, although this can be compensated for by reducing RBC count. Here, we demonstrate that raising MCHC by 75 g/L halves D CO2 (Fig. 4B), an effect that is quantitatively similar to that of other macromolecules (e.g. rhodamine green diffusivity halves with every 88 g/L rise in Ficoll70 concentration) 36 . Thus, a physiological need to pack more hemoglobin into a RBC (e.g. observed in some altitude-adapted species, such as llamas and alpacas) comes at the cost of more restricted gas diffusivity. Cells adapted in this manner would need to be appropriately thinner in order to support efficient gas exchange. To investigate this hypothesis, we compiled data on RBC half-thickness (i.e. the shortest cytoplasmic path-length) and MCHC from > 250 cold-and warm-blooded species (see Table S2; Fig. S8). Figure 5C shows a two-dimensional frequency histogram for RBC half-thickness and MCHC, superimposed with simulations for the time-to-reach 90% pCO 2 equilibrium (τ 90 ), determined using the model, shown in Fig. 5A, with appropriately varied MCHC (hence D CO2 , D HCO3 ) and half-thickness. The majority of naturally-occurring MCHC/ half-thickness combinations fell in the τ 90 range of 0.03− 0.3 s, i.e. compatible with near-complete equilibration during typical capillary transit times. This model was also used to simulate τ 90 for pH RBC equilibration in response to ambient acidosis. The majority of naturally-occurring MCHC/half-thickness combinations were constrained to meet the criterion of τ 90 < 1 s (Fig. 5D). We postulate that thick RBCs with high MCHC (i.e. upper right quadrant in Fig. 5C,D) are not normally found in nature because these would be associated with unacceptably slow gas exchange (τ 90 for pCO 2 equilibration > 0.3 s) and an incomplete manifestation of the Bohr Effect (τ 90 for pH RBC equilibration > 1 s). Membrane tension and, in some species, the presence of a nucleus limit the extent to which RBCs could be made thinner; we propose that this imposes a cap on MCHC. According to Fig. 4B, MCHC greater than 500 g/L would restrict CO 2 diffusion by a factor of > 100, requiring RBCs to be thinner than 1 μ m; this may explain why such high hemoglobin levels are rare, despite pertinent selection pressures for increasing O 2 -carrying capacity. This line of reasoning may also explain why mature RBCs in mammals have lost their nucleus in a bid to become thinner.
The regulatory means by which normal RBCs meet the favorable MCHC/half-thickness criterion is likely to involve interactions within the set of gene loci (e.g. seventy-five in humans) 43 that collectively determine RBC shape and hemoglobin content. Failure to attain a favorable combination of cell thickness and MCHC may impact on gas exchange efficiency. The analyses shown in Fig. 5C,D highlight a potential disadvantage of attaining spherical symmetry in diseases such hereditary spherocytosis 44 , where mean cytoplasmic path-length increases due to shape and also MCHC tends to be higher due to cellular dehydration 45,46 . These circumstances predict a less complete gas exchange and only a partial manifestation of the Bohr effect at higher blood flows, which may contribute towards the reduced exercise tolerance commonly observed in patients with hereditary spherocytosis 46 .
Restricted diffusion in RBC cytoplasm may have important physiological implications besides gas exchange at capillaries. Slow diffusion of nitric oxide in RBC cytoplasm, in addition to the immobilizing effect of hemoglobin nitrosylation, may be a requirement for confining its signaling cascade to a sub-membranous domain 47 . A dense network of proteins may restrict gas diffusion in non-erythroid cell types or subcellular structures 48 . The magnitude of macromolecule-restricted diffusion highlights the potential for cytoplasm to regulate gas fluxes, a feat considered thus far to be in the remit of cell membranes only.
In conclusion, we propose that the efficiency of gas exchange depends critically on RBC thickness because the high density of hemoglobin restricts the movement of small solutes. This restriction is substantially greater in intact RBCs, compared to cell-free hemoglobin solutions or hemolysates, which explains why the cytoplasmic mobility of gases, HCO 3 − and H + ions has previously been overlooked as a rate-limiting step in the gas exchange cascade. In addition to increasing the cell's surface area/volume ratio for greater trans-membrane solute traffic and providing mechanical advantages for blood flow, the evolutionarily-conserved flattening of RBCs is an essential adaptation to minimize delays owing to slow cytoplasmic diffusion.

Methods
Red blood cells. Alpaca (Vicugna pacos) and chicken (Gallus gallus) blood were collected by trained staff from Seralabs (U.K.) and delivered on ice within 24 hours of collection. Xenopus laevis blood was collected by ventricular puncture of animals that have been sacrificed humanely by pithing in accordance with Schedule I of Animal (Scientific Procedures) Act 1986 carried out in a licensed facility at Oxford University. Human blood was obtained from two volunteers who gave formal and informed consent, in accordance with Central University Research Ethics Committee (CUREC) guidelines (reference: R46540/RE001, approved procedure #24), and data were fully anonymized and not traceable to the donor. All methods involving the use of blood were performed in accordance with ethical guidelines set by Oxford University, with appropriate risk assessments in place. All experimental protocols performed on collected human red blood cells are performed in accordance with Oxford University's Central University Research Ethics Committee (CUREC) guidelines (reference: R46540/RE001, approved procedure #24). Additionally, all experimental protocols perform on animal and human red blood cells were performed in accordance with Oxford University's Health and Safety regulations (OHS Policy Documents 1/03, 1/01, OHS 2/03, UGN S1/95). Clinical tests confirmed that mean corpuscular volume (MCV) and mean corpuscular hemoglobin concentration (MCHC) of donor blood were in the normal range (MCV: 88.2-90.9 fL, MCHC: 325-334 g/L). Blood samples were collected in tubes treated with heparin (Xenopus) or EDTA (all other species), spun-down at 4 °C (10,000 for 5 min for mammalian blood; 5,000 for 5 min for chicken blood; 1,000 for 10 min for Xenopus blood) to remove the supernatant and buffy coat. A sample of the fraction containing packed red blood cells was re-suspended in Hepes-buffered NT and used for confocal imaging or flow cytometry; the remainder was freeze-thawed twice to break up cell membranes, and used for measurements of hemoglobin concentration (HemoCue Hb 201plus), histidyl-containing small molecules (Pauly's assay) and carbonic anhydrase activity.
Confocal imaging. To image intracellular pH (pH i ), cells were loaded with the acetoxymethyl (AM) ester of the pH-reported dye cSNARF1 for 60 minutes (Xenopus) or 10 minutes (all other species), and allowed to settle on a poly-L-lysine pretreated coverslip at the base of a superfusion chamber mounted on an inverted Zeiss Axiovert microscope that was coupled to a Zeiss LSM700 confocal system. Superfusates were delivered at 4 mL/min at 25 °C (Xenopus) or 37 °C (all other species). Cells were imaged with a x100 objective and stored as stacks of 256 × 256 pixel 8-bit bitmaps (53 nm pixel length). cSNARF1 was excited by a 555 nm laser line and fluorescence was collected at 580 ± 10 and 640 ± 10 nm. The fluorescence ratio was converted to pH using a calibration curve, obtained by a published method 20 . To image calcein, cells were loaded with calcein-AM for 10 minutes, and fluorescence (> 515 nm) was excited with low intensity 488 nm laser line. Offline analysis was performed using ImageJ.
H + uncaging and calcein bleaching. Photolytic uncaging of H + ions from 6-nitroveratralhyde 23 (NVA; added to solutions at 1 mM) was evoked by scanning a region of interest (ROI) at one end of the RBC (ROI width equal to 1/10 th of RBC diameter) with 405 nm laser light, as shown in Fig. 1C,D. By alternating between regional H + uncaging and whole-field pH i -imaging, the spatial dissipation of H + ions from the uncaging site can be followed. Fluorescence maps were analyzed using ImageJ. Two stacks of images (580 nm and 640 nm fluorescence) were background-offset and fluorescence signal was averaged in ten ROIs of width equal to 1/10 th of the RBC diameter along x-axis and height equal to the RBC diameter in the y-axis. For Xenopus and chicken, the fluorescence from the nucleus was excluded from analysis. ROI1 represented the uncaging region. The ratio of ROI fluorescence at 580 and 640 nm was converted to [H + ] time courses, which were fitted with a diffusion equation to obtain the apparent diffusion coefficient (D H app ), according to an algorithm described previously 26 and in the Supplement. Bleaching of calcein fluorescence was performed by a similar protocol to H + uncaging. The excitation protocol alternated between high intensity laser for localized calcein bleaching and low intensity 488 nm laser for imaging cytoplasmic calcein. Time courses of calcein fluorescence, measured in ten ROIs, were fitted with a diffusion equation, similar to that described in the Supplement for analyzing D H app .