Assessment of transient changes in oxygen diffusion of single red blood cells using a microfluidic analytical platform

Red blood cells (RBCs) capability to deliver oxygen (O2) has been routinely measured by P50. Although this defines the ability of RBCs to carry O2 under equilibrium states, it cannot determine the efficacy of O2 delivery in dynamic blood flow. Here, we developed a microfluidic analytical platform (MAP) that isolates single RBCs for assessing transient changes in their O2 release rate. We found that in vivo (biological) and in vitro (blood storage) aging of RBC could lead to an increase in the O2 release rate, despite a decrease in P50. Rejuvenation of stored RBCs (Day 42), though increased the P50, failed to restore the O2 release rate to basal level (Day 0). The temporal dimension provided at the single-cell level by MAP could shed new insights into the dynamics of O2 delivery in both physiological and pathological conditions. Kevin Ziyang Chng et al. develop a microfluidic platform using a fluorescence quenching approach to assess transient changes in oxygen release rate of single red blood cells. They observe that biologically older red blood cells have a faster O2 release rate than younger cells and that storage of blood enhances this release rate.

O xygen (O 2 ) is required for maintaining various vital cellular functions and its supply to tissues is maintained by the cardiovascular system that ramifies throughout every organ in the body. Deprivation of O 2 leads to tissue hypoxia, ultimately inducing cell dysfunction and apoptosis. Delivery of O 2 is facilitated by red blood cells (RBCs) which constitute the majority of blood cells. Owing to its densely packed hemoglobin (Hb), RBCs can bind with O 2 and release it on demand under deoxygenated conditions. Yet, the ability of RBCs to deliver O 2 has been routinely characterized solely by their HbO 2 affinity (P50), defined as the partial pressure of O 2 (PO 2 ) required to saturate Hb to 50% in a thermodynamic equilibrium process 1 . This metric, however, neglects the dynamic release rate of O 2 from RBCs, and hence may be insufficient to assess the efficacy of RBC O 2 delivery in the microcirculation. In contrast, assessing the temporal aspect of O 2 delivery using the dynamic O 2 release rate of RBCs would afford a more direct measure of the O 2 transport by RBCs 2 .
O 2 transport in the microvasculature typically occurs within a time frame of 4-7 s 3 . During the unloading process, O 2 diffuses across the RBC membrane after dissociating from Hb. It has been shown that the internal cytoplasmic resistance of the RBC to O 2 transport is not negligible 4,5 . Consequently, the time required for an RBC to unload O 2 is dependent on both this internal resistance and the external milieu. Moreover, O 2 uptake and release in RBCs are known to depend on their morphological and physiological properties 6 . Since RBCs exist as a non-homogenous entity of various biological ages with different biophysical and biochemical properties [7][8][9] , how these variations ultimately affect the O 2 release rate remains unclear. Furthermore, a single-cell measurement is particularly important for RBC studies since there is vast cellular heterogeneity in the RBC population, which could result in the masking of functionally important subpopulations. Hence, profiling individual cells is essential to provide a representation of cellular level events in lieu of stochastic averages from bulk measurements 10 . Our motivation for this study is echoed by a recent study that investigated alterations in the O 2 release rate of RBCs with varying cytoplasmic diffusivity 11 . Specifically, they have shown that an increase in pathlength or tortuosity of RBCs under engineered and diseased conditions can lead to a decrease in O 2 transport. However, their measurements failed to isolate single cells from intercellular diffusion, unlike the current study.
Here, we use a diffusion parameter, D50, to better explain how fast each RBC can diffuse O 2 to the surrounding tissues. For the isolation of a single RBC, we developed a microfluidic analytical platform (MAP) to examine the O 2 diffusion from each single RBC, from which we derived the population variability of the O 2 release dynamics. Using this MAP, we examined the effect of in vivo aging on the O 2 release rate of RBCs and found that biologically older RBCs exhibited higher O 2 release rates. We also demonstrated the effect of RBC storage lesion on the O 2 release rate, which revealed a faster O 2 release rate in stored RBCs than fresh cells. Moreover, we confirmed that the P50 value alone does not substantiate how fast RBCs can diffuse O 2 even at the bulk level.

Results
MAP design and validation. The MAP was designed with an array of microwells for isolating single RBCs in the main microchannel. The microwells were then deoxygenated by nitrogen (N 2 ) flowing in the parallel gas microchannel, which is separated from the microwells by 25-µm polydimethylsiloxane (PDMS) layer (Fig. 1a). Using fluorescence microscopy (Fig. 1b), we determined the rate of RBC deoxygenation from the temporal variations of fluorescence intensity in the microwells (Fig. 1c). As the presence of RBCs in the microwells provided an additional source of O 2 , higher rates of O 2 release by the RBCs would thus reduce the rate of fluorescence decay in the microwells. This fluorescence quenching approach allows for the direct measurement of local O 2 concentration. To validate the MAP, we established the decay rate of the O 2 depletion in the microwells without RBCs, which corroborated the decay profile obtained from our computational simulation (R 2 = 0.97) ( Supplementary  Fig. 2). To quantitatively assess the O 2 release rate from single RBCs, we first fitted the experimental data using nonlinear regression, PO 2 ¼ Ae ÀKt , where A is the initial PO 2 in the microwell (155.3 mmHg), K denotes the decay constant, and t represents time. The obtained decay constant was then used to compute the time taken for the O 2 level in the microwell to decrease by 50% from the initial equilibrium concentration (D50 ¼ lnð2ÞK À1 ). Therefore, a higher D50 value indicates a faster O 2 release rate from the RBC. Our results revealed that microwells with single RBCs showed a significantly larger D50 (1.17 ± 0.52 s) than those without RBCs (0.80 ± 0.29 s) (Fig. 1d).
In vivo aging enhances the O 2 release rate of RBCs. To investigate the effects of in vivo aging on the O 2 release rate of RBCs ( Fig. 2), freshly drawn RBCs from healthy donors (n = 5) were density-fractionated into two groups: younger (top 20%) and older (bottom 20%). Older RBCs (41.3 ± 4.3 g/dL) had a significantly higher mean corpuscular hemoglobin concentration (MCHC) than younger RBCs (35.9 ± 1.7 g/dL). Absolute 2,3-DPG was also found to be substantially lower in older RBCs (1.76 ± 0.26 mM) than that in younger RBCs (2.80 ± 0.81 mM). As a result, normalized 2,3-DPG of older RBCs (4.32 ± 0.9 µmol/g Hb) was correspondingly lower than younger RBCs (7.78 ± 2.24 µmol/ g Hb). Consequentially, the P50 of older RBCs (29.9 ± 1.0 mmHg) was lower than younger RBCs (32.0 ± 1.7 mmHg). The process of deoxygenation induces a conformational change in Hb from the relaxed (R) to the tense (T) structure, where Hb in the R state has a higher affinity for O 2 than that in the T state. The 2,3-DPG binds to the central cavity of the deoxygenated Hb and anchors the molecular configuration of Hb in the T state. Hence, a lower 2,3-DPG level, as observed in the older RBCs, increases the Hb affinity for O 2 .
Interestingly, despite the increase in O 2 affinity as determined by the P50 level (Fig. 2d), the diffusion rate (D50) of older RBCs (1.17 ± 0.32 s) was significantly higher than younger RBCs (1.03 ± 0.27 s) (Fig. 2e), implying a faster O 2 release rate. D50 is governed by O 2 diffusion through the cytoplasm and across the cell membrane. MCHC increases with RBC age through the reduction in cell volume via microvesiculation 12 . This results in a substantial tortuosity to the movement of solutes in the cytoplasm due to molecular crowding 13 . However, this is countered by a reduction in the diffusion path-length 2,14 . We observed an overall increase in D50 with RBC age, which implies the dominance of the reduced path-length over cytoplasmic diffusivity in regulating O 2 diffusion. In probing the O 2 release rates at the single-cell level, we were able to obtain and compare the distributions of D50 between older and younger RBCs (Fig. 2f), which showed that the full width at half maximum of the D50 distribution in older RBCs (0.75) was also larger than younger RBCs (0.64). In addition, the skewness for the D50 distributions of younger and older RBCs was determined to be 0.17 and 0.48, respectively, indicating a stronger deviation from normality toward higher D50 values in older RBCs.
Blood storage enhances the O 2 release rate of RBCs. To examine the effect of in vitro aging on the O 2 release rate of RBCs, fresh RBCs were stored for 42 days following the standard blood banking protocol. Figure 3 shows changes in the O 2 affinityrelated factors of the stored RBCs. No significant alteration in the MCHC was observed during the storage duration. In contrast, absolute (Day 0: 2.77 ± 0.52 mM, Day 7: 1.10 ± 0.32 mM) and normalized 2,3-DPG concentrations (Day 0: 8.61 ± 1.51 µmol/g Hb, Day 7: 3.49 ± 0.99 µmol/g Hb) dropped drastically by Day 7 and came close to depletion on Days 28 and 42. Consequently, P50 of stored RBCs decreased significantly with storage duration, indicating an increase in O 2 affinity (Day 0: 33.2 ± 0.87 mmHg, Day 42: 24.9 ± 0.16 mmHg) (Fig. 3d). Detailed experimental results can be found in Supplementary Table 1.
As opposed to the decrease in the P50, the D50 of stored RBCs (Fig. 4) was significantly higher than that of the fresh RBCs (Day 0: 0.87 ± 0.30 s, Day 42: 1.08 ± 0.26 s). The D50 distribution at Day 0 assumed a Gaussian profile. However, the distribution deviated from Gaussian at Day 1 and tended toward a bimodal distribution by Day 14; this bi-modal distribution seemingly reverted to normality again by Day 42. Concomitantly, the distribution was initially right-shifted toward higher D50 but became left-shifted from Day 14. The full width half maximum (FWHM) for each of the storage durations was also calculated ( Supplementary Fig. 3). We found that Day 14 (1.32) had an ∼86% increase in FWHM as compared to Day 0 (0.71). The FWHM then decreased after Day 14 and by Day 42, the FWHM reduced to ∼13% less than Day 0 (Day 28: 1.14, Day 42: 0.62). Considering these observations, we hypothesized that different RBC subpopulations undergo differential rates of hemolysis during storage, which would account for the emergence of bimodal distributions and gradual left-shift with longer storage durations. To confirm this, we measured the hemolysis level of density-fractionated stored RBCs (for 42 days) in a separate experiment. Notably, older (denser) RBCs (1.81 ± 0.36%) showed a~35% higher hemolysis than younger RBCs (1.34 ± 0.39%) after 42 days.

Rejuvenation of 2,3-DPG does not improve the O 2 release rate.
To determine the effect of 2,3-DPG rejuvenation on the O 2 release rate of stored RBCs, Day 42 RBCs were incubated in 10 mM inosine and pyruvate solutions. The rejuvenation process increased the MCHC of the stored RBCs by~5% (Day 42: 30.8 ± 1.1 g/dL, Rejuvenated RBCs: 32.4 ± 1.1 g/dL) (Fig. 5a). As The microfluidic device consists of two fluid channels: RBC loading (red) and N 2 flow (yellow). The RBC loading channel consists of an array of microwells for isolating single RBCs. A 25-µm PDMS wall separates the microwells from the N 2 channels. The sample comprising RBCs suspended in tris(2,2′-bipyridyl) dichlororuthenium(II) hexahydrate (RTDP) solution was first loaded into the MAP. Mineral oil was then flushed into the microchannel to remove the cells in the main channel and isolate single RBCs in each microwell. N 2 gas was then pumped into the gas channel to initiate the RBC deoxygenation process. b Trans-(i) and epiillumination (ii) images of a microwell with a single RBC are shown. The red arrow indicates a single RBC isolated in the microwell. Scale bar = 10 µm. c Comparison of PO 2 decay between microwells with (blue) and without (red) RBCs. The blue and red dots represent the raw experimental data while the corresponding solid lines represent the curves fitted using nonlinear regression. Blue line: PO 2 ¼ 155:3e À1:07t ; ðR 2 ¼ 0:85Þ. Red line: PO 2 ¼ 155:3e À0:63t ; ðR 2 ¼ 0:86Þ. d The decay constant obtained from the fitted curves was subsequently used to calculate the D50 (D50 ¼ lnð2ÞK À1 ), with microwells loaded with RBCs (n = 82 of biologically independent RBCs) showing a significantly higher D50 (*P < 0.0001) than without RBCs (n = 127 of biologically independent RBCs). The central mark of the boxplot indicates the median, and the bottom and top edges of the box indicate the 25th and 75th percentiles, respectively. The whiskers extend to the most extreme data points not considered outliers.  (Fig. 5b, c). Rejuvenation of 2,3-DPG led to a significant increase (~20%) in the P50 of stored RBCs (Day 42: 24.9 ± 0.16 mmHg, Rejuvenated RBCs: 29.2 ± 0.99 mmHg) (Fig. 5d), although it was still significantly lower than the fresh RBCs (Day 0: 33.2 ± 0.87 mmHg). On the other hand, the measured D50 values (Fig. 5e) before and after rejuvenation were not significantly different (Day 42: 1.08 ± 0.26 s, Rejuvenated RBCs: 0.98 ± 0.28 s).

Discussion
The fundamental role of RBCs is to deliver O 2 to the tissues to sustain key metabolic processes that are required to maintain life.
Hence, determining the efficacy of RBC O 2 delivery is crucial in explaining a wide range of cellular phenomena in both physiological and pathological conditions. Many previous studies have emphasized the significance of the temporal aspect of O 2 delivery 15,16 . Specifically, a decrease in RBC residence time or higher RBC flow velocity in the microcirculation could lead to tissue hypoxia due to reduced O 2 release from the RBCs 17 .
The cellular heterogeneity that exists within the RBCs population demands single-cell analyses, which can provide new insights into RBCs O 2 delivery functionally 18,19 . Di Caprio et al. 18   release dynamics. In contrast, the present study localized the O 2 -sensitive fluorescent dye in the suspending medium to avoid disrupting the native state of the RBCs. Compared with the previous studies, a distinctive feature of our MAP is the array of microwells, which was designed to limit the diffusion to only between the suspending medium and the isolated single RBCs. Thus, the microwells effectively allow us to obtain the diffusion kinetics of the O 2 unloading without interference from intercellular diffusion, along with the diffusion of the ambient gases into the suspending medium. O 2 diffusion in RBCs. O 2 uptake and release by RBCs involve complex molecular processes that occur over a span of length and time scales. RBCs uptake O 2 in the alveoli in the lungs, within which O 2 diffuses through blood plasma and the RBC cell membrane to reach the intracellular Hb. Within the RBCs, Hbfacilitated diffusion of O 2 exhibits a biphasic trend with Hb concentration 20 . Moreover, increasing the amount of Hb molecules is likely to impose a substantial tortuosity to the movement of solutes in the cytoplasm 2 . This is supported by a previous study 11 that showed a decrease in O 2 transport rate across the cytoplasm due to an increase in MCHC. Nonetheless, our experimental results showed that older RBCs exhibited significantly higher O 2 release rates (higher D50) than their younger counterparts. This suggests that within the physiological range of MCHC (i.e., younger: 35.9 ± 1.7 g/dl; older: 41.3 ± 4.3 g/dL), the concentration of Hb does not have a dominant effect on the O 2 diffusion rate from RBCs.
During the aging process, whether in vivo or in vitro, RBCs experience a decline in the activities of vital metabolic enzymes and altered redox metabolism, leading to reduced ATP production 21 , oxidative damage to Band 3 22 and membrane vesiculation 23,24 . Consequently, older RBCs experience a decrease in mean cell volume (MCV) 9,25 , which increases the MCHC, while reducing the path-length and the time required for O 2 to diffuse across the cytoplasm to the inner side of cell membrane 2 , thus enhancing the rate of O 2 release from RBCs 11 . The reductions in MCV and total surface area of the membrane in older RBCs could also lead to a decrease in the amount of membrane cholesterol [26][27][28] , which was previously shown to modulate O 2 permeability in RBC membrane [29][30][31][32][33] . Therefore, these factors could cumulatively contribute to the elevation of D50 observed in older RBCs. While Richardson et al. 11 demonstrated the effects of isolated changes in RBCs, our study suggests that there is a myriad of factors within the scheme of O 2 diffusion of single RBCs. Concomitantly, diffusion of NO and CO 2 could also be modulated through altered deoxygenation conditions 34,35 and compromised Band 3 36,37 as a consequence of the aging process. However, the effect of the two gases on our O 2 measurements would be negligible since the RBCs are fully saturated with O 2 and the NO produced by the NOS within the RBCs would be insignificant.

2,3-DPG. 2,3-DPG is an intermediary metabolite in the
Embden-Meyerhof glycolytic pathway that binds specifically to the central cavity of deoxyhemoglobin. This stabilizes the tense state of Hb, which reduces its O 2 affinity as confirmed in our results (Figs. 2d and 3d) 38 . RBCs have a typical lifespan of 100-120 days during which they are constantly exposed to various insults from the external environment, leading to cumulative changes in their physical and chemical properties 39 . RBC senescence is associated with the exponential decline in glycolytic enzymes with RBC age, leading to reduced glycolysis and thus intracellular 2,3-DPG 40 . During in vitro blood storage, a similar decrease in glycolysis is attributed to the accumulation of lactic acid that lowers the pH of the storage medium 41 . While our results confirm that both in vivo (biological) and in vitro Fig. 3 Effect of in vitro aging during blood storage on O 2 affinity of RBCs. a-d MCHC, absolute 2,3-DPG concentration, normalized 2,3-DPG, and P50 during the storage period. All the results are based on bulk sample measurements (n = 5 of biologically independent samples). The significant difference relative to Day 0: *P < 0.05, **P < 0.0001. The significant difference relative to Day 1: † P < 0.01, † † P < 0.0001. The significant difference relative to Day 7: § P < 0.001, § § P < 0.0001. The significant difference relative to Day 14: ❖ P < 0.01. Error bars correspond to the standard deviation of the data.
(storage), aging processes decrease the 2,3-DPG concentration and P50 of RBCs, we found no decline in their actual O 2 diffusion rate. Interestingly, in both older and stored RBCs, D50 was significantly larger (O 2 diffusion was faster) than in the younger (Fig. 2e) and fresh RBCs (Fig. 4) 42 . Therefore, the use of both D50 (dynamic factor) and P50 (static factor) would provide a more comprehensive assessment of the O 2 delivery capability of RBCs.
Higher vs. lower D50. Oxygenated RBCs are constantly exposed to a large O 2 gradient across the tissues before returning to the pulmonary circulation. Thus, it is likely that the intrinsic variability and distribution of D50 in RBCs are favorable in contributing to uniform O 2 supply in the highly heterogeneous vascular networks. There exist various compensatory mechanisms in the cardiovascular system for regulating O 2 delivery. For instance, a hypoxic environment could elicit a considerable increase in the cardiac output 43 . However, these mechanisms may be undermined under pathological conditions. Hence, RBCs with lower D50 could be better adapted to transport O 2 to distal regions. We speculate that excessive variability in the distribution of D50, as shown with our blood storage study, may reduce the overall population O 2 transport efficacy due to the detrimental contributions of the poorer functioning cells and may be explored as a potential indicator of hematological disorders. RBC distribution width (RDW), which measures the coefficient of variation of RBC size, has been associated with various deleterious clinical conditions such as higher mortality in older adults 44 , The significant difference relative to Day 0: *P < 0.05, **P < 0.0001. The significant difference relative to Day 1: † P < 0.001, † † P < 0.0001. The significant difference relative to Day 7: § P < 0.001, § § P < 0.0001. Error bars correspond to the standard deviation of the data.
critical illnesses 45 , and increased odds of prevalent dementia 46 . Drawing a parallel between D50 and RDW, any major deviation from homeostatic distribution could have undesirable physiological consequences.
Implication on the efficacy of blood transfusion. Transfusion of RBCs has been a stalwart therapeutic intervention for the treatment of acute and chronic anemia 47,48 . Its ultimate goal is to restore the tissue oxygenation level to maintain aerobic metabolism 49 . Currently, there is no consensus on the efficacy of stored RBC transfusion in restoring tissue oxygenation [50][51][52] . In this study, we found that the O 2 diffusion rate of RBCs (D50) was enhanced with blood storage. RBCs undergo a myriad of biophysical and biochemical changes during storage, collectively defined as the storage lesion. Our P50 results show an increase in O 2 affinity with storage duration that has mostly been interpreted as a potentially negative outcome. In contrast, the increase of D50 in stored RBCs suggests the apparent benefit of faster O 2 supply to the tissues. RBC storage has been shown to increase the MCV of RBCs in some studies [53][54][55][56][57][58] , albeit marginally (<5%) especially when stored at 4°C. Yet other studies showed no appreciable changes with storage [59][60][61] . These suggest that MCV changes with RBC storage have minimal impact on the O 2 diffusion dynamics unlike in biological aging. Instead, we surmise that increased membrane permeability caused by oxidation and precipitation of membrane proteins and cellular dehydration (resulting from calcium influx and potassium efflux) 62,63 may have contributed to the enhanced diffusion of solutes across the RBCs. However, it should be noted that this short-term, immediate increase in microvascular oxygenation might be offset by the more deleterious increase in hemolysis arising from the increased cellular mechanical fragility and impaired deformability 64 . Various research groups have investigated the potential of prolonging the duration of stored blood beyond the 42 days period through rejuvenating techniques 65,66 . In this study, RBCs were rejuvenated after 42 days of storage with pyruvate and inosine to Fig. 5 Effect of intracellular 2,3-DPG rejuvenation on the O 2 release rate of RBCs stored for 42 days. a-d MCHC, absolute 2,3-DPG, normalized 2,3-DPG, and P50 before and after rejuvenation (n = 5 of biologically independent samples). e D50 of RBCs before (n = 65 of biologically independent RBCs) and after rejuvenation (n = 93 of biologically independent RBCs). *P < 0.05, **P < 0.001. The central mark of the boxplot indicates the median, and the bottom and top edges of the box indicate the 25th and 75th percentiles, respectively. The whiskers extend to the most extreme data points not considered outliers. restore the depleted intracellular 2,3-DPG so as to investigate the potential effect on the altered O 2 release rate. Our results show that the restoration of 2,3-DPG to basal levels (Day 0) has no significant effect on the O 2 release rate. This is supported by a previous clinical study 67 which showed no improvement in tissue oxygenation in hypoxic patients with increased 2,3-DPG concentrations in RBCs. Apart from the assessment of stored blood, we envision our proposed D50 metric to be used clinically to complement the existing measure of P50 in disease diagnosis, monitoring, or even intervention 1 . This may be especially pertinent in diseases associated with altered hemoglobin (such as sickle cell anemia 68 and beta thalassemia 69 ) or impaired perfusion (such as diabetic microangiopathies 70 and neurodegeneration 71 ). It is of note that our proposed D50 metric, similar to the conventional P50, does not consider the effects of RBC perfusion on the overarching tissue oxygenation efficacy. Aging of RBCs has been known to be associated with the reduction of cellular deformability 8 . The impaired RBC deformability may lead to reduced perfusion at the tissue level and this is well documented in diseases such as malaria 72 , sickle cells anemia 73 , and hereditary RBC membrane disorders 74 . This reduction in perfusion, corresponding to longer capillary transit time, could complement the augmented D50 in enhancing tissue O 2 delivery, possibly as a compensatory mechanism to maintain PO 2 homeostasis.
In summary, this paper presents a MAP for determining the O 2 release rate of RBCs (D50) at the single-cell level. In addition to P50 to assess the RBC's capability of delivering O 2 , we use D50 to complement our assessment of the dynamic process of O 2 delivery. Interestingly, based on our D50 results, biologically older RBCs exhibited a faster O 2 release rate than younger RBCs. We further demonstrated that blood storage enhanced the RBC O 2 release and confirmed that 2,3-DPG rejuvenation had no significant impact on restoring the O 2 release rate. The single-cell level data acquired with the MAP provides additional information (D50) that fills the knowledge gap in the O 2 delivery of RBCs.

Methods
Blood collection and sample preparation. This study was approved by the National University of Singapore Institutional Review Board (H-18-007) and all procedures were performed in accordance with the approved guidelines. Healthy male volunteers aged between 21 and 50 years old, without any known pathological conditions that could affect the properties of the collected blood, were recruited. The number of samples in each experiment was derived by equally distributing the total number of participants recruited. Informed consent was obtained from all recruited participants. No data from research participants were excluded.
All blood collection and storage processes were performed under sterile conditions. Whole blood (20 mL) was collected from the donors into citratephosphate-dextrose (CPD) (Terumo) in the ratio of 7:1. The blood was then centrifuged at 2500g for 10 min at 4°C (Sigma 2-6, Goettingen, Germany) to remove the plasma and buffy coat. The top 2-mm layer of the packed RBCs was also removed to minimize the presence of leukocytes. To examine the effect of in vivo aging, packed RBCs were further density fractionated at 4600g for 30 min at 4°C to obtain younger (top 20% fraction) and older (bottom 20%) RBCs 75 . To examine the effect of storage duration, the packed RBCs were resuspended in saline-adenine-glucose-mannitol (SAGM) additive solution (Terumo) to a hematocrit of 50-70% and stored under standard blood bank conditions (2-6°C). RBC samples were aliquoted aseptically for measurements on Days 0, 1, 7, 14, 28, and 42. To restore intracellular 2,3-DPG, RBCs were incubated in a 10 mM solution of pyruvate and inosine for 1 h at 37°C 76 .
Biochemical assay. MCHC and Hb were quantified using the direct cyanmethemoglobin 77 with Drabkin's reagent (D5941, Sigma), and the absorbance was recorded at 540 nm (Spark ® , Tecan). Absolute 2,3-DPG was measured with a commercial kit (2,3-DPG, Roche). Normalized 2,3-DPG was obtained by normalizing the measured 2,3-DPG level by the corresponding MCHC from the same sample. Blood gases and pH were measured with a commercial blood gas analyzer (Epoc, Alere) and subsequently, the P50 values were determined 78 . Measurements for each sample were repeated successfully with similar results.
Microfluidic analytical platform. The MAP was fabricated from PDMS (Dow Corning, MI) by standard soft lithography and replica molding techniques. The MAP consists of an array of microwells for trapping single RBCs, which are separated from two flanking gas microchannels by a thin PDMS wall (Fig. 1a). The detailed information of the channel design has been provided in Supplementary  Fig. 1. On-chip O 2 concentrations in the microwells were determined by an O 2 sensitive probe, tris(2,2′-bipyridyl)dichlororuthenium(II) hexahydrate (RTDP) (544981, Sigma), dissolved in 1× phosphate-buffered saline (PBS) at 1 mg/mL (pH 7.2). The fluorescence intensity (I) was monitored using our microscope system (IX71, Olympus) under epi-fluorescence (U-MWB2, Olympus). The collisional quenching of this probe by O 2 is described by the first-order Stern-Volmer equation: where I 0 is the intensity at 0% O 2 , K q is the Stern-Volmer constant, and [O 2 ] is the O 2 concentration. The experimental K q in our setup was determined as 2.2 × 10 −3 µM −1 , corresponding to previously reported values 79,80 .
RBC samples were washed with PBS twice before resuspending in RTDP at a hematocrit of 5%. The RBC suspension was then flowed into the MAP at 2 μL/min using a syringe pump (70-4504, Harvard Apparatus). After filling the microwells with RBCs, heavy mineral oil (330760, Sigma) was used to flush the residual RBCs and to isolate single RBCs in the microwell array. We introduced air flows in the two gas channels to initialize the O 2 concentrations in the microwells to 100% saturation before starting the deoxygenation process. N 2 gas was subsequently pumped through the gas channels under constant pressure (400 mbar) using a precision pressure controller (MCFS-EZ, Fluigent) to deoxygenate the microwells. Concurrently, the temporal variations of fluorescence intensity in the microwell were obtained from the fluorescent images (640 × 540 pixels, 4 × 4 binning) recorded with an sCMOS camera (Pco.edge, PCO AG, Kelheim, Germany) at 100 frame/s and a 40× objective (LCPlanFl, NA = 0.60, Olympus).
Statistics and reproducibility. Statistical analyses were performed using a commercially available software package (Graphpad Prism 4). Results are presented as means ± SD. The Mann-Whitney test (two-tailed) was used to assess the statistical significance between microwells with (n = 82 of biologically independent RBCs) and without RBCs (n = 127 of biologically independent RBCs). Unpaired Student's t tests (two-tailed) were used to assess the statistical difference between younger and older RBCs for MCHC, absolute 2,3-DPG, normalized 2,3-DPG, and P50 (n = 5 of biologically independent samples). Mann-Whitney tests (two-tailed) was used to assess the statistical significance between younger (n = 93 of biologically independent RBCs) and older (n = 89 of biologically independent RBCs) RBCs for D50. One-way ANOVA was used to assess the statistical significance across the storage duration for MCHC, absolute 2,3-DPG, normalized 2,3-DPG, and P50 (n = 5 of biologically independent samples). Kruskal-Wallis test was used to assess the statistical significance across the storage duration for D50 (Day 0: n = 72, Day 1: n = 71, Day 7: n = 60, Day 14: n = 55, Day 28: n = 75 and Day 42: n = 65 of biologically independent RBCs). When appropriate, post hoc analyses were performed with Tukey's honestly significant difference test. Paired Student's t test (two-tailed) was used to analyze the data before and after RBC rejuvenation for MCHC, absolute 2,3-DPG, normalized 2,3-DPG, and P50 (n = 5 of biologically independent samples). Wilcoxon test was used to analyze the data before (n = 65 of biologically independent RBCs) and after (n = 93 of biologically independent RBCs) RBC rejuvenation for D50. Single-cell distribution profiles were created using the kernel probability distribution. To describe the distribution profiles, full width at half maximum (FWHM = 2.355×σ) and skewness ð n ðnÀ1ÞðnÀ2Þ P x j Àμ σ À Á 3 Þ were used, where µ and σ represent the mean and SD, respectively, n is the total number of observations and x j is the observation of j. The sample size for each experiment was reported in the figure caption. Data presentations were performed using MATLAB 2019a (MathWorks, USA).