Development of a flow standard to enable highly reproducible measurements of deformability of stored red blood cells in a microfluidic device

Abstract

BACKGROUND:

Great deformability allows red blood cells (RBCs) to flow through narrow capillaries in tissues. A number of microfluidic devices with capillary-like microchannels have been developed to monitor storage-related impairment of RBC deformability during blood banking operations. This proof-of-concept study describes a new method to standardize and improve reproducibility of the RBC deformability measurements using one of these devices.

STUDY DESIGN AND METHODS:

The rate of RBC flow through the microfluidic capillary network of the microvascular analyzer (MVA) device made of polydimethylsiloxane was measured to assess RBC deformability. A suspension of microbeads in a solution of glycerol in phosphate-buffered saline was developed to be used as an internal flow rate reference alongside RBC samples in the same device. RBC deformability and other in vitro quality markers were assessed weekly in six leukoreduced RBC concentrates (RCCs) dispersed in saline-adenine-glucose-mannitol additive solution and stored over 42 days at 4°C.

RESULTS:

The use of flow reference reduced device-to-device measurement variability from 10% to 2%. Repeated-measure analysis using the generalized estimating equation (GEE) method showed a significant monotonic decrease in relative RBC flow rate with storage from Week 0. By the end of storage, relative RBC flow rate decreased by 22 ± 6% on average.

CONCLUSIONS:

The suspension of microbeads was successfully used as a flow reference to increase reproducibility of RBC deformability measurements using the MVA. Deformability results suggest an early and late aging phase for stored RCCs, with significant decreases between successive weeks suggesting a highly sensitive measurement method.

Red blood cells (RBCs) are responsible for O2 and CO2 transport through blood vessels.¹ The deformability of healthy discoidal RBCs allows their passage through much narrower blood capillaries.^2,3 Over their 120-day lifespan, aging RBCs are prone to various alterations that are countered by beneficial enzymatic activity and shedding of damaged cell components via microvesiculation. RBCs with impaired deformability are trapped in the spleen and liver, and senescent markers are recognized by macrophages, leading to RBCs engulfment and degradation. The senescent RBCs are removed daily resulting in a heterogeneous age distribution of the RBCs in the bloodstream.

In blood banking operations, RBCs are separated from other blood components before being dispersed in an additive solution and stored at temperature (T) = 2–6°C, allowing preservation of RBC concentrates (RCCs) for up to 42 days prior to transfusion.⁴ During storage, RBCs undergo many changes, which include intracellular [K⁺] and [Na⁺] imbalance, ATP and 2,3-DPG depletion, cytoskeleton and membrane component alteration, decrease of cellular antioxidant capability, and exposure of markers of senescence.⁵ These changes (collectively labeled “storage lesions”) result in generation of microvesicles, impairment of deformability and hemolysis,^5–7 and are mostly not recoverable after transfusion.⁸ The transfusion of RCC units stored for longer durations is known to be less effective in some aspects, namely the reduced ability to perfuse microvasculature,^9–11 reduced RBC survival after transfusion¹² and release of free hemoglobin via intravascular hemolysis.¹³ These observations suggest a link between higher RBC clearance rates for recipients of RCC units with impaired RBC deformability.^7,14 Therefore, a significant research effort has been focused on understanding the mechanisms involved in storage-related impairment of deformability, to eventually improve RCC quality.⁸ Direct in vitro measurement methods, such as filterability,¹⁵ rheoscopy, and ektacytometry¹⁶ exist and have been described extensively.^17,18 The apparent lack of sensitivity of these conventional techniques to properly characterize storage-dependent loss of RBC deformability motivated our present effort in development of microfluidic devices and methods.^17,19–23

Microfluidics enables creating fluidic structures at the same scale as even the smallest blood vessels.²⁴ Various microfluidic designs have been developed to interpret RBCs deformability.^25–32 In some designs, transit of a single RBC through a channel slightly larger and wider than cell diameter is quantified, and deformability is inferred from the geometric shape of the transiting RBC.^26,27 In other cases, RBCs are passed through channels or slits smaller than the cell diameter and RBC deformability is estimated from the transit time required to go through the restrictive area²⁸ or the pressure needed for individual RBCs to successfully cross the slits.²⁹ All those single-cell analysis methods use specialized analysis software allowing relatively high throughput. Other microfluidic devices strictly interpret overall RCC deformability, with measurements of hemolysis generated by the transition of less deformable RBCs³⁰ or from overall RBCs velocity^31,32 through restrictive channels. Single-cell analysis methods are considered more informative but also labor intensive compared to bulk-analysis methods, which lack information on individuals beneath the population.

The MVA (Microvascular Analyzer, Hemanext Inc.) has been developed and used as an in vivo-like bulk analysis strategy to assess RBCs deformability.³³ Fabrication of the microfluidic chip involves molding a precise pattern in polydimethylsiloxane (PDMS), a process described extensively before.^31,34 To enhance chip production efficiency, a mold can hold multiple replicas of the same pattern. PDMS lots, slight variation over patterns on a single mold, manual fabrication and assembly steps, and pressure system and manual operation are factors that may affect reproducibility between devices. This phenomenon underlines the need to enable real-time chip quality control, allowing ease of chip production while keeping the great precision and repeatability seen with the MVA approach using chips from a single mold replica.³³ We hypothesized that side-by-side measurement of RBC sample flow rate and a reference fluid flow rate would display similar fluctuations related to inherent factors, while distinct fluctuations would result from RBC-specific behavior. The aim of the study was to develop an internal standard (IS) composed of suspension of microbeads and test its utility to enhance the repeatability and precision of the MVA approach to assess RBCs deformability. Initially, RBCs deformability of one freshly stored RCC was measured on multiple chips with and without the IS as a proof-of-concept. Then, storage-related impairment of stored RBCs was measured weekly in terms of deformability, filterability, hemolysis, and other RBC quality markers.

MATERIAL AND METHOD

Blood donation

RBCs for method validation:

One unit of leukoreduced RBCs suspended in AS-3 additive solution, collected from a volunteer donor who signed consent for research use, was procured from Rhode Island Blood Center. This RCC was sampled aseptically to characterize the IS utility for compensating MVA chip variability.

RBCs for storage study:

Six (n = 6) volunteers signed an ethical consent form to give whole blood donation (V = 450 mL) at Héma-Québec which was processed using the Atreus system (Terumo BCT). Leukoreduced RCCs were suspended in SAGM (saline-adenine-glucose-mannitol) additive solution and stored at 4 ± 2°C. Those six RCCs were sampled aseptically for analysis at the end of processing (Week 0) and weekly from Weeks 1 to 6 during the hypothermic storage. RBC deformability, biochemical markers ([K+], [Na+], [ATP], [lactate], and [glucose]), mean cell volume (MCV), mean corpuscular hemoglobin concentration (MCHC), hemolysis, and phosphatidylserine (PS) exposure were measured on all samples, as described below.

MVA method

Each single-use MVA chip comprises two identical microchannel networks with separate inlets, both connected to a common outlet (Fig. 1). The classical MVA approach involves deposing RBC samples of controlled Hct (40 ± 1%) into the inlet wells while a movable water column, acting as a negative pressure system, is connected to the outlet. The negative pressure aspirates RBC samples through the restrictive microchannels (3 μm wide at their narrowest, 5 μm deep throughout the network). A specialized software simultaneously assesses flow rates over the last segment of both networks using brightfield microscopy.³³ From 10 images captured at a rate of 100 frames per second (fps), the software measures the shift of the cell pattern between subsequent images using cross-correlation analysis. Displacement measured along the channel is multiplied by the channel cross-sectional area and divided by the time difference between successive frames to calculate the average flow rate in nL/s. Flow rates were measured this way every 10 seconds for a total of 240 seconds, making 24 measurement points per acquisition for each micro-channel network of a device (Fig. 2B).

Fig. 1.

MVA microfluidic network. A negative pressure is applied (▼a) to aspirate the internal standard (IS) (▼b) and the 40% Hct RCCs sample (▼c) through their respective network. Flowrates are digitally measured in the region of interest. Due to the restrictive size of the capillary channels (▼d), relative flowrate measurement is directly influenced by RBC deformability.

Fig. 2.

MVA signal and analysis. (A) Section of the MVA as seen by the CCD camera. This image represents 1 of 10 images taken at 100 fps (both inlets are filled with IS). (B) Analytical signal detected by the software from 40 to 240 seconds for both channels showed in (A). Every slope spikes represent the averaged flow rate from 10 images taken at a single time point. (C) The NMF ability to correct for device-to-device analytical signal variation was determined with 18 different chips using IS in left channels and the same RBC sample in right inlets. (D) Analytical signal detected by the software from 40 to 240 seconds for both channels showed in (C). Since the IS was developed to reach flow rate of averaged fresh RCC samples, it is expected to obtain fresh RCCs flow rate values a little bit higher or lower than IS flow rate. The flow rate direction is indicated by rows.

Development and characterization of the IS

Red dyed-doped carboxyl polystyrene microspheres of d = 1 μm (DCCR004, Bangs Laboratories, Inc.) were chosen because of their opacity and their size expected to easily flow through 3 × 5 μm channels without causing obstructions. Dilution in PBS and glycerol was conducted to adjust viscosity to reach a detectable flow rate similar to those of fresh RBC, as obtained on previous assays (data not shown). The concentration of microsphere was adjusted to minimize channel obstruction while still being detectable by the analysis software. The IS solution was prepared by adding 50 μL of microspheres (5% w/v) to 950 μL of a glycerol (16.6%v/v)-PBS stock solution. Although this formula was developed to cope with obstruction while achieving an optimal flow rate, it also prevented sedimentation of the microspheres due to the similarity of their density with that of the aqueous glycerol solution. The software ability to track either IS and RBC sample was tested on single devices over nine successive acquisitions. Left and right inlets were filled with IS and RBCs at Hct = 40 ± 1%, respectively. Pressure was adjusted between successive acquisitions on the single device to 20, 10, or 5 cm of H₂O. Each pressure was used three times. This part of the study was done with fresh RBCs (RBC_D0) and samples from the same RCC after 2 (RBC_D2) and 5 days (RBC_D5) of storage at room temperature (RT, T = 23°C). Output signal variability was assessed by injecting 25 μL aliquots of IS in both inlets of single-use MVA (Fig. 2A). Under RT conditions, perfusion was triggered by applying a negative pressure using a 20 cm-high water column. Flow rate measurements were compared within the same chip and between different chips (n = 9) to assess the IS utility for correcting the flow rate signal variation.

The MVA fabrication process used a master mold with nine identical patterns. Two MVA chips per pattern were used to measure deformability of fresh RBCs (suspended in AS-3), to evaluate the IS utility to correct the variation of the signal of RBCs flow rate. This RBC sample was diluted in PBS to 40 ± 1% Hct and was run on 18 chips as left and right inlets were filled with IS and RBCs, respectively. The perfusion initiation, length, and flow rate acquisition were conducted as described above and depicted in Fig. 2. The normalized MVA flow rate (NMF) corresponded to the relative flow rate with respect to the internal reference and was defined as NMF = (ƒ_RBCs/ƒ_IS)*ƒ_ISav, where ƒ_RBCs is the sampleʼs flow rate, ƒ_IS is the IS flow rate, and ƒ_ISav is the average flow rate of the IS, determined from previous IS passage on 62 chips. The ƒ_ISav allows data interpretation as flow rate values instead of ratios of flow rates with arbitrary units.

Measurement of RBC deformability over storage

Deformability was assessed weekly by the MVA method and by an adaptation of the filterability method. For NMF measurements, RBCs samples from six SAGM RCCs were diluted in PBS to a 40 ± 1% Hct. All RBC samples were run on two different single-use MVA chips as described above. We used an adaptation of the classical methods to assess RBCs filterability¹⁵ and RBCs mechanical fragility.³⁵ The concentration of RBCs in the samples was adjusted to 1.48 ± 0.07 × 10¹² cells/L with PBS, and total hemoglobin (Hb_Tot) is measured using a standard method (HemoCue Hb 301 System). Under constant pump-induced pressure, RBC aliquots were filtered through d = 3 μm and d = 5 μm pore sized filters (WHA110412, WHA110413; Sigma Diagnostics). Free Hb was measured on filtrates using a standard method (fHb, HemoCue Plasma/Low HB). RBC deformability was then calculated by defining a microfiltration-related mechanical fragility index (MMFI):

$$\text{MMFI}=\left[1-\frac{\text{fHb}3\mu \text{m}-\text{fHb}5\mu \text{m}}{\text{HbTot}-\text{fHb}5\mu \text{m}}\right]$$

Measurement of biochemical and hematological parameters

At each time point, RBC concentration, Hct level, Hb_Tot, MCV, and MCHC were assessed using a hematology analyzer (Coulter AcT 5diff AL, Beckman Coulter). The hemolysis was calculated with the following equation: Hemolysis = ([fHb]/[Hb_Tot]) * (100 – Hct). The [ATP] was measured for RCCs using the ATPlite kit (Perkin Elmer). The [glucose], [lactate], [K⁺], and [Na⁺] were measured using a blood analyzer (ABL90 FLEX PLUS). Serial dilutions in saline were carried out if the results were greater than the detection limits of the apparatus. Alexafluor 488 AnnexineV (A13201, Invitrogen) and Anti-CD235a (551336, BD Biosciences) were used to detect PS exposure by flow cytometry (Accuri C6, BD Biosciences).

Statistical analysis

To calculate the IS variability of the MVA method, coefficient of variation (CV) was used. The changes in RBC deformability and other in vitro quality metrics during storage were compared using generalized estimating equation (GEE) repeated measure analysis, with exchangeable covariance matrix, to account for small sample size and correlation of outcomes within RBC units over time. The changes over time were compared using the beta coefficient, with a 95% confidence interval (CI). p values were based on two-tailed tests, and 95% CI limits were calculated using Waldʼs test.³⁶ Trend tests, to estimate monotonic ordered significance over time for outcomes, were assessed by the rank-based, nonparametric Jonckheere–Terpstra test.

RESULTS

Characterization of the MVA method

To determine the MVA methodʼs channel-to-channel variability, IS flow rates (ƒ_IS) were measured through both micro-channel networks of nine different chips. The flow rate in the right (ƒ_IS,r) and left (ƒ_IS,l) networks of the same chip displayed low CV, with minimal and maximal values of 0.2 and 2%, respectively. The flow rate for different chips had a much larger variation, with minimal and maximal CV values of 0.3 and 12%, respectively. Meanwhile, the ratio ƒ_IS,r/ƒ_IS,l displayed a much more monotonic character over time, with CV < 2% for all measurements with an average 0.99 ± 0.01 a.u. value calculated from all nine ratios measured.

To establish the utility of using the IS for correcting the device variability, the MVA flow rate was measured on 18 chips (two per pattern) for a sample of RBCs stored in AS-3. Figure 3 shows the comparison of variability in raw flow rates, and in NMF data. Raw RBC and IS flow rates had a CV of 10 and 8%, respectively. The CV for the NMF values was 5%.

Fig. 3.

The IS ability to correct for method variability. The MVA method was applied on 18 chips for which left and right inlets were filled with IS and 40 ± 1% Hct RBC sample respectively. (A) RBC raw flow rate and (B) IS raw flow rate were measured by the software and show a CV of 10 and 8% respectively. (C) The normalized MVA flow rate shows gathered data with a CV of 5%.

To determine whether the software could measure the MVA flow rate for both IS and RBC samples simultaneously, we performed the experiments while intentionally varying the driving pressure. For IS, flow rates measured under 50% (10 cm of H₂0) and 25% (5 cm of H₂0) of the maximal driving pressure (20 cm of H₂0) showed values of 50 ± 1% and 25 ± 1% of the flow rate measured under the maximal driving pressure, respectively (see Table S1, available as supporting information in the online version of this paper). For RBC_D0, flow rates measured under 50% and 25% of the maximal driving pressure showed values of 44 ± 5% and 17 ± 2% of the flow rate measured under the maximal driving pressure, respectively. The RBC_D2 and RBC_D5 results suggest that the gap between the proportional drinving pressure and its related proportional flow rate was enhanced by RT storage. Additionally, to evaluate the effect of pressure on the interpretation of RCC deformability, NMF was calculated per pressure value over RT storage. NMF dropped by 22, 35, and 43% between fresh RBCs and RBC_5d units for pressure values of 20, 10, and 5 cm of H₂O, respectively.

Deformability analysis – NMF and MMFI

RBCs deformability was monitored using the MVA method for n = 6 RCC units over the six-week storage duration (Fig. 4). Panel A depicts the RBC average raw flow rate while Panel B shows NMF. RBCs deformability was also monitored using the filterability methods (Fig. 5). Over the whole duration of storage, the MMFI (filterability) and the NMF (MVA perfusion rate) revealed a significant impairment in the average RBCs deformability of 9 ± 3% and 22 ± 6%, respectively (Table 1). The MMFI week-to-week statistical analysis revealed no significant differences between any successive weeks. The GEE analysis showed a significant monotonic decrease in NMF over the whole storage duration (β = −0.07 [95% CI: −0.09–0.04], −0.13 [95% CI: −0.15-.010], −0.11 [95% CI: −0.13–0.08], −0.14 [95% CI: −0.17–0.11], −0.21 [95% CI: −0.25–0.17], −0.23 [95% CI: −0.270.19] for weeks 1, 2, 3, 4, 5, and 6 versus Week 0, respectively. More specifically, a post-hoc analysis adjusted for multiple comparisons using Tukey–Kramer method revealed a significant NMF decrease between all successive weeks except between Weeks 1 and 2, and Weeks 2 and 3 (Table 2).

Fig. 4.

RBCs raw and relative data analysis. Weekly results for all RCCs units (n = 6) as interpreted with the classical MVA method (A) and with the new MVA approach using the relative analysis (B). Generally, trends are clearer and standard errors are smaller with the relative analysis. It is also possible to denote the inter-donor variability on panel B, especially after 6 weeks of storage.

Fig. 5.

Deformability measurements. (A) The MMFI results suggest a slow and steady alteration of RBCs deformability while the MVA results rather suggest a more abrupt and significant (*) impairment at the beginning (Weeks 0 to 1) and at the end (Weeks 3 to 6) of the storage with a latency phase from Weeks 1 to 3. (B) The NMF average impairment of 22 ± 6% after 6 weeks of storage suggests that the MVA method is more sensitive than the filtration method that shows an average MMFI impairment lower than 10% after 6 weeks of storage.

TABLE 1.

RCCs quality markers evolution during storage

Parameters	Week 0	Week 1	Week 2	Week 3	Week 4	Week 5	Week 6
NMF (nL/s)	0.24 ± 0.01	0.22 ± 0.01	0.21 ± 0.01	0.21 ± 0.01	0.21 ± 0.01	0.19 ± 0.02	0.19 ± 0.02
MMFI (a.u.)	99 ± 1	98 ± 1	96 ± 2	95 ± 2	94 ± 2	93 ± 3	91 ± 3
Hemolysis (%)	0.02 ± 0.05	0.05 ± 0.04	0.08 ± 0.04	0.12 ± 0.04	0.20 ± 0.05	0.23 ± 0.06	0.35 ± 0.09
ATP (umol/Hg)	3.8 ± 0.7	3.9 ± 0.5	3.3 ± 0.5	3.1 ± 0.6	2.6 ± 0.5	2.4 ± 0.6	2.3 ± 0.5
pH	6.79 ± 0.04	6.63 ± 0.04	6.52 ± 0.04	6.43 ± 0.05	6.36 ± 0.05	6.31 ± 0.05	6.29 ± 0.06
Lactate (mmol/L)	7 ± 1	13 ± 1	20 ± 1	21 ± 2	22 ± 1	22 ± 2	17 ± 1
GLucose (mmol/L)	26 ± 1	22 ± 1	18 ± 2	18 ± 1	17 ± 1	15 ± 1	15 ± 2
MCV (fL)	92 ± 3	93 ± 4	93 ± 4	94 ± 4	93 ± 4	93 ± 4	94 ± 4
MCHC (g/L)	336 ± 4	337 ± 4	335 ± 4	336 ± 4	333 ± 3	336 ± 6	333 ± 4
pO2 (mm Hg)	40 ± 10	40 ± 10	50 ± 10	60 ± 20	70 ± 30	70 ± 30	90 ± 40
K+ (mmol/L)	4.5 ± 0.3	17 ± 1	26 ± 1	32 ± 2	32 ± 2	28 ± 1	30 ± 1
Na+ (mmol/L)	142 ± 1	129 ± 1	120 ± 2	114 ± 3	114 ± 3	118 ± 3	101 ± 3
PS (%)	0.01 ± 0.01	0.02 ± 0.01	0.12 ± 0.04	0.1 ± 0.1	0.2 ± 0.1	0.30 ± 0.2	0.8 ± 0.5

^*Statistical measures of variations are given as standard deviation. The precision about standard deviation apply to all given data in Table 1.

TABLE 2.

NMF week-to-week significant differences

Weeks	B coeff.	Adj* 95%CI	Adj* p value
0 versus 1	−0.069	−0.114–0.023	0.0002
1 versus 2	−0.058	−0.008–0.125	0.1324
2 versus 3	0.019	−0.020–0.061	0.7825
3 versus 4	−0.032	−0.056–0.009	0.0012
4 versus 5	−0.066	−0.098–0.033	<0.0001
5 versus 6	−0.023	−0.035–0.011	<0.0001

^*Analyses were adjusted for multiple comparisons using Tukey-Kramer method.

Biochemical and hematological quality metrics

Our statistical analysis showed that NMF decreases were significantly related to degree of hemolysis and [lactate] increase (Table 3). The [ATP] and [glucose] were positively related to NMF. Other markers (K+, Na+, MCV, MCHC, and PS exposure) were not significantly associated with NMFs. Finally, Jonckheere-Terpstra test demonstrated a strong negative monotonic trend between hemolysis and NMF over the duration of storage (p < 0.0001).

TABLE 3.

NMF related markers

Variables	B coeff.	Wald 95% CL		Pr > ChiSq
Hemolysis	−0.5448	−0.7003	−0.3893	<.0001
MMFI	−0.0200	−0.0255	−0.0145	<.0001
Lactate	−0.0037	−0.0071	−0.0004	0.0300
ATP	0.0004	0.0003	0.0005	<.0001
Glucose	0.0051	0.0019	0.0083	0.0019

DISCUSSION

Measurement of RBC flow velocity to determine deformability

In a typical MVA experiment, RCC sample flows through the MVA microfluidic network under a constant driving pressure. The observed RBCs flow rate is recorded by specialized software as an indicator of the overall sample deformability (Fig. 1). In a microfluidic device under constant pressure, flow rate is mainly a function of sample viscosity. Viscosity of a RBCs sample is determined by sample Hct and RBCs deformability and aggregability.^37,38 Aggregability is not a significant factor under the high shear stress environment induced by the microscopic scale of the channels,³⁹ and due to a relatively low concentration of plasma proteins in RCC.⁴⁰ The RCC Hct level is adjusted to the physiological value of 40 ± 1% during sample preparation to normalize the impact of hematocrit on flow rate between samples. Preliminary assays (Fig. S1, available as supporting information in the online version of this paper) has shown that the flow rate of RCC sample was inversely influenced by its Hct level, which is in line with observations in similar microfluidic systems.^41,42 The RBCs settling behavior of slow initial sedimentation phase followed by fast sedimentation phase was depicted by an initial steady flow rate followed by a flow rate decreases. The length of the slow sedimentation phase was related to Hct level, which is in line with expectations.^43,44

As for any non-Newtonian fluid, RCC viscosity varies according to the shear stress applied. The RCC shear thinning behavior is caused by the ability of single RBCs to deform and align with the current.⁴⁵ Therefore, RBCs flow rate through the MVA is mainly related to deformability. This relation was demonstrated within a similar microfluidic device in a previous study using glutaraldehyde treatments to alter RBCs deformability,²³ and in a subsequent study of storage-induced RBC deformability with the MVA device.³³

MVA method characterization

The software capacity to track either IS and RBC samples was demonstrated during the pressure variation test. The relation between relative driving pressure and IS flow rates is due to the Newtonian rheological behavior of the IS.⁴⁶ RBC samples present relative flow rates lower than the relative pressures applied. As a shear thinning fluid, RBC sample viscosity is lower at higher shear stress, which explains the greater flow rate drops under lower pressure. The characterization of IS and RBC samples as Newtonian and non-Newtonian fluid, respectively, confirm the software capacity to appropriately track samples flow rate. The shear thinning effect of RBC samples is enhanced by storage at RT, which could be due to the fast deformability impairment occurring under these conditions.⁴⁷ NMF results per pressure value indicate the device could be more sensitive with an applied pressure of 5 cm of H₂O than the set value of 20 cm of H₂O. Although reducing the driving pressure to enhance sensitivity is tempting, such conditions could greatly amplify the plugging of the capillary microchannels thus reducing reproducibility of the MVA flow rate measurements.

Relative flow rate analysis only becomes possible if both microchannel networks of a single chip can reasonably be considered identical, which was demonstrated by the very low CV values between channels with IS in both inlets. Results demonstrated that, while significant variability (CV up to 12%) from different MVA chips can be observed in raw flow rate measurements, the relative flow rate (i.e., NMF) displayed a greater repeatability characterized by much lower CV (~2%). The value of CV encompasses both fluctuations in IS and in device properties, enabling quantitative characterization of RBC deformability using the MVA as suggested by the normalized MVA flow rate obtained for 18 chips. Raw flow rate data (Fig. 4A) are significantly more dispersed, while the NMF (Fig. 4B) yields more homogenous data and allows better visibility to trends over time.

The ideal flow reference should have the following features: flow velocity being software-detectable and similar to 40% Hct RBCs; chemical and rheological stability; producing a repeatable signal over the duration of the experiment while minimizing channel obstructions. Those features were mostly fulfilled by the IS formulation developed in this study, although obstruction events caused analysis rejection and instant remeasurement of NMF on a new chip, underlining the need to improve the IS formula. However, results shown in this study suggest that the use of an IS-like flow standard could help users of microfluidic devices to enhance repeatability of results between devices from multiple molds or replicas from a single mold.

Storage lesions

RBC deformability impairment has previously been reported as part of storage lesions, while the moment of occurrence is still a matter of debate.^48–52 Our results suggest that changes unfold in two phases bridged by a 2-week plateau (between Days 7 and 21) when changes in RBC deformability are not statistically significant. Previous reports of deformability impairment occurring before Day 7 or after Day 21 are partly in line with our observations.^19–53 Bennett-Guerrero et al.²² concluded that RBC deformability measured using ektacytometry was gradually altered over the 42-day storage period. Berezina et al.⁵² found significant early (Weeks 1–2) and late (Weeks 5–6) deformability impairment using RBC filterability. A recent study using microfluidic devices suggests early and late deformability impairment with a still phase between Weeks 1 and 3, just as depicted by NMF results.⁴⁰ Another microfluidic-based study found similar results with significant early (Weeks 1 to 2) and late (Weeks 5 to 6) deformability impairment.²⁹ Compared to those latest studies using sensitive single-cell analysis microfluidic methods, the MVA showed the ability to detect similar deformability impairment, despite being a bulk analysis method. Compared to the previous study using the MVA without the IS, the actual NMF results depict a similar overall deformability impairment while, in the present study, the results were obtained from chips molded on a multiple-patterned mold.³³

An extensive review identifies two main distinct phenomena as responsible for RBC storage lesions.⁵⁴ Storage conditions, namely temperature and additive solution replacing natural plasma environment are believed to be associated with early reversible alterations, while later effects of oxidative damages and ATP depletion would cause irreversible deformability impairments.^55,56 In our study, early [K⁺] and [Na⁺] changes could not be significantly related to deformability impairment. However, the relation between RBCs deformability and both lactate released, ATP depletion and hemolysis found in our study is in line with the scientific literature.⁷ Deformability impairment detected after the third week of storage might be related to those more permanent storage lesions.

Monitoring the ability of RBCs to perfuse through capillary-like channels in a bulk-analysis manner could allow relevant RCC characterization from a blood bank perspective. The MVA approach combined with the IS offers high sensitivity and reproducibility, allowing assessment of subtle changes in RBC deformability. The IS could find applications in real-time quality control testing of PDMS microfluidic devices.

Supplementary Material

Table S1

Table S1. Determination of the software tracking capacity by flow rate measurements under various driving pressures.

Figure S1

Fig. S1. Hematocrit effect on flow rate. RCC sample was diluted in PBS at 20%, 40%, and 60%. HCT effect on sample viscosity is shown by the inverse relationship between flow rate and HCT. Important flow rate decreases at lower HCT suggest RBCs tend to settle quickier at low HCT. Stand-still phase is longer at higher HCT levels, with a non-significant impairment of initial flow rate after more than 4 minutes for 40% HCT samples.