Microfluidic assessment of PO₂-regulated RBC capillary velocity in ME/CFS Open Access

Myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) is a complex, debilitating disease affecting >20 million people of all ages and races worldwide.^1-3 ME/CFS is characterized by profound fatigue that is not alleviated by rest and persistent postexertional malaise. Additional symptoms include impaired memory, orthostatic intolerance, tender lymph nodes, unrefreshing sleep, headaches, and polyarthralgia, all of which are related to cognitive, immune, and autonomous dysfunctions.^1,4-6 Approximately 70% of patients with ME/CFS cannot work, and at least 25% become bed- or house-bound for extended periods. Furthermore, up to 40% of COVID-19 survivors exhibited persistent, debilitating symptoms that resemble ME/CFS,^7-10 raising concerns about a potential ME/CFS-like pandemic in this population (ie, long COVID).

The causes of ME/CFS are inconclusive, and no clinically accepted treatment and diagnostic biomarkers have been established.¹¹ However, ME/CFS has been consistently associated with abnormal cerebral blood flow (CBF) and impaired blood flow regulation. Studies have shown that patients with ME/CFS exhibited reduced global^12-14 and regional¹⁵ baseline CBF compared to healthy controls (HCs). Reduced baseline regional CBF in the anterior cingulate cortex was associated with the overall symptom severity score of ME/CFS,¹⁵ demonstrating a tight relation between impaired CBF and ME/CFS progression. Additionally, decreased CBF during orthostatic intolerance (a head-up tilt testing) has been observed in ME/CFS.^16-20 The degree of CBF reduction during the tilt testing is also related to ME/CFS severity scores.²¹ Furthermore, functional magnetic resonance imaging and blood oxygenation level–dependent imaging have shown reduced brain responsiveness,²² delayed reaction time,²³ and extended blood oxygenation level–dependent activation regions²³ in ME/CFS, suggesting abnormal CBF regulation, that is, neurovascular coupling (NVC). Using an invasive cardiopulmonary exercise testing, peripheral neurovascular dysregulation, and impaired oxygen extraction were also observed in ME/CFS.²⁴ 

The underlying mechanisms of impaired CBF and NVC in ME/CFS remain unclear. We have previously demonstrated that cerebral NVC in mouse brains is initiated in brain capillaries. In these capillaries, activity-induced decreases in local oxygen tension (PO₂) trigger an increase of red blood cell (RBC) capillary velocity.²⁵ This PO₂-regulated RBC capillary velocity was further confirmed in ex vivo microfluidic capillaries, where reduced PO₂ increases RBC deformability via a hemoglobin–band 3–based molecular switch on the RBC membrane, consequently increasing RBC capillary velocity.²⁶ Thus, RBCs are active players in capillary blood flow, promptly increasing O₂ delivery in response to activity-induced local changes in PO₂. Because ME/CFS exhibits increased levels of inflammation and oxidative stress^27-33 and RBCs are known to interact with cytokines^34-37 and are susceptible to oxidative modifications,^38-41 it is likely that RBC functions in ME/CFS are impaired, contributing to the widely observed CBF impairment. Herein, we measured ex vivo PO₂-regulated capillary velocity of RBCs from patients with ME/CFS and HCs. We explored the use of PO₂-regulated RBC capillary velocity for ME/CFS classification and tested salmeterol xinafoate and xanomeline as potential drugs to improve RBC velocity in ME/CFS. The results presented here unveil a previously unrealized role of RBC in ME/CFS and suggest a potential RBC-based approach for ME/CFS diagnosis and treatment.

Microfluidic devices were fabricated using standard photolithography and polydimethylsiloxane soft-lithography processes. One-milliliter whole blood sample obtained from HCs or patients with ME/CFS (diagnosed based on the Canadian Consensus Criteria⁴²) was centrifuged at 300g at room temperature for 1.5 minutes. The resulting RBC pallets were washed 3 times in phosphate-buffered saline (PBS; Thermo Fisher, Gibco; 20012-PBS; pH = 7.2). Nitrogen-purged PBS was prepared by introducing nitrogen to PBS for 3 hours on days receiving the blood samples. Then, 10 μL of packed RBCs were resuspended into 1 mL of nitrogen-purged PBS and introduced to the microfluidic device through a polyethylene tube (PE-20) under a consistent pressure (1.6 psi) provided by a nitrogen tank. An infusion pressure of 1.6 psi was selected to mimic the physiological arteriovenous pressure gradient, which is ∼90 mm Hg (equivalent to ∼1.7 psi).⁴³ The microfluidic device had a 170-μm long constriction channel with dimensions of 5.05 ± 0.1 μm (width) × 5.94 ± 0.33 μm (height). The PO₂ level inside the microfluidic channel was controlled by immersing the microfluidic device inside a sulfite sink (sodium sulfite; Millipore Sigma; S0505-250G) at concentrations of 0.0, 0.01, 0.1, and 1.0 M. PO₂ in the microfluidic channel was colorimetrically quantified as described previously²⁵ and in the supplemental Methods. RBC movements at each PO₂ level were recorded by a high-speed camera (Phantom Miro 320), and RBC capillary velocity was calculated from image analysis using Image J and the Phantom Camera Control software. Each sample was analyzed using the same microfluidic chip across all 4 PO₂ levels, provided the channel remained undamaged. These experiments were approved by the Stanford Institutional Review Board (IRB) and the University of California Davis IRB, and written consent was obtained before testing or analysis. This study was conducted in accordance with the Declaration of Helsinki and was approved by the University of California Davis IRB (protocol number 1332137-4).

Salmeterol xinafoate (Millipore Sigma; S5068-10MG) or xanomeline (Cayman Chemical; 10790) was dissolved in dimethyl sulfoxide (DMSO; Sigma-Aldrich; 472301; 100 mL) to prepare a stock solution of 1 μM or 10 mM, respectively. The stock solution was then added to the RBC suspension in PBS (Thermo Fisher, Gibco; 20012-PBS; pH = 7.2) to achieve a concentration of 1 nM or 10 nM salmeterol xinafoate or 1 μM or 10 μM xanomeline. These RBC suspensions with salmeterol xinafoate or xanomeline were then incubated for 15 minutes at room temperature before testing.

Leave-one-subject-out (LOSO) evaluation was used to assess accuracy; details of the calculation were described in the supplemental Methods. To evaluate different feature sets, we conducted forward-feature selection, backward-feature selection, and domain knowledge–driven feature selection. The results are generated using the Scikit-learn 1.3.0 software package.

Data are reported as mean ± standard error of the mean. Statistical significance was set at a confidence level for all tests (P < .05), and Prism (GraphPad) was used for statistical analysis. Data shown in supplemental Figure 3 were compared using z tests. This manuscript will be deposited in PubMed Central, and data are available from the corresponding contacts upon request.

To examine PO₂-regulated RBC capillary velocity in individuals with ME/CFS, we extracted RBCs from 35 patients with ME/CFS (aged 17-68 years; 25 females and 10 males) and 23 HC participants (aged 25-77 years; 14 females and 9 males; supplemental Tables 1 and 2) and measured RBC velocity at reduced PO₂ levels. RBCs, suspended in N₂-purged PBS, were injected under a consistent pressure (1.6 psi) into a microfluidic device immersed in an oxygen scavenger sink (sodium sulfite; Figure 1A). The microfluidic device had a constriction channel with dimensions of 5.05 ± 0.1 μm (width) × 5.94 ± 0.33 μm (height), mimicking the size of a typical capillary blood vessel.⁴⁴ These variations in channel size did not lead to significant changes (∼3.9%) in volumetric flow rate. The oxygen scavenger sink contained 0.0, 0.01, 0.1, or 1.0 M of sodium sulfite, resulting in PO₂ levels of 34 , 25, 12, or 0 mm Hg inside the microfluidic capillary, respectively.²⁶ RBC movements under these PO₂ levels were recorded by a high-speed camera and analyzed to obtain RBC velocity. The PO₂ levels inside the microfluidic capillary are within a rage similar to that in the brain during normoxia (20-60 mm Hg) and hypoxia (2-18 mm Hg).⁴⁵ 

The results demonstrated that RBC velocity increased in response to decreasing PO₂ for both HCs and individuals with ME/CFS (Figure 1B), aligning with findings from mouse studies.^25,26 However, at low PO₂ levels (0 and 12 mm Hg), ME/CFS RBC velocities were significantly lower than those of HCs, whereas at elevated PO₂ levels (eg, 25 and 34 mm Hg), ME/CFS RBC velocities were comparable to or even exceeded those of HCs. A linear correlation between RBC velocity and PO₂ was observed for both HCs (R² = 0.99) and patients with ME/CFS (R² = 0.92). The RBC velocity slope, representing the sensitivity of RBC velocity to changes in PO₂, was significantly smaller in ME/CFS than HCs (Figure 1B), indicating that RBCs from patients with ME/CFS were less sensitive to local PO₂ changes. These data demonstrate that PO₂-regulated RBC capillary velocity is impaired in ME/CFS, with a reduced magnitude of capillary velocity at low PO₂ and dampened sensitivity to PO₂ changes.

We further examined whether the participants’ age and sex affect the magnitude and calculated slope of PO₂-regulated RBC capillary velocity. When the calculated slopes were plotted against participants’ age, small correlation coefficients (R²) were observed for both HCs and patients with ME/CFS (for HCs, R² = 0.042; for ME/CFS, R² = 0.00005; Figure 1C), indicating a weak linear correlation between the calculated slopes and participants’ age. The correlation coefficient in ME/CFS was 3 orders of magnitude lower than that in HCs. The magnitude of RBC velocity exhibited a similar change with age, in which a small R² was observed for both HCs (R² = 0.040) and patients with ME/CFS (R² = 0.059). No significant difference was observed in terms of fitting slopes between HCs and ME/CFS (for velocity, P = .32; for the calculated slope, P = .52). On the contrary, capillary velocity of RBCs from female participants was significantly higher than that from male participants for both HCs and ME/CFS (Figure 1D-E). The calculated slopes between male and female participants were not significantly different, suggesting that sex affects the magnitude of RBC capillary velocity but not RBC sensitivity to local PO₂ changes. Within the same sex, calculated slopes in HCs were still significantly higher than that in ME/CFS (for female, P < .001; for male, P < .01; supplemental Figure 1). These results indicate that, in the current cohort, participants’ age has nonsignificant effects on RBC velocity, whereas sex affects the magnitude but not the calculated slope of PO₂-regulated RBC capillary velocity.

To evaluate the diagnostic performance of our metric in distinguishing patients with ME/CFS from HCs, we generated receiver operator characteristic (ROC) curves using the absolute RBC velocity under hypoxia (0 mm Hg) and the calculated slope of RBC velocity (Figure 2A). ROC analysis yielded an area under the curve (AUC) of 0.62 (95% confidence interval [CI], 0.47-0.77; P > .1) for velocity-based calculation and 0.82 (95% CI, 0.71-0.93; P < .0001) for calculated slope–based calculation. Optimal sensitivity and specificity were achieved with slope thresholds between –0.142 mm/s per mm Hg (sensitivity = 86%; specificity = 70%) and –0.069 mm/s per mm Hg (sensitivity = 57%; specificity = 87%). These findings suggest that the slope of RBC velocity is a more robust biomarker than absolute velocity alone.

We further examined whether the severity of ME/CFS affected PO₂-regulated RBC capillary velocity and plotted RBC velocity based on severity scored of participants with ME/CFS, using Bell's disability scale ranging from moderate (M) to moderate-to-severe (M/S).⁴⁶ The results showed that the M group (ie, aged 35-68 years; 10 females and 2 males) exhibited a higher magnitude of RBC velocity than the M/S group (ie, aged 36-68 years; 10 females and 6 males; Figure 2B). The calculated slopes of the M and M/S groups were not significantly different. Furthermore, ROC curve analysis showed an AUC of 0.66 (95% CI, 0.44-0.88; P > .1) and 0.52 (95% CI, 0.30-0.74; P > .5) for velocity slope– and calculated slope–based calculations, respectively, indicating limited significant discriminative power, consistent with the observed lack of statistical difference between the 2 severity groups.

We further evaluated different informative features of PO₂-regulated RBC capillary velocity as a potential biomarker for ME/CFS classification. In addition to RBC velocity and slope, these informative features could also include participants’ sex or age, mean RBC velocity at different PO₂ levels, subslopes (slopes based on 2 adjacent PO₂ levels), and the maximum subslope. Using feature selection techniques and domain knowledge, a range of machine language (ML) algorithms can be trained to identify a specific feature or a combination of features as a classifier that exhibits the highest accuracy, sensitivity, and specificity for ME/CFS classification (Figure 3A)

To identify such a classifier, we first computed the correlation between these informative features and ME/CFS. The calculated RBC velocity slope exhibited the strongest correlation with ME/CFS (Figure 3B), whereas other slope-related features, such as slope-range, (slope)², slope-young, and slope-old, demonstrated high to close-to-highest correlations (supplemental Figure 2A). The magnitudes of velocity at various PO₂ levels did not exhibit comparable correlations with ME/CFS as the aforementioned features, and age and sex showed no significant correlation with ME/CFS (supplemental Figure 2A). We then proceeded to train algorithms using ≥1 feature with high correlations to ME/CFS, choosing 7 ML algorithms that are commonly used for classification: logistic regression, random forest, k-nearest neighbors (KNN), support vector machine (SVM), gradient boost, LightGBM, and AdaBoost. We assessed the accuracy, sensitivity, and specificity of these algorithms using the LOSO evaluation methodology, an established approach in classification.⁴⁷ For each algorithm and feature set combination, all participants except 1 were used for training, and then the left-out participant was used for testing. LOSO simulates evaluation on an independent data set by leaving each participant out of training, allowing validation on unseen data and ensuring robust performance across individuals. This process was iteratively applied to each participant and the overall accuracy, sensitivity, and specificity derived from the average results of all iterations.

Through these evaluations, we demonstrated that the combination of all slope features, that is, features 2 to 6, exhibited the highest average accuracy (71.4%), sensitivity (70.3%), and specificity (70.8%) among all the informative features tested (Figure 3C; supplemental Figure 2B-C). Features 2 to 6 exhibited an accuracy of 77.8% in KNN, a sensitivity of 76% in AdaBoost, and a specificity of 90% in SVM, indicating that these features may contain the most predictive information for this classification task. Meanwhile, across multiple feature sets, SVM showed the highest average accuracy (71.1%) and specificity (79%), whereas LightGBM had the best sensitivity results (70.9%). KNN was the only algorithm that produced accuracy (77.8%), sensitivity (75.8%), and specificity (81.2%) all >75%. Other combinations of algorithm and feature set combinations, feature selection algorithms, and ensemble learning were also experimented with but generated no significant differences and thus were not reported here. Collectively, these results demonstrate that the combination of slope features in PO₂-regulated RBC capillary velocity exhibits the highest accuracy to differentiate ME/CFS from HCs.

Because RBC membrane deformability played a critical role in PO₂-regulated RBC capillary velocity,²⁶ we hypothesized that drugs that can increase RBC membrane deformability could improve PO₂-regulated RBC capillary velocity in ME/CFS.^16,18 To test this hypothesis, we treated RBCs from participants with ME/CFS with salmeterol xinafoate or xanomeline and measured RBC velocity accordingly. Salmeterol xinafoate and xanomeline, used to treat asthma⁴⁸ and schizophrenia,⁴⁹ respectively, are known to increase RBC deformability.^50-53 We isolated RBCs from 9 participants with ME/CFS and incubated them with 0, 1, or 10 nM salmeterol xinafoate (which was first dissolved in DMSO) in PBS at room temperature for 15 minutes. The treated RBCs were then injected into the microfluidic capillary to measure RBC capillary velocity at reduced PO₂ levels. As depicted in Figure 4A, RBCs treated with 1 nM salmeterol exhibited the highest capillary velocity, whereas RBCs treated with 10 nM salmeterol had the lowest velocity. The slope calculated from each participant was not significantly different between salmeterol treatments but trended toward a higher slope in 1 nM salmeterol treatment. Linearity analysis showed a similar trend, with the highest R² for 1 nM salmeterol treatment (R² = 0.81, 0.91, and 0.85 for 0, 1, and 10 nM salmeterol, respectively). These data demonstrate that salmeterol treatments improve the magnitude of RBC capillary velocity in a dose-dependent manner but do not significantly affect RBC velocity slope.

Similarly, we incubated RBCs from another 8 participants with ME/CFS with 0-, 1-, or 10-μM xanomeline and measured RBC capillary velocity after treatment. Given that a previous study demonstrated a significant response in Chinese hamster ovary cells with 1-μM xanomeline,⁵⁴ we chose to treat RBCs with 1- or 10-μM xanomeline. Compared to the 0-μM xanomeline treatment, RBCs treated with 1- and 10-μM xanomeline exhibited increased capillary velocity with increasing xanomeline concentration (Figure 4B). RBCs treated with 10-μM xanomeline had the highest capillary velocity, whereas RBCs treated with 0-μM xanomeline had the lowest velocity. The linear relations between RBC velocity and PO₂ were not significantly different among the xanomeline treatments (R² = 0.92, 0.91, and 0.91 for 0-, 1-, and 10-μM xanomeline, respectively). The slope calculated from each participant increased with increasing xanomeline concentration and exhibited a significant increase between 0- and 10-μM xanomeline treatments. These results demonstrate that xanomeline treatments improve both the magnitude of RBC capillary velocity and RBC velocity slope in a dose-dependent manner. Incubation of RBCs from a HC with salmeterol or xanomeline showed similar trends as observed in ME/CFS, except that the RBC velocity was higher in the HC (supplemental Figure 3A-B), consistent with the results in Figure 1. DMSO had neglectable effects on PO₂-regulated RBC capillary velocity in both HC and ME/CFS (supplemental Figure 3C-D).

ME/CFS is a complex disease affecting multiple cerebral processes, including cerebral hypoperfusion and cognitive dysfunctions. The underlying mechanisms of impaired CBF in ME/CFS remain unclear and are likely caused by a combination of factors.⁵⁵ The impaired PO₂-regulated RBC velocity in ME/CFS observed here suggests compromised oxygen delivery, aligning with previous studies showing altered blood flow patterns, prolonged recovery times for oxygen saturation, and tissue hypoxia.^56,57 The impaired RBC responses to changes in PO₂ (ie, calculated RBC velocity slopes) may result from ME/CFS-induced cellular and molecular abnormality in RBCs. For example, reduced numbers of discoid RBCs,^58,59 increased RBC distribution width,⁶⁰ and reduced RBC deformability⁶¹ have been observed in ME/CFS. Such cellular impairments in RBCs are known to negatively affect RBC capillary flow⁶² and thus likely contribute to the impaired RBC responses to low PO₂ in ME/CFS. Furthermore, the amount of 2,3-diphosphoglyceric acid and oxidized hemoglobin (ie, methemoglobin) in RBCs is increased in patients with ME/CFS.^27,63 Elevated levels of RBC 2,3-diphosphoglyceric acid attenuate PO₂-regulated RBC capillary velocity,²⁶ and increased methemoglobin reduces the amount of unoxidized hemoglobin available for hemoglobin–band 3 interaction at reduced PO₂, which may underlie the molecular characteristics of dampened RBC responses to PO₂ changes in ME/CFS.

Our findings also indicate that capillary velocities of RBCs derived from female cohorts (both HCs and patients with ME/CFS) are higher than those from male cohorts, consistent with prior research showing increased RBC deformability in females.⁶⁴ Female participants also typically displayed a lower mean red cell volume than male participants.⁶⁵ The smaller RBC volume in female participants could facilitate RBC transport across the constriction channel, supporting our data in Figure 1D-E. Although statistical analyses yield a smaller P value for female participants (supplemental Figure 1; for females, P < .001; for males, P < .01), suggesting a potentially more pronounced compromise among female participants, average calculated slopes do not demonstrate a significant difference between male and female participants for both ME/CFS and HCs.

Velocity slope–based ROC analysis yielded a decent AUC of 0.82, which is comparable or higher than that of recent studies in ME/CFS diagnosis. Eguchi et al demonstrated that circulating extracellular vehicles (EVs) containing actin-network proteins, specifically talin-1 and filamin-A, showed promise as blood-based biomarkers for ME/CFS.⁶⁶ EV counts alone yielded an AUC of 0.80 in the ROC analysis. This result supports the view that noninvasive, liquid biopsy–based markers, particularly those reflecting cytoskeletal or structural cell changes, can effectively distinguish patients with ME/CFS from healthy individuals. In another study, circulating microRNA signatures yielded an AUC of 0.65 when distinguishing participants with ME/CFS from HCs (sensitivity = 70%; specificity = 60%).⁶⁷ Sato et al evaluated specific immunoglobulin heavy variable genes as diagnostic markers for ME/CFS and reported an AUC of 0.907 in cohort 1 and 0.846 in cohort 2.⁶⁸ The 2-day cardiopulmonary exercise test, a commonly used functional marker, yielded an AUC value of ∼0.88 when analyzing postexertional drops in ventilatory threshold.⁶⁹ Notably, we observed that absolute RBC velocity under hypoxia (0 mm Hg) yielded a lower AUC (0.62; P > .1), reinforcing the concept that stress-induced responses, rather than static measurements, provide more sensitive and specific diagnostic insights in ME/CFS. This supports the emerging idea that dynamic physiological perturbations, such as hypoxia or exertion, are necessary to reveal subtle pathophysiological dysfunction characteristics of ME/CFS.

The ML approach further confirms the ROC analysis, demonstrating that slope-based RBC features are of the highest accuracy (77.8%), sensitivity (76%), and specificity (90%) for ME/CFS classification among 7 classic classification algorisms. Alternations in RBC shape and deformability under capillary flow have been observed in diseases such as COVID-19, suggesting that examining RBC shape changes in capillary flow could aid diagnosis in other conditions.^70,71 For ME/CFS diagnosis, a recent study by Xu et al^72,73 measured Raman spectral data of peripheral blood mononuclear cells (PBMCs) from patients with ME/CFS and trained the data using an ensemble of multiple classification algorithms (ie, ensemble learning). The classification accuracy reported in Xu et al was ∼91%. However, unlike the LOSO evaluation methodology used here, the training and testing data sets in Xu et al’s study⁷³ were derived from the same pool of participants, potentially leading to information leakage and inflated accuracy. Another related study by Esfandyarpour et al developed a novel nanoelectronic assay to measure impedance of PBMCs in patients with ME/CFS and demonstrated that impedance patterns from PBMCs could be used to train a SVM to establish a classifier for ME/CFS.⁷⁴ However, they did not report classification accuracy in that study. The ML techniques demonstrated here analyzed novel RBC velocity features and offered reliable and robust data characterization that extends beyond classification accuracy.

Lastly, we showed that salmeterol xinafoate and xanomeline treatments improved PO₂-regulated RBC capillary velocity in ME/CFS. Previous studies have demonstrated that stimulating the adenylyl cyclase–cAMP–protein kinase A pathway, either via β-adrenergic receptors or direct activation, increases RBC deformability. Exposure of human RBCs to β-adrenergic agonists such as adrenaline (epinephrine) and isoproterenol resulted in up to 45% increase in the maximal amplitude of cell membrane fluctuations, measured by time-dependent light scattering microscopy (ie, point dark field microscopy), as well as shorter transmit time through a polycarbonate filter (a pore size of 5 μm), suggesting increased RBC deformability.^75,76 This effect was amplified under hypoxic conditions and mediated through a cAMP-dependent pathway. In addition, human RBCs treated with forskolin (direct adenylyl cyclase activator) exhibited increased RBC deformability measured by RBC elongation index using laser ektacytometry.⁷⁶ These effects are modulated by shear stress, demonstrating that β-adrenergic stimulation affects RBC properties under physiological flow conditions. Furthermore, RBCs treated with pentoxifylline, a hemorheological agent, exhibited increased deformability.⁷⁷ Pentoxifylline is a phosphodiesterase inhibitor and leads to increased intracellular cAMP. IV injection of pentoxifylline (1200-1500 mg/d) in patients for 10 days led to a marked decrease in RBC transmit time through an ex vivo 5-μm pore filtration system, indicating improved RBC deformability. Salmeterol xinafoate is a long-acting β₂-adrenergic agonist that acts on β₂ receptors on the RBC membrane, resulting in elevated cAMP levels in RBCs⁵³ and increased RBC deformability,⁷⁵ which is expected to lead to an increased capillary velocity.⁷⁸ RBCs treated with β-adrenergic agonists, that is, isoproterenol,⁷⁵ were known to exhibit a bell-shaped, dose-dependent cell membrane fluctuation, in which RBC deformability increased monotonically when 0.1 to 10 nM β-adrenergic agonists were used. Although the concentration of salmeterol xinafoate used here (1 nM and 10 nM) fell into this concentration range, PO₂-regulated RBC capillary velocity increased in 1 nM but not in 10 nM salmeterol-treated RBCs. This discrepancy is likely because isoproterenol is a nonselective β receptor agonist, whereas salmeterol is a β₂-adrenergic agonist. The bell-shaped, dose-dependent RBC cell membrane fluctuation may shift to a lower concentration when salmeterol was used to treat RBC.

Xanomeline, a muscarinic agonist, activates muscarinic receptors on RBCs, leading to increased cGMP production⁵² and RBC deformability.⁵¹ RBCs express muscarinic acetylcholine (ACh) receptors, particularly the M1 and M3 subtypes, and activation of these receptors has been associated with increased RBC deformability and decreased aggregation. For instance, ACh, a key neurotransmitter that activates muscarinic receptors, can bind to RBCs, increasing RBC deformability and reducing RBC aggregation.⁵¹ In this case, the ACh-muscarinic receptor interaction triggers G-protein signaling, altering downstream protein kinase A activity (eg, band 3 phosphorylation) and enhancing RBC deformability. Thus, incubating RBCs with xanomeline potentially initiates a cascade of events, leading to the phosphorylation of tyrosine residues on band 3 protein within RBCs,⁵⁰ resulting in the detachment of band 3 from its ankyrin tethering on the spectrin-actin cytoskeleton,²⁶ and, consequently, increasing RBC deformability. Because RBC hemoglobin–band 3 interaction regulates PO₂-regulated RBC capillary velocity, the effects of xanomeline on band 3 phosphorylation may explain the improved responsiveness of RBC to reduced PO₂. Such improved RBC capillary velocity upon salmeterol and xanomeline treatments may provide new pharmaceutical approaches to rescue impaired blood flow in ME/CFS.^16,18,56 

In summary, we demonstrated that RBCs from individuals with ME/CFS exhibited significantly impaired capillary velocity at reduced PO₂ ex vivo using microfluidics. This PO₂-regulated RBC capillary velocity provides a new and effective approach for ME/CFS classification, monitoring, and assessment. Furthermore, RBCs treated with salmeterol and xanomeline showed improved capillary velocity and responsiveness to reduced PO₂ in ME/CFS. Given the essential role of RBCs in maintaining physiological function, RBC dysfunction or altered biomechanics may significantly contribute to blood flow abnormalities observed in ME/CFS. We acknowledge the limitations of this study, which is being continued with a larger patient cohort while simultaneously improving detection accuracy, sensitivity, and ease of use. Despite the limited sample size of this study, this RBC-based microfluidic approach interrogates RBC responses to reduced PO₂, revealing previously unrealized roles of RBCs in pathophysiology of ME/CFS and offering a novel approach toward ME/CFS diagnosis.

The authors thank Anna Okumu for participant recruitment and Amrit Shahzad for helpful discussions.

This study was supported by a National Institute of Health grant 1R21AI175960-01A (J.W.) and by the Open Medicine Foundation (A22-2417; J.W.).

Contribution: J.W., R.W.D., and Y.G. designed research; J.W., Y.G., X.L., M.N.-G., S.Z., and M.G. wrote the manuscript; Y.G. and S.Z. performed ex vivo microfluidic experiments; S.R., X.L., and Y.G. performed machine learning experiments; and Y.G., J.W., S.R., X.L., and S.Z. conducted data analysis and calculation.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Jiandi Wan, Department of Chemical Engineering, University of California Davis, CA 95616; email: jdwan@ucdavis.edu.