Individuals with sickle cell disease (SCD) face ongoing risk of multi-organ ischemia resulting in chronic disability, frequent hospitalizations, and early mortality.¹ The relationship between hemoglobin (Hb) S polymerization, erythrocyte sickling, and tissue ischemia has been of great interest. Oxygen off-loading and increasing deoxy-Hb concentration promote HbS polymerization, the latter, which has been linked to early erythrocyte deformation or “reversible” sickling.² Eventually, severe polymerization weakens the cell membrane, leading to “irreversibly” sickled cells. Whether the degree of polymerization and the cell's morphologic state, in turn, influence oxygen binding and thus, tissue oxygen availability has been of interest, but technically challenging to study in patients.² Over two decades, Wang and colleagues³ developed an imaging platform, photoacoustic microscopy (PAM), which offers two unique aspects compared to other intravital microscopy systems: (1) high resolution in vivo human imaging using the cuticle as the window to a highly organized vascular bed capable of imaging single capillary loops, and (2) measurements of oxy- and deoxy-Hb levels within single capillaries and single erythrocytes. In this study, we aimed to: (1) characterize capillary morphology and hemodynamic/oxygen metabolic properties in the cuticle nailbed of individuals with SCD compared to healthy controls and (2) track single erythrocytes along the capillary loop to obtain measurements of erythrocyte elongation (ellipticity index, EI) and oxygen saturation before and after tissue oxygen exchange. We hypothesized that erythrocyte EI, as an index of HbS polymerization, would be associated with decreased arteriolar oxygen saturation and/or increased oxygen extraction fraction (OEF) across capillaries—representing compromised “oxygen reserve.”

Adult participants with SCD (HbSS) and controls (HbAA) were prospectively enrolled and excluded for recent hospitalization, chronic transfusion therapy, and history of stem cell transplant. Controls were excluded for any chronic medical disorder. Hb type was confirmed by peripheral blood electrophoresis. Written informed consent was obtained from all participants. PAM is a dual-wavelength optical resolution system with 3 μm lateral and 15 μm axial resolution (Figure S1). The blood absorption spectra from individuals with SCD previously have been found to be similar to that of healthy controls.⁴ The nailbed cuticle imaging procedure consisted of both wide-field and high-speed dynamic imaging. Capillary measurements included density, diameter, and tortuosity. Number and duration of erythrocyte pauses were measured from the spatiotemporal image and its frequency domain image. Multiple hemo-metabolic parameters, calculated from both time-averaged, capillary measurements and single erythrocytes, included: blood velocity, oxygen saturation (sO₂), OEF, and relative metabolic rate of oxygen utilization (MRO₂). Single-cell PAM additionally yielded measurements of arteriolar (sO_{2 in}) and venular (sO_{2 out}) oxygen saturation, from which single-cell OEF was calculated. Elongation of single-cells, ellipticity index (EI), was measured as the mean EI of six frames for each flowing erythrocyte (Figure S2). Details of PAM imaging and statistical methods are described in the Appendix S1.

Ten adults with SCD (HbSS) and healthy controls (HbAA) underwent PAM cuticle imaging sessions, totaling 97 capillaries and 180 erythrocytes imaged (Table S1). Hb and hematocrit measured from peripheral blood correlated with cuticle Hb and hematocrit using PAM (Hb: ρ = 0.825, p = .002; Hct: ρ = 0.800, p = .003). Capillary diameter, density, and tortuosity were statistically increased in SCD versus healthy controls. Time-averaged capillary blood velocity was decreased in SCD versus controls: 62.5 μm/s (51.0, 74.3) versus 69.8 μm/s (63.6, 77.4), respectively (p = .013); capillary OEF was increased in SCD versus controls: 0.205 (0.150, 0.246) versus 0.147 (0.121, 0.188), respectively (p = .049); and capillary MRO₂ was similar in SCD versus controls: 49.9 arbitrary units (a.u.) (21.3, 60.9) versus 39.0 a.u. (26.1, 50.6), respectively (p = .394) (Figure S3).

Animal models of SCD have demonstrated the presence of red blood cell pauses, although the etiologies of pauses such as mechanical obstruction or endothelial adhesion are incompletely understood. Erythrocyte pause count was higher in SCD versus controls: 2 (2, 3) versus 1.5 (1, 2), respectively, (p < .0001). Pause duration was also higher in SCD versus controls: 14.6 s (10.3, 19.9) versus 7.0 s (4.5, 9.0), respectively, (p < .0001). Using a linear mixed-model to account for repeated measures within individuals, we evaluated the relationship between pause duration and capillary morphological and hemo-metabolic properties. Pause duration was inversely associated with blood velocity (p = .009), but not associated with OEF or capillary diameter. We also observed a direct association between pause duration and capillary tortuosity that approached significance (p = .09) (Figure S4).

PAM imaging can resolve individual erythocytes, permitting a metric of elongated erythrocytes using “EI” (Figure S2). EI was increased in SCD versus controls: 0.201 (0.133, 0.245) versus 0.112 (0.075, 0.153), respectively (p < .0001) (Figure 1A). While the SCD cohort demonstrated a wider dynamic range of EI values, substantial overlap was seen across lower EI between the two cohorts, suggesting that SCD patients have a population of normal or near-normal shaped erythrocytes, consistent with the literature on peripheral blood.⁵ Moreover, the population of cells in the SCD cohort with increased EI extended beyond the EI distribution in controls, an observation consistent with HbS polymerization and sickling.

In contrast to velocity measurements using the time-averaged approach, mean single-cell blood velocity was higher in SCD versus controls: 79.8 μm/s (69.4, 90.4) versus 68.9 μm/s (64.7, 79.7), respectively (p = .003) (Figure 1B). During single-cell image acquisition, stalled erythrocytes were not included; therefore, this result likely represents an increase in cell velocity without accounting for the effect of pauses. Consistent with this, the measured velocities from the single-cell method for both control and SCD cohorts were higher than the velocities measured with the time-averaged method for capillaries.

Arteriolar oxygen saturation (sO_{2 in}) was lower in SCD versus controls (p = .014, Figure 1C) suggesting decreased oxygen availability to the capillary bed in SCD. Consistent with time-averaged results, single-cell OEF was increased in the SCD cohort versus controls: 0.101 (0.077, 0.133) versus 0.067 (0.051, 0.091), respectively (p < .0001) (Figure 1D). The finding of higher single-cell OEF is consistent with the observed anemia, diminished erythrocyte velocities, and reduced arteriolar oxygen saturation in patients with SCD, which suggests that greater oxygen extraction from individual erythrocytes is required to meet oxygen metabolic demand.

To examine the relationship between EI, as an index of Hb polymerization and sickling, and measures of erythrocyte velocity and oxygen metabolism, we performed mixed-model linear regression to examine EI as a predictor of velocity, sO_{2 in}, and OEF (Figure 1E–H). In SCD, but not in controls, an increase in EI was associated with slower erythrocyte velocity (β = −98.0, p = .001). Consistent with this, erythrocytes with greater EI were associated with increased OEF (β = 21.5, p = .013). Further, increased EI was associated with decreased sO_{2 in} (β = −17.3, p = .038). Erythrocyte velocity was also modeled in relation to OEF. Both in controls and SCD, individual erythrocyte OEF was inversely proportional to velocity indicating that erythrocyte OEF was increased with prolonged capillary transit time.

In this study, we used cuticle PAM to examine microstructural and physiological measures within capillaries and single-cells to advance our understanding of the pathophysiology underlying tissue ischemia in SCD. In addition to altered capillary architecture, individuals with SCD demonstrated reduced bulk flow velocity, increased OEF, more frequent erythrocyte pauses, and prolonged pause duration compared to controls. In single-cell measurements, we found that erythrocyte elongation (EI) was much higher in SCD versus controls. Moreover, EI distribution in patients with SCD was much broader, suggesting that a large subset of cells are elongated due to HbS polymerization (Figure 1A, above red dashed line). Previous studies have noted a spectrum of erythrocyte morphology and wide variation in the proportion of sickled cells (29%–43%).⁵ Indeed, we found higher EI in the SCD cohort predicted slower erythrocyte velocity, lower sO_{2 in}, and higher OEF, while these relationships were absent in controls. These findings suggest that sickled erythrocytes exhibit lower oxygen reserve than normally shaped erythrocytes, as indicated by both lower sO_{2 in} and increased OEF across the capillary loop, thus severe reductions in blood flow or sO_{2 in} would potentially place the tissue at risk of ischemia.

In summary, we found that increased EI, as a proposed index of polymerized HbS, was associated with distinct single-cell characteristics (velocity, arteriolar oxygen binding, and oxygen off-loading). This study was conducted in SCD patients who were not actively symptomatic, suggesting that baseline oxygen reserve (decreased sO_{2 in} and increased OEF) in patients with SCD may be compromised in a subset of erythrocytes. Increased OEF resulted in preserved MRO₂, suggesting overall, a well-compensated metabolic state. Future PAM studies examining patients during vaso-occlusive crises may reveal an elevated proportion of “sickled” erythrocytes with elevated EI and decreased oxygen reserve. Such patients may be on the precipice of tissue infarction and could be identified for early intervention using PAM technology at the bedside.

The current work represents a proof-of-concept study suggesting that PAM technology may improve our understanding of the relationship between HbS polymerization, erythrocyte morphology, and tissue oxygen transport in SCD. Larger studies will be required to confirm our findings and examine covariates, which could further define these relationships. While our results demonstrate a link between EI and tissue oxygen availability, these relationships do not prove causality. Finally, we did not measure erythrocyte deformability, blood viscosity, or shear rates, which are known to be altered in patient with SCD and likely play a role in thrombosis and hemostasis.⁶ The potential influence of erythrocyte rheology on oxygen availability, however, should not impact the accuracy of hemo-metabolic measurements, nor minimize the cohort differences identified between SCD and controls.

镰状细胞病 (SCD) 患者面临多器官缺血的持续风险，导致慢性残疾、频繁住院和早期死亡。¹血红蛋白 (Hb) S 聚合、红细胞镰状化和组织缺血之间的关系一直备受关注。氧气卸载和增加脱氧 Hb 浓度促进 HbS 聚合，后者与早期红细胞变形或“可逆”镰状化有关。²最终，严重的聚合会削弱细胞膜，导致“不可逆转”的镰状细胞。聚合度和细胞的形态状态是否反过来影响氧结合，从而影响组织氧的可用性一直很有趣，但在患者身上进行研究在技术上具有挑战性。²二十多年来，Wang 及其同事³开发了一种成像平台，即光声显微镜 (PAM)，与其他活体显微镜系统相比，它具有两个独特的方面：(1) 高分辨率体内人体使用角质层作为高度组织化血管床的窗口进行成像，能够对单个毛细血管环进行成像，以及 (2) 测量单个毛细血管和单个红细胞内的含氧和脱氧 Hb 水平。在这项研究中，我们的目的是：(1) 与健康对照者相比，表征 SCD 患者甲床角质层的毛细血管形态和血液动力学/氧代谢特性，以及 (2) 沿着毛细血管环追踪单个红细胞以获得红细胞伸长的测量值（椭圆率指数，EI）和组织氧交换前后的氧饱和度。我们假设红细胞 EI 作为 HbS 聚合的指标，与小动脉血氧饱和度降低和/或毛细血管的氧提取分数 (OEF) 增加有关——代表受损的“氧气储备”。

患有 SCD (HbSS) 和对照组 (HbAA) 的成年参与者被前瞻性地纳入，并因近期住院、长期输血治疗和干细胞移植史而被排除。排除任何慢性疾病的控制。Hb型通过外周血电泳确定。所有参与者都获得了书面知情同意书。PAM 是一种双波长光学分辨率系统，具有 3 μm 的横向分辨率和 15 μm 的轴向分辨率（图 S1）。先前已发现患有 SCD 的个体的血液吸收光谱与健康对照者的血液吸收光谱相似。^4个甲床角质层成像程序包括广域和高速动态成像。毛细管测量包括密度、直径和曲折度。从时空图像及其频域图像测量红细胞暂停的次数和持续时间。根据时间平均、毛细血管测量值和单个红细胞计算的多个血液代谢参数包括：血流速度、氧饱和度 (sO ₂ )、OEF 和氧利用的相对代谢率 (MRO ₂ )。单细胞 PAM 还产生了小动脉 (sO _{2 in} ) 和静脉 (sO _{2 out}) 氧饱和度，从中计算单细胞 OEF。单细胞的伸长率、椭圆率指数 (EI) 被测量为每个流动的红细胞的六帧的平均 EI（图 S2）。PAM 成像和统计方法的详细信息在附录 S1 中进行了描述。

10 名患有 SCD (HbSS) 的成年人和健康对照者 (HbAA) 接受了 PAM 角质层成像，总共成像了 97 个毛细血管和 180 个红细胞（表 S1）。使用 PAM 从外周血测量的 Hb 和血细胞比容与角质层 Hb 和血细胞比容相关（Hb：ρ  = 0.825，p  = .002；Hct：ρ = 0.800，p  = .003）。与健康对照组相比，SCD 的毛细血管直径、密度和曲折度在统计学上有所增加。与对照组相比，SCD 组的时间平均毛细血管血流速度降低：分别为 62.5 μm/s (51.0, 74.3) 和 69.8 μm/s (63.6, 77.4) ( p  = .013)；SCD 与对照组相比毛细血管 OEF 增加：分别为 0.205 (0.150, 0.246) 和 0.147 (0.121, 0.188) ( p = .049); 和毛细管 MRO ₂在 SCD 中与对照相似：分别为 49.9 任意单位 (au) (21.3, 60.9) 和 39.0 au (26.1, 50.6) ( p  = .394)（图 S3）。

SCD 的动物模型已经证明存在红细胞暂停，尽管暂停的病因如机械阻塞或内皮粘附尚不完全清楚。SCD 组的红细胞暂停计数高于对照组：分别为 2 (2, 3) 和 1.5 (1, 2) ( p  < .0001)。SCD 的暂停持续时间也高于对照组：分别为 14.6 秒 (10.3, 19.9) 和 7.0 秒 (4.5, 9.0) ( p  < .0001)。使用线性混合模型来解释个体的重复测量，我们评估了暂停持续时间与毛细血管形态和血液代谢特性之间的关系。暂停持续时间与血流速度呈负相关（p = .009)，但与 OEF 或毛细管直径无关。我们还观察到停顿持续时间与毛细血管迂曲度之间的直接关联接近显着性 ( p  = .09)（图 S4）。

PAM 成像可以分辨单个红细胞，允许使用“EI”测量细长的红细胞（图 S2）。SCD 与对照组相比 EI 增加：分别为 0.201 (0.133, 0.245) 和 0.112 (0.075, 0.153) ( p  < .0001)（图 1A）。虽然 SCD 队列显示出更宽的 EI 值动态范围，但两个队列之间较低的 EI 存在大量重叠，这表明 SCD 患者具有正常或接近正常形状的红细胞群，这与外周血文献一致。⁵此外，SCD 队列中 EI 增加的细胞群超出了对照组的 EI 分布，这一观察结果与 HbS 聚合和镰状化一致。

与使用时间平均方法的速度测量相比，SCD 的平均单细胞血流速度高于对照组：分别为 79.8 μm/s (69.4, 90.4) 和 68.9 μm/s (64.7, 79.7) ( p  = . 003) (图 1B)。在单细胞图像采集过程中，不包括停滞的红细胞；因此，这个结果可能代表细胞速度的增加，而不考虑暂停的影响。与此一致，对于对照和 SCD 队列，单细胞方法测量的速度高于毛细血管的时间平均方法测量的速度。

SCD 中的小动脉氧饱和度 (sO _{2 in} ) 低于对照组 ( p  = .014，图 1C)，表明 SCD 中毛细血管床的氧气可用性降低。与时间平均结果一致，与对照组相比，SCD 组的单细胞 OEF 增加：分别为 0.101 (0.077, 0.133) 和 0.067 (0.051, 0.091) ( p  < .0001)（图 1D）。更高的单细胞 OEF 的发现与观察到的 SCD 患者贫血、红细胞速度降低和小动脉氧饱和度降低一致，这表明需要从单个红细胞中提取更多的氧气来满足氧代谢需求。

为了检查 EI 作为 Hb 聚合和镰状化的指标与红细胞速度和氧代谢的测量值之间的关系，我们进行了混合模型线性回归来检查 EI 作为速度、sO _{2 in}和 OEF 的预测因子（图 1E -H）。在 SCD 中，而不是在对照组中，EI 的增加与较慢的红细胞速度相关（β  = −98.0，p  = .001）。与此一致，具有更高 EI 的红细胞与 OEF 增加相关（β  = 21.5，p  = .013）。此外，增加的 EI 与降低的 sO _{2 in} ( β  = −17.3, p =.038）。还根据 OEF 对红细胞速度进行了建模。在对照组和 SCD 中，单个红细胞 OEF 与速度成反比，表明红细胞 OEF 随着毛细血管传输时间的延长而增加。

在这项研究中，我们使用角质层 PAM 来检查毛细血管和单细胞内的微观结构和生理指标，以促进我们对 SCD 中组织缺血的病理生理学的理解。除了改变毛细血管结构外，与对照组相比，患有 SCD 的个体表现出体积流速降低、OEF 增加、红细胞停顿更频繁以及停顿持续时间延长。在单细胞测量中，我们发现 SCD 的红细胞伸长 (EI) 比对照组高得多。此外，SCD 患者的 EI 分布范围更广，表明大部分细胞因 HbS 聚合而延长（图 1A，红色虚线上方）。先前的研究已经注意到一系列红细胞形态和镰状细胞比例的广泛变化 (29%–43%)。⁵事实上，我们发现 SCD 队列中较高的 EI 预示着红细胞速度较慢、sO _{2 in}较低和 OEF 较高，而这些关系在对照组中不存在。这些发现表明，镰状红细胞的氧气储备低于正常形状的红细胞，毛细血管环中的 sO _{2 in较低和 OEF 增加表明，因此血流量或 sO 2 in 的}严重减少可能会使组织面临缺血风险.

总之，我们发现增加的 EI 作为聚合 HbS 的建议指标，与不同的单细胞特征（速度、小动脉氧结合和氧卸载）相关。这项研究是在没有明显症状的 SCD 患者中进行的，这表明 SCD 患者的基线氧储备（sO _{2 in}降低和 OEF 升高）可能在红细胞亚群中受损。增加的 OEF 导致保留的 MRO ₂, 表明总体上是一种补偿良好的代谢状态。未来对血管闭塞危象期间患者进行的 PAM 研究可能会揭示“镰状”红细胞比例升高，EI 升高和氧储备减少。这些患者可能处于组织梗死的边缘，可以在床边使用 PAM 技术进行早期干预。

目前的工作代表了一项概念验证研究，表明 PAM 技术可以提高我们对 SCD 中 HbS 聚合、红细胞形态和组织氧输送之间关系的理解。将需要更大规模的研究来证实我们的发现并检查协变量，这可以进一步定义这些关系。虽然我们的结果表明 EI 和组织氧气可用性之间存在联系，但这些关系并不能证明因果关系。最后，我们没有测量红细胞变形能力、血液粘度或剪切率，这些已知在 SCD 患者中会发生改变，并且可能在血栓形成和止血中发挥作用。^6个然而，红细胞流变学对氧气可用性的潜在影响不应影响血液代谢测量的准确性，也不应最小化 SCD 和对照组之间确定的队列差异。