In this study, we characterized red blood cell deformability (RBCD) parameters pertinent to diabetes, predictive demographic and clinical covariates, and diabetes-specific RBC changes associated with altered RBCD. The diabetes cohort had significantly reduced RBCD, even after adjustments for potential confounding demographic and clinical variables that were either different at baseline (Table 1) or independently associated with reduced RBCD (Fig. 3, Supplementary Table 2). Reduced RBCD was strongly associated with aberrant osmotic-gradient ektacytometry parameters suggestive of RBC dehydration in diabetes, with the most profound associations observed with reduced osmotic fragility (Figs. 4 and 5). A structure-function relationship was observed with reduced RBCD, reduced osmotic fragility and increased hemoglobin-oxygen dissociation (Figs. 5 and 6). Findings reveal pathophysiologic RBC changes in diabetes that are inconsistent with hypothesis linking reduced RBCD to increased RBC rigidity and increased osmotic fragility. However, findings provide strong evidence for disordered oxygen release as a functional consequence of reduced RBCD in diabetes [6, 16,17,18].

In characterizing RBCD in diabetes, we investigated whether specific ektacytometry measurement approaches might be more sensitive and applicable to a diabetes population. We focused only on shear stress gradient parameters—the basis for recent RBCD studies in diabetes and found significant differences in all raw and derived shear-stress gradient measures (Fig. 2) [1, 2, 4]. Given that several ektacytometry-based studies of populations with diabetes used shear stress values ranging from 1.13 Pa to 3pa [1, 2, 4, 31], we investigated differences across shear stresses using two separate analyses (Fig. 2A inset and Fig. 2B), and found the magnitude of between-group differences were highest at 0.95 Pa, 1.69 Pa and 3.0 Pa, validating use of lower shear stress values in investigating RBCD in diabetes [1, 2, 4, 31]. These data indicate that comparative rheological differences in diabetes vs. controls are more discernable at lower, rather than higher shear stresses. An explanation is that shear stress in the microvasculature is predominantly impacted by forces of viscosity rather than forces of high flow (inertia) [32]. Thus, increased viscosity from less deformable RBCs will require comparatively higher shear stresses to maintain flow/perfusion in the microcapillaries.

In contrast to shear stress gradient ektacytometry in which RBCD indices are obtained over a spectrum of shear stresses and at fixed osmolality, osmotic gradient ektacytometry involves dynamic changes to osmolality (hypotonic, isotonic, and hypertonic conditions) at fixed shear stress values [14, 15, 23]. Osmotic gradient ektacytometry provides a powerful yet not previously described means to investigate RBC structural changes in diabetes. Our study showed significantly reduced osmolality measures (Omax, Omin and Ohyper) in the diabetes cohort vs. nondiabetic controls (Fig. 4), consistent with a “leftward shift” along the osmotic axis in the osmotic-gradient curve (Fig. 7: scheme), and indicative of disordered RBC dehydration [23]. While similar “leftward shift” has been described in hematological disorders like sickle cell disease and hereditary xerocytosis [23, 25], a distinguishing feature in the diabetes cohort was the similar maximum elongation index (O.EImax) relative to nondiabetic controls (Fig. 4), findings that suggest structural membrane abnormalities (e.g. defects in alpha or beta spectrin) are less likely a factor in diabetes [23]. While precise underlying mechanisms that lead to dysregulation in RBC hydration are unknown, several possibilities include glucose-mediated dysregulation of RBC membrane channels that regulate volume and cation flux, including Na/K ATPase pump, aquaporins, PIEZO1, K-Cl cotransport (KCC) and Gardos channel [18, 25, 33].

Scheme of osmotic-gradient profile changes in diabetes

It was notable that while the maximum elongation indices derived using osmotic-gradient ektacytometry [O.EImax] were similar in the diabetes cohort and nondiabetic controls (Fig. 4B), the equivalent measure derived from shear stress ektacytometry (EImax) was significantly different in both groups (Fig. 2C). Analyses of shear stress-gradient measures only in participants who obtained osmotic-gradient studies, showed decreased EImax and increased SS1/2, consistent with data from the full cohorts (Fig. 2C, Supplementary Fig. 1) that showed reduced RBCD in diabetes. Despite prior assumptions that osmotic gradient and shear stress gradient data provide similar information, our findings provide evidence that osmotic gradient measures may not be an appropriate or equivalent substitute for shear stress gradient measures in characterizing RBCD in diabetes, or in clinical studies evaluating cardiometabolic risk outcomes [34].

With an understanding of the osmotic gradient profile in diabetes, we investigated the relationship between RBCD (using SS1/2) and these aberrant osmolality measures. Among the osmolality measures, only Omin was significantly associated with reduced RBCD. While Omin values are said to correspond RBC osmotic fragility—the osmolality at which 50% of cells are hemolyzed [13, 23, 26], it was unknown whether this correlation would be applicable to a diabetes population, as studies establishing this correlation were based on modified cell studies unrelated to diabetes. In the diabetes cohort, we found significant association between Omin and osmotic fragility measured using the traditional approach[NaCl 50% hemolysis]26, (Fig. 5A, Methods). Taken together, these data recapitulate a relationship between reduced RBCD and reduced osmotic fragility (Omin), likely resulting from dysregulated RBC hydration status [23].

One potential consequence of reduced RBCD in diabetes and its predisposition to diabetic microvascular complications is impaired hemoglobin-oxygen delivery, a critical function of RBCs. Increased hemoglobin-oxygen dissociation (higher p50 values) suggests weaker hemoglobin affinity for oxygen, facilitating increased oxygen release to the tissues. Our data indicated a structure-function relationship between reduced RBCD (higher SS1/2), lower osmotic fragility (lower Omin) and higher hemoglobin-oxygen dissociation (higher P50), in subgroups within the diabetes cohort and nondiabetic controls subgroups stratified by SS1/2 tertiles (Figs. 5E and 6D). In contrast with more constrained subgroups of the nondiabetic controls, subgroups within the diabetes cohort were more divergent and distinct, with the high-tertile subgroup showing the most extreme values, and the low-tertile diabetic subgroup having similar values as the nondiabetic controls (Figs. 5E and 6D). Independent of RBCD, lower osmotic fragility (Omin) was significantly associated with higher p50 values in the diabetes cohort (p = 0.038, Supplementary Fig. 2A), recapitulating structure-function relationships.

While studies have described changes in p50 as a function of hyperglycemia, our data suggest p50-RBCD relationship may be independent of hyperglycemia for two reasons: First, p50-RBCD association was significant in the normoglycemic nondiabetic controls (Corr.0.28, p = 0.016), albeit with weaker correlation than the diabetes cohort (Corr.0.35, p = 0.008, Fig. 6A). Second, the significant difference in p50 values observed in high- vs. low-tertile diabetic subgroups (Fig. 6B) was not accompanied with differences in glycemic parameters (Supplementary Fig. 2B). Additionally, we found no significant correlation between p50 and either fasting glucose or hemoglobin A1c in the diabetes cohort (Supplementary Fig. 2C and 2D). These data imply that p50 changes may be more a consequence of the structural RBC changes such as osmotic fragility, rather than hyperglycemia. Long-term consequences of p50 changes and the predisposition to vascular complications in diabetes will require dissecting new mechanisms revealed here in combination with longitudinal clinical studies in diabetic subjects.

Our study revealed notable demographic and clinical associations with reduced RBCD, including older age, Black race, male sex, and hyperglycemia–all risk factors for diabetic vascular complications [35,36,37,38,39,40] (Fig. 3). Reduced RBCD was also associated with higher hemoglobin concentrations, likely due to physiologically higher hemoglobin concentrations in males–that is independent of diabetes status (Supplementary Fig. 3A). To further explore the how sex differences and diabetes status variably impact RBC physiology, we compared subgroup differences in mean corpuscular hemoglobin concentration (MCHC), a measure of RBC hydration and viscosity as evidenced by the significant association with Ohyper, an osmotic gradient measure of cellular hydration (Supplementary Fig. 3B) [15, 22, 23, 25]. Within the diabetes cohort, there was comparatively higher MCHC concentrations male vs. female participants, but not within the nondiabetic controls (Supplementary Fig. 3C). Findings suggest association between male sex and reduced RBCD (Fig. 3), may be a combination of pathologically higher MCHC in males with diabetes—perhaps from dehydrated RBCs, and physiologically higher hemoglobin concentrations in males independent of diabetes. These findings highlight the complex physiological and pathological factors that likely modulate RBCD changes independently of diabetes status. Beyond diabetes, findings may offer mechanistic insights in conditions such as RBC storage lesion, where pathophysiological RBC changes and donor-related factors potentially impacts storage quality and possibly RBC transfusion efficacy [41,42,43].

There are several strengths of the study presented here. This study included the first description of the demographic and clinical covariates predictive of reduced RBCD in diabetes, and the first known use of osmotic gradient ektacytometry to characterize structural RBC changes in diabetes, with mechanistic insights that point towards disordered RBC dehydration in diabetes. While cross-sectional study design provides a critical first step in evaluating relationships, limitations include the inability to characterize dynamic and longitudinal relationships between clinical parameters such as glycemic changes, and their effect on RBCD measures. Additionally, this study was not powered to investigate subgroup differences based on race and demographic parameters. The scope of this study did not include other potential factors implicated in RBCD, including genetics, chronic inflammation, oxidative stress [44, 45]. Another limitation is that not all participants obtained all RBC physiology studies (Fig. 1B) due to availability of supplies and/or instrument, however analysis for potential selection bias indicated no marked differences in demographic characteristics other than a single comparison (out of two dozen) that exceeded the ‘moderate’ or ‘medium’ levels (Supplementary Fig. 4) [46].

Lastly, the diabetes cohort was a relatively well controlled cohort with an average A1C of 7.9%, only 24% had microvascular complications and excluded smokers and individuals with known cardiovascular disease, limiting our understanding in populations with advanced complications such as coronary artery disease and chronic kidney diseases While it is possible our findings underestimate the scope of RBCD-related dysregulation in diabetes, this relatively well controlled diabetes cohort offers a mechanistic window in early pathophysiological changes that may precede overt clinical manifestations of diabetic vascular complications.

In conclusion, this study showed that aberrant RBCD in diabetes was independent of significant demographic and clinical covariates, with potentially early pathophysiological RBC changes, as assessed by shear stress ektacytometry, osmotic gradient ektacytometry, and p50. In addition to expanding our understanding of RBC pathophysiology in diabetes, these data provide a critical foundation for future studies aimed at understanding how RBCD measures can be utilized in the early recognition, mitigation, and management of vascular complications in diabetes. Comprehensive larger and longitudinal studies are needed to better understand mechanisms and clinical implications of early RBCD changes in diabetes.