Detection of human disease conditions by single-cell morpho-rheological phenotyping of whole blood

Blood is arguably the most important bodily fluid and its analysis provides crucial health status information. A first routine measure to narrow down diagnosis in clinical practice is the differential blood count, determining the frequency of all major blood cells. What is lacking to advance initial blood diagnostics is an unbiased and quick functional assessment of blood that can narrow down the diagnosis and generate specific hypotheses. To address this need, we introduce the continuous, cell-by-cell morpho-rheological (MORE) analysis of whole blood, without labeling, enrichment or separation, at rates of 1,000 cells/sec. In a drop of blood we can identify all major blood cells and characterize their pathological changes in several disease conditions in vitro and in patient samples. This approach takes previous results of mechanical studies on specifically isolated blood cells to the level of application directly in whole blood and adds a functional dimension to conventional blood analysis.


Introduction
Blood is responsible for the distribution of oxygen and nutrients, and centrally involved in the immune response. Consequently, its analysis yields crucial information about the health status of patients. The complete blood count, the analysis of presence and frequency of all major blood cells, constitutes a basic, routine measure in clinical practice. It is often accompanied by analysis of blood biochemistry and molecular markers reflecting the current focus on molecular considerations in biology and biomedicine.
An orthogonal approach could be seen in the study of the overall rheological properties of blood. It is evident that the flow of blood throughout the body will be determined by its physical properties in the vasculature, and their alterations could cause or reflect pathological conditions (1)(2)(3). In this context, blood is a poly-disperse suspension of colloids with different deformability and the flow properties of such non-Newtonian fluids have been the center of study in hydrodynamics and colloidal physics (4). Probably due to the dominant importance of erythrocytes for blood rheology, at the expense of sensitivity to leukocyte properties, this approach has not resulted in wide-spread diagnostic application, maybe with the exception of blood sedimentation rate (5).
Focusing on the physical properties of individual blood cells has suggested a third possibility to glean maximum diagnostic information from blood. Various cell mechanics measurement techniques, such as atomic force microscopy (6)(7)(8), micropipette aspiration (1,(9)(10)(11) or optical traps (12)(13)(14), have been used to show that leukocyte activation, leukemia, and malaria infection, amongst many other physiological and pathological changes, lead to readily quantifiable mechanical alterations of the major blood cells (6,12,(15)(16)(17)(18)(19). These proof-ofconcept studies have so far been done on few tens of specifically isolated cells. This line of research has not progressed towards clinical application for lack of an appropriate measurement technique that can assess single-cell properties of sufficient number directly in whole blood.
This report aims to close this gap by presenting a novel approach for high-throughput singlecell morpho-rheological (MORE) characterization of all major blood cells in continuous flow.
Mimicking capillary flow, we analyse human blood without any labeling or separation at rates of 1,000 cells/sec. We show that we can sensitively detect MORE changes of erythrocytes in spherocytosis and malaria infection, of leukocytes in viral and bacterial infection, and of malignant transformed cells in myeloid and lymphatic leukemias. The ready availability of quantitative morphological parameters such as cell shape, size, aggregation, and brightness, as well as rheological information of each blood cell type with excellent statistics might not only inform further investigation of blood as a complex fluid. It also connects many previous reports of mechanical changes of specifically isolated cells to a measurement now done directly in whole blood. As such, it adds a new functional dimension to conventional blood analysis -a MORE complete blood count -and, thus, opens the door to a new era of exploration in investigating and diagnosing hematological and systemic disorders

Establishment of MORE analysis
In order to establish the normal MORE phenotype of cells found in whole blood, we obtained venous, citrate-anticoagulated blood of healthy donors, of which 50 µl was diluted in 950 µl of measurement buffer with a controlled elevated viscosity, but without any additional labeling, sorting, or enrichment. The cell suspension was then pumped through a microchannel not unlike micro-capillaries in the blood vasculature (Fig. 1A). Brightfield images of the cells, deformed by hydrodynamic shear stresses in the channel (20), were obtained continuously by RT-DC (21) (see Methods; Movie S1). These images revealed distinct differences in overall morphology, brightness, and amount of deformation between all major cell types found in blood (Fig. 1B). RT-DC further enabled the continuous, real-time quantification of the cross-sectional area and of the deformed shape (see detailed description in Methods and Fig. S1) of an, in principle, unlimited number of cells at measurement rates of 100 -1,000 cells/sec (Fig. 1C). For each cell detected and analyzed, an image was saved and the average pixel brightness within the cell determined (Fig. 1D, Fig. S1). This single-cell MORE analysis of whole blood revealed distinct and well-separated cell populations in the space spanned by the three parameters (Movie S2). Notably, size and brightness aloneparameters not unlike those accessible by light scattering analysis in standard flow cytometers -were sufficient for the identification of the cell types (Fig. 1D), so that deformation as additional, independent parameter was available for assessing their functional changes. The identity of the individual cell populations by size and brightness was established by magnetic cell sorting, controlled by fluorescence immunophenotyping, and subsequent MORE analysis brightness. This feature gives access to leukocyte properties directly in whole blood, without the potentially detrimental effects of hemolysis or other separation steps, which are required for analysis with cell mechanics techniques with lower specificity and throughput, or noncontinuous measurement. In extensive tests of the variability of this approach, MORE phenotyping yielded identical results in repeated measurements of blood from the same donor, with sodium citrate added as an anti-coagulant and for different storage times (Fig. S3), between different donors of both sexes (Fig. S4), and blood samples taken at different times during the day (Fig. S5). This robustness served to establish a norm for the different cell types (Fig. 1E). MORE analysis provided the identity and frequency of all major white blood cells as with a conventional differential blood count ( Fig. 1F; Table S1) -obtained from a single drop of blood, with minimal preparation, and within 15 min. Going beyond this current gold-standard of routine blood cell analysis, and importantly also beyond all other single-blood-cell mechanical analysis studies to date, MORE phenotyping allowed the sensitive characterization of pathophysiological changes of individual cells directly in whole blood. In the following, we exemplarily demonstrate, in turn for each of the blood cell types, the new possibilities of gaining MORE information from an initial blood test as a time-critical step in generating specific hypotheses and steering further investigation enabled by this approach.

MORE analysis of erythrocytes
Spherocytosis is a prototypical hereditary disease in humans in which genetic changes (here ankyrin and spectrin mutations) cause abnormal shape and mechanical properties of erythrocytes. Current diagnosis based on shape detection in a blood smear, osmotic fragility assessed by Acidified Glycerol Lysis Time (AGLT) or by osmotic gradient ektacytometry, flow-cytometric determination of staining with Eosin-5-Maleimide (EMA test), or direct detection of the mutation by PCR takes time, requires specific preparation, is costly and does not lend itself to screening. MORE analysis of whole blood of patients with spherocytosis directly revealed significantly stiffer and smaller erythrocytes than normal ( Fig. 2A-C). The differences are so clear that this analysis can serve as a fast primary and cheap screening test for spherocytosis to be followed up by more elaborate analysis. A change in RBC deformability has also been implicated in malaria pathogenesis, since single cells infected by parasites have been reported to be stiffer (18). This insight has not progressed towards clinical application and the gold standard in malaria diagnosis is still a manual thick blood smear analysis. To evaluate whether MORE analysis could provide a sensitive, automated alternative, we analyzed populations of RBCs infected in vitro with Plasmodium falciparum (P.f.) with a parasitemia of 7 -8 % at time points over the 2 day parasite life cycle. We found a clear, significant, and increasing reduction in the deformation of the entire exposed RBC population detectable after 4 h (Fig. 2D-F; Fig. S6). Inspection of the individual cell images revealed the appearance of characteristic features likely associated with the maturation of parasite inside a subset of RBCs (Fig. 2D, E insets; Fig. S6). These features permitted the direct identification of positively infected cells, whose relative frequency peaked at 36 h (Fig. S6). The separate assessment of overtly infected cells showed an even greater deformation reduction than observed in the entire exposed population ( Fig.   2F; Fig. S6), which relates to the possibility of clearance of stiff, infected cells from the circulation by the spleen (22,23). However, this small fraction of stiffer cells alone cannot account for the reduced deformation of the whole population, so that a bystander stiffening of non-infected cells seems involved (14). Inhibition of palmitoylation of membrane proteins by 2-bromo-palmitate (2BP) led to less deformation than in controls and in RBCs treated with palmitic acid (PA; Fig. 2G), with a concurrent reduction in P.f. infectivity (Fig. 2H). While a previous report found no change in infectability of RBCs treated with 2BP (24), the difference could stem from the different RBC receptors involved in invasion by the different parasite clones (3D7 vs. HB3), which in turn are differentially affected by palmitoylation. Thus, MORE analysis has the potential to not only simplify, automate, and speed up malaria diagnosis, but also to provide additional quantitative information aiding research into the pathogenesis of the disease (25).

MORE analysis of leukocytes
While RBC mechanics has already been used for clinical diagnostics using rheoscopes and ektacytometers for over 40 years (15,22,26), leukocyte mechanics has not been utilized for diagnostic purposes. This is likely due to their increased stiffness compared to RBCs and a lack of convenient techniques capable of sufficiently deforming them in suspension -their physiological state. Until recently, techniques with sufficient throughput, obviating the need for specifically isolating the relevant cells of interest, which always bears the potential of inadvertent cell change, did not exist. In this sense, the mechanical phenotyping of diagnostic changes of leukocytes directly in whole blood is the most transformative application area of MORE analysis. For example, there have been proof-of-concept studies on the mechanical changes associated with neutrophil activation, with older reports showing a stiffening, in line with the pronounced actin cortex that is a hallmark of neutrophil activation (6,17). These studies are seemingly in conflict with recent findings reporting a softening of neutrophils with activation (19). While some of the discrepancy could stem from accidental activation by the cell preparation required, the different modes of mechanical testing (11)   Congruently, blood taken from patients with an acute lung injury (ALI) of most likely bacterial origin had larger and more deformed neutrophils compared to healthy controls ( Fig.   3C, E, H). The same neutrophil response was found in blood samples from patients hospitalized with viral respiratory tract infections (RTI; Fig. 3D, E, H). Also monocytes responded by a size increase in both RTI and ALI patients and after in vitro stimulation with S. aureus, but only in viral RTI showed a significantly increased deformation, while blood lymphocytes did not show any consistent response (Fig. 3F-H; Fig. S8). The lymphocyte response changed when analyzing blood of patients with acute Epstein-Barr-virus (EBV) infection, which is known to also stimulate the lymphatic system, where both monocytes and lymphocytes showed an increase in cell size and deformation, while neutrophils showed less of a response (Fig. 3I-L). These results suggest that MORE blood analysis might be sufficiently sensitive to distinguish bacterial from viral infections, and potentially other inflammatory diseases, by the differential response of the selective blood leukocyte populations. This possibility will be followed up in future specific trials. Importantly, MORE blood analysis is of special interest for blood tests in neonatology with patients at high risk of infections but only minute amounts of blood available for diagnostics, or to characterize neutrophils in neutropenic patients, as it merely requires longer data acquisition periods. 4A-C). Cell populations gated for AML revealed less deformed cells but at about the same size compared to healthy and fully differentiated myeloid cells (Fig. 4D, Fig. S9), in line with previous results (7,9,10,12,13). ALL blast cells were larger in size compared to mature lymphocytes, but did not show any consistent trend in deformation ( Fig. 4D; Fig. S9). Since larger cells of identical stiffness should deform more in RT-DC (20,21,31), these findings together imply that mature lymphocytes, ALL blasts, mature myeloid cells, and AML blasts have decreasing levels of deformability, consistent with the composite findings of previous reports (1, 7-10, 12, 13, 17). This is quite different to the general trend in solid tumors, where cancer cells are found to be more deformable than their healthy counterparts (27)(28)(29). Sensibly, the differential stiffness of AML and ALL blasts, and its potential further increase with chemotherapy, has been implicated in the occurrence of leukostasis (8,17,32). MORE analysis might not only permit screening for novel therapeutic targets to soften cells (19,33,34), but also assessing the risk of leukostasis directly in each patient. In addition, by following the ALL blast population in a patient over 12 days of methylcortisone treatment we could monitor the return to the normal morpho-rheological fingerprint of blood as mature lymphocytes successfully replaced the lymphoblasts (Fig. 4E-H). Thus, MORE blood analysis can be used to monitor morphorheological effects of chemotherapy in a quantitative manner. This last finding also touches upon the study of hematopoietic differentiation of cells in the bone marrow, which is an obvious further potential area of application of this approach.

Discussion and Conclusion
MORE phenotyping allows individual blood cell mechanics to be studied in a range of human diseases and takes cell mechanical phenotyping to an entirely new level. While established techniques such as micropipette aspiration (1,(9)(10)(11), indentation by cell pokers and atomic force microscopes (6-8), or optical trapping (12)(13)(14) have provided important proof-ofconcept insight over the last decades, the recent advent of microfluidic techniques approaching the throughput of conventional flow cytometers (19,21,30,35,36) has finally brought mechanical phenotyping close to real-world applications (29,37). Amongst the latter techniques, RT-DC stands out because it can continuously monitor an unlimited number of cells, which enables the direct sensitive assessment of the state of all major blood cell types directly in whole blood. A volume as small as 10 µl can be analyzed cell-by-cell, with only minimal dilution and no labeling, enrichment or separation, which could otherwise cause inadvertent activation of blood cells. The conventional blood count is extended by information about characteristic, and diagnostic, morpho-rheological changes of the major cell types. Cell mechanics and morphology are inherent and sensitive markers intimately linked to functional changes associated with the cytoskeleton (38)(39)(40)(41)(42) and other intracellular shape-determining and load-bearing entities (43,44). Thus, label-free, disease-specific

Real-time deformability cytometry
Real-time deformability cytometry (RT-DC) was carried out as described previously (21).

Data processing in MORE analysis
The raw data obtained from RT-DC measurements consisted of the following information of every detected cell: a bright field image of the cell, the contour of the cell, its deformation value, and the cell size as the cross-sectional area of the cell in the image (Fig.   S1). The deformation was calculated from the convex hull contour of the cell -a processed contour, where all points contributing to concave curvature were removed: where A is the area enclosed by the convex hull contour and l is the length of the convex hull contour. Therefore, deformation is the deviation from a perfectly circular cell image. It describes the change of the cell's shape by the hydrodynamic forces in the measurement channel but may also contain pre-existing shape deviations from a sphere. Image brightness analysis was carried out using the contour information and the image of the cell. The mean brightness of the cell was determined from all pixel values within the cell's contour (Fig. 1D).
With this information the distinction of leukocyte subpopulations was possible in the space spanned by cell size and mean brightness ( Fig. 1D  thrombocytes and microerythrocytes were still found within this gate as seen in Fig. 1C and D. Mechanical analysis of these events constitutes an interesting challenge in that they can be detected and counted, but at present not tested for activation via their deformation given their very small size compared to the channel size, which was chosen to accommodate all cells found in blood. Measurements in the leukocyte gate were carried out over a timespan of 15 min, followed by a separate measurement in the erythrocyte gate for a few seconds until data of 5,000 -10,000 cells were acquired. Measurements for establishing the normal MORE blood phenotype in healthy human volunteers (Fig. 1E), and all measurements directly compared to this norm, e. g., blood samples derived from patients, were carried out at a   (50)

Data availability
The data supporting the findings of this study are available from the corresponding author upon reasonable request.