2.1.

Digital Holographic Microscopy

Measurements were performed with a transmission off-axis holographic microscope provided by Ovizio Imaging Systems, Belgium, which has been described before.^14,16 It is a double-differential DHM and is equipped with a $40\times$ NA 0.55 Nikon CFI LWD $40\times$ Cremove objective, a 528-nm Oslon PowerStar SLED (Osram), and a PointGrey Grashopper GS3U332S4 camera (105 fps, exposure time $5\mu \mathrm{s}$) for high-throughput imaging (Fig. 4, Appendix A). By using a low-coherent light source instead of a coherent laser, the image degradation is eliminated and the image quality is improved.^17,18 The light beam emitted from the light-emitting diode first passes a Köhler illumination path and then passes through the sample, which is located in the back focal plane of the microscope objective (Fig. 6, Appendix A). Either a blazed or Ronchi diffraction grating is used to split the incident beam into a diffracted (reference beam) and a nondiffracted beam (object beam). A lens located at focal distance of the grating allows a reshaping of both beams in parallel orientation. A wedge at the optical path of the reference beams induces a small shift of the images produced by diffracted and nondiffracted beams in order to obtain differential interferogram, as described in detail by Dubois and Yourassowsky.¹⁹ In parallel to the wedge, compensation means are introduced in the optical path of the nondiffracted light beam compensating the phase shift introduced by the wedge in the diffracted beam. In a last step, the object and reference beam are focused on a detector, resulting in an interferogram. A detailed description of the working principle and microscope setup is described by Dubois and Yourassowsky.^19,20 Reconstruction of the phase and amplitude images out of the holograms were performed using the OsOne Software Version 5.1 (Ovizio Imaging Systems, Belgium).

2.2.

Microfluidics

A channel with a height of $50\mu \mathrm{m}$, a width of $500\mu \mathrm{m}$, and a total length of $\mathrm{50,000}\mu \mathrm{m}$ was used for precise alignment of a submonolayer of blood cells. The microfluidic poly(methyl methacrylate) (PMMA) system (Fraunhofer ICT-IMM, Mainz) contained five inlets with a top and bottom flows ($z$-sheaths), two side flows ($x$ and $y$ sheaths), and one sample flow. The samples were diluted in 0.9% polyvinylpyrrolidone (PVP, molecular weight $M_w=1.3\mathrm{MD}$, Alfa Aesar) in phosphate-buffered saline (PBS). Sheath flow conditions were established using PBS as a sheath flow buffer. A total flow rate of $1{\mu \mathrm{l}}^{-1}$ was adjusted by the neMESYS Base 120 pump system with five modules (cetoni GmbH). Each slot was equipped with a 2.5-ml gas tight syringe (VWR). During the measurements, all pumps were adjusted to a flow rate of $0.2{\mu \mathrm{l}}^{-1}$. Microfluidic components (tubes, connections) and a six-port injection valve (V-451 six-port medium pressure injection valves bulkhead version .040 Black) were purchased from IDEX Health & Science. Maximum Reynolds numbers in the single digit range ($\mathrm{Re}\approx 4.1$) have been calculated for the measurement state. Therefore, measurements were performed in the laminar flow regime.²¹

2.3.

Data Processing

Data processing and analysis of leukocyte data were performed as described by Ugele et al.¹⁴ In short, floating point phase-shift pixel values for each recorded phase image within the interval of [0,8] were converted to grayscale values. By calculating the median gray value from the first 11 images of the resulting grayscale image, a background image was determined. By subtracting this background image from each phase image, a correction of the background and, therefore, a reduction of the background noise was achieved. For each corrected image, a binary picture was generated by thresholding at the gray level of 28. This resulted in a removal of holes in binary images. As every image contained more than one cell, all objects were segmented. For each object, parameters based on the phase values inside the contour and based on a gray-level co-occurrence matrix²² were calculated. Gating strategies for the differentiation of leukocytes were developed in Kaluza flow cytometry analysis (v.2.1, Beckman Coulter).

2.4.

Sample Preparation

Leukocytes were isolated from the healthy donors by two different preanalytical methods. All human samples were collected from volunteers with informed consent.

3.1.

Comparison of Sample Preparations Methods

To determine the effects of sample preparation on leukocytes, the two parameters, namely, “optical phase delay maximum” (OPDM, maximal phase delay value of the segmented object) and “equivalent diameter” (ED, mean diameter of cell contour area), were investigated (Fig. 1), because changes in these parameters represent changes in the outer and/or inner shape of the cells. OPDM and ED of leukocyte raw data were plotted against the different SPMs, SPM 1 [Fig. 1(a)] and for each application time of SPM 2 [Figs. 1(b)–1(e)]. Initially, four main populations, consisting of cell debris, lymphocytes, granulocytes and monocytes, and aggregates were separated by gating [Figs. 1(a)–1(e)]. First, the amount of cell debris between SPM 1 and SPM 2 was compared. By the use of SPM 1, about 32% of cell debris occurred, which is significantly lower than SPM 2. This can be explained by the use of the erythrocyte depletion kit, which removed a high amount of the remaining erythrocyte fragments. However, the magnetic depletion is costly and time-consuming (5 min incubation time, 10 min magnetic separation, 10 min centrifugation) and, therefore, SPM 1 does not provide a suitable solution for a low-cost application at the point of care. In contrast, a high amount of 81% of cell debris for 5-min application time of SPM 2 was observed. With increased incubation time, a reduction to 51% (10 min), 61% (15 min), and 53% (20 min) was detected. This observation was confirmed by the reduction of cell debris from 85% (5 min) to 40% (10 min), 71% (15 min), and 35% (20 min) in a second measured sample. A higher amount of cell debris at 15 min than at 10 and 20 min occurred in both samples, which was quite unexpected. Nevertheless, the overall observation that an incubation time longer than 5 min is necessary to effectively reduce cell debris was confirmed. Furthermore, the total number of measured leukocytes was significantly higher for SPM 2. With SPM 2, nearly 3.8 as much cells as with SPM 1 were acquired (1870 cells on average for SPM 1 and 7126 cell on average for SPM 2). It seems likely that the magnetic depletion of erythrocytes leads to cell loss, due to sedimentation effects and physical interactions of leukocytes with erythrocyte–magnetic bead complexes induced by the high amount of added magnetic beads ($600\mu \mathrm{l}$ for 1 ml blood) to ensure efficient depletion. Consequently, if samples with low leukocyte numbers are analyzed, SPM 2 should be favored.

Fig. 1

Influence of lysis buffer incubation time on the amount of cell debris and lymphocytes: (a) label-free density plots of leukocytes purified by hypotonic shock followed by magnetic depletion of erythrocytes. (b)–(e) Label-free density plots of leukocytes purified by erythrocyte lysis with the commercial lysis buffer with incubation times of (b) 5 min, (c) 10 min, (d) 15 min, and (e) 20 min. The plots (a)–(e) are divided into four sections indicating cell debris (Db), lymphocytes (L), granulocytes and monocytes (G), and duplets (D). The use of hypotonic shock followed by magnetic depletion of erythrocytes resulted in the lowest amount of cell debris, with greater 5-min incubation time of the lysis buffer significantly reducing the amount of cell debris. Parameters ED and OPDM are plotted. Each density plot shows representative single-cell data from one sample. Percentages of cells are indicated for each gate. (f) The changes in the parameters ED and OPDM of lymphocytes [gate L in (a)–(e)] for the different SPMs, respectively, incubation times are plotted. Overall, only slight changes were observed, indicating that lymphocytes are not affected by different treatment. Data points show the pooled data from two samples. Error bars represent the first standard deviation of the grouped population.

To determine the qualitative impact of sample preparation on leukocytes, the populations of lymphocytes (gate L) and granulocytes and monocytes (gate G) were evaluated. For each, OPDM and ED of the grouped populations of two samples from different donors were plotted against the applied method [Figs. 1(f) and 2(f)]. For lymphocytes, no significant change in OPDM and ED was observed, while the granulocyte and monocyte population showed a clear tendency of decreasing OPDM and increasing ED with increased SPM 2 application time compared to SPM 1. To better analyze the changes in inner and outer cell shapes, the granulocyte and monocyte population was divided into four sections [a1 to a4, Figs. 2(a)–2(e)]. Initially, a shift in the percentage of each section toward lower OPDM was observed, when comparing SPM 1 and SPM 2. Most important, the percentage of cells in gate a1 was significantly reduced from about 16% in SPM 1 to below 2% in SPM 2. With increased incubation time of SPM 2, the percentage of cells in a1 decreased from 1.7% (5 min) to about 0.1% (15 and 20 min). The amount of cells in gate a2 decreased from 50.8% to 23.0%, and the percentage of cells in gates a3 and a4 increased from 45.6% to 65.1% and 1.8% to 11.8%, respectively. These changes in the gates a1 to a4 confirmed the previously observed changes to lower OPDM with increased incubation time of SPM 2.

Fig. 2

Influence of lysis buffer incubation time on granulocytes and monocytes: (a) label-free density plots of granulocytes and monocytes purified by hypotonic shock followed by magnetic depletion of erythrocytes. Only cells from gate G in Fig. 1 are shown. (b)–(e) Label-free density plots of granulocytes and monocytes purified by lysis of erythrocytes by the use of lysis buffer with incubation times of (b) 5 min, (c) 10 min, (d) 15 min, and (e) 20 min. Only cells from gate G in Fig. 1 are shown. The plots (a)–(e) are divided into four sections a1, a2, a3, and a4. In comparison to (a) a shift to lower percentage of cells in gate a1 was observed in (b)–(e). With increased incubation time, a shift to a higher amount of cells in gates a3 and a4 and a shift to a lower amount of cells in gate a2 were observed. Parameters ED and OPDM are plotted. Each density plot shows representative single-cell data from one sample. Percentages of cells are indicated for each gate. (f) The changes in the parameters ED and OPDM of granulocytes and monocytes (gate G in Fig. 1) for the different SPMs, respectively, incubation times are plotted.

As discussed above, more than 5 min incubation time of SPM 2 should be applied to avoid high amounts of cell debris. However, increasing incubation time seems to affect the leukocytes more strongly. An uptake of lysis buffer inside the cells and an efflux of intracellular components could possibly be the reason for the increase in ED and the decrease in OPDM in granulocytes and monocytes over time. An explanation why lymphocytes were less affected could be the higher ratio of nucleus to plasma in the cells, which limits an uptake of lysis buffer. On the contrary, with a smaller ratio of nucleus to plasma, granulocytes and monocytes were more affected, as a higher uptake of lysis buffer might lead to swelling of cells and, therefore, to increased ED and decreased OPDM. To confirm these findings, more samples should be measured, including the analysis of other parameters describing the inner consistency, such as contrast, refractive index, or dry mass. Furthermore, the impact of the sample preparation on leukocyte differentiation should be analyzed to guarantee a reliable classification of leukocytes, as described before.¹⁴ These experiments will be addressed by planned future studies.

3.2.

Impact of Lysis Buffer Incubation Time on Leukocyte Subpopulations

Finally, the impact of SPM 2 on leukocyte subpopulations was investigated by multicolor fluorescence flow cytometry. It was of special interest, if an increased incubation time of the lysis buffer would induce cell loss. This could be directly observed by a change in the relative amount of lymphocytes, granulocytes, and monocytes. Therefore, leukocytes were isolated and labeled as described in Sec. 2. Three blood samples of one healthy donor were independently labeled and each measured one time. As a reference, the leukocyte differential of the Sysmex XN-350 hematology analyzer was used. Granulocytes between 55.1% and 57.9%, lymphocytes between 39.1% and 41.2% and monocytes between 2.9% and 3.8% were determined by flow cytometry (Fig. 3), which are in good agreement with the results from the Sysmex XN-350 (56.2% granulocytes, 38.2% lymphocytes, and 5.6% monocytes). The observed differences of up to 4% between flow cytometry and Sysmex XN-350 measurements are most likely due to the different methods used for leukocyte isolation (SPM 2 for flow cytometry, fully automated lysis and staining of cells by the Sysmex XN-350). Incubation times up to 25 min of SPM 2 did not significantly influence the relative percentage of the different subpopulations, which show the reliability and applicability of the applied leukocyte preparation method also for DHM. Nevertheless, cell loss might occur in SPM 2, but the obtained results strongly indicate that the loss of cells occurs equally distributed across all leukocyte subpopulations.

Fig. 3

Impact of lysis buffer incubation times on leukocytes subpopulations: flow cytometry measurements of one sample for five different lysis buffer incubation times are plotted. By labeling the leukocytes subpopulations with fluorescent markers, the relative concentrations of monocytes, lymphocytes, and granulocytes were determined. Reference measurements of leukocyte subpopulations in whole blood were performed with the Sysmex XN-350 (dashed lines). Each data point shows average values of three measurements. All samples were from one healthy donor and were labeled independently from the others.