1. Introduction

Leukemia is one of the most common types of cancer in children and adolescents [1, 2]. The incidence of leukemia is increasing in recent years, especially in developing countries. Fortunately, the majority of leukemia patients can be cured by early detection and successful treatments [1]. Bone marrow smears have been widely used to detect leukemia cells by observing the stained cells under a microscope [3]. However, this method is labor intensive and cannot provide quantitative analysis for early detection of leukemia. Immunophenotyping by flow cytomtry or immunohistochemistry requires adequate antibodies for labeling [4, 5]. Cytogenetic approaches, either the conventional karyotyping or fluorescence in situ hybridization (FISH), are also labor intensive and time-consuming [5, 6]. Although polymerase chain reaction (PCR) can be used for early detection of the leukemia cells with high specificity by labeling the cells with immunomagnetic beads [7], it is limited to those subtypes of leukemia in which the molecular alterations are clearly defined. A method that can perform leukemic cell analysis without the staining or bio-labeling of the cells could be of great interest.

The diagnosis of leukemia using the method of bone marrow smears is based on leukemic cell clinical features or morphology. The staining of the cells in bone marow smears is mainly to enhance the image contrast. This could mean that the cell physical properties such as cell size or structure may be used for label-free analysis of leukemic cells [8–10]. It has been reported recently that the leukemic cells have different mechanical properties as compared with lymphocytes by microfluidic measurements [11]. The microfluidic technique for the diagnosis of leukemia may provide high-throughput measurements that are comparable to conventional flow cytometry with miniaturized, economic devices [12–15]. Compared with the measurements of mechanical properties, an optical method for label-free analysis of leukemic cells based on the cell physical properties may serve as a better alternative.

Light scattering occurs when an incident wave interacts with the heterogeneous biological cells. Salzman et al. identified the lymphocytes, monocytes and granulocytes from human peripheral blood by simultaneously measuring forward scattering (FSC) and side scattering (SSC) [16]. Mourant et al. have reported that the scattered light mainly depends on the size of cells for small angle FSC, while at wide angle the organelles will dominant the light scattering [17]. The measurements of scattered light at FSC and SSC, or in a certain 1D polar angular range have been shown as an effective method for label-free cell analysis [16–22]. Su et al. have recently developed a two dimensional (2D) light scattering technique for single cell analysis [23, 24]. The 2D light scattering patterns contain richer information of cells [23–29], and may serve better for label-free cell classifications.

Flow cytometry is a powerful method for the study of leukemic cells based on fluorescence or Raman scattering measurements [4, 15, 30, 31]. The 2D light scattering technique has been incorporated into microfluidic flow cytometry by Su et al. [23, 24, 26], which not only provides a label-free method for cell analysis but also shows the promise for the miniaturization of flow cytometer. In this report, we obtain 2D light scattering patterns from single microspheres, normal granulocytes and leukemic cells by using a label-free static cytometer [32]. This cytometer can obtain 2D light scattering patterns from unlabeled cells without the complex microfluidic fabrications or flow control. Measurements of standard microspheres with a mean diameter of 4.19 μm validate the operation of our label-free static cytometer. Label-free classification of HL-60 cells and K562 cells from normal granulocytes is achieved with high accuracy rates by using our 2D light scattering static cytometric technique, which may provide a label-free method for early screening of leukemia, especially for chronic leukemia.

2. Methods

2.1 Materials and experimental setup

Figure 1(a) shows a schematic diagram of the 2D light scattering label-free cytometer. The essential components of the experimental setup include an optical fiber coupling system, a single scatterer excitation system, and a 2D light scattering pattern recording system. Collimated light from a diode pumped solid state (DPSS) laser (Frankfurt Laser Company, 100 mW, Germany) with a wavelength of 532 nm is coupled into a multimode optical fiber (Thorlabs, 105/125 μm) via a 4 × microscope objective. The fiber localizes and illuminates single static microspheres or cells on chip, and the chip is mounted on a 3-axis translation stage of a microscope (Olympus, BX 53, Japan). To capture the 2D light scattering patterns, the microscope is set to work in a positive defocusing mode [32], where the single scatterers are positioned about 200 μm away from the microscope focal plane, namely increasing the distance between the detector and the scatterers. The 2D light scattering patterns from single scatterers are recorded by using a 14-bit CMOS detector (Canon, APS-C, Japan) via a 10 × microscope objective with a numerical aperture (NA) of 0.25. The 2D light scatting patterns are acquired with an exposure time of 1/25 s.

 

Fig. 1 A schematic diagram of the experimental setup and the determination of the scattered light angular range in the 2D label-free static cytometer. In Fig. (a), the incident beam is coupled into an optical fiber via a microscope objective to illuminate the single static microspheres or cells, and the scattered light is collected by a CMOS sensor. Figure (b) demonstrates the propagation of the scattered light from a scatterer on chip to the microscope objective (as shown by the red circle in Fig. (a)). Figure (c) shows the intersections between the CMOS sensor plane and the cones of the scattered light at fixed polar angles.

Download Full Size | PDF

Figure 1(b) shows the propagation of the scattered light from an excited static scatterer on chip to the microscope objective (as shown by the red circle in Fig. 1(a)). The scattered light goes through a sheet of phosphate buffered saline (PBS) or ultrapure water (80 μm, refractive index 1.334), a glass slide (1.0 mm, refractive index 1.5) and the air (refractive index 1.0) onto the 10 × microscope objective. The angular range of the polar angle $\theta$ is determined by the media that the scattered light propagates and the NA of the microscope objective. Figure 1(c) shows that the scattered light from a homogeneous microsphere can form a cone at a fixed polar angle $\theta$. The intersection of this cone and the CMOS sensor plane (parallel to $xz$ plane) is a parabola. For simplicity, only the scattered light above the $xy$ plane is shown in Fig. 1(c). In the configuration of Fig. 1, scattered light in the polar angle range of ${79}^o\le \theta \le {101}^o$ and azimuthal angle range of ${79}^o\le \phi \le {101}^o$ can be obtained by using our label-free static cytometry.

2.2 Microsphere and cell sample preparation

The standard microspheres are with a mean diameter of 4.19 μm and a standard deviation of 0.27 μm (PS05N, Bangs Laboratories, USA). The microsphere suspension is with a concentration of approximately 2000 microspheres/mL in ultrapure water, and the suspension was ultrasonicated for 5 minutes before use.

Three types of human cells (normal granulocytes, HL-60 acute leukemia cells and K562 chronic leukemia cells) are used with the label-free static cytometer. Normal granulocytes were isolated from human peripheral blood (diluted with 1 × PBS solution) by density gradient centrifugation method, where the diluted blood suspension was layered over Lymphocyte Separation Medium (LTS 1077, TBDScience, China). Leukemic HL-60 and K562 cell lines (Cell Bank of Chinese Academy of Sciences, China) were maintained in Iscove's Modified Dulbecco's Medium (IMDM, Gibco, Invitrogen, USA), supplemented with 10% fetal calf serum, 100 U/mL penicillin, and 10 mg/mL streptomycin in a humidified 5% CO₂ and 95% air atmosphere at 37°C. Both cell lines were cultured in growth medium for less than 15 passages. To avoid potential biological hazards, these three types of living cells were fixed with Immunology Staining Fix Solution (P0098, Beyotime, China) for 30 minutes at room temperature, at a concentration of 10M cells/mL. The normal granulocytes, HL-60 and K562 cells were then suspended in PBS, centrifuged at 1400, 800 and 800 RPM for 10, 3 and 3 minutes, respectively. Then the supernatants were removed. At last, the fixed cells were re-suspended in PBS at a concentration of approximately 2000 cells/mL for single cell analysis.

3. Results and discussion

3.1 Measurements of microspheres by the 2D light scattering static cytometer

The performance of the 2D light scattering label-free static cytometer is verified by using standard microspheres. Figures 2(a) and 2(b) are microscopic images from the 4.19 μm microspheres, and their corresponding experimental 2D light scattering patterns with 4 or 5 fringes are shown in Figs. 2(c) and 2(d). For the randomly collected 148 experimental 2D patterns of 4.19 μm microspheres, there are 83 patterns with 4 fringes and 65 patterns with 5 fringes.

 

Fig. 2 Representative 2D light scattering patterns with 4 and 5 fringes from 4.19 μm (mean diameter) microspheres. Figures (a) and (b) are microscopic images of two different microspheres, and their corresponding 2D patterns with 4 and 5 fringes are shown in Figs. (c) and (d), respectively. Differences in the number of fringes are primarily due to size variations of the microspheres. Numbers of 2D patterns with 4 or 5 fringes of experimental and simulation results are shown in Fig. (e).

Download Full Size | PDF

The sizes of 4.19 μm microspheres (standard deviation 0.27 μm) follow a Gaussian distribution with a mean of 4.19 μm and a variance of 0.0081. Based on this normal distribution, the probabilities of microspheres with diameters changing from 3.92 to 4.46 μm at a step size of 0.03 μm (19 sampling points) can be calculated. Simulated 2D light scattering patterns of these 19 microspheres were obtained by using our Mie theory based algorithm, where all microspheres (refractive index 1.591) were suspended in water. The wavelength of the excitation beam is assumed to be 532 nm, and the polar angle range of the simulated 2D patterns is from ${79}^o$ to${101}^o$. To keep consistent with the number of experimental 2D patterns, the total number of simulated patterns is also 148. The number of 2D patterns with 4 fringes in the simulated results is 85, and is 63 for the 2D patterns with 5 fringes. This is in good agreement with the experimental results as shown in Fig. 2(e).

Previous work for microsphere size differentiation has concentrated on particles with different mean diameters [24, 32]. Here we observe that the microspheres with a given mean diameter (4.19 μm) give rise to different fringe numbers in their 2D patterns, which may correlate to different sizes of microspheres. The validation by using standard microspheres with a given mean diameter demonstrates the capability of our 2D light scattering static cytometry for acquiring certain optical information from cells.

3.2 Label-free classification of normal and leukemic cells

Figure 3 shows the representative 2D light scattering patterns from normal and leukemic cells. Figures 3(a)-3(c) are microscope images from a normal granulocyte, an HL-60 cell and a K562 cell, respectively. Figures 3(d)-3(f) are the corresponding 2D light scattering patterns of Figs. 3(a)-3(c), respectively. In this work, we are interested in light scattering from single cells as in conventional flow cytometry. The cells are not in contact with each other, and the 2D light scattering patterns from a single cell is not affected by other cells. However, our label-free static cytometry can be used for the study of multiple cells [29].

 

Fig. 3 Two dimensional light scattering patterns obtained from normal granulocytes and leukemia cells. Figures (a)-(c) show the microscopic images from a normal granulocyte, an HL-60 cell and a K562 cell, and their corresponding 2D light scattering patterns are shown in Figs. (d)-(f), respectively.

Download Full Size | PDF

The 2D light scattering patterns of both normal and leukemic cells are dominated by speckles. Statistical analysis of these speckles could be used for the classification of different cells. Two parameters, namely the number of speckles and their average area in each 2D pattern (260 by 260 pixels), were extracted by employing watershed algorithm and particle analysis algorithm. Before analysis, the patterns were processed with Gaussian filtering and normalized. A speckle is considered only if its area is greater than 14 pixel². The statistical results of these two parameters for randomly selected 53 normal granulocytes, 53 HL-60 and 53 K562 cells are shown as black squares, red circles and blue triangles in Fig. 4, respectively.

 

Fig. 4 Classification of normal and leukemic cells using 2D light scattering label-free cytometry. Two parameters (the total number of speckles and the average area of these speckles in each 2D light scattering pattern) are extracted from 53 normal granulocytes, 53 HL-60 cells, and 53 K562 cells. Figures (a) and (b) show the classification of HL-60 cells and K562 cells from normal granulocytes, respectively.

Download Full Size | PDF

Figures 4(a) and 4(b) show the results of classification of HL-60 cells and K562 cells from normal granulocytes, respectively. The majority of normal granulocytes, HL-60 and K562 cells, here 75%, are clustered in the ellipsoidal areas in Fig. 4. The number of speckles falls in the ranges from 22 to 29, from 31 to 38, and from 44 to 56, for these 75% normal granulocytes, 75% HL-60 and 75% K562 cells, respectively; and the corresponding average area varies from 785 to 1060, from 555 to 719, and from 390 to 514 pixel². For the purpose of leukemia screening, preference is made that all leukemia cells can be detected and identified from normal granulocytes. The two linear equations (Eqs. (1) and (2)) are drawn for rapid identification of HL-60 acute leukemic cells and K562 chronic leukemic cells, respectively. In Fig. 4(a), the accuracy rate (AR) for the identification of HL-60 cells is 92.5% as determined by Eq. (1). Although some normal granulocytes are determined as “HL-60 cells”, all the HL-60 cells are identified correctly. Figure 4(b) shows that K562 cells are classified well from the normal granulocytes by Eq. (2), and the AR for the identification of K562 cells is 100%. Thus the acute leukemic cells and the chronic leukemic cells can be differentiated from the normal granulocytes by our 2D light scattering static cytometry with high accuracy rates.

It is important to notice that the differentiation of normal granulocytes and leukemic cells in this work is for well controlled and fixed cells. Recently, we have shown that the orientation of a platelet causes changes to its 2D light scattering patterns [26]. The cell shape or structure variations due to different cell states may affect its 2D light scattering patterns, especially for live cell study. In Fig. 4, although the normal granulocytes are well differentiated from the leukemic cells, we notice that the two parameters for cells from the same group could be different. Other parameters, such as parameters for texture character of the 2D light scattering patterns [33] may be studied for the characterization of cells with complex physiological or biological states. The differentiation of cells by using 2D light scattering patterns can also be performed automatically with the well-developed machine learning techniques [29].

4. Conclusion

We demonstrated here a label-free technique based on 2D light scattering static cytometry for the classification of normal granulocytes and leukemic cells (HL-60 acute leukemic cells and K562 chronic leukemic cells). The principle of our cytometric technique for the obtaining of 2D light scattering patterns from static scatterers was introduced. It was shown experimentally that the 2D light scattering patterns of standard microspheres with a mean diameter of 4.19 μm (standard deviation 0.27 μm) have either four or five fringes in the polar angular range from ${79}^o$ to ${101}^o$, and the statistical results of these 2D patterns agree well with Mie theory simulations. This shows that our 2D light scattering static cytometry could acquire detailed information from cells. The accuracy rate for the differentiation of HL-60 cells from normal granulocytes is 92.5%, and is 100% for the differentiation of K562 cells from normal granulocytes by using our label-free static cytometry. Our results show that the 2D light scattering static cytometry may provide a label-free alternative for leukemia screening in clinics.