SEARCH

SEARCH BY CITATION

Keywords:

  • light scattering;
  • flow cytometry;
  • microfluidics;
  • diagnostics;
  • leukemia;
  • minimal residual disease;
  • white blood cells

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED

The prognostic value of assessing minimal residual disease (MRD) in leukemia has been established with advancements in flow cytometry and PCR. Nonetheless, these techniques are limited by high equipment costs, complex, and costly cell processing and the need for highly trained personnel. Here, we demonstrate the potential of exploiting differences in the relative intensities of backscattered light at three wavelengths to detect the presence of leukemic cells in samples containing varying mixtures of white blood cells (WBCs) and leukemic cells flowing through microfluidic channels. Using 405, 488, and 633 nm illumination, we identify distinct light scattering intensity distributions for Nalm-6 leukemic cells, normal mononuclear (PBMC) and polymorphonuclear (PMN) white blood cells and red blood cells. We exploit these differences to develop cell classification algorithms, whose performance is evaluated based on simultaneous acquisition of light scattering and fluorescence flow cytometry data. When this algorithm is used prospectively for the analysis of samples consisting of mixtures of PBMCs and leukemic cells, we achieve an average specificity and sensitivity of leukemic cell detection of 99.6 and 45.2%, respectively. When we consider samples that include leukemic cells along with PMNs and PBMCs, which can be acquired using a simple red blood cell lysis step following venipuncture, the specificity and sensitivity of the approach decreases to 91.6 and 39.5%, respectively. On the basis of the performance of these algorithms, we estimate that 42 or 71 μL of blood would be adequate to confirm the presence of leukemia at an 80% power level in samples containing 0.01% leukemia to either PBMCs or PBMCs and PMNs, respectively. Therefore, light scattering-based flow cytometry in a microfluidic platform could provide a low cost, highly portable, minimally invasive approach for detection and monitoring of leukemic patients. This could offer significant improvements especially for pediatric patients and for patients in developing countries. © 2011 International Society for Advancement of Cytometry

Leukemia is characterized by the overproduction of immature and malignant white blood cells (WBCs) that are unable to carry out normal hematopoietic functions (1). This disease is expected to affect more than 44,000 adults and ∼3,500 children under the age of 15 in the United States (2). Although more adults are affected by this disease compared to children, leukemia is the most common pediatric cancer, accounting for 25% of cancer cases, and the leading cause of death in children under 20 years of age (3, 4). Of the different types of leukemia, acute lymphoblastic leukemia (ALL) is the most common form in children affecting ∼80% of patients. Progress in the development of effective treatment and supportive care has improved the cure rate of ALL, particularly in children (5, 6). However, clinical and biological parameters, such as age and WBC count, may not be sufficient for individualized treatment stratification or risk assessment, resulting in the over-treatment or insufficient treatment of some patients (7, 8). Studies of minimal residual disease (MRD) have aimed to estimate leukemic burden after initial therapy, providing clinicians with an indication of the aggressiveness of the disease and the efficacy of treatment (9–11). Such information could be used to optimize treatment strategy, minimizing patients' risk for relapse and ultimately improving cure rates.

At diagnosis, patients with ALL may have more than 1012 leukemic cells (12). Patients are considered to be in complete remission, after therapy, if fewer than 5% of the cells in the bone marrow are identified as blasts according to morphological standards. However at this level, patients may still harbor as many as 1010 leukemic cells (12). As a result, more than 25% of ALL patients may eventually experience a relapse after induction therapy (13). Advances in flow cytometry and PCR techniques have been able to improve the detection of these residual leukemic cells, which represent MRD, at neoplastic cell proportions of <1% relative to normal nucleated blood cells. A number of flow cytometry and PCR studies have shown that patients with MRD levels of 0.01–0.1% (i.e., 1 leukemic cell to 1,000–10,000 normal nucleated cells) during or after therapy have higher chances of relapse (9, 14–16). However, neither technique is applicable to all patients (17, 18). Specific immnophenotypic or molecular probes may not be found in all patients and the sensitivity and specificity of the technique may vary with the choice of marker (6, 17–19). Furthermore, phenotypic switches or changes in antigen- receptor gene rearrangements that can occur during the course of the disease cause false-negative results (6, 18, 20, 21). Although most MRD assessments are performed on bone marrow aspirates or biopsies, few MRD studies showed that peripheral blood MRD monitoring in patients with ALL is possible and may even provide strong prognostic value in comparison to bone marrow MRD assessment (18, 22). For example, peripheral blood-based approaches can improve the frequency with which residual leukemia can be monitored, especially in children, for whom bone marrow aspirations are particularly difficult and painful. Therefore, improved methods in the detection of leukemic cells within bone marrow aspirates or peripheral blood could significantly impact the treatment of this disease.

Light scattering techniques in conjunction with microfluidic-based devices may be a novel approach suitable for monitoring MRD. Microfluidic-based platforms have the advantage of allowing for single cell analysis and enumeration, as conventional bench top flow systems, but at a lower cost, better portability and higher throughput. Additionally, since light scattering is a natural source of signal contrast, it obviates the need for cell labeling, minimizing the perturbation of biological samples as well as the need for complicated and time consuming processing, as in PCR and standard flow cytometry. Though light scattering in the forward and perpendicular direction are routinely used in standard flow cytometry, light scattering in the backward direction has not been applied thus far. Studies of light scattered in the backwards direction from cell monolayers and tissue samples have shown the sensitivity of backscattering to morphological differences that occur during neoplastic transformation, thus allowing for the discrimination of normal from cancerous cells (23–25). In addition, we have previously shown characteristic differences between the back-scattered light from ALL cells (Nalm-6) and that from red blood cells (RBCs), peripheral blood mononuclear (PBMCs) and polymorphonuclear (PMNs) white blood cells normally found in the blood (Ref. 26; Greiner et al. in press, Cytometry A). Here, we describe the work extending our light scattering spectroscopy studies on static, layered cell samples to flowing leukemic and normal blood cell samples, further assessing the potential applicability of this approach for MRD assessment of ALL. Specifically, leukemic cells, in various concentrations from 0.01% to >20%, are combined with normal blood cells and flowed in single microfluidic channels for light scattering measurements. Classification equations, based on discriminant analysis, are applied to identify leukemic cells from normal blood cells based on differences in their backscattering characteristics. Finally, statistical analyses are performed to prospectively assess the diagnostic performance of our approach.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED

Primary Cell Culture

Nalm-6 cells, which are a pre-B ALL cell line, were cultured in RPMI-1640 medium containing 10% fetal bovine serum (FBS) and 1% Penicillin-Streptomycin (PSF). The cells were incubated at 37°C, 5% CO2, and 85% humidity and split twice a week.

Preparation of Blood Cell Samples

WBCs were isolated from whole blood acquired from healthy human adults (20- to 50-years old) via venipucutre into vacutainers containing sodium heparin. PBMCs were isolated by a standard density gradient technique, as described previously (Ref. 26; Greiner et al., manuscript submitted). PMNs were isolated using a rosette sedimentation technique (27). After the final wash, isolated PBMCs and PMNs were re-suspended in RPMI supplemented with 2% FBS and 1% PSF. RBCs were acquired from the Massachusettes General Hospital blood bank. All procedures were approved by the Tufts University Institutional Review Board.

Sample Preparations for Flow Measurements

For all flow measurements, Nalm-6 cells were stained with a dye solution of 1 mg calcein AM dissolved in 1 mL of anhydrous DMSO (Invitrogen). Briefly, cultured Nalm-6 cells were harvested, centrifuged, and re-suspended in 500 μL of RPMI-1640 without FBS. To this solution, 5 μL of calcein AM solution was added to every 1.0 × 106 Nalm-6 cells for 30 min, and then washed two to three times with RPMI-1640 to remove excess dye.

Both isolated and mixed cell samples were prepared for light scattering measurements. Isolated Nalm-6 cells, PBMCs, PMNs, and RBCs were diluted with RPMI-1640 supplemented with 2% FBS and 1% PSF to a concentration of 3.0 × 105–5.0 × 105 cells mL−1. Mixed cell samples were prepared combining different proportions of Nalm-6 cells, PBMCs, and PMNs at final cell concentrations similar to those of the isolated cell samples. Specifically, we prepared mixed samples with 0.01, 0.1, 1, 2, 5, and 20–35% Nalm-6 cells. The WBC concentration was varied accordingly, but maintaining a 60:40 ratio for PMN:PBMC populations to simulate physiological conditions (Greiner et al., manuscript submitted). For a portion of the 1% Nalm-6 mixed cell samples, we added 10–30% RBCs and 89–69% leukocytes. Prepared samples were placed in 2 mL eppendorf tubes and maintained in ice prior to light scattering measurements.

Microfluidic Chip Design

Pure and mixed cell samples were flowed in microfluidic devices, each of which had three to five single 30 μm (width) × 30 μm (height) × 10 mm (length) channels (Fig. 1a). Microfluidic devices were made of polydimethysiloxane (PDMS), by standard soft lithography techniques, and bonded to 1-mm glass slides using oxygen plasma treatment. To flow fluid through the channels, flexible tubings were added to access ports that were made using commercially available punch tools. Finally, blunt gauge needles were added at the inlet port tubing to interface with standard luer lock syringes.

thumbnail image

Figure 1. (a) Sample microfluidic device with five single 30 μm × 30 μm channels. (b) Schematic of the backscattering system with slit illumination. Legend: D = dichroic, L = lens, M = mirror, BS = beam splitter, Pol = polarizer, and CL = cylindrical lens. (c) Drawing showing the channel dimensions relative to those of the illumination slit. Channel dimensions are: 30 μm × 30 μm × 10 mm (width × height × length). Slit width is 5 μm. Slit length ranged from 29 to 56 μm and the axial resolution varied from 60 to 80 μm depending on the laser. Drawing is not to scale.

Download figure to PowerPoint

Instrumentation Details

Light scattering measurements of flowing cell samples were obtained with a system described previously (Greiner et al., manuscript submitted). The illumination and detection configurations were modified to match the width of the microfluidic channel, in order to detect all the cells flowing through the channel. The optical schematic of the system is shown in Figure 1b. Three laser sources, emitting light at 405, 488, and 633 nm, were combined into a single illumination path by mirrors (M1–M4) and dichroic filters (D1–D2). The beams went through a linear polarizer (Pol1) before passing through a cylindrical lens, which focused the beam along a line. This slit was re-imaged across the microfluidic channel by an achromatic lens (L1; f = 150 mm), and an infinity corrected 40× microscope objective, (NA 0.6; Olympus). The slit width, as assessed by comparing the image of the slit against that of a graticule was ∼5 μm and the slit lengths were 56, 29, and 37 μm for 405, 488, and 633 nm illumination, respectively, determined from knife edge measurement. The variability in the slit length was due to differences in the divergence and diameter of the laser beams. The axial resolution of the system, determined from the full width at half maximum of the intensity distribution recorded when scanning a mirror along the axial direction, was 60 μm for 405 nm, and 80 μm for 488 and 633 nm. The microfluidic device was oriented such that the length of the slit traversed the channel width (Fig. 1c). The alignment of the channel across the illumination slit was verified by imaging the microfluidic device onto a CCD, using the microscope objective and a 150 mm focal length lens (L2), with a 530 nm LED providing sample illumination.

The backscattered light from a flowing cell that traversed the illumination slit was collected by the microscope objective. The collected backscattered light was directed towards the analyzer (Pol2) in the detection path by a 50/50 beam splitter (BS1) and imaged onto a 150 μm width × 3000 μm length slit aperture. This detection slit was placed confocal to the excitation slit, to minimize detection of out-of-focus scattering. To further reduce the contribution from specular reflections, a beam block was placed so as to cover over half of the microscope objective back aperture. In addition, the illumination beam was titled by ∼18° by laterally moving mirror M5, which steered the input beam prior to entering the back of the objective, so the objective collected light scattered at angles between 0° and 18° about the exact backscattering direction. Backscattered light that made it through the pinhole was spatially separated by dichroic filters (D3–D4) and detected by three photomultiplier tubes (PMTs; Hamamatsu), each placed behind a 405/10 nm (center wavelength/bandpass width), a 488/10 nm, or a 633/10 nm bandpass filter, to detect scattered light at each one of the illumination wavelengths. To assist in the evaluation of results acquired from light scattering measurements, a fourth PMT was added to detect fluorescence in the 500 to 590 nm range from labeled cells excited with the 488 nm laser. The scattered and fluorescence light intensities were digitally sampled at 25 KHz and recorded onto a computer for analysis.

Flow Measurement Details

To flow cells in a microfluidic channel for light scattering measurements, the outlet tubing was placed into eppendorf tubes and the cell suspension was drawn by a syringe pump (Harvard Apparatus) at a flow rate of 3 μL min−1. For each sample, a new microfluidic channel was used to minimize contamination. Approximately 15 min of time traces were recorded from each PMT for each sample. At the end of the experiment, the backscattering from a spectralon on a microscope slide was measured as a calibration standard for the angular and spectral dependence of the system. Since it was typically necessary to use neutral density filters to change the illumination power for spectralon measurements to prevent PMT saturation, differences in the illumination power during cell sample and spectralon measurements were recorded and accounted for during calibration. Independent flow measurements of each type of isolated and mixed cell sample were repeated at least three times, on different days, to assess reproducibility.

Data Analysis

To analyze flow measurements, we used a previously developed peak counting Matlab program to detect and characterize fluorescence signals from flowing cells (28, 29). The detected peaks from all the PMTs were compared to determine correlated peaks, which were peaks that occurred within a time tolerance of 6.5 × 10−4 seconds. Correlated peaks from the entire data trace were calibrated by the average light scattering signal from the spectralon and Pfactor, which is the ratio of the illumination power between the cell and spectralon measurements, to calculate the calibrated scattering intensity at each wavelength, i.e., equation image. In our experiments, time correlated peaks in all three scattering wavelengths with no corresponding fluorescence signal (“3λ correlated peaks”) were considered to be those from PBMCs, PMNs, or RBCs. Correlated peaks in all three scattering wavelengths with associated fluorescence peaks (“4λ correlated peaks”) were attributed to calcein labeled Nalm-6 cells.

To visualize the light scattering signal distribution from the cells, the calibrated light scattering intensities at two wavelengths were displayed as density plots, in RGB. Each cell type was assigned a particular color with the color intensity used to represent the signal distribution, analogous to a histogram, against the two wavelength parameters. For mixed samples, the light scattering peak intensities from PBMCs and RBCs were identified as red, those from PMNs as green and the ones from Nalm-6 cells as blue.

Statistical analysis was performed to determine the diagnostic potential of these backscattering microfluidic-based measurements. In all our cell samples, fluorescence was used as a gold standard for assessing the accuracy of the classification achieved based on the light scattering properties of the cells at 405, 488, and 633 nm. Specifically, light scattering peaks correlated with a fluorescence peak were assumed to be emanating from cells that were truly Nalm-6 cells, while light scattering peaks that did not have a correlated fluorescence peak were assumed to result from normal white or red blood cells. We should note that, unfortunately, this is not an ideal gold standard, since on the one hand it is possible that some Nalm-6 cells were not adequately stained to yield a fluorescence peak, while, on the other hand, calcein could also potentially leak from the Nalm-6 cells and stain white or red blood cells.

Fisher classification equations from discriminant analysis, using IBM SPSS software, were used to classify the cells based on differences in their light scattering characteristics. The classification equations were determined from discriminant analysis on a training data set, which consisted of the first data set of measurements from 0.01 to 20% mixed samples, including 49,678 PMNs, 55,493 PBMCs, and 3,935 Nalm-6 cells. Two sets of Fisher classification equations were derived to discriminate leukemic cells from either PBMCs or PBMCs and PMNs. The derived Fisher classification equations were then applied prospectively to the validation data set, which consisted of the remaining data from mixed samples, which included 261,083 PMNs, 307,944 PBMCs, and 24,083 Nalm-6. Additionally, we calculated the minimum number of leukemic cells that would need to be detected (Ncancer) to conclusively diagnose the presence of cancer in the sample at a certain level of confidence, using G*Power software. Finally, from Ncancer, we estimated the volume of blood that we would need to sample assuming acertain concentration of cancer cells within a given blood volume.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED

Light Scattering Evaluation with Pure Cell Samples

Figures 2a–2c show the density plots of calibrated light scattering intensities from isolated normal blood cell samples. Density plots are useful in showing the presence of more than one type of cell in the same sample, as in the case of isolated PMN samples, clearly showing two sub-populations of cells with different scattering intensities (Fig. 2c). Typically, over 80% of detected cells in the isolated PMN samples have an associated high backscattering intensity, which we expect to be from true PMNs. The remaining detected cells with low scattering intensities are likely contaminating PBMCs and/or RBCs, based on the similarly low scattering intensities observed from pure PBMC and RBC samples (Figs. 2a and 2b)

thumbnail image

Figure 2. Density plots of calibrated light scattering intensities from flowing pure cell samples of: (a) RBCs, (b) PBMCs, and (c) PMNs. Also shown are the corresponding density plots of calibrated scattering intensities from labeled Nalm-6 cells, showing the scattering distributions from (d) the proportion of detected cells with time correlated scattering and fluorescent signals, shown in blue, compared to (e) the proportion of detected cells with correlated scattering peaks only, shown in red. (f) Density plots show the overlap in the scattering distribution, in magenta, between (d) and (e). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

Figures 2d–2f show the density plots of calibrated light scattering from calcein-labeled Nalm-6 samples. Scattering peaks of leukemic cells with and without an associated fluorescence signal are represented by blue and red colors, respectively. A majority of detected peaks, ∼90% are 4λ correlated. Interestingly, 62% of the peaks without fluorescence, i.e., 3λ correlated, have similar backscattering as those from cells that are 4λ correlated. This suggests that ∼6.5% of Nalm-6 cells in the sample were not stained properly. As mentioned previously, this would lead to an artificial decrease of the true positives and a corresponding increase of false negatives in our classification. The remaining 3.5% of detected cells with 3λ correlated peaks but significantly lower scattering intensities may be dead leukemic cells (Figure 2f).

Comparing the density plots of light scattering of 4λ correlated Nalm-6 cells to normal blood cells, we observe that the scattering intensity from Nalm-6 is lower in comparison to the scattering from PMNs but higher compared to those from MNs and RBCs. Additionally, we note that the light scattering distribution of PMNs is much broader compared to the other cell types.

Light Scattering Evaluation with Mixed Cell Samples

From the light scattering density plots of PMN samples, we can determine a 3λ correlated peak signal threshold that best separates true PMNs from contaminating PBMCs and/or RBCs. Specifically, peaks that are < 0.95 × 10−3 at 405 nm and < 0.6 × 10−3 at 633 nm can separate the PMN signals from contaminating PBMCs/RBCs. In addition, this threshold can separate pure PBMCs and RBCs from true PMNs. Therefore, in our analysis of scattering from mixed samples, 3λ correlated peaks may be identified as either PBMCs/RBCs or PMNs and represented with two different colors in RGB density plots. It is not necessary to further differentiate PBMCs from RBCs, since the overall goal is to differentiate leukemic cells from normal blood cells.

In Figure 3, we show representative light scattering density plots from mixed cell samples. Cells with 4λ correlated peaks are shown in blue (i.e., Nalm-6). Cells with 3λ correlated peaks below and above the 3λ threshold are shown in red (i.e., PBMCs/RBCs) and green (i.e., PMNs), respectively. The different populations of cells in mixed samples can be easily visualized in these light scattering intensity RGB density plots. PMNs have the highest and broadest intensity distribution, followed by the leukemic Nalm-6 cells and then by the PMBCs/RBCs.

thumbnail image

Figure 3. Density plots of calibrated light scattering intensities from flowing mixed cell samples with (a) 0.01%, (b) 1%, and (c) 5% Nalm-6 cells relative to PBMCs and PMNs, with Nalm-6 in blue, PBMC in red, and PMN in green. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

From our mixed cell samples, we can estimate the number of expected Nalm-6 cells to be detected based on the overall cell concentration of the sample, the proportion of Nalm-6 cells and the flow rate. On average, the system detects ∼94% of the expected number of Nalm-6 cells. In addition, we investigated the possible effects of RBC contamination in our samples by comparing the backscattering from mixed samples with 1% Nalm-6 and 99% WBC and from those with 1% Nalm-6, 69% WBC, and 30% RBCs. As shown in Figures 4a and 4b, the light scattering distributions for 405 and 488 nm appear to be similar. This same observation is noted for the other combinations of wavelength parameters and, therefore, is not shown. However, corresponding zoomed-in density plots (Figs. 4c–4f) show a V-shape distribution of scattering from mixed samples containing RBCs not apparent in the mixed samples without RBCs when the detected scattering at 633 nm is one of the axis parameters (Figs. 4c–4f). The zoomed-in density plots are similar when comparing the backscattering between 405 and 488 nm (Figs. 4g and 4h). This suggests that the V-shape distribution of scattering intensity can be attributed to a larger spectral difference between RBCs and the other cell types. Using the backscattering characteristics from pure RBC samples as a guide, we can deduce the portion of the light scattering distribution that likely corresponds to light scattering from RBCs, which is circled in Figures 4d and 4f.

thumbnail image

Figure 4. Density plots of calibrated light scattering intensities from flowing mixed cell samples with 1% Nalm-6 (a) with WBC and (b) 1% Nalm-6 with WBC and 30% RBC. Scattering from Nalm-6 are shown in blue, PBMC and RBC in red, and PMN in green. Close view of scattering distribution for PBMC and RBC (cd) 405 compared to 488 nm scattering, (ef) 405 compared to 633 nm scattering, and (gh) 488 compared to 633 nm scattering. Circled region indicate the likely scattering signal from RBC. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

Discriminant Analysis and Statistical Evaluation of Mixed Cell Samples with PBMC and Leukemic Cells

Two sets of Fisher classifications are derived considering samples consisting of: a) PBMCs and leukemic cells, relevant for microfluidic analysis applications following initial blood sample processing via a density gradient isolation/enrichment approach, and b) PBMC, PMN, and leukemic cells, relevant for microfluidic analysis applications following red blood cell lysis, which can occur within the same microfluidic device (30, 31).

Statistical results from the training and validation data sets are summarized and compared in Table 1. Similar performance levels of the algorithms for the training and prospective validation set indicate that the algorithms developed are fairly robust. As expected, the sensitivity and specificity of our classification of leukemic cells decreases when both PMNs and PBMCs are included in the flowing samples. Specifically, while the specificity and sensitivity achieved in the validation set are on average 99.6% ± 0.1% and 45.2% ± 2.1%, respectively, for samples containing PBMCs and leukemic cells, they decrease to 91.6% ± 1.5% and 39.5% ± 2.58%, when PMNs are also present. Given that no labeling or expensive reagents are needed for this type of diagnostic, these results are quite promising.

Table 1. Sensitivity and specificity from the training and validation data sets from different leukemic cell proportions relative to WBCs
 %Leukemia cells to WBCSample: PBMC + ALLSample: PBMC + PMN + ALL
Sensitivity (%)Specificity (%)Sensitivity (%)Specificity (%)
  1. Calculated values are for samples with PBMC and leukemia cells and for samples with PBMC, PMN and leukemic cells.

Training data0.01–>2069.6096.4058.182.40
Validation data0.0150.0 ± 0.0199.99 ± 0.0137.5.0 ± 12.597.85 ± 1.31
0.146.43 ± 0.2299.55 ± 0.2248.28 ± 7.8879.95 ± 4.72
153.11 ± 0.3899.55 ± 0.3835.44 ± 8.7996.57 ± 1.01
237.75 ± 0.0699.99 ± 0.0637.11 ± 6.1792.18 ± 2.09
540.60 ± 0.3699.49 ± 0.3638.89 ± 4.0691.51 ± 2.27
1038.99 ± 0.5399.59 ± 0.5343.11 ± 7.7496.24 ± 1.53
>2039.73 ± 0.0399.27 ± 0.0333.92 ± 1.6393.48 ± 1.17

On the basis of the prospective performance of these diagnostic algorithms, we estimated the minimum number of cells that would need to be identified by our test as leukemic cells, Ncancer, in order for us to ascertain the presence of leukemia with a power level of 0.80. The significance level used for these calculations was based on the prospective performance of our diagnostic algorithm. From this number, we can estimate the amount of blood that we would need to sample from a patient, assuming we didn't lose any cells during sample processing. Results, shown in Table 2, are for varying leukemic cell concentrations, assuming an average detection sensitivity to leukemic cells of 45%. The ranges of blood volumes correspond to expected ranges of PBMC concentrations in blood of 0.1–0.4 × 107 cells mL−1 and of PBMC+PMN concentrations varying between 0.4 and 1.0 × 107 cells mL−1 (32, 33). For samples that undergo an initial density gradient isolation and consist of mononuclear cells only (i.e., PBMC and leukemic cells), the initial volume of blood that we would need to draw and process to confirm the presence of leukemia would be less than 42 μL, even if the concentration of leukemic cells is 0.01% relative to PBMCs. For blood samples undergoing only simple RBC lysis, and, thus, also containing PMNs, we would require to draw and process less than 71 μL of blood, at leukemic cell proportions of less than 0.01% relative to WBCs. These volumes decrease if our sensitivity target drops from 0.01 to 5% to a few μL of blood.

Table 2. Volume of whole blood required to ascertain the presence of leukemia in samples with varying concentrations of Nalm-6
Mix (% Nalm)PBMCs + leukemic cellsPBMCs + PMNs + leukemic cells
αaNcancerBlood volume to sample (μL)αaNcancerBlood volume to sample (μL)
  • Results assume either an initial density gradient isolation, yielding samples with PBMCs and leukemic cells, or an RBC lysis step, yielding samples with leukemic, PBMCs and PMNs

  • a

    α = statistical significance (or false positive rate).

0.010.00125316.5–41.680.02151128.21–70.51
0.10.00404147.78–19.450.1759143.59–8.97
10.00448150.83–2.080.0453140.36–0.90
50.00944260.28–0.720.0719180.09–0.23

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED

Flow cytometry and PCR have improved the assessment of MRD of ALL compared to morphological analysis. Nonetheless, these techniques are limited by the large volume of blood that needs to be sampled and extensive sample processing that not only increases cost, but also can result in the loss of already rare cells. Microfluidics is a fast growing technology that can potentially overcome some of these limitations. The reduction in scale leads to significant reductions in the volume of needed reagents and sample and often allows for a degree of automation. In combining microfluidics with backscattering, sample processing can be minimized and reliability maybe improved. This unique combination of capabilities makes for an appealing candidate for the development of a portable, simple, and cost-effective alternative for MRD assessment in leukemia patients, particularly appealing in developing countries where material resources and qualified personnel are limited and cure rates are lower (34). In resource-rich countries, it can additionally result in more continuous monitoring of patients, allowing for more effective therapeutic interventions. Further benefits may also be realized in the simultaneous monitoring of WBCs and leukemia cells. For example, absolute lymphocyte counts during chemotherapy have recently been shown to be an independent predictor of ALL patient relapse and survival (34). But most importantly, a diagnostic or monitoring test for leukemia that would be based on the use of a small volume of blood would be beneficial to pediatric patients throughout the world.

In this study, we demonstrate that, using the backscattered light at 405, 488, and 633 nm, we can detect characteristic light scattering distributions from flowing RBCs, PBMCs, PMNs, and leukemic cells. These results are in agreement with previous light scattering studies performed with similar, but static cell samples (Ref. 26; Greiner et al., manuscript submitted). Specifically, the scattering intensity from PMNs is highest in comparison to Nalm-6 and PBMCs. RBC scattering is the lowest at 405 nm, but higher than that of PBMCs and leukemic cells at 633 nm. Thus, measurements performed at 405 nm and 633 nm are particularly sensitive to hemoglobin absorption changes, allowing for discrimination of RBCs from the other cell types. This was also noted by Ost et al. in their flow cytometric measurements of forward and perpendicular scattering of diluted blood samples with a 413 nm laser source (35). A steeper decrease in the wavelength dependence of the backscattered light from Nalm-6 cells relative to that from the normal PBMCs and PMNs, as well as differences in the scattering intensities, forms the basis for separating leukemic from non-leukemic WBCs.

To assess the diagnostic potential of our approach, we evaluated prospectively blood samples with mixtures of RBCs, WBCs, and varying concentrations of leukemic cells in the range of 0.01–35% (leukemic:WBC). We used calcein fluorescence labeling of the leukemic cells, to assess the validity of the classification results acquired based on using the backscattered light alone. Even though this is not a perfect gold standard, it provides a good measure of performance.

Our results indicate that the proposed light scattering based diagnostic would perform better for the analysis of blood samples that undergo an initial density gradient isolation step to collect mostly the mononuclear cells of the sample. This is still a much simpler sample preparation requirement in comparison to multi-labeling for flow cytometry or to RNA/DNA extraction and multiple gene amplification for PCR. Even though a portion of cells could be lost during this step, with one study as an example indicating ∼67% cell recovery rate, we could ascertain the presence of leukemia at a power level of 80% using less then 63 μL of whole blood (for 0.01% leukemia cells relative to PBMCs) (36). This is a smaller blood volume in comparison to the volume needed for other MRD assessment techniques, which can range from 2 to 5 mL of bone marrow aspirates or 10 mL of peripheral blood samples with leukemic cell proportions of 0.01% or more relative to normal mononuclear cells (37).

When only a simple RBC lysis processing step is expected prior to light scattering assessment, which would result in the presence of both PBMCs and PMNs along with leukemic cells, the blood volume required to have a positive test for the presence of leukemic cells with a 0.80 power level increases to 71 μL. Even if 10% of leukemic cells are lost during this initial processing, we would still be able to perform the test using very small volumes of blood that could be isolated with a simple finger prick (38). Since in this case, both the RBC lysis and the light scattering based measurements could be performed using microfluidics, the potential for developing a truly minimally invasive, cheap and highly portable lab-on-a-chip type device would be highly feasible (38).

Additionally, we should note that the classification algorithm developed in this study is simple and can be implemented for real-time data analysis. A decrease in the sensitivity in our validation data compared to our training data suggests that further improvements in classification performance may be possible by considering larger data sets or other, slightly more complex algorithms, such as discriminant analysis based on quadratic instead of linear classification optimization or discriminant analysis with kernels (39). Nonetheless, the application of a standard classification algorithm presents improvements over the sophisticated data analysis software and personnel training needed to correlate the various parameters in multidimensional space for proper cell classification, when performing multiparameter flow cytometry (9, 40). Finally, automated data analysis in combination with the capability for parallel processing in a microfluidic device offers the potential for processing times that are significantly lower than the 24-h time frame typically offered by standard PCR and flow cytometry analysis (9, 41).

To the best of our knowledge, our backscattering microfluidic-based platform is the first to be applied for MRD assessment. Nonetheless, we expect to easily apply our microfluidic chip to other forms of cancer, since spectroscopy studies have noted characteristic backscattering from other types cancer cells (23–25). As the number of studies increasingly supports the clinical value of enumerating circulating tumor cells in blood, researchers are developing microfluidic-based platforms to capture circulating epithelial cancer cells (42, 43). One example is a microfluidic chip developed to capture epithelial tumor cells, without blood pre-processing, using antibody coated microposts (42). This device can process a large volume of blood (1–2.5 mL h−1) compared to most microfluidic chips, but the capture rate varies from 20 to 60% depending on the flow rate (42). Moreover, the average purity of the captured cells is ∼50% and therefore, it requires further cell characterization procedures via cell staining and fluorescence imaging (42). In comparison, our microfluidic platform technique is simpler and may be easier to apply to different types of cancer.

In conclusion, we have demonstrated and assessed the potential of back scattering in combination with a microfluidic chip platform for MRD assessment of ALL. We demonstrate that this approach could lead to a simple, cost-effective, minimally invasive, portable alternative that allows for automated and real-time diagnosis compared to conventional MRD assessment techniques.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED

The authors acknowledge Dr. Robert White for providing with the 30 μm channel patterned silicon wafer.

LITERATURE CITED

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED
  • 1
    Cripe LD. Adult acute leukemia. Curr Probl Cancer 1997; 21: 164.
  • 2
    Leukemia Classifications [Online]. 2010 [cited 2010 May 13]; Available at: URL: http://cancer.org.
  • 3
    Jemal A,Siegel R,Ward E,Murray T,Xu J,Thun MJ. Cancer statistics, 2007. CA Cancer J Clin 2007; 57: 4366.
  • 4
    Ries LAG,Smith MA,Gurney JG,Linet M,Tamra T,Young JL,Bunin GR. Cancer incidence and survival among children and adolescents: United States SEER Program 1975–1999; Pub. no. 99–4646. Bethesda, MD: National Cancer Institute, 1999;. p. 189.
  • 5
    Pui CH,Evans WE. Acute lymphoblastic leukemia. N Engl J Med 1998; 339: 605615.
  • 6
    Campana D,Coustan-Smith E. Detection of minimal residual disease in acute leukemia by flow cytometry. Cytometry 1999; 38: 139152.
  • 7
    Felix CA,Lange BJ,Chessells JM. Pediatric acute lymphoblastic leukemia: Challenges and controversies in 2000. Hematol Am Soc Hematol Educ Program 2000; 1: 285302.
  • 8
    Pui CH,Campana D,Evans WE. Childhood acute lymphoblastic leukaemia—Current status and future perspectives. Lancet Oncol 2001; 2: 597607.
  • 9
    Campana D. Status of minimal residual disease testing in childhood haematological malignancies. Br J Haematol 2008; 143: 481489.
  • 10
    Flohr T,Schrauder A,Cazzaniga G,Panzer-Grumayer R,van der Velden V,Fischer S,Stanulla M,Basso G,Niggli FK,Schafer BW,Sutton R,Koehler R,Zimmermann M,Valsecchi MG,Gadner H,Masera G,Schrappe M,van Dongen JJ,Biondi A,Bartram CR,International BFM Study Group (I-BFM-SG). Minimal residual disease-directed risk stratification using real-time quantitative PCR analysis of immunoglobulin and T-cell receptor gene rearrangements in the international multicenter trial AIEOP-BFM ALL 2000 for childhood acute lymphoblastic leukemia. Leukemia 2008; 22: 771782.
  • 11
    Coustan-Smith E,Sancho J,Behm FG,Hancock ML,Razzouk BI,Ribeiro RC,Rivera GK,Rubnitz JE,Sandlund JT,Pui CH,Campana D. Prognostic importance of measuring early clearance of leukemic cells by flow cytometry in childhood acute lymphoblastic leukemia. Blood 2002; 100: 5258.
  • 12
    Campana D,Pui CH. Detection of minimal residual disease in acute leukemia: Methodologic advances and clinical significance. Blood 1995; 85: 14161434.
  • 13
    Henze G,Fengler R,Hartmann R,Kornhuber B,Janka-Schaub G,Niethammer D,Riehm H. Six-year experience with a comprehensive approach to the treatment of recurrent childhood acute lymphoblastic leukemia (ALL-REZ BFM 85). A relapse study of the BFM group. Blood 1991; 78: 11661172.
  • 14
    Robillard N,Cave H,Mechinaud F,Guidal C,Garnache-Ottou F,Rohrlich PS,Avet-Loiseau H,Garand R. Four-color flow cytometry bypasses limitations of IG/TCR polymerase chain reaction for minimal residual disease detection in certain subsets of children with acute lymphoblastic leukemia. Haematologica 2005; 90: 15161523.
  • 15
    Cave H,van der Werff ten Bosch J,Suciu S,Guidal C,Waterkeyn C,Otten J,Bakkus M,Thielemans K,Grandchamp B,Vilmer E. Clinical significance of minimal residual disease in childhood acute lymphoblastic leukemia. European Organization for Research and Treatment of Cancer—Childhood Leukemia Cooperative Group. N Engl J Med 1998; 339: 591598.
  • 16
    Coustan-Smith E,Sancho J,Hancock ML,Boyett JM,Behm FG,Raimondi SC,Sandlund JT,Rivera GK,Rubnitz JE,Ribeiro RC,Pui CH,Campana D. Clinical importance of minimal residual disease in childhood acute lymphoblastic leukemia. Blood 2000; 96: 26912696.
  • 17
    Neale GA,Coustan-Smith E,Pan Q,Chen X,Gruhn B,Stow P,Behm FG,Pui CH,Campana D. Tandem application of flow cytometry and polymerase chain reaction for comprehensive detection of minimal residual disease in childhood acute lymphoblastic leukemia. Leukemia 1999; 13: 12211226.
  • 18
    Coustan-Smith E,Sancho J,Hancock ML,Razzouk BI,Ribeiro RC,Rivera GK,Rubnitz JE,Sandlund JT,Pui CH,Campana D. Use of peripheral blood instead of bone marrow to monitor residual disease in children with acute lymphoblastic leukemia. Blood 2002; 100: 23992402.
  • 19
    van der Velden VH,Hochhaus A,Cazzaniga G,Szczepanski T,Gabert J,van Dongen JJ. Detection of minimal residual disease in hematologic malignancies by real-time quantitative PCR: Principles, approaches, and laboratory aspects. Leukemia 2003; 17: 10131034.
  • 20
    Beishuizen A,Verhoeven MA,van Wering ER,Hahlen K,Hooijkaas H,van Dongen JJ. Analysis of Ig and T-cell receptor genes in 40 childhood acute lymphoblastic leukemias at diagnosis and subsequent relapse: Implications for the detection of minimal residual disease by polymerase chain reaction analysis. Blood 1994; 83: 22382247.
  • 21
    Steward CG,Goulden NJ,Katz F,Baines D,Martin PG,Langlands K,Potter MN,Chessells JM,Oakhill A. A polymerase chain reaction study of the stability of Ig heavy-chain and T-cell receptor delta gene rearrangements between presentation and relapse of childhood B-lineage acute lymphoblastic leukemia. Blood 1994; 83: 13551362.
  • 22
    Brisco MJ,Sykes PJ,Hughes E,Dolman G,Neoh SH,Peng LM,Toogood I,Morley AA. Monitoring minimal residual disease in peripheral blood in B-lineage acute lymphoblastic leukaemia. Br J Haematol 1997; 99: 314319.
  • 23
    Wax A,Yang C,Backman V,Badizadegan K,Boone CW,Dasari RR,Feld MS. Cellular organization and substructure measured using angle-resolved low-coherence interferometry. Biophys J 2002; 82: 22562264.
  • 24
    Hunter M,Backman V,Popescu G,Kalashnikov M,Boone CW,Wax A,Gopal V,Badizadegan K,Stoner GD,Feld MS. Tissue self-affinity and polarized light scattering in the born approximation: A new model for precancer detection. Phys Rev Lett 2006; 97: 138102.
  • 25
    Mujat C,Greiner C,Baldwin A,Levitt JM,Tian F,Stucenski LA,Hunter M,Kim YL,Backman V,Feld M,Munger K,Georgakoudi I. Endogenous optical biomarkers of normal and human papillomavirus immortalized epithelial cells. Int J Cancer 2008; 122: 363371.
  • 26
    Hsaio A,Hunter M,Greiner C,Gupta S,Georgakoudi I. Non-invasive identification of subcellular organization and nuclear morphology features associated with leukemic cells using light scattering spectroscopy. J Biomed Opt 2011; 16: 037007.
  • 27
    Peters CE,Woodside SM,Eaves AC. Isolation of subsets of immune cells. Methods Mol Biol 2005; 302: 95116.
  • 28
    Novak J,Georgakoudi I,Wei X,Prossin A,Lin CP. In vivo flow cytometer for real-time detection and quantification of circulating cells. Opt Lett 2004; 29: 7779.
  • 29
    Boutrus S,Greiner C,Hwu D,Chan M,Kuperwasser C,Lin CP,Georgakoudi I. Portable two-color in vivo flow cytometer for real-time detection of fluorescently-labeled circulating cells. J Biomed Opt 2007; 12: 020507.
  • 30
    Davis JA,Inglis DW,Morton KJ,Lawrence DA,Huang LR,Chou SY,Sturm JC,Austin RH. Deterministic hydrodynamics: Taking blood apart. Proc Natl Acad Sci USA 2006; 103: 1477914784.
  • 31
    Sethu P,Moldawer LL,Mindrinos MN,Scumpia PO,Tannahill CL,Wilhelmy J,Efron PA,Brownstein BH,Tompkins RG,Toner M. Microfluidic isolation of leukocytes from whole blood for phenotype and gene expression analysis. Anal Chem 2006; 78: 54535461.
  • 32
    Young B,Heath JW. Wheater's Functional Histology: A Text and Color at Last. London: Elsevier Science Limited; 2000.
  • 33
    Sethu P,Sin A,Toner M. Microfluidic diffusive filter for apheresis (leukapheresis). Lab Chip 2006; 6: 8389.
  • 34
    De Angulo G,Yuen C,Palla SL,Anderson PM,Zweidler-McKay PA. Absolute lymphocyte count is a novel prognostic indicator in ALL and AML: Implications for risk stratification and future studies. Cancer 2008; 112: 407415.
  • 35
    Ost V,Neukammer J,Rinneberg H. Flow cytometric differentiation of erythrocytes and leukocytes in dilute whole blood by light scattering. Cytometry 1998; 32: 191197.
  • 36
    Hamburger AW,Dunn FE,White CP. Percoll density gradient separation of cells from human malignant effusions. Br J Cancer 1985; 51: 253258.
  • 37
    Scheuring UJ,Pfeifer H,Wassmann B,Bruck P,Atta J,Petershofen EK,Gehrke B,Gschaidmeier H,Hoelzer D,Ottmann OG. Early minimal residual disease (MRD) analysis during treatment of Philadelphia chromosome/Bcr-Abl-positive acute lymphoblastic leukemia with the Abl-tyrosine kinase inhibitor imatinib (STI571). Blood 2003; 101: 8590.
  • 38
    Sethu P,Anahtar M,Moldawer LL,Tompkins RG,Toner M. Continuous flow microfluidic device for rapid erythrocyte lysis. Anal Chem 2004; 76: 62476253.
  • 39
    Mika S,Ratsch G,Weston J,Scholkopf B,Mullers KR. Fisher discriminant analysis with kernels. Neural Networks for Signal Processing IX, 1999 Proceedings of the 1999 IEEE Signal Processing Society Workshop 1999: 4148.
  • 40
    Stelzer GT,Shults KE,Loken MR. CD45 gating for routine flow cytometric analysis of human bone marrow specimens. Ann N Y Acad Sci 1993; 677: 265280.
  • 41
    Bjorklund E,Mazur J,Soderhall S,Porwit-MacDonald A. Flow cytometric follow-up of minimal residual disease in bone marrow gives prognostic information in children with acute lymphoblastic leukemia. Leukemia 2003; 17: 138148.
  • 42
    Nagrath S,Sequist LV,Maheswaran S,Bell DW,Irimia D,Ulkus L,Smith MR,Kwak EL,Digumarthy S,Muzikansky A,Ryan P,Balis UJ,Tompkins RG,Haber DA,Toner M. Isolation of rare circulating tumour cells in cancer patients by microchip technology. Nature 2007; 450: 12351239.
  • 43
    Mohamed H,Murray M,Turner JN,Caggana M. Isolation of tumor cells using size and deformation. J Chromatogr A 2009; 1216: 82898295.