Name: label-free identification of lymphocyte subtypes using three-dimensional quantitative phase imaging and machine learning
Uploaded: 2018-11-19T13:00:00.000+00:00
Duration: 8 min 58 s
Description: Watch this Scientific Journal Video about Label-Free Identification of Lymphocyte Subtypes Using Three-Dimensional Quantitative Phase Imaging and Machine Learning at JoVE.com

This work was supported by the KAIST BK21+ Program, Tomocube, Inc., and the National Research Foundation of Korea (2015R1A3A2066550, 2017M3C1A3013923, 2018K000396). Y. Jo acknowledges support from the KAIST Presidential Fellowship and Asan Foundation Biomedical Science Scholarship.

Animal care and experimental procedures were performed under the approval of the Institutional Animal Care and Use Committee of KAIST (KA2010-21, KA2014-01, and KA2015-03). All the experiments in this study were carried out in accordance with the approved guidelines.
1. Lymphocyte Isolation from Mouse Blood
<ol>
	<li>Once a C57BL/6J mouse is euthanized via CO2&#160;inhalation, insert a 26-G needle into the mouse heart and collect 0.3 mL of blood. Directly put blood into a tube with 100 U/mL heparin solution diluted with 1 mL of phosphate-buffered saline (PBS). 
	NOTE: Alternatively, lymphocytes from the spleen can be isolated.</li>
	<li>Centrifuge the tube at 400 x g for 5 min at 4 &#176;C.</li>
	<li>Add 0.5 mL of ammonium-chloride-potassium lysing buffer to the tube and gently invert it a few times to mix the solution.</li>
	<li>Incubate the tube at room temperature (RT) for 5 min.</li>
	<li>Wash the cells by adding 4.5 mL of PBS and centrifuging at 400 x g for 5 min at 4 &#176;C, twice.</li>
	<li>Remove the supernatant and resuspend the cell pellet in 100 &#181;L of fresh RPMI-1640 medium with 10% fetal bovine serum (FBS).</li>
	<li>Add 0.1 &#181;g of CD16/32 (2.4G2) antibody to the tube for blocking.</li>
	<li>Keep the tube on ice.</li>
</ol>
2. Flow Cytometry and Sorting of Lymphocyte Subtypes
NOTE: Sorting lymphocytes depending on cell type is essential for establishing the ground-truth (i.e., correct) cell type labels to train and test a cell type classifier in supervised learning. Flow cytometry, a gold standard method, is used to identify and separate lymphocytes42.
<ol>
	<li>Make a mixture of surface marker staining antibodies in 100 &#181;L of fresh RPMI-1640 medium [with 10% FBS, 0.1 &#181;g of CD3e (17A2), CD8a (53-6.7), CD19 (1D3), CD45R (B220, RA3-6B2), and NK1.1 (PK136)] and 0.25 &#181;g of CD4 (GK1.5) antibodies to target B, CD4+ T, and CD8+ T lymphocytes.</li>
	<li>Add 100 &#181;L of the antibody mixture to the cell suspension (obtained in step 1.8).</li>
	<li>Incubate for 25 min on ice.</li>
	<li>Wash the cells by adding 5 mL of PBS and centrifuging at 400 x g for 5 min at 4 &#176;C, twice.</li>
	<li>Resuspend the cell pellet in 5 mL of fresh RPMI-1640 medium with 10% FBS and 2.5 &#181;g of DAPI (4&#39;,6-diamidino-2-phenylindole).</li>
	<li>Collect each lymphocyte type separately with flow cytometry using the fluorescence levels of the markers described above. Simultaneously exclude dead cells using the DAPI signals. 
	NOTE: Detailed protocols regarding flow cytometry-based cell sorting have been described previously42.</li>
</ol>
3. 3D Quantitative Phase Imaging
<ol>
	<li>Keep the sorted lymphocytes on ice throughout the imaging procedures, which should be completed within 5 h (since lymphocyte isolation from the mouse) to avoid cell damage and biochemical alterations.</li>
	<li>Select a sorted cell type (among B, CD4+ T, and CD8+ T lymphocytes) and dilute the sample (obtained in step 2.6) to 180 cells/&#181;L&#160;with RPMI medium for optimal imaging condition (i.e., one cell per single field-of-view).</li>
	<li>Load 120 &#956;L of the diluted sample into an imaging chamber by slow injection. Thoroughly check on the presence of bubbles in the imaging chamber with the sample. If there are bubbles, carefully remove them, as they will compromise the quality of the measurements.</li>
	<li>Acquire 3D RI tomograms using a commercial 3D quantitative phase microscope, or holotomography, and its imaging software. 
	NOTE: Detailed information about the experimental setup can be found in the original manuscript5.
	<ol>
		<li>Place a drop of distilled water on top of the objective lens of the microscope.</li>
		<li>Place the imaging chamber with the sample on the translation stage of the microscope and adjust its location so that the sample aligns with the objective lens.</li>
		<li>Adjust the axial positions of the objective and condenser lenses by clicking Focus and Surface, respectively, on the &#34;Calibration&#34; tab of the &#34;Microscope&#34; perspective of the imaging software.</li>
		<li>Align the objective and condenser lenses by clicking Auto Mode. Alternatively, use Scanning Mode and manually adjust the lenses so that the illumination patterns are localized at the central region of the field-of-view.</li>
		<li>Return to Normal Mode and adjust the translation stage to locate a cell in the field-of-view.</li>
		<li>Find the focal plane by adjusting the axial position of the objective lens. Perfect focusing makes the sample boundary visualized in the screen almost invisible.</li>
		<li>Adjust the translation stage to find a location without a cell.</li>
		<li>Click Calibrate to measure multiple 2D holograms with varying illumination angles.</li>
		<li>Adjust the translation stage to locate a cell at the center of the field-of-view.</li>
		<li>Move to the &#34;Acquisition&#34; tab and click 3D Snapshot to measure the holograms of the cell with the identical illumination angles as done in step 3.4.8.</li>
		<li>When the acquired data is presented on the &#34;Data Management&#34; panel, right-click on the acquired data and click Process to reconstruct a 3D RI tomogram from the holograms measured in steps 3.4.8 and 3.4.10, using the diffraction tomography algorithm9,10&#160;implemented in the imaging software.</li>
		<li>Repeat steps 3.4.5-3.4.11 to measure more than 100 cells to ensure statistical power for machine learning.</li>
		<li>All images that are processed through the step 3.4.11. can be visualized. On the &#8220;Data Management&#8221; panel, right-click the data and click Open to visualize the data. Click RI Tomogram on the &#8220;Data Manager&#8221; panel. On the &#8220;Preset&#8221; tab, click Load and double click &#8220;lymphocyte.xml&#8221;, which is a predefined transfer function provided by the imaging software to visualize the tomogram according to the 3D RI distributions.&#160;</li>
	</ol>
	</li>
	<li>Repeat steps 3.2-3.4 to measure 3D RI tomograms of all lymphocyte subtypes.</li>
</ol>
4. Quantitative Morphological and Biochemical Feature Extraction from 3D RI Tomograms
<ol>
	<li>Place all the tomographic data measured above in a single folder. Split the cell types in the subfolders of this folder. Prepare each tomogram to be a single .mat file.</li>
	<li>Open Supplementary File 1 - Feature Extraction&#160;(written for an image processing software).</li>
	<li>Edit line 14 to designate the tomogram folder prepared in step 4.1.</li>
	<li>Edit line 15 to designate a folder to save the extracted feature data.</li>
	<li>Optionally, edit line 17 to adjust the RI threshold parameters for feature extraction. The default option is 20 RI thresholds from 1.340-1.378, with an increment of 0.002 as described previously5.</li>
	<li>Execute the code. For every tomogram in the dataset, the code calculates five features: surface area (SA), cellular volume (CV), sphericity (SI), protein density (PD), and dry mass (DM), per RI threshold. The detailed algorithms for feature extraction are described elsewhere5.</li>
	<li>In order to monitor the feature extraction, during the execution check on the screen visualizing RI threshold-based cell segmentation.</li>
	<li>Check on the extracted feature data, as .mat file per tomogram, saved in the folder designated in step 4.4.</li>
</ol>
5. Supervised Learning and Identification
<ol>
	<li>Randomly split the feature data obtained in step 4.8 to training (70%) and test (30%) sets with separate folders.</li>
	<li>Open Supplementary File 2 - Train&#160;(written for an image processing software).</li>
	<li>Edit line 14 to designate the training set prepared in step 5.1.</li>
	<li>Edit line 16 to designate a folder to save the trained classifier.</li>
	<li>Edit line 17 to set a file name for the classifier.</li>
	<li>Optionally, edit line 19 to select the features for training. The default option, as specified previously5, was used to obtain the representative results below.</li>
	<li>Execute the code. Using the selected features of the training set, the code trains a classifier with the k-nearest neighbor algorithm (k-NN; k = 4) and then saves the classifier (as named in step 5.5) in the folder designated in step 5.4.</li>
	<li>Check on the screen visualizing the classifier performance and the cross-validation accuracy.</li>
	<li>Optionally, train multiple classifiers with different feature combinations by repeating steps 5.5-5.7. Then choose the classifier with the highest cross-validation accuracy.</li>
	<li>Open Supplementary File 3 - Test&#160;(written for an image processing software).</li>
	<li>Edit lines 14-15 to designate the trained classifier to be tested.</li>
	<li>Edit line 17 to designate the test set prepared in step 5.1.</li>
	<li>Execute the code. The classifier described above identifies the cell types of the individual lymphocytes in the test set.</li>
	<li>Check on the screen visualizing the identification performance and the test accuracy.</li>
</ol>

Prof. Y. Park, Y. Jo, Y. S. Kim, and S. Lee have financial interests in Tomocube, Inc., a company that commercializes optical diffraction tomography and quantitative phase imaging instruments and is one of the sponsors of the work.

We present a protocol that enables the label-free identification of lymphocyte types exploiting 3D quantitative phase imaging and machine learning. Critical steps of this protocol are quantitative phase imaging and feature selection. For the optimal holographic imaging, the density of cells should be controlled as described above. Mechanical stability of the cells is also important to obtain a precise 3D RI distribution because floating or vibrational cellular motions will disturb hologram measurements upon illumination angle changes. We, therefore waited several minutes until the sample became stable and static in the imaging chamber before measuring holograms. Lastly, bubbles inside the imaging chamber are problematic when measuring holograms due to RI differences between air and the sample; thus, the sample should be carefully loaded to the imaging chamber.
Feature extraction and selection help determine the identification performance of the classifier. We calculated 5 quantitative morphological (CV, SA, SI) and biochemical (PD, DM) features from 3D RI distribution at 20 different RI threshold values; thus, we extracted a total of 100 features. We exhaustively searched optimal feature and classifier combinations, which show that the best cross-validation accuracy was selected. We tested 6 different machine learning algorithms, including k-NN (k = 4 and k = 6), linear discrimination analysis, quadratic discrimination analysis, nai&#776;ve Bayes, and decision tree, and we found that k-NN (k = 4) showed the best identification performance. However, there is a chance to improve identification accuracy using other machine learning methods, including support vector machine and neuronal networks.
This protocol measures intrinsic optical properties via 3D quantitative phase imaging in order to identify lymphocyte types; thus, it does not require a labeling process based on antigen-antibody reactions used in fluorescence or magnetic bead-based cell-sorting techniques, which have risks of altering cellular function by modifying membrane protein structures. Moreover, the present method measures 3D RI distribution and provides 3D morphological and biochemical information about the cell, which cannot be obtained by a single-shot holography method46; therefore, the identification performance of the protocol is more accurate due to high-dimensional information.
A minor limitation of this protocol is the manual adjustment of the sample stage and required labeling process for supervised machine learning. We searched a lymphocyte by adjusting the manual translational stage and measured holograms, which are the most time-consuming steps. This limitation would be improved by employing an automated motorized stage or microfluidic channel devices. Regarding supervised learning, the known lymphocyte types are required to establish the optimal classifier; thus, we had to first isolate and identify lymphocyte cell types based on the antigen-antibody-based sorting technique. Nonetheless, this protocol still uses the intrinsic optical contrast of lymphocytes, and the labeling agents used to specify antibodies have negligible effects on the measured 3D RI signal. Therefore, the established classifier may be used for identifying lymphocytes in a label-free manner.
Although this protocol mainly utilizes phenotypes of lymphocytes by measuring 3D RI tomograms of individual cells, these 3D RI data can also be used in combination with other modalities addressing genotypes or proteomic information for better classification of subtypes. Recently, correlative microscopy techniques combining fluorescence imaging and QPI have been introduced47,48,49. The approach presented in this protocol can also be extended to these correlative imaging methods.
Label-free identification of lymphocyte types can be applied to studying pathophysiology or diagnosing disease by detecting abnormal lymphocytes or ratios among lymphocyte types. Furthermore, this protocol can be applied to whole blood analysis by identifying various cells including red blood cells, platelets, and white blood cells.

Lymphocytes can be classified into various subtypes including B, helper (CD4+) T, cytotoxic (CD8+) T, and regulatory T cells. Each lymphocyte type has a different role in the adaptive immune system; for example, B lymphocytes produce antibodies, whereas T lymphocytes detect specific antigens, eliminate abnormal cells, and regulate B lymphocytes. Lymphocyte function and regulation is tightly controlled by and related to various diseases including cancers1, autoimmune diseases2, and viral infections3. Thus, the identification of lymphocyte types is important to understand their pathophysiological roles in such diseases and for immunotherapy in clinics.
Currently, methods for classifying lymphocyte types rely on antigen-antibody reactions by targeting specific surface membrane proteins or surface markers4. Targeting surface markers is a precise and accurate method to determine lymphocyte types. However, it requires expensive reagents and time-consuming procedures. Furthermore, it carries risks of the modification of membrane protein structures and the alteration of cellular functions.
To overcome these challenges, the protocol described here introduces the label-free identification of lymphocyte types using 3D quantitative phase imaging (QPI) and machine learning5. This method enables the classification of lymphocyte types at a single-cell level based on morphological information extracted from label-free 3D imaging of individual lymphocytes. Unlike conventional fluorescence microscopy techniques, QPI utilizes refractive index (RI) distributions (intrinsic optical properties of live cells and tissues) as optical contrast6,7. The RI tomograms of individual lymphocytes represent phenotypic information specific to subtypes of lymphocytes. In this case, to systemically utilize 3D RI tomograms of individual lymphocytes, a supervised machine learning algorithm was utilized.
Using various QPI techniques, the 3D RI tomograms of cells have been actively used for the study of cell pathophysiology because they provide a label-free, quantitative imaging capability8,9,10,11,12,13. Also, the 3D RI distributions of individual cells can provide morphological, biochemical, and biomechanical information about cells. 3D RI tomograms have been previously utilized in the fields of hematology14,15,16,17, infectious diseases18,19,20, immunology21, cell biology22,23, inflammation24, cancer25, neuroscience26,27, developmental biology28, toxicology29, and microbiology12,30,31,32.
Although 3D RI tomograms provide detailed morphological and biochemical information of cells, the classification of lymphocyte subtypes is difficult to achieve by simply imaging 3D RI tomograms5. To systematically and quantitatively exploit the measured 3D RI tomograms for the cell type classification, we utilized a machine learning algorithm. Recently, several works have been reported in which quantitative phase images of cells were analyzed with various machine learning algorithms33, including the detection of microorganisms34, classification of bacterial genus35,36, rapid and label-free detection of anthrax spores37, automated analysis of sperm cells38, analysis of cancer cells39,40, and detection of macrophage activation41.
This protocol provides detailed steps to perform label-free identification of lymphocyte types at the individual cell level using 3D QPI and machine learning. This includes: 1) lymphocyte isolation from mouse blood, 2) lymphocyte sorting via flow cytometry, 3) 3D QPI, 4) quantitative feature extraction from 3D RI tomograms, and 5) supervised learning for identifying lymphocyte types.

<table><tbody><tr><td>Mouse</td><td>Daehan Biolink</td><td>C57BL/6J mice&amp;nbsp;</td><td>gender and age-matched, 6 &amp;ndash; 8 weeks</td></tr><tr><td>Falcon conical centrifuge tube</td><td>ThermoFisher Scientific</td><td>14-959-53A</td><td>15 mL</td></tr><tr><td>Phosphate-buffered saline&amp;nbsp;</td><td>Sigma-Aldrich</td><td>806544-500ML</td><td></td></tr><tr><td>Ammonium-chloride-potassium lysing buffer&amp;nbsp;</td><td>ThermoFisher Scientific</td><td>A1049201</td><td></td></tr><tr><td>RPMI-1640 medium</td><td>Sigma-Aldrich</td><td>R8758</td><td></td></tr><tr><td>Fetal bovine serum</td><td>ThermoFisher Scientific</td><td>10438018</td><td></td></tr><tr><td>Antibody</td><td>BD Biosciences</td><td>553140 (RRID:AB_394655)</td><td>CD16/32 (clone 2.4G2)</td></tr><tr><td>Antibody</td><td>BD Biosciences</td><td>555275 (RRID:AB_395699)</td><td>CD3&amp;epsilon; (clone 17A2)</td></tr><tr><td>Antibody</td><td>Biolegnd</td><td>100734 (RRID:AB_2075238)</td><td>CD8&amp;alpha; (clone 53-6.7)</td></tr><tr><td>Antibody</td><td>BD Biosciences</td><td>557655 (RRID:AB_396770)</td><td>CD19 (clone 1D3)</td></tr><tr><td>Antibody</td><td>BD Biosciences</td><td>557683 (RRID:AB_396793)</td><td>CD45R/B220 (clone RA3-6B2)</td></tr><tr><td>Antibody</td><td>BD Biosciences</td><td>552878 (RRID:AB_394507)</td><td>NK1.1 (clone PK136)</td></tr><tr><td>Antibody</td><td>eBioscience</td><td>11-0041-85 (RRID:AB_464893)</td><td>CD4 (clone GK1.5)</td></tr><tr><td>DAPI&amp;nbsp;</td><td>Roche</td><td>10236276001</td><td>4,6-diamidino-2-phenylindole</td></tr><tr><td>Flow cytometry&amp;nbsp;</td><td>BD Biosciences</td><td>Aria II or III&amp;nbsp;</td><td></td></tr><tr><td>Imaging chamber</td><td>Tomocube, Inc.</td><td>TomoDish</td><td></td></tr><tr><td>Holotomography</td><td>Tomocube, Inc.</td><td>HT-1H</td><td></td></tr><tr><td>Holotomography imaging software</td><td>Tomocube, Inc.</td><td>TomoStudio</td><td></td></tr><tr><td>Image professing software</td><td>MathWorks</td><td>Matlab R2017b</td><td></td></tr></tbody></table>

label-free identification of lymphocyte subtypes using three-dimensional quantitative phase imaging and machine learning

We describe here a protocol for the label-free identification of lymphocyte subtypes using quantitative phase imaging and machine learning. Identification of lymphocyte subtypes is important for the study of immunology as well as diagnosis and treatment of various diseases. Currently, standard methods for classifying lymphocyte types rely on labeling specific membrane proteins via antigen-antibody reactions. However, these labeling techniques carry the potential risks of altering cellular functions. The protocol described here overcomes these challenges by exploiting intrinsic optical contrasts measured by 3D quantitative phase imaging and a machine learning algorithm. Measurement of 3D refractive index (RI) tomograms of lymphocytes provides quantitative information about 3D morphology and phenotypes of individual cells. The biophysical parameters extracted from the measured 3D RI tomograms are then quantitatively analyzed with a machine learning algorithm, enabling label-free identification of lymphocyte types at a single-cell level. We measure the 3D RI tomograms of B, CD4+ T, and CD8+ T lymphocytes and identified their cell types with over 80% accuracy. In this protocol, we describe the detailed steps for lymphocyte isolation, 3D quantitative phase imaging, and machine learning for identifying lymphocyte types.

Figure 1&#160;shows the schematic process of the entire protocol. Using the procedure presented here, we isolated B (n = 149), CD4+ T (n = 95), and CD8+ T (n = 112) lymphocytes. To obtain phase and amplitude information at various angles of illumination, multiple 2D holograms of each lymphocyte were measured by changing the angle of illumination (from -60&#176; to 60&#176;). Typically, 50 holograms can be used to reconstruct a 3D RI tomogram, but the number of 2D holograms can be adjusted considering the imaging speed and quality. Amplitude and phase information of the measured holograms are retrieved using a field retrieval algorithm based on Fourier transform43,44. The 3D RI tomogram of each lymphocyte was reconstructed from multiple 2D retrieved phase and amplitude information at various angles of illumination using optical diffraction tomography algorithm. Details of image process and 3D RI tomogram reconstruction method can be found elsewhere21,45.
Figure&#160;2A-2C shows representative 3D rendered RI tomograms of B, CD4+ T, and CD8+ T lymphocytes by allocating different color schemes according to RI values via the imaging software. From the RI values, quantitative morphological (SA, CV, and SI) and biochemical (PD and DM) features were calculated (Figure&#160;2A-2C). This result clearly demonstrates that 3D RI distribution enables quantitative analysis of morphological as well as biochemical information of lymphocytes.
Supervised machine learning was exploited to identify lymphocyte types at a single-cell level. The measured 3D RI tomograms were randomly split into 70% and 30% of training (B: 104, CD4+ T: 66, and CD8+ T: 77) and test (B: 45, CD4+ T: 29, and CD8+ T: 35) datasets, respectively. We optimized the classifiers to maximally utilize the cell-type-specific fingerprints encoded in the feature space. The total accuracy, sensitivity (true positive), and specificity (true negative) were calculated by comparing the classifier-predicted results and ground-truth cell types.
In order to demonstrate proof-of-concept of the proposed protocol, we performed supervised machine learning on three different cases: binary classification of (i) B and T lymphocytes and (ii) two T lymphocyte subtypes (CD4+ and CD8+), and (iii) multiclass classification of all lymphocyte types.
Figure 3&#160;shows identification performance of optimized classifiers for training and test stages. The accuracy of the T and B lymphocyte classification was 93.15% and 89.81% for the training and test cases, respectively. The CD4+ and CD8+ T lymphocytes were statistically classified, and the accuracy was 87.41% and 84.38% for the training and test sets, respectively. Lastly, the accuracy of the multiclass cell type classifier was 80.65% and 75.93% for the training and test stages, respectively.
<img alt="Figure 1" class="xfigimg" src="/files/ftp_upload/58305/58305fig1.jpg" /> 
Figure 1: Schematic diagrams of the label-free identification of lymphocyte types exploiting 3D quantitative phase imaging and machine learning. <a href="https://www-jove-com.remotexs.ntu.edu.sg/files/ftp_upload/58305/58305fig1large.jpg" target="_blank">Please click here to view a larger version of this figure.</a>
<img alt="Figure 2" class="xfigimg" src="/files/ftp_upload/58305/58305fig2.jpg" /> 
Figure 2: Representative 3D rendered RI tomograms of each lymphocyte cell type with quantitative morphological and biochemical features. (A) B cell, (B) CD4+ T cell, and (C) CD8+ T cell. Scale bar = 2 &#181;m. SA, surface area; CV, cellular volume; SI, sphericity; PD, protein density; DM, dry mass. This figure is modified with permission5. <a href="https://www-jove-com.remotexs.ntu.edu.sg/files/ftp_upload/58305/58305fig2large.jpg" target="_blank">Please click here to view a larger version of this figure.</a>
<img alt="Figure 3" class="xfigimg" src="/files/ftp_upload/58305/58305fig3.jpg" /> 
Figure 3: Identification of individual lymphocyte types via supervised machine learning (A) binary classification of B and T cells, (B) binary classification of CD4+ and CD8+ T cells, and (C) multiclass classification of all three lymphocyte cell types; for both training and test sets. Note the small difference between the training and test cases, suggesting nice generalization of the established classifiers. The numbers below the names of each cell type indicate the number of cells used. This figure is modified with permission5. <a href="https://www-jove-com.remotexs.ntu.edu.sg/files/ftp_upload/58305/58305fig3large.jpg" target="_blank">Please click here to view a larger version of this figure.</a>
<img alt="Supplementary File 1" src="/files/ftp_upload/58305/58305supfig1.jpg" /> 
Supplementary File 1: Feature extraction code. Extracting features (SA, CV, SI, PD, and DM) after RI threshold-based segmentation of each tomogram. Implemented in an image processing software. <a href="https://www-jove-com.remotexs.ntu.edu.sg/files/ftp_upload/58305/Supplementary_File_1.m" target="_blank">Please click here to download this file.</a>
<img alt="Supplementary File 2" src="/files/ftp_upload/58305/58305supfig2.jpg" /> 
Supplementary File 2: Training code. Training a k-NN classifier training based on selected features. Implemented in an image processing software. <a href="https://www-jove-com.remotexs.ntu.edu.sg/files/ftp_upload/58305/Supplementary_File_2.m" target="_blank">Please click here to download this file.</a>
<img alt="Supplementary File 3" src="/files/ftp_upload/58305/58305supfig3.jpg" /> 
Supplementary File 3: Testing code. Testing a trained k-NN classifier for a new dataset (i.e., test set). Implemented in an image processing software. <a href="https://www-jove-com.remotexs.ntu.edu.sg/files/ftp_upload/58305/Supplementary_File_3.m" target="_blank">Please click here to download this file.</a>

Watch this Scientific Journal Video about Label-Free Identification of Lymphocyte Subtypes Using Three-Dimensional Quantitative Phase Imaging and Machine Learning at JoVE.com

Label-Free Identification of Lymphocyte Subtypes Using Three-Dimensional Quantitative Phase Imaging and Machine Learning

We describe a protocol for the label-free identification of lymphocyte subtypes using quantitative phase imaging and a machine learning algorithm. Measurements of 3D refractive index tomograms of lymphocytes present 3D morphological and biochemical information for individual cells, which is then analyzed with a machine-learning algorithm for identification of cell types.

We describe here a protocol for the label-free identification of lymphocyte subtypes using quantitative phase imaging and machine learning. Identification ...

label-free-identification-lymphocyte-subtypes-using-three-dimensional

Nanyang Technological University

Research

JoVE Journal

Immunology and Infection

12.4K Views.  University of Cambridge. Deze methode biedt een alternatief voor conventionele fluorescentie etikettering en flow cytometrie analyse procedure, die tijdrovend, kostbaar, en het risico van het veranderen van de cellulaire functie van de monsters. Het belangrijkste voordeel van deze techniek is de driedimensionale brekingsindextomografie en machine learning zijn labelvrije en kwantitatieve methoden die een snelle en nauwkeurige lymfocytenidentificatie mogelijk maken. De implicaties van deze techniek breiden zich uit to...

Video: Label-Free Identification of Lymphocyte Subtypes Using Three-Dimensional Quantitative Phase Imaging and Machine Learning

Automated Quantification of Hematopoietic Cell &#8211; Stromal Cell Interactions in Histological Images of Undecalcified Bone

Confocal microscopy is the method of choice for the analysis of localization of multiple cell types within complex tissues such as the bone marrow. However, the analysis and quantification of cellular localization is difficult, as in many cases it relies on manual counting, thus bearing the risk of introducing a rater-dependent bias and reducing interrater reliability. Moreover, it is often difficult to judge whether the co-localization between two cells results from random positioning, especially when cell types differ strongly in the frequency of their occurrence. Here, a method for unbiased quantification of cellular co-localization in the bone marrow is introduced. The protocol describes the sample preparation used to obtain histological sections of whole murine long bones including the bone marrow, as well as the staining protocol and the acquisition of high-resolution images. An analysis workflow spanning from the recognition of hematopoietic and non-hematopoietic cell types in 2-dimensional (2D) bone marrow images to the quantification of the direct contacts between those cells is presented. This also includes a neighborhood analysis, to obtain information about the cellular microenvironment surrounding a certain cell type. In order to evaluate whether co-localization of two cell types is the mere result of random cell positioning or reflects preferential associations between the cells, a simulation tool which is suitable for testing this hypothesis in the case of hematopoietic as well as stromal cells, is used. This approach is not limited to the bone marrow, and can be extended to other tissues to permit reproducible, quantitative analysis of histological data.

A strategy to quantitatively analyze histological data in the bone marrow is presented. Confocal microscopy of fluorescently labeled cells in tissue sections results in 2-dimensional images, which are automatically analyzed. Co-localization analyses of different cell types are compared to data from simulated images, giving quantitative information about cellular interactions.

Confocal microscopy is the method of choice for the analysis of localization of multiple cell types within complex tissues such as the bone marrow. ...

Developmental Biology

A Time-lapse, Label-free, Quantitative Phase Imaging Study of Dormant and Active Human Cancer Cells

The acquisition of the angiogenic phenotype is an essential component of the escape from tumor dormancy. Although several classic in vitro assays (e.g., proliferation, migration, and others) and in vivo models have been developed to investigate and characterize angiogenic and non-angiogenic cell phenotypes, these methods are time and labor intensive, and often require expensive reagents and instruments, as well as significant expertise. In a recent study, we used a novel quantitative phase imaging (QPI) technique to conduct time-lapse and labeling-free characterizations of angiogenic and non-angiogenic human osteosarcoma KHOS cells. A panel of cellular parameters, including cell morphology, proliferation, and motility, were quantitatively measured and analyzed using QPI. This novel and quantitative approach provides the opportunity to continuously and non-invasively study relevant cellular processes, behaviors, and characteristics of cancer cells and other cell types in a simple and integrated manner. This report describes our experimental protocol, including cell preparation, QPI acquisition, and data analysis.

Dormant and active cancer cell phenotypes were characterized using quantitative phase imaging. Cell proliferation, migration, and morphology assays were integrated and analyzed in one simple method.

The acquisition of the angiogenic phenotype is an essential component of the escape from tumor dormancy. Although several classic in vitro assays (e.g., ...

Cancer Research

Morphology-Based Distinction Between Healthy and Pathological Cells Utilizing Fourier Transforms and Self-Organizing Maps

The appearance and the movements of immune cells are driven by their environment. As a reaction to a pathogen invasion, the immune cells are recruited to the site of inflammation and are activated to prevent a further spreading of the invasion. This is also reflected by changes in the behavior and the morphological appearance of the immune cells. In cancerous tissue, similar morphokinetic changes have been observed in the behavior of microglial cells: intra-tumoral microglia have less complex 3-dimensional shapes, having less-branched cellular processes, and move more rapidly than those in healthy tissue. The examination of such morphokinetic properties requires complex 3D microscopy techniques, which can be extremely challenging when executed longitudinally. Therefore, the recording of a static 3D shape of a cell is much simpler, because this does not require intravital measurements and can be performed on excised tissue as well. However, it is essential to possess analysis tools that allow the fast and precise description of the 3D shapes and allows the diagnostic classification of healthy and pathogenic tissue samples based solely on static, shape-related information. Here, we present a toolkit that analyzes the discrete Fourier components of the outline of a set of 2D projections of the 3D cell surfaces via Self-Organizing Maps. The application of artificial intelligence methods allows our framework to learn about various cell shapes as it is applied to more and more tissue samples, whilst the workflow remains simple.

Here, we provide a workflow that allows the identification of healthy and pathological cells based on their 3-dimensional shape. We describe the process of using 2D projection outlines based on the 3D surfaces to train a Self-Organizing Map that will provide objective clustering of the investigated cell populations.

The appearance and the movements of immune cells are driven by their environment. As a reaction to a pathogen invasion, the immune cells are recruited to ...

Quantitative Multispectral Analysis Following Fluorescent Tissue Transplant for Visualization of Cell Origins, Types, and Interactions

With the desire to understand the contributions of multiple cellular elements to the development of a complex tissue; such as the numerous cell types that participate in regenerating tissue, tumor formation, or vasculogenesis, we devised a multi-colored cellular transplant model of tumor development in which cell populations originate from different fluorescently colored reporter gene mice and are transplanted, engrafted or injected in and around a developing tumor. These colored cells are then recruited and incorporated into the tumor stroma. In order to quantitatively assess bone marrow derived tumor stromal cells, we transplanted GFP expressing transgenic whole bone marrow into lethally irradiated RFP expressing mice as approved by IACUC. 0ovarian tumors that were orthotopically injected into the transplanted mice were excised 6-8 weeks post engraftment and analyzed for bone marrow marker of origin (GFP) as well as antibody markers to detect tumor associated stroma using multispectral imaging techniques. We then adapted a methodology we call MIMicc- Multispectral Interrogation of Multiplexed cellular compositions, using multispectral unmixing of fluoroprobes to quantitatively assess which labeled cell came from which starting populations (based on original reporter gene labels), and as our ability to unmix 4, 5, 6 or more spectra per slide increases, we&#39;ve added additional immunohistochemistry associated with cell lineages or differentiation to increase precision. Utilizing software to detect co-localized multiplexed-fluorescent signals, tumor stromal populations can be traced, enumerated and characterized based on marker staining.1

Complex tissue masses, from organs to tumors, are composed of various cellular elements. We elucidated the contribution of cellular phenotypes within a tissue utilizing multi-labeled fluorescent transgenic mice in combination with multiparameter immunofluorescent staining followed by spectral unmixing to decipher cell origin as well as cell characteristics based on protein expression.

With the desire to understand the contributions of multiple cellular elements to the development of a complex tissue; such as the numerous cell types that ...

Medicine

Highly Resolved Intravital Striped-illumination Microscopy of Germinal Centers

Monitoring cellular communication by intravital deep-tissue multi-photon microscopy is the key for understanding the fate of immune cells within thick tissue samples and organs in health and disease. By controlling the scanning pattern in multi-photon microscopy and applying appropriate numerical algorithms, we developed a striped-illumination approach, which enabled us to achieve 3-fold better axial resolution and improved signal-to-noise ratio, i.e. contrast, in more than 100 &#181;m tissue depth within highly scattering tissue of lymphoid organs as compared to standard multi-photon microscopy. The acquisition speed as well as photobleaching and photodamage effects were similar to standard photo-multiplier-based technique, whereas the imaging depth was slightly lower due to the use of field detectors. By using the striped-illumination approach, we are able to observe the dynamics of immune complex deposits on secondary follicular dendritic cells &#8211; on the level of a few protein molecules in germinal centers.

High-resolution intravital imaging with enhanced contrast up to 120 &#181;m depth in lymph nodes of adult mice is achieved by spatially modulating the excitation pattern of a multi-focal two-photon microscope. In 100 &#181;m depth we measured resolutions of 487 nm (lateral) and 551 nm (axial), thus circumventing scattering and diffraction limits.

Monitoring cellular communication by intravital deep-tissue multi-photon microscopy is the key for understanding the fate of immune cells within thick ...

Qualitative and Quantitative Analysis of the Immune Synapse in the Human System Using Imaging Flow Cytometry

The immune synapse is the area of communication between T cells and antigen-presenting cells (APCs). T cells polarize surface receptors and proteins towards the immune synapse to assure a stable binding and signal exchange. Classical confocal, TIRF, or super-resolution microscopy have been used to study the immune synapse. Since these methods require manual image acquisition and time-consuming quantification, the imaging of rare events is challenging. Here, we describe a workflow that enables the morphological analysis of tens of thousands of cells. Immune synapses are induced between primary human T cells in pan-leukocyte preparations and Staphylococcus aureus enterotoxin B (SEB)-loaded Raji cells as APCs. Image acquisition is performed with imaging flow cytometry, also called In-Flow microscopy, which combines features of a flow cytometer and a fluorescence microscope. A complete gating strategy for identifying T cell/APC couples and analyzing the immune synapses is provided. As this workflow allows the analysis of immune synapses in unpurified pan-leukocyte preparations and hence requires only a small volume of blood (i.e., 1 mL), it can be applied to samples from patients. Importantly, several samples can be prepared, measured, and analyzed in parallel.

Here, we describe a complete workflow for the qualitative and quantitative analysis of immune synapses between primary human T cells and antigen-presenting cells. The method is based on imaging flow cytometry, which allows the acquisition and evaluation of several thousand cell images within a relatively short period of time.

The immune synapse is the area of communication between T cells and antigen-presenting cells (APCs). T cells polarize surface receptors and proteins ...

Light-sheet Microscopy for Three-dimensional Visualization of Human Immune Cells

In vivo, activation, proliferation, and function of immune cells all occur in a three-dimensional (3D) environment, for instance in lymph nodes or tissues. Up to date, most in vitro systems rely on two-dimensional (2D) surfaces, such as cell-culture plates or coverslips. To optimally mimic physiological conditions in vitro, we utilize a simple 3D collagen matrix. Collagen is one of the major components of extracellular matrix (ECM) and has been widely used to constitute 3D matrices. For 3D imaging, the recently developed light-sheet microscopy technology (also referred to as single plane illumination microscopy) is featured with high acquisition speed, large penetration depth, low bleaching, and photocytotoxicity. Furthermore, light-sheet microscopy is particularly advantageous for long-term measurement. Here we describe an optimized protocol how to set up and handle human immune cells, e.g. primary human cytotoxic T lymphocytes (CTL) and natural killer (NK) cells in the 3D collagen matrix for usage with the light-sheet microscopy for live cell imaging and fixed samples. The procedure for image acquisition and analysis of cell migration are presented. A particular focus is given to highlight critical steps and factors for sample preparation and data analysis. This protocol can be employed for other types of suspension cells in a 3D collagen matrix and is not limited to immune cells.

Here, we present a protocol to visualize immune cells embedded in a three-dimensional (3D) collagen matrix using light-sheet microscopy. This protocol also elaborates how to track cell migration in 3D. This protocol can be employed for other types of suspension cells in the 3D matrix.

In vivo, activation, proliferation, and function of immune cells all occur in a three-dimensional (3D) environment, for instance in lymph nodes or ...

A Rapid Method for Multispectral Fluorescence Imaging of Frozen Tissue Sections

Multispectral fluorescence imaging on formalin-fixed paraffin-embedded (FFPE) tissues enables the detection of multiple markers in a single tissue sample that can provide information about antigen coexpression and spatial distribution of the markers. However, a lack of suitable antibodies for formalin-fixed tissues may restrict the nature of markers that can be detected. In addition, the staining method is time-consuming. Here we describe a rapid method to perform multispectral fluorescence imaging on frozen tissues. The method includes the fluorophore combinations used, detailed steps for the staining of mouse and human frozen tissues, and the scanning, acquisition, and analysis procedures. For staining analysis, a commercially available semiautomated multispectral fluorescence imaging system is used. Through this method, up to six different markers were stained and detected in a single frozen tissue section. The machine learning analysis software can phenotype cells that can be used for quantitative analysis. The method described here for frozen tissues is useful for the detection of markers that cannot be detected in FFPE tissues or for which antibodies are not available for FFPE tissues.

We describe a rapid staining method to perform multispectral imaging on frozen tissues.

Multispectral fluorescence imaging on formalin-fixed paraffin-embedded (FFPE) tissues enables the detection of multiple markers in a single tissue sample ...

Discrimination and Characterization of Heterocellular Populations Using Quantitative Imaging Techniques

Cellular processes are complex and result from the interplay between multiple cell types and their environment. Existing cell biology techniques often do not allow for accurate interpretation of this interplay. Using a quantitative imaging-based approach, we present a high-content protocol for characterizing the dynamic phenotypic responses (i.e. morphology changes, proliferation, apoptosis) of heterogeneous cell populations to changes in environmental stimuli. We highlight our ability to distinguish between cell types based upon either fluorescence intensity or inherent morphology features depending on the application. This platform allows for a more comprehensive characterization of subpopulation response to perturbation while utilizing shorter time, smaller amounts of reagents, and lower likelihood of error than traditional cell biology assays. However, in some cases, cell populations may be difficult to identify and quantitate based on complex cellular features and will require additional troubleshooting; we highlight some of these circumstances in the protocol. We demonstrate this application using response to drug in a cancer model; however, it can easily be applied more broadly to other physiological processes. This protocol allows one to identify subpopulations within a co-culture system and characterize the particular response of each to external stimuli.

We have developed a novel protocol for studying heterocellular population dynamics in response to perturbations. This manuscript describes an imaging-based platform that produces quantitative datasets for simultaneous characterization of multiple cellular phenotypes of heterocellular populations in a robust manner.

Cellular processes are complex and result from the interplay between multiple cell types and their environment. Existing cell biology techniques often do ...

Biology

3D Visualization of Immune Cell Populations in HIV-Infected Tissues via Clearing, Immunostaining, Confocal, and Light Sheet Fluorescence Microscopy

Human Immunodeficiency Virus (HIV), the causative agent of Acquired Immune Deficiency Syndrome (AIDS), is a major global health concern with nearly 40 million individuals infected worldwide and no widely accessible cure. Despite intensive efforts, a detailed understanding of virus and host cell interactions in tissues during infection and in response to therapy remains incomplete. To address these limitations, water-based tissue clearing techniques CUBIC (Clear, Unobstructed Brain/Body Imaging Cocktails and Computational analysis) and CLARITY (Clear Lipid-exchanged Acrylamide-hybridized Rigid Imaging/Immunostaining/in situ-hybridization-compatible Tissue hYdrogel) are applied to visualize complex virus host-cell interactions in HIV-infected tissues from animal models and humans using confocal and light sheet fluorescence microscopy. Optical sectioning of intact tissues and image analysis allows rapid reconstruction of spatial information contained within whole tissues and quantification of immune cell populations during infection. These methods are applicable to most tissue sources and diverse biological questions, including infectious disease and cancer.

Tissue clearing, combined with immunofluorescence microscopy, allows spatial visualization and quantification of immune-cell populations and virus proteins within intact tissues. Optical sectioning of cleared tissues with confocal and light sheet fluorescence microscopy can generate 3D models of complex tissue environments and reveal spatial heterogeneity exhibited during HIV infection.

Human Immunodeficiency Virus (HIV), the causative agent of Acquired Immune Deficiency Syndrome (AIDS), is a major global health concern with nearly 40 ...

Label-Free Identification of Lymphocyte Subtypes Using Three-Dimensional Quantitative Phase Imaging and Machine Learning

Samenvatting

Meer video's verkennen

Hoofdstukken in deze video