Here, we describe a combined flow cytometric cell sorting and low-input, next-generation library construction protocol designed to produce high-quality, whole-exome data from the Hodgkin Reed-Sternberg (HRS) cells of classical Hodgkin lymphoma (CHL).
The Hodgkin Reed-Sternberg cells of classical Hodgkin lymphoma are sparsely distributed within a background of inflammatory lymphocytes and typically comprise less than 1% of the tumor mass. Material derived from bulk tumor contains tumor content at a concentration insufficient for characterization. Therefore, fluorescence activated cell sorting using eight antibodies, as well as side- and forward-scatter, is described here as a method of rapidly separating and concentrating with high purity thousands of HRS cells from the tumor for subsequent study. At the same time, because standard protocols for exome sequencing typically require 100-1,000 ng of input DNA, which is often too high, even with flow sorting, we also provide an optimized, low-input library construction protocol capable of producing high-quality data from as little as 10 ng of input DNA. This combination is capable of producing next-generation libraries suitable for hybridization capture of whole-exome baits or more focused targeted panels, as desired. Exome sequencing of the HRS cells, when compared against healthy intratumor T or B cells, can identify somatic alterations, including mutations, insertions and deletions, and copy number alterations. These findings elucidate the molecular biology of HRS cells and may reveal avenues for targeted drug treatments.
Advancements in cancer genomics as a result of next-generation sequencing have led to significant breakthroughs in the identification of therapeutic targets and in prognostication for many hematologic and non-hematologic neoplasms. New individualized treatment strategies based on specific genomic alterations are rapidly being introduced in many tumor types (reviewed in references1,2). Despite significant progress in lymphoma genomics, the genome of the neoplastic HRS cells in classical Hodgkin lymphoma (CHL) had been underexplored. The investigations have been hampered by the scarcity of neoplastic HRS cells within a reactive microenvironment, making it difficult to isolate purified HRS cell populations3.
The method for isolating viable HRS cells from primary tumors was developed by Fromm et al.4,5,6. The method uses an eight-antibody cocktail, consisting of CD30, CD15, CD40, CD95, CD45 CD20, CD5, and CD64, to unequivocally identify HRS cells from a CHL tumor suspension. Using this methodology, we are able to isolate at least 1,000 viable HRS cells from fresh or frozen cell suspensions from tumor biopsies consisting of at least 107 cells (approximately 10 mg of tissue). The purity is greater than 90% by flow cytometric analysis and is estimated to be least 80% by exome genomic analysis of ten consecutive cases.
We have refined a flow cytometric cell isolation technique that has greatly eased the process, allowing for the rapid isolation of thousands of viable HRS cells from primary CHL tumors7. We have utilized the technique to produce what is believed to be the first whole-exome sequence of the tumor cells in primary cases of Hodgkin lymphoma. Our studies demonstrate the feasibility of high-throughput, genome-wide studies of individual CHL cases and have already led to the identification of novel genomic alterations with the potential to explain aspects of CHL pathogenesis.
We further developed a pipeline to utilize the extracted DNA for high-throughput genomic studies. In order to achieve reliable results from as few as 1,000 sorted HRS cells (thhe minimum obtained from sequential cases), we further developed a modified next-generation DNA library construction procedure8 that allowed us to increase adaptor ligation efficiency and to generate DNA fragment libraries without excessive amplification. This method allows for the analysis of routine clinical samples and the detection of recurrent mutations and chromosomal alterations7.
1. Tissue Processing and Freezing
2. Preparing Cell Suspensions for Cell Sorting
NOTE: Each lot of antibody must be properly titered using 10 million cells in a 300- µΛ staining volume. Peripheral blood may be used for all antibodies except CD30, for which the KMH2 cell line spiked into peripheral blood may be used for titration10. We generally begin with the manufacturer-recommended volume of antibody and perform two two-fold dilutions and one two-fold increase (four data points) for each lot of antibody titration. For example, if the manufacturer recommends a 10-µL volume, we perform the titration using 2, 5, 10, and 20 µL volumes.
3. (Optional Protocol) T Cell Rosette Blocking
NOTE: HRS cells are rosetted by T cells in tissue sections and cell suspension, and these T cells may potentially contaminate the sorted HRS fraction. These interactions are mediated by CD54 and CD58 on the HRS cell binding to LFA-1 and CD2 on the T cells4,11. These interactions can be blocked with unlabeled antibodies to these adhesion molecules.
4. HRS-, B-, and T-cell Isolation Using Cell Sorting
NOTE: Although we used a special research order instrument using 5 lasers (see the Materials spreadsheet), any sorter with the capability of detecting the fluorochromes used in the antibody panel should be sufficient. The execution of the steps below requires a familiarity with software12 function and a basic knowledge of cell sorter operations. Please refer to the online software manual for detailed instructions.
5. DNA Extraction
6. Library Construction
7. Exome Hybridization
8. Multiplexed Sequencing
9. Analysis (Can be Substituted with Alternative Pipeline(s) if Desired)
A bioanalyzer plot should be taken after library amplification and 0.8x bead cleanup. One should see a "normal-like" distribution of fragment sizes in the desired range (Figure 2a). Deviations from this shape, such as a visible "shoulder" in the curve, indicate the presence of a high or low molecular weight artifact. For example, Figure 2b-2d shows examples of libraries containing visible artifacts that should ideally not be sequenced. If a library is significantly compromised, it may be worthwhile to repeat the library construction if DNA is available and/or the cell sorting to start fresh.
An adapter dimer may sometimes carry over through the first 0.8x bead cleanup. If this occurs, it will be visible as a sharp peak centered around 125 – 130 bp (see Figure 2c). In this case, it is advisable to repeat an additional 0.8x bead cleanup and to repeat the bioanalyzer to ensure the successful removal of the dimer.
Using the specifications described in this protocol and multiplexing the sequencing at the level of four samples per lane in the sequencer listed in the Materials spreadsheet, a median depth of coverage of 50 – 100x across the target (after the bioinformatics removal of PCR duplicate reads) is achievable (see Figure 3a). In addition, libraries that were not overamplified and that resulted in libraries with adequate coverage should produce clean copy number plots. During early work, this library construction protocol was tested over three orders of magnitude of input mass, and variable results were discovered (see the Discussion section). Copy number plots generated using a library generated from 1 ng, a sub-optimal amount of input DNA, resulted in spurious copy number gains and losses (Figure 3b, bottom).
Approximately 70% of cases of CHL carry B2M and A20 mutations and/or deletions7. B2M mutations predominantly involve a translational start site, a first exon stop site, and a first exon stop site. A20 mutations are a mix of copy number losses and indels. The mutations appear to be clonal and can be used to estimate relative tumor content. If T cells are used as somatic controls, HRS sequencing will show significant copy number gains in the positions corresponding to T-cell receptor alpha/delta (14q11 – 12) and beta positions (7q32 – 35), as well as significant losses in the B-cell receptor heavy chain (14q32) and kappa light chain positions (2p11) (see Figure 4). Overall sequencing shows a range from 100 to 400 somatic mutations per case. Copy number variation is highly variable from case to case, with some cases showing numerous segmental gains and losses and other cases showing only a few alterations.
Figure 1: Multiparameter Gating for Flow-sorting of Viable HRS, B, and T cells. Population names are shown on plots and indicate the parent population of any gate. (A)shows all cells: On the FSC-H versus FSC-A histogram, draw a SINGLETS gate that includes populations with proportional FSC-H and FSC-A increases. Exclude doublets that have high FSC-A and lower FSC-H. Hodgkin cells are very large; expand the gate to include high FSC-H and FSC-A events. (B) Exclude debris, dead cells, and unlysed RBCs that normally have lower FSC-A and slight increases in SSC-H without cutting out viable populations (VIABLE gate). (C) Gate MONONUCLEAR cells on FSC-A vs. SSC-H to exclude granulocytes with high SSC-H; note that the gate is only used for the identification of T and B cells. (D) Gate T CELLS positively identified by CD5 (green) without CD20 and B CELLS (blue), which have CD20 expression without CD5. (E) Draw a LARGER gate on the CD45/SSH plot, including 10 – 20% of the lymphoid cells and all granulocytic cells. (F) Gate CD30-positive events with dim to negative CD64/FITC autofluorscence. (G) Among CD30 positive events, gate CD40- and CD95-positive events (CD30/40/95 HRS gate, red). (H-L) Observe the confirmatory immunophenotypic features on HRS cells; HRS cells generally do not express CD20 (plots H and L). The expression of CD5 is correlated with CD45 due to T-cell rosetting (I); most cases will show expression of CD15 and will be bright for CD71 (J). CD40 and CD95 expression is generally above the level of B and T cells, respectively (K). This research was originally published in Blood7. Please click here to view a larger version of this figure.
Figure 2: Size Distribution Plots of Samples after Adapter Ligation and Amplification. Ideally, libraries with visible artifacts should be repeated and not sequenced. (a) Well-formed library; no visible artifacts. (b) Asymmetry in the curve reveals a problem. (c) Significant adapter dimer (peak at 128) and a high molecular weight artifact (>400 bp). The dimer may be cleaned with an additional 0.8x bead cleanup. (d) Significant high molecular weight artifact. Please click here to view a larger version of this figure.
Figure 3: Bioinformatic Quality Control Measurements for Higher-quality Libraries made from 10 ng of Input DNA Compared to Lower-quality Libraries made from 1 ng of Input DNA. The magnitude and distribution of the unique read depth of coverage expectations. (a) A median depth of coverage of at least 50 is expected when using 10-100 ng of input DNA. (b) Each of 2 panels depicts copy number variation analysis results comparing data between 2 sequenced libraries. Exonic probe segments (x-axis) versus copy number change on a log2 scale (y-axis) are plotted for a single representative chromosome (chr 6). Comparing data from a 10 ng low-input library from intratumoral T-cell DNA to a 100 ng normal-input library from intratumoral T-cell DNA from the same case showed no significant false-positive results; that is, low-input and normal-input libraries are copy-neutral (top). Numerous false-positive segmental copy number alterations were called when data from a 1 ng low-input library from intratumoral T-cell DNA was compared to a 100 ng normal-input library from intratumoral T-cell DNA from the same case (bottom). This research was originally published in Blood7. Please click here to view a larger version of this figure.
Figure 4: Recurrent Copy Number Gains and Losses, with Tumor Cells (Primary HRS and Cell Lines) Compared Against Healthy Intratumor T Cells from Primary Cases. If T cells are used as the somatic control against B cell-derived HRS and cell line populations, recurrent gains can be seen in T-cell receptors alpha/delta and beta. Likewise, recurrent losses are detectable in immunoglobulin heavy and kappa chains. This research was originally published in Blood7. Please click here to view a larger version of this figure.
DNA input mass (n) | 1-10 ng | 25 ng | 100 ng |
Adapter : insert molar ratio (r) | 65:1 | 25:1 | 15:1 |
Table 1: DNA Input Mass and the Corresponding Adapter:Insert Molar Ratio. The values in this table can be used as a guide to calculate the amount of adapter oligo to add during the adapter ligation step of library construction.
Future applications or directions after mastering this technique
This work allows for exome sequencing from samples containing at least 10 ng of DNA. In the clinical context, this limit excludes most fine-needle aspiration samples due to insufficient material, but it includes adequate core biopsies and excisional biopsy samples. This will enable the acquisition of data from a larger set of possible samples.
Critical steps within the protocol
Proper freezing and dissociation techniques are critical to the success of the experiment. The entire sample should be dissociated, including fibrotic sections, as much as possible. Utilizing a 130 µm nozzle for cell sorts appears to be critical to maximize the yield of the large HRS cells and DNA. While rigorous controlled experiments with multiple nozzle sizes were not performed, significant increases in cell yields were observed using the larger nozzle compared to the 100-µm, and virtually no cells were detected on the slide post 85-µm nozzle sort (unpublished observation).
Modifications and troubleshooting
When working with less than 100 ng masses of input DNA for library construction, care should be taken to reduce all unnecessary processing and cleanup steps, especially prior to PCR amplification, since every lost unique molecule results in a reduction of the complexity of the final library. The protocol deliberately recommends eluting extracts in 50 µL because this value is compatible with both sonication and end-repair using the equipment recommended. In this way, it is not necessary to reduce the elution volume using a centrifugal evaporator or a column for the sonication step or to reduce the sonication volume for the end-repair step. Additional modifications to this protocol are possible; for example, new library construction kits may improve adapter ligation efficiency. One could troubleshoot/optimize by comparing adapter-mediated library amplification curves using real-time amplification on libraries made with different kits from the same quantity and quality of input DNA.
Limitations of the technique
At least 10 ng of purified HRS DNA (approximately 1,000 sorted cells after losses in the sorting and extraction steps) appears to be the minimal amount for high-quality results. In pilot experiments, DNA from a single sample of purified T cells was first used to make libraries spanning a range of input masses. 100 ng (the standard recommended amount) of input DNA was compared to libraries generated using this low-input protocol for 10 ng and 1 ng of input DNA. The library generated from 10 ng of input DNA had no significant differences from the 100-ng input library in magnitude and distribution of coverage across the target, or in PCR duplication fraction. However, the library generated using 1 ng of input DNA resulted in a significantly higher PCR duplicate fraction and a corresponding decrease (approximately 3-fold) in mean coverage across the target. Moreover, the 1 ng library also exhibited deviation from uniformity in coverage, which created false-positive copy number alterations with respect to the 100 ng library, unlike the 10 ng input library, which was copy-neutral in the same assessment. For this reason, exert caution using less than 10 ng of input DNA using this method.
Significance of the technique with respect to existing/alternative methods
Following the protocol will allow full exome sequencing of HRS cells from primary Hodgkin lymphoma samples. CHL lymph node samples that contain at least 2 x 107 cells are suitable for this procedure. This was not previously available to researchers, who relied on techniques such as laser capture microdissection and whole genome amplification. It is predicted that genomic approaches to primary CHL cases will continue to be valuable for further understanding CHL pathogenesis. Data also points to significant genome-level heterogeneity within this disease entity, with potentially at least two defined subsets that loosely follow morphological stratification between nodular sclerosis and mixed cellularity CHL. Finally, it is believed that genome-level data will lead to the development of targeted therapy for difficult-to-treat relapsed/refractory CHL cases.
The authors have nothing to disclose.
The development of this project method was funded by the Department of Pathology and Laboratory Medicine of Weill Cornell Medical College. We acknowledge the Tri-Institutional Training Program in Computational Biology and Medicine for partial funding. We would like to thank the scientists who shared their time and knowledge with us, especially Maryke Appel; Dan Burgess; Iwanka Kozarewa; Chad Locklear; and everyone from the Weill Cornell Medical College Genomics Core Facility, including Jenny Zhang, Xiaobo (Shawn) Liang, Dong Xu, Wei Zhang, Huimin Shang, Tatiana Batson, and Tuo Zhang.
Petri or Cell Culture Dish (sterile) | |||
RPMI-1640 Media | Roswell Park Memorial Institute | ||
Fetal Calf Serum (FCS), (heat inactivated) | |||
Freezing Media (RPMI, 20% FCS, 10% dimethylsulfoxide (DMSO))-make fresh and keep sterile | |||
RPMI with 2% FCS (make fresh or store for up to 1 month) | |||
scalpel with fresh blade | |||
10 ml syringe (no needle) | |||
Cryogenic vials | |||
50 ml conical centrifuge tubes, force | |||
Centrifuge | capable of handling 50 ml conical centrifuge tubes and providing 400g | ||
Hepes buffer(1M, cell culture grade) | |||
phosphate buffered saline (PBS) | |||
Pluoronic-F68 | Thermo-Fisher | 24040-032 | |
DNAase-I | Sigma-Aldrich, St. Louis, MO | D4527-10KU | store as 5mg/ml in RPMI in -200C |
Bovine Serum Albumin (BSA) | |||
Sort Media (PBS+2%BSA+25mM HEPES+ Pluoronic –F68 (1X)) | |||
CD64-FITC (22) | Beckman Coulter, Miami, FL | 20 uL suggested starting volume; Titering is suggested | |
CD30-PE (BerH83) | BD Biosciences, San Jose, CA | 20 uL suggested starting volume; Titering is suggested | |
CD5-ECD (BL1a) | Beckman Coulter, Miami, FL | 10 uL suggested starting volume; Titering is suggested | |
CD40-PerCP-eFluor 710 (1C10) | Ebiosciences, San Diego, CA | 5 uL suggested starting volume; Titering is suggested | |
CD20-PC7 (B9E9) | Beckman Coulter, Miami, FL | 10 uL suggested starting volume; Titering is suggested | |
CD15-APC (HI98) | BD Biosciences, San Jose, CA | 20 uL suggested starting volume; Titering is suggested | |
CD45 APC-H7 (2D1) | BD Biosciences, San Jose, CA | Can be substituted with 10 uL suggested volume of CD45-Krome Orange (J.33, Beckman Coulter); Titering is suggested | |
CD95-Pacific Blue (DX2) | Life Technologies, Grand Island, NY | 5 uL suggested starting volume; Titering is suggested | |
CD2 (5 μg; clone RPA-2.10) | Biolegend, San Diego, CA | For optional protocol; Titering is suggested | |
CD54 (10 μg; clone 84H10) | Serotec, Oxford, United Kingdom | For optional protocol; Titering is suggested | |
CD58 (10 μg; clone TS2/9) | eBioscience, San Diego, CA | For optional protocol; Titering is suggested | |
LFA-1 (12 μg; clone MHM23) | Novus Biologicals, Littleton, CO | For optional protocol; Titering is suggested | |
BD CS&T Beads | BD Biosciences, San Jose, CA | ||
BD Accudrop Beads | BD Biosciences, San Jose, CA | ||
BC Versa Comp antibody capture beads | Beckman Coulter, Miami, FL | Compensation Beads | |
BD-FACS ARIA special research order instrument using 5 lasers | BD Biosciences, San Jose, CA | any BD-FACS aria with capabilities to detect the fluorochromes in the antibody panel should be sufficient | |
Wizard | Promega | A2360 | |
10 mM Tris-Cl buffer | NA | ||
Qubit dsDNA HS Assay kit | Life Technologies, Carlsbad, CA | ||
S2 Sonicator | Covaris, Woburn, MA | Alternatives may be substituted | |
microTUBE | Covaris, Woburn, MA | ||
Low-Throughput Library Preparation Kit | Kapa Biosystems, Wilmington, MA | KK8221 | |
Sybr Green | Sigma-Aldrich, St. Louis, MO | S9430 | |
Agencourt AMPure XP Beads | Beckman Coulter, Miami, FL | ||
Bioanalyzer | Agilent Technologies, Santa Clara, CA | ||
SeqCap EZ Exome v.3.0 | Roche Nimblegen | 6465684001 | |
HiSeq | Illumina | ||
TruSeq-style Universal adapter | Integrated DNA Technologies (IDT), Coralville, Iowa | HPLC purification; AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGAT*C*T | |
TruSeq-style index adapter | Integrated DNA Technologies (IDT), Coralville, Iowa | HPLC purification; /5Phos/GATCGGAAGAGCACACGTCTGAACTCCAGTCACNNNNNNATCTCGTATGCCGTCTTCTGCTTG | |
TruSeq-style PCR primer 1 | Integrated DNA Technologies (IDT), Coralville, Iowa | AATGATACGGCGACCACCGAGA | |
TruSeq-style PCR primer 2 | Integrated DNA Technologies (IDT), Coralville, Iowa | CAAGCAGAAGACGGCATACGAG | |
Nuclease Free Duplex Buffer | Integrated DNA Technologies (IDT), Coralville, Iowa | ||
BD FACSDIVA software | BD Biosciences, San Jose, CA | ||
BD Falcon Tubes | BD Biosciences, San Jose, CA | ||
BD Flow Tubes | BD Biosciences, San Jose, CA |