Original Article

Does FACS Perturb Gene Expression? Graham M. Richardson,1 Joanne Lannigan,2 Ian G. Macara1*

1

Department of Cell and Developmental Biology, Vanderbilt University Medical Center, Nashville, Tennessee 37232

2

Flow Cytometry Core, University of Virginia, Charlottesville, Virginia 22904

Received 25 June 2014; Revision Received 27 October 2014; Accepted 27 November 2014 Grant sponsor: National Cancer Institute; Grant number: CA132898 Grant sponsor: National Institutes of Health Additional Supporting Information may be found in the online version of this article. *Correspondence to: Ian Macara; Department of Cell and Developmental Biology, Vanderbilt University Medical Center, 465 21st Avenue S., U 3209 MRB III, Nashville TN 37232-8240, USA. E-mail: [email protected] Published online 13 January 2015 in Wiley Online Library (wileyonlinelibrary. com) DOI: 10.1002/cyto.a.22608 C 2015 International Society for V

Advancement of Cytometry

 Abstract Fluorescence activated cell sorting is the technique most commonly used to separate primary mammary epithelial sub-populations. Many studies incorporate this technique before analyzing gene expression within specific cellular lineages. However, to our knowledge, no one has examined the effects of fluorescence activated cell sorting (FACS) separation on short-term transcriptional profiles. In this study, we isolated a heterogeneous mixture of cells from the mouse mammary gland. To determine the effects of the isolation and separation process on gene expression, we harvested RNA from the cells before enzymatic digestion, following enzymatic digestion, and following a mock FACS sort where the entire cohort of cells was retained. A strict protocol was followed to minimize disruption to the cells, and to ensure that no subpopulations were enriched or lost. Microarray analysis demonstrated that FACS causes minimal disruptions to gene expression patterns, but prior steps in the mammary cell isolation process are followed by upregulation of 18 miRNA’s and rapid decreases in their predicted target transcripts. VC 2015 International Society for Advancement of Cytometry  Key terms FACS; mammary epithelial cells; miRNA

MURINE mammary glands are highly branched tubular organs that develop mostly postnatally, within subcutaneous fat pads. The ducts that comprise the mammary gland arise from multipotent stem cells that generate all of the lineages found in the mature ductal tree (1,2). Differentiated mammary epithelial cells (MECs) can be broadly classified into two populations: luminal epithelial cells that line the ducts, and myoepithelial cells that form a layer surrounding the luminal cells. However, there are multiple cell types within each of these two populations; for example some luminal cells express the estrogen receptor and some do not. To identify the mechanisms that regulate lineage commitment, researchers have conducted transcriptional analysis of discrete mammary cell populations isolated by fluorescence activated cell sorting (FACS) (3,4). These studies all rely on the assumption that robust differences in gene expression between separate lineages are due to endogenous differences in gene expression. However, to our knowledge, this assumption has never been tested. Our study now demonstrates that FACS produces minimal short-term transcriptional artifacts, but also emphasizes the need for care in the methods used to isolate primary cells.

MATERIALS AND METHODS Mice A 8–9 week old C3H mice were ordered from Harlan laboratories. Our handling and use of these mice strictly adhered to ACUC-approved protocol (M/12/080). Cell Preparation and RNA Extraction Mammary tissue from the 3rd, 4th, and 5th mammary fat pads was collected (minus the lymph nodes). Each replicate consisted of pooled tissue from three mice. Each mock sorting experiment contained three biological replicates from nine mice. Cytometry Part A  87A: 166 175, 2015

Original Article Tissue from each replicate was minced using surgical scissors and placed into digestion media (25 mL 50:50 DMEM/Ham’s F12 containing: 12.5 lL human insulin (Sigma, I9278-5ML) and 0.05 g collagenase A (Roche 11088793001). Tissue was digested for 45 min at 37oC (600 rpm, shaking) with pipetting to break up clumps of tissue every 15 min. Digested tissue was spun down and washed in 5 mL serum-free DMEM/Ham’s F12 media. Tissue was re-suspended in 1 mL of serum-free medium. A total of 2 U/mL of DNAse I (Thermo, EN0521) was added to the tissue and the tissue was swirled gently at room temperature for 1 min. A total of 3 mL of serum-free medium and 1 mL of heat-inactivated calf serum (Gibco) was added to the tissue. The tissue was pipetted to resuspend it in the full 5 mL volume and then spun at 1,500 rpm for 15 s. The pellet was washed in 5 mL of serum-free medium and then spun down for 15 s at 1,500 rpm four more times (for a total of five differential centrifugations). The tissue was resuspended in 1 mL of serum-free medium. A fraction of the cells from each replicate were lysed and RNA was extracted (“untrypsinized” condition) using mirVana RNA isolation kit (total RNA extraction protocol) (Ambion, AM1560). Remaining tissue was spun down again, and re-suspended in 1 mL of 0.05% trypsin/EDTA (Gibco). Tissue was incubated at 37oC for 2.5 min (shaken at 500 rpm). A total of 1 mL of heatinactivated calf serum was added to the digestion to inactivate trypsin, and then cells were mixed and passed through a 40 lM filter to produce a single-cell suspension. Cells were once more spun down and then re-suspended in 1 mL of FACS buffer (1x PBS, 1% calf-serum, 1 mM EDTA). A fraction of cells from each replicate was lysed and RNA was extracted as before (“trypsinized” condition). Cells were kept on ice as much as possible after this point. RNA was extracted from a further fraction of cells from each replicate following the mock FACS sort (“mock sorted” condition). RNA Extraction RNA was extracted from 2.5–5 3 105 cells (per sample) using mirVana RNA isolation kit (total RNA extraction protocol). At each step in the cell isolation process where RNA was collected, the cells were lysed and RNA was immediately collected in parallel for each biological replicate. Cells were lysed for RNA extraction within 20 min of treatment with trypsin and within 30 min of mock sorting. To minimize differences due to any batch variation, each experiment used reagents from the same kit. Mock Sorting A total of three mock sorts were performed using a Becton Dickinson FACSVantage SE Turbo sorter or a Becton Dickinson FACSAria II to sort unstained, trypsinized cells. Each mock sort contained three biological replicates isolated as described above. To minimize cell stress, cells were sorted at 33 PSI through an 80 micron nozzle on the FACSVantage SE Turbo sorter. On the FACSAria II, comparable conditions were selected; cells were sorted at 35 PSI through an 85-micron nozzle. For the mock sort using the FACSVantage SE Turbo sorter, sample and collection tubes were kept at room temperature for the duration of the sort for each biologCytometry Part A  87A: 166 175, 2015

ical replicate (20 min/replicate) and then returned to storage on ice until the mock sort was completed. Sample and collection tubes during one of the mock sorts using the FACSAria II were kept at room temperature during one mock sort and at 4oC in a separate mock sort experiment. Cells were briefly vortexed immediately before loading into the sorter. Sort gates with >99.9% of cells were used for the sort logic. Cells before sorting were in FACS buffer and then sorted into tubes containing 1x PBS, 5% calf-serum, 1 mM EDTA, and 1 U/mL of recombinant RNasin inhibitor (Promega, N2511). Ice and Room Temperature Controls To examine the effects of temperature on changes in gene expression, we kept fractions of cells from the FACSAria II mock sort (where sample/collection tubes were held at room temperature) on ice during the mock sort. The “ice control” condition was kept on ice during the hour where three biological replicates were mock sorted. The “room temperature control” cells were taken off ice and held at room temperature for 20 min (but not otherwise manipulated). These cells were then returned to ice storage until completion of the mock sort. RNA extraction for the ice and room temperature control cells was performed in parallel with the mock-sorted cells as described above, following completion of the mock sort. Flow Cytometry and Cytometry Analysis Cells from the FACSVantage SE Turbo mock sort experiment were analyzed by flow cytometry for changes in subpopulation composition before, and following immediately, the mock FACS sort. Cells were blocked in 10% rat serum on ice for 5 min before staining for mammary lineage markers. Cells were stained 1:250 with a-CD24-FITC (eBioscience, clone M1/69), a-CD31-PE (BioLegend, clone MEC 13.3), a-CD45PE (BD, clone 30-F11), a-Ter119-PE (eBioscience, clone TER119), and a-CD49f-PerCP-Cy5.5 (BioLegend, clone GoH3). A fraction of the trypsinized (unsorted) cells were stained and analyzed on a BD FACSCalibur. Unstained, or single stained cells were used as compensation controls. Compensation, population analysis, and quantitation were performed using Flowjo software (version 9.6). Cell Viability Assay Trypsinized and mock-sorted cells from the FACSVantage SE Turbo mock sort experiment were analyzed for cell viability using the countess automated cell counter (Life Technologies, C10227) system. Reported values are an average of two readings taken immediately before and after mock sorting. Immunofluorescence Staining and Analysis A fraction of each replicate from the FACSVantage SE Turbo mock sort experiment (unsorted and mock sorted) was fixed in 4% paraformaldehyde (PBS). Cells were permeabilized in 0.5% Triton X-100. Cells were blocked/stained in 1x western blocking reagent (Roche, 11921673001). Cells were stained with rat a-cytokeratin 8 (1:250) (DSHB; (U of Iowa), TROMA-1) and rabbit a-cytokeratin 14 (1:500) (Covance, PRB-155P). Cells were mounted and then imaged using a Zeiss 710-LSM. A total of 25 fields of view were recorded 167

Original Article (per-replicate/condition) and then counted for CK81 and CK141 cells. Real Sorting Mammary epithelial cells were isolated as previously described. Cells were sorted using the same nozzle and pressure as described in the Mock sort, using the gating strategy described in Ref. [3), using a-CD24-FITC (eBioscience, clone M1/69), a-CD31-PE (BD, clone MEC 13.3), a-CD45-PE (eBiosciences, clone 30-F11), a-Ter119-PE (eBioscience, clone TER-119), and a-CD49f-PerCP-Cy5.5 (BioLegend, clone GoH3), and a-Sca1-APC-Cy7 (BioLegend, Clone D7) to separate basal, luminal estrogen receptor positive, and luminal estrogen receptor negative cells. Staining for flow cytometry was performed as previously described. Following sorting, RNA was harvested as previously described. Quantitative Reverse Transcription Polymerase Chain Reaction (qRT-PCR) cDNA was generated from isolated RNA as previously described in Ref. (5). Transcript levels were analyzed by qPCR using a Biorad C1000 Thermocycler (with a Biorad CFX96 Real-Time PCR detection system) and Bioline’s sensimix containing SYBR and Fluorescein. We used U6 and 18S as housekeeping genes for miRNA gene normalization and Gapdh, Rpl4, Rpl27, and Rps29 as housekeeping genes for mRNA normalization. Gene expression values from each mock sort were internally normalized such that the average value of the three biological replicates for the “trypsinized” condition was equal to one. Data were analyzed using the 2–DCt method. See Supporting Information File S1 for primer sequences. Microarray Analysis Microarray analysis was performed on cells from the FACSVantage SE Turbo mock sort experiment. cDNA was made using the Ambion WT gene expression kit and labeled using the Affymetrix WT genechip WT terminal labeling kit using manufacturer protocols. cDNA was hybridized to Affymetrix Mouse Gene 1.0 ST Arrays (same chip batch to avoid batch variation) and analyzed using the Affymetrix GeneChip system (Fluidics Station 450, GeneChip Scanner 3000 7G). Normalization of the CEL files was performed using the “Oligo” package in R-BioConductor (6). The RMA subpackage was used to normalize the individual chips at the probe level and then at the transcript level. Linear models for microarray data (7) was used to get the differential expression in terms of the “top genes”, in R-BioConductor packages. Variations across each condition were considered to be statistically significant when a two-tailed t-test gave a P-value

Does FACS perturb gene expression?

Fluorescence activated cell sorting is the technique most commonly used to separate primary mammary epithelial sub-populations. Many studies incorpora...
1MB Sizes 5 Downloads 8 Views