Isolating pulmonary microvascular endothelial cells ex vivo: Implications for pulmonary arterial hypertension, and a caution on the use of commercial biomaterials

Transcriptomic analysis of pulmonary microvascular endothelial cells from experimental models offers insight into pulmonary arterial hypertension (PAH) pathobiology. However, culturing may alter the molecular profile of endothelial cells prior to analysis, limiting the translational relevance of results. Here we present a novel and validated method for isolating RNA from pulmonary microvascular endothelial cells (PMVECs) ex vivo that does not require cell culturing. Initially, presumed rat PMVECs were isolated from rat peripheral lung tissue using tissue dissociation and enzymatic digestion, and cells were cultured until confluence to assess endothelial marker expression. Anti-CD31, anti-von Willebrand Factor, and anti-α-smooth muscle actin immunocytochemistry/immunofluorescence signal was detected in presumed rat PMVECs, but also in non-endothelial cell type controls. By contrast, flow cytometry using an anti-CD31 antibody and isolectin 1-B4 (from Griffonia simplicifolia) was highly specific for rat PMVECs. We next developed a strategy in which the addition of an immunomagnetic selection step for CD31+ cells permitted culture-free isolation of rat PMVECs ex vivo for RNA isolation and transcriptomic analysis using fluorescence-activated cell sorting. Heterogeneity in the validity and reproducibility of results using commercial antibodies against endothelial surface markers corresponded to a substantial burden on laboratory time, labor, and scientific budget. We demonstrate a novel protocol for the culture-free isolation and transcriptomic analysis of rat PMVECs with translational relevance to PAH. In doing so, we highlight wide variability in the quality of commonly used biological reagents, which emphasizes the importance of investigator-initiated validation of commercial biomaterials.

Limited reproducibility of published scientific results has led to an emerging initiative among funding sponsors, including the National Institutes of Health, that emphasizes data quality [21,22]. The widespread availability of commercial biomedical products has simplified reagent preparation and improved laboratory efficiency. However, inconsistent product quality-for example, uncertain binding epitopes among some commercial antibodies-may contribute to variability in experimental biology. In turn, data validating these biotechnologies is likely to improve rigor of scientific findings, but is rarely reported independently by investigators [4,[23][24][25][26].
Microvascular endothelial dysfunction has been implicated in multiple aspects of PAH pathobiology including angiogenesis, proliferation, apoptosis, and adaptation to shear stress [5,27,28]. In this report, we describe a practical method for isolating high-quality mRNA for transcriptomic analysis from rat pulmonary microvascular endothelial cells (PMVECs) ex vivo without cell passaging. In doing so, we also demonstrate wide variability in the quality of purchased laboratory reagents. Together, the current work outlines a cell culture-free approach to studying PMVECs, and reinforces the use of commercial biomaterials without on-site validation as a modifiable step toward enhancing the reproducibility of data in PAH.

Cell culture and reagents
Human pulmonary artery endothelial cells (HPAECs), human lung fibroblasts (HLFs), and human pulmonary artery smooth muscle cells (HPASMCs) were purchased from Lonza. Rat pulmonary artery endothelial cells (RPAECs), rat pulmonary artery smooth muscle cells (RPASMCs), and rat lung fibroblasts (RLFs) were purchased from Cell Biologics. The details of each cell type are provided in the S1 Supporting Information. Rat PMVECs were harvested in our laboratory from the peripheral region of the lung, expressed CD31 (Santa Cruz 376764 [Ab #1] for immunofluorescence and immunohistochemistry; R & D Biosystems FAB3628P [Ab #20] for flow cytometry), and, in some experiments were confirmed by co-labeling with isolectin 1-B 4 from Griffonia simplicifolia (GS-IB 4 ) (Thermo Fisher, Catalog #I21411). Cell culture was performed under standard conditions (37˚C, 5% CO 2 , 90% humidity) in vendorrecommended media (S1 Supporting Information). Cell passages 1-8 were used for experiments. Details on commercially purchased antibodies (Abs #1-22) used for experiments are presented in Table 1.

Statistical methods
All statistical analyses were performed using Origin Pro 2015 version b9.2.272. The unpaired Student's t-test was used for experiments involving comparisons between two samples and one-way analysis of variance (ANOVA) was used for comparisons involving three or more samples. Post-hoc analyses were performed using the method of Tukey. Data are presented as mean ± SE. P < 0.05 was used to define statistical significance. Initial data review of immunocytochemistry, immunofluorescence, and flow cytometry experiments was performed by a blinded investigator whenever possible. At least three technical or biological replicates were used per experiment. For colocalization, data were measured in at least 3 cells/field in at least 3 randomly-selected fields/technical replicate.

Cell isolation from experimental PAH in vivo
The overarching objective of this project was to isolate high-quality mRNA from rat PMVECs acquired ex vivo without tissue culturing. We chose to study rats based on severe histopathological remodeling and pulmonary hypertension that are reported for experimental PAH in this species compared to other rodent models [4]. Animals were handled in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and all procedures were approved by the Brigham and Women's Hospital Animal Care and Use Committee. A summary of the overall experimental workflow is provided in Fig 1. We first focused on immunophenotyping cells harvested from normal rats to demonstrate endothelial cell type (Fig 1A). Male Sprague Dawley rats (200-225g, Charles River) were anesthetized with ketamine (50 mg/kg)/xylazine (10 mg/kg), sacrificed by exsanguination under general anesthesia, and the outer~4 mm of peripheral lung tissue was resected [6,12]. The whole lung segments were stored in cold Dulbecco's Modified Eagle Medium (Life Technologies, catalog #11995-065) briefly until digestion was performed by submerging the tissue in type II collagenase (1 mg/mL, Worthington Biochemical, catalog #LS004176) for 20 min at 37˚C, with 5% CO 2 and 90% humidity.
Next, the digested lung fragments were transferred to a culture dish containing endothelial cell-specific culture medium (Vasculife with EnGS Life Factors Kit, Lifeline Cell Technology, catalog #LL-0004) supplemented with 1% (vol/vol) penicillin/streptomycin (Thermo Fisher, catalog #15140122)/amphotericin B (catalog #A2942-50ML), minced with sterile scissors (~100 times), and the tissue homogenate was filtered through a 70 μm cell strainer (Falcon, catalog #352350). The filtrate was then passed through a 20 μm strainer (Pleuriselect, catalog #43-50020-01). Trapped cells were eluted into culture medium and centrifuged at 330 x g for 5 min to generate a cell pellet. The pellet was resuspended and plated on a gelatin-coated P60 An endothelial phenotype was confirmed in confluent cells by phase contrast microscopy. These were presumed to be rat PMVECs (referred to as 'presumed rat PMVECs' throughout). Rat PMVECs used in ICC, IF, and pilot flow cytometry experiments were cultured (Fig 1A), and rat PMVECs used for RNA isolation were not cultured (Fig 1B). Determining cell type by immunocytochemistry Immunocytochemistry (ICC) was performed on presumed rat PMVECs using the 3,3´-diaminobenzidine substrate (DAB) method, as reported by our laboratory previously, with some modifications [1]. In addition, we confirmed with the vendors that the tested antibodies were compatible with this method. Cells were fixed in ice cold acetone or methanol, blocked in 10% bovine serum albumin (BSA), and then incubated with anti-CD31 primary antibodies #1 or #2, or anti-IgG (Ab #3) as control (dilution 1:100). The secondary antibodies (Abs #4,5) were incubated at a dilution 1:500. Cells were imaged using an Olympus BX51 microscope, Retiga 3000 camera (1-5 randomly-selected fields/condition). The DAB substrate luminosity was quantified using Image J (NIH) and expressed in arbitrary units (a.u.) normalized to IgG. There was no meaningful difference in CD31 or α-SMA signal between acetone and methanol fixation (S1 Fig).

Cell identification by flow cytometry: Establishing a positive control
Our experiments using ICC and IF aimed to confirm the identity of cells isolated from rats ex vivo using two standard lineage markers for ECs, but produced conflicting results. Therefore, flow cytometry was performed next (Fig 1B). To establish a positive control for the identification of pulmonary endothelial cells by flow cytometry, cultured HPAECs, HPASMCs, and HLFs were grown to confluence, lifted with Accutase (Thermo Fisher, catalog #A1110501), and blocked with 10% BSA in PBS for 15 min at 4˚C. Next, cells were incubated with directlyconjugated fluorescent antibody against CD31 (Ab #12, 3 μg/mL), CD144 (Ab #13, 4.5 μg/ mL), or isotype control (Ab #15, 3.0 μg/mL; Ab #14 4.5 μg/mL) for 45 min at 4˚C, washed, and resuspended in 250 μL 1% BSA in PBS for immediate analysis (details on flow cytometry methods are provided in the S1 Supporting Information).

Cell identification by flow cytometry: Confirming isolated rat PMVEC identity
Validation of endothelial surface markers for flow cytometry. There are limited data on the use of an optimal antibody labeling strategy for the immunophenotyping of rat lung endothelial cells by flow cytometry. Therefore, we tested a range of endothelial surface marker antibodies on presumed rat PMVECs. Commercially purchased RPAECs served as a positive control.
Presumed rat PMVECs isolated from normal rat lungs (as described in the S1 Supporting Information) were labelled with antibodies against CD31 (Ab #12 or #16, 18 μg/mL), CD144 (Ab #13, 18 μg/mL) or isotype control (Ab #15, Ab #14, 18 μg/mL), as these were the antibodies that effectively identified HPAECs by flow cytometry. Although rat is not listed as a target species for Ab #13 by its vendor, we nonetheless included it in this experiment because we observed wide variability in antibody reactivity across species for many antibodies. For example, Ab #12 is not recommended by the vendor to label human endothelial cells but did so effectively (Fig 3A). In presumed rat PMVECs, no meaningful labeling was observed for CD31 or CD144 by Ab # 12, 16, and 13, respectively, compared to isotype control (Fig 4A). Similarly, Ab #12 and #16 did not result in meaningful labeling of commercial RPAECs compared to isotype control (Figs 4B and 4C). In the case of Ab #19, a high degree of false-positive, but not true-positive, target labeling was observed despite using a conservative blocking step to attenuate non-specific signal (10% BSA) (Fig 4D).
Preparing rat PMVECs isolated by FACS for transcriptomic analyses. Overall, CD31-positive cells comprised 86.7 ± 2.9% (N = 3) of the total cell population isolated from rat lungs by magnetic bead selection. Although this is consistent with published data on other endothelial cell types, it is possible that including cells that were not positive for CD31 could adversely influence results of subsequent transcriptomic analyses [16]. Thus, we confirmed rat PMVEC identity by virtue of CD31 + GS-IB 4 double-positivity sorted by FACS. Only that cell population was used for isolating RNA in preparation for transcriptomic analyses. The FAC-SAria Special Order flow cytometer was used for cell sorting, as detailed in the S1 Supporting Information.

Isolating pulmonary microvascular endothelial cells ex vivo
To demonstrate that immunomagnetic bead selection for CD31 followed by FACS isolation of CD31 + GS-IB 4 double-positive cells is a valid method of isolating high-quality RNA for transcriptomic profiling in experimental PAH, male Sprague Dawley rats were administered a single intraperitoneal injection of monocrotaline (MCT) (Sigma C2401-1G) or normal saline control on day 0 of the protocol. Confirmed rat PMVECs comprised 78.1 ± 5.8% and 61 ± 7.3% of viable presumed PMVECs, in control and MCT-PAH, respectively (p = 0.1) (S3 Fig). RNA was isolated from confirmed MCT-PAH PMVECs on day 23 and assayed for quality (Agilent 2100 Bioanalyzer). RNA integrity number (RIN) (Agilent) was 9.2 ± 0.1 vs. 8.7 ± 0.1 (p = 0.003) for control vs. MCT-PAH, respectively (N = 6 rats/condition) ( Table 2). Full RNA electropherograms are available for all samples in S2 Supporting Information.

Discussion
In this report, we detail a novel approach to isolating PMVECs directly from rat lungs using flow cytometry. Specifically, magnetic bead selection for CD31 followed by FACS identified a high rate of double positive cells for CD31 and GS-IB 4 , which enhanced the specificity of cell recovery and permitted isolation of high-quality RNA from PMVECs without culturing cells. This allows for next generation transcriptomic analyses on PMVECs ex vivo without passaging, which is associated with changes in genomic and proteomic programing that may limit translational relevance of results [20]. Our data also illustrate discrepancies in the quality of commercially available and commonly referenced biomaterials that are used to identify cell type. Thus, a second major finding of this work relates to the critical importance of investigator-initiated validation of reagents used in PAH and other experimental biology fields.
Studying PAECs isolated from patients is ideal for experimental research, but PAH is a rare disease, limiting donor availability from explanted lungs, and accessing distal pulmonary arterials using minimally invasive methods is associated with risk. Isolating and phenotyping primary human or murine pulmonary endothelial cells has been reported previously using serial passaging with endothelial cell-selective media, anti-endothelial antibody-coated magnetic beads, or FACS [5][6][7][9][10][11][12]15,17,19]. However, experimental murine models do not recapitulate many features of PAH observed in patients, particularly plexogenic vascular lesions or severe pulmonary hypertension. By contrast, there are few reports focusing on methods for isolating pulmonary artery endothelial cells in rats despite important advantages of PAH models in this species. This may partly reflect inconsistent quality for rat-compatible commercially available anti-endothelial antibodies. For example, the anti-CD31 monoclonal antibody clone TLD-3A12 is commonly referenced as a rat lung endothelial marker [10,19,32,33]. But in our experience, this clone was ineffective for profiling presumed PMVECs by ICC or flow cytometry.
Our findings show that Ab #12, which is also derived from clone TLD-3A12, had excellent sensitivity and specificity for the detection of CD31 on HPAECs by flow cytometry. Importantly, Homo sapiens is not listed as a target species by the antibody vendor. This observation is consistent with limited access to (or availability of) data on the target antigen or derivative epitope for that clone specifically, as well as other antibodies more generally, which ultimately confounds predicting relevant biophysical interactions that may explain experimental results. Other antibodies against endothelial surface antigens were associated with poor specificity. For example, in some experiments we observed strong CD31 and vWF expression in HPASMCs. This trend was not limited to ICC or IF analytical methods: a false detection rate for CD144 positive cells was observed in >20% RPASMCs analyzed by flow cytometry, which is consistent with published reports from others [5]. Taken together, these data suggest that the recognition of pulmonary endothelial antigens by commercially available antibodies is variable and assay-dependent. This experience is in concert with accumulating data implicating reagent quality in the synthesis of low fidelity or irreproducible findings [24]. Our data expand this field by cataloguing the pervasiveness and extent of this problem across a wide spectrum of methodologies and antibodies that are purported to share the same target. Although these findings clarify the importance of using valid reagents for fundamental experimentation (i.e., cell type identification), a simple solution to this problem appears less certain. Antibody validation imposes a substantial financial and time burden on scientific investigators. For example, over the course of this project, antibody validation experiments consumed an estimated $69,096 in reagent, personnel, and FACS expenses-$9,545 (13.8%) of which was spent on the purchase of commercial antibodies (Fig 6). Reagent validation may, thus, warrant greater consideration when considering laboratory budgets by investigator and sponsor alike.
The intent of this project was to isolate high-quality mRNA from rat PMVECs successfully, and, therefore, an exhaustive evaluation of all commercially available products and labeling conditions was not performed. For example, antibodies against intracellular antigens were not tested since this necessitates cell fixation and/or permeabilization for FACS. However, others have reported that efficient RNA recovery from fixed cells may be feasible [34][35][36]. Although most cells and biomaterials were fresh, our experiments did not control for age and variability in storage conditions, which may have affected our results. We recognize that antibody labeling experiments typically require optimization; it is possible that different extracellular matrix coatings in vitro, cell culture conditions, and methods of fixation or immunolabeling could influence the results of the IHC, IF, and flow cytometry analyses. Significant RECA-1 expression was not observed by IF in presumed rat PMVECs. The mechanism of this observation was not explored, as we ultimately favored a FACS-based rat PMVEC isolation strategy that was not contingent on RECA-1 labeling. In general, we followed vendor-recommended antibody labeling protocols, and performed additional attempts at optimization when possible. Budgetary and time constraints precluded an independent evaluation of the purity of each lot of commercial primary cells, [37] although each primary cell type was analyzed by flow cytometry.
Further experiments are needed to confirm that our methodology isolated PAECs exclusively. Others have reported CD31 and α-D-galactosyl surface residue presence in non-endothelial cells, and distinguishing endothelial cells of vascular versus lymphatic origin was not a specific focus of this study [16,[38][39][40]. However, it has been shown previously that GS-IB 4 does not co-localize with lymphatic endothelial-specific cells (defined by LYVE-1 expression) in the lung, providing indirect evidence to cell endothelial populations isolated in our report are predominately vascular in origin [41,42]. Additionally, we used flow cytometry gating settings that were conservative, and focused on populations with the highest CD31 and GS-IB 4 signals to minimize the chance for detection of other cell types [16,38]. Nonetheless, alternative proteins, such as endoglin and vascular endothelial growth factor receptor-2, may provide enhanced endothelial specificity and should be considered in future research.
In summary, we present a methodology for isolating PMVECs from rats ex vivo using flow cytometry that does not require cell culture or passaging prior to transcriptomic analysis. We identified poor sensitivity and specificity of commercially available antibodies for pulmonary endothelial antigens. These collective findings have important implications for future work in experimental PAH, particularly translational endeavors that involve interrogating the 'omic' profile for PMVECs, and underscores the need for investigator-driven validation of key immunological reagents. (DOCX) S1 Fig. Detection of CD31 and α-smooth muscle actin by immunocytochemistry was similar with acetone or methanol cell fixation. Peripheral rat lung tissue was treated with mechanical and enzymatic dissociation, and the cell pellet was cultured in endothelial-selective medium. Presumed rat PMVECs were fixed in acetone or methanol and analyzed using anti-CD31 Ab #1 and anti-α-smooth muscle actin Ab # 7 immunocytochemistry. Luminosity was normalized to IgG (Ab #3). Representative images shown. a.u., arbitrary units. Student's

S2 Fig. CD31 and von Willebrand Factor colocalize in human pulmonary arterial endothelial cells and presumed rat pulmonary microvascular endothelial cells. (A)
Human pulmonary artery endothelial cells were labeled with either anti-CD31 Ab #1 or anti-von Willebrand Factor Ab #6 and analyzed using confocal microscopy to determine colocalization thresholds. (B) Peripheral rat lung tissue was subjected to mechanical and enzymatic dissociation, and the cell pellet was cultured in endothelial-selective medium. Presumed rat PMVECs, human pulmonary artery endothelial cells, human pulmonary artery smooth muscle cells, and human lung fibroblasts were fixed in acetone and co-labeled with anti-CD31 Ab #1 and anti-von Willebrand Factor Ab #6 and colocalization was measured using the thresholds established in panel (A). To enhance visualization, regions of colocalization are emphasized using a false-colored yellow overlay.