Improved quantitative analysis of primary bone marrow megakaryocytes utilizing imaging flow cytometry

Authors

  • Lisa M. Niswander,

    1. Department of Pediatrics, Center for Pediatric Biomedical Research, University of Rochester Medical Center, Rochester, New York
    2. Department of Pathology and Laboratory Medicine, University of Rochester Medical Center, Rochester, New York
    Search for more papers by this author
  • Kathleen E. McGrath,

    1. Department of Pediatrics, Center for Pediatric Biomedical Research, University of Rochester Medical Center, Rochester, New York
    Search for more papers by this author
  • John C. Kennedy,

    1. Department of Pediatrics, Center for Pediatric Biomedical Research, University of Rochester Medical Center, Rochester, New York
    Search for more papers by this author
  • James Palis

    Corresponding author
    1. Department of Pediatrics, Center for Pediatric Biomedical Research, University of Rochester Medical Center, Rochester, New York
    • Correspondence to: James Palis, Department of Pediatrics, Center for Pediatric Biomedical Research, University of Rochester Medical Center, 601 Elmwood Ave. Box 703, Rochester, NY 14642. E-mail: james_palis@urmc.rochester.edu

    Search for more papers by this author

Abstract

Life-threatening thrombocytopenia can develop following bone marrow injury due to decreased platelet production from megakaryocytes (MKs). However, the study of primary MKs has been complicated by their low frequency in the bone marrow and by technical challenges presented by their unique maturation properties. More accurate and efficient methods for the analysis of in vivo MKs are needed to enhance our understanding of megakaryopoiesis and ultimately develop new therapeutic strategies for thrombocytopenia. Imaging flow cytometry (IFC) combines the morphometric capabilities of microscopy with the high-throughput analyses of flow cytometry (FC). Here, we investigate the application of IFC on the ImageStreamX platform to the analysis of primary MKs isolated from murine bone marrow. Our data highlight and address technical challenges for conventional FC posed by the wide range of cellular size within the MK lineage as well as the shared surface phenotype with abundant platelet progeny. We further demonstrate that IFC can be used to reproducibly and efficiently quantify the frequency of primary murine MKs in the marrow, both at steady-state and in the setting of radiation-induced bone marrow injury, as well as assess their ploidy distribution. The ability to accurately analyze the full spectrum of maturing MKs in the bone marrow now allows for many possible applications of IFC to enhance our understanding of megakaryopoiesis and platelet production. © 2014 International Society for Advancement of Cytometry

Platelets are anucleate, circulating blood cells essential for clot formation and hemostasis. Thrombocytopenia is a potentially life-threatening problem that afflicts patients with a wide range of conditions. Following chemotherapy and radiation therapy, thrombocytopenia can develop due to a loss of megakaryocytes (MKs), the cells in the bone marrow responsible for the steady-state production of more than 1 × 1011 platelets each day [1-3]. Accordingly, the study of megakaryopoiesis and its response to injury is critical for our understanding of platelet production and the development of new strategies for the effective treatment of thrombocytopenia.

Megakaryopoiesis in the bone marrow proceeds through the transition of lineage-committed MK progenitors to maturing MKs that undergo abortive replication cycles known as endomitosis [4]. Accordingly, MK maturation is characterized by progressive increases in DNA ploidy and cell size, followed by cytoplasmic maturation [5]. Ultimately, polyploid MKs, which comprise less than 0.5% of nucleated marrow cells and can reach over 50 µm in diameter in mice, extrude proplatelet extensions through the sinusoidal endothelium to shed platelets in the peripheral blood [6-9]. This process results in “exhausted” MKs, which have also been referred to by the misnomer “naked nuclei” comprised of a nucleus with a thin layer of cytoplasm surrounded by a cell membrane [10, 11].

The study of primary MKs has proven more problematic than many other hematopoietic cell types due to their low frequency in the marrow and to technical difficulties associated with their distinctive maturational characteristics. The majority of MK studies utilize transformed cell lines or in vitro production of an enriched MK population through maturation of progenitor cells isolated from murine adult marrow, murine fetal liver, human umbilical cord blood, or human peripheral circulation [12]. As in vitro differentiation does not fully recapitulate normal megakaryopoiesis, it is often necessary to validate these findings using primary uncultured MKs. Additionally, the study of primary in vivo MKs facilitates the elucidation of the complex processes of megakaryopoiesis and thrombopoiesis that occur in specialized marrow niches.

The current standard for the accurate quantification of MK number in the marrow is immunohistochemical analysis of bone marrow sections with antibodies raised against MK surface antigens, including components of the integrin complexes CD41/CD61 (GpIIb/IIIa) and CD42a–d (GPIX/GP1Bα/GP1Bβ/GPIV) [13, 14]. Immunohistochemistry (IHC) allows for the identification of MKs throughout their maturational continuum and is ideal for analysis of large, fragile MKs as they remain immobilized in the bone marrow. However, on sectioned tissue, it is not possible to accurately assess DNA content, a key indicator of MK maturation, and the evaluation of sufficient numbers of MKs for statistical analyses is extremely labor intensive.

To overcome the obstacle of low frequency in the marrow, investigators have increasingly turned to flow cytometric analysis of MKs [15-19]. Although flow cytometry (FC) permits the rapid sampling of many cells, there are concerns about the accurate enumeration and characterization of primary MKs. The MK lineage presents two main challenges for flow cytometric analysis. First, anucleate platelets, which share a surface immunophenotype with parental MKs, can frequently become activated and adhere to other cells during sample preparation [20, 21]. Accordingly, false positive events stemming from the inclusion of cells with attached platelets could result in overestimation of MK frequency and inaccurate ploidy measurements. Second, due to the wide range of cellular size, cytoplasmic complexity, and DNA content within the MK maturational continuum, it is necessary to include large (Forward scatter - area high), granular (Side scatter - area high) events in analysis. This broader gate may also encompass multicellular events, including MKs associated with other cells, which could potentially complicate the analysis of DNA ploidy.

To circumvent these challenges that limit the study of primary MKs, we established an assay using imaging flow cytometry (IFC; ImageStreamX, Amnis Corporation), which unites surface immunophenotype signal intensity measurements with morphometric quantifications and the localization of signal intensity. IFC is ideal for the rapid and quantitative analysis of cells traditionally defined morphologically, and our laboratory has previously applied this technology to murine erythropoiesis [22-26]. Here, we describe the accurate and reproducible analysis of primary MKs isolated directly from murine bone marrow using IFC and explore potential applications of this technology for the study of murine megakaryopoiesis.

Material and Methods

Mice

Seven- to 9-week-old female C57BL/6J mice (The Jackson Laboratory, Bar Harbor, ME) were used for all experiments. For total body irradiation (TBI) experiments, unanesthetized mice were confined in a plexiglass restraint and exposed to 4 Gy TBI at a dose rate of 1.6 Gy/min using a Shephard Irradiator with a 6,000 Ci 137Cs source and collimating equipment. Mice were euthanized by CO2 narcosis, and femoral marrow or intact hind limbs prepared as described. Unirradiated age-matched control mice were analyzed in parallel with each experiment. The University of Rochester Committee on Animal Resources approved all animal experiments.

Bone Marrow Preparation

Femoral marrow was flushed into PB2 (Dulbecco's phosphate-buffered saline and 0.1% glucose, Invitrogen, Carlsbad, CA; 0.3% bovine serum albumin, Gemini Bio-Products, Sacramento, CA) with 25 µg/ml heparin. Single-cell suspensions were prepared by trituration followed by passage through a 45-µm cell strainer (BD Biosciences, Caanan, CT). Marrow cell counts were obtained by hemacytometer.

For live cell IFC experiments, 107 marrow cells were blocked in 25% rat whole serum (Invitrogen) prepared in PB2 and stained in 12.5% rat whole serum with 1:100 FITC-CD9, phycoerythrin(PE)Cy7-Kit, eFluor450-CD41, and biotinylated Lin antibodies (Ter119, CD3, B220, Gr1, Mac1, CD105, Sca1) all from eBioscience (San Diego, CA). Secondary staining was performed with 1:500 streptavidin-conjugated AlexaFluor594 (Invitrogen). Immediately prior to analysis on the ImagestreamX (Amnis Corporation, Seattle, WA), cells in 100 µl of PB2 were stained with 2.5 µM DRAQ5 (BioStatus, Shepshed, UK).

For fixed cell IFC experiments, 107 marrow cells were stained as above except eFluor450-CD41 was replaced with PE-CD41 (BioLegend, San Diego, CA). After staining, samples were fixed with 1% methanol-free formaldehyde (PolySciences, Warrington, PA) for 10 min, washed in PB2, and permeabilized for 10 min in 0.1% TritonX (Sigma-Aldrich, Saint Louis, MO). Finally, 4′,6-diamidino-2-phenylindole (DAPI, Invitrogen) was added to a final concentration of 1 µg/ml.

For live cell conventional FC experiments, 107 marrow cells were stained as above except using FITC-CD41 (BioLegend) and the biotinylated Lin antibodies referenced above. Secondary staining was performed with 1:500 streptavidin-conjugated Alexa Fluor 594 (Invitrogen). Last, cells were stained with 2.5 µM DRAQ5 and 5 µg/ml DAPI.

Immunohistochemistry

Hindlimbs were removed, fixed in 4% paraformaldehyde for 24 h, decalcified in 10% EDTA for 10–14 days, and embedded in paraffin. Femoral sections (5 µm) underwent heat-based antigen retrieval (Dako, Denmark), blocking with 5% normal goal serum (Vector Lab, Burlingame, CA), primary staining with 1:300 anti-Gp1Bβ (Emfret, Germany) and secondary staining with 1:500 biotinylated goat anti-rat antibody (BD Biosciences). Visualization was performed with Vectastain ABC alkaline phosphatase (Vector Labs) followed by 45 µg/ml BCIP (Roche, Germany) and 175 µg/ml NBT (Roche) prepared in NTMT (100 mM NaCl, 100 mM Tris pH 9.5, 50 mM MgCl2, 1% Tween20). Gp1Bβ+ MKs per field was quantified for ≥10 40× objective fields, and normalized to MK/40× field in the untreated control for each experiment.

Imaging Flow Cytometry

IFC was performed on a two-camera ImageStreamX with INSPIRE acquisition software (Amnis). Cell files (50,000) were collected with a cell classifier applied to the brightfield (BF) channel to capture only events larger than 25 µm2, which includes erythrocytes but excludes background calibration beads. The fluor panels detailed above were captured in channels as follows for the live and fixed panels, respectively (Supporting Information Table 1): FITC—Channel 2, PECy7—Channel 6, eFluor450—Channel 7, AlexaFluor594—Channel 10, DRAQ5—Channel 11; FITC—Channel 2, PE—Channel 3, PECy7—Channel 6, DAPI—Channel 7, and AlexaFluor594—Channel 10. For both live and fixed panels, BF was set in Channel 1 for the first camera and Channel 9 for the second camera. Excitation lasers used with typical intensity settings include 405 nm (80 mW), 488 nm (100 mW), 594 nm (20 mW), and 658 nm (40 mW). All images were captured with the 40× objective and acquired at a rate of ∼200–250 cell images per second. With the 40× objective, image pixels are 0.5 µm2.

IFC data analysis was performed with IDEAS software (version 4.0, Amnis). Images were compensated based on a matrix generated by single-stained samples acquired with identical laser settings in the absence of BF illumination. Following compensation, the number of nucleated events in each 50,000-cell file was determined by the presence of DRAQ5 signal intensity (DNA), and a smaller file containing only the subset of events with CD41-positivity was created. Multiple (range 2–5) CD41+ files for each biologic sample were merged. The gating strategy beginning with a CD41 merged file is detailed in Results and in Figure 2. Descriptions of some of the masks are changed in the text to more clearly convey meaning; however, BF measurements are all within the default channel mask, measurements within a mask of CD41 signal area refer to the CD41 object mask, the “Area of CD9 and not CD41” refers to the object mask of CD9 where there is no overlap with the object mask of CD41 (Supporting Information Table 2). The features used to make measurements are labeled on each displayed plot and described in Results: intensity, area, aspect ratio intensity, aspect ratio, circularity, and perimeter (Supporting Information Table 3). For each experiment, the gates were set on bone marrow from an untreated control mouse run in parallel, and this template was used to analyze experimental marrow samples.

For each biologic sample, the frequency of MKs per nucleated bone marrow cell was determined by IFC. As red blood cells are preferentially lost in the staining procedure, and the nucleated and enucleated cell populations vary in the setting of in vivo marrow injury, a file of unmanipulated marrow with DRAQ5 DNA dye was run for each sample to accurately determine the frequency of nucleated bone marrow cells. Finally, MKs were normalized to per femur values utilizing the total cell counts on flushed marrow samples by hemacytometer.

Flow Cytometry

Data were acquired on an LSR II flow cytometer (BD Biosciences). FITC and AlexaFluor594 were excited by a 488-nm laser and collected with 525(50)-nm and 610(20)-nm band pass filters, respectively. DRAQ5 was excited by a 633-nm laser and collected with a 660(20)-nm band pass filter. DAPI was excited by a 405-nm laser and collected with a 450(50)-nm band pass filter. Compensation and data analysis were performed with FlowJo version 9.6.3.

Results

Our overall goal was to determine if IFC could be used to quantitatively assess the frequency and ploidy distribution of primary murine MKs. Investigators have used a number of standard FC approaches to study MKs including surface marker positivity, DNA content measurements, and parameters approximating size (Forward scatter; FSC), and cellular complexity (Side scatter; SSC) [19, 27]. Given the issues presented by the unique maturational characteristics of rare MKs and their shared surface phenotype with abundant platelets, we first utilized the imaging capabilities of IFC to visually evaluate the composition of a population defined solely with total signal intensity parameters standardly used to delineate MKs with conventional FC. This population was defined as events having positivity for CD41 (GpIIb), a commonly used immunophenotypic marker for the FC identification of cells committed to the MK lineage [13, 28], as well as DNA positivity with DRAQ5 staining. When we use standard FC to similarly delineate DNA+CD41+ events, approximately half of this population is additionally positive for markers of other hematopoietic lineages (Supporting Information Fig. 1); however, with standard FC, there is no way to ascertain whether these events are legitimate MKs in multicellular events or unwanted platelets adherent to nucleated cells. Conversely, with IFC the identity of these DNA+CD41+ events can be determined visually. Strikingly, 55% of the DNA+CD41+ population was comprised of platelets attached to other cell types rather than true MKs (Fig. 1A, blue). Additionally, 70% of MKs were found attached to other cells (Fig. 1A, green), with only 15% of the DNA+CD41+ population containing single MK events (Fig. 1A, red). These results demonstrate that the DNA+CD41+ MK population is markedly heterogenous and suggest the benefit of an IFC-based approach both to discriminate platelet contamination and to include MKs even when physically associated with other cells.

Figure 1.

IFC demonstrates contamination in the standard flow cytometric analysis of MKs. (A) Representative plot of a FC-based, total signal intensity-defined MK population (DNA+CD41+) using IFC, indicating more than half of the events contain CD41+ platelets attached to non-MK nucleated cells (blue, n = 6, ±SD). The majority of MKs are present in multicellular events (green) and not as single MKs (red). Classification of events in the DNA+CD41+ population was performed through a combination of the IFC gating strategy presented in Figure 2 and manual, visual discrimination. (B) Representative plot of the subclassified DNA+CD41+ population described in panel A demonstrating brightfield area versus brightfield aspect ratio (width/height). A “single” gate set on the total collected event population, which identifies single cells by small area and high aspect ratio, is also shown. With this strategy each 50,000-event file contains 65.7 ± 2.8% single cells (black gate, n = 10, ±SD), however, only 3.1 ± 2.1% (n = 10, ±SD) of MKs (red or green) are included. Additionally, 55.5 ± 6.1% of all MKs (red or green, n = 3, ±SD), and 88.8 ± 7.8% of single MKs (red, n = 3, ±SD) have the same aspect ratio as cells in the “single” gate, regardless of area. Examples of events (40× objective: Brightfield, eFluor450-CD41, DRAQ5) are shown with location on the plot indicated by an arrow. All of the images in Figure 1 were obtained with identical display settings.

Figure 2.

Identification of MKs by IFC. Single cell suspensions of flushed murine bone marrow cells were stained with FITC-CD9, PECy7-Kit, EF450-CD41, and the DNA stain DRAQ5. Events containing CD41 positivity were pooled from five 50,000-event files and analyzed by the gating strategy depicted in panels A through E and described in the text. Arrows matching the color of a gated population indicate successive gating of that population in the next panel. (A) Platelets are excluded based on little DNA-positivity within the CD41 mask. Inset depicts total DRAQ5 intensity of the single cell population (gated as in Fig. 1B) used to set DNA+ gates and to define 2N DNA content. (B) Further elimination of platelets based on small area and shape irregularity (aspect ratio intensity). (C) Exclusion of events in which the CD9 signal area is larger than the CD41 signal area. (D+E) Circularity measurements further refine the MK population. (F) Examples of IFC-defined MKs, which are comprised of events in the “round” gate (D, blue) and those “reclaimed” from the oddly shaped gate (E, blue). This population contains all MKs, regardless of whether or not there are contaminating cells within the same event. (G) Compensated images depicting examples of events eliminated or retained in each panel. From left to right (40× objective): Brightfield, FITC-CD9, eFluor450-CD41, DRAQ5. The location of each example is indicated with the corresponding number on the plots in panels A–E. All images in Figure 2 were obtained with identical display settings.

Light scatter parameters, including size approximations with FSC and cell complexity measurements with SSC, are routinely used as a surrogate for size detection in conventional FC analysis to exclude multicellular events. In addition, area versus height plots of these pulse parameters can be used to help distinguish round, single cells from doublets. Although the IFC platform does not have parallel correlates for these measurements, it allows for the direct determination of image area and aspect ratio (width/height), which are routinely used to discriminate single cells (small area and high aspect ratio) from multicellular events (large area and low aspect ratio) [29]. In our hands, 65.7 ± 2.8% (n = 10, ±SD) of all collected events from live bone marrow samples are categorized as analyzable single cells by these criteria; however, only 3.1 ± 2.1% (n = 10, ±SD) of MKs fall into this single-cell gate, representing the smaller, most immature MKs (Fig. 1B, black gate). We find that using the same aspect ratio but eliminating area criteria captures 88.8 ± 7.8% (n = 3, ±SD) of single MKs (Fig. 1B, red), but only 55.5 ± 6.1% (n = 3, ±SD) of all IFC-defined MKs (Fig. 1B, red and green), which also include MKs associated with other cells. Additionally, 43.7 ± 7.1% (n = 3, ±SD) of the DNA+CD41+ population with the same aspect ratio as single cells consists of rounder multicellular events with platelets attached (Fig. 1B, blue). Consistent with these data, following refinement of the DNA+CD41+ population with area/height FSC and SSC parameters in conventional FC, approximately half of the population remains positive for markers of other hematopoietic lineages (Supporting Information Fig. 1). With conventional FC, we cannot be certain if these events represent contaminating platelets or MKs, but our IFC data strongly suggest that this population includes both (Fig. 1A). Taken together, these results highlight the benefit of utilizing IFC to visualize and localize signal intensity as a strategy to include MKs associated with other cells and to exclude platelet contamination. Additionally, as our goal was to achieve accurate MK frequency measurements, these data led us to bypass typical, size-based exclusion criteria in our IFC approach for the analysis of MKs.

The IFC gating strategy we developed for the enumeration of primary bone marrow MKs is shown in Figure 2 for an untreated 7–9-week-old female C57BL/6 mouse. The marker panel was designed to specifically delineate MKs utilizing [1] CD41 and CD9, which are coexpressed throughout the MK lineage but are each individually expressed by a minority of other hematopoietic cells, [2] Kit (CD117), which is expressed by proliferating MK progenitors but decreases with cell maturity, and [3] live cell permeable DRAQ5 to detect DNA [28, 30-32]. Along with measurements of total fluorescent intensity standard with FC, IFC additionally allows for the use of signal-based masks that restrict measurements to specified regions within an image (Supporting Information Tables 2 and 3). The overlap between masks of CD41 signal and CD9 signal was used to ensure these markers were coexpressed on the same cell within a multicellular image. Subsequently, the CD41 signal mask was used to specifically restrict measurements (“features” in the IDEAS software) to MKs, even when segregated with non-MKs within a multicellular event. For example, measurement of DRAQ5 DNA signal intensity was limited to that within CD41-expressing MKs. Similarly, morphologic measurements including area, and shape-based measurements, such as aspect ratio (width/height) and circularity (average distance from center of mask to the edge of mask/variance in distance), were restricted to the CD41 signal mask.

To accurately quantitate MK frequency in the bone marrow and to analyze sufficient cells for maturational studies, it was necessary to collect several (range 2–5) 50,000-event files for each biologic sample and then concatenate events with CD41 positivity from multiple parent files. The technical limitations posed by large file size and processing time sets 50,000 events as the upper limit for each collected file. With this concatenation approach, we combine the approximately 100–150 MKs in each 50,000-event collection file for a final population that is more robust for analysis with typically 200–400 MKs. Starting with a CD41-enriched concatenated file, anucleate CD41-expressing platelets were excluded by selecting cells positive for DRAQ5 signal (DNA dye) within the CD41 mask (Fig. 2A). To set the lower limit of this positive gate, which includes cells with 2N and above DNA intensity within the CD41 mask, single cells were gated as in Figure 1B, and the 2N DNA intensity determined for each sample (Fig. 2A, inset). As some platelet-cell aggregates are positioned such that the CD41 area and DRAQ5 area overlap, we further refined the selected population by plotting CD41 area against shape-based feature CD41 aspect ratio intensity (intensity width/intensity height; Fig. 2B), to eliminate any small, irregularly shaped platelets. To limit analysis to MKs, MK progenitors with Kit signal positivity within the CD41 mask were eliminated (gate not shown). Although CD41 and CD9 have both been used as markers of the MK lineage, other marrow cell populations can express each surface marker singly, with CD9 expressed more broadly than CD41 [26]. Consequently, a Boolean logic mask was used to highlight regions of CD9 signal where CD41 signal was absent, and cells with more promiscuous CD9 signal were eliminated as a tactic to ensure these markers were coexpressed on the same cell within a multicellular image (Fig. 2C).

Following the application of these gates, the cell population was relatively enriched for MKs; however, further shape-based discriminations were required to eliminate large cytoplasmic fragment debris or irregular platelet clumps. To this end, the aspect ratio of the CD41 mask was plotted against its circularity, allowing for the separation of “round” MKs from cells with “oddly shaped” CD41 signal (Fig. 2D). Although most MKs fall under the “round” classification, a small number of MKs with cytoplasmic extensions were reclaimed from the “oddly shaped” population based on the circularity of an eroded CD41 mask (seven pixels removed from the edges of the mask; Fig. 2E). The final population of MKs is comprised of both the “round” cells and those “reclaimed from oddly shaped” and contains MKs throughout the size and DNA content continuum both as single cells and adherent to other non-MK cells (Fig. 2F).

An important aspect of these analyses is the inclusion of a control to reliably set gates for morphometric measurements. For each experiment, the bone marrow from an untreated, age-matched control mouse is run in parallel and used to set gate boundaries. The gating template is then applied to experimental samples. As described, the “DNA in CD41” gate (Fig. 2A) is set based on the 2N ploidy peak of the single-cell population. The CD41 area used to eliminate small platelets (Fig. 2B) is set at 60 µm2 for each experiment, based on area measurements we made by IFC on platelet preparations prepared from peripheral blood (data not shown). As MKs represent a continuum of cells of varying size and DNA content, the “CD41 same as CD9” gate (Fig. 2C) and CD41 circularity-based gates (Fig. 2D,E) are set based on visual observation of cells surrounding the gate boundaries in the untreated control sample. Although there is a degree of subjectivity in these final gates, we find very little variation in MK frequency determined for many biologic and experimental replicates.

IFC-defined MKs comprise 0.29 ± 0.05% of nucleated bone marrow cells in C57BL/6 mice (Fig. 3A, n = 45, ±SD). The maturing MK population can be further characterized by distinguishing exhausted MKs, high ploidy nuclei with a thin rim of cytoplasm that remain following platelet production [10]. Exhausted MKs are rarely studied as low levels of surface marker expression make them difficult to discriminate by standard FC; however, the morphologic analysis capacities afforded by IFC allow for the identification of this population [33]. Subtracting the DRAQ5-defined nuclear mask from the CD41-based cell mask with Boolean logic creates a “cytoplasm” mask. Exhausted MKs with a thin rim of cytoplasm can be differentiated from the cytoplasmically rich MK population by plotting the area of the cytoplasm mask versus the cell perimeter (Fig. 3B). This exhausted MK population comprises 0.046 ± 0.016% of nucleated marrow cells and contains cells with high DNA content and low levels of CD41 and CD9 surface marker expression (Fig. 3A,B).

Figure 3.

Quantification and characterization of the IFC-defined MK population. (A) Frequency of IFC-defined MKs (red, n = 45, ±SD) and exhausted MKs (EM, purple, n = 14, ±SD) in murine bone marrow. (B) Exhausted MKs can be distinguished by plotting the perimeter versus the area of a cytoplasm mask (CD41 signal mask minus the DRAQ5 signal mask). Images of example exhausted MKs (1 and 2) and a MK with more cytoplasm for comparison [3] are shown with the plot location indicated by number. Images (obtained with identical display settings) from left to right (40x objective): Brightfield, FITC-CD9, EF450-CD41, DRAQ5. (C) Representative histogram of ploidy distribution within the IFC-defined MK population obtained by measuring the intensity of DRAQ DNA dye within the CD41 signal mask for live cells. (D) The ploidy distribution based on DNA signal intensity within the CD41 mask of live cells with DRAQ5 (red, n = 6, ±SEM) and fixed cells with DAPI (blue, n = 9, ±SEM). (E) The ploidy distribution of exhausted MKs measured in live cell analysis with DRAQ5 (n = 6, ±SEM).

Limiting analysis of DNA content to within the CD41+ region allows for accurate measurement of ploidy, an indicator of progressive MK maturation, even when MKs are physically associated with other nucleated cells. However, the low incidence of MKs in the marrow coupled with the sampling of fewer cells with IFC than with standard FC results in less distinct DNA intensity peaks for each ploidy class. To circumvent this challenge, we set the 2N and 4N ploidy gates based on all single cells (small BF area and high BF aspect ratio gated as shown in Fig. 1B) and subsequent ploidy gates based on DNA intensity doubling (Figs. 2A insert and 3C). The clearest definition of ploidy classes was achieved with DAPI following sample fixation and permeabilization (Fig. 3D, “Fixed MK”; Supporting Information Fig. 2); however, reproducible results were obtained with less sample manipulation using DRAQ5 as a live cell DNA dye (Fig. 3C,D, “Live MK”). Live cell analysis with DRAQ5 consistently yielded an increase in MKs in the immature 2N and 4N as well as the mature 32N ploidy classes, which could be due to differential preservation of these cell populations or more precise DNA staining with DAPI and/or fixation and permeabilization. With both methods, ploidy distribution analysis demonstrates that the majority of IFC-defined MK precursors are 8N or above (Fig. 3D). Exhausted MKs are preserved with DRAQ5 analysis of live cells, and are almost exclusively 16N or above, consistent with having completed thrombopoiesis (Fig. 3E).

To validate our IFC approach for the quantification of primary bone marrow MKs, we compared IFC with IHC in the setting of in vivo injury following a sublethal dose of TBI. For this comparison, surface marker Gp1Bβ was used to define MKs in paraformaldehyde-fixed, paraffin-embedded femur sections, as detection of this MK lineage-specific antigen is more robust than CD41 in these preparations [14]. MKs defined by Gp1Bβ-positivity with femoral IHC decline in the marrow to 30% of unirradiated control levels by days 6 and 9 following 4 Gy TBI (Fig. 4A, grey). Importantly, independent analysis of IFC-defined MKs from flushed bone marrow reveals no significant difference compared to IHC in the timing or degree of MK injury following 4 Gy TBI (Fig. 4A, black).

Figure 4.

Comparison of IFC to standard IHC and FC for analysis of MKs. (A) MKs enumerated by IFC from flushed marrow cells (black, n ≥ 3, ±SEM) or by Gp1Bβ IHC of paraffin-embedded femur sections (grey, n ≥ 3, ±SEM) isolated from mice at the designated time points following 4 Gy total body irradiation (TBI). With both methods, MK numbers were normalized to an unirradiated control mouse for each independent experiment. Results are presented as percent of untreated control. Statistical analysis comparing IFC to IHC at each day postTBI was performed using a 2-tailed Student's t test (NS indicates not significant, p > 0.13 for each time point). (B) The frequency of two populations analyzed with IFC using total signal intensity measurements (approximation of conventional FC-based gating) in live cell preparations of flushed bone marrow: DNA+CD41+ (grey, n = 6, ±SEM) and DNA+CD41+ cells negative for markers of other hematopoietic lineages (Lin-DNA+CD41+, black, n = 6, ±SEM). This lineage marker set includes: Ter119, CD3, B220, Gr1, Mac1, CD105, and Sca1. Both populations are “DNA+” based on events with ≥2N total DNA content. (C) The ploidy distribution of the above described FC-based DNA+CD41+ (n = 6, ±SEM) population determined by total signal intensity of DRAQ5 for each event as would be obtained with standard FC (grey, “Total DNA”) and by the intensity of DRAQ5 only within the CD41 signal mask for each event which determines the DNA within the MK present in each event (black, “MK DNA”). (D) The ploidy distribution for the FC-based Lin-DNA+CD41+ population (n = 6, ±SEM) determined by “total DNA” (grey) and “MK DNA” (black) for each event as described for panel C.

Having validated the accuracy of IFC as a method to enumerate primary bone marrow MKs, we next compared IFC-defined MKs to two standard FC definitions for the MK lineage: CD41+ cells either with or without excluding cells expressing markers of other hematopoietic lineages (Lin). As in Figure 1, we used IFC to gate DNA+CD41+ cells with or without the exclusion of Lin+ cells using only the total signal intensity parameters available with standard FC. The FC-defined DNA+CD41+ population is over 10-fold more frequent than IFC-defined MKs and reported values (Fig. 4B). In contrast, the more stringent Lin-DNA+CD41+ population falls within reported ranges for bone marrow MK frequency though is statistically lower than the IFC-defined MK frequency (Fig. 4B black vs. Fig. 3A red, p = 0.037) [6]. DNA content measurements with DRAQ5 were used to characterize the composition of the FC-defined populations (Fig. 4C,D). In the DNA+CD41+ population, total DNA intensity measurements comparable to standard FC analysis indicated a shift toward >32N ploidy events (Fig. 4C, “Total DNA,” grey). However, more accurate measurement of DRAQ5 restricted to the CD41 mask within the same events revealed that the majority of this DNA+CD41+ population contains events with <2N DNA content within the CD41+ region (Fig. 4C, “MK DNA,” black), representing platelets attached to other cells (Fig. 1B). Accordingly, without morphometric refinements, the CD41 intensity-defined population overestimates both the number and true ploidy of MKs by including platelets and MKs associated with cells of other lineages. Addition of lineage exclusion criteria to the DNA+CD41+ population results in the elimination of MKs attached to other cells and more accurate ploidy measurements of the identified MKs, as the total event DNA content and the DNA content within the CD41 mask are consistent (compare black and grey in Fig. 4D). However, there is a significant underrepresentation of higher ploidy (≥8N) MKs with this approach relative to the IFC analysis (compare ≥8N black in Fig. 4D with red in Fig. 3D, p = 2.7 ×10−5), suggesting that higher ploidy MKs may be associated with lineage positive cells that are eliminated. Thus, while the Lin-DNA+CD41+ population does not contain an accurate ploidy distribution of bone marrow MKs, it represents a more pure MK population using standard FC than can be achieved with CD41-positivity alone.

Discussion

Here, we detail a novel assay utilizing IFC to investigate primary MKs isolated from murine bone marrow. This technology is ideal for the analysis of the rare, morphologically distinct MK population. IFC uniquely unites advantages of two standard techniques for MK lineage analysis: IHC and FC. Like standard FC, IFC allows for analysis of a large number of cells more rapidly than IHC methods. Furthermore, the cells analyzed with flow-based IFC are intact rather than sectioned, permitting the measurement of true cell size and DNA content as well as other standard flow cytometric assays including intracellular cell signaling and apoptosis determination. Like IHC, the imaging capabilities afforded by IFC provide the opportunity to use morphologic criteria to aid in MK identification and analysis. This facilitates both inclusion of the full spectrum of maturing MKs as well as the elimination of platelet contaminants, which share a surface phenotype with parental MKs and adhere to the surface of other cells (Fig. 1A). Recently, the propensity of platelets and released microparticles to complicate the pure analysis of other cell types has gained attention [20, 21]. This is especially relevant for the study of megakaryopoiesis, given the immunophenotypic overlap between MKs and contaminating platelets adherent to the surface of other nucleated cells.

An important strength of IFC is the ability to limit analysis of one stain to only the region of an image that is positive for a second stain. This strategy enables the specific analysis of MKs even when associated with other cells in a single event image (Fig. 1A). Bypassing an initial multicellular exclusion step eliminates an upfront loss of large MKs and those MKs segregating in cell clumps. Instead, the presence of DNA within the CD41 signal area, the colocalized surface expression of CD9 and CD41, as well as specific size and shape criteria, exclude contaminating platelets and allow for the targeted analysis of bona fide MKs (Fig. 2). Although the inclusion of multicellular events creates inaccurate ploidy determination when using standard FC measurements of total signal intensity, IFC permits restricting measurements (size, DNA content, etc.) to within the CD41 signal mask.

To optimize MK frequency determination, we found it essential to avoid the differential cell loss that accompanies the sample fixation and permeabilization required to use most vital DNA dyes. To overcome this obstacle, our standard MK panel includes DRAQ5, which rapidly stains dsDNA in live cells. Although fixation and permeabilization with DAPI staining creates cleaner ploidy peaks, equivalent MK ploidy distribution can be achieved with DRAQ5 by defining ploidy classes based on the sequential doubling of DNA content from the 2N peak set on total nucleated cells (Fig. 2A, inset). For experiments where the measurement of MK number in the marrow is not a priority, fixation and permeabilization with DAPI may be a better option for ploidy determination. Additionally, IFC will facilitate the application of other assays to primary, uncultured MKs, for example, phosphoprotein detection or DNA double-strand break identification with γH2AX staining, by accurately restricting analysis to the full spectrum of the MKs in the bone marrow [34, 35].

Ultimately, the combination of these IFC-based strategies results in the accurate and reproducible determination of MK frequency and nuclear maturation in the bone marrow. Our studies indicate that IFC-defined MKs comprise 0.29% of nucleated marrow cells (Fig. 3B), which closely agrees with the range of 0.1–0.5% reported in the literature with different quantification techniques [6]. One critical component of this IFC-based analysis is the inclusion of bone marrow from an age-matched untreated mouse in each experiment, which is analyzed in parallel with marrow from experimental mice. As many of the described features represent a continuum, gates for defining MKs are set based on this control. This has been described for IFC as a “morphometrically relevant biologic control” [29]. Importantly, the MK-defining gates in each experiment are appropriately adjusted based on visual examination of the cells along gate boundaries in this untreated sample. In addition to its value in setting morphologic gates on an experiment-to-experiment basis, the ability to visually examine each cell was imperative to the development of several strategies to delineate MKs with IFC. The IDEAS software is equipped to aid in identifying the features that will best separate two user-identified populations, for example MKs versus platelet contamination or MKs versus exhausted MKs.

The ploidy distribution of IFC-defined MKs also resembles published values for C57BL/6 mice, with the majority of MKs having DNA content ≥8N and modal ploidy of 16N [6, 36]. In addition, the relative frequency of low ploidy MKs is slightly higher with IFC than reports with histochemical techniques [37]. One explanation may be that the IFC approach has increased sensitivity for the detection of smaller MKs in the 2N and 4N ploidy classes. Ploidy of IFC-defined MKs provides a more accurate representation of the ploidy distribution of primary bone marrow MKs than standard FC-defined ploidy, which can either result in a falsely high proportion of 2N/4N cells or ≥32N cells depending on whether markers of other hematopoietic lineages are included (Fig. 4C,D) [18].

Significantly, IFC also facilitates the analysis of past-maturity exhausted MKs, a rarely studied component of the MK lineage which is impossible to identify with FC. This population was only analyzable in live samples and was not reliably preserved following fixation and permeabilization. Though DNA content is not part of the IFC-definition of exhausted MKs, cells that fall into the “cytoplasm”-related exhausted MK gates are at least 8N, with the vast majority being ≥16N (Fig. 3E). This agrees well with reports of exhausted MKs as having 8N or greater ploidy [5, 15]. In addition, our studies confirm electron micrograph studies indicating the exhausted MKs that remain following platelet production are not “naked nuclei,” but rather retain an intact surface membrane and thin layer of cytoplasm [10]. IFC affords a unique ability to analyze exhausted MKs, potentially providing new insights into the final stages of megakaryopoiesis and thrombopoiesis.

Although we demonstrate the validity of IFC as a tool for the rapid analysis of MKs, this is a new technology requiring the use of an instrument that may not be widely available, and FC is still required for cell sorting experiments. The studies reported here are also informative for investigators analyzing the MK lineage with standard FC. First, we highlight contamination in the MK population defined by FC that is necessary to recognize for the proper interpretation of results. Our data indicate that for flow cytometric analysis of the MK lineage, the exclusion of cells expressing markers of other hematopoietic lineages is imperative to decrease unwanted contamination of both platelets attached to non-MK nucleated cells and MKs associated with other lineage cells (Fig. 4B–D). For our studies, we included lineage markers beyond those standardly used in studies of hematopoiesis (see Methods), and for FC recommend expanding the markers of “Lin” to include as many non-MK markers as materials allow. Although the resulting population will be an underrepresentation of the maturing MK compartment, especially the large, higher ploidy MKs, it will represent a more pure population of MKs (Fig. 4D). Thus, this approach will make the determination of other flow-based measurements such as ploidy more accurate.

In summary, we detail a novel IFC strategy for the analysis of primary, uncultured MKs from murine bone marrow and provide evidence supporting the accuracy and reproducibility of this technique. Although IHC has the benefit of minimal physical cell manipulation and provides valuable spatial data within the marrow environment, it is labor intensive and lacks the ability to accurately measure cellular DNA content. Consequently, FC offers the advantages of rapid analysis and cell sorting based on immunophenotype. However, the data presented here demonstrate the FC-defined MK populations must include measures to ensure purity, which consequently prohibit acquisition of the full range of maturing MKs. IFC merges the ability to visualize cells with the speed of FC and thus provides a valuable, efficient new tool for the quantitative analysis of megakaryopoiesis. Through overlapping morphometric and flow cytometric analyses, the described IFC-based examination of MKs has the potential to significantly increase our ability to investigate the MK lineage throughout maturation in its in vivo setting. This technology will be especially relevant to the study of megakaryopoiesis in knockout mouse models or in models of systemic manipulation where it is essential that MKs develop in their native microenvironment rather than be matured in vitro from progenitors.

Acknowledgment

The authors gratefully acknowledge Tim Bushnell and the Flow Cytometry Core Facility at the University of Rochester Medical Center for technical support and critical review of this manuscript.

Ancillary