Cellular and Molecular Life Sciences

, Volume 71, Issue 19, pp 3841–3857

Cell surface antigen profiling using a novel type of antibody array immobilised to plasma ion-implanted polycarbonate

Authors

    • Department of Cell and Molecular BiologyKarolinska Institutet
    • Applied and Plasma Physics (A28)University of Sydney
  • Jelena Radenkovic
    • Department of Cell and Molecular BiologyKarolinska Institutet
  • Elena Kosobrodova
    • Applied and Plasma Physics (A28)University of Sydney
  • David McKenzie
    • Applied and Plasma Physics (A28)University of Sydney
  • Marcela Bilek
    • Applied and Plasma Physics (A28)University of Sydney
  • Urban Lendahl
    • Department of Cell and Molecular BiologyKarolinska Institutet
Research Article

DOI: 10.1007/s00018-014-1595-2

Cite this article as:
Main, H., Radenkovic, J., Kosobrodova, E. et al. Cell. Mol. Life Sci. (2014) 71: 3841. doi:10.1007/s00018-014-1595-2

Abstract

To identify and sort out subpopulations of cells from more complex and heterogeneous assemblies of cells is important for many biomedical applications, and the development of cost- and labour-efficient techniques to accomplish this is warranted. In this report, we have developed a novel array-based platform to discriminate cellular populations based on differences in cell surface antigen expressions. These cell capture microarrays were produced through covalent immobilisation of CD antibodies to plasma ion immersion implantation-treated polycarbonate (PIII-PC), which offers the advantage of a transparent matrix, allowing direct light microscopy visualisation of captured cells. The functionality of the PIII-PC array was validated using several cell types, resulting in unique surface antigen expression profiles. PIII-PC results were compatible with flow cytometry, nitrocellulose cell capture arrays and immunofluorescent staining, indicating that the technique is robust. We report on the use of this PIII-PC cluster of differentiation (CD) antibody array to gain new insights into neural differentiation of mouse embryonic stem (ES) cells and into the consequences of genetic targeting of the Notch signalling pathway, a key signalling mechanism for most cellular differentiation processes. Specifically, we identify CD98 as a novel marker for neural precursors and polarised expression of CD9 in the apical domain of ES cell-derived neural rosettes. We further identify expression of CD9 in hitherto uncharacterised non-neural cells and enrichment of CD49e- and CD117-positive cells in Notch signalling-deficient ES cell differentiations. In conclusion, this work demonstrates that covalent immobilisation of antibody arrays to the PIII-PC surface provides faithful cell surface antigen data in a cost- and labour-efficient manner. This may be used to facilitate high throughput identification and standardisation of more precise marker profiles during stem cell differentiation and in various genetic and disease contexts.

Keywords

Antibody arrayPlasma ion immersion implantation (PIII)Cluster of differentiation (CD)Embryonic stem (ES) cellCell capture profilingNeural differentiation

Introduction

To identify and sort subpopulations of cells from an initially more complex mixture has proven invaluable for progress in many areas of biology and medicine. Identification and sorting based on a specific repertoire of cell surface markers has, for example, been critical for the identification of heematopoietic stem cells [1, 2] and for elucidation of tumour-initiating cells within a larger tumour mass [3]. The success of future stem cell therapies is critically dependent on identification and standardisation of surface antigen profiles for a vast number of possible cell types, and will require high throughput methods for cell surface profiling. A number of strategies have been developed to establish cell surface marker profiles. Gene and protein expression arrays predict cellular expression in an unbiased manner, though do not guarantee cell surface display, and require sourcing of appropriate antibodies for further analysis. The use of mass spectrometry proteomics for identification of cell surface proteins provides important information [47], but is not only biased by the labelling method used but is expensive, critically dependent on the quality of lysate processing, and, similarly, dependent on the availability of suitable antibodies for validation and subsequent analysis. The possibility of arraying large numbers of antibodies in parallel is an alternative and interesting approach and has been carried out in high-throughput platforms including 96-well plate flow cytometry screens and nitrocellulose cell capture formats. These platforms have been used for experimental determination of cellular marker profiles [8, 9], and the diagnosis of human disease through Peripheral Blood Mononuclear Cells (PBMC) analysis [1012]. These methods screen over 100 antigens per assay and inherently identify antibodies suitable for further analysis. While 96-well flow cytometry screening is a powerful technique, it requires large numbers of cells and advanced flow cytometry facilities.

Cell capture arrays are easy to use, requiring only dissociation of cells and hybridisation for antigen analysis by light scattering [13]. Conventionally used nitrocellulose platforms are not optimal for cell capture applications due to their soft and porous nature which restricts contact printing, requires immobilisation to a glass surface and loss of antibody activity due to adsorption. Alternatively, surface modification of solid plastic surfaces with linker technologies has proven suitable for printing of protein- and oligo-based array applications, but current antibody immobilisation technologies require complex chemical treatment of these surfaces or antibody modification and may demonstrate species-specific immobilisation bias [1418]. In contrast, Plasma Ion Immersion Implantation (PIII) technology demonstrates ‘universal’ covalent binding to small molecules, proteins, DNA, and even viable cell surface proteins in a dense molecular monolayer without substrate modification [19]. Briefly, PIII treatment involves immersion of polymers in plasma where they undergo surface activation through substrate bias induced ion implantation. The energetic ions create collision cascades as they interact with the polymer surface, breaking bonds in polymer macromolecules. The formation of new bonds in the ion-implanted sub-surface region results in an amorphous carbon surface layer, which contains a significant density of unpaired electrons associated with radical groups in the structure. Radicals randomly migrate to the surface becoming available for covalent immobilisation of macromolecules (see Fig. 1 for schematic). The high density of radicals on PIII-treated surfaces enables high protein binding capacity for complete surface coverage with a dense, covalently bound, protein molecular monolayer, as opposed to island-like absorptive binding on untreated surfaces [20]. Furthermore, covalent immobilisation of enzymes and ECM molecules to PIII-treated polymers retains protein activity [2123]. Pill treatment also reduces the contact angle of polymer surfaces making them more suitable for biological applications and provides a long shelf life with covalent immobilisation possible at least 1 year after treatment.
https://static-content.springer.com/image/art%3A10.1007%2Fs00018-014-1595-2/MediaObjects/18_2014_1595_Fig1_HTML.gif
Fig. 1

Schematic of production, hybridisation and analysis of plasma ion immersion implantated polycarbonate (PIII-PC) cluster of differentiation (CD) antibody microarrays. a Ions from a nitrogen plasma are accelerated to collide with the polycarbonate surface. b Collision cascades under the polycarbonate surface create radicals (R*). c Radicals diffuse to the surface and covalently couple to printed antibodies physically adsorbed to the surface (left). d The unprinted surface binds milk protein during blocking (right). e During hybridisation cells expressing the respective CD antigens bind to immobilised antibody (left). f Specific binding is visualised through DAPI staining of DNA in cell nuclei (blue) and printed antibody is visualised by hybridisation of secondary antibody conjugated to Cy3 (red). Note that schematised images are not to scale

Advances in cell surface profiling have been largely aided by the use of CD (cluster of differentiation) antigens. Historically classified in leukocytes, CD antigens are defined as a cell surface antigen to which two monoclonal antibodies are validated to bind. Unique combinations of expression of CD antigens are widely used to, for example, trace cell lineage, identify diseased versus healthy tissues and to distinguish various developmental stages [8, 9, 24, 25]. While a single CD antigen is rarely sufficient to mark a single cell type, combinations of antigens can be used to distinguish specific cell types. Regardless of the profiling method used, most studies identify the expression of CD antigens [5, 24], and, in many situations, CD antigen status alone will be sufficient for differential characterisation of cell types [6, 7].

Stem cell research represents an area where an improved understanding of cell type heterogeneity is warranted. A detailed characterisation of various stem cell types and their differentiated progeny is essential to cancer stem cell applications and for understanding cell lineage differentiation. This is important for identifying subsets of cells and standardising surface profiles for high throughput pharmaceutical testing, disease modelling and personalised stem cell applications. Embryonic Stem (ES) cells represent a stem cell population capable of differentiating to every cell type of the adult body and are used as an in vitro model of embryonic development. Controlling media components directs ES cell lineage decisions, for example, to neural lineages using monolayer differentiation in chemically defined N2B27 media, which results in production of neural precursors, neural stem cells and neurons by 8 days of differentiation [2628]. Without exception, differentiation of ES cells occurs in a non-synchronous and heterogeneous manner resulting in mixed populations of terminally differentiated cell types as well as populations of cells at delayed states of differentiation along various lineages. This heterogeneity is a critical roadblock to regenerative medicine applications due to the tumourigenic nature of residual undifferentiated cells and impedes progress in lineage determination analyses. While analysis of a restricted number of cell types has begun [9, 2931], varied platforms, differentiation conditions, purification procedures and stages of analysis can complicate interpretation of results [9, 29, 32]. Standardised characterisation of the vast magnitude of possible cell types from ES differentiation with existing cell profiling technologies represents a labour-intensive and financially daunting undertaking, and would benefit from new profiling technologies.

In this report, we have developed and used PIII-PC CD antibody arrays to analyse the cell surface antigen profiles of various primary mouse cell types and genetically modified ES cell lines. Mouse antibodies are used to not only develop better cell characterisation capabilities but also for the advanced possibilities for specific genetic alterations in the mouse both in vivo and in ES cells. We compare PIII-PC array results with nitrocellulose arrays and flow cytometry to confirm the validity of this novel approach and identify changes in surface profiles between cell lines and also between undifferentiated and neurally differentiated ES cells. We also further characterise changes in differentiated cell type proportions in ES cells deficient for Notch signalling. In conclusion, the low cost, stability and ease of use of the PIII-PC platform makes this technology ideally suited to array applications including the standardised high throughput analysis of cell surface marker profiles.

Materials and methods

PIII polycarbonate treatment

Polycarbonate (PC) sheet with a nominal thickness of 1 mm was purchased from Goodfellow, Cambridge, UK. The PC sheet was cut into samples with a surface area of 25 × 75 mm2 and placed on an electrically isolated stainless steel holder with a mesh, electrically connected to the holder and held in front of the sample parallel to its surface. The distance between the sample and the mesh was 5 cm. The slide holder was then immersed in inductively coupled radio-frequency (13.56 MHz) nitrogen plasma, which was used as a source of ions for PIII. The forward power was 100 W with reverse power of 12 W when matched. The base pressure of the vacuum system was 10−6 Torr (10−4 Pa) and the pressure of nitrogen during implantation was 2 × 10−3 Torr (4.4 × 10−2 Pa). Plasma ions were accelerated by the application of high voltage (20 kV) bias pulses of 20-μs duration at a frequency of 50 Hz and drawing a current of 1.1 mA to the substrate holder and its mesh. The PC samples were treated for 40–1,600 s corresponding to ion implantation fluences of 5 × 1014–2 × 1016 ions/cm2.

Covalency testing

Untreated and PIII-treated PC sheets were cut into four 1 cm × 2.5 cm stripes each. Two of four samples were placed into Falcon tubes with 1 mM phosphate-buffered saline (PBS), pH 7.4, stored overnight at room temperature and washed for 10 s in mQ-water (Millipore). One of these two samples was dried overnight. The second sample was boiled in 1 % SDS for 30 min, washed for 30 min in mQ-water and dried overnight. The other two samples were incubated with mouse IgG (100 ug/ml) overnight at room temperature, washed in PBS for 30 min and for 10 s in mQ-water to remove unbound IgG. Then, one of these samples was dried overnight at room temperature and the second one was boiled in 1 % SDS for 30 min, washed in mQ-water for 30 min and dried overnight. FTIR ATR spectra were measured from the samples using a Digilab FTS7000 FTIR spectrometer fitted with a multibounce ATR accessory with a trapezium germanium crystal at an incidence angle of 45°. To obtain sufficient signal/noise ratio and resolution of spectral bands, 500 scans were taken at a resolution of 4 cm−1. The presence of mouse IgG on the polymer surface was inferred by the appearance of absorption-associated characteristic amide vibrations in the spectra of samples incubated in IgG after subtraction of spectra taken from the equivalent (boiled/not boiled in SDS) samples incubated in PBS. The spectra were analysed using Digilab software.

Construction of CD antibody arrays; blocking and storage of slides

Slides were spotted using drop (Biodot AD5000 Gantry System; Biodot, Irvine, CA, USA) or contact (VersArray Chip Writer Compact System; BioRad Laboratories, Ontario, Canada) printing robots. A total of 103 unique flow cytometry-tested CD antibodies generated against mouse antigens and their respective isotype controls (listed in Supplementary Table 1) were printed in duplicate onto PIII-PC. Antibodies were purchased from the following companies: BD Biosciences (North Ryde, NSW, Australia) and Abcam (Waterloo, NSW, Australia). Drop-spotting deposited 5 nL and contact spotting as little as 1 nL of antibody at concentrations as supplied. Antibody dots were air-dried and then blocked with 5 % (w/v) skim milk (Diploma; Bonlac Foods, Melbourne, Victoria, Australia) in PBS (90 min at room temperature), washed with water, dried at RT and stored at 4 °C with desiccant for up to 1 year.

Cell culture and neural monolayer differentiation

Mouse embryonic fibroblasts and cells from the 3T3-L1 adipose-like cell line, the C2C12 myoblast cell line and the RAW macrophage-like cell line were cultured in DMEM with 10 % FCS. ES cell lines used included 129 Sv cells (kind gift from Stuart Fraser) and CSL+/ and CSL/ ES cells (kind gift from Timm Schroeder and Tasuku Honjo [48]). ES cells were maintained on gelatin (Sigma)-coated dishes (Corning) in knockout DMEM medium (Gibco) supplemented with 5 % ESC tested FCS (Sigma), 5 % KSR (Gibco), glutamine (Gibco), non-essential amino acids (Gibco), beta-mercaptoethanol and LIF (Millipore) at 5 % CO2 at 37 °C. Cells were passaged the day prior to initiation of differentiation. The following day, cells were plated in N2B27 medium on gelatin at 1–2 × 10cells per well in 6-well plates or 0.25–0.5 × 105 cells per well in 24-well plates (Corning). N2B27 medium consists of a 1:1 ratio of DMEM/F12 (Gibco) and neurobasal medium (Gibco) supplemented with 0.5 % N2 (Gibco) 0.5 % B27 (Gibco), 0.1 mM beta- mercaptoethanol and 1× glutamine (Gibco). Medium was changed every day.

Primary tissue isolation of spleen-derived PBMCs

ACK lysis buffer was prepared by adding 8.29 g NH4Cl (0.15 M), 1 g KHCO3 (10.0 mM) and 37.2 mg Na2EDTA (0.1 mM) to 800 mL H2O, adjusted to pH 7.2–7.4 with 1 N HCl, made up to 1L with H2O and following filter sterilisation through a 0.2-μm filter was stored at room temperature. Mice were sacrificed with isofluran. Spleens were collected into separate 15-ml tubes containing 5 ml DMEM and kept on ice. Spleens were mashed with the plunger of a 10-ml syringe in a cell strainer on top of a 50-ml Falcon tube. The cell strainer was flushed with DMEM and cells were spun at 300 g for 5–7 min at 4 °C. Cells were resuspended in 3–4 ml ACK lysis buffer, and incubated for 1–2 min at RT. Twenty milliliters of 10 % FCS/PBS was added and cells were spun at 300g for 5–7 min at 4 °C. If the cell pellet was red, the ACK lysis was performed again. If the cell pellet was pink or white, the cells were resuspended in 10 % FCS/PBS for hybridisation.

Primary tissue isolation of hepatocytes

A two-step collagenase perfusion technique was performed. In the first step, 1× Hank’s buffered salt solution (Invitrogen #14175-095) with 0.5 mM EDTA and 25 mM HEPES was perfused through the liver to remove calcium ions. In the second step, 100 CDU/mL Collagenase Type I was perfused in digestion medium [DMEM-low glucose with 25 mM HEPES (#22320-022), 1.8 mM CaCl2] to disrupt supporting extracellular matrix. The procedure was carried out in situ using the animal’s own circulatory system to perfuse the liver via the portal vein or vena cava. The liver was then removed and triturated in digestion medium before filtering through a 70- to 75-µ membrane. Cells were centrifuged at 4°C for 2 min at 50g then washed in ice-cold 10 % FCS/PBS twice before resuspension in 10 % FCS/PBS and hybridisation.

Hybridisation protocol

Hydrophobic pap-pen was used to border antibody arrays. Cells were sequentially rinsed with DPBS, then dissociation buffer (1 mM EDTA/DPBS or PBS Based cell dissociation buffer; Gibco) and finally incubated for 4 min in dissociation buffer at 37 °C. Following pipetting from the culture surface, cells were diluted 1:1 in 10 % FCS/PBS and centrifuged for 3 min at 260g. Cells were resuspended in 10 % FCS/DPBS and gently pipetted to a single cell solution. Cells were counted using a burker chamber. Previously pap-penned array slides were dipped in DPBS to wet the surface and then laid flat, face up, in a humidified box. Two million cells were diluted to 400 uL in 10 % FCS/DPBS and applied evenly to the array surface. Slides were incubated at 37 °C for 1 h in a humidified box. The cell solution was poured off and the slides were gently dipped 3× in DPBS. Cells were fixed with 4 % PFA for 20 min at RT. To visualise spots and cells for counting, 400 uL of 1:100 diluted Cy3-conjugated anti-rat secondary antibody (or Cy3-conjugated protein-G) plus 1:1,000 DAPI in 10 % FCS/DPBS was applied per slide for 20 min at RT to visualise the antibody spots and cells, respectively. Slides were washed 3 × 2 min in DPBS then coverslipped with 50 % glycerol/DPBS for analysis and/or storage.

Cell counting analysis and normalisation

Fluorescent images were captured on a Nikon Eclipse E800 equipped with a Spot RT3 camera and Spot 5.0 software. Images were analysed using ImageJ for cell counting analysis. Briefly, cells were counted by inverting the DAPI image using the threshold option. A fixed area of the stained antibody spots was counted. Selection of the Process-Binary-Watershed separated the joined cells with the following parameters for Analyse–Analyse Particles (Size 0–200, Circularity 0.85–1.00, Show nothing, Check summaries), and the number of cells in the fixed area was determined.

CD antibody post staining and confocal analysis

Hybridised cells were blocked 1 h in 10 % FCS/DPBS at RT. Primary antibodies were diluted 1:100 in 10 % FCS/DPBS and incubated on cells either 1 h at RT or overnight at 4 °C. Slides were washed three times in DPBS, and secondary antibodies, diluted 1:100 in 10 % FCS/DPBS, were applied for 1 h at room temperature. The slides were then washed again, and coverslipped in 50 % glycerol/DPBS. For post-staining of internal antigens 0.1 % TritonX was added to the 10 % FCS/PBC solution. Images were acquired on a Zeiss LSM 510 confocal microscope, using the same settings for all slides. A ×20 confocal objective was used for confocal analysis of cells in 12-well tissue culture plates.

Immunocytochemistry of ES cells

Adherent cells were fixed for 10 min RT with 4 % PFA, washed 2× PBS and left in PBS overnight at 4 °C. Cells were blocked with 5 % FCS/PBS for 1 h and stained 1 h at 4 °C with primary antibodies diluted in 5 % FCS/PBS. Cells were then washed 3 × 5 min in PBS, and secondary antibodies, diluted in 5 % FCS/PBS, were applied for 1 h at RT. Cells were washed 3 × 15 min in PBS. All images were acquired with ×40 objective on a Zeiss Axiovert 200 M or Zeiss Observer Z1 inverted microscope using Openlab 3.1.7 software. Differentiation of ES cells on glass for confocal analysis was performed using poly-ornithine/laminin coating. Culture slides (BD Falcon) were coated for 2 h with 0.01 % polyornithine (SIGMA), washed 2 × 5 min with PBS and then coated overnight with 4 ug/mL laminin (Invitrogen). Slides were mounted in glycerol/PBS 1:1. Confocal images were acquired on a Zeiss LSM 510 (Zeiss, Germany), with Zeiss software.

Antibodies

Cluster of differentiation antibodies used throughout the study are listed in Supplementary Table 1. Other antibodies used for intracellular staining were anti-nestin (DSHB) and anti-nanog (Cell Signalling #8822SS). Anti-rat and anti-Armenian hamster secondary antibodies were sourced from Australian Biosearch.

Flow cytometry

One million cells were dissociated by rinsing first with PBS, then with PBS-based cell dissociation buffer (GIBCO) and then a 4-min incubation with CDB at 37 °C. Cells were washed down and diluted at least 1:4 in PBS. Cells were centrifuged 3 min at 1,200 rpm and resuspended in 100 uL of primary antibody diluted 1:100 in 10 % fetal calf serum (FCS)/DPBS. Following a 15 min primary antibody incubation at 4 °C, cells were centrifuged at 1,000 rpm for 1 min and washed twice by resuspension in 10 % FCS/DPBS. Following washing, cells were resuspended in 100 uL 1:100 secondary antibody diluted in 10 % FCS/DPBS for 15 min at 4 °C and then washed three times in 10 % FCS/DPBS. Cells were resuspended in 500 uL 10 % FCS/DPBS for analysis and analysed with a FACSCalibur and CellQuest Pro software.

Results

Production of PIII-PC arrays

We designed a method for production, hybridisation and visualisation of PIII-PC CD antibody microarrays (shown schematically in Fig. 1). Ions from a nitrogen plasma are accelerated towards the PC surface (Fig. 1a). Collision cascades of the impacting ions in the PC result in the formation of radicals (R*) (Fig. 1b), which diffuse to the surface and react with antibodies loaded during printing (Fig. 1c, left) or milk protein, applied on the unprinted surface during blocking (Fig. 1d, right). Following blocking, slides are hybridised with cells, which bind antibody if they express the respective antigen (Fig. 1e, left) but do not bind to milk protein (Fig. 1e, right). Specific binding is visualised through DAPI staining of DNA in cell nuclei (blue) and printed antibody is visualised by hybridisation of secondary antibody conjugated to Cy3 (red) (Fig. 1f).

In order to produce the array slides, we first assessed the intensity of PIII treatment required for maximal covalent immobilisation of macromolecules to treated PC surface. PC was PIII-treated between 0 and 1,600 s followed by coating with mouse IgG, washing with SDS and quantification of protein with Attenuated Total Reflection Fourier Transform Infrared Spectroscopy. The characteristic peaks of amide I, amide II, and amide A bands, at 1,650, 1,540 and 3,300 cm−1, respectively, indicate the presence of protein on the surface. Untreated PC (0 s) was able to bind IgG; however, upon SDS washing, the protein was lost from the surface, indicating that the protein was adsorbed rather than covalently immobilised. The fraction of covalently bound protein is the fraction of the initially bound protein remaining after SDS and is shown as the % covalency (Fig. 2a, triangles) indicating 320–800 s of PIII treatment gives optimal covalent protein binding to PC (Fig. 2a, squares). From these data, it was decided that a 400-s PIII-PC treatment would be used to ensure maximal covalent protein loading for production of all arrays used in this study. To allow replication of these data in other treatment systems, the fluence, defined as the total number of plasma ions accelerated to the surface per unit area of PC, was determined and is shown in Fig. 2. The integral absorbance (a measure of the darkening of the material) is also shown as ordinate on the graph. For maximal covalent protein loading, a fluence of 5 × 1015 ions/cm2 or 11.2 % integral absorbance gave optimal covalent protein binding to PC (Fig. 2a, dark and light grey bars as indicated). The treated surface is hereafter referred to as PIII-PC.
https://static-content.springer.com/image/art%3A10.1007%2Fs00018-014-1595-2/MediaObjects/18_2014_1595_Fig2_HTML.gif
Fig. 2

Development and optimisation of PIII-PC arrays. a The properties of polycarbonate treated for 0, 40, 80, 160, 320, 800, and 1,600 s with PIII are shown. The percentage of loaded protein that is covalently bound (triangles) and the amount of covalently bound protein (squares) indicates that 320–800 s of PIII treated gives optimal covalent protein binding to PC. The fluence and integral absorbance of PIII treatment in our system is shown (dark and light grey bars as indicated). b Cells bound to the PIII-treated surface following 1 % BSA and 5 % milk inactivation and stored under various conditions. The lowest non-specific binding occurred with 5 % milk treatment overnight at 4 °C. Bar graphs depict means from three experiments performed in biological triplicate, error bars indicate standard deviations. c Cells stained with DAPI (blue) specifically bound to CD9 antibody stained with anti-rat secondary antibody conjugated to Cy3 (red)

Prior to hybridisation of cells, activated surfaces must be blocked to scavenge remaining radicals from non-specifically binding to cell surface proteins. PIII-PC slides were blocked with 5 % skim milk or 1 % BSA. We found that 5 % skim milk incubation at 4 °C overnight was optimal for reducing non-specific cell binding (Fig. 2b) and was therefore used post-printing for blocking all slides used in this study. To visualise specific binding, cells were stained with DAPI and antibody spots stained with fluorophore conjugated antibodies (Fig. 2c).

Hybridisation of diverse cell lines to PIII-PC antibody arrays gives unique cell surface antigen profiles

A total 103 different CD antibodies, specific for mouse CD antigens were printed as an array onto PIII-PC slides, which were then blocked with skim milk. Isotype controls for all antibody species and subtypes of antibody were included on arrays to control for non-specific antibody binding. Details of these antibodies can be found in Supplementary Table 1. To validate the activity of covalently immobilised antibodies, we hybridised a range of mouse cell types and screened for attachment. Undifferentiated ES cells, PBMCs from spleen, cells from dissociated liver, mouse embryonic fibroblasts (MEF), 3t3-L1 adipose-like cells, C2C12 myoblast cells and the RAW macrophage-like cells were hybridised to separate PIII-PC arrays, printed with duplicates of all antibodies. Following hybridisation, cells were fixed with 4 % paraformaldehyde (PFA) and stained with DAPI and Cy3-conjugated secondary antibodies to allow clear visualisation of cells and printed spots, respectively. Signals were counted as positive when cells were specifically bound to antibody printed regions on both replicates of one slide with at least a doubling over background levels and with a minimum of 20 cells bound. Any positive signal determined for species- and subtype-specific isotype controls was deleted from corresponding primary antibody signals. A total of 52 of the 103 CD antibodies were found to bind cells from at least one tested cell population (listed in green in Supplementary Table 1). The lack of activity of the remaining 51 antibodies did not correlate with antibody subtype or the concentration and format (±BSA) of the original antibody solution (details listed in Supplementary Table 1), and is most likely due to lack of expression of these antigens on the restricted number of unique cell types we have used.

Table 1 shows positive signals for each cell line only when reproducible in all experimental (n) and technical duplicates. Due to the historical nature of CD antigen characterisation in lymphocytes, as expected, biological triplicate of spleen PBMCs hybridisation gave the greatest number of positive signals, reproducibly expressing 19 antigens, followed by RAW cells expressing 10 antigens and ES cells 7 antigens (Table 1). From triplicate experiments, distinct CD expression patterns were determined reproducibly with CD11b [33] and CD11c [33] specifically expressed in RAW cells, CD2 [34], CD19 [35], CD23 [36], CD27 [37], CD35 [37], CD45R [37], CD62L [38] and CD180 [39] in spleen PBMCs, and CD29 [40], CD71, CD98 [41] and CD326 [40] in undifferentiated ES cells (Table 1). While most of these expression data could be verified by literature as indicated, expression of CD11b may be expected in spleen-derived PBMCs, although the percentage of positive cells is <6 % and is known to decrease with age [35], possibly explaining why we do not see this expression on our array. Also, expression of CD71 has not been previously reported in ES cells, but we later verify this by flow cytometry (see Fig. 4a below). Combined with the lack of non-specific binding gauged by isotype controls, these results indicate maintenance of the accuracy and selectivity of antibodies for reproducible results when covalently immobilised to PIII-PC.
Table 1

Positive signals recorded for every tested replicate of hybridisation of RAW macrophage-like cell line (RAW), Mouse embryonic fibroblasts (MEF), undifferentiated C2C12 myoblast cells (C2C12), 3t3-L1 adipose-like cells (3T3-L1), cells from dissociated liver (Hepatocytes), undifferentiated ES cells (Undiff ESC) and peripheral blood mononuclear cells from spleen (Spleen), Antibodies are listed in the far left column

 

RAW n = 4

MEF n = 3

C2C12 n = 3

3t3 li n = 1

Hepatocytes n = l

Undiff ESC n = 3

Spleen n = 3

1d

    

+

  

2

      

+

9

+

  

+

+

+

+

11b

+

      

11c

+

      

16/32

+

   

+

 

+

18

+

     

+

19

      

+

21/35

      

+

23

      

+

27

      

+

29

   

+

+

+

 

31

      

+

35

      

+

36

+

   

+

 

+

44

+

+

+

+

+

 

+

44v6

   

+

   

45

+

     

+

45. R

      

+

49d

+

     

+

49e

 

+

+

+

+

+

 

51

   

+

   

54

+

   

+

+

+

61

 

+

+

+

  

+

62L

      

+

71

     

+

 

73

    

+

  

79b

    

+

  

90/90.1

    

+

  

93

    

+

  

98

   

+

 

+

 

105

   

+

   

106

 

+

+

    

180

      

+

326

     

+

 

A plus sign (+) indicates that this marker was positive in this cell line for all tested replicates as indicated (n)

PIII-PC array data are consistent with nitrocellulose array data

With our interest in determining developmental changes in surface antigens and to further validate our platform against previously characterised platforms, we continued our analysis with undifferentiated ES cells, which represent a homogenous cell population capable of differentiation to all adult lineages. Nitrocellulose-based CD antibody arrays have been validated to determine changes in antigen profiles between normal and diseased cell types. To determine if the behaviour of CD antibodies on PIII-PC is consistent with that on nitrocellulose, we printed antibodies and blocked arrays in the same manner and compared antigen detection. While existing reports detect cell capture data by light absorption, at this stage we chose to count DAPI-positive cells specifically bound to an antibody spot of fixed area. Compared to light absorption, this removes artifactual signals from debris and surface damages. PIII-PC arrays reproducibly demonstrated lower levels of non-specific binding to the unprinted surface (Fig. 3a). As demonstrated in Table 1, CD9, CD29 (Intb1), CD49e (Inta5), CD54 (ICAM1), CD71 (transferrin Receptor), CD98 and CD326 were reproducibly positive and bound similar cell numbers on PIII-PC and nitrocellulose-based arrays (Fig. 3b). These results are in line with published studies, which separately describe expression of markers in undifferentiated ES cells [4, 5, 4146]; however, once again CD71 expression has not been previously reported but is seen on both nitrocellulose and PIII-PC platforms. Post-staining of unpermeabilised cells and confocal analysis showed that cells bound to immobilised CD326 expressed CD54 and, vice versa, cells bound to immobilised CD54 expressed CD326 indicating that post-staining can be used to verify co-expression (Fig. 3c). Together, these results indicate that our PIII-PC array platform detects real antigen expression at a level consistent with nitrocellulose platforms; however, with the advantages of lower non-specific binding, a transparent platform, lack of requirement for advanced manufacturing and minimal risk of breakage or associated harm to users. Also, the solid surface of PIII-PC allows contact printing and does not allow absorption of antibody, unlike the porous structure of nitrocellulose that absorbs antibody away from the surface obscuring a proportion of potential antibody binding. Combined, these results indicates that PIII-PC arrays retain similar activity to nitrocellulose-based cell capture arrays, demonstrating that covalent immobilisation of antibodies to polycarbonate is not detrimental to their specificity/activity. Further, there are both commercial- and user-associated benefits to the PIII-PC platform over nitrocellulose platforms.
https://static-content.springer.com/image/art%3A10.1007%2Fs00018-014-1595-2/MediaObjects/18_2014_1595_Fig3_HTML.gif
Fig. 3

PIII-PC array results are compatible with results from nitrocellulose arrays for undifferentiated ES cells. a DAPI staining (white) indicates cells bound to antibody-free surface (background) or CD9 antibody on either PIII-PC or nitrocellulose. b Quantification of cells bound to CD antibodies on nitrocellulose or PIII-PC. Bar graph depicts means from three biological replicates, error bars indicate standard deviation. c Confocal images of unpermeabilised cells bound to CD326 and stained with the CD54 antibody (green) or cells bound to CD54 stained with the CD326 antibody (red). Cells are visualised with DAPI staining (blue) and merged with antibody staining to show surface protein expression

Cell number per spot correlates with the intensity of antigen expression by flow cytometry

Flow cytometry is the most commonly used method for cell surface antigen expression verification. Flow cytometry analysis, with antibodies identified as positive in ES cells by PIII-PC arrays, showed that almost 100 % of undifferentiated ES cells expressed CD9, CD29, CD54 and CD98, while levels of CD49e and CD326 were variable around 70 % and the levels of CD71 were very low (Fig. 4a; gating is shown for Armenian hamster antibodies and rat antibodies, respectively in Fig. 4b). The low level of expression of CD71 not only verifies the novel expression of this marker on ES cells but also demonstrates the high degree of sensitivity of microarray cell capture techniques. To verify the specificity and selectivity of PIII-PC arrays, antibodies determined as negative in ES cells but positive in spleen were verified by flow cytometry analysis. CD44, CD62L and CD180 were found to be positive in spleen cells while demonstrating background levels of expression in undifferentiated ES cells (Fig. 4c). CD54 was determined as positive in both cell types and was also verified by flow cytometry (Fig. 4c). By graphing biological triplicates of the number of cells per spot against the mean intensity of antigen expression by flow cytometry, an R2 value of 0.54646 was obtained by linear regression (Fig. 4d). This value indicates a moderate positive correlation between cell number per spot and surface antigen expression intensity. Cell surface expression of CD antigens on Nanog-positive ES cells was verified by post staining (Fig. 4e). We also see that, compared to a typical flow cytometry protocol, where cells are dissociated and stained with a fluorophore-conjugated primary antibody, the procedure of PIII-PC hybridisation is advantageous because, for the equivalent number of cells, up to at least 5× less antibody is required for a positive signal, primary antibody fluorophore conjugation is not required for analysis, expression of over 100 antigens can be determined in a single hybridisation and analysis can be performed with any fluorescent microscope in the absence of advanced cytometry facilities, which would be required for an equivalent flow cytometry analysis. Together, these results demonstrate the usefulness of this technique for quantitative assessment of antigen expression for homogenous populations and the power of the platform for identification of antibodies for further analysis by flow cytometry and morphological analysis.
https://static-content.springer.com/image/art%3A10.1007%2Fs00018-014-1595-2/MediaObjects/18_2014_1595_Fig4_HTML.gif
Fig. 4

PIII-PC array data for undifferentiated ES cells correlate with data from flow cytometry. a Flow cytometry analysis of undifferentiated embryonic stem cells using antibodies identified from PIII-PC arrays. The percentage of positive cells compared torat or Armenian hamster isotype controls is shown. Bar graph depicts means from three experiments performed in biological triplicate, error bars indicate standard deviation. b Gating for Armenian hamster and rat antibodies are shown. The red line indicates the isotype control and the blue line the shift in the population due to specific primary antibody binding for CD29 (left) and CD71 (right). c Expression of CD44, CD54, CD62L and CD180 in undifferentiated ES cells and spleen by flow cytometry. Anti-armenian hamster (AH-488) and anti-rat (rat-Cy3) secondary antibody only controls indicate non-specific signal. Values are relative to the highest level of expression in each population. d Means from biological triplicate experiments showing mean population intensity by flow cytometry (x-axis) versus cell number per spot on PIII-PC arrays (y-axis). Correlation by linear regression is shown with line equation and R2 value = 0.54646 indicating correlation. e Immunofluorescent staining of undifferentiated ES cells with CD54, CD326, CD29, CD49e, CD9 and Nanog antibodies

Changes in cell surface marker expression during ES cell differentiation can be detected with PIII-PC arrays

The ability to determine quantitative changes in antigen expression indicates that PIII-PC array may be used to monitor changes in cell type proportions that occur during ES cell differentiation, evident as up- or downregulation of surface antigens. To test if our combination of antibodies is sufficient to detect changes in cell type proportions that occur during neural differentiation, we hybridised undifferentiated ES cells and N2B27-differentiated neural monolayers onto separate PIII-PC arrays. Differentiations resulting from 8 days of N2B27 monolayer culture contain a number of neural cell types including neural progenitors, neural stem cells and neurons [26]. Hybridisation of biological triplicates of undifferentiated and day 8 neural differentiations to PIII-PC arrays demonstrated a clear decrease in CD54 and CD326 and a corresponding increase in CD71 upon differentiation (Fig. 5a). Cell numbers per spot were comparable for CD9, CD29 and CD98 in both undifferentiated and differentiated cell populations, whereas CD49e was higher in undifferentiated cells and CD44 and CD71 higher in differentiated cells (Fig. 5b). While almost 100 % of cells in undifferentiated cells were positive for CD9, CD29, CD54 and CD98 by flow cytometry, these numbers decreased to roughly 17, 26, 13 and 43 % of cells in the differentiated population (Fig. 5c). Neurally differentiated cultures hybridised to CD9, CD29, CD71 and CD98 were post-stained with nestin, a marker for neural progenitors [47]. Subsets of these cells stained positive for nestin, indicating a neural identity and not simply that undifferentiated cells remain in differentiated cultures (Fig. 5d). Interestingly, staining of intact neural differentiations demonstrated that CD9 was expressed specifically in the apical centres of neural rosettes (Fig. 5e, top panel) and also at high levels in nestin negative, non-neural cell types demonstrating CD9 is not only specifically expressed in undifferentiated embryonic stem cells but in neural stem cells and in, as yet uncharacterised, non-neural cell types (Fig. 5e, lower panel). These results demonstrate that PIII-PC arrays alone are sufficient to quantitatively identify changes in cell surface marker profiles during cellular differentiation.
https://static-content.springer.com/image/art%3A10.1007%2Fs00018-014-1595-2/MediaObjects/18_2014_1595_Fig5_HTML.gif
Fig. 5

Differential expression of CD antigens in undifferentiated and differentiated cells determined with PIII-PC arrays. a Cells bound to CD54, CD71 and CD326 and stained with DAPI for undifferentiated cells (white) and neurally differentiated ES cells (blue). b Cell number bound to CD antibodies immobilised to PIII-PC for undifferentiated ES cells (black bars) and cells differentiated for 8 days in N2B27 monolayer differentiation (grey bars). c Percentage of cells in day 0 undifferentiated ES cells and day 8 neural differentiations expressing CD antigens by flow cytometry. d Cells bound to CD9, CD29, CD71 and CD98 immobilised to PIII-PC, permeabilised and stained for nestin (red), cells visualised with DAPI (blue). e Immunocytochemistry of day 8 neural differentiations with nestin (green) and CD9 (red). The upper panel shows neural rosettes, and the lower panel shows non-neural cells. Cell nuclei are visualised using DAPI staining. Bar graphs depict means from biological triplicates, error bars indicate standard deviation

Characterisation of non-neural differentiation occurring in the absence of Notch signalling

We have previously identified changes in neural differentiation in the absence of Notch signalling: specifically, neural progenitors decrease and non-neural differentiation increases after 8 days of culture [27]. This was analysed in cells in which the CSL gene was deleted [48], eliminating canonical CSL-dependent Notch signalling [49]. As we have shown that our arrays can detect changes in cell type proportions, we hypothesised that CD antigens enriched in CSL+/− over CSL−/− differentiations will mark neural stem cells and those enriched in CSL−/− over CSL+/− will define non-neural differentiation. While no changes in cell surface profile were seen between undifferentiated CSL+/− and CSL−/− ES cells (Fig. 6a), consistent with lack of an overt phenotype in the CSL−/− cells at this stage [27, 50], analysis of day 8 differentiations showed that higher numbers of cells bound to CD49e and CD117 in CSL−/− compared to CSL+/− differentiations, indicating that these markers may be expressed on non-neural cells (Fig. 6b). Immunocytochemistry analysis of day 8 differentiations showed that CD49e and CD117 were indeed expressed in cells with a non-neural morphology (Fig. 6c). These results demonstrate that our antibody array allows for identification of novel markers for hitherto uncharacterised differentiated cell types.
https://static-content.springer.com/image/art%3A10.1007%2Fs00018-014-1595-2/MediaObjects/18_2014_1595_Fig6_HTML.gif
Fig. 6

Elucidation of antigen expression through comparison of CSL+/− and CSL−/− ES cell neural differentiations. a Cell number per spot for undifferentiated CSL+/− and CSL−/− ES cells. b Cell number per spot for CSL+/− and CSL−/− differentiated for 8 days in N2B27 monolayer neural-inducing conditions. c Immunofluorescence of day 8 differentiation of CSL−/− ES cells, permeabilised and post-stained with CD117 and CD49e. Bar graphs depict means from three experiments performed in triplicate, error bars indicate standard deviation

Discussion

There is a clear need to improve technologies for characterisation and sorting of specific sub-populations of cells out of initially more complex cell ensembles. This is important for human cells, for example for cancer stem cell, pharmaceutical testing and personalised stem cell applications, but also for other species, and notably for the mouse, given the increasing number of genetically modified mice and mouse ES cell lines. To address this need, we here report on the development of a novel mouse CD antibody array, which allows for profiling mouse cell surface CD antigens and offers several advantages over existing antibody-based profiling platforms.

The method is based on transparent polycarbonate slides and offers efficient analysis of cell surface antigen expression on live, intact cells. We show that immobilisation of CD antibodies to PIII-PC antibody arrays does not disturb antibody function, and that cells can be post-stained for surface and intracellular markers allowing not only confirmation of the immunoreactivity but also simultaneous characterisation of other cell surface markers on the cells. With the panel of cell lines tested, we were able to verify activity of 52 CD antibodies. With these antibodies, we defined unique profiles for almost all cell types tested, validated in ES cells to represent true protein expression. PIII-PC array results were consistent with nitrocellulose arrays and flow cytometry data demonstrating that PIII-PC antibody arrays faithfully record cell surface antigen expression. Furthermore, both high and low expression levels could be detected by the PIII-PC arrays; for example, the low level of CD71 expression in undifferentiated ES cells was confirmed by flow cytometry.

The PIII-PC array shows distinct advantages over existing platforms. First, the PIII-PC array is transparent, which allows light microscopy analysis for multiple antigen post-staining. The transparency of PIII-PC along with the increased sensitivity will be useful for upscaling and high throughput analysis of antibody arrays, which is currently hampered by the autofluorescence and opacity of nitrocellulose. Second, when analysed in both wet and dry formats, PIII-PC gave very low background signals with regard to both autofluorescence and non-specific cell binding as compared to nitrocellulose arrays. The lack of requirement for advanced manufacturing, a solid surface that allows contact printing and does not absorb antibody, and the minimal risks of breakage or associated harm to users are important considerations in development of commercial products. Compared to flow cytometry, for the equivalent number of cells analysed PIII-PC arrays use at least five times less antibody, do not require primary antibody conjugations (broadening the scope of suitable antibodies), detect expression of over 100 antigens in a single hybridisation and can be analysed in the absence of advanced cytometry capabilities. This is particularly important for analyses where the number of cells is restricted, for example in biopsy and embryo analyses, in order to minimise handling and maximise results. Use of this platform for identification of antibodies for further morphological or flow cytometry analysis was demonstrated to be a successful method for characterisation of cell types and in the formation of biologically interesting questions like the role of CD9 in the apical domain of neural stem cells (NSC), known to control symmetric and asymmetric NSC cell divisions [27].

The data also demonstrate that new insights into cellular differentiation and the consequences of targeting key regulatory genes can be gained from the PIII-PC arrays. The in vitro differentiation of ES cells to neural fates is a well-established strategy to explore neural differentiation, but we still have a limited repertoire of markers to monitor the steps in the differentiation process and the diversity of resultant cell types. Additional markers are needed to better understand neural differentiation and for characterisation of asynchronous differentiation and non-neural fates. These PIII-PC array experiments have added a number of new markers, including loss of CD54 and CD326 during neural differentiation, and contrasting increases in CD71. CD9, CD29 and CD98 were expressed both during undifferentiated and differentiated conditions, whereas CD49e was higher in undifferentiated cells and, conversely, CD44 and CD71 were more highly expressed in differentiated cells. Post-staining analysis for intracellular and extracellular markers have been used to further extend the expression profile of nestin-positive neural progenitors, for which a subset express CD9, CD29, CD71 or CD98. While expression of CD9, CD29 and CD71 has been identified in neural progenitors in separate studies [29, 51, 52], we are the first to identify expression of CD98 in a neural subset, and also the first to show neural rosette luminal accumulation of CD9, which indicates an unidentified role for this marker in NSC polarity and maintenance [27, 53].

The PIII-PC array data also contribute to a better understanding of the consequences of specific genetic modifications, and we analysed the effects of genetic ablation of the CSL gene, which encodes the DNA-binding protein CSL mediating all conventional Notch signalling [49]. It has been demonstrated that loss of CSL alters the neural differentiation of ES cells, such that neuronal differentiation is accelerated, coupled with a large reduction of the number of neural rosettes at an intermediate differentiation stage [27]. Expression of CD49e and CD117 was higher at day 8 of neural differentiation in CSL−/− differentiations compared to CSL+/− differentiations, and was indeed expressed in cells with non-neural morphology that are known to be enriched in CSL−/− differentiations. The expression of CD117 under ectoderm-promoting culture conditions suggests an epidermal origin of these cells [54, 55].

Studies using classical nitrocellulose arrays have concentrated only on PBMC surface profiles, enabling use of CD44, expressed on all lymphocytes, to control for both cell number and antibody activity between arrays. The PIII-PC arrays used in this study are suitable as a broad application technology for analysing complex cellular compositions and even uncharacterised cell types, both of which are essential for characterisation of ES cell differentiation paradigms. This broad potential makes this technology very powerful and, as demonstrated, allowed determination of novel markers for characterised and uncharacterised cell types. Inclusion of a positive control, broader then CD44 as used for lymphocyte analysis, would allow for greater precision of analysis between slides and slide batches, decreasing experimental variation. A marker expressed on all ‘possible’ cell types may be used to control for slide and experimental variation. This may take the form of ‘anti-mouse’ or ‘anti-human’ antibodies, which recognise ubiquitously expressed cell surface protein(s) in a species-dependent manner, and will be developed for future studies using PIII-PC arrays. Normalisation for experimental variation by addition of positive controls would no doubt increase the positive correlation we have detected between cell number per spot and expression intensity by flow cytometry (Fig. 4d), by increasing the quantitative power of this platform.

The orientation of antibody binding is important to ensure antigen recognition sites are not masked. Binding to both nitrocellulose and PIII-PC may occur in a random fashion (schematised in Fig. 1). This will mean that a proportion of antigen recognition sites will be unavailable for binding on both platforms. It will be interesting to pursue methods for orienting antibodies to increase PIII-PC assay sensitivity. This may be achieved by several methods including the use of polarised antibodies, for example where the antibody backbone carries a positive charge and thus preferentially orients to a negatively charged PIII-PC surface. Appropriate charge distributions may also be achieved by modifying the pH of the loading solution. It is possible that some of the 103 antibodies we have used are at least weakly polarised, which may provide a basis for lack of activity of some antibodies in our assay and/or explain why some antibodies demonstrate a stronger signal than others. A second possibility for masking of antigen expression may exist if surface antigens are unable to traverse the distance to immobilised antibody or if antigens are retained in membrane pockets such as caveolae. Our immunostaining and flow cytometry results are yet to identify either scenario.

In conclusion, the new PIII-PC array is cost- and time-efficient and offers several additional advantages over traditional nitrocellulose- and FACS-based approaches. Addition of new antibodies to our array is likely to further improve determination of cellular profiles and for discrimination of additional cell types and subpopulations. This will be particularly useful in the area of stem cell research where high throughput identification of marker profiles and standardisation of populations for pharmaceutical and personalised cell therapy applications is critical to success. As the vast majority of CD antigens were characterised from lymphocytes, we have seen a bias in data output towards the haematopoietic system. This was evident in our analysis with almost four times as many positive spots for spleen PBMCs compared to ES cells. Expansion of our antibody complement to not only the missing CD antibodies but also to other cell surface antigens will improve further our ability to define non-blood lineages. Furthermore, manipulation of the orientation status of the antibodies of the PIII surface may even further increase the sensitivity [15, 17]. Given the current pace in stem cell research, with an increasing number of protocols for cell type-specific differentiation, the access to rapid and robust technology to profile both intermediate and terminally differentiated cell types is important, and we hope that the PIII-PC array will be a valuable tool in this regard. Finally, the covalent nature of immobilisation in the PIII-PC system suggests that this technology would also be suited to high-throughput immunophenotyping in medical diagnosis [18], and the universal activity of PIII surfaces presents great potential for the application of PIII-treated substrates to small molecule-, protein- and DNA-based arrays.

Acknowledgments

This work was financially supported by Karolinska Institutet KID funding (H.M.), the Swedish Cancer Society, the Swedish Research Council (project Grant; D.B.R.M.; and the Strategic Research Initiative in Stem Cells and Regenerative Medicine), Karolinska Institutet (Distinguished Professor Award; U.L.) and Knut och Alice Wallenbergs Stiftelse (WIRM). Development of the PIII-PC microarray platform was funded by the Australian Research Council (D.M., M.B.).

Supplementary material

18_2014_1595_MOESM1_ESM.pdf (74 kb)
Supplementary material 1 (PDF 74 kb)

Copyright information

© Springer Basel 2014