The use of four-colour immunofluorescence techniques to identify mesenchymal stem cells

Abstract

In stem-cell research a major difficulty is caused by the lack of distinctive features that allow the identification of human mesenchymal stem cells (hMSC). Until now, there has been no specific marker and the most common way to identify hMSC is by their characteristic stem-cell properties: self-replication and differentiation potential. However, these findings can only be revealed retrospectively, and, once differentiated, hMSC lose their stem-cell character. The aim of this study was to establish four-colour immunofluorescence of several markers simultaneously in order to address the problem of how to identify hMSC on the single-cell level. The four markers collagen-I, collagen-IV, fibronectin and CD44 are known to be expressed by hMSC. Antibody binding was detected using secondary antibodies conjugated to FITC, Alexa546, TexasRed and AMCA. Because the distinction between Alexa546 and TexasRed was not possible on conventional digital images using standard filter sets, we performed spectral image acquisition. The image was subsequently decomposed into its pure spectral components, which permitted linear unmixing. Using this procedure we were able to demonstrate four-colour immunofluorescence on hMSC. With the possibility of using more sophisticated marker profiles and/or additional markers, four-colour immunofluorescence offers the opportunity of identifying hMSC on the single-cell level without performing differentiation assays.

Introduction

Mesenchymal stem cells are defined by their ability to self-replicate and differentiate towards different cell lineages. Characterization of human mesenchymal stem-cell (hMSC) populations has been extensively performed by flow cytometry analysis (Conget & Minguell, 1999; Pittenger et al. 1999; Majumdar et al. 2003). It has become apparent that the antigenic phenotype of hMSC is not unique and that so far no single marker is known to be specific exclusively for hMSC. In addition, it has been shown that mesenchymal stem cells represent a heterogeneous cell population consisting of several different cell types that can be distinguished by morphology and immunocytochemistry (Colter et al. 2001; Vogel et al. 2003). To understand the molecular basis of the heterogeneity in hMSC, the simultaneous detection of several antigens on a single cell is necessary. However, only two-colour immunofluorescence has been used in attempts to identify hMSC (Jiang et al. 2002). The limitations are both the uncharacteristic expression profile of hMSC (Bianco & Robey, 2001) and the optical discrimination of more than three fluorochromes with the naked eye (Liu et al. 1997). One way to overcome the latter is to use spectral image acquisition. The advantage of a spectral image is that it creates a precise database of the spectral information of each pixel of the image, which allows a demarcation of features from multipoint spectral information (Malik et al. 1996).

In order to establish a suitable approach to characterize attached hMSC at the single-cell level we established a four-colour immunofluorescence on hMSC and a spectral image analysis system.

Materials and methods

Cells

Human mesenchymal stem cells were purchased from Cambrex (USA). The cells were cultivated according to the supplier's recommendations in hMSC-growth medium (Cambrex, USA) containing mesenchymal cell growth supplement, L-glutamin and antibiotics (penicillin, streptomycin) but no stimulatory supplements or vitamins. Cell culture was performed in T75 flasks (Nunc, USA) in a humidified incubator at 37 °C, using a standard mixture of 95% air and 5% CO₂.

Immunofluorescence

For immunofluorescence, cells were seeded on uncoated glass slides before the 7th passage at approximately 5000 cells cm². After 24 h, cells were fixed at −20 °C in cold methanol for 8 min and subsequently washed in phosphate-buffered saline (PBS). No enzyme treatment was performed. Non-specific binding of the secondary antibody was reduced with an appropriate serum block. Primary antibodies were raised in different species against fibronectin (mouse), collagen-I (rabbit), collagen-IV (goat) and CD44 (rat) (Table 1). All secondary antibodies were raised in donkey to allow a simultaneous incubation step; each was labelled with a different fluorochrome (Texas Red, FITC, Alexa546 and AMCA, respectively, Table 1). Non-specific binding of secondary antibodies was controlled by omitting the primary antibody. As an additional control, an identical cell population on the same slide was also labelled separately with each primary antibody at the same dilution. Nuclear counterstaining with DAPI at a dilution of 1 : 5000 generated a fifth fluorescent spectrum. All slides were mounted with a polymerizing hydrophilic mounting medium containing an anti-fade reagent (Molecular Probes, USA).

Table 1. Primary antibodies and fluorochrome-conjugated secondary antibodies used for immunofluorescence; dilutions, antibody hosts and sources.

Host and antigen	Antibody	Dilution	Source
Mouse anti-human fibronectin	HFN7.1	1 : 5	DSHB
Rabbit anti-human collagen-I	AB745	1 : 5	Chemicon
Rat anti-human CD44	Hermes-1	1 : 5	DSHB
Goat anti-human collagen-IV	YMPS063	1 : 5	Accurate Chemical
TexasRed-conjugated donkey anti-mouse IgG	715-075-151	1 : 250	Dianova
Fluorescin (FITC)-conjugated donkey anti-rabbit IgG	711-095-152	1 : 250	Dianova
AMCA-conjugated donkey anti-rat IgG	712-155-153	1 : 25	Dianova
Alexa Fluor546-labelled donkey anti-goat IgG	A-11056	1 : 125	Molecular Probes

DSHB, Developmental Studies Hybridoma Bank.

Data acquisition

The fluorescent spectra were acquired using a Sagnac-type interferometer SpectraCube SD-200 (Applied Spectral Imaging (ASI), Israel) installed on an Axioskop 2 microscope (Zeiss, Germany) attached to a charge-coupled device (CCD) camera (Hamamatsu CCD 5880-C, Japan) and a personal computer. Two different triple-bandpass filter sets for red, green and blue spectra (F61 002, AHF, Germany), and for green, red and infrared spectra (SKY, ASI, Israel) as well as appropriate standard filter sets for red, green and blue spectra (#15, #10 and #01, Zeiss, Germany) were used. Details of the principle for data acquisition have been published elsewhere (Malik et al. 1996; Rothmann et al. 1998). Briefly, the fluorescent spectra from the cells passing through the barrier filter is split in the interferometer in opposite directions and is recombined by reflecting mirrors at the exit with an optical path difference (OPD). The OPD is dependent on the angle between the incoming beam and the interferometer itself. The OPD arises from the different optical paths travelled by the two beams in the beam-splitter for non-zero angles. The high mechanical stability of the interferometer allows a Fourier technique to be used to analyse the visible spectral region. This is achieved by synchronizing the recording of successive CCD frames with the steps of the motor used to rotate the collimated beam, so that the instantaneous OPD is known for every pixel in every recorded frame and can be used in the fast Fourier transform calculation. This technique enables demarcation of wavelength ranges of 10 nm and less (Malik et al. 1996; Rothmann et al. 1998). For image analysis, SpectraView Software (ASI, Israel) enabled linear unmixing based on decomposition of the image in its pure spectral components. The reference spectra used to analyse the multicolour image were derived from the single colour labelling of each antibody on the same slide as described above. As controls, all microscope images acquired with the spectral interferometer were additionally acquired with a conventional digital camera (Cybershot DSC S 75, Sony, Japan).

Results

Human mesenchymal stem cells revealed heterogeneous morphology and heterogeneous protein expression profiles for the investigated proteins collagen-I, collagen-IV, fibronectin and CD44. Digital image acquisition using a triple band filter for red, green and blue spectra (F61002, AHF) allowed simultaneous visualization of all four spectra of FITC, Alexa546, TexasRed and AMCA, respectively. Labelling patterns of the single antigens could only be discriminated in regions where fluorescent signals did not overlap. No differentiation was possible in regions where, in particular, green (FITC), orange (Alexa546) and red (TexasRed) spectra were located, i.e. in the perinuclear region of hMSC (Fig. 1a). Using different standard filter sets for blue (#01, Zeiss), green (#10, Zeiss) and red spectra (#15, Zeiss), we could discriminate between AMCA (CD44), FITC (collagen-I) and Alexa546 (TexasRed), respectively. CD44 revealed homogeneous labelling of the cell with accentuation of the cell membrane. DAPI and AMCA spectra could not be separated (Fig. 1c). Collagen-I was found in the cell with a granular labelling pattern as well as in the extracellular matrix (Fig. 1e). Spectra for Alexa546 (collagen-IV) and TexasRed (fibronectin) were not distinguishable with standard filter sets and digital image acquisition (Fig. 1d). Resolution of these two spectra was only possible in the spectral image by linear unmixing (Fig. 1f,g). Fibronectin revealed granular intracellular labelling and a fibrous extracellular labelling. Collagen-IV showed a granular intracellular pattern and different cells labelled inhomogeneously. Labelling patterns of all investigated proteins were identical to the pattern for each antibody in single-colour fluorescence. For Alexa546 we found a shift of the spectra to longer wavelength regions using the triple-band filter F61002. The difference between the peaks of Alexa546 and TexasRed was very minor (Fig. 2a,b). Using the triple-band filter for green, red and infrared spectra (SKY, ASI) the wavelength shift of Alexa546 was smaller and the peak could be separated from TexasRed more clearly (data not shown). Linear unmixing for FITC spectra (collagen-I) showed the same results as with digital image acquisition (Fig. 1h). DAPI and AMCA spectra could not be resolved using the triple-band filter F61002 (Figs 1i and 2a,b). Decomposition into different blue spectra was achieved by spectral image acquisition using a long-pass emission filter (#01) (Fig. 2c,d). Although fluorescent signals for CD44/AMCA were located mainly in the cell membrane, we also acquired signals in the nuclear region (Fig. 1k). DAPI spectra were exclusively found in the nucleus but not in the cell membrane (Fig. 1l).

Fig. 1.

Four-colour immunofluorescence plus DAPI nuclear staining of two single human mesenchymal stem cells (hMSC) with different morphology, with digital image acquisition (a, c–e) and with spectral image acquisition (b, f–l). (a) Digital image of a four-colour immunofluorescence using a red, green and blue triple-band filter set F61002 (AHF) of collagen-I (FITC, green spectra), collagen-IV (Alexa546, orange spectra), fibronectin (TexasRed, red spectra) and CD 44 (AMCA, blue spectra). (b) Spectral image with the same triple-band filter set (F61002, AHF). Although each pixel has been decomposed into its spectral components in the superimposed figure, the individual spectra could not be resolved. (c) Digital image of CD44 (AMCA) and DAPI spectra with a standard filter set for blue spectra (#01, Zeiss). (d) Digital image of TexasRed (fibronectin) and Alexa546 (collagen-IV) spectra with a standard filter set for red spectra (#15, Zeiss). Resolution of TexasRed and Alexa546 was not possible. (e) Digital image of collagen-I (FITC) with a standard filter set for green spectra collagen-I/FITC (#10, Zeiss). (f–i) Spectral images with a triple-band filter for red, green and blue spectra (F61002, AHF) after linear unmixing for (f) Alexa546 (collagen-IV, pseudo-coloured yellow), (g) TexasRed (fibronectin), (h) FITC (collagen-I) and (i) AMCA (CD44)/DAPI, respectively. Note that the spindle-shaped cell reveals a more intense labelling for collagen-IV compared with the spread cell. (i) Using the triple-band filter F61002, AHF, DAPI and AMCA spectra could not be decomposed. (k,l) Linear unmixing for (k) AMCA and (l) DAPI was possible with a long-pass emission filter (#01, Zeiss). Note that DAPI was only found in the nucleus. The stripes in figures f, g and k are due to spectral image acquisition of fluorescent signals with high-intensity differences. All scale bars = 50 µm.

Fig. 2.

(a,c) Details of the filter sets used for spectral image acquisition with excitation, beamsplitter and emission spectra, and (d,b) the corresponding applied spectra analysed with SpectraView software. (a,b) Spectral decomposition with the triple-band filter set for red, green and blue spectra (F61002, AHF) of Alexa546 and TexasRed was possible, but only small peak shifts were found. Discrimination of DAPI and AMCA was not possible. Linear unmixing for AMCA spectra resulted in background signal. (c,d) Linear unmixing for DAPI and AMCA spectra was possible using a long-pass emission filter set (#01, Zeiss).

Discussion

The aim of the present study was to establish a four-colour immunofluorescence of characteristic markers of hMSC at the single-cell level.

hMSC represent a heterogeneous cell population containing several subpopulations that can be distinguished morphologically as well as immunocytochemically (Colter et al. 2001; Vogel et al. 2003). Whether these subpopulations consist of different differentiation stages of hMSC or different cell types is still unclear. A major problem in the approach to characterizing these cells is that to date no single specific marker exists that exclusively allows the identification of hMSC. To understand more about the basis of this heterogeneity, a suitable approach is the simultaneous detection of several characteristic markers on the single cell. Simultaneous detection of more than ten different antigens by flow cytometry has been established for cells in suspension, especially for lymphocytes – revealing a characteristic marker profile (De Rosa et al. 2003). Because flow cytometry is a more sensitive method than conventional immunofluorescence on attached cells or tissue (Zola et al. 1990), the number of antigens that can be detected by immunocytochemistry is consequently lower. For cryosections of mouse lymphatic tissue, simultaneous detection of seven different antigens by immunofluorescence has been reported (Tsurui et al. 2000). For hMSC, however, only a two-colour immunofluorescence has been reported to date (Jiang et al. 2002). This is because of various limitations. The most apparent is based on the expression properties of hMSC. In contrast to lymphatic tissue or leucocytes, the antigenic phenotype of hMSC is not unique and lacks prominent markers (Conget & Minguell, 1999; Bianco & Robey, 2001).The four markers used in this study are not specific for hMSC alone, but each marker contributes to the expression profile of hMSC. Collagen-I and CD44 are very early markers that are expressed as soon as hMSC attach to a surface (Zohar et al. 1997; Shur et al. 2002; Wexler et al. 2003). Collagen-IV is only present in hMSC and very early progenitors but not in older differentiation stages of hMSC (Chichester et al. 1993; Deschaseaux & Charbord, 2000). Fibronectin is a matrix protein that is strongly expressed by hMSC (Vogel et al. 2003). Although each single marker is expressed by other cell types as well, the expression of all four markers on an individual cell represents the typical labelling profile of hMSC. In particular, the heterogeneous labelling results for collagen-IV in confluent hMSC cultures indicate that only a part of the cell population reveals the labelling profile of hMSC or early progenitors. A further specification of the marker profile together with simultaneous detection of additional markers will be useful to discriminate more precisely between the various cellular (sub) populations.

We performed indirect multicolour immunofluorescence using fluorochrome-conjugated secondary antibodies, because the binding step of the secondary antibody leads to an amplification and a more intense fluorescent signal (Lamvik et al. 2001). Particularly for the detection of low levels of surface antigens such as CD44 in the present study, signal amplification by indirect immunofluorescence was necessary. With this technique, however, the availability of characteristic primary antibodies from different species is restricted. In a different study, seven-colour immunofluorescence was performed with fluorochrome-labelled primary antibodies for five out of seven antibodies (Tsurui et al. 2000). This technique allows the application of primary antibodies raised in the same species, and thus the choice of antibodies is almost unlimited. The disadvantage is that antigens are required that are abundantly expressed, because direct labelling results in a much weaker fluorescent signal intensity compared with indirect labelling with fluorochrome-conjugated secondary antibodies (Lamvik et al. 2001).

Another limitation of multicolour immunofluorescence is the optical resolution of fluorescent spectra of more than three colours (two antibodies plus nuclear stain). Our approach to this problem was to use an interferometer that can distinguish wavelength ranges of less than 10 nm in combination with a triple-bandpass filter. This technique makes it possible to distinguish blue, green, orange and red fluorescent spectra in one single spectral image. Because the image acquisition time of spectral images varies between 2 and 3 min, data acquisition from a single image protects the samples against photo bleaching. Although photo bleaching was not quantified in this study, it did not disturb several image acquisitions from the same microscopic field of view. However, from the spectral image acquired with the triple-band filter it was not possible to distinguish between the two different spectra of DAPI and AMCA blue. For this purpose, it was necessary to acquire an additional spectral image using a long-pass emission filter (#01, Zeiss). Although it is postulated with regard to pixel accuracy and noise that spectral image acquisition is superior to image acquisition using different broad-band filter sets only (Garini et al. 1999), based on our data we suggest the combination of appropriate filter sets together with spectral image acquisition for achieving optimal results. The spectra of DAPI, FITC and TexasRed were close to the reference spectra given by the supplier. For AMCA, and especially for Alexa546, we found spectra that were shifted to longer wavelength regions (to the right in the diagrams). This could be due to the blocking of emission signals in shorter wavelength regions by the triple-band filter for red, green and blue (Tsurui et al. 2000). The shift for Alexa546 was smaller but still apparent using a triple-band for green, red and infrared. Another reason for the shift in the fluorescent spectra of Alexa546 might be interference with the other fluorescent dyes during simultaneous incubation of the secondary antibodies (Tsurui et al. 2000). Interference in the sense of blocking of one antibody by another cannot be excluded. This hypothesis is supported by the fact that in the single-colour immunofluorescence the fluorescent signal appeared to be slightly brighter compared with the multicolour application. However, using spectral analysis we did not find co-localization of the different antigens and the respective labelling patterns in multicolour and single-colour fluorescence were identical, indicating that these effects did not disturb the multicolour immunofluorescence.

In conclusion, we present four-colour immunofluorescence on single hMSC with a representative labelling profile. The combination of appropriate fluorescent filter sets and a spectral camera makes it possible to discriminate wavelength ranges of less than 20 nm. With this method, hMSC are shown to be a heterogeneous cell population especially with respect to collagen-IV expression. Expanding this labelling profile with other suitable markers at single-cell level can provide a suitable approach to identify hMSC without the need to perform differentiation assays.