SP定量观察法

q 2007International Society for Analytical Cytology Cytometry Part A 71A:251–257

(2007)

Combination of Quantiﬁcationand Observation Methods for Study of ‘‘SidePopulation’’Cells

in Their ‘‘InVitro’’Microenvironment

Rachid Benchaouir, 1 Julien Picot, 1 Nicolas Greppo, 1Philippe Rameau, 1Daniel Stockholm, 1

1*Luis Garcia, 1Andras Paldi, 1,2and Corinne Laplace-Builh e

GENETHON ÀCentre National de la Recherche ScientiﬁqueUMR 8115, 1bis, rue de l’Internationale91002Evry, France

Ecole Pratique des Hautes Etudes, 46rue de Lille, 75007Paris, France

Received 15June 2006; Revision received 27October 2006; Accepted 28November 2006

Background:Qualitative and quantitative analyses of the rare phenotypic variants in in vitro culture systems is necessary for the understanding of cell differentiation in cell culture of primary cells or cell lines. Slide-based cyto-metry combines image acquisition and data treatment, and associates the power of ﬂowcytometry (FCM)and the resolution of the microscopic studies making it suita-ble for the analysis of cells with rare phenotype. In this paper we develop a method that applies these principles to a particularly hot problem in cell biology, the study of stem cell like cells in cultures of primary cells, cancer cells, and various cell lines.

Methods:The adherent cells were labeled by the ﬂuores-cent dye Hoechst 33342. The images of cell populations were collected by a two-photon microscope and pro-cessed by a software developed by us. The software allows the automated segmentation of the nuclei in a very dense cell environment, the measurement of the ﬂuorescencein-tensity of each nucleus and the recording of their position

in the plate. The cells with a given ﬂuorescenceintensity can then be located easily on the recorded image of the culture plate for further analysis.

Results:The potential of our method is illustrated by the identiﬁcationand localization of SP cells in the cultures of the C2C12cell line. Although these cells represent only about 1%of the total population as calculated by ﬂowcytometry, they can be identiﬁedin the culture plate with high precision by microscopy.

Conclusion:Cells with the rare stem-cell like phenotype can be efﬁcientlyidentiﬁedin the undisturbed cultures. Since the ﬂuorescenceintensity of rare events and the position of thousands of surrounding cells are recorded at the same time, the method associates the advantage of the FCM analysis and the microscopic observation. q 2007

International Society for Analytical Cytology

Key terms:image processing; rare phenotype; confocal microscopy; stem cells; myogenic

The understanding of cell differentiation and development of complex biological systems is largely facilitated by the use of in vitro cell culture systems of primary cells or cell lines capable of undergoing differentiation under artiﬁcialcondi-tions. The qualitative and quantitative analysis of the rare phenotypic variants in such systems is a major challenge. On one hand, the study of cellular patterns and cell to cell interactions requires morphological, typical microscopic analysis of the undisturbed cell culture. On the other hand, quantitative analysis of various phenotypic characteristics of a large number of individual cells typically requires invasive methods, such as ﬂowcytometry (FCM),that disrupt the spatial patterns of the population. Especially, collecting quantitative measures on cells with rare phenotype repre-senting only a few percent of the whole population requires the analysis of a large number of cells. Slide-based or micros-copy-based cytometric methods associate the advantage of the FCM analysis and the microscopic observation (1,2).Technically, this approach requires acquisition, treatment, and analysis of images of cell cultures. Since individualization

of nuclei or cells in aggregates or complex tissue organiza-tion is a prerequisite for subsequent image analysis, the seg-mentation method is one of the critical steps of the image treatment. Previous efforts made available efﬁcientimage segmentation methods for cell nuclei and whole cells into two and even three dimensional biological objects. Most of them were obtained by combining image treatment with confocal microscopy technology (3À8).

Automated acquisition and segmentation of the images, measurement of the ﬂuorescenceintensity of each cell or cell nucleus, and the recording of their position in the plate made possible to associate the resolution of micro-These authors contributed equally to this work. *Correspondenceto:Corinne Laplace-Builh e , GENETHON ÀCentre National de la Recherche ScientiﬁqueUMR 8115, 1bis, rue de l’Inter-nationale 91002Evry, France.

E-mail:[email protected]

Published online 5February 2007in Wiley InterScience (www.interscience.wiley.com).

DOI:10.1002/cyto.a.20367

252

BENCHAOUIR ET AL.

scopic analysis with the potential of FCM. In the present paper, we have adapted the principles of the microscopy-based cytometry to a speciﬁcbiological problem, the identiﬁcationof rare cells with stem-cell like phenotype in culture. Pluripotential cells can be identiﬁedin many cell lines, primary cell cultures established directly from tissues or even from tumors on the basis of their high ac-tivity of the ATP-binding cassette (ABC)transporters, which allows the efﬁcientexclusion of the ﬂuorescentdye Hoechst 33342(9,10).As a result of their high dye exclusion capacity, the ﬂuorescenceintensity of the nuclei in these cells is lower than that in the nuclei of the popula-tion. This property allows the identiﬁcationand isolation of these so called ‘‘SidePopulation’’(SPcells) versus the ‘‘MainPopulation’’(MP)(10).Whereas, SP cells are routi-nely isolated for molecular analysis from a cell suspension using a cell sorter, their in situ examination in culture could help to elucidate what interactions with their neigh-bors they have. The aim of our work was to develop a method that allows the rapid and reliable identiﬁcationof stem-cell like SP cells in culture. The method was set up and validated on cultures of the myogenic C2C12cell line as a model system. The populations of the C2C12cells contain a small subpopulation of cells with stem cell-like phenotype (11),which represent only 1À2%of the whole culture. By using the principles of slide-based cytometry our method overcomes the difﬁcultiesassociated with the standardization of Hoechst labeling in adherent cells, the precise measurement of the ﬂuorescenceintensities and the identiﬁcationand localization of the SP cells in culture.

MATERIALS AND METHODS

Cell Culture

C2C12, a sub clone derived from the original C2mouse myoblast cell line was obtained from the American Type Culture Collection (CRL1772). C2C12cells were routinely propagated in proliferation Dulbecco’sModiﬁedEagle’sMedium (DMEM,Gibco BRL, Gaithersburg, MD) with 4.5g/mlof glucose, supplemented with 20%(v/v)foetal calf serum (FCS,Hyclone), 100U/mlpenicillin, and 100l g/mlstreptomycin. The cultures were performed at 37°C under a humidiﬁedatmosphere of air with 7%CO 2. Initial plating density was between 23103and 33103cell/cm2and the cells were cultured for 6days. Thirty-ﬁvemillimeter glass bottom culture dishes (MatTek Corporation, Ash-land, MA) were used for cell culture and confocal analyses.

Hoechst 33342Staining and Cell Sorting This protocol was adapted from the staining of haemato-poietic originating cell populations (9).C2C12cells were trypsinized (GibcoBRL Àbatch 3101673) and suspended at a working concentration of 106cells/mlin ice-cold PBS con-taining 2%FCS, or kept adherent in tissue culture dishes in their proliferation medium. In each case, the DNA dye Hoechst 33342(Sigma-AldrichB-2261, batch 123K4080, St. Louis, MO) was added to a ﬁnalconcentration of 11l g/mland cells were incubated 90min at 37°C under mild shak-ing. The controls were incubated in the presence of 100l M

of verapamil (Sigma-Aldrich,V-106batch 28H4699). After staining, adherent cells were trypsinized. Both cultures were spun down and cell suspensions were suspended in fresh 2%FCS ÀPBS with 2l g/mlpropidium iodide (PI,Sigma) for the partition of dead vs. living cells. Samples were kept on ice until ﬂowcytometry (FCM)analysis.

Analysis and sorting were performed on a dual-laser MoFlo ﬂowcytometer (Dako,Glostrup, Denmark). An ar-gon laser tuned to 350À360nm (100mW) was used to excite the Hoechst dye. Fluorescence emission was col-lected with a 450/65band pass (BP)ﬁlterfor the Hoechst ÔDMLP blue Õand was a used 570/40to separate BP ﬁlterthe for the emission Hoechst wavelengths. Ôred Õ. A 510A second 488nm argon laser (100mW) was used to excite PI ﬂuorescence.Fluorescence emission was then meas-ured with a 630/30BP ﬁlter.PI-positive dead cells were excluded from the analysis. Results were analyzed using the Summit v3.1software (Cytomation,Dako, Glostrup, Denmark). The Ôblue Õemission 5f(Ôred Õemission) diagram revealed a side population (SP)displaying low staining with Hoechst and a main population (MP)of cells that were more brightly stained.

Microscopy and Image Processing

The two-photon confocal microscope consisted of a Radiance 2100MP scan head (Bio-Rad,Hercules, CA) equipped with a mode-locked Titanium ÀSapphire laser system (CoherentVerdi-Mira) pumped at 5W and tuned to a 800nm excitation with 100fs pulses at 76MHz. On-line checking of the excitation parameters was done through the use of beam conditioning and Pockel Cell units (Bio-Rad) that enabled pulse length to be controlled and output power levels to be attenuated below a range so that no damage can be observed on the tissue (typically10À46mW depending on the objective lens). Power measure-ments exiting the microscope objective were made with a Coherent LaserMate model 33-0191power meter. Two-photon ﬂuorescencewas detected by a nondescanned de-tector (Bio-Rad)after passing optical ﬁlters.The Dual-emis-sion of Hoescht dye used a dichroic mirror 500DCLPXR (Bio-Rad)and two band pass emission ﬁlters:HQ450/80for the blue emission and 575D150for the red emission (FRETﬁlter;Omega Optical, Brattleboro, VT). The detector gain settings were ﬁxedto 50%for each detector and the laser power with a 103objective was set to 10mW . The micro-scope was an inverted Nikon TE300(NikonInstech Co., Kanagawa, Japan) with Nikon objectives, dry CFI Plan APO, 103, NA0.50. The acquired images were analyzed by our own software written in C language.

Two images are collected simultaneously by a two photon confocal microscope. They correspond respectively to the blue and red ﬂuorescenceemitted by the Hoechst located in the nucleus of the C2C12cells (Figs.2A and 2C). The ﬁrstimage process used is a basic threshold (Figs.2B and 2D), which leaves a noise characterized by clusters of a few pix-els. This noise is removed by means of a median ﬁlterwith a kernel of 333pixels (Figs.2E and 2G). The nuclei in contact of the frame borders on the ﬁlteredimage were removed (Figs.2F and 2H). An algorithm settled by Seraﬁni

Cytometry Part A DOI 10.1002/cyto.a

COMBINATION OF METHODS FOR STUDY OF SIDE POPULATION CELLS IN THEIR MICROENVIRONMENT

253

F IG . 1. Hoechst-proﬁleof the C2C12cells labeled in suspension (A , B ) or attached to the culture dish (C , D ). Both labeling methods give identical proﬁlesat 11l g/mlof Hoechst 33342concentration. When the cells are treated with verapamil (Band D), the ﬂuorescenceof the SP cell fraction is shifted to a higher level that speci-ﬁesthe Mdr activity of the SP population.

is then used to attribute a label to each cell of the treated bi-nary image (12).The user set a minimal size so that the objects smaller to this size are recognized, selected, and eliminated (Figs.2G and 2H). A distance map is then obtained as described by Shih and Wu (13).Each white pixel is replaced by a grey pixel, where value represents the square of the Euclidian distance from the nearest black pixel (Figs.2I and 2K). This distance map emphasizes the maxima regions:pixels within the cells that are the far form the borders (notshown). These maxima are used as labels for the segmentation. We used the unbiased watershed algo-rithm with the ‘‘waitingqueue’’data structure developed by Meyer (14)and improved by Beucher (15)(http://cmm.ensmp.fr/$beucher/publi/LPE_sans_biais_V2.pdf).It enables a correct individualization of the cellular nuclei by tracing a watershed line at the border between the con-nected nuclei (Figs.2J and 2L). Finally, to prevent over-seg-mentation, the size of each segmented object is calculated and the objects with a size smaller to the user-deﬁnedsize are considered as being a watershed line.

RESULTS

SP/MPProﬁleof C2C12on Adherent Cells First, we have set up the conditions for the labeling of ad-herent C2C12cells with the Hoechst 33342dye so that ﬂuo-rescence proﬁlesgiven by ﬂowcytometry are equivalent to those obtained when cells are stained in suspension. To avoid the misidentiﬁcationof the SP cells (16),the optimal Hoechst staining conditions were determined by using differ-ent concentrations of the dye. A Hoechst concentration of 11l g/mlfor 90min, was found to be optimal for both adher-Cytometry Part A DOI 10.1002/cyto.a

ent cells and cells in suspension. Controls incubated with verapamil, an inhibitor of the Mdr1activity, were also ana-lyzed to check the ‘‘speciﬁcity’’of the SP population.

The Figure 1shows the proﬁlesobtained by ﬂowcyto-metry of cells labeled under ‘‘insitu’’or ‘‘insuspension’’conditions. There were no signiﬁcantdifferences between the SP/MPproﬁlesobtained in the two conditions (Figs.1A and 1C). When the exclusion of the dye is inhibited by ve-rapamil, the SP cells cannot be discriminated from the MP cells (Figs.1B and 1D). This result indicates that the capa-city of the SP fraction of C2C12cells to exclude the Hoechst 33342is not affected by their adhesion to the sub-strate of the culture plate.

Image Acquisition

The next step was to adapt the method for the SP cell identiﬁcationbased on the measurement of the ﬂuorescenceintensity of the cell nuclei by the use of a non invasive ima-ging system. We chose two-photon confocal microscopy , because it allows quantiﬁcationof the ﬂuorescenceunder conditions of Hoechst excitation similar to that of our ﬂowcytometer. Moreover, two-photon imaging reduces substan-tial damage in the living cells (17).Images of cell nuclei col-lected with a two-photon confocal microscope were then processed. As the confocal instrument produces optical sec-tions, we checked that the Z position did not affect the rela-tive ﬂuorescenceintensity of the nuclei.

Image Processing

The different steps of the image processing are shown on Figure 2. The segmentation algorithm described

here

254

BENCHAOUIR ET AL.

F IG . 2. Major steps of the image treatment. The images of the red and blue ﬂuorescentcell nuclei are acquired on a two-photon confocal microscope (A ). After binarization and threshold selection (B and E ) the background noise is removed from the image by a median ﬁlter(E).Cells in contact with the border are removed by conditional dilatation (F ). A distance map is constructed by replacement of the white pixels by a grey (I ) leaving to the individualization of the cell nuclei (J).A zoom of the same part of the global image is repre-sented on the side (C , D , G , H , K , and L of the A, B, E, F , I, and J panels, respectively). [Colorﬁgurecan be viewed in the online issue, which is avail-able at www.interscience.wiley.com.]

allowed us to individualize the majority of the nuclei in a dense cell culture without over-segmentation and elimi-nate artifacts corresponding to the background noise on the original image. The efﬁciencyof the algorithm was evaluated by manual counting of the over-and under-seg-mented nuclei of an image containing $1,000nuclei. The number of successfully individualized cell nuclei was 90%with our algorithm at the threshold of 33/255of grey level and 94%with a threshold of 50/255.Both threshold levels were far below the minimal ﬂuorescentintensity of the nuclei.

Validation of the Method on Sorted Populations To validate the method, we isolated the subpopulations of the SP and MP cells by cell sorting, then they were put in a culture dish and maintained at 4°C. The two subpo-

pulations were imaged separately. The images were pro-cessed as described. The results show that the SP cell nuclei can be differentiated from those of the MP cells (Figs.3A and 3B). Therefore, our method that associates a speciﬁcstaining procedure, standardized image acquisi-tions and a new software development allows us to detect weak differences between SP and MP cells as by FCM.

Identiﬁcationand localization of SP Cells in an

In Vitro Growing C2C12Cell Culture We have evaluated the capacity of our system to detect SP cells in a growing population of C2C12cells. The low incidence of SP cells expected in a growing cell popula-tion required the acquisition of a large number of images for collecting the information from thousands of

cell

Cytometry Part A DOI 10.1002/cyto.a

COMBINATION OF METHODS FOR STUDY OF SIDE POPULATION CELLS IN THEIR MICROENVIRONMENT

255

F IG . 3. SP/MPproﬁlereconstructed from segmented images. Sorted SP and MP cells are plated at equal den-sity. The SP nuclei (A ) are less ﬂuo-rescent than MP nuclei (B ) as also shown on the red and blue ﬂuores-cence intensity plots (C and D , respectively). Some intensely ﬂuores-cent cells in the SP preparation are likely to be MP contaminants. E is the superposition of C and

nuclei. We have acquired series of up to 30images of 800À1,200nuclei of each cell culture and analyzed them with our method. The comparison of the plots (Fig.4) of ﬂuorescencevalues from the normal and verapamil-trea-ted cell populations revealed that a fraction of cells had low ﬂuorescencedue to the high Hoechst 33342exclu-sion activity inhibited by verapamil. The proportion of these cells—onaverage of 1%—isnot different from those usually obtained by ﬂowcytometry for the SP cells by similar criteria. We consider therefore the cells with low ﬂuorescencevalues as bona ﬁdeSP cells.

Since the graphical coordinates of each cell are recorded we could localize the rare SP cells dispersed within the culture (Fig.5). Therefore, it is possible to obtain quantitative information on a large number of cells as by ﬂowcytometry, but also to localize the cells in the culture and make direct observations in their in vitro environment by microscopy.

DISCUSSION

The method reported here follows the principles of the slide-based cytometry by joining the advantage of the ﬂowcytometry to obtain quantitative information on a large number of individual cells with the morphological analysis of the cell cultures by microscopy. Our method represents a speciﬁcapplication of these principles for the identiﬁcationvia Hoechst efﬂuxof putative stem cells in cultures of adherent cells of a model cell line. This approach required the adaptation of the cell labeling method to adherent cells and the development of efﬁ-cient image acquisition and processing methods speciﬁcto the biological material investigated. The application of our method to other culture systems—asprimary cell populations containing stem cells or neoplastic cell lines with cancer stem cells—mayhelp elucidate the nature of interactions the putative stem cells have with the sur-rounding other cells in the

culture.

F IG . 4. The SP/MPproﬁlof a typical C2C12population. Six day cell cultures untreated (A ) or treated (B ) with verapamil were labeled as above. The aver-age ﬂuorescenceintensity of the nuclei treated by verapamil is higher. The superposition of the two plots (C ) reveals the SP cell fraction in the untreated cul-ture (A).The plot is similar of that obtained by FCM.

Cytometry Part A DOI 10.1002/cyto.a

256

BENCHAOUIR ET AL.

F IG . 5. Localization of the SP cells in vitro. Series of images treated by the segmentation software (A ) or in phase contrast (B ). The weaker ﬂuorescentintensity of the SP nuclei is evident on the high power pictures shown on C and D with the SP nuclei indicated by an arrow. Low power bright ﬁeldﬂuorescent(E , F ) and phase contrast (G and H ) images with the SP cells (inyellow, indicated by an arrow).

Cytometry Part A DOI 10.1002/cyto.a

COMBINATION OF METHODS FOR STUDY OF SIDE POPULATION CELLS IN THEIR MICROENVIRONMENT

257

The precise localization of the SP cells in a cell culture is for a great part due to the efﬁciencyof the computer tool to treat dozens of thousands of events. With our algo-rithm we could achieve segmentation of the original images, individualizing each cell nucleus, even those from complex cell clusters. The successful reproduction of the SP/MPproﬁlessimilar to those typically obtained by ﬂowcytometry provides a proof of concept of our method. The software developed here is fully automated and does not require user initiation except for setting up the back-ground noise level, which can vary according to the condi-tion of staining or acquisition, as well as the determination of the minimal size of the events to be considered. This setting is greatly facilitated by a real time viewer that dis-plays new images when changes are operated in threshold or minimum size levels.

The time required to process an image of 1,02431,024pixels is about 20min, when using a ‘‘meanPc conﬁgura-tion’’,and allows a large number of nuclei to be examined and quantiﬁedso that statistical accuracy can be achieved in a reasonable period of time.

The capacity to identify rare cellular phenotypes by mi-croscopy opens the way toward the analysis of rare bio-logical events as interactions in the microenvironment of the stem cell-like cells in heterogeneous cultures of large cell populations of primary cells isolated from normal or neoplastic tissues without the disruption of the popula-tion morphology and intercellular interactions.

Future extensions of the method will include the use of ﬂuorescentlylabeled markers, as antibodies or direct ﬂuorescentlabels of various cellular phenotypes. Quanti-tative data can ﬁrstbe obtained from the recorded images with the same statistical robustness as usually provided by ﬂowcytometry analyses. Then, the cells with rare phe-notypes can be studied by microscopy in their original microenvironment in the cell culture. Our method also opens the way to the time-lapse analysis of cells with rare phenotypes. Such a method will largely contribute to the understanding of dynamic phenomena like cell differen-tiation or emergence of heterogeneity in cell cultures.

Cytometry Part A DOI 10.1002/cyto.a

LITERATURE CITED

1. Ta

`rnok A. Slide-based cytometry for cytomics—Aminireview. Cyto-metry A 2007;69:555À562.

2. Ecker C, Steiner G. Microscopy-based multicolor tissue cytometry at the single-cell level. Cytometry A 2004;59:182À190.

3. Baggett D, Nakaya MA, McAuliffe M, Yamaguchi TP , Lockett S. Whole cell segmentation in solid tissue sections. Cytometry A 2005;67:137À143.

4. Krtolica A, Ortiz de Solorzano C, Lockett S, Campisi J. Quantiﬁcationof epithelial cells in coculture with ﬁbroblastsby ﬂuorescenceimage analysis. Cytometry 2002;49:73À82.

5. Litle VR, Lockett SJ, Pallavicini MG. Genotype/phenotypeanalyses of low frequency tumor cells using computerize image microscopy. Cytometry 1996;23:344À349.

6. Lockett SJ, Herman B. Automatic detection of clustered, ﬂuorescent-stained nuclei by digital image-based cytometry. Cytometry 1994;17:1À12.

7. Lockett SJ, Sudar D, Thompson CT, Pinkel D, Gray JW . Efﬁcient,inter-active, and three-dimensional segmentation of cell nuclei in thick tis-sue sections. Cytometry 1998;31:275À286.

8. Wahlby C, Sintorn IM, Erlandsson F , Borgefors G, Bengtsson E. Com-bining intensity, edge and shape information for 2D and 3D segmen-tation of cell nuclei in tissue sections. J Microsc 2004;215(Part1):67À76.

9. Goodell MA, Brose K, Paradis G, Conner AS, Mulligan RC. Isolation and functional properties of murine hematopoietic stem cells that are replicating in vivo. J Exp Med 1996;183:1797À1806.

10. Zhou S, Schuetz J, Bunting K, Colapietro A, Sampath J, Morris J,

Lagutina I, Grosveld G, Osawa M, Nakauchi H, Sorrentino BP . The ABC transporter Bcrp1/ABCG2is expressed in a wide variety of stem cells and is a molecular determinant of the side-population pheno-type. Nat Med 2001;7:1028À1034.

11. Benchaouir R, Rameau P , Decraene C, Dreyfus P , Israeli D, Pietu G,

Danos O, Garcia L. Evidence for a resident subset of cells with SP phenotype in the C2C12myogenic line:a tool to explore muscle stem cell biology. Exp Cell Res 2004;294:254À268.

12. SeraﬁniN.Implantation d’unelibrairie de morphologie math e matique

sur un processeur DSP . Thesis work, Ecole d’ing e nieurs de Gene

`ve. HES 2001:110.

13. Shih FY , Wu Y -T. Fast Euclidean distance transformation using a 333

neighbourhood. Comput Vis Image Underst 2003;93:195À205.

14. Meyer F . Topographic distance and watershed lines. Signal Process-ing. 1994;38:113À125.

15. Beucher S.Algorithmes sans biais de ligne de partage des eaux. Cen-tre de Morphologie Math e matique, Thesis work, Ecole des Mines de Paris 2004. 35. Available at http://cmm.ensmp.fr/$beucher/publi/LPE_sans_biais_V2.pdf.

16. Petersen T, Ibrahim S, Diercks A, van den Engh G. Chromatic shifts in

the ﬂuorescenceemitted by murine thymocytes stained with Hoechst 33342. Cytometry A 2004;60:173À181.

17. Bestvater F , Spiess E, Stobrawa G, Hacker M, Feurer T, Porwoll T,

Berchner-Pfannschmidt U, Wotzlaw C, Acker H. Two-photon ﬂuores-cence absorption and emission spectra of dyes relevant for cell ima-ging. J Microsc 2002;208:108À115.