In a constant effort to produce more and more pertinent in vitro models for cancer studies, the importance of studying cancer cells in three-dimensions (3D) is increasingly recognized 1 2 3 4 5 . For this purpose, spheroids can be produced in vitro with some permanent cancer cell lines, including colon cancer cell lines, when cultured in non-adherent conditions 6 7 . These spheroids are known to mimic microtumours more closely than cancer cell line monolayers and have been mainly used in chemo- and radio-resistance studies. Indeed, tumour spheroids represent quite realistically the three-dimensional growth and organization of solid avascular tumours and, consequently, simulate much more precisely the cell-cell interactions and microenvironmental conditions found in tumours, especially nutrient and oxygen gradients. Another 3D cancer cell model, requiring a specific culture protocol, aims to promoting in vitro expansion of cancer stem cells (CSCs) from solid tumour tissue as is the case with neurospheres 8 9 , mammospheres 10 11 or colon cancer spheres 12 13 . CSCs are defined as a rare subset of tumour cells, which have the unique capability to form tumours in serial xenotransplantation assays, and to reestablish, at each in vivo passage, the hierarchical cell organization and heterogeneity of the parental tumour. In vitro methods have been developed to grow and study these cells in sphere-forming assays. The phenotypic cell isolation strategy that relies on the immunotargeting of cell surface proteins coupled with cell sorting by flow cytometry (FC) is now widely used to isolate CSCs, after enzymatic dissociation of tumour samples or cancer spheres.

Up to now, protein expression by these sphere forming cells was mostly studied by FC on single cells. Immunostaining techniques performed on the whole intact spheres are traditionally done on cytospin preparations 14 15 and paraffin 16 17 or frozen sections 13 17 because of the common failure of the antibodies to penetrate fully into compact tumour spheres, leading to a loss of information. Besides, the cytospin approach is limited to the investigation of the more external cells, while physical sections do not allow the analysis of the entire 3D structure, unless time-consuming serial sectioning is performed. Another way to gain access to the whole structure has been to use confocal imaging of three dimensional tumour spheroids expressing genetically encoded fluorescent reporters fused to the proteins of interest 18 19 . However, this approach requires a cell transfection step and does not allow visualizing endogenous proteins at their physiological levels.

In the present work, we describe a new immunostaining protocol in which spheroids are simultaneously fixed and permeabilized in a mixture of paraformaldehyde and Triton X100 and then gradually dehydrated in methanol before incubation with labelling antibodies. We show that this results in the complete permeabilization of the spheroids while preserving their structures and the localization of the proteins of interest. Whole spheroids can then be imaged in the confocal microscope to reveal the spatial distribution patterns of several endogenous proteins in situ.

The method presented here aims to visualizing in situ the expression of proteins in three-dimensional tumour models using immunostaining coupled with confocal imaging. We used HT29 and CT320X6 colon cancer cell lines because of their ability to form compact multicellular tumour spheroids when cultured according to the liquid overlay technique 6 . All spheroids used had a diameter comprised between 120 and 150 μm.

One of the main problems arising in confocal imaging of thick specimens is the poor penetration of the antibodies within the whole structure. Therefore, we first addressed this issue by performing confocal imaging after immunostaining of the spheroids against a marker expressed evenly by all cells within the spheroids. We chose Epithelial Cell Adhesion Molecule (EpCAM), a membrane protein also known as Epithelial-Specific Antigen (ESA) since all cells of the HT29 and CT320X6 spheroids are known to be positive for EpCAM, as confirmed by FC analysis (Figure 1A). After immunostaining with an anti-EpCAM monoclonal antibody conjugated to FITC, confocal images were recorded in single optical planes located 40 μm and 70 μm below the surface of the coverslip.

As explained below, the main problem we had to deal with was the penetration of the antibody through the whole spheroid. The fixation and permeabilization steps appeared to be the most critical ones to achieve this goal.

A classical two steps method, namely fixation in 4% PFA (3 h, 4°C) followed by permeabilization in 1% Triton X100 (1 h, RT) did not yield satisfactory results. This procedure resulted in a bright anti-EpCAM staining of cell plasma membranes in the periphery of the spheroid but left the centre of the spheroid without any detectable labelling (Additional file 1A). This was in contrast to staining with TOPRO-3, a small cell-impermeant molecule, which stained all cell nuclei in the spheroids. Changing the experimental conditions for the fixation (2-4% PFA, 1-12 h, RT or 4°C) or the permeabilization (0.5-3% Triton X100, 1-6 h, RT or 4°C), using other detergents (saponine, SDS) or increasing incubation times in antibodies (up to 72 h) did not lead to any improvement (data not shown). The reverse protocol consisting in extraction (0.1% Triton X100, 5 min, 4°C) prior to fixation (4% PFA, 3 h, 4°C) resulted in cell membrane damage in the periphery of the spheroids with seemingly no permeabilization in the centre (Additional file 1B). Simultaneous fixation/permeabilization in a single step led to a striking increase of the staining inside the spheroids with no apparent alteration of the structure (Additional file 1C). Indeed, when spheroids were incubated in PBS containing 4% PFA and 1% Triton X-100 at 4°C for 3 h before immunostaining, all cells were evenly labelled down to a depth of 40 μm. However, the very center of the spheroid still appeared devoided of any staining as evidenced in images taken at 70 μm depth (Additional file 1D).

Dehydration in alcohols has been reported to efficiently permeabilize thick specimens, thus allowing full penetration of antibodies or fluorescent probes. For example, dehydration in methanol followed by rehydration allowed whole-mount immunostaining of large insect brains 22 or microvasculature imaging in whole-mount organs 23 . Moreover, only minimal, and more importantly isotropic, shrinkage occurs when alcohol concentration is gradually increased 24 . After simultaneous fixation/permeabilization as described above we then passed the spheroids through a series of baths of increasing methanol concentration at 4°C (see Methods). The spheroids were then rehydrated through the same descending series of baths and resuspended in PBS before immunostaining.

As can be seen in Figure 1B, this procedure resulted in a strong and uniform membrane staining with an anti-EpCAM antibody in both HT29 and CT320X6 spheroids. The same staining pattern was observed at 40 μm and 70 μm depth, thus demonstrating the full penetration of the antibody down to the center of the spheroids. The dramatic increase in permeability observed after methanol treatment likely results from two effects. Indeed, methanol is both a strong permeabilizing agent which 'dissolves' the membranes more efficiently than nonionic detergents like Triton X100 and a coagulative fixative 25 which probably makes the cytosol matrix less gelatinous and thus more permeable to antibodies. However, a simplified procedure consisting in a single step of fixation/permeabilization in cold methanol or methanol/acetone without prior aldehydic fixation led to a massive delocalization of membrane proteins (data not shown). This is in contrast with a previous work reporting the successful fixation and permeabilization of embryoid bodies by cold methanol/acetone treatment without observation of membrane damage 26 . Antigen delocalization is a known possible artefact in immunostaining experiments 27 28 . It can occur during the fixation and/or permeabilization steps and lead to a wrong conclusion regarding the subcellular localization of the labelled proteins. In the current study, we found that aldehydic fixation before methanol treatment allowed EpCAM to be correctly localized at the plasma membrane. This was also the case for all antigens examined (see below) whatever the subcellular location (cytosolic, nuclear, cytoskeleton- and membrane-associated proteins). Nevertheless the possibility of this artefact should be kept in mind when addressing a new target since it is dependent of both the tissue and the antigen examined. It is also noteworthy that, although the present protocol implies long processing times, it can be performed in several steps. In particular, after dehydration/rehydration in methanol, spheroids could be stored for up to one month at 4°C in PBS without any alteration in the staining pattern of the antigens examined. This protocol was used in all subsequent immunostaining experiments.

To further validate our protocol, in situ expression of other proteins was studied and compared to the expression profiles obtained by FC or paraffin immunochemistry (pIHC). We selected several antigens for their different subcellular distribution (cytoplasm and/or nucleus and/or plasma membrane) and their relevance in the field of colon cancer.

Cytokeratin 20 (CK20) is an intermediate filament protein whose presence is essentially restricted to differentiated cells from gastric and intestinal epithelium and urothelium 29 . CK20 is routinely used in combination with CK7 in distinguishing colon carcinomas (CK20⁺) from ovarian, pulmonary and breast carcinoma 30 . Both HT29 and CT320X6 spheroids are formed with well-differentiated human colon adenocarcinoma cell lines and cultured in serum-containing medium. The CK20 expression profile observed here is consequently in agreement with these data. Indeed, almost all cells in HT29 and CT320X6 spheroids were found positive in confocal imaging and also FC analysis (Figure 2).

The cytosolic carbohydrate antigen CA 19-9 is a glycoprotein related to the monosyalilated Lewis antigen, which is expressed in the cell membrane and the cytosol of human colorectal carcinoma as well as in the normal mucosa 31 . CA 19-9 is frequently used as a tumour marker, particularly for colon and pancreas cancer and CA 19-9 serum level is considered a reliable indicator for patient follow-up 32 . The three immunostaining techniques revealed both CA 19-9 positive and negative cell populations within HT29 spheroids (Figure 3). However, these techniques were not quantitatively equivalent even when the same antibody was used. Unsurprisingly, FC was the most sensitive technique with 75 ± 2.6% CA 19-9 positive cells detected while confocal microscopy (62 ± 6.1%) turned out to have a higher sensitivity than pIHC (51 ± 0.6%, p < 0.05). Besides, the location pattern of CA 19-9 was qualitatively similar in confocal imaging (Figure 3A) and in pIHC (Figure 3C), thus confirming the validity of our protocol for this marker.

β-catenin plays a dual role in cells as a structural component of cell-cell adherens junctions and as a key player in the Wnt signalling pathway 33 . This pivotal protein thus displays a complex localization pattern, being present at the plasma membrane (associated with E-cadherin), the cytosol (within a multiprotein complex containing notably GSK3β, axin, and APC) and the nucleus (where it acts as a co-transcriptional activator of target genes and promotes tumour progression). The nuclear accumulation of the β-catenin is considered the initial step in colon carcinogenesis and the transcription of β-catenin target genes becomes constitutively activated in APC-deficient colon cancers 34 . However, most tumours do not display a homogeneous subcellular expression pattern but a nuclear accumulation of β-catenin predominantly localized at the tumour's periphery 35 . All cells from HT29, an APC-deficient cell line, were positive for β-catenin, as illustrated by FC analysis (Figure 3B). This was confirmed by confocal microscopy (Figure 3A) which gave additional information about the subcellular distribution of the protein. Besides being strongly expressed at the level of the plasma membrane, soluble β-catenin was often found in cytosolic and nuclear compartments in confocal images, as confirmed by pIHC (Figure 3C). This nuclear expression was randomly detected within the spheroids. More recently, nuclear β-catenin has been also presumed to mark colon cancer stem cells 13 36 37 .

Because they are suspected to play a crucial role in the initial steps of tumorigenesis, cancer stem cells are the subject of intensive investigations. Floating multicellular spheres (including colon cancer spheres, mammospheres or neurospheres) have been described as being valuable models for in vitro CSC expansion and are topologically similar to the spheroids used in the present study. The expression of specific CSC markers within these microspheres is traditionally investigated by FC after enzymatic dissociation. This is a powerful approach allowing to rapidly and quantitatively assess the presence of CSCs in the spheres and also to isolate CSCs which can then be tested for in vivo tumourigenicity. However, it can not address the precise distribution of CSC markers within the spheres, for example in clustered or isolated cells.

We applied the method described in the present work to study the localization of putative CSC marker-expressing cells within HT29 and CT320X6 spheroids. It has to be noted that the liquid overlay technique in conventional culture medium used here is not a method dedicated to CSC expansion. However, this technique allows the obtaining of reproducible and compact cancer cell spheroids, mimicking strong cell-cell interactions occurring in CSC spheres. First, CD44, a widely distributed transmembrane cell adhesion molecule and the major cell surface receptor for hyaluronic acid, is reported for isolating colorectal cancer initiating cells 38 39 40 . Confocal imaging, FC and pIHC studies revealed that all cells within HT29 spheroids displayed CD44 expression (Figure 4). These results are in line with previous phenotyping of HT29 cells 41 . ALDH1, largely reported as breast CSC marker 42 43 , is also a potential stem cell marker for colon cancer 44 45 . In the present study, a large population of ALDH1⁺ cells has been detected within HT29 spheroids both in confocal imaging (86% ± 2) and with pIHC (85% ± 3) (Figure 4). These data are consistent with strong ALDH1 protein staining observed in patient specimens of colon adenocarcinoma, while only few cells were found in FC assays to display ALDH enzymatic activity, a potential characteristic of CSCs 44 .

CD133 (prominin-1) is a five-transmembrane domain glycoprotein localized in membrane protrusions. It is suspected to be a CSC marker in brain 9 46 or colon 12 47 tumours, although this view has been challenged 48 49 . FC analysis (Figure 5B) showed that about 5.0% ± 0.4 of the cells were CD133 positive in CT320X6 spheroids. This percentage was close to 1.1% (n = 2) in HT29 spheroids, a value consistent with a previous report in HT29 cells 50 . Confocal microscopy also identified a small fraction of CD133 positive cells in CT320X6 and HT29 spheroids (Figure 5A) and double staining experiments additionally revealed that CD133 was colocalized with EBP50, a protein found at the apical membrane of polarized epithelia and involved in microvilli formation 51 . At high magnification, EBP50 staining also revealed the presence of glandular acini filled with CD133 positive material. Interestingly, a similar observation was made recently in human colorectal cancer where CD133 was found at the glandular-luminal surface of cancerous cells and in shed cellular debris inside the glandular lumina 52 . Our method is therefore able to reveal fine features in the distribution of specific markers, even with a very low expression, throughout the entire spheroids.

Different methods have been reported to study protein expression within tumour spheres. Immunostaining detected by peroxidase or fluorophore-conjugated antibodies has been traditionally performed using cytospin preparations 14 15 , paraffin sections 16 17 or frozen sections 13 17 . Nevertheless, cytospin slides allow study of only external cells of the structure while paraffin and frozen sections do not permit in situ visualization of the target protein in the entire structure.

We have developed a protocol allowing successful antibody penetration in tumour spheres of 150 μm in diameter with excellent preservation of the structure and without artefactual delocalization of the proteins examined. This allowed us to perform whole mount immunostaining of these compact structures and to use the optical sectioning capability of confocal microscopy to analyse the distribution of several proteins of interest in the entire spheroids. Besides, 3D localization of the antigens of interest can be easily obtained from a z-series of confocal images collected through the spheroid (see Additional File 2). Our method is simple, has a high sensitivity and offers the possibility to screen the whole structure. It also makes multi-staining and colocalization of several antigens possible. Therefore, it should prove useful as a new tool to image the subcellular distribution of relevant markers or to localize rare positive cells, like CSCs, within tumour spheres.

The authors declare that they have no competing interests.

LBW contributed to the conception of the method, performed all confocal experiments in the laboratory, and co-drafted the manuscript. JMG and SR participated in method development, performed paraffin immunostaining and cytometry analysis respectively, and revised the manuscript. DB contributed with data analysis and participated in revision of the manuscript. BS participated in method development, contributed with data analysis and revision of the manuscript. VDM initiated the project, contributed with data analysis and co-drafted the manuscript. All authors read and approved the final manuscript.