3D polyelectrolyte scaffolds to mimic exocrine glands: a step towards a prostate-on-chip platform

The membrane was formed using a layer-by-layer process that utilised the polyelectrolytes poly-sodium 4-styrenesulphonate (PSS) and poly-allylamine hydrochloride (PAH) (10). The optimisation of this membrane for the UroLOC device proceeded using 2 protocols. The first protocol constructed the membrane inside 3D hemispherical cavities, with the purpose to grow the exocrine epithelial cells on these 3D hemispherical membranes and to optimize the analytical method for the small volumes of secretions from each 3D hemispherical cavity. The second protocol was to construct the membrane on a 3D acinar-mimic sacrificial mould so as to create a free-standing porous membrane, with the purpose to mimic the biological acinar architecture with epithelial cells on the inside of the membrane (luminal side) and stromal cells on the outside (basal side).

(a) First Protocol - Mould design, fabrication and optimisation

The 3D hemispherical cavities were produced from a master mould. Fig. 1 illustrates the steps required for the assembly of the master mould. An inket printer was used to spot resin drops of about 1 nl onto a resin substrate so as to create an array of 200 resin hemispheres of 100 μm diameter. This procedure was optimized to avoid the viscous resistance of the resin in the injector. Once the master mould was assembled, it was replicated by slowly pouring liquid PDMS onto it (Two-part Sylgard silicon 10:1 v/v) at RT and allowing the PDMS to polymerize overnight at 60°C (Fig. 1.A). From that master mould, a derived mould was formed for the routine production of the 3D hemispherical cavities. The derived mould consisted of 200 PDMS hemispherical multi-wells (100 μm x 50 μm) formed across a chip with a surface area of 2 cm². Characterisation of the surface and the hemispheres of the PDMS device was performed by scanning electron microscopy (SEM) (Fig. 1.B) and by Z stack and 3D reconstruction on a confocal microscopy (LSM 710 confocal scanning laser microscope (Carl Zeiss, Jena, Germany) (Fig. 1.D). The Z-stack acquisition was carried out using a C-apochromat water-immersion objective (magnification 63x, NA 1.2, Carl Zeiss).

Figure 1

The derived mould was used to form 3D hemispherical cavities in a PDMS chip. Multi-layered polyelectrolyte films were assembled inside these 3D hemispherical cavities by alternate coating of PSS (poly (sodium 4-styrenesulphonate; Sigma 243051)) and PAH (poly (allylamine hydrochloride; Sigma 283223)) following the LbL (Layer-by-Layer) process we have previously described (10, 11) (see also Fig. SI.1). The 3D hemispherical cavities were coated with 16 polyelectrolyte (PE) layers in order to obtain a membrane of around 32 nm in thickness that was easily visualised with fluorescein (Fig. 1.C). For that visualization, the PDMS chip containing the 3D hemispherical cavities was incubated for 2 hours at RT in a solution of Fluorescein (10 μM in PBS) and washed twice in PBS before observation. Fluorescein was excited at 488 nm and fluorescence was observed using a 510–560 nm filter. 13 successive optical slices were captured along the cell z-axis, with a step of about 3.5 μm. For 3D reconstruction, PE-Fluorescein was scanned with about 50 slices. Images were processed using the ZEN software. A color code was established to indicate the depth of the well.

(e) Second Protocol - Development of 3D self-supporting PE membrane scaffolds

he concept was based on the modular assembly of 3D printed pieces that provided a system to shape the self-supporting “acinar-like” PE scaffolds. The methodology relied on using alginate as a sacrificial core for PE coating from a shape of either alginate “finger-like” structures (Fig. 2, first approach) or alginate “wells” (Fig. 3, second approach). In both cases, 3D printed pieces were manufactured by 3D Fab platform (Villeurbanne, France) using an Objet Pro 3D printer. The material used was acrylate resin, also known as plexiglass, (generally transparent to allow microscopy) which was UV photo cured after each layer had been deposited.

Figure 2

Figure 3

For the first approach (Fig. 2), the detail of the individual 3D pieces are illustrated in Fig. SI.3. To create the “finger-like” structure that provided the sacrificial core for the PE membrane, alginate (1.5 %) was flowed through the different UroSCAF pieces. Three different pieces were assembled and clamped together to form the UroLOC device, with PDMS gaskets used to avoid leaks in the system. The free end of the UroFLU piece was then fit into a three-way valve and connected to a syringe that had previously been filled with alginate (1.5-3%). The flow of alginate was achieved by a six-channel pump (SDX Tethered Membranes Pty Ltd), which was under software control to achieve flow-rates of μL/min so as to slowly and precisely control the formation of the “finger-like” extrusions. Alginate polymerization was achieved by adding CaCl₂ (100 mM) in the upper chamber once alginate had started to be extruded. PE coating was done over several substrates to which the general layer by layer deposition protocol was adapted (see Fig. SI.1). PSS and PAH were always prepared in a 1 mg/mL concentration solution, dissolved in either CaCl₂ (5, 100 or 500 mM), NaCl (250 or 500 mM), MgCl₂ (500 mM) or KCl (500 mM). Then, alginate dissolution was done with EDTA.

In the second approach (Fig. 3), sacrificial alginate moulds were created from 3D printed pieces containing pillars (Fig. 3.A). The feasibility of this approach was tested using a protocol inspired from a previously published method (14). Square moulds were filled with alginate and covered with a piece of Whatman paper soaked with CaCl₂ (100 mM). A support containing a chamber was then placed on top of the paper and the system was hold together with clamps. The resulting chamber was then filled with CaCl₂ (100 mM). The system was then incubated for 4h at room temperature to allow alginate gelation. That protocol produced “finger-like” alginate structures from which a free-floating PE membrane was formed (Fig. SI.4.1). To go beyond that feasibility of forming a polyelectrolyte membrane and to better control the attachment of the polyelectrolyte membrane to a substrate, new pieces namely UroBOXES were designed and manufactured by 3D printing (Fig. 3.B). PE coating over the UroBOXES was done by assembling the green piece with the black one containing the corresponding alginate mould. The “cheddar-like” piece (orange one) was then used to keep the green piece in tight contact with the alginate mould. The whole system was hold together with clamps. PSS and PAH solutions were alternatively added to the free surface on the green piece and incubated for 5 minutes (Fig. SI.1). A washing step with milliQ water followed by a liquid removal from the piece with an absorbent paper was done between each layer (see Fig. SI.1). This procedure was repeated until the desired number of layers was obtained. All incubation steps were done under gentle agitation on a rocking platform (70 rpm) and at room temperature.

Some technical issues were addressed during the assembly procedure. First, it was required to have a perfect alignment between the wells created on the alginate mould and the holes in the second piece. This issue was solved by designing the blue and green piece with the pillars/holes respectively in the exact same position. Secondly, the addition of liquids to the system was necessary for alginate gelation, polyelectrolyte coating and cell culture. Alginate needs to be contained in a support and a non-leaky connection needs to be created between the black and blue piece. Therefore, both pieces contain walls to avoid these drawbacks. Finally, the green piece was originally created with symmetrical walls on both sides of the horizontal plane. However, this design prevented liquid from entering the bottom chamber, and placed the horizontal plane outside the normal working distance for microscope objectives. This design was not appropriate for cell culture nor for imaging by inverted microscopy. For these reasons, a new (grey) piece was created with a castellation shape to enable the passage of liquid to the bottom chamber as well as to provide the opportunity to remove almost completely one of the walls in the green piece, which placed the horizontal plane within the working distance of the objectives for inverted microscopy. After several trials had been done, an extra set of pieces was designed to evaluate different diameter sizes as well as the height of the pillars. This set consisted in a blue and green piece containing four different conditions (h = 0.4 or 1 mm and Ø = 0.5 or 1 mm) with three replicas for each condition.

RWPE-1 and WPE1-int were growth in KSFM (Life Technologies, Carlsbad) supplemented with 5 ng/mL of epidermal growth factor (EGF) and 50 mg/mL of bovine pituitary extract (Gibco) and 1% Penicillin-Streptomycin (Gibco). Cells were cultured up to 70% confluence and seeded in T25 Flasks at a density of 2 × 10⁴ viable cells/cm². PNT2 (SV40-immortalized normal adult prostate epithelial cells), PC3 and LNCaP (cancerous prostatic cell lines) previously described (11) were used for MALDI-based analysis of the secretome from normal and cancerous cells. Among the cancerous cells, PC3 are androgen receptor-negative while LNCaP, as androgen receptor-positive cells, have the property to secrete PSA. PNT-2 cells were obtained from ECACC (N_95012613) and were routinely cultured in RPMI Glutamax culture media supplemented with 10% foetal calf serum and 1% penicillin/streptomycin. PC3 cells (ATCC, CRL-1435) were initiated from a bone metastasis of a grade IV prostatic adenocarcinoma from a 62-year-old male Caucasian. PC3 cells were routinely cultured in RPMI Glutamax culture media (Invitrogen, Ref. 61870-010) supplemented with 10% foetal calf serum (PAA, Ref. A15-101) and 1% penicillin/streptomycin (Invitrogen, Ref. 15140-122). LNCaP (ATCC #1740) were grown in 2D in 25 cm²flasks with 5 mL of RPMI 1640 medium supplemented with 10% SVF penicillin-streptomycin. Pre-treated SVF was used (charcoal filtered) to remove all residues of hormones. All cell lines were grown in an incubator at 37°C, 5% CO₂, and 95% humidity. For different passages, the cells were washed twice with PBS followed by the addition of trypsin–EDTA and incubation for 3–10 min depending on the cell line. Trypsinisation was stopped by adding 6 to 8 ml of 0.1% soybean trypsin inhibitor. Cells counting was performed following centrifugation (5 min at 1600 rpm) and after adding 10 μl Trypan Blue to 10 μl of cell suspension.

In order to investigate the invasiveness processes, LNCaP cells were labelled with the Molecular Probes_CellTracker™_Green CMFDA (5-chloromethylfluorescein diacetate) (Life Technologies C7025) (3 μM, for 30 minutes) and then introduced into a monolayer of PNT-2 epithelial cells once they reached 90% of confluency. For immunostaining assays, cells were then incubated for 20 min at room temperature in PFA 4% in PBS then rinsed in PBS for 10 minutes and permeabilized 5 minutes in 0.5% Triton X100 in PBS. Cultures were finally incubated with Phalloidin-Cy3 (1:400) in PBS for 1 hour then washed 3x5 minutes and then nuclei were stained with Hoechst (2 μg/ml) for 10 minutes. Stained cells were imaged by fluorescence microscopy using a confocal spinning-disk inverted microscope (Nikon TI-E Eclipse) with a 20x air objective equipped with a Yokogawa CSU-1 Unit for z-stack acquisitions and an Evolve EM-CCD camera. Primary Renal Proximal Tubule Epithelial Cells (RPTEC) were kindly provided by I. Villegrain (CEA Grenoble) and were used as control for ELISPOT assay as non-PSA secreting epithelial cells.

Biochemical studies were carried out on secreted supernatants from cells grown in 2D to validate the use of our cell lines and the possibility to detect secreted proteins in the supernatants without having to apply purification or concentration steps. The 2D format offers very well controlled culture conditions to evaluate the minimum number of cells required to perform a secretomics-based Maldi analysis. Moreover, this number of cells and the corresponding volume of secretions need to be compatible with further microfluidic integration and sampling. PSA was chosen as “housekeeper” protein for the secretome.

Western Blot was performed with the same supernatant of cell culture used for MALDI analysis. Different quantities of supernatant were separated by electrophoresis on a NuPAGE™ Novex™ 4-12% Bis-Tris Protein Gel (Life Technologies) in MES buffer and then transferred onto a nitrocellulose membrane (Sigma-Aldrich). The membranes were blocked in 5% non-fat milk in Tris-buffered saline, 0.1% Tween 20 (TBS-T, Sigma) for 1 hour at room temperature and incubated with primary antibodies diluted at 1:1000 in 5% non-fat milk in TBS-T overnight at 4°C. Rabbit polyclonal antibodies directed against PSA (PSA/KLK3 (D6B1) XP® Rabbit mAb ref 5365, cell signalling technology) and GAPDH (Santa Cruz Biotechnology) were used. Membranes were washed three times for 10 min in TBS-T and incubated with secondary horseradish peroxidase-conjugated antibodies (anti-rabbit, eBiosciences) at a dilution of 1:10000 in TBS-T for 50 min. Membranes were washed three times for 10 min in TBS-T and revealed with ChemiDoc™ MP systems (BioRad). Chemiluminescence was analysed using a Chemi-Doc Touch Imaging System (BioRad).

ELISA tests were performed by using a Human Free PSA ELISA Kit from Anogen (EL10050) for the in vitro quantitative determination of free Prostate-Specific Antigen (fPSA) concentrations in the supernatant. Washing steps were performed with PBS, Tween 0.05% and milk 5% during 15 min. Incubation with IgG anti-PSA-ncam was performed during 30 min and incubation with a rabbit HRP (1:1000) during 30 min. Washing step was done with PBS Tween 0.05% during 15 min, twice.

The “Proteome Profiler Human XL Cytokine Array Kit” was used with 200 μl of the same supernatants used for biochemistry and MALDI analysis and following the supplier protocol and revealed with ChemiDoc™ MP systems (BioRad) with an exposure time of 150 seconds. For the calculation of the intensity of each of the spots, we used the Fiji software with the “Microarray profile” plugin. Each protein was spotted twice on the microarray and the histogram corresponds to the mean of the integrated density obtained for these two spots to which the value of the negative control has been subtracted.

Spectra were preprocessed (tophat baseline removal and smoothing) and normalized by total ion current (TIC) in Bruker Clinprotool software v3.0. 2D peak distribution and Principal Component Analysis (PCA) was performed in Clinprotool. Unsupervised hierarchical clustering was performed on exported spectra from Clinprotool, using log transformation of the intensities and clustering via uncentered correlation similarity metric (Gene cluster v3.0). For each condition, 3 biological replicates were analysed 3 times.

Cells at day 8 were well-organized at the bottom of the PE scaffolds and form a monolayer with intercellular tight junctions revealed by immunofluorescence (Fig. 4). Moreover, a protocol to collect secretions from cells grown in the wells was established. The volume of secretions was determined to be 1.76 μl per 100 wells, according to the bromophenol blue volume which has been transferred to Whatman paper and then measured from the calibration curve (Fig. 4.B). This volume of 1.76 μl per well is consistent with a microfluidic handling.

In order to validate the exocrine function of prostatic cells grown at the bottom of the wells, we verified that cells did secrete PSA. For this, the medium present above 3 moulds was harvested in 3 conditions. WPE1-int unstimulated with DHT, WPE1-int stimulated with DHT, but also 3 PDMS mould with RPTEC cells (renal cells, as negative control). A positive control was made with 3 ng of pure PSA (Fig. 4.C). We observed that the medium of the stimulated WPE1-int present PSA. The ELISPOT analysis confirms that it is possible to detect PSA secreted by cells growing in small wells that have an hemisphere-shape that mimics the dimensions of acinar-like structures.

To validate our hypothesis that cancerous cell lines can be discriminated on the basis of their secretome, we performed biochemical analysis from their supernatant. As expected and confirmed with ELISA test, PSA secretions were detected from LNCaP, with or without stimulation (15) by contrast to the other cell lines that are described as not secreting PSA (16) (Fig. SI.5). With the same LNCaP supernatant used for ELISA, we performed ELISPOT (Fig. SI.5.B) and Western Blot (Fig. SI.5.C) assays as second confirmation. The protease microarrays showed overexpression of PSA and many other known proteins (e.g. Urokinase, Kallikrein 10, 7, 6, 5, Cathepsin D) in the LNCaP cell-line compared to the others (Fig. SI.5.D).

To avoid any interference of FCS within MALDI analysis, FCS was replaced with BSA 2g/L, as previously described in the literature (17, 18), during the incubation phase used to recover the secretory fluid. In the final point after five days, a Live/Dead staining showed no cell mortality and no alteration in cell morphology (data not shown). After biochemistry validations, supernatants from LNCaP, PNT2 and PC3 cell cultured as described above were analyzed by MALDI MS in the range of 1000 to 20000 Da for peptides/small proteins profiling. The amount of supernatant spotted was normalized to approximately 300 cells per sample.

Figure 4

Fig. 5A shows PCA of the 3 different cell lines. Spectra are projected in a coordinate system defined by the PCs (95 % of the total variance). The best segregation between normal and cancerous cells is reached by representing data in a PC1-PC2 plane. The score plots reveal a trend of separation of MALDI spectra corresponding to the different cell lines. The discrimination of the supernatants from the different cell types was further confirmed by the unsupervised hierarchical clustering analysis (Fig. 5.B). The obtained dendrogram indicates that MALDI MS data of the supernatants are specific of each cell lines. Fig. 5.C shows the two peaks that have the most significant p-value to differentiate the 3 cell types (see also Fig. SI.6). The 2D peak distribution resulted in a clear segregation of the 3 cell lines.

Figure 5

To assess the ability to differentiate a healthy PNT2 culture from a PNT2 and LNCaP co-culture, different numbers of LNCaP cells (1,000, 10,000 and 50,000) have been seeded in a PNT2 culture and MALDI profiling analysis has been performed on the supernatant. We first observed that spiked cancerous cells were properly incorporated in the non-cancerous cell monolayer (Fig. 6.A). Tumour cells settled down, proliferated, and gradually populated the device, to possibly replace healthy basal cells and eventually affect the overall secretory profile of the tissue. Fig. 6.B-D illustrates the 2D peak distribution of MALDI profiling of supernatants from PNT2, LNCaP and PNT2 loaded with 1,000, 10,000 or 50,000 LNCaP cells. While LNCaP culture is clearly separated, the differentiation of PNT2 and PNT2 co-cultured with LNCaP cells is less obvious. However, there is a tendency to separate the PNT2 co-culture condition when the amount of LNCaP is increased to 50,000 cells (Fig. 6.B)

On the basis of the results obtained with PE-coated hemispheres, we developed incremental systems to construct self-supporting finger-like scaffolds providing an access to both side of the acinus, basal and luminal. Alginate flowing was used to create a sacrificial 3D shaped structure in top of which polyelectrolyte membrane would be created. The idea was then to obtain a homogeneous region containing several alginate pillars. Moreover, the use of a programmable software controlled syringe pump would allow a fine optimisation and automatisation of the process. Nevertheless, alginate flowing through the 3D printed UroSCAF pieces resulted in irregular and non-uniform extrusions of alginate from the holes; sometimes extrusions appeared from only one hole. Moreover, when the procedure was repeated with the same piece, alginate came out from a different hole. These results may have been due to microfluidic physical flow-resistance issues, which might be due to slight differences in the shape between different holes in the same piece that was 3D-printed.

As this first device (see Fig. 2) didn’t allow the assembly of an array of hollow scaffolds in a reproducible manner, we then developed another approach by assembling 3D-printed modules composed of “finger-like” structures instead of microholes (see Fig. 3). Alginate gelation was achieved by following a published protocol (19) that was adapted to our system. The optimisation of the protocol worked out well and the gelation and shaping of the alginate was reproducible (data not shown). Thus, polyelectrolyte membranes were obtained from alginate moulds containing wells. The dissolution of PE coated alginate moulds by introduction in an EDTA bath resulted in a visible fragile floating membrane shaped to the desired structure (Fig. SI.4.A) However, these floating structures could not be manipulated and the protocol was then adapted to provide a membrane that could be attached to a glass surface, which would enable the manipulation of the obtained polyelectrolyte membrane and its optimisation on both the holographic and confocal microscopes. However, this procedure induced folding in the polyelectrolyte membrane originated by an uneven attachment of the membrane to the bottom of the petri dish. Nevertheless, as expected, this procedure allowed us a good optimisation of the shape. The optimisation of those membranes showed a difficulty of the finger-like shape to remain self-supporting (Fig. SI.4.B) although they seemed to hold their shape in the floating membranes. Moreover, when the polyelectrolyte membrane was attached to a surface, most of the fingers seemed to have exploded or collapsed and were principally found in three conformations: volcano like, side collapsed or central collapsed. We hypothesise that this could be due either to the strong electrostatic force created between the glass (coated or not) and the polyelectrolyte membrane; or to the electrostatic forces produced between the walls of the finger itself.

Figure 6

New UroBOXES pieces were then carefully designed (see Fig. 3) and created to have a polyelectrolyte membrane accessible from both sides and easily manipulated. Promising results were obtained from the first trials where almost closed fingers could be seen (Fig. SI.4.C). Further optimisations of the protocol resulted in the reduction of the hole size found in the polyelectrolyte membrane although the presence of those defects were not eliminated. Several hypotheses were then set for the existence of this holes which could either be due to: (i) the difficulty to remove completely the liquid from the holes due to their small size or (ii) the lack of mixing inside each one of the wells due to their size.