Biomedical Microdevices

, Volume 12, Issue 3, pp 465–475

Spontaneous formation of stably-attached and 3D-organized hepatocyte aggregates on oxygen-permeable polydimethylsiloxane membranes having 3D microstructures

Authors

    • Laboratory for Integrated Micro-Mechatronic Systems (LIMMS/CNRS-IIS), Institute of Industrial ScienceUniversity of Tokyo
  • Teruo Fujii
    • Institute of Industrial Science (IIS)University of Tokyo
  • Yasuyuki Sakai
    • Institute of Industrial Science (IIS)University of Tokyo
Article

DOI: 10.1007/s10544-010-9403-8

Cite this article as:
Evenou, F., Fujii, T. & Sakai, Y. Biomed Microdevices (2010) 12: 465. doi:10.1007/s10544-010-9403-8

Abstract

In order to enhance the viability and the differentiated functions of primary hepatocytes in cultures, it appears important to have them organized within a three-dimensional (3D) structure which promotes extensive cell-cell contacts, but also to be adequately supplied with oxygen. Here, we report a simple methodology satisfying these two fundamental but sometimes conflicting issues: primary rat hepatocytes were cultured on polydimethylsiloxane (PDMS) membranes with 3D-pillared microstructures with various dimensions, so that the cells could organize themselves around the pillars into various kinds of 3D multicellular aggregates, while being continuously supplied with oxygen by diffusion through the PDMS membrane. As expected, under such conditions, hepatocyte cultures exhibited higher albumin secretion and urea synthesis rates. It appeared then that as the spacing decreased between the pillars, the cells were more stably organized into smaller spherical aggregates and displayed the highest albumin secretion rates. Such a simple design is likely to be included in a new drug/chemical screening system in a practical microplate format, but also appears applicable to microfluidic devices.

Keywords

Rat hepatocyteOxygen supplyPDMS membrane3D structure3D aggregate

1 Introduction

Conventional liver cell cultures present some limitations as they do not mimic properly in vivo situations, thus failing in maintaining high liver-specific functions. When culturing liver cells, it appears important to recover three-dimensional (3D) tissue organization and cell-cell contacts to retain the differentiated functions of the hepatocytes. Indeed, when cultured as a cellular monolayer in conventional tissue culture plates, primary hepatocytes rapidly lose their specific functions (Clayton et al. 1985; Strain and Neuberger 2002). Furthermore, in vivo, blood capillaries ensure an adequate delivery of various nutrients to the cells, as well as waste removal (Martin and Vermette 2005). Oxygen supply appears particularly crucial since it is one of the most limiting parameters of hepatocytes’ functions (Martin and Vermette 2005): a decrease of both their viability and function under low-oxygen tensions has thus been reported (Tilles et al. 2001). Hepatocytes are indeed highly metabolic and they exhibit elevated oxygen consumption rates (OCR), especially during the first hours of culture as they require a lot of energy to attach and spread on the surfaces (Rotem et al. 1992, 1994). However, oxygen delivery to the cells remains critical since it is poorly soluble in culture media, and it has been estimated that in standard tissue cultures, hepatocytes’ requirements were not matched (Nahmias et al. 2006).

Hepatocytes organized into 3D aggregates such as spheroids have been demonstrated to exhibit greater differentiated functions, with a longer duration, than 2D monolayers (Landry et al. 1985). Several methods have thus been developed in order to induce hepatocyte spheroids formation on various kinds of substrates such as non-adherent plastic substrates (Landry et al. 1985), proteoglycan-coated dishes (Koide et al. 1989), positively-charged surfaces (Koide et al. 1990; Hansen et al. 1998; Tzanakakis et al. 2001; Peshwa et al. 1994; Sakai and Suzuki 1991), PDMS surfaces (Nakazawa et al. 2009), or macroporous scaffolds (Ijima et al. 1998; Glicklis et al. 2004; Dvir-Ginzberg et al. 2004). Hepatocyte spheroids were also formed under rotational conditions in spinner vessels (Wu et al. 1996; Sakai et al. 1992) or in the microcavities of either a polystyrene (Fukuda and Nakazawa 2005) or a PDMS chip (Nakazawa et al. 2006). However, these methods present several drawbacks: as their formation is promoted on surfaces which provide a weak cell-adherent environment, the spheroids formed are not kept attached to the substrates but are floating, thus leading to a poor applicability in microfluidic devices or in microplate-based drug/chemical screening where easy separation of cells and culture medium is required. Moreover, such compact 3D multicellular organizations sometimes raise the issue of mass transfer limitations, notably with regards to the oxygen supply. Indeed, large spheroids generally exhibit a necrotic core, and it has been estimated that in order to maintain 90 % cell viability, the spheroids’ size should not exceed 100 µm in diameter (Glicklis et al. 2004; Dvir-Ginzberg et al. 2004).

Polydimethylsiloxane (PDMS) is a polymer which presents numerous advantages: it is biocompatible, transparent, relatively cheap, and it has already been used for conducting liver cell culture in microdevices (Leclerc et al. 2003, 2004; Ostrovidov et al. 2004; Mehta et al. 2007). It has also been reported that hepatocyte spheroids could be formed on either hydrophilic or hydrophobic non-coated PDMS surfaces (Nakazawa et al. 2009). Moreover, since it is highly permeable to gases (Charati and Stern 1998), this polymer appears particularly attractive for cell culture. By partly focusing on this property, our group has recently reported the static culture of liver cells (human hepatoma Hep G2 cell line or primary-cultured rat hepatocytes) on smooth PDMS membranes in bottomless polycarbonate multi-well frames (Evenou et al. 2010; Nishikawa et al. 2008a, b). In the case of primary rat hepatocyte culture, we showed that direct oxygenation through the PDMS membrane together with appropriate chemical surface modification (collagen covalent binding) enabled the formation of stably-attached and functional hemispheroids (Nishikawa et al. 2008a).

In the present study, we investigated the effects of 3D PDMS microstructures consisting of arrays of macropillars, with or without direct oxygen supply through similar PDMS membranes, on the morphology and the functionality of primary rat hepatocyte cultures in a simple microplate format. Because it presents a high surface/volume ratio, such kind of structure appears suitable for culturing cells at a high density, thus promoting cell-cell interactions. Moreover, we expected that such 3D topographical features would promote the 3D organization of the cultured hepatocytes.

As a result, we were able to show that the combination of 3D microstructures and direct oxygen supply through a PDMS membrane synergistically induced the spontaneous formation of 3D multicellular aggregates or spheroids which were firmly anchored to the pillars. Accordingly, hepatocyte cultures exhibited higher albumin secretion and urea synthesis rates.

2 Materials and methods

2.1 Fabrication process of 3D PDMS macropillars

PDMS microstructures were fabricated according to a replica moulding process against a 3D-micropatterned silicon wafer. SU-8 2100 photoresist (MicroChem) was spin-coated on a silicon wafer at 2,500 rpm for 30 s to produce a layer about 100 µm-thick, then pre-baked at 95°C for 20 min. It was exposed for 35 s under UV light (290 mJ/cm2) through a photomask, post-baked 1 min at 65°C followed by 10 min at 95°C, then developed 45 min in MicroChem`s SU-8 Developer and rinsed with isopropanol. It was finally hard-baked 10 min at 150°C to remove cracks. The mould master was then treated by CHF3 plasma using a Reactive Ion Etching machine (RIE-10NR, Samco, Kyoto, Japan) to enable easy separation of the PDMS polymer from the mould.

A 10:2 mass ratio mixture of PDMS prepolymer and curing agent (Silpot 184 W/C, Dow Corning Toray, Tokyo, Japan) was poured on the mould, degassed in a vacuum chamber, cured for 45 min at 100°C in an oven, then the PDMS replica was peeled off from the master. 3D microstructures were analyzed using a Field-Emission Scanning Electron Microscope (FESEM JEOL JSM-7400F): PDMS macropillars were octagonal and they were arranged in a honeycomb pattern; they were around 100 µm-high, 40 µm-wide and they were edge-to-edge spaced out every 20, 50, 100 or 200 µm (sp20, sp50, sp100, sp200) (Fig. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs10544-010-9403-8/MediaObjects/10544_2010_9403_Fig1_HTML.gif
Fig. 1

PDMS membrane-based cell culture system with 3D-microstructured PDMS substrates: (a) PDMS macropillars are arranged in a honeycomb pattern, the spacing between two adjacent pillars (sp) varying between 20, 50, 100 and 200 µm; they are 40 µm-wide and 100 µm-high. (b) Scanning electron micrograph of PDMS pillars with a 50 µm-spacing

2.2 Set up of high- and low-oxygen cell culture systems

In the low-oxygen cell culture system, O2-, the 3D-micropatterned PDMS substrates were transferred into the wells of a standard tissue culture treated 24-well plate (Iwaki, Funabashi, Japan), whereas in the high-oxygen cell culture system, O2+, they were bound to a 1.5 mm-thick PDMS membrane assembled to a bottomless polycarbonate 24-well frame, according to a procedure that has been previously described (Nishikawa et al. 2008a, b; Evenou et al. 2010). Briefly, the PDMS membranes were prepared from a 10:1 mass ratio mixture of PDMS pre-polymer and curing agent which was poured inside a large polystyrene box, degassed in a vacuum chamber and cured for 2 hours at 75°C in an oven. PDMS 3D substrates were covalently bound to the PDMS membranes by applying oxygen plasma exposure for 7 s using the RIE machine. They were then assembled to the bottomless frames: PDMS membranes were fixed by screws between a polycarbonate plate and a stainless steel pierced blade.

Prior to cell culture, PDMS substrates were oxidized by a 5 s O2 plasma exposure using the RIE machine, then sterilized with ethanol 70 % for 45 min and rinsed three times with phosphate buffered saline (PBS; Sigma, Tokyo, Japan). PDMS surfaces were subsequently coated with 0.3 g/l type I collagen (Nitta Gelatin, Osaka, Japan) for one hour, then rinsed once with water and twice with PBS.

2.3 Cells and culture medium

A male Wistar rat aged 5–8 weeks was anaesthetized and its hepatocytes were isolated by the collagenase perfusion method (Seglen 1976). Viability was above 85 % as determined by trypan blue exclusion. All the wells were inoculated with the same number of hepatocytes: 4.0 × 105 hepatocytes/well (this corresponds to a cell density of 2.0 × 105 cells/cm2 with regards to smooth PDMS surfaces (2 cm2)). The cells were cultured for 14 days in an incubator at 37°C, in a humidified atmosphere of 5 % CO2 and 95 % air. Culture medium was replaced within 24 h after hepatocyte inoculation, and then every one or two days. Samples were collected and stored at 20°C for various biochemical assays. Morphologies of the cells were routinely observed through an inverted Microscope (Leica DM IRB).

Culture medium was Williams’ medium E (Sigma) supplemented with 4 mM hydroxyethylpiperazine-N’-2-ethanesulfonic acid (HEPES), 1 % non-essential amino acids (100× solution for MEM; Gibco), 1 % antibiotic and antimicotic solution (which contains penicillin, streptomycin and amphotericin B; Wako, Osaka, Japan), 0.5 mM ascorbic acid diphosphate (from magnesium salt n-hydrate; Wako), 10 ng/ml mouse epidermal growth factor (EGF; Takara, Otsu, Japan), 0.1 µM insulin (Takara) and 0.1 µM dexamethasone (Wako).

2.4 Vertical cross-section analysis

At the end of the culture period, cell-loaded PDMS substrates were fixed in paraformaldehyde (Sigma) solution at 4°C. After dehydration, they were embedded in paraffin for cross-section analysis: 5 µm-thin vertical sections were cut off, deparaffinised, hydrated, and then stained with hematoxylin and eosin (HE). Images were taken with a transmitted light microscope (Olympus BX50).

2.5 Measurements of liver-specific functions

Albumin secreted in the culture medium was measured by a sandwich-type enzyme-linked immunosorbent assay (ELISA) using anti-rat albumin goat antibody (MP Biomedicals-Cappel Products, Irvine, CA) and peroxidase-conjugated anti-rat albumin rabbit antibody (MP Biomedicals-Cappel Products) respectively as 1st and 2nd antibody.

To evaluate urea synthesis, the cells were incubated for two hours with the culture medium supplemented with 2 mM NH4Cl. Urea concentrations were then measured by the diacetyl monoxime method.

2.6 Statistical analysis

All the measurements were performed in triplicate wells in each culture and in each of the two independent hepatocyte isolations. A single factor ANOVA test was performed to assess the statistical significance of the results: differences with P < 0.05, P < 0.01 and P < 0.001 were considered significant.

3 Results

3.1 Morphological changes of the cells under the different culture conditions

Under both O2 and O2+ conditions, primary rat hepatocytes adhered well onto collagen-coated PDMS surfaces; however, whereas in the O2 system the cells almost did not spread, neither on smooth nor on 3D-microstructured PDMS surfaces, but remained round throughout the culture period (Fig. 2(a, g and l)), from Day 1, under O2+ conditions, the cells spread and tended to stretch and to bridge across the pillars (Fig. 2(b-e)). On the one hand, hepatocytes cultured in the O2 system did not organize themselves during the culture but remained independent. Moreover, although some cells started to die from Day 2, most of them remained attached to the surface. On the other hand, in the O2+ system, between Day 5 and Day 14, hepatocytes progressively arranged into different kinds of 3D aggregates, from multilayered canopy-like structures on top of the pillars (Fig. 2(h-j)), to spherical aggregates or spheroids (Fig. 2(k, m-p)). The more the pillars were spaced out (sp100, sp200), the more the cells tended to organize themselves at first into a stretched and stringy network. Since such structures appeared quite unstable, subsequently they tended to break more easily, thus leading to a higher cellular detachment and, on Day 14, to a small number of large 3D aggregates (Fig. 2(k, o and p)). In contrast, as the spacing decreased between the pillars (sp20, sp50), the cells organized into a larger number of more compact and smaller 3D aggregates or spheroids (Fig. 2(m and n)). Figure 3 shows the aggregate size distribution for each condition on pillared substrates in the O2+ system: results indicated that the distribution varied with the spacing between adjacent pillars. The size range was wider for 3D substrates with the 200 µm-spacing (sp200), where particularly large aggregates were formed, with about 30 % of them measuring more than 250 µm in diameter. In contrast, a higher number of smaller aggregates were formed on sp20 and sp50 microstructures: 71 % and 51 % respectively measured less than 150 µm in diameter, and 88 % and 90 % respectively were less than 200 µm in size.
https://static-content.springer.com/image/art%3A10.1007%2Fs10544-010-9403-8/MediaObjects/10544_2010_9403_Fig2_HTML.gif
Fig. 2

Morphological changes of primary rat hepatocyte cultures under O2- (a, g and l) and O2+ conditions (b-f, h-k and m-p), on the different PDMS substrates: flat surfaces (F) vs. 3D-pillared structures, the spacing between two adjacent pillars (sp) varying between 20, 50, 100 and 200 µm. Pictures were taken at Day 1, Day 5 and Day 14 at different magnifications: all scale bars indicate 100 µm

https://static-content.springer.com/image/art%3A10.1007%2Fs10544-010-9403-8/MediaObjects/10544_2010_9403_Fig3_HTML.gif
Fig. 3

Aggregate size distribution of rat hepatocyte cultures on 3D-pillared structures with 20, 50, 100 and 200 µm-spacing (sp) between adjacent pillars and under O2+ conditions (Day 14)

When cultured on smooth PDMS surfaces under O2+ conditions (Fig. 2(f)), hepatocytes tended to arrange themselves at first into cellular monolayers which almost detached from the surfaces after Day 5. However, they sometimes did not detach completely, but wrapped themselves and remained anchored to the wells, thus explaining the very low reproducibility of functional assays as discussed later.

3.2 Histological analysis of vertical cross-sections on Day 14

Histological analysis of vertical cross-sections of rat hepatocyte cultures on 3D-microstructured PDMS substrates under O2+ conditions showed that the cells were organized in 3D around the pillars (Fig. 4). They had arranged into either spheroid-like 3D aggregates which were firmly anchored to the pillars (Fig. 4(a, c and f)), or multilayered canopy-like structures on top of the pillars (Fig. 4(b, d, e and g)). In the first case, the size of the aggregates strongly depended on the spacing between the pillars, the larger ones being formed on sp200 substrates (Fig. 4(f)). Moreover, when organized into such spherical 3D aggregates, individual cells exhibited a cubical in vivo-like morphology. On the other hand, in order to maintain the cohesion of the canopy-like 3D structures, the cells exhibited a more stretched shape as the distance increased between the pillars (Fig. 4(e and g)). Finally, within the different 3D structures, although some cells located in the centre of the aggregates lacked nuclei, most hepatocytes seemed to be alive with a distinct round nucleus.
https://static-content.springer.com/image/art%3A10.1007%2Fs10544-010-9403-8/MediaObjects/10544_2010_9403_Fig4_HTML.gif
Fig. 4

Hematoxylin and eosin staining of vertical cross-sections of primary rat hepatocyte cultures on the different PDMS 3D substrates under O2+ conditions (Day 14). Scale bars indicate 50 µm

3.3 Liver-specific functions

The metabolic activity of hepatocyte cultures was assessed by measuring the amounts of albumin secreted in the culture medium. Time course changes of albumin secretion rates in each of the two independent hepatocyte isolations, here named as 1st and 2nd runs, are reported separately (Fig. 5(a and b)): indeed, although the general trends were almost similar between the two runs, quantitative measurements did not fit well. First of all, hepatocytes exhibited much higher albumin secretion rates in O2+ conditions than in O2 conditions: in the latter case, from Day 2, the average production always remained less than 1.5 µg/well/day for both smooth and 3D-micropatterned PDMS substrates, albumin being even no longer detectable in the culture medium for some experiments in O2 conditions (left charts on Fig. 5(a and b)).
https://static-content.springer.com/image/art%3A10.1007%2Fs10544-010-9403-8/MediaObjects/10544_2010_9403_Fig5_HTML.gif
Fig. 5

Time course changes of per-well based albumin secretion rates of rat hepatocyte cultures under both O2- (charts on the left) and O2+ conditions (charts on the right), and on the different PDMS substrates: flat surfaces (F) vs. 3D-pillared structures, the spacing between two adjacent pillars (sp) varying between 20, 50, 100 and 200 µm. Data for each of the two independent hepatocyte isolations (1st run and 2nd run) are presented separately (a and b). The scales of the vertical axes are different for the results obtained in O2- and O2+ conditions. Columns and error bars represent the means±SD of three wells in each of the two independent experiments

Under O2+ conditions, from Day 5 then throughout the culture period, it appeared that in both two independent experiments, hepatocytes cultured on 3D substrates with the 20 µm-spacing (sp20) exhibited markedly higher albumin secretion rates than hepatocytes cultured either onto sp50, sp100 and sp200 3D substrates or on smooth PDMS surfaces (F) (P < 0.001 except for the results on Day 9 in the 2nd run) (Fig. 5(a and b), Table 1). When comparing these latter four surfaces, it appeared that in the first run, the cells cultured on the pillared substrates (sp50, sp100 and sp200) secreted significantly higher amounts of albumin than those cultured on the smooth PDMS surfaces (P < 0.001) (Fig. 5(a), results for the 1st run in Table 2); however, in the 2nd run, such differences were not always significant (notably on Day 5 between F and sp200, on Day 9, and on Day 13 between F and sp100, and F and sp200 ) (Fig. (5b), results for the 2nd run in Table 2). As already mentioned in paragraph 3.1, the reason for this inconsistency between the two runs is the very low stability of the cellular monolayers formed on the smooth PDMS membranes: whereas in the 1st run they had completely detached from the surfaces on Day 5, in the 2nd run, although the cell layers tended to detach, they wrapped themselves and remained anchored to one side of the well. These observations suggest that the incorporation of such pillared microstructures is favorable for obtaining stable cultures, and accordingly more reproducible results.
Table 1

Statistical significance (single factor ANOVA) of the results of albumin secretion measurements, in each of the two independent runs: were considered the differences under O2+ conditions, between rat hepatocyte cultures on PDMS 3D structures with the 20 µm-spacing (O2+,sp20) and the four other conditions (O2+,sp50 ; O2+,sp100 ; O2+,sp200 and O2+,F)

 

1st run

2nd run

O2+, sp20>

Day2

Day5

Day7

Day9

Day11

Day13

Day2

Day5

Day7

Day9

Day11

Day13

O2+, sp50

**

***

***

***

***

***

**

***

***

<,P = 0.705

***

***

O2+, sp100

***

***

***

***

***

***

***

***

***

<,P = 0.086

***

***

O2+, sp200

***

***

***

***

***

***

**

***

***

<,**

***

***

O2+, F

***

***

***

***

***

***

***

***

***

<,*

***

***

Differences with P < 0.05 (*), P < 0.01 (**) and P < 0.001 (***) were considered significant

Table 2

Statistical significance (single factor ANOVA) of the results of albumin secretion measurements, in each of the two independent runs: were considered the differences under O2+ conditions, between rat hepatocyte cultures on smooth PDMS surfaces (O2+,F) and the four other conditions on 3D-microstructures (O2+,sp20 ; O2+,sp50 ; O2+,sp100 and O2+,sp200)

 

1st run

2nd run

O2+, F<

Day2

Day5

Day7

Day9

Day11

Day13

Day2

Day5

Day7

Day9

Day11

Day13

O2+, sp20

***

***

***

***

***

***

***

***

***

>,*

***

***

O2+, sp50

**

***

***

***

***

***

***

***

***

>, P = 0.080

***

*

O2+, sp100

P = 0.319

***

***

***

***

***

***

***

***

>, P = 0.030

**

P = 0.407

O2sp200

*

***

***

***

***

***

**

P = 0.062

***

P = 0.493

***

P = 0.609

Differences with P < 0.05 (*), P < 0.01 (**) and P < 0.001 (***) were considered significant

As for the albumin secretion, the results of urea synthesis measurements in each of the two independent hepatocyte isolations are reported separately (Fig. 6(a and b)). Overall, the results suggested less dependency of this function on both oxygenation and the geometry of PDMS patterns than the albumin secretion, and a sharp decrease was observed in most cultures between Day 5 and Day 14. However, when focusing on the results on pillared structures on Day 5, it appeared that urea synthesis was far much improved when the cells were subjected to direct oxygen supply through the PDMS membranes (O2+) (P consistently < 0.001 in the first run, and in the 2nd run except for the differences between O2+,sp200 and {O2-,sp50; O2-,sp100; O2-,sp200} with P < 0.05, and between O2+,sp200 and O2-,F with P < 0.01) (Table 3). Moreover, in the latter case, although we could not observe any significant differences between the cultures on pillared substrates with the different spacings (sp20 to sp200), all the 3D cultures exhibited much higher rates than those on smooth PDMS surfaces (F) (P consistently < 0.001 in the first run, and in the 2nd run except for the differences between F and sp200 with P < 0.01) (Table 4).
https://static-content.springer.com/image/art%3A10.1007%2Fs10544-010-9403-8/MediaObjects/10544_2010_9403_Fig6_HTML.gif
Fig. 6

Urea synthesis rates on Day 5 and Day 14 of rat hepatocyte cultures under both O2- and O2+ conditions and on the different PDMS substrates: flat surfaces (F) vs. 3D-pillared structures, the spacing between two adjacent pillars (sp) varying between 20, 50, 100 and 200 µm. Data for each of the two independent hepatocyte isolations (1st run and 2nd run) are presented separately (a and b). Columns and error bars represent the means±SD of three wells in each of the two independent experiments

Table 3

Statistical significance (single factor ANOVA) of the results of urea synthesis measurements on Day5, in each of the two independent runs: were considered the differences between rat hepatocyte cultures in O2+ and in O2- conditions, on the different PDMS substrates (smooth surfaces and 3D-pillared microstructures)

 

1st run

2nd run

O2+,sp20>

O2+,sp50>

O2+,sp100>

O2+,sp200>

O2+,sp20>

O2+,sp50>

O2+,sp100>

O2+,sp200>

O2-,sp20>

***

***

***

***

***

***

***

***

O2-,sp50>

***

***

***

***

***

***

***

*

O2-,sp100>

***

***

***

***

***

***

***

*

O2-,sp200>

***

***

***

***

***

***

***

*

O2-,F

***

***

***

***

***

***

***

**

Differences with P < 0.05 (*), P < 0.01 (**) and P < 0.001 (***) were considered significant

Table 4

Statistical significance (single factor ANOVA) of the results of urea synthesis measurements on Day5, under O2+ conditions, in each of the two independent runs: were considered the differences between rat hepatocyte cultures on smooth PDMS substrates (O2+,F) and the four other conditions on 3D-microstructures (O2+,sp20 ; O2+,sp50 ; O2+,sp100 and O2+,sp200)

O2+,F<

1st run

2nd run

O2+,sp20

***

***

O2+,sp50

***

***

O2+,sp100

***

***

O2+,sp200

***

**

Differences with P < 0.05 (*), P < 0.01 (**) and P < 0.001 (***) were considered significant

4 Discussion

In order to maintain hepatocytes in culture viable and in a highly-differentiated state, it appears crucial to have them organized into a tissue-like 3D structure which promotes extensive cell-cell contacts, but also to be adequately supplied with nutrients and particularly with oxygen. Here, we report a simple methodology satisfying these two fundamental but conflicting issues: primary rat hepatocytes were cultured on various 3D-microstructured PDMS substrates on a highly gas-permeable PDMS membrane in a practical microplate format. We showed that the combination of direct oxygen supply and 3D pillars induced the spontaneous organization of the cultured hepatocytes into stably-anchored and functional 3D multicellular aggregates, whose size and number depend on the dimensions of the pattern.

First, it appeared that direct oxygenation through an oxygen-permeable PDMS membrane enables both a quicker and a stronger attachment of the cells onto the different kinds of PDMS substrates. Indeed, during the first 24 hours of culture, hepatocytes spread and started to organize themselves around the pillars into 3D cellular networks. On the other hand, the cells cultivated in conventional tissue culture plates did not spread nor organized themselves, but most of them even died within one week. Indeed, it has been previously reported that hepatocytes required a lot of energy to attach to and spread on the surfaces (Rotem et al. 1992, 1994) and that cells with such a high oxygen demand were rapidly oxygen-deprived under conventional culture conditions (Nahmias et al. 2006; Camp and Capitano 2007; Metzen et al. 1995). In addition, in order to increase cell-cell contacts, hepatocytes were here cultured at a high inoculum density. In contrast, the cells cultured in the O2+ system were continuously supplied with oxygen by diffusion through the PDMS: oxygen diffusion coefficient in PDMS was reported as 3.4 × 10-9 m2/s (Merkel et al. 2000) against 2.1 × 10-9 m2/s in culture medium (Mehta et al. 2007). Moreover, the particularly high oxygen concentration in PDMS of 2 mM is 10 times that in aqueous media (Shiku et al. 2006). By using the highly proliferative human hepatoma Hep G2 cell line, we recently showed that such a PDMS membrane-based cell culture system provided sufficient oxygen to such kind of high-oxygen consuming cells (Evenou et al. 2010).

Moreover, it appeared that 3D structures and direct oxygen supply had synergistic effects: under such conditions, the cultures were better stabilized and 3D-organized. So far, most of the methods that have been developed in order to induce hepatocyte spheroids formation have led to freely-floating spheroids. Indeed, spheroid formation is generally promoted on surfaces that provide a weak cell-adherent environment (Koide et al. 1990). However, there is a growing interest in maintaining the spheroids stably-attached onto the substrates for use in microfluidic devices or just for practical reasons regarding sampling and periodic culture medium replacement in a static culture. Moreover, it appears important to maintain to some extent cell-substrate interactions and notably with some extra-cellular matrix components. Stably-attached spheroids from either primary hepatocytes or Hep G2 cells were formed according to quite laborious procedures which used other cell types such as 3T3 fibroblasts and substrates with particular cell-adhesive properties to stabilize the pre-formed hepatocyte aggregates (Lu et al. 2005; Fukuda et al. 2006). Hepatocyte spheroids have yet been formed onto either hydrophilic or hydrophobic non-coated PDMS surfaces, but they were reported to detach during the culture (Nakazawa et al. 2006, 2009); using surface-modification techniques, we recently reported stably-attached hepatocyte hemispheroid formation on PDMS membranes with covalently-immobilized collagen (Nishikawa et al. 2008a). In the present study, although we just applied a simple collagen coating, it appeared that topographical features such as PDMS macropillars together with a high oxygen environment also promoted stably-attached 3D multicellular aggregate formation. Whereas hepatocytes cultured on smooth PDMS substrates under O2+ conditions spread and formed monolayers which rapidly detached from the surfaces after a few days, the cells spontaneously organized themselves around the pillars into stable 3D structures. The less the pillars were spaced out, the more the cells tended to organize into compact 3D aggregates, whereas for the larger spacings, they displayed a more stretched configuration. At first, they organized into multilayered canopy-like 3D structures on top of the pillars. Then, depending on the spacing between them, subsequent cell retraction resulted in the formation of 3D multicellular clumps or spheroids which were firmly anchored to the pillars.

However, the formation of such compact 3D multicellular aggregates may raise the problem of mass transfer limitation of nutrients and particularly of oxygen. Thus, since nutrients are supplied by diffusion through the cell layers, large spheroids often exhibit a necrotic centre. In the present study, it appeared that cultured hepatocytes preferably organized around the pillars which served as oxygen providers; accordingly, one could expect a higher viability and function. Indeed, histological analysis of vertical cross-sections indicated that most of the cells within the 3D aggregates looked alive. Moreover, according to previous studies with both theoretical and experimental approaches, no oxygen limitation should occur in spheroids up to 100 µm in diameter, but it has been hypothesized that bile accumulation might then be responsible for cell death at the centre of the spheroid (Glicklis et al. 2004; Dvir-Ginzberg et al. 2004). In the present study, the size of the aggregates formed depended on the spacing between the pillars (Figs. 3 and 4), the larger ones being obtained on PDMS substrates with the 200 µm-spacing (30 % were more than 250 µm in diameter). In contrast, the aggregates formed on 3D substrates with the 20 µm- and 50 µm-spacing were in majority less than 150 µm in size. In order to avoid restrictions related to mass transfer, we may thus be able to control the size of the spheroids just by adapting the geometry of the PDMS patterns.

As in our previous studies on liver cell culture on similar PDMS membranes (Nishikawa et al. 2008a, b; Evenou et al. 2010), the high dependency of albumin secretion on oxygen was here again confirmed, with markedly greater amounts of albumin being secreted in O2+ conditions (Fig. 5). Indeed, the biosynthesis of albumin appears to be mainly expressed under high oxygen concentrations (Camp and Capitano 2007). Moreover, among the different cultures on PDMS membrane, hepatocyte 3D cultures obtained on 20 µm-spacing PDMS substrates displayed the highest secretion rates on a per well basis, with a better maintenance. Higher function can first be explained by the lower cellular detachment on those substrates. Indeed, as the spacing increased between the pillars up to 200 µm, the multicellular canopy-like 3D structures which were formed between them may have been prone to physical tensions as assessed by the stretched morphology of the constituting cells (Fig. 4(g)), thus consequently breaking and resulting in a higher cellular detachment. In contrast, hepatocytes on sp20 substrates exhibited a less-stretched but more in vivo-like cuboidal morphology. In addition, as the aggregates formed were smaller than for the other conditions on PDMS 3D structures (Fig. 3), mass transfer limitations of both nutrients and wastes should have been reduced and anyway, should have been less than in the case of the very large aggregates (those more than 200 µm and up to 350 µm in size) formed on sp100 and sp200 3D substrates.

5 Conclusions

Here we demonstrated that the combination of 3D PDMS pillars and direct oxygenation through a gas-permeable PDMS membrane allows the spontaneous organization of cultured rat hepatocytes into 3D multicellular aggregates which are firmly anchored to the pillars. Accordingly, such 3D cultures exhibited higher albumin secretion and urea synthesis rates. The highest amounts of albumin were secreted in the case where the culture was better stabilized, with hepatocytes forming the smaller aggregates, on 3D microstructures with a 20 µm-spacing between the pillars. Such a simple cell culture system, where various favourable conditions for liver cell culture are simultaneously realized, is likely to be included in a new drug/chemical screening system in a practical microplate format, or used in microfluidic applications.

Acknowledgments

We thank N. Yamamoto for her assistance in rat hepatocyte isolations. We thank Dr. S. Couderc for her assistance in SEM analyses and Pr. H. Toshiyoshi for allowing us to use FESEM equipment. We thank Saipaso Research Center (Tokyo, Japan) for preparing histological samples. This work was performed in the framework of LIMMS (Laboratory for Integrated Micro-Mechatronic Systems), a joint laboratory between the CNRS (Centre National de la Recherche Scientifique) and IIS (Institute of Industrial Science), the University of Tokyo. It was supported by the Japanese Society for the Promotion of Science (JSPS).

Copyright information

© Springer Science+Business Media, LLC 2010