Epithelial morphogenesis of MDCK cells in three-dimensional collagen culture is modulated by interleukin-8

Abstract

Epithelial morphogenesis is dependent upon a variety of factors, many of which involve complex interactions between cells and their surrounding environments. We analyzed the patterns of differential gene expression associated with Madin-Darby canine kidney (MDCK) renal epithelial cells grown within a collagen gel in three-dimensional (3D) culture compared with those grown atop a collagen gel in two-dimensional (2D) culture. Under these conditions, MDCK cells spontaneously formed either hollow spherical cysts or flat monolayer sheets, respectively. Microarray analysis of gene expression revealed a twofold or greater expression difference in 732 gene sets from MDCK cysts compared with monolayers (false discovery rate or FDR-adjusted P values <0.05). Interleukin-8 (IL-8) was reproducibly found to be among the genes whose expression was most dramatically upregulated, and this behavior was verified through real-time PCR analysis. The level of IL-8 protein expression was significantly increased in 3D MDCK cultures compared with that detected in cells in 2D culture. Hepatocyte growth factor (HGF) induces MDCK cells in 3D culture to form linear tubule-like structures. We found that HGF stimulation caused MDCK cells in 3D culture to decrease the expression of IL-8 at both the mRNA and protein levels. Furthermore, the addition of recombinant IL-8 to HGF-stimulated 3D MDCK cultures was sufficient to partially reverse the tubulogenic effects of HGF, resulting in the formation of cystic structures. These data suggest that IL-8 participates in the formation of cystic structures by MDCK cells in 3D culture and that HGF may stimulate tubulogenesis through the suppression of IL-8.

in tissues such as the lungs, intestines, and kidneys, epithelial cells form barriers between an organism and its external milieu. The surface membranes of these epithelial cells are exposed to disparate environments and are divided into discrete apical and basolateral domains (4, 8, 37, 39). These membrane domains are established and maintained through the directed trafficking and selective retention of cellular components. In addition to this capacity to generate polarized plasma membrane domains with distinct biochemical and physiological properties, epithelial cells must also be able to organize themselves into complex multicellular architectures. The formation and maintenance of these structures is dependent both upon intercellular interactions between epithelial cells and upon cell-matrix interactions.

Interactions between an individual cell and its surrounding environment are essential in the activation of signaling networks that subsequently regulate polarity and the determination of epithelial structural morphology (6, 7, 9, 40, 48). As epithelial cells establish polarity and interactions with their environment they can begin to function together to generate organized multicellular structures. This property is exemplified by comparing the behaviors of epithelial cell lines, such as Madin-Darby canine kidney (MDCK) cells, grown in two-dimensional (2D) vs. three-dimensional (3D) culture conditions. MDCK cells form flat monolayer sheets when grown atop tissue culture plastic, permeable filter supports, and various extracellular matrix (ECM) substrates in traditional 2D culture conditions. When grown in 3D culture, in which cells are embedded within an ECM substrate, MDCK cells spontaneously form cysts and can be induced to form tubules (59, 60). These structures constitute the primary building blocks in the development and growth of epithelial tissues in vivo (13, 25, 43). The application of 3D culture systems has facilitated the identification of factors that influence epithelial cystogenesis and tubulogenesis (14, 32, 33, 50, 52, 63). In both of these 3D geometries, cellular monolayers can encapsulate a central fluid-filled lumen. Whereas cysts are generally spherical, tubules are usually defined as cylindrical structures whose length exceeds their width. In the presence of hepatocyte growth factor (HGF), MDCK and other epithelial cell types grown under 3D in vitro culture conditions are stimulated to undergo tubulogenesis (32, 33, 50, 52). The mechanisms controlling the regulation of cystogenesis and tubulogenesis appear to be complex and dependent on both the nature and the composition of the growth environment.

Chemokines and their receptors have recently been discovered to play a significant part in regulating epithelial cell morphology (22, 35, 55–57). In 1987, Yoshimura and colleagues described interleukin-8 (IL-8) as the first inflammatory cytokine (65). IL-8 is generally thought to be secreted by mononuclear cells of the immune system leading to the activation of the signaling cascade that triggers the chemotaxis of neutrophils (12, 31). In addition, however, IL-8 has been found to be secreted by, and to target to, a vast number of cell types, including epithelial cells such as those of the intestines, lungs, and kidneys as well as retinal pigment epithelial cells and hepatocytes (15, 23, 24, 55, 61, 66). Recent studies have highlighted the importance of chemokines in a variety of cellular morphogenetic processes and diseases of morphogenesis, including angiogenesis, tumorigenesis and metastasis, cystic fibrosis, and polycystic kidney disease (PKD) (2, 11, 17, 29, 34, 36, 56–58, 62, 64). Along these lines, it has been shown that exposing murine-derived inner medullary collecting duct (mIMCD) cells to HGF induces the expression of keratinocyte chemoattractant (KC), a murine functional homolog of IL-8. This expression of KC appeared to stimulate cell proliferation and to downregulate branching morphogenesis and cell migration (57). A gene encoding IL-8 is absent from the genomes of mice or rats and was most likely lost in a common ancestor (31), but KC appears to mediate at least some of the functions of IL-8 in these organisms (3).

To gain further insight into epithelial morphogenesis and the mechanisms regulating spontaneous cyst formation in MDCK epithelial cells, we explored the influence of culture geometry on gene expression patterns. We performed a comparative analysis of gene expression between MDCK cells grown in two distinct morphologies: flat sheets and cysts. A large number of genes were found to be differentially regulated between these two conditions, and prominent among this collection were IL-8 and other genes of immune and inflammatory response pathways, in addition to genes integrally involved in cell-matrix interactions and ECM remodeling. Our data suggest that HGF may induce tubulogenesis rather than cystogenesis at least in part by virtue of its ability to suppress IL-8 production.

MATERIALS AND METHODS

2D and 3D MDCK collagen culture and treatment with HGF and anti-IL-8.

The comparative gene microarray analysis was carried out with MDCK type II cells stably transfected with an empty Tet-off construct supplied by Dr. Richard Lifton. Tet-off MDCK type II cells were grown in DMEM supplemented with 10% FBS, 1% penicillin/streptomycin, 1% l-glutamine, and G418 (Invitrogen, Carlsbad, CA). All subsequent experiments employed untransfected MDCK cells, which were grown in α-minimal essential medium (Invitrogen) supplemented with 10% FBS, 1% penicillin/streptomycin, and 1% l-glutamine. Cells were maintained at 37°C with 5% CO₂ atmosphere in a humidified incubator.

For 2D monolayer cultures, collagen gels were prepared at a 1:2 dilution with growth medium and liquid rat tail type I collagen (Becton-Dickinson, Franklin Lakes, NJ) and incubated overnight in the wells of a tissue culture plate to assure complete solidification. The following day, cells were plated atop presolidified collagen gels. The classic 3D MDCK cyst culture conditions were reproduced by adding cells and media to liquid rat tail type I collagen (Becton-Dickinson) at a 1:2 dilution, and this mixture was then deposited into the wells of tissue culture plates and incubated for ∼30 min or until solidified. To maximize the yield of total RNA for the Affymetrix Microarray analysis, a cell density of ∼8.0 × 10⁵ and a total gel volume of 3 ml were plated in six-well plates and bathed in 5 ml of culture media. For all remaining experiments, the size of individual collagen gels was reduced to 500 μl, and they were plated in 24-well plates with a cell density of 1.6 × 10⁵ and bathed in 1 ml of media. Samples were grown for 8 days with media collection and changes taking place on days 3, 6, and 8. In some experiments, the media was supplemented with 40 ng/ml of HGF (Sigma-Aldrich, St. Louis, MO) to stimulate MDCK cells grown in 3D culture to form tubule-like structures. In addition, 3D cultures were grown in the presence of polyclonal goat anti-canine IL-8 (0.8 μg/ml) (R&D Systems, Minneapolis, MN) to assess the consequences for cyst development of neutralizing the activity of secreted IL-8.

Immunofluorescence in two dimensions and three dimensions.

Cells grown on 2D collagen gels were washed two times with cold PBS with 1 mM MgCl₂ and 100 μM CaCl₂ (PBS2+) and fixed for 30 min in 4% paraformaldehyde (PFA) at room temperature. Cells were subsequently washed with PBS2+ and then permeabilized with permeabilization buffer (0.3% Triton X-100/0.1% BSA in PBS). Cells were incubated for 1 h with anti-GP58 (gift of Ira Mellman), which detects the Na-K-ATPase β-subunit (10), followed by incubation with rhodamine-conjugated anti-mouse IgG (Molecular Probes). Immunofluorescence on 3D cysts was performed essentially as previously described by O'Brien et al. (42). Briefly cysts were fixed with 4% PFA for 30 min, permeabilized in 0.025% saponin for 30 min, quenched for 10 min with 7 mM NH₄Cl, 20 mM glycine in PBS2+ (pH 8), and then incubated overnight with anti-gp58 antibody. Cysts were washed for 2–4 h with saponin in PBS2+ and incubated with secondary antibodies overnight. Cells were visualized using confocal microscopy (Zeiss LSM 780) at ×40 and ×63. Images were processed using an LSM Image Viewer (Carl Zeiss).

Total RNA isolation.

2D and 3D MDCK collagen cultures were digested with a solution of 0.01% collagenase (Sigma-Aldrich) and RNase OUT (Invitrogen) in phosphate-buffered saline solution (PBS) at 4°C for ∼1 h. A 5-ml volume of collagenase mixture was added to the larger gels used for microarray analysis, whereas 1 ml was used for all other smaller gels. Cells were collected after two PBS washes with low-speed spins (500 rpm) at 4°C. Total RNA was isolated using the TRIzol reagent (Invitrogen) according to the manufacturer's instructions, employing the standard suggested volumes of solutions. To summarize briefly, samples were incubated with 1 ml of TRIzol reagent for 5 min, after which 0.2 ml chloroform was added to samples and they were spun at 12,000 g for 15 min at 4°C to allow for phase separation. The upper aqueous phase was isolated and incubated with 0.5 ml isopropanol for 10 min. Samples were then centrifuged at 12,000 g for 10 min at 4°C. Supernatant was removed, and the pellet was washed with 1 ml 75% RNase-free ethanol, vortexed, and spun at 7,500 g at 4°C for 5 min. Supernatant was again removed, and the pellet was dried at room temperature for 10 min. Pellets were redissolved in 20 μl RNase-free water and incubated at 55°C for 10 min. Purification of total RNA was accomplished with the RNeasy kit (Qiagen, Germantown, MD) following the manufacturer's protocol for RNA Cleanup at room temperature.

Affymetrix GeneChip microarray hybridization and analysis.

Total RNA samples were submitted to the W.M. Keck Facility at Yale for hybridization to Affymetrix GeneChip Canine Genome 2.0 arrays (Affymetrix, Santa Clara, CA). Because of the limited yield of total RNA, four individual RNA samples were pooled for hybridization to each microarray. Six biological replicates were performed for each growth condition. A minimum of 10 μg of total RNA per individual sample was submitted for analysis to assure adequate quantity and quality. Before being employed in the array analysis, the quality of total RNA was assessed by electrophoresis on an Agilent Bioanalyzer as well as by assuring that the ratio of absorbance at 260 nm to that at 280 nm was at least 1.9. Four individual samples were pooled, and 5 μg of pooled total RNA were used for each individual array. The remaining procedures were performed according to the protocols outlined in the Affymetrix technical manual. To summarize briefly, double-stranded cDNA was created, and a one-cycle target labeling kit (Affymetrix) was used to synthesize biotin-labeled cRNA. Labeled cRNA was purified with the GeneChip Cleanup Module and fragmented through incubation in a buffer comprised of 40 mM Tris-acetate, 100 mM potassium acetate, and 30 mM magnesium acetate at pH 8.1 for 35 min at 94°C. Fragments of 35–200 bases were hybridized to GeneChip arrays at 45°C for 16 h, after which the arrays were washed with an Affymetrix fluidics station, stained with streptavidin-phycoerythrin, and scanned with an Affymetrix GeneChip Scanner 3000.

The scanned output files from each array were inspected for hybridization artifacts and individually analyzed with Affymetrix GeneChip Operating Software by the W.M. Keck Biostatistics Resource at Yale. Comparative analysis between MDCK cells grown in 2D and 3D collagen cultures was done with guanine and cytosine-robust multiarray analysis (GC-RMA), identifying genes that were differentially expressed by a factor of twofold or greater with false discovery rate (FDR) P values <0.05. FDR is a statistical measure that determines the expected proportion of false positives within a defined group of significant results (45). All spiked-in GeneChip controls had similar values with no fold change between the two conditions. A complete microarray expression dataset has been submitted to the National Center for Biotechnology Information's (NCBI) Gene Expression Omnibus (GEO) database (GEO submission GSE28381, NCBI tracking system 16045009). Signaling pathway analysis was obtained by importing the dataset into MetaCore.

Real-time PCR analysis.

Sample concentrations of total RNA were determined by ultraviolet-visible spectrometry at a 1:100 dilution in 200 μl RNase-free water. A quantity of 0.4 μg of total RNA/sample was reverse transcribed following the manufacturer's recommended protocol for first-strand cDNA synthesis using SuperScript II RT (Invitrogen). Each sample of total RNA was combined with 1 μl oligo(dT) (W.M. Keck Facility at Yale Oligonucleotide Synthesis Service) and 1 μl dNTPs (Invitrogen) and brought to a final volume of 12 μl with RNase-free water. Samples were incubated at 65°C for 5 min and then placed on ice for 2 min. A volume of 7 μl of master mix (4 μl 5× First Strand Buffer, 2 μl 0.1 M dithiothreitol, and 1 μl RNase Out; all from Invitrogen) was added to samples, and they were incubated at 42°C for 2 min. A 1-μl volume of Superscript II (Invitrogen) was then added, and the samples were incubated at 42°C for 50 min followed by a 15-min incubation at 70°C. In 96-well plates, 1 μl of the synthesized first-strand cDNA was combined with target gene primers at a final concentration of 10 μM and QuantiTect SYBR Green (Qiagen) master mix at a total volume of 25 μl/well and run in duplicate. Canine β-actin PCR primers were designed using the Primer3 website and were used to measure the expression of actin, which served as the control gene. The β-actin forward primer sequence was 5′-ggcatcctgaccctcaagta-3′ and the reverse sequence was 3′-aaggcgtacccctcgtagat-5′. The sequences for canine IL-8 PCR primers were previously published (46). Real-time PCR analysis was performed with a Stratagene Mx 3000P machine. After an initial denaturation step at 95°C for 5 min, 40 cycles consisting of a denaturation step at 95°C for 1 min, an annealing step at 57°C for 30 s, and an extension step at 72°C for 1 min were performed. Upon completion, the temperature was ramped from 55°C to 95°C to generate a melting curve. The efficiency for the primer pairs of each target gene was calculated using the equation: [10^(1/−S) − 1] × 100%, where S equals the slope of the linear regression trendline. The trendline was determined from the count threshold (C_t) values plotted against the serial dilution of known concentrations of cDNA. The equation of the line and the correlation coefficients were also determined. C_t values obtained through real-time PCR were used to assess the relative abundance of IL-8 in experimental conditions compared with control conditions with normalization to the control gene, canine β-actin, using the equation 2^ΔΔC_t. Average C_t values for IL-8 were normalized to average β-actin C_t values, representing the first ΔC_t. The normalized experimental condition (3D MDCK culture) was then subtracted from the normalized control condition (2D MDCK culture) to obtain the value of ΔΔC_t. Student's t-test was used for statistical analysis. Replicates were done in triplicate for each experimental condition.

Canine IL-8 luminex assay.

A custom-ordered canine IL-8 cytokine/chemokine Lincoplex kit (Millipore, Billerica, MA) was used based on company specifications to quantify the amount of IL-8 in media samples collected from 2D and 3D MDCK collagen cultures. Samples of media (1 ml) were collected, spun at 12,000 rpm to remove cell debris, and stored at −80°C. To summarize the assay briefly, 50-μl media samples (mixed at a 1:2 dilution with the supplied Luminex buffer) and the supplied canine IL-8 standard (at a final volume of 100 μl) were mixed with 25 μl of the supplied fluorescein isothiocyanate-conjugated beads specific for canine IL-8 on the supplied 96-well microfiltration plates and incubated overnight at 4°C with agitation. After two washes with 200 μl Luminex buffer, 25 μl of supplied detection antibody cocktail were added to wells and incubated for 1 h at room temperature with agitation. A 25-μl volume of supplied streptavidin-phycoerythrin solution was used to label the samples for 30 min at room temperature with agitation. A series of three washes with 200 μl Luminex buffer was followed by incubation with 125 μl Luminex buffer at room temperature for 5 min. Samples were analyzed on a Luminex instrument with BioPlex Manager 4.0 software. Based on the supplied canine IL-8 standard, a standard curve with the best-fitting linear trendline was generated. The equation for the line was then used to estimate IL-8 protein concentrations in samples done in triplicate. Statistical analysis was done using Student's t-test.

Sandwich enzyme-linked immunosorbent assay.

A checkerboard titration was performed to determine the optimal conditions for the generation of the most linear standard curve. Ninety-six-well plates were coated overnight at 4°C with 2.0 μg/ml canine IL-8 capture antibody (clone 258901; R&D Systems) in PBS. All steps were carried out at room temperature, and solutions were added at a final volume of 100 μl/well unless otherwise noted. Plates were washed three times with 0.05% Tween 20 (Roche) in PBS and then blocked with 200 μl/well of 5% BSA for 2 h. After another series of washes, a serial dilution of recombinant canine IL-8 (RcIL-8; R&D Systems) was added as the standard. A volume of 1 ml of culture media was collected from cultured cells on days 3, 6, and 8. These samples were spun at 12,000 rpm to remove cell debris and then stored at −80°C until use in enzyme-linked immunosorbent assays (ELISA). Media samples at a 1:2 dilution in diluent buffer (1% BSA, 0.05% Tween 20 in PBS) were added to wells plated in duplicate for 2 h. After another series of washes, 0.5 μg/ml of biotinylated anti-canine IL-8 detection antibody (clone 258911; R&D Systems) in diluent buffer was applied for 2 h with agitation. Three washes were done before the addition of horseradish peroxidase-streptavidin conjugate (Invitrogen) at a 1:4,000 dilution for 30 min with agitation. After four washes, plates were dried, and 100 μl of chromogenic 1-Step Ultra TMB-ELISA substrate (Thermo Fisher Scientific, Waltham, MA) were added per well for 20–30 min to allow for the development of color. The chromogenic reaction was stopped with sulfuric acid, and the optical absorbance of each sample well was measured for 1 s at 450 nm with Wallac 1420 software on a Perkin Elmer Plate Reader. With the use of the average absorbance values for the known concentrations of RcIL-8, a standard curve was plotted, and a best-fitting linear trend line was generated. With the use of the equation for the line, the estimated IL-8 concentrations were calculated for experimental samples. All sample values were normalized to blank media samples. Three individual repetitions were performed for each experiment. Statistical analysis was done using Student's t-test.

Nuclear staining and image quantification.

3D MDCK collagen gels were fixed in 1 ml 4% PFA for 30 min at room temperature and washed three times with PBS. Samples were stained with 1 ml Hoechst 33342 (Invitrogen) at 1:2,500 dilution in PBS for ∼5 min. Gels were placed on microscope slides and briefly dried on 55°C heat block to remove excess fluid. Samples were then mounted with Vectashield anti-fade mounting media (Vector Laboratories, Burlingame, CA). Images of MDCK cells grown in 3D culture were captured using a Zeiss Axiophot microscope with a charge-coupled device camera with a ×20 objective. Before each image was obtained, structures within the randomly chosen field of view were examined to determine if they had four nuclei or more and to ensure that the plane of focus was selected to reveal optimal views of the cross sections. Images were calibrated in Image J (National Institutes of Health, Bethesda, MD). With use of the line tool, the width and length of multicellular MDCK structures were measured and labeled. The length-to-width ratio was calculated for each individual structure, and, for each experimental condition, 25 MDCK structures were averaged per experimental repetition. Three individual repetitions were performed for each experiment. Statistical analysis was done using Student's t-test.

RESULTS

Comparative analysis of differentially expressed genes in MDCK cells grown in 2D and 3D collagen culture.

MDCK renal epithelial cells cultured in 3D collagen gels spontaneously form spherical cystic structures (42, 59, 60). To compare the differential expression of genes between MDCK cells grown as flat monolayer sheets (2D culture) or cysts (3D culture), a microarray analysis was performed. For the 2D culture condition, MDCK cells were plated atop solidified collagen gels and allowed to form confluent monolayers that grew as flat sheets on the surfaces of the gels. For 3D culture, MDCK cells were embedded within collagen gels before gel solidification, where they proliferated to form spherical cystic structures. As revealed by the confocal images of immunocytochemical staining for Na-K-ATPase β-subunit that are depicted in Fig. 1A, the cells grown under both conditions are fully polarized. Under the 2D conditions, the cells form a flat monolayer epithelium with Na-K-ATPase staining limited to their lateral membranes. Cells grown under the 3D condition form cysts in which the Na-K-ATPase is excluded from their lumen-facing apical membrane domains. Total RNA was isolated from samples representing both epithelial morphologies and reverse transcribed to generate cDNA. Following biotinylation, fragments were hybridized to second-generation Affymetrix Canine Microarrays. The final dataset was obtained from six biological replicates for each of the two culture conditions. Genes exhibiting a twofold or greater differential expression in 3D culture, compared with 2D culture, with FDR adjusted P values <0.05 were scored as hits.

Fig. 1.

Immunofluorescence analysis of two-dimensional (2D) and three-dimensional (3D) cultures and quantification of interleukin-8 (IL-8) protein produced in 2D and 3D Madin-Darby canine kidney (MDCK) culture by Luminex assay. A: MDCK cells grown in 2D culture on top of a collagen gel (left) or in 3D culture embedded in a collagen gel (right) were stained with an antibody directed against the β-subunit of the Na-K-ATPase. In both culture conditions, immunostaining is limited to the basolateral domains of the epithelial cells' plasma membranes. B: a customized Luminex assay was performed using media samples collected from MDCK cells grown in 2D culture (dark gray) and 3D culture (light gray). A significant increase in the level of IL-8 protein in culture media was detected on day 3, and this increase continued throughout day 8 (average IL-8 pg/ml ± SE; n = 3 experiments; *P values <0.05).

The dataset was analyzed using GC-RMA and uncovered 732 gene probe sets with a twofold or greater expression difference in MDCK cells grown in 3D culture compared with those grown in 2D culture [Supplemental Table S1 (Supplemental data for this article may be found on the American Journal of Physiology: Cell Physiology website.)]. Interestingly, the comparative analysis showed that a substantial majority of the differentially regulated genes (630 genes) were upregulated in MDCK cysts. Signaling pathway analysis revealed the enrichment of 37 pathways in MDCK cysts compared with MDCK flat sheets (FDR P values <0.05). Of these, a large number were pathways associated with immune responses and with cell adhesion and ECM remodeling (Supplemental Table S2). Transgelin, or SM22α, was the most highly upregulated gene, with increases of 110.9-, 43.0-, and 18.2-fold detected in three separate hits. Transgelin is a cytoskeletal protein primarily expressed in smooth muscle cells but also found to be upregulated in injured podocytes (27) and endothelial cells undergoing endothelial-mesenchymal transformation (18). Both matrix metalloproteinase (MMP) 13 (12.1- and 8.5-fold in two separate hits) and MMP9 (9.6-fold) were upregulated, along with tissue inhibitor of matrix metalloproteinase (TIMP) 1 (4.5-fold) and TIMP3 (5.0-, 4.9-, 4.8-, and 3.0-fold in four separate hits). MMPs are involved in ECM degradation, and their activities are regulated by TIMPs. Several proteins involved in ECM interactions, such as integrin α₆ (4.7-fold) and ADAM metallopeptidase domain 28 (2.0-fold), were also differentially expressed in MDCK cysts. The analysis also detected the upregulation of fibronectin 1 (3.5-fold), an ECM linker protein that binds to integrin receptors. Additionally, proteins known to contribute to the cytoskeleton or to play a role in mediating cell-cell adhesion, such as filamin A (2.7-, 2.0-, and 2.0-fold in three separate hits), actinin α₁ (3.0- and 2.9-fold in two hits), and claudin 1 (2.1-fold), were found to be upregulated.

Among the collection of differentially expressed genes and significantly enhanced pathways were genes of immune and inflammatory response pathways. Included on this list, for instance, were several major histocompatibility complex class II molecules, monocyte chemoattractant protein-1, as well as CCL and CXC chemokines, all of which were dramatically upregulated. IL-8 was one of the most highly upregulated genes with duplicate hits on each biological replicate. In two separate hits, IL-8 showed a 7.8-fold increase (3D 201.1 vs. 2D 25.7; P = 9.81 × 10⁻⁹) and a 2.5-fold increase (3D 86.5 vs. 2D 34.5; P = 4.05 × 10⁻⁶) in MDCK cysts compared with 2D monolayers.

Validation of IL-8 upregulation in 3D MDCK collagen culture.

To validate the upregulation of IL-8 expression in MDCK cysts, total RNA samples from MDCK cysts and flat sheets were analyzed by real-time PCR. The relative abundance of IL-8 mRNA in MDCK cells cultured under 3D vs. 2D conditions was determined by real-time PCR analysis to be 1,232.0 ± 396.5-fold. The independent demonstration of IL-8 upregulation through both of these methods serves to validate the differential expression of IL-8 in MDCK cells grown in 3D vs. 2D culture.

Increased IL-8 protein expression in 3D MDCK collagen culture.

After determining that IL-8 mRNA expression was dramatically upregulated by the cells that constitute MDCK cysts, we asked if this increase in mRNA levels was accompanied by increased IL-8 expression at the protein level. To answer this question, we applied two different molecular techniques. First, a canine specific Luminex Assay was performed with media samples collected from MDCK cells grown in 2D and 3D culture. As seen in Fig. 1B, we found that IL-8 levels present in culture media bathing MDCK 3D cultures, compared with those present in media bathing 2D cultures, were significantly increased beginning on day 3 postplating [3D average (avg) 244.5 ± 37.3 pg/ml vs. 2D avg 1.5 ± 15.5 pg/ml; P value <0.05] and remained elevated throughout day 8 (3D avg 695.2 ± 68.6 pg/ml vs. 2D avg 13.5 ± 20.1 pg/ml; P value <0.05).

While Luminex is a highly sensitive assay for the quantification of IL-8 protein levels, it was necessary to develop an additional method that would permit us to accurately quantify protein levels from a large number of media samples. For this purpose, an ELISA designed specifically to detect canine IL-8 was used to quantify the levels of protein in culture media collected from MDCK cysts and flat sheets. Figure 2 illustrates that a significant increase in IL-8 was detected in 3D MDCK cultures on all days of media collection (3D MDCK culture vs. 2D MDCK culture day 3: 2,002.8 ± 231.5 vs. 66.5 ± 12.5 pg/ml; day 6: 2,826.7 ± 662.9 vs. 67.3 ± 6.3 pg/ml; day 8: 3,298.4 ± 701.2 vs. 102.5 ± 23.7 pg/ml; P value <0.05). It is of note that, while the levels of IL-8 were always dramatically higher in 3D MDCK culture compared with 2D culture, there was often large variation among the absolute values of the IL-8 concentrations detected in individual samples from the same condition. It is also worth noting that, while both the Luminex and ELISA assays consistently demonstrated dramatic increases in IL-8 production by cells grown in 3D culture, the absolute quantities of IL-8 measured varied among these techniques, perhaps because different anti-IL-8 antibodies were employed.

Fig. 2.

Quantification of IL-8 protein in 2D and 3D MDCK cultures by IL-8 sandwich enzyme-linked immunosorbent assay (ELISA). The levels of IL-8 present in media samples collected from 2D (dark gray) and 3D (light gray) MDCK cultures were quantified by sandwich ELISA. Consistent with the results of the Luminex assay, the levels of IL-8 protein were found to be significantly higher in 3D culture media compared with 2D culture media (average IL-8 pg/ml ± SE; n = 4; *P values <0.05).

Determination of IL-8 expression in HGF-stimulated 3D MDCK culture.

After determining that IL-8 was highly upregulated at both the mRNA and protein levels in MDCK cysts, we wondered whether it plays a role in regulating the morphogenetic processes that lead to cystogenesis and tubulogenesis. When 3D MDCK cells are grown in the presence of HGF they do not form cysts but instead undergo tubulogenesis, forming elongated cylindrical structures (32, 33, 49, 50). To assess whether there is a relationship of IL-8 production to these processes, we began by stimulating 3D MDCK cultures with 40 ng/ml HGF to induce tubule formation. As seen in Fig. 3, real-time PCR analysis revealed that the abundance of IL-8 mRNA was decreased by almost 85% compared with the levels produced by untreated 3D MDCK cultures (IL-8 expression in HGF-stimulated 3D MDCK culture vs. that detected in untreated 3D MDCK cultures = 0.17 ± 0.07; P values <0.05). The addition of 2,000 pg/ml RcIL-8 to HGF-stimulated 3D MDCK cultures had no effect on the suppression of IL-8 mRNA expression (HGF + 2,000 pg/ml RcIL-8 vs. untreated 3D MDCK cells = 0.14 ± 0.05; P value <0.05).

Fig. 3.

Real-time PCR analysis of IL-8 gene expression in hepatocyte growth factor (HGF)-treated and recombinant canine IL-8 (RcIL-8)-treated 3D MDCK cultures. When grown in 3D culture and stimulated with HGF, MDCK cells undergo tubulogenesis. Real-time PCR analysis was performed with total RNA isolated from 3D cultures maintained in the presence or absence of 40 ng/ml HGF. The relative abundance of IL-8 mRNA was suppressed in the presence of HGF (HGF treated/untreated = 0.17 ± 0.07). RcIL-8 (2,000 pg/ml) did not relieve the suppression of IL-8 mRNA expression induced by the presence of HGF (HGF treated/untreated = 0.14 ± 0.05). (n = 3; *P values <0.05).

Protein levels of IL-8 were also assessed by ELISA assay performed on media samples collected from HGF-stimulated 3D MDCK cultures. Figure 4 illustrates the decreased levels of IL-8 protein detected in media samples collected from HGF-stimulated 3D MDCK cultures compared with untreated 3D culture (HGF-stimulated 3D MDCK culture vs. untreated 3D MDCK culture day 3: 386.2 ± 119.4 vs. 827.4 ± 214.5; day 6: 1,327.9 ± 330.8 vs. 3,016.2 ± 634.6; day 8: 479.9 ± 110.5 vs. 1,581.2 ± 301.5; P values <0.05). RcIL-8 added to 3D MDCK cultures was detectable by ELISA, and the sum of the exogenous plus the endogenously produced IL-8 was reduced in HGF-treated 3D MDCK cultures, consistent with the observation that endogenous IL-8 production in 3D-grown MDCK cells was reduced by exposure to HGF.

Fig. 4.

HGF exposure reduces the quantity of IL-8 protein in the media bathing 3D MDCK cultures. The levels of IL-8 protein in media collected from 3D MDCK cultures were measured by sandwich ELISA. The levels present in the media bathing cells stimulated with 40 ng/ml HGF were found to be significantly decreased. HGF-stimulated MDCK cultures treated with 2,000 pg/ml RcIL-8 continued to manifest an apparent decrease in the level of endogenous IL-8 present, as evidenced by the reduced total IL-8 levels quantified in these media samples (average IL-8 pg/ml ± SE; n = 3; *P values <0.05).

RcIL-8 induces cystogenesis in HGF-stimulated 3D MDCK cultures.

Our data indicate that MDCK cells form cysts in 3D culture that produce and secrete IL-8 and that this IL-8 production is suppressed when cells are treated with the tubulogenic factor HGF. These observations suggest the interesting possibility that increased levels of IL-8 promote cyst development and that HGF induces tubulogenesis at least in part by suppressing IL-8 production. Were this the case, then addition of exogenous IL-8 to HGF-treated cultures should be sufficient to at least partially suppress tubulogenesis and to favor cyst development. To assess this possibility, 3D MDCK cultures grown in the presence or absence of 40 ng/ml HGF and 2,000 pg/ml RcIL-8 were stained with Hoechst 33342 nuclear stain. To quantify the differences in the structures that developed under these different culture conditions, captured images were imported to ImageJ (National Institutes of Health). With the use of the line tool, measurements of length and width were obtained for multicellular structures in the different experimental conditions. The average ratio of length to width was calculated for each experimental condition. The length-to-width ratio of spherical cysts should be ∼1, whereas cylindrical tubules should manifest ratios greater than 1 because of their extended lengths. It should be noted that, in the present study, no attempt was made to assess whether tubule structures had undergone lumen formation. As shown in Fig. 5A, we found that untreated 3D MDCK cultures and those treated with 2,000 pg/ml RcIL-8 both produced spherically shaped cystic structures with hollow central lumens and corresponding length-to-width ratios of close to unity that did not differ significantly among these two conditions (untreated 3D MDCK culture 1.1 ± 0.03 vs. 2,000 pg/ml RcIL-8 3D culture 1.1 ± 0.02). The spontaneous formation of cysts in 3D MDCK culture was inhibited in the presence of 40 ng/ml HGF. The presence of HGF stimulated the formation of elongated tubule-like structures, as depicted in Fig. 5A. The length-to-width ratio increased to 3.3 ± 0.25 and differed significantly from the values measured in the untreated 3D MDCK cultures (P < 0.05; Fig. 6B). In contrast, the reintroduction of 2,000 pg/ml RcIL-8 in HGF-stimulated 3D MDCK cultures resulted in the formation of both spherical cysts and elongated tubule-like structures. Quantitation of these images (Fig. 5B) indicates that the length-to-width ratio of the cellular structures present in this condition remained higher than those found in untreated 3D MDCK culture (2.5 ± 0.2; P < 0.05) but was significantly lower than that measured in 3D MDCK cultures treated with HGF alone (P < 0.05). This decreased length-to-width ratio was in part due to the reappearance of cysts in 3D MDCK cultures treated with both HGF and RcIL-8. These data indicate that the addition of exogenous IL-8 at least partially overcame the tubulogenic influence of HGF, resulting in the formation of a mixed population of cystic and tubule-like structures.

Fig. 5.

HGF stimulates tubulogenesis, and RcIL-8 partially overcomes this effect to induce cystogenesis in 3D MDCK culture. MDCK 3D culture samples were labeled with Hoechst 33342 nuclear stain after 8 days of growth. Representative images (A) illustrate that untreated MDCK cells spontaneously form cysts in 3D collagen culture (thick arrows). In the presence of 40 ng/ml HGF, however, the proportion of cystic structures is substantially reduced, and instead linear tubule-like structures, as indicated by the thin arrows, are detected. In cultures treated with both HGF and 2,000 pg/ml RcIL-8, both tubule-like and cystic structures (thick arrows) were observed. Images were imported into Image J software, and the length-to-width ratio of the cellular structures from each culture condition were calculated. Scale bars (bottom right corners) correspond to 80 μm. B: the ratios calculated for untreated 3D MDCK cultures, as well as for those treated with 2,000 pg/ml RcIL-8 alone, are ∼1, consistent with the fact that the cells form spherical cystic structures under these conditions. HGF-stimulated 3D MDCK cultures exhibited much higher ratios (3.25 ± 0.25), indicating the presence of linear structures (P value <0.05). The addition of 2,000 pg/ml RcIL-8 to HGF-stimulated 3D MDCK culture resulted in cultures characterized by ratios intermediate between those measured in untreated and HGF-treated cultures. Taken together with the representative images (A), these data suggest that MDCK cells grown in the presence of both HGF and IL-8 condition form both cysts and tubules. Scale bar = 50 μM. Average length-to-width ratio ± SE; *P values <0.05. Quantification values were obtained from the measurements of 25 individual 3D MDCK structures over three individual experiments.

Fig. 6.

Addition of anti-IL-8 antibody to MDCK 3D cultures perturbs cyst formation. MDCK cells were grown for 6 days in 3D culture in the absence or presence of anti-IL-8 antibody (0.8 μg/ml). Cells maintained in the absence of anti-IL-8 antibody primarily formed cysts with sharply defined edges, indicating that they possessed fluid-filled lumens (A, thin arrows). When grown in the presence of anti-IL-8 antibody, the MDCK cells formed a variety of structures, including spherical cell aggregates that appeared to lack central lumens (B, arrowhead), elongated and branched structures (B, thick arrows), and collections of individual cells or small cell aggregates (B, asterisk). Scale bars (bottom right corners) correspond to 40 μm.

Finally, if IL-8 secreted by MDCK cells grown in 3D culture contributes to cyst formation, then interfering with the capacity of this IL-8 to interact with its receptors should alter the size or nature of the structures produced by cells grown under these circumstances. MDCK cells were grown for 6 days in 3D culture in the absence or presence of anti-IL-8 antibody (0.8 μg/ml). As can be seen in Fig. 6, cells maintained in the absence of anti-IL-8 antibody primarily formed cysts with sharply defined edges, indicating that they possessed fluid-filled lumens (Fig. 6A). When grown in the presence of anti-IL-8 antibody, the MDCK cells formed a variety of structures, including spherical cell aggregates that appeared to lack central lumens (Fig. 6B), elongated and branched structures (Fig. 6B), and collections of individual cells or small cell aggregates (Fig. 6B). Thus, it appears that the propensity of MDCK cells to form lumen-containing cysts in 3D culture depends at least in part upon the secretion of functionally active IL-8.

DISCUSSION

The primary goal of this study was to gain insight into the influence of culture geometry on gene expression patterns and epithelial morphogenesis. Mauchamp et al. originally described the culture of thyroid epithelial cells within an ECM substrate (28), in what has now become a classic in vitro model for the study of epithelial cystogenesis and tubulogenesis (5, 59, 60). We began our study by utilizing this model to perform a comparative gene expression analysis between MDCK cells grown in two different morphologies. MDCK cells cultured in 2D were plated atop a presolidified collagen gel. Cells cultured under these conditions grew as flat monolayer sheets. In 3D culture, cells were embedded within type I collagen gel. Under 3D culture conditions, MDCK cells formed large spherical cysts with hollow central lumens. An Affymetrix microarray analysis uncovered 732 gene probe sets whose expression varied by twofold or greater, with FDR P values <0.05, in MDCK cysts compared with flat sheets. The large number of differentially expressed genes identified in this dataset provides an indication of the complexity of signaling networks and cellular processes involved in responding to environmental cues during the establishment and growth of epithelial structures.

The gene whose expression was most dramatically upregulated by growth in the 3D culture conditions encodes the transgelin protein. Transgelin is an actin cross-linking protein that is enriched in smooth muscle cells (21) and whose expression is reduced in a variety of epithelial carcinomas (54). Transgelin expression also correlates with cell shape changes and with migration (53). Recently, microarray and immunolocalization studies have revealed that transgelin expression is a marker of epithelial-to-mesenchymal transition (EMT) in human polycystic kidneys and in other renal epithelial pathologies (27, 44, 51).

The importance of cell-cell and cell-matrix interactions in the determination of epithelial morphology has been well established (59, 60, 67). It was, therefore, not surprising to find that our dataset of differentially expressed genes was enriched in genes that encode proteins involved with cell adhesion and with the formation or remodeling of the ECM. In fact, microarray analysis of HGF-treated MDCK cells has shown that this treatment alters the expression of a large number of genes, particularly those involved in ECM associations (1, 20). Additionally, an Affymetrix microarray analysis of HGF-stimulated MDCK cells revealed an upregulation of the expression of MMP13 and TIMP1 (14). The proteins encoded by these genes mediate and regulate, respectively, the degradation of the ECM. Subsequently, it was shown that the expression of the proteins encoded by both of these genes was essential for late stages of morphogenesis following HGF treatment, which are characterized by tubulogenesis. Analysis of our dataset revealed that both of these molecules were also substantially upregulated in the 3D growth condition (see Supplemental Table S1). Interestingly, transgelin appears to act as a repressor of the expression of the matrix metalloproteinase MMP9 (38). The expression of the ECM component fibronectin was also upregulated in MDCK cysts grown in 3D culture. It has previously been shown that fibronectin plays an essential role in the partial EMT that occurs early in the process of HGF-stimulated tubule formation. In addition, it has been previously shown that disrupting fibronectin expression impairs cyst growth and prevents the initiation of tubulogenesis in MDCK cells grown in 3D culture (16, 26, 47, 50). Taken together, these findings suggest that proteins involved in remodeling the ECM play an important role in the development and expansion of multicellular epithelial structures in 3D culture.

An analysis of significantly enriched signaling pathways revealed that, in addition to cell adhesion and ECM remodeling, many of the differentially expressed genes were involved in pathways regulating immune and inflammatory response processes (FDR P values <0.05). The upregulation of immune and inflammatory response genes was an unexpected observation. Recent studies, however, have suggested that these pathways may participate in orchestrating normal and pathological epithelial morphogenesis. Chemokines have been shown to play a role in the regulation and determination of endothelial and epithelial morphogenesis in a variety of tissues (22, 35, 55–57). An upregulation of the expression of genes related to immune and inflammatory processes that was qualitatively similar to the pattern that we observed in the present study was also found in a microarray analysis that compared severely affected kidneys vs. mildly affected kidneys from the cpk mouse model of PKD (34). Increased expression of cytokines has also been detected in the cystic fluid of human tissues affected by PKD and in the blood plasma and the cystic fluid of mouse models of PKD (29, 30, 41). Perhaps of greatest relevance to the present work, the murine IL-8 functional homolog KC was found to be upregulated in HGF-stimulated mIMCD cells grown in 3D culture at both the mRNA and protein levels (57). This upregulation of KC appears to inhibit the migration, branching, and extension that accompany HGF-stimulated tubulogenesis.

We found that IL-8 expression was upregulated at both the message and protein levels in MDCK cells grown as cysts in 3D culture. This IL-8 expression was dramatically inhibited through the addition of the tubulogenic factor HGF to the media bathing 3D MDCK cultures. As expected, when MDCK cells were exposed to HGF in 3D culture, the cells organized themselves into linear tubule-like structures. The addition of exogenous RcIL-8 was sufficient to result in the reappearance of cystic structures in the HGF-treated cultures. Conversely, addition of anti-IL-8 antibody to 3D cultures resulted in the development of fewer fluid-filled cysts and instead was associated with the development of spherical, elongated, and branched cell aggregates. These data suggest that autoctine or paracrine IL-8 signaling plays an important role in driving cyst formation in 3D cultures of MDCK cells.

IL-8, renamed CXCL8, is the ligand for the G protein-coupled receptors CXCR1 and CXCR2. Skin epithelial cells have been shown to upregulate IL-8 expression in response to wounding. These cells express both CXCR1 and CXCR2, and recombinant IL-8 increased both migration and proliferation of cultured keratinocytes, suggesting an autocrine pathway for chemokine-mediated wound repair (19). As opposed to epithelial cells in the skin, which organize on a primarily flat surface, tubular cells must achieve a 3D shape with a central lumen. Furthermore, that shape must be predominantly linear rather than spherical. Although many factors have been suggested to regulate the process of linear tube formation, it is clear that coordination of the rate of proliferation with that of cellular elongation is required to achieve a tubular shape. Thus, by inhibiting IL-8 expression, HGF may alter this balance in a manner sufficient to shift the balance from cystic to tubular morphogenesis. Taken together, our data suggest the interesting possibility that HGF may stimulate tubulogenesis and suppress cystogenesis by virtue of its ability to inhibit IL-8 production. According to this model, IL-8 serves as an important driver of cyst formation in 3D culture, and the tubulugenic influence of HGF is mediated, at least in part, by suppressing the dominant cystogenic influence of IL-8.

GRANTS

This research was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) under Ruth L. Kirschstein National Research Service Award F31-DK-076207 (E. K. Wells), NIDDK Grant DK-17433 (M. J. Caplan and R. P. Lifton), and the Leducq Foundation Transatlantic Network for Hypertension (M. J. Caplan and R. P. Lifton).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

Author contributions: E.K.W., O.Y., R.P.L., L.G.C., and M.J.C. conception and design of research; E.K.W. and O.Y. performed experiments; E.K.W., O.Y., R.P.L., L.G.C., and M.J.C. analyzed data; E.K.W., R.P.L., L.G.C., and M.J.C. interpreted results of experiments; E.K.W. prepared figures; E.K.W. and M.J.C. drafted manuscript; E.K.W., O.Y., R.P.L., L.G.C., and M.J.C. edited and revised manuscript; E.K.W., R.P.L., L.G.C., and M.J.C. approved final version of manuscript.