A Gene Expression Profile That Defines Colon Cell Maturation in Vitro

INTRODUCTION

Cellular proliferation, lineage-specific differentiation, migration, and apoptosis are highly coordinated processes that occur in a sequential and spatially organized manner along the colonic crypt axis. Stem cells at the base of the crypt give rise to progenitor cells, which expand by rapid proliferation before undergoing cell cycle arrest followed by differentiation along one of three cell lineages, absorptive, goblet, or enteroendocrine, as they migrate along the crypt axis toward the lumenal surface (1) . The factors that regulate these processes are only partially understood, although a number of biochemical events including E-cadherin-mediated cell-cell and integrin-mediated cell-substratum adhesion, chemotactic gradients, ECM 3 and mesenchymal components, cytokines, hormones and growth factors, have been implicated (2, 3, 4, 5) . Downstream of these stimuli, signaling pathways and transcription factors, including Tcf-4, MATH-1, the homeobox genes cdx-1 and cdx-2, kruppel-like factor 4, and several members of the forkhead family of transcription factors, have been shown to play a role in the coordination of colonic cell maturation (5, 6, 7, 8, 9) .

Whereas the genetic reprogramming induced by these stimuli to bring about the differentiated phenotype are likely to be extensive and complex, understanding these events would enhance both our knowledge of the mechanisms of maintenance of colonic epithelial cell homoeostasis, as well as the pathogenesis of colorectal tumorigenesis. Differentiation of colonic epithelial cells also remains a poorly defined process at the biochemical and molecular level. Typically, differentiation of absorptive cells is described as a cessation of DNA synthesis, followed by morphological changes such as cellular polarization, and the formation of well-developed tight junctions and microvilli (1 , 10) . On a biochemical level, increased expression of brush border hydrolases such as alkaline phosphatase, sucrase isomaltase, and dipeptidylpeptidase IV have been described (11) . However, other changes likely to occur as colonic epithelial cells migrate upwards along the crypt axis, including changes in signaling pathways, interactions with the extracellular matrix, metabolic processes, adaptive responses to lumenal contents, and apoptotic pathways, remain only partially characterized.

Advances in cDNA microarray technologies have enabled the definition of global changes in gene expression (12) . In the present study we have used this technology, in combination with a unique model system, the Caco-2 colon carcinoma cell line, to gain additional insight into the genetic reprogramming that accompanies colonic epithelial cell maturation. The Caco-2 cell line spontaneously undergoes contact inhibition-dependent cell cycle arrest and differentiation along the absorptive cell lineage as a function of time in culture, modeling the phenotypic changes normal colonic epithelial cells undergo as they migrate along the crypt axis. This model has been used extensively in the study of colon cell maturation (13, 14, 15, 16) .

Microarray analysis of 17,280 sequences revealed that maturation of Caco-2 cells is characterized by extensive reprogramming at the molecular level. Changes in expression occurred in an organized manner, with genes in subsets of functional categories highly enriched in patterns of coordinate regulation. Consistent with the induction of cellular differentiation, several previously defined markers of absorptive cell differentiation were up-regulated in expression. Classification of named genes in the database into functional categories revealed that, in general, genes involved in cell cycle regulation and nucleic acid synthesis were down-regulated, consistent with the parallel induction of cell cycle arrest. Coordinate down-regulation of genes involved in RNA processing, translation, protein folding, and degradation was also evident. Conversely, genes involved in xenobiotic and drug detoxification, extracellular matrix deposition, and lipid metabolism were up-regulated. A number of signal transduction pathways and transcription factors potentially involved in the coordination of these responses were identified. The complete database can be accessed at our laboratory website. 4

MATERIALS AND METHODS

Cell Culture.

The Caco-2 human colon cancer cell line was obtained from the American Type Culture Collection. Cells were cultured in MEM (Life Technologies, Inc., Grand Island, NY), supplemented with 10% FCS, 0.1 mm nonessential amino acids, 10 mm HEPES buffer, 100 units/ml penicillin, 100 μg/ml streptomycin, and 0.25 μg/ml amphotericin (Life Technologies, Inc.). Cell cultures were maintained at 37°C in 95% air and 5% CO₂.

RNA and Probe Preparation.

Caco-2 cells were cultured to confluence (day 0), or for 2, 5, 14, or 21 days post confluence, with medium changed every second day. Cells were harvested and RNA isolated using the RNeasy kit (Qiagen, Valencia, CA). For each hybridization two separate probes were prepared: one labeled with Cy3 (day 0) and the other with Cy5 (days 2, 5, 14, and 21). The probe preparation, hybridization conditions, and slide scanning procedure were as described previously (17) . Arrays used in this report were prepared by the microarray facility at the Albert Einstein College of Medicine (18) . For each time point, arrays were performed on two independent chips generated from independent clone sets. The first chip contained 8,064 sequences and the second 9,216 sequences, resulting in 17,280 sequences analyzed for each time point.

Data Analysis.

The Scanalyze software of Eisen et al. (19) was used to generate signal and background intensities for each channel at each spot on the microarray. Data were transferred to a Microsoft Excel spreadsheet where the signal:background ratio for each channel was calculated as well as the ratio between these ratios (i.e., green signal:background ratio was divided by the red signal:background ratio). Data were then normalized among arrays by expressing this value relative to the average of these values for all of the expressed genes on the array. The data were log (2) transformed and transferred to Microsoft Access where genes changing in expression during the maturation program were identified. For graphical representation, data were clustered and displayed using the Eisen’s Cluster and Treeview programs (19) . A gene was included for additional analysis if the signal:background ratio was >1.25 for either the red or green channel for at least two of the four time points. A gene was considered differentially expressed if the ratio of the red:green signal was >25% in at least two of the final three time points. These criteria are similar to those used previously (17) and are based on the 95% confidence interval determined from self-hybridization experiments.

Gene Classification and Functional Group Analysis.

Of the 17,280 genes on the arrays, 39% were named genes, and 61% genes of unknown function or ESTs. Named genes were sorted in alphabetical order and categorized into 1 or more of 25 functional categories. Gene classification was based on searches of the Genecards 5 or Medline 6 databases. Functional group analysis was performed as described previously by Muller et al. (20) and is based on the hypothesis that if a particular biological process is modulated during the differentiation program, a disproportionately higher percentage of genes involved in the regulation of that process will be altered in expression (20) . The level of significance was calculated from the binomial distribution, using the following algorithm in Microsoft Excel: BINOMDIST (X, Y, Z, FALSE), where X is the number of genes regulated in a given functional group, Y is the number of genes in that functional group, and Z is the overall proportion of genes regulated in the experiment.

Measurement of Apoptosis.

Caco-2 cells cultured to confluence (day 0) or for 2, 5, 7, 14, or 21 days postconfluence were treated with 10 μm of cisplatin (Sigma, St. Louis, MO) or 1 μm camptothecin (Calbiochem, San Diego, CA) for 72 h. Adherent and nonadherent cell populations were pooled and stained overnight with propidium iodide [50 mg/ml in 0.1% (w/v) sodium citrate and 0.1% (v/v) Triton X-100; Sigma]. The extent of apoptosis was determined by quantification of the sub-diploid cell fraction by fluorescence-activated cell sorter analysis, as described previously (21) .

Clonogenic Assay.

Caco-2 cells cultured to confluence (day 0) or for 2, 5, 7, 14, or 21 days postconfluence were treated with 10 μm cisplatin (Sigma) or 1 μm camptothecin (Calbiochem) for 9 h. Medium was removed, cells trysinized, counted, and reseeded in triplicate into six-well plates at a density of 500 cells/well. Colonies formed over ∼3 weeks and were visualized by staining with 1% crystal violet for 30 min.

Western Blotting.

Protein isolation, Western blotting, and signal detection were performed as described previously (16) . Anti-p21^WAF-1/CIP-1, cyclin A, cyclin E, c-myc, PCNA, HMG-1, topoisomerase I, eIF2β, eIF4E, and E2A were obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-thymidylate synthase was obtained from Lab Vision Corporation (Fremont, CA), and anti-β-actin from Sigma.

[³H]Thymidine and [¹⁴C]Leucine Uptake Experiments.

For measurement of [ 3 H]thymidine and [¹⁴C]leucine uptake, Caco-2 cells were grown to confluence (day 0), or for 2, 5, 7, 14, and 21 days postconfluence, in 24-well plates. At each time point, cells were pulsed with 0.5 μCi/well [¹⁴C]leucine or 0.5 μCi/well [³H]thymidine (Amersham, Piscataway, NJ) for 8 h. Cells were washed three times in cold PBS, harvested, and centrifuged at 1,500 rpm for 5 min at 4°C. Pellets were resuspended in 0.5 ml of 10% trichloroacetic acid and 1% phosphotungstic acid (Sigma), vortexed, and centrifuged at 15,000 rpm for 15 min. The supernatant was discarded, the process repeated on two more occasions, and the pellet air-dried overnight. The dried pellet was resuspended in 0.3 ml of NCS tissue solubilizer (Amersham), mixed with 5 ml of liquid scintillant (Sigma), and counted in a β counter. Counts were corrected for total cellular protein measured in parallel samples.

Boyden Chamber Assay of Chemotactic Endothelial Cell Migration.

HUVECs were isolated from umbilical cords and cultured as described previously (22) . Confluent HUVEC monolayers (passages 2–4) were harvested with cell dissociation solution and suspended at 1 × 10⁶/ml in M199 medium (Life Technologies, Inc.) supplemented with 1% serum. HUVECs (10⁵ cells) were seeded into 8.0-μm pore transwell inserts (Costar, Cambridge, MA) precoated with 10 μg/ml fibronectin. Inserts containing HUVECs were placed into a 24-well plate (Costar) containing M199 medium supplemented with 1% serum and incubated for 1 h at 37°C. HUVEC migration was stimulated by coculture with Caco-2 cells grown to confluence (day 0), or for 2, 5, 7, 14, and 21 days postconfluence. After 5 h, HUVECs were fluorescently stained with 10 μm of cell tracker green (Molecular Probes, Eugene, OR) and the upper surface of the insert swabbed to remove nonmigrated cells. Inserts were washed three times with PBS, fixed in 3.7% formaldehyde, and mounted on microscope slides. HUVEC migration was quantified by counting the number of cells in three random fields (100× total magnification) per insert. Data are expressed as endothelial cells per field (mean ± SE)/1 × 10⁶ Caco-2 cells.

RESULTS

Changes in Gene Expression during Caco-2 Cell Differentiation.

Caco-2 cells spontaneously undergo cell cycle arrest and differentiation along the absorptive cell lineage as a function of time in culture (15) . Initial analysis revealed that 13,638 of the 17,280 sequences analyzed (78.9%) were expressed to a significant degree in Caco-2 cells (ratio of signal:background was >1.25 in at least one of the two channels). By our criteria (see “Materials and Methods”), 2,286 of these sequences (13.2%) were altered in expression during the maturation program. This comprised 697 genes that were increased and 1,589 genes that were down-regulated in expression. The percentage of genes expressed in Caco-2 cells (78.9%) is consistent with what we have observed in a number of other colon cancer cell lines, 7 whereas the percentage of genes altered in expression (13.2%) confirms previous reports that the extent of alteration of gene expression during differentiation or transformation is on the order of 10% (23 , 24) .

Functional Group Analysis.

To understand the biological significance of the overall changes in gene expression, genes in the database were categorized into 1 or more of 25 predefined functional groups (Table 1) ⇓ . As expected, the number of genes comprising the different functional categories varied significantly. For example, 504 genes were linked to a role in metabolism, whereas only 30 genes with a role in neurotransmission were identified (Table 1) ⇓ . Functional groups that showed significant enrichment during Caco-2 cell maturation were determined by comparison of the percentage of genes altered in expression in a particular functional group to the overall percentage of change during the maturation program (13.2%). Functional groups that showed highly significant alterations of expression (P < 0.001) included cell cycle, DNA synthesis and repair, RNA processing, translation, protein processing and transport, protein degradation, xenobiotic and drug detoxification, ECM-associated factors, metabolism, kinases and phosphatases, and transcription factors (Table 1) ⇓ .

In contrast, we observed only a minimal enrichment of apoptosis-related genes consistent with the fact that despite undergoing cell cycle arrest and differentiation, the basal level of apoptosis remains unchanged during Caco-2 cell differentiation (Ref. 15 ; Fig. 5I ⇓ ). Furthermore, and consistent with the fact that Caco-2 are epithelial cells, we found no enrichment of genes involved in immunoinflammatory events or in neurotransmission (Table 1) ⇓ .

Of the 17,280 sequences analyzed, 10,549 were sequences of unknown function or ESTs. Interestingly, only 9% of these sequences (948) were altered in expression, which was significantly less than the overall level of change (13.2%). These results indicate that named genes were more likely to be altered in expression than genes of unknown function and may reflect the existence of a bias in terms of gene discovery, with genes with more visible phenotypic effects tending to be discovered and characterized more quickly than those with more subtle biological effects.

Finally, to additionally validate the functional group analysis, we randomly chose 222 genes from the database. This number was selected as it reflected the average size of the functional groups. Thirty-three of the 222 (14.9%) randomly chosen genes were altered in expression during the maturation program, which was not significantly different from the overall percentage of genes changed, demonstrating that functional groups that showed significant enrichment truly reflected an underlying biological change.

Differentiation.

The morphological changes that characterize colon cell differentiation encompass changes in cell polarization, with the development of well-formed tight junctions and microvilli (13) . Consistent with this and a number of previous reports (16 , 25) , we observed increased expression of the intestinal brush border-associated proteins dipeptidylpeptidase IV and villin, the adherens junction protein E-cadherin, and the tight junction protein claudin 7 (See Fig. 1 ⇓ ). 8 Importantly, we also observed reduced expression of the goblet cell-specific markers mucin-2 and intestinal trefoil factor 3 in accordance with Caco-2 cell differentiation occurring specifically along the absorptive cell lineage (26) . Increased ion-transport, and associated water absorption, is an additional characteristic of differentiated colonic epithelial cells and is a feature that manifests itself in differentiated Caco-2 cells through the formation of dome-like structures (13) . Consistent with these changes, we observed increased expression of a number of genes involved in the regulation of Na⁺, Cl⁻, K⁺, H⁺ and HCO₃⁻ exchange (Fig. 1 ⇓ ; supplementary data).

Cell Cycle Genes.

We and others have demonstrated previously that Caco-2 cell differentiation is accompanied by a parallel decrease in cell proliferation reflected in a reduction of the percentage of cells in S phase (14 , 15) . This observation was confirmed by a 67% reduction in the rate of [³H]thymidine incorporation during the maturation program (Fig. 1H) ⇓ . To dissect the molecular mechanisms and pathways involved in the induction of cell cycle arrest, we analyzed the expression profiles of a panel of 148 genes that we had identified previously as playing a role in cell cycle regulation. Thirty-eight of these (25.7%) were altered in expression during Caco-2 cell maturation, with the majority down-regulated in expression. Relative to the overall level of change in gene expression (13.2%), this represented a significant enrichment in this functional class (P < 0.0001). Specific genes down-regulated included the cyclins A, B, B1, D1, D2, E, and F, the cell division cycle proteins cdc2, cdc6, cdc10, cdc25C, and cdc28, and the cyclin-dependent kinases, cdk-1 and cdk-2. Conversely, three genes with a role described previously in cell cycle arrest, p21^WAF-1/CIP1, transducer of ERBB2 1 (TOB), and the fork head/winged helix family transcription factor checkpoint suppressor 1 (27 , 28) were up-regulated. The changes in expression of cyclin A, E, and p21^WAF1/CIP1 were confirmed by Western blot (Fig. 1, A–C) ⇓ .

Caco-2 cell differentiation was also accompanied by reduced expression of the E2F transcription factor family members, E2F-1, E2F-3, and E2F-5, as well as the E2F transcriptional coactivator DP-1 (Fig. 1D) ⇓ . This family of transcription factors plays a central role in the regulation of gene expression at the G₁-S phase transition of the cell cycle by regulating the expression of genes of which the products are required for nucleotide biosynthesis, DNA replication, and cell cycle progression (20) . Consistent with their down-regulation, there was a parallel decrease in a number of E2F target genes reported previously including cyclin A, E, and D1, dihydrofolate reductase, DNA polymerase α, minichromosome maintenance deficient proteins 2, 3, and 6, cdc 2 and 6, origin recognition complex 1, and c-myc (Ref. 29 ; Figs. 1 ⇓ and 2 ⇓ ).

Finally, and consistent with the inhibition of cell proliferation, Caco-2 cell maturation was accompanied by a concomitant down-regulation of several M phase phosphoproteins and proto-oncogenes, including K-ras, tpr, c-fes, met (HGF receptor), c-myc, v-ets, and v-yes (Fig. 1, E–G) ⇓ .

Nucleic Acid Synthesis, DNA Replication and Repair, Chromosome Segregation, and Chromatin Assembly.

Consistent with the reduction in expression of cell cycle regulatory genes and the reduction in cell proliferation, there was a concomitant reduction of genes involved in DNA synthesis, replication, and repair. Down-regulated genes with a role in nucleic acid biosynthesis included thymidylate synthase, dihydrofolate reductase, UMP kinase, adenosine kinase, adenylosuccinate lyase, and phosphoribosyl PP_I synthetase 2 (Fig. 2A) ⇓ . Genes with a role in DNA replication included MCM2, MCM3, MCM6, origin recognition complex subunits 1 and 3, topoisomerases I, and II, DNA ligase 1, DNA polymerase α and δ, PCNA, and replication factor c (activator 1) 4 (Fig. 2B) ⇓ . Similarly, and consistent with DNA repair being most active in rapidly proliferating cells, DNA repair genes including DNA polymerase epsilon, mutS homologue 3, BRCA1 and BRCA2 were maximally expressed in undifferentiated, proliferating Caco-2 cells (Fig. 2C) ⇓ . Furthermore, genes involved in spindle formation and chromosome segregation, including several dyneins, microtubule-associated proteins (MAPs), tubulin, and kinesin family members, were down-regulated during the maturation program (Fig. 2D) ⇓ . Ran, Ran-GAP1, and several Ran binding proteins, which in addition to their role in nuclear transport, play a role in microtubule polymerization and spindle formation (30) , were similarly down-regulated in expression (see Fig. 4E ⇓ , protein transport). Finally, genes involved in chromatin assembly were progressively down-regulated including nucleosome assembly protein 1-like 1, H4 histone family member G, histone H2A, and H2A histone family member O (Fig. 2E) ⇓ . The down-regulation of PCNA, thymidylate synthase, and topoisomerase 1 were confirmed by Western blot analysis (Fig. 2) ⇓ .

RNA Processing, Splicing, and Transport.

Consistent with an overall reduction in gene expression during the maturation program, genes with a role in RNA processing, including splicing, transport, and maturation of RNA, were coordinately down-regulated in expression (Fig. 2 ⇓ ; supplementary data). Of the 121 predefined genes in this category, 35 (28.9%) were altered in expression by our criteria, with the majority down-regulated in expression. This included several splicing factors, hnRNP family members, and RNA helicases, including BAT-1 and the DEAD/H box containing polypeptides 9, and 20 (Fig. 4, A–D) ⇓ . Furthermore, poly (rC) binding proteins 1, and 2 (also known as hnRNPE1 and E2), which are RNA binding proteins implicated in translational silencing, were up-regulated in expression (31) . However, two members of the serine/arginine family of splicing factors, SRp20 and SC35 were up-regulated during Caco-2 cell differentiation, as were the DEAD/H box containing polypeptides 19 and hPrp18 (Fig. 2 ⇓ ; supplementary data). These observations raise the possibility that different splicing factors may regulate the post-transcriptional processing of different groups of mRNA transcripts, which may be differentially expressed during colon cell maturation.

Translation, Protein Processing, Trafficking, and Degradation.

Caco-2 cell maturation was also associated with a significant enrichment of genes involved in protein translation (Table 1) ⇓ . Paradoxically, however, this was caused by both increased and decreased expression of genes involved in protein translation (Fig. 3, A–C) ⇓ . First, with the exception of 3 sequences, a number of genes involved in translation initiation and elongation were down-regulated. Examples included subunits of translation initiation factors 1–5, and translation elongation factor 1, epsilon 1 (Fig. 3A) ⇓ . The down-regulation of eIF4E and eIF2β was confirmed by Western blotting. Furthermore, the Pkr gene, the product of which catalyzes the phosphorylation of the α subunit of eIF2 leading to an inhibition of protein synthesis initiation (32) , was also down-regulated (Fig. 3A) ⇓ . Finally, genes encoding 9 of the 20 tRNA synthetases, and exportin, which mediates the nuclear export of all of the tRNAs, were coordinately down-regulated during the maturation program (Fig. 3B) ⇓ . Consistent with these changes, the overall rate of protein synthesis was reduced by 80% in fully differentiated cells (Fig. 3D) ⇓ .

We also observed reduced expression of RNA polymerase I (M_r 16,000 subunit) and upstream binding transcription factor (see Fig. 7C ⇓ , transcription factors). These enzymes play a key role in ribosomal RNA expression, and their down-regulation is suggestive of a down-regulation of rRNA gene expression (33) . However, in apparent contrast we observed a general, although modest, up-regulation of ribosomal protein genes, a trend that was consistent for 20 family members (Fig. 3C) ⇓ . Whether this reflects the continued need for translation in differentiated cells, a compensatory response to the down-regulation of other components of the translational machinery, or simply a discordance in the model system, remains to be determined. Nevertheless, the general up-regulation of genes encoding ribosomal proteins is consistent with their coordinate regulation (34) .

After translation, nascent polypeptides that exit the ribosome must be folded into their correct tertiary structure, a task performed by a number of proteins collectively termed chaperones (35) . In accordance with the down-regulation of protein synthesis there was a concomitant down-regulation of expression of several such genes, including members of the major chaperone classes, hsp60, hsp70, and hsp90 (Fig. 4, A and B) ⇓ . Similarly, several cyclophilins and FK506-binding protein 5, involved in accelerating protein folding (36) , were also down-regulated (Fig. 4C) ⇓ . Interestingly, we observed an increase in expression of the heat shock transcription factor HSF4 (Fig. 7I) ⇓ , which has been shown to be a negative regulator of heat shock gene expression (37) , suggesting a possible mechanism for the down-regulation of heat shock gene expression.

Upon synthesis and folding, a large percentage of proteins must be translocated to specific cellular organelles, particularly the nucleus. Transport of proteins >M_r 40,000 into and out of the nucleus occurs via the nuclear pore complex. To do so, proteins containing a nuclear localization signal first attach to soluble carriers of the importin-β/karyopherin-β family either directly or via an adapter such as importin α (38) . Translocation into the nucleus is an energy-dependent process and requires the cotranslocation of Ran, a Ras family GTPase (38) . A number of genes involved in the regulation of these processes were down-regulated, suggesting a reduction in nuclear protein trafficking during Caco-2 cell differentiation. These included four members of the importin-α (2, 3, 4, and 6) and four members of the importin-β family (1, 2, 2b, and 3) of protein carriers (Fig. 4D) ⇓ . Furthermore, Ran, RanGAP1, and 5 Ran binding proteins (1, 2, 2-like-1, 7, and 16) were also concomitantly down-regulated in expression (Fig. 4E) ⇓ . Finally, a number of components of the nuclear pore complex itself were down-regulated, including the nucleoporins 88, 153, 155, and 214, and the proto-oncogene tpr (Fig. 4F) ⇓ . There was also decreased expression of genes involved in protein translocation to other organelles. Examples included, adaptor-related protein complex μ 2 subunit, and metaxin-1, active in protein trafficking to lysosomes and mitochondria, respectively (Fig. 4G) ⇓ . Overall, of the 216 predefined genes in this functional class, 54 (25%) were altered in expression (P < 0.0001).

Several components of the ubiquitin-26S proteasome protein degradation pathway were also down-regulated (Fig. 4, H–L) ⇓ . This pathway involves the conjugation of ubiquitin to a protein substrate, a process executed by a series of well-defined enzymatic reactions, involving ubiquitin-activating enzyme (E-1), ubiquitin-conjugating enzymes, and ubiquitin ligase complexes (39) . Caco-2 cell differentiation was characterized by the down-regulation of E-1 enzyme, three ubiquitin-conjugating enzymes (E2G 1, E2L 3, and E2M), and several members of ubiquitin ligase complexes including cullins 1, 3, 4A, and the 1B F-box protein (Fig. 4, H–J ⇓ , respectively). There was also a down-regulation of five UBPs, UBP 4, 8, 10, 12, and 14 (Fig. 4K) ⇓ , which help regulate the ubiquitin-26S proteolytic pathway by generating free ubiquitin monomers from their initial translational products, by recycling ubiquitins during the breakdown of ubiquitin-protein conjugates, and/or by removing ubiquitin from specific targets (40) .

Ubiquitinated proteins are recognized by the 26S proteasome and targeted for destruction. The 26S proteasome comprises two subunits, the 20S proteolytic core and the 19S regulatory subunit. The 20S core is composed of 7 different α and 7 different β subunits arranged as a cylindrical α7β7β7α7 complex in four stacked rings (41 , 42) . Eight of these 14 subunits were present on the arrays, but none were altered in expression during the maturation program (data not shown). In contrast, several components of the 19S regulatory subunit were down-regulated in expression. The 19S regulatory subunit, which delivers protein substrates to the 20S subunit in an ATP-dependent manner, comprises at least 18 different proteins (42) . Of these, 3 ATPases (1, 2, and 3) and 3 non-ATPases (5, 11, and 12) were down-regulated during the maturation program (Fig. 4L) ⇓ . Furthermore, proteasome activator subunit 3 (PA28, γ, and Ki), an 11S regulator capable of activating the proteolytic activity of the 20S proteasome (43) , was also down-regulated in expression (Fig. 4L) ⇓ .

Collectively, of the 158 predefined genes in this functional category, 37 (23.4%) were altered in expression during Caco-2 cell differentiation, a significant enrichment relative to the overall level of change (P < 0.0001). The down-regulation of this pathway is mediated both at the level of protein ubiquitination and also at the proteasomal level, particularly through down-regulation of components of the 19S subunit.

Xenobiotic and Drug Detoxification.

Among the most highly induced genes during Caco-2 cell differentiation, and consistent with previous reports (44) , were four members of the glutathione S-transferase family, GST A1-1, A3, A4, and Ha subunit 2 (Fig. 5A) ⇓ . Glutathione S-transferases are Phase II conjugating enzymes that play an important role in xenobiotic efflux and in the prevention of oxidative stress (45) . Similarly, and consistent with the presence of common regulatory elements in the promoter regions of genes encoding Phase II enzymes (46) , members of two other families of Phase II conjugating enzymes, the sulfotransferases and UDP glycosytransferases, were also up-regulated (Fig. 5, B and C) ⇓ .

Expression of several metallothioneins (which are involved in cell protection against toxic metals and reactive oxygen species; Ref. 47 ), ATP binding cassette sub-family C (CFTR/MRP), cytochrome p450, and aldehyde dehydrogenase family members were also up-regulated during the maturation program (Fig. 5, C–G) ⇓ . Furthermore, epoxide hydrolase, which catalyzes the hydration of reactive expoxide species that arise from the metabolism of endogenous as well as xenobiotic compounds (48) , was also up-regulated (Fig. 5H) ⇓ .

Colonic cell exposure to xenobiotic agents is likely to increase as cells migrate along the crypt axis to the lumenal surface. Therefore, the increased expression of these genes during differentiation may reflect the activation of defense mechanisms in colonic epithelial cells as they mature in relation to their altered position in the crypt. Increased expression of many of these genes has also been linked to increased resistance of tumor cells to chemotherapeutic agents. For example, the active metabolite of camptothecin, SN-38, is detoxified through glucuronidation (49) , whereas platinum compounds such as cisplatin are eliminated through glutathione conjugation (50) , two sets of enzymes up-regulated during Caco-2 cell maturation. To test whether the increase in expression of these genes (glycosyltransferases and glutathione S-transferases) is linked to reduced sensitivity of differentiated Caco-2 cells to these agents, we treated Caco-2 cells at various stages of differentiation with camptothecin and cisplatin. As shown in Fig. 5I ⇓ and consistent with the increased expression of genes involved in drug detoxification, the ability of camptothecin and cisplatin to induce apoptosis decreased significantly as Caco-2 cells underwent differentiation. However, it remained possible that this reduced sensitivity was a consequence of the reduced rate of cell proliferation during Caco-2 cell differentiation. Therefore, we performed an additional, clonogenic assay in which cells were treated with camptothecin or cisplatin for 9 h, after which the drug was removed, and cells were trypsinized, reseeded, and allowed to form colonies. Differentiated Caco-2 cells had a substantially greater ability to form colonies after drug treatment compared with their undifferentiated counterparts. Treatment of undifferentiated Caco-2 cells with 1 μm camptothecin or 10 μm cisplatin resulted in 20% and 0% colony formation, respectively, in comparison to untreated controls. In comparison, identical treatment of differentiated Caco-2 cells resulted in 69% and 33% colony formation after camptothecin and cisplatin treatment, respectively (Fig. 5J) ⇓ .

Extracellular Matrix and Cell Migration.

An additional functional category that showed significant enrichment during Caco-2 cell maturation were genes active in ECM formation. Genes up-regulated included laminin β2, collagen type III α1, and type VI α1, fibrinogen β polypeptide, fibriongen-like 2, biglycan, and vitronectin, suggesting increased ECM deposition is a feature of colon cell differentiation (Fig. 6A) ⇓ . There was also increased expression of lysyl oxidase, a cuproenzyme that stabilizes ECMs by catalyzing the enzymatic cross-linking of collagen and elastin (51) . Similarly, bone morphogenetic protein 1, a metalloproteinase that processes pro-lysyl oxidase into active lysyl oxidase, as well as procollagens I-III into collagen (52) , was also up-regulated (Fig. 6B) ⇓ . Consistent with the increased expression of extracellular matrix genes, the integrin α3 subunit, which binds laminin and collagen (as well as fibronectin), integrin α5, which binds fibrinogen, and syndecan-4, which acts as a coreceptor in the adhesion of integrins to a range of ECM components (53) , were all up-regulated during Caco-2 cell differentiation (Fig. 6C) ⇓ . In contrast, fibronectin was down-regulated during the maturation program (Fig. 6A) ⇓ , an observation consistent with a previous report (54) . Therefore, whereas these results indicate increased secretion of ECM components may be a feature of colon cell differentiation, they also suggest that proliferating and differentiated cells secrete different combinations of ECM components.

Also included in this functional category were genes that have been implicated in cell migration. First, a number of proteases were altered in expression. Those down-regulated included MMP-2 and tPA, whereas MMP-14 and cathepsin B were up-regulated in expression. On the other hand, several protease inhibitors were up-regulated, including TIMP3, PAI III, α-1-antitrypsin, and kallistatin (Fig. 6, D and E) ⇓ . We have demonstrated previously that the basal rate of cell migration decreases ∼2-fold during Caco-2 cell differentiation (21) . The general increase in expression of genes associated with ECM deposition and protease inhibitors are consistent with this reduced rate of cell migration.

The ECM also plays a role in regulating the rate of migration of other cell types. For example, in the context of a growing tumor, the ECM can act as a reservoir for the storage of proangiogenic factors, such as TGF-β and basic fibroblast growth factor, secreted by tumor cells (55) . There was increased expression of several such factors during Caco-2 cell differentiation. Proangiogenic factors up-regulated included VEGF-B (but not VEGF or VEGF-C), TGF-β, TGF-β2, and angiogenin. We also observed a decrease in BAI-associated protein 1, a gene of which the product interacts with the angiogenesis inhibitor BAI-1 (56) . On the other hand, the proangiogenic factor endothelin 1, and endothelin 1 converting enzyme were down-regulated in expression (Fig. 6F) ⇓ .

These observations prompted us to ask whether the ability of tumor cells to induce angiogenesis was dependent on their differentiation status. To determine this, we used an experimental coculture system to test the ability of Caco-2 cells at progressive stages of differentiation to induce endothelial cell migration, an initial step in the angiogenic process. Endothelial cell migration was maximal when cocultured with undifferentiated Caco-2 cells, an effect that decreased significantly as the cells underwent differentiation (Fig. 6G) ⇓ . As expected, therefore, these observations suggest that undifferentiated, proliferating tumor cells have a greater ability to induce angiogenesis than their differentiated counterparts. Therefore, despite the differentiation-associated increase in expression of several proangiogenic genes, this is outweighed by the reduced expression of others or by the increased expression of inhibitors of chemotaxis, resulting in an overall reduction in endothelial cell migration.

Metabolism.

In addition to the absorption of water and ions, factors that have escaped absorption in the small intestine, such as bile acids and cholesterol, can be taken up by the colonic epithelium (57) . Furthermore, colonic epithelial cells also readily absorb and metabolize short-chain fatty acids produced from the bacterial fermentation of dietary carbohydrate (58) . Consistent with the likelihood that exposure of colonic cells to these substrates is likely to vary according to location along the crypt axis, several genes involved in lipid uptake and metabolism were altered in expression during Caco-2 cell maturation, including muscle fatty acid binding protein, lipase A, hepatic lipase, acyl-CoA dehydrogenase (very long chain), and C-2 to C-3 short chain acyl-CoA dehydrogenase (Fig. 3 ⇓ ; supplementary data).

The increased expression of genes involved in lipid catabolism during colon cell differentiation is consistent with several previous observations. First, levels of fatty acid binding protein, which is involved in fatty acid uptake, increases in vivo as enterocytes migrate toward the lumen along the crypt axis (59) . Similarly, alkaline phosphatase, a brush border protein consistently up-regulated during colon cell differentiation (21) , is thought to play a role in fat absorption (60) . Second, PPARγ, a member of a family of nuclear hormone receptors involved in the regulation of the catabolism and storage of fatty acids, is expressed in an incremental manner up the crypt axis and during Caco-2 cell differentiation (61 , 62) . Finally, increased lipid uptake and metabolism has been linked to differentiation status of colonic epithelial cells in vitro (21 , 63) .

Dietary fats absorbed by enterocytes are assembled into chylomicrons and secreted into the mesenteric lymphatics. Consistent with the increased expression of genes involved in lipid metabolism and uptake, expression of a number of apolipoproteins, a key component of chylomicrons, was concomitantly up-regulated in expression. Among those were apolipoproteins A-1, B, C-III, and H, which are consistent with previous reports (Refs. 64 , 65 ; See Fig. 3 ⇓ ; supplementary data). Collectively, therefore, these observations suggest that increased lipid uptake, metabolism, and packaging are characteristics of differentiated colonic epithelial cells.

In contrast, and possibly reflecting the reduced need for new membrane synthesis after the cessation of cell division, there was a general down-regulation of genes with a role in lipid and cholesterol synthesis, including acetyl CoA carboxylase α, ATP citrate lyase, fatty acid synthase, fatty acid CoA ligase, pyruvate carboxylase, mevalonate decarboxylase, mevalonate kinase, and sterol isomerase, during the maturation program (See Fig. 3 ⇓ ; supplementary data).

Mechanisms.

To gain additional insight into the mechanisms that underlie the reprogramming events described, we examined the intracellular signaling and transcription factor functional classes. The signal transduction class showed only a slight enrichment in the percentage of genes altered in expression relative to the overall changes. This may be because in many cases, the activity of signaling molecules is regulated by post-translational modification events, particularly phosphorylation. Therefore, signaling molecules not altered in expression at the mRNA level may still play critical roles in mediating the differentiated phenotype. However, one pathway that was regulated at the mRNA level was the MAPK signaling pathway with a number of components, including Ras, MAPK6, MAP kinase kinase kinase 2, 8, and 12, MAPK8-IP1, and JAK kinase 1 down-regulated during the maturation program (Fig. 7A) ⇓ .

On the other hand, and consistent with the extensive genetic reprogramming that characterizes Caco-2 cell maturation, a number of transcription factors were both up and down-regulated. Among those down-regulated were members of the high mobility group family of transcription factors (HMG-1, 2, and 14; Fig. 7B ⇓ ). HMG1 and 2 have no known specific DNA recognition sequence but are able to induce bends in DNA, suggesting they may function as architectural factors in processes that require transient alteration of DNA structure such as DNA repair, recombination, replication, and transcription (66) . Therefore, their down-regulation is consistent with the concomitant down-regulation of many of these processes during Caco-2 cell maturation. Similarly, a number of genes involved in alteration of chromatin structure, including several histone acetyltransferases and histone deacetylases, were down-regulated (Fig. 7C) ⇓ . A role for these proteins in regulating colon cell maturation is consistent with previous observations that inhibitors of histone deacetylase activity, such as sodium butyrate, are potent inducers of colon cell maturation (67) .

Several components of the RNA polymerase I, I, and III transcriptional complexes were also down-regulated, an observation consistent with the overall reduction in transcription during Caco-2 cell maturation (Fig. 7, D–F) ⇓ . Finally, several basic helix-loop-helix proteins (HTF4A, ITF-2, and E2A; Fig. 7G ⇓ ) and a number of zinc finger proteins were down-regulated during the maturation program.

Among the transcription factors up-regulated were two members of the retinoid receptor family of transcription factors, RXRα and β (Fig. 7H) ⇓ . RXRs heterodimerize with activated PPARs and alter the transcription of numerous target genes, including several involved in lipid metabolism and transport. One such target, Apo A-1, was shown to be highly up-regulated in the present study (See Fig. 3 ⇓ ; supplementary data). Previous studies have shown that PPARγ is up-regulated during Caco-2 cell differentiation and as colonic epithelial cells migrate upwards along the crypt axis (61 , 62) . Furthermore, overexpression of PPARγ stimulates markers of cell differentiation in colon cancer cell lines, and loss of function mutations in the PPARγ gene is associated with colon cancer (68 , 69) , although studies to the contrary have also been reported (70) . LXRα was an additional nuclear receptor up-regulated during Caco-2 cell differentiation. LXRs heterodimerize with RXRs and regulate the expression of several genes involved in cholesterol metabolism and efflux (71) . Collectively, these changes may be responsible for the alterations in expression of genes involved in lipid and cholesterol metabolism, and transport, during Caco-2 cell differentiation.

A number of other transcription factors up-regulated during Caco-2 cell differentiation were identified (Fig. 7I) ⇓ , each of which may potentially contribute to the maturation program. Of these, the increase in CCAAT/enhancer binding protein-α is particularly interesting, as, consistent with the changes in the present study, it has been implicated in the regulation of p21^WAF-1/Cip-1, and genes involved in lipid uptake and metabolism through induction of PPARγ (72) . Among the other transcription factors, the increase in kruppel-like factor 4 and Id-2 are consistent with their increased expression in differentiated enterocytes of the intestinal epithelium (8 , 73 , 74) .

DISCUSSION

The study of normal colon cell maturation has been hindered by the inability to maintain normal colonic epithelial cells in culture (5) . However, model systems such as the Caco-2 cell line have proven to be a suitable and highly informative alternatives for such studies (13) . On contact inhibition, Caco-2 cells spontaneously undergo cell cycle arrest, with cells accumulating in the G₀/G₁ phase of the cell cycle, followed by differentiation along the absorptive cell lineage (15) . In the current study we used microarray analysis to profile the changes in gene expression that accompany these processes.

As expected, the alterations in gene expression during spontaneous Caco-2 cell differentiation were extensive. By our criteria, 13.2% of all of the genes were altered in expression during the maturation program. The majority of these changes (70%) were reductions in gene expression, a finding consistent with cell differentiation reflecting increased cellular specialization.

The regulation of colon cell differentiation has been suggested to be modulated by a number of stimuli, including contact inhibition, chemotactic gradients, interactions with the ECM, a variety of growth factors, cytokines, and luminal contents (5) . Whereas all of these factors may play significant roles, the extent of the genetic reprogramming observed in the present study suggests that contact inhibition may be the most fundamental of these, and that the ability to undergo differentiation is essentially an inherent feature of colon cells, requiring only the proper trigger to initiate a preprogrammed cascade of events.

One likely trigger is the β-catenin-TCF signaling pathway. We have shown recently that this pathway is down-regulated during Caco-2 cell differentiation as a consequence of both increased E-cadherin and reduced TCF-4 expression (16) . Consistent with this, in the present study we observed a concomitant down-regulation of two well-characterized β-catenin-TCF signaling genes, cyclin D1 and c-myc. Our previous studies showed that down-regulation of this pathway resulted in increased promoter activities in two of the four markers of cell differentiation tested (16) , suggesting at least part of the differentiation response can be attributed to the down-regulation of this pathway. There is also considerable in vivo evidence implicating this pathway in the regulation of intestinal cell maturation. First, down-regulation of β-catenin-TCF signaling through targeted inactivation of the TCF-4 gene results in reduced cell proliferation and the premature onset of differentiation in the small intestine (7) . Similarly, down-regulation of the pathway through forced expression of E-cadherin results in reduced cell proliferation along the crypt axis (75) .

Organization of the microarray database into functional groups enabled changes in biological processes to be more clearly visualized. This approach revealed that colon cell maturation is a highly organized process. First, in parallel with the cessation of cell proliferation, genes involved in cell cycle progression, DNA synthesis, replication, and repair were down-regulated. Second, and reflecting the reduced size of the transcriptome in differentiated cells, components of the basal transcriptional apparatus, and genes involved in RNA splicing and transport, were down-regulated. Third, and consistent with the reduced rate of [¹⁴C]leucine uptake, genes involved in protein translation, folding, and degradation were coordinately down-regulated.

Despite the general down-regulation of components of the translational machinery, including translation initiation and elongation factors, translation does continue in differentiated cells, with several proteins expressed at high levels. One explanation may be the high degree of variability in translational efficiency that exists among different mRNA species. Typically, mRNAs coding for proteins positively involved in growth control (e.g., cyclin D1) are poorly translated in resting cells, attributable in part to the presence of a high degree of secondary structure in the 5′ untranslated region in the cyclin D1 mRNA. The translation of such messages is particularly sensitive to the activity of the CAP-dependent unwinding machinery, components of which (e.g., eIF4E) were down-regulated during Caco-2 cell differentiation (32) . Therefore, it is possible that mRNA species that are up-regulated in expression and which continue to be translated in differentiated cells may be less reliant on eIF4E complexes because of less highly ordered secondary structures.

Our analyses also revealed that colon cell differentiation was associated with increased expression of genes encoding ECM-associated proteins, as well as proteins involved in xenobiotic detoxification, lipid metabolism, and transport. The increase in ECM-associated genes suggests increased deposition of an ECM may be a function of differentiated colonic epithelial cells. Indeed, interaction of epithelial cells with ECM components is known to modulate differentiation (76) . For example, culture of Caco-2 cells on different matrix substrates, such as laminin or collagen IV, enhanced the expression of markers of cell differentiation (77) . The stimulation of cell differentiation may be brought about in several ways: integrin-mediated adhesion of epithelial cells to ECM components can trigger signaling events leading to alterations in cell growth and differentiation. The ECM can also serve as a reservoir for the sequestration of latent forms of growth factors such as TGF-β, which along with TGF-β2, was up-regulated during Caco-2 cell differentiation. TGF-β is secreted as an inactive “latent” complex, which can be subsequently activated by a number of proteases contained within the ECM. On activation, TGF-β can induce a number of changes in epithelial cell growth and differentiation (78) . Whether some of the pathways of Caco-2 cell maturation are mediated by enhanced TGF-β signaling requires additional investigation.

The increased expression of xenobiotic and drug detoxification genes may reflect an important defense mechanism that colon cells have developed to manage their increased exposure to lumenal compounds as they migrate up the crypt axis. However, in the context of a tumor, increased expression of these genes may lead to increased resistance to specific chemotherapeutic agents. Indeed, differentiated Caco-2 cells showed a high degree of resistance to cisplatin and camptothecin-induced apoptosis. Importantly, the ability of gene array analysis to simultaneously examine the expression of large numbers of such genes illustrates the potential use of this methodology for the prediction of response of tumor cells to chemotherapeutic agents.

Finally, the analyses revealed a general up-regulation of genes involved in lipid metabolism and transport, likely reflecting a role for differentiated enterocytes in the absorption and metabolism of lumenal lipids, and their packaging into chylomicrons for transport to the liver.

The analyses also revealed a number of pathways that may drive these reprogramming events, including chromatin-modifying enzymes, signaling pathways such as TGF-β, MAPK, and RXR, as well as many other transcription factors. The challenge of future studies will be to dissect the contribution of these individual pathways, and more important, their interaction to the overall maturation program.

In conclusion, this study demonstrates that the genetic reprogramming that accompanies colon cell maturation occurs in a highly organized and coordinate manner. Most important, the complexity of change, even within some functional groups of genes, and along a single differentiation lineage emphasizes that cell maturation reflects an interplay and balance among many pathways.