Proteomic characterization of peroxisome proliferator-activated receptor-γ (PPARγ) overexpressing or silenced colorectal cancer cells unveils a novel protein network associated with an aggressive phenotype

Abstract

Peroxisome proliferator-activated receptor-γ (PPARγ) is a transcription factor of the nuclear hormone receptor superfamily implicated in a wide range of processes, including tumorigenesis. Its role in colorectal cancer (CRC) is still debated; most reports support that PPARγ reduced expression is associated with poor prognosis. We employed 2-Dimensional Differential InGel Electrophoresis (2-D DIGE) followed by Liquid Chromatography (LC)-tandem Mass Spectrometry (MS/MS) to identify differentially expressed proteins and the molecular pathways underlying PPARγ expression in CRC progression. We identified several differentially expressed proteins in HT29 and HCT116 CRC cells and derived clones either silenced or overexpressing PPARγ, respectively. In Ingenuity Pathway Analysis (IPA) they showed reciprocal relation with PPARγ and a strong relationship with networks linked to cell death, growth and survival. Interestingly, five of the identified proteins, ezrin (EZR), isoform C of prelamin-A/C (LMNA), alpha-enolase (ENOA), prohibitin (PHB) and RuvB-like 2 (RUVBL2) were shared by the two cell models with opposite expression levels, suggesting a possible regulation by PPARγ. mRNA and western blot analysis were undertaken to obtain a technical validation and confirm the expression trend observed by 2-D DIGE data. We associated EZR upregulation with increased cell surface localization in PPARγ-overexpressing cells by flow cytometry and immunofluorescence staining. We also correlated EZR and PPARγ expression in our series of CRC specimens and the expression profiling of all five proteins levels in the publicly available colon cancer genomic data from Oncomine and Cancer Genome Atlas (TCGA) colon adenocarcinoma (COAD) datasets. In summary, we identified a panel of proteins correlated with PPARγ expression that could be associated with CRC unveiling new pathways to be investigated for the selection of novel potential prognostic/predictive biomarkers and/or therapeutic targets.

Keywords: 2-D DIGE; Colorectal cancer; Ingenuity pathway analysis; Mass spectrometry; Peroxisome proliferator-activated receptor gamma; Proteomics.

Figure 1

Western blot analysis of HT29 and HCT116 silenced or overexpressing PPARγ protein. In HT29 and HCT116 cell lines PPARγ was stably silenced or overexpressed to generate HT29 sh‐PPARγ and HCT116 PPARγ cells, respectively, as compared to cells transfected with empty vectors, the HT29‐cc and HCT116‐cc cells (Pancione et al., 2010). The western blot illustrated is representative of three independent experiments (**p < 0.05). β‐actin ensures equal loading of samples in each lane.

Figure 2

Schematic representation of the 2‐D DIGE procedure from sample preparation and labeling to gel analysis. A representative image of a DIGE gel with extracts from HT29‐cc plus HT29 sh‐PPARγ and HCT116‐cc plus HCT116 PPARγ is shown and the number of proteins differentially expressed is reported for each cell model (A). PCA plots and 2‐D DIGE spot maps. Unsupervised multivariate analysis showed good experimental reproducibility as demonstrated by the close relation between the three biological replicates in the 2‐D DIGE results. In detail, the six spot maps for each experimental group clearly clustered into two groups corresponding to HT29 sh‐PPARγ (red) vs HT29‐cc (black) (B) and HCT116 PPARγ (red) vs HCT116‐cc (black) (C). In order to be included in the analysis, protein spots had to be found in 80% of the spot maps and to display an expression variation of at least 1.4‐fold with a Student's t‐test (p < 0.05). Representative Deep Purple™‐stained spot maps of HT29 sh‐PPARγ vs HT29‐cc (D) and HCT116 PPARγ vs HCT116‐cc (E). All the detected differences between the two experimental groups are visualized by circles. For MS‐identified protein spots, the spot numbers match those listed in Tables 1 and 2.

Figure 3

Functional distribution of proteins identified by 2‐D DIGE/MS. Pie chart of HT29 sh‐PPARγ vs HT29‐cc (A) and HCT116 PPARγ vs HCT116‐cc (B). Cellular component annotation of the identified proteins was performed by STRAP software (C).

Figure 4

2‐D DIGE protein spots quantification. The graph view, representing the normalization procedure generated by DeCyder software, reports the corresponding log‐transformed standard abundance of the protein spot of interest (y‐axis) identified in the gels obtained from HT29 sh‐PPARγ vs HT29‐cc and HCT116 PPARγ vs HCT116‐cc cells extracts (x‐axis) in three replicates. A 3D view of the 2‐D DIGE quantification for each spot is illustrated below. The numbers in each graph indicate the fold change values in spot levels from 2‐D DIGE data relative to control cells as reported in Tables 1 and 2.

Figure 5

Visual representation of the network generated by Ingenuity Pathway Analysis (IPA). The network included the 5 identified/validated proteins and PPARg with direct and indirect interactions (EZR, LMNA, ENOA, PHB, and RUVBL2 orange stained) with PPARg (red stained) and localized to a specific cell compartment. Network proteins are visualized by proper symbols, which specify the functional nature of the protein. Each node represents a protein, the solid and dotted lines represent the direct and indirect associations among proteins, respectively. Proteins with no background color were undetected in the study but were inserted by IPA to produce a highly connected network.

Figure 6

Western blotting validation of proteins identified as differentially expressed in the 2‐D DIGE/MS analysis. Protein lysates from HT29 sh‐PPARγ, HT29‐cc, HCT116 PPARγ and HCT116‐cc cells were immunostained with anti‐PHB, anti‐RUVBL2, anti‐EZR, anti‐LMNA and anti‐ENOA antibodies. γ‐tubulin immunoblotting or ponceau staining ensured equal loading of samples in each lane. Densitometric analysis of western blotting was obtained using the software ImageJ (Schneider et al., 2012), normalized on loading control and reported as relative protein levels compared to corresponding control cells considered equal to 1 (dotted line).

Figure 7

Correlation of protein abundance with mRNA expression levels. . PHB, RUVBL2, EZR and LMNA mRNA levels were detected by qRT‐PCR from HT29 sh‐PPARγ vs HT29‐cc and HCT116 PPARγ vs HCT116‐cc (A). 18S mRNA was used as control. The amplified products were run on a 2% agarose gel and stained with ethidium bromide. Expression levels in HT29 (B) and HCT116 (C) cells were normalized to 18S and calculated as fold change (2−ΔΔCT) with respect to control cells (*, p ≤ 0.05 value; **, p ≤ 0.01).

Figure 8

PPARγ enhances EZR localization on the surface of CRC cells. Flow cytometry assessment of EZR surface exposure. CRC cells incubated with fluorescence‐labeled secondary antibodies alone served as background controls (Gray). Dotted lines mark the MFI (mean fluorescence intensity) of the EZR‐FITC signal observed in CRC cells modified for PPARγ expression (A). The histograms report the calculated MFI values of EZR‐FITC in the different CRC cells lines investigated (*, p ≤ 0.05). (B). Immunofluorescence analysis of EZR surface localization (red) in HCT116‐cc, HCT116 PPARγ, HT29‐cc, and HT29 sh‐PPARγ cells; nuclei are stained with DAPI. Scale bar, 5 μm (C). Representative immunoblots for EZR in frozen CRC specimens (T) with high or low PPARγ expression matched with normal mucosa (N) identified in the same cohort of patients (Pancione et al., 2010). Ponceau red was used as a loading control (D). Densitometric analysis (histograms on the right) was performed by ImageJ software (National Institute of Health, USA) and EZR protein expression levels were normalized to Ponceau Red staining and reported as mean of arbitrary units. Error bars in the figure indicate SEM (*, p ≤ 0.05).

Figure 9

Tumor vs normal distributions of PPARγ, RUVBL2, EZRIN, PHB, LMNA A/C and ENOA. The box plots depict the mRNA levels in normal (n = 21) vs tumor tissues (n = 210) in the TCGA COAD dataset.