This study has been published [76]: Publication IV
Riehl A, Bauer T, Brors B, Busch H, Mark R, Németh J, Gebhardt C, Bierhaus A, Nawroth P, Eils R, König R, Angel P, Hess J: Identification of the Rage-dependent gene regulatory network in a mouse model of skin inflammation. BMC Genomics 2010 Oct 5;11:537.
5.1. Motivation
Throughout the previous studies I showed that analyzing gene expression and transcriptional regulation by TFs grants insight into biological mechanisms of tumorigenesis and provides a better understanding of the nature of cancer entities, specifically of neuroblastomas. In the project described in this chapter, we worked together with Dr. Astrid Riehl and colleagues to address the questions how signaling triggered by the receptor for advanced glycation end products (Rage) contributes to the establishment and maintenance of a pro-inflammatory microenvironment that supports neoplastic transformation and malignant progression of skin cancer.
Unresolved inflammation is thought to foster multiple hallmarks of cancer [65] by providing a tumor-promoting environment. A better insight into molecular mechanisms underlying tumorigenesis in the context of inflammation, such as signaling and gene regulatory networks, is implicitly required. Rage is a signaling protein that mediates and maintains the strength of inflammatory responses in a mouse model of skin carcinogenesis upon inflammation [77]. Rage functions as a pattern recognition receptor with multiple ligands and is expressed at increased levels at sites of inflammation. Several downstream target genes have been identified that are expressed context-specifically [78] and have been implicated in neoplastic cell transformation and tumor progression [79-81]. However, in- between mediators of Rage signaling and involved TFs remain mostly unknown. In this project, my cooperation partners conducted time-resolved gene expression profiling of skin samples taken from Rage-/- and wild-type (wt) mice treated with tetradecanoyl phorbol acetate (TPA), a potent inducer of inflammation and tumor promoter, to identify genes that are affected by Rage signaling. Subsequently, I applied comprehensive TFBS scans to predict associations of TFs with gene sets exhibiting Rage-dependent gene expression patterns. I further divided the extracted Rage-responsive genes into clusters and predicted associations of TFs with these more specific gene sets. My analysis provided candidate TFs that were investigated on the protein level in skin samples and found to be involved in the gene regulatory network downstream of Rage signaling.
5.2. Main results
5.2.1. Rage-dependent gene expression profiles exhibit two temporal phases in response to TPA stimulation
Rage-/- and wt mice (three biological replicates each) were treated with TPA. Gene expression profiles of skin samples were measured at 6, 12, 24, and 48 hours after treatment and compared to TPA-untreated controls. Transcripts were ranked by average and peak expression relative to controls of respective genotypes, yielding 341 common transcripts
Figure 5.1 | Gene expression dynamics of wt and Rage-/- mice upon TPA treatment. A Back skin was isolated 6, 12, 24, or 48 hours after TPA stimulation. Non-treated and acetone-treated mice served as controls (0). Microarray global gene expression analysis of RNA samples was performed (three replicates for each genotype and time point). B Following quantile normalization, wt genes were ranked according to high mean and peak expression separately to filter for TPA- responsive genes. A common subset of 341 genes was identified among the 1000 top-ranked genes of three experiment series. C K-means clustering produced six clusters of which cluster 3 and 6 were the largest. The kinetic of these clusters
showed a response
independent of the Rage genotype at t=6h, but the stimulus response of both repressed (cluster 3) and induced (cluster 6) transcripts was only sustained in wt mice.
among the three replicate series (Figure 5.1). These transcripts were divided into six clusters by k-means clustering. Expression profiles, particularly of the two largest clusters (cluster 3, n=125; cluster 6, n=84), were in line with the previously described function of Rage in sustaining inflammatory stimuli: cluster 3 contained transcripts that were repressed six hours after TPA induction in both genotypes. The repression was persistent throughout later time-points in wt, but not in Rage-deficient mice. Vice versa, cluster 6 transcripts were induced transiently in Rage-/- mice, but maintained at induced levels in wt animals. These results indicate that gene expression dynamics of genes targeted by Rage can be divided into two phases. The first phase is characterized by an early Rage-independent response to the inflammatory stimulus, whereas Rage is essential for sustaining the changed transcription levels in the subsequent second phase.
Taken from Publication IV (Riehl et al., 2010)
5.2.2. Rage-dependent differential expression after TPA stimulation
We aimed at identifying genes that were differentially expressed between Rage-/- and wt mice in the second phase after the TPA stimulus. (This analysis was conducted independently of the clustering described in the previous passage.) A linear model was applied that revealed differential expression at a significant level (corrected P<0.05) only at time point t=24h upon TPA administration. A total of 122 transcripts (representing 114 different genes) were differentially expressed at this time point, including induced and repressed genes.
5.2.3. Predicting TFs of the Rage-dependent gene regulatory network
I applied hierarchical clustering employing Pearson correlation distances to the gene expression values, thereby revealing three clusters with distinct expression profiles (Figure 5.2). Then I assessed which TFs were likely to regulate the individual clusters as well as the whole set of differentially expressed genes. I conducted PWM scans on promoter regions (±2 kb of the TSS) with all PWMs available for TFs in Transfac database [2]. After mapping TFBSs from PWMs to TFs, I calculated over-representation of TFBSs in the gene sets by Fisher’s exact tests. Within all differentially expressed genes, putative binding sites of 17 TFs and
Cluster size 42 49 23 Cluster label 3 1 2 0 6 12 24 48 0 6 12 24 48 0 6 12 24 48
Corr
ela
tion
dis
tance
0 0.5 1.0 1.5 2.0 0 -2 -4 2 4 0 -2 -4 2 4 0 -2 -4 2 4Time [h] Time [h] Time [h]
R ela tiv e lo g2 e xp ressio n ○○wtRage-/- ○wt ○Rage-/- ○wt ○Rage-/- Figure 5.2 | Hierarchical clustering of differentially expressed transcripts
between wt and Rage-/- mice
24 h after TPA treatment. Three distinct clusters were derived from the dendrogram. Clusters 1 and 2 contained significantly up-regulated transcripts at 24h in wt mice, whereas cluster 3 transcripts
were significantly down-
isoforms were significantly enriched (corrected P<0.05, Table 5.1). Some TFs could thereby be associated with individual clusters: trans-acting transcription factor 1 (Sp1) and 4 (Sp4), hepatic nuclear factor 4 (Hnf4), and CAC-binding protein (CAC-bp) with cluster 1, Wilms tumor 1 homolog (Wt1) with cluster 2, and E2F transcription factor (E2f) with cluster 3. TFBSs for MAZ related factor (Mazr) were enriched in both clusters 1 and 2.
Table 5.1 | Enriched TFBS in differentially expressed genes 24 hours after TPA stimulation.
Clusters TF Fischer test
P-value Corrected P-value Cluster genes with PWM without PWM all Sp1 5.33E-07 1.06E-04 94 3
Sp1 isoform 1 5.33E-07 1.06E-04 94 3
Sp4 1.72E-06 2.27E-04 83 14
AP-2beta 1.24E-06 1.22E-03 77 20
AP-2alpha 1.60E-05 1.27E-03 79 18
AP-2gamma 2.13E-05 1.41E-03 79 18
MAZR 6.03E-05 3.41E-03 74 23
CAC-binding protein 1.32E-04 6.52E-03 81 16
Egr-1 3.56E-04 1.56E-02 85 12
Egr-3 4.38E-04 1.73E-02 78 19
E2F 4.85E-04 1.75E-02 58 39
c-Myc 7.31E-04 2.41E-02 67 30
Egr-2 9.69E-04 2.94E-02 80 17
COUP-TF1 1.06E-03 2.94E-02 89 8
WT1 1.19E-03 2.94E-02 67 30
WT1-isoform1 1.19E-03 2.94E-02 67 30
COUP-TF2 1.45E-03 3.38E-02 48 49
cluster 1
Sp4 2.01E-06 7.93E-04 40 2
Sp1 6.13E-05 6.86E-03 42 0
Sp1 isoform 1 6.13E-05 6.86E-03 42 0
MAZR 6.93E-05 6.86E-03 36 6
HNF-4alpha7 1.84E-04 1.46E-02 29 13
CAC-binding protein 3.12E-04 2.06E-02 38 4
cluster 2
MAZR 2.14E-04 7.64E-02 19 1
WT1 5.79E-04 7.64E-02 18 2
WT1-isoform1 5.79E-04 7.64E-02 18 2
cluster 3 E2F 1.88E-05 7.45E-03 28 8
E2F-1 8.50E-05 1.68E-02 28 8
Taken together, the enrichment analyses predicted several TFs that had not been reported previously in connection to Rage signaling. These TFs were therefore promising
candidates for further investigation. In particular, TFs associated with cell cycle and tumor pathology, such as E2f and Wt1 proteins, are of interest in the context of carcinogenesis.
5.2.4. Expression of E2f TFs upon induction of Rage signaling by TPA stimulation We wanted to test the hypothesized involvement of members from the E2f family of TFs in Rage signaling. No significant changes in gene expression levels were detected in the expression data, so we considered post-transcriptional alterations that would affect the activity of these TFs. Protein levels of E2f1, a transcriptional activator, and E2f4, a transcriptional repressor, were quantified on Western blots and further visualized by immunohistochemical staining.
Indeed, protein levels of the transcriptional activator E2f1 were induced in keratinocytes of both Rage-/- and wt samples at time points t=6h and t=12h after TPA stimulation. While the expression level was still kept up high in wt keratinocytes 24 hours after TPA stimulation, it was not increased in Rage-deficient skin lysate. E2f4 protein was induced and a constant increase was observed until 24 hours after TPA stimulation in wt samples. In contrast, E2f4 levels were not induced and remained constant in Rage-/- samples. The findings are consistent with the two phase model of Rage signaling described earlier and corroborate a potential impact of Rage signaling on E2f activity. Whether this impact is direct or indirect could not be concluded at this time, however. It is worth mentioning that the E2f- Rb pathway is critical for strict regulation of cell cycle progression and often directly targeted in carcinogenesis. It is therefore plausible that downstream targeting of this pathway by Rage signaling may provide a molecular link between inflammation keratinocyte hyperproliferation supporting skin cancer development. The results from the immunohistochemical staining were in line with the protein dynamics observed on the Western blots and provided an additional highly informative visualization.