HDACis were reported to influence the response towards EGFR-inhibition in several cancer models.225,226,227,228,229 To test, if these findings might be adapted to our settings the cell lines SW480, RKO, LIM1215 and CaCO2 were treated with HDACis and af-terwards treated with Erlotinib or Gefitinib. LIM1215 cells showed the strongest effect on sensitivity towards Erlotinib and Gefitinib after Valproat treatment (see figures 47 and 51). Since AREG and EREG expression increased upon HDACi-treatment, a con-sequence was to test, if AREG or EREG are responsible for the increased sensitivity.
LIM1215 cells were transfected transiently or by lentiviral infection with an AREG -sequence containing plasmid. As shown in section 3.10, in none of the experiments, an increased sensitivity towards EGFR-inhibition was achieved althoughAREG protein amounts and mRNA-expression increased in a similar way than after HDACi-treatment (compare figures 6, 7 and 10 with 61). Direct application of recombinant AREG into the
supernatant did also not lead to an increased sensitivity towards Erlotinib or Gefitinib (data not shown). Although LIM1215 cells were described to be a KRAS-wildtype cell line,275 in our laboratory a mutation within the KRAS gene was detected in exon 4 (see table 12). It might be possible thatAREG over-expression alone will not affect sensitiv-ity due to thisKRAS mutation. It could also be possible that the effect of the HDACis onto sensitivity is independent of AREG. To clarify this problem, a low AREG ex-pressing cell line would be necessary without any mutations within the MAPK-pathway components. However, such a cell line was not available during this work. Therefore, the direct contribution of AREG to the sensitivity towards EGFR inhibitors could not be proven by the experiments presented here. Nevertheless, in a publication by Ferraros et al., the authors verified a causal relationship between AREG and EREG expression and response to Cetuximab in KRAS-wildtype cells.276 In contrast to the experiments presented here, they studied the effects of AREG and EREG expression on sensitivity towards Cetuximab by decreasing the levels of these proteins. They showed that tar-geting either AREG or EREG gene expression with shRNAs led to decreased response rates towards Cetuximab.
A further interesting issue is that in LIM1215 cells HDACi-treatment increases sensitivity towards EGFR-targeted therapeuticsKRAS mutation independently. A con-sequence of this experiment could be that HDACis might have the potential to overcome KRAS mutation dependent resistance towards EGFR-targeting therapeutics. But, if this is true, it is unclear why the HDACi-treatment did not affect the other KRAS-mutated, or even the KRAS-wildtype cell lines, in a similar way. One reason might be that the cell lines tested have different response rates towards the HDACis used. Also, the target HDACs might account for the dissimilar effectiveness of the HDACis. In LIM1215 cells, Cambinol, for example, did not increase sensitivity towards EGFR-inhibition, although Trichostatin A and Valproat did (see figure 47). Nevertheless, while TSA and Valproat are mainly targeting HDACs class I and II, Cambinol targets the HDAC Sirtuin I, a HDAC class III.277 SAHA, in contrast, was not tolerated by LIM1215 at all. Addition-ally, it was tested if these findings can be applied in vivo. LIM1215 cells were injected in mice to create xenografts. But, in contrast to the in vitro results, HDACi treatment did not lead to an increase of senstitivity towards EGFR inhibitors (see figure 54).
As a second approach the mice, injected with LIM1215 cells, were treated with the DNA methyltransferase inhibitor 5-Azacytidine. By using the DNA methyltransferase inhibitor DAC inin vitroexperiments,AREG andEREG expression increased (see figure 4). But, as already described, post-treatment procedures were not possible, since the cells died after replating. Nevertheless, in mice 5’Azacytidine treatment, which is also a DNA methyltransferase inhibitor, led to an increased response to Erlotinib (see figure 54). Therefore, in in vivo systems global epigenetic changes might also play a role for EGFR-treatment response. But, when testing theAREG and EREG expression in the xenografted cells, it was observed that treatment did not influence mRNA-expression of
AREG orEREG (see figure 55). Additionally, IHC-experiments showed that the AREG protein expression varied strongly among the xenografted cells treatment independently (see table 21 and figure 56). As a consequence AREG or EREG mRNA-expression as well as AREG protein amounts might not correlate with the increase of sensitivity in this experimental model or cannot be measured precisely. It was, for example, hardly possible to analyze the 5-Azacytidine treated samples, because the tumor growth was reduced dramatically upon treatment. Nevertheless, a decreased methylation index was observed in theAREG intragenic CpGs p150 and p220 as well as in theEREG intragenic CpG p297 upon 5-Azacytidine treatment alone or in combination with Erlotinib (see figure 57). Therefore, the epigenetic regions might function as predictive markers in combination therapies, when the expression of AREG cannot be evaluated. The lack between methylation and expression in the xenograft model has to be further addressed.
Overall, the results indicate for clinical settings that patients, which do not respond to EGFR-targeted therapies might become sensitive after treatment with epigenetically interfering compounds by different, yet largely unknown mechanisms.
4.7 AREG CpG p150 and CpG p220 are differentially methyl-ated in human tumor samples
Subsequent experiments were performed in the ColoNET-consortium to evaluate if the results of this work could also be applied to human tissues. As already described,KRAS -wildtype and mutant tumors were grown on mice and the xenografts were measured for their AREG mRNA and protein expression and for their response towards Cetuximab (see supplementary material 8.11). Similar to the cell lines, AREG was differentially expressed in human carcinoma-derived xenografts. Additionally the xenografts also re-sponded differently to Cetuximab according to their mutation status andAREG expres-sion (see supplemetary tables 27 and 28).
As a next approach, methylation of the identified intragenic CpGs in the AREG gene was measured in 5 different human tissue samples. Each tissue was separated by Florian Rossner (Charit´e Berlin, Institute for Pathology) into 4 parts using microdi-section: normal epithelium, normal stroma, tumor epithelium and tumor stroma. The results are shown in supplementary material 8.12. Three of the 4 normal epithelium samples (1 sample failed) and 3 of 5 normal stroma samples showed high methylation within the tested CpGs (MI >0.5). Four of 5 tumor stroma samples showed also high methylation. Interestingly only 1 of 5 tumor epithelium samples showed high methy-lation for one CpG (CpG p220), which means that the tumor epithelium samples were mainly unmethylated at the tested CpGs, whereas in the other compartments the CpGs were mainly methylated. To verify the results, the approach was repeated with 20 addi-tional human tissues. Addiaddi-tionally to the AREG CpGs p150 and p220 also the EREG CpGs p143 and p297 were tested. The boxplots of all methylation indices observed in
this experiment are shown in supplementary material 8.13. Similar to the results before, the methylation at theAREG CpGs were in general lower in the tumor epithelium com-partments compared to the tumor stroma, the normal stroma or the normal epithelium.
This was also seen for the methylation of the EREG CpGs. However, looking in detail to the samples, there are also 5 samples, which show a high methylation index greater than 0.6 at the AREG CpG p150 and CpG p220 in the tumor epithelium. In contrast, for EREG the highest methylation index is only 0.6. Nevertheless, it needs to be clari-fied to what extent the differential methylation of the AREG and EREG CpGs within the tumor epithelium samples and between the different tissue types are influenced by remaining tumor stroma cells within the tumor epithelium samples. It would also be interesting to clarify with corresponding patient data, if the methylation ofAREG CpG p150 and CpG p220 within the tumor epithelium might also predict the outcome of EGFR-targeted therapies, as it was seen for the AREG expression in in vivo experi-ments using xenografts (see section 4.5). However, to test that, the number of tested samples has to be increased dramatically to ensure significant results.