• No se han encontrado resultados

11. Discusión

11.2. El conflicto en el marco de las dinámicas de convivencia

11.2.3. El chisme y el rumor en las redes sociales

The aim the experiments outlined in this chapter was to further characterise loci identified as differentially methylated between sensitive and resistant A2780 human ovarian cell lines, by DMH using a 12K CGI microarray, and attempt to identify candidate biomarkers of acquired drug resistance or identify key genes or pathways involved in ovarian cancer development.

Methylation of NR2E1 and LMX1A was increased in resistant cell lines and matched tumour pairs and therefore could represent loci that are selected for during the course of chemotherapy. This would however require prospective validation within a clinical trial setting and the number of matched tumour pairs was both limited and not annotated in terms of survival– it is not known whether the matched samples represent platinum sensitive or resistant disease and this is clearly important. In addition as discussed the biological subtypes of ovarian cancer clearly have now been identified as having a key role in the response to chemotherapy (Lalwani, Prasad et al. 2011), However at the time of analysis the subtypes of EOC used in these samples was not determined.

As only 3 from an expected 6 genes were identified as potential novel markers of acquired drug resistance in epithelial ovarian cancer, from an original list of 41 sequences, this was deemed to be a relatively low yield. There could have been various reasons for this. Since the original DMH experiment (Heisler, Torti et al. 2005) larger and better annotated arrays are now available. The human genome is thought to consist of approximately 45000 CGIs and this library included only 12000 of these (Cross, Charlton et al. 1994). The annotation of this library frequently changed so it is possible that genes were missed that should have been investigated or vice versa. Cross et al demonstrate in the initial validation of DMH

which uses an MSe1 digest in combination with an MBD binding column that 77% were likely to be CGI. Ten percent represented rDNA and 10% were bulk DNA (Cross, Charlton et al. 1994). In the table at the beginning of this chapter it is shown that 13 of 41 sequences fulfilled our criteria of being a CGI and having a 5‟ location and this would argue that there are significantly more false positives in the library than proposed.

In addition the statistical method used to extract candidate loci from the DMH data was originally developed for the interpretation of gene expression microarray experiments and it is possible that vital information is lost by not taking into account the unique biological differences between changes in methylation and changes in expression. In expression experiments a small number of losses and gains in expression are seen whereas when analysing methylation a larger number of changes are seen and these tend to be asymmetrical – with more loci showing an increase in methylation in the resistant cell lines than a decrease. The normalisation that is necessary in RNA microarray experiments could be detrimental when analysing methylation data and valuable information could be lost.

The 34 cell lines panel is weighted towards the A2780 cell line series which is in vitro generated and it is possible that sequences and hence genes identified from this panel could have less biological relevance. Of note for all genes in this chapter although a clear increase in methylation was seen in the A2780 sensitive and resistant cell lines it was minimal in the other cell line pairs raising the question as to whether this is an A2780 effect only, and therefore of less biological relevance.

MSP was initially used to screen the 6 genes that were selected for further characterisation. For PTTG a decrease in methylation was expected but instead no methylation in any cell line was seen. For CNTNAP it was not possible to optimise primers and for CR2 no increase in methylation was seen in the resistant cell lines. MSP is non-quantitative and optimisation and analysis of the gels relatively subjective. Since the start of this thesis pyrosequencing of bisulphite modified DNA has effectively replaced this technique.

With both MSP and pyrosequencing it is possible to gain a false negative result because the primers were not designed to amplify exactly the same area where the maximal difference in methylation was identified by PAM and similarly both techniques only investigate a few CpG residues so it is possible to miss increased methylation by a relatively small change in primer location. Using MSP false positive results can be obtained from incompletely methylated DNA (although a calponin PCR had been preformed to limit the effect of this).

After confirming an increase in methylation in the ovarian cisplatin resistant cell lines I investigated for methylation in primary tumours. The assumption was that if methylation was playing a role in these tumours that a heterogeneous pattern of methylation would be seen in the tumours – with the anticipation that this would then be increased in resistant tumours had these been available. However DLC1 showed no methylation in primary tumours and was excluded from further analysis. There is a risk with this as if this gene was a marker of acquired resistance it would not necessarily need to show this heterogeneous pattern of methylation at presentation that we predicted.

In Chapter 7 the methylation status of these loci are re-examined in primary tumours on the OGT customised array. This gave an additional chance to investigate the loci for which MSP or pyrosequencing primers were not optimised or where differential methylation had not been observed (CNTNAP and CR2).

The tumour pairs that were available were obtained from patient‟s pre- and post- chemotherapy at the time of surgery for residual disease. If the patients had had a sufficient response to be considered eligible for such an approach it is possible that the disease that was left behind was not drug resistant and therefore these tumour pairs would not be suitable for detecting changes in methylation associated with resistant disease. No clinical details were available for these samples to indicate response to treatment. At the time of subsequent chapters the relapsed disease pairs were available and it would be

interesting to go back and pyrosequence these (assuming they represent chemoresistant relapse).

The correlation between methylation and expression was next examined. This was done in a limited number of cell lines and not performed in tumours as RNA was not available. For these candidates to be of true biological relevance it needs to be demonstrated that increased methylation at the promoter consistently correlates with decreased mRNA gene expression and that this can be reversed with a demethylating agent – both in cell lines and in tumours. In addition it would be important to assess the effect of methylation on protein expression using for example western blot analysis. Another important series of experiments would assess the functional impact of increased or decreased expression of the genes. As NR2E1 and LMX1A have both been associated with embryonic stem cells and perhaps tumour sustaining cells it would be very interesting to investigate this in the ovarian cancer setting.

4 Characterisation of loci showing differential

methylation in cisplatin resistant lines identified

by methylation linear discrimination analysis

(MLDA).

4.1 Background and aims

The aim of the experiments in this chapter was to assess the ability of a novel statistical package, MLDA (Dai, Teodoridis et al. 2008), developed in our laboratory, to identify differentially methylated loci from the previously described 12k array. The 16 A2780- based sensitive and resistant cell lines (chapter 2.18.4) were used and candidates investigated by MSP +/- pyrosequencing. It was hoped that these candidates could also represent potential markers of acquired cisplatin resistance.

As outlined in Chapter 2.24.4 and 2.24.5, PAM and SAM were originally developed to interrogate microarray expression data rather than methylation data. They rely on some important assumptions in order to normalise the data. Firstly, that only a small percentage of genes will change expression, or in this case methylation status, and secondly that these changes will show symmetry, i.e. as many will show an increase in methylation as a decrease when comparing, for example, sensitive and resistant ovarian cancer cell lines.

However this is not usually the case when analysing methylation data where many loci are predicted to change their status and this change is likely to be asymmetrical because an „enrichment‟ is seen with more sequences gaining methylation in resistant cell lines than losing it.

It was predicted that MLDA could therefore have advantages in detecting loci that optimally discriminate these isogenically matched ovarian cancer cell lines, without using

arbitrarily chosen cut offs or losing vital information as a result of over normalisation of the data.

As this was the first time sequences generated using MLDA had been examined in the laboratory we wished to do this in the most unbiased manner possible. Using the MLDA score, PAM and SAM an extensive list of candidate sequences was generated when comparing the more limited panel of 16 sensitive and resistant epithelial surface ovarian cancer cell lines, referred to as the A2780 cell lines in future text. These included A2780, A2780 p3, A2780 p5, A2780 p6, A2780 p13 and A2780 p14 (6 sensitive) and A2780 CP70 and MCP 1-9 (10 resistant). Nine sequences were chosen entirely at random, which it was hoped reflected samples that ranked high and low by MLDA/PAM/SAM, from the generated list and firstly examined in the same 16 ovarian cancer cell lines, by MSP. The MLDA ranking for each sequence is shown in table 26 below:

Table 26 Candidate loci from MLDA analysis of 16 A2780 cell lines.

Loci Gene Chromosome Sensitive

Score* Resistant Score** MLDA Ranking*** 3 A 11 MLLT6 17 -0.86 0.01 27 17 H 9 HRASLS3 11 0.10 0.92 34 20 F 11 NTN4 12 -0.81 0.03 41 21 A 11 NTN4 12 -0.86 0.25 9 24 D 3 SP5 12 0.25 0.71 75 38 D 7 AGBL2 11 -0.91 0.18 11 41 D 12 GLS2 12 -0.81 0.00 36 101 G 6 GLS2 12 -0.91 0.025 21 121 D 9 CRABP1 15 -0.65 0.65 2

Sensitive score 1* is the average MLDA score in the sensitive cell lines , Resistant score 2** is the average MLDA score in the resistant cell lines. MLDA ranking*** is the rank of standardised residuals to the robust regression line constructed by the averaged sensitive scores against averaged resistance scores. Further information is available in table 2, page 9, Dai et al (Dai, Teodoridis et al. 2008).

4.2 Examination of candidate loci in ovarian

Documento similar