While microRNA expression profiling has been shown to be highly specific for cancer typing, tumor diagnosis and identifying tumors of unknown origin, it is not yet clear if microRNAs can perform as markers of disease progression – a crucial requirement for identifying those conditions which require treatment. Ideally, such markers of progression would show systematic changes in expression in precancerous tissues, which correlate with disease progression to invasive cancer. Clearly, this requires access to large numbers of clinical samples which cover the range of histologically defined tissue types involved in disease progression. In this study we investigated the expression levels of two key MicroRNAs identified in the literature (and their validated protein targets) in cervical tissue stratified by histological diagnosis in order to determine their utility as biomarkers for early cancer diagnosis. We found that mir-21 expression was significantly increased only in ICC with respect to other clinical samples and that mir-143 expression did not correlate with histology at all. Cervical cancer cell line results, while consistent with previous studies on the same cell lines, appeared to give discordant microRNA expression results in comparison to clinical samples. Interestingly, PDCD4 protein expression increased in precancerous lesions then decreased significantly in ICC, possibly as an effect of mir-21-induced silencing. These results highlight the need for further studies focusing on tissue profiling in preference to cell lines, where expression levels of candidate MicroRNA biomarkers can be assessed in pre-cancerous disease, where their utility in cancer diagnostics would prove most beneficial. We identified key inconsistencies in microRNA expression profiles in our clinical specimens as compared to cervical cancer cell lines. However, recent investigations show significant differences in microRNA expression based on cell culture conditions suggesting that cell line experiments may not be particularly informative when searching for clinically relevant microRNAs 
. Cheng et al.
showed significant increase in HeLa cell growth and proliferation in response to inhibition of mir-21, whereas Yao et al
showed significant decrease in cell growth and proliferation for the same treatment. Wang et al
further showed that changing culture conditions from a monoculture to a suspension raft culture led to a completely different set of significantly deregulated microRNAs. Similar discordant results have been seen in other studies investigating the relevance of cell culture models in comparison to in vivo
tissue of spheroidal cultures 
. Also, other gene/protein expression differences have recently been identified, including the presence of WNT5A expression in gastric cancers (not expressed in cell lines) 
and differences in DNA methylation 
and gene expression 
signatures between colorectal carcinomas and colon cancer cell lines. Clearly, a key problem with the cell line approach is that potential biomarkers are identified in cells immortalized from invasive cancers – it is often completely unknown if these biomarkers can be detected in pre-cancerous disease, which requires analysis of clinical samples. A further complication in the current study may be the racial background (West African) in comparison to cancer cell lines which are Caucasian – however little information is known about such variation.
Increased mir-21 expression has been identified in a large number of tumor types 
and in this study has shown significant association with the development of histologically-defined ICC. Furthermore, our results show that mir-21 may be of some utility in predicting the progression of pre-cancerous lesions into ICC, as women with high mir-21 values may be expected to progress. However, further studies involving larger sample sizes and patient follow-up are required to investigate this aspect, as there are clearly insufficient data for proper testing of this hypothesis by splitting into training and testing sets. Furthermore, while the mir-21 expression levels of women with normal pathology with and without HRHPV infection are not statistically different, the HRHPV- group showed approximately double the number of samples above the optimal cut-off compared to the HRHPV+ group. The significance of this result is again difficult to quantify due to the relatively low sample sizes involved, but it may indicate that a panel of biomarkers may be necessary to help reduce false positives.
A number of studies have identified mir-143 as significantly underexpressed in cancer, including breast and cervical cancers 
. Whilst we also observed significant underexpression in cervical cancer cell lines with respect to normal (HRHPV-) samples, we did not find any statistically significant associations with respect to clinical samples from women across a wide spectrum of neoplastic disease. Previous studies investigating microRNA expression in cervical samples again show conflicting results. Lee et al
did not include either mir-21 or mir-143 in their initial screening experiments and as such did not identify these species as being involved in cancer development; this highlights a clear challenge in identifying potential biomarkers, in cases where leading candidates are ignored. However, Wang et al
suggested that reduction of mir-143 was required for tumor development whilst Lui et al
showed mixed results for mir-143, mainly due to the low level of expression detected in either normal or cancerous samples. Inconsistencies may be due to the relatively low level mir-143 expression in cervical tissues, which may require a more precise extraction method (e.g.
laser microdissection) for tumor extraction – none of the studies listed here used such methods.
Previous studies revealed that PDCD4, PTEN and TPM1 are all transcripts targeted by mir-21 in various tissues 
and that PTEN is progressively underexpressed from normal tissue through to cervical cancer 
. One recent study has also reported reduced PDCD4 expression in cervical cancer cell lines 
which is in agreement with our findings in the clinical samples examined in this study. In fact recent studies have demonstrated not only a direct relation between PDCD4 expression and tumor progression (Wei et al
, 2009 
) but also the key role of PDCD4 and mir-21 in apoptotic inflammatory response (Sheedy et al
, 2009 
). Moreover, recent cell model studies support suggest a role for loss of PDCD4 expression in increased sensitivity of cells to agents that cause DNA damage (Singh et al
, 2009 
), all of which would be expected to provide favourable conditions for tumor growth. Mice expressing high levels of PDCD4 are also more resistant to tumor growth in comparison to normal controls. We hypothesise that cPDCD4 expression (and, to a lesser extent, nPDCD4 expression) is increased in pre-cancerous specimens due to increased apoptosis, followed by a sharp decrease in expression for ICC specimens, where apoptotic mechanisms are subverted. Similar trends have been observed for other tumor suppressor genes in relation to oncogenesis and apoptosis 
. Interestingly, we did not identify any correlation between PTEN expression and histological diagnosis; loss of expression has been noted previously, and we have observed lower expression in a breast cancer study using the same methods (data not shown).
In this study we have found that the expression level of mir-21 is significantly correlated with histological diagnosis of ICC. A known target of mir-21, PDCD4 was also downregulated but was not significantly correlated with mir-21 expression. Expression of mir-143 was not correlated with histological diagnosis of clinical samples, however it was significantly down-regulated in cervical cancer cell lines. These findings are important in the quest to gauge the utility of MicroRNA biomarkers for diagnostic applications, a key component of which is to measure expression levels in precancerous tissues as well as normal and invasive cancers. Clearly, future work must focus on clinically relevant samples, in which predictive markers expressed in pre-cancerous tissue can be identified for further validation.