This study shows that CIN biopsies shed a complex mixture of proteins into a cell culture medium for 24 hours at 4°C. Supernatants from 20 patient samples were analyzed using a bottom-up shotgun proteomics approach  in which the proteins were digested into smaller peptides using trypsin. The peptide mixture was then analyzed using uni-dimensional LC-MS/MS. Despite the depletion of seven high abundance proteins including immunoglobulins and albumin, peptides from these proteins were detected, while transferrin was not found at all after depletion. In addition, not unexpectedly haemoglobins constitute a relatively large part of the identifications (cervical tissues with CIN2-3 are usually richly vascularised) and should be included in future depletion work. The protein mixtures still are of such complexity after depletion that it would be advantageous to do further fractionations by for example using two-dimensional separations like 2D gels or MudPit . Selective enrichments of for example phosphorylated kinases that are important components in regulation of the cell cycle  would reduce the complexity of the samples while at the same time enrich interesting proteins. The glycosylation pattern of proteins is another interesting topic [24, 25] that could be further elucidated using this sample set.
The gene ontology bar diagram in figure 2 shows that 33% of the proteins identified are annotated to metabolic processes, 35% to signal transduction in CIN2 and 27% in CIN3, 9% annotated to cell cycle processes in CIN2 and 15% in CIN3, and annotated to trafficking/transport. In agreement with our results, Panicker et al  reported that 33% of the identified proteins in cervical mucus samples were related to metabolism, while Dasari et al  reported 32% after analysis of cervical-vaginal fluid. The fraction of the proteins involved in signal transduction processes (35 and 27%) in this study is much higher than the results from the studies by Panicker et al  and Dasari et al , who reported 1% and 3%, respectively. This may be due to the higher efficiency of the exudation process of a biopsy (which in principle has a much larger direct stroma contact surface with the RPMI than intact cervical epithelium has with cervical mucus or vaginal fluid) and the long incubation time we have used (24 hours).
Several of the proteins found in this study have been reported by others in cervical mucus samples , and also in plasma samples from patients with CIN  or in cervical tissue samples [29–32]. Vimentin was found down-regulated in vaginal and cervical carcinoma compared to normal tissue [31, 32]. Actin, transthyretin, lamin a/c, fibrinogen and apolipoprotein A-I are all proteins identified in one or more of the mentioned studies, and several of these proteins have been connected to cancer or used as cancer markers [33–35]. Transthyretin has been used as a biomarker for nutritional status and inflammation, but post-translational modified forms have also been reported as part of a biomarker panel for early detection of ovarian cancer [36, 37]. Transthyretin was found in 6 of the 10 CIN3 samples and none of the CIN2 samples. Alpha-1-acid glycoprotein (AGP) is known to increase during acute-phase response, , and has also been identified in other studies [26, 27] using cervical vaginal fluid or mucus. The heat shock proteins are involved in a range of cell processes. They are induced under stress conditions and known to be over-expressed in human cancers. Some of them are used as biomarkers for carcinogenesis and some as signals for aggressiveness of some cancers . Two studies [30, 31] found the level of heat shock protein 1 (Hsp27) to be up-regulated in carcinoma samples. However, another study discovered a decline  and related this to the presence of HPV oncoproteins with a negative effect on the ability of lesions to undergo terminal differentiation. We found no significant difference between CIN2 and CIN3 samples with regards to Hsp27 based on spectral count comparison.
A conserved 7 amino acid sequence found in lipocalins, including the lipocalin-type prostaglandin D synthase (L-PGDS) was identified in 17 of the 20 samples in this study and was recently demonstrated to modulate cell survival . Lipocalins are normally present at low concentrations, but the expression can increase due to physiological conditions . The tumor suppressor p53 has been found to suppress the expression of L-PGDS . A decrease in the p53 level, as found in most CIN2-3 lesions , combined with a local inflammation, might explain the frequent occurrence of L-PGDS in the CIN samples.
Proteins from the intermediate filament (IF) protein family is highly represented among the proteins identified in both the CIN2 and CIN3 group. The cytokeratins belong to the type I and II of the IF's, desmin and vimentin belong to the type III and lamin to type V. This group of proteins have been widely used as markers of different cancers .
Special attention should be paid to the late epithelial cell differentiation marker Cytokeratin 2 (CK-2), which was found to have highest discriminatory power between CIN2 and CIN3, with 90% correct classification. The protein was identified using 5 different peptides (2 unique), and only in samples from the CIN2 group (in 8 of the 10 samples).
This protein is expressed late in the differentiation process in the uppermost epidermal layers of the normal skin , and has normally been associated with the skin disorder ichthyosis bullosa of Siemens . The protein was characterized by Collin et al [45, 46], and was found overexpressed in patients with head and neck squamous cell cancer . The fact that CK-2 is identified in CIN-2 lesions only indicates that the epithelial cells have a greater tendency to high-end differentiation than CIN3 lesions, which is biologically well understandable. To our knowledge, this is the first report of C ytokeratin 2 associated with differences between CIN2 and CIN3, or with other neoplasia .
Although the number of samples is small, the proteomics results have been confirmed by immunohistochemical evaluation, when performed, and further validation is in progress.
A differentiation between a CIN2 and CIN3 diagnosis has at the moment no consequence for patients with regards to follow-up or treatment as all these patients in principle will undergo surgical cone excision. However, the fact that there are differences at the protein level amongst high grade CIN lesions with microscopically different epithelial appearances, as detectable by expert pathologists supported by p16 and MIB-1 immunohistochemistry, make clear that these differences are real and may have biological impact. These additional novel markers could help pathologists in differentiating CIN2 and CIN3. This may be especially of interest as high grade CIN 3 lesions have a lower likelihood to regress spontaneously and a higher probability to progress to invasive cancer, than CIN 2 lesions . The fact that this well-known knowledge currently is not used in therapeutic decision making, is due to the lack of reproducibility amongst pathologists when classifying CIN 2 and CIN3 lesion with conventional Hematoxyllin and Eosin stained sections alone. However, with the advent of new molecular immunohistochemical biomarkers, and possibly also proteins and peptides isolated with the water soluble strategy as described in the current study, the distinction could become of clinical interest.