Proteomic patterns of cervical cancer cell lines, a network perspective
© Higareda-Almaraz et al; licensee BioMed Central Ltd. 2011
Received: 4 February 2011
Accepted: 22 June 2011
Published: 22 June 2011
Skip to main content
© Higareda-Almaraz et al; licensee BioMed Central Ltd. 2011
Received: 4 February 2011
Accepted: 22 June 2011
Published: 22 June 2011
Cervical cancer is a major mortality factor in the female population. This neoplastic is an excellent model for studying the mechanisms involved in cancer maintenance, because the Human Papilloma Virus (HPV) is the etiology factor in most cases. With the purpose of characterizing the effects of malignant transformation in cellular activity, proteomic studies constitute a reliable way to monitor the biological alterations induced by this disease. In this contextual scheme, a systemic description that enables the identification of the common events between cell lines of different origins, is required to distinguish the essence of carcinogenesis.
With this study, we sought to achieve a systemic perspective of the common proteomic profile of six cervical cancer cell lines, both positive and negative for HPV, and which differ from the profile corresponding to the non-tumourgenic cell line, HaCaT. Our objectives were to identify common cellular events participating in cancer maintenance, as well as the establishment of a pipeline to work with proteomic-derived results. We analyzed by means of 2D SDS-PAGE and MALDI-TOF mass spectrometry the protein extracts of six cervical cancer cell lines, from which we identified a consensus of 66 proteins. We call this group of proteins, the "central core of cervical cancer". Starting from this core set of proteins, we acquired a PPI network that pointed, through topological analysis, to some proteins that may well be playing a central role in the neoplastic process, such as 14-3-3ζ. In silico overrepresentation analysis of transcription factors pointed to the overexpression of c-Myc, Max and E2F1 as key transcription factors involved in orchestrating the neoplastic phenotype.
Our findings show that there is a "central core of cervical cancer" protein expression pattern, and suggest that 14-3-3ζ is key to determine if the cell proliferates or dies. In addition, our bioinformatics analysis suggests that the neoplastic phenotype is governed by a non-canonical regulatory pathway.
The definition of cancer has evolved according to the knowledge and perspective of the scientific context in which it is conceived. It has changed from a highly heterogeneous disease seen from a cell type and tissue of origin point of view, to the conception of cancer as an illness that involves the deregulation of various pathways that govern key, and somewhat common, cellular processes . Particularly, in 2000 Hanahan and Weinberg suggested that all cancer types represent a manifestation of six essential alterations in cell physiology that collectively coordinate the malignant phenotype: self-sufficiency in growth signals, insensitivity to growth inhibitors, evasion of programmed cell death, increase of the replicative potential, sustained angiogenesis and tissue invasion and metastasis . Furthermore, in a recent review published by the same two authors, they proposed two emerging hallmarks: reprogramming of energy metabolism and evading immune destruction; besides suggesting genomic instability and mutations, as well as tumor-promoting inflammation, as enabling characteristics .
Regardless of the latter, several cancer types have been more intensively studied due to their penetrance in the human population, such as prostate and breast cancer. However, we see in cervical cancer a unique opportunity to study the malignant transformation because of its common origin: 90.7% of the cases arise as a consequence of High-Risk Human Papilloma Virus (HR-HPV) infection, according to a study carried out in nine countries of diverse cervical cancer incidences . HR-HPVs encode the E6 and E7 oncoproteins, which interact with the well known and very common, tumor suppressor proteins p53 and pRB, respectively; among several other proteins that together impart a very strong oncogenic potential to the virus [5–7].
To date 120 types of human papillomaviruses have been identified, which can be subdivided into low-risk types, found mainly in genital warts, and high-risk types, which are frequently associated with invasive cervical cancer . Among the high-risk types HPV16 and HPV18 are the most prevalent, present in 54.6% and 11% of squamous cervical carcinomas, respectively . This is part of the reason why, cervical cancers derived from patients infected with those viral types, have been intensively studied, and one of the best studied human cell line, HeLa, is positive for HPV18 . Likewise, it is pertinent to mention that there are other cell lines which originated from HPV negative cervical cancers that have also been widely studied and therefore, enable the visualization of alterations in protein expression, common to many cervical cancer cell lines independently of their origin.
Cervical cancer is one of the most common types of cancer and a major mortality factor of women worldwide . Because most of the cases are a consequence of viral infections, cervical cancer is a disease that has been successfully addressed in developed countries thanks to preventive medicine . Nonetheless, when left unattended, a persistent viral infection combined with a strong and constitutive expression of viral oncoproteins E6 and E7 are highly inductive steps towards the malignant transformation of cervical epithelium . However, owing to a frequent and spontaneous elimination of viral sequences, not all the patients infected with HR-HPV develop cervical cancer . This indicates that most HPV infections are subclinical and only a small fraction of HR-HPV infections produce early epithelial lesions, and a more modest fraction of those lesions will develop into cancer. Consequently, even if infection by HR-HPV can be considered as the initial hit that gives rise to cervical cancer, the superseding steps that enable cancer development have not yet been described .
Currently, an accurate prediction of the evolution of the tumor is one of the biggest challenges for clinical oncology. Because of this, the conception of an integrative model that enables the prediction of future states of a system has become of vital importance. We consider that one of the best approaches currently available to accomplish this task are protein-protein interactions (PPI), because they encompass the scaffold of molecular pathways and cellular processes, besides being capable of revealing the dynamic and interactive function of human proteins . Furthermore, available databases of some organisms have promoted the construction of networks which become the starting point to explore and infer the fundamental principles by which the cell orchestrates its response to different kinds of perturbations.
In this work, we developed a pipeline for the functional analysis of differentially expressed proteins in cervical cancer compared with a non-cancerous control and obtained from 2D SDS-PAGE. We used PPI networks, pathways and GOs enrichment analysis, succeeded by an analysis of the promoter gene sequences to get a more systemic point of view of cervical cancer. We consider that the main advantage of following this approach is that it enabled to explore and locate important proteins for the biological phenomenon, which could have been overlooked as a consequence of the experimental methodology used, because some proteins are present in low concentrations or posses isoelectric points or molecular weights located outside the electrophoretic resolution range utilized in 2D SDS-PAGE.
Proteins identified as members of the "central core of cervical cancer".
78 kDa glucose-regulated protein
91 - 46
82 - 34
79 kDa glucose-regulated protein
64 - 32
Heat shock cognate 71 kDa protein
70 - 39
64 - 33
82 - 34
77 - 30
65 - 31
Tubulin beta chain
73 - 32
50 - 25
Elongation factor Tu, mitochondrial
44 - 22
42 - 17
62 - 23
Phosphoglycerate mutase 1
41 - 18
ATP synthase subunit beta, mitochondrial
84 - 44
Actin, cytoplasmic 2
50 - 20
Bifunctional purine biosynthesis protein PURH
49 - 22
Creatine kinase B-type
50 - 19
Elongation factor 2
67 - 26
Hypoxia up-regulated protein 1
38 - 19
Creatine kinase B-type
50 - 19
Peptidyl-prolyl cis-trans isomerase A
45 - 16
46 - 19
WD repeat-containing protein 1
46 - 16
60 - 26
Heat shock 70 kDa protein 1A/1B
42 - 21
Tubulin alpha-1C chain
45 - 13
53 - 18
Protein disulfide-isomerase A3
47 - 15
Stress-70 protein, mitochondrial
81 - 34
Succinate dehydrogenase flavoprotein subunit
39 - 17
40S ribosomal protein SA
37 - 11
Heat shock protein HSP 90-beta
44 - 19
25 - 11
52 - 19
L-lactate dehydrogenase B chain
43 - 20
Retinal dehydrogenase 1
58 - 19
22 - 9
ATP-dependent RNA helicase DDX3X
23 - 13
Glutamate dehydrogenase 1, mitochondrial
38 - 13
55 - 13
Aconitate hydratase, mitochondrial
54 - 23
Myosin regulatory light chain 12A
32 - 11
Ribonucleoside-diphosphate reductase large subunit
28 - 15
Elongation factor 2
Cytochrome b-c1 complex subunit 1, mitochondrial
45 - 15
51 - 13
Poly(rC)-binding protein 1
39 - 14
Mitochondrial inner membrane protein
34 - 13
Macrophage capping protein
41 - 11
Translationally-controlled tumor protein
19 - 8
Proteasome subunit beta type-4
20 - 7
Transitional endoplasmic reticulum ATPase
49 - 15
Keratin, type II cytoskeletal 1
46 - 13
41 - 8
14-3-3 protein zeta/delta
43 - 14
37 - 11
Heterogeneous nuclear ribonucleoprotein L
27 - 13
T-complex protein 1 subunit eta
36 - 15
Oxysterol-binding protein-related protein 8
56 - 13
T-complex protein 1 subunit epsilon
30 - 11
Hypoxia up-regulated protein 1
28 - 13
Proteasome subunit alpha type-5
17 - 6
Forkhead box protein P3
41 - 12
Since this study pointed to a "central core of cervical cancer", it was necessary to understand the nature of the proteins involved in the process. To this end, 66 proteins were classified via GO (Gene Ontology) using QuickGo of EMBL-EBI database . The identities of the proteins of the central core confirm that these are important in cancer because several studies have linked their aberrant expression to the neoplastic phenotype. With reference to their assigned functions, these proteins can be divided into at least three groups:
The first group includes proteins related to cell migration and metastasis, like anexin 2 that plays a crucial role in establishing metastasis of prostate cancer (PCa) by regulating the adhesion and migration of PCa to osteoblasts and endothelial cells ; protein disulfide-isomerase which is strongly expressed by invasive glioma cells, and its inhibition led to reduced glioma cell migration and invasion ; vimentin, an intermediate filament cytoskeletal protein and a marker of Epithelial Mesenchymal Transition (EMT), contributes to invasion and metastasis in prostate cancer ; ezrin is required for invasion and metastasis of mammary carcinoma ; and finally vinculin, which facilitates contractile force generation, enhancing cell invasion . This group of proteins could be a sign of invasion and metastasis in cervical cancer, similar to their role in other cancer types.
In a second group, we placed proteins related to evasion of apoptosis: GRP78, HSP71, HSP7C, HS90B and GRP75. These proteins are activated as part of the Unfolded Proteins Response (UPR), which has been largely related to survival, cell proliferation and angiogenesis . Likewise, it has been shown that overexpression of GRP78 is sufficient to confer apoptosis resistance in at least two cell types, regardless of its function within the UPR . We believe these proteins could be a link between different hallmarks of cancer, like apoptosis evasion, angiogenesis and proliferation.
In a third group, we find proteins involved in or associated with central metabolism like glyceraldehyde 3 phosphate dehydrogenase, phosphoglycerate mutase 1, enolase A, triosephosphate isomerase and L-lactate dehydrogenase B. We consider that the expression of this group of proteins could be due to the cancerous cells' need to divide their incoming nutrients between energy production and macromolecular biosynthesis to support cell growth and DNA replication. This increase in cell metabolism has been described in cancer cells, which exhibit a much higher intake of glucose than normal cells, as well as increased rates of glycolysis and lactate production even in the presence of oxygen [23–26].
Another protein detected by us as a member of the "central core of cervical cancer" was galectin-1. Remarkably, very recently it was reported that cells can stimulate tumor angiogenesis by secretion of this protein, and it was previously found in the vasculature of many human tumors, including colon, head and neck, lung, prostate, and oral cancers [27, 28]. All of the proteins identified as members of the "central core of cervical cancer" are summarized in Table 1.
Afterwards, we expanded the PPI network as a way to integrate all the proteins of the central core, in order to obtain a coherent output which could be related in its entirety to the neoplastic phenotype. It also served the purpose of broadening our perspective and strengthening the statistical significance of our results.
We considered that the network expansion would allow us to place the entities belonging to our central core in a broader context, specially because the amount of proteins that were identified represent a small fraction of all the entities responsible for the phenotype. We also expected to observe the same topological phenomena in the expanded network as we did in the network composed solely of cervical cancer-specific proteins.
A connectivity distribution analysis allowed us to identify those proteins that highly interact with others, defined as the hubs of the expanded network. This pointed once again to the multifunctional regulator of cell signal transduction and adaptor protein, 14-3-3ζ, followed by 14-3-3γ with less than half the number of interactions. Also, we calculated the betweenness of all the proteins in the network, because it is a feature which allows us to explore the connection between local network structure and global network topology, besides representing a measure of a node's influence over information transfer [35, 36]. The proteins with the highest betweenness were 14-3-3ζ, vimentin and vinculin, part of our central core of cervical cancer proteins, which is consistent with their role as bait proteins. However, there were also non-bait nodes that featured among the proteins with highest betweenness, such as ubiquitin and 14-3-3γ; thus, supporting the notion that the betweenness metric, though biased in this case, is still a valid metric of essentiality and robustness.
Enriched gene ontology level 3 categories of biological processes.
gene ontology term
response to unfolded protein
inclusion body assembly
generation of precursor metabolites and energy
maintenance of protein location in cell
carbohydrate catabolic process
maintenance of location in cell
programmed cell death
regulation of cell death
histamine production involved in inflammatory response
regulation of secondary metabolic process
leukocyte mediated immunity
cellular catabolic process
negative regulation of immune response
cellular macromolecule localization
maintenance of location
Enriched KEGG pathway-based sets.
Small cell lung cancer
Protein processing in endoplasmic reticulur
Pathways in cancer
Neurotrophin signaling pathway
T cell receptor signaling pathway
Pathogenic Escherichia coli infection
Non-small cell lung cancer
Bacterial invasion of epithelial cells
B cell receptor signaling pathway
Leukocyte transendothelial migration
Using the information generated by the ENCODE project, we searched among the transcription factors (TF) that had chromatin immunoprecipitation-sequencing (ChIP-Seq) data available, for overrepresented binding sites of a given TF in the genes of our expanded network (Figure 1). We evaluated for enrichment by considering binding sites that were located up to 700 bp upstream or 300 bp downstream of all the transcription start sites reported in the GENCODE database and that had a score larger than 750. We considered these 1,000 bp as the promoter sequence, because it has been reported to be the region of highest ChIP-Seq peak concentration in previous studies .
Significantly overrepresented transcription factors identified by ENCODE ChIP-Seq peaks based on the GENCODE Database
tissue of origin
peaks near TSS of the gene of the extended network
peaks near TSS of all the
Significantly overrepresented transcription factors identified by ENCODE ChIP-Seq peaks based on the TRED Database
tissue of origin
peaks near TSS of the genes of the extended network
peaks near TSS reported on SwitchGear
Among the overrepresented TFs, we consider that the presence of c-Myc is remarkable because it is in agreement to what is currently known about cancer cells in general. Also, c-Myc's primary binding partner for gene activation is Max, so that the presence of this protein is complementary to c-Myc's activity. The mechanisms that make of c-Myc a very powerful oncogene are not very clear; however, there is evidence that indicates that the activation of c-Myc correlates with approximately 70% of human cancers .
The first possible explanation for c-Myc's prevalence in cancer is that the genes it induces represent the primary response of nearly all signal transduction pathways known to be involved in cancer . Likewise, Nilsson et al,  have suggested that a loss in c-Myc's ability to induce apoptosis can convert c-Myc into a pure promoter of cell growth and transformation. Concordantly, the abrogation of p53 function, which is the protein responsible for executing c-Myc's apoptotic response, has been substantiated in all HPV positive cell lines, as well as the HPV negative cell line, C33-A .
The last factor, E2F1, controls cell-cycle progression and DNA replication and it has been shown to be induced by c-Myc's activity . E2F1 has also been implicated in c-Myc mediated down regulation of p27KIP1, which is a cyclin-dependent kinase inhibitor . Finally, polymerase 2 featured as the DNA-binding protein with the most significant p-value, thereby providing us with a notion of the genes of the network that are being transcribed (Tables 4 and 5).
We performed a comprehensive analysis that stems from data obtained by proteomic techniques, using perspectives of biological networks and the annotated data of the human genome and large-scale sequencing techniques available to the public as a tool for building models to infer biological behaviors (Figure 1).
As defined in this study, the "central core of cervical cancer" does not describe the immortalization process directly; rather, it represents processes involved in maintaining a viable and proliferating cell. Table 2 presents the biological processes that are enriched in our core set of proteins and as can be seen, these are suggestive of a malignant phenotype, where the main GOs are those related to UPR, metabolism, catabolic processes, maintenance of cell-location and cell death. These observations are consistent with our initial conjectures, since we considered that using the HaCaT cell line as a control would allow us to distinguish between immortalization and transformation, because HaCaT is an immortalized cell line with a stable kariotype and phenotype, which is non-tumorigenic.
One of the studies that uses the HaCaT cell line as a control was performed by Choi, et al.. The main difference between the latter study and our work, is that they used cervix biopsies; this implies that there are important differences in the intracellular and extracellular contexts of the biological model. Likewise, even though the methodology they utilized only considers samples with the same pathologic diagnosis, it is likely that there is still an important amount of heterogeneity in their samples due to inter-individual differences. As expected, we only found one common protein among the results of both studies: Vimentin. However, they were able to identify a number of proteins that strengthen our findings such as Amy-1, which stimulates c-Myc's E-box dependent transactivation activity ; and Miz-1, which must be inhibited in order for c-Myc to mediate apoptosis . Another result that supports our findings is that they observed a downregulation of 14-3-3σ, which plays an opposite role in cell growth, compared to 14-3-3ζ . Finally, they also found a clear overexpression of chaperone proteins, different from the ones in our central core.
The acquisition of the reported interactions between the identified proteins provided a first insight of the key elements in the neoplastic dynamics, and became the meter that allowed us to judge the validity of later analysis, by seeking an overall agreement between the original and the extended network. It is noteworthy that for the small network the highest interconnected node was 14-3-3ζ, and this was not lost when we expanded the network. Also, both networks are enriched in similar processes.
When analyzing the extended network it is apparent that 6 of the members of the 14-3-3 protein family (ζ, γ, β, τ, σ and ε) feature among the highest interconnected nodes. This is consistent with the biological role of the members of this protein family as signal transduction molecules . The identification of 14-3-3ζ and its role as a hub and as a bottleneck in the PPI network is particularly relevant because it is known to be involved in three important cellular processes: cell cycle regulation, signal transduction and regulation of apoptosis [47–50].
It was previously reported by Jonsson and Bates  that cancer proteins exhibit a significantly different network topology when compared to proteins unrelated to cancer; as well as a higher ratio of promiscuous structural domains. This is in agreement with what is known about the members of the 14-3-3 protein family, which have been described as signal integrators, amplifying strong signals and filtering out weaker conflicting ones to achieve a meaningful, coordinate biological output, such as cell death or survival .
Also, there is strong evidence that the overexpression of 14-3-3ζ promotes p53 degradation by the proteosomal route , which is increased by the action of E6 in HPV-infected cells . An enhanced turn-over rate of the p53 protein and the pleiotropic effects consequence of the increased number of 14-3-3ζ proteins, which are involved in a large number of cellular processes, could be the reason why not all infections develop into cancer and most of them are subclinical: c-Myc could be activating 14-3-3ζ in a differential manner. Therefore, we consider that we posses sufficient evidence to suggest the role of 14-3-3ζ in cervical cancer cell lines as the determining factor that can ultimately dictate the fate of a cell, regardless of it being infected with HPV; and cause it to undergo cell cycle deregulation onto malignant transformation (Figure 5).
Likewise, when we performed a biological pathway enrichment, which we consider the best way to perform a functional analysis, we found consistent results. To do this, we used the ConsensusPathDB site to find overrepresented KEGG pathways in the proteins of our extended network. The enrichment analysis showed an overrepresentation of survival factors of apoptosis pathways and focal adhesion that is related to invasion and metastasis, two of the most important and well-documented hallmarks of cancer; as well as various established cancer pathways like small cell lung cancer, prostate cancer, endometrial cancer and pathways in cancer (Table 3). This was obviously expected and helps to strengthen our analysis, indicating that the central core of cervical cancer, and its subsequent expansion proposed in the PPI interaction network, still retain those proteins that are closely related to the neoplastic process.
However, there are other pathways that do not present such a straightforward connection to cancer, such as the neurotrophin and T-cell receptor signaling pathways, bacterial infections and proteasome pathways. We can only speculate about their meaning, as the cellular stress, chronic inflammation, turnover of proteins, cell survival and evasion of the immune system are other features related to the progression and maintenance of tumorigenesis.
As a way to understand the underlying regulatory dynamics of the system, we also performed an enrichment analysis of transcription factor binding sites among the promoters of the genes of our extended network. We focused on the three transcription factors that complied with both of the following: those which were significantly overrepresented among the promoter regions of the genes of the proteins of our network and those in which the corresponding ChIP-Seq assays were carried out on the HeLa-S3 cell line http://genome.ucsc.edu/. The resulting transcription factors are: E2F1, c-Myc and Max. We believe that together, these transcription factors bring an overall feeling of coherence to the network because the overexpression of c-Myc is widespread in human cancers, and specifically in cervical cancer . Moreover, there is evidence that E2F1 is activated by the E7 protein of HPV-16 in a pRB-independent manner , apart from its pRB-dependent activation, which in the case of HPV infected cells is also promoted by pRB inactivation by viral oncoprotein E7. In the context of cervical cancer, it has been reported that an overexpression of E2F1 can drive quiescent cells through G1 into S-phase of the cell cycle, ultimately leading to apoptosis or neoplastic transformation.
E2F1 has been shown to be an inhibitor of c-Myc's activator, β-catenin, by means of the Wnt pathway [55, 56]. We also found a clear overrepresentation of the TCF4 transcription factor in the genes of our network. This contradicts our previous finding because TCF4 functions in the activation of the Wnt pathway, and therefore, of c-Myc. However, we had to consider that all the published works that point to E2F1 as an inhibitor of c-Myc have been performed in colorectal cancer. Likewise, the evidence of enrichment of TCF4 binding sites was performed in the HCT-116 cell line and so we do not posses information about the function of this transcription factor in cervical cancer.
It has been observed in human foreskin keratinocyte and fibroblast (HFK and HFF) cells that the genetic background of the cell type can be determinant for the outcome of altered gene expression. This is the case of HFF which cannot be immortalized by the mere addition of oncoproteins E6 and E7, unlike the HFK cells, due to a differential turnover of c-Myc . Moreover, as pointed by Bernards  in the context of colorectal cancer pRB is more likely to be acting as an oncoprotein than as a tumor suppressor, which is clearly not the case in cervical cancer. Considering all the previous evidence, we can only infer that the proteins pRB and E2F1 are acting with opposing roles in cervical cancer, as compared to colorectal cancer.
On the other hand, there are reports of E2F1 and c-Myc overexpression in cervical cancer, which also correlate their expression with advanced states of the disease . Also, genetic studies of the E2F promoters have shown that these genes are induced by c-Myc, dependent on the E box sites. Coordinately, all data point to a model where c-Myc is activated by means different from the canonical pathways and is working with E2F1 to promote the neoplastic phenotype.
The possible scenario of molecular events in cervical cancer suggested by our study portrays the evasion of apoptosis mediated in two ways, first, the sequestration of the pro-apoptotic protein BAD by 14-3-3ζ  and the overexpression of the anti-apoptotic protein Bcl-2 managed by the interaction of c-Myc and E2F1 . Evasion of apoptosis by these routes can be supported by the interaction of viral oncoproteins E6 and E7 with tumor suppressor proteins p53 and pRb, when there is persistent infection with high-risk HPV. Simultaneously, these oncoproteins are able to promote significant increase in proliferation, and immortalize cells through the activation of hTERT .
The expression of E2F factors and cyclins facilitated by c-Myc, together E2F1 leads to a quick transition from the G1 to the S phase of the cell cycle, and the consequent boost in growth and cell proliferation .
Finally, c-Myc overexpression allows a significant increase in cell proliferation by key roads, such as protein biosynthesis, central metabolism, the expression of transcription and cell cycle factors  and expression of 14-3-3ζ. All of these contribute to facilitate the transition from G1 to S phase . c-Myc's mediated expression of Ezrin  also promotes epithelial mesenchyma transition, facilitated by the overexpression of vimentin  (Figure 5).
Our results suggest a phenotype shared by the six cervical cancer cell lines as a result of the overexpression of c-Myc, helped by E2F1, which in turn allows the overexpression of 14-3-3ζ and other proteins of the "central core of cervical cancer". This signal transduction protein has been reported in other models of cancer as being responsible for malignant transformation and the decision between life and death of cells (Figure 5).
The CaSki, HeLa, SiHa, C-33A, ViBo and CaLo cell lines were provided by the oncology laboratory of the Centro Medico Siglo XXI which belongs to the Instituto Mexicano del Seguro Social. The HaCaT cell line was donated by the Centro de Investigación Sobre Enfermedades Infecciosas, which belongs to the Instituto Nacional de Salud Pública. All cell lines were cultured in RPMI-advanced 1640 serum-free media (Gibco BRL, USA) with red phenol and antibiotic-antimycotic solution (10,000 units penicillin, 10 mg streptomycin, and 25 μg amphotericin B per mL), supplemented with 1% fetal bovine serum (Invitrogen, Carlsbad, CA) and 200 mM of GlutaMAX (Invitrogen). The cells were incubated in 5% of CO2 and humidity saturation at 37°C in culture flasks of 75 cm2 (Nalge Nunc International, Rochester, NY). Cells were harvested at 70% confluence with Verseno solution (Tris base 25 mM, NaCL 136.8 mM, KCl 5.36 mM, EDTA 1 mM pH7.7) and washed 3 times in phosphate buffer saline (0.1 M sodium phosphate and 0.15 M NaCl in one liter, pH 7.2).
Protein extraction and two dimensional gel electrophoresis were done as previously described . Gels were dyed in colloidal coomasie  and scanned in a GS-800 densitometer (Bio-Rad, Hercules, CA). Digital images were analyzed and compared using the PDQuest 8.0.1 software (Bio-Rad). Each experiment was done in triplicate. Once the digital image of each gel was compared against the rest, the electrophoretic entities of interest were cut, alkylated, reduced, digested and automatically transferred to a MALDI analysis target by a Proteineer SP II and SP robot using the SPcontrol 220.127.116.11 v software (Bruker Daltonics, Bremen, Germany), with the aid of a DP Chemicals 96 gel digestion kit (Bruker Daltonics) and processed in a MALDI-TOF Autoflex (Bruker Daltonics) to obtain a mass fingerprint. We performed 100 satisfactory shots in 20 shotsteps, the peak resolution threshold was set at 1,500, the signal/noise ratio of tolerance was 6, and contaminants were not excluded. The spectrum was annotated by the flexAnalysis 1.2 v SD1 Patch 2 (Bruker Daltonics). The search engine MASCOT  was used to compare the fingerprints against the UNIPROT  release 2010_09 database with the following parameters: Taxon-Human, mass tolerance of up to 500 ppm, one miss-cleavage allowed, and as the fixed modification Carbamidomethyl and oxidation of methionine as the variable modification.
Antibodies for immunoblotting were as follows: anti-14-3-3ζ human polyclonal (Imgenex, San Diego, CA) and anti-α-tubulin mouse monoclonal (Zymed Laboratories, South San Francisco. CA). Equal amounts of protein samples were subjected to SDS-PAGE, and transferred to a polyvinylidene difluoride membrane (Immobilon-P, Millipore Corp., Billerica, MA, USA). After blocking with 5% skimmed milk, the membrane was washed in TBS-tween 20%, and incubated with a primary antibody, followed by a AP-conjugated second antibody anti-rabitt or anti-mouse (Zymed). Blots were detected by chromogenic substrate BCIP/NBT (Zymed).
The network reconstruction was performed with the aid of the Cytoscape  Plugin, BisoGenet , using the identified proteins as bait nodes and adding edges with the following parameters: Organism> Homo sapiens, protein identifiers only; Data Settings>protein-protein interactions; all data sources and all experimental methods; method> By adding edges connecting input nodes and as Output> Proteins.
The primary network was extended by performing the bioinformatic bait technique, using the Cytoscape Plugin Bisogenet with the expand network option and the previous parameters, except that the method used was By adding neighbors of input nodes to a distance of one. The resulting network was subjected to a centrality measure analysis with the Plugin CentiScaPe1.1 .
We performed an enrichment analysis of pathway-based sets of proteins considering all the nodes of our extended network. Enrichment was done employing ConsensusPathDB , of the Max Planck Institute for Molecular Genetics, by using the overrepresentation analysis online tool. As input, we uploaded the UNIPROT protein identifiers of all the elements of the extended network. We searched against pathways as defined by KEGG , with a minimal overlap with the input list of 5 and a p-value cutoff of 0.0001.
Also, employing the same website and the same analysis tool, we performed an enrichment analysis based on Gene Ontology  level 3 category of biological processes. For this analysis, we considered only the identified core proteins and set the p-value cutoff on 0.00001.
To help us uncover some of the regulatory dynamics underlying cervical cancer we downloaded the Gencode Genes-ENCODE Gencode Gene Annotations table (wgEncodeGencodeManual V3) [78, 79] from the University of California Santa Cruz Genome Browser website . From this table we retrieved the transcription start site of all GENCODE genes with a score greater than 750, as well as its associated ENSEMBL Transcript ID and gene name. Because some transcripts present several nearby start sites, we also set a filter that only considers start sites located at a distance greater than 500 bp, this reduced the amount of entries considered to 18,534. Promoter sequences were defined as 700 bp upstream and 300 bp downstream of the transcription start site. The promoter locations corresponding to the genes of the proteins of our network were identified based on the ENSEMBL transcript ID as well as the gene names and synonyms stated in the UNIPROT data of those proteins.
Peak information was downloaded from the Yale transcription factor binding site (TFBS) track of the ENCODE Project  with a preference for the assays from the HeLa-S3 cell line. A peak was considered to be inside the promoter if the mid-point of the peak was inside the 1,000 bp reported as the promoter sequence. Hypergeometric tests were performed in order to assess the significance of our findings. Sample size was considered as the number of promoters that contain TFBS. All methodology is summarized in Figure 1.
The data associated with this manuscript may be downloaded from ProteomeCommons.org Tranche using the following hash:
High Risk Human Papillomavirus
Human foreskin fibroblast
Human foreskin keratinocyte
Transcription factor binding site
Unfolded protein response
Epithelial Mesenchymal Transition
Kyoto Encyclopedia of Genes and Genomes
Universal Protein Resource
Encyclopedia Of DNA Elements
European Bioinformatics Institute
European Molecular Biology Laboratory.
We would like to thank Mauricio Salcedo-Vargas to provide the cervical cancer cell lines, Gabriel Martínez-Batallar for his advice in PDQuest handling, Alberto Checa-Rojas for his help with cell culture, and Alberto Ramírez-Torres during the elaboration of 2D SDS-PAGE.
Part of this work was supported by CONACyT grant 152167 and DGAPA-PAPIIT grant IN-216260. Juan Carlos Higareda-Almaraz is a recipient of a PhD Studentship from the CONACyT, CVU 176426.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.