It is well described that persistent infection with HR-HPV is the causal factor in the development of CC [4]. Several reports have described that HPV16 is more frequent in SCCs and HPV18 in ADCs [5], [6], [7]. Hence, it is to be expected that there are molecular and clinical differences in both histologic subtypes. In this regard, several reports have described controversial data about the prognosis of CC patients with ADC and SCC. For instance, some of them have reported that ADC is associated with a poorer prognosis than SCC [9], [10]. However, other studies described that ADC and SCC displayed equivalent survival outcomes [11], [12]. In this respect, researchers have focused on identifying molecular differences unique to one or another histological subtype so that they can be used as prognostic biomarkers for these patients. Nonetheless, this field needs to be investigated further. In this work, through the collection of the transcriptome data of all CC patients from the TCGA, we identified a molecular profile of DEGs between ADC and SCC, which was validated in the Mexican-Mestizo independent cohort. Despite the fact that the TCGA database includes patients from different ethnic groups and all clinical stages, we detected that of the DEGs, 70 overlapped with genes from the Mexican-Mestizo database and even displayed the same log2FoldChange values, showing a high concordance in the molecular differences between the two histological types in the datasets. Likewise, Lin W.E. and collaborators analyzed TCGA data and reported 1733 DEGs between ADC and SCC from different tumors, such as those of the esophagus, lung, and cervix [13]. Chao A. and collaborators, through microarray assays, revealed a profile of DEGs between 9 ADC and 16 SCC samples from CC patients; TSPAN-3, CEACAM5, TACSTD1, MSLN and S100P were of particular interest, as they were overexpressed in ADC [15]. These findings suggested that there are molecular differences between ADC and SCC.
The molecular differences between ADC and SCC also have an impact on signaling pathways that promote disease. In this work, the KEGG, WikiPathways and Reactome databases were employed to elucidate these differences. We found that SCC tumors have a higher level of activation of critical cancer pathways, such as IL-17, JAK/STAT and Ras signaling, compared to ADC tumors. In this respect, it is well known that IL-17 is the central cytokine of the Th17 response, during persistent infection of the with HR-HPV, this response triggers chronic inflammation of long duration and production of IL-17 among other pro-inflammatory cytokines, hence creating a favorable environment for tumor development associated with associated with poor CC prognosis. Interestingly, higher IL-17 expression has been observed in patients infected with HPV 16 and 18 in relation to other genotypes of the virus [24], [25], [26]. Additionally, Punt S. and collaborators, reported that the predominant cell type expressing IL-17 in SCC CC is the neutrophilic granulocyte and directly contribute to tumorigenesis [27]. Moreover, it has been shown that the expression of Th17 cells is related with stage, tumor size, lymph node metastases, and vasoinvasion [28]. Although there are no studies that evaluate the correlation between IL-17 and histological subtypes (ADC and SCC), we observed that SCC has a higher level of IL-17 pathway activation. It is recognized that cytokines can regulate the local immune state of the cervix. For instance, modifications in some cytokines, such as IL-6 and IL-2, and even IL-17 produced by persistent HR-HPV infection can induce hyperplasia of the local vascular epithelium and activation of the JAK/STAT pathway, promoting local epithelial cell proliferation and disease progression [29]. Thus, HR-HPV may also have an impact on different components of the JAK/STAT pathway and its inhibitors [30]. Strikingly, the JAK/STAT signaling, which is involve in proliferation, invasion, survival, inflammation, and immunity and has been reported to be altered in CC patients [31], was another pathway enriched in our analysis. There is a connection between JAK/STAT pathway and TH17 cells, since JAK/STAT pathway is necessary for the differentiation of Th cells [32]. The aberrant signaling of JAK/STAT pathway has not been associated with any histological subtype. Given that ADC and SCC have frequently HPV 18 and 16, respectively, it is expected that this pathway is altered in both subtypes, nonetheless we were able to distinguish that it is enriched in the SCC subtype, which is an interesting fact that has yet to be validated. However, Wu S. and collaborators, reported that STAT1 expression, a member of this pathway, was positively correlated with HPV 16 viral load [33]. Therefore, if the SCC subtype has a higher HPV16 load, an increase activation of this pathway is expected. Interestingly, JAK/STAT pathway has become an attractive therapeutic target in modern CC treatment [34]. Nevertheless, clinical trials are needed to determine whether targeted therapy can benefit CC patients. Another pathway enriched in our study was Ras signaling. Aberrant activation of this pathway is common in several cancers, including CC, which often results from the presence of mutations and amplifications of KRAS. For instance, Zou Y. and collaborators, reported that the frequency of the KRAS mutation ranged from 8.0–17.5% in cervical ADC [35], to absent or rare in SCC, which suggested that KRAS mutations are frequent and might be a driving factor for the development of cervical ADC but not SCC [36], [37]. Although, in our study we observed that Ras pathway has a higher level of activation in SCC than ADC, which may be due to mutations or alterations in other elements of the pathway [35]. Moreover, Both HPV E6 and E7 oncoproteins can alter cellular RAS signaling pathways to induce cancer [38]. Thus, our findings indicate that SCC is more aggressive than ADC; however, the candidate pathways and genes for distinguishing the subtypes need to be further verified.
A main focus of this work was to identify potential genes associated with the prognosis of ADC or SCC, and we generated Kaplan-Meier survival curves with the Mexican-Mestizo and TCGA clinical data. We noticed that GABRB2 and TSPAN8 overexpression in ADC was associated with favorable OS, while TMEM40 overexpression in ADC was associated with poor OS in the Mexican-Mestizo CC cohort. There is only one report associating the GABRB2 protein with cancer; however, GABRB2 overexpression is correlated with lymph node metastasis in thyroid cancer [39]. TSPAN8 has been studied more than GABRB2 in cancer, although it has an oncogenic role because its expression is correlated with a poor prognosis in breast cancer [40], renal cell carcinoma [41] and pancreatic cancer [42]. However, its role in CC has not yet been studied. On the other hand, TMEM40 has been poorly studied in cancer, but a study proposed that this protein plays a crucial role in proliferation and apoptosis via the p53 signaling pathway [43] and may be a potential diagnostic biomarker for bladder cancer [44]. The expression levels of GABRB2, TSPAN8 and TMEM40 were not associated with the OS of TCGA CC patients (Supplementary Fig. 1). In this respect, it is well known that the TCGA database contains incomplete or no information regarding the survival of many patients. This often occurs because almost all of the patients are untreated and thus have no response data and short follow-up periods [16]. It is possible that for this reason, we did not find any association between the expression of these genes and OS. Lin W.E. noted that the TCGA database is interesting because it includes patients of different ethnicities; however, although TCGA datasets are generally large, they may not be representative of the general population [13]. Despite this, they reported that in TCGA ADC samples, the increased expression of the TPX2, KIF4A, IGF2BP1, and HSPA6 genes was correlated with poor survival; in contrast, the expression of MS4A1 (CD20), SUSD2, and CX3CL1 was related to a favorable prognosis [13]. On the other hand, the study by Chao A. and collaborators reported that positive expression of CEACAM5 and TACSTD1 was associated with poor OS in ADC CC patients, and they suggested both genes as independent prognostic factors [15]. Overall, it is obvious that when searching for a prognostic biomarker for ADC or SCC, there were inconsistencies in the studies performed by us, Lin W. E. [13], and Chao W. [15]. The candidate genes differed in each study. This may be due to the lack or incompleteness of survival data, as previously mentioned, which can consequently cause variations in the results regarding the associations of genes with OS. However, the MS4A1, SUSD2, CX3CL1, CEACAM5, and TACSTD genes, which were associated with OS in the studies by Lin W.E. [13] and Chao W. [15], were also identified as DEGs in our study (Supplementary table 3). Thus, it is reasonable to assume that there is a consistent molecular profile for each histological subtype. However, the correlation of these genes with the survival or prognosis of patients’ needs to be studied meticulously.