Competing Endogenous RNAs in Cervical Carcinogenesis: A New Layer of Complexity

MicroRNAs (miRNAs) regulate gene expression by binding to complementary sequences within target mRNAs. Apart from working ‘solo’, miRNAs may interact in important molecular networks such as competing endogenous RNA (ceRNA) axes. By competing for a limited pool of miRNAs, transcripts such as long noncoding RNAs (lncRNAs) and mRNAs can regulate each other, fine-tuning gene expression. Several ceRNA networks led by different lncRNAs—described here as lncRNA-mediated ceRNAs—seem to play essential roles in cervical cancer (CC). By conducting an extensive search, we summarized networks involved in CC, highlighting the major impacts of such dynamic molecular changes over multiple cellular processes. Through the sponging of distinct miRNAs, some lncRNAs as HOTAIR, MALAT1, NEAT1, OIP5-AS1, and XIST trigger crucial molecular changes, ultimately increasing cell proliferation, migration, invasion, and inhibiting apoptosis. Likewise, several lncRNAs seem to be a sponge for important tumor-suppressive miRNAs (as miR-140-5p, miR-143-3p, miR-148a-3p, and miR-206), impairing such molecules from exerting a negative post-transcriptional regulation over target mRNAs. Curiously, some of the involved mRNAs code for important proteins such as PTEN, ROCK1, and MAPK1, known to modulate cell growth, proliferation, apoptosis, and adhesion in CC. Overall, we highlight important lncRNAmediated functional interactions occurring in cervical cells and their closely related impact on cervical carcinogenesis.


Introduction
Cervical cancer (CC) is the fourth leading cause of death by cancer among women, with over 311,000 deaths in 2018 worldwide [1], even though it can be prevented by vaccination programs [2]. Generally, CC develops slowly, and there is a great opportunity to detect initial cervical alteration through cytology screening programs and HPV testing, which were showed to impact CC mortality [3]. Indeed, higher CC mortality rates are observed in poor countries [4] where screening programs are less effective [5]. Therefore, understanding the pathogenesis and exact molecular mechanisms underlying the emergence and progression of CC is crucial for defining promising biomarkers and an opportunity to develop better diagnostic and therapeutic interventions for CC.
Recently, emerging evidence has suggested promising applications of noncoding RNAs (ncRNAs)-including microRNAs (miRNAs) and long noncoding RNAs (lncRNAs)as novel candidate biomarkers and/or potential targets of treatment in multiple cancers. Different networks engaged by different ncRNAs influence numerous molecular targets, acting as key regulators of several biological programs in developmental and disease contexts, being particularly relevant in cancer [6].
For their important roles in cancer, both miRNAs and lncRNAs have been extensively described in the past few years as master regulators of cancer themselves. Most recently, considerable evidence has shown that besides regulating messenger RNA (mRNA) expression through independent mechanisms, these ncRNAs may 'work' together in complex networks of interactions, fine-tuning their expression [7]. Among different types of miRNA-lncRNA crosstalk, lncRNAs can act as ceRNAs, being a sponge for miRNAs and obstructing their interaction with target mRNAs, consequently favoring the expression of these coding RNAs [8,9]. In the ceRNA hypothesis, ceRNAs refer to all transcripts that have sequence similarity with protein-coding transcripts and act in competition for regulatory molecules, such as lncRNA, pseudogenes, and circular RNA (circRNAs) [8,10].
Different RNAs, expressed concurrently and with a similar set of miRNA recognition elements (MREs), are capable of indirectly regulating one another by competing for a shared, limited pool of miRNA molecules [8,9]. LncRNAs and mRNAs, for example, may compete for their common pair(s)-their 'matching' miRNA(s). Consequently, as miRNAs act as negative post-transcriptional regulators, any activity that decreases their interaction with their target mRNAs may positively contribute to gene expression regulation [7,11,12]. Therefore, the lncRNA-miRNA-mRNA crosstalk seems to crucially influence the overall activity and functional balance of gene networks in a cell [13]. Although still relatively scant, some studies investigating the lncRNA-miRNA crosstalk in CC have already demonstrated the importance of these networks in cervical tumorigenesis, defining specific ncRNAs involved in CC development as well as unveiling new promising biomarkers and potential therapeutic targets for CC. Thus, in the present review, we aim to summarize previously well-described functional interactions between pairs of lncRNA and miRNA (focusing on ceRNA networks) involved in cervical carcinogenesis and to present the current progress of research in this field. Understanding their role in the cervical microenvironment might be of great benefit to better comprehend the pathogenesis of CC and its progression. Additional goals are to discuss their use in CC screening as well as their application as therapeutic strategies for CC treatment.

Cross-Regulatory Mechanisms Involving lncRNAs and miRNAs
Several cross-regulatory mechanisms allow the lncRNA-miRNA crosstalk. LncRNAs may act as ceRNAs, being a sponge for miRNAs and obstructing miRNA action on the target mRNA, consequently favoring its expression [8]. In addition, lncRNAs may be precursors of miRNAs as part of their biogenesis-e.g., H19 lncRNA is the precursor of miR-675 [14]; miRNAs can decrease lncRNA stability through sequence-specific interaction, as occurs between MALAT1 and miR-9 [15]; lncRNAs may compete with miRNAs for shared target mRNAs-e.g., BACE1AS lncRNA, BACE1 mRNA, and miR-485-5p [16]. Such mechanisms highlight the wide range of molecular strategies involved in the lncRNA-miRNA crosstalk in eukaryotic cells. Figure 1 schematically presents such cross-regulatory mechanisms involving lncRNAs and miRNAs.

Figure 1.
Cross-regulatory mechanisms involving lncRNAs and miRNAs. LncRNAs can act as ceR-NAs, sponging miRNAs and obstructing their interaction with target mRNAs, consequently allowing for mRNA translation (upper A); alternatively, the respective miRNA may bind to a target mRNA, inducing mRNA degradation or translation repression (lower A). LncRNAs can be precursors of miRNAs as part of their biogenesis (B). MiRNAs can also decrease lncRNA stability through sequence-specific interaction, leading to lncRNA degradation (C). Finally, lncRNAs can compete with miRNAs for shared target mRNAs (D). (Image source: Mind the Graph).

The ceRNA Hypothesis: A miRNA-Mediated Interplay through Complex Networks of Interactions between Coding and Noncoding RNAs
Initially, miRNAs were shown to inhibit their target genes through binding to MREs of complementary coding RNA transcripts, leading to decreased expression of the respective proteins, either by mRNA degradation or translation inhibition [17]. Once mature, miRNAs guide the miRNA-programmed RNA-induced silencing complex (miR-RISC) to MREs, leading to mRNA destabilization or posttranslational repression, which results in inhibition of gene expression [18]. MREs are normally located in 3′ untranslated regions (3′UTRs) of various types of RNA transcripts, such as of mRNAs and ncRNAs [19].
Despite the important implications of miRNA-mediated regulation in homeostasis/disease, miRNA-mediated repression is surprisingly less impacting than most welltolerated, intrinsic variations in gene expression [20,21]. Through computational analysis, several miRNA-binding sites that can titrate miRNAs and, thereby, regulate miRNA availability were identified, redefining the initial concept about miRNA targets [21]. In such a context, the hypothesis of different transcripts 'talking' to each other using MREs as letters of a new language-known as the ceRNA hypothesis-has emerged [8]. Therefore, several transcripts such as mRNAs, transcribed pseudogenes, and lncRNAs, expressed concurrently and with a similar set of MREs, are capable of indirectly regulating one another by competing for a shared, limited pool of miRNA molecules [8,10].
More recently, experimental evidence started to unveil several ceRNA axes in CC, especially lncRNA-mediated ceRNA networks, highlighting their importance in cervical carcinogenesis. As lncRNAs (i) constitute a large fraction of the transcriptome (with thousands of new lncRNAs reported each year) [22], (ii) exert multiple functions [23], and (iii) Figure 1. Cross-regulatory mechanisms involving lncRNAs and miRNAs. LncRNAs can act as ceRNAs, sponging miRNAs and obstructing their interaction with target mRNAs, consequently allowing for mRNA translation (upper A); alternatively, the respective miRNA may bind to a target mRNA, inducing mRNA degradation or translation repression (lower A). LncRNAs can be precursors of miRNAs as part of their biogenesis (B). MiRNAs can also decrease lncRNA stability through sequence-specific interaction, leading to lncRNA degradation (C). Finally, lncRNAs can compete with miRNAs for shared target mRNAs (D). (Image source: Mind the Graph).

The ceRNA Hypothesis: A miRNA-Mediated Interplay through Complex Networks of Interactions between Coding and Noncoding RNAs
Initially, miRNAs were shown to inhibit their target genes through binding to MREs of complementary coding RNA transcripts, leading to decreased expression of the respective proteins, either by mRNA degradation or translation inhibition [17]. Once mature, miRNAs guide the miRNA-programmed RNA-induced silencing complex (miR-RISC) to MREs, leading to mRNA destabilization or posttranslational repression, which results in inhibition of gene expression [18]. MREs are normally located in 3 untranslated regions (3 UTRs) of various types of RNA transcripts, such as of mRNAs and ncRNAs [19].
Despite the important implications of miRNA-mediated regulation in homeostasis/disease, miRNA-mediated repression is surprisingly less impacting than most welltolerated, intrinsic variations in gene expression [20,21]. Through computational analysis, several miRNA-binding sites that can titrate miRNAs and, thereby, regulate miRNA availability were identified, redefining the initial concept about miRNA targets [21]. In such a context, the hypothesis of different transcripts 'talking' to each other using MREs as letters of a new language-known as the ceRNA hypothesis-has emerged [8]. Therefore, several transcripts such as mRNAs, transcribed pseudogenes, and lncRNAs, expressed concurrently and with a similar set of MREs, are capable of indirectly regulating one another by competing for a shared, limited pool of miRNA molecules [8,10].
More recently, experimental evidence started to unveil several ceRNA axes in CC, especially lncRNA-mediated ceRNA networks, highlighting their importance in cervical carcinogenesis. As lncRNAs (i) constitute a large fraction of the transcriptome (with thousands of new Processes 2021, 9,991 4 of 23 lncRNAs reported each year) [22], (ii) exert multiple functions [23], and (iii) tend to be longer than other ncRNAs [24] (which may reflect in an abundance of multiple MREs), such molecules may actively participate in cervical cell ceRNA interactome. Here, we especially present and discuss the recently identified lncRNA-mediated ceRNA axes in the cervical microenvironment. Figure 1A schematically exemplifies a lncRNA acting as a ceRNA.

LncRNA-Mediated ceRNA Axes in Cervical Tissue: Unveiling Important Networks in CC
Despite the importance of all the aforementioned cross-regulatory mechanisms, the final impact of different ceRNA networks in global gene expression in many different cellular contexts highlights the impact of such complex molecular networks of interactions on both normal physiology and tumorigenesis. In such a context, the perturbation of crucial interactions in ceRNA networks may potentially contribute to carcinogenesis by affecting the balance of the cellular regulatory system [25,26]. Several ceRNA networks involving different transcripts seem to play essential roles in CC. Up to date, only a few studies have focused on ceRNAs involving circRNAs [27,28]; none have evaluated networks involving pseudogenes, while a considerable number of original articles have already described several ceRNA networks led by different lncRNAs-described here as lncRNA-mediated ceRNAs.
Through an extensive literature search, we compiled different lncRNA-mediated ceRNA networks involved in CC. Despite all the potentially relevant studies initially identified, only those strongly evidencing the proposed molecular interactions and their final impact on cervical carcinogenesis were included and discussed in the present review. Table 1 summarizes the main characteristics of the included studies (and Table S1 presents information on the quality assessment of all potentially relevant papers). Additionally, in order to highlight the major impacts of such dynamic molecular changes over multiple cellular processes, we carefully focused on the different adopted strategies (in silico, in vitro, and in vivo approaches, as well as experiments performed using patients' samples), as presented through the following sections.
According to the score obtained in the quality assessment analysis, articles were classified into A high-quality assessment or B moderatequality assessment (see Table S1). CeRNA = competing endogenous RNA; [ref] = reference; -= not applicable; X = not performed; ✓ ✓ = performed. * BBOX1-AS1 upregulates HOXC6 expression through miR-361-3p but also through ELAV-like RBP 1 (HuR). ** Apart from sponging miR-505 and regulating the expression of ZEB2, CTS was shown to activate the TGF-β/SMAD signaling pathway. *** ceRNA interaction, but with an additional mechanism described (as MKL1 exerting a positive loop over the HOTAIR gene and EZH2 exerting a negative loop over the miR-214 gene). 1 The transcriptional expression of POU3F3 is activated by transcription factor SP1. Here, miRNA nomenclature was reproduced exactly as mentioned by the author in the respective article (thus, not all follow miRBase's most recent IDs). For further information, consult the miRBase database. All the abovementioned miRNAs are hsa-miRNAs (Homo sapiens miRNAs). The lncRNA nomenclature applied here was reproduced exactly as mentioned by the author in the respective article and is better described in the list of abbreviations (except for some lncRNAs such as RP11-552-M11.4 and TCONS_0026907, which correspond to transcript codes).

In Silico Predictions and In Vitro Strategies Reveal Molecular Interactions between Different Components of the Proposed ceRNA Networks
Basically, in silico analysis explored the prediction of ceRNA interactions based on the identification of MREs within the transcripts of interest, representing an interesting 'start' in order to suggest potential downstream targets of distinct lncRNAs and miRNAs. Almost all studies performed in silico analysis, aiming to predict the potential miRNAs targeted (sponged) by distinct lncRNAs as well as potential target mRNAs of the previously predicted miRNA.
After predicting potential lncRNA-miRNA and miRNA-mRNA molecular interactions, in order to validate such results, in vitro approaches were performed. Essentially, the adopted techniques focused on observing two major aspects: (1) evaluating the patterns of expression of the different components of the investigated ceRNA axes (lncRNA/miRNA/mRNA) as well as positive/negative correlations between them, and (2) validating potential molecular interactions within such components.
Therefore, most studies firstly investigated the expression of the candidate lncRNA (and miRNA/mRNA) in different CC cell lineages compared to control lineage(s). Overall, for almost all ceRNA axes, the following pattern of expression was observed: the involved lncRNAs and mRNAs were significantly upregulated in CC cell lines in comparison to control cell lineage(s), while their 'matching' miRNA was significantly downregulated. For instance, HOTAIR expression was significantly upregulated in different CC cell lines in comparison to that in control lineages in different studies as well as the involved mRNAs [37][38][39][40]. Contrarily, the expression of the involved miRNAs was significantly downregulated [37][38][39][40]. Curiously, an opposed pattern was observed for nontumoral cells in comparison to normal cells.
As mentioned above, such pattern was reproducible in almost all the included studies except for the references investigating ceRNA axes led by tumor-suppressive lncRNAs such as GAS5 [35], PTENP1 [56], STXBP5-AS1 [65], TUSC8 [70], WT1-AS [71], and ZNF667-AS1 [74]. In those cases, an opposite pattern of expression was observed (with the lncRNA and mRNA being downregulated in CC, while the miRNA was upregulated). For instance, TUSC8 and PTEN were significantly downregulated in different CC cell lines in comparison to that in control lineages [56], while miR-641 was significantly upregulated in the respective studies [37][38][39][40]. Likewise, an opposed pattern was observed for nontumoral cells in comparison to normal cells.
Additionally, based on those results, several studies also highlighted important correlations between lncRNAs × miRNAs and/or miRNAs × mRNAs involved in distinct ceRNA axes, proposing potential direct interactions between some of these molecular targets. For all ceRNA axes, a significantly positive correlation was observed between the involved lncRNA and mRNA, while a negative correlation was observed between the lncRNA and the miRNA (or between the miRNA and the mRNA). As expected, such patterns of correlation were also observed between GAS5, PTENP1, STXBP5-AS1, TUSC8, WT1-AS, and ZNF667-AS1 and the particular molecular components involved in the respective axes.
Furthermore, the lncRNA-miRNA and/or miRNA-mRNA crosstalk was further validated. Here, we only reported on ceRNA axes validated by the luciferase reporter assay, considered the 'gold standard' technique in validating miRNA targets [75], which may be extended for validating lncRNA-miRNA targets. Details on the molecular interactions of different lncRNA-mediated ceRNA axes involved in CC are provided in Table S2. Essentially, all interactions previously predicted by in silico methodologies and suggested by the observed patterns of expression/correlations were ultimately validated.

In Vitro and In Vivo Experiments Demonstrate the Final Biological Impacts of the Respective Molecules/ceRNA Axes over Important Cellular Processes and Tumor Growth
Despite the remarkable diversity of neoplastic diseases, several biological capabilities acquired during the multistep development of human tumors-well known as 'the hallmarks of cancer'-seem to be present in all cancer types. Such capabilities include sustaining proliferative signaling, enabling replicative immortality, resisting cell death, and activating invasion and metastasis [76]. In such a context, the effects of the involved lncRNA (and miRNA/mRNA) over crucial cellular processes closely related to the aforementioned hallmarks (such as cell proliferation, cell cycle, apoptosis, migration/invasion, and metastasis) were investigated. Through in vitro gain-and/or loss-of-function strategies evaluating the effect of the lncRNA and/or miRNA (silenced and/or overexpressed), it was possible to observe the biological effects induced by these molecules and their respective ceRNA network.
Essentially, almost all described ceRNA axes led to an increase in cell proliferation and/or stimulated colony formation and/or cell cycle progression, as well as invasion and/or migration; contrarily, apoptosis was inhibited (see Table 2). For instance, HO-TAIR led to an increase in cell proliferation [37,39,40], stimulated cell migration and invasion [37,38], and inhibited apoptosis [37,39,40]. Likewise, MALAT1-, NEAT1-, OIP5-AS1, and XIST-mediated ceRNA axes were reported by independent studies to affect distinct cellular processes [46][47][48]50,53,54,72,73]. Therefore, these ceRNA axes present oncogenic roles as they favor cervical carcinogenesis. On the other hand, a few lncRNAmediated axes led by tumor-suppressive lncRNAs, such as GAS5 [35], PTENP1 [56], STXBP5-AS1 [65], TUSC8 [70], WT1-AS [71], and ZNF667-AS1 [74], present a tumorsuppressive role in CC. For instance, GAS5, STXBP5-AS1, WT1-AS, and ZNF667-AS1 overexpression suppressed cell growth and/or cell cycle progression and migration/invasion [35,65,71,74], while GAS5-and PTENP1-mediated ceRNA axes markedly increased apoptosis in CC cells [35,56]. It is actually important to highlight that such observed effects are mainly exerted through the lncRNA involved (and in a condition where they are 'overexpressed'). As they are downregulated in CC cells/tissue/patients' samples, the endogenous competitions through such axes are not likely to occur.    For example, some references investigated the silencing of the respective lncRNA implicated in decreasing or increasing cellular proliferation, while others investigated such an effect using both strategies of silencing and overexpressing the involved lncRNA. Differences like these were observed within strategies to evaluate lncRNA action over these different biological parameters as well as, in some cases, of the different miRNAs and/or mRNAs effects. In general, the authors present such effects as a consequence of the respective lncRNA, although, sometimes, the observed effects are presented as a consequence of the respective miRNA (such as for DANCR/miR-665/TGFBR1 and SNHG12/miR-125b/STAT3) or mRNA (such as for MALAT1/miR-202-3p/POSTN)). Despite that, based on all adopted experiments and strategies, we may consider the whole ceRNA axis involved. The arrows represent cellular proliferation and apoptosis (↑ = increases; ↓ decreases); for colony formation, cell cycle progression, migration, invasion, and metastasis (↑ = promotes; ↓ inhibits). * Some references mentioned that the respective lncRNA had an influence over metastasis based on in vitro experiments, although such effects were not fully demonstrated by the performed experiments. ** Studies providing evidence that metastasis is, in fact, occurring (based on patients with lymph node metastasis occurrence and/or in vivo evidence). *** The observed biological effects were related to the overexpression of the involved lncRNAs. Those ceRNA axes seem not to be occurring in CC cell lineages since these lncRNAs and mRNAs are downregulated in such context; contrarily, they seem to be predominantly occurring in normal cell lines. 1 Cell proliferation was also evaluated in ME180 and SiHa cells. 2 Cell proliferation was also evaluated in CaSki, and C33A cells. Here, miRNA nomenclature was reproduced exactly as mentioned by the author in the respective article (thus, not all follow miRBase's most recent IDs). For further information, consult the miRBase database. All the abovementioned miRNAs are hsa-miRNAs (Homo sapiens miRNAs). The lncRNA nomenclature applied here was reproduced exactly as mentioned by the author in the respective article and is better described in the list of abbreviations (except for some lncRNAs such as RP11-552-M11.4 and TCONS_0026907, which correspond to transcript codes).
Further on, it is important to mention that the capability of inducing metastasisanother hallmark of cancer-is another evaluated biological parameter, although this is not quite correct in some cases. Several authors reported that 'the respective lncRNA stimulated invasion and migration and also increased metastasis', although the performed in vitro experiments did not fully support such effect. In fact, because some of the involved ncRNAs (and their respective axes) presented a pattern of expression of epithelial-mesenchymal transition (EMT)-related proteins consistent with the occurrence of this biologic process, the authors inferred their effect over metastasis. Such inference should be presented as a suggestion since the deregulated expression of EMT-related components may indeed suggest their influence over metastasis but not as a closed statement supporting such an effect. Some references provided proper evidence that metastasis is, in fact, occurring (based on lymph node metastasis occurrence in patients and/or in vivo evidence, using appropriate animal models of metastasis) [31,42,43] or is somehow prevented by tumorsuppressive lncRNA-mediated axes [56,65,71,74].
Finally, in vivo experiments have confirmed the involvement of the respective ceRNA axes in modulating cell proliferation, cell cycle, apoptosis, and tumor growth. The observed expression patterns of classical proliferation-related (such as Ki-67), cell-cycle-related (such as C-Fos, Cyclin D1, and Bcl-2), and apoptosis-related (such as cleaved caspase-3) markers corroborated such effects. For instance, an increased percentage of Ki-67-positive cells was observed in tumors induced by the SC injection of cells overexpressing the respective lncRNA (or silencing the involved target miRNA) [57,67,73]. Likewise, different lncRNAmediated ceRNA axes were found to promote tumor growth in vivo, both independently as well as through common molecular networks. Additionally, the expression of different EMTrelated (such as VEGF, vimentin, E-cadherin) markers was also evaluated, suggesting the involvement of some lncRNA-mediated axes in inducing EMT. Details on the investigated targets as well as the adopted strategies and the main findings are provided in Table S3.

Evaluating Patients: Corroborating the Involvement of lncRNA-Mediated ceRNA Networks in Cervical Carcinogenesis
In general, the aforementioned strategies are worthy to better propose potentially involved ceRNA axes and confirm the direct interaction(s) among their components but have no clinical value if the obtained results are not somehow confirmed in the patients' biological samples. Therefore, experiments involving patients' samples-such as methods evaluating the expression profiles of the involved ncRNAs-represent an important strategy to ultimately suggest and/or confirm the involvement of the lncRNA-miRNA (or ceRNA axis) in cervical carcinogenesis. Almost all references adopted this strategy. Table 3 is an overview of sample description, performed methods, and major results.
In agreement, the respective involved mRNAs were found upregulated under the same conditions, while their 'matching' miRNAs were found downregulated. Consequently, a significantly positive correlation between the involved lncRNA and mRNA as well as a negative correlation between the lncRNA and miRNA (or between the miRNA and mRNA, when applicable) were also reported through experiments involving patient's samples. On the other hand, GAS5, PTENP1, STXBP5-AS1, TUSC8, WT1-AS, and ZNF667-AS1 were downregulated in CC samples in comparison to NTT [35,56,65,70,71,74], and opposing patterns/correlations between them and the components of the different ceRNA axes that they take part in were observed in comparison to the ones mentioned above. Therefore, this, once again, highlights their tumor-suppressive role. Table 3. Studies validating the involvement of the lncRNA-mediated ceRNA networks in patients with CC.

Sample Description
Observed Results  Some references do not use adjacent NTT but, instead, normal cervix (or healthy controls). * Other candidate RNAs were tested by RT-qPCR, although not validated. No reference has evaluated the expression of the involved RNAs at the serum/plasmatic level (nor in any other biofluid). ** The article mentions a serum measurement, although not clearly explained/presented. *** Those ceRNA axes seem not to be occurring in CC tissue since these lncRNAs and mRNAs are downregulated in such context; contrarily, they seem to be predominantly occurring in NTT. 1 The expression of WT1-AS in cervical cancer tissues was checked by in situ hybridization. Here, miRNA nomenclature was reproduced exactly as mentioned by the author in the respective article (thus, not all follow miRBase's most recent IDs). For further information, consult the miRBase platform. All the abovementioned miRNAs are hsa-miRNAs (Homo sapiens miRNAs). The lncRNA nomenclature applied here was reproduced exactly as mentioned by the author in the respective article and is better described in the list of abbreviations (except for some lncRNAs such as RP11-552-M11.4 and TCONS_0026907, which correspond to transcript codes).

CC (n) SCC (n) AD (n) NTT (n) Normal (n) SIL/CIN (n) Expression Levels of Target Molecules (RNAs by RT-qPCR/Proteins by WB)
Interestingly, some references also measured the final 'mRNA product'-the expression of the respective protein. In CC samples, in comparison to NTT, HLA-G [39], KIF23 [42], MAPK1 [67], and periostin (POST) [46] were found overexpressed, while PTEN was found underexpressed [56,70], at the protein level, in agreement with what had been seen at the mRNA level. No reference assessed the serum/plasma expression level of the involved lncRNA in the CC cases and age-matched controls.
Collectively, we observed that different lncRNA-mediated ceRNA axes seem to be closely associated with CC. Together, such observations demonstrate the abnormal patterns of gene expression of the different molecular components involved in the diverse lncRNA-mediated ceRNA networks described here, suggesting their important role in cervical carcinogenesis.

Important Cancer-Related RNAs Working through Different lncRNA-Mediated Axes in CC
Apart from unveiling important ceRNA networks involved in cervical carcinogenesis, we also aimed to highlight some important molecules that work through distinct lncRNAmediated ceRNA networks. Some are very well described oncogenic RNAs in different cancers, while others have emerged as novel cancer-related RNAs, highly deregulated in a variety of cancers. Table 4 summarizes important details about the involved RNAs.

OIP5-AS1
OIP5-AS1 maintains cell proliferation in embryonic stem cells and can bind to and negatively regulate the activity of multiple cellular RNAs and microRNAs, including cyclin G associated kinase and HuR Oncogene Oncogene [84,88] XIST Besides regulating X chromosome inactivation, XIST is a classic example of how lncRNAs can exert multilayered and fine-tuned regulatory functions by acting as a molecular scaffold for the recruitment of distinct protein factors as well as acting as a transcriptional regulator for numerous genes Oncogene Oncogene [84,89,90] miRNA miR-140-5p Inhibition of cell proliferation, migration, and invasion, as well as induction of apoptosis TS TS [63,73,[91][92][93][94][95]

MAPK1/MAPK1
MAPK1 is part of the MAPK pathway-an important bridge in the switch from extracellular signals to intracellular responses-that is ultimately involved in stimulating cell proliferation and differentiation as well as modulating transcription regulation and development Oncogene Oncogene [106,107] PTEN/PTEN PTEN is a very well-described tumor-suppressive protein, regulating proliferation and cell survival TS TS/ Oncogene ** [108][109][110]

ROCK1/ROCK1
ROCK1 is part of the ROCK family, which plays a central role in the organization of the actin cytoskeleton and is involved in a wide range of fundamental cellular functions. Stimulates tumor growth, angiogenesis, and metastasis.

Oncogene
Oncogene [111][112][113] CC = cervical cancer; LSD1 = lysine-specific demethylase 1; PRC2 = polycomb repressive complex 2; HOXD = homeobox D Cluster; HuR = ELAV-like RNA-binding protein 1; ROCK = Rho-associated coiled-coil kinase; TS = tumor suppressor. * miR-148a and miR-148a-3p represent the same miRNA. ** PTEN's overexpression (as a consequence of NEAT1 sponging miR-9-5p) ultimately promoted CC cell growth, therefore presenting an oncogenic role. Despite its classical ability to regulate proliferation and cell survival, PTEN also seems to play different intriguing roles under different cellular contexts, which may, in part, explain such contradictory patterns. However, as only one study reported such an effect, more experiments are necessary in order to corroborate this result. Here, miRNA nomenclature was reproduced exactly as mentioned by the author in the respective article (thus, not all follow miRBase's most recent IDs). For further information, consult the miRBase platform. All the abovementioned miRNAs are hsa-miRNAs (Homo sapiens miRNAs). The lncRNA nomenclature applied here was reproduced exactly as mentioned by the author in the respective article and is better described in the list of abbreviations (except for some lncRNAs such as RP11-552-M11.4, and TCONS_0026907, which correspond to transcript codes).

HOTAIR, MALAT1, NEAT1, OIP5-AS1, and XIST
Among several lncRNAs, HOTAIR, MALAT1, NEAT1, OIP5-AS1, and XIST stood out as important players in different lncRNA-mediated ceRNA networks in CC. Interestingly, HOTAIR was the most described lncRNA, leading different ceRNA axes. All references investigating HOTAIR's role in CC performed in silico predictions and pointed to interesting molecular targets, which were confirmed by in vitro experiments. Furthermore, HOTAIR overexpression was found to increase cell proliferation [37,39,40], migration/invasion [37,38], and prevent apoptosis [37,39,40] through different ceRNA axes. In vivo experiments also confirmed HOTAIR's ability to promote CC tumor growth [37,39,40]. Additionally, HOTAIR expression in CC tissues, in comparison to adjacent normal tissues, was found upregulated in all studies. Therefore, HOTAIR seems to be closely related to CC development. In fact, HOTAIR is a very well-described oncogenic lncRNA in different cancers [114] and by far the most studied lncRNA involved in CC [79,84,115].
NEAT1 and XIST were also reported by independent studies as participating in more than one ceRNA network, exerting protumorigenic effects. NEAT1 has emerged as a novel cancer-related lncRNA [86], highly de/upregulated in a variety of cancers [87]. Similarly, the high expression of XIST has been associated with poor prognosis and metastasis in different cancer types, representing a predictive factor of poor prognosis in human cancers [89,90]. In fact, both have been described as crucial modulators in several cellular processes, influencing cervical oncogenesis [84]. Through in silico predictions and in vitro experiments, the molecular interactions between NEAT1-and XIST-mediated axes were confirmed. As for HOTAIR, such interactions were associated with an increase in cell proliferation, colony formation, and invasion and/or migration [48][49][50]72,73], together with the inhibition of apoptosis [49,50,72,73]. NEAT1 and XIST also promoted CC tumor growth [50,73] in vivo and were found upregulated in CC tissues in comparison with adjacent normal tissues [48,49,72,73], confirming the oncogenic role of both lncRNAs in CC.
Additionally, MALAT1 and OIP5-AS1 also stood out as important lncRNAs involved in different lncRNA-mediated ceRNA axes. Likewise, they were found to stimulate cell proliferation, invasion and/or migration [46,47,53,54], and prevent apoptosis [47,54] through different ceRNA axes, and were found upregulated in CC tissues in comparison with adjacent normal tissues [46,47,53,54]. Both seem to have an important role in cancer [88,116] and cervical carcinogenesis [84]. Differently from HOTAIR, NEAT1 and XIST, the protumorigenic effects of MALAT1 and OIP5-AS1 on cancer, in general, and on CC are still emerging.
Similarly, mir-148a-3p (or miR-148a) was reported as an important molecular target of HOTAIR and SNHG4, consequently blocking HLA-G and MET expression modulation [39,60], which ultimately led to CC progression. Despite its tumor-suppressive role in CC, mir-148a-3p was found to be aberrantly expressed in various cancers, having been identified as a tumor suppressor and also as an oncogenic miRNA, depending on the molecular context [119]. Other miRNAs such as let-7a, and miR-9-5p were involved in distinct ceRNA axes, although this was investigated by the same reference [48,58]. Despite the evidence, the role of such miRNAs in CC development is not quite established.

MAPK1, PTEN, and ROCK1
Likewise, coinciding mRNAs participating in more than one ceRNA axis were also observed, such as MAPK1, PTEN, and ROCK1. For instance, MAPK1 was reported as competing for different miRNAs with HOTAIR [37] and TDRG1 [67]. As HOTAIR and TDRG1 sponge miR-23b and miR-326, respectively, MAPK1 translation was expected to be stimulated. In fact, MAPK1 was upregulated (at both the mRNA and protein levels) in CC in comparison to matched controls [67]. Similarly, ROCK1 also caught our attention, taking part in two different ceRNA networks and being upregulated in CC in comparison to normal samples [32,54]. Accumulating evidence supports MAPK1 and ROCK1 oncogenic roles, which corroborates the results previously presented.
Additionally, PTEN was reported as participating in different axes [35,48,56,65,70]. As for MAPK1, PTEN was also deregulated at both the mRNA and protein levels, although exhibiting a different expression pattern (PTEN mRNA/PTEN protein were downregulated in CC in comparison to their matched controls) [56,70], suggesting that such ceRNA axes occurs primarily in normal tissue (see Section 6). Classically, PTEN is a well-described tumor-suppressive protein [108][109][110], and PTEN underexpression in CC cells/tissue corroborates such a statement. Intriguingly, PTEN overexpression (as a consequence of NEAT1 sponging miR-9-5p) ultimately promoted CC cell growth [48]. Despite its classical ability to regulate proliferation and cell survival, PTEN also seems to play different intriguing roles under different cellular contexts [120], which may, in part, explain such opposing patterns. However, as only one study reported such an effect, more experiments are necessary in order to corroborate this result.

Protumorigenic and Tumor-Suppressive lncRNA-Mediated ceRNA Axes Involved in Cervical Carcinogenesis
Ultimately, after carefully looking at individual lncRNA-mediated ceRNA networks as well as important molecules working through those axes, it is also relevant to define general patterns. Although quite a challenge, after carefully gathering all the results, two major patterns were observed: one with particular molecular endogenous competitions occurring predominantly in CC, highly associated with a protumorigenic final effect (Figure 2), and another with such molecular interactions most likely occurring in a normal cervical cell, with the lack of such competitions being associated with CC ( Figure 3).
Mainly, we observed lncRNA-mediated axes with a protumorigenic role, with their respective lncRNAs (and mRNAs) being significantly upregulated in CC cell lines in comparison to control cell lineages, while their 'matching' miRNA was significantly downregulated under the same conditions. Predominantly, such lncRNA-mediated ceRNA axes led to a final state of sustained proliferative signaling, allowed replicative immortality, resisted cell death, and showed 'signs' of presenting the ability to activate (or somehow being associated with) migration/invasion and metastasis-classical cancer hallmarks. Furthermore, they were found to stimulate cell proliferation and cervical tumor growth (in vivo) and were ultimately detected in patients' biological samples. HOTAIR-, MALAT-1-NEAT1-, OIP5-AS1, and XIST-mediated ceRNA axes were widely reported as presenting protumorigenic roles in CC.
Contrarily, some axes led by GAS5 [35], PTENP1 [56], STXBP5-AS1 [65], TUSC8 [70], WT1-AS [71], and ZNF667-AS1 [74] seemed to be 'inactive' in CC cells/tissue/patients' samples. In fact, such lncRNAs (and mRNAs participating in their respective ceRNA axes, when applicable) were scarce in CC cells, impairing the 'endogenous competition'. Overall, when overexpressed, they inhibited cell proliferation and/or colony formation and/or invasion/migration and increased apoptosis. In addition, GAS5, WT1-AS, and ZNF667-AS1 were found to inhibit tumor growth in vivo [35,71,74]. The patterns of correlation between the involved molecules were the same as for 'protumorigenic' axes, as expected, considering the ceRNA hypothesis. Together, such results suggest that these particular molecular competitions have the potential to occur, although they do not seem to be occurring (or broadly occurring) in CC cells/tissue/patients' samples since the involved lncRNAs are downregulated (and consequently 'unable' to sponge their 'matching' miRNAs). Accordingly, such specific endogenous competitions seem to primarily take place in normal tissue, being particularly important for cervical homeostasis. sses 2021, 9, x FOR PEER REVIEW 16 of different intriguing roles under different cellular contexts [120], which may, in part, e plain such opposing patterns. However, as only one study reported such an effect, mo experiments are necessary in order to corroborate this result.

Protumorigenic and Tumor-Suppressive lncRNA-Mediated ceRNA Axes Involved in Cervical Carcinogenesis
Ultimately, after carefully looking at individual lncRNA-mediated ceRNA networ as well as important molecules working through those axes, it is also relevant to defi general patterns. Although quite a challenge, after carefully gathering all the results, tw major patterns were observed: one with particular molecular endogenous competitio occurring predominantly in CC, highly associated with a protumorigenic final effect (F ure 2), and another with such molecular interactions most likely occurring in a norm cervical cell, with the lack of such competitions being associated with CC ( Figure 3). LncRNA-mediated ceRNA crosstalk in tumoral cervical tissue. Some lncRNA-mediated ceRNA axes (led by upregulated lncRNAs, as HOTAIR, MALAT1, NEAT1, OIP5-AS1, and XIST) seem to primarily take place in tumoral tissue, leading to a state of constant cell proliferation and cell cycle progression, inducing migration/ invasion, and preventing apoptosis. Note that as tumor-suppressive lncRNAs such as GAS5, PTENP1, STXBP5-AS1, TUSC8, WT1-AS, and ZNF667-AS1 are downregulated in cervical tumoral tissue, they are impaired from exerting a final tumor-suppressive effect through different lncRNA-mediated ceRNA axes. The protein FOS is represented by a question mark (?) in order to emphasize that, although it has not yet been investigated, it probably presents a deregulated pattern of expression, as observed for the respective gene that codes this specific protein.
Mainly, we observed lncRNA-mediated axes with a protumorigenic role, with th respective lncRNAs (and mRNAs) being significantly upregulated in CC cell lines in co parison to control cell lineages, while their 'matching' miRNA was significantly downre ulated under the same conditions. Predominantly, such lncRNA-mediated ceRNA ax LncRNA-mediated ceRNA crosstalk in tumoral cervical tissue. Some lncRNA-mediated ceRNA axes (led by upregulated lncRNAs, as HOTAIR, MALAT1, NEAT1, OIP5-AS1, and XIST) seem to primarily take place in tumoral tissue, leading to a state of constant cell proliferation and cell cycle progression, inducing migration/invasion, and preventing apoptosis. Note that as tumor-suppressive lncRNAs such as GAS5, PTENP1, STXBP5-AS1, TUSC8, WT1-AS, and ZNF667-AS1 are downregulated in cervical tumoral tissue, they are impaired from exerting a final tumor-suppressive effect through different lncRNA-mediated ceRNA axes. The protein FOS is represented by a question mark (?) in order to emphasize that, although it has not yet been investigated, it probably presents a deregulated pattern of expression, as observed for the respective gene that codes this specific protein.
(in vivo) and were ultimately detected in patients' biological samples. HOTAIR-, MALAT-1-NEAT1-, OIP5-AS1, and XIST-mediated ceRNA axes were widely reported as presenting protumorigenic roles in CC. Consequently, through such interactions, they seem to inhibit abnormal cell proliferation and migration/ invasion and to stimulate apoptosis, ultimately maintaining cervical tissue homeostasis. Note that as oncogenic lncRNAs such as HOTAIR, MALAT1, NEAT1, OIP5-AS1, and XIST are downregulated in normal cervical tumoral tissue, they are impaired from exerting a protumorigenic effect through different lncRNA-mediated ceRNA networks. Potential target mRNAs are represented by a question mark (?) in order to emphasize that, although they have not yet been investigated by the included references, they probably present a deregulated pattern of expression, as observed for the respective proteins coded by these genes.
Contrarily, some axes led by GAS5 [35], PTENP1 [56], STXBP5-AS1 [65], TUSC8 [70] WT1-AS [71], and ZNF667-AS1 [74] seemed to be 'inactive' in CC cells/tissue/patients samples. In fact, such lncRNAs (and mRNAs participating in their respective ceRNA axes when applicable) were scarce in CC cells, impairing the 'endogenous competition'. Overall, when overexpressed, they inhibited cell proliferation and/or colony formation and/or invasion/migration and increased apoptosis. In addition, GAS5, WT1-AS, and ZNF667-AS1 were found to inhibit tumor growth in vivo [35,71,74]. The patterns of correlation between the involved molecules were the same as for 'protumorigenic' axes, as expected considering the ceRNA hypothesis. Together, such results suggest that these particular molecular competitions have the potential to occur, although they do not seem to be occurring (or broadly occurring) in CC cells/tissue/patients' samples since the involved lncRNAs are downregulated (and consequently 'unable' to sponge their 'matching' miR-NAs). Accordingly, such specific endogenous competitions seem to primarily take place in normal tissue, being particularly important for cervical homeostasis.
Additionally, despite primarily occurring in tumoral tissue or nontumoral tissue driving the development of CC or guaranteeing cervical homeostasis, we understand that . LncRNA-mediated ceRNA crosstalk in normal cervical tissue. Some lncRNAs, such as GAS5, PTENP1, STXBP5-AS1, TUSC8, WT1-AS, and ZNF667-AS1, are abundant in a normal cervical cell and may compete for 'matching' miRNA(s), acting as molecular sponges, indirectly regulating protein-coding gene expression by increasing translation. Consequently, through such interactions, they seem to inhibit abnormal cell proliferation and migration/invasion and to stimulate apoptosis, ultimately maintaining cervical tissue homeostasis. Note that as oncogenic lncRNAs such as HOTAIR, MALAT1, NEAT1, OIP5-AS1, and XIST are downregulated in normal cervical tumoral tissue, they are impaired from exerting a protumorigenic effect through different lncRNA-mediated ceRNA networks. Potential target mRNAs are represented by a question mark (?) in order to emphasize that, although they have not yet been investigated by the included references, they probably present a deregulated pattern of expression, as observed for the respective proteins coded by these genes.
Additionally, despite primarily occurring in tumoral tissue or nontumoral tissue, driving the development of CC or guaranteeing cervical homeostasis, we understand that different intensities of endogenous competition may occur in different cellular microenvironments since we are not dealing with a 'binary condition'. Thus, intermediate patterns of expression may be associated with greater or lesser competition, considering different ceRNA axes. In such a context, the relative concentration of the involved ceRNAs and their target miRNAs seem to be crucial for the occurrence (or not) of a particular ceRNA network [8]. Moreover, depending on the molecules involved, different responses may also occur (as a consequence of the similarity between the molecules and their availability to allow for such competition). It is important to consider aspects such as the available number and sets of MREs (since different transcripts may host different numbers and types of MREs in their 3 UTR) and the accessibility of those MREs (for instance, lncRNAs may present different structural forms with different binding partners, which may also influence the accessibility of specific miRNA molecules). Another interesting observation is the fact that different axes are comprised of different components. Such diversity is already expected once one ncRNA may regulate different other ncRNAs and/or mRNAs (proteins).
At the cellular level, an influence of the subcellular localization of the involved ceRNAs and miRNAs may be of great importance. Since different ceRNAs and miRNAs present different patterns of expression between different subcellular compartments in a given cell type at a given moment [121][122][123][124], subcellular localization stands as one of the many important factors influencing the occurrence (or not) of different ceRNA networks.

Remarks and Perspectives
Despite the recent advances toward understanding the molecular basis of CC, solid knowledge on the role of ncRNAs in such context and their involvement in important molecular networks and biological processes is still scarce. Once considered 'junk' sequences, ncRNAs have recently emerged as 'treasure' [125], standing as important key players in gene regulatory networks. Such molecules work in integrated networks, together with other biomolecules, including coding and other ncRNAs, DNA, and proteins [24]. Among them, ceRNAs have emerged as important post-transcriptional regulators that alter gene expression through a miRNA-mediated mechanism [13], playing a crucial role in both homeostasis and tumorigenesis [25]. In fact, the disruption of the equilibrium of ceRNAs and miRNAs seems to be critical for the activity of multiple ceRNA networks, influencing the outcome of several diseases, including cancer [126].
In the present review, we have comprehensively summarized different lncRNAmediated ceRNA axes and promising ceRNA networks involved in cervical carcinogenesis as well as highlighted their major impacts on dynamic molecular changes over important cellular processes. Despite such effort, we emphasize the urgent need for more studies investigating ceRNA networks in CC. Collectively, with new studies investigating overlapping ncRNAs (and mRNAs) and consolidating the molecular basis of the ceRNA theory, it will be possible to reinforce some of the evidence presented here and better understand the importance of these networks in CC development and/or progression.
Consequently, the potential application of such molecules in CC screening (diagnosis, prognosis) will become more achievable. Indeed, it is important to highlight that the ceRNA model not only suggests the existence of a complex regulatory network in cancer but also implies the possibility of using a panel of network molecules to diagnose and predict cancer [127]. Additionally, the future use of such targets, focusing on therapeutic strategies for CC treatment, is also likely to occur. Different RNA-based strategies have emerged in the past few years [128], with several RNA-based products, mostly ncRNAs, successfully approved for clinical use [129,130]. Therefore, the use of the ceRNA concept, focusing on therapeutics, is also a reasonable possibility. For example, the use of artificial miRNA sponges containing various miRNA binding sites, mimicking the characteristics of ceRNAs as miRNA inhibitors, could be an interesting strategy for 'silencing' axes with upregulated miRNAs. Depending on the final desired effect, different strategies might be adopted. Together, they could represent new platforms for RNA-based therapeutic applications.
Finally, as limitations observed through the careful analysis of the included articles, it is important to mention that, with the exception of one, all other studies included patients' samples from the Chinese population. This represents an important limitation in the field of ceRNAs in CC. For sure, it is of great importance to study different populations, especially considering the existence of differential genetic backgrounds among them. Genetic variants may lead to different individual responses that may impact the course of the disease. Additionally, not discriminating among CC histological types, as well as the involved HPV type, may also be of important concern since we understand that different ceRNA axes may be associated with these different parameters. The lack of information on the ceRNAs' and miRNAs' subcellular localization and their relative abundance (i.e., the molecular ratio between them) also represents a limitation; it would be very enlightening to access such parameters. Overall, we observed problems concerning the adequate description of the employed methods and samples. All the abovementioned problems may be easily addressed, and we encourage future studies to more carefully present methods/results. This is important in order to standardize the included strategies, facilitating a better overview of the topic.

Conclusions
The involvement of ceRNA networks in cancer is an emerging topic, and their role in CC progression/development is becoming clearer. In the ceRNA hypothesis, lncRNAs and mRNAs act as 'players' competing for their common pair(s)-their 'matching' miRNA(s). Subsequently, changes in gene expression at the mRNA level may occur, affecting the overall activity and functional balance of gene networks in a cell. Essentially, this is a recent research field, with few studies investigating some of the many possible ceRNA axes involved. Despite that, the importance of these ceRNA networks in cervical carcinogenesis has already been demonstrated. Through an extensive literature search, we summarized previously described functional interactions between lncRNAs and miRNAs, focusing on lncRNA-mediated ceRNA networks involved in CC. Several axes seem to contribute to CC progression/development, while others seem to be important for cervical homeostasis maintenance. In the future, such molecular networks may present a potential application in CC screening (diagnosis, prognosis), and they may become important therapeutic targets for CC. Despite the enthusiasm, it is still an enormous challenge to fully understand the ceRNA regulatory networks in CC, and it will be quite a long journey, although a worthy and exciting one.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/pr9060991/s1. Table S1: Quality assessment form used to evaluate the quality of the included studies. Table S2: Molecular interactions of different lncRNA-mediated ceRNA networks involved in CC (in vitro strategies). Table S3  Institutional Review Board Statement: Ethical review and approval were waived for this study, due to the fact that the present work is a review article, and approval has already being declared in each of the particular studies cited here.
Informed Consent Statement: Patient consent was waived due to the fact that the present work is a review article, and formal consent has already being obtained and declared in each of the particular studies cited here.

Data Availability Statement:
The data presented in this study are available in the supplementary material.