Adenosine is produced by canonical and non-canonical pathways

Purinergic signalling is characterized by the activity of ectoenzymes called ectonucleotidases, which control the levels of nucleotides and nucleosides present in the extracellular microenvironment (Ref. Reference Zimmermann, Zebisch and Sträter9). There are different families of ectonucleotidases described, with varied hydrolysis ability and affinity potential for purines (ATP, adenosine diphosphate (ADP), adenosine monophosphate (AMP) and ADO) and pyrimidines (uridine-5′-triphosphate (UTP) and uridine diphosphate (UDP)) (Ref. Reference Zimmermann, Zebisch and Sträter9). Among these families, we can mention E-NTPDases (ectonucleoside triphosphate-diphosphohydrolases), CD73, E-NPPs (ectonucleotide pyrophosphatase/phosphodiesterases), alkaline phosphatase and adenosine deaminase (ADA). These enzymes, alone or in combination, will exert an essential role in regulating purinergic signal transmission through hydrolysis of nucleotides/nucleosides, thereby controlling persistence, conclusion or restart of the signalling. CD73, specifically, is an ectoenzyme encoded by the NT5E gene, able to catabolize AMP into ADO at the extracellular level (Ref. Reference Zimmermann, Zebisch and Sträter9).

Extracellular nucleotides and nucleosides exert their effects through interactions with specific membrane receptors, known as purinergic receptors (Ref. Reference Burnstock10). The P2-type receptors are subdivided into ionotropic (P2X 1–7) and metabotropic (P2Y 1, 2, 4, 6, 11–14) receptors and exhibit different ranges of responsiveness to ATP, UTP and its diphosphate analogues (Ref. Reference Burnstock10). On the other hand, the P1-type receptors (A1, A2A, A2B and A3) are responsive mainly to ADO (Figure 1 – (6)) (Ref. Reference Burnstock10).

Figure 1. Cellular pathways regulating adenosinergic signalling. (1) The canonical pathway is the main way of extracellular ADO production, in which extracellular ATP is first hydrolyzed by the NTPDase1/CD39 to AMP. The AMP molecules can be further hydrolyzed by the CD73/5′-NT, thereby generating ADO. (2) The non-canonical cascade is an alternative ADO-generating pathway, which allows the production of ADO independent of CD39. (3) First, NAD+ is hydrolyzed by CD38, generating adenosine diphosphate ribose (ADPR), (4) which, in turn, is hydrolyzed by NPP1/CD203a, producing AMP. CD203a can also hydrolyze NAD+ or ATP directly generating AMP which can be hydrolyzed by CD73 to ADO. (5) Extracellular ADO can either be catabolized to inosine (INO) by extracellular adenosine deaminase (ADA), (6) activate the type 1 purinergic (P1) receptors (A1, A2a, A2b and A3) or (7) be transported into cells by equilibrative (ENT1/2) and/or concentrative (CNT1/2) nucleoside transporters. (8) Once in the cytosol, ADO can be degraded to inosine (INO) by cytosolic adenosine deaminase (ADA) or (9) re-phosphorylated to AMP by adenosine kinase (ADK-S in the cytoplasm and ADK-L in the nucleus).

Purinergic signalling is an important mechanism used by cells to control their internal events and interact with the external environment (Ref. Reference Alcedo, Bowser and Snider11). ATP is released from different cell types and is considered a key component of the purinergic system mainly because the sequential removal of its phosphate groups results in the formation of the nucleoside ADO (Ref. Reference Alcedo, Bowser and Snider11). Extracellular ATP and ADO, as well as purinergic enzymes, receptors and transporters, are part of this signalling system, which links cellular metabolism to other cellular processes, including proliferation, differentiation and cell death (Ref. Reference Alcedo, Bowser and Snider11).

The canonical pathway of ADO production involves E-NTPDase 1 (NTPDase1/CD39), which hydrolyzes ATP, producing AMP. Under the action of CD73, AMP is then hydrolyzed to ADO, completing the adenosinergic loop (Figure 1 – (1)) (Refs Reference Galgaro, Beckenkamp and van den M Nunnenkamp12, Reference Chiarella, Ryu and Manji13). The non-canonical pathway is an alternative adenosine production route, independent of CD39 (Figure 1 – (2)). This pathway involves the production of ADP-ribose from NAD+ through CD38 (Figure 1 – (3)), which is then processed to AMP by ENPP1/CD203a (Figure 1 – (4)) (Refs Reference Galgaro, Beckenkamp and van den M Nunnenkamp12, Reference Horenstein, Chillemi and Zaccarello14). Finally, the AMP is metabolized to ADO by CD73, which, in turn, acts as a link between the canonical and non-canonical pathways (Figure 1) (Ref. Reference Galgaro, Beckenkamp and van den M Nunnenkamp12). The extracellular catalytic activity of CD73 can be performed by a membrane-bound form of CD73 or a soluble form derived from the glycosylphosphatidylinositol (GPI)-anchored CD73 by cleavage (Ref. Reference Richart15). Although the main focus of this work is to explore the role of CD73 in cervical cancer, we must also mention that the ability to metabolize AMP is not an exclusiveness of CD73. Other enzymes, such as tissue-specific alkaline phosphatases and tissue-non-specific alkaline phosphatases (TNAPs), can also hydrolyze AMP to ADO (Figure 1) (Ref. Reference Jackson, Cheng and Verrier16).

At the end of the purinergic signalling, ADO can be degraded to inosine by ADA (Figure 1 – (5)) or be uptaken into the intracellular environment through both equilibrative (ENT1/2) and concentrative (CNT1/2) nucleoside transporters (Figure 1 – (7)) (Ref. Reference Zimmermann17). In the intracellular compartment, ADO can be degraded to INO by cytosolic ADA (Figure 1 – (8)) or re-phosphorylated to AMP by adenosine kinase (ADK) (Figure 1 – (9)). The ADK can be found in a cytoplasmic (ADK-S) and a nuclear (ADK-L) isoform (Figure 1 – (9)), but both drive the phosphorylation of ADO to form AMP (Ref. Reference Boison and Yegutkin18). The cytoplasmic short isoform ADK-S provides the main metabolic way of ADO clearance under normal conditions and, consequently, regulates intra- and extracellular levels of ADO by phosphorylating ADO in AMP (Ref. Reference Zhang and Gao19). In contrast, the long isoform ADK-L acts as a regulator of DNA methylation. The activity of ADK-L decreases the concentration of ADO, favouring the S-adenosylmethionine (SAM)-dependent transmethylation pathway, which drives DNA and histone methylation. Thereby, high levels of ADK-L are associated with increased DNA methylation in the nucleus and, consequently, contribute to the regulation of gene expression (Refs Reference Boison and Yegutkin18, Reference Bertoni, Valandro and Brasil20, Reference Jeong, Oh and Choi21). An overview of this signalling cascade is summarized in Figure 1.

Adenosinergic signalling in cancer

The CD73/5′-NT hydrolysis activity leads to ADO production, which is related to an immunosuppressed environment, commonly associated with cancer development and progression (Ref. Reference Allard, Longhi and Robson22). The importance of the CD73/ADO pathway in cancer has been extensively demonstrated (Refs Reference Gao, Dong and Zhang6, Reference Young, Mittal and Stagg23, Reference Chen, Wainwright and Wu24). The ADO generated by the action of the CD73 enzyme can lead to decreasing activity of several immune cell lineages, including CD4 and CD8 T cells, natural killer cells and antigen-presenting cells. On the other side, it promotes the enhancement of the function of Treg cells, MDSCs and tumour-associated macrophages (TAM), further reinforcing immunosuppression (Refs Reference Antonioli, Blandizzi and Pacher25, Reference Häusler, Del Barrio and Diessner26). Furthermore, extracellular ADO regulates the growth and dissemination of cancer by direct actions in tumour cell migration, invasion, metastasis and proliferation (Refs Reference Antonioli, Blandizzi and Pacher25, Reference Sheth, Brito and Mukherjea27). Also, it stimulates the formation of cancer stem cells, which have a direct role in tumour heterogeneity and resistance to therapy (Ref. Reference Albini, Bruno and Gallo28).

In human breast cancer cells, the treatment with adenosine 5′-(α,β-methylene)diphosphate (APCP), a competitive CD73 inhibitor, decreased proliferation rate and cell viability induced by CD73 overexpression (Ref. Reference Zhi, Wang and Zhou29). In glioma, the APCP treatment reduced glioma cell proliferation, while ADO treatment increased it in a similar way (Ref. Reference Bavaresco, Bernardi and Braganhol30). Azambuja et al. showed that CD73 downregulation by small interfering RNA (siRNA) or APCP treatment decreased glioma cell migration, invasion and proliferation in vitro, as well as rat glioblastoma progression in vivo (Refs Reference Azambuja, Gelsleichter and Beckenkamp31, Reference Azambuja, Schuh and Michels32). In agreement, the therapy using an anti-CD73 monoclonal antibody in mice can initiate the adaptive antitumour immunity, leading to inhibition of breast tumour growth and metastasis (Ref. Reference Stagg, Divisekera and McLaughlin33). Similar results were observed in a melanoma model (Ref. Reference Perrot, Michaud and Giraudon-Paoli34). Besides, the non-enzymatic role of CD73 has been recently explored (Refs Reference Gao, Wang and Lin35, Reference Sadej and Skladanowski36), highlighting the adhesive-related properties of this protein. The protein structure can control cell interaction with extracellular matrix (ECM) components, such as laminin and fibronectin. Therefore, it can mediate cancer invasion, migration and metastasis processes (Refs Reference Gao, Dong and Zhang6, Reference Sadej and Skladanowski36, Reference Cappellari, Vasques and Bavaresco37).

In contradiction to evidence showing the inductive impact of CD73 in tumour development and progression, some studies have unveiled the role of the CD73/ADO pathway in inhibiting tumour development in a context-dependent way (Ref. Reference Borea, Gessi and Merighi38). Studies have shown that ADO, through its receptors, can promote tumour cell death (Refs Reference Merighi, Mirandola and Milani39, Reference Dietrich, Figueiró and Filippi-Chiela40), reduce tumour cell proliferation (Refs Reference Hosseinzadeh, Jaafari and Shamsara41–Reference Hajiahmadi, Panjehpour and Aghaei45), inhibit cell invasion and migration (Refs Reference Gessi, Sacchetto and Fogli46–Reference Martínez-Ramírez, Díaz-Muñoz and Battastini48) and enhance chemotherapy sensitivity (Ref. Reference Sureechatchaiyan, Hamacher and Brockmann49) in different tumour types. Cappellari et al. demonstrated in an experimental xenograft model of medulloblastoma that implanted cells overexpressing CD73 formed smaller tumours with reduced vascularization and enhanced apoptosis rates when compared to controls (Ref. Reference Cappellari, Pillat and Souza50). Similar results were observed using agonists of adenosine receptors in animal models of melanoma (Refs Reference Madi, Bar-Yehuda and Barer51, Reference Fishman, Madi and Bar-Yehuda52), colon (Refs Reference Bar-Yehuda, Madi and Silberman53, Reference Ohana, Bar-Yehuda and Arich54), prostate (Ref. Reference Fishman, Bar-Yehuda and Ardon55) and hepatocellular carcinoma (Ref. Reference Bar-Yehuda, Stemmer and Madi56). The A3 receptor agonist has also been used in a clinical trial to treat hepatocellular carcinoma, showing to be safe and well tolerated by patients (Ref. Reference Stemmer, Benjaminov and Medalia57).

Bowser et al. showed that CD73 is markedly downregulated in poorly differentiated and advanced-stage endometrial human carcinoma compared to normal endometrium and low-grade tumours. Using animal model, they found that ADO, via A1 receptor, promotes epithelial integrity protection by promoting cortical actin polymerization, being essential to preventing migration, invasion and metastasis (Ref. Reference Bowser, Blackburn and Shipley58). Downregulation of CD73 in ovarian carcinoma and advanced breast cancer has also been reported (Refs Reference Supernat, Markiewicz and Welnicka-Jaskiewicz59, Reference Oh, Sin and Choi60).

Therefore, it can be observed that adenosinergic signalling in the tumour process is dependent on the target tissue and specific characteristics of tumour cells. In this context, the role of CD73/ADO in cancer is not clear. The results observed in different experimental models are complex puzzles that involve many factors to be still understood.

Cervical cancer

Cervical cancer is one of the most common diseases in women around the world, being the fourth main cause of cancer death in women (Refs Reference Tian, Gong and Gao61, Reference Bray, Ferlay and Soerjomataram62). Histologically, there are different types of cervical carcinoma. The most common type is squamous cell carcinoma (SCC), followed by adenocarcinoma (AC) (Ref. Reference Hu, Wang and Liu63). Several histological and cytological nomenclature systems have been developed (Ref. Reference Schiffman, Doorbar and Wentzensen64). In the late 1980s, the Bethesda Classification of the cervical cytology categorized the precursor lesions in the uterine cervix in low-grade (LSIL) or high-grade squamous intraepithelial lesions (HSIL) (Ref. Reference Bhatla, Berek and Cuello Fredes65). The cervical histology grading was classified by cervical intraepithelial neoplasia (CIN) system, comprising low-grade lesions (CIN1) and high-grade lesions (CIN2/3) (Refs Reference Richart15, Reference Richart66).

HPV infection is caused by a DNA non-enveloped virus from the Papillomaviridae family (Ref. Reference Burley, Roberts and Parish67). HPV is a group of more than 200 related viruses, but approximately 54 of them are able to infect the reproductive tract mucosa and 12 of them are oncogenic types classified as high-risk HPV (Ref. Reference Huh, Joura and Giuliano68). The HPV infection, in particular, the oncogenic types HPV 16 and 18, is detected in the vast majority of cervical cancers (approximately 99% of cervical tumours), being the most important cause of this tumour type (Ref. Reference Walboomers, Jacobs and Manos69). Most HPV infections clear up on their own without treatment, but some of them progress to HSIL and invasive cervical cancer (<1% of HPV infections) (Refs Reference Ho, Bierman and Beardsley70, Reference Monsonego, Cox and Behrens71).

Two oncoproteins known as E6 and E7 are encoded by HPVs, and they are directly responsible for the development of HPV-induced carcinogenesis (Ref. Reference Tomaić72). The activity of E6 and E7 viral genes is present in high- and low-risk types of HPV, but only in high-risk types, their action is sufficient to trigger preneoplastic lesions and cancer, deregulating normal cell cycle, promoting telomerase activity, disrupting immune cells response and providing cellular factors necessary for productive viral replication (Refs Reference de Sanjosé, Brotons and Pavón73, Reference Moody74).

In face of the complexity of cervical carcinogenesis, over the last years, numerous studies have focused in providing an understanding of the pathophysiologic mechanisms of this disease, underlying molecular mechanisms and cellular pathways, such as Wnt/β-catenin, PI3K/AKT, RAF/MEK/ERK and adenosinergic pathway (Refs Reference Zhang and Gao19, Reference Manzo-Merino, Contreras-Paredes and Vázquez-Ulloa75).

The RAF/MEK pathway is involved in the regulation of cancer proliferation, differentiation, angiogenesis and survival. RAS, which is a member of the RAF family, is activated in 20% of human cancers, including cervical cancer (Ref. Reference Yoshida, Kajitani and Satsuka76). The activation of the RAF /MEK/ERK pathway promotes proliferation and invasion of cervical cancer cells (Ref. Reference Li, G-X and Yang77), as well as promotes malignant conversion of HPV-infected keratinocytes, contributing to cancer invasiveness (Ref. Reference Yoshida, Kajitani and Satsuka76). Interestingly, the transformation and carcinogenesis of human keratinocytes expressing HPV also requires the activation of the Wnt pathway (Ref. Reference Uren, Fallen and Yuan78). The Wnt signalling pathway plays critical roles in cell fate, proliferation, migration and tissue homeostasis (Ref. Reference Manzo-Merino, Contreras-Paredes and Vázquez-Ulloa75). In cervical cancer, aberrant Wnt/β-catenin pathway activation promotes proliferation and inhibits apoptosis (Ref. Reference Zhang and Gao19). Moreover, this pathway contributes to cervical cancer cell migration and invasion by regulating epithelial-to-mesenchymal transition (EMT) (Ref. Reference Xue, Yang and Chen79). Another pathway that is often dysregulated in cervical malignancy is PI3K/AKT. This pathway is considered a biomarker for poor prognosis in cervical cancer (Ref. Reference Schwarz, Payton and Rashmi80), being related with metastasis and tumour recurrence (Ref. Reference Hou, Liu and Wheler81). It has been shown that oncoproteins of HPV are associated with the PI3K/AKT pathway, contributing to the tumour initiation, cell proliferation, metastasis, angiogenesis and resistance to therapy (Ref. Reference Bhattacharjee, Das and Biswal82). The signal transduction pathways present a central role in tumour development and progression, being the focus in several studies, including preclinical and clinical approaches (Ref. Reference Manzo-Merino, Contreras-Paredes and Vázquez-Ulloa75). The influence of purinergic signalling in cervical cancer progression has also been explored since this pathway is implicated in several tumoural mechanisms, including immune response, inflammation, cell proliferation, differentiation and cell death (Ref. Reference Pfaffenzeller, Franciosi and Cardoso83).

Adenosinergic pathway in cervical cancer: the example of dual face of CD73/ADO

CD73 is a major enzyme producing ADO from extracellular AMP. Furthermore, the influence of CD73 in tumour growth is not only limited to its enzymatic function but also related to its non-enzymatic action (Ref. Reference Gao, Dong and Zhang6). The protein structure of CD73 can control cell–cell adhesion properties and cell interaction with ECM components, mediating cancer invasion, migration and metastasis (Ref. Reference Gao, Dong and Zhang6).

CD73/ADO plays a role in several physiological and pathological cell processes, including vasodilatation, neurotransmission, tissue homeostasis and acute and chronic inflammation (Ref. Reference Zimmermann84). More recently, CD73 has gained considerable attention as a target for cancer treatment (Ref. Reference Gelsleichter, Azambuja and Rubenich85). CD73 is upregulated in the majority of human solid tumours in comparison to normal tissues. Among these tumours, we can mention glioblastoma (Ref. Reference Gelsleichter, Azambuja and Rubenich85), thyroid carcinoma (Ref. Reference Bertoni, Bracco and de Campos86), pancreatic carcinoma (Ref. Reference Tang, Huang and Lu87), renal cell carcinoma (Ref. Reference Zhou, Jiang and Chu88) and colorectal carcinoma (Ref. Reference Bendell, LoRusso and Overman89). However, in general, in cancers of the genitourinary system, such as endometrial, ovarian, uterine and prostate, a lower expression of CD73 is observed (Ref. Reference Tang, Zhang and Cao90).

In cervical tumours, the role of CD73/ADO is a bit controversial (Ref. Reference Tang, Zhang and Cao90). Our research group has already demonstrated that different cervical cancer cell lines express ectonucleotidase members at different levels and show different hydrolysis patterns of adenine nucleotides (Ref. Reference Beckenkamp, Santana and Bruno91). Whereas SiHa (HPV+) and HeLa (HPV+) cancer cell lines presented similar levels of CD73 expression and AMP hydrolysis, in C33A (HPV-) cells, the expression of this gene and its enzymatic activity were almost undetectable (Ref. Reference Beckenkamp, Santana and Bruno91).

Just like CD73 expression can fluctuate among cell types, the impact of the adenosinergic pathway on cervical cancer development and progression is not completely clear. Studies targeting this issue are presented below.

CD73/ADO axis inhibiting cervical cancer progression

In contrast to the data addressed above, there is evidence showing that, unlike the majority of solid cancer types, an inverse correlation between the adenosinergic axis and cervical tumour development and progression has been reported (Refs Reference Allard, Allard and Buisseret3, Reference Yang, Yao and Davis98–Reference Mello, Filippi-Chiela and Nascimento100). Curiously, the same is observed in other tumours of the genitourinary system, such as ovarian serous cystadenocarcinoma, testicular germ cell tumours, endometrial carcinoma, uterine carcinosarcoma, bladder carcinoma, kidney and prostate adenocarcinoma (Ref. Reference Yang, Yao and Davis98).

Our research group investigated the expression and activity of CD73 in sphere-forming cells from cervical cancer in comparison to monolayer cells in vitro. It is known that low adherent tumourspheres are enriched in stem-like cancer cells (Ref. Reference Fotinós, Marks and Barberis101), which have tumour-initiating ability and play a critical role in tumour metastasis, relapse and chemoresistance (Ref. Reference Tang102). Our results evidenced a reduction in CD73 expression and enzymatic activity in cervical spheres when compared to monolayers (Ref. Reference Iser, de Andrade Mello and Davies99). Interestingly, our in silico analyses have supported our in vitro results, showing that three-dimensional spheres derived from cervical, thyroid and breast cancer presented decreased expression of CD73, when compared to their adherent counterparts (Ref. Reference Iser, de Andrade Mello and Davies99).

Gao et al. showed that ADO treatment inhibited the migration and invasion of Hela and SiHa cells via repressing the EMT program. The authors showed that after treatment of tumour cells with ADO, the epithelial marker E-cadherin was significantly increased, while the mesenchymal markers N-cadherin and fibronectin were decreased. In addition, ADO induced cervical cancer cell apoptosis, as confirmed by analysing the expression levels of apoptosis-related molecules (Ref. Reference Gao, Wang and Dong103). Another study reported that the treatment with ADO analogue decreases the viability of HeLa and SiHa cells through induction of necrosis and apoptosis, respectively (Ref. Reference Sinha, Srivastava and Prusty104).

In accordance, Mello et al. showed that ADO uptake produced from ATP degradation by ectonucleotidases plays a crucial role in inducing apoptosis in the human cervical cancer cell line (Ref. Reference Mello, Filippi-Chiela and Nascimento100). ADO uptake promoted AMPK activation, deoxy-ATP (dATP) accumulation and increased p53 level, culminating in induction of apoptosis and autophagy (Ref. Reference Mello, Filippi-Chiela and Nascimento100). The authors confirmed the results by treatment of cervical tumour cells with dipyridamole (an adenosine transporter inhibitor), which induced almost complete recovery of cell viability (Ref. Reference Mello, Filippi-Chiela and Nascimento100). Indeed, in the same work, the authors observed that when SiHa cells were treated with different ADO concentrations (from 0.1 mM to 5 mM), only the higher concentrations of ADO (2 mM and 5 mM) reduced cell viability (Ref. Reference Mello, Filippi-Chiela and Nascimento100). In the same way, Monroy-Mora showed that high ADO levels in the TME of cervical cancer cells, caused by ADA inhibition, decreased the viability of cell lines by inducing apoptosis. When cells were treated with ADO at a concentration of 1000 μM in the presence of EHNA, a specific ADA inhibitor, the viability of C33A, CaSki and HeLa cells decreased by 60%, 83% and 96%, respectively. On the other hand, the presence of ADA conferred protection to cervical tumour cells against the cytotoxic effect of high ADO content (Ref. Reference Monroy-Mora, de Lourdes Mora-García and Alheli Monroy Mora105). These results reinforce the idea that nucleoside concentration may be critical to the effects produced on cervical tumour cells (Table 1).

Table 1. Studies using different conditions of treatment and concentrations of adenosine (ADO) in cervical cancer cells

Analysis of modulation of CD73 and other adenosinergic pathway components in cervical cancer

In spite of the considerable number of studies linking the roles of CD73/ADO to the development process or prognosis of cervical cancer, as presented in this review, the results obtained by different authors are even contradictory. In this context, comprehensive bioinformatics methods have made possible to find an association among other genes and cellular pathways. Given that, to clarify how the expression of CD73 is in cervical cancer and how this enzyme is related to cancer progression, we conducted bioinformatics analysis using various transcriptomic databases. Additional materials and methods information is provided in Supplementary Material.

Expression of CD73 in cervical cancer cell lines

First of all, we decided to investigate the expression of CD73 in different cervical cancer cell lines from the GSE9750 dataset, obtained from the Gene Expression Omnibus (GEO) database. In Figure 2, it is possible to observe that the expression of CD73 diverges widely among the different cells analysed. The C-33A lineage showed lower expression of CD73, while SW756 presented higher expression levels. SiHa and HeLa cells, which are the most common cell lines used in studies about cervical cancer, presented a very similar level of CD73 expression. In accordance, the enzymatic activity for AMP hydrolysis evaluated on the cell surface of SiHa, HeLa and C-33A cells showed that C-33A has smaller degradation rate levels compared with SiHa and HeLa that are very similar (Ref. Reference Beckenkamp, Santana and Bruno91). The wide range of variation in the expression levels of CD73 found in this analysis can explain, at least in part, the contrasting results observed in studies using cancer cell lines, both in vitro and in vivo.

Figure 2. Profile of CD73 (NT5E) expression on cervical cancer cell lines from the GSE9750 dataset.

It is important to highlight that the cell lines whose CD73 expression is lower (C-33A and HT-3) in our analysis are negative for HPV DNA or RNA. In contrast, the cells with the highest CD73 expression are positive for HPV (Refs Reference Sykes, Whitescarver and Jernstrom106–Reference Seedorf, Oltersdorf and Krämmer108). This observation comes together with the results obtained by Mora-García et al., which showed that cells positive for HPV-16 obtained from cervical samples of CIN1 patients presented higher CD73 expression when compared to HPV-16 negative samples (Ref. Reference de Lourdes Mora-García, López-Cisneros and Gutiérrez-Serrano109). The same result was observed when comparing the expression of CD73 in cervical carcinoma cell lines positive or negative for HPV (Ref. Reference Mora-García, Ávila-Ibarra and García-Rocha93). All together, these findings could indicate that HPV infection is able to modulate CD73 expression in cervical cancer.

Expression of CD73 in tumour tissue

Our next step was identifying whether there are differences in the expression of CD73 in human tissues of cervical cancer in comparison to non-tumour cervix tissue. For that, we analysed ten microarray datasets (GSE29570, GSE39001, GSE67522, GSE7410, GSE9750, GSE7803, GSE52903, GSE63514, GSE6791 and GSE27678) obtained from the GEO database and The Cancer Genome Atlas Cervical Squamous Cell Carcinoma and Endocervical Adenocarcinoma (TCGA-CESC) database (Supplementary Table 1). Our analysis revealed a significant downregulation of CD73 in samples of human cervical cancer tissue when compared to non-tumour tissue in the majority (8/11; 72.7%) of cohorts analysed (Figure 3). Among them, two datasets (GSE6791 and GSE63514) did not present differences between cancer tissue and non-tumour tissue, and one of them presented upregulation of CD73 (GSE27678) (Figure 3).

Figure 3. Expression of CD73 (NT5E) in cervical samples and non-tumour tissues. Relative expression of CD73 mRNA in cervical cancer and non-tumour tissues is shown in scatter dot-plots from GSE29750 (A), GSE39001 (B), GSE67522 (C), GSE7410 (D), GSE9750 (E), GSE7803 (F), GSE52903 (G), GSE63514 (H), GSE6791 (I), GSE27678 (J) and TCGA-CESC (K) datasets. Data presented as mean ± standard deviation and statistical significances (P values) between groups were determined by the Mann–Whitney U-test or two-tailed Student’s t-test.

CD73 expression × FIGO staging system for cervical cancer

The FIGO (International Federation of Gynecology and Obstetrics) staging system is used for staging classification of carcinoma of the cervix uteri. The cervical cancer stage ranges from stages I (1) through IV (4) (Ref. Reference Bhatla, Berek and Cuello Fredes65). Thus, next, we evaluated the expression of CD73 in higher FIGO stages compared to lower FIGO stages. The presence or absence of lymph node metastasis (LNM) is essential to the FIGO statement (Ref. Reference Bhatla, Berek and Cuello Fredes65) since it facilitates prognosis evaluation. Therefore, in the present study, we included this variable in our analysis. The data analysed here were obtained from GEO databases and the TCGA-CESC database. Interesting, there was no correlation between the CD73 expression and FIGO staging (Supplementary Figure 1 and Supplementary Table 2). Similarly, there was no difference comparing the expression of CD73 in samples with the presence or absence of LNM (Supplementary Figure 2 and Supplementary Table 3).

In sequence, we used the GSE7803 dataset, with samples of high-grade squamous intraepithelial lesions (HSIL) of the cervix and invasive squamous cell carcinomas (SCC) to investigate whether CD73 expression could be linked to cervical cancer progression. We observed a significant downregulation of CD73 in SCC compared with normal squamous cervical epithelium (NC) or with HSIL (Figure 4), reinforcing the idea of a negative correlation between CD73 expression and tumour development.

Figure 4. Differential CD73 (NT5E) expression in normal cervix, preinvasive and invasive cervical lesions. Each dot in the scatter plot represents the individual sample from GSE7308 stratified as normal cervix (NC; n = 10), high-grade squamous intraepithelial lesions (HSIL; n = 7) and invasive squamous cell carcinomas of the cervix (NSIL; n = 21). Data presented as mean ± standard deviation and statistical significance (P values) between groups were determined by two-tailed Student’s t-test.

CD73 expression and survival rate

In our conception, the reduced expression of CD73 in cervical cancer may be associated with cancer development or with the transition from a healthy cell to a tumour cell, passing through different degrees of lesions, without necessarily impacting the outcome of disease. Survival curves reinforce this idea indicating that CD73 expression did not significantly affect disease-free survival (HR = 1.2, p(HR) = 0.58) or overall survival (HR = 1.6, p(HR) = 0.071) in patients from the TCGA-CESC database when samples were categorized into two groups of low (n = 146) and high (n = 146) CD73 expression, based on the median level of CD73 expression (Figure 5A and B, respectively).

Figure 5. Estimated overall survival curve for CD73 (NT5E). (A) Kaplan–Meier curves of disease-free survival and (B) overall survival of patients with cervical cancer based on CD73 mRNA expression (low (n = 146) versus high (n = 146) transcripts of per million (TPM); hazard ratio (HR) = 1.6). The solid lines indicate the survival curve, and the dotted lines indicate the 95% confidence interval. Overall survival analyses were performed using the GEPIA online platform (http://gepia.cancer-pku.cn/). (C) Forest plot of the multivariate Cox regression analysis showing the risk score for clinical factors and NT5E expression.

The multivariate Cox regression analysis results showed that clinical stage, age, histological grade, CD73 expression and methylation sites were not prognostic factors, not influencing the patient’s risk of death (Figure 5C).

To evaluate whether CD73 expression could be related to other outcome factors, we performed univariate analysis, and the results showed that CD73 expression was not associated with age ≥ 50 years old (p = 0.1137), keratinizing versus non-keratinizing squamous cell carcinoma (p = 0.1421), presence of regional lymph node (p = 0.856), distant metastases (p = 0.681) or between histological grades G1/G2 versus G3/G4 (p = 0.3489) (Supplementary Table 4). Although clinical stages I/II versus III/IV and tumour stages (T1, T2 and T3/T4) have not shown statistical differences (p = 0.1038 and p = 0.122, respectively), we can observe a reduction in CD73 expression associated with disease progression.

It is essential to highlight a limitation of this analysis since they were performed using only one dataset (TCGA-CESC). The TCGA cervical cohort is composed mainly of patients with FIGO stage I disease, being not comparable to patients with advanced cervical cancer. Thus, for a more accurate result, it would be necessary to evaluate samples from patients with more advanced tumour stages. In addition, as shown in Figure 5, we considered confounding factors such as clinical stage, patient age and tumour stage/grade, but not others like treatment and HPV type, which might lead to potential bias. Other datasets either have none or very few cervical cancer-related data or often lack critical clinical information about survival time. Additionally, the analyses conducted are exclusively in silico. While valuable for initial insights, they require further validation through experimental studies using fresh tumour samples. Future research should focus on addressing these gaps to provide a more thorough validation of the results presented.

The modulation of other adenosinergic components in the cascade

It is essential to keep in mind that the purinergic system accounts for a complex network of enzymes and receptors responsible for the generation, recognition and degradation of extracellular ATP and ADO. However, although we are focusing our attention on CD73, other factors contribute to regulating extracellular ADO levels. For example, ADK is an intracellular enzyme able to metabolize intracellular ADO. Thereby, the modulation of ADK activity by different factors can interfere with extracellular levels of ADO. If ADK is inhibited, there is an increase in intracellular concentrations of ADO, favouring the release of ADO to extracellular space. Moreover, the downregulation of nucleoside transporters can enhance the concentrations of extracellular ADO (Ref. Reference Allard, Allard and Buisseret3).

ADO can also not bind to its receptors to exert their effects but instead be metabolized to inosine by secreted ADA (ecto-ADA) or yet be transported into cells by equilibrative and concentrative transporters (ENTs and CNTs), which can also mediate nucleoside efflux (Ref. Reference Allard, Allard and Buisseret3). All of these alternatives, taken together, can regulate ADO metabolism. Therefore, we decided to analyse other components that could regulate ADO concentrations.

In our in silico analysis, we observed an imbalance of adenosinergic pathway components (Figure 6). Among nine datasets analysed from GEO, in six of them (66.7%) (GSE67522, GSE52903, GSE39001, GSE29570, GSE9750 and GSE63514), the tumour samples presented enhanced expression of at least one member of CNT (SLC28A1, A2 and A3 genes) and ENT protein families (SLC29A1, A2, A3 and A4 genes) when compared to normal tissue samples (Figure 6). These transporters can transfer ADO into cells, reducing the levels of this nucleoside in extracellular space. Additionally, all datasets analysed presented a reduction of the surface glycoprotein dipeptidyl peptidase 4 (DPP4/CD26) expression, and eight of them (88.9%) presented enhanced ADA expression (Figure 6).

Figure 6. Gene expression modulation of adenosinergic-related components in normal cervix and cervical cancer. The boxplots show the expression levels of CD73 (NT5E), concentrative nucleoside transporters (CNTs: SLC28A1, SLC28A1 and SLC28A1), equilibrative nucleoside transporters (ENTs: SLC29A1, SLC29A2, SLC29A3 and SLC29A4), adenosine deaminase (ADA), the surface glycoprotein CD26 (DPP4), tissue-nonspecific alkaline phosphatase (TNAP) and adenosine receptors (ADORA1, ADORA2A, ADORA2B and ADORA3) genes. The relative expressions were collected from the GEO database through microarray data using (A) GSE67522, (B) GSE52903, (C) GSE39001, (D) GSE29570, (E) GSE6791, (F) GSE7803, (G) GSE7410, (H) GSE9750 and (I) GSE63514 datasets. Data were expressed as mean, and the upper and lower whiskers represent the maximum and minimum values of gene expression. The P value was considered non-significant (ns) if equal to or less than 0.050 and statistically significant when *P = 0.049–0.01, **P = 0.01–0.001, ***P < 0.0001, ****P < 0.00001.

ADA catabolizes ADO to inosine, thus downregulating the biologic effects of extracellular ADO. However, ADA does not have its own transmembrane domain and is associated with CD26. CD26 works as a binding protein for extracellular ADA, anchoring it to the cell membrane (Ref. Reference Mandapathil, Szczepanski and Harasymczuk110). Interestingly, CD26 also binds ECM proteins such as fibronectin and collagen. Consequently, the decreased expression of CD26 has been linked to increases in tumour invasion and metastasis (Refs Reference Tan, Mujoomdar and Blay111, Reference Beckenkamp, Davies and Willig112). Additionally, we observed that the expression of TNAP, a membrane-bound phosphatase, is decreased in tumour samples of 55.5% of datasets analysed (Figure 6). This result reinforces that in addition to CD73, other enzymes able to hydrolyze AMP to ADO, such as TNAP, are also downregulated in cervical cancer.

CD73 is epigenetically modulated in cervical cancer

DNA methylation is a common epigenetic signalling mechanism that cells use to modulate gene expression. Therefore, it is an important component in numerous cellular processes, including many cancer types. DNA methylation is related to suppression of gene expression by blocking the promoters (Ref. Reference Greenberg and Bourc’his113). However, it is also associated with gene bodies or repeated sequences and, in some cases, with gene activation or heterogeneous gene expression (Refs Reference Greenberg and Bourc’his113, Reference Carter and Zhao114). Mechanistically, the downregulation of NT5E mRNA levels was shown to be modulated by hypermethylation in the cytosine-phosphate-guanine (CpG) island located in the regulatory region of the CD73 gene in breast cancer (Ref. Reference Lo Nigro, Monteverde and Lee115) and in both primary and metastatic melanomas (Ref. Reference Wang, Lee and Nigro116).

Using an artificial intelligence tool, our research group successfully demonstrated the potential of combining mRNA levels of NT5E with its DNA methylation levels as a differentiating factor between thyroid cancer and normal thyroid tissues (Ref. Reference Bertoni, Valandro and Brasil20). Recent studies have shown genes that might be downregulated due to hypermethylation in cervical cancer samples in comparison with normal cervical samples (Refs Reference Liu, Wang and Li117, Reference Xu, Xu and Wang118). Interestingly, one of these genes is the DPP4 gene, which is downregulated in cervical cancer samples, as can be observed in our in silico analysis (Figure 6).

Considering these studies, we hypothesized that the CD73 downregulation in cervical samples could be a result of abnormal methylation in distinct genomic regions of the NT5E gene, which encodes CD73. As shown in the schematic representation of the CD73 structure in Figure 7A and Supplementary Table 5, we collected 17 probes from publicly available datasets that covered the transcription start site (TSS) in the promoter region, first exon, body, and three prime untranslated regions (3′-UTR) of CD73.

Figure 7. Correlation between CD73 (NT5E) mRNA expression and methylation levels from TCGA-CESC. (A) Schematic presentation of cytosine-phosphate-guanine (CpG) site distribution in the CD73 promoter, 1st exon and body. (B) Scatter plot showing the Pearson’s correlation coefficient (r) between methylation levels (Y-axis) and relative expression of CD73 (X-axis) in samples of cervical cancer from TCGA. The solid lines indicate a linear fit, and the dotted lines indicate the 95% confidence interval for the correlation.

First, the methylation analysis and gene expression data from TCGA-CESC revealed a significant inverse correlation between NT5E mRNA expression and methylation levels of 10 out of 15 potential methylation sites (p < 0.0001, n = 306; Figure 7B and Supplementary Table 5), with most strong correlation within the promoter region ((2) cg27039625: ρ = −0.453; (4) cg13315970: ρ = −0.544; (5) cg21730993: ρ = −0.532; (6) cg10663055: ρ = −0.529) and another one site in the CD73 (NT5E) body ((11) cg27297263: ρ = −0.569).

Next, we compared the CpG methylation status of normal and cervical cancer in methylation array in the TCGA-CESC, GSE20080, GSE30760, GSE36637, GSE99511, GSE37020, GSE41384, GSE134772 and GSE46306 datasets separately considering its β-values processed. Our results from TCGA-CESC, GSE46306 and GSE134772 datasets indicate that CpG sites within on the NT5E promoter were significantly hypermethylated in cervical cancer samples in comparison to standard cervical samples (adjusted p-value < 0.05) (Figure 8). There is no significant difference between cancer and standard cervical samples in the GSE99511 dataset. The probes targeting the CpG sites with the NT5E promoter were not covered on the HumanMethylation27K array used for analyses of the GSE20080, GSE30760, GSE36637, GSE37020 and GSE41384 datasets.

Figure 8. DNA methylation profile in normal cervix and cervical cancer. Bars show the DNA methylation β-values (Y-axis) from (A) GSE46306, (B) GSE134772, (C) GSE99511 datasets and from (D) The Cancer Genome Atlas (TCGA) cervical squamous cell carcinoma and endocervical adenocarcinoma database. Data represent mean ± standard deviation. Adeno: adenocarcinoma; CIN: cervical intraepithelial neoplasia; CIN3+: CIN grade 3; TSS: transcription start sites; TSS200: 0–200 nucleotides upstream of the TSS; TSS1500: 200–1500 nucleotides upstream of the TSS; 3′-UTR: three prime untranslated region; *P = 0.05–0.01, **P = 0.01–0.001, ***P < 0.0001.

cg10663055, cg21730993, cg27039625, cg13315970 and cg27297263 CpGs showed a significant relation with CD73 expression (p ≤ 0.05 or Pearson’s ≥ 0.4). We used these sites to generate a linear regression model for predicting the impact of methylation on the levels of CD73 expression. As a result, the coefficient of correlation (R) was 0.558, and the adjusted R2 was 0.312, indicating that about 31.2% of the decreased expression of CD73 could be explained by the hypermethylation of CD73 in these sets of five CpGs.

ADK controls the clearance of intracellular ADO by its conversion to AMP, using ATP as a phosphate donor. Studies have shown that a higher expression of ADK-L and, consequently, low levels of ADO lead to an increase in global DNA methylation (Ref. Reference Oh, Sin and Choi60). Conversely, low levels of ADK-L and consequently high levels of ADO prevent DNA methylation and lead to global DNA hypomethylation. Thus, we analysed if there is a relationship between levels of NT5E and ADK from the TCGA-CESC cohort. As a result, the levels of NT5E exhibit a significant and weak negative correlation with the levels of ADK (r = −0.1364; p = 0.0166; n = 308) (Supplementary Figure 3A). We also noticed that ADK expression was higher in cervical cancer than in normal cervix samples (0.4845 ± 0.0369, n = 305 vs −1.257 ± 0.142, n = 3; p < 0.0001) (Supplementary Figure 3B). Further studies could better investigate this question considering the expression of both isoforms of ADK, nuclear (ADK-L; long isoform) and cytoplasmic (ADK-S; short isoform), since ADK-S is involved mainly in the purine salvage pathway, while ADK-L drives the intranuclear methylation reactions (Ref. Reference Boison and Yegutkin18).

In other cancer types, such as breast cancer and melanoma, the relation between higher gene methylation and downregulation of the NT5E gene has already been shown (Refs Reference Jeong, Oh and Choi21, Reference Wang, Lee and Nigro116). In breast cancer, the methylation level of the NT5E gene was significantly higher when compared with normal breast tissues and associated with poor prognostic factors, such as large tumour size, high histologic grade, negative oestrogen receptor expression and negative Bcl-2 expression (Ref. Reference Jeong, Oh and Choi21). In the same way, the NT5E gene was downregulated by methylation in melanoma cell lines and patient melanoma samples (Ref. Reference Wang, Lee and Nigro116).

Clinical applications

Cervical cancer is preventable through regular screening strategies, such as Pap test and colposcopy (Ref. Reference Olusola, Banerjee and Philley119). These tests in combination can detect very early stages of disease. In addition, prophylactic HPV vaccinations raise as the best choice to prevent cervical cancer (Ref. Reference Olusola, Banerjee and Philley119). Although the use of these screening and prophylactic strategies has decreased the incidence of cervical cancer, significant occurrences of this cancer still remain in certain groups, especially Hispanic/Latina and African American women (Ref. Reference Olusola, Banerjee and Philley119). Interestingly, Hispanic/Latina women have the highest incidence of cervical cancer, with the worst outcome and higher mortality rate compared to other populations (Ref. Reference Olusola, Banerjee and Philley119). Thus, the development of new treatment strategies for these patients is imperative to improve cervical cancer survival rates.

In the early stages of cervical cancer, surgery and/or radiotherapy are the treatment choice. In most aggressive cases, the standard of care involves chemotherapy and radiotherapy in combination. Unfortunately, disease recurrence can reach 30% in the early stages and increase up to 70% in the locally advanced disease (Ref. Reference Valdivia, Grau-Béjar and García-Durán120). Luckily, in recent years, new therapeutic strategies have been studied, including antiangiogenic agents, immunomodulatory vaccination and immunotherapy with monoclonal antibodies (Ref. Reference Valdivia, Grau-Béjar and García-Durán120). Several clinical trials have already been completed or are ongoing to evaluate with the use of anti-CD73 monoclonal antibody (mAb) for different types of tumours and cancers, such as MEDI9447 (oleclumab), AB680 (quemliclustat), TJ004309 (uliledlimab), JAB-BX102, PT199, AK119, BMS-986179, IBI325 and IPH5301 (for details of clinical trials, see Supplementary Table 6). However, only two of them are available to cervical cancer patients (NCT03454451: CPI-006 (mupadolimab) and NCT04672434: Sym024 – https://www.clinicaltrials.gov Accessed on 08/03/2024).

A clinical study involving cervical cancer patients treated with mupadolimab has been concluded. Although the study does not specifically mention patients with cervical cancer, it included 34 patients with various types of advanced cancers, who received intravenous doses ranging from 1 to 24 mg/kg every 21 days. Immunological analysis of patient blood samples revealed a slight decline in circulating CD73-positive B cells immediately after infusion, with levels returning to baseline by day 21. However, the researchers observed that this reduction was not due to blockade by the administered antibody since reactivity with another non-blocking anti-CD73 antibody, AD2, was similarly reduced. Interestingly, patients with non-small cell lung, head and neck and prostate cancers exhibited the most significant changes in circulating B cells and were more likely to experience tumour volume reduction. This observation suggests that mupadolimab may serve as an effective immunotherapy for cancer, owing to its capacity to activate B cells, generate memory B cells and stimulate the production of antigen-specific antibodies. Moreover, the treatment was generally well tolerated, with only a few patients experiencing grade 3 adverse events. This phase 1 trial is now evaluating mupadolimab in combination with adenosine A2A receptor blockade, in combination with pembrolizumab, and in a triplet regimen (Ref. Reference Miller, Luke and Hu121).

Although the clinical study with Sym024 is not yet complete, the preliminary results presented at the 2024 Cancer Research Annual Meeting are promising. The authors reported minimal adverse effects among the 43 patients treated as from October 12, 2023. Target engagement assessed both peripherally (free soluble CD73) and within tumours (enzyme activity), demonstrating a sustained reduction in free soluble CD73 in the blood at dose levels of ≥1500 mg. Additionally, dose-dependent CD73 modulation was observed, with over 80% inhibition achieved in tumour biopsies at doses of ≥1500 mg, suggesting that Sym024 may effectively prevent adenosine-mediated tumour evasion and support further clinical investigation (Ref. Reference Spreafico, Ahnert and Sommerhalder122).

However, further efforts in basic and translational research are still needed to reach significant advances in anti-CD73 mAb-based immunotherapies, particularly for cervical cancer. This is important because clinical studies with oleclumab, one of the most extensively evaluated anti-CD73 mAbs to date, have yielded conflicting results. In a phase I study, oleclumab demonstrated a manageable safety profile, with minimal treatment-related adverse events. Moreover, antitumour activity was observed in cancer types that are generally resistant to immunotherapy (Ref. Reference Bendell, LoRusso and Overman123). On the other hand, in a Japanese study with advanced solid tumours, although the treatment with oleclumab was well tolerated, no disease control was achieved after 8 weeks and all six patients developed progressive disease (Ref. Reference Kondo, Iwasa and Koyama124). Subsequent studies, such as the COAST and SYNERGY trials, indicated that the combination of oleclumab with durvalumab may enhance objective response rates and prolong progression-free survival compared to durvalumab alone, though without a significant increase in clinical benefit rate in certain scenarios, such as in triple-negative breast cancer (Refs Reference Herbst, Majem and Barlesi125–Reference Buisseret, Loirat and Aftimos127). Additionally, the Neo-CheckRay trial suggested the safety of combining oleclumab with stereotactic body radiation therapy and chemotherapy, warranting further investigation in future studies (Ref. Reference Caluwe, Romano and Poortmans128).