It was a collaborative work between University of Foggia-ITALY and the HP2 lab at Grenoble-Alpes University-FRANCE.

Written informed consent was obtained from all subjects and an ethical approval was obtained from two institutional review board (Ethics Committee, Policlinico Riuniti of Foggia, Italy and Comité de Protection des Personnes Sud-Est V, Grenoble, France). Protocols conformed to the principles of the Declaration of Helsinki.

A total of 78 blood samples from patients and healthy controls were evaluated: (i) healthy controls (n=13); (ii) cancer patients (n=26); (iii) OSA patients (n=26); (iv) patients with OSA and cancer (ONCO-OSA, n=13).

Participants’ baseline characteristics are reported in Table 1.

Participants were matched for age and BMI. Human fasted sera were collected and stored at −80°C until use.

Main data collected were as follows: comorbidities (hypertension, cardiovascular disease, diabetes, chronic obstructive pulmonary disease, asthma) and sleep studies data including apnoea–hypopnoea index (AHI), mean nocturnal SpO2, time with SpO2 lower than 90% (TS90) and oxygen desaturation index (ODI). OSA was defined as AHI>5.

Almost all ONCO-OSA and cancer patients had colorectal cancer (22 of 39 patients). Others exhibited ovarian cancer (n=4), breast cancer (n=4), lung cancer (n=2), gallbladder cancer (n=2), pancreatic cancer (n=4) or prostate cancer (n=1). Supplementary Table S1 depicts types of cancer affecting cancer and ONCO-OSA patients.

Human colorectal adenocarcinoma cell lines (CaCo2) were purchased from American Type Culture Collection (ATCC). CaCo2 were maintained in Dulbecco's modified Eagle's medium (DMEM) 4500mg/l glucose (Gibco™ – Sigma–Aldrich, USA) supplemented with 10% fetal bovine serum (Sigma–Aldrich, USA), 100U/ml penicillin and 100μg/ml streptomycin (Gibco™ – Sigma–Aldrich, USA) at 37°C in a humidified atmosphere containing 5% CO2 and submitted to different duration of N or IH exposure.

An MTT assay was used to determine the cytotoxic effect of acriflavine (ACF, Gibco™ – Sigma–Aldrich, USA) on CaCo2 cell lines and 0.5μM ACF was considered optimal (data not shown). CaCo2 cells (1.5×105) were cultured in 6 well semipermeable plates pre-coated with type I collagen in complete medium for 24h. Just before exposing the cells to N/IH, the culture medium was replaced with DMEM supplemented with ACF 0.5μM to avoid toxicity.

CaCo2 cells (1.5×105) were seeded in 6 well semipermeable plates (Zell-Kontakt Imaging FC plates, Germany) pre-coated with type I collagen in complete culture medium. Just before exposing the cells to N/IH, the culture medium was replaced with DMEM supplemented with 1% FBS, 100U/ml penicillin and 100μg/ml streptomycin.

Cultured CaCo2 cells were exposed to intermittent hypoxia (IH) as described before,19 alternating cycles of 5min of normoxia (16% PO2) and 5min of hypoxia (2% PO2) using a custom-made plate holder connected to two gas blenders (Gas Blender 100, MCQ Instruments, Rome, Italy) and located inside a standard cell culture incubator (SANYO, MCO-15AC). Control cells were exposed to constant normoxia (N, 16% PO2). PCO2 was maintained at 5% throughout exposure. The cells were subjected to this intermittent hypoxia system for up to 8h. A long exposure of 32h was also performed, consisting of 8h of IH followed by 16h of normoxia and another 8h of IH.

See Supplementary methods.20-22

See Supplementary methods.

CaCo2 cells were transfected with commercially available miRNA inhibitors for miR-21, miR-23b, miR-26a and miR-210 (anti-miR™ miRNA inhibitor: hsa-miR-21-5p, hsa-miR-23b-5p, hsa-miR-26a-5p and hsa-miR-210-3p – Invitrogen™) using Lipofectamine™ 2000 Reagent (Invitrogen™, Carlsbad, CA, USA) according to manufacturer's instructions.

RNA extraction and functional assays were performed 24h post-transfection. See Supplementary methods for more details.

Transfected CaCo2 cells (5×103) were seeded in 96-well semipermeable plates pre-coated with type I collagen and submitted to 32h of N/IH exposure. Cell viability was then assessed using MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) staining as described in Supplementary methods.

See Supplementary methods.

Following miRNA inhibition, total RNA was extracted as described above and reverse-transcribed by using iScript™ Reverse Transcription Supermix (Bio-Rad, Hercules, CA, USA), according to the manufacturer's protocol.

Then, the expression of HIF-1α gene was evaluated by qRT-PCR as described in Supplementary methods.

Nuclear extract lysates were obtained from CaCo2 cells by using a Nuclear Extraction Kit (ab221978, Abcam). The protein concentrations of nuclear fractions were determined by Bradford assay using CLARIOstar® Plus plate reader (BMG LABTECH, Germany).

The activities of HIF-1α in nuclear extract lysates were detected using the HIF-1α Transcription Factor Assay Kit (ab133104, Abcam) according to the manufacturer's protocol.

Comparisons between groups were performed by ANOVA or Kruskal–Wallis test depending on whether the data were normally distributed or not. The data are presented as mean±standard deviations (SD) or median±range, depending on the normality of values.

Results were considered significant when p values were ≤0.05.

All the statistical analyses were performed using GraphPad Prism software (version 9.0, GraphPad Software).

Spearman's correlation was used to assess relationships between miRNAs expression levels and main nocturnal respiratory data in OSA and ONCO-OSA patients. Cluster analysis was performed on miRNAs’ expression value in order to identify their relationships. A p-value below 0.05 has been considered statistically significant. GraphPad Software (version 9.0, GraphPad Software) and Orange (version 3.0, University of Ljubljana, Slovenia) were used for the analysis.

The expression of miR-21, miR-26a and miR-210 was significantly higher in OSA and in ONCO-OSA patients compared to control and cancer non OSA patients (Fig. 1). miR-23b instead, was significantly lower in cancer patients than in controls, and higher in ONCO-OSA than in cancer and control patients.

Fig. 1.

miRNA expression in patients. (A) miRNA's expression in patients. Quantitative real-time PCR analysis of differentially expressed microRNAs in healthy subjects compared to patients with cancer, OSA and both cancer and OSA (ONCO-OSA). RNU-6B was used as endogenous control. Comparisons between groups were performed by Kruskal–Wallis test. Individual data and median are plotted. *p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001. (B) Clustering of miRNAs. The results from the four patients’ groups were considered together for this analysis. The strength of correlation is inversely indicated by black line's length. Ward's method was applied in cluster analysis. Height ratio: 25% is shown.

(0.17MB).

miR-27b and miR-145 were significantly increased in both cancer and OSA groups compared to controls, but expression was reduced in ONCO-OSA patients compared to cancer patients (Fig. 1A).

Cluster analysis showed that miR-21, miR-23b, miR-26a and miR-210 had similar behavior (Fig. 1B).

The correlation between the level of miRNAs and indices of hypoxia severity are reported in Supplementary Table S3. SpO2 correlated negatively with miR-21, miR-23, miR-26a and miR-210, and positively with miR-27b and miR-145. AHI and ODI correlated positively with miR-21, miR-23, miR-26a and miR-210, and negatively with miR-145. Finally, TS90 correlated positively with miR-21 and miR-23.

Considering that the majority of our patients (23/39) had colorectal cancer, we subdivided them further considering only oncology patients with CRC. Supplementary Fig. S1A shows results considering only CRC patients. In addition, we assessed the difference of miRNA expression in CRC patients versus all other cancer types (Supplementary Fig. S1B).

Considering only CRC patients, the total number was smaller, but the results are more or less the same as using all patients (Supplementary Fig. S1A). However, making a comparison between CRC versus other types of cancer in cancer and ONCO-OSA patients, no significant differences were found, except for miR-21 which showed a difference in the cancer group between CRC and others (Supplementary Fig. S1B). This suggests that the miRNA variations are globally robustly found in all patients, and are not depending on the cancer type.

MiRNAs expression was also tested in colorectal cancer cell line (CaCo2) subjected to normoxia versus intermittent hypoxia for 2, 4, 8 and 32h.

The expression of miR-23b, miR-26a and miR-210 at 2, 4 and 32h was highest after IH exposure (Fig. 2). MiR-21 increased in cells exposed to IH for 4 and 32h compared to normoxia. MiR-27b and miR-145 showed no significant variation, whatever the exposure time tested (Fig. 2).

Fig. 2.

Cellular expression of different microRNAs in normoxia and IH. Quantitative real-time PCR analysis of differentially expressed microRNAs in CaCo2 cells in normoxia vs intermittent hypoxia. Comparisons between groups were performed by two-way ANOVA. Individual data and median are plotted (n=4). *p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001.

(0.13MB).

We tested the impact of IH on cell proliferation and migration. We observed that after 32–48h, IH stimulated both cell proliferation and migration (Fig. 3A–C).

Fig. 3.

Proliferation and migration test in CaCo2 cells both in normoxia and IH after miRNA inhibition. (A) Proliferation test. CaCo2 cells were transfected with various miRNA inhibitor and then incubated in normoxia or IH for 32h. Cells were then assayed for viability by the MTT assay. Cells in normoxia transfected with negative control represented 100% of viability. Individual data and median are plotted (n=3). *p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001. (B) Transwell migration assay. Transwell experiments were performed to analyze the cell migration and invasion in CaCo2 cancer cells exposed to normoxia or intermittent hypoxia for 48h and transfected with miRNA inhibitors. Individual data and median are plotted (n=3). *p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001. (C) Pictures of the lower side of transwells after migration test and crystal violet staining, taken under a light microscope coupled to a camera at ×10 magnification.

(0.31MB).

To investigate the functional role of miRNAs, we used transient transfection of antisense inhibitors of miR-21, miR-23b, miR-26a and miR-210, to silence their expression in CaCo2 cells. We confirmed that all these four miRNA expressions can be significantly suppressed by their respective inhibitors (Supplementary Fig. S2). Inhibition of each microRNA reduced proliferation (Fig. 3A) and migration (Fig. 3B and C) compared to IH negative control (p≤0.0001) and to normoxia.

Then, we aimed to evaluate the links between IH, HIF-1 and miRNA. First, we observed that HIF-1α activity was increased by up to 9-fold after 4h of IH exposure (p≤0.0001). As expected, this HIF-1 activity was abolished after treatment with 0.5μM ACF (p≤0.0001) (Fig. 4A).

Fig. 4.

IH and miRNAs regulate HIF-1. (A) Treatment with ACF abolished IH-induced HIF-1α activity. Measurement of HIF-1α activity in CaCo2 cells exposed to normoxia or intermittent hypoxia for 2 and 4h, in presence or absence of 0.5μM ACF. Individual data and median are plotted (n=3). ****p≤0.0001. (B) MiRNA inhibition suppresses HIF-1α gene expression. qRT-PCR of HIF-1 gene expression in CaCo2 cells after transfection with miRNA inhibitors. Comparisons between groups were performed by Kruskal–Wallis test. Individual data and median are plotted (n=3). *p≤0.05, **p≤0.01.

(0.11MB).

Then, HIF-1α gene expression was tested following miRNA inhibition. HIF-1α gene expression was 2–3-fold lower with all inhibitors used (Fig. 4B), suggesting that miRNAs are involved in the regulation of HIF-1α gene expression.

On the other hand, we tested if HIF-1 is mediating the regulation of miRNAs by IH. The increase of miR-21, miR-23b, miR-26a and miR-210 expression in CaCo2 after 4h of IH was significantly reversed following 0.5μM ACF treatment (Fig. 5). Similar effects were observed at other time points (2, 8 and 32h) (Supplementary Fig. S3). No differences were found between normoxic conditions with and without ACF. MiR-27b and miR-145 expression did not show any significant difference following treatment with ACF.

Fig. 5.

IH-induced expression of miRNAs is reversed by ACF treatment in CaCo2 cells. Cells were incubated in presence or absence of 0.5μM ACF for 4h in condition of normoxia or IH. Individual data and median are plotted (n=4). *p≤0.05, **p≤0.01, ****p≤0.0001.

(0.11MB).

Proliferation (Fig. 6A) and migration (Fig. 6B and C) of CaCo2 cells were significantly increased in IH in cells treated with DMSO 0.1% as a control (p=0.0014 and p≤0.0001). This increase of proliferation and migration was abolished by 0.5μM ACF (p=0.0004 and p≤0.0001).

Fig. 6.

ACF reverses IH-induced proliferation and migration of CaCo2 cells. (A) Viability of CaCo2 cells assessed by MTT assay in condition of normoxia and intermittent hypoxia with 0.1% DMSO or 0.5μM of acriflavine (ACF) for 32h. Cells in normoxia treated with DMSO 0.1% represented 100% of viability. Individual data and median are plotted (n=3). **p<0.005. (B) Transwell migration assay were performed to analyze the cell migration after exposure to normoxia or intermittent hypoxia for 48h, in presence or in absence of 0.5μM ACF. Individual data and median are plotted (n=4). **p<0.005. (C) Pictures of the lower side of transwells after migration test and crystal violet staining, taken under a light microscope coupled to a camera at ×10 magnification.

(0.2MB).

The results of this study elucidated that miR-21, miR-23b, miR-26a and miR-210 expressions were elevated in ONCO-OSA patients compared to non-OSA cancer patients and healthy controls. The same miRNAs were up-regulated in CaCo2 cells following IH exposure. In addition, miR-21, miR-26a and miR-210 had a higher expression in OSA compared to cancer and healthy controls. miRNAs expression correlated with the severity of sleep apnoea as evaluated by AHI, ODI and SpO2.

Clustering analysis allows the identification of miRNAs with similar expression profiles, that are generally involved in the same cellular functions or the same regulatory pathways.23 We showed that miR-21, miR-23b, miR-26a and miR-210 clustered together, while miR-27b and miR-145 were in a different cluster. These results are consistent with the fact that the 4 first miRNAs positively correlate with sleep parameters (AHI, ODI, SpO2) while the other 2 negatively correlate with the same parameters. It was also consistent with their regulation in CaCo2 cells, and with the finding that inhibiting miR-210, miR-21, miR-23b or miR-26a led to the reversion of IH-induced cell proliferation and migration.

It is known that miR-210 mediates important processes associated with tumorigenesis such as proliferation or angiogenesis. In fact, it targets many other miRNAs encoding proteins that play key roles in proliferation, DNA and RNA binding and repair, differentiation, development and apoptosis.24,25 miR-210 has been reported as involved in colorectal cancer migration and invasion through HIF-1.18 Our results about miR-210 expression agree with previous studies showing an upregulation of miR-210 in OSA patients.10

miR-21 and miR-23b have been described as up-regulated in many tumors15,26,27 and were elevated in one study investigating their expression in OSA patients.9 miR-21 was up-regulated in mice models of OSA28,29 but it was not consistent with another study reporting that miR-21 and miR-23 expression was down-regulated in OSA patients.30

MiR-26a is induced by hypoxia and its expression is up-regulated during cell differentiation. Previous studies have demonstrated its involvement in different types of cancers, and its target genes are involved in cellular process such as proliferation, differentiation, apoptosis, invasion and metastasis.16 miR-26a was associated with OSA severity in patients,31 although conflicting data showed that miR-26a was up-regulated by sustained but not intermittent hypoxia.10

Interestingly, although these 4 miRNAs are usually associated with cancer, in our cohort they were not elevated in cancer patients without OSA. By contrast, they were elevated in OSA patients (except for miR-23) and in ONCO-OSA patients, suggesting that OSA rather than cancer is responsible for the serum expression of these miRNAs.

On the other hand, miR-145 and miR-27b are generally down-regulated in various types of cancer and considered as inhibitors of tumor cell proliferation and migration,17,32,33 although conflicting data showed that miR-27 stimulated cell proliferation and invasion.34 In our study, these 2 miRNAs were found elevated in cancer patients and OSA patients but not in ONCO-OSA patients and in CaCo2 cells exposed to IH, suggesting a specific regulatory mechanism when OSA and cancer are combined.

In this study, we wanted to investigate the involvement of HIF-1 in the expression of specific miRNAs and its role in proliferation and migration of cancer cells exposed to intermittent hypoxia. In this regard, we used acriflavine, an inhibitor of HIF-1 dimerization, which decreased HIF-1 transcriptional activity and showed anticancer efficacy in vivo.35 We showed that inhibiting HIF-1 with acriflavine abolished the IH-induced increase in miRNA expression. This suggests that HIF-1 is necessary for miRNA upregulation. This is consistent with data showing that miR-210,18 miR-21,36 miR-26a16 are targets of HIF-1. On the other hand, we observed that inhibition of miR-21, miR-23b, miR-26a and miR-210 in turn led to significant reduction of HIF-1α gene expression. HIF-1 has indeed been described as a target of miR-21,14 miR-23b15 and miR-26a.37 Our data thus support the hypothesis of a hypoxia-triggered feedback loop involving the expression of HIF-1 and several miRNAs (Fig. 7).

We showed that acriflavine inhibited proliferation and migration of colorectal cancer cell exposed to IH acriflavine treatment. This is line with existing data showing the regulatory role of acriflavine in tumors35,38 and with the known role of HIF-1 in tumor progression induced by IH.39 Altogether, our results highlight a loop between HIF-1 and some miRNAs induced by IH/OSA and contributing to cancer development and progression. This hypothesis is recapitulated in Fig. 7.

Despite the relatively small population analyzed, we evidenced statistically significant differences in miRNA expression, suggesting that the variations are strong enough to be detected in this small cohort. Certainly, further studies on a larger and more carefully matched population will be needed to confirm our results.

Moreover, due to the predominance of colorectal cancer in the cancer and ONCO-OSA groups, we chose colorectal cancer cells (CaCo2) to investigate the cellular mechanisms induced by IH. Further studies will be required to confirm these mechanisms in other colorectal cancer cell lines and other types of cancer.