DNA methylation of a NF-κB binding site in the aquaporin 5 promoter impacts on mortality in sepsis

Altered aquaporin 5 (AQP5) expression in immune cells impacts on key mechanisms of inflammation and is associated with sepsis survival. Since epigenetic regulation via DNA methylation might contribute to a differential AQP5 expression in sepsis, we tested the hypotheses that DNA methylation of the AQP5 promotor (1) influences AQP5 expression, (2) is associated with the 30-day survival of septic patients, and (3) alters the nuclear transcription factor NF-κB binding. AQP5 mRNA expression was quantified by real-time PCR in whole blood samples of 135 septic patients. In silico computer analysis of the AQP5 promoter (nt-567 to nt-975) revealed seven putative inflammatory transcription factor binding sites and methylation of these sites was analyzed. Electrophoretic mobility shift assays were performed to assess the binding of nuclear NF-κB to the AQP5 promoter region nt-937. After adjustment for multiple testing, a greater methylation rate was found at cytosine site nt-937 in the AQP5 promoter linked to NF-κB binding in non-survivors compared to survivors (p = 0.002, padj = 0.014). This was associated with greater AQP5 mRNA expression in non-survivors (p = 0.037). Greater (≥16%) promoter methylation at nt-937 was also associated with an independently increased risk of death within 30 days (HR: 3.31; 95% CI: 1.54–6.23; p = 0.002). We detected a functionally important AQP5 promoter cytosine site (nt-937) linked to the binding of the inflammatorily acting nuclear transcription factor NF-κB, with increased methylation in sepsis non-survivors. Thus, nt-937 APQ5 promoter methylation, presumably related to NF-κB binding, is prognostically relevant in sepsis and demonstrates that epigenetic changes impact on sepsis outcome.

Sepsis is a grave medical condition and its mortality remains high 1 . Thus, identification of diagnostic and therapeutic targets is a corner-stone of current research. Since the wide variability regarding sepsis outcome cannot be explained solely by patients' comorbidities and severity of the triggering infection, some of this variability may be influenced by genetic variations that, in turn, may provide insights into relevant sepsis mechanisms.
Aquaporins seem to be a promising diagnostic and therapeutic target 2 , specifically the gene encoding aquaporin 5 (AQP5) 3 . AQP5 mediates key mechanisms of inflammation, including immune cell migration and proliferation 4,5 , activity of the renin-angiotensin-aldosterone system 6 , and the transport of water across cellular membranes 7 . Thus, AQP5 is involved in a lot of pathophysiological properties that prevail in sepsis and its altered expression seems to represent a crucial regulatory mechanism 2 . In this context, previous studies have shown that inflammatory mediators can induce the downregulation of AQP5 protein and mRNA expression 8,9 . In particular, there is a growing body of evidence showing that the activation of the proinflammatory NF-κB pathway attenuates AQP5 expression [10][11][12] . Notably, lower AQP5 expression attributable to the AQP5 -1364A/C single nucleotide promoter polymorphism (SNP; rs3759129) increased survival significantly in the acute respiratory distress syndrome (ARDS) 13 and in sepsis 14 . In addition, the AA-genotypes of the AQP5 -1364A/C associated with greater AQP5 expression showed increased pulmonary inflammation and an increased risk of acute kidney injury in ARDS 13,15 . Accordingly, these results suggest a protective impact of lesser AQP5 expression in proinflammatory diseases, and mechanisms linked to an altered AQP5 expression are of great interest.
Methylation of the cytosine residue in the sequence 5′-cytosine-phosphate-guanine-3′ (CpG) is a frequent epigenetic modification involved in the regulation of gene expression 16 and the degree of promoter methylation can also influence AQP5 expression 17,18 . Greater overall methylation of the AQP5 promoter diminished reporter gene transcription, whereas demethylation by 5-azacytidine evoked greater AQP5 expression 18 .
In the context of sepsis, there is growing evidence that epigenetic modifications can affect protein expression [19][20][21] ; hence, AQP5 promoter methylation might be a mechanism influencing AQP5 expression. However, it is unknown whether AQP5 expression under septic conditions might be epigenetically regulated by an altered promoter methylation. Accordingly, we tested the hypotheses that DNA methylation at a specific AQP5 promoter binding site is associated (1) with altered AQP5 expression, (2) 30-day survival of septic patients and (3) altered NF-κB binding. Table 1 shows the characteristics upon ICU admission of the 135 study patients (78 men [58%], 57 women [42%], mean age: 57.5 yrs. ± 16 SD) with sepsis who were admitted to the intensive care unit (ICU). The 30-day survival observed was 65% (88/135) and the median duration of ICU stay was 25 days [IQR: 12-36 days]. All patients were white Germans of Caucasian ethnicity. As expected, some differences were noted in baseline characteristics between sepsis survivors and non-survivors, such as the SOFA score (p = 0.003), platelet count (p = 0.025), and bilirubin concentration (p = 0.011; Table 1). Moreover, non-survivors were more frequently mechanically ventilated at baseline (79.5%, 35/47) compared to survivors (44.3%, 39/88; p = 0.001). By contrast, no evidence for significant associations of 30-day survival was found for age (p = 0.757), sex (p = 0.855), body mass index (p = 0.128), necessity for continuous hemofiltration/dialysis (p = 0.129) and Simplified Acute Physiology Score II (p = 0.119). Moreover, there were no mortality-dependent patterns regarding infection type (p = 0.581) or comorbidities (p = 0.969).
To verify binding of NF-κB to the AQP5 promoter region nt-937 an EMSA analysis was conducted covering the AQP5 promoter region from nt-922 to nt-940. Utilizing nuclear extracts of LPS-treated U937, HeLa, and THP-1 cells resulted in formation of a specific shift, and addition of an NF-κB antibody attenuated specific binding, probably due to the affection of the DNA binding site by a NF-κB (p65) antibody (Fig. 5).
Next, the prognostic value of nt-937 methylation and AQP5 mRNA expression to predict 30-day mortality was then evaluated by receiver operating characteristic curve analysis. The analysis revealed an area under the curve (AUC) of 0.717 (95% CI: 0.629-0.804; Fig. 6) for nt-937 methylation but only 0.586 (95% CI: 0.445-0.707; Fig. 6) for AQP5 mRNA expression, suggesting a superior prognostic accuracy for nt-937 methylation. A cut-off ratio of 16% nt-937 methylated cells was determined as an optimum regarding 30-day survival using Youden's J statistics. The related cut-off ratio of 16% showed a sensitivity of 76.6% and a specificity of 58.0% to discriminate between 30-day survivors and non-survivors. Using this cut-off, 30-day survival could be calculated as 82.3% for patients with an nt-937 AQP5 promoter methylation rate <16% but only 50.7% for those with an nt-937 methylation rate ≥16% (p < 0.001, Fig. 7).
Multivariable Cox regression analysis referring to the methylation of nt-937 when jointly considering age, sex, SOFA score and mechanical ventilation revealed that patients with 16% or more blood cells methylated at nt-937 had a hazard ratio for death of 3.31 (95% CI 1.54-6.23, p = 0.001; Table 2). Therefore, methylation of the APQ5 promoter site nt-937 had a significant and independent impact on 30-day survival prediction, independent of other important prognostic factors, such as the SOFA score (Table 2).

Discussion
This study shows that CpG methylation at the AQP5 promoter position "nt-937" in septic patients is associated with substantially greater 30-day mortality (estimated HR 3.3) and, therefore, represents an important and independent prognostic factor. This study also provides new evidence suggesting a significant impact of NF-κB binding to the AQP5 promoter on AQP5 gene expression. In this regard, the binding of NF-κB seems to be diminished by CpG methylation at AQP5 promoter position nt-937.
The AQP5 expression in immune cells seems to impact on key mechanisms of inflammation and immune cell migration 2,22 . Notably, diseases with potentially exaggerated inflammatory responses, such as sepsis or ARDS, are associated with an attenuated AQP5 expression, as demonstrated in different sepsis settings in human and rodent models 9,23-26 . In fact, downregulation of AQP5 expression may be an adaptive reaction of the immune system www.nature.com/scientificreports www.nature.com/scientificreports/ that may dampen inflammation-induced harm, especially in infections with an overwhelming inflammation or a rather "dysregulated" immune response 2 .
However, regulation of AQP5 expression varies in sepsis patients, and greater (or "less suppressed") AQP5 expression was accompanied by an increased 30-day mortality 13,14 . In line with these results, our present study also demonstrated a higher initial AQP5 mRNA expression in blood cells of patients dying from sepsis compared to survivors. However, regulatory mechanisms that explain altered expression levels of AQP5 in inflammatory diseases are still elusive.
We considered DNA methylation within the AQP5 promoter as a possible mechanism for differential AQP5 expression. It was shown previously that the AQP5 gene is regulated by CpG methylation in the AQP5 promoter region 18 . Nomura et al. demonstrated that a highly methylated murine cell rate was associated with a repressed www.nature.com/scientificreports www.nature.com/scientificreports/ AQP5 expression 18 , while cells in a hypomethylated state expressed great levels of AQP5 18 . In addition, they could demonstrate that the binding of the transcription factor SP1 in cultured mouse cells was increased in the hypomethylated state and contributed to basal AQP5 expression 18,27 .
In this study, we identified a functionally important CpG site at the AQP5 promoter position nt-937. The latter showed an increased methylation in non-surviving septic patients and was associated with greater AQP5 expression compared to sepsis survivors. Thus, we speculate that the binding of an "inhibitory" transcription factor to the specific promoter region around nt-937 might decrease promoter activity and subsequent gene expression. Notably, we could already demonstrate that the promoter region around this CpG site probably depicts an important promoter region and, therefore, may harbor silencer motives 6 . Strikingly, we could also confirm the binding of NF-κB to this specific AQP5 promoter region, as the specific band was attenuated by the addition of NF-κB (p65) antibody, supporting the evidence that NF-κB signaling acts in an inhibitory fashion on regulation of AQP5 expression 10,11,28,29 . Interestingly, several studies referring to an altered AQP5 expression in ARDS or sepsis could not find differential NF-κB or cytokine expression between survivors and non-survivors 5,13,14 . In this context, our findings may also provide a rationale that dissimilar AQP5 expression in sepsis may be facilitated by an altered NF-κB binding ability attributable to CpG methylation rather than NF-κB protein expression. Thus, it seems     www.nature.com/scientificreports www.nature.com/scientificreports/ appropriate to suggest that CpG methylation at the AQP5 promoter position nt-937 attenuates the inhibitory effects of NF-κB on AQP5 expression and may explain the relationship between inflammation and reduced AQP5 expression. Therefore, our findings may reconcile hitherto conflicting results and provide a plausible explanation for the altered AQP5 expression in sepsis even with impact on the patients' outcome. However, further experiments also with regard to the AQP5 -1364 A/C single nucleotide polymorphism genotype are needed to fully elucidate and understand the potential impact of epigenetic alterations on AQP5 expression and outcome in sepsis 14,15 .
One of the important questions to be addressed in follow-up investigations is how overexpression of different subunits and the subsequent change of NF-κB dimer composition affects AQP5 promoter activity containing methylated or unmethylated CpG nt-937. In addition, assessment of the kinetics and impact of NF-κB dimer formation and their impact on AQP5 expression is warranted. Finally, it seems prudent to analyze the methylation of nt-937 in different isolated immune cells, such as neutrophils, monocytes, and lymphocytes, so as to discover potential cell-type specific differences. However, we could not demonstrate a relevant association of different blood cell counts with AQP5 promotor methylation at CpG site nt-937 ( Supplementary Fig. 1). Simultaneous examination of the AQP5 promoter methylation and expression in different immune cells and at sequential time points could provide additional insight into molecular mechanisms. Apart from the mechanistic insights, our aim was to identify a potential prognostic marker for sepsis, that could easily be measured in DNA samples obtained from whole blood since DNA is a stable and abundant material and is therefore, ideal for bedside tests. Furthermore, altering methylation at the CpG-site nt-937 in the AQP5 promoter may constitute a potential therapeutic target, but robustness of this assumption should be assessed in further studies.
Limitations. Some limitations should also be mentioned. Although all sepsis patients were treated with a rather standardized multimodal regimen, we cannot exclude the possibility that unknown and potentially confounding factors exist. While repeated measurements during the further course of sepsis may have expanded our insights, associations between promoter methylation and prognosis in this study are limited to day 1 predictions upon ICU admission. In addition, some experiments of this study were performed solely on cell lines and not in cells from organs of septic patients due to ethical and methodical concerns. Therefore, results of our cell culture experiments may not allow translation to human pathology. Furthermore, the differences in promoter  www.nature.com/scientificreports www.nature.com/scientificreports/ methylation at nt-937 in cells of survivors and non-survivors and the methylation rate generally may appear small, since other studies have shown a very high AQP5 promoter methylation of more than 80% 30 . However, these values were only seen in leukemia or other cancer cells, and the promoters of several genes in leukemia patients are hypermethylated [31][32][33] . Furthermore, all these latter cells were cultured, which is also linked to a greater degree of AQP5 promoter methylation 18 . Nevertheless, other studies have indicated that methylation rate differences below 10% can be decisive for the magnitude of gene expression 34,35 . Furthermore, even on an epigenetic regulatory level, i.e. micro-RNA, gene methylation and acetylation, and histone rearrangement, in addition to interactions with other regulatory mechanisms, the regulation of AQP5 protein expression in sepsis is likely to be very complex and certainly cannot be clarified by a single study.

conclusions
We detected a functionally important AQP5 promoter CpG site (nt-937) linked to the binding of the inflammatorily acting nuclear transcription factor NF-κB, with increased methylation in sepsis non-survivors. Thus, nt-937 APQ5 promoter methylation, presumably related to NF-κB binding, is prognostically relevant in sepsis and demonstrates that epigenetic changes contribute to sepsis outcome.

Materials and methods
Patients. This study was reviewed and approved by the Ethics Committee of the Medical Faculty of the University of Duisburg-Essen (Essen, Germany; Protocol No. 06-3078). Written informed consent was obtained from all 135 participating patients or their guardians, according to the Declaration of Helsinki, good clinical practice guidelines and local regulatory requirements, for study inclusion between 2009 and 2014. Patients were considered eligible if they fulfilled the criteria for severe sepsis as defined by Bone et al. 36 Blood samples were taken within the first 24 h after criteria for severe sepsis were met. In addition, all patients included also met the criteria of the current SEPSIS-3 definition 37 . Patients' samples were gathered from a bio-database and some of these data had been utilized for previous studies 5, 38,39 .
All patients were followed up for 30-day survival calculated on the diagnosis of sepsis. Clinical and demographic data, including the Simplified Acute Physiology Score II (SAPS II) and Sequential Organ Failure Assessment (SOFA) score, were gathered within the first 24 h after the criteria mentioned above had been met. Patients were treated using a multimodal concept, including analgesia and sedation, fluid administration, protective mechanical ventilation and hemodynamic, antibiotic and diagnostic management, as described previously 14,39 . Blood sample collection, preparation and storage. Blood samples were collected and processed within 30 min. The DNA and RNA were each immediately extracted from whole blood using the QIAamp ® DNA Blood Mini Kit (Qiagen, Hilden, Germany) or RNeasy Mini Kit (Qiagen, Hilden, Germany), respectively, according to the manufacturer's instructions. The DNA and RNA samples were shock frozen and stored at −80 °C until analysis. Separate sample aliquots were stored and thawed later for analysis of the current hypotheses, so as to avoid multiple freezing and thawing procedures.
Real-time polymerase chain reaction (PCR) for expression analysis. The RNA samples collected from whole blood were considered eligible for analysis when their concentration was greater than 50 ng/µl, according to nanodrop measurement (ND-1000, PEQLAB Biotechnologie GmbH, Erlangen, Germany). The RNA samples from 98 septic patients were available in acceptable quality. First-strand cDNA was synthesized from 0.5 µg of total RNA using a QuantiTect Reverse Transcription Kit (Qiagen, Hilden, Germany). The qPCR reaction was performed using GoTaq ® qPCR Master Mix (Promega, Mannheim, Germany), as described previously 5 . A cDNA dilution series for AQP5 confirmed a PCR efficiency >95%, which was comparable to the efficiency of ß-actin (data not shown). Relative AQP5 mRNA expression was measured by two-step qPCR and expressed as 2 -∆CT .
Overall DNA methylation analysis by pyrosequencing. Bisulphite conversion was performed with the EZ DNA Methylation-Gold ™ Kit (Zymo Research, Irvine, CA, USA), according to the manufacturer's instructions for DNA methylation analysis. Briefly, 500 ng DNA was treated with bisulphite. Subsequently, DNA was diluted to a concentration of 10 ng/µl and pyrosequencing DNA methylation analysis was performed using ZymoTaq ™ PreMix (Zymo Research, Irvine, CA, USA) with the primer sets AQP5 (forward primers) F1 + F2 and AQP (reverse primers) R1 + R2 (Supplementary Table 1). The pyrosequencing reaction was performed using the sequencing primers AQP5 S1 + S2 (Supplementary Table 1) with the PSQ HS 96 Gold Reagent Kits (Qiagen, Hilden, Germany) spanning the promoter region from nucleotide (nt) -1081 to nt-547.