Pharmacovigilance of nephrotoxic drugs in neonates: the Pottel method for acute kidney injury detection in ELBW neonates.

BACKGROUND: Extremely low birth weight (ELBW) neonates (birth weight ≤ 1000 g) are at high risk to develop drug-induced acute kidney injury (AKI). However, we lack a pragmatic detection tool to capture their time-dependent (patho)physiologic serum creatinine (Scr) patterns. Pottel et al. suggested rescaling Scr by dividing Scr with the mean Scr value of the age- and sex-specific reference population. We explored if this Pottel method can detect drug-related nephrotoxicity in ELBW neonates. METHODS: A previously reported dataset on Scr changes in ELBW neonates exposed to ibuprofen, amikacin, or vancomycin was updated to calculate Pottel scores for every available Scr value in the first 28 postnatal days. We hereby used previously published postnatal age-specific 50th centile values in an ELBW population. Linear mixed models were applied, analyzing Pottel scores as response variable and continuous time (day), drug exposure, and interaction thereof in the explanatory model. RESULTS: Serum creatinine (n = 3231) observations in 201 ELBW neonates were collected. A statistically significant rise of Pottel scores was observed with ibuprofen starting from postnatal day 4. In addition, a cumulative effect of treatment with mean Pottel scores on day 0 of 1.020 and on day 3 during treatment of 1.106 (95% CI 1.068-1.145, p < 0.001) was observed, corrected for effect of antibiotics. Antibiotic administrations showed a small but statistically significant difference up to postnatal day 5. CONCLUSIONS: As rescaled Scr biomarker, the Pottel method showed a clear association with ibuprofen-exposed ELBW neonates, suggesting its applicability as a pragmatic bedside alternative tool to assess nephrotoxicity.

Estimated glomerular filtration rate: applicability of creatinine-based equations in African children.

BACKGROUND: The Schwartz equation is the most widely used serum creatinine (SCr)-based formula to estimate the glomerular filtration rate (GFR) in children of European descent, but whether this applies to African children is unclear. METHODS: In a cross-sectional study, 513 apparently healthy African children aged 6 to 16 years were randomly recruited in school area of Kinshasa, the Democratic Republic of Congo (DRC). SCr was measured using calibrated enzymatic method. SCr was normalized using Q-values designed for European descent children, due to the absence of Q-values for African children. Commonly used eGFR equations were applied in this population. RESULTS: Normalization of SCr using Q-values for European descent children was effective in this cohort. The majority of African children (93.4%) have normalized SCr (SCr/Q) values within the reference interval (0.67-1.33) of children of European descent. The bedside-Schwartz equation was associated with significant age and sex dependency. However, the FAS-Age formula showed no sex and age dependency. The new CKiDU25 equation did not show a significant sex dependency. The recently introduced EKFC and LMR18 equations also showed no age and sex dependency, although the distribution of eGFR-values was not symmetrical. On the other hand, the FAS-Height and the Schwartz-Lyon equations showed significant sex dependency but no age dependency. CONCLUSIONS: The reference interval for SCr designed for European descent children can be applied to African children. Of all the equations studied, FAS-Age performed best and is most suitable because no height measurements are required. Establishment of specific Q-values for the widespread Jaffe-measured creatinine in Africa can further broaden applicability.

Glomerular filtration rate estimation in adults: myths and promises.

BACKGROUND: In daily practice, glomerular filtration rate (GFR) is estimated with equations including renal biomarkers. Among these biomarkers, serum creatinine remains the most used. However, there are many limitations with serum creatinine we will discuss in the current review. We will also discuss how creatinine-based equations have been developed and what we can expect from them in terms of performance to estimate GFR. SUMMARY: Different creatinine-based equations have been proposed. We will show the advantages of the recent European Kidney Function Consortium (EKFC) equation. This equation can be used in children and adults. This equation can also be used with some flexibility in different populations. KEY MESSAGES: GFR is estimated by creatinine-based equations, but the most important for nephrologists is probably to know the limitations of these equations.

Estimating Glomerular Filtration Rate in China: Is the European Kidney Function Consortium (EKFC) Equation the Solution?

The new European Kidney Function Consortium (EKFC) creatinine-based equation has been developed to be applicable over the entire age range (from 2 to 100 years) without any loss of performance in young adults and without loss of continuity in estimating glomerular filtration rate (GFR) between adolescents and adults. This goal is obtained by better taking into account the relationship between serum creatinine (SCr) and age in the estimating GFR model. This is accomplished by rescaling SCr, namely, dividing SCr by so-called Q value which is the median normal value of SCr concentration in a given healthy population. The better performance of the EKFC equation, compared to the current equations, has been shown in large European and African cohorts. Such good results are also suggested in cohorts from China, including in the current issue of Nephron. The good performance of the EKFC equation is observed, especially when the authors used a specific Q value for their populations notwithstanding GFR was measured by a controversial method. Using a population-specific Q value could make the EFKC equation universally applicable.

Rescaling creatinine makes GFR estimation equations generally applicable across populations - validation results for the Lund-Malmö equation in a French cohort of sub-Saharan ancestry.

OBJECTIVES: To make glomerular filtration rate (GFR) estimating equations applicable across populations with different creatinine generation by using rescaled serum creatinine (sCr/Q) where sCr represents the individual creatinine level and Q the average creatinine value in healthy persons of the same population. METHODS: GFR measurements (mGFR, plasma clearance of 51Cr-EDTA) were conducted in 964 adult Black Europeans. We established the re-expressed Lund-Malmö revised equation (r-LMR) by replacing serum creatinine (sCr) with rescaled creatinine sCr/Q. We evaluated the r-LMR equation based on Q-values of White Europeans (r-LMRQ-white; Q-values females: 62 µmol/L, males: 80 µmol/L) and Black Europeans (r-LMRQ-Black; Q-values females: 65 µmol/L, males: 90 µmol/L), and the European Kidney Function Consortium equation (EKFCQ-White and EKFCQ-Black) regarding bias, precision (interquartile range, IQR) and accuracy (percentage of estimates within ±10 % [P10] and ±30 % [P30] of mGFR). RESULTS: Median bias of r-LMRQ-White/r-LMRQ-Black/EKFCQ-White/EKFCQ-Black were -9.1/-4.5/-6.3/-0.9 mL/min/1.73 m2, IQR 14.7/14.5/14.5/15.6 mL/min/1.73 m2, P10 25.1 %/34.8 %/30.3 %/37.2 % and P30 74.2 %/84.1 %/80.6 %/83.6 %. The improvement of bias and accuracy when using proper Q-values was most pronounced in men. Similar improvements were obtained above and below mGFR 60 mL/min/1.73 m2 and at various age and BMI intervals, except for BMI<20 kg/m2 where bias increased, and accuracy decreased. CONCLUSIONS: GFR estimating equations may be re-expressed to include rescaled creatinine (sCr/Q) and used across populations with different creatinine generation if population-specific average creatinine concentrations (Q-values) for healthy persons are established.

Extending the cystatin C based EKFC-equation to children - validation results from Europe.

BACKGROUND: A new cystatin C based European Kidney Function Consortium (EKFCCysC) equation was recently developed for adults, using the same mathematical form as the previously published full age spectrum creatinine based EKFC-equation (EKFCCrea). In the present study the cystatin C based EKFC-equation is extended to children, by defining the appropriate cystatin C rescaling factor QCysC. METHODS: Rescaling factor QCysC for cystatin C was defined as: a) 0.83 mg/L, exactly as it was defined for young adults in the adult equation, and b) a more complex QCysC-age relationship based on 4th degree cystatin C-age polynomials after evaluation of data from Uppsala, Stockholm and Canada and aggregated data from Germany. The EKFCCysC equation was then validated in an independent dataset in European children (n = 2,293) with measured GFR, creatinine, cystatin C, age, height and sex available. RESULTS: The EKFCCysC with the simple QCysC-value of 0.83 had a bias of -7.6 [95%CI -8.4;-6.5] mL/min/1.73 m2 and a P30-value of 85.8% [95%CI 84.4;87.3] equal to the EKFCCysC with the more complex 4th degree QCysC-value. The arithmetic mean of the EKFCCrea and EKFCCysC with the simple QCysC of 0.83 had a bias of -4.0 [95%CI -4.5;-3.1] mL/min/1.73 m2 and P30 of 90.4% [95%CI 89.2;91.6] similar to using the more complex 4th degree QCysC-polynomial. CONCLUSION: Using exactly the same QCysC of 0.83 mg/L, the adult EKFCCysC can easily be extended to children, with some bias but acceptable P30-values. The arithmetic mean of EKFCCrea and EKFCCysC results in bias closer to zero and P30 slightly over 90%.

Performance of the European Kidney Function Consortium (EKFC) creatinine-based equation in United States cohorts.

Estimating glomerular filtration rate (GFR) is important in daily practice to assess kidney function and adapting the best clinical care of patients with and without chronic kidney disease. The new creatinine-based European Kidney Function Consortium (EKFC) equation is used to estimate GFR. This equation was developed and validated mainly in European individuals and based on a rescaled creatinine, with the rescaling factor (Q-value) defined as the median normal value of serum creatinine in a given population. The validation was limited in Non-Black Americans and absent in Black Americans. Here, our cross-sectional analysis included 12,854 participants from nine studies encompassing large numbers of both non-Black and Black Americans with measured GFR by clearance of an exogenous marker (reference method), serum creatinine, age, sex, and self-reported race available. Two strategies were considered with population-specific Q-values in Black and non-Black men and women (EKFCPS) or a race-free Q-value (EKFCRF). In the whole population, only the EKFCPS equation showed no statistical median bias (0.14, 95% confidence interval [-0.07; 0.35] mL/min/1.73m2), and the bias for the EKFCRF (0.74, [0.51; 0.94] mL/min/1.73m2) was closer to zero than that for the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI2021) equation (1.22, [0.99; 1.47]) mL/min/1.73m2]. The percentage of estimated GFR within 30% of measured GFR was similar for CKD-EPI2021 (79.2% [78.5%; 79.9%]) and EKFCRF (80.1% [79.4%; 80.7%]), but improved for the EKFCPS equation (81.1% [80.5%; 81.8%]). Thus, our EKFC equations can be used to estimate GFR in the United States incorporating either self-reported race or unknown race at the patient's discretion per hospital registration records.

Diagnostic standard: assessing glomerular filtration rate.

Creatinine-based estimated GFR (eGFR) is imprecise at individual level, due to non-GFR-related serum creatinine determinants, including atypical muscle mass. Cystatin C has the advantage of being independent on muscle mass, a feature that led to the development of race- and sex-free equations. Yet, cystatin C-based equations do not perform better than creatinine-based equations to estimate GFR, unless both variables are included together. The new race-free Chronic Kidney Disease Epidemiology (CKD-EPI) equation, had slight opposite biases between Black and Non-Black subjects in USA, but performs poorer than that the previous version in European populations. The European Kidney Function Consortium (EKFC) equation developed in 2021 can be used both in children and adults, is more accurate in young and old adults, and is applicable to non-white European populations, by rescaling the Q factor, i.e. population median creatinine, in a potentially universal way. A sex- and race-free cystatin C-based EKFC, with the same mathematical design, has also be defined. New developments in the field of GFR estimation would be standardization of cystatin C assays, development of creatinine-based eGFR equations that would incorporate muscle mass data, implementation of new endogenous biomarkers, and the use of artificial intelligence. Standardization of different GFR measurement methods would also be a future challenge, as well as new technologies for measuring GFR. Future research is also needed on discrepancies between cystatin C and creatinine, which is associated with high risk of adverse events: standardize the definition of discrepancy, and understand its determinants.

Efficacy and safety of in-hospital treatment of Covid-19 infection with low-dose hydroxychloroquine and azithromycin in hospitalized patients: A retrospective controlled cohort study.

Objectives: In this study we evaluate the efficacy and safety of a treatment protocol with standard dose of hydroxychloroquine plus azithromycin in patients hospitalized with COVID-19 infection. Methods: We conducted a retrospective analysis to compare the 28-day mortality rate in 352 patients treated with hydroxychloroquine with or without azithromycin (HCQ-group) in our hospital with a contemporary control group of 3533 patients receiving standard of care from the Belgian Collaborative Group on COVID-19 Hospital Surveillance. Results: All patients who received at least one dose of treatment were included in the analysis. A statistically significant reduction in crude mortality rate at 28 days was observed in the HCQ-group compared to standard of care (16.8% vs 25.9%,p â€‹= â€‹0.001).Patients in the treatment group were on average younger (69,7 vs73,1 years, p â€‹= â€‹0,0002), were less likely to smoke or to have malignancy and more likely to be male. Patients in the treatment group were more likely to be obese, immunocompromised or to have arterial hypertension, liver disease and lung disease.After adjustment for these variables the OR for mortality was 0.635 (95%CI 0.464-0.875). Patients who did not receive HCQ had a 57% higher risk of mortality. A survival benefit in the treatment group was consistent across all age groups. 13 patients discontinued treatment due to side effects (4 with QTc-prolongation>60msec (1.1%) and 9 because of gastro-intestinal symptoms (2.55%)). No episodes of ventricular arrhythmia or torsade de pointes were recorded during treatment. Conclusion: Treatment of COVID-19 using a combination of hydroxychloroquine plus azithromycin was safe and was associated with a statistically significant mortality benefit in the treatment of COVID-19 infection in hospitalized patients. Our findings do not support the current negative recommendations regarding this treatment.

New and old GFR equations: a European perspective.

Glomerular filtration rate (GFR) is estimated in clinical practice from equations based on the serum concentration of endogenous biomarkers and demographic data. The 2009 creatinine-based Chronic Kidney Disease Epidemiology Collaboration equation (CKD-EPI2009) was recommended worldwide until 2021, when it was recalibrated to remove the African-American race factor. The CKD-EPI2009 and CKD-EPIcr2021 equations overestimate GFR of adults aged 18-30 years, with a strong overestimation in estimated GFR (eGFR) at age 18 years. CKD-EPICr2021 does not perform better than CKD-EPI2009 in US population, overestimating GFR in non-Black subjects, and underestimating it in Black subjects with the same magnitude. CKD-EPICr2021 performed worse than the CKD-EPI2009 in White Europeans, and provides no or limited performance gains in Black European and Black African populations. The European Kidney Function Consortium (EKFC) equation, which incorporates median normal value of serum creatinine in healthy population, overcomes the limitations of the CKD-EPI equations: it provides a continuity of eGFR at the transition between pediatric and adult care, and performs reasonably well in diverse populations, assuming dedicated scaling of serum creatinine (Q) values is used. The new EKFC equation based on cystatin C (EKFCCC) shares the same mathematical construction, namely, it incorporates the median cystatin C value in the general population, which is independent of sex and ethnicity. EKFCCC is therefore a sex-free and race-free equation, which performs better than the CKD-EPI equation based on cystatin C. Despite advances in the field of GFR estimation, no equation is perfectly accurate, and GFR measurement by exogenous tracer clearance is still required in specific populations and/or specific clinical situations.

Predicting outcomes of acute kidney injury in critically ill patients using machine learning.

Acute Kidney Injury (AKI) is a sudden episode of kidney failure that is frequently seen in critically ill patients. AKI has been linked to chronic kidney disease (CKD) and mortality. We developed machine learning-based prediction models to predict outcomes following AKI stage 3 events in the intensive care unit. We conducted a prospective observational study that used the medical records of ICU patients diagnosed with AKI stage 3. A random forest algorithm was used to develop two models that can predict patients who will progress to CKD after three and six months of experiencing AKI stage 3. To predict mortality, two survival prediction models have been presented using random survival forests and survival XGBoost. We evaluated established CKD prediction models using AUCROC, and AUPR curves and compared them with the baseline logistic regression models. The mortality prediction models were evaluated with an external test set, and the C-indices were compared to baseline COXPH. We included 101 critically ill patients who experienced AKI stage 3. To increase the training set for the mortality prediction task, an unlabeled dataset has been added. The RF (AUPR: 0.895 and 0.848) and XGBoost (c-index: 0.8248) models have a better performance than the baseline models in predicting CKD and mortality, respectively Machine learning-based models can assist clinicians in making clinical decisions regarding critically ill patients with severe AKI who are likely to develop CKD following discharge. Additionally, we have shown better performance when unlabeled data are incorporated into the survival analysis task.

Validated risk prediction models for outcomes of acute kidney injury: a systematic review.

BACKGROUND: Acute Kidney Injury (AKI) is frequently seen in hospitalized and critically ill patients. Studies have shown that AKI is a risk factor for the development of acute kidney disease (AKD), chronic kidney disease (CKD), and mortality. METHODS: A systematic review is performed on validated risk prediction models for developing poor renal outcomes after AKI scenarios. Medline, EMBASE, Cochrane, and Web of Science were searched for articles that developed or validated a prediction model. Moreover, studies that report prediction models for recovery after AKI also have been included. This review was registered with PROSPERO (CRD42022303197). RESULT: We screened 25,812 potentially relevant abstracts. Among the 149 remaining articles in the first selection, eight met the inclusion criteria. All of the included models developed more than one prediction model with different variables. The models included between 3 and 28 independent variables and c-statistics ranged from 0.55 to 1. CONCLUSION: Few validated risk prediction models targeting the development of renal insufficiency after experiencing AKI have been developed, most of which are based on simple statistical or machine learning models. While some of these models have been externally validated, none of these models are available in a way that can be used or evaluated in a clinical setting.

Posterior pelvic ring involvement detected with CT taken within a week of admission in acute fragility fractures of the pelvis (FFP) does not predict failure of conservative treatment: a retrospective cohort study.

BACKGROUND: Acute low energy pubic rami fractures in the elderly receive primarily conservative treatment. There is debate to what extent posterior ring involvement, which is detected superiorly by CT compared to X-ray, has an impact on outcome and may require modified treatment. We want to demonstrate if posterior ring involvement has an influence on different types of outcome in primarily conservatively treated acute FFP, questioning the usefulness of early CT. Additionally we analysed the early fracture pattern in cases where conservative treatment failed with need for secondary surgery. METHODS: A retrospective cohort study of 155 consecutive patients, recruited between 2009 and 2016, aged over 65 years diagnosed with an acute LE-PFr on X-ray at the emergency department of a single, level-one trauma centre and receiving an early CT. A set of outcome parameters was compared between patients with an isolated pubic rami fracture (CTia) and patients who had a combined posterior pelvic ring fracture (CTcp). RESULTS: There were 155 patients of whom 85.2% were female with a mean age of 83 years. 76.8% of patients living at home returned home and 15.5% moved to a nursing home. Mortality rate during hospitalisation was 6.4% and 14.8% at one year post-trauma. Secondary fracture displacement occurred in 22.6%. Secondary surgery was performed in 6 cases (3.9%). Median hospitalisation length of stay was 21 days (range 0 to 112 days). There was no significant association between the subgroups and change in residential status (p = 0.65), complications during hospitalisation (p = 0.75), mortality rate during admission (p = 0.75) and at 1 year (p = 0.88), readmission within 30 days (p = 0.46) and need for secondary surgery (p = 0.2). There was a significant increased median length of stay (p = 0.011) and rate of secondary displacement (p = 0.015) in subgroup CTcp. Secondary displacement had no impact on in-hospital complications (p = 0.7) nor mortality rate during admission (p = 0.79) or at 1 year (0.77). Early CT in patients who underwent secondary surgery showed stable B2.1 lesions in 4 of 6 cases. CONCLUSIONS: Our data suggest that early CT in patients with conservatively treated acute LE-PFr in order to detect posterior lesions, has limited value in predicting failure of conservative treatment.

Low agreement between various eGFR formulae in pediatric and young adult ADPKD patients.

BACKGROUND: Young autosomal dominant polycystic kidney disease (ADPKD) patients are becoming the new target population for the development of new treatment options. Determination of a reliable equation for estimated glomerular filtration rate (eGFR) from early stages is needed with the promising potential interventional therapies. METHODS: Prospective and longitudinal study on a cohort of 68 genotyped ADPKD patients (age range 0-23 years) with long-term follow-up. Commonly used equations for eGFR were compared for their relative performance. RESULTS: The revised Schwartz formula (CKiD) showed a highly significant decline in eGFR with aging (- 3.31 mL/min/1.73 m2/year, P < 0.0001). The recently updated equation by the Schwartz group (CKiDU25) showed a smaller (- 0.90 mL/min/1.73 m2/year) but significant (P = 0.001) decline in eGFR with aging and also showed a significant sex difference (P < 0.0001), not observed by the other equations. In contrast, the full age spectrum (FAS) equations (FAS-SCr, FAS-CysC, and the combined) showed no age and sex dependency. The prevalence of hyperfiltration is highly dependent on the formula used, and the highest prevalence was observed with the CKiD Equation (35%). CONCLUSIONS: The most widely used methods to calculate eGFR in ADPKD children (CKiD and CKiDU25 equations) were associated with unexpected age or sex differences. The FAS equations were age- and sex-independent in our cohort. Hence, the switch from the CKiD to CKD-EPI equation at the transition from pediatric to adult care causes implausible jumps in eGFR, which could be misinterpreted. Having reliable methods to calculate eGFR is indispensable for clinical follow-up and clinical trials. A higher resolution version of the Graphical abstract is available as Supplementary information.

Cystatin C-Based Equation to Estimate GFR without the Inclusion of Race and Sex.

BACKGROUND: The accuracy of estimation of kidney function with the use of routine metabolic tests, such as measurement of the serum creatinine level, has been controversial. The European Kidney Function Consortium (EKFC) developed a creatinine-based equation (EKFC eGFRcr) to estimate the glomerular filtration rate (GFR) with a rescaled serum creatinine level (i.e., the serum creatinine level is divided by the median serum creatinine level among healthy persons to control for variation related to differences in age, sex, or race). Whether a cystatin C-based EKFC equation would increase the accuracy of estimated GFR is unknown. METHODS: We used data from patients in Sweden to estimate the rescaling factor for the cystatin C level in adults. We then replaced rescaled serum creatinine in the EKFC eGFRcr equation with rescaled cystatin C, and we validated the resulting EKFC eGFRcys equation in cohorts of White patients and Black patients in Europe, the United States, and Africa, according to measured GFR, levels of serum creatinine and cystatin C, age, and sex. RESULTS: On the basis of data from 227,643 patients in Sweden, the rescaling factor for cystatin C was estimated at 0.83 for men and women younger than 50 years of age and 0.83 + 0.005 × (age - 50) for those 50 years of age or older. The EKFC eGFRcys equation was unbiased, had accuracy that was similar to that of the EKFC eGFRcr equation in both White patients and Black patients (11,231 patients from Europe, 1093 from the United States, and 508 from Africa), and was more accurate than the Chronic Kidney Disease Epidemiology Collaboration eGFRcys equation recommended by Kidney Disease: Improving Global Outcomes. The arithmetic mean of EKFC eGFRcr and EKFC eGFRcys further improved the accuracy of estimated GFR over estimates from either biomarker equation alone. CONCLUSIONS: The EKFC eGFRcys equation had the same mathematical form as the EKFC eGFRcr equation, but it had a scaling factor for cystatin C that did not differ according to race or sex. In cohorts from Europe, the United States, and Africa, this equation improved the accuracy of GFR assessment over that of commonly used equations. (Funded by the Swedish Research Council.).

Relation between obesity-related comorbidities and kidney function estimation in children.

BACKGROUND: The current childhood obesity pandemic is likely to result in an increased risk of chronic kidney disease (CKD) later in life. Correlations between obesity-related comorbidities and kidney function can be found, but it is unclear to what extent this is caused by bias due to different mathematical forms of the estimated glomerular filtration rate (eGFR) equations. The present study aimed to analyze correlations between obesity-related comorbidities and different eGFR equations and to investigate whether rescaled serum creatinine (SCr/Q) for sex and age or height might be an alternative biomarker for kidney function estimation. METHODS: This cross-sectional cohort study included 600 children with overweight and obesity. Mean age was 12.20 ± 3.28 years, 53.5% were female, and mean BMI z-score was 3.31 ± 0.75. All children underwent a comprehensive assessment that included anthropometrical and blood pressure measurements, laboratory examination, air displacement plethysmography, and polysomnography. Qage and Qheight polynomials were used to rescale SCr and multiple creatinine-based eGFR equations were compared. RESULTS: SCr/Q and almost all GFR estimations significantly correlated with a waist-to-hip ratio, fat mass, homeostasis model assessment for insulin resistance, and triacylglyceride, HDL cholesterol, alanine transaminase, and serum uric acid concentrations. Multiple correlations, however, were not confirmed by all equations, which suggests dependency on the mathematical form of the different eGFR equations. CONCLUSIONS: Correlations between obesity-related comorbidities and creatinine-based eGFR are present in children with overweight and obesity, but depend to a large extent on the eGFR equation of choice. SCr/Q might be an alternative biomarker for assessing correlations between obesity-related comorbidities and kidney function in children with overweight and obesity. A higher resolution version of the Graphical abstract is available as Supplementary information.