Fibroblast growth factor 21 as a biomarker for long-term complications in organic acidemias

A total of 17 patients (adults and children) attending the outpatient clinic of the metabolic center of the Erasmus University Medical Center, were selected. Thereafter FGF-21 measurements were done in leftover plasma samples previously collected and stored in the laboratory biobank for research purposes. Of the 17 patients included: 11 were known with PA (cases 1–11), four with MMA (cases 12–15), and two with IVA (cases 16 and 17). The clinical courses and other laboratory findings were obtained from the medical files. The study was assigned as non-Medical Research Involving Human Subjects Act (WMO) by the Medical Ethical Committee (METC) of the Erasmus University Medical Center. Patients and/or parents of the patients gave permission for publication of their anonymous clinical and laboratory data. Patients and/or parents were informed and aware of the use of leftover material for research to which none objected. Cardiomyopathy, prolonged QT interval, renal failure, and optic neuropathy were defined as long-term complications. Severe mental retardation and growth failure were not defined as long-term complications. A severe decompensation was defined as a blood pH level below 7.1 and/or ammonia levels above 400 μmol/L (Baumgartner et al. 2014). Total protein intake was defined as natural plus synthetic protein intake.

For diagnostic purposes, propionyl-CoA carboxylase activities were measured in fibroblasts at the Section Genetic Metabolic Diseases of the Erasmus University Medical Center, Rotterdam, The Netherlands. The propionyl-CoA carboxylase activity measurement is based upon the incorporation of ¹⁴C-labeled CO₂ (given in the form of ¹⁴C-labeled sodium carbonate) within the substrate propionyl-CoA. The excess CO₂ that was not incorporated was removed by acidification. At the same time, the incorporation of CO₂ was also measured with pyruvate and with methylcrotonoyl as the substrate. The isovaleryl-CoA dehydrogenase activities were measured in lymphocytes based upon the anaerobic electron transfer flavoprotein fluorescence reduction assay at the Laboratory for Genetic Metabolic Diseases, Academic Medical Center, Amsterdam, The Netherlands. Serum levels of FGF-21 were determined in duplicate using a commercial ELISA kit (Millipore), which is also applied in our diagnostic process. To establish reference ranges, two hundred control subjects were studied, covering all ages, including newborns. Positive controls were 50 patients with a mitochondrial disorder known with an increased plasma FGF-21 level. We established a reference range of 0–200 pg/ml for healthy subjects, which coincides with that of Suomalainen et al. (2011). The samples were measured in duplicate and no significant differences between samples (inter-sample variability) were found. Plasma FGF-21 levels up to 3000 pg/ml were linear, and measurements were reproducible up to 4000 pg/ml. Plasma FGF-21 levels higher than 4000 pg/ml were therefore cut off to a value of 4000 pg/ml. Total protein intake estimated at several time points was based on dietary prescription by a trained metabolic dietitian during standard care, results were compared to the WHO recommended daily allowance (RDA) (Joint FAO/WHO/UNU Expert Consultation on Protein and Amino Acid Requirements in Human Nutrition (2002: Geneva et al. 2007)). Plasma amino acid levels measured during the standard of care were obtained from the medical records to determine nutritional status, and based on reference levels expressed as low (below normal reference level), normal (within reference level) or elevated levels (above reference level). The patients with MMA and PA were selected to assess the association of plasma amino acid levels and plasma FGF-21 levels. Data were summarized by frequencies, median, and means with standard deviation (SD). Normal distribution was examined by Kolmogorov-Smirnov tests and quantile-quantile (Q-Q) plots. Student’s t-tests were performed to compare means in the case of Gaussian and Wilcoxon rank sum tests (Ws) of non normal distributions. Pearson’s correlation coefficient was used to evaluate the correlation between two Gaussian distributed continuous variables and Kendall’s tau rank correlation (r_τ) in the case of non normal distributions. Paired sample t-tests were performed to compare plasma FGF-21 levels in stable metabolic periods to levels in decompensations within each patient. In analyzing positive predictive value, negative predictive value, and odds ratios on the occurrence of long-term complications, we included plasma FGF-21 measurements during stable metabolic periods prior to the development of long-term complications in MMA and PA patients. In defining an optimal cut off point value for FGF-21 we used a receiver operating characteristic (ROC) curve (defining the plasma FGF-21 level with optimal sensitivity and specificity). Since protein restriction may infer FGF-21 increases, and since especially leucine and methionine restriction are associated with FGF-21 increase (Perez-Marti et al. 2016), Kendall’s tau rank correlation was calculated to test the correlation of total protein intake and methionine or leucine concentrations with plasma FGF-21 levels for each patient individually. Furthermore, we tested the association of other nutritional parameters, namely valine, isoleucine, alanine, phenylalanine, proline, and threonine, with plasma FGF-21 levels. P-values smaller than 0.05 were considered to indicate statistical significances. SPSS data manager (version 24) was used for statistical analyses.

Patient characteristics of all 17 patients with organic acidemia are presented in Table 1. Four patients were diagnosed by family or selective newborn screening and one patient was diagnosed prenatally. Eight patients (six with PA and two with MMA) presented in the neonatal period. Two of these patients were initially on mechanical ventilation and one was dialyzed. A total of four patients presented in childhood. These patients presented either with complaints of vomiting, malaise, and weight loss; with gastro-esophageal reflux, with encephalopathy or coma.

Sixteen patients showed a mild clinical course with no severe decompensations after the initial presentation (Table 1). One patient (case 10) died in the neonatal period during the first severe decompensation due to a refractory hemodynamic shock and hyperammonemia. Despite good metabolic control, eight of the remaining 16 patients developed long-term complications (Table 2). The mean age at the first long-term complication was 8.0 years (± SD: 3.4). Five PA patients developed cardiac complications, four of them (cases 1, 2, 4, and 8) developed cardiomyopathy between 8 and 13 years of age. In one patient (case 2) this was unexpected as she was diagnosed prenatally by selective screening and she always showed good metabolic control. A rapidly progressive dilated cardiomyopathy lead to cardiac failure and death at eight years of age, four months after the diagnosis of cardiomyopathy. Case 8 had mild cardiomyopathy, but died during a severe metabolic decompensation at the age of 12 years. One patient had prolonged QT interval without signs of cardiomyopathy. Four patients developed renal failure between the age of 11 months and 12 years of age. Two patients developed optic neuropathy, both at the age of 16 years. One of them was also known with reduced left ventricular function, and the other patient was already known with renal failure.

A total of 62 FGF-21 measurements were performed in the 17 patients. Plasma FGF-21 levels were determined before the onset of long-term complications in seven of the eight patients with long-term complications (Fig. 1a). In the remaining patient, FGF-21 measurements were performed in samples obtained after long-term complications had occurred (Fig. 1a). All FGF-21 levels from samples retrieved before long-term complications developed, were elevated (Fig. 1a). In two patients (cases 1 and 2), plasma FGF-21 levels were already elevated, up to ten times the upper level of normal, from the neonatal period onward and remained at the same level with older age. In one patient (case 3), plasma FGF-21 levels increased rapidly from the neonatal period up to eight months of age and remained high thereafter. In another patient (case 8), plasma FGF-21 levels were extremely high before complications and decreased thereafter, as previously mentioned, she eventually died of dilated cardiomyopathy. One patient (case 11) demonstrated high plasma FGF-21 levels during stable disease, at that time point without long-term complications. In an IVA patient without long-term complications (case 16), the highly elevated plasma FGF-21 levels taken during a metabolic decompensation normalized thereafter (Fig. 1b).

In MMA and PA patients, median plasma FGF-21 levels (of each patient) during stable metabolic periods were significantly higher in patients with long-term complications during follow up (median FGF-21 = 2556.0 pg/ml) compared to those who did not develop complications (median FGF-21 = 287.0 pg/ml) (Ws = 29.00, Z = −2.066, p = 0.039) (Fig. 2a). In MMA and PA patients, the individual mean plasma FGF-21 levels during stable metabolic periods (M = 2600, SD = 843) did not differ from the individual median plasma FGF-21 levels during decompensations (M = 2309, SD = 619) (t(6) = 0.566, p = 0.592). None of the patients with a median plasma FGF-21 below 1500 pg/ml during the stable metabolic period (n = 5) developed long term complications. All, but one patient, developed long-term complications in those with a median plasma FGF-21 above 1500 pg/ml (n = 5) (samples taken during a stable metabolic period prior to the occurrence of long-term complications) in MMA and PA patients (Suppl. Fig. 1). A median plasma FGF-21 level above 1500 pg/ml during a stable metabolic period, measured before the occurrence of long-term complications, has a positive predictive value of 0.83 on long-term complications and a negative predictive value of 1.00 in MMA and PA patients.

Plasma creatinine levels did not correlate with plasma FGF-21 levels within the individual patient; nor did the median plasma FGF-21 level of each patient correlate with the median plasma creatinine level of each patient.

Total protein intake (g/kg/day) did not correlate with plasma FGF-21 levels in any patient, nor did median total protein intake (of each patient) correlate with median plasma FGF-21 levels (of each patient). Furthermore, total protein intake as %RDA was not associated with plasma FGF-21 levels, except in one patient (case 3) (r_τ(9) = 0.957, p < 0.001) (Suppl. Fig. 2). In MMA and PA, plasma valine levels were low in the majority of the patients (Suppl. Fig. 3). In MMA and PA, plasma FGF-21 levels were not lower in patients with low plasma amino acids levels; valine, isoleucine, leucine, methionine, threonine, alanine, phenylalanine, and proline, respectively (Suppl. Fig. 3). Individual plasma amino acid levels were not associated with plasma FGF-21 levels, except for one patient (case 16). In this patient, all three branched chain amino acids (BCAA) were associated with plasma FGF-21 levels (valine (r(8) = −0.643, p = 0.026), isoleucine (r(8) = −0.618, p = 0.034), leucine (r(8) = −0.618, p = 0.034)).

There is no clear association of plasma FGF-21 levels and age of the patient, in patients with long-term complications (data not shown) as well as in patients without long-term complications (Fig. 1b). In 16 of 17 patients, growth charts were available. Six out of these 16 patients had moderate to severe growth retardation (Table 2). In all of the patients, height z-score did not correlate with plasma FGF-21 levels. The median plasma FGF-21 levels did not differ significantly in patients with growth retardation (z-score < -2SD) compared to patients without growth retardation (data not shown).