Plasma glial fibrillary acidic protein levels in children with sickle cell disease

To determine if glial fibrillary acidic protein (GFAP) is associated with brain injury in children with sickle cell disease (SCD), we measured plasma GFAP among cross-sectional groups of unselected children with SCD, subsets of children with SCD and normal brain MRI or MRI evidence of cerebral infarct, healthy pediatric controls, and adults with brain injury. Children with SCD had higher plasma GFAP than healthy pediatric controls (mean concentrations 0.14 ± 0.37 vs. 0.07 ± 0.08 ng/mL; P = 0.003); also, 16.0% (16/100) of children with SCD and cerebral infarct had GFAP elevations above the 95th percentile of healthy pediatric controls (P = 0.04). Although not statistically significant, more children with SCD and cerebral infarct had elevated GFAP levels than with SCD, and no infarct (16/100, 16.0% vs. 14/168, 8.3%; P = 0.07) was observed. Children with SCD and acute brain ischemia had a higher proportion of elevated GFAP than SCD children with normal MRI (3/6, 50% vs. 8.3%; P = 0.01). GFAP was associated with elevated systolic blood pressure in the preceding year and correlated positively with white blood cell count and negatively with age and performance IQ. Plasma GFAP is elevated among children with SCD and may be associated with subclinical brain injury. To determine if glial fibrillary acidic protein (GFAP) is associated with brain injury in children with sickle cell disease (SCD), we measured plasma GFAP among cross-sectional groups of unselected children with SCD, subsets of children with SCD and normal brain MRI or MRI evidence of cerebral infarct, healthy pediatric controls, and adults with brain injury. Children with SCD had higher plasma GFAP than healthy pediatric controls (mean concentrations 0.14 ± 0.37 vs. 0.07 ± 0.08 ng/mL; P = 0.003); also, 16.0% (16/100) of children with SCD and cerebral infarct had GFAP elevations above the 95th percentile of healthy pediatric controls (P = 0.04). Although not statistically significant, children with SCD and cerebral infarct had more elevated GFAP levels than with SCD and no infarct (16/100, 16.0% vs. 14/168, 8.3%; P = 0.07). Children with SCD and acute brain ischemia had a higher proportion of elevated GFAP than SCD children with normal MRI (3/6, 50% vs. 8.3%; P = 0.01). GFAP was associated with elevated systolic blood pressure in the preceding year and correlated positively with white blood cell count and negatively with age and performance IQ. Plasma GFAP is elevated among children with SCD and may be associated with subclinical brain injury. Brain injury in sickle cell disease (SCD) occurs on a spectrum of severity. On one end of the spectrum, children are normal or manifest subtle deficits only on sensitive neurocognitive testing; alternatively, on the severe end of the spectrum, people with SCD can acquire devastating neurologic deficits [1-3]. Cerebral infarct is the most commonly observed form of brain injury, and it may be detected as silent cerebral infarct (SCI) or overt infarct with neurologic deficits. In cases of overt stroke, traumatic brain injury, and other forms of CNS injury, plasma levels of neuronal and glial proteins reflect cellular leak from areas of injury [4-6]. Biomarkers in this group can be very specific for brain injury. This approach of using brain-specific proteins has been taken by investigators to identify biomarkers of overt stroke in adults [7-11]; however, the identification of brain injury from brain-specific proteins leaked into the plasma of SCD patients, or any patient with clinically silent infarct or ischemia, has not been described. GFAP is a highly brain-specific intermediate filament protein, which is a known biomarker of acute stroke and head trauma in adults [6, 12-15]. In ongoing studies of SCD plasma for proteomic discovery of brain injury biomarkers, we noted that GFAP was detectable in the plasma of some patients with SCD; therefore, we tested the hypothesis that elevations in plasma GFAP are associated with brain injury in children with SCD. Baseline characteristics for healthy pediatric controls and children with SCD screened for SCI as part of the SIT Trial with baseline plasma GFAP levels are shown in Supporting Information Table I. Adjudicated MRI readings were available on 268/295 (90.8%) of SIT trial subjects with plasma GFAP measurements. Figure 1A shows GFAP levels for healthy pediatric controls (n = 60), children with SCD (n = 295), and adults with overt brain injury 1–22 days prior (n = 28). To validate that our GFAP assay detected elevated plasma GFAP in the context of brain injury, a positive control population composed of plasma samples from adults after acute stroke (n = 12), brain biopsy (n = 3), or partial brain resection (n = 13) was assayed and found to have mean plasma GFAP levels of 0.906, 1.42, and 5.10 ng/mL, respectively. As the plasma concentration of GFAP in children is unknown, we established the normal range using controls from a general pediatric outpatient clinic. Healthy pediatric controls had a mean plasma GFAP concentration of 0.072 ± 0.083 ng/mL, and children with SCD had a mean of 0.144 ± 0.368 ng/mL (P = 0.003). The 95th percentile for GFAP in healthy pediatric controls was 0.227 ng/mL, with 3/60 healthy pediatric controls (5.0%) above this threshold. Log plot of plasma GFAP concentrations in children with SCD and controls. (A) Displayed are GFAP concentrations in healthy pediatric controls (n = 60), children 5–14 years old with HbSS and HbSβ°–thalassemia (n = 295); and adults with overt brain injury: stroke (n = 12), brain biopsy (n = 3), or partial brain resection (n = 13) 1–22 days prior. There is one GFAP observation per subject. Bars represent mean values. (B) Log plot of plasma GFAP concentrations in subsets of SCD children with no infarct by brain MRI (n = 168), cerebral infarct by brain MRI (n = 100), and MRI (DWI) evidence of acute infarct (n = 6) and plasma drawn within 21 days of the MRI. The dashed line marks the 95th percentile value among 60 healthy pediatric controls, and percentages list the proportions above the 95th percentile of healthy controls. Children with SCD and cerebral infarct on MRI had more elevated GFAP levels than healthy pediatric controls, as defined by the 95th percentile cutoff (16/100, 16.0%; P = 0.04). SCD subjects with normal brain MRI had a similar proportion of elevated GFAP to healthy pediatric controls (14/168, 8.3%; P = 0.6). The proportion of SCD subjects with cerebral infarct and elevated GFAP (16.0%) was higher than SCD subjects with normal MRI and elevated GFAP (8.3%), although the difference was not statistically significant (P = 0.07, Fig. 1B). To determine the temporal relationship between GFAP levels and cerebral infarct, we analyzed GFAP levels in all children with acute brain ischemia by MRI, as defined by restricted diffusion on diffusion-weighted imaging (DWI) and a corresponding decrease in signal intensity on the apparent diffusivity coefficient map. Six children with SCD had plasma samples drawn within 21 days of a brain MRI that showed ischemia and DWI positivity. Among children with SCD and positive DWI, 3/6 had elevated GFAP (50%; Fig. 2), which is greater than the proportion of children with SCD and elevated GFAP and cerebral infarct of undetermined age (13.8%; P = 0.05) and SCD children without infarct (8.3%; P = 0.01). Temporal relationship of plasma GFAP levels to diagnosis of acute cerebral infarct. Diffusion weighted brain MRI imaging (DWI) was positive in six children with SCD. Blood samples were drawn either before or after brain MRI. The dashed line marks the 95th percentile value among 60 healthy pediatric controls. The shaded area indicates the approximate age of infarct, as estimated by DWI. The child with a sample drawn 15 days prior to MRI was ultimately determined to have an overt stroke. All children were clinically asymptomatic at the time of MRI screening and sample acquisition. SCI has previously been correlated with higher WBC count and lower hemoglobin [16]. A positive correlation was seen with WBC count, and a negative correlation, although not statistically significant, was seen with hemoglobin (Supporting Information Table II). When stratified by cerebral infarct status, the association between WBC count and GFAP persisted among children without cerebral infarct (Spearman ρ = 0.18, P = 0.02) and children with cerebral infarct (Spearman ρ = 0.20, P = 0.05). Wechsler Abbreviated Scale of Intelligence [17] performance IQ negatively correlated with GFAP levels (Spearman ρ = −0.29, P = 0.04). The verbal IQ assessment showed a nonstatistically significant negative correlation (Spearman ρ = −0.19, P = 0.19). Systolic blood pressure (SBP) has been associated with SCI and overt stroke in adults [18, 19], and we postulated that a relationship may exist between elevated SBP and elevated GFAP in SCD. There was little correlation between GFAP and peak SBP in the year prior to study screening (Supporting Information Table II). To address the possibility that GFAP elevations occur above a threshold SBP, we compared peak SBP over the year before study screening to the presence of elevated GFAP levels. Among children with SCD and peak SBP >2 standard deviations above the sex, height, and age-adjusted mean, GFAP was elevated (>95th percentile of controls) more frequently than among children with SCD and normal peak SBP (17.7% vs. 8.6%, P = 0.04). When stratified by cerebral infarct status, a weak association between peak SBP and GFAP remained among children without cerebral infarct (Spearman ρ = 0.11, P = 0.15) but not children with cerebral infarct (Spearman ρ = 0.01, P = 0.9). There was a negative correlation between age and plasma GFAP concentrations (Supporting Information Table II), which was statistically significant among children with SCD. A negative correlation was also observed among the age-matched healthy pediatric controls (Spearman ρ = −0.19, P = 0.15). Age or developmental-dependent levels of circulating GFAP have not been described previously. We report the distribution of plasma GFAP concentrations and their associations with the spectrum of brain injury in SCD, as assessed by MRI and other measures associated with cerebral infarct. As a group, patients with SCD showed elevated levels of GFAP when compared to healthy pediatric controls, and there was a nonstatistically significant difference between children with and without cerebral infarct. In a subset of six children enriched for acute cerebral ischemia, GFAP was most strongly correlated. The observation that these children had the highest proportion of elevated GFAP (50%), followed by children with SCD and cerebral infarct of undetermined age (13.8%), and, finally, children with SCD and no cerebral infarct (8.3%) suggests that elevated plasma GFAP concentrations may identify some children with cerebral infarct, particularly acute infarct; however, this study design precludes any meaningful conclusions about overall sensitivity and specificity. Plasma GFAP is not elevated in most patients with SCD, even those with a prior cerebral infarct. Thus, GFAP is only weakly associated with cerebral infarct of undetermined age and would not be useful clinically on a cross-sectional basis. The apparent temporal relationship between cerebral infarct and elevated GFAP posed a particular challenge for this study. Because this study was cross sectional, the ages of cerebral infarcts on MRI were unknown, with the exception of six subjects who were serendipitously determined to have acute ischemia. After acute stroke in adults, blood GFAP levels are elevated but return towards baseline within approximately 7 days [6, 14]. The reduction in plasma concentrations of GFAP after brain injury likely represents a combination of the extent and time-course of injury, as well as clearance of plasma proteins by renal and proteolytic pathways; furthermore, clearance may be uniquely affected by SCD pathophysiology, for example, renal hyperfiltration or hemolysis, but the mechanisms of GFAP clearance from blood have yet to be characterized. With our cross-sectional study design, linking an episodic, subclinical event to a transiently elevated plasma biomarker would be a chance event. Longitudinal studies including MRI are necessary to define how long GFAP is elevated in children with SCD and to define further the relationship of GFAP levels to the spectrum of brain MRI findings in SCD. The most striking finding of this study may be that 11.2% of children with SCD and, no overt evidence of neurologic symptoms had elevated GFAP. Children with SCD and elevated GFAP may have ongoing brain injury, regardless of MRI status, as many of these children have extremely high levels that may be consistent with brain injury, and we observed a negative correlation between IQ and GFAP levels. At completion of the SIT Trial, longitudinal studies of serial GFAP measurements and 3-year follow up MRI and neurocognitive studies will be evaluated to determine if elevated plasma GFAP levels precede evidence of brain injury. This longitudinal analysis will be especially important for the children with normal brain MRIs and elevated GFAP, because elevated GFAP may be a harbinger of cerebral infarct or future neurocognitive decline. In summary, we demonstrate for the first time that elevations of plasma GFAP, a known marker of CNS injury, can be detected in the plasma of children with SCD who did not show clinical signs of brain injury, and elevations of GFAP correlate with acute cerebral infarct in SCD, albeit weakly in those of undetermined age. Exploratory associations find that GFAP levels may correlate positively with WBC count and SBP and negatively with age and performance IQ. Further studies of brain-specific blood biomarkers in SCD should increase our understanding of the epidemiology of brain injury and enhance detection, prevention, and treatment of CNS injury in SCD. Four groups of patients were studied: children with SCD, subsets of children with SCD and normal brain MRI or MRI evidence of cerebral infarct, healthy pediatric controls, and adults with overt brain injury and no SCD. The study population of children with SCD was selected from those screened for the SIT Trial (ClinicalTrials.gov NCT00072761) [20]. The SIT Trial is a randomized controlled trial in which children 5 to 14 years old with either HbSS or HbSβ0-thalassemia (n = 1,211) were screened for SCI, and eligible children were randomly assigned to receive blood transfusion therapy or conventional therapy for 36 months. Plasma sample acquisition from the SIT trial was optimized for proteomic studies starting February 2007. For this nested analysis, children with SCD, who were screened in the SIT trial from February 2007 to May 2009, were included. An additional three samples were selected from outside this time-period to include all SIT subjects with known acute brain ischemia on MRI. MRI imaging sequences for acute ischemia (DWI) was not required by the SIT trial, but a subset of study sites elected to perform these sequences. The final sample size included 295 SCD children. Cerebral infarct is defined as MRI signal abnormality visible on two views on T2-weighted images. The signal abnormality must measure at least 3 mm in one dimension, and a determination of cerebral infarct status is adjudicated by a panel of neuroradiologists [20, 21]. The SIT trial is supervised by the Institutional Review Boards at each participating institution. All SIT trial participants signed informed consent. Healthy pediatric controls (5–16 years, n = 60) and adults with overt brain injury (≥18 years, n = 28 with acute stroke, brain biopsy, or partial brain resection) were selected from Johns Hopkins Hospital clinics and inpatient units. For healthy pediatric controls, clinic notes were reviewed to exclude patients with any acute illness, neurologic disorder, or chronic illness other than asthma, obesity, and behavior/mood disorders. Blood from healthy pediatric controls was drawn for routine testing. Deidentified blood samples and clinical data on healthy pediatric controls and adults with overt brain injury were obtained through a separate IRB approved study at Johns Hopkins Hospital. Data were extracted from a central data-coordinating center. SBP was analyzed using gender, height, and age-specific z-scores as values for these measurements. IQ testing was performed only on those children who were eligible by screening criteria and consented to be allocated randomly to treatment assignment in the SIT trial. Blood was collected into ACD or EDTA tubes, spun at 1,500g for 8 min at room temperature per the SIT trial protocol, and stored at −80°C in the biologic repository for the SIT trial at Johns Hopkins University School of Medicine until analysis. Two plasma samples were inadvertently diluted 3.3-fold through a systematic error, and GFAP levels for these samples were adjusted accordingly. The error was reproduced in an experiment to confirm the dilution factor of 3.3. The GFAP assay and performance has been described previously [4]. The primary analysis was to assess whether elevations in plasma GFAP concentrations were different among the following three groups: children with SCD and cerebral infarct, children with SCD and no cerebral infarct, and healthy pediatric controls. GFAP values from adults with overt brain injury are presented to calibrate our GFAP assay with the spectrum of brain injury, and no statistical comparisons are made between these adults and the various pediatric groups. Among children with SCD, we also compared GFAP levels among those with cerebral infarct of undetermined age and a subset of children with positive (DWI) scans by brain MRI, indicating acute or subacute cerebral infarction (within 10–14 days [22, 23]). Twenty-seven children did not have cerebral infarct determination because of withdrawal from the study from screening before MRI (n = 21) or inadequate MRI studies (n = 6). Children were included in analyses if the relevant clinical and laboratory data were available. Means between groups were compared using two-tailed t tests with unequal variance. We compared proportions of elevated concentrations in each SCD group to pediatric healthy controls (abnormal defined by >95% percentile of controls). Tests of proportion were performed using a two-tailed Fisher's exact test. Correlations between GFAP concentrations and other continuous variables were assessed using the Spearman correlation coefficient. Statistical significance was defined as P < 0.05. Values of GFAP below the lower limit of quantification of the assay (0.040 ng/mL) were recorded as zero. Analyses were conducted using Stata v11.1 (Stata Corporation, College Station, TX). W.S. designed and performed research, analyzed data, and wrote the article. E.B.C. designed and performed research. Z.F. designed and performed research. P.D. performed research. L.W. performed research. B.C. designed research. D.W. performed and designed research and analyzed data. J.J. analyzed data and wrote the article. J.V.E. designed research and contributed vital research tools. M.D. designed research, analyzed data, and wrote the article. A.D.E. designed research, analyzed data, and wrote the article. J.C. designed research, performed research, analyzed data, and wrote the article. Additional Supporting Information may be found in the online version of this article. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.