|Home | About | Journals | Submit | Contact Us | Français|
Intraventricular hemorrhage (IVH) is the most frequent, severe neurological complication of prematurity and is associated with post-hemorrhagic hydrocephalus (PHH) in up to half of cases. PHH requires lifelong neurosurgical care and is associated with significant cognitive and psychomotor disability. Cerebrospinal fluid (CSF) biomarkers may provide both diagnostic information for PHH and novel insights into its pathophysiology.
To explore the diagnostic ability of candidate CSF biomarkers for PHH.
Concentrations of amyloid precursor protein (APP), soluble APPα, sAPPβ, NCAM-1, L1CAM, tau, phosphorylated tau, and total CSF protein (TP) were measured in lumbar CSF from neonates in six groups: (a) no known neurological disease (n=33); (b) IVH Grades I-II (n=13); (c) IVH Grades III-IV (n=12); (d) PHH (n=12); (e) ventricular enlargement without hydrocephalus (n=10); and (f) hypoxic ischemic encephalopathy (n=13). CSF protein levels were compared using analysis of variance, and logistic regression was performed to examine the predictive ability of each marker for PHH.
Lumbar CSF levels of APP, sAPPα, L1CAM, and TP were selectively increased in PHH compared to all other conditions (all p<0.0001). The sensitivity, specificity, and odds ratios of candidate CSF biomarkers for PHH were determined for APP, sAPPα, and L1CAM; cutpoints of 699, 514, and 113 ng/ml yielded odds ratios for PHH of 80.0, 200.0, and 68.75, respectively.
Lumbar CSF APP, sAPPα, L1CAM, and TP were selectively increased in PHH. These proteins, and sAPPα in particular, hold promise as biomarkers of PHH and provide novel insight into PHH-associated neural injury and repair.
Intraventricular hemorrhage (IVH) is a common, severe neurological complication of prematurity, occurring in one-third of infants born ≤28 weeks post-menstrual age (PMA).1,2 Up to one half of preterm infants with IVH develop post-hemorrhagic hydrocephalus (PHH),3–5 a condition associated with substantial neurological disability and complex, lifelong neurosurgical care.1,6Despite the morbidity associated with PHH and its treatment, there is no consensus regarding the diagnosis or treatment of PHH, largely due to a dearth of quantifiable PHH metrics.
At present, clinical assessment of infants with PHH includes physical parameters such as head circumference, tenseness of the anterior fontanel, and splaying of cranial sutures, as well as changes in vital signs, which occur late in the disease course. As a result, clinicians rely in large part on semiquantitative imaging-based measurements of ventricular size for clinical decision-making. However, ventricular size is an imperfect metric for the treatment of neurological disease in preterm infants. Neurological injury (e.g. IVH, infarct, hypoxia-ischemia), white matter volume loss, and impaired brain development are all common among preterm infants and impact ventricular size, irrespective of PHH. New tools are needed to complement existing measures of ventricular size and inform clinical research aimed at improving the diagnosis, treatment, and outcomes of infants with PHH.
Cerebrospinal fluid (CSF) proteins have been investigated as candidate biomarkers of PHH by our group and others.7–13 In particular, CSF levels of amyloid precursor protein (APP) isoforms, L1 cell adhesion molecule (L1CAM), and neural cell adhesion molecule-1 (NCAM-1) have been shown to be elevated in PHH and to normalize after initiation of clinical treatment.7 Further, CSF APP levels have been shown to be associated with ventricular size and possibly intracranial pressure throughout the active PHH treatment interval.14 Thus, CSF levels of APP and related proteins represent promising candidate biomarkers of PHH. The potential role of CSF biomarkers in PHH should not be limited to diagnosis of the condition, however, nor even monitoring therapeutic efficacy, but may extend to the prediction of long-term neurological outcome and developmental impairment.
Prior investigations into CSF biomarkers of PHH and other types of pediatric hydrocephalus, including those from our group, share the limitation of using CSF acquired via lumbar puncture (LP) (in subjects without hydrocephalus) as the comparator for ventricular CSF obtained from children with hydrocephalus.7,8 Comparing ventricular to lumbar CSF introduces the potential for sampling artifact, as rostro-caudal protein gradients (RCGs) exist between ventricular and lumbar CSF and may vary both by protein and by neurological disease.15,16 For hydrocephalus, RCGs may also vary by hydrocephalus etiology.16
In the current study, we account for RCGs in PHH by comparing the levels of candidate biomarkers (APP, soluble APP-α, soluble APP-β, Aβ1–42, NCAM-1, L1CAM, tau, and phosphorylated tau (p-tau)) in CSF obtained from human infants exclusively via LP. Using this approach, candidate CSF biomarkers were compared not just between control and PHH but across a number of newborn neurological conditions, including IVH alone, hypoxic ischemic encephalopathy (HIE), and ventricular enlargement without hydrocephalus (VEWOH).
Approval from the Washington University Human Research Protection Office was acquired prior to initiation of this study (WU HRPO #s 201101887 and 201203126). As part of routine clinical care, all preterm infants ≤34 weeks estimated PMA admitted to the St. Louis Children’s Hospital Neonatal Intensive Care Unit (NICU) underwent routine head ultrasound examinations between 0 and 5 days of age to screen for IVH or other neurological injuries. IVH was graded according to Papile’s criteria,17 and, if present, ultrasounds were repeated weekly in order to monitor for progressive ventricular dilatation. PHH was diagnosed in infants who demonstrated a frontal-occipital ratio ≥0.55 on ultrasound or magnetic resonance imaging (MRI),18 progressive increase in occipitofrontal circumference, splaying of the sagittal suture ≥2 mm in the mid-parietal region, and palpation of the anterior fontanel above the level of the surrounding bone.19 The current study includes subjects who were diagnosed with PHH and underwent LP prior to neurosurgical implantation of a ventricular access device (VAD) for either diagnostic purposes (eg to rule out infection prior to VAD implantation) or therapeutic purposes (eg to enable ventricular decompression while awaiting VAD) (Table 1). None of the 12 infants with PHH in this study had loculated hydrocephalus or a trapped fourth ventricle at the time of acquisition of the CSF sample, and none went on to later develop a trapped fourth ventricle or require treatment for loculated hydrocephalus (shunt system with >1 ventricular catheter or fenestration procedure) at a median follow-up 27.2±15.4 months. Additional subjects with clinical findings and radiological evidence (ultrasound and/or MRI) consistent with HIE or VEWOH were recruited for the study. MRIs of HIE subjects involved injury to the deep gray nuclei, cerebral cortex, and subcortical white matter. Those with VEWOH showed ventriculomegaly from cerebral malformation, ‘arrested hydrocephalus,’ or other unidentifiable cause but did not develop macrocephaly (occipitofrontal circumference >97th centile) or require neurosurgical treatment. A cohort of infants diagnosed with IVH who did not go on to develop PHH were also included in the study. Infants without any known neurological insult or injury who underwent LP for routine sepsis evaluation (where all cultures were deemed negative and sepsis was ruled out) were recruited and termed ‘controls’ (Table 1).
Lumbar CSF samples were collected from February 1, 2012–March 1, 2015. During this same time interval, the rate of post-hemorrhagic ventricular dilatation at our institution was 5–7 cases/year (total of 17 cases). Twelve of these patients (65%) were recruited and included in this study. Lumbar punctures were performed by the clinical team in the NICU under sterile conditions and for clinical purposes only. Immediately after LP, the samples were transported directly to the St. Louis Children’s Hospital clinical laboratory where they were frozen at −20°C for a maximum of two days. They were then transported on ice to the Washington University Neonatal CSF Repository, where they were stored at -80°C. Prior to experimental analysis, samples were thawed and centrifuged at 2500 rpm for six minutes, and the supernatant was analyzed for protein levels.
In all cases, CSF microbiological cultures sent at the time of the LP remained sterile (cultures were monitored for 3.68±0.13 days); per our NICU practice, anaerobic CSF cultures were not routinely performed. Data for CSF cell count and differential analysis were recorded from the electronic medical record. Using Bonadio et al’s20 proposed threshold red blood cell count of >1000/mm3, “traumatic” LPs were noted in 15/56 (26.8%) of the CSF samples (IVH and PHH samples excluded), which is consistent with or lower than previously reported data for LPs in neonates.21,22 For neonatal plasma samples, blood was collected from newborn infants in evacuated tubes with anticoagulant. The plasma was separated within two hours and aliquoted and stored at −80°C until experimental analysis.
Commercially available Enzyme-Linked ImmunoSorbent Assays (ELISAs) were used to measure the concentration of APP, soluble APP alpha and beta (sAPPα and sAPPβ), Aβ42, L1CAM, NCAM-1, tau, and p-tau. Sandwich Duoset ELISA development systems (R&D Systems, catalog #DY-850, and DY-2408 respectively; Minneapolis, MN) were used to measure APP and NCAM-1 as previously described.14 L1CAM levels were measured using a commercially available kit (DRG, catalog #EIA5074; Mountainside, NJ) and have also been described previously.14 Soluble APPα and sAPPβ were measured using ELISAs (IBL-International, catalog #27734 and 27732, respectively; Toronto, ON, Canada). The levels of tau, p-tau, and Aβ42 were also determined by ELISAs (Fujirebio, Ghent, Belgium, catalog #80226, 80062, 80177). In each instance, ELISAs were run according to the manufacturer’s protocol. In brief, plates were coated with a primary capture antibody and blocked prior to incubation with CSF or plasma samples. After washing, the secondary antibody was added, followed by streptavidin-HRP and tetramethylbenzidine. The reaction was then stopped with sulfuric acid. The plates were washed between each of the prior steps except the final one. All ELISAs were run in duplicates and the 96-well plates were read at 450nm on a Versamax microplate reader (Molecular Devices; Sunnyvale, CA). Protein concentrations were then determined using a four parameter logistic standard curve as detailed by the manufacturers.
The Pierce Bicinchoninic Acid protein assay kit (Thermo Scientific; Waltham, MA) was used to estimate the total protein (TP) level in each sample. Briefly, samples and bovine serum albumin standards were placed into microplate wells in duplicate. The working reagent was then added and the plate was incubated at 37°C for 30 minutes. The plate was then cooled to room temperature and the absorbance was measured at 562nm on a plate reader. TP levels were determined using a four parameter logistic standard curve.
One-way ANOVA was used to test for association of CSF proteins with PHH and alternate outcome categories (IVH, VEWOH, HIE, and control). A Bonferroni corrected threshold (0.05/9 =.0055) was calculated for multiple comparisons. Proteins APP, sAPPα, and L1CAM showed association and were used in simple logistic regressions of PHH on candidate biomarker concentration to examine their predictive ability. Sensitivity, specificity, and odds ratios were used to evaluate candidate cut point concentrations. Significance was set at a P-value <.05. SAS 9.4 was used for all analyses.
Six groups of research subjects were identified based on clinical history and events and radiologic findings (ultrasound and/or MRI): 1) control subjects (no known neurological insult or injury, n=33); 2) IVH Grades I-II (n=13); 3) IVH Grades III-IV (n=12); 4) PHH (n=12); 5) HIE (n=13); and 6) VEWOH (n=10). Please refer to Table 1 for summary characteristics of subject groupings. There was no difference among the first four groups for birth weight, PMA at the time of birth, or PMA at the time of acquisition of the CSF sample; however, compared with PHH subjects, subjects with HIE and VEWOH tended to have higher birth weights (P<.0001 and P=.0065, respectively) and be older at birth (P<.0001 and P=.0004, respectively).
Lumbar CSF levels of APP, sAPPα, sAPPβ, Aβ42, NCAM-1, L1CAM, tau, p-tau, and TP were measured as detailed in Methods (Table 2). Among APP and its derivatives, CSF APP and sAPPα were increased in PHH compared to all other conditions (P<.0001 and P<.0001, respectively; Figure 1A–B). CSF sAPPβ was different between PHH and control (P=.0436), and Aβ42 was not significantly different among any groups. CSF L1CAM was elevated in PHH relative to all other conditions (P<.0001; Figure 1C). NCAM-1 was elevated in PHH relative to the other conditions with the exception of IVH III-IV (P=.1365). Total tau was not different among groups (P=.0951), but p-tau was variable, greater in PHH than control (P=.0153), IVH I-II (P=.0221), HIE (P<.0071), and VEWOH (P<.0061). In order to account for multiple comparisons, the Bonferroni threshold was calculated as 0.05/9 or 0.0055. Three candidate CSF biomarkers, APP, sAPPα, and L1CAM met the corrected threshold for significance (P<.0001 for each) and were selective for PHH.
Total CSF protein was also elevated in PHH (P<.0001; Table 2), consistent with findings from other groups studying various forms of hydrocephalus.9 While it is possible that blood or blood breakdown products contributed to this difference, this is unclear since TP was significantly lower in IVH III-IV compared with PHH (P=.0052). In order to examine whether blood or blood breakdown products could be contributing to the increases in CSF APP, sAPPα, L1CAM, and NCAM-1, the levels of these proteins were measured in age-matched neonatal plasma samples. With the exception of NCAM-1, which trended towards higher levels in the plasma samples compared with PHH CSF (P=.1395), APP, sAPPα, and L1CAM were significantly lower in plasma than in PHH CSF (P=.0086, P=.0034, and P=.0028, respectively). These relationships extended to plasma versus control CSF; NCAM-1 was higher in plasma (P<.0001), while APP and L1CAM trended towards lower levels in plasma (P=.0845 and P=.1223, respectively); sAPPα was significantly lower in plasma (P=.0147). Thus, it is unlikely that blood or blood breakdown products are responsible for the observed PHH-associated increases in CSF APP, sAPPα, and L1CAM.
Logistic regression was used to evaluate the relationship between each of the candidate CSF biomarkers (APP, sAPPα, sAPPβ, Aβ42, NCAM-1, L1CAM, tau, p-tau) and CSF cell count parameters (total cells, nucleated cells, red blood cells, neutrophils, lymphocytes, monocytes, eosinophils, and macrophages). None of the candidate biomarkers demonstrated a significant relationship with any of the cell count parameters, with the exception of CSF L1CAM with CSF neutrophils (P=.0223) and CSF monocytes (P=.0216).
The predictive ability of lumbar CSF APP, sAPPα, and L1CAM levels for PHH were examined by first generating a logistic plot for each candidate biomarker using the measurements for CSF samples from infants with PHH and those without PHH (control, IVH Grades I-II, IVH Grades III-IV, HIE, VEWOH) (Figure 2A–C). Simple logistic regression was performed on each biomarker to determine the strength of prediction for PHH and cut points were chosen based on combinations of sensitivity and specificity (Table 3–4). CSF APP (cutpoint=699.67 ng/ml, sensitivity=0.91, specificity=0.89), sAPPα (cutpoint=514.43 ng/ml, sensitivity=0.91, specificity=0.95), and L1CAM (cutpoint=113.33 ng/ml, sensitivity=0.91, specificity=0.89), yielded odds ratios for PHH of 80.00, 200.00, and 68.75, respectively.
In recent years, biomarkers have been sought to facilitate the diagnosis of PHH and improve clinical management of this devastating condition (reviewed in Mehrar, 20128). Previous work from our group has shown that ventricular CSF levels of APP, L1CAM, and other protein mediators of neurodevelopment are increased in PHH and resolve after initiation of neurosurgical treatment.7 CSF APP in particular also appears to be associated with ventricular size, and possibly intracranial pressure, in infants undergoing treatment for PHH.14 Along with much of the work in this field, these studies share the limitation of using lumbar CSF as the comparator for ventricular CSF obtained from children with PHH. Such a comparison introduces the potential for sampling artifact due to RCGs, which have been observed for specific proteins in a number of diseases.15,16,23,24 The objective of the current study was to eliminate the potential bias of RCGs and examine the levels of APP, L1CAM, and related proteins using lumbar CSF from infants with PHH and other neurological conditions relevant to the neonate.
Studies into RCGs suggest that the levels of brain-derived proteins (e.g. neuron-specific enolase, or NSE, and S100B) remain consistent regardless of access site, while blood-derived proteins (e.g. albumin and β–trace) tend to be higher in lumbar CSF.25 These findings are supported by recent proteomic analysis.26 However, at least one study found differential abundance of amyloid isoforms (lumbar levels higher) and tau isoforms (ventricular levels higher) in normal pressure hydrocephalus, with albumin levels highest in lumbar samples.16 The etiology of RCGs remains uncertain, though investigators have speculated that the gradients reflect diffusion or transport of serum proteins into the CSF, or CSF proteins into the blood, as a function of distance from the brain (the primary site of brain-derived proteins).16,23
The findings presented in this study provide convincing evidence that lumbar CSF levels of APP, sAPPα, and L1CAM are elevated in untreated PHH compared with neonates without neurological injury and those with IVH alone (Grades I-II and Grades III-IV), HIE, and VEWOH. Related APP derivatives sAPPβ and Aβ42 as well as tau and p-tau did not demonstrate the same degree of selectivity for PHH. NCAM-1 was increased in both PHH and Grade III-IV IVH, possibly due to its high concentration in plasma. Thus, in this study of relatively small sample size, CSF APP, sAPPα, and L1CAM appeared the most appropriate CSF biomarkers of PHH. The strong predictive ability of APP, L1CAM, and particularly sAPPα, for PHH is an important and entirely novel finding.
The high levels of TP observed in PHH in our lumbar CSF samples will not surprise clinicians who treat PHH regularly. In many centers, serial LPs are employed as a temporizing measure for PHH until the condition resolves or the infant undergoes a more definitive neurosurgical procedure, and large increases in TP are expected in this setting. To our knowledge, a comparison of TP in lumbar CSF of infants with IVH alone, PHH (or other neurological conditions), or adults with hydrocephalus, has not been reported previously.
In addition to any potential role in PHH diagnosis or clinical management, CSF biomarkers of PHH give important insight into the pathophysiology of this condition. PHH afflicts preterm infants during the critical developmental period corresponding to 23 weeks PMA to term equivalent age and beyond. Seminal neurodevelopmental events, including precursor migration and differentiation, synaptogenesis, and the generation of neural networks occur during this interval and are almost certainly affected by PHH. As noted previously, many of the CSF biomarkers involved in PHH are protein mediators of neurodevelopment.7 For example, L1CAM is involved in neuronal migration, axonal growth, and synaptogenesis.27–32 Indeed, mutations in L1CAM have been linked to the debilitating conditions of MASA (mental retardation, aphasia, spastic paraplegia, adducted thumbs) syndrome, X-linked hydrocephalus, and CRASH (Corpus callosum hypoplasia, Retardation, Adducted thumbs, Spastic paraplegia and Hydrocephalus) syndrome.33–37
Amyloid precursor protein has long been known to be increased in the setting of axonal injury or stretch,38 but accumulating evidence also strongly supports that APP is an important trophic factor, with roles in neural stem cell development, neurite outgrowth and synaptogenesis, and neural repair.39 Post-translational APP processing is complex and may occur at a number of stages, each of which affects the ultimate fate and biological activity of APP and its derivatives. Perhaps the most critical processing point is the cleavage of cell surface APP by α- or β-secretase. If cleaved by α-secretase (a disintegrin and metalloproteinase, or ADAM), sAPPα is generated; if cleaved by β-secretase (β-site APP-cleaving enzyme 1, or BACE1), sAPPβ results.40 While β-site cleavage and sAPPβ are associated with downstream production of amyloid-β1–42, the primary component of neuritic plaques in Alzheimer’s disease, sAPPα appears to function as a trophic factor for neural precursor cell development, neurite growth and synapse formation, and even neuronal growth, survival, and repair (reviewed in Dawkins and Small39). The results from this study clearly demonstrated specific increases in CSF sAPPα in PHH, and while it is possible that sAPPα may simply result from the neurological injury itself, it is intriguing to consider that sAPPα may be serving a neuronal survival or repair function.
It is worth noting that in adults, sAPPα and sAPPβ, but not tau and p-tau, are also diagnostic biomarkers for distinguishing between idiopathic normal pressure hydrocephalus and Alzheimer’s disease.41 Specifically, the cutoff value for predicting cognitive outcomes in Miyajima et al’s study41 was 198ng/ml, which is considerably lower than our value of 514ng/ml for sAPPα. In a related study, Pyykko et al42 concluded that increases in APP isoforms were independent of neuroinflammation mechanisms in patients with idiopathic normal pressure hydrocephalus. This finding lends further support to our contention that the trophic effects of APP are important for functional outcome.
The role of neuroinflammation in PHH is an active area of study for our lab and others.43–48 Intraventricular hemorrhage with or without PHH may incite inflammatory processes within the CSF and within the central nervous system more broadly, potentially creating a chemical meningitis. In order to explore the influence of neuroinflammation in PHH, we analyzed the relationship between our candidate CSF biomarker levels and the CSF cell count and differential analysis results (including total cells, nucleated cells, RBCs, neutrophils, lymphocytes, monocytes, eosinophils, and macrophages) in the CSF samples tested. In this relatively small series, no significant relationships existed between CSF APP, sAPPα, sAPPβ, Aβ42, NCAM-1, L1CAM, tau, p-tau, and differential cell counts in CSF, with the exception of CSF L1CAM and neutrophils (P=.0223) and monocytes (P=.0216). This is not entirely surprising, as the biomarkers investigated in this study are primarily associated with neurodevelopment and neurological injury, rather than inflammation. One relevant exception to this generalization is L1CAM, which has recently been shown to be induced by transforming growth factor (TGF)-β149 and have a signaling role in inflammation via nuclear factor (NF)-κB.50
A number of limitations in this study must be acknowledged. In this study of human newborn infant lumbar CSF samples, there were no true “control” infants; even those subjects with no known neurological injury were born prematurely and subject to the influence of myriad clinical factors that could potentially confound the results (e.g. clinical status concerning for sepsis, long periods of ventilation). Additionally, in utero or developmental events and pre- and peri-natal complications (preterm labor, rupture of membranes, breech presentation, placental abruption) could potentially influence CSF protein levels in these very preterm infants. In the analysis of our specimens, spectrum bias may have attenuated the comparison between conditions, particularly between IVH III-IV and PHH. Similarly, the degree of ventricular enlargement in our VEWOH samples was modest, again introducing the possibility of spectrum bias and limiting the ability to rigorously test the effect of ventricular enlargement alone (no hydrocephalus) on CSF biomarker levels. Of note, VEWOH and HIE infants had greater PMA at birth and higher birth weights than the other groups, limiting stringent comparisons between groups. Finally, while it would be ideal to examine biomarker levels in both ventricular and lumbar CSF simultaneously, clinical and ethical challenges prevent this from being performed routinely in these preterm infants.
Future studies will focus on validating the results of this study in a larger collection of CSF samples and examining the levels of these candidate biomarkers of PHH (APP, sAPPα, and L1CAM) in parallel blood samples. We will also explore the relationship of these biomarkers to neurosurgical (shunt malfunction or infection, need for surgical revision, association with multiloculated hydrocephalus) and neurodevelopmental outcomes. Similarly, we will also investigate the levels of these and related biomarkers in hydrocephalus of other etiologies.
In the current study, we report the novel observation that the lumbar CSF levels of APP, sAPPα, and L1CAM are specifically and significantly elevated in untreated PHH. In addition to their potential role as diagnostic biomarkers, these proteins provide novel insights into the pathophysiology of PHH and possible mechanisms of neural injury and neural repair.
We would like to express our profound gratitude to the patients and families who graciously participated in this study and the teams of physicians and nurses that cared for them
This project was supported by a career development award to David D. Limbrick Jr (NIH/NINDS K23 NS075151), the Washington University Intellectual and Developmental Disabilities Research Center (NIH/NICHD P30 HD0627171), and the Washington University Institute for Clinical and Translational Research (NIH CTSA award UL1 RR024992). Additional support was provided through the Children’s Surgical Sciences Institute, St. Louis Children’s Hospital (David D. Limbrick Jr) and the Washington University Dean’s Fund (Shawgi A. Silver). Clinton D. Morgan was supported through a Howard Hughes Medical Institute fellowship and Rakesh Rao was supported by the Women’s and Infant’s Health Specimen Consortium.
Drs. Limbrick and McAllister receive research funds and/or research equipment for unrelated research projects from Medtronic, Inc. (Minneapolis, MN), Karl Storz, Inc. (Tuttlingen, DE), and Aesculap, Inc. (Center Valley, PA). Dr. Limbrick has received philanthropic equipment contributions for humanitarian relief work from Karl Storz, Inc., and Aesculap, Inc.
The authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article.