|Home | About | Journals | Submit | Contact Us | Français|
The amyloid-β peptide (Aβ) plays a central pathophysiological role in Alzheimer’s disease, but little is known about the concentration and dynamics of this secreted peptide in the extracellular space of the human brain. We used intracerebral microdialysis to obtain serial brain interstitial fluid (ISF) samples in 18 patients who were undergoing invasive intracranial monitoring after acute brain injury. We found a strong positive correlation between changes in brain ISF Aβ concentrations and neurological status, with Aβ concentrations increasing as neurological status improved and falling when neurological status declined. Brain ISF Aβ concentrations were also lower when other cerebral physiological and metabolic abnormalities reflected depressed neuronal function. Such dynamics fit well with the hypothesis that neuronal activity regulates extracellular Aβ concentration.
Aβ is the principal constituent of the hallmark amyloid plaques found in Alzheimer’s disease and is the target of many potential treatments for the disease (1). However, little is known about the concentration and dynamics of this secreted peptide in the extracellular space of the human brain where these plaques form. In vitro and animal studies have shown that neuronal and synaptic activity dynamically regulate soluble extracellular Aβ concentrations (2–4). Whether similar regulation of Aβ levels occurs in the human brain is unknown.
We used intracerebral microdialysis (5) to obtain serial brain interstitial fluid (ISF) samples in 18 intensive care unit (ICU) patients who had sustained acute brain injury and were undergoing invasive intracranial monitoring for clinical purposes. In all patients, Aβ1−x was detected in hourly or bihourly intracranial microdialysis samples. None had a diagnosis of Alzheimer’s disease or dementia, demonstrating that Aβ is a normal constituent of human brain extracellular fluid (6). The Aβ1−x enzyme-linked immunosorbent assay (ELISA) used detects Aβ species from amino acid 1 to amino acid 28 or greater (3, 7). There were rising trends in brain ISF Aβ concentrations over several hours to days in most patients, though the specific pattern of these trends was variable (Fig. 1, B, D, and F; Fig. 2B; and Fig. 4, A to D). Median brain ISF Aβ1−x at 60 to 72 hours was 59% higher than at 0 to 12 hours (Fig. 1G) (P = 0.0002, Wilcoxon signed rank test). Urea concentrations in the same samples, which control for the stability of the microdialysis catheter function (8), remained stable over the same time frame (Fig. 1H) (median 14% lower, P = 0.06, Wilcoxon signed rank test). Thus, the observed Aβ dynamics are likely to be of cerebral origin and not an artifact of the measurement procedure.
Aβ1−x concentrations were lower in microdialysate than in concomitantly sampled ventricular cerebrospinal fluid (CSF) (Fig. 2, B and D). However, at the flow rate used (0.3 μl/min), the microdialysate is not in complete equilibrium with the surrounding extracellular space (5). To calculate the true brain extracellular concentrations, we used two methods to estimate the fractional recovery of Aβ under these conditions (Fig. 2C). First, we performed in vitro recovery experiments with known concentrations of Aβ. Second, we used in vivo zero-flow extrapolation, which relies on the principle that as flow rate approaches zero, equilibration across the membrane will approach completion (9). These two methods agreed well; recovery of Aβ was ~30% at 0.3 μl/min (Fig. 2C and figs. S1 and S2). After correction for this recovery, the concentrations in ventricular CSF and brain ISF were, on average, very similar (Fig. 2D). However, the range of concentrations in brain ISF was significantly wider (Fig. 2D), and the hour-by-hour dynamics of Aβ changes correlated poorly (Fig. 2, B and E).
The 42–amino acid form of Aβ (Aβ1–42) is of special interest; this form appears to have the greatest propensity to deposit into insoluble plaques, one of the pathological hallmarks of Alzheimer’s disease, as well as to aggregate into oligomeric Aβ species, reported to be especially neurotoxic (10–12). Current ELISA methods are not sensitive enough to measure Aβ1–42, Aβ1–40, and Aβ1−x in individual hourly or bihourly microdialysis samples, so 8-hour pools of microdialysis samples were combined for five patients. In these pooled samples, the Aβ1–40/Aβ1–42 ratio was 14 ± 5.2. Aβ1–42 concentrations were lower than Aβ1−x concentrations by a factor of 35 ± 12, and Aβ1–40 concentrations were lower than Aβ1−x concentrations by a factor of 2.5 ± 0.9. Thus, much of the Aβ measured by the Aβ1−x assay does not appear to be either Aβ1–40 or Aβ1–42. Brain ISF Aβ1–42 concentrations were consistently proportional to those of Aβ1−x (Fig. 2F). The relation was similar in ventricular CSF (Fig. 2F), though greater variability was observed. Thus, the observed trends in brain ISF Aβ1−x levels over time likely reflect changes in Aβ1–42 levels as well.
To address potential mechanisms underlying these dynamics, we compared brain ISF Aβ1−x concentrations from single samples with various microdialysis parameters and physiological measures obtained concurrently as part of clinical care in the intensive care unit (Fig. 3 and figs. S3 and S4). A positive correlation with brain ISF glucose (Fig. 3A), as well as negative correlations with brain ISF lactate/pyruvate ratio (Fig. 3B), elevated intracranial pressure (Fig. 3C), and extremes of cerebral temperature (Fig. 3D), were the most salient.
Thus, brain ISF Aβ increased as overall physiology normalized. Indeed, in several patients, brain ISF Aβ concentrations tracked the patients, global neurological status, as assessed with the Glasgow Coma Score (Fig. 4, A to D); brain ISF Aβ concentrations rose as patients improved, remained stable in clinically stable patients, and appeared to decline when clinical status worsened. The Glasgow Coma Score (13) is a crude but reliable measure of neurological status in severely brain-injured patients that ranges from 3 (deeply comatose) to 15 (eyes open spontaneously, following commands, and speaking appropriately).
To quantify this relation across patients, we plotted change in brain ISF Aβ1−x from single samples versus change in Glasgow Coma Score (Fig. 4E). The clinical importance of a one-point change in the Glasgow Coma Score is often unclear, especially in critically injured, sedated patients. So for our primary analysis, we assessed changes in brain ISF Aβ only at times when Glasgow Coma Score had changed by two or more points from baseline (Fig. 4E). The correlation in this analysis was markedly strong (Spearman r = 0.82, P < 0.0001). The correlation, including one-point changes in Glasgow Coma Score, was still quite robust (Spearman r = 0.52, P < 0.0001). Such correlations were present in both traumatic brain injury patients and subarachnoid hemorrhage patients (r = 0.72, P < 0.0001 and r = 0.39, P = 0.0004, respectively; fig. S5). A much weaker correlation with neurological status was seen for ventricular CSF Aβ(r = 0.20, fig. S6).
What underlies this seemingly surprising association between local brain ISF Aβ levels and global neurological status (Fig. 4F)? Our findings in the human brain are consistent with previous results that indicate a direct causal relation between neuronal activity and extracellular Aβ in vitro (2) or brain ISF Aβ concentrations in mice (3, 4). From this perspective, the correlations between brain ISF Aβ concentrations and the other measured parameters fit well; low glucose and high lactate/pyruvate ratio indicative of abnormal brain metabolism, high intracranial pressure, and extremes of brain temperature all would likely result in impaired neuronal function (5). The absence of a correlation with glutamate levels, brain oxygenation, or cerebral perfusion pressure should not be taken as evidence against the neuronal activity hypothesis, because these parameters were rarely outside of their normal ranges in our patients (fig. S3).
Our findings complement the evidence from lesion studies in mice (14) and imaging studies in humans linking neuronal activity and Aβ deposition (15). However, based on prior results, the speed and magnitude of the dynamics seen here were unexpected. Smaller fluctuations in Aβ levels over hours to days have been observed in lumbar CSF from normal subjects (16).
Global neurological status was better correlated with brain ISF Aβ measurements in our patients than with ventricular CSF Aβ. This was surprising because ISF sampled by microdialysis comes from a relatively restricted area of the brain near the catheter, whereas ventricular CSF is derived from diffuse periventricular brain regions. On the other hand, many of the processes involved in acute brain injury and subsequent ICU care (e.g., sedation, anesthesia, etc.; fig. S7) likely affect neocortical neuronal and/or synaptic activity relatively homogeneously. We hypothesize that sampling from white matter adjacent to any cortical region gives a reliable indicator of global neuronal and/or synaptic function. Changes in global neuronal and/or synaptic function are in turn likely to be directly reflected in changes in the Glasgow Coma Score. In contrast, the relatively weak relation with ventricular CSF in these patients may indicate that factors other than neuronal/synaptic activity contribute much of the variance in ventricular CSF Aβ concentrations. The relation between brain ISF Aβ and CSF Aβ is clearly complex and requires further study. Lumbar CSF (16) may be derived from cortical ISF to a greater extent than ventricular CSF, and future studies will be required to address this issue because lumbar CSF was not obtained in these patients.
It is unlikely that there is a direct relation between the reported deposition of insoluble Aβ after traumatic brain injury (17–19) and the changes in brain ISF Aβ levels that we observed. Aβ deposits have been observed in a minority (approximately one-third) of subjects, whereas the rise in ISF Aβ over time was a nearly universal phenomenon (Fig. 1G). Thus, the relation between our findings and the reported epidemiological link between acute brain injury and the later development of Alzheimer’s disease (20, 21) is unclear.
Similarly, it is unlikely that these dynamics are related to axonal injury. If Aβ were being released into the extracellular space as a consequence of immediate axonal injury, we would have expected to see initially high brain ISF Aβ values, which then would have been expected to fall over time. This is the opposite of what we observed. We recognize that an increase of total tissue or intracellular Aβ levels (22, 23) would not have been detected, because microdialysis samples derive only from the extracellular space. Likewise, an early, transient rise in brain ISF Aβ concentrations would not have been detected, because catheters were implanted 12 to 48 hours after injury.
Our results provide direct information on the true extracellular fluid Aβ levels in the living human brain. To our knowledge, all but one of our patients had an acute brain injury, and none had dementia or Alzheimer’s disease. Nonetheless, this information is useful when considering whether levels of exogenous (10, 12, 24, 25) or endogenous Aβin experimental animals (26) are likely to be physiologically relevant in humans.
Figs. S1 to S7