Fax: (216) 445-1466
A noninvasive marker of blood-brain barrier function and brain lesions
Article first published online: 19 MAY 2003
Copyright © 2003 American Cancer Society
Volume 97, Issue 11, pages 2806–2813, 1 June 2003
How to Cite
Kanner, A. A., Marchi, N., Fazio, V., Mayberg, M. R., Koltz, M. T., Siomin, V., Stevens, G. H. J., Masaryk, T., Ayumar, B., Vogelbaum, M. A., Barnett, G. H. and Janigro, D. (2003), Serum S100β. Cancer, 97: 2806–2813. doi: 10.1002/cncr.11409
- Issue published online: 19 MAY 2003
- Article first published online: 19 MAY 2003
- Manuscript Accepted: 11 FEB 2003
- Manuscript Revised: 31 DEC 2002
- Manuscript Received: 4 NOV 2002
- Sangtec Medical Inc. (Bromma, Sweden). Grant Numbers: HL51614, NS43284, NS38195
- blood-brain barrier disruption;
- brain damage;
- magnetic resonance imaging;
- metastatic brain tumor
S100β protein is expressed constitutively by brain astrocytes. Elevated S100β levels in cerebrospinal fluid and serum reported after head trauma, subarachnoid hemorrhage, and stroke were correlated with the extent of brain damage. Because elevated serum S100β also was shown to indicate blood-brain barrier (BBB) dysfunction in the absence of apparent brain injury, it remains unclear whether elevation of serum levels of S100β reflect BBB dysfunction, parenchymal damage, or both.
The authors conducted a prospective study of serum S100β levels in six patients who underwent hyperosmotic BBB disruption (BBBD) with intraarterial chemotherapy for primary central nervous system lymphoma. In addition, 53 serum S100β samples were measured in 51 patients who had a variety of primary or metastatic brain lesions at the time of neuroimaging.
S100β was correlated directly with the degree of clinical and radiologic signs of BBBD in patients who were enrolled in the hyperosmotic study. In patients with neoplastic brain lesions, gadolinium enhancement on a magnetic resonance image was correlated with elevated S100β levels (n = 45 patients; 0.16 ± 0.1 μg/L; mean ± standard error of the mean) versus nonenhancing scans (n = 8 patients; 0.069 ± 0.04 μg/L). Primary brain tumors (n = 8 patients; 0.12 ± 0.08) or central nervous system metastases also presented with elevated serum S100β levels (n = 27 patients; 0.14 ± 0.34). Tumor volume was correlated with serum S100β levels only in patients with vestibular schwannoma (n = 6 patients; 0.13 ± 0.10 μg/L) but not in patients with other brain lesions.
S100β was correlated directly with the extent and temporal sequence of hyperosmotic BBBD, further suggesting that S100β is a marker of BBB function. Elevated S100β levels may indicate the presence of radiologically detectable BBB leakage. Larger prospective studies may better determine the true specificity of S100β as a marker for BBB function and as an early detection or follow-up marker of brain tumors. Cancer 2003;97:2806–13. © 2003 American Cancer Society.
Serum S100β is a low-molecular-weight Ca2+-binding protein composed of two isomeric subunits and is found predominately in astroglial and Schwann cells.1–5 S100β normally is low or is not detectable in serum; however, elevated serum levels have been detected in a number of neuropathologic conditions.6–15 Intraoperative measurements have demonstrated temporary elevations in serum S100β levels during cardiac and carotid artery surgery that investigators believed was caused by cerebral emboli.16, 17 Both the time course and the degree of elevation were associated strongly with neuropsychological outcome after cardiac surgery and head trauma.9, 10, 12 Serum S100β levels also have been correlated with the volume of infracted brain in patients with stroke.18, 19 Although it is recognized that S100β originates from the central nervous system (CNS) and increases in cerebrospinal fluid (CSF) after injury, it remains unclear whether elevation of serum levels of S100β is a sign of blood-brain barrier (BBB) dysfunction or neuronal/parenchymal damage. Our previous experience suggested that S100β levels reflected BBB dysfunction,11, 20, 21 whereas recent work suggested that both BBB leakage and parenchymal lesions can be detected using this serum marker.11 In these studies, BBB leakage was not assessed independently, although we hypothesized that contrast enhancement on neuroimaging would correlate with breaching of the BBB and, consequently, should demonstrate elevated serum S100β levels. This was tested directly in the work presented herein, in which patients who underwent routine or selective magnetic resonance imaging (MRI) or who participated in the intraarterial osmotic BBB disruption (BBBD) protocol for the treatment of primary central nervous system lymphoma (PCNSL) were enrolled to demonstrate a correlation between S100β and clinical and radiologic evaluations of BBB function.
MATERIALS AND METHODS
This prospective study included two patient populations. All patients signed an informed consent form according to Institutional Review Board protocols at The Cleveland Clinic Foundation. The first group consisted of six patients who participated in a treatment protocol that included BBBD22–25 for PCNSL, primitive neuroectodermal tumors, certain gliomas, CNS germinomas, and metastatic brain tumors (breast, small cell lung, or germ cell). These patients were treated with an intraarterial injection of mannitol that caused a temporarily disruption of the BBB; this was followed by a selective, intracarotid methotrexate injection. Patients were studied with serial blood samples at each treatment session. Blood samples were drawn before (10 minutes prior to mannitol injection), during (45 seconds, 1 minute, and 10 minutes after mannitol injection), and after (up to 6 hours) iatrogenic, selective, intraarterial osmotic opening of the BBB.11, 20, 21 A total of 54 BBBD procedures in 6 patients were studied.
The second group consisted of 51 patients. All patients underwent diagnostic or volumetric MRI studies at The Cleveland Clinic Foundation between September 2001 and September 2002. Fourteen consecutive patients had MRI scans for known neurologic disorders or brain lesions. Thirty-nine consecutive patients had volumetric MRI studies before undergoing radiosurgical treatment at The Cleveland Clinic Gamma Knife Center. Fifty-three blood samples were drawn concurrently with MRI studies.
S100β and neuron specific enolase (NSE) levels were measured using techniques described elsewhere.26 Routine MRI studies included T1-weighted sequences with and without gadolinium-diethylene diamine tetraacetic acid, T2-weighted images, and fluid-attenuated inversion recovery (FLAIR) images. Contrast-enhanced, three-dimensional volume acquisition with 1-mm slice intervals and supplementary 2-mm spin echo images through the area of interest were obtained for volumetric MRI studies in patients who were scheduled for radiosurgical therapy. The data sets of volumetric MRI studies were then loaded for tumor volume calculation using Elekta Gamma Knife planning software (Leksell GammaPlan®; Elekta, Atlanta, GA). All MRI scans were analyzed for contrast-enhancement patterns on T1-weighted gadolinium studies and for T2 and FLAIR abnormalities.
Routine immunocytochemical techniques were used to visualize S100β in 30-μm sections that were obtained at biopsy. Biochemical and neuroradiologic analyses were conducted in a blinded and independent manner. Statistical analysis was performed with an analysis of variance using Origin Microcal software (version 6.1).
Six patients (one male patient and five female patients) underwent a total of 54 BBBD procedures. The patients ranged in age from 22 years to 74 years (median, 65.6 years). In all, 324 blood samples were drawn over the treatment/study period. The 51 patients who were enrolled on the MRI-S100β study (20 male patients and 31 female patients) had blood samples drawn at the time of their MRI studies. Patients ranged in age at the time of analysis from 24 years to 80 years (median, 55 years). Tumors among this group of patients included 8 glial tumors, 27 metastatic tumors, 15 extraaxial lesions (8 meningiomas, 6 vestibular schwannomas, and 1 chondroma), 1 patient with a history of PCNSL, and 2 patients with trigeminal neuralgia. Patient data and MRI presentation are summarized in Table 1.
|Patient||S100β (ng/mL)||NSE (mg/L)||Age (yrs)||Gender||Weight (lbs)||Diagnosis|
|3||0.069||10.7||67||F||147||Meningeoma, no tumor|
|6||0.120||17.0||35||F||145||Glioma, tectal, hydrocephalus|
|7||0.063||10.0||71||F||165||Meningeoma, no tumor|
|15||0.280||4.8||49||F||138||Metastasis, breast carcinoma|
|16||0.026||27.2||38||F||140||Metastasis, breast carcinoma|
|17||0.038||12.3||64||F||160||Metastasis, breast carcinoma|
|20||0.001||4.5||71||M||163||Metastasis, esophageal carcinoma|
|23||0.001||8.3||46||M||183||Metastasis, lung carcinoma|
|24||0.350||10.9||55||F||160||Metastasis, ovary carcinoma|
|25||0.300||11.6||62||M||174||Metastasis, lung carcinoma|
|26||0.100||18.8||35||M||160||Glioma, mixed oligoastrocytoma|
|29||0.068||14.4||43||M||206||Metastasis, lung adenocarcinoma|
|30||0.046||6.4||46||M||180||Metastasis, lung carcinoma|
|32||0.019||7.3||42||F||154||Glioma, anaplastic astrocytoma|
|33||0.160||7.9||58||M||175||Metastasis, lung carcinoma|
|34||0.300||14||52||F||170||Metastasis, breast carcinoma|
|35||0.290||5.7||67||F||200||Metastasis, lung carcinoma|
|38||0.180||56.2||42||F||214||Metastasis, breast carcinoma|
|42||0.160||6.1||53||F||141||Metastasis, lung carcinoma|
|43||0.140||5.8||53||F||126||Metastasis, lung carcinoma|
|44||0.210||4.2||59||F||134||Metastasis, lung carcinoma|
|45||0.170||6.9||83||F||155||Trigeminal neuralgia, infarct|
|46||0.060||1.5||62||F||161||Metastasis, endometrial carcinoma|
|47||0.079||4.9||51||F||152||Metastasis, lung carcinoma|
|48||1.700||10.5||48||F||146||Metastasis, breast carcinoma|
|49||0.033||13.9||53||M||173||Metastasis, lung carcinoma|
|52||0.51||24.8||24||F||162||Petrous bone tumor, chondroma|
In patients who underwent iatrogenic BBBD, serum S100β levels increased significantly from baseline after intraarterial mannitol infusion and peaked with the methotrexate infusion thereafter (Fig. 1). No significant changes were observed in serum levels of NSE, which is a marker of brain damage that remained stable over the entire sampling period (data not shown).20 Figure 1A shows the mean results from all procedures performed. Detailed analysis of the differential increase in S100β from baseline levels to BBBD levels revealed that, although most procedures led to significant elevations in S100β levels after mannitol infusion, a small but significant number of BBBD attempts failed to cause comparable changes in serum S100β levels (Fig. 1B). The efficacy of the BBB opening was defined as the serum S100β level after mannitol infusion subtracted from the baseline S100β level and plotted against the baseline S100β level. Note that increases in S100β levels induced by osmotic BBBD were achieved only when basal, pre-BBBD levels were equal to or less than normal the normal level (S100β < 0.1 μg/mL). To underscore this, the data shown in Figure 1A were replotted after purging nonsuccessful attempts to emphasize that, when it was effective, the BBBD procedure led to rapid elevation of S100β levels (Fig. 1C). Based on these findings, we concluded that normal S100β levels were correlated with the efficacy of mannitol to disrupt the BBB in patients with PCNSL. This may have been a consequence of the fact that low S100β levels are indicators of adequate BBB integrity.
Because baseline values for S100β were variable in patients with PCNSL, we sought to determine whether this variability was due to tumor size. Quantitative evaluation of tumor enhancement and subsequent comparison with serum S100β levels did not reveal any significant correlation (data not shown). This may be in disagreement with our hypothesis linking BBB dysfunction and radiologic enhancement with S100β levels. PCNSL tumors, however, do not produce S100β, as demonstrated in immunocytochemical staining for S100β (Fig. 2); thus, the tumor itself does not contribute significantly to serum S100β levels (see also below and Discussion). In contrast, metastatic brain lesions were characterized by normal or elevated S100β expression in perivascular glia, suggesting that, in these patients, BBB leakage may cause abnormal S100β levels. This hypothesis was tested further, as described below.
Patients undergoing routine MRI scans were recruited for the study, and S100β levels were measured in blood samples that were drawn immediately before radiologic observations. In six patients who had a history of previous brain surgery or treatment for a brain tumor, no abnormal findings were found on contrast-enhanced MRI studies. In these patients, serum S100β levels were close to normal (Fig. 3A,A2). Among two patients who were treated for trigeminal neuralgia, one patient had a normal MRI study, and the other patient showed lacunar infarcts and signs of microvascular infarct-related white matter changes. This second patient demonstrated elevated S100β levels. The remaining patients demonstrated gadolinium enhancement and significantly elevated basal S100β levels.
When data were replotted according to tumor pathology, it was found that patients who were diagnosed with metastasis from breast carcinoma (n = 7 patients) (Fig. 3A, A3), lung carcinoma (n = 12 patients), renal cell carcinoma (n = 3 patients), adenocarcinoma of an unknown primary origin, melanoma, ovarian carcinoma, and endometrial and esophageal malignant tumors (n = 1 patient each) had elevated S100β levels, but no significant differences were discovered between patients with tumors of different etiologies. It was found that patients with meningioma had normal S100β levels, except in patients who presented with considerable mass effect and edema in the compressed brain tissue (0.036–0.32 μg/L) (Fig. 3A, A1). One patient with biopsy-proven glioblastoma multiforme had three blood samples taken over a 4.4-month period during which radiologic tumor progression was apparent. The samples showed increases in S100β levels from 0.12 μg/L to 0.27 μg/L (Fig. 4). These changes correlated well with tumor progression.
This is the first systematic, prospective, clinical study to investigate the link between iatrogenic or pathologic disruption of the BBB and serum S100β levels. The results confirm that elevated serum S100β levels are not necessarily associated with brain damage but, rather, reflect ongoing failure of the BBB. An important corollary of this study is the possible use of this serum marker for the early detection of primary and metastatic brain tumors.
The objective of this investigation was to test the hypothesis that S100β represents a marker of BBB integrity independent of structural neuronal/brain tissue damage. Two patient groups were elected to evaluate this hypothesis. In one set of experiments, we correlated S100β levels with MRI findings. The presence of contrast-enhancement MR scans of brain lesions reflects an increased permeability of the BBB, which leads to passage of the paramagnetic agent gadolinium into the perivascular region and shortening of relaxation times.27–29 It seems likely that S100β expressed by astrocytes and present in the perivascular space extravasates into the systemic circulation in areas of BBB impairment. Thus, elevation of this low-molecular-weight protein may be a useful marker of BBB function. However, there are several factors that may influence serum S100β levels, including the availability/secretion of S100β by the adjacent normal brain tissue and the availability/production of this marker by the lesion itself.
The current results demonstrated that enhancement by tumors like PCNSL does not cause elevated S100β levels. In fact, the majority of patients with PCNSL had normal S100β levels before they underwent iatrogenic BBBD (Fig. 1). This fact may appear to contradict our hypothesis, because PCNSL tumors are characterized by substantial enhancement in the brain, presumably due to altered BBB integrity. However, immunocytochemical experiments (Fig. 2) demonstrated that PCNSL tumors are characterized by a remarkably uniform cytology that is devoid of normal astrocytic markers, glial fibrillary acidic protein, and S100β. Thus, it is not surprising that elevations in S100β levels were not common in these patients. It remains to be elucidated whether PCNSL specific markers are extravasated under these pathologic conditions. In summary, our data suggest that brain tissue invaded by PCNSL cannot release S100β, because relatively little is left of the normal glial population. Thus, although, in patients with PCNSL, the BBB indeed is leaky, S100β levels are low. Osmotic disruption verisimilarly allows increased brain-blood interfacing, thus triggering S100β release in plasma. Incidentally, we would like to note that this hypothesis also is supported by the fact that PCNSL tumors frequently are drug-resistant despite the leaky BBB associated with most abnormal brain regions. The peritumoral area, therefore, may be reached by the drug only after further and distal BBBD.
If PCNSL is not the origin of serum S100β, then what accounts for the elevated levels in a subpopulation of patients or after intraarterial mannitol? Or, more generally, what are the mechanisms of S100β extravasation into plasma? In the context of infarcted brain tissue, it has been postulated that necrotic cell death of astroglia and membrane instability in the penumbra region around the ischemia lead to the leakage of (brain) cytosolic S100β into the extracellular space, increasing serum concentrations.30–32 This may not necessarily be the sole mechanism of S100β production by brain tumors. Neoplasms in general, and primary brain tumors in particular, actively secrete S100β.33 Furthermore, neoplasms may affect the BBB by mechanisms different from ischemia, including secretion of a number of cytokines (e.g., vascular permeability factors).34, 35
These considerations are the bases for two different scenarios, one in which the tumor and surrounding tissue actively produce and extravasate S100β (e.g., glioblastoma) and another in which only adjacent normal tissue produces S100β in regions with normal BBB permeability (e.g., lymphoma). BBBD with intraarterial mannitol will cause the breakdown of this normal BBB and will lead to increased serum S100β levels. In the first scenario, the sources of S100β will include both normal and abnormal glia, and the extravasation will be limited only by regions of intact BBB. A third scenario relates to non-CNS tumors that lack a BBB but that express S100β. We examined patients with vestibular schwannoma and found that, as expected, the size of their tumors was correlated with S100β plasma levels. Similar conclusions were drawn in patients with malignant melanoma.6, 8
A number of routes of S100β leakage into the peripheral blood circulation have been suggested. One possible route consists of disruption of the brain-CSF interface, leading to increased levels of S100β in CSF that are reabsorbed into the cerebral venous system. A second, more direct route is provided by disruptions on the capillary level that allow drainage of perivascular S100β directly into the circulation.11, 21 The second route is more likely in patients with brain tumors or other lesions. Finally, markers other than S100β may indicate CSF-to-BBB dysfunction better.
All of the current findings suggest that serum S100β represents a marker of BBB integrity rather than brain-neuronal damage. In conclusion, S100β may be useful as a noninvasive, peripheral measurement of BBB function in patients with brain lesions. To date, no accepted, accurate, and predictive method has been defined to determine BBB function. Alternative methods include invasive CSF sampling or more expansive neuroimaging methods. Other useful applications for a panel of BBB integrity could include screening for brain lesions, for example, in patients with systemic malignancies, or follow-up of patients with known primary brain tumors.
Serum S100β is related directly to the extent and temporal sequence of BBBD and may qualify as a potential marker for BBB function. S100β and NSE levels may be used concomitantly to differentiate brain damage from BBB leakage. S100β could be used as an objective measure for the evaluation of therapeutic BBBD and also may be useful as a screening tool for further evaluation of patients who are at risk for brain tumors associated with BBBD. Larger prospective studies may better determine the true specificity of S100β as a marker for BBB function and early detection of brain tumors.
The authors acknowledge the continuous support of Sangtec Medical Inc. (Bromma, Sweden).
- 11Peripheral markers of brain damage and blood-brain barrier dysfunction. Restorative Neurosci. In press., , , , , .