| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
* Departments of Neuroscience and Pharmacology, Mayo Clinic, Mayo Clinic College of Medicine, Jacksonville, Florida, USA; and
Department of Chemical and Materials Engineering, Arizona State University, Tempe, Arizona, USA
1Correspondence: Department of Neuroscience, Mayo Clinic Jacksonville, Birdsall 210, 4500 San Pablo Rd., Jacksonville, FL 32224, USA. E-mail: tgolde{at}mayo.edu
ABSTRACT
A number of hypotheses regarding how anti-Aß antibodies alter amyloid deposition have been postulated, yet there is no consensus as to how Aß immunotherapy works. We have examined the in vivo binding properties, pharmacokinetics, brain penetrance, and alterations in Aß levels after a single peripheral dose of anti-Aß antibodies to both wild-type (WT) and young non-Aß depositing APP and BRI-Aß42 mice. The rapid rise in plasma Aß observed after antibody (Ab) administration is attributable to prolongation of the half-life of Aß bound to the Ab. Only a miniscule fraction of Ab enters the brain, and despite dramatic increases in plasma Aß, we find no evidence that total brain Aß levels are significantly altered. Surprisingly, cerebral spinal fluid Aß levels transiently rise, and when Ab:Aß complex is directly injected into the lateral ventricles of mice, it is rapidly cleared from the brain into the plasma where it remains stable. When viewed in context of daily turnover of Aß, these data provide a framework to evaluate proposed mechanisms of Aß attenuation mediated by peripheral administration of an anti-Aß monoclonal antibody (mAb) effective in passive immunization paradigm. Such quantitative data suggest that the mAbs are either indirectly enhancing clearance of Aß or targeting a low abundance aggregation intermediate.—Levites, Y., Smithson, L. A., Price, R. W., Dakin. R. S., Yuan, B., Sierks, M. R., Kim, J., McGowan, E., Reed, D. K., Rosenberry, T. L., Das, P., Golde, T. E. Insights into the mechanisms of action of anti-Aß antibodies in Alzheimer’s disease mouse models.
Key Words: immune complex • amyloid deposits
THERE IS COMPELLING evidence that aggregation and accumulation of Aß play a pivotal role in the development of Alzheimer’s disease (AD). Numerous strategies to prevent Aß aggregation and accumulation are being evaluated as ways to treat or prevent AD, and a select number of these are now entering the clinic (1)
. Preclinical studies in APP transgenic mice demonstrate the therapeutic potential of altering Aß deposition by inducing a humoral immune response to fibrillar Aß42 (fAß42) or passively administering anti-Aß mAbs (2
, 3)
. A human clinical trial of active immunization with fAß42+QS-21 adjuvant (AN-1792) was halted due to a meningio-encephalitic-like presentation in 
6% of individuals (4
5
6)
. No definitive data regarding the nature of the meningio-encephalitic presentation have been published, but the leading hypothesis, supported by some recent experimental data, is that it was attributable to an autoreactive T cell response against Aß (7)
. Reports of individuals enrolled in the now discontinued phase II trial suggest that those subjects who developed robust anti-Aß amyloid Ab titers did show some clinical benefit relative to subjects that did not develop robust titers (4
, 8
, 9)
. Moreover, an anecdotal report of a small phase II study of AD patients administered human intravenous infusion of immunoglobulin containing anti-Aß Abs showed slight improvement in ADAS-cog after administration (10)
. Because of fears of the possible side effects of active vaccination, passive immunization with humanized anti-Aß mAbs is being vigorously pursued as an alternative approach. One humanized anti-Aß mAb is in a phase II trial (http://www.elan.com/research%5Fdevelopment/Alzheimers), and it is likely that additional humanized anti-Aß mAbs will be tested in humans in the very near future. Thus, animal modeling studies and preliminary human data suggest that efforts to develop better active vaccination and passive immunization strategies are warranted.
Despite multiple studies examining various parameters that may predict the efficacy of active or passive anti-Aß immunotherapy in mice, there is still no consensus on how either form of Aß immunotherapy works (11
, 12)
. As passive administration of anti-mAb antibodies works as effectively as active immunization in APP mice, it is generally acknowledged that it is the anti-Aß Ab response that mediates the effects of active immunization. The polyclonal response to active vaccination with fibrillar Aß peptides generates multiple Abs with varying degrees of binding specificity for soluble Aß, preamyloid aggregates, and Aß amyloid (3
, 13
14
15
16
17)
. Thus, it is potentially misleading to use active vaccination studies to predict precisely which type of Ab is mechanistically associated with efficacy. Nevertheless, in certain active and passive immunization studies, the efficacy of immunization correlates with the ability of the immune serum or anti-Aß mAbs to recognize Aß deposited as amyloid (3
, 14
, 18)
. In contrast, certain mAbs that are effective at reducing Aß loads and other AD-like pathologies do not bind Aß amyloid (19)
. Such data raise the possibility that the binding properties of the mAbs with respect to various forms of Aß have not been sufficiently characterized to enable identification of the common target or that there may be multiple ways in which anti-Aß antibodies can influence amyloid deposition and other AD-like pathologies.
The amount of Aß deposited when immunization is initiated, the APP mouse model used, the methodology used to measure differences in amyloid loads, and the properties of the mAbs used for passive immunization all affect the outcome of passive immunization (2
, 3
, 13
, 14
, 20)
. In PDAPP mice, different types of Aß plaques are reported to be more easily altered by immunization (21)
. After immunization of PDAPP and other APP mice with abundant diffuse plaques, much larger reductions in immunohistochemical amyloid loads are reported than biochemical loads as measured by ELISA, suggesting a preferential reduction of diffuse Aß deposits (2
, 19
, 22)
. Moreover, at least in terms of percent reduction in amyloid loads, most studies show much more dramatic effects in mice initially treated when they have only minimal plaque deposition (13
, 18
, 23)
. Even more puzzling, some mAbs that work well when administered before plaque deposition do not work at all once plaque deposition has begun, whereas other mAbs work in both settings (18)
. Finally, in some reports it can be shown that intact mAb is not required for efficacy: Fab and scFv fragments work (24
, 25
; Levites and Golde, unpublished observations), as well as Aß binding proteins/gangliosides (26)
. Such data suggest that the binding of Aß by Ab, but not effector functions of the Ab, are required for efficacy.
A great deal of debate with respect to mechanism also centers on peripheral vs. central action of an anti-Aß Ab (19
, 27)
. The peripheral sink hypothesis proposes that binding of Aß in the blood enhances efflux from the brain. Several studies have shown that plasma Aß levels increase dramatically after both active and passive Aß immunotherapy and that at least some of the Aß in the plasma is complexed to the anti-Aß antibodies (19
, 27)
, but it is not clear whether this peripheral binding accounts for a decrease in brain deposition. Another issue relates to whether microglia activation contributes to efficacy; current data on the role of microglia are inconsistent. Several groups report transient or stable enhancements of microglial activation associated with Aß removal; others do not (2
, 28
29
30)
. At least in Tg2576 APP mice, a role for enhanced phagocytosis of mAb:Aß complexes via the FcR has been ruled out (30)
. In postmortem human tissue from several AD patients who had received the AN-1792 vaccine, Aß laden microglia are noted in areas where Aß clearance is hypothesized to have occurred (6
, 31)
.
To provide a quantitative framework in which to analyze proposed mechanisms of action of anti-Aß immunotherapy, we have measured the acute effects on plasma, cerebrospinal fluid (CSF), and brain Aß after a single intraperitoneal dose of several anti-Aß mAbs. For the initial studies, we chose an anti-Aß1–16 IgG2a, (mAb9; ref 18
). mAb9 recognizes both monomeric and aggregated Aß and Aß amyloid. We also performed a more limited set of studies using an anti-Aß1–16 IgG1 (mAb3), an anti-Aßx-40 IgG1 (mAb40.1), and an anti-Aßx-42 (mAb42.2) (18)
. To avoid potential confounds introduced by the presence of Aß deposits, these studies were conducted in young APP or BRI-Aß42 transgenic mice prior to Aß deposition (32
, 33)
. We also measured mAb9 levels and estimated the half-life of the mAb9:Aß complex in nontransgenic mice. We chose these mAbs for study, as we have previously shown that 500 µg of each chronically administered intraperitoneally every 2 wk can significantly attenuates Aß deposition (18)
. These data show that in mice 1) binding of mAbs to Aß significantly prolongs the half-life of plasma Aß from minutes to days, 2) very little free anti-Aß mAb actually enters the brain or CSF, 3) anti-Aß mAb:Aß complexes are rapidly cleared from the brain, 4) that passive administration of these anti-Aß mAbs has little effect on total steady-state predeposition brain Aß levels, and 5) that neither the amount of mAb relative to the amount of Aß or type of anti-Aß mAb significantly influence the overall changes in Aß induced acutely after passive immunization. Such data, when viewed in the context of the total amount of Aß turnover in a human or mouse per day, have important implications with respect to design and interpretation of future studies using active or passive immunotherapy as a treatment or preventive strategy for AD.
MATERIALS AND METHODS
Antibodies
The anti Aß1–16 specific mAb9 (IgG2a) and mAb3 (IgG1) used for immunizations as well as anti-Aß40 specific mAb40.1 (IgG1) and anti-Aß42 specific mAb42.2 (IgG1) used for ELISAs were characterized previously (18)
. Biotinylation was performed according to the manufacturer. Briefly, 0.27 µmol of sulfo-NHS-LC-biotin (Pierce, Rockford, IL) was added to 2 mg mAb9 or mouse IgG and incubated for 2 h at RT, followed by purification of labeled protein over desalting column. 4G8, human Aß17–14 epitope was obtained from Signet (Dedham, MA). Mouse IgG was obtained from Equitech-Bio Inc. (Kerrville, TX).
Mice
All animal husbandry procedures performed were approved by the Mayo Clinic Institutional Animal Care and Use Committee in accordance with National Institutes of Health guidelines under protocol A34602. Tg2576 mice and BRI-Aß42B mice were generated and confirmed by genotyping as described previously (32
, 33)
. All animals were housed 3–5 to a cage and maintained on ad libitum food and water with a 12 h light/dark cycle.
Binding kinetics
Affinity measurements were performed using a BIAcore X biosensor (BIAcore Inc., Piscataway, NJ). A CM5 sensor chip (BIAcore) was activated as recommended by the manufacturer using an equimolar mix of N-hydroxysuccinimide (NHS) and N-ethyl-N'-(dimethylaminopropyl)carbodiimide (EDC), immobilized with 50 µl of a capture Ab (BR100514, 100 µg/ml in 10 mM NaAcatate, pH 4.8), and then blocked with ethanolamine; 70 µl of the mAb (diluted in running buffer (HBS-EP) at 100 µg/ml) were injected onto the immobilized chip. The association and dissociation rate constants (ka and kd) were determined using an Aß concentration range with HBS-EP [0.01 M HEPES, 0.15 M NaCl, 3 mM EDTA, 0.005% (v/v) surfactant P20, pH 7; BIAcore, Uppsala, Sweden] as a running buffer at a flow rate of 10 µl/min. The sensor surface was regenerated using 10 mM Glycine-HCl, pH 1.5. Kinetic parameters were evaluated using BIA evaluation 3.1 software (BIAcore).
Passive immunizations
Young Tg2576 mice or nontransgenic controls (3 month old, n=4 per group) were given a single intraperitoneal dose of 500 µg (1600 pmol) biotinylated mAb9. Control mice received biotinylated mouse IgG or PBS.
Intracerebroventricular injections
For stereotactic intracerebroventricular injections, nontransgenic mice (females, 3 monthold, n=2 per group) were injected with preformed complex of 50 µg (160 pmol) of biotinylated mAb9 and 320 pmol of Aß in the left cerebral ventricle. On the day of the surgery, mice were anesthetized with isoflurane (5% induction and 3% maintenance) and placed in a stereotactic apparatus. A midsagittal incision was made to expose the cranium, and a hole was drilled to the after coordinates taken from bregma: A/P, –0.4 mm; L, –1.0 mm. A 26 gauge needle attached to a 10 µl syringe was lowered 1.8 mm dorsoventral, and a 4 µl injection was made over 10 min. The incision was closed with surgical staples, and the mice were killed at various time points after the surgery.
Measurement of mAb9, Aß or Aß40:mAb9 complexes in plasma
Groups of female Tg2576 mice or their nontransgenic littermates were immunized with biotinylated mAb9 and plasma was collected at various time points. Control mice received biotinylated mouse IgG or PBS. To measure the Aß40-biotinylated mAb9 complex in the plasma capture, ELISA was used with an Ab against free end of Aß40 peptide, mAb40.1 (2.5 µg/well), as capture and Neutravidin-HRP, 1:2000, as detection. For standards, we saturated mAb40.1-coated plate with Aß (5 µg/well), applied increasing amounts of biotinylated mAb9, and detected with Neutravidin-HRP. Control PBS injected plasma was spiked with 500 µg mAb9 to determine the basal levels of Aß capable to bind mAb in the plasma. To determine the level of total Aß40, we used mAb40 as capture and 4G8, 1:2000, as detection. In non-Tg mice, levels of biotinylated mAb9 were determined by direct ELISA with Aß40 (5 µg/well) as capture and Neutravidin-HRP as detection. Additionally, 1 ml plasma pooled from 3 mice 24 h after the administration of biotinylated mAb9 or biotinylated mouse IgG was fractionated on a 1 x 30 cm Superose 6 PC 3.2/30 column (Amersham Biosciences, Piscataway, NJ). Superose columns were routinely pretreated with a bolus of BSA (50 mg) in running buffer to block nonspecific binding followed by a wash with at least 4 column volumes of running buffer. Aß40 in each fraction was measured using capture ELISA as described above.
ELISA analysis of extracted Aß from the brain
At the time of death, the brains of mice were divided by midsagittal dissection, and both hemibrains were used for biochemical analysis. One hemibrain was homogenized in TBS with CompleteTM protease inhibitors (150 mg/ml wet wt) while the other hemibrain was homogenized in radio-immunoprecipitation assay (RIPA; 50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% Triton x-100, 1% Sodium deoxycholate, 0.1% SDS) with CompleteTM protease inhibitor. Homogenates were than centrifuged at 20,000 g for 1 h at 4°C, the resultant supernatant was collected, representing the TBS- or RIPA-soluble fraction, respectively. Additionally, a hemibrain was homogenized in guanidinium extraction buffer (GuHCl, 5M Guanidine and 50 mM Tris-HCl) and incubated at room temperature for 4 h, representing GuHCl fraction. The after mAbs against Aß were used in the sandwich capture ELISA: for brain Aß40, mAb40.1 capture and 4G8-HRP detection; for brain Aß42, mAb42.2 capture and 4G8-HRP detection. To determine the amount of biotinylated mAb in the brain, direct ELISA with Aß40 as capture and Neutravidin-HRP as detection was used.
Collection of cerebrospinal fluid
The procedure was performed according to that described by Vogelweid et al. (34)
. Briefly, mice were anesthetized with 2.5% Avertin IP. The fur of the animal was clipped and placed in ventral recumbence over a gauze roll (attached to a 13x10x6 cm support) allowing the head to lie at a 45 degree angle. A small strip of transpore tape was used to hold the head in place. A midline incision starting at the base of the pinnae and continuing for 1 cm caudal was made with a #10 blade. Iris scissors were used to separate the muscle layers of the "pocket" 2 mm below the caudal edge of the occipital bone down to atlas. The underlying layers were bluntly separated with microdissecting forceps and retracted with bull clamps to visualize the dura mater, an opaque triangular-shaped membrane. If microhemorrhaging occurred during dissection, the window was blotted gently with an absorbent triangle to clear the area. An 18 gauge needle was guided to gently pierce the dura mater over the cisterna magna followed by immediate replacement with a pulled pipette (and aspirating bulb) to collect the CSF. The CSF was transferred to a gas tight screw cap vial and stored at –80C.
Measurement of Aß and Aß40-monoclonal Ab complex in CSF
To measure the Aß40-biotinylated mAb9 complex in the CSF capture, ELISA was used with an Ab against free end of Aß40 peptide, mAb40.1 as capture, and Neutravidin-HRP as detection. To determine to level of total Aß we used mAb40.1 as capture and 4G8-HRP as detection.
Statistical analysis
One-way ANOVA followed by the Dunnet’s multiple comparison test was performed using the GraphPad Prism version 4 software.
RESULTS
Peripheral administration of anti-Aß mAb creates a stable mAb:Aß complex in the plasma
Previous studies have established that Aß has a very short half-life in the plasma. When free Aß is injected intravenously into the animal, it is cleared with a half-life of <10 min (35
, 36)
. Such data are consistent with our finding that intraperitoneal administration of a single 20 mg/kg of dose of a 
-secretase inhibitor to Tg2576 mice can reduce plasma Aß by 80% within 1 h and by >98% within 5 h, indicating that even endogenous plasma Aß has a short half-life (data not shown). To study changes in Aß levels induced by passive immunization with an anti-Aß mAb as well as the in vivo binding properties and plasma half-life of the mAb itself, 500 µg (1600 pmol) of biotinylated mAb9 was administered intraperitoneally to 3 month old nondepositing female Tg2576 mice. Plasma Aß levels were analyzed by capture ELISA over an extended time course. To ensure that the biotinylated mAb9, which recognizes Aß1–16, did not interfere with detection of Aß by ELISA, Aß was captured with end specific anti-Aß mAbs and detected with HRP-conjugated 4G8, which recognizes a nonoverlapping epitope on Aß. In pilot studies with synthetic Aß standards, mAb9 did not interfere with Aß detection in end specific capture 4G8 detection ELISAs. After biotinylated Ab9 administration, within 1 day after administration, Aß40 in the plasma increased 15-fold, from 50 pmol/ml in untreated mice to almost 750 pmol/ml and and Aß42 levels increased 25-fold, from 2 pmol/ml in untreated mice to almost 55 pmol/ml, respectively. Plasma Aß levels then slowly decreased over an extended period of time to near basal levels by 14 days (Fig. 1
A). To examine the extent to which mAb binding of Aß causes an increase in plasma Aß, we detected biotinylated mAb9:Aß complexes in plasma using a modified ELISA. The biotinylated mAb9:Aß complex is captured with an Aß40 specific mAb and the complex detected with Neutravidin-HRP. The amount of the biotinylated mAb9:Aß40 complex reached its highest value of 450 pmol mAb9 bound to Aß40 per ml of plasma after 6 h (Fig. 1B
). The complex appears to be quite stable with a half-life of 7 days. Although the difference in standardization methods between ELISA measurements of plasma Aß and plasma biotinylated mAb9:Aß complexes introduce some uncertainty with respect to the levels of "total" plasma Aß relative to the level of biotinylated mAb9:Aß, a comparison of the peak levels of total Aß and mAb:Aß complex would suggest that the majority of Aß is bound to the mAb. Consistent with these data, we find that preclearing the plasma with protein A/G removes >90% of the ELISA signal (data not shown). Size exclusion column chromatography of mouse plasma collected 1 d post mAb9 injection shows that most of the plasma Aß that accumulates after mAb9 treatment is present in a high molecular weight fraction with a peak level in a fraction that corresponds to the peak fraction in which unbound mAb9 elutes (Fig. 1C
). In the plasma from the mice injected with biotinylated mouse IgG, the levels in most fractions are much lower and Aß appears to be broadly distributed presumably because, as previously reported, it is bound to numerous serum proteins (37
38
39)
. Finally, when plasma from biotinylated mAb9 injected mice is precipitated with Streptavidin beads and subjected to Western blot analysis, an increase in a 4 kDa Aß species is observed (Fig. 1D
)
|
The half-life of an IgG2a Ab in mouse plasma has been reported to be 1 wk (40)
. When 500 µg (1600 pmol) of biotinylated mAb9 are administered to 3 month old female nontransgenic littermates of the Tg2576 mice, 800 pmol mAb9/ml plasma can be detected in the plasma 1 d later. The biotinylated mAb9 is quite stable and appears to have a half-life of 5–7 d (Fig. 1E
). Collectively, such data suggest that the increase in Aß levels is attributable to binding and stabilization of Aß by the anti-Aß mAb. To directly determine if binding of the mAb9 to Aß prolongs the half-life of Aß, we administered via i.p. injection a preformed complex of biotinylated mAb9 (500 µg, 1600 pmol) and human Aß40 (3200 pmol) into young nontransgenic mice. The mAb9:Aß40 complex was detected as described previously. As mAb9 does not recognize mouse Aß, these studies are not confounded by mAb interaction with endogenous mouse Aß. Within 6 h, 500 pmol/ml of the complex is detected and the complex, like the unbound Ab, is cleared slowly with a half-life of 5–7 d (Fig. 1F
). Thus, in contrast to endogenous Aß, the mAb9:Aß40 complex has a prolonged half-life. In addition, these studies suggest that the binding of the mAb to Aß does not result in the formation of a classic immune complex that would be rapidly cleared. Finally, such data suggest that in plasma the tight binding of mAb9 to Aß (Kd is estimated by surface plasmon resonance to be 3.5e-9 M) prevents the bound Aß from being rapidly turned over.
Effects of acute immunization with anti-Aß mAb on Aß levels in the brains of Tg2576 and BRI-Aß42B mice
To determine if alterations in brain Aß occur after peripheral immunization, we examined the effects on brain Aß in young female Tg2576 mice for up to 2 wk after intraperitoneal administration of 500 µg of biotinylated mAb9. To reduce interference from vascular Aß and mAb9, we extensively perfused the mice with PBS prior to brain harvest. Aß40 and Aß42 levels were measured by ELISA in separate TBS, RIPA, and 5 M guanadinium hydrochloride (GuHCl) fractions. In these studies and as previously reported, GuHCL extracts the highest levels of Aß from the brain (41)
, and despite the marked accumulation of plasma Aß at the 6 and 24 h time points, there is no appreciable change in the levels of GuHCl-extractable brain Aß40 or Aß42 (Fig. 2
A). TBS extracts presumably reflect levels of soluble Aß and contain much lower amounts of Aß than are present in the GuHCl extract (Aß40 6–7% and Aß42 2–3% of GuHCL extract Aß levels). TBS-extractable Aß40 and Aß42 increase slightly after peripheral administration, though the absolute level of increase is small, 5–10% of control values, and does not reach statistical significance by ANOVA (Fig. 2B
). RIPA, a moderately denaturing detergent mix, extracts a higher level of Aß40 and Aß42 than TBS but lower levels than GuHCl (Aß40 25–30% and Aß42 8–10% of the GuHCL extract Aß levels). RIPA-extractable Aß decreases slightly after immunization by 20% of control or 10 pmol/g (Fig. 2C
). No statistically significant decrease in RIPA-soluble Aß40 levels is detected up to 14 d after a single mAb administration (Fig. 2D
). Moreover, the slight decrease observed 24 h after the single mAb administration is not additive, since continuous weekly administration of 500 mg mAb for 4 wk results in similar slight but not significant decrease in RIPA-soluble Ab levels (Fig. 2E
).
|
Tg2576 mice make large amounts of Aß both peripherally and in the brain. In nondepositing Tg2576 mice, this Aß is rapidly turned over. The half-life of Aß in brain is estimated to be 1–2 h (42
43
44)
. Indeed, our studies on mAb9 binding to plasma Aß in Tg2576 mice suggest that after peripheral immunization, mAb9 is saturated with Aß within 6–12 h of administration. Thus, the small changes in brain Aß observed in Tg2576 mice immediately after mAb9 administration might be amplified if more mAb were administered or if the same amount of mAb was administered to a transgenic mouse, which produces much lower levels of Aß. Because we were already delivering an amount of mAb that was near the maximal tolerated dose, we administered the same amount of biotinylated mAb9 to a low expressing BRI-Aß42B line (32)
. This line of BRI-Aß42B mice only expresses Aß42 and has 5-fold lower levels of total brain Aß and 100 fold lower plasma levels relative to Tg2576 mice. At 3 months of age, these mice do not have detectable Aß deposits. After mAb9 administration we again observe a rapid increase in Aß levels in the plasma from 0.5 pmol/ml in untreated mice to 7 pmol/ml at 3 h and 30 pmol/ml 1 d after immunization (Fig. 3
A). The amount of the biotinylated mAb9:Aß42 complex increases in parallel (data not shown). There was no significant change in total brain Aß42 levels extracted by GuHCl (Fig. 3B
), a slight, nonsignificant increase in TBS-extractable brain Aß42 levels (total increase 15%; Fig. 3C
), and a slight nonstatistically significant decrease in RIPA-extractable Aß42 (25%; Fig. 3D
). The magnitude of these changes is similar to that seen in Tg2576 mice, indicating that the small effects induced by mAB9 administration are not influenced to any great extent by the relative amount of plasma or brain Aß in the different transgenic lines.
|
Brain levels of mAb9 after acute peripheral administration of anti-Aß mAb
In previous studies we failed to detect anti-Aß mAb binding to plaques after peripheral anti-Aß mAb administration using immunohistochemical techniques. Others, however, have reported that, consistent with previous reports of blood brain barrier (BBB) penetrance of Abs, a small fraction of anti-Aß mAbs can penetrate the BBB (if quantified levels are <0.1% of total dose; refs 3
, 27
, 45
). After administration via intraperitoneal injection of 500 µg (1600 pmol) biotinylated mAb9 to nontransgenic mice, we can detect 1.0 ± 0.08 fmol/mg of biotinylated mAb9 6 h postinjection, which is 300 fmol per brain or 0.02% of the total amount of the Ab administered. The levels of Ab fall by 24 h to 0.53 ± 0.06 fmol/mg and by 2 wk the levels are 0.06 ± 0.01 fmol/mg. Even lower levels of mAb9 were detected in the Tg2576 brain (data not shown). Despite extensive perfusion, it is impossible to determine whether these trace amounts of mAb9 are truly in the brain or simply stuck to the cerebral vessels; multiple attempts to detect the mAb in situ in the brain sections using immunohistochemical techniques gave negative results. In any case, such data place an upper limit on the amount of mAb9 present in the brain at the time the plasma mAb levels are near maximal.
Effects of anti-Aß mAb on CSF Aß and clearance of mAb9:Aß complexes from the brain
We also examined the levels of Aß and biotinylated mAb9:Aß complexes in the CSF after intraperitoneal administration of mAb9 to Tg2576 mice. Six hours post-mAb injection, a 6-fold increase in Aß40 and a 2-fold increase in Aß42 levels is observed in CSF collected from the cisterna magna. This result contrasts with plasma Aß levels, which peak at 6 h post mAb injection and remain at a relatively stable baseline over 24–72 h (Fig. 1A
); CSF Aß levels decrease rapidly toward control levels by 24 h (Fig. 4
A). Low levels of biotinylated mAb9:Aß complexes are also detected in the CSF, and change in parallel with Aß levels (Fig. 4B
). Unlike in plasma, where Aß levels are roughly comparable to the levels of mAb9 bound to Aß, in CSF there is 50-fold more Aß than mAb bound to it. One possible explanation for this high ratio of Aß to mAb would be that mAb is bound to an Aß aggregate in CSF. The concentration of the mAb9:Aß complex in the plasma remain unchanged during this period, suggesting there may be rapid export of the mAß9:Aß complex from the CSF. To explore this possibility, we injected intracerebroventricular a preformed complex of 5 µg (160 pmol) of biotinylated mAb9 and 320 pmol of Aß. After injection into the ventricles, the biotinylated mAb9:Aß complex is detected in CSF collected from the cisterna magna within 30 min. By 3 h, the levels are dramatically decreased and at 24 h no complex is detectable (Fig. 4C
). In contrast, the low levels of complex appear in plasma by 30 min and appear relatively stable up to 72 h post injection. Such data suggest that even though the anti-Aß mAb:Aß complex has a long-half-life in the plasma, the complex is rapidly cleared from the CSF, and at least some of this clearance is via export into the vasculature.
|
Additional anti-Aß mAbs have similar effects on Aß levels in plasma, brain and CSF of Tg2576 mice
To determine if the observed dynamics in plasma, CSF, and brain after an acute dose of mAb in TG2576 mice are common to the other anti-Aß mAb characterized in our previous studies and shown to reduce Aß deposition after peripheral administration, we injected 500 µg biotinylated anti-Aß1–16 mAb3, anti-Aß42 mAb 42.2, and anti-Aß40 mAb40.1 to 3 month old Tg2576 mice (18)
. Like mAb9, mAb3 administration results in an 7-fold increase in Aß40 and 20-fold increase in Aß42 levels in plasma (Fig. 5
A), but only a slight, nonsignificant decrease in Aß40 levels in RIPA-soluble brain extracts and no effect on RIPA-soluble Aß42 levels (Fig. 5B
). mAb40.1 and mAb42.2 are end-specific antibodies that have been shown to selectively bind Aß40 and Aß42, respectively, in vivo. To avoid interference by the end-specific mAbs present in the plasma, in the ELISAs we only measured total Aß levels using mAb9 as capture and mAb 4G8-HRP as detection. Both end-specifc mAbs caused an increase in total Aß levels in plasma 6 and 24 h after the injection. Higher levels of plasma Aß accumulated after administration of mAb40.1, then mAb42.2, presumably because the mAbs are end-specific; thus the "total" Aß level reflects the relative abundance of these species in the plasma. No effect was observed on the brain RIPA-soluble Aß (Fig. 5D
and E). Aß levels in CSF were also increased on administration of all three mAbs, although the dynamics of this increase vary between the antibodies (Fig. 5C
and F).
|
DISCUSSION
These data provide a quantitative framework in which to consider possible mechanisms of action of an anti-Aß mAb with respect to decreasing Aß deposition. Before fully exploring the possible implications of the current findings, it is important to consider the data in light of estimates of the total amounts of Aß that are synthesized and cleared in Tg2576 mice in a similar time period (see Table 1
). There is extensive evidence that Aß is rapidly metabolized in vivo (35
, 36
, 42
43
44)
. In the brain of nondepositing Tg2576 and other APP mice, the half-life is estimated to be <2 h, and in plasma the half-life is estimated to be 5–10 min. As measured by our ELISA systems, the steady-state levels of Aß in the GuHCL extract from Tg2576 brains is >100 pmol/gm and the steady-state level in plasma is 50 pmol/ml. Assuming a conservative 2 h half-life for brain Aß and total brain wt of 0.4 g, 10 pmol of Aß are cleared from the brain every hour and 240 pmol of brain Aß are cleared every day. Again, assuming a conservative 10 min half-life and a plasma vol of 1 ml, 25 pmol of Aß are cleared every 10 min and 3600 pmol of plasma Aß are cleared per day. Thus, in Tg2576 mice we estimate total Aß turnover at 4 nmol per day, with a great deal more Aß cleared in the periphery.
|
The 500 µg dose of anti-Aß mAbs that we have used in these studies and our previous long-term studies represents 1600 pmol of mAb and would be predicted, if 100% bioavailable, to maximally bind 3200 pmol of Aß (18)
. A more likely estimate is that after intraperitoneal injection no more than 50% of the anti-Aß mAb is bioavailable in the plasma. Because the binding of mAb9 to Aß significantly extends the half-life of plasma Aß without altering the half-life of the mAb, it is clear that mAb9 binding of plasma Aß could sequester the total amount of Aß normally cleared in the periphery for a maximum of 12–24 h. Our data showing that the biotinylated mAb9:Aß complex reaches peak levels (of 500 pmol of mAb9) 6–12 h after intraperitoneal administration are entirely consistent with these estimates. Thus, the large increase in plasma Aß observed after peripheral administration is easily attributable to peripheral binding of Aß by the mAbs.
Consistent with previous studies of mAb penetrance into the brain, our current studies indicate that the amount of anti-Aß mAb that enters the brain is <0.1% of the total mAb injected (3
, 27
, 45)
. If we assume 50% bioavailability, then the maximum amount of anti-Aß mAb that can reach the brain using the 500 µg dose is 1.6 pmol. To be able to influence Aß deposition, it is reasonable to assume that the mAb must be free and not have previously bound Aß in the periphery. As mAb9, which is our most effective anti-Aß mAb in long-term peripheral immunization studies, is saturated in vivo by Aß within 6–12 h and appears to have a half-life of 3 h in the CSF, it would seem that the upper limit of Aß bound by the mAb in the brain before saturation would be 6.4 pmol, which is equal to the amount of mAb cleared from the brain in the initial 12 h after mAb dosing (assuming two binding sites for Aß per mAb). If, as we calculate above, the amount of Aß cleared from the brain is 240 pmol per day, then the maximum amount of Aß the mAb could sequester and potentially clear is <3–4% of the brain Aß produced in a day. As our empirical data show that the amount of anti-Aß mAb in the brain is in fact even less then the 1.6 pmol estimate, it is likely that the mAb would bind an even smaller percentage of the total brain Aß produced per day. Although it is more difficult to precisely determine the level of anti-Aß mAb present after active immunization, our previous studies suggest that the steady-state level of anti-Aß IgG in mice immunized with fibrillar Aß42 is 10 µg/ml (13)
. As long as the half-life of the endogenously produced anti-Aß antibodies is not altered by Aß binding, the relative binding capacity of the total pool of polyclonal anti-Aß after active immunization will be at least an order of magnitude less then the binding capacity of the 500 µg bolus of mAb9. Thus, only trace amounts of free anti-Aß Ab are likely to cross the BBB after active immunization and would be predicted to have minimal impact on steady-state Aß levels in nondepositing mice.
Anti-Aß mAb binding to Aß in the plasma has been proposed to create a peripheral sink that enhances clearance of Aß from the brain (19
, 27)
. If such a mechanism was at work, it should be possible to see reduced levels of total Aß in the brain after peripheral administration of the mAb. Moreover, the decrease would be maximal during the time in which free anti-Aß mAb is present in the plasma. In Tg2576 mice dosed with 500 µg of mAb9, the mAb is fully bound with Aß by 6–12 h. At no time point post-dosing do we observe significant changes in total GuHCl extractable Aß levels in the brain. Perplexingly, we do note subtle, but not statistically significant, changes in the TBS- and RIPA-extractable Aß levels at 6 h post-dosing with mAb9, mAb3 and mAb40.1. TBS-extractable Aß levels seem to increase whereas RIPA-extractable levels transiently decrease at 6 h, but return to control levels by 24 h and remain stable for 2 wk. Moreover, the effect of mAb9 administration is not additive, repeated dosing does not result in a larger decrease in brain levels. When an identical dose of mAb9 was delivered to BRI-Aß42B mice, which, before deposition, have much lower plasma and brain Aß levels then Tg2576 mice, we observed similar trends: a small decrease in the RIPA-extractable pool of Aß at 6 and 24 h and a small increase in the TBS extractable pool of Aß. Given that the stoichiometry between the injected mAb and Aß is quite different in Tg2576 and BRI-Aß42B mice, one would expect that if the subtle alterations in brain Aß were attributable to a simple mass action effect of the mAb on total Aß, much larger changes in the RIPA and TBS pools of brain Aß would be observed in the BRI-Aß42B mice.
The dynamic changes in Aß levels and mAb levels are somewhat different between the CSF and plasma, and also differ between the mAbs. Plasma Aß sequestered by mAb9 rises rapidly and then stays stable for several days. In contrast, after mAb9 dosing, CSF Aß rises rapidly and declines to near baseline levels by 24 h. Similarly, the mAb9:Aß complex levels remain stable in the plasma between 6 and 24 h, but the levels of the complex in the CSF decrease. Moreover, brain mAb9 levels decrease by 50% between 6 and 24 h of time. Such data are consistent with a recent report showing that intraparenchymal injections of the anti-Aß mAb 4G8 result in rapid clearance of Aß from the brain (46)
. Our data, at least with mAb9, suggest that there is not necessarily a simple constant equilibrium between the mAb:Aß complex in plasma and brain or CSF. If this was the case, then the CSF and brain changes should parallel changes in plasma. The fact that the initial rise in CSF Aß and mAb:Aß complex is observed at the 6 h time point but that both Aß and mAb:Aß are decreased in the CSF by 24 h could be explained by several mechanisms. It is possible that 1) the permeability of the BBB is altered acutely after the mAb dosing allowing more mAb to penetrate the brain and CSF initially or that 2) the presence of a bound or unbound mAb in the brain activates a clearance mechanism that alters the equilibrium between the mAb in the CSF and blood. It is also apparent that the dynamics between plasma Aß and CSF Aß may differ from one mAb to another. In contrast to mAb9, other mAbs cause CSF Aß levels to continue to increase from 6 to 24 h or remain relatively constant over this time period. In any case, these data, together with data from others, showing that mAbs injected into the brain or CSF are rapidly cleared into the periphery suggest that an unbound anti-Aß mAb could enter the brain, bind Aß, and then clear it into the periphery (45
, 46)
. Given the tiny amounts of mAb that get into the brain, it remains difficult to envisage how such a mechanism could influence "total" Aß levels sufficiently to alter Aß deposition.
Long-term peripheral anti-Aß mAb delivery has been shown to reduce Aß deposition in the brain of multiple different APP mouse models (3
, 18
, 19
, 22
, 23)
. Because of differences in transgene expression levels and patterns, there are large differences in the relative levels of plasma and brain Aß in these different mouse models. For example, in Tg2576 mice, plasma Aß levels are over 100-fold higher then in PDAPP mice, whereas the differences in brain Aß levels, prior to deposition, are thought to be <3-fold. Despite these differences, long-term studies of mAbs with similar binding properties have similar overall effects on the extent of Aß reduction. Such data would argue that binding of plasma Aß by the Ab is unlikely to have a significant impact in terms of the ability of the mAb to reduce Aß deposition.
It is not possible to use the current data to completely exclude certain hypothesis as to how passive immunization with anti-Aß reduces amyloid deposition. The data do, however, provide a framework in which to consider the potential contribution of various mechanisms that have been proposed to account for attenuation of Aß deposition. Very little mAb gets in the brain; thus, if the mAbs are acting centrally, they are doing so at substoichiometric levels. There is simply insufficient mAb in the brain to significantly influence bulk metabolism of Aß, and we find no evidence that total Aß levels are influenced by mAb administration. If anti-Aß mAbs work directly on Aß, they must either alter some select pool or species of Aß that is present at low abundance and critical for deposition. Perhaps the consistent but small, nonsignificant decrease in RIPA soluble Aß reflects changes in this pool? Soluble preamyloid aggregates (oligomers, ADDLs, and protofibrils) are present at low levels and may represent critical intermediates in the aggregation pathway (47
48
49
50)
. Although there is no direct evidence that anti-Aß antibodies influence preamyloid aggregates in vivo, there is a lot of circumstantial evidence to suggest they do, including our current data (20
, 51
52
53)
. Furthermore, mAbs can influence Aß aggregation in vivo at substoichiometric levels, suggesting that they could preferentially recognize aggregated or even unaggregated Aß present in certain conformations that are critical for subsequent aggregate formation (54)
. It is also possible that small amounts of mAb binding to Aß in the brain or vasculature could trigger a change that indirectly influences Aß clearance. There is evidence that in certain APP mouse models, some anti-Aß mAbs can enhance phagocytosis of Aß by microglia, but in other studies there is no evidence that this is the case (14
, 22
, 30)
. Our current data do seem to indicate that there are some potential alterations in the transport of mAb:Aß out of the brain after peripheral administration; it is unclear, though, how this alteration could influence Aß deposition. It is certainly possible that the subtle, and opposite, shifts in the RIPA- and TBS extractable pools of Aß seen 6–24 h post administration are attributable to experimental variance, although the fact that they are seen in two different models would suggest this is not the case. If real, these changes do suggest that anti-Aß mAbs are likely to be altering intracerebral Aß metabolism in some enigmatic fashion.
Anti-Aß binding to plasma Aß results in the formation of a stable mAb:Aß complex. Though the binding does greatly extend the half-life of the bound Aß, it does not appear to significantly impact the half-life of the mAb, indicating that a classic immune complex is not formed and rapidly cleared. During the time period (0–12 h post anti-Aß mAb9 administration) when significant amounts of free anti-Aß mAb are presumably entering the blood, there is no detectable effect on total brain Aß levels in a GuHCL extract. In fact, CSF levels of Aß rise, an effect possibly attributable to mAb binding Aß and prolongation of its half-life in the CSF. Such data would suggest that peripheral anti-Aß mAbs do not enhance total bulk flow of Aß across the BBB. As noted previously, it is of course impossible to discount the possibility that the peripheral effects of the mAb binding to plasma Aß alter clearance of some select pool or species of Aß that is critical for Aß accumulation in the brain.
Anti-Aß mAbs can improve behavioral deficits in APP transgenic mice in the absence of significant effects on amyloid deposition and also improve behavioral deficits within a time period that would likely preclude major effects on Aß loads (20
, 51
, 55)
. The mAb-induced behavioral improvement has been postulated as attributable to alterations in Aß efflux from the brain or binding of small "neurotoxic" oligomers within the brain. However, given the large effects on Aß in the plasma and the potential vasoactive properties of plasma Aß, it is possible that sequestering of plasma Aß by the mAb could attenuate purported vasoconstrictive effects of Aß within the cerebral vasculature, thereby improving cerebral blood flow and performance in certain cognitive tasks (56
, 57)
. If anti-Aß antibodies are in fact working through some low abundance aggregation intermediate, it may be that mAb binding to nonaggregated Aß in the blood or brain might actually reduce the efficacy of passive immunotherapy. Most of the anti-Aß mAb will bind unaggregated Aß in the plasma and be unable to further influence Aß metabolism.
At least one humanized monoclonal anti-Aß mAb is in human clinical trials, and it is likely that additional anti-Aß mAbs will be tested in humans over the next several years. These studies in mice serve to highlight some issues that could significantly influence the outcome of the human trials. Assuming that humans metabolize Aß in the brain and plasma at similar rates to mice, it is possible to estimate both the daily turnover of Aß in the brain and the plasma of humans. As shown in Table 1
, these estimates indicate that humans turn over 50 nmol and 20 nmol of Aß in the plasma and brain, respectively, per day. If one were to administer an anti-mAb with the same binding properties as the mAb used in these mouse studies, and scale dosing to account for differences in body size (2000 fold increase), one would administer at least 1 g of mAb per dose to humans. If one scaled dosing based on the relative levels of Aß turnover in the brain of humans per day (80-fold more than in Tg2576 mice), one would administer 40 mg of mAb. However, if the mAb worked by binding plasma Aß, one might need only increase the relative dose by 6–7 fold, as the Tg2576 mice relative to their plasma volume have much higher levels of Aß.
An additional issue to be considered when thinking about the levels of mAb needed to confer efficacy has to do with the amount and type of deposited Aß present in the brain. In mice, mAbs that bind Aß amyloid are effective at preventing Aß deposition but are less effective at clearing preexisting Aß deposits, especially amyloid deposited in neuritic dense core plaques (13
, 18
, 21)
. In 21 month old Tg2576 mice, it is possible to estimate that the amount of Aß accumulated in the brain is 10 nmol (equivalent to the Aß turned over in the brain in 40 d) (58)
. If only a few pmoles of free mAb reach the brain per dose of anti-Aß mAb, then it would seem virtually impossible for the mAb to have any significant direct effect on the deposits themselves. In the typical human AD brain, on average 10 µmol of Aß accumulate (equivalent to the Aß turned over in 250 d) (59)
. Thus, there is 1000 times the amount of Aß present in the AD brain as in a mouse brain with extensive plaque pathology. Does this mean that to effectively dose a human with AD one would need to deliver 1000 times more mAb to the brain than is delivered to a mouse?
In summary, this quantitative evaluation of the acute effects of anti-Aß mAb delivery to AD mouse models prior to deposition serves to illustrate that the mechanisms by which such mAbs might alter amyloid deposition and reverse behavioral deficits remain enigmatic. Though it remains possible that very small acute changes in total Aß levels (<10%), which are difficult to measure experimentally, account for the long-term effects of immunization, it seems more likely that the mAbs are either 1) working on a select pool or conformer of Aß required for deposition or 2) indirectly influencing Aß deposition in the brain. Given the potential clinical promise of both passive and active approaches, additional studies that attempt to understand how mAbs, or active vaccination with Aß, attenuates AD-like pathologies are warranted.
ACKNOWLEDGMENTS
These studies were funded by the NIH/NIA to T. Golde (AG18454) and to E. McGowan. (AG022595–01). Additional resources from the Mayo Foundation provided by a gift from Robert and Clarice Smith were used to support the Tg2576 mouse colony that provided the mice used in these studies. Y. Levites was supported by John Douglas French Alzheimer’s Foundation fellowship. P. Das and Y. Levites were supported by a Robert and Clarice Smith Fellowship.
Received for publication May 8, 2006. Accepted for publication July 24, 2006.
REFERENCES
