Let's assume that some of the patients show improvement and this is correlated with antibody levels. Can we rule out some nonspecific immunological reactions that cause improvement independent of the ability of the antibodies to bind to Aβ? If these were experimental animals, one would be able to test the effects of immunizing with different forms of synthetic peptides. This is clearly not possible with human subjects. I am also concerned about the different results that are reported for the ELISA tests and the authors' tissue amyloid plaque assay. It is possible that they are looking at different conformational epitopes, as the authors suggest, but one...  Read more

Let's assume that some of the patients show improvement and this is correlated with antibody levels. Can we rule out some nonspecific immunological reactions that cause improvement independent of the ability of the antibodies to bind to Aβ? If these were experimental animals, one would be able to test the effects of immunizing with different forms of synthetic peptides. This is clearly not possible with human subjects. I am also concerned about the different results that are reported for the ELISA tests and the authors' tissue amyloid plaque assay. It is possible that they are looking at different conformational epitopes, as the authors suggest, but one should not overlook the fact that the tissue assay involves "fixed" tissue (they don't specify how) that is embedded in paraffin. It is not stated whether the Aβ peptides were similarly treated. If they were not, I would look first at the differences in antigenicity related to antigen preparation before concluding that conformational differences explain differences in immunoreactivity.

I find it puzzling that serum antibodies against Aβ remain high in the patients, without changes in circulating Aβ levels.

View all comments by Vincent Marchesi

Janus C, Pearson J, McLaurin J, Mathews PM, Jiang Y, Schmidt SD, Chishti MA, Horne P, Heslin D, French J, Mount HT, Nixon RA, Mercken M, Bergeron C, Fraser PE, St George-Hyslop P, Westaway D. A beta peptide immunization reduces behavioural impairment and plaques in a model of Alzheimer's disease. Nature. 2000 Dec 21-28;408(6815):979-82. Abstract

Morgan D, Diamond DM, Gottschall PE, Ugen KE, Dickey C, Hardy J, Duff K, Jantzen P, DiCarlo G, Wilcock D, Connor K, Hatcher J, Hope C, Gordon M, Arendash GW. A beta peptide vaccination prevents memory loss in an animal model of Alzheimer's disease. Nature. 2000 Dec 21-28;408(6815):982-5. Abstract

Nicoll JA, Wilkinson D, Holmes C, Steart P, Markham H, Weller RO. Neuropathology of human Alzheimer disease after immunization with amyloid-beta peptide: a case report. Nat Med. 2003 Apr;9(4):448-52. Abstract

Pfeifer M, Boncristiano S, Bondolfi L, Stalder A, Deller T, Staufenbiel M, Mathews PM, Jucker M. Cerebral hemorrhage after passive anti-Abeta immunotherapy. Science. 2002 Nov 15;298(5597):1379. Abstract

Schenk D. Opinion: Amyloid-beta immunotherapy for Alzheimer's disease: the end of the beginning. Nat Rev Neurosci. 2002 Oct;3(10):824-8. Abstract

Schenk D, Barbour R, Dunn W, Gordon G, Grajeda H, Guido T, Hu K, Huang J, Johnson-Wood K, Khan K, Kholodenko D, Lee M, Liao Z, Lieberburg I, Motter R, Mutter L, Soriano F, Shopp G, Vasquez N, Vandevert C, Walker S, Wogulis M, Yednock T, Games D, Seubert P. Immunization with amyloid-beta attenuates Alzheimer-disease-like pathology in the PDAPP mouse. Nature. 1999 Jul 8;400(6740):173-7. Abstract

View all comments by Dave Morgan

During the first year following the appearance of memory deficits in Tg(APPNL)2576 mice, neurons and synapses are largely intact (Irizarry et al., 1997). During the second year, postsynaptic markers decline, while presynaptic markers and neurons remain unchanged (G. Cole and B. Hyman, personal communication). We have proposed that soluble Aβ assemblies...  Read more

During the first year following the appearance of memory deficits in Tg(APPNL)2576 mice, neurons and synapses are largely intact (Irizarry et al., 1997). During the second year, postsynaptic markers decline, while presynaptic markers and neurons remain unchanged (G. Cole and B. Hyman, personal communication). We have proposed that soluble Aβ assemblies impair memory in Tg(APPNL)2576 mice (Ashe, 2001; Westerman, 2002), and have suggested that the rapid restoration of memory by passive immunization against Aβ indicates that Aβ assemblies disrupt memory by altering neuronal function, but not neuronal structure.

Patients with Alzheimer’s disease differ from Tg(APPNL)2576 mice because they have substantial plaque and tangle deposition as well as significant cell loss in vulnerable brain regions important for memory. The relative benefit conferred by Aβ immunization in the Hock et al. paper may reflect the inhibition of the disruptive effects of Aβ assemblies on cognitive function or the improvement of certain aspects of amyloid pathology taking place in the setting of ongoing neurodegeneration. Achieving in humans the dramatic results observed in mice is more likely to occur if interventions are administered in earlier stages of disease. Understanding how to improve cognitive function in later stages of Alzheimer’s disease will require a new generation of mouse models to study.

View all comments by Karen Hsiao Ashe

Site-directed antibodies induced by various immunological approaches are aimed at treatment of a disease that is caused by abnormal conformational changes or folding of a peptide or protein, as presented in Alzheimer’s disease and other amyloidosis disorders (Solomon, 2002). However, any effective immunization strategy must identify not only the specific nature of the antigen or the epitope, but also address the formulation and method of delivery of the antigen or antibodies as a major and critical parameter.

Unfortunately, humans may develop self-antibodies when immunized with whole or fragments of AβPP. These antibodies are capable of binding to a variety of Aβ species in the brain; thus, immunization could have beneficial effects, such as inhibition of amyloid fibril formation, while microglial overactivation may lead to neuroinflammation. The consequence of this on inflammatory pathology in AD brains needs to be considered before immunization is used as a strategy for treating AD. As recently reported, interactions of human microglia with antibody-opsonized amyloid showed increased inflammation (Lue et al., 2002).

Several strategies directed towards prevention of neuroinflammation are under investigation. Active immunization with synthetic Aβ1-42 peptide reduces β-amyloid plaques in AβPP-transgenic mice without detectable toxicity, but the extension of this approach to AD patients induced a neuroinflammatory reaction in some of the study subjects, precluding further testing of the preparation. Vaccination with nontoxic, small antiaggregating epitopes of AβPP may partially avoid the undesirable effects of neuroinflammation, e.g., by preventing T cell activation (Frenkel et al., 2003).

Administration of intravenous immunoglobulin (IVIG), which has well-recognized antiinflammatory activities independent of the antigen-specific effect, may modulate the inhibitory FcR pathway, thus controlling autoantibody-mediated inflammation induced by self-antigens or antibodies in immunotherapeutic strategies for treatment of AD. Another approach may be passive immunization with antibodies devoid of Fc, which may prevent overactivation of microglia and, thus, attenuation of autoantibody-triggered neuroinflammation. Progress in vector development for brain delivery of such antibodies, as well as clearance of immunocomplex devoid of Fc region, was recently reported (Frenkel and Solomon, 2002).

Many important questions remain open. Is the reported improvement in the behavior of AD patients caused by dissolving existing plaques or preventing formation of new plaques, or is it caused by sequestration of soluble AβPP? How many antibodies are required? How can inflammation and/or overactivation of microglia be prevented? In spite of these questions, the immunotherapeutic approach towards amyloid peptide remains the most fascinating therapeutic target for generating agents potentially able to modify the natural history of AD.

References:
Solomon B. Immunological approaches as therapy for Alzheimer's disease. Expert Opin Biol Ther. 2002 Dec;2(8):907-17. Abstract

Lue LF, Walker DG. Modeling Alzheimer's disease immune therapy mechanisms: interactions of human postmortem microglia with antibody-opsonized amyloid beta peptide. J Neurosci Res. 2002 Nov 15;70(4):599-610. Abstract

Frenkel D, Dewachter I, Van Leuven F, Solomon B. Reduction of beta-amyloid plaques in brain of transgenic mouse model of Alzheimer's disease by EFRH-phage immunization. Vaccine. 2003 Mar 7;21(11-12):1060-5. Abstract

Frenkel D, Solomon B. Filamentous phage as vector-mediated antibody delivery to the brain. Proc Natl Acad Sci U S A. 2002 Apr 16;99(8):5675-9. Abstract

View all comments by Beka Solomon

HASTA LA VISTA, AMYLOID CASCADE HYPOTHESIS, OR WILL ACADEMIC DISHONESTY YIELD ALZHEIMER'S CURE?

Please see my letter to Neuron editor regarding the article by Hock et al.
Science SAGE KE (26 May 2003) [ Full Text ] [ No-registration access link ] .


View all comments by Alexei R. Koudinov

First, there was no control group that was not immunized, and the rate of increase in antibody levels was not studied in the patients prior to baseline. Thus, since some patients had endogenous antibodies to β amyloid, there was no evidence that the immunizations actually caused the increase in antibodies.

Since a large portion of the patients did not generate a significant increase in antibody levels, and these patients did worse than the literature average on some cognitive tests, this calls into question several things:

1) Perhaps the immunization is doing harm in some and is benefiting others,

or, what I consider more likely:

2) The level of antibody increase may simply be gauging the health of the immune system, and thereby gauging general health. Those who have the lowest response, then, have worse-than-average general health and thus would suffer a worse-than-average cognitive decline.

A control of non-immunized subjects is absolutely necessary to judge how the immunization...  Read more

First, there was no control group that was not immunized, and the rate of increase in antibody levels was not studied in the patients prior to baseline. Thus, since some patients had endogenous antibodies to β amyloid, there was no evidence that the immunizations actually caused the increase in antibodies.

Since a large portion of the patients did not generate a significant increase in antibody levels, and these patients did worse than the literature average on some cognitive tests, this calls into question several things:

1) Perhaps the immunization is doing harm in some and is benefiting others,

or, what I consider more likely:

2) The level of antibody increase may simply be gauging the health of the immune system, and thereby gauging general health. Those who have the lowest response, then, have worse-than-average general health and thus would suffer a worse-than-average cognitive decline.

A control of non-immunized subjects is absolutely necessary to judge how the immunization actually influenced antibody levels.

Finally, the authors conclude that there is no evidence that the β amyloid was being bound and transported out of the brain; therefore, it is quite clearly not a clinical confirmation that β amyloid has a central role in AD, as they write.

In fact, it seems to be suggestive that a third factor not involving β amyloid was causing the rise in antibodies as well as the increased performance—namely, the multiplicity of factors involved in immune function that are also involved in general health.

I cite this study and discuss it in my article, "Myth: Cholesterol Causes Alzheimer's Disease | Part I: Debunking the Myth," which is located at http://www.cholesterol-and-health.com/Cholesterol-Alzheimers.html.

View all comments by Chris Masterjohn

Importantly, the study also showed considerable variability in the antibody response across the individuals studied. Combined with the adverse events noted in 5 percent of the patient population, this would suggest passive immunization approaches—a reversible therapy with known amounts of antibody—would be the most prudent next step in testing the immunotherapy approach to AD. It might be most appropriate to only include patients with a...  Read more

Importantly, the study also showed considerable variability in the antibody response across the individuals studied. Combined with the adverse events noted in 5 percent of the patient population, this would suggest passive immunization approaches—a reversible therapy with known amounts of antibody—would be the most prudent next step in testing the immunotherapy approach to AD. It might be most appropriate to only include patients with a patent blood brain barrier.

The McLaurin paper was mischaracterized in the US press (AP) as suggesting that a "safe" form of the vaccine had been tested. What was shown was that, using the standard Aβ1-42 vaccine preparation initially described by the Elan group and similar to that used in the human trials, the primary epitope recognized was the N-terminal amino acids 4-10. This replicated our own work (Dickey et al, DNA and Cell BIol. 11:723) and that of the Elan group and Beka Solomon's group, albeit using considerably more sophisticated technology. However, these authors did not test an alternative vaccine formulation, such as was described by Sigurdsson et al (see related news story) or Nicolau et al (see related news story). While McLaurin et al. did not observe CNS inflammation in young transgenic mice vaccinated against Aβ1-42, it would be intriguing to find if the vaccine might cause CNS inflammation in aged transgenic mice, or mice with opening of the blood-brain barrier, given the Hock et al results.

View all comments by Dave Morgan

A potential reason why some AD patients developed an adverse reaction to the vaccine is that immunization with Aβ42 might have elicited a T-helper cell 1 (Th1)-mediated response, which in turn might have stimulated a pro-inflammatory reaction (Munch and Robinson, 2002b). McLaurin and colleagues (2002) reasoned that if an immunization regime could instead activate a Th2 response, which aids B-cells, a pro-inflammatory response could be avoided. In an earlier paper, the same team had immunized TgCRND8 mice with protofibrillar/oligomeric assemblies of Aβ42 and found that these mice...  Read more

A potential reason why some AD patients developed an adverse reaction to the vaccine is that immunization with Aβ42 might have elicited a T-helper cell 1 (Th1)-mediated response, which in turn might have stimulated a pro-inflammatory reaction (Munch and Robinson, 2002b). McLaurin and colleagues (2002) reasoned that if an immunization regime could instead activate a Th2 response, which aids B-cells, a pro-inflammatory response could be avoided. In an earlier paper, the same team had immunized TgCRND8 mice with protofibrillar/oligomeric assemblies of Aβ42 and found that these mice showed a reduction in Ab plaque burdens and an improved performance in the Morris water maze test (Janus et al., 2000). Notably, the brains of these mice did not exhibit appreciable levels of microglial activation. Could it be that they had found a way to generate antibodies to Aβ without provoking a neuroinflammatory Th1 response? The present study suggests that they have. They demonstrate that their immunization approach produces antibodies with a high affinity for residues 4-10 of human Aβ42. The antibodies bind selectively to human Aβ42 protofibrils and can prevent fibrillogenesis of Aβ in vitro. Importantly, the Aβ42-immunized sera predominantly expressed IgG2b, which is the least effective activator of the complement system. Furthermore, no increased inflammatory response was observed in splenic T-cells from these mice, and when challenged with Aβ, their pattern of cytokine production was typical of a Th2 response.

These results hold out the exciting promise of a less inflammatory approach to the treatment of AD, but a key question remains: are transgenic mice a reliable model of AD? We have commented elsewhere (Check, 2002; Munch and Robinson, 2002a,b; Smith et al., 2002a) that antibodies against human Aβ42 merely assist in removing a redundant protein that is being overexpressed in the brains of these transgenic mice. Extracellular accumulation of this protein will impede the flow of solutes in the extracellular space and will impair brain function (Robinson and Bishop, 2002). Therefore, even if the deposits of human Aβ are inert, their clearance from the brain would be expected to improve cognition. Aβ endogenous to the mouse brain is not recognized by antibodies to human Aβ because it has a different amino acid sequence, so these animals are able to maintain a normal complement of mouse Aβ. By contrast, immunization in humans aims to deplete the brain of its endogenous Aβ. Can we really be sure that this will not produce an adverse inflammatory response, or prevent Aβ from performing its important (and hitherto unrecognized) functions (Atwood et al., 2001; 2002a,b,c; Robinson, 2002; Robinson and Bishop, 2002)?

For some time now we have been warning about the potential dangers of a vaccine approach to AD (Perry et al., 2000; Smith et al., 2000; Joseph et al., 2001). Our concerns have been based on our findings that Aβ and APP may act as antioxidants (Bush et al., 2000), and also that oxidative stress not only precedes Aβ deposition in AD, but that the appearance of Aβ plaques is associated with a decrease in oxidative stress (Nunomura et al., 2001). The present results of McLaurin and colleagues (2002) or Hock et al. (2002) do nothing to address these concerns.

Hock and colleagues (2002) provided a cohort of 30 patients for the Elan AN1792 trials. They have now tested the sera from patients immunized with Aβ42 and they report that the majority contain antibodies which bind to fibrillar plaques, diffuse Aβ deposits and vascular amyloid in sections from human brains. These antibodies do not recognize cell surface APP, or full length APP in Western blots. This important finding will help to allay fears that antibodies directed against Aβ could inadvertently impair the functions of APP, or worse still, stimulate an autoimmune response against neurons (Munch and Robinson, 2002a,b).

The study also examined whether the cerebrospinal fluid (CSF) from six patients contained appreciable titers of antibodies to Aβ. Half of the patients were found to have high CSF albumin levels, indicative of a leaky blood-brain-barrier (BBB), and all of these patients had high Aβ antibody titers in their CSF. Of the three patients with an intact BBB, only one had a detectable Aβ antibody titer. These results are interesting for several reasons. First, the proportion of patients with a leaky BBB appears to be relatively high, considering that they are mildly or moderately demented (eg. Wada, 1998). Second, the presence of detectable antibodies in the CSF (and hence the brain) appears to be associated with the presence of a leaky BBB. Third, the patient diagnosed with meningoencephalitis had, by far, the highest levels of Aβ antibodies in their CSF.

The patient sample is small, but the data are consistent with the possibility that the immunization technique increases permeability of the BBB. This might occur because cytokines that increase BBB permeability, such as TNFα (Tsao et al., 2001), are released from activated macrophages in response to intravascular and perivascular deposits of Aβ. An additional mechanism relates to our speculation that perivascular Aβ deposits may help to seal leaks in the BBB. We have predicted that clearance of these protective perivascular deposits by antibody-mediated phagocytosis of Aβ could facilitate the entry of pathogens into the brain, leading to an increased risk of meningoencephalitis and other brain infections (Atwood et al., 2002a,b; see also Bishop and Robinson, 2002). If this prediction is correct, a successful immune response to Aβ could exact a high price.

The paper by Hock and colleagues is also remarkable for what it doesn't say. Here was a golden opportunity to provide answers to whether Aβ immunization in humans reduces Aβ levels in the CSF, and whether it triggers a Th1-mediated neuroinflammatory response, yet no mention was made of the Aβ levels or cytokine levels in the CSF of these patients. We wonder at the reason for this omission, and we continue to wonder whether the AD patients with high antibody titers in their CSF did in fact exhibit a slower rate of cognitive decline then their less immune counterparts. Stephen Robinson, Monash University, Clayton, Australia; Glenda Bishop, Mark Smith, George Perry, Craig Atwood, all at Case Western Reserve University, Cleveland, Ohio.

References:

Atwood CS, Bishop GM, Perry G. and Smith MA (2002a). Don't pick the scab: Ab as a vascular sealant that protects against hemorrhage. Commentary for the Alzheimer Research Forum live chat session: Alzheimer Immunotherapy Trial Grounded - Time to Reassess Safety and Vaccine Design?

Atwood CS, Bishop GM, Perry G. and Smith MA (2002b) Amyloid-beta: a vascular sealant that protects against hemorrhage. J. Neurosci. Res., (in press).

Atwood, C.S., Robinson, S.R. and Smith, M.A. (2002c). Amyloid-beta: redox-metal chelator and antioxidant. J. Alzheim. Dis., 4, 203-214.

Bishop GM, Robinson SR, Smith MA, Perry G, and Atwood CS. (2002) Call for Elan to publish AlzheimerÕs trial details. Nature, 416:677.

Bishop GM, and Robinson SR (2002) The amyloid hypothesis: let sleeping dogmas lie? Neurobiol Aging, 23: in press.

Bush AI, Atwood CS, Goldstein LE, Huang X, and Rogers J. (2000) Could Abeta and AbetaPP be Antioxidants? J Alzheim. Dis., 2: 83-84.

Check E. (2002) Nerve inflammation halts trials for Alzheimer's drug. Nature, 415: 462.

Janus C. et al (2000) Ab peptide immunization reduces behavioral impairment and plaques in a model of Alzheimer's disease. Nature, 408: 979-982.

Joseph J, Shukitt-Hale B, Denisova NA, Martin A, Perry G, Smith MA. (2001) Copernicus revisited: amyloid beta in AlzheimerÕs disease. Neurobiol. Aging, 22:131-146.

Munch G, and Robinson SR. (2002a) Alzheimer's vaccine: a cure as dangerous as the disease? J Neural Transm., 109: 537-9.

Munch G, and Robinson SR. (2002b) Potential neurotoxic inflammatory responses to Abeta vaccination in humans. J Neural Transm., 109 : 1081-7.

Nunomura A, Perry G, Aliev G, Hirai K, Takeda A, Balraj EK, Jones PK, Ghanbari H, Wataya T, Shimohama S, Chiba S, Atwood CS, Petersen RB, Smith MA. (2001) Oxidative damage is the earliest event in Alzheimer disease. J Neuropathol Exp Neurol., 60: 759-767.

Perry G, Nunomura A, Raina AK, Smith MA. (2000) Amyloid-§ junkies. Lancet, 355: 757.

Robinson SR (2002) Alzheimer vaccine: lest we forget. Commentary for the Alzheimer Research Forum live chat session: Alzheimer Immunotherapy Trial Grounded - Time to Reassess Safety and Vaccine Design?

Robinson SR, and Bishop GM (2002) Ab as a bioflocculant: implications for the amyloid hypothesis of Alzheimer's disease. Neurobiol Aging, 23: in press.

Smith MA, Joseph JA, Perry G. (2000) Arson: tracking the culprit in AlzheimerÕs disease. Ann NY Acad Sci., 924:35-38.

Smith MA, Atwood CS, Joseph JA, and Perry G. (2002a) Predicting the failure of amyloid-b vaccine. Lancet, 359: 1864-5.

Smith MA, Atwood CS, Joseph JA, and Perry G. (2002b) Ill-fated amyloid-b vaccine. J. Neurosci. Res., 69: 285.

Smith MA, Joseph JA, Atwood CS, and Perry G. (2002c) Dangers of the amyloid-b vaccination. Acta Neuropathol., 104: 110.

Tsao N, Hsu HP, Wu CM, Liu CC, Lei HY (2001) Tumour necrosis factor-alpha causes an increase in blood-brain barrier permeability during sepsis. J Med Microbiol., 50: 812-21.

Wada H. (1998) Blood-brain barrier permeability of the demented elderly as studied by cerebrospinal fluid-serum albumin ratio. Intern Med., 37: 509-13.

View all comments by Craig Atwood
View all comments by Glenda Bishop
View all comments by George Perry
View all comments by Stephen Robinson
View all comments by Mark A. Smith

The paper by McLaurin et al. describes a very careful, in-depth analysis of the antibodies generated against protofibrillar Aβ1-42 preparations. The paper shows that with this preparation, the predominant epitope targeted in the mice is Aβ 4-10. Furthermore, this purified antibody preparation is highly effective in both cellular toxicity models as well as in vivo. They also demonstrate that the predominant immunological response is Th2, which is often considered to be non-inflammatory in nature.

In the simplest interpretation, these findings suggest that antibodies against a small region of Aβ are sufficient to elicit all of the benefits seen with the Aβ1-42 immunization studies that have been previously reported (Schenk D., et al. Nature 400: 173 [1999], Morgan, D. et al. Nature 408: 982 [2000], Janus C. et al. Nature 408:979 [2000]. It is also important to note that immunization...  Read more

The paper by McLaurin et al. describes a very careful, in-depth analysis of the antibodies generated against protofibrillar Aβ1-42 preparations. The paper shows that with this preparation, the predominant epitope targeted in the mice is Aβ 4-10. Furthermore, this purified antibody preparation is highly effective in both cellular toxicity models as well as in vivo. They also demonstrate that the predominant immunological response is Th2, which is often considered to be non-inflammatory in nature.

In the simplest interpretation, these findings suggest that antibodies against a small region of Aβ are sufficient to elicit all of the benefits seen with the Aβ1-42 immunization studies that have been previously reported (Schenk D., et al. Nature 400: 173 [1999], Morgan, D. et al. Nature 408: 982 [2000], Janus C. et al. Nature 408:979 [2000]. It is also important to note that immunization with a smaller fragment of Aβ peptide has already been reported to be effective (Sigurdsson E. et al. Am. J. Pathol. 159: 439 [2001]. These findings are also consistent with previously published results showing that monoclonal antibodies against Aβ (which, by definition, are only against a single epitope as well) also work very well for efficacy in vivo (Bard, F. et al. Nature Med. 6: 916 [2000]; De Mattos et al. Proc. Natl. Acad. Sci 98: 8850 [2001], Kotilinek, L. et al. J. Neurosci. 22:6331 [200]).

The paper by Hock, C. et al. describes the first of several reports regarding the much-publicized phase II study involving Aβ1-42 immunization that was halted after 1-3 doses due to meningo-encephalitis.

In this paper, the authors examined the specificity of the antibodies generated against the Aβ peptide as well as AβPP following several doses of AN 1792. They concluded that, first, the patients clearly did make detectable antibodies specific to Aβ peptide found in pathological structures, second, anti-Aβ antibodies were specific to amyloid deposits and did not cross-react with other brain components nor with APP and, third, antibody titer was detectable in CSF of some of the patients as well as plasma of the patients.

Collectively these findings are cautiously encouraging in the context of the immunotherapy approach. They suggest that antibodies elicited by Aβ are not likely causing side effects due to spurious cross-reactivity and also show that antibodies alone do not result in side effects. Importantly, the results parallel findings in the mice where it has been shown that antibodies to Aβ can be found in both plasma and CSF. The fact that they can be seen in humans, as well, is important.

Clearly there is much still to be learned about the mechanisms by which Aβ immunotherapy (including both active and passive administration of anti-Aβ antibodies) elicits significant pathological and cognitive performance efficacy in APP-transgenic animal models of Alzheimer's disease towards the ultimate goal of clinical utility.

There appears to be a growing consensus that while immunization with Aβ1-42 (i.e. AN 1792) has resulted in unacceptable level of adverse events, i.e. encephalitis in a subset of patients, additional research and efforts are clearly warranted. At the very least, findings such as those described by McLaurin et al. and Hock et al. suggest that second-generation strategies that optimize efficacy while improving upon the initial safety profile of holo-Aβ1-42 should be pursued for clinical development for Alzheimer's disease. Specifically, these include both immunoconjugates and monoclonal antibodies to Aβ peptide. My colleagues and I at Elan and Wyeth—as well as many other scientists around the globe—remain committed to this goal for future treatment of Alzheimer's disease.

View all comments by Dale Schenk

Whether the antibodies would immunoprecipitate Aβ under physiological conditions from body fluids such as CSF was not tested, so it is still possible that the antibodies could see soluble Aβ in solution. The titers of the antibodies in the plasma in some individuals were sometimes high > 1:10,000, and in patients without encephalitis with an intact BBB, the CSF titers were all 1:50 or less. This is consistent with the fact that the IgG that crosses that BBB is by passive diffusion and its amount is proportional to size of the molecule (usually   Read more

Whether the antibodies would immunoprecipitate Aβ under physiological conditions from body fluids such as CSF was not tested, so it is still possible that the antibodies could see soluble Aβ in solution. The titers of the antibodies in the plasma in some individuals were sometimes high > 1:10,000, and in patients without encephalitis with an intact BBB, the CSF titers were all 1:50 or less. This is consistent with the fact that the IgG that crosses that BBB is by passive diffusion and its amount is proportional to size of the molecule (usually <0.1 percent of plasma levels). Measurements of Aβ in the plasma of the humans was not reported.

The take-home message is that many humans immunized with Aβ42 in the protocol did develop antibodies to β and that the antibodies do not appear to cross-react with AβPP. The results do not yet explain the side effects seen in a subset of subjects who developed meningoencephalitis. It is certainly possible that the side effects are due to T-cells, though more studies need to be done. It is useful to know that humans can generate anti-Aβ antibodies at higher titers and this information may be helpful for future work in this area.

McLaurin et al.show that immunization of TgCRND8 mice or control mice with protofibrillar aggregates of Aβ induced formation of antibodies against Aβ. The minimal peptide that the antibodies in this paradigm were generated against was sequence 4-10. The authors nicely show that these antibodies have potent effects in vitro in blocking Aβ toxicity, preventing Aβ fiber formation, and in partially deaggregating pre-formed fibrils. The data suggest that one mechanism by which such antibodies could work in vivo is by the effects seen in vitro. The authors suggest that the effects they observe are caused by the effects of the antibodies on Aβ protofibrils/oligomers. Whether this is the way such antibodies work in vivo may be difficult to ultimately prove. There are many different features of anti-Aβ antibodies, not only to sequences 4-10 but to other regions, which could potentially have similar effects as the antibodies studied here. One may need to study a variety of antibodies, not only ones that identify different regions but also some that have different affinities or other properties, in order to understand whether the sequence is what is absolutely critical for these effects.

Nonetheless, the data show convincingly that the antibodies generated in the mice studied have potent effects in several different assays. Whether a similar protocol as studied here in mice will be useful in humans is an interesting question. The fact that there are positive clinical effects and a lessening (not an increase) in microgliosis is promising. To know whether there will side effects in humans with this type of approach will require further testing.

View all comments by David Holtzman

HSV1 is of particular interest because we detected HSV1 DNA (in so-called latent virus form) in brain of a high proportion of elderly people by PCR (Jamieson et al., 1991, et seq.). We subsequently implicated this virus, when in brain of ApoE4 carriers, as a strong risk factor for AD (Itzhaki et al., 1997; Lin et al., 1998). This has been strengthened by our findings that ApoE determines the severity of damage caused by a different virus, hepatitis C virus in liver (Wozniak et al., 2002), that it determines whether HSV1 in the peripheral nervous system causes cold sores (Itzhaki et al., 1997; Lin et al., 1998; Corder et al. 1998), and whether pre-AIDS HIV causes dementia and peripheral neuropathy.

We suggested that HSV1 in brain can reactivate from latency and that it then causes damage that is greater in APOE-e4 carriers, leading to AD. Reactivation could occur during stress, immunosuppression, or inflammation, and in turn would cause further inflammation. Thus, relevant information about the patients in the Elan trial—both those with and those without overt brain inflammation—would be of great interest. We, therefore, urge Elan to divulge the numbers and the ApoE genotypes of patients with brain inflammation, and the numbers of those in whose CSF viral DNA or virus was found. Further, we urge them to consider using antiviral agents to treat patients who developed brain inflammation.

Secondly, as far as we know, no information has been disclosed about the rate of cognitive decline in the whole group before brain inflammation occurred. Prevention of such decline is surely the "gold standard" for treatment success. Also, have the affected patients recovered or have they declined further, and just how many were affected? Again, we urge Elan to disclose the relevant information. (See also Current Hypothesis)

View all comments by Curtis Dobson
View all comments by Ruth Itzhaki

When the authors gave 21-month-old APP23 mice a monoclonal antibody directed to Aβ3-6 once a week for five months, they saw that the...  Read more

When the authors gave 21-month-old APP23 mice a monoclonal antibody directed to Aβ3-6 once a week for five months, they saw that the rate of hemorrhages increased about twofold above baseline. The severity of the hemorrhages also increased approximately 30 percent above levels seen in the untreated group. No control antibody group was tested, and thus, we cannot be sure that the effect was specific for the particular anti-Aβ antibody monoclonal used.

As the authors correctly point out, these types of findings have not been seen in other APP-transgenic mouse models that have been actively or passively immunized with Aβ. Though other APP transgenes, such as the PDAPP mouse that we routinely have used in our studies, do show CAA, as well (Kimchi, 2001), the amount of this type of pathology is significantly less than that seen in the APP23 mice, and this might be the reason for the novelty of the new report.

Aβ immunotherapy for Alzheimer’s disease remains an important new approach for potential treatment of this devastating disease. Clinical progress with immunization of Aβ42 (AN 1792) recently suffered a setback when a subset of treated patients developed meningoencephalitis—a condition distinct from the hemorrhagic stroke described in the APP23 mouse. This recent finding, nevertheless, adds to a growing body of literature on the subject. As with all new preclinical observations, additional experiments will be required to understand whether these new findings, in this particular animal model, will have a clinical correlate in humans or not.

See response by Alexei Koudinov: Amyloid was never clearly implicated in Alzheimer's disease, so look at Aβ from a different angle. Koudinov AR. British Medical Journal (30 November 2002).

View all comments by Dale Schenk

The Hock et al. paper is very important and timely, as it represents the first insight into the patients who received the vaccine. I don't agree with the data that these sera did not recognize the soluble monomers or oligomers of BAP. They bind to all conformations of AAβ, but not to APP.

Reference:

Frenkel D., I. Dewachter I, F. Van Leuven, B. Solomon. Reduction of ß-amyloid plaques in brain of transgenic mice, model of Alzheimer's disease, by EFRH-phage immunization. Vaccine, 2002, in press.

View all comments by Beka Solomon

It does seem a likely hypothesis that the adverse events observed in the Elan trial were related to a Th1-type response against Aβ induced by the adjuvant (Weiner and Selkoe, 2002). Of note, T cell reactivity to Aβ has not been documented in patients with AD. However, in extensive ongoing studies in our laboratories, we have found such T cell reactivity and are in the process of characterizing...  Read more

It does seem a likely hypothesis that the adverse events observed in the Elan trial were related to a Th1-type response against Aβ induced by the adjuvant (Weiner and Selkoe, 2002). Of note, T cell reactivity to Aβ has not been documented in patients with AD. However, in extensive ongoing studies in our laboratories, we have found such T cell reactivity and are in the process of characterizing it—something that will be required should further vaccination trials be attempted in AD. The human result does not mean that Aβ vaccination cannot be attempted if modifications of the immunization protocol are made. This includes immunizing with fragments of Aβ to obtain only antibody responses, or immunizing in a fashion that does not induce Th1 type responses. We are planning the latter approach in humans by using the mucosal route to administer Aβ, inducing both antibodies and Th2/Th3 responses rather than Th1 responses (Monsonego et al., 2000). We also postulate that, in addition to antibodies, T cells which secrete cytokines such as TGF-β may also be of benefit in AD.

References:
Monsonego A, Maron R, Zota V, Selkoe DJ, Weiner HL. Immune hyporesponsiveness to amyloid beta-peptide in amyloid precursor protein transgenic mice: implications for the pathogenesis and treatment of Alzheimer's disease. Proc Natl Acad Sci U S A. 2001 Aug 28;98(18):10273-8. Abstract

Weiner HL, Selkoe DJ. Inflammation and therapeutic vaccination in CNS diseases. Nature. 2002 Dec 19-26;420(6917):879-84. Abstract

Weiner HL, Lemere CA, Maron R, Spooner ET, Grenfell TJ, Mori C, Issazadeh S, Hancock WW, Selkoe DJ. Nasal administration of amyloid-beta peptide decreases cerebral amyloid burden in a mouse model of Alzheimer's disease. Ann Neurol. 2000 Oct;48(4):567-79. Abstract

View all comments by Howard Weiner

It is good to see that more attention is being given to the immunogenicity of the peptide/HL-A complexes. A better outcome would be wonderful, but active immunization of AD patients at this stage of our understanding of the disease remains risky.

View all comments by Harvey Cantor

The article by Roberto Furlan and colleagues indicates that vaccination with Aβ in Freund’s adjuvant may induce...  Read more

The article by Roberto Furlan and colleagues indicates that vaccination with Aβ in Freund’s adjuvant may induce encephalomyelitis in mice when coadministered with pertussis toxin (PT). Based on the known properties of PT, this vaccine combination could be expected to generate a T-1-type immune response which, among other things, results in the activation of macrophages, cytotoxic T-lymphocytes (CTLs), and high levels of IgG2a antibodies. The authors point out that the mouse encephalitis observed in their studies might resemble the cerebral inflammation seen in some of the AD patients receiving the Elan vaccine AN-1792, in which QS-21 served as an adjuvant. Although the cause of the adverse effects in the clinical trial remains to be determined, it is interesting to note similarities in the properties of these two adjuvants. Like PT, QS-21 promotes a T-1 type response and, in mice, both are associated with an increase in complement-fixing antibodies, such as IgG2a (Kensil et al., 1995). Related to this type of response, PT induces CD4+ T-cells as shown by Furlan and colleagues, whereas QS-21 elicits both a CD4+ and a CD8+ effect in various species (Kensil et al., 1995). Interestingly, most CTLs are CD8+, suggesting that QS-21 may have a more pronounced CTL effect than PT. In designing vaccines for promoting Aβ clearance, it is obviously important to avoid a CTL induction, and any detrimental complement activation. As we discussed previously, adjuvants stimulating a T-2 type response such as alum may be more appropriate for human vaccination with Aβ derivatives (Sigurdsson et al., 2001). These adjuvants promote predominantly a humoral immune response, which is more likely to lead to Aβ clearance without the potentially destructive effects of CTLs.

References:
Sigurdsson E, Wisniewski T, Frangione B. Infectivity of amyloid diseases. Trends Mol Med. 2002 Sep ; 8(9):411. Abstract

Kensil CR, Wu J-Y, Soltysik S. Structural and Immunological Characterization of the Vaccine Adjuvant QS-21, Pharm Biotechnol. 1995;6:525-41. (No abstract available)

Sigurdsson EM, Scholtzova H, Mehta PD, Frangione B, Wisniewski T. Immunization with a nontoxic/nonfibrillar amyloid-beta homologous peptide reduces Alzheimer's disease-associated pathology in transgenic mice. Am J Pathol. 2001 Aug ; 159(2):439-47. Abstract

View all comments by Blas Frangione
View all comments by Einar Sigurdsson

Conversely,...  Read more

Conversely, an excessive T cell response is precisely what Elan claims is responsible for the problem they observed in the five percent of vaccinated patients who exhibited signs of CNS inflammation. For this reason they are planning further study with vaccines that have less of a T cell response.

Another option would be to evaluate passive antibody immunotherapy, as this should totally lack T cell activation. Until Elan releases more information about the cases that had adverse reactions, it will be difficult to know if the model described here is relevant to human AD cases.

View all comments by Dave Morgan

Using a phage-peptide library composed of filamentous phage displaying three million random combinatorial peptides, we defined the EFRH residues located at positions 3-6 of the N-terminal Aβ as the epitope of anti-aggregating antibodies within Aβ (Frenkel et al., 1998; Frenkel et al., 1999). The EFRH is not only the epitope of anti-aggregating antibodies but acts as a regulatory site controlling both the formation and disaggregation process of the amyloid fibrils. Locking of this epitope by highly specific antibodies affects the dynamics of the entire Aβ molecule, preventing self-aggregation as well as enabling resolubilization of already formed aggregates. This conclusion was reached from experimental data with different lengths of Aβ peptides or similar peptides with one or two mutations in EFRH region (Frenkel et al., 1998; Frenkel et al., 1999).

Antibodies resulting from EFRH immunization are similar in their anti-aggregating properties to antibodies raised by direct injection with whole Aβ and/or mAbs directed to this region (Frenkel et al., 2000). Such antibodies at low titer (1-100—1-1000) are enough to reduce the amyloid plaques to the same extent as passive immunization with larger amounts of antibodies directed to EFRH (Frenkel et al., 2003). Antibodies that bind to the epitope containing only a few amino acids from EFRH, such as mAb 3D6 (Bacskai et al., 2002), are less effective compared to mAb 10D5, which binds to the whole sequence.

However, not all the antibodies that bind to EFRH exhibit anti-aggregating properties. Mab 2H3, whose epitope is located between amino acids 1-7, binds with a higher binding constant (10-⁹M) to the whole epitope, but only with (10-⁴M) to EFRH and did not have anti-aggregating properties, highlighting the importance of the high affinity of the antibodies to this specific sequence on the behavior of whole Aβ (Frenkel et al., 1999).

Unfortunately, immunization could have contradictory effects; besides disaggregating amyloid plaques it could trigger also microglial overactivation, which might lead to neuroinflammation. Mabs that bind to available epitopes of Aβ in brain (passive or active immunization) activate the Fc receptors which may initiate the inflammatory response. Modulation of FcR activation, using antibodies devoid of the Fc region, or partial FcR blockage, may be efficient practical therapeutic approaches for controlling autoantibody-mediated inflammation induced by self-antigens or antibodies in immunotherapeutic strategies for treatment of AD (Solomon, 2002).

References:
Solomon B. Immunological concept in the treatment of Alzheimer’s disease. Drug Development Research. 2002;56:163-167. (No abstract available)

Silen JL, Agard DA. The alpha-lytic protease pro-region does not require a physical linkage to activate the protease domain in vivo. Nature. 1989 Oct 5;341(6241):462-4. Abstract

Solomon B, Schwartz F. Chaperone-like effect of monoclonal antibodies on refolding of heat-denatured carboxypeptidase A. J Mol Recognit. 1995 Jan-Apr;8(1-2):72-6. Abstract

Frenkel D, Balass M, Solomon B. N-terminal EFRH sequence of Alzheimer's beta-amyloid peptide represents the epitope of its anti-aggregating antibodies. J Neuroimmunol. 1998 Aug 1;88(1-2):85-90. Abstract

Frenkel D, Balass M, Katchalski-Katzir E, Solomon B. High affinity binding of monoclonal antibodies to the sequential epitope EFRH of beta-amyloid peptide is essential for modulation of fibrillar aggregation. J Neuroimmunol. 1999 Mar 1;95(1-2):136-42. Abstract

Peretz D. Antibodies inhibit prion propagation and clear cell cultures of prion infectivity. Nature. 2001 Aug 16; 412(6848):739-43. Abstract

Hanan E, Goren O, Eshkenazy M, Solomon B. Immunomodulation of the human prion peptide 106-126 aggregation. Biochem Biophys Res Commun. 2001 Jan 12;280(1):115-20. Abstract

Hanan E and Solomon B. Protective effect of monoclonal antibodies against Alzheimer’s beta-amyloid aggregation. Amyloid: Int. J. Exp. Clin. Invest. 1996;3:130-133. (No abstract available)

Solomon B. Disaggregation of Alzheimer beta-amyloid by site-directed mAb. Proc Natl Acad Sci U S A. 1997 Apr 15. Abstract

Bacskai BJ. Non-Fc-mediated mechanisms are involved in clearance of amyloid-beta in vivo by immunotherapy. J Neurosci. 2002 Sep 15; 22(18):7873-8. Abstract

Frenkel D. Immunization against Alzheimer's beta -amyloid plaques via EFRH phage administration. Proc Natl Acad Sci U S A. 2000 Oct 10; 97(21):11455-9. Abstract

Frenkel D. Reduction of beta-amyloid plaques in brain of transgenic mouse model of Alzheimer's disease by EFRH-phage immunization. Vaccine. 2003 Mar 7; 21(11-12):1060-5. Abstract

Solomon B. Immunological approaches as therapy for Alzheimer's disease. Expert Opin Biol Ther. 2002 Dec ; 2(8):907-17. Abstract

View all comments by Beka Solomon

By administering 2 mg of anti-prion antibodies twice a week, White et al. achieved a substantially better therapeutic effect than we did by injecting 50 μg once a week. Although extrapolation of an effective dose in a mouse to a human dose is not an exact science, 2 mg/20 g mouse corresponds to a 6 g/60 kg individual. Hopefully, a...  Read more

By administering 2 mg of anti-prion antibodies twice a week, White et al. achieved a substantially better therapeutic effect than we did by injecting 50 μg once a week. Although extrapolation of an effective dose in a mouse to a human dose is not an exact science, 2 mg/20 g mouse corresponds to a 6 g/60 kg individual. Hopefully, a clinically effective prophylactic dose in humans will be closer to the dose we administered.

To avoid any misunderstanding, we would like to point out that in one of our effective treatment paradigms, we initiated active immunization 24 hours after scrapie infection (Sigurdsson et al., 2002). This would result in detectable anti-prion antibodies probably over a month following exposure. Although this rescue approach had a less dramatic effect than our prophylactic treatment, it cannot be interpreted as a simple neutralization of the inoculum, as stated in the White et al. paper. Rather, it indicates that the antibodies may in some way be interfering with PrP^Sc-mediated conversion of PrP^C to PrP^Sc, and/or increasing clearance of endogenous PrP^Sc.

View all comments by Blas Frangione

Alzheimer's disease vaccine danger: take it straightforward, not double-edged.

Alexei R. Koudinov

BMJ online (23 March 2003) [ FullText ]

View all comments by Alexei R. Koudinov

Dr. Marchesi provides a thoughtful summary of our study and calls attention to an issue that's of central concern to us—specificity. AD is so memory-specific, especially early on, that one would hope to identify molecular pathogens capable of explaining this key feature of the disease. It's turning out that the property of specificity is a most intriguing aspect of ADDL nerve cell biology.

As Dr. Pascale Lacor will show in her poster at SFN-New Orleans, those hot spots of ADDL binding are neither random nor nonspecific. They actually are just what the doctor ordered—synapses. And when ADDLs get lodged in those synapses, they disrupt particular molecular mechanisms essential for memory. (Since Dr. Lacor's study is out for review, I won't comment further on its details.) The bottom line is that the specific manner in which ADDLs attack neurons can provide a synaptically localized mechanism to account for memory loss in AD.

However, even with these further interesting findings, we would, of course, agree that as...  Read more

Dr. Marchesi provides a thoughtful summary of our study and calls attention to an issue that's of central concern to us—specificity. AD is so memory-specific, especially early on, that one would hope to identify molecular pathogens capable of explaining this key feature of the disease. It's turning out that the property of specificity is a most intriguing aspect of ADDL nerve cell biology.

As Dr. Pascale Lacor will show in her poster at SFN-New Orleans, those hot spots of ADDL binding are neither random nor nonspecific. They actually are just what the doctor ordered—synapses. And when ADDLs get lodged in those synapses, they disrupt particular molecular mechanisms essential for memory. (Since Dr. Lacor's study is out for review, I won't comment further on its details.) The bottom line is that the specific manner in which ADDLs attack neurons can provide a synaptically localized mechanism to account for memory loss in AD.

However, even with these further interesting findings, we would, of course, agree that as always, more evidence is desirable.

View all comments by William Klein

These results differ from those reported by Hock et al., where humans vaccinated against Aβ did not reveal detectable increases in circulating Aβ, suggesting that the antibodies generated in humans did not create a peripheral sink for Aβ. However, it is important to recognize that measurement of serum Aβ and anti-Aβ antibodies may be complicated when both agents are present in the sample to be evaluated. Certainly, if an antibody against...  Read more

These results differ from those reported by Hock et al., where humans vaccinated against Aβ did not reveal detectable increases in circulating Aβ, suggesting that the antibodies generated in humans did not create a peripheral sink for Aβ. However, it is important to recognize that measurement of serum Aβ and anti-Aβ antibodies may be complicated when both agents are present in the sample to be evaluated. Certainly, if an antibody against Aβ is bound to circulating Aβ peptide before placing the serum into an ELISA assay, the antibody cannot bind to additional Aβ tethered to the ELISA plate. For high-affinity antibody-antigen interactions, the off rate may be too slow for dissociation to occur during the period of incubation on the ELISA plate, and the antibody concentration will be underestimated. We have evidence that this sort of antibody masking does occur in transgenic mice when antibody titers are not in excess of circulating Aβ (Li et al., in review).

Conversely, measurement of Aβ may also be modified in sandwich ELISA assays by the presence of anti-Aβ antibodies derived from the serum. First, if the circulating anti-Aβ antibody and the capture antibody have overlapping epitopes, they may compete and prevent the Aβ from being captured and thus detected by the ELISA. However, if the two epitopes do not overlap, permitting capture of Aβ still bound to the circulating host antibody, and the detection antibody can also bind the Aβ, there is an opportunity for magnification of the signal. Assuming a secondary antibody binding the detection antibody can cross-react with the circulating host antibody, the apparent signal may be doubled, relative to a standard curve made from Aβ without attached antibody.

These complications make direct comparisons between papers difficult. Often, manuscripts do not provide the detailed steps used for the ELISAs measuring Aβ and anti-Aβ antibodies, as these are viewed as standard techniques within the respective laboratories. However, the antibodies used and their extent of cross-reactivity and epitope overlap may be important to the overall results obtained. Even the time that sera are in a diluted state may influence the results, depending upon antibody-Aβ dissociation rates To avoid these problems, we have recently started dissociating serum antibody-bound Aβ with a mild acid denaturation step (pH 2.5) followed by centrifugation through a size sieving filter to separate Aβ and antibody prior to ELISA. Obviously, other techniques may be used that accomplish the same result.

Thus, the question regarding a peripheral sink for Aβ remains with regard to humans vaccinated against Aβ. Our view of the literature, coupled with our own data, finds support for at least three mechanisms by which immunotherapy lowers Aβ in transgenic mouse models of amyloid deposition (Wilcock et al., 2003; Wilcock et al., 2004) It would be surprising if all three were not also at work in humans vaccinated against Aβ. The work from Gandy et al. would suggest that more detailed and controlled analyses will be needed to reach a final conclusion.

View all comments by Dave Morgan

The changes in treated monkeys of plasma levels of Aβ, similar to those found in young AD transgenic mice before plaque appearance, may support the peripheral sink theory (2). Treatment with intravenous immunoglobulin (IVIG), containing natural anti-Aβ antibodies, of elderly people suffering from neurological diseases other than AD (such as multiple sclerosis, peripheral neuropathy, LEMNS, dermatomyositis) showed a similar pattern of reduction of CSF Aβ and Aβ42 and an increase of CSF anti-Aβ antibodies as compared to the baseline. Total serum Aβ and anti-Aβ antibodies both increased, with a nonsignificant trend toward increased serum Aβ42 after treatment,...  Read more

The changes in treated monkeys of plasma levels of Aβ, similar to those found in young AD transgenic mice before plaque appearance, may support the peripheral sink theory (2). Treatment with intravenous immunoglobulin (IVIG), containing natural anti-Aβ antibodies, of elderly people suffering from neurological diseases other than AD (such as multiple sclerosis, peripheral neuropathy, LEMNS, dermatomyositis) showed a similar pattern of reduction of CSF Aβ and Aβ42 and an increase of CSF anti-Aβ antibodies as compared to the baseline. Total serum Aβ and anti-Aβ antibodies both increased, with a nonsignificant trend toward increased serum Aβ42 after treatment, suggesting the possibility of increased antibody-mediated clearance of Aβ from CSF to serum (3) unrelated to Alzheimer's disease.

In the absence of AD brain pathology, antibodies bind to soluble Aβ and may interfere with the equilibrium between the brain and peripheral Aβ peptide, which supports the sink theory. However, immunotherapy of AD patients who show plaque pathology did not support this theory (4). Therefore, it seems that this research, done on only four monkeys exhibiting no signs of AD, cannot support the sink theory, as appealing as it is.

References
1. Walker LC, Cork LC. The neurobiology of aging in nonhuman primates. In: Terry RD, et al., eds. Alzheimer's Disease. Philadelphia: Lippincott Williams and Wilkins, 1999: 233-243.

2. DeMattos RB, Bales KR, Cummins DJ, Dodart JC, Paul SM, Holtzman DM. Peripheral anti-A β antibody alters CNS and plasma A β clearance and decreases brain A β burden in a mouse model of Alzheimer's disease. Proc Natl Acad Sci U S A. 2001:98(15):8850-5. Abstract

3. Dodel R, Hampel H, Depboylu C et al. Human antibodies against amyloid β peptide: a potential treatment for Alzheimer's disese. Annal Neurol, 2002: 52(2) 253-256. Abstract

4. Hock C, Konietzko U, Streffer JR et al. Antibodies against β-amyloid slow cognitive decline in Alzheimer's disease. Neuron. 2003: 38(4) 547-54. Abstract

View all comments by Beka Solomon

An important effect of immunization that has not been reported on in the current study is cerebral amyloid angiopathy (CAA) and microhemorrhage. It has been shown that passive immunotherapy increases CAA in transgenic mice (Wilcock et al., 2004) and causes increased incidence of microhemorrhage (Pfeifer et al., 2001, Wilcock et al., 2004, Racke et al., 2005). We also reported that these adverse events occurred with active vaccination (Wilcock et al., 2007). Indeed, the authors of the current report use CAA and Aβ accumulation around capillaries as histopathological factors used to determine the degree of amyloid clearance. It also seems the microhemorrhage occurrence will be difficult to overcome. The recent report from Schroeter et al. showed that even low doses of antibody, which were associated with essentially no amyloid removal, resulted in an apparent subtle increase in microhemorrhage (Schroeter et al., 2008; control mice had no animals with microhemorrhage rated 2 or 3 while the lowest dose of 3D6 had three mice rated 2 or 3). Accumulation of CAA and associated microhemorrhage likely contributes significantly to the clinical progression of disease. Additionally, as the authors suggest, a change in inflammatory state could certainly contribute to further cognitive decline. Recent data show that the inflammatory profile of Alzheimer’s and transgenic mouse brain is highly complex (Colton et al., 2006). It is likely that Fcγ receptor activation affects the inflammatory state.

These data highlight the significant differences between human and mouse studies. Since neurodegeneration is not abundant in the majority of mouse models, it has not been possible, to date, to study this. It is likely that while amyloid may initiate the cascade, neurodegeneration may be self-perpetuating and neuroprotection may also be critical for successful anti-amyloid therapeutics. It has been suggested that passive immunization will overcome some of the limitations of active vaccination, and we certainly eagerly anticipate the data from Elan’s passive immunization trial of bapineuzumab.

References:
Colton CA, Mott RT, Sharpe H, Xu Q, Van Nostrand WE, Vitek MP. Expression profiles for macrophage alternative activation genes in AD and in mouse models of AD. J Neuroinflammation. 2006 Sep 27;3:27. Abstract

Hock C, Konietzko U, Streffer JR, Tracy J, Signorell A, Müller-Tillmanns B, Lemke U, Henke K, Moritz E, Garcia E, Wollmer MA, Umbricht D, de Quervain DJ, Hofmann M, Maddalena A, Papassotiropoulos A, Nitsch RM. Antibodies against beta-amyloid slow cognitive decline in Alzheimer's disease. Neuron 2003 May 22;38(4):547-554. Abstract

Pfeifer M, Boncristiano S, Bondolfi L, Stalder A, Deller T, Staufenbiel M, Mathews PM, Jucker M. Cerebral hemorrhage after passive anti-Abeta immunotherapy. Science 2002 Nov 15;298 (5597):299. Abstract

Racke MM, Boone LI, Hepburn DL, Parsadainian M, Bryan MT, Ness DK, Piroozi KS, Jordan WH, Brown DD, Hoffman WP, Holtzman DM, Bales KR, Gitter BD, May PC, Paul SM, DeMattos RB. Exacerbation of cerebral amyloid angiopathy-associated microhemorrhage in amyloid precursor protein transgenic mice by immunotherapy is dependent on antibody recognition of deposited forms of amyloid beta. J Neurosci 2005 Jan 19;25(3):629-636. Abstract

Schroeter S, Khan K, Barbour R, Doan M, Chen M, Guido T, Gill D, Basi G, Schenk D, Seubert P, Games D. Immunotherapy reduces vascular amyloid-β in PDAPP mice. J Neurosci 2008 Jul 2; 28(27): 6787-6793. Abstract

Wilcock DM, Rojiani A, Rosenthal A, Subbarao S, Freeman MJ, Gordon MN, Morgan D. Passive immunotherapy against Abeta in aged APP-transgenic mice reverses cognitive deficits and depletes parenchymal amyloid deposits in spite of increased vascular amyloid and microhemorrhage. J Neuroinflammation 2004 Dec 8;1(1):24. Abstract

Wilcock DM, Jantzen PT, Li Q, Morgan D, Gordon MN. Amyloid-beta vaccination, but not nitro-nonsteroidal anti-inflammatory drug treatment, increases vascular amyloid and microhemorrhage while both reduce parenchymal amyloid. Neuroscience. 2007 Feb 9;144(3):950-60. Abstract

View all comments by Donna Wilcock

We suggest a number of possible explanations for our findings:

1. The...  Read more

We suggest a number of possible explanations for our findings:

1. The presence of Aβ plaques is required to initiate, but not to maintain the progressive neurodegeneration in AD.

2. Amyloid plaques are an epiphenomenon, and extracellular soluble/oligomeric or intraneuronal forms of Aβ are responsible for the neurodegeneration in AD.

3. Immunization activates microglia, which may be beneficial (by removing plaques) but at the same time neurotoxic.

4. The plaques could have been removed shortly before the patients died, after their cognitive function had declined—this seems rather unlikely.

A major driver of the immunization strategies currently in clinical trials has been to avoid a T lymphocyte reaction in the belief that this is what underlay the side effect noted in the second study of AN1792. Passive immunization, in particular, should theoretically not be able to provoke a T cell response and has the additional benefit that the bioavailability of the antibodies can be controlled. However, it is not clear by what mechanism lymphocytes in the leptomeninges, identified in patients with the side effect, can cause changes in the cerebral white matter (a consistent feature on brain imaging of the affected patients). An alternative explanation for the side effect is that it was due to disaggregation and solubilization of plaque Aβ which then tracks to the cerebral vasculature, increasing the severity of cerebral amyloid angiopathy (CAA). We know from previous studies that severe cerebral amyloid angiopathy is associated with abnormalities in the white matter. Interestingly, the new information from the current Elan trial of passive immunization (bapineuzumab) seems to be showing evidence of white matter abnormalities which are occurring more frequently in patients with ApoE4—known to be associated with more severe CAA.

On the basis of our findings, we would predict that other immunization protocols (e.g., passive immunization and active immunization with truncated versions of the Aβ peptide) will also be effective in clearance of plaques. A number of current studies have before and after immunization in vivo plaque imaging, for example, with PIB, built into their design. We would predict that these will demonstrate plaque clearance following immunization. However, on the basis of our findings, we would speculate that plaque removal will not correlate well with any changes in cognitive function.

It is possible that some of the new immunization protocols will have a different balance of effects on the different forms of Aβ (e.g., plaque, soluble, oligomeric, intraneuronal) and may therefore have different effects on cognitive function. One of the approaches being trialled involves passive immunization with an Fc-truncated antibody, and this may have the potentially beneficial effect of not provoking microglial activation.

Using immunization as prevention rather than treatment would likely avoid these complications which seem to be due to the presence of substantial quantities of Aβ already being present in the brain. On the basis of the animal studies, immunization at a young age can prevent the formation of plaques in later life. Of course, we don’t yet know if this can be done safely in humans—we don’t know the physiological function of Aβ and if immunization might interfere with this function. A study to determine if Aβ immunization at a young age could prevent the development of AD later in life would be the ultimate test of the Aβ hypothesis.

View all comments by Delphine Boche
View all comments by Clive Holmes
View all comments by James Nicoll

The Next Phase: Prevention. Where Do I Sign Up?
The Aβ vaccination strategy failed because it was not used early enough in the course of the disease.

Come again?

We know this apparently because Aβ oligomers, which are artifacts of ultracentrifugation, when injected into the ventricles of mice, cause mice to navigate water mazes poorly, and press levers inappropriately. We know this because when hippocampal slices are bathed in a suspension of the artifact, they demonstrate electrophysiological abnormalities. And we know this because transgenic mice, which are engineered to overproduce Aβ, and then administered antibodies against it, improve in their ability to navigate water mazes and press the appropriate levers.

We apparently also must set aside the ad hoc revisions and contortions of the amyloid cascade hypothesis over the years (1-3), and the plethora of problems with experimental AD models, from lack of cognitive dysfunction, to lack of...  Read more

The Next Phase: Prevention. Where Do I Sign Up?
The Aβ vaccination strategy failed because it was not used early enough in the course of the disease.

Come again?

We know this apparently because Aβ oligomers, which are artifacts of ultracentrifugation, when injected into the ventricles of mice, cause mice to navigate water mazes poorly, and press levers inappropriately. We know this because when hippocampal slices are bathed in a suspension of the artifact, they demonstrate electrophysiological abnormalities. And we know this because transgenic mice, which are engineered to overproduce Aβ, and then administered antibodies against it, improve in their ability to navigate water mazes and press the appropriate levers.

We apparently also must set aside the ad hoc revisions and contortions of the amyloid cascade hypothesis over the years (1-3), and the plethora of problems with experimental AD models, from lack of cognitive dysfunction, to lack of neuronal loss, to necessity of multiple mutations, to hyperphysiologic production of a target protein. We set this aside because we apparently now know that synaptic damage, a process never directly assessed, and which probably has the same specificity as gliosis, is the pathological substrate for this laboratory artifact in AD (1).

So a strategy, founded in the analysis of a pathological lesion (once said to be toxic and now discarded as a distraction, except of course for the two subjects who found to be “cleared” of plaques at autopsy), based on an ad hoc modification of a hypothesis that a laboratory artifact specifically causes nonspecific damage that has never been analyzed directly, verified in a transgenic mouse construct that generally does not lose neurons, and which was tested and failed in human disease subjects, must now be used on normal people. Where do I sign up?

References:
1. Castellani RJ, Lee HG, Zhu X, Perry G, Smith MA. Alzheimer disease pathology as a host response. J Neuropathol Exp Neurol. 2008;67:523-531. Abstract

2. Castellani RJ, Lee HG, Zhu X, Nunomura A, Perry G, Smith MA. Neuropathology of Alzheimer disease: pathognomonic but not pathogenic. Acta Neuropathol. 2006;111:503-509. Abstract

3. Smith MA, Casadesus G, Joseph JA, Perry G. Amyloid-beta and tau serve antioxidant functions in the aging and Alzheimer brain. Free Radic Biol Med. 2002;33:1194-1199. Abstract

View all comments by Rudy Castellani
View all comments by Hyoung-gon Lee
View all comments by George Perry
View all comments by Mark A. Smith
View all comments by Xiongwei Zhu

I have often used the analogy that anti-Aβ therapy for AD is analogous to treating patients whose coronary arteries are 99 percent clogged with a statin and hoping for a clinical effect. These new data raise the possibility that anti-Aβ immunotherapy is more like trying to treat somebody with massive myocardial contraction deficits following multiple MIs with a statin and a bypass. So much damage has been done that targeting the trigger, by itself, is simply too little too late. Indeed, we would not approach the treatment of a patient...  Read more

I have often used the analogy that anti-Aβ therapy for AD is analogous to treating patients whose coronary arteries are 99 percent clogged with a statin and hoping for a clinical effect. These new data raise the possibility that anti-Aβ immunotherapy is more like trying to treat somebody with massive myocardial contraction deficits following multiple MIs with a statin and a bypass. So much damage has been done that targeting the trigger, by itself, is simply too little too late. Indeed, we would not approach the treatment of a patient in severe cardiac dysfunction that has resulted from multiple MIs as a result of long-standing atherosclerotic disease with a statin alone. It simply is not going to work, though it might have some benefit in combination with other therapeutic agents.

Though a small and vociferous group of colleagues are publicly using such data to refute the role of Aβ aggregation in AD and thus indirectly attempting to invalidate it, Aβ or Aβ aggregates, as a target, I think a more parsimonious approach and one discussed to an extent by the authors is to really think carefully about these data and how we as a field might modify our approach to AD therapy and research based on such studies. Although there are numerous potential implications of these data, I will limit myself to a few issues that I see as most important. Obviously, the following comments may be tempered somewhat by any future demonstration of efficacy in Phase 3 studies of anti-Aβ therapies, but I think they will likely hold even in that event.

From a basic research point of view, this ups the ante on two critical issues.

In order to enable better preclinical studies, we still need better animal models of AD that fully recapitulate all the features of the human disease—especially neuronal loss. Given that this appears very difficult to do in APP mice, we probably need to consider looking at other species. Indeed, this report suggests the AN1792 trial appears to have “worked” in humans as it did in mice. Of course, APP mice are good models of Aβ deposition but not real models of AD. If we had a complete animal model of AD, we might be better able to evaluate therapeutic paradigms for impact on neurodegeneration. Tau mice might be better predictors for effects on neuronal loss, but obviously aren’t much use for testing anti-Aβ therapies. Hopefully they will be predictive of clinical outcomes when novel anti-tau therapies are moved into the clinic.

We need a real understanding of why neurons die in AD, and we need to identify additional therapeutic targets that will protect or restore neuronal function. Indeed, though my own research is Aβ-centric, I believe it is of paramount importance to identify targets beyond Aβ and, for that matter, tau. I think that it is more important to explicitly state that we need additional targets than to try and invalidate current ones.

From a clinical perspective, I think this reinforces our need to figure out how to prophylactically treat AD. We need to directly confront and overcome the challenges that distinguish therapeutic trials from prevention trials. We also need to figure out whether a trial of MCI of the AD type to AD conversion is really a prevention trial or just a very early therapeutic trial. Current predictive AD biomarker initiatives will certainly help to frame and define some aspects of the problem in more detail, but we also need to find common ground on how to actually execute a prophylactic trial that is economically feasible, ethical, and appropriately powered. Such trials will almost certainly require the joint efforts of academic, government, and commercial sectors, and of course, “safe agents.” Indeed, the true test of the Aβ “aggregate/amyloid” hypothesis of AD is a trial to prevent Aβ deposition in humans, not a therapeutic treatment of patients with clinical symptoms.

On a final, more technical note, following the initial report (Nicoll et al., 2003) of plaque clearance in one patient, I was less than convinced that there was clearance. The new data do make me more convinced. However, I would like to see some rigorous biochemical analysis of Aβ levels in the brains of these subjects. Even in mouse models, “plaque loads” seem to overestimate reductions in Aβ as compared to biochemical measures. I am also struck by what appears to be patchy clearance. I find it hard to rationalize how patchy clearance can occur with an antibody-mediated mechanism and wonder whether cellular immune responses play some role in the actual clearance.

References:
Golde TE. Alzheimer disease therapy: can the amyloid cascade be halted? J Clin Invest. 2003 Jan;111(1):11-8. Abstract

Golde TE. Disease modifying therapy for AD? J Neurochem. 2006 Nov;99(3):689-707. Abstract

Nicoll JA, Wilkinson D, Holmes C, Steart P, Markham H, Weller RO. Neuropathology of human Alzheimer disease after immunization with amyloid-beta peptide: a case report. Nat Med. 2003 Apr;9(4):448-52. Abstract

View all comments by Todd E. Golde

There are a few issues that I’d like to comment on. I find it noteworthy that seven out of eight cases had MMSE scores of zero when last screened. The authors point out that these were “end stage” AD cases—and judging from the MMSE scores, that’s an understatement. I agree with Todd Golde...  Read more

There are a few issues that I’d like to comment on. I find it noteworthy that seven out of eight cases had MMSE scores of zero when last screened. The authors point out that these were “end stage” AD cases—and judging from the MMSE scores, that’s an understatement. I agree with Todd Golde that AD immunotherapy in this small, severely affected cohort is not a robust test of the amyloid cascade hypothesis in humans. But, I don’t believe that this detracts at all from the provocative nature of the findings, and from the message to keep an open mind and to critically consider the etiological contribution of Aβ to AD. More than likely, what these data are telling us is that there is a cutoff beyond which severe neuronal damage/loss has already occurred, and removing Aβ from the equation will have little if any effect clinically. This has prompted a number of researchers to conclude that prevention by immunotherapy is a more viable strategy. That may be true, but when should vaccination be initiated—five, 10, 20, or more years before symptoms manifest? Also, what biomarkers should be used to determine those at risk: APOE genotype, CSF Aβ, CSF tau, plasma Aβ? At this stage, a preventative Aβ vaccine seems much more viable for the <5 percent of individuals genetically predisposed to familial AD, but again—when would treatment need to be initiated and how often would it need to be given to be efficacious and safe?

It is unfortunate that an adjuvant-alone (placebo) treatment group could not be evaluated side-by-side with the AN1792-treated cases, and that historical non-vaccinated AD cases had to be used as controls. It is possible that the inflammatory side effects of the Th1-biasing vaccine adjuvant (QS-21) negatively impacted cognitive function and/or survival independently of the synthetic Aβ42 peptide. Along those lines, the authors comment that “only one patient had clinical features of meningoencephalitis….” Did the authors evaluate CD4+ T cells in these vaccinated cases, and if so, were they present in greater quantity than in the historical non-vaccinated AD cases?

In summary, this paper represents a timely, thought-provoking examination of the clinical and pathological correlates of Aβ vaccination. As we move forward in this exciting time of AD therapeutics, it will be important to view the results of such clinical trials with open eyes and without bias toward whichever AD pathogenic hypothesis we hold close to our hearts.

References:
Holmes C, Boche D, Wilkinson D, Yadegarfar G, Hopkins V, Bayer A, Jones RW, Bullock R, Love S, Neal JW, Zotova E, Nicoll JA. Long-term effects of Abeta42 immunisation in Alzheimer’s disease: follow-up of a randomized, placebo-controlled phase I trial. Lancet 2008 July 19;372:216-223. Abstract

Nicoll JA, Wilkinson D, Holmes C, Steart P, Markham H, Weller RO. Neuropathology of human Alzheimer disease after immunization with amyloid-beta peptide: a case report. Nat Med. 2003 Apr;9(4):448-52. Abstract

View all comments by Terrence Town

Data derived from our ongoing clinical molecular pathologic investigations of MCI using the cholinotrophic basal forebrain system as a model for neuronal selective vulnerability has shown that these neurons display a...  Read more

Data derived from our ongoing clinical molecular pathologic investigations of MCI using the cholinotrophic basal forebrain system as a model for neuronal selective vulnerability has shown that these neurons display a myriad of biochemical and molecular alterations, which appear to be unrelated to amyloid deposition (Counts and Mufson, 2004). For example, cholinergic neurons are simultaneously undergoing re-expression of cell cycle markers, alterations in neurotrophic support and the ratio of tau epitopes but not changes in APP or presenilin expression. The molecular signature of these neurons is commensurate with a hypothesis related to multiple cellular and connectivity-based dysregulation, which probably begins several decades before the onset of clinical symptoms. What initiates neuronal dysfunction remains unknown, and merits serious research in relevant animal models as well as in well-characterized postmortem human brain tissues. In this regard, it would be of interest to examine the molecular pathology of the cholinergic basal forebrain (CBF) neurons in the same vaccine treated brains examined by Holmes et al. to determine whether amyloid removal from cortical and hippocampal parenchymal projection sites of the CBF neurons rejuvenates these cell bodies.

To anyone who has ever examined the brain of a patient with AD, it is evident that the disease is not simply an amyloidosis. AD is a multi-neuronal system disconnection syndrome of unknown etiology, with pronounced selective cell loss, synaptic dysfunction, atrophy, vascular pathology, tau pathology, in addition to intracellular Aβ disturbances and extracellular amyloid deposition, among other problems that may yet be discovered. It is not our intention to advocate any singular hypothesis of AD, rather to suggest that other treatment approaches and modalities should be pursued with a solid federal and private funding base in addition to amyloid-based clinical trials. An effective treatment will ultimately be a poly-pharmaceutical approach that targets both mechanisms underlying neurodegeneration as well as symptoms of cognitive decline until the etiology of AD is revealed.

References:
Forman, M.S. Mufson, E.J., Leurgans, S., Pratico, D., Joyce, S., Leight, S., Lee, V.M.-Y. and J.Q. Trojanowski: Cortical Biochemistry in MCI and Alzheimer Disease, Neurology, 68: 757-763, 2007. Abstract

Mufson, E. J., Chen, E-Y., Cochran, E. J., Beckett, L. A., Bennett, D. A. and Kordower, J. H.: Entorhinal cortex beta amyloid load in individuals with mild cognitive impairment. Exp. Neurol., 158, 469-490, 1999. Abstract

Counts, S.E. and Mufson, E. J.: The role of nerve growth factor receptors in cholinergic basal forebrain degeneration in prodromal Alzheimer’s disease, J. Neuropath. Exper. Neurol., 64, 263-272, 2005. Abstract

View all comments by Stephen D. Ginsberg
View all comments by Elliott Mufson