Iturria-Medina Y, Sotero RC, Toussaint PJ, Mateos-Pérez JM, Evans AC, Alzheimer’s Disease Neuroimaging Initiative. Early role of vascular dysregulation on late-onset Alzheimer's disease based on multifactorial data-driven analysis. Nat Commun. 2016 Jun 21;7:11934. PubMed.

Recommends

Please login to recommend the paper.

Comments

  1. I’m not an expert on the mathematical modeling done here. While it’s clear that vascular abnormalities can contribute to cognitive decline, I am very concerned about the conclusion drawn from this data, namely that vascular dysregulation is the earliest event leading to LOAD.

    One concern is that if you look at the results of their data analysis in Figure 3, it shows that Aβ42 changes after amyloid deposition detected by amyloid imaging using florbetapir. Florbetapir is a good amyloid imaging agent but it is not sensitive. We know that CSF Aβ42 begins to drop right at or likely before amyloid imaging positivity as assessed in longitudinal and cross-sectional studies done with PiB (a more sensitive amyloid imaging agent than florbetapir). It is also clear from several longitudinal and cross-sectional studies that functional decline does not come before decreases in CSF Aβ42 (shown in Figure 3). I am concerned that either the modeling used or the way the analysis from the ADNI dataset was done has led to the wrong conclusions. The ADNI dataset has more individuals who are impaired than non-impaired. Also, the subjects are relatively old at entry. It would be better to try to model this type of data from longitudinal studies that start with people at younger ages (40-70), when AD pathology is usually beginning, and who develop clinical disease in their 70s and 80s. This is being done in a number of data sets.

    View all comments by David Holtzman

  2. Iturria-Medina and colleagues report a new analysis of the ADNI data that intriguingly suggests an early role for vascular dysregulation in AD pathophysiology. Their approach is a tour de force in the field of progression tracking. Although others have used similar techniques for aligning biomarkers along a common age axis, this study stands out in its thoroughness and in the amount of data sets and biomarkers used.

    That vascular effects precede other abnormalities is easily the most interesting finding. This idea has of course been proposed before and this study provides a strong statistical support to it. Practically, these results might persuade investigators to propose the inclusion of high quality MR perfusion sequences in future research and clinical protocols.

    To me what is even more interesting is that this study allows new mathematical models of progression to be encoded and tested. For instance, vascular maps could be used as starting patterns that could drive models of subsequent brain-wide spread using the Network Diffusion framework or other models of networked spread. 

    View all comments by Ashish Raj

  3. The paper by Iturria-Medina et al. reminds us how little we know about the biological predictors, modulators, correlates, and causes contributing to late-onset Alzheimer type dementia (LOAD). Their multimodality study based on the entire ADNI cohort reveals that among the multiple domains of acquired data, vascular factors carry the greatest weight in characterizing the trajectory of cognitive decline. Moreover, these data challenge existing views of the primacy of Aβ lesions as the first detectable feature of the LOAD process. Rather, the authors point us toward vascular-mediated defects that have a role in the clearance of brain waste products, which of course includes products of Aβ metabolism. This paper should encourage us to consider the subject composition from large-scale national projects, the biological measures to invest in, and the evolution of high dimensional data analytic procedures such as first presented here.

    View all comments by Mony de Leon

  4. Iturria-Medina and colleagues provide definitive confirmation of two facts about the Alzheimer's pathophysiological cascade that have been underappreciated, despite cumulative evidence from several groups: First, that memory dysfunction may occur prior to abnormalities in biomarkers for amyloid and tau, the classic pathophysiological hallmarks of Alzheimer's disease. This suggests that another, earlier pathology may initiate cognitive decline. This leads us to their second finding, which is that changes in the cerebrovascular system (vascular dysregulation) precede cerebral amyloidosis and tau-mediated neurodegeneration, providing a potential explanation for early cognitive changes. Their findings concur very well with our recent observations that loss of cerebrovascular integrity and blood-brain barrier breakdown are amongst the first changes in the brain during normal aging and in individuals with mild cognitive impairment (Montagne et al., 2015). Given that findings by Iturria-Medina et al. are from one of the largest and most comprehensive biomarker analyses ever conducted, it is probably good time for us to begin thinking how to update our model of the Alzheimer's pathophysiological cascade by including vascular dysfunction as a driver of disease pathogenesis.

    References:

    Montagne A, Barnes SR, Sweeney MD, Halliday MR, Sagare AP, Zhao Z, Toga AW, Jacobs RE, Liu CY, Amezcua L, Harrington MG, Chui HC, Law M, Zlokovic BV. Blood-brain barrier breakdown in the aging human hippocampus. Neuron. 2015 Jan 21;85(2):296-302. PubMed.

    View all comments by Daniel Nation

  5. In the paper Iturria-Medina et al. analyzed 7,700 multimodality brain images and tens of different plasma and CSF biomarkers from 1,171 healthy patients and from patients with late-onset Alzheimer's disease (LOAD). A data-driven approach revealed multiple factors responsible for the progression of Alzheimer's disease. One of the major conclusions is that vascular dysregulation may be the earliest and strongest pathological factor associated with LOAD, followed in order by deposition of Aβ, dysregulation of glucose metabolism, functional impairment, and atrophy of the gray matter

    Such an approach promises to be extremely valuable, particularly as it is difficult to determine the causative factors in Alzheimer's disease from those factors that are secondary effects. The data-driven approach using an evaluation of misfolded amyloid proteins, glucose metabolism, cerebral blood flow, functional activity, and neuroimaging revealed a characteristic trajectory for each biological factor during the development of Alzheimer's disease.

    Valuable data have been gathered from the pathological examination of the brains, CSF, and blood over the last 100 years. The present data-driven approach is a valuable addition for those seeking to determine the major factors that initiate Alzheimer's disease and result in its progression. The hypothesis that vascular dysregulation may be the earliest and strongest pathological factor in Alzheimer's disease requires further investigation. Arteries and capillaries in the brain not only deliver blood and nutrients, they are also the route for the elimination of fluid and soluble metabolites from the brain in the maintenance of tissue homoeostasis. Although the authors of the paper recognized that failure of elimination of Aβ is a major factor in the pathogenesis of Alzheimer's disease, they concentrate on absorption of Aβ into the blood. In assessing the role of arteries and capillaries in the brain, the perivascular route for elimination of fluid and solutes, including Aβ, should not be ignored. Experimental studies have shown that perivascular elimination of fluid and solutes is impaired with increasing age, and this is reflected in the accumulation of Aβ in the walls of capillaries and arteries as cerebral amyloid angiopathy in Alzheimer's disease (Carare et al., 2013; Weller et al., 2015). 

    The current paper is a significant step forward in our understanding of the pathogenesis and progression of Alzheimer's disease. Its value will be increased by correlating the data with previous observations, particularly on the pathophysiology of cerebral arteries and capillaries in the elimination of soluble metabolites, including Aβ in the maintenance of homoeostasis of the brain. 

    References:

    Carare RO, Hawkes CA, Jeffrey M, Kalaria RN, Weller RO. Cerebral amyloid angiopathy, Prion angiopathy, CADASIL and the spectrum of Protein Elimination-Failure Angiopathies (PEFA) in neurodegenerative disease with a focus on therapy. Neuropathol Appl Neurobiol. 2013 Mar 13; PubMed.

    Weller RO, Hawkes CA, Carare RO, Hardy J. Does the difference between PART and Alzheimer's disease lie in the age-related changes in cerebral arteries that trigger the accumulation of Aβ and propagation of tau?. Acta Neuropathol. 2015 May;129(5):763-6. Epub 2015 Mar 27 PubMed.

    View all comments by Roy Weller

  6. One remarkable finding of this interesting study is the observation that the CSF Aβ1-42 abnormality index only increases to 30 percent at the late AD stage. However, amyloid PET binding (which correlates around 0.8 with CSF Aβ1-42 in ADNI [Toledo et al., 2015]) reaches a much higher abnormality index than CSF Aβ1-42, namely around 90 percent. One possible explanation could be that CSF Aβ1-42 plateaus earlier than amyloid PET. As pointed out above, a more detailed definition of the abnormality index would be useful to better understand these findings.

    References:

    Toledo JB, Bjerke M, Da X, Landau SM, Foster NL, Jagust W, Jack C Jr, Weiner M, Davatzikos C, Shaw LM, Trojanowski JQ, Alzheimer’s Disease Neuroimaging Initiative Investigators. Nonlinear Association Between Cerebrospinal Fluid and Florbetapir F-18 β-Amyloid Measures Across the Spectrum of Alzheimer Disease. JAMA Neurol. 2015 May;72(5):571-81. PubMed.

    View all comments by Pieter Jelle Visser

  7. I tend to agree with David Holtzman. One thing unclear to me is why the authors did not distinguish between ApoE4 carriers and non-carriers in their mathematical modeling. This is because ApoE4 genotype is an independent risk factor for vascular dysfunction and also for CAA. The errors that arose during computation might have been smaller and could have given different conclusions.

    View all comments by Takaomi Saido

  8. This study is very impressive. The list of consortium members is gigantic, and the amount of work amazing. There are, however, a couple of points that trigger me to comment on such an impressive piece of work and the role it assigns to early vascular dysregulation on LOAD.

    The authors themselves are very careful and warn not to over-interpret the results. They state this clearly: “Although the obtained abnormality trajectories may be reflecting a tentative ordering in which pathophysiological events occur, our results should be interpreted more in terms of biomarker sensitivity to disease progression than in terms of causal pathologic interactions conducive of LOAD.”

    Nevertheless, you immediately read in a comment on this paper that “memory dysfunction may occur prior to abnormalities in biomarkers for amyloid and tau”—something that is in direct contradiction to many other findings, as pointed out by another commenter. Many readers of this paper will make the same kind of mistake to find support in this large study for their alternative hypotheses to explain LOAD.

    However, detecting a fever before the causative pathogen can been detected does not mean that the body temperature increases before the infection. It just means that you can measure the increase in body temperature before you can detect the causative infectious agent, but the infection always occurs first.

    Many scientists just want to believe that AD is a highly complex disorder that must have a high number of complex mechanisms underlying the disease. Although I do not reject this hypothesis and I am convinced that to some extent it is true, it is always interesting to remember that AD is one of the few common diseases that can be explained (in some familial cases) by the simplest mechanism known in medicine. One single base substitution in a single allele of one single gene (that is, a single base substitution in the 6 billion bases of a diploid genome) can be sufficient to cause an autosomal-dominant form of AD that is basically indistinguishable from the more common LOAD.

    This is, of course, in the case of a mutation of the APP gene. We also know that an extra copy of the wild-type APP gene that leads to a 50 percent increase in APP production without any other alteration can also cause a form of AD that is indistinguishable from LOAD. Actually, we do not know many common complex disorders that can be explained by such a minimal, simple, and straightforward etiology.

    The big problem with AD is that, although in this familiar form of the disease everything (and I mean really everything) about the disease can be explained by just an overexpression of APP by 50 percent—the plaques, tangles, neuronal cell death, cognitive decline, etc., etc., etc.—it seems amazing that when you ask what could be causing the overproduction of APP in LOAD, a convincing answer is nowhere to be found. On the contrary, the authors of this impressive paper state that even if their approach “does not reveal causal pathologic interactions, concordant evidence suggests that in LOAD Aβ deposition is mainly caused by a deficiency in the Aβ clearance system rather than by an Aβ overproduction.” They probably favor such an explanation to justify the important role they assign to the early vascular dysregulation described in their impressive piece of work.

    There is, however, nothing concordant about Aβ deposition being caused mainly by a deficiency in clearance rather that production. As indicated above, everything about this disease can be explained by a simple overproduction of APP and thus also the observed decrease in clearance. The authors themselves indicate that “Aβ has vascular destructive activity”.

    The reason so many studies overlook the increase in APP production is because they simply forget to take into account that, in AD brains, an apparently similar level of APP is produced (in some regions) by less than half the number of neurons. On a per-neuron basis, the production may well have doubled. Note that even if this is overlooked, not taken into account, or not detected by a majority of studies, the logic is indisputable. If in AD brains fewer neurons produce the same amount of APP, then the production at the individual neuron level has to increase. It seems reasonable to assume that overproduction at the individual neuron level is toxic to the neuron itself or its direct neighbors.

    In my opinion the AD field too often focuses on the consequences of the disease instead of the causes. I have to admit, from analyzing a plane wreck after a crash it would be difficult to deduce what caused the accident. If the black box indicates a leak in the fuel tank, however, you have a strong clue. The experiments of nature that clearly provide us clues as to what has to happen first and what should follow in AD, irrespective of the sensitivity of the assay used, should be more consistently used in the interpretation of complex experimental results if we are going to find a cure soon for this devastating disorder. There is not much time to lose, as we are unfortunately not getting any younger.

    View all comments by Torik Ayoubi

  9. This is an interesting report, and should be considered in the context of a similar observation that was recently published on FAD subjects (Lee et al., 2016), in which white matter hyperintensities were observed as more prominent in FAD mutation carriers within the DIAN cohort than in related non-carriers, increasing approximately six years before expected symptom onset, with divergence from non-mutation-carrying family members as much as 22 years before symptom onset.

    The findings in that publication were interpreted as suggesting that WMHs are a core feature of AD that should be integrated into pathogenic models of the disease, as well as a potential therapeutic target. Together, these reports speak to vascular abnormalities as contributing to early disease pathogenesis, perhaps as early as the onset of amyloid accumulation. An alternative view might be that amyloid is still driving the disease in the FAD mutation carriers, but with an early prodromal outcome being induction of vascular lesions.

    References:

    Lee S, Viqar F, Zimmerman ME, Narkhede A, Tosto G, Benzinger TL, Marcus DS, Fagan AM, Goate A, Fox NC, Cairns NJ, Holtzman DM, Buckles V, Ghetti B, McDade E, Martins RN, Saykin AJ, Masters CL, Ringman JM, Ryan NS, Förster S, Laske C, Schofield PR, Sperling RA, Salloway S, Correia S, Jack C Jr, Weiner M, Bateman RJ, Morris JC, Mayeux R, Brickman AM, Dominantly Inherited Alzheimer Network. White matter hyperintensities are a core feature of Alzheimer's disease: Evidence from the dominantly inherited Alzheimer network. Ann Neurol. 2016 Jun;79(6):929-39. Epub 2016 Apr 27 PubMed.

    View all comments by Barry Greenberg

  10. I read the paper by Iturria-Medina et al. and the discussions here with great interest, as the research in my laboratory has been focused on the connections between cardiovascular disease and Alzheimer’s disease. While most of the molecules/biomarkers described are clear, there may be some confusion about apo(a) and apoA-I. Please note that these are two different proteins; they are encoded by different genes and are different in structure and function (McLean et al., 1987; Segrest et al., 2000). To avoid any potential confusions, apo(a) has been called the “apo little a” as written in the cardiovascular field. However, apo(a) is called Apo A in the paper by Iturria-Medina et al. Apo(a) associates with low-density lipoproteins (LDL), not high-density lipoproteins (HDL) as described by Iturria-Medina and colleagues. Apo(a) forms lipoprotein particles called lipoprotein(a) [Lp(a)]. ApoA-I is the major protein component of HDLs in the plasma and determines most of HDL functions.

    It has been well established that high levels of apoA-I/HDL reduce the risk of cardiovascular disease, whereas high levels of apo(a)/Lp(a) increase the risk of cardiovascular disease (Davidson and Toth, 2007; Emerging Risk Factors et al., 2009). The cardiovascular risk associated with apo(a)/Lp(a) is further complicated by the size variations of apo(a), due to the existence of polymorphic repeats in its gene (Erqou et al., 2010). In the paper by Iturria-Medina et al., it was the level of apo(a) in the CSF that was associated with the risk of Alzheimer’s disease.

    Previously, other clinical studies have shown that high levels of plasma apoA-I/HDL are associated with better cognitive function and a reduced risk of Alzheimer’s disease (reviewed in Hottman et al., 2014). We and others have demonstrated that genetic manipulation of apoA-I affects the extent of cerebral amyloid angiopathy (CAA) and neuroinflammation in animal models of Alzheimer’s disease (Lefterov et al., 2010; Lewis et al., 2010; Oct 2010 news). However, few studies have investigated the role of apo(a)/Lp(a) in cognition and AD. Since apo(a) and apoA-I play opposite roles in cardiovascular disease and most likely also in Alzheimer’s disease, we should avoid any confusions when these two proteins are discussed.

    References:

    Davidson MH, Toth PP. High-density lipoprotein metabolism: potential therapeutic targets. Am J Cardiol. 2007 Dec 3;100(11 A):n32-40. PubMed.

    Emerging Risk Factors Collaboration, Erqou S, Kaptoge S, Perry PL, Di Angelantonio E, Thompson A, White IR, Marcovina SM, Collins R, Thompson SG, Danesh J. Lipoprotein(a) concentration and the risk of coronary heart disease, stroke, and nonvascular mortality. JAMA. 2009 Jul 22;302(4):412-23. PubMed.

    Erqou S, Thompson A, Di Angelantonio E, Saleheen D, Kaptoge S, Marcovina S, Danesh J. Apolipoprotein(a) isoforms and the risk of vascular disease: systematic review of 40 studies involving 58,000 participants. J Am Coll Cardiol. 2010 May 11;55(19):2160-7. PubMed.

    Hottman DA, Chernick D, Cheng S, Wang Z, Li L. HDL and cognition in neurodegenerative disorders. Neurobiol Dis. 2014 Dec;72 Pt A:22-36. Epub 2014 Aug 13 PubMed.

    Lefterov I, Fitz NF, Cronican AA, Fogg A, Lefterov P, Kodali R, Wetzel R, Koldamova R. Apolipoprotein A-I deficiency increases cerebral amyloid angiopathy and cognitive deficits in APP/PS1DeltaE9 mice. J Biol Chem. 2010 Nov 19;285(47):36945-57. Epub 2010 Aug 25 PubMed.

    Lewis TL, Cao D, Lu H, Mans RA, Su YR, Jungbauer L, Linton MF, Fazio S, LaDu MJ, Li L. Overexpression of human apolipoprotein A-I preserves cognitive function and attenuates neuroinflammation and cerebral amyloid angiopathy in a mouse model of Alzheimer disease. J Biol Chem. 2010 Nov 19;285(47):36958-68. Epub 2010 Sep 16 PubMed.

    McLean JW, Tomlinson JE, Kuang WJ, Eaton DL, Chen EY, Fless GM, Scanu AM, Lawn RM. cDNA sequence of human apolipoprotein(a) is homologous to plasminogen. Nature. 1987 Nov 12-18;330(6144):132-7. PubMed.

    Segrest JP, Li L, Anantharamaiah GM, Harvey SC, Liadaki KN, Zannis V. Structure and function of apolipoprotein A-I and high-density lipoprotein. Curr Opin Lipidol. 2000 Apr;11(2):105-15. PubMed.

    View all comments by Ling Li

  11. A brief response to the interesting comments by Torik Ayoubi: An earlier paper by our group (Iturria-Medina et al., 2014) looked at a causal model of Aβ propagation through the white-matter pathways. The results suggest that accumulation of Aβ is more a consequence of reduced clearance than overproduction. Moreover, that clearance deficit showed a strong gene-dose dependence upon the number of APOE4 alleles.

    References:

    Iturria-Medina Y, Sotero RC, Toussaint PJ, Evans AC, Alzheimer's Disease Neuroimaging Initiative. Epidemic spreading model to characterize misfolded proteins propagation in aging and associated neurodegenerative disorders. PLoS Comput Biol. 2014 Nov;10(11):e1003956. Epub 2014 Nov 20 PubMed.

    View all comments by Alan Evans

Make a Comment

To make a comment you must login or register.

This paper appears in the following:

News

  1. LOAD of Data Place Vascular Malfunction as Earliest Event in Alzheimer’s