|Home | About | Journals | Submit | Contact Us | Français|
The historical roots of Alzheimer’s disease provide a sound conceptual basis for linking the behavioral and neurological symptoms of the disease with the frequently associated pathology of amyloid plaques and neurofibrillary tangles. Out of these roots has grown the ‘amyloid cascade hypothesis’ – a vision of the etiology of Alzheimer’s that has spurred the discovery of many important insights into the neurobiology of the disease. Despite these successes, the wealth of new data now available to biomedical researchers urges a full review of the origins of Alzheimer’s and such a reconsideration is offered here. It begins with the most widely accepted risk factor for developing Alzheimer’s disease: age. Then, for an individual to progress from normal age-appropriate cognitive function to a condition where the full palette of clinical symptoms is expressed, three key steps are envisioned: 1) an initiating injury, 2) a chronic neuroinflammatory response and 3) a discontinuous cellular change-of-state involving most, if not all, of the cell types of the brain. The amyloid cascade is integrated into this sequence, but reconfigured as an amyloid deposition cycle. In this way, the pathology of amyloid plaques is envisioned as highly correlated with, but mechanistically distinct from, the three obligatory steps leading to Alzheimer’s disease. The implications of this new model are discussed with respect to our current diagnostic criteria and suggestions are put forward for expanding our future research efforts.
In 1906 Alois Alzheimer reported the first description of the dementing illness that now bears his name. He documented the progression of symptoms that beset a farmer’s wife, Auguste D., as her mental status deteriorated through a complex series of behavioral and cognitive changes that left her aggressive, delusional and unable to remember recent events. After her death, Alzheimer drew on his interest in the emerging techniques of histochemistry. He stained sections from the autopsied brain and discovered the presence of “miliar foci, which are caused by deposition of a peculiar substance in the cortex” (now recognized as neuritic or senile plaques). He also reported “very peculiar changes in the neurofibrils” (now recognized as paired helical filaments or tangles). By concerning himself with the structure of the diseased brain and the abnormal deposits that he found, he was among the early pioneers whose studies linked brain structure to function.
In considering the biology of Alzheimer’s disease over 100 years later, a few aspects of this case study deserve note. Auguste D. became ill in her early 50’s, meaning her symptoms emerged from a familial (i.e., genetic) form of Alzheimer’s rather than the sporadic form that makes up over 90% of prevalent cases (Yu et al., 2010). Second, part of what made the case noteworthy for its era was the inclusion of the neuropathological examination and the proposal that the abnormal behavior of the patient was the consequence of the abnormal deposits in her brain. As a result, from the very beginning, Alzheimer’s disease research and diagnosis has been based on a tight association between the dementia we now know as Alzheimer’s and the peculiar deposits we now recognize as plaques and tangles. That the presence or absence of these deposits is considered the gold standard of Alzheimer’s disease (AD) diagnosis only serves to underline the broad acceptance in the field of the importance of the association.
A major advance in the study of AD came with the sequencing of the main constituent of the senile plaque – the amyloid β peptide (Aβ) (Glenner and Wong, 1984). This led in rapid sequence to four key discoveries. First, the Aβ peptide is a part of a large type I membrane protein, the amyloid precursor protein (APP), which is encoded by the APP gene on chromosome 21. Second, the APP gene is mutated in a significant fraction of the cases of familial Alzheimer’s disease. Third, individuals with Down’s syndrome, who have three copies of chromosome 21 and hence three copies of the APP gene, develop clinical and pathological signs of early onset Alzheimer’s. And fourth, mutations in the presenilin-1 (PSEN1) and presenilin-2 (PSEN2) genes can behave as dominant familial AD genes. Presenilin is the catalytic subunit of the γ-secretase activity that liberates the Aβ peptide from the C-terminus of APP.
These findings led to the elaboration of a theory of AD known as the amyloid cascade hypothesis (Hardy and Selkoe, 2002; Citron, 2004). In familial AD (Figure 1 – top left), mutations in either APP or one of the PSEN genes lead to the brain accumulation of a 42-amino acid form of the amyloid peptide that has a high tendency to form β-pleated sheet structures. Amyloid aggregates form – first small oligomers and finally plaques. The amyloid cascade hypothesis proposes that these Aβ aggregates lead in turn to a series of downstream events ranging from synapse loss to plaque deposition to inflammation to the triggering of tau hyperphosphorylation to the death of susceptible neurons. The hypothesis also proposes that sporadic AD develops when the natural history of an individual accelerates a normal age-dependant process of Aβ accumulation (Figure 1 – top right). At some point, sufficient Aβ becomes deposited that the amyloid cascade is engaged. Subsequently, the sporadic disease follows the same pathway to dementia as the familial form.
In recent years, our knowledge base has broadened considerably, affording us the luxury of being able to revisit our hypotheses about causes of AD. Based on its prevalence, it now makes sense to begin this exercise with the more common sporadic form of the disease rather than with the rare familial forms. The precise etiology of sporadic AD is not known in detail. We do know, however, that the most critical risk factor by far is age. This makes intuitive, if not mechanistic sense. In all species, age brings a progressive slowing of brain function in virtually every domain. Our cognition slows; our ability to form new memories is reduced; our motor functions deteriorate; even our brain’s homeostatic functions become less and less robust. This functional decline is correlated with the loss of structural complexity of our brain cells. Neuronal dendrites become less complex, spine and synapse density decrease, the cell bodies of the larger neurons accumulate lipofuscin, astrocytic function declines (e.g., glutamate uptake and recycling), our immune systems become less responsive, etc. And as this list of insufficiencies grows, the brain’s defenses against many diseases, Alzheimer’s included, is weakened. The question is: how do the signature symptoms of AD emerge from this state. The new model envisions 3 key events that occur sequentially when an individual develops Alzheimer’s disease (Figure 2). The first is a precipitating injury that is begins the pathogenic process. This injury in turn triggers the second key event: a chronic inflammatory process that adds additional relentless stress to brain cells already weakened by age. The third event is a major shift in the cellular physiology of the brain cells. Described in more detail below, it represents a tipping point that marks the beginning of a cell’s degenerative process and leads to major synaptic dysfunction and neuronal death – the final and most direct cause of Alzheimer’s dementia.
Although aging gradually takes its toll on our brains, the hypothesis stipulates that some event – a physical head trauma, a major illness or infection, a vascular event (possibly so small as to be clinically undetectable), the metabolic stress associated with adult-onset diabetes or even the stress associated with a major life event such as a death in the family – is required to initiate the disease process. A genetic mutation can be such an injury, but only if it must interact with the aging process to be expressed. The injury triggers a protective response among the cells of the brain, but the age-related failure of the normal homeostatic mechanisms means that the response continues, even if after injury itself abates. The key concept is that it is the nature of the response, not the nature of the injury, that determines the outcome of Alzheimer’s disease.
A useful analogy to consider is hip fracture. For a wide variety of reasons, the risk of breaking our hipbone increases dramatically with age. Bone density decreases; osteoporosis becomes more likely; balance is less sure; reaction times slow; muscles weaken; visual acuity fades; etc. Each of these is a risk factor, but the factors themselves do not cause the hipbone to break; there has to be a precipitating injury (usually a fall). Applied to AD, the analogy is meant to suggest that while any of the changes in the brain that come with advancing age may increase our risk of Alzheimer’s, without an injury none can cause dementia.
The idea that AD begins with an initiating injury has both theoretical and practical relevance. Theoretically, it means that Alzheimer’s is not a part of normal aging any more than breaking your hip is a part of normal aging. They are both pathological events with an underlying biology. The practical relevance is that, if research can identify the most common sources of injury, we may be able to intervene proactively and delay disease onset. Currently, it is not possible to identify a single candidate for this precipitating injury. But the frequent co-occurrence of vascular pathology with AD, and the protective effects of genetic and environmental factors that improve cardiovascular health suggest that a common if not exclusive initiating injury would be a vascular event such as a head trauma or microstroke.
A cardinal feature of the neuropathology of most AD brains is the evidence for a chronic neuroinflammatory process. Many recent reviews have summarized this topic in some detail (McGeer et al., 1996; Akiyama et al., 2000; Bamberger and Landreth, 2002; Wyss-Coray and Mucke, 2002; Wyss-Coray, 2006; Glass et al., 2010). While it is true that an inflammatory response accompanies tissue damage in many brain diseases including Parkinson’s (McGeer et al., 2001; Nagatsu and Sawada, 2005), ALS (Henkel et al., 2009; Ilieva et al., 2009), and others, AD is unique in the intimate association found between chronic inflammation and disease. There is also solid epidemiological evidence that inflammation serves as a cause of AD. Long-term use of high doses of certain non-steroidal anti-inflammatory drugs (NSAIDs) lowers the lifetime risk of AD from 30–60% (McGeer et al., 1996; Stewart et al., 1997; Vlad et al., 2008), and there is biochemical evidence for elevated levels of cytokines such as Il-1, Il-6, TNFα and S100β in human AD brain (Griffin et al., 1998; Akiyama et al., 2000). Microgliosis as well as astrocytosis are prevalent, and most plaques are surrounded by activated astrocytes and invaded by activated microglia (McGeer et al., 1987; Heneka and O'Banion, 2007; El Khoury and Luster, 2008; Rodriguez et al., 2009).
Discussions of brain inflammation tend to focus on the microglial cell; however, a variety of cell types participate in the AD inflammatory response. These cell:cell interactions have been reviewed for other diseases (Ilieva et al., 2009) and evidence for their importance appears in the AD literature as well (Griffin et al., 1998; Mucke et al., 2000; Wegiel et al., 2001; Giri et al., 2002; Griffin, 2006). Through a variety of feed-forward loops, the microglial cells are assisted by the responses of astrocytes and brain vascular endothelial cells (Giri et al., 2002) in maintaining a chronic shift in the inflammation status of the brain. The result of this network-like response is a chronic stress on neurons and their function. Cell cycle proteins are activated (Wu, Q. et al., 2000), reactive oxygen species are produced (Keller et al., 1999; Fuller et al., 2010), mitochondrial function is reduced (Chen and Chan, 2009), dendritic/axonal transport is impaired (Wu, H. Y. et al., 2010), etc. A second tenet of the new model, therefore, is that a chronic immune response, persisting over months and years, creates the unique chemistry and cellular physiology that results in the core symptoms we recognize as dementia of the Alzheimer’s type.
Thus far, the re-envisioning of AD has included two important tenets. The first is that Alzheimer’s must be triggered by an injury. The second is that the establishment of unique type of chronic inflammation is required to set the brain’s chemistry on the path to AD. To fully capture the neurobiology of the disease, however, a third tenet must be introduced: the progression to full Alzheimer’s disease requires a functional discontinuity between the physiology of the brain cells in early and late stages of the disease. This dramatic change-of-state results in a ‘new normal’ biology, primed towards neurodegeneration and dementia. The exact meaning of this transition in biological terms is only beginning to be understood but some of its consequences are already becoming apparent. It is envisioned as a one-way cellular door; once a cell crosses the threshold, it can never return to its earlier state.
The best example of this change-of-state concept can be found in the paradoxical association of neuronal cell cycle events with the process of neurodegeneration. Neurons are generally considered to be permanently post-mitotic cells. But when they are stressed, fully differentiated neurons can and do re-initiate the enzymatic cascades of a normal cell cycle. Curiously, the cycle stalls after DNA replication such that the neurons can neither proceed into late G2/M-phase and complete division, nor can they reverse and turn off the cell cycle proteins. This new neuronal ‘state’ is of more than academic interest as it is dramatically elevated in populations of neurons that are at risk for death in several neurodegenerative diseases, the best studied of which is AD. The at-risk neurons re-express cell cycle proteins (Arendt et al., 1996; Vincent, I et al., 1996; McShea et al., 1997; Vincent, I. et al., 1997; Busser et al., 1998; Nagy et al., 1998), accompanied by true DNA replication (Yang et al., 2001; Kingsbury et al., 2005; Yang et al., 2006; Mosch et al., 2007). These unscheduled neuronal cell cycles appear during early disease stages (Yang et al., 2003; Arendt et al., 2010), which argues that they are an integral part of the disease process. The full mechanistic implications of the appearance of cell cycle events (CCEs) in a ‘post-mitotic’ adult neuron are not yet understood, but certainly the doubling of the DNA content of any cell would seem to qualify as a profound and irreversible change-of-state.
Findings in the mouse models of AD enhance this view. In one carefully studied AD model, both the anatomical and temporal appearance of the CCEs follow the course of the neuropathology seen in human AD (Yang et al., 2006; Varvel et al., 2008). Further, the CCEs in this model do not appear in a slow progressive manner. Rather, in any one population of neurons, the percentage of ‘cycling’ neurons rises rapidly from near zero to final values and then remains stable – a rapid change-of-state. A similar progression is likely to occur in the human AD brain (Arendt et al., 2010). The results of anti-inflammatory treatments of AD mouse models are also consistent with a change-of-state model. While 3 months of NSAIDs treatment can block new CCEs from appearing, even 6 months of NSAIDs treatment do not reverse the cell cycle protein expression pattern once it has begun (Varvel et al., 2009).
Neurons are not the only cells of the brain whose cell biology is changed during the progression of AD. Astrocytes become activated in regions of AD neuropathology, as do the microglial cells. Equally intriguing from the change-of-state perspective, the microglia can apparently adopt a phenotype found in macrophages known an “alternate activation state” (Colton et al., 2006; Cameron and Landreth, 2010). This state is accompanied by a shift from an acute pro-inflammatory reaction to a chronic state of activation more suited towards vascular growth and tissue repair (Allavena et al., 2008). Though not yet fully documented for microglia in AD, evidence from spinal cord injury studies suggest that this alternate activation can be achieved in CNS (Kigerl et al., 2009)
Hypothesizing that AD involves a cellular change-of-state from early to late disease has theoretical as well as practical importance. At the theoretical level, it encourages us to revisit the cellular events that occur after the change. The prediction is that the biology of early AD differs in qualitative ways from the biology that ultimately produces the dementia. As an analogy, if someone were to stop smoking after they developed lung cancer, they would not be likely to alter the progression of the cancer. The biology of the cells involved has changed and the process is now independent of the initiating injury and transformation. It also offers a theoretical explanation for the failure of the prospective human trials of NSAIDs: the trials were all begun after AD symptoms were manifest. With the disease already in progress, it is likely that many of the neurons in the subjects’ brains had gone through their change of state. Their biology no longer required chronic inflammation to sustain their abnormal state. Coupled with the evidence cited above that the change of state may be quite abrupt in entire cohorts of neurons (Varvel et al., 2009; Arendt et al., 2010), the human NSAIDs trail data are consistent with the new model. At the practical level the model predicts that there is a post-amyloid, post-inflammatory biology that offers important new areas for neuroprotective drug discovery.
A significant case has been made that the loss of synaptic structure and function is an integral part of the advancement of AD (Selkoe, 2002, 2008; Arendt, 2009; Palop and Mucke, 2010). The failure of these uniquely neuronal structures has been seen at the level of the single cell and also in the neural networks of the region. There is a loss of dendritic mass a decrease in spine density in post-mortem AD brain (Uylings and de Brabander, 2002). There is also a strong correlation between the sites of synaptic loss and the regions of the most dramatic neurodegenerative changes (Terry et al., 1991). The use of biochemical and immunocytochemical markers of both pre- and post-synaptic structures has validated this observed decrease (Hamos et al., 1989; Lippa et al., 1992). In vitro, electrophysiological studies in slice preparations have shown that when Aβ levels are elevated, they interfere with both LTP and LTD (Shankar et al., 2008; Li et al., 2009) and the connection has recently been made between the altered chemistry of the AD brain and the triggering of electrical activity consistent with seizure (Palop et al., 2007; Palop and Mucke, 2009).
Hyperphosphorylated forms of tau – modifications that weaken its affinity for microtubules – are found as the main protein constituent of the neurofibrillary tangle (NFT). This is the second of the two archetypal lesions described by Alzheimer in 1906, and the one that has proven the more reliable partner of neurodegenerative change (Gomez-Isla et al., 1997; Mitchell et al., 2002; Bennett et al., 2004). Indeed the tight correlation between the anatomical location of the NFTs and sites of greatest neuronal cell loss in AD argues for a central role of tau phosphorylation in the disease. In support of this concept, transgenic mice carrying human tau mutations develop a late-onset neurodegenerative phenotype that includes neuronal loss (Andorfer et al., 2005; Ramsden et al., 2005). The role of tau in AD is likely a complex one, however, as mutations in the tau gene itself (MAPT) have not been identified in familial forms of AD. Rather, MAPT mutations have been identified as leading to a separate late-onset form of dementia known as FTDP-17 (Hutton et al., 1998; Poorkaj et al., 1998). The simplest way of incorporating these findings into the new model is to place the role of tau in the final stages of the disease process, following the change-of-state. This envisions the phosphorylation of tau as a mechanistic part of the cell death program. Given the complexity of the disease, however, and the early appearance of NFTs in the progress of AD, it is highly likely that tau plays one or more additional roles in setting up the change-of-state in the affected neurons.
Autophagy involves the coordination of a number of vesicle populations culminating in a process that assists the cell in degrading long-lived proteins and spent organelles. It is stimulated rapidly in response to various stresses, and results in the formation of an autophagosome, which then fuses with a lysosome and leads to the degradation of the contents. This process is of particular interest in AD research because if it is overstimulated, it can lead to cell death (Nixon, 2007). As might be expected, neurons at risk for death in AD show marked defects in their autophagic function (Cataldo et al., 1996). The presenilin mutations may also play a role here. Recent evidence has shown that one of the normal functions of presenilin is to facilitate the acidification of the cell’s lysosomes, a requirement for efficient autophagy (Lee et al., 2010). As autophagy is one of the main defense mechanisms for clearing failed organelles and large protein aggregates from a cell, any compromise at this stage would only hasten the loss of cellular integrity and make the cell death process more rapid and more sure.
It is gratifying that the new model comfortably incorporates the amyloid cascade as a contributing cause of AD pathogenesis. Extracellular Aβ naturally accumulates with age and with time the non-neuronal cells of the brain would be expected to sense its presence and react (Matsuoka et al., 2001). The associated cytokines then enhance the production of the Aβ peptide. This creates a feed forward reaction described previously in the writings of Griffin (Griffin et al., 1998; Griffin, 2006). Other factors such as excessive synaptic activity of the type found during epilepsy or excitotoxic injury can also enhance Aβ production. Thus, driven by one or more of these means, Aβ aggregates stimulate the immune response, and the immune response stimulates more Aβ production. In this way a cycle is created – the amyloid deposition cycle (Figure 3). While this description requires no change in the well-described chemistry of APP metabolism, recasting the amyloid cascade as an amyloid deposition cycle presents a new view of the linkage between this chemistry and the biology of AD. In this view, the deposition of amyloid is tightly linked to, but mechanistically distinct from, the forces that drive the development of dementia. Thus, the model predicts that while Aβ deposition and chronic inflammation each renders the other more likely, neither one has an absolute requirement for the other to be present in the brain.
Note that the proposed close linkage between chronic inflammation and the Aβ deposition cycle offers a pathway by which familial AD genes can lead to disease. By accelerating the deposition of Aβ, these mutations enhance inflammation and thus drive the amyloid deposition cycle earlier and harder than normal, strongly favoring the development of a chronic inflammation. The ApoE protein is likely to play an important role in this process at this stage. ApoE is closely involved in mediating the clearance of Aβ from brain (Shibata et al., 2000; Deane et al., 2008). Thus any change that reduced ApoE-dependent clearance of Aβ, such as the E4 variant of the gene, would enhance deposition, encouraging the establishment of a chronic inflammation and driving the amyloid deposition cycle. ApoE genotype is also recognized for its impact on the cardiovascular system and since the first step of the new model is postulated to involve a vascular injury, ApoE genotype might independently act to alter the probability of the occurrence of an initiating injury.
This recasting of the role of Aβ in AD has both theoretical and practical significance. At a theoretical level, the prediction that the amyloid deposition cycle can run independently offers a fresh way of looking at the 30% of elderly individuals who are cognitively normal but are found to have significant plaque deposits in their brains. Rather than characterizing them as pre-clinical Alzheimer’s, the new hypothesis suggests that it would be more accurate to say that they are cognitively normal, but their high plaque burden places them at enhanced risk for developing AD in the future. At a practical level, the model makes another important statement: the presence of plaques in the brain, while highly correlated with AD, should not be an essential part of an Alzheimer’s diagnosis. This is a major departure from current thinking and thus deserves careful consideration, as it will impact both our pre- and post-mortem diagnostic criteria in substantial ways. One attractive feature of adopting this idea is that it removes a troubling piece of circular logic from our current models. Presently, we are comfortable identifying individuals as having plaques-without-Alzheimer’s but we have stipulated that it is impossible to have Alzheimer’s-without-plaques. There is no inherent biological reason for this; we have simply defined away this category. If an individual presents with a classical behavioral and neurological course of AD symptoms, but their brain does not contain the requisite plaque burden and tangle density at autopsy, the dementia that they had was not Alzheimer’s disease – by definition. Note that the new model predicts that this total discordance between plaques and AD will be rare in practice since the amyloid deposition cycle is stimulated by the same immune response that drives the neurodegeneration. This encourages us to continue to use Aβ and plaque based outcome measures in our diagnoses. But just as there can be Parkinson’s-without-Lewy-bodies (Gomez and Ferrer, 2010), it may be wise for us to consider the possibility that there can be Alzheimer’s-without-plaques.
Viewed in its entirety (Figure 4), the new model is a complex web of interactions. This complexity makes viewing the diagram difficult, but it probably represents an accurate reflection of the biology underlying Alzheimer’s disease, arguably one of the most complex diseases of the human nervous system. A key strength of the new framework is that it begins not with pathology, but with age, a well correlated risk factor. In doing so it encourages us to draw on the rapid advances that have been made in the biology of aging and incorporate these insights into the design of new therapeutic approaches to Alzheimer’s. The three tenets of the new model (Figure 2; Figure 4) can each be seen as enlarging the number of targets for future intervention. The requirement for an initiating injury may not offer specific pharmaceutical targets, but it suggests that adjunct therapies and treatments may well be used to improve the action of more targeted drugs.
The requirement for chronic inflammation opens a wide array of well-studied targets for potential intervention. Many labs are already exploring this pathway but there is much to learn. A critical unanswered question, indeed a key challenge for future research, is how does a process as non-specific and prevalent as brain inflammation reproducibly create the unique set of symptoms we recognize as Alzheimer’s disease. While part of the answer is likely to lie in the persistent nature of the response, the bulk of the answer is most probably found in an Alzheimer-specific quality to the ‘marinade’ of cytokines and chemokines that are produced by the aging astrocytes, microglia, endothelial cells and others. A useful context in which to view this challenge can be found in the observation that experimental allergic encephalomyelitis (EAE) and multiple sclerosis (MS) are both demyelinating conditions caused by brain inflammation. Yet, EAE is now recognized as an imperfect model of MS with treatments that function as cures in the model proving ineffective or worse in the human disease (Sriram and Steiner, 2005; Baxter, 2007). This emphasizes the diversity of possible neuroinflammatory responses in different conditions and urges us to learn their details in the context of the AD brain.
The proposal for a cellular change-of-state is a call to explore the cell biological changes that move the degenerative process after the change-of-state has occurred. The model proposes that cell cycle events, defects in autophagy and alterations of tau are part of this process, but because they exert their actions after the change-of-state, they must now be re-examined with the assumption that the processes with which they are involved can be independent of both amyloid and inflammation.
Finally, beyond Alzheimer’s, the new model contains the seeds of a re-examination of other diseases. The proposition of an initiating injury that is required to begin the process of Alzheimer’s disease comes with the strong implication that different injuries to the same age-weakened brain will lead to different responses of the cells of the brain and in so doing begin different disease processes (Figure 4, light blue arrows). The most clear-cut example of this would be Parkinson’s disease. In this case, the ability of paraquat and MPTP to mimic the symptoms of the disease hints that Parkinson’s may represent the brain’s reaction to oxidative damage and mitochondrial malfunctions rather than inflammatory changes. What makes this last notion particularly appealing is the prediction that if different late-onset neurodegenerative diseases can evolve from a common origin but take different pathways to neuronal loss, then the late life dementias should often appear as mixed dementias. This follows because it is unlikely that one type of injury precludes a second. The occurrence of these mixed dementias vexed researchers performing the early studies of Alzheimer’s disease, but can be viewed in the context of the new model as a predictable consequence of how the diseases begin.
The author wishes to thank his friends and colleagues, including Jianmin Chen, Jiali Li Jie Zhang and Haley Fitzpatrick for their helpful comments during the preparation of this manuscript. Special thanks go to the contributions of Gary Landreth for insightful and critical comments; to Maria Carrillo for helpful discussions and suggestions; and to Gina Kolata, whose beautifully written series of articles in the NY Times triggered the urge to present this model in a formal way. Financial support from Rutgers University, the Alzheimer’s Association and the NIH (AG029449; NS20591) is also gratefully acknowledged.