College of American Pathologists
Printable Version

Digging deep for the roots of dementia

July 2003
Cover Story

William Check, PhD

Not all facial tissues are Kleenex. Not all photocopiers are Xeroxes.
Not all adhesive bandages are Band-Aids. And not all dementias are Alzheimer’s disease. Like the brand names that are virtually synonymous with the products they represent, Alzheimer’s has become the form of dementia that has greatest awareness in the public mind.

“We are definitely in the era where Alzheimer’s dementia rules the day,” says Howard Crystal, MD, professor of neurology and pathology at the State University of New York Downstate, Brooklyn.

To some extent, this is understandable, since Alzheimer’s disease, or AD, comprises about 70 percent of dementias. “Sometimes people start thinking about Alzheimer’s and dementia as the same thing,” Dr. Crystal says. “But they are not.” Vascular disease contributes to an important fraction of dementia cases as well, though that fraction is difficult to estimate.

Distinguishing among causes of dementia assumes greater urgency as basic research begins to illuminate the biochemical pathology underlying common, late-onset AD and as epidemiological and longitudinal cohort studies attempt to define more precisely how vascular disease contributes to dementia. Two critical practical benefits are expected from this work: early diagnosis of dementias and more effective treatment.

“Genetic advances in Alzheimer’s are really very interesting from the point of view of delineating the pathogenesis of disease,” says Barbara Crain, MD, PhD, associate professor of pathology at Johns Hopkins Medical Institutions. Autosomal dominant mutations in three genes—presenilin-1 (PS1), presenilin-2 (PS2), and amyloid precursor protein (APP)—that cause early-onset AD have been known for a number of years. Evidence is accumulating that the processes these mutations affect also play an essential role in the predominant sporadic cases.

“These three genes are intimately associated with the central AD pathology— deposition of beta-amyloid [Abeta] in the brain,” Dr. Crain says. “If we could figure out what their gene products were doing, and then figure out how to block them, we could perhaps prevent the formation of amyloid. To me, that is where we are heading in terms of treatment.”
One prominent researcher in this area is Dennis Selkoe, MD, the Coates professor of neurologic diseases at Harvard Medical School and Brigham and Women’s Hospital. “PS1 and PS2 genes are implicated in all forms of Alzheimer’s disease,” Dr. Selkoe says. “That is because while they are mutated in only a subset of early-onset cases of Alzheimer’s disease, their gene products, the proteins presenilin-1 and presenilin-2, are the active site of an enzyme called gamma-secretase, one of two key enzymes that make Abeta throughout life. We believe that without Abeta protein buildup, people wouldn’t get Alzheimer’s disease.”

Important discoveries about beta-secretase, the other enzyme that processes APP into Abeta, have been made by Philip Wong, PhD, associate professor of pathology and neuroscience at Johns Hopkins Medical Institutions, with professor of neurology Donald Price, MD, and colleagues. They have proved the enzyme called BACE-1 to be the beta-secretase, which also qualifies as a major therapeutic target. “Knowing these enzymes,” Dr. Wong says, “we could come up with drugs that would inhibit their activity, inhibit production of Abeta protein, and decrease
plaque deposition.”

While AD is certainly “the gorilla in the room,” as Dr. Crystal puts it,
in the past few years more research has focused on cerebrovascular disease as a cause of dementia, says David Desmond, PhD, associate professor of neurology and pathology at SUNY Downstate.

“Stroke in the elderly increases the risk of dementia by fourfold relative to elders who don’t have stroke,” Dr. Desmond says. “And in patients with stroke and dementia, cognitive decline is more rapid even than in Alzheimer’s disease.”

Research is also ongoing on the combination of cerebrovascular disease and AD. “The Nun Study helped us to tease out the relative importance of cerebrovascular disease and Alzheimer’s disease,” Dr. Desmond says.

A multicenter program to study vascular dementia, funded by the National Institute of Aging, is headed by Helena Chui, MD, professor of neurology at the University of Southern California, Los Angeles. “Our goal is to improve diagnosis,” Dr. Chui says, “and to understand how small-vessel disease, and brain injury resulting from small-vessel disease, can lead to cognitive impairment by itself or, as commonly occurs, in combination with Alzheimer’s.” The idea that vascular pathology alone can cause dementia is “an assumption that is still open to testing,” she says. “My bias is that small-vessel disease can cause slowly progressive dementia by itself. But that is still for us to prove.”

Ultimately, to make the greatest therapeutic impact on AD, early diagnosis will be necessary. “We think there will be more effective medications in the next decade or two,” says Marilyn Albert, PhD, director of the Division of Cognitive Neuroscience in the Department of Neurology at JHMI. Dr. Albert studies the transition from mild cognitive impairment to dementia. “We want to intervene with these medications as early as possible. And likely they will not be benign,” she says. “So we want to select the correct patients. For that we need early and accurate diagnosis.” As experience with cholinesterase inhibitors has shown, by the time AD becomes established, it is difficult to affect the disease course substantially. “Right now there is no single simple test that allows you to diagnose Alzheimer’s disease at any point, much less early,” Dr. Albert says.

Diagnostic tests for analytes in cerebrospinal fluid were proposed several years ago. However, Dr. Crain says, “Right now CSF testing tells normals from demented patients as a group and, in some studies, tells Alzheimer’s from non-Alzheimer’s dementia. But there is a lot of overlap.”

Elucidation of the activity of the protein products of the PS1 and PS2 genes has shown that each functions as part of a four-protein complex that executes the gamma-secretase function, which is to make the second cut that releases the 42-amino acid Abeta peptide from its 770-amino acid precursor, APP. (The other three proteins in the complex are called nicastrin, aph-1, and pen-2.) Gamma-secretase is a novel aspartyl protease that cuts the intramembrane domain of the transmembrane protein APP within the lipid layer—“a strange place for a protease to cut a peptide bond,” Dr. Selkoe says. Mutations in PS genes change the relative activity of gamma-secretase on its two cleavage sites. Wild-type gamma-secretase cuts the 42-43 peptide bond 10 percent of the time and the 40-41 bond 90 percent of the time. Mutated PS proteins cut the 42-43 bond 20 percent or more of the time, so the amount of 42-amino acid Abeta—which forms the first amyloid plaques—is doubled.

Even normal PS proteins produce Abeta throughout life, albeit at a low level. “So presenilins are also players in the lifetime risk of sporadic Alzheimer’s disease,” Dr. Selkoe says. How amyloid formation is increased when PS genes are normal is not known, but it is clear that a relatively small increase in the amount of Abeta can dramatically increase the risk of AD.

“I think it is likely for a few percent of common late-onset Alzheimer’s disease that environmental factors will be found that change the rate of the events [around APP],” Dr. Selkoe says. “But I think a more common basis of late-onset disease will be genetic mutations that alter the rate of these functions, especially clearance of Abeta. My guess is that late-onset Alzheimer’s disease is mostly a clearance failure.”

Dr. Selkoe studies one enzyme that degrades Abeta, called insulin degrading enzyme, or IDE. “Our colleagues Rudy Tanzi and Lars Bertram at MGH [Massachusetts General Hospital] have found evidence of genetic linkage and even allelic association between SNPs in IDE and late-onset Alzheimer’s disease,” Dr. Selkoe says. “They have indirect evidence implicating IDE, but they have not yet found any single base change that tracks unequivocally with Alzheimer’s disease.”

Dr. Wong’s work on beta-secretase, the enzyme that makes the first cut in the formation of Abeta, confirmed that this activity resides in a protein called BACE-1. “To prove that BACE-1 is really the key beta-secretase, we had to genetically ablate it,” Dr. Wong says. So he and his colleagues made a mouse in which the gene encoding BACE-1 was absent. Nerve cells from the brain of this knockout mouse did not make any Abeta peptide.

Dr. Wong and his colleagues next worked with transgenic mice carrying mutated PS or APP genes. These strains “recapitulate aspects of AD pathology, including amyloid deposition,” Dr. Wong says. When the BACE-1 gene was ablated in this model of AD, the mice no longer showed amyloid deposition in their brains. “So if you could find compounds that inhibit BACE-1, you would have a good chance of preventing buildup of amyloid plaque in animal models, and we think in humans as well,” Dr. Wong says. Pharmaceutical companies are now screening for compounds that inhibit this enzyme.

In principle, compounds that inhibit presenilin might ameliorate plaque formation. However, the gamma-secretase complex also processes proteins that are essential in development, so inhibiting this activity might have serious side effects. A particularly important developmental protein processed by gamma-secretase, called Notch, is involved in cell fate determination, especially in the embryo. “Notch must be cut to regulate transcription of certain genes in the nucleus,” Dr. Selkoe says. Even affecting Notch later in life could affect such functions as hematopoesis. Could partial inhibition of gamma-secretase—by 20 to 30 percent—decrease Abeta deposition to a desirable degree while allowing adequate Notch function? That’s an open question.

In this regard, it is interesting that anti-inflammatory agents such as ibuprofen—which, in epidemiological studies, are associated with a decreased risk of AD—have recently been found by Edward Koo, MD, and Todd Golde, MD, PhD, to act as gamma-secretase inhibitors. “Ibuprofen can somehow affect presenilin and its ability to form the 42-43 peptide bond,” Dr. Selkoe says.

Because nicastrin is also essential for functioning of the gamma-secretase complex, one might try to reduce Abeta production by inhibiting nicastrin. Theoretically, this maneuver would also inhibit Notch, with deleterious effects on development. In fact, a nicastrin knockout mouse looks like the Notch knockout phenotype, Dr. Wong says. However, nicastrin remains a potential therapeutic target for AD: Mice heterozygous for nicastrin develop normally. “Perhaps a compound that reduces nicastrin activity by 50 percent would have a significant effect on production of Abeta without a substantial impact on the Notch signaling pathway,” Dr. Wong says.
Another strategy for reducing the amount of Abeta in plaques was the anti-Alzheimer’s vaccine that was taken out of clinical trials recently because it caused encephalitis. “Everyone was quite disappointed when those early trials had to be stopped,” Dr. Crain says. “It looked like the vaccine worked against Alzheimer’s changes. Why it caused encephalitis in humans but not in animals isn’t clear.” Dr. Crain believes that the vaccine’s impact provides hope that this approach, in an altered form, may eventually work.

Dr. Wong and his colleagues are moving toward generating a mouse model in which researchers could simulate a drug trial to define a therapeutic time window for effective treatment. This model would have inducible expression systems in which a gene of interest, such as BACE-1, could be turned on or off. First, researchers would allow Abeta to form in the brains of mice. “Then we would selectively turn off the gene,” Dr. Wong says, “which would be analogous to giving a drug to these animals, and watch to see if Abeta plaque goes away.” What is the time window in which “therapy” will be effective? When does plaque deposition become irreversible? “This information can guide human trials,” Dr. Wong says. Studies such as this are in their infancy and will take several years to mature.

Mutations in APP genes cluster in and around the Abeta peptide region. Mutations in the N terminus region of Abeta promote increased cutting by beta-secretase, while mutations at the C terminus increase cutting by gamma-secretase. Thus, both types of mutations generate more Abeta peptide. A cluster of mutations in the middle of the Abeta region is thought to increase the amyloidogenic property, or aggregability, of Abeta peptide. So in effect, all APP mutations can promote Abeta plaque formation.

Alpha-secretase is actually an anti-amyloidogenic enzyme: It cuts APP between the cleavage sites of beta- and gamma-secretase, rendering Abeta peptide formation impossible.

No kindreds have yet been found with mutations in either beta- or alpha-secretase. Nor have humans been found with mutant forms of the three other protein components of the gamma-secretase complex. The reasons for this are not clear. On the flip side, Dr. Selkoe says, “there are still quite a few genetic traits to be discovered. There are any number of folks with a familial tendency to Alzheimer’s disease in which none of the three proteins known to bear mutations in AD is responsible.”

In addition to mutations in PS1, PS2, and APP, the likelihood of having AD is affected by one other known genetic trait: carrying the E4 allele of the apolipoprotein E gene. Dr. Wong calls apoE “one of the key genes involved in garden-variety AD.” People with the E4 allele have an increased level of Abeta plaque in their brains. “ApoE is a tendency or risk factor for AD, but is not determining like the autosomal dominant mutations,” Dr. Crain says. ApoE makes a protein that carries lipid around the body. The apoE protein comes in three isoforms. Inheritance of the E4 allele increases the risk of getting AD relative to the E2 and E3 alleles—two- to fivefold for the heterozygous state and three- to 10-fold for the homozygous state. “ApoE is an an important risk factor for Alzheimer’s because it is relatively common,” Dr. Selkoe says. Roughly 17 percent of Americans carry one apoE4 allele, and about two percent are homozygous. Preliminary evidence suggests that apoE also affects clearance of Abeta, perhaps binding and internalizing it into cells, with the E4 allele having reduced clearance ability. “Another theory is that apoE4 may somehow stabilize Abeta in extracellular deposits, so you get more mature and less immature or diffuse plaques,” Dr. Selkoe says.

Alzheimer’s disease did not always dominate thinking about
dementia. In fact, Dr. Crystal says, its current prominence reflects a major turnaround. “Until the 1960s, most dementia cases in older persons were thought to be due to vascular causes,” he says. For about 60 years after Alois Alzheimer described his eponymous disease in 1906, it was appreciated as a cause of dementia but thought to be rare and to principally appear in people under age 65. (Alzheimer’s three cases were 56, 56, and 61 years old, much younger than our current notion of the typical AD patient.) Clinico-pathological studies among patients in their 60s, 70s, and 80s in England in the 1960s, in which mental status tests were correlated with brain autopsies, revealed the surprising fact that half of people with dementia had only Alzheimer’s pathology postmortem. Most of the other half had mixed Alzheimer’s as well as vascular pathology. Overnight the number of people in the United States thought to have AD went from a few thousand to perhaps 4 million. “Now you could say this condition was a public health problem,” Dr. Crystal says.

Now vascular dementia is the poor cousin. Several studies from the mid-’70s to mid-’90s sought to determine where vascular disease fits into the dementia picture. In one, only six of 2,000 cases qualified as pure vascular dementia on autopsy. However, any degree of Alzheimer’s pathology was taken as the cause of the patient’s dementia. “There are all different degrees of Alzheimer’s pathology,” Dr. Crystal notes. And Alzheimer’s lesions evolve over 20 to 30 years. “If you have 10 years of lesions, is that enough to make you demented?” he asks. “Perhaps not. But in some studies that would be enough to say the dementia was not caused by vascular lesions.” Clean cases of pure vascular disease are rare. “How much does vascular disease contribute to cognitive impairment in cases of mixed pathology?” Dr. Crystal says. “No one knows.”

Determining how many people with dementia have Alzheimer’s pathology only or mixed Alzheimer’s-vascular dementia is “a very contentious issue,” Dr. Desmond says. “Experts in cerebrovascular disease believe that cerebrovascular disease is a major contributor to dementia,” he adds. “Experts in the world of Alzheimer’s believe cerebrovascular disease to be of minor importance.” Before anyone can say how common vascular dementia is versus AD and combined dementia, uniform criteria for cerebrovascular disease as a cause of dementia are needed.
Dr. Desmond conducted a large epidemiologic study to evaluate the contribution of CVD to dementia. He recruited more than 500 patients who were at least 60 years old and admitted with ischemic stroke, evaluated them at three months, then performed annual tests of cognitive abilities and functional independence. At three months after stroke, 26.3 percent of patients met the criteria for dementia. Among patients who were not experiencing dementia at three months, new dementia was diagnosed in 72 (21.6 percent) over five to 10 years of followup. “In this study, ischemic stroke was a very potent risk factor for delayed dementia,” Dr. Desmond says. “I think we can generalize that finding to people who experience stroke at an older age.”

Brain autopsies were not done in this study. However, Dr. Desmond speculates that some patients had Alzheimer’s changes—a reasonable inference given the age of the cohort—and that some late dementia was due to mixed neuropathology.

He cites data from the Nun Study as evidence. A cohort of Catholic nuns was recruited and followed with serial measures of cognitive function. Autopsies were performed on those who died. Among nuns who met the neuropathological criteria for AD, those who had suffered strokes that damaged important parts of the brain, such as the thalamus, had dramatically worse performance on cognitive tests than nuns who simply had Alzheimer’s pathology at autopsy. “Perhaps silent strokes or infarcts increase the risk or severity of dementia,” Dr. Desmond suggests.

How these two conditions might interact is a focus of study. Dr. Desmond raises hypotheses. Perhaps a person has a predisposition to AD and then experiences a stroke, which may initiate the onset of Alzheimer’s. “If stroke did not occur, would that person’s decline to Alzheimer’s disease have begun?” he asks. “Can stroke unmask Alzheimer’s disease?” Another possibility is that a person with AD is declining slowly and stroke accelerates the course.

“Certainly vascular disease can impair cognitive functioning,” Dr. Crain says. And autopsies frequently show small vascular lesions in people with dementia. “If a patient has Alzheimer’s lesions that are not enough to cause dementia, and on top of that vascular lesions develop that also would not be enough to cause dementia, can you get additive effects from that mixed pathology?” she asks. “Proving that would be extremely difficult.”

Intriguing observations related to the importance of mixed pathology have been made in people with risk factors for vascular disease. Dr. Crystal cites studies in which tens of thousands of people with hypertension took neuropsychological tests. “You find that hypertensives do a little worse than nonhypertensives over time,” he says. “The difference is very subtle. You wouldn’t notice it in individuals, but you do see it in group studies.” On imaging, people with hypertension or diabetes have more cortical atrophy, more white matter disease, and less white matter volume. Again, differences are small. In Dr. Desmond’s study, diabetes increased the risk of dementia. “This has also been found in other cohorts of stroke patients,” he says.

Along this same line, epidemiological studies have, after controlling for many AD risk factors, showed that people taking statins are less likely to have Alzheimer’s disease. Dr. Crystal calls the evidence that statins reduce the risk of Alzheimer’s “very compelling.” The Alzheimer’s Disease Cooperative Study is conducting prospective studies of statins among people with AD.

In summary, Dr. Crystal says, it appears from epidemiological studies that patients with vascular diseases, including coronary artery disease, do suffer more dementia. “My first response,” he says, “is that they have vascular dementia, not Alzheimer’s dementia. But the data are compelling that they do have true Alzheimer’s disease.” It is not clear why this is so. However, Dr. Crystal emphasizes, “that makes it more important to be able to untangle whether a given case of dementia is Alzheimer’s or vascular.”

“To take care of patients, we have to know what we are treating,” Dr. Desmond adds. Dementia is not a disease; it is an outcome of a variety of diseases. “In an individual, we need to know whether to treat Alzheimer’s disease or cerebrovascular disease or Lewy body disease, or perhaps to treat more than one etiology,” he says. Treating risk factors for CVD can dramatically reduce the incidence of stroke and possibly of dementia.

To improve ways of estimating the contribution of vascular disease
to dementia, Dr. Chui, along with Harry Vinters, MD, professor of neuropathology at UCLA, and others, has enrolled a cohort of people with cerebral small-vessel disease, AD, or mixed pathology and a group of normal age-matched controls. Cognitive testing is repeated annually and MRI scans biennially. Imaging hyperintensity on T2-weighted sequences in deep white matter and deep gray matter is used as a surrogate for vascular disease. Imaging changes in the hippocampal region serve as an Alzheimer’s surrogate. All 570 subjects will be followed to death, and brain autopsy will be performed to detect AD pathology and estimate the extent of vascular injury.
A number of valuable observations have already emerged. The initial measure of small-vessel disease—a single score—“doesn’t correlate very well” with cognitive dysfunction, Dr. Chui says. Alzheimer’s pathology does explain cognitive impairment. On imaging, cognitive decline is fairly well predicted by hippocampal plus cortical atrophy, not so much hyperintensity but overall severity of gray matter pathology. “We have also found that cortical gray matter atrophy is not specific for Alzheimer’s,” Dr. Chui says. Some cases of pure vascular disease show cortical atrophy of the whole surface area. “This is surprising,” Dr. Chui says. “It was previously thought that general atrophy is related to Alzheimer’s and that vascular disease only causes focal atrophy. But that is not what we have found so far.”

Dr. Chui says that one distinctive finding in the 45 people who have gone to autopsy so far is that hippocampal atrophy, an important correlate of cognitive impairment, cannot be completely explained by AD. A subset—about 10 percent—have hippocampal sclerosis (no plaques or tangles), both focally and, in some cases, extending over most of the hippocampus. “That also thickens the plot,” Dr. Chui says. What causes hippocampal sclerosis is controversial. “Perhaps small-vessel injury,” Dr. Chui speculates.

She acknowledges that some experts say vascular pathology causes dementia only in rare Binswanger-type cases (deep white matter pathology). “In our 45 cases, what I have appreciated is that probably the truth is going to be less extreme,” Dr. Chui says. “A lot of those cases have some early Alzheimer’s and some vascular disease, and the two are working additively.”

To improve early diagnosis, Dr. Albert and other researchers are following people who have mild cognitive impairment, or MCI, but who do not yet have dementia. “We have had some success in predicting outcomes,” she says. “But I think what is increasingly clear is that each research group is looking at a slightly different cohort based on referral patterns and selection method.” Some groups are looking at people 80 and older, while others take only people who have been referred because they complain of memory loss or a primary care physician notices cognitive difficulty. So different research groups are studying people with differing degrees of severity who are further from or closer to the endpoint—conversion to dementia. “You want people with a higher probability of converting,” Dr. Albert says, “especially if you are testing medications.” However, taking people too close to conversion
means you won’t get useful predictors for those in the early stages of cognitive decline.

One problem is that there is no fixed definition of MCI. For an ongoing drug study, the Alzheimer’s Disease Cooperative Study defined MCI fairly narrowly: people who reported memory problems themselves plus a collateral source to confirm progressive difficulty with memory. “They are getting a conversion rate of 15 percent per year,” Dr. Albert says, “but those people have an awful lot of trouble when they start out.” If you don’t define MCI so strictly, you get large numbers of people who don’t progress in a short time and large numbers with normal aging who remain stable. “It makes sense to have subjects with a range of difficulty,” Dr. Albert says, “to represent a range of pathological change in the brain.” Dr. Albert and her former colleagues in Boston formed such a cohort, and the conversion rate has been relatively low. “We started 10 years ago,” she says. “At that time I don’t think anyone understood that MCI has as broad a range as it has turned out to have.” However, this has turned out to be an advantage, Dr. Albert believes. “We are now in a good position, because we have persons who have mild MCI and people who are impaired, and we can study the whole spectrum of severity.”

Memory tests have proved to be good predictors at the cohort level of who will develop AD. “The most powerful memory measures are those that are somewhat difficult in our hands,” Dr. Albert says. “We have found that measures of new learning that require verbal information are particularly sensitive. However, these measures turn out to be exceedingly difficult from the perspective of being accurate on an individual basis.” As a result, memory tests alone are inadequate. “Even with tests of executive function, we still have a long way to go,” Dr. Albert says.

Imaging tests are useful at increasing accuracy, presumably because they measure not only mental function but also brain lesions related to that function. Anatomic imaging provides good agreement with conversion. “If you measure the volume of regions that are part of the medial temporal lobe memory system,” Dr. Albert says, “you improve your accuracy in predicting who will develop Alzheimer’s.” Some apparently crucial areas are the entorhinal cortex, the hippocampus, and other regions such as the superior temporal sulcus. Plaques can be imaged in persons with AD, but they can’t yet be imaged in those with MCI to predict development of disease. With functional imaging, glucose metabolism has been seen to be decreased in those with the apoE4 allele. “It is unclear how that relates to diagnosis,” Dr. Albert says.

The National Institute of Aging is trying to organize a large study with several centers that would follow normal people, those with MCI, and AD patients with imaging measures, as well as biological measures, to look for a relationship between these parameters and development or progression of disease. “Drug companies are very interested in using imaging or other yet to be identified biomarkers to shorten the time over which they would have to follow people to know if a drug is effective,” Dr. Albert says.

Dr. Desmond agrees: “Identifying Alzheimer’s early remains a hot issue.” He thinks that expanding MCI to nonmemory impairment would be useful. “People with cerebrovascular disease often present with impairment in executive functions—initiation of behavior, decisionmaking, inhibition of behavior—that are less concrete than memory or language,” he says. “Criteria that look for mild cognitive impairment in any domain would be most worthwhile.”

While many issues surrounding dementia remain unsettled, our knowledge has advanced substantially over the past several years. Under the probing of contemporary molecular genetic techniques, the neuropathology of AD, which initially looked like an impenetrable tangle, has yielded several crucial secrets. Therapeutic trials with specifically targeted drugs are underway. A greater understanding of the relevance of vascular dementia has emerged, and the question of the significance of mixed pathology dementia is being directly addressed. Longitudinal studies are seeking the keys to detecting which people with MCI will progress to dementia. Combined, these initiatives may lead to optimal application of potent anti-dementia therapy, so that late-life cognitive decline will become a manageable condition.

William Check is a medical writer in Wilmette, Ill.