The Gene Conundrum in Alzheimer’s Disease

Alzheimer’s disease (AD) is the most common cause of dementia, currently affecting an estimated 5.8 million Americans. It has been over a century since AD was first described, but it is still not sufficiently well understood to enable development of drugs to treat it. As lifespan continues to rise and for myriad other reasons, the number of AD cases per state in the US is predicted to increase 12 to 43 percent over the next five years.

The lack of disease-modifying treatments may reflect a feature of AD pathology that was first noted in its initial description: the vast heterogeneity of the hallmark “senile plaques” that are found in all AD brains. When Alois Alzheimer and Oskar Fischer described the first cases of AD, they noted plaque accumulations of a protein called amyloid that builds up in between brain cells and interrupts cell-to-cell communication; amyloid plaques vary in size, shape, abundance, and location within the brain. “Among the plaques in the cerebral cortex many were of an extraordinary size, such as I have never seen,” Alois Alzheimer stated. “Some evidently arose from the fusion of smaller ones since they contained several central cores, but others had one exceptionally big central core and uncommonly large halo.”

Disease heterogeneity extends to behavior and includes varying age of onset, symptoms, and disease progression. Some variability may be explained by genetic heterogeneity, since more than 33 AD risk factor genes have been identified via a technique called “genome wide association studies” (GWAS), which broadly samples DNA from cells outside of the brain to identify mutations that are present in every cell of the body. None of these genes, however, are considered to cause AD.

However, mutations in three genes are causal for AD: APP, PSEN1, and PSEN2 (the genes for amyloid precursor protein and presenilin 1 & 2, respectively). Causal gene mutations account for rare familial forms that affect about one percent of people with AD and produce disease earlier, during mid-life. Notably, these genes were not identified by GWAS but by different, classical methods. All three genes support the dominant theory in AD known as the “amyloid cascade hypothesis.” This hypothesis emerged with the identification of amyloid-b (Aβ) that led to identification of the amyloid precursor protein (APP) from which Aβ arises (PSEN1 and 2 help to create Aβ from APP). Aβ is thought to be the major protein of the diverse senile plaques first described by Alois Alzheimer and Oskar Fischer.

In addition to mutations, gene dosage—the number of gene copies in our genome that produce the APP protein (normally 2; one from each parent)—plays an important role in the etiology of AD: rare families (report 1 and report 2) with three copies of APP, and Down syndrome patients with three copies of chromosome (21) on which APP resides, uniformly get AD. Moreover, rare cases of losing an APP copy in Down syndrome protect against AD. All of these genetic observations support essential roles for APP in the development of at least the rare familial and Down syndrome forms of the disease. Despite this genetic causality, major puzzles remain towards explaining the most common form of AD, “sporadic” AD, where the three Aβ-related genes are viewed as non-causal. Moreover, uniform failures of clinical trial designed to block Aβ and amyloid plaque formation to treat AD have led many to question the validity of the amyloid cascade hypothesis.

Can these puzzling disparities be reconciled? Recent findings on the fundamental molecular biology of the brain could explain the documented heterogeneity of amyloid plaques, the genetic importance of APP in familial AD and Down syndrome, and the high-profile clinical trial failures. The possible solution may be through unappreciated, enormous APP DNA sequence diversity produced by changes within single brain cells via a process known as somatic gene recombination (SGR). SGR contributes to the creation of “genomic mosaicism;” just like a tile mosaic, each tile can be different in size, color, and shape—yet all come together to form a cohesive image. Similarly, genomic mosaics consist of single cell “tiles” that vary in their DNA blueprint yet come together to form our brain. SGR can modify the blueprint “tiles,” thus changing the brain over time. In so doing, SGR “records” new DNA information in a stable way, which may represent a form of long-term cellular memory. SGR warrants a rethinking of how genes function in the complex organization of the brain under normal as well as pathological conditions.

Variability in the Very Blueprint of Our Brain Cells

In Gregor Mendel’s famous pea experiment, traits (like the color of pea flowers) were discovered to be heritable. We now know that such characteristics are encoded by genes located within the double strands of DNA: the blueprint of our genomes. Conventional genomic science generally assumes that all cells within an individual have identical and immutable genomes.

Your genome is composed of 46 chromosomes, 22 paired autosomes and two sex chromosomes, with one copy of each inherited from your mother and your father during fertilization. The fertilized egg, with a complete genomic blueprint, undergoes many cell divisions, giving rise to every cell in your body, all with the same genome…or so it was thought. However, genomic variations among cells in the immune system produced by SGR were discovered in the mid-1970s by Susumu Tonegawa, through a process of cutting and pasting DNA gene segments to produce immunological gene recombination (known as VDJ recombination). Immunological SGR “mixes and matches” gene segments to generate an astronomical repertoire of different antibody and T-cell receptor sequences that encode proteins protecting us from a universe of external and internal pathogens.

Could SGR occur in the brain? Scientists have speculated since the 1960s that the cellular diversity observed in the nervous system may arise from similar changes to the genome, but evidence for SGR in the brain eluded scientists for decades. Molecular hints of such a process emerged in 1991 when part of the machinery behind immunological SGR (the recombination activating gene 1 (RAG1) that is necessary for VDJ recombination) was identified in the brain. However, no corresponding genomic changes could be found, which in retrospect was due to technological limitations and unappreciated genomic mosaicism. Over the last 20 years, however, a vast range of single-cell DNA alterations have emerged to define genomic mosaicism, beginning with aneuploidies (the gain and/or loss of whole chromosomes) and now covering the gamut of DNA sequence alterations. These discoveries indicate that any study of SGR in the brain must ultimately interrogate single cells.

In 2015, my lab identified the first link between somatic genomic mosaicism and AD, showing increased amounts of total DNA in AD disease neurons. These DNA gains averaged more than 500 megabases—nearly twice the size of the largest human chromosome (Chr 1)—and were accompanied by copy number increases in APP, one of the three causal familial AD genes, with as many as 13 copies per cell. Somatic increases in APP offered a new explanation for common AD. However, the genomic structure of these copy number gains was quite unclear, including whether or not they were intact or partial copies.

Normally, the genomic structure of a gene contains alternating exons and introns: exons are short DNA sequences that contain the information used to encode a protein, while introns are long stretches of DNA sequence between exons, which are removed to produce messenger RNAs (mRNAs) that are the molecular intermediate required to produce (translate) protein. mRNAs contain information copied from gene exons without the introns. Interestingly, data from studying APP copies suggested that an inexact copy mechanism might exist to produce variant APP gene copies.

Several years later, we acquired proof for the production of inexact APP gene copies. We identified thousands of previously undetected sequences containing combinations of exons and partial exons that were more diverse and abundant in AD neuronal genomes compared to controls. They resembled mRNAs by lacking introns, yet were present in the genomic DNA blueprint. Such DNA sequences that reflect mRNAs are known as complementary DNAs or “cDNAs” since they “complement” a strand of mRNA. We therefore called these DNA blueprint copies “genomic complementary DNAs” or “gencDNAs.” gencDNAs were found in normal and AD brains. However, in AD neurons, the APP sequence heterogeneity increased and remarkably, included mutations that were already known to cause disease in the rare genetically inherited familial AD forms, yet were present in the common sporadic AD brains.

Could APP copy number gains and sequence changes occurring in single cells lead to the toxic accumulation of Aβ proteins and senile plaques seen in AD? SGR has the potential to impact all levels of protein structure—from amino acid sequence to 3-D conformations—that could have many functional effects on APP and Aβ proteins, including the production of prion-like molecules. As a first proof-of-concept my lab has demonstrated that APP gencDNA sequences can produce proteins that are toxic to cells. To identify possible points of intervention, we also modeled APP SGR in vitro, which required three conditions: APP gene transcription into mRNA, production of DNA strand breaks, and reverse transcription. Reverse transcription requires specialized protein enzymes called reverse transcriptases (RTs) that can generate cDNA from mRNA (discussed further below).

The generation of myriad APP gene forms could explain clinical trial failures whereby the tested experimental agents that are known to recognize only a single form of Ab simply miss the universe of other APP and Aβ-like variants. Notably, SGR of APP preserves components of the amyloid cascade hypothesis while also accounting for many of its experimental and therapeutic shortcommings. Critically, the requirement for reverse transcription implicates a new therapeutic approach to treating AD by shutting-down reverse transcription in the brain.

Reverse Transcriptase Inhibitors (RTis) as AD Therapeutics

RTs were first discovered in retroviruses by David Baltimore and independently by Howard Temin and Satoshi Mizutani in 1970. Unlike a traditional DNA or RNA polymerase, RTs synthesize DNA from RNA templates (retroviruses use RTs as famously known for human immunodeficiency virus (HIV) that is a retrovirus). Scientists have now identified other sources of RTs, including myriad RT-like genes embedded in our genomic blueprint, which may represent a reservoir of new therapeutic targets.

The involvement of RT activity in SGR suggests that its inhibition could be a preventative and/or therapeutic intervention for AD. Agents for the task are already available: reverse transcriptase inhibitors (RTis) have been approved by the Food and Drug Administration for treatment of HIV and as a class, have over three decades of continuous use with acceptable safety data. Thanks to effective antiretroviral therapy that includes RTis, many people with HIV have reached an age group at risk for AD, providing an opportunity to look for beneficial effects of RTis on sporadic AD.

Actually, this notion is contrary to an anticipated increase in AD incidence and co-morbidity for people with HIV, which was predicted over a decade ago. However, amongst the estimated 80,000-100,000 HIV patients aged 65 or older in the US with an expected AD prevalence of around 10 percent entailing thousands of expected AD cases, one AD/HIV case has been reported (by positive amyloid pet scan) in the peer-reviewed literature. This result is consistent with a link between inhibition of SGR and reduced risk of AD, and prospective clinical trials with RTis to prevent or treat AD in defined patient groups are needed.

In the meantime, physicians treating patients might consider at least for some patients the legal possibility of prescribing these relatively safe medications off-label in the absence of effective medicines which do not exist now nor will they in the patients’ lifetimes because of the many years required to test new agents and demonstrate that they are safe and effective. On a scientific level, ongoing studies of SGR and its role in AD (and other brain disorders as well) will open new vistas towards understanding the normal and diseased brain.