Can we refute a role for infections in Alzheimer's disease pathogenesis?

Abstract

While a growing body of literature suggests a role for infections in Alzheimer's disease (AD), microbial contributions to AD remains a contentious topic, in part due to challenges in reconciling the positive evidence with studies reporting null findings. Here, we examine the evidence that argues against a role for infections in AD, while offering mechanistic hypotheses that may account for both the negative and positive findings, including dysregulated host immunity and gene–environment interactions of AD‐associated genes.

Highlights

• A growing body of evidence has implicated infections in AD.

• Studies reporting null associations are often cited to refute an association between infection and AD.

• Here, we discuss these null findings and examine why they are insufficient to refute a role for infections in AD.

• We propose mechanistic hypotheses that reconcile both the null and positive findings, including dysregulated host immunity and gene–environment interactions of AD‐associated genes.

1. PERSPECTIVE

A growing body of evidence, from epidemiologic studies to cell culture and animal models of the disease, ¹ , ² has implicated infections in Alzheimer's disease (AD). Associated pathogens are diverse, including viruses, bacteria, fungi, and parasites. ¹ , ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ Despite accumulating literature suggesting an association between infections and AD risk, the scientific community has called on studies reporting null associations to dismiss the findings supporting this dogma‐challenging hypothesis. However, many of these null associations suffer from one or more pitfalls, including (1) controls can be infected and not affected and (2) suboptimal methods for measuring microbial associations. Here, we examine why these null reports are insufficient for arguing against the role of infections in AD. Instead, we posit how these findings can enhance our understanding of the circumstances under which infection may contribute to AD risk.

It is frequently hypothesized that if infections are associated with AD, then AD cases should have a higher frequency of infections or greater microbial loads compared to controls. Several studies have investigated this idea by leveraging electronic health records (EHRs), antibody serologies, or post mortem brain tissue. Some have found no specificity for pathogens in AD cases compared to controls, including HHV6A and HHV6B in brain tissue, or infection frequency of HSV‐1, varicella zoster virus (VZV), cytomegalovirus (CMV), and Epstein–Barr virus (EBV), as reported in EHRs or measured by serum antibodies. ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ While these studies are used to support an argument against an association between infections and AD, this argument is flawed. Not all individuals who have been exposed to infectious agents will necessarily have increased risk for disease, as infection itself does not confer disease or produce clinical manifestations in everyone. For example, most individuals have been exposed to herpes simplex viruses (HSVs) and often present asymptomatically (i.e., due to the unique latent life cycle of these viral species), and only some develop clinical symptoms associated with this ubiquitous pathogen, ranging from cold sores to encephalitis. Thus, as controls can be infected but without associated symptoms, similar frequencies of infection or comparable microbial loads in AD cases and controls does not necessarily negate a role for infection in AD risk. Furthermore, if exposure to a risk factor is homogeneous across the population (e.g., HSV‐1 exposure), cohort methods will likely fail to detect it but rather detect susceptibility factors, potentially leading to null findings. ¹⁴ Rather, other critical questions arise, including the risk factors that may be driving clinical manifestations in some individuals but not others and the circumstances under which these risk factors may be more likely to exert their effects.

Genetic predispositions and gene–environment interactions may play a crucial role in determining who may be susceptible to infection as a risk factor for dementia, and by which pathogens. In particular, AD genome‐wide association studies (GWASs) have identified several risk variants that also regulate the host response to pathogens. ¹⁵ For example, apolipoprotein E (APOE) genotype mediates the association between HSV‐1 and AD. ⁷ , ¹⁶ , ¹⁷ As most of the population is infected with HSV‐1 but not every HSV‐1‐infected individual develops AD, it is clear that HSV‐1 infection alone does not confer strong risk for AD. However, in combination, HSV‐1 presence in brain tissue and carriage of an APOE ε4 allele confer strong risk for developing AD. ⁷

Several other risk variants may similarly interact with particular pathogens to confer risk for AD, associations that might not be detected when investigating populations in a genotype‐agnostic manner. For example, PILRA, identified as a genetic risk variant for AD through GWAS, has also been identified as a receptor for glycoprotein B of HSV‐1. ¹⁸ , ¹⁹ The AD‐protective variant of PLIRA results in reduced ligand binding, which has been shown to limit viral entry into host cells in macrophages. ¹⁸ In this way, individuals carrying the AD‐protective variant of PLIRA may have reduced HSV‐1 entry into microglia and, thus, be less susceptible to HSV‐1‐mediated risk for AD. HLA DR15, associated with AD risk, has been shown to lead to a higher viral load and attenuated immune control of EBV in a model of multiple sclerosis. ²⁰ Another risk variant associated with AD through GWASs, OAS1, is an interferon response gene and part of the innate immune antiviral response. ²¹ OAS1 has been shown to increase susceptibility to critical illness with viruses including COVID‐19 and may confer AD risk through gene–environment interactions such as viral infections. ²² Similarly, the protective variant of MS4A4A, a genetic risk factor for AD also prioritized through GWASs, has been shown to regulate an interferon‐response population of microglia while the risk variant suppresses this interferon state. ²³ As such, the MS4A4A risk variant may increase susceptibility to poorly controlled central nervous system (CNS) viral infections. Together, these examples showcase molecular avenues grounded in genetics that likely mediate viral contributions to AD.

Thus, rather than refuting an association between infections and AD, the null associations may instead suggest that genetic predispositions that dysregulate host immunity may interact with specific pathogens to unveil AD susceptibility. These gene–environment effects also account for why a diverse range of pathogens has been associated with AD, as genetic risk factors for AD may impair the response to specific pathogens or canonical pathogen–response pathways.

Another essential issue that arises in this connection has to do with the circumstances under which infection interacts with risk factors, including genetics, to increase disease risk. For example, co‐infections, reactivation of viral infections, or local CNS infections, rather than peripheral infections, may be relevant circumstances for increasing dementia risk. The methodologies of many cohort studies and cross‐sectional studies are not designed to account for such factors, potentially leading to erroneous null findings.

For example, one relevant factor for AD risk may be the presence of local CNS infection, rather than peripheral pathogen exposure. In the case of HSV‐1, detection in brain tissue specifically appears to be critical in conferring AD risk particularly in individuals carrying an APOE ε4 allele. ⁷ Some studies have investigated the association between infection and AD by measuring whether serum antibodies against pathogens differ in individuals with AD compared to controls, several of them producing null results. ⁹ , ¹¹ These studies, however, do not negate an association between infection and AD risk but raise the question of how peripheral infection versus local CNS pathogen exposure modulates risk for developing AD. As serum antibodies do not always clearly distinguish CNS exposure, these studies may not provide a relevant measure to the relationship between some infections and AD risk. Moreover, direct detection of viruses in brain tissue is also technically challenging and requires sensitive methods and specialized protocols following autopsy, posing an added challenge. ²⁴

Similarly, co‐infections and viral reactivation may play an important role in determining subsequent risk for AD and are rarely accounted for. Infections, such as severe COVID‐19 infections, have been found to reactivate latent infections, including HSV‐1, VZV, CMV, EBV, and, in some cases, co‐reactivation of more than one virus. ²⁵ , ²⁶ In the case of HSV‐1, recurrent reactivations are thought to contribute to cumulative damage, increasing the risk for AD. ¹⁶ , ²⁷ , ²⁸ While some pathogens might reactivate latent viral infections, the reactivating pathogen itself might not be associated with increased risk for AD but merely reactivate an associated pathogen (such as HSV‐1). Many studies have not been designed to investigate the interaction of pathogens or assess reactivation of other viral infections, which would not necessarily be apparent in comparing pathogen exposure in cases versus controls. Additionally, most studies leveraging EHRs only account for severe infections, resulting in hospitalization, rather than more common, subliminal infections that more people encounter. As most reactivated herpesviruses are mild or asymptomatic, these methods again present suboptimal study design for detecting associations between infection and AD.

Animal studies investigating the role of infection in AD are also hampered by similar experimental design limitations, including those investigating the antimicrobial properties of amyloid beta. This antimicrobial activity may link amyloid pathology with the infectious hypothesis of AD, as reactivation of CNS viral infections may incite amyloid deposition, and amyloid plaques may act to contain virus. ²⁹ , ³⁰ , ³¹ While several studies support this antimicrobial activity, one study in 5xFAD mice found that following HSV‐1 infection, HSV‐1 neither was present within amyloid plaques nor induced their formation. ³² However, this study has two important caveats. First, it used a substantially lower viral dose compared to other studies demonstrating a protective effect and, second, it examined viral presence within plaques with immunohistochemistry for glycoproteins that are thought to be hidden by amyloid binding, rather than a fluorescently tagged virus as did studies reporting positive associations. ³¹ These null findings are insufficient to refute antimicrobial activity of amyloid in the setting of these key methodologic differences.

Collectively, many of the negative findings cited to refute the association between infection and AD are insufficient to do so. Rather, they highlight that understanding this association requires nuanced assessment of interaction of pathogens with other risk factors across the human lifespan, including genetics and polymicrobial infections.

Importantly, a lack of evidence for a unifying hypothesis should not be misconstrued as evidence against an association between infection and AD. While there is currently no mechanistic understanding of the diverse pathogen associations with AD, one factor that likely connects the large body of evidence thus far is dysregulated host immunity and response to infection, including genetic susceptibility resulting in an impaired host response. The questions of who might be susceptible to increased risk for AD associated with infection and what circumstances might increase risk point to the complex gene–environment interactions that might underlie the association of infection with AD.

We have focused on studies reporting null associations between infection and AD to critically examine the arguments used to refute a role for infection in this disease, ultimately to address the controversy that persists on this topic. However, it is important to recognize the large body of rigorous evidence supporting the hypothesis that dysregulated host immunity and response to infection is an important risk factor for AD. Indeed, several studies have identified associations between infection and AD pathogenesis and pathology. ³ , ⁷ , ²⁴ , ³¹ , ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² Other risk factors contributing to impaired immunity, including the aging immune system, environmental exposures including air pollution, and genetic background have also been identified as risk factors for AD. ¹⁵ , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷ One particularly strong body of evidence suggesting a dysregulated immune response itself contributes to AD risk includes studies demonstrating protective effects of vaccines against developing AD, including the shingles vaccine. ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ Another compelling piece of evidence that a productive immune system is protective against AD is that immune stimulation with granulocyte‐macrophage colony‐stimulating factor reduces serum markers of neuronal death in humans, reduces amyloid pathology, and provides neuroprotection in animal models of AD. ⁵² , ⁵³ , ⁵⁴ Overall, the evidence strongly suggests that a productive host immune pathogen response is protective against AD.

While challenging to do at a population level, future studies that investigate unique combinations of risk factors (genetic and environmental), as well as how these exposures interact over time, may ultimately clarify the link between infection and AD. As such, we invite the scientific community to rigorously test the infectious hypothesis of AD, addressing the factors and limitations raised here, to collectively improve our understanding of the etiological factors of AD and identify relevant therapeutic approaches.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the Supporting Information.

Supporting information

Supporting information