The connection is probably that ApoE4 carriers clear out amyloid beta deposits less efficiently than ApoE3/2 carriers (see e.g. [1] among many other sources), whereas microbial infection can induce amyloid deposits to form in the first place (see e.g. [2]), although it's not the only mechanism which can induce such deposits.
[2] Eimer et al (2018). Alzheimer’s Disease-Associated β-Amyloid Is Rapidly Seeded by Herpesviridae to Protect against Brain Infection. https://doi.org/10.2139/ssrn.3155923
That's a natural a priori hypothesis but there's extremely strong evidence that amyloid is causally upstream.
For example, mutations in the enzymes or proteins involved in the production of amyloid-β (APP, PSEN1, and PSEN2), which structural analysis show have the direct effect of producing either more total amyloid-β, or more of amyloid-β 42 (the specific peptide implicated in Alzheimer's), deterministically cause Alzheimer's. That puts amyloid-β causally upstream at least in those cases, the so-called autosomal dominant variety, which are ~1% of Alzheimer's cases.
The remaining ~99% of cases are less straightforward to interpret, but the disease looks essentially the same as the ~1% of cases described above: first amyloid-β builds up in the default mode network, then it spreads throughout the brain, eventually reaching the medial temporal region where primary age-related tauopathy (PART) is waiting, only then does the tau pathology spread, and neurodegeneration follows the tau pathology in lockstep. The way the amyloid-β and tau proteins misfold in autosomal dominant and sporadic Alzheimer's are the same, too, even though other tau pathologies have different tau folds. So basically, there's every reason to believe sporadic Alzheimer's is essentially the same disease as the one which we're basically certain is caused by amyloid-β.
The current most plausible perspective is amyloid-β pathology causes tau pathology which causes neurodegeneration. Thus it's not surprising that amyloid-β pathology can exist without neurodegeneration, if there is no tau pathology (or none yet).
However, I have not heard of the opposite: people with cognitive decline and amyloid-β pathology but no tau pathology.
Amyloid-targeting drugs have not failed. There are currently three — aducanumab (phase 2 trial successful, one of two phase 3 trials successful), donanemab (successful phase 2 trial), and lecanemab (promising phase 2 trial which technically failed but which showed a 64% Bayesian posterior probability of being at least 25% better than placebo) — that have shown a tendency to a roughly ~20-35% reduction in the pace of cognitive decline compared to placebo.
That said, amyloid is very upstream in the causal chain, which begins about 15-20 years before clinically detectable symptoms. It's believed by some (including myself) that amyloid therapy would be preventative at that early stage, but you won't do much better than a ~35% slowdown at the stage Alzheimer's is typically diagnosed. At that stage, tau therapy is more promising, since it's the more proximate cause of neurodegeneration.
[1] is probably the study you're thinking of. There's also [2] regarding herpesviridae, among others.
In short, there are good reasons at this point to believe that amyloid-β's primary function is as antimicrobial peptide, and thus various infections may cause the seeding of amyloid-β deposits. These deposits may then persist for years beyond any useful benefit, especially if the brain's clearance mechanisms are impaired.
Note that, per [2]:
Importantly, in the antimicrobial protection model, neurodegeneration is not mediated by pathogen activities that directly kill neurons. Rather, Aβ innate immune pathways targeting pathogens mediate the AD [Alzheimer's Disease] neuropathogenesis that leads to widespread neurodegeneration. Thus, our model is consistent with the amyloid cascade hypothesis and overwhelming data showing the primacy of Aβ in AD pathology.
[1]. Dominy et al (2019). Porphyromonas gingivalis in Alzheimer’s disease brains: Evidence for disease causation and treatment with small-molecule inhibitors. https://doi.org/10.1126/sciadv.aau3333
[2]. Eimer et al (2018). Alzheimer’s Disease-Associated β-Amyloid Is Rapidly Seeded by Herpesviridae to Protect against Brain Infection. https://doi.org/10.1016/j.neuron.2018.06.030
I've spent about the last two months reading the scientific literature on Alzheimer's, and while there is some dissent, the amyloid hypothesis is currently by far the dominant one, and for good reason IMO. The below is adapted from a comment I wrote on Aug 20 on the ACX Substack:
---
It is natural to ask: "sure, we see amyloid-β plaques with Alzheimer's Disease, but could there be a confounder, rather than amyloid-β being the cause?" However, we have strong evidence that it's the cause.
In fact, in a subset of cases, we have smoking gun evidence that amyloid is the cause: certain mutations or duplications of the APP (amyloid precursor protein), PS1 & PS2 (presenilin 1 & presenilin 2, parts of the enzyme γ-secretase involved in cleaving APP to make amyloid-β) genes guarantee that one gets Alzheimer's, and typically quite early (between the age of 30 and 60 for the onset of clinical symptoms). We have mapped out the structure and function of the corresponding proteins extremely well, and we know how, functionally, those specific mutations affect the behavior of those proteins: either to increase total amyloid-β production, or to increase production of the specific peptide (amyloid-β 42) implicated in Alzheimer's Disease. [1] I am not aware of another plausible effect of those mutations besides this one, and the mutations guarantee you get Alzheimer's Disease.
Now, this represents approximately 1% of Alzheimer's cases, the so-called autosomal dominant variety, so it's a priori conceivable that the other cases have a different cause. In the remaining cases, we have evidence consistent with an impairment of amyloid clearance mechanisms, however that evidence is more circumstantial and in some cases compatible with other hypotheses. But the disease looks like exactly the same disease as the 1% of cases in which we have smoking gun evidence of amyloid being the cause: we still see the same progression of amyloid-β, followed by a progression of hyperphosphorylated tau, followed by neurodegeneration and cognitive decline, and with the same sequence of brain regions and cognitive symptoms.
Suppose there are two bank robberies. In the first, we have smoking gun evidence of the culprit: a video camera showing a guy getting out of his car, with a clear image of his face and the license plate, and then of him walking into the bank, pointing a gun at the teller, the teller handing over a bag of cash, and him walking out with that bag of cash. In the second: we also have footage of the same guy and the same license plate at the scene of the crime, but an occlusion prevents clear footage of the exact moment of the robbery. However, the robbery occurred in the same town and on the same day, and eyewitness reports are that the robbery was conducted in basically the same manner. In that case, is there much question as to the identity of the robber?
**
As for the track record of amyloid-targeting therapy for Alzheimer's, I think it's fair to say that it's been less successful than hoped for but that also:
1) There has been a mixture of benefit and no effect; rather than the mixture of benefit, no effect, and harm which you'd expect to see if the drugs really were useless. For example:
A) The recent phase 2 trial of the similar drug donanemab showed success in its primary endpoint of reducing cognitive decline, by 32% (p = 0.04). [2]
B) The recent phase 2b trial of another similar drug lecanemab had promising results, although it didn't pass its primary endpoint, arguably due to a needlessly convoluted Bayesian statistical analysis. It showed reductions in cognitive decline of 26%, 30%, or 47% (p = 0.125, 0,034, or 0.017), depending on endpoint. [3]
C) With respect to aducanumab [4], the first phase 3 trial, EMERGE, passed its primary endpoint and all pre-designated secondary cognitive endpoints, with p-values for the high dose arm between 0.0006 and 0.0493 (0.0120 on the primary endpoint), and effect sizes ranging from an 18% slowdown to a 40% slowdown in cognitive decline (22% on the primary endpoint).
The second phase 3 trial, ENGAGE, did not show statistically significant benefits on any pre-designated endpoints, but the effect sizes still ranged from -3% to 18% (-2% on the primary endpoint). So given only pre-designated endpoints, we have two trials, one of which showed a clear benefit, and the other of which showed no effect, or very slight benefit if we give some weight to pre-designated secondary endpoints. From a Bayesian perspective, this has to be seen as weak evidence of benefit.
Lastly, both trials showed a substantial and significant (p < 0.001 in EMERGE, p < 0.01 in ENGAGE) reduction in phosphorylated tau, a neuropathology both regionally and chronologically highly correlated with volume loss and cognitive decline in Alzheimer's Disease.
2) The amyloid cascade hypothesis offers explanations for past failures of amyloid-targeting therapy, for example poor target specificity (some therapies such as semagacestat were not proven to engage amyloid in the first place, and were also known to produce other toxic effects unrelated to reduction in amyloid), deficient screening of participants (in some cases, patients were selected only for cognitive symptoms and not for the presence of amyloid, thereby reducing statistical power), and not intervening early enough (no one believes amyloid is the main proximate cause of neurodegeneration, so if you intervene late, it might be too late to prevent the cascade of problems it's believed to cause).
By contrast, I'm not aware of alternative hypotheses which even offer a story which can account for the known facts, such as the aforementioned evidence in the APP and PS1/PS2 genes.
**
For those interested in learning more, a good starting point for the current state of the science is [5]. A more up-to-date and comprehensive review paper is [6].
[3] Swanson et al (2021). A randomized, double-blind, phase 2b proof-of-concept clinical trial in early Alzheimer's disease with lecanemab, an anti-Aβ protofibril antibody. https://doi.org/10.1186/s13195-021-00813-8
[1] Castellano et al (2011). Human apoE Isoforms Differentially Regulate Brain Amyloid-β Peptide Clearance. https://doi.org/10.1126/scitranslmed.3002156
[2] Eimer et al (2018). Alzheimer’s Disease-Associated β-Amyloid Is Rapidly Seeded by Herpesviridae to Protect against Brain Infection. https://doi.org/10.2139/ssrn.3155923