Cancer-Alzheimer inverse comorbidity from epidemiology to protein self-assembly.

Cancer–Alzheimer inverse comorbidity from epidemiology to protein self-assembly

Cancer and Alzheimer’s disease (AD) are common age-associated disorders characterized by contrasting cellular outcomes, yet epidemiological studies repeatedly report an apparent, modest bidirectional inverse association between cancer and dementia, including AD. This Mini Review examines the extent to which this pattern reflects biological antagonism versus competing mortality, diagnostic bias, survivor selection, and treatment exposure. Breast and prostate cancers are considered as endocrine models because estrogen- and androgen-receptor signaling supports tumor growth while also regulating neuronal metabolism, mitochondrial function, vascular homeostasis, and inflammation. Population studies show heterogeneous neurological associations for breast cancer hormone-modulating therapy and prostate cancer androgen-deprivation therapy, emphasizing the need to distinguish tumor biology from systemic endocrine treatment. Mechanistically, cancer and AD engage overlapping metabolic pathways, including insulin/IGF–PI3K–AKT, mTOR, and AMPK; cell-fate programs involving cell-cycle control and senescence; and proteostatic systems involving molecular chaperones and autophagy–lysosomal function. These networks produce divergent outcomes: proliferative and proteotoxic adaptation in tumors versus energetic failure, impaired proteostasis, and neuronal degeneration in AD. Protein self-assembly provides an additional, context-dependent connection: amyloidogenic proteins can modify cancer-cell fitness, while recent preclinical evidence identifies a tumor-derived cystatin C–TREM2 pathway that enhances cerebral Aβ clearance. Whether these mechanisms contribute to human cancer–AD inverse comorbidity remains unproven. The observed cancer–AD inverse comorbidity therefore likely reflects interacting epidemiological, endocrine, metabolic, proteostatic, and protein-assembly mechanisms rather than a single causal pathway. Treatment-aware longitudinal studies incorporating AD biomarkers and physiologically relevant models are needed to distinguish genuine biological antagonism from observational bias.

1Introduction

Cancer and Alzheimer’s disease (AD) are prevalent age-associated disorders but represent markedly different cellular trajectories. Cancer cells maintain proliferation through metabolic adaptation, resistance to apoptosis, and enhanced tolerance of genomic and proteotoxic stress. AD, by contrast, is characterized by synaptic dysfunction, impaired energy metabolism, failure of protein quality control, and progressive loss of post-mitotic neurons. Epidemiological studies have repeatedly reported an apparent bidirectional inverse association: individuals with cancer have a modestly lower subsequent incidence of all-cause dementia and, in some studies, AD, whereas people with dementia are less frequently diagnosed with cancer (1, 2). This pattern is neither universal nor necessarily causal, however, because its magnitude varies across cancer types, dementia subtypes, populations, and analytical designs and is influenced by competing mortality, diagnostic surveillance, survivor selection, and the long preclinical phases of both diseases. Hormone-dependent cancers provide particularly informative models for separating intrinsic tumor biology from treatment-related effects. Estrogen- and androgen-receptor signaling promotes the growth of breast and prostate tumors, respectively, while also regulating neuronal metabolism, mitochondrial function, synaptic activity, vascular homeostasis, and inflammation. Endocrine therapies impose sustained systemic hormone perturbations that may affect cognitive and neurodegenerative outcomes independently of the tumor itself. The heterogeneous associations reported for breast cancer hormone-modulating therapy and prostate cancer androgen-deprivation therapy therefore demonstrate why cancer diagnosis, disease severity, and treatment exposure must be evaluated separately (3, 4).

Recent reviews have primarily synthesized epidemiological and Mendelian-randomization evidence or quantified associations between cancer treatments and dementia risk (2, 5). Here, we instead propose a treatment-aware, multilevel framework that separates three distinct questions: the population-level inverse association, the systemic consequences of cancer and its treatment, and the molecular mechanisms that may connect cancer and AD. Our central conclusion is that cancer–AD inverse comorbidity should not be viewed as a single antagonistic biological relationship. Rather, its direction and interpretation depend on the level of analysis: competing mortality and ascertainment influence population estimates; tumor biology and endocrine therapy can produce different systemic effects; and shared metabolic, proteostatic, and protein-assembly pathways generate strongly context-dependent outcomes. Two observations constrain interpretation further. First, cancer associations are not uniform across neurodegenerative disorders or proteinopathies, arguing against a generic cancer–neurodegeneration antagonism. Second, protein self-assembly is not uniformly protective or deleterious: amyloidogenic proteins can modify cancer-cell fitness in experimental models, whereas tumor-derived cystatin C has been shown to enhance TREM2-dependent Aβ clearance in AD models. The latter currently provides the most direct experimentally demonstrated tumor-to-brain mechanism linking cancer with AD pathology. The strength of evidence nonetheless differs across these connections: the inverse associations rest primarily on human epidemiological and neuropathological evidence, whereas most molecular links derive from experimental models, and whether any materially contributes to population-level inverse comorbidity remains unproven. Thus, the key question is not whether one mechanism explains the cancer–AD paradox, but which mechanisms operate in particular cancer, treatment, tissue, and disease contexts. Accordingly, this Mini Review proceeds across four levels of analysis: Section 2 evaluates the population-level epidemiological evidence and its interpretive limitations; Section 3 separates intrinsic tumor biology from treatment-associated endocrine effects; Section 4 examines shared metabolic, cell-fate, and proteostatic pathways; and Section 5 considers protein self-assembly and emerging cancer-to-brain mechanisms.

2Epidemiological evidence and interpretive limits

Population-based evidence generally supports a modest inverse association between cancer and dementia, although effect estimates differ across all-cause dementia, AD, and vascular dementia. In a cohort of 3,021,508 adults aged ≥60 years, a previous cancer diagnosis was associated with a 25% lower subsequent dementia risk (HR 0.75, 95% CI 0.74–0.76) after adjustment for demographic, lifestyle, and cardiometabolic factors. The association persisted after accounting for competing mortality (SHR 0.56, 95% CI 0.55–0.56), while Mendelian-randomization analysis yielded a much smaller inverse association between genetically predicted cancer susceptibility and AD, consistent with—but not proving—a biological contribution beyond selective survival (1). A bidirectional meta-analysis similarly found that cancer survivors had 8–14% lower risks of all-cause dementia, AD (RR 0.88, 95% CI 0.78–0.99), and vascular dementia (RR 0.86, 95% CI 0.75–0.99), whereas individuals with dementia had an approximately 25% lower subsequent cancer risk. These associations varied substantially by cancer site, dementia subtype, treatment exposure, and study quality (5). Nevertheless, the observed association is highly sensitive to disease ascertainment and analytical design. In the UK Biobank, neither prevalent nor incident cancer was significantly associated with AD. All-cause dementia diagnoses increased sharply during the first year after an incident cancer diagnosis but subsequently attenuated to the null, suggesting that intensified healthcare contact, treatment-related cognitive symptoms, or diagnostic investigation can create short-term detection bias (6). In the Rotterdam Study, estimated dementia risk ratios ranged from 0.71 to 1.10 depending on how cancer exposure, confounding, selective survival, and death were handled (7). Competing mortality, reduced cancer screening after dementia, delayed dementia diagnosis after cancer, and the long preclinical phases of both diseases therefore complicate causal interpretation.

Evidence beyond diagnostic records nevertheless suggests that the inverse pattern may contain a biological component. In a community-based autopsy cohort, a history of cancer was associated with lower odds of advanced Braak neurofibrillary tangle stages, neuritic and diffuse plaques, and cerebral amyloid angiopathy, but not with TDP-43, α-synuclein, or cerebrovascular pathology (8). Recent integration of local genetic correlations with transcriptome-wide association analysis identified negatively correlated genomic regions and genes with opposing associations across AD and multiple cancers, converging on cell-cycle regulation, apoptosis, DNA-damage responses, and metabolism (9). Overall, the inverse pattern is modest, cancer-specific, and sensitive to survival, ascertainment, treatment exposure, and the definition of AD outcomes. Inverse cancer associations have also been reported in other neurodegenerative disorders, but they are not uniform. For example, a Korean cohort found a lower overall cancer incidence in patients with Parkinson’s disease (PD), whereas a meta-analysis identified an increased association between PD and melanoma (10, 11). These contrasting patterns argue against a universal cancer–neurodegeneration antagonism. Neuropathological evidence is likewise heterogeneous: although the community autopsy study discussed above found lower AD-type pathology but no association with TDP-43 or α-synuclein (8), a recent NACC analysis reported lower TDP-43 pathology and weaker evidence for lower Lewy body pathology among individuals with cancer (12). Thus, the specificity of the inverse association across neurodegenerative disorders and proteinopathies remains unresolved.

3Hormone-dependent cancers and endocrine therapy

Breast and prostate cancers are emphasized here not as representatives of all malignancies, but as tractable hormone-dependent models in which estrogen- and androgen-receptor signaling is integral to tumor biology and endocrine therapies impose sustained systemic hormonal perturbations. They therefore allow tumor-associated and treatment-associated effects to be considered separately. Broader metabolic and proteostatic mechanisms discussed below may extend across multiple cancer types, whereas sex-steroid-specific mechanisms and endocrine-treatment effects should not be generalized to all malignancies.

3.1Breast cancer: estrogen signaling and endocrine therapy

Breast cancer illustrates how estrogen-dependent tumor biology and estrogen-directed treatment can produce distinct neurological consequences. Estrogen receptor signaling drives the growth and survival of hormone receptor–positive breast tumors, whereas estrogen supports glucose metabolism, mitochondrial function, synaptic plasticity, and inflammatory regulation in the brain. Tumor biology, estrogen deprivation, and treatment-related cognitive symptoms should therefore be distinguished. Population-based studies provide inconsistent evidence that breast cancer survivorship is associated with a lower risk of AD. In a Korean nationwide cohort of 70,701 breast cancer survivors and 180,265 cancer-free controls, survivors had an 8% lower adjusted risk of AD (SHR 0.92, 95% CI 0.86–0.98). However, the association disappeared in landmark analyses restricted to women surviving more than five years, arguing against a durable inverse association. Within the survivor cohort, neither tamoxifen nor aromatase-inhibitor treatment was independently associated with AD, whereas radiotherapy was associated with lower risk (13). In contrast, a Swedish register study of 26,741 five-year survivors and 249,540 cancer-free women found no overall association with all-cause dementia, AD, or vascular dementia. Women diagnosed after age 65 nevertheless had higher risks of all three outcomes, including a 35% higher risk of AD (SHR 1.35, 95% CI 1.05–1.75) (14). Together, these findings indicate that any survivorship association is modest and sensitive to age, follow-up duration, competing mortality, and survivor selection.

Studies focused on endocrine treatment are similarly heterogeneous. In a US claims-based cohort of 57,843 women with breast cancer, exposure to hormone-modulating therapy was associated with an 18% lower incidence of AD (RR 0.82, 95% CI 0.75–0.90) and a 12% lower incidence of dementia (RR 0.88, 95% CI 0.83–0.93), with the strongest inverse associations observed for tamoxifen and the steroidal aromatase inhibitor exemestane (15). A subsequent SEER–Medicare analysis of 18,808 women aged 65 years or older found a smaller 7% reduction in AD and related dementias (HR 0.93, 95% CI 0.88–0.98) over approximately 12 years. The association varied substantially by age, race, and treatment class: it was strongest among Black women aged 65–74 years and was absent among White women aged 75 years or older (3). Conversely, a UK cohort of 14,214 postmenopausal survivors found no difference in dementia risk between aromatase-inhibitor and tamoxifen users (16). Importantly, cancer-related cognitive impairment should not be equated with AD. Breast cancer chemotherapy and endocrine therapy can affect memory, attention, executive function, and processing speed, but these heterogeneous and often partly reversible changes do not establish progressive AD pathology (17). Overall, current evidence does not support a consistent association of breast cancer or estrogen-directed therapy with either lower or higher AD risk.

3.2Prostate cancer: androgen-deprivation therapy

Prostate cancer extends this distinction to androgen signaling, for which the neurological consequences of profound and sustained hormone suppression may differ from any association attributable to the tumor itself. Androgen receptor signaling is a central driver of prostate cancer, whereas androgens contribute to neural, metabolic, and vascular homeostasis. Prostate cancer diagnosis and androgen-deprivation therapy (ADT) must therefore be evaluated separately. For prostate cancer itself, the apparent inverse association may reflect altered detection as much as an underlying biological association. A Swedish nationwide study compared 178,746 men with prostate cancer with 1,787,460 age-matched controls and also examined subsequent prostate cancer diagnoses among men with pre-existing dementia. Men with low- or intermediate-risk prostate cancer had little or no excess risk of AD or vascular dementia. By contrast, men with high-risk disease treated with ADT alone had higher risks of AD and vascular dementia, with hazard ratios of 1.37 and 1.51, respectively. In the reverse direction, men with pre-existing AD had substantially lower odds of subsequently diagnosed high-risk or low- to intermediate-risk prostate cancer. However, reduced prostate-specific antigen screening and diagnostic investigation after dementia may partly explain this apparent inverse association (18).

Several large observational studies associate ADT with increased dementia risk. Among 100,414 men aged ≥66 years with localized or locally advanced prostate cancer, pharmacological ADT was associated with 17% higher all-cause dementia risk and 23% higher AD risk. Exposure for ≥7 months was associated with increases of 25% and 34%, respectively (19). ADT modality may also be important. In a Taiwanese cohort of 23,651 patients, antiandrogen monotherapy was associated with increased risks of dementia and AD, whereas GnRH agonist treatment and orchiectomy showed no significant differences from no ADT (20). Other well-controlled cohorts have not reproduced these associations. Among 45,218 US veterans receiving definitive radiotherapy for nonmetastatic prostate cancer, ADT was not significantly associated with all-cause dementia, vascular dementia, or AD, including after treatment exceeding one year (21). Nevertheless, a meta-analysis of 28 studies comprising approximately 2.54 million participants found that ADT was associated with 20% higher all-cause dementia risk and 26% higher AD risk (4). The substantial variation among individual studies indicates that these pooled associations do not establish causality. Taken together, pooled observational evidence suggests a modest association between ADT exposure and all-cause dementia and AD, but substantial heterogeneity among studies precludes a causal conclusion (4). Differences in treatment modality and duration may contribute: recent meta-analyses report variation in dementia risk across ADT modalities and a nonlinear relationship with exposure duration (4, 22). A recent nationwide Korean cohort likewise found no significant association between ADT and overall dementia but a higher risk of AD, further illustrating outcome-specific heterogeneity (23). Differences in cancer stage and treatment indication may further introduce confounding by disease severity, while age and cardiovascular or metabolic comorbidities influence both ADT selection and baseline dementia risk. Finally, studies vary in follow-up duration and dementia ascertainment, ranging from administrative diagnostic codes to clinically defined outcomes, which may contribute to inconsistent estimates. Thus, current evidence supports a possible ADT-associated neurological risk rather than a protective or harmful effect attributable to prostate cancer itself. The major epidemiological studies discussed above, including differences in study population, cancer type, treatment exposure, neurological outcome, follow-up duration, and effect estimates, are summarized in Table 1.

Table 1Selected epidemiological studies of cancer, dementia/Alzheimer disease, and endocrine treatment.

Population Cancer type Sample size Exposure/treatment Outcome/ascertainment Key effect estimate (95% CI) Follow-up Major limitations Ref

UK adults ≥60 y; both sexes Multiple cancers 3,021,508; 412,903 with cancer Previous/incident cancer diagnosis; MR analysis Dementia/ADRD from EHR, hospital/death records or dementia medication Dementia: HR 0.75 (0.74–0.76); competing-risk SHR 0.56 (0.55–0.56); genetically predicted cancer–AD: OR 0.97 (0.95–0.99) Median 7.9 y; observation up to 30 y Possible dementia under-reporting/misclassification; no biomarker-defined AD; missing covariates and residual confounding (1)

International observational cohorts; mixed sex/age Multiple cancers 16 studies; 3,615,269 cancer survivors Cancer diagnosis; also endocrine therapy, chemotherapy, radiotherapy All-cause dementia (ACD), AD, VaD ACD: RR 0.92 (0.83–1.02); AD: RR 0.88 (0.78–0.99); VaD: RR 0.86 (0.75–0.99) Varied; analyses stratified by follow-up duration Very high between-study heterogeneity; observational evidence; English/published literature only; limited VaD evidence (5)

UK Biobank participants ≥55 y; predominantly White Multiple cancers 140,959 Prevalent or incident cancer All-cause dementia, AD, VaD from linked health records Incident cancer–all-cause dementia: HR 1.14 (1.02–1.29); prevalent cancer: HR 1.04 (0.92–1.17); AD not significantly associated; first year after cancer HR 1.83 (1.42–2.36) Median 12.3 y Healthy-volunteer/White-population selection; relatively young cohort; limited dementia cases; EHR diagnostic bias and incomplete time-varying health-care utilization (6)

Rotterdam Study; adults 60–70 y, cognitively normal at baseline Multiple cancers 3,634 Incident cancer diagnosis treated as time-varying proxy exposure Dementia diagnosis from medical records/registries Dementia RR ranged from 0.71 (0.49–0.95) to 1.10 (0.79–1.30), depending on exposure definition and bias-control assumptions Up to 20 y Primarily a causal-methodology case study; cancer is an imperfect proxy for proposed mechanisms; estimates sensitive to assumptions about death/confounding; second subcohort had 15-y follow-up extrapolated to 20 y (7)

Community autopsy cohort; mean age at death 83.8 y; 60.1% female Multiple cancers 785; 190 with cancer Registry-confirmed cancer history MCI/dementia and neuropathology at autopsy MCI/dementia: OR 0.45 (0.32–0.65); Braak V/VI: OR 0.38 (0.26–0.55); neuritic plaques: OR 0.52 (0.36–0.75) Longitudinal cohort; autopsy at death Only 190 cancer cases; unable to evaluate site, stage or treatment; autopsy cohort predominantly White and well educated (8)

Korean women undergoing breast cancer surgery; mean age 53.1 y Breast 70,701 survivors; matched cancer-free controls Breast cancer and subsequent treatments AD from national claims; competing-risk models Breast cancer vs controls: SHR 0.92 (0.86–0.98); radiotherapy: adjusted HR 0.77 (0.68–0.87); inverse association did not persist beyond 5 y Median 7.3 y No detailed stage/radiation-dose data; ICD-based AD ascertainment; selection and residual confounding; maximum follow-up 11 y (13)

Swedish women ≥50 y who survived breast cancer ≥5 y Breast 26,741 survivors; 249,540 controls Breast cancer survivorship All-cause dementia, AD, VaD using ICD codes Overall association null; diagnosis after age 65: ACD SHR 1.30 (1.07–1.58), AD 1.35 (1.05–1.75), VaD 1.64 (1.11–2.43) Up to 15 y after survivor entry Registry codes have limited sensitivity/onset precision; less severe dementia may be missed; race/ethnicity unavailable; survivor selection (14)

US women ≥45 y with breast cancer; mean age ≈76 y Breast 57,843 Hormone-modulating therapy (SERMs, ER antagonists, aromatase inhibitors) Claims-defined neurodegenerative disease, AD and dementia AD: RR 0.82 (0.75–0.90); dementia: RR 0.88 (0.83–0.93) Mean 5.5 y Retrospective claims data; residual confounding; filled prescriptions do not ensure adherence; limited pathology/treatment detail (15)

US women ≥65 y with newly diagnosed breast cancer Breast 18,808 Hormone-modulating therapy vs no HMT ADRD from SEER–Medicare HMT: HR 0.93 (0.88–0.98); association varied substantially by age and race Mean ≈12 y Medicare population limits age generalizability; no APOE or AD biomarkers; HMT duration not modeled; residual treatment confounding (3)

UK postmenopausal women ≥54 y with breast cancer Breast 14,214 Aromatase inhibitor vs tamoxifen Incident dementia from primary-care records/prescriptions AI vs tamoxifen: HR 1.04 (0.83–1.30); no significant difference 57,102 person-years Chemotherapy, stage, grade and ER status incompletely available; undiagnosed dementia missed; residual confounding (16)

Swedish men with prostate cancer; median age 67 y (low/intermediate risk), 75 y (high risk) Prostate 178,746 PC; 1,787,460 matched controls PC diagnosis, risk category and treatment including ADT AD and VaD from national registers High-risk PC treated only with ADT: AD HR 1.37 (1.19–1.58); VaD HR 1.51 (1.29–1.78). Prior AD associated with lower odds of high-risk PC: OR 0.39 (0.35–0.45) Median 7 y low/intermediate risk; 4 y high-risk ICD-based dementia may be underdiagnosed; possible detection bias; Swedish population limits generalizability (18)

US men ≥66 y; median age 73 y; localized/locally advanced PC Prostate 100,414 Pharmacologic ADT; duration 0, 1–6, ≥7 mo All-cause dementia and AD from Medicare claims Any ADT: all-cause dementia HR 1.17 (1.07–1.27); AD HR 1.23 (1.11–1.37) 36 mo Claims misclassification/underreporting; differential diagnostic surveillance with ADT; potential exposure-time alignment bias; relatively short follow-up (19)

Taiwanese men; median age 73 y Prostate 23,651 GnRH agonist, orchiectomy, antiandrogen monotherapy vs no ADT All-cause dementia and AD from linked claims/registries Antiandrogen: dementia HR 1.34 (1.16–1.55), AD HR 1.52 (1.13–2.04); GnRH and orchiectomy not significantly different from no ADT Median 3.46 y Claims-based diagnosis; residual confounding (education, lifestyle, family history); potential detection bias (20)

US male veterans with nonmetastatic PC receiving definitive radiotherapy Prostate 45,218 Radiotherapy with vs without ADT Any dementia, VaD, AD Any dementia: SHR 1.04 (0.94–1.16); VaD 1.20 (0.97–1.50); AD 1.11 (0.91–1.36); all nonsignificant Median 6.8 y Restricted to veterans receiving definitive radiotherapy; limited generalizability to other PC populations/treatments (21)

International studies of men with prostate cancer Prostate 28 studies; 2,543,483 participants reported across 27 studies ADT vs no ADT; multiple ADT modalities/durations Dementia, AD, VaD, PD Dementia: HR 1.20 (1.11–1.29); AD: HR 1.26 (1.10–1.43); VaD: HR 1.30 (0.97–1.73) Varied across included studies Substantial methodological heterogeneity and residual confounding; differences in disease stage/treatment; possible overlapping databases (4)

4Mechanisms of inverse comorbidity

The mechanisms considered here concern tumor- and tissue-level biology and should be distinguished from population-level associations and from systemic effects of anticancer treatment. These levels need not operate in the same direction, and the pathways discussed below should therefore be interpreted as candidate biological contributors rather than direct explanations of the epidemiological inverse association.

4.1Endocrine–metabolic signaling

The insulin/insulin-like growth factor (IGF)–PI3K–AKT axis provides a central metabolic intersection between cancer and AD, but its consequences depend on tissue context. In many cancers, hyperinsulinemia, IGF signaling, and oncogenic PI3K activation support glucose uptake, anabolic metabolism, proliferation, and resistance to apoptosis. Loss of PIP4K2C, for example, sensitized melanoma cells to insulin-dependent PI3K–AKT signaling and enabled them to exploit the insulin-rich liver environment for organ-specific metastasis (24). In the AD brain, by contrast, postmortem tissue shows impaired insulin-receptor–IRS-1–PI3K and IGF1-receptor–IRS-2–PI3K responses, with inhibitory IRS-1 phosphorylation correlating with Aβ pathology and cognitive decline (25). Systematic evidence further identifies brain insulin resistance and impaired glucose metabolism as recurrent features of AD, with metabolic dysfunction showing a more consistent association with tau than with amyloid biomarkers (26, 27). In rodent models, increasing circulating IGF-1 facilitated cerebral Aβ clearance by promoting the brain entry of Aβ carrier proteins, illustrating that peripheral IGF-1 actions can differ from impaired central insulin signaling (28). These shared endocrine and metabolic pathways help explain how cancer and AD can emerge from overlapping regulatory networks yet diverge into opposite biological outcomes (Figure 1).

The mechanistic target of rapamycin (mTOR) promotes protein synthesis and biosynthetic growth in tumors, whereas dysregulated mTOR activity in AD may aggravate metabolic and proteostatic stress. Rapamycin enhanced synaptic and mitochondrial metabolism in AD-model mice, but produced distinct responses in APOE3 and APOE4 backgrounds, indicating that the consequences of mTOR inhibition depend on genetic and metabolic context (29). AMP-activated protein kinase (AMPK) can restrain anabolic mTOR signaling, yet established tumors may exploit AMPK to survive nutrient limitation. Breast cancer cells adapted to glutamine deprivation through AMPK-dependent upregulation of PSAT1 and serine-pathway-derived α-ketoglutarate (30), while AMPK-driven phosphorylation of INF2 promoted mitochondrial fission and growth in endometrial cancer (31). Conversely, activation of neuronal and microglial AMPK by GLP-1 receptor agonists reduced amyloidogenic APP processing, enhanced Aβ phagocytosis, and improved memory in AD-model mice (32).

Mitochondrial metabolism provides a parallel distinction between tumor adaptation and neuronal energetic failure. In aging and amnestic mild cognitive impairment, synaptic and intramitochondrial Aβ42 accumulation correlated with impaired synaptic mitochondrial bioenergetics (33). AD brain tissue also exhibited increased methylation of mitochondrial ND5 mRNA, reduced complex I expression, and impaired respiration (34). Restoring NAD+ with nicotinamide mononucleotide enhanced the ATF4-dependent mitochondrial stress response and reduced synaptic and neuronal injury across AD models, further implicating loss of metabolic reserve in neurodegeneration (35). Sex-steroid signaling intersects with these metabolic pathways rather than acting independently. Particularly after menopause, obesity and insulin resistance increase estrogen availability and insulin–PI3K–AKT–mTOR signaling and are associated with poorer outcomes and possible endocrine-therapy resistance in ER-positive breast cancer (36). In neural systems, estrogen supports glucose utilization, mitochondrial respiration, and ATP production, although these effects depend on age, timing, and the metabolic health of the target cells (37). In prostate cancer, reciprocal AR–IGF1 signaling supports both androgen-dependent and androgen-independent growth through epithelial and stromal mechanisms (38). Taken together, these data establish that cancer and AD engage overlapping endocrine–metabolic networks, but they do not demonstrate that these pathways generate the epidemiological inverse association. Their contribution remains context-dependent and inferential, particularly because endocrine therapy can perturb the same pathways independently of tumor biology.

4.2Cell fate and chaperone-regulated proteostasis

Cancer cells bypass checkpoint arrest and apoptosis to sustain proliferation, whereas post-mitotic neurons cannot complete productive cell division. In aging and AD, persistent DNA-damage responses can reactivate cell-cycle machinery and drive neurons toward senescence or degeneration (39). Aβ oligomers can induce tau-dependent neuronal cell-cycle re-entry (40). Under oxidative stress, tau promotes S-phase entry as a transient pro-survival response (41), whereas single-cell analyses link neuronal cell-cycle re-entry more broadly to a senescent state (42). PIN1 illustrates how the same conformational regulator can promote tumor growth while preserving neuronal protein homeostasis. In breast cancer, PIN1 prevents the ubiquitin-dependent proteasomal and lysosomal degradation of METTL3, thereby increasing m6A-dependent translation of TAZ and EGFR and promoting tumorigenesis (43). Conversely, reduced PIN1 expression was detected in neocortical and limbic regions of individuals with mild cognitive impairment or AD (44), particularly in women, and was associated with greater neurofibrillary tangle burden and poorer cognition.

Cellular senescence is similarly bidirectional. Senescence initially restrains abnormal proliferation, but persistent senescent cells acquire a senescence-associated secretory phenotype that can promote chronic inflammation, tumor progression, and age-related tissue dysfunction (45). In postmortem AD brains, imaging and single-nucleus transcriptomics of more than 200,000 nuclei revealed increased senescence markers predominantly in glia; senescent microglia were associated with greater Aβ burden and reduced expression of phagocytic pathways required for amyloid clearance (46). A pilot trial of dasatinib plus quercetin demonstrated cerebrospinal-fluid penetration of dasatinib and short-term feasibility, although the small sample did not establish cognitive efficacy (47). The HSF1–heat-shock-protein system links cellular stress adaptation with proteostasis capacity. In breast cancer cells, BAG1 overexpression increased HSF1 phosphorylation and elevated HSP70 and HSP27, supporting survival and stabilization of tumor-promoting signaling; HSP90 abundance itself was not altered (48). By chaperoning steroid hormone receptors, including ER and AR, and stabilizing diverse oncogenic clients, HSP90 contributes to the non-oncogene dependence of hormone-driven tumors (49). HSF1 also acts beyond cancer cells: its activation in tumor-infiltrating natural killer cells directly suppresses cytotoxicity and cytokine production, thereby weakening antitumor immunity (50). In the aging brain, by contrast, the stress-inducible HSF1–HSP response becomes attenuated, reducing chaperone reserve and increasing vulnerability to Aβ, tau, and other aggregation-prone proteins (51–53).

Autophagy–lysosome function determines whether proteotoxic stress is successfully buffered or progresses toward aggregate accumulation. Although autophagy can limit early tumor development, established cancers exploit it for nutrient recycling, mitochondrial quality control, metabolic adaptation, and survival under hypoxia or therapeutic stress (54, 55). In tauopathy, disruption of TFEB–vacuolar ATPase signaling impaired lysosomal acidification, increased tau pathology, and weakened the microglial response (56). APOE4 can further suppress transcription of lysosomal-autophagy genes, whereas blocking this interaction restores aggregate-clearance pathways in cellular and animal models (57). Autophagy is also non-uniform in AD: autophagic induction may be preserved or increased while downstream lysosomal degradation becomes inefficient. In AD mouse models, impaired vATPase-dependent autolysosomal acidification precedes extracellular amyloid deposition and is accompanied by accumulation of Aβ and APP-βCTF in poorly acidified autolysosomes (58). Thus, rather than a simple opposition between enhanced autophagy in cancer and impaired autophagy in AD, the consequences of autophagy depend on disease stage and the integrity of downstream lysosomal degradation. Autophagy can suppress early tumorigenesis yet support established tumors, while in AD increased autophagic activity may coexist with ineffective lysosomal clearance (Figure 1). More broadly, these studies support a contrast between enhanced stress tolerance and proteostatic adaptation in established tumors and progressive cell-fate and proteostasis failure in AD. However, most evidence derives from disease-specific human tissue or experimental models rather than direct cancer–AD systems. These pathways therefore provide biological plausibility for divergent cellular states, but they do not by themselves establish a mechanism for inverse comorbidity.

5Amyloidogenic hormones and self-assembly at the cancer–brain interface

Protein self-assembly provides a potential mechanistic bridge between endocrinology, cancer, and AD, although the evidence ranges from direct experimental observations to still-unproven connections with population-level inverse comorbidity. Rather than being exclusively pathological, amyloid-like conformations can support the dense and stable storage of peptide and protein hormones within secretory granules (Figure 2A). Cross-β-rich assemblies have been detected for prolactin, adrenocorticotropic hormone, growth hormone, β-endorphin, oxytocin, and vasopressin, and these deposits can release biologically active monomers following secretion into the extracellular environment (59–61). This physiological use of controlled aggregation places endocrine tissues near a boundary between reversible hormone storage and pathological deposition. IAPP was originally identified in amyloid isolated from a pancreatic islet-cell tumor and is co-stored and co-secreted with insulin, whereas full-length calcitonin constitutes the amyloid deposited in medullary thyroid carcinoma (62, 63). Pathogenic provasopressin mutants likewise form amyloid-like assemblies in familial neurohypophyseal diabetes insipidus, linking abnormal hormone self-assembly to defective secretory-granule sorting and endocrine dysfunction (64) (Figure 2A). The presence of calcitonin-derived amyloid in human medullary thyroid tumors has also enabled selective imaging with an amyloid-binding PET tracer, demonstrating that hormone-derived amyloid is a clinically detectable component of an endocrine tumor microenvironment (65).

Importantly, amyloidogenic peptides and their assembly states are not biologically equivalent. Molecular-dynamics simulations of Aβ- and IAPP-derived fragments showed distinct homotypic aggregate architectures and heterotypic Aβ–IAPP assemblies with altered β-barrel propensity, while Aβ42–hIAPP co-oligomers produced a more advanced AD-like phenotype with markedly increased neuronal death in human cerebral organoids (66, 67). These findings illustrate that the biological consequences of self-assembly depend on peptide identity, assembly state, and heterotypic interactions; functional hormone-storage assemblies therefore should not be equated with disease-associated toxic species. In cancer, aggregation of tumor suppressors may instead promote malignant survival. Destabilized wild-type or mutant p53 can form amyloid-like oligomers and fibrils in breast-cancer models and glioblastoma, sequestering wild-type p53 and its paralogs p63 and p73 to produce dominant-negative and gain-of-function phenotypes (68–71) (Figure 2B). Recent experiments further showed that oncogenic p53 converts p63 and p73 condensates into β-sheet-rich aggregates (72). Nevertheless, this behavior should be described cautiously: some studies support seeded, prion-like conformational conversion, whereas others indicate that mutant and wild-type proteins primarily associate through direct coaggregation rather than classical catalytic seeding (73). The functional importance of the aggregated state is supported by ReACp53, a designed peptide that inhibited p53 aggregation, restored p53-dependent transcription and apoptosis in ovarian carcinoma cells and patient-derived organoids, and suppressed tumor growth in vivo (74) (Figure 2B). Thus, p53 aggregation is included here as an independent cancer-side example of context-dependent protein self-assembly rather than as a mechanistic bridge between cancer and AD. We therefore do not interpret p53 aggregation as evidence for a cancer-to-AD pathway or an amyloid-mediated protective effect against cancer. Its relevance is conceptual: protein self-assembly can produce distinct pathological outcomes depending on protein identity and cellular context, as illustrated by loss or gain of p53 function in cancer versus neurotoxicity of Aβ and tau in AD. In AD, aggregation of Aβ and tau is associated with synaptic dysfunction and neuronal degeneration (Figure 2C). Recent evidence provides a more direct cancer-to-AD mechanistic connection. In AD mouse models, peripheral lung, prostate, and colon tumors attenuated amyloid pathology and improved cognition through secretion of cystatin C, which interacted with Aβ and activated TREM2-dependent microglial clearance. Genetic disruption of TREM2 or mutation of cystatin C abolished these effects, establishing a causal tumor-to-brain mechanism in experimental models. Notably, this effect was observed across lung, prostate, and colon tumor models, suggesting that the mechanism is not restricted to hormone-dependent cancers, although its relevance across human cancer types remains unknown. However, whether this pathway contributes to the epidemiological cancer–AD association in humans remains unknown (75).

Conversely, preclinical studies indicate that neurodegenerative and endocrine amyloids can alter cancer-cell fitness. Aβ inhibited angiogenesis and reduced the vascularization and growth of glioblastoma and lung-cancer xenografts, while an Aβ-derived antiangiogenic fragment suppressed breast-tumor growth (76, 77). Oligomeric Aβ also preferentially impaired breast cancer stem cells by disrupting lysosomal iron homeostasis and increasing iron-dependent oxidative stress, whereas tumor-targeted KLVFF motifs assembled into extracellular nanofiber networks that restricted breast-cancer migration, vascular passage, and spontaneous metastasis (78, 79). Extending these observations beyond Aβ, hIAPP, and human calcitonin inhibited PANC-1 proliferation and migration in sequence-, concentration-, and aggregation-state-dependent manners (80) (Figure 2D). For hIAPP, aggregation is strongly sequence- and processing-dependent: β-sheet-prone regions include residues 12–17, 23–27, and 32–37, with the FGAIL motif contributing prominently to amyloid formation; moreover, unprocessed prepro-IAPP can also aggregate and induce neuronal degeneration in vivo (81, 82). Comparable short amyloidogenic/self-recognition motifs have been identified in human calcitonin (DFNKF) and Aβ (KLVFF) (83, 84). These findings emphasize that biological effects observed for a particular peptide preparation should not be generalized across monomeric, oligomeric, fibrillar, intracellular, or extracellular states. Monomer-initiated conditions predominantly elevated intracellular reactive oxygen species, whereas oligomeric preparations produced greater membrane leakage and membrane-disruption-associated cell death.

The physiological relevance and cellular selectivity of these anticancer effects remain uncertain. In the PANC-1 study, amyloid peptides were tested over concentrations of 0.1–160 μM, from which IC25, IC50, and IC75 values were determined; whether comparable concentrations and aggregation states occur within human tumor microenvironments has not been established (80). Circulating Aβ concentrations are orders of magnitude below the concentrations typically used in these cell-based assays, and local concentrations within tumor microenvironments have not been established (85, 86). Moreover, experimentally prepared monomeric and oligomeric species may not reproduce the heterogeneous and transient assemblies present in vivo. Evidence for cancer-cell selectivity is also limited: oligomeric Aβ preferentially inhibited breast cancer stem cells relative to bulk cancer cells (79), but this does not establish selective toxicity relative to matched nonmalignant cells, and ROS generation and membrane disruption can represent general proteotoxic mechanisms. A potential basis for differential susceptibility exists because many malignant cells expose phosphatidylserine and other anionic components on the outer membrane, features exploited by some membrane-active anticancer peptides (87, 88). Whether these properties account for the effects of Aβ, hIAPP, or calcitonin remains untested. Human circulating-Aβ data also do not support a simple systemic protective model: plasma Aβ40 and Aβ42 were elevated in a case-control study of patients with several cancers, and higher plasma Aβ40 and Aβ42 were associated with higher subsequent cancer incidence in a population-based cohort (85, 86). These observations do not address local peptide concentration or aggregation state within tumors, but they further argue against interpreting circulating Aβ as a straightforward anticancer factor. These limitations generate specific testable predictions. If amyloid self-assembly contributes biologically to cancer suppression, inhibitory effects should persist at concentrations measured in relevant tumor microenvironments, vary reproducibly with aggregation state, be attenuated when aggregation is prevented, scale with outer-leaflet phosphatidylserine exposure if membrane-mediated, and show greater effects in malignant cells than in matched nonmalignant cells. Failure to meet these criteria would favor nonspecific proteotoxicity rather than a cancer-selective mechanism.

Together, these preclinical studies show that protein self-assembly can influence both neurodegenerative and cancer phenotypes in experimental systems, but they do not establish that amyloid-mediated suppression of cancer-cell fitness explains inverse comorbidity in humans. Indeed, melanoma-secreted Aβ can promote brain metastasis by suppressing local neuroinflammatory responses, further demonstrating that amyloid effects are strongly context dependent (85, 89). Thus, the contribution of amyloid self-assembly to the population-level cancer–AD association should presently be regarded as a testable mechanistic hypothesis rather than an established causal mechanism. The major metabolic, cell-fate, proteostatic, and protein-self-assembly mechanisms discussed above are summarized in Table 2.

Table 2Major mechanistic pathways at the cancer–Alzheimer’s disease interface.

Pathway/protein Cancer context AD context Key mechanistic connection Experimental model

Insulin/IGF–PI3K–AKT Supports glucose uptake, anabolic metabolism, proliferation, and apoptosis resistance Brain insulin resistance and impaired insulin/IGF signaling Shared metabolic network produces divergent tissue-specific outcomes Cancer cell models; human AD brain tissue; rodent models

mTOR/AMPK Supports biosynthesis and metabolic adaptation under stress Modulates neuronal metabolism, APP processing, Aβ clearance, and proteostasis Nutrient-sensing pathways influence adaptation versus dysfunction Cancer cell models; AD mouse models

Mitochondrial metabolism Enables metabolic flexibility and survival under stress Mitochondrial energetic failure and oxidative stress Different capacities to maintain metabolic reserve under chronic stress Human AD tissue; cellular and animal models

Cell cycle/PIN1/senescence Promotes proliferation and tumor progression; senescence may support tumor-promoting inflammation Aberrant neuronal cell-cycle re-entry and glial senescence contribute to neurodegeneration Opposing regulation of cell fate and stress responses Cancer cell models; human AD tissue; single cell/transcriptomic studies

Chaperones/autophagy–lysosome Supports proteotoxic stress tolerance and survival in established tumors Reduced chaperone reserve and inefficient lysosomal degradation promote aggregate accumulation Proteostasis capacity determines whether misfolded proteins are cleared, tolerated, or become pathogenic Cancer cell models; human AD tissue; tauopathy and AD animal models

p53 aggregation Aggregated p53 can produce dominant-negative and gain-of-function phenotypes that promote malignancy No established direct AD pathway Illustrates context-dependent pathological consequences of protein self-assembly Cancer cells; patient-derived organoids; tumor models

Cystatin C–TREM2 Peripheral tumors secrete cystatin C Promotes TREM2-dependent microglial Aβ clearance Direct tumor-to-brain signaling pathway affecting amyloid pathology Lung, prostate, and colon tumor models in AD mice

Aβ/hIAPP/calcitonin Can alter angiogenesis, proliferation, migration, and cancer-cell fitness Amyloid assembly contributes to neurodegenerative pathology Effects depend on peptide identity, concentration, and aggregation state Cancer cell assays; xenografts; peptide-assembly models

6Conclusions and future outlook

The inverse association between cancer and dementia, including AD, is unlikely to arise from a single causal pathway. Rather, the observed pattern likely reflects the combined effects of competing mortality, diagnostic ascertainment, systemic endocrine and metabolic states, anticancer treatment, and oppositely regulated cellular programs. Hormone-dependent cancers illustrate this complexity: tumor-associated proliferative and metabolic states may differ fundamentally from the systemic consequences of endocrine therapy. Accordingly, breast cancer hormone-modulating therapy and prostate cancer androgen-deprivation therapy can show distinct neurological associations. Genetic and transcriptomic studies further identify opposing associations across AD and several cancers, implicating cell-cycle control, apoptosis, metabolism, and proteostasis. This framework is centered on Alzheimer-type amyloid–tau pathology and should not be assumed to generalize to other neurodegenerative proteinopathies. Cancer associations differ across disorders—for example, melanoma shows a positive association with Parkinson’s disease—and neuropathological associations with TDP-43 and α-synuclein appear less consistent than those observed for AD-type pathology. Preclinical evidence that amyloid assemblies can suppress aspects of cancer-cell fitness under defined experimental conditions provides biological plausibility for one connection, while the recently identified tumor-derived cystatin C–TREM2 pathway establishes a cancer-to-brain mechanism in mice; whether either contributes materially to inverse comorbidity in humans remains unknown.

Future studies should therefore separate cancer subtype and stage from endocrine treatment class, dose, and duration while applying time-varying exposure and competing-risk analyses. Incorporating plasma Aβ42/Aβ40 ratio, phosphorylated tau, neuroimaging, endocrine markers, and metabolic profiles could determine whether cancer or its treatment modifies preclinical AD biology rather than only dementia diagnosis. Mechanistic studies should test amyloid–cancer interactions at physiologically relevant concentrations, aggregation states, and exposure durations in patient-derived organoids, immune-competent cocultures, and animal models. Therapeutic interpretation must remain cautious: pathways revealed by inverse comorbidity may identify useful targets, but neither amyloid toxicity nor anticancer treatment should be considered intrinsically beneficial. Integrating treatment-aware epidemiology with physiologically relevant models will be essential to distinguish genuine biological antagonism from observational bias.

References

1Bassil DT Zheng B Su B Kafetsouli D Udeh-Momoh C Tzoulaki I .
Lower incidence of dementia following cancer diagnoses: Evidence from a large cohort and Mendelian randomization study . J Prev Alzheimer's Dis . (2024 ) 11 :1397–405. doi: 10.14283/jpad.2024.135 39350386
PMC11436397

2Aung KZ Zin SS Wu X Myint ZW Karanth S Estus S .
Disentangling the inverse relationship between cancer and Alzheimer's or Parkinson's disease: A systematic review on Mendelian randomization studies . Neurobiol Dis . (2025 ) 217 :107190. doi: 10.1016/j.nbd.2025.107190 41241298
PMC12766696

3Cai C Strickland K Knudsen S Tucker SB Chidrala CS Modugno F .
Alzheimer disease and related dementia following hormone-modulating therapy in patients with breast cancer . JAMA Netw Open . (2024 ) 7 :e2422493. doi: 10.1001/jamanetworkopen.2024.22493 39012631
PMC11252894

4Hinojosa-Gonzalez DE Zafar A Saffati G Kronstedt S Zlatev DV Khera M .
Androgen deprivation therapy for prostate cancer and neurocognitive disorders: a systematic review and meta-analysis . Prostate Cancer Prostatic Dis . (2024 ) 27 :507–19. doi: 10.1038/s41391-023-00785-w 38167924

5Ma L Tan ECK Goudey B Jin L Pan Y .
Unraveling the bidirectional link between cancer and dementia and the impact of cancer therapies on dementia risk: A systematic review and meta-analysis . Alzheimer's Dementia . (2025 ) 21 :e14540. doi: 10.1002/alz.14540 39807644
PMC11851135

6Wang J Buto P Ackley SF Kobayashi LC Graff RE Zimmerman SC .
Association between cancer and dementia risk in the UK Biobank: evidence of diagnostic bias . Eur J Epidemiol . (2023 ) 38 :1069–79. doi: 10.1007/s10654-023-01036-x 37634228
PMC10854217

7Rojas-Saunero LP van der Willik KD Schagen SB Ikram MA Swanson SA .
Towards a clearer causal question underlying the association between cancer and dementia . Epidemiology . (2024 ) 35 :281–8. doi: 10.1097/ede.0000000000001712 38442423
PMC11022995

8Karanth SD Katsumata Y Nelson PT Fardo DW McDowell JK Schmitt FA .
Cancer diagnosis is associated with a lower burden of dementia and less Alzheimer’s-type neuropathology . Brain . (2022 ) 145 :2518–27. doi: 10.1093/brain/awac035 35094057
PMC9612796

9Debnath D Housini M Sariya S Phillips NR Pathak GA Barber RC .
Locus- and gene-level insights into the inverse association between Alzheimer’s disease and cancer . Int J Mol Sci . (2026 ) 27 :2900. doi: 10.3390/ijms27062900 41898760
PMC13026669

10Kim SY Choi HG Kim YH Kwon MJ Kim J-H Lee HS .
Longitudinal study of the inverse relationship between Parkinson’s disease and cancer in Korea . NPJ Parkinson's Dis . (2023 ) 9 :116 . doi: 10.1038/s41531-023-00562-5 37481603
PMC10363116

11Huang P Yang X-D Chen S-D Xiao Q .
The association between Parkinson’s disease and melanoma: a systematic review and meta-analysis . Trans Neurodegeneration . (2015 ) 4 :21 . doi: 10.1186/s40035-015-0044-y 26535116
PMC4631109

12Teipel S Akmatov M Michalowsky B Junghanss C Holstiege J Bohlken J .
Association of cancer with neuropathological markers of Alzheimer's disease and related dementias . Alzheimers Dement (Amst) . (2025 ) 17 :e70222. doi: 10.1002/dad2.70222 41367891
PMC12682591

13Jeong S-M Jung W Cho H Choi HL Jeon KH Nam K-W .
Alzheimer disease in breast cancer survivors . JAMA Netw Open . (2025 ) 8 :e2516468. doi: 10.1001/jamanetworkopen.2025.16468 40540273
PMC12181787

14Wennberg A Ding M Feychting M Modig K .
Risk of all-cause dementia, Alzheimer disease, and vascular dementia in breast cancer survivors . Neurol Clin Pract . (2023 ) 13 :e200173. doi: 10.1212/cpj.0000000000200173 37188244
PMC10176959

15Branigan GL Soto M Neumayer L Rodgers K Brinton RD .
Association between hormone-modulating breast cancer therapies and incidence of neurodegenerative outcomes for women with breast cancer . JAMA Netw Open . (2020 ) 3 :e201541. doi: 10.1001/jamanetworkopen.2020.1541 32207833
PMC7093781

16Bromley SE Matthews A Smeeth L Stanway S Bhaskaran K .
Risk of dementia among postmenopausal breast cancer survivors treated with aromatase inhibitors versus tamoxifen: a cohort study using primary care data from the UK . J Cancer Survivorship . (2019 ) 13 :632–40. doi: 10.1007/s11764-019-00782-w 31321612
PMC6776493

17Oliva G Giustiniani A Danesin L Burgio F Arcara G Conte P .
Cognitive impairment following breast cancer treatments: an umbrella review . Oncologist . (2024 ) 29 :e848–63. doi: 10.1093/oncolo/oyae090 38723166
PMC11224991

18Crump C Stattin P Brooks JD Sundquist J Wei J Sieh W .
Associations between prostate cancer and dementia: A nationwide study in Sweden . Eur Urol Oncol . (2025 ) 8 :857–65. doi: 10.1016/j.euo.2025.04.001 40280786
PMC12790758

19Krasnova A Epstein M Marchese M Dickerman BA Cole AP Lipsitz SR .
Risk of dementia following androgen deprivation therapy for treatment of prostate cancer . Prostate Cancer Prostatic Dis . (2020 ) 23 :410–8. doi: 10.1038/s41391-019-0189-3 31784699

20Huang W-K Liu C-H Pang S-T Liu J-R Chang JW-C Liaw C-C .
Type of androgen deprivation therapy and risk of dementia among patients with prostate cancer in Taiwan . JAMA Netw Open . (2020 ) 3 :e2015189. doi: 10.1001/jamanetworkopen.2020.15189 32865575
PMC7489824

21Deka R Simpson DR Bryant AK Nalawade V McKay R Murphy JD .
Association of androgen deprivation therapy with dementia in men with prostate cancer who receive definitive radiation therapy . JAMA Oncol . (2018 ) 4 :1616–7. doi: 10.1001/jamaoncol.2018.4423 30325986
PMC6248075

22Zhu S Ni J Long S Teng Y Yao Y Shi J .
Androgen deprivation therapy and dementia risk: An updated and dose-response meta-analysis . J Alzheimer’s Dis . (2025 ) 104 :547–62. doi: 10.1177/13872877251318003 39956977

23Jeon H-L Choo E Jeong S Yun J Jeong CW Lee H .
Risk of Alzheimer’s disease and Parkinson’s disease following androgen deprivation therapy in a real world nationwide cohort . Sci Rep . (2025 ) 15 :23490 . doi: 10.1038/s41598-025-08279-6 40603530
PMC12223037

24Rogava M Aprati TJ Chi W-Y Melms JC Hug C Davis SH .
Loss of Pip4k2c confers liver-metastatic organotropism through insulin-dependent PI3K-AKT pathway activation . Nat Cancer . (2024 ) 5 :433–47. doi: 10.1038/s43018-023-00704-x 38286827
PMC11175596

25Talbot K Wang H-Y Kazi H Han L-Y Bakshi KP Stucky A .
Demonstrated brain insulin resistance in Alzheimer’s disease patients is associated with IGF-1 resistance, IRS-1 dysregulation, and cognitive decline . J Clin Invest . (2012 ) 122 :1316–38. doi: 10.1172/jci59903 22476197
PMC3314463

26Andrade LJO Oliveira LM Bittencourt AMV Lourenço LGC Oliveira GCM .
Brain insulin resistance and Alzheimer’s disease: a systematic review . Dementia Neuropsychol . (2024 ) 18 :5 –14 . doi: 10.1590/1980-5764-dn-2023-0032 38425702
PMC10901561

27van Gils V Rizzo M Côté J Viechtbauer W Fanelli G Salas-Salvadó J .
The association of glucose metabolism measures and diabetes status with Alzheimer’s disease biomarkers of amyloid and tau: A systematic review and meta-analysis . Neurosci Biobehav Rev . (2024 ) 159 :105604. doi: 10.1016/j.neubiorev.2024.105604 38423195

28Carro E Trejo JL Gomez-Isla T LeRoith D Torres-Aleman I .
Serum insulin-like growth factor I regulates brain amyloid-β levels . Nat Med . (2002 ) 8 :1390–7. doi: 10.1038/nm1202-793 12415260

29Sanganahalli BG Mihailovic JM Vekaria HJ Coman D Yackzan AT Flemister A .
mTOR inhibition enhances synaptic and mitochondrial function in Alzheimer's disease in an APOE genotype-dependent manner . J Cereb Blood Flow Metab . (2024 ) 44 :1745–58. doi: 10.1177/0271678x241261942 38879800
PMC11494852

30Qiu Y Stamatatos OT Hu Q Ruiter Swain J Russo S Sann A .
The unique catalytic properties of PSAT1 mediate metabolic adaptation to glutamine blockade . Nat Metab . (2024 ) 6 :1529–48. doi: 10.1038/s42255-024-01104-w 39192144
PMC11490312

31Ding Y Lv Z Cao W Shi W He Q Gao K .
Phosphorylation of INF2 by AMPK promotes mitochondrial fission and oncogenic function in endometrial cancer . Cell Death Dis . (2024 ) 15 :65 . doi: 10.1038/s41419-024-06431-0 38233384
PMC10794193

32Zhang Y Chen H Feng Y Liu M Lu Z Hu B .
Activation of AMPK by GLP-1R agonists mitigates Alzheimer-related phenotypes in transgenic mice . Nat Aging . (2025 ) 5 :1097–113. doi: 10.1038/s43587-025-00869-3 40394225

33Jia K Tian J Wang T Guo L Xuan Z Swerdlow RH .
Mitochondria-sequestered Aβ renders synaptic mitochondria vulnerable in the elderly with a risk of Alzheimer disease . JCI Insight . (2023 ) 8 :e174290. doi: 10.1172/jci.insight.174290 37991017
PMC10721326

34Jörg M Plehn JE Kristen M Lander M Walz L Lietz C .
N1-methylation of adenosine (m1A) in ND5 mRNA leads to complex I dysfunction in Alzheimer’s disease . Mol Psychiatry . (2024 ) 29 :1427–39. doi: 10.1038/s41380-024-02421-y PMC11189808 38287100

35Xiong X Hou J Zheng Y Jiang T Zhao X Cai J .
NAD+-boosting agent nicotinamide mononucleotide potently improves mitochondria stress response in Alzheimer’s disease via ATF4-dependent mitochondrial UPR . Cell Death Dis . (2024 ) 15 :744 . doi: 10.1038/s41419-024-07062-1 39394148
PMC11470026

36Javed SR Skolariki A Zameer MZ Lord SR .
Implications of obesity and insulin resistance for the treatment of oestrogen receptor-positive breast cancer . Br J Cancer . (2024 ) 131 :1724–36. doi: 10.1038/s41416-024-02833-1 39251829
PMC11589622

37Rettberg JR Yao J Brinton RD .
Estrogen: A master regulator of bioenergetic systems in the brain and body . Front Neuroendocrinol . (2014 ) 35 :8 –30 . doi: 10.1016/j.yfrne.2013.08.001 23994581
PMC4024050

38Adzavon YM Culig Z Sun Z .
Interactions between androgen and IGF1 axes in prostate tumorigenesis . Nat Rev Urol . (2025 ) 22 :268–75. doi: 10.1038/s41585-024-00942-3 39375467

39Wong GC Chow KH .
DNA damage response-associated cell cycle re-entry and neuronal senescence in brain aging and Alzheimer's disease . J Alzheimers Dis . (2023 ) 94 :S429–51. doi: 10.3233/JAD-220203 PMC10473156 35848025

40Seward ME Swanson E Norambuena A Reimann A Cochran JN Li R .
Amyloid-β signals through tau to drive ectopic neuronal cell cycle re-entry in Alzheimer's disease . J Cell Sci . (2013 ) 126 :1278–86. doi: 10.1242/jcs.1125880 23345405
PMC3635465

41Denechaud M Geurs S Comptdaer T Bégard S Garcia-Núñez A Pechereau L-A .
Tau promotes oxidative stress-associated cycling neurons in S phase as a pro-survival mechanism: Possible implication for Alzheimer’s disease . Prog Neurobiol . (2023 ) 223 :102386. doi: 10.1016/j.pneurobio.2022.102386 36481386

42Wu D Sun JK-L Chow KH-M .
Neuronal cell cycle reentry events in the aging brain are more prevalent in neurodegeneration and lead to cellular senescence . PloS Biol . (2024 ) 22 :e3002559. doi: 10.1371/journal.pbio.3002559 38652714
PMC11037540

43Bhattarai PY Kim G Lim S-C Mariappan R Ohn T Choi HS .
METTL3 stabilization by PIN1 promotes breast tumorigenesis via enhanced m6A-dependent translation . Oncogene . (2023 ) 42 :1010–23. doi: 10.1038/s41388-023-02617-6 36755057

44de Ávila C Suazo C Nolz J Nicholas Cochran J Wang Q Velazquez R .
Reduced PIN1 expression in neocortical and limbic brain regions in female Alzheimer’s patients correlates with cognitive and neuropathological phenotypes . Neurobiol Aging . (2024 ) 141 :160–70. doi: 10.1016/j.neurobiolaging.2024.06.007 38964013

45McHugh D Durán I Gil J .
Senescence as a therapeutic target in cancer and age-related diseases . Nat Rev Drug Discov . (2025 ) 24 :57 –71 . doi: 10.1038/s41573-024-01074-4
39548312

46Fancy NN Smith AM Caramello A Tsartsalis S Davey K Muirhead RCJ .
Characterisation of premature cell senescence in Alzheimer’s disease using single nuclear transcriptomics . Acta Neuropathol . (2024 ) 147 :78 . doi: 10.1007/s00401-024-02727-9 38695952
PMC11065703

47Gonzales MM Garbarino VR Kautz TF Palavicini JP Lopez-Cruzan M Dehkordi SK .
Senolytic therapy in mild Alzheimer’s disease: a phase 1 feasibility trial . Nat Med . (2023 ) 29 :2481–8. doi: 10.1038/s41591-023-02543-w 37679434
PMC10875739

48Kizilboga T Özden C Can ND Onay Ucar E Dinler Doganay G .
Bag-1-mediated HSF1 phosphorylation regulates expression of heat shock proteins in breast cancer cells . FEBS Open Bio . (2024 ) 14 :1559–69. doi: 10.1002/2211-5463.13843 PMC11492399 39049197

49Whitesell L Lindquist SL .
HSP90 and the chaperoning of cancer . Nat Rev Cancer . (2005 ) 5 :761–72. doi: 10.1038/nrc1716 16175177

50Hockemeyer K Sakellaropoulos T Chen X Ivashkiv O Sirenko M Zhou H .
The stress response regulator HSF1 modulates natural killer cell anti-tumour immunity . Nat Cell Biol . (2024 ) 26 :1734–44. doi: 10.1038/s41556-024-01490-z 39223375
PMC12312658

51Neef DW Jaeger AM Thiele DJ .
Heat shock transcription factor 1 as a therapeutic target in neurodegenerative diseases . Nat Rev Drug Discov . (2011 ) 10 :930–44. doi: 10.1038/nrd3453 22129991
PMC3518299

52Khatatneh S Sőti C Somogyvári M .
HSF1 in aging: The output selection model . Mech Ageing Dev . (2026 ) 233 :112225. doi: 10.1016/j.mad.2026.112225 42419688

53Liu AY Minetti CA Remeta DP Breslauer KJ Chen KY .
HSF1, aging, and neurodegeneration . In:
Turksen K , editor. Cell Biology and Translational Medicine, Volume 18: Tissue Differentiation, Repair in Health and Disease .
Springer Nature Switzerland, Cham (2023 ). p. 23 –49 .
10.1007/5584_2022_733 35995906

54Debnath J Gammoh N Ryan KM .
Autophagy and autophagy-related pathways in cancer . Nat Rev Mol Cell Biol . (2023 ) 24 :560–75. doi: 10.1038/s41580-023-00585-z 36864290
PMC9980873

55Assi M Kimmelman AC .
Impact of context-dependent autophagy states on tumor progression . Nat Cancer . (2023 ) 4 :596 –607 . doi: 10.1038/s43018-023-00546-7 37069394
PMC10542907

56Wang B Martini-Stoica H Qi C Lu T-C Wang S Xiong W .
TFEB–vacuolar ATPase signaling regulates lysosomal function and microglial activation in tauopathy . Nat Neurosci . (2024 ) 27 :48 –62 . doi: 10.1038/s41593-023-01494-2
PMC12203451 37985800

57Balasubramaniam M Narasimhappagari J Liu L Ganne A Ayyadevara S Atluri R .
Rescue of ApoE4-related lysosomal autophagic failure in Alzheimer’s disease by targeted small molecules . Commun Biol . (2024 ) 7 :60 . doi: 10.1038/s42003-024-05767-9 38191671
PMC10774381

58Lee J-H Yang D-S Goulbourne CN Im E Stavrides P Pensalfini A .
Faulty autolysosome acidification in Alzheimer’s disease mouse models induces autophagic build-up of Aβ in neurons, yielding senile plaques . Nat Neurosci . (2022 ) 25 :688 –701 . doi: 10.1038/s41593-022-01084-8 35654956
PMC9174056

59Maji SK Perrin MH Sawaya MR Jessberger S Vadodaria K Rissman RA .
Functional amyloids as natural storage of peptide hormones in pituitary secretory granules . Science . (2009 ) 325 :328–32. doi: 10.1126/science.1173155 19541956
PMC2865899

60Riek R Eisenberg DS .
The activities of amyloids from a structural perspective . Nature . (2016 ) 539 :227–35. doi: 10.1038/nature20416 27830791

61Fowler DM Koulov AV Balch WE Kelly JW .
Functional amyloid – from bacteria to humans . Trends Biochem Sci . (2007 ) 32 :217–24. doi: 10.1016/j.tibs.2007.03.003 17412596

62Stridsberg M Wilander E .
Islet Amyloid Polypeptide (IAPP) A short review . Acta Oncol . (1991 ) 30 :451–6. doi: 10.3109/02841869109092400 1854502

63Khurana R Agarwal A Bajpai VK Verma N Sharma AK Gupta RP .
Unraveling the amyloid associated with human medullary thyroid carcinoma . Endocrinology . (2004 ) 145 :5465–70. doi: 10.1210/en.2004-0780 15459123

64Beuret N Hasler F Prescianotto-Baschong C Birk J Rutishauser J Spiess M .
Amyloid-like aggregation of provasopressin in diabetes insipidus and secretory granule sorting . BMC Biol . (2017 ) 15 :5. doi: 10.1186/s12915-017-0347-9 28122547
PMC5267430

65Li C Zhang P Nie R Gong X Xie J Yu Z .
Targeting amyloids with [18F]AV-45 for medullary thyroid carcinoma positron emission tomography/computed tomography imaging: A pilot clinical study . Mol Pharmaceutics . (2022 ) 19 :584–91. doi: 10.1021/acs.molpharmaceut.1c00680 34982563
PMC8826757

66Cramer NA Kawecki G Bevan DR Brown AM .
Insight into amyloid interactions: molecular dynamics simulations of model peptide fragments . Biophys J . (2019 ) 116 :437a . doi: 10.1016/j.bpj.2018.11.2353 42574925

67Yan J Tang Z Luo Y Liu B Huang X Wei J .
Islet amyloid polypeptide modelled to simulate diabetes co-oligomerized with β-amyloid 1-42 reproducing the pathological cascade of alzheimer's disease in human cerebral organoids . Adv Sci . (2026 ) 13 :e16837. doi: 10.1002/advs.202516837 41572435
PMC13042975

68Ano Bom APD Rangel LP Costa DCF de Oliveira GAP Sanches D Braga CA .
Mutant p53 Aggregates into Prion-like Amyloid Oligomers and Fibrils: Implications for Cancer . J Biol Chem . (2012 ) 287 :28152–62. doi: 10.1074/jbc.m112.340638 22715097
PMC3431633

69Palanikumar L Karpauskaite L Al-Sayegh M Chehade I Alam M Hassan S .
Protein mimetic amyloid inhibitor potently abrogates cancer-associated mutant p53 aggregation and restores tumor suppressor function . Nat Commun . (2021 ) 12 :3962 . doi: 10.1038/s41467-021-23985-1 34172723
PMC8233319

70Ghosh S Salot S Sengupta S Navalkar A Ghosh D Jacob R .
p53 amyloid formation leading to its loss of function: implications in cancer pathogenesis . Cell Death Differ . (2017 ) 24 :1784–98. doi: 10.1038/cdd.2017.105 28644435
PMC5596421

71Pedrote MM Motta MF Ferretti GDS Norberto DR Spohr TCLS Lima FRS .
Oncogenic gain of function in glioblastoma is linked to mutant p53 amyloid oligomers . iScience . (2020 ) 23 :100820. doi: 10.1016/j.isci.2020.100820 31981923
PMC6976948

72Petronilho EC de Andrade GC de Sousa GS Almeida FP Mota MF Gomes AVS .
Oncogenic p53 triggers amyloid aggregation of p63 and p73 liquid droplets . Commun Chem . (2024 ) 7 :207 . doi: 10.1038/s42004-024-01289-x 39284933
PMC11405828

73Wang G Fersht AR .
Propagation of aggregated p53: Cross-reaction and coaggregation vs. seeding . Proc Natl Acad Sci . (2015 ) 112 :2443–8. doi: 10.1073/pnas.1500262112 25675527
PMC4345553

74Soragni A Janzen DM Johnson LM Lindgren AG Thai-Quynh Nguyen A Tiourin E .
A Designed Inhibitor of p53 Aggregation Rescues p53 Tumor Suppression in Ovarian Carcinomas . Cancer Cell . (2016 ) 29 :90 –103 . doi: 10.1016/j.ccell.2015.12.002 26748848
PMC4733364

75Li X Tang X Zeng J Duan L Hou Z Li L .
Peripheral cancer attenuates amyloid pathology in Alzheimer's disease via cystatin-c activation of TREM2 . Cell . (2026 ) 189 :853–71. doi: 10.1016/j.cell.2025.12.020 41576952

76Paris D Townsend K Quadros A Humphrey J Sun J Brem S .
Inhibition of angiogenesis by Aβ Peptides . Angiogenesis . (2004 ) 7 :75 –85 . doi: 10.1023/b:agen.0000037335.17717.bf 15302999

77Paris D Patel N Ganey NJ Laporte V Quadros A Mullan MJ .
Anti-tumoral activity of a short decapeptide fragment of the alzheimer’s Aβ Peptide . Int J Pept Res Ther . (2010 ) 16 :23 –30 . doi: 10.1007/s10989-010-9198-8
PMC2867080 20473341

78Luo S Feng J Xiao L Guo L Deng L Du Z .
Targeting self-assembly peptide for inhibiting breast tumor progression and metastasis . Biomaterials . (2020 ) 249 :120055. doi: 10.1016/j.biomaterials.2020.120055 32315863

79Qin H Cui T Liu Z Zhou Y Niu J Ren J .
Engineering amyloid aggregation as a new way to eliminate cancer stem cells by the disruption of iron homeostasis . Nano Lett . (2021 ) 21 :7379–87. doi: 10.1021/acs.nanolett.1c02734 34436904

80Tang Y Zhang D Robinson S Zheng J .
Inhibition of pancreatic cancer cells by different amyloid proteins reveals an inverse relationship between neurodegenerative diseases and cancer . Adv Biol . (2023 ) 7 :2300070. doi: 10.1002/adbi.202300070 37080947

81Sharma K Rai P Maurya SK Tapadia MG .
Anti-diabetic drug pioglitazone reduces Islet amyloid aggregation overload in the Drosophila neuronal cells . Naunyn-Schmiedeberg’s Arch Pharmacol . (2025 ) 398 :6031–41. doi: 10.1007/s00210-024-03632-4 39636405

82Jaikaran ETAS Higham CE Serpell LC Zurdo J Gross M Clark A .
Identification of a novel human islet amyloid polypeptide [beta]-sheet domain and factors influencing fibrillogenesis . J Mol Biol . (2001 ) 308 :515–25. doi: 10.1006/jmbi.2001.4593 11327784

83Reches M Porat Y Gazit E .
Amyloid fibril formation by pentapeptide and tetrapeptide fragments of human calcitonin . J Biol Chem . (2002 ) 277 :35475–80. doi: 10.1074/jbc.m206039200 12095997

84Tjernberg LO Näslund J Lindqvist F Johansson J Karlström AR Thyberg J .
Arrest of beta-amyloid fibril formation by a pentapeptide ligand . J Biol Chem . (1996 ) 271 :8545–8. doi: 10.1074/jbc.271.15.8545 8621479

85van der Willik KD Ghanbari M Fani L Compter A Ruiter R Stricker BHC .
Higher plasma amyloid-β Levels are associated with a higher risk of cancer: A population-based prospective cohort study . Cancer Epidemiology Biomarkers Prev . (2020 ) 29 :1993 –2001 . doi: 10.1158/1055-9965.epi-20-0167 32727725

86Jin W-S Bu X-L Liu Y-H Shen L-L Zhuang Z-Q Jiao S-S .
Plasma amyloid-beta levels in patients with different types of cancer . Neurotoxic Res . (2017 ) 31 :283–8. doi: 10.1007/s12640-016-9682-9 27913965

87Utsugi T Schroit AJ Connor J Bucana CD Fidler IJ .
Elevated expression of phosphatidylserine in the outer membrane leaflet of human tumor cells and recognition by activated human blood monocytes . Cancer Res . (1991 ) 51 :3062–6.
2032247

88Gaspar D Veiga AS Castanho MARB .
From antimicrobial to anticancer peptides. A review . Front Microbiol . (2013 ) 4 :294. doi: 10.3389/fmicb.2013.00294 24101917
PMC3787199

89Kleffman K Levinson G Rose IVL Blumenberg LM Shadaloey SAA Dhabaria A .
Melanoma-secreted amyloid beta suppresses neuroinflammation and promotes brain metastasis . Cancer Discov . (2022 ) 12 :1314–35. doi: 10.1158/2159-8290.cd-21-1006 35262173
PMC9069488