The concept of cognitive resilience (CR) refers to the phenomenon in which individuals show no cognitive impairment despite the presence of neurobiological hallmarks of Alzheimer’s disease (AD) (Negash et al., 2013). Indeed, an increasing number of studies have demonstrated that a significant proportion of subjects with cerebral amyloid deposits and tau pathology remained individuals living without dementia over the course of aging, thereby demonstrating CR to the pathology (Negash et al., 2013). The result is a high degree of inter-individual cognitive variability between AD patients. Such heterogeneity presents a major challenge for the development of effective therapeutic strategies (Perneczky et al., 2019; Stern et al., 2023). Studying CR in clinical practice, however, is constrained by several limitations, including limited access to post-mortem samples and time constraints imposed by the progressive nature of the disease, which complicate longitudinal follow-up. In this context, animal models represent an invaluable resource for advancing our understanding of CR and its underlying mechanisms.

In the context of AD, two main approaches are employed to identify resilient animals. The first approach utilizes the enriched environment (EE) to model CR. Indeed, a slowdown in the spread of the disease was observed in rodents placed in EE (Lazarov et al., 2005), associated with a maintenance of cognitive performances (Verret et al., 2013). EE promotes social interaction, cognitive stimulation, novelty and exercise that contribute to these beneficial effects. These behavioral observations are linked to enhanced hippocampal neurogenesis driven by EE (Salmin et al., 2017) and by the regulation of the balance between BDNF and its precursor, pro-BDNF. While BDNF promotes hippocampal plasticity and improves cognitive performance in AD (de Pins et al., 2019; Hsiao et al., 2014), pro-BDNF is associated with neurotoxic signaling in AD (Fleitas et al., 2018). The second approach involved identifying resilient animals within a given population. For instance, cohorts of Tg2576 mice, a well-established model of AD, can be used to study CR. By aging these animals, resilient and non-resilient subpopulations can be identified. While non-resilient aged Tg2576 mice exhibit impaired spatial memory performance in the Morris water maze (MWM) test as early as 16 months of age, resilient mice maintain performance comparable to that of age-matched control mice (Pérez-González et al., 2020, 2021), providing a suitable tool to investigate the underlying mechanisms of CR. Resilient Tg2576 mice showed increased levels of proteins involved in different processes such as synaptic plasticity and integrity or reduced neuroinflammation (Pérez-González et al., 2020, 2021). Besides, these mice showed preserved dendritic spine plasticity and synaptic machinery in the hippocampus, which would play a crucial role in the mechanisms of CR (Pérez-González et al., 2020, 2021). Other studies have used contextual fear conditioning to identify resilient 5xFAD mice, detected at 8 months of age. Researchers have demonstrated that resilient 5xFAD mice differentially express more than 100 hippocampal proteins compared to their non-resilient counterparts, including proteins involved in neuronal excitability and synaptic plasticity (Neuner et al., 2017).

Overall, the majority of preclinical studies on AD are based on genetic models that represent only 1% of all AD cases. These models exacerbate the amyloid pathology and do not fully reflect the complex pathophysiology of AD (Xu et al., 2023). In addition, the identification of resilient animals within such models requires several months and is currently based solely on prediction, since it is not possible to predict in advance the number of animals that will be resilient to a disease within a cohort. It may therefore be advantageous to use sporadic AD models to more closely reflect clinical reality, combined with a characterized model of CR in order to limit the drawbacks of these studies linked to the assumption of the appearance of CR.

One of the most commonly used sporadic models is the intracerebroventricular (ICV) administration of streptozotocin (STZ). Originally developed in Wistar (WIS) rats, a dose of 3 mg/kg of STZ has been shown to disrupt the brain glucose metabolism (Mayer et al., 1989). One month after STZ administration, hallmarks of AD, including β-amyloid accumulation, phosphorylated tau protein, and neuroinflammation were observed in both the hippocampus and cortical regions (Moreira-Silva et al., 2019). These cerebral changes were associated with significant behavioral impairments. Enhanced anxiety-like behaviour has been observed between 2 and 4 weeks post-injection, as assessed using the elevated plus maze test (Amani et al., 2019; Mehla et al., 2013). Furthermore, deficits in recognition memory (Moreira-Silva et al., 2018), spatial memory (Amani et al., 2019; Tiwari et al., 2021), as well as working memory performance have been reported using the novel object recognition test (NORT), the MWM, and spontaneous alternation in Y-maze test (Amani et al., 2019), respectively. Despite its relevance in mimicking AD-like features, this sporadic model has not yet been used to investigate CR.

Genetically derived from WIS rats, LOU/c/jall (LOU) rats have been described as a model of successful aging. Subjecting themselves spontaneously to caloric restriction, LOU rats exhibit low levels of body fat throughout life, which may explain their longer median lifespan than WIS rats (Alliot et al., 2002). Regarding behavior, reduced anxiety, increased activity, and preserved memory performance during aging have been observed in this strain (Alliot et al., 2002; Kollen et al., 2010; Ménard et al., 2014). Genetic analysis has identified a set of genes in the hippocampus and in the frontal cortex that may be associated with the preservation of the cognitive abilities during aging (Paban et al., 2013). These findings led us to hypothesize that the LOU rat may serve as a model of CR in the context of AD.

Therefore, the aim of our study was to model the heterogeneity of CR by combining a model of successful aging with a rat model of sporadic AD. Using a behavioural and biochemical approach, we hypothesized that LOU rats would demonstrate CR to the STZ treatment compared to WIS rats. Using proteomic analysis, we sought to identify proteins of interest potentially involved in this CR observed in the LOU strain. This approach may facilitate the elucidation of the neuroprotective mechanisms underlying the CR, thereby informing future therapeutic strategies.

By combining for the first time a model of sporadic AD with a rat model of successful aging, we demonstrated that the LOU strain exhibited CR in the context of streptozotocin-induced cognitive impairment. Indeed, we showed that the LOU strain maintained cognitive performance after ICV injection of STZ compared to WIS, in the presence of similar brain amyloidogenesis.

Locomotor activity and anxiety-like behavior were assessed by measuring the total distance traveled in the OF and the time spent in the center of the arena. Consistent with the literature using the same paradigm on the WIS strain (Amani et al., 2019; Andrade et al., 2023), STZ injection did not impair locomotor activity or anxiety like behavior. It is worth noting that an increase in anxiety-like behavior following STZ injection has been reported in some studies using the elevated plus maze test (Amani et al., 2019). However, we did not observe such effect in our experiments (Supplementary Figure 1 and Supplementary Data 2). In LOU rats, we observed no changes in locomotor activity or anxiety-like behavior following STZ injection. Regardless of the treatment, LOU rats spent more time in the center of the OF arena than WIS rats, supporting the low anxiety phenotype previously described (Ménard et al., 2014). However, such strain-dependent variation in anxiety-like behavior is unlikely to affect memory performance assessments, as it was found to be independent of STZ treatment.

We observed significantly elevated beta-amyloid peptide concentrations in the hippocampus and cortex of WIS-treated animals, which is consistent with the literature on this model (Choi et al., 2014; Salkovic-Petrisic et al., 2011). Interestingly, STZ treatment induced a similar increase in brain beta-amyloid peptide concentrations in LOU rats, supporting our CR hypothesis in this strain (See also Supplementary Figure 2).

In order to identify resilient subjects, the majority of studies mainly use two behavioral tests which are the MWM and the contextual fear conditioning (Aguado et al., 2024; Neuner et al., 2017; Pérez-González et al., 2020, 2021). Unfortunately, the MWM protocol we used did not allow us to identify resilient animals among those treated with STZ, even if the STZ is known to induce spatial memory impairments in this task (Bavarsad et al., 2020; Bokare et al., 2018; Mehla et al., 2013).Moreover, our research has shown that STZ model was not referenced to induce deficits in the contextual fear conditioning (Moreira-Silva et al., 2018; Qi et al., 2021; Wei et al., 2022). Consequently, we chose to use two other less aversive experimental protocols in which STZ injection was known to induce deficits. Spatial working memory and recognition memory were assessed through the measurement of spontaneous alternation (Y-Maze) and the spontaneous objects exploration (NORT), respectively. For both tasks, our results indicated that the aCSF groups exhibited intact memory performance, as their percentage of alternation and novel object discrimination were significantly higher than the chance level (Grayson et al., 2015; Paban et al., 2013). As anticipated in this AD model (Amani et al., 2019; Moreira-Silva et al., 2019), STZ treatment reduced both working and recognition memory performances in both strains. Nevertheless, the percentage of alternation in the LOU-STZ group remained above the chance level, suggesting some preservation of memory, compared to WIS rats. Taken together, our results from the Y-maze and the NORT suggest that LOU rats displayed better memory preservation than WIS rats following STZ treatment.

Our results therefore showed that cognitive performances were maintained in LOU-treated rat in the presence of neuropathological hallmarks similar to those found in the WIS-treated rat confirming our initial hypothesis that the LOU strain could be a CR strain in the context of cognitive impairment induced by STZ.

One hypothesis we have considered to explain this CR involves the levels of hippocampal and cortical BDNF. This neurotrophine is known to promote immune processes, axonal growth, synaptic and dendritic plasticity, as well as learning and memory (Mu et al., 1999; Neumann et al., 1998). Moreover, one study showed an increase in hippocampal and cortical BDNF concentration in LOU rats compared to WIS, suggesting that this observation could be a hypothesis explaining why the LOU strain is considered as a successful aging strain (Silhol et al., 2008). Relevantly, authors highlighted that the difference between both strains was that during aging WIS strain developed more pro-BDNF form compared to LOU strain that is involved in neurotoxic pathways related to AD (Fleitas et al., 2018). Our findings contrast with the existing literature, showing an increase in BDNF concentration in STZ-treated rats. Indeed, a few studies reported a reduction in hippocampal and cortical BDNF concentrations in STZ-treated WIS rats (Bokare et al., 2018; Tiwari et al., 2021). However, the number of injections and the concentration of STZ injected in rat differed from our study. It would also be relevant in future studies to evaluate the levels of pro- and pre-pro BDNF isoforms in order to assess their distribution following STZ treatment and determine whether there is an increase in the pathogenic pro-BDNF isoform, which is known to be involved in neurotoxic pathways associated with AD. Moreover, unlike Silhol’s study, we did not find an increase in basal BDNF concentration in LOU strain compared to WIS, although levels in WIS rats were similar between our studies. However, some clinical studies have reported an increase in BDNF levels in the cortex and hippocampus of AD patients (Laske et al., 2006), as well as elevated plasma BDNF levels in the early stages of AD (Song et al., 2015). A similar increase in cerebral BDNF levels has been observed in the APP23 mouse model of AD, where BDNF was co-located with amyloid plaques (Burbach et al., 2004). These results suggested that in the context of progressive neurodegeneration, the brain may initiate compensatory functional changes and repair mechanisms, involving the upregulation of specific neurotrophic factors in vulnerable regions such as the hippocampus and cortex (Arendt et al., 1995). Unfortunately, we are not aware of any studies evaluating BDNF concentrations in STZ-injected rats under conditions strictly similar to our own.

In order to explore more hypothesis that could explain this CR observed in LOU rats, we then carried out a proteomic study to identify the different proteins potentially involved. We decided to focus our study on the proteins differentially regulated between the both strains and that have been linked to AD. Proteomic analysis revealed many mains findings. First, there were almost 4 times more differentially regulated proteins between the both strains in the hippocampal region (49 proteins) than in the cortical region (13 proteins). In addition, STZ injection mainly down regulated proteins in both strains. Indeed, the percentages of down-regulated proteins were around 50 and 56% in the cortex of WIS and LOU and around 64% and 71% in the hippocampus of the both strains respectively. Finally, LOU rats exhibited proteins that were differentially regulated compared to WIS rats, suggesting potential candidates underlying the observed CR in this strain.

Among the proteins listed, Ufsp2 protein was the only one to be differentially regulated between the both strains and the both structures (hippocampus and cerebral cortex) of STZ-treated animals. Ufsp2 protein, also known as UFM1-specific peptidase 2, is a cysteine protease involved in the UFMylation pathway. This pathway is crucial for various cellular processes including DNA damage response (Qin et al., 2019), immune response (Zhou et al., 2024) or brain development (Wang X. et al., 2023; Zhou et al., 2024). It has recently been shown that the concentration of soluble Ufsp2 specifically decreases in the temporal cortex of AD patients, while the concentration of insoluble Ufsp2 increases in temporal and frontal cortices (Yan et al., 2024).

Interestingly, our proteomic results, based mainly on the analysis of soluble forms, indicate an up-regulation of the Ufsp2 protein in the cortex and hippocampus of STZ-injected WIS rats, whereas a down-regulation was observed in STZ-injected LOU rats. Considering that some have shown ufsp2 KO neurons to be less sensitive to DNA damage in the context of AD (Yan et al., 2024), the down-regulation of ufsp2 in STZ-injected LOU may therefore reflect a protective factor against AD-like pathology.

Another protein of interest was the Usp11 protein. Located on the X chromosome, this protease involved in cell cycle regulation specifically cleaves ubiquitin, regulating the ubiquitination of associated proteins. Among these proteins, Usp11 is involved in the ubiquitination of Tau protein, inhibiting its degradation and exacerbating its accumulation in the formation of neurofibrillary tangles in the context of AD (Guo et al., 2024). Inhibition of Usp11 activity, whether genetic (Yan et al., 2022) or drug-induced (Guo et al., 2024), led to a decrease in pTau, Aβ₄₂ burden and an increase in synaptic plasticity and spatial memory in transgenic mouse models of AD, particularly in females. To our knowledge, Usp11 protein expression has never been studied in WIS strain, particularly in the context of AD. Our results, demonstrating an up-regulation in WIS-STZ rats, are in line with the literature showing an improvement of cognitive performances in AD models in which Usp11 protein has been inhibited. Interestingly, our results indicated a down-regulation of Usp11 protein in the hippocampus of STZ-injected LOU rats. Once again, these results are consistent with the maintenance of cognitive performance observed in STZ-injected LOU rats and suggest that this protein may be involved in the CR observed in this strain. With regard to TAU pathology, studies of STZ are more contradictory, as some have shown an increase in TAU genes expression in hippocampus of WIS-treated rats (Rostami et al., 2017) and other strains of rats (Lu et al., 2017; Tiwari et al., 2021), while others have shown a decrease (Moreira-Silva et al., 2018) or no modification of the expression of TAU and p-TAU in hippocampus and cortex (Motzko-Soares et al., 2018). The results of our proteomic analysis showed no variation in the regulation of the neurofibrillary tangle protein in our different groups (see all identified proteins in Supplementary Data 1).

Hemoglobin β (Hbb) undergoes dynamic alterations over the course of AD. In plasma, Ashraf et al. (2020) reported elevated Hbb levels in AD patients, associated with greater amyloid burden, hippocampal atrophy, and cognitive decline, suggesting its potential as a peripheral biomarker. In contrast, post-mortem analyses revealed marked neuronal depletion of hemoglobin subunits (α, β) in cortical and hippocampal neurons containing amyloid-β and tau aggregates, especially at advanced Braak stages (Ferrer et al., 2011). Mechanistic insight from APP/PS1 mice shows that Hbb expression rises with age and amyloid pathology, where it binds amyloid-β via its heme group and promotes plaque formation (Chuang et al., 2012). Collectively, these findings support a biphasic role for Hbb: early upregulation as a compensatory response to hypoxia and metabolic stress, followed by sequestration into amyloid aggregates and neuronal loss in advanced AD. Consistent with these observations, we found that Hbb was upregulated in the cortex of WIS-STZ rats but downregulated in LOU-STZ rats. The upregulation observed in WIS-STZ rats may represent a compensatory response during the early phase of β-amyloid accumulation induced by STZ administration, whereas the downregulation in LOU-STZ rats may reflect strain-specific mechanisms underlying their cognitive resilience.

In addition to the proteins mentioned above, the proteomic study showed a differential regulation of proteins involved in microtubule stability, such as the Map1a protein, which was down-regulated in WIS hippocampus and up-regulated in LOU hippocampus. Similarly, proteins involved in the stability of ribosomal and mitochondrial complexes showed opposite regulation between WIS and LOU strains, with Rpl39 and Dhodh being down-regulated, while Coq4 was up-regulated in WIS hippocampus. It is now clearly established that microtubule degradation occurs in the context of AD, and is partly mediated by Tau pathology (Cai et al., 2023). Other studies have highlighted the role of amyloid pathology in the degradation of the Map1a protein, which is thought to be an early event leading to synaptic dysfunction (Clemmensen et al., 2012; Fifre et al., 2006). In the same way, the dysregulation of both the ribosome and the mitochondrial complexes have been characterized in the context of STZ-injected WIS (da Silva et al., 2024). The overall down-regulation of these proteins in WIS rats following STZ treatment may thus underlie the memory deficits observed.

Taken together, the differential protein expression observed between WIS and LOU rats reinforces the interest in pursuing our efforts to determine whether these proteins are involved in the CR reported here in the LOU strain. In addition to focusing on proteins specifically associated with AD pathology, we have identified other pathways of interest, characterized by differently expressed proteins. This opens up new prospects for the characterisation of proteins involved in CR to AD.