One persistent criticism is that basic research on sepsis has failed to yield translational benefits (Cavaillon et al., 2020). Rodent models are limited by their intrinsic resistance to endotoxins and robust recovery from infections (Warren et al., 2010). Moreover, genetic differences among inbred strains lead to variable immune responses (Sellers et al., 2012). While humanized mice have been proposed as a potential solution (Laudanski, 2021), their complexity and cost hinder widespread adoption.

Common sepsis models include bacterial infection, intraperitoneal injection of LPS (endotoxemia model), cecal ligation and puncture (CLP), and cecal slurry. A recently proposed alternative is the fecal suspension and intraperitoneal injection model (Tsuchida et al., 2022). Each has strengths and limitations (Cai et al., 2023). Endotoxemia models poorly replicate clinical sepsis, and LPS results vary depending on type and lot (Beyer et al., 2020). Even CLP, the so-called “gold standard,” is subject to variability based on cecum size, needle gauge, extruded fecal volume, and fat handling (Buras et al., 2005; Niiyama et al., 2016). As Gonnert et al. (2011) emphasized, standardized models using clinical severity scores are essential for reproducible original laboratory research.

In contrast to humans, encephalopathy in animals cannot be assessed using GCS or CAM-ICU. As a result, behavioral tests evaluating cognition—such as learning and memory—are employed. The hippocampus is considered the most vulnerable brain region in sepsis (Semmler et al., 2005), and hippocampus-dependent tasks are preferred. However, contextual fear conditioning and the Morris water maze require equivalent baseline locomotor activity between groups—a condition not met in sepsis models. Sepsis induces profound behavioral alterations, including sickness behavior during the acute phase, marked by reduced activity, food intake, weight loss, and sleep disturbance. This is followed by depression-like behavior, attributed to cytokine-induced activation of indoleamine 2,3-dioxygenase (IDO), leading to anhedonia and helplessness (Dantzer et al., 2008; Pereira De Souza Goldim et al., 2020). These overlapping behavioral states complicate data interpretation (Dantzer et al., 2008). Indeed, even endotoxemia models show that such behaviors persist for up to 14 days after intraperitoneal LPS injections (Sohroforouzani et al., 2022), corresponding to the clinical ICU phase, during which behavioral testing may not reliably reflect cognitive function. Furthermore, general limitations of behavioral tests include experimenter bias, stress induction, low ecological-validity, and circadian mismatches (Lang et al., 2023). When coupled with sepsis-specific behavioral abnormalities, these factors introduce substantial confounds.

Neuroinflammation resulting from sepsis leads to various forms of synaptic dysfunction, including neurotransmitter imbalance, synaptic deficiency, myelin damage, and inhibition of synaptic plasticity (Tang et al., 2022; Yang et al., 2025). Synaptic plasticity is broadly categorized into structural and functional types. Among the latter, long-term potentiation (LTP)—first described by Bliss and Collingridge (1993)—is regarded as a cellular substrate for learning and memory (Takeuchi et al., 2014).

Historically, SAE research relied heavily on behavioral assessments. However, due to limitations in such tests during the acute phase of sepsis, electrophysiological approaches have gained traction. In 2000, Vereker et al. (2000) demonstrated impaired LTP in the hippocampal Schaffer collateral (SC)–CA1 pathway 3 h after intraperitoneal LPS injection in rats. Imamura et al. (2011) later reported reduced LTP in the CA1 region 24 h post-CLP.

Table 1 summarizes key studies evaluating hippocampal LTP in rodent SAE models. Many studies report reduced LTP during the acute phase (3 h−14 days post-insult), highlighting its utility as a surrogate marker for SAE. However, outcomes vary depending on stimulation protocols and animal age. Chapman et al. (2010) found no reduction in early-phase LTP (E-LTP) following 100 Hz stimulation in either young (3 months) or aged (24 months) rats 4–5 days after intraperitoneal E. coli infection. Interestingly, only aged rats exhibited reduced late-phase LTP (L-LTP) in response to theta-burst stimulation (TBS) 8 days after E. coli infection (Tanaka et al., 2018), suggesting that the impairment is both age-specific and dependent on the stimulation protocol. Kakizaki et al. (2023) showed that immature rats (4–5 weeks old) subjected to CLP displayed no significant reduction in E-LTP at 8 or 16 h post-insult. These findings emphasize the need to account for stimulation protocol and animal age when interpreting LTP outcomes in SAE models. Although most studies summarized in Table 1 have been conducted exclusively in male rodents, recent evidence indicates that hippocampal synaptic plasticity exhibits sex-dependent mechanisms. For instance, N-methyl-D-aspartate (NMDA) receptor contributions to LTP and spatial memory differ between sexes even in juvenile animals (Narattil and Maroun, 2024), and estrogen receptor signaling critically regulates LTP in females (Wang et al., 2018). Future studies should address sex as a biological variable to enhance translational relevance.

In NMDA receptor-dependent LTP, as seen in hippocampal SC–CA1 and perforant path (PP)–dentate gyrus (DG) pathways, calcium influx via NMDA receptors—elicited by TBS or high-frequency stimulation (HFS)—activates calcium–calmodulin-dependent protein kinase II (CaMKII) and enhances alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) trafficking, thereby potentiating synaptic transmission (Nicoll and Roche, 2013).

A key mechanism by which SAE impairs LTP involves the activation of the NLRP3–IL-1β pathway in microglia. Although physiological levels of IL-1β are thought to be important for the maintenance of hippocampal LTP (Ross et al., 2003), its overexpression has long been known to disrupt hippocampal LTP (Bellinger et al., 1993), particularly the NMDA receptor-dependent form (Hoshino et al., 2017a). This impairment is mediated via activation of p38 mitogen-activated protein kinase (MAPK; Coogan et al., 1999). IL-1β also promotes the neuronal surface expression of GABA_A receptors—especially those containing the α5 subunit—leading to increased tonic inhibitory currents and further LTP impairment through p38 MAPK-dependent mechanisms (Serantes et al., 2006; Wang et al., 2012). Additionally, IL-1β suppresses NMDA-induced outward currents via p38 MAPK (Zhang et al., 2010), offering another potential mechanism of LTP disruption.

Furthermore, microglia-driven synaptic pruning, discussed earlier in the context of SAE pathophysiology, may also impair hippocampal plasticity, warranting further investigation (Yang et al., 2025).

Grünewald et al. (2024) reported significant downregulation of Arc/Arg3.1—a master regulator of synaptic plasticity (Shepherd and Bear, 2011)—in the hippocampus of mice subjected to peritoneal contamination and infection. Hippocampal overexpression of Arc with adeno-associated virus vectors restored impaired E-LTP. Similarly, environmental enrichment during the post-septic period enhanced hippocampal Arc expression via the brain-derived neurotrophic factor (BDNF)/TrkB pathway, rescuing LTP and spatial memory. These findings suggest that directly targeting impaired LTP could offer therapeutic benefits.

In clinical studies, cerebrospinal fluid (CSF) from infected patients with delirium showed reduced levels of proteins involved in synapse formation and function, compared to non-delirious patients (Peters Van Ton et al., 2020), lending further support to this hypothesis. Moreover, postoperative cognitive dysfunction—believed to share pathophysiological features with SAE—has been linked to impaired E-LTP to L-LTP transformation mediated by hippocampal BDNF downregulation (Liu et al., 2025), underscoring the growing relevance of LTP-focused interventions. To date, no clinical studies have directly examined the relationship between hippocampal LTP and standard cognitive assessments such as the Mini-Mental State Examination (MMSE) or Montreal Cognitive Assessment (MoCA) in patients with sepsis or SAE. However, indirect evidence from other neurological conditions supports a potential link between LTP-related plasticity and cognitive performance. For instance, in patients with multiple sclerosis, reduced transcranial magnetic stimulation (TMS)-induced cortical plasticity—considered a human proxy for LTP—was associated with both lower CSF amyloid-β_{1 − 42} levels and poorer attention and concentration scores on cognitive tests (Mori et al., 2011). Although these data reflect neocortical rather than hippocampal plasticity, they suggest that LTP-like measures may serve as functional indicators of cognitive integrity. Given the role of hippocampal LTP in memory formation, future studies are warranted to determine whether electrophysiological markers of hippocampal plasticity in animal models of SAE correspond to clinical memory assessments such as MMSE or MoCA in human patients. Establishing such links would enhance the translational value of LTP as a surrogate marker and may provide a measurable endpoint for therapeutic trials targeting cognitive dysfunction in SAE.