Hippocampal Long Axis Differentiation in Memory and Seizure Networks

Molecular, Synaptic, and Connectivity Gradients

This basic tri-synaptic circuit is thought to be conserved across the curved longitudinal axis. ¹⁴ However, differences in intrinsic and extrinsic connectivity, molecular expression, and neuromodulatory receptor profiles have been described along the long axis.^1–4,15 In rodents, dorsal hippocampus predominantly interacts with visuospatial, and sensorimotor networks responsible for spatial memory and navigation, whereas ventral hippocampus is coupled to affective, autonomic, and associative networks supporting emotion, salience, and context.^2,3 Lesions in rat dorsal hippocampus produce spatial memory deficits, especially in detailed spatial representations. On the other hand, damage to ventral hippocampus inhibits fear-driven behaviors^16,17 (Figure 1, BOTTOM).

OPEN IN VIEWER

Figure 1.

Long-axis organization in human and rodent hippocampus. the hippocampus is schematized within the human and rodent brains and unfolded along its long axis. in each schematic, color gradients run from the septal/dorsal/posterior pole (blue) to the temporal/ventral/anterior pole (orange). Functional, physiological, and anatomical features (listed on the left) are shown as varying along the hippocampal long axis. Findings derived primarily from human studies are summarized in the TOP panel, rodent studies in the BOTTOM panel, and features reported across both species in the MIDDLE panel. Horizontal bars indicate the structure and extent of variation along the long axis; graded slopes denote continuous, monotonic changes, arrows indicate direction of movement, capped, solid bars indicate discrete domains, and dotted segments indicate unclear boundaries and/or intervals that have not been tested for a given feature. *Theta frequencies are highly modulated by specific task demands. ^{18 **} Functional connectivity and functional association with episodic recall shift from posterior to anterior hippocampus over development.^19,20 † In humans, the direction and magnitude of propagation may vary with behavior and cognitive demands. ²¹ ‡ Enriched expression of genes associated with cognition and affect has been shown in rodents and humans, however smooth gradients of their expression levels are present along the human long axis,^22,23 while discrete gene expression domains have been identified in rodents, including a distinct integrative domain in the intermediate hippocampus. ² .

Longitudinally organized projections from entorhinal, septal, thalamic, and hypothalamic sources, together with topographic outputs to retrosplenial and prefrontal cortex, as well as the amygdala, nucleus accumbens and other subcortical targets, establish distinct dorsal-to-ventral network affiliations within the hippocampus.^3,12 Gene-expression profiling reveals molecularly distinct domains along CA1's LA,^2,24 with graded differences in receptors and signaling pathways.^3,15,25,26 These receptors are implicated in long-term plasticity, particularly in fear-learning contexts. In general, synapses in dorsal hippocampus undergo greater potentiation than those in ventral hippocampus in response to the same stimulation, though this effect reverses following acute stress. ²⁷ Ventral principal cells also exhibit longer and more complex dendritic arborization. ²⁸ Neurogenesis, which is restricted to the hippocampal DG in adult mammals, ²⁹ occurs across the LA, typically at a higher rate in the dorsal than ventral regions. ³⁰ Axis-specific differences in neurogenesis have been associated with distinct behavioral tendencies ³¹ ; for example, increased neurogenesis in dorsal and ventral DG have been linked to continued use of outdated spatial rules in reversal tasks, and reduced exploration of novel environments, respectively.^32,33 Neurogenesis in ventral DG is also differentially inhibited by stress. ³⁴ At a histological level, dorsal hippocampal pyramidal CA1 cell layers are more tightly packed compared to the more dispersed ventral counterpart.^24,35,36 Although basic circuit architecture is conserved along the LA, numerous cellular, anatomical, and synaptic properties vary along this axis—some as smooth gradients and others as more discrete domains, motivating consideration of how hippocampal functions and dynamics are organized along the LA.

Representational Variation

Despite the surgical challenges of recording from rodent intermediate and ventral hippocampus, a few studies suggest a longitudinal gradient in how neural activity encodes spatial information and other variables. Compared with canonical place cells in dorsal hippocampus, intermediate CA1 contains fewer place cells, with broader place fields and reduced spatial selectivity^37–40—consistent with a coarser spatial code at the population level.^40,41 Intermediate CA1 neurons were shown to increase remapping in response to changes in reward location. ³⁸ In ventral hippocampus, representations are even coarser: in ventral CA3, one study has reported neurons with very large place fields on a long linear track, ³⁹ while another has demonstrated more abstract, spatially discontinuous representations on an 8-arm maze (eg, inward vs outward trajectories). ⁴² Beyond spatial coding, an odor-reward association task has shown that dorsal CA1 neurons encoded odor identities, while ventral CA1 neurons combined odor identity with predicted reward following learning. ⁴³ Imprecise definitions of “dorsal,” “intermediate,” and “ventral” hippocampus, subfield recording sites (CA1 vs CA3), and task designs preclude a unitary model of long axis organization. Nonetheless, there is general consensus that spatial coding becomes coarser from dorsal to ventral poles.^39,40 Yet, the nature of spatial representation change (eg, scale vs context), and whether these changes are gradual or abrupt (Figure 1, MIDDLE), remain unclear.

Spatiotemporal Organization in Oscillatory Dynamics

Hippocampal theta oscillations are low-frequency oscillations (8-10 Hz in rodents) seen during active behavior that increase in frequency and power with locomotion speed.^44–46 Theta oscillations phase-modulate gamma oscillations and unit spiking to organize item, sequence, and content associations. ⁴⁷ Theta waves travel from the dorsal to ventral pole with phase lags approaching a half-cycle (180 degrees) across the full length,^48,49 although an earlier report suggests a larger full cycle shift. ⁵⁰ The mechanism underlying this traveling wave is not fully understood. Theta generation depends on reciprocal connections between CA1 and the medial septum (cholinergic, glutamatergic and GABAergic septal projections to CA1, with GABAergic feedback ⁵¹ ), which are topographically organized along the LA. A phase delay across medial septal inputs could hypothetically drive theta propagation along LA. Additionally, local circuits and neuronal resonance properties at theta frequency may contribute to traveling waves. Likely, a combination of long-range drive and local resonance produces the observed propagation.^51,52 Theta power and phase coherence have also been observed to decrease from dorsal to ventral poles, although these features may vary by region or behavioral context.^3,49

Sharp-wave ripples (SPW-Rs) also demonstrate heterogeneous features (eg, slow-wave phase preference) along the LA.^53–57 SPW-Rs are high-frequency oscillations (100-200 Hz) seen during awake rest and NREM sleep. SPW-Rs appear in the local field potential as a negative sharp-wave in the CA1 stratum radiatum followed by a burst of high-frequency oscillations in the CA1 pyramidal layer. ⁵⁸ In rodents, place cell sequences acquired during exploratory behavior are replayed in SPW-Rs during quiet rest and NREM sleep.^59,60 SPW-Rs may be local or global; they may propagate toward either pole, split, or remain confined locally.^61,62 Together, these findings raise the question of whether the rodent hippocampus operates as a single computational module with graded parameters or as multiple, partially segregated modules that route distinct computations to different downstream targets.^3,4

Translational Challenges From Rodent to Human Brain

Species differences in neuroanatomy, behavior, and recording practices complicate translation. The lissencephalic rodent brain features relatively enlarged olfactory and sensorimotor cortices. The enlarged dorsal hippocampus is surgically accessible and thus heavily studied in tasks emphasizing locomotion and spatial navigation. ⁶³ By contrast, the human brain is over 1000 times larger than the rodent brain. The human cortex is expanded and folded, especially in association cortex (prefrontal, temporal, and parietal).^64,65 These anatomical differences and their inputs to entorhinal cortex and hippocampus likely explain why the anterior hippocampus is significantly enlarged compared to the posterior end and the rodent ventral axis. Enlarged frontal and temporal cortex and their involvement in over 90% of adult focal epilepsies also explain why anterior mesial temporal structures are frequently sampled in epilepsy surgery.^3,4,66 These divergent sampling strategies probably contribute to reported differences in theta-band properties, sharp-wave ripple prevalence and coupling, and place-like coding across species. Finally, a unique feature of the human brain is hemispheric specialization. Thus, human hippocampal LA functional differentiation may differ by laterality. This functional specialization is mainly driven by speech and language functions localizing to the dominant hemisphere (or the left hemisphere in 96% of right-handers ⁶⁷ ).

Theta Heterogeneity Across the Long Axis in Humans

In humans, the anterior hippocampus is expanded and preferentially connected with anterior temporal and frontal association cortices, while posterior hippocampus interfaces with occipital and parietal visuospatial systems—an organization reminiscent of a coarse-to-fine representational gradient with differential roles in gist versus detailed memory and navigation.^{4,5,72,73,78,100–104}

Human hippocampal theta has been debated because of observed variability in frequency, coherence, and task dependence across studies and, critically, across recording sites along the LA (Figure 1 TOP). Early intracranial work during virtual navigation demonstrated hippocampal theta near 8 Hz. ⁸ Subsequent direct recordings across a wide range of tasks and locations indicated that human hippocampal theta is often slower than in rodents, with prominent 2-4 Hz activity that can dissociate functionally and anatomically from faster 6-10 Hz components.^9,11,18,105 It is unclear whether slower and faster oscillations reflect true longitudinal gradients, as intracranial electroencephalogram sampling has primarily been in hippocampal head and body, under varying task conditions. However, there is some evidence that faster ambulation is associated with higher frequency theta activity (8 Hz), akin to rodent theta activity. ¹⁸

Real-world ambulatory recordings have further shown theta bouts elicited by movement and speed. ¹⁰ Similar to rodents, theta appears to be modulated at decision points ¹⁰⁶ and boundaries. ¹⁰⁷ Theta bouts appear to be modulated in real-world, observed, and imagined ambulation.^106,107 Consistent with rodents, human hippocampal theta can manifest as traveling waves with systematic phase gradients along the LA, although the direction and magnitude of propagation can vary with behavior and cognitive demands. ²¹ Reports of reduced coherence between hippocampus and surrounding medial temporal structures in some contexts may reflect heterogeneity in sampling along the LA, task-related state differences, and analytic choices.^11,105 Accumulating evidence thus supports the view that distinct slow and fast theta components differentially engage the hippocampus and map onto nonspatial/associative versus spatial/precise mnemonic operations, respectively.^4,18,101

Hippocampal Tail Onset as an Underrecognized Source of Seizures

Focal onset seizures arise primarily from the frontal (30%) or temporal (65%) lobes, ⁶ with a majority involving the anterior hippocampus. However, emerging evidence suggests that posterior hippocampal structures, including the hippocampal tail and the fasciola cinereum (FC), play an underrecognized role in seizure generation and propagation. ¹⁰⁸ In rodents, the FC, a small strip-like region at the dorsal/posterior extreme of the hippocampal formation, shows context- and space-related coding and can be recruited during interictal and ictal events; manipulating FC excitability with modern activity-integrator tagging and optogenetic tools can modulate seizure dynamics in experimental epilepsy^109–111 (Figure 1 MIDDLE). In clinical populations, case-based SEEG evidence indicates that posterior hippocampal contacts can show early ictal involvement in a subset of patients—including those with persistent seizures after anterior mesial temporal interventions—supporting the rationale for more systematic posterior sampling and, in select cases, tailored posterior ablations.^{66,108,112,113} Given the posterior hippocampus’ strong coupling to occipital and medial parietal networks in humans, seizures arising from or rapidly engaging the tail could present with atypical semiology and elude detection when posterior contacts are absent.^100,102 Systematic anatomical targeting that includes the hippocampal tail/FC in SEEG hypotheses—guided by imaging, semiology, and network models—may improve localization and outcomes in otherwise “nonlesional” or surgically refractory cases.