The tumor was observed in the hilus of the right lung (Figure 4A). Microscopically, it was composed of solid growth of small carcinoma cells suggesting SCLC (Figure 4B). Immunohistochemically, the tumor cells tested positive for chromogranin A, synaptophysin, INSM1 (Figures 4C–E), and TTF-1. The rate of positive cell rate for Ki67 (Figure 4F) was 46%, all compatible with SCLC.

The brain weighed 880 g and showed symmetric atrophy. The arachnoid appeared mildly cloudy (Figure 5A). Bilateral hippocampi were mildly atrophic on the surface (Figure 5B). Histologically, the hippocampus (Figure 6) and amygdala were the most severely affected, with severe neuronal loss on H&E (Figure 6A) and astrogliosis evident by GFAP (Figure 6B), along with increased CD68-positive microglia (Figure 6C), MHC class I immunostaining (Figure 6D), and infiltration of CD8-positive cytotoxic T cells. CD8+ CTLs were found in the parenchyma and around pericapillary areas; CD3-positive T cells localized similarly to CD8+ cells. CD20-positive B cells were detected in pericapillary regions, and MHC class I antigens were positive in a subset of neuronal cells (Figure 6D). Double immunostaining with NeuN and immune markers supported direct neuron–CTL interactions, with CTLs attached to neurons in elongated or ameboid forms (Figure 7).

Semiquantitative evaluation of neuronal loss was first performed using H&E staining. The results showed that the left CA1 (Figure 8A), CA4 (Figure 8D), and right CA4 (Figure 8J) regions exhibited complete neuronal loss and gliosis. The left CA2 (Figure 8B), CA3 (Figure 8C), right CA1 (Figure 8G), and right CA3 (Figure 8I) regions demonstrated severe neuronal loss and gliosis, with only a few neurons preserved. The right CA2 (Figure 8H) displayed moderate neuronal loss and gliosis. Both the left and right dentate gyri showed dispersion with marked loss of granule cells and gliosis in the granular and molecular layers (Figures 8F, 8L). Neurons were well-preserved in the bilateral subiculum (Figures 8E, 8K).

Immunostaining for NeuN showed that only a small portion of neurons was preserved in CA1 of the anterior portion of the left hippocampus and in CA2 of the posterior portion of the left hippocampus. NeuN immunostaining was not observed in the right hippocampus, possibly due to delayed brain sectioning of the right hippocampus, although GFAP and CD68 immunostaining were successful in the same right hippocampus. The degree of neuronal loss was evaluated only by H&E staining. The degrees of astrogliosis (GFAP) and microgliosis (CD68) paralleled the neuronal loss estimated by H&E and NeuN staining. Finally, the degrees of damage to the hippocampus, as evaluated by neuronal loss (Figure 9) and gliosis, according to the international consensus of hippocampal sclerosis in temporal lobe epilepsy (30), are summarized in Table 1.

No obvious vascular pathology, including microinfarctions, was observed in the hippocampi. Furthermore, no ischemic red neurons were observed in the hippocampus or in other brain regions, suggesting that the severe neuronal loss was not primarily attributable to acute hypoxic injury secondary to status epilepticus. Immunohistochemically, staining for phosphorylated tau, phosphorylated TDP-43, α-synuclein, and amyloid-β was negative in the brain tissue.

As listed in Table 2, semiquantitative microscopic examination across the whole brain revealed that severe neuronal loss (score 2) was strictly confined to the limbic system (hippocampus and amygdala). In the neocortex (frontal and lateral lobes) and the putamen, mild neuronal loss (score 1) was observed, accompanied by mild T-cell infiltration (score 1). Conversely, several regions, including the occipital lobe, subthalamic nucleus, thalamus, dentate nucleus, pons, and medulla oblongata, exhibited mild T-cell infiltration (score 1) but showed no obvious neuronal loss (score 0). No neuronal loss or T-cell infiltration (score 0/0) was evident in the globus pallidus, substantia nigra, or cerebellar hemisphere. Furthermore, lymphocytes positive for CD3 (Figure 10B) and CD8 infiltrated the meninges (Figure 10A), spinal cord (Figure 10C), and sciatic nerve (Figure 10D). In the spinal cord, they were observed in the white matter and around perivascular areas. There was evidence of interaction between spinal neuronal cell bodies and CTLs. The psoas muscles showed mild atrophy with CD68-positive macrophages (Figures 10E, F); however, no lymphocytic infiltration was detected in this tissue.

Anti-SOX1 antibodies target Sry-like high mobility group box 1 (SOX1) proteins, key transcription factors involved in neural lineage commitment and differentiation (23, 24). In adults, SOX1 expression persists mainly in neural stem cell niches, such as the subventricular zone and dentate gyrus (25, 26). These antibodies are strongly associated with SCLC and paraneoplastic syndromes, most notably Lambert–Eaton myasthenic syndrome (LEMS), cerebellar degeneration, and limbic encephalitis (27–29). Our case aligns with this, presenting with SCLC and anti-SOX1 antibody-positive limbic encephalitis.

The pathological findings overwhelmingly support a T cell-mediated cytotoxic mechanism for neuronal damage. Encephalitides caused by antibodies against intracellular antigens (such as anti-Hu) typically involve neurodegeneration mediated by CTLs (22). Since SOX1 is an intracellular antigen, its associated encephalitis is expected to involve robust T-cell cytotoxicity. We demonstrated this by showing direct neuron–CTL interactions using CD8/NeuN double immunostaining, observing CD8+ CTLs attached directly to shrunken neurons. The CTLs also exhibited shape changes—becoming elongated or amoeboid—indicating active engagement. Importantly, residual neurons expressed MHC class I antigens on their surfaces, a necessary recognition step for CTLs, providing strong evidence that neuronal loss was mediated by CD8+ CTLs (32).

The pathological assessment confirmed HS, characterized by severe neuronal loss, gliosis, and granule cell dispersion (GCD), predominantly affecting hippocampal subfields CA1, CA3, CA4, and the dentate gyrus (33). Given the patient’s long-standing history of schizophrenia (over two decades), this prompts an analysis of the relationship between these conditions. Schizophrenia is widely associated with hippocampal pathology, including volume reduction and interneuron loss (34, 35). The resulting astrogliosis observed in our case could reflect chronic schizophrenia-related changes. However, several findings support anti-SOX1 encephalitis as the primary, acute cause of the severe neuronal destruction:

Detailed semiquantitative scoring of the hippocampal formation (Table 1), which employed criteria derived from the international consensus classification of hippocampal sclerosis in temporal lobe epilepsy (30), revealed a distinct and widespread pattern of neuronal vulnerability. Specifically, severe neuronal loss (score 2) was observed consistently across the CA1, CA3, and CA4 subfields in all four quadrants examined (left anterior, left posterior, right anterior, and right posterior). Notably, the CA2 subfield also showed significant involvement, registering moderate (score 1) to severe (score 2) neuronal loss. This degree of damage in CA2 deviates from the classic mesial temporal lobe epilepsy (mTLE)-related HS pattern, where CA2 is often relatively preserved (30). Furthermore, the dentate gyrus exhibited severe granule cell loss/dispersion (score 2) in the left anterior and posterior sections, and moderate dispersion (score 1) in the right sections. Crucially, this widespread pyramidal and granule cell pathology sharply contrasted with the complete preservation of the subiculum (score 0) across all assessed sections, indicating a non-classic pattern of injury that is spatially inconsistent with typical mTLE pathology (30).

The pathological process was not confined to the limbic system. Our brain-wide analysis (Table 2) demonstrates widespread CD3-positive T-cell infiltration across the CNS, including the putamen, thalamus, and neocortex. Crucially, we found involvement of the peripheral nervous system, with CTL infiltration observed in the sciatic nerve and meninges (Figure 10). This peripheral finding aligns with the known clinical associations between anti-SOX1 antibodies and peripheral syndromes, such as LEMS and sensory neuronopathy (28, 29, 36, 37). While myasthenic symptoms remain uncertain due to the patient’s long institutionalization, the presence of CTLs in the sciatic nerve strongly suggests a generalized, multifocal paraneoplastic process extending beyond the brain. Although CD20-positive B cells were detected in pericapillary regions, the central mechanism remains predominantly cytotoxic T cell-mediated.

There are some limitations in this report. First, no banked serum prevented cell-based assays (CBA) for neuronal surface antibodies (e.g., anti-GABABR, anti-AMPAR1/2) and anti-SOX1 serology; P/Q-type VGCC testing to exclude subclinical LEMS/ataxia was not possible. Second, dorsal root ganglia (DRG) tissue was unavailable for pathology, hindering confirmation of CD8/NeuN cytotoxicity to sensory neurons. Third, no contrast-enhanced brain MRI to exclude SCLC brain metastases. Fourth, as our facility is a community hospital without research funding, we were unable to procure certain pathological reagents. Therefore, overall data/autopsy sampling constraints limit interpretability of peripheral and immunopathogenic aspects. Had these results been included, the presentation would have been more compelling.