Introduction

Paraneoplastic neurological syndromes (PNS) with Hu antibodies are strongly associated with small-cell lung cancer (SCLC), but show a heterogenous clinical presentation, ranging from isolated peripheral (e.g., sensory neuropathy) or central (e.g., limbic encephalitis) nervous system disorders, to, very often, multifocal involvement known as encephalomyelitis [1, 2]. The reason underlying this clinical diversity, and whether it reflects different pathophysiological pathways, is yet unknown. Moreover, why immune tolerance breaks down in SCLC and develops anti-Hu PNS is not understood. Indeed, SCLC constantly expresses Hu antigens, but only ~ 15% of patients harbor Hu antibodies and even fewer manifest with PNS [3]. Recent studies in other PNS have described particular genetic features of the associated tumors, but, conversely, no mutations in Hu genes have been identified in SCLC to date [4]. Investigation into whether the genetic characteristics of the patients themselves could play a role in the pathogenesis of anti-Hu PNS has so far been limited to a few small case series focusing on the human leukocyte antigen (HLA) [5, 6]. Herein, we conducted an HLA association analysis in a large cohort of anti-Hu PNS, exploring potential differences according to clinical presentation and cancer status.

Methods

Patients and clinical classification

From a total of 466 patients with anti-Hu PNS from the French Reference Center for PNS and Autoimmune Encephalitis (1990–2022), 100 (21%) had available DNA and were included in the study. The clinical picture was characterized according to: (1) general nervous system involvement (i.e., central nervous system, peripheral nervous system or combined) and (2) PNS phenotype (i.e., limbic encephalitis, brainstem/cerebellar, sensory neuropathy that included sensory neuronopathy and other less specific patterns after exclusion of alternative causes, motor neuropathy, Lambert–Eaton myasthenic syndrome or multifocal whenever two or more phenotypes existed). We coded patients as having "sensory neuropathy" following the approach of previous researchers who used this term for patients with a predominant large fiber sensory neuropathy, even with mild motor symptoms [1]. This decision stems from the difficulty of obtaining consistent electrophysiology evaluations in a retrospective cohort and evidence showing that most anti-Hu PNS patients with clinically pure sensory neuronopathy also had motor abnormalities [7]. The classification of cancer status was based on three categories: whether an established cancer diagnosis was obtained, or alternatively, if the length of follow-up exceeded or was less than 2 years.

HLA analysis

Patients and 508 healthy controls provided by the Stanford Center for Sleep Sciences and Medicine were genotyped using the Affymetrix PMRA array. Genotypes were processed using PLINK (version 2.0) as previously reported [8], and a principal component analysis (PCA), Euclidean distance-based measure was used to match patients to the closest controls (ratio 1:10, Supplementary Figure). HLA imputation was performed using HLA Genotype Imputation with Attribute Bagging [9]. HLA genotypes with an imputation probability lower than 0.3 were excluded. Allele carrier frequencies were first compared between patients (entire cohort and according to clinical presentation and cancer status) and PCA-matched controls using logistic regression controlled by the three main PCs. Secondly, a logistic regression controlled by the three main PCs and the significant alleles found in the first analysis was performed. Additional analyses comprised a zygosity analysis for the effect of DR3 ~ DQ2 haplotype dosage and a logistic regression in non-DR3 ~ DQ2 carriers. Multiple comparisons were corrected by Bonferroni’s method, and corrected p values < 0.05 were considered statistically significant. HLA analyses were performed using R Studio (version 2023.12.1 + 402).

Ethics approval

The study was approved by the Institutional Review Boards of Stanford University (IGNITE, IRB-65073) and Université Claude Bernard Lyon 1 and Hospices Civils de Lyon (ICARE-II, NCT04823728). Written informed consent was obtained from all participants for the storage and use of biological samples and clinical information for research purposes. The study was performed in accordance with the ethical standards framed by the Declaration of Helsinki and its later amendments.

Results

Demographic and clinical features

The main demographic and clinical features of the 100 patients with anti-Hu PNS are summarized in the Table 1. Notably, clinical presentation involved the central (28, 28%), peripheral nervous system (36, 36%) or both combined (36, 36%); the diversity and overlap of PNS phenotype are depicted in Fig. 1. As expected, among the 75 (75%) patients with a diagnosis of cancer, 62 (83%) were SCLC.

Table 1 Main demographic and clinical features of the cohort
Full size table
Fig. 1
figure 1

Upset plot of the isolated and combined phenotypes in 100 patients with anti-Hu PNS. Each row represents a clinical involvement, and the total number of patients with each of them are represented in the left horizontal bar chart. The columns represent the number of patients with isolated or combined clinical pictures. The cells are filled in black whenever a phenotype is present, and the dots are connected by a line when several phenotypes overlap

Full size image

HLA analysis in the whole cohort

Patients with anti-Hu PNS were significantly more frequently carriers of DPB1*01:01 (OR = 2.56 [1.40–4-70], corrected p value = 0.03), DQA1*05:01 (OR = 2.80 [1.74–4.49], corrected p value = 0.00019), DQB1*02:01 (OR = 2.88 [1.79–4.64], corrected p value = 0.00015, DRB1*03:01 (OR = 2.92 [1.80–4.73], corrected p value = 0.00031), DRB3*01:01 (OR = 2.20 [1.38–3.50], corrected p value = 0.02) and DRB4*01:01 (OR = 2.01 [1.16–3.48], corrected p value = 0.00031), in comparison to healthy controls (Fig. 2A, Supplementary Table 1). Noteworthy, DPB1*01:01 ~ DQA1*05:01 ~ DQB1*02:01 ~ DRB1*03:01 ~ DRB3*01:01 constitutes a common conserved haplotype (DR3 ~ DQ2) [10]. Additionally, protective effects were identified for DQB1*06:02, DRB4*01:03 and DRB5*01:01 (Fig. 2A, Supplementary Table 1).

Fig. 2
figure 2

HLA association study in the whole anti-Hu PNS cohort and according to clinical involvement. Forest plot depicting the significant HLA allele associations after performing logistic regression controlling for the three main principal components on: A the whole cohort of patients with PNS and Hu antibodies (n = 100), B patients with peripheral involvement (n = 36), patients with combined involvement (n = 36) and those with exclusively central involvement (n = 28)

Full size image

Logistic regression, including as covariates the significant alleles, showed that the DPB1*01:01 effect disappeared when controlling for the rest of the alleles of the DR3 ~ DQ2 haplotype, but was unable to determine whether a DR or DQ effect is responsible for the DR3 ~ DQ2 predisposition (Supplementary Table 2). In addition, the predisposing effect of DRB4*01:01, an allele carried by a subset of DRB1*07:01 ~ DQB1*02:02 haplotypes, remained after controlling for the alleles of the DR3 ~ DQ2 haplotype (Supplementary Table 3). Furthermore, no zygosity effect was observed for the DR3 ~ DQ2 haplotype (Supplementary Table 4). Logistic regression in non-DR3 ~ DQ2 carriers also showed a predisposing effect for DRB4*01:01, along with DRB3*02:02; DRB4*01:03 was still identified as a protective allele in this subgroup (Supplementary Table 5).

HLA analysis according to clinical presentation

The association with DQA1*05:01 ~ DQB1*02:01 ~ DRB1*03:01 ~ DRB3*01:01 was also observed when patients with anti-Hu PNS and peripheral involvement were analyzed separately (Fig. 2B, Supplementary Table 6). Conversely, among patients with combined involvement, only DRB3*01:01 remained statistically significant (Fig. 2C, Supplementary Table 7), whereas the association with the DR3 ~ DQ2 haplotype was absent in patients with central involvement (Fig. 2D, Supplementary Table 8). Since patients with peripheral involvement comprised cases with either sensory neuropathy, motor neuropathy or Lambert–Eaton myasthenic syndrome, we then analyzed only those presenting with sensory neuropathy, confirming an association with DQA1*05:01 ~ DQB1*02:01 ~ DRB1*03:01 (Fig. 3, Supplementary Table 9).

Fig. 3
figure 3

HLA-DR3 ~ DQ2 associates with sensory neuropathy in anti-Hu PNS. Forest plot depicting the significant HLA allele associations after performing logistic regression controlling for the three main principal components in patients with sensory neuropathy (n = 28)

Full size image

HLA analysis according to cancer status

Patients with anti-Hu PNS and a diagnosed cancer showed similar HLA associations than those observed when the entire cohort was analyzed, with predisposing effects for A*34:02, DPB1*01:01, DQA1*05:01, DQB1*02:01, DQB1*02:02, DRB1*03:01, DRB3*01:01 and DRB4*01:01, and a protective effect for DRB4*01:03 (Supplementary Table 10). The results were almost identical when only patients with SCLC were investigated, with a risk effect for A*34:02, DQA1*02:01, DQA1*05:01, DQB1*02:01, DQB1*02:02, DRB1*03:01, DRB3*01:01 and DRB4*01:01 (Supplementary Table 11). Logistic regression performed on the 12 patients with no cancer after > 2 years of follow-up did not identify any significant HLA association (data not shown), but it is noteworthy that only 2 (17%) were DR3 ~ DQ2 carriers.

Discussion

Herein, we confirm an association between anti-Hu PNS and the HLA haplotype DR3 ~ DQ2 previously reported in 53 patients [6], which was carried by nearly 40% of the patients of the present cohort of 100 individuals. This association is not related to the underlying cancer as SCLC does not show any HLA association [11]. Most interestingly, we also found that this association is more specific to patients with peripheral involvement, and, particularly, those manifesting with sensory neuropathy. Conversely, cases with exclusive central involvement lacked the DR3 ~ DQ2 association.

Along with the DR3 ~ DQ2 association, we also found a secondary association with DQA1*02:01 ~ DQB1*02:02 ~ DRB4*01:01, which was more evident when only patients with cancer or SCLC were analyzed. Remarkably, DRB4*01:01 was the single predisposing allele detected in patients with central involvement. Although the epitope reactivity of Hu antibodies has not been observed to vary according to the clinical presentation [12, 13], the different immunogenetic profiles exhibited herein by the clinical phenotypes likely reflect that the underlying pathophysiological mechanisms might be, at least partially, distinct. Similarly, patients without an identified cancer lacked the DR3 ~ DQ2 association, which could also suggest pathogenic differences compared to those with SCLC; however, it is noteworthy that the number of non-paraneoplastic cases analyzed was considerably small. Such clinical and oncological correlations with HLA have already been described in other PNS and related autoimmune encephalitis, like Lambert–Eaton myasthenic syndrome [14] or syndromes with contactin-associated protein-like 2 (CASPR2) antibodies [15]; nevertheless, in the aforementioned diseases, the non-paraneoplastic subtype is the one associated with HLA.

The haplotype DR3 ~ DQ2 (and the extended ancestral haplotype 8.1, which also includes HLA class I alleles HLA-A*01:01, B*08:01, C*07:01 and class II DRB3*01:01) has been associated with many autoimmune diseases in populations of European descent, such as type 1 diabetes mellitus, celiac disease and myasthenia gravis [16]. The lack of antigen specificity suggests that other mechanisms different from altered peptide presentation could also be involved in the predisposition to autoimmunity conferred by the haplotype 8.1, and, accordingly, several immune dysfunctions have been reported in its carriers [16]. Furthermore, the haplotype DR3 ~ DQ2 has been related to a few autoimmune encephalitis and PNS, principally to neurological syndromes with antibodies against glutamic acid decarboxylase 65 (GAD65) and limbic encephalitis with adenylate kinase 5 (AK5) antibodies [17, 18]. Interestingly, anti-Hu PNS, anti-GAD65 neurological syndromes and anti-AK5 limbic encephalitis seem to share a mostly CD8+ T cell-mediated pathogenesis [18,19,20]. It is therefore striking that no HLA class I association has been found for these diseases, although this could be due to the small sample size of the cohorts analyzed. In addition, CD4+ T cells might also be relevant to the immunopathogenesis of such disorders, as suggested by some neuropathological studies that showed important CD4+ T cell infiltrates accompanying CD8+ T cells [18, 21]. Remarkably, peripheral blood mononuclear cells (PBMCs) from patients with anti-Hu PNS stimulated with recombinant HuD protein (the main Hu antigen) exhibited an intense proliferation of CD4+ T cells, but not of CD8+ T cells [22]. These findings suggest that the activation of auto-reactive anti-Hu CD4+ T cells might be a necessary and early step in the pathogenesis of anti-Hu PNS, whereas the contribution of cytotoxic CD8+ T cells might occur later in the disease [22].

The main limitation of our study is the relatively small sample size, particularly when considering subgroups according to clinical involvement and phenotype. This point might also have hindered the identification of HLA associations in the non-paraneoplastic subset of patients.

In conclusion, we confirm an association with HLA-DR3 ~ DQ2 and show that it preferentially associates with paraneoplastic sensory neuropathy and Hu antibodies, suggesting pathophysiological heterogeneity. Larger studies are warranted to better define the HLA association in anti-Hu PNS, as well as to explore the role of other non-HLA genes.