Between 2021 and 2023, patients diagnosed with psoriasis (based on medical history, physical examination and skin biopsy) were recruited from the Dermatology Department at University Hospitals in Rio de Janeiro, Brazil. All participants signed an Informed Consent Form (ICF) to confirm their willingness to participate. Inclusion criteria included age 18–70 years, sex, ethnicity, and disease duration of more than 10 years. Patients with psychiatric disorders, intellectual disabilities, or other visible skin diseases were excluded from the study. The Psoriasis Activity Score Index (PASI) was also completed during consultation.

All participants read and signed an informed consent form for inclusion in the study, which was approved by the Research Ethics Committee (CAAE: 41957320.3.0000.5259).

Peripheral blood samples were collected in 4 mL EDTA tubes (purple top) from all participants. The samples were transported to a thermal cooler using industrial ice packs.

Genomic DNA was extracted from blood samples using the Biopur Mini Spin Plus Kit (Biopur, Biometrix, Curitiba, Brazil) (Catalog: BP100-50, Batch: 32200739). DNA concentration was measured using a NanoDrop spectrophotometer.

The authors included previous HLA-typed with medium resolution SSOP HLA-A, -B, -C, -DRB1 and -DQB1 typed results.12 The allelic level (4-digit resolution) was imputed by selecting the most common allele for each MAC code in the Brazilian population.19 Samples of 85 patients included in the cohort after 2021 were HLA-typed (HLA-A, -B, -C, -DRB1, -DQB1 and -DPB1) using a commercial NGS-based HLA kit (Holotype HLA™ NGS Assay; Omixon Inc., Budapest, Hungary) on an Illumina MiSeq sequencer (Illumina, Inc., San Diego, USA).

Genotypic and allelic frequencies were assessed and compared with those in the general Brazilian population. Controls were randomly selected from the Brazilian Bone Marrow Donor Registry (REDOME) and matched to cases based on sex, self-reported color-race (ethnic descent), and geographic region at a ratio of 5:1. In total, 720 healthy controls were included.

HLA alleles within P or G groups were denoted by a lower case ‘g’.20

Allelic frequencies, as well as the Hardy-Weinberg balance test, and Class I haplotypes were determined based on conventional expectation maximization, analyzed for HLA class I and II results, and evaluated for their presence in psoriasis patients and controls using Arlequin 3.5.2.2.21 The HLA frequencies and HLA Class I A∼C∼B haplotypes, with counts greater than 4 were compared: a) between patients with psoriasis and the control group, b) PASI score, and c) psoriasis age of onset was determined using the EpiInfo software. The odds ratio (OR) and confidence interval (CI; 5%–95%) were calculated, and p (significance level Fisher test, < 0.05), or analysis of variance (ANOVA) was applied when appropriate, the pC (p correction - Bonferroni factor)

The loading plots were created on the platform MetaboAnalyst 5.0 Website using log10 normalized data, auto-scale features, Euclidean distance measure, and Ward clustering method, with the top 10 different alleles selected after the t-test results.22,23 The authors also sought to identify alleles among psoriatic patients with differences between mild vs. moderate to severe according to the PASI score (Psoriasis was defined as mild if PASI than 10 and moderate to severe if ≥ 10), and/or previous or current use of systemic treatment (classified as moderate to severe disease), and if the disease appeared before or after 30 years of age.

The study included 144 patients with psoriasis (84 men and 60 women) and 720 healthy individuals (420 men and 300 women). The mean age of the disease group was 29 ± 15.5 (5–68) years-old and 34% self-declared as branca (white-European descent), 13.9% as preta (black -sub-Saharan African descent), and 52.1% as parda (admixed) in both the case and control groups (Table 1).

Psoriasis HLA loci pairwise population was significantly different from the control (FST = 0.03485, p < 0.05). When considering the Hardy-Weinberg equilibrium of the studied population (Table S1), HLA-DQB1 was the only locus that was not in equilibrium. Linkage disequilibrium (LD) analysis among HLA loci (Table S2) demonstrated that, in the psoriasis group, most pairs of loci were in linkage disequilibrium, particularly across Class I regions (A∼C, A∼B, C∼B) and between Class I and Class II loci (A∼DRB1, B∼DRB1, C∼DRB1).

The alleles associated with disease protection (in crescent OR order) and risk were presented in Table 2. HLA-DRB1*15:03 g was found to be protective, and C*06:02 g is the strongest risk allele for psoriasis. Allele frequencies in controls and in psoriasis patients are detailed in Tables S3 and S4, for class I (HLA- A, -B and C) and class II (HLA-DRB1, DQB1 and DPB1) alleles, respectively.

The present matrix data included 144 patients and 720 controls with 160 variables: sex, PASI, Onset age, self-reported color, and HLA alleles with a frequency greater than 4 HLA-A (n = 29), -B (n = 54), -C (n = 23), -DRB1 (n = 35), and -DQB1 (n = 15). The alleles were scored as 1 (absent), 10 (heterozygosity), or 100 (homozygosity) and log transformed. Principal Component Analysis (PCA) was performed to explore the contribution of HLA alleles to the genetic differentiation between psoriasis patients and healthy controls. The loading plot (Fig. 1) demonstrates that most alleles are clustered near the origin, indicating low discriminative power across groups. In contrast, specific Class II alleles, particularly HLA-DRB1*07:01 g, HLA-DQB1*02:01g, and HLA-DQB1*03:01g, showed higher loading distances, suggesting a stronger influence in distinguishing psoriasis cases from controls.

Fig. 1.

Loading plot of Principal Component Analysis (PCA) applied to HLA alleles comparing psoriasis patients and healthy controls based on HLA frequencies. The loading plot displays the contribution of individual HLA alleles to the first two principal components. Most alleles cluster near the origin, indicating low discriminatory power between groups, while a few alleles ‒ such as HLA-DRB1*07:01 g, HLA-DQB1*02:01 g, and HLA-DQB1*03:01 g ‒ show greater loading distances, suggesting stronger influence in distinguishing psoriasis cases from controls.

Univariate analysis of the association between HLA alleles and psoriasis severity (Table 3) revealed that A*34:02 g (p = 0.010) and B*50:01 g (p = 0.011) were more frequently observed in patients with mild disease, suggesting a possible protective effect against moderate-to-severe forms. In contrast, the allele DRB1*01:01 g (p = 0.062) was identified exclusively in patients with moderate-to-severe psoriasis, representing a risk marker for greater clinical severity. Regarding age at onset, A*23:01 g (p = 0.006), A*30:01 g (p = 0.003), B*15:03 g (p = 0.078), and DQB1*03:03 (p = 0.026) were associated with late-onset psoriasis (>30 years), whereas B*44:03 g (p = 0.009) and C*07:02 g (p = 0.015) were more frequent in early-onset cases (<30 years), the latter being up to five times more common in this group, thus representing an important marker of risk for type 1 psoriasis (Table 3).

HLA-C*06:02, although strongly associated with psoriasis susceptibility, did not show significant differences between severity subgroups (Table 3), corroborating previous findings in the literature.

When the Principal Component Analysis (PCA) stratified by disease severity (PASI < 10 versus PASI ≥ 10) was made, it revealed that most HLA alleles clustered centrally, indicating minimal contribution to the differentiation between mild and moderate-to-severe psoriasis (Fig. 2). Nonetheless, HLA-DRB1*07:01g, HLA-DQB1*02:01g, and HLA-C*06:02g, showed greater loading distances, suggesting a potential role in modulating disease phenotype and inflammatory intensity.

Fig. 2.

Loading plot from Principal Component Analysis (PCA) comparing psoriasis patients with mild disease (PASI < 10) and those with moderate to severe disease (PASI ≥ 10). The PCA loading plot displays the contribution of HLA alleles to the genetic differentiation between psoriasis subgroups stratified by disease severity. Most alleles are concentrated near the center, indicating limited discriminative impact on clinical severity. However, HLA-C*06:02 g, HLA-DRB1*07:01 g, and HLA-DQB1*02:01 g appear further from the origin, reflecting a stronger contribution to variance and suggesting potential modulation of disease phenotype. The dispersion pattern supports a complex genetic architecture in which certain Class II alleles may influence not only susceptibility but also the clinical expression of psoriasis.

Among the 81 A∼C∼B haplotypes with count greater than 4, 02:01g∼06:02g∼13:02g (3.5% × 0.7% OR = 5.14 [1.47–18.91]), 02:01g∼06:02g∼57:01g (3.5% × 1.0; OR-3.66 [1.15–11.71]) and 26:01g∼12:03g∼38:01g (4.9% × 1.1% OR = 4,55 [1.62–12.75]) were more frequent among individuals with psoriasis.

HLA-A*34:02g and HLA-B*50:01g had lower frequencies in patients with moderate to severe PASI than in controls, whereas DRB1*01:01g was detected only in cases of moderate to severe PASI. The discrepancy between these results and those in the literature may be due to the smaller sample size of mild disease cases (17.4%) compared to moderate to severe cases (82.5%).

Another Brazilian group found an association of more severe disease in male patients with alleles B*37, C*06, C*12, and DRB1*07, whereas B*57 was associated with mild disease.14

A larger study in the Brazilian population with a larger sample size of mild disease cases is warranted for further research.

Nonetheless, on loading plot (Fig. 2), HLA-DRB1*07:01g, HLA-DQB1*02:01g, and HLA-C*06:02g, showed greater loading distances, suggesting a potential role in modulating disease phenotype and inflammatory intensity. These alleles, previously associated with susceptibility in the overall cohort, also appear to influence the genetic structure of patients with more severe clinical forms. The pattern supports the hypothesis that while several loci contribute to disease risk, specific Class I and II alleles may further shape the phenotypic spectrum and therapeutic responsiveness of psoriasis in the Brazilian population.

Univariate analysis was performed and showed that alleles associated with psoriasis onset after 30 years of age (type 2 psoriasis) were: A*23:01g (p = 0.006), A*30:01g (p = 0.003), B*15:03g (p = 0.078), and DQB1*03:03 (p = 0.026), in contrast, B*44:03g (p = 0.009) and C*07:02g (p = 0.015) were detected mainly before 30y and C*07:02g was 5 times more frequent in these young patients (14.5% × 3.3%).

Choonhakarn et al., in a Thai population, identified the alleles HLA-A*01, A*02:07, A*30, B*08, B*13, B*46:01, B*57, C*01, C*06:02 (the strongest association), and DRB1*07 as being associated with type I psoriasis (onset before 30 years of age), while higher frequencies of A*02:07, HLA-A*30, C*01 and DRB1*14:01 were significantly associated with psoriasis onset after 30 years of age.35 In a Turkish population, Atasoy et al. found that alleles B*57, HLA-Cw6, and DRB1*07 are significantly associated with type I psoriasis.36

Kim et al. studied the Korean population and found that the haplotype HLA-A*30-B*13-C*06:02-DRB1*07-DQA1*02-DQB1*02 is a high-risk factor for the disease, particularly at an early age in females.37 The haplotype HLA-A*33-B*44-C*14:01-DRB1*13-DQA1*01-DQB1*06-DPB1*04:01 was identified as a protective haplotype for psoriasis, whereas the extended haplotype HLA-A1-B37-Cw0602-DRB1*10-DQA1*01-DQB1*05 was found to be a high-risk factor for psoriasis in Koreans.37 The Pakistani population previously discussed had HLA alleles B*57, C*06:02, and DQB1*03:03:02 that were strongly associated with early-onset psoriasis, whereas alleles B*15, DRB1*13:02, and DQB1*03:03:02 were associated with late-onset psoriasis.24

This study had some limitations that should be acknowledged. The relatively small sample size, particularly within the mild disease subgroup, may have limited the statistical power to detect subtle genetic associations and affected the generalizability of the findings. Additionally, the cross-sectional design precludes longitudinal assessment of disease progression and long-term outcomes associated with specific HLA alleles. The reliance on the self-reported age of onset introduces potential recall bias, and the use of a single PASI assessment does not account for intra-individual variability over time. However, this limitation was mitigated by incorporating additional variables, such as previous or current use of systemic treatment, to classify patients with moderate to severe disease. Furthermore, given the high degree of admixture in the Brazilian population, undetected population substructures may have influenced allele frequency distributions, contributing to potential confounding factors. Despite these limitations, this study provides important and novel data regarding HLA associations in an admixed Brazilian cohort, a population that is underrepresented in global psoriasis research. The use of comprehensive statistical approaches, including univariate and principal component analyses, enhances the robustness of the findings and underscores the potential of certain HLA alleles to serve as biomarkers for disease susceptibility and severity in this unique population. Future larger, longitudinal studies are warranted to confirm these associations and further explore their clinical implications.

In conclusion, the findings of this study underscore the complexity and diversity of HLA allele associations in psoriasis across different populations. Although several alleles have been consistently linked to psoriasis in various populations, the present study adds to the growing body of evidence suggesting population-specific variations in genetic risk factors. The differences observed between the present results and those reported in other populations highlight the need for a more nuanced understanding of how genetic susceptibility varies across ethnicities.