Regional brain iron deposition and functional alterations in thalassemia: a multimodal human study with murine validation

Abstract

Background and objectives

Chronic anemia and systemic iron overload in thalassemia major (TM) patients may result in neurological deficits, although researchers have not completely elucidated their specific impacts on the brain. This study employed a multimodal research methodology to investigate the characteristics and intrinsic connections of cerebral iron deposition, cerebral dysfunction, and cognitive impairment in TM patients, providing theoretical support for brain monitoring and neuroprotective interventions in this condition.

Methods

We prospectively enrolled 88 patients with TM, including 46 transfusion-dependent thalassemia (TDT) and 42 non-transfusion-dependent thalassemia (NTDT) patients, alongside 79 healthy controls (HCs), between September 2022 and September 2025. All participants underwent quantitative susceptibility mapping (QSM), resting-state functional MRI (rs-fMRI), and assessments using the Mini-Mental State Examination (MMSE) and the Montreal Cognitive Assessment (MoCA). Serum biomarkers including S100 calcium-binding protein B (S100B) and neuron-specific enolase (NSE) were quantified. A parallel murine model of iron overload was created for validation purposes.

Results

QSM revealed localised iron deposition in the red nucleus, substantia nigra, subthalamic nucleus, and globus pallidus of patients with TM (F = 3.294–13.305, P = 0.001–0.040). The susceptibility values in these basal ganglia regions demonstrated a negative link with cognitive scores (r = − 0.27 to − 0.43, P < 0.05), suggesting that localised iron buildup correlates with cognitive decline.Rs-fMRI revealed significant irregularities in spontaneous brain activity across the frontal lobe-limbic system-cerebellar network, which correlated with deficits in attention, memory, and language functions as assessed by the cognitive evaluation scale. A murine iron overload model further demonstrated that cerebral iron deposition originates in the choroid plexus.Serum S100B and NSE concentrations increased progressively from HCs to NTDT and TDT groups (P < 0.05), reflecting gradual loss of astrocytic and neuronal integrity driven by worsening systemic iron overload.

Conclusions

TM patients, especially those with TDT, exhibit measurable cognitive impairment and cerebral functional remodeling, which are tightly associated with regional brain iron deposition. This study’s findings establish a theoretical foundation for the early detection of thalassemia-associated neurological problems and the formulation of focused therapeutic strategies.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s00277-026-07251-2.

Introduction

TM is an inherited hemolytic anemia characterized by chronic anemia and systemic iron overload [1, 2]. The accumulation of iron significantly compromises several essential organs, including the heart, liver, and endocrine glands [3–5]. Previous studies described systemic iron toxicity across multiple organs, yet iron-mediated cerebral damage remains understudied [6, 7]. Conventional structural MRI studies have detected subclinical white matter lesions and mild cerebral atrophy in neurologically asymptomatic patients [8, 9]. Nevertheless, these discoveries do not properly illuminate the cognitive and emotional difficulties that are being increasingly recorded in TM patients [10–12]. A thorough evaluation incorporating brain iron accumulation, functional connectivity, and domain-specific cognitive performance is currently absent [13, 14].

Earlier neuroimaging investigations of TM only evaluated isolated structural lesions or individual imaging biomarkers, lacking integrated anatomical, functional, and biochemical analyses [15, 16]. Rs-fMRI has emerged as a valuable tool for evaluating spontaneous brain activity and functional connectivity without requiring active task participation, making it highly advantageous in neuroscience research [14, 17, 18]. QSM—an advanced MRI technique that identifies magnetic susceptibility sources like iron—offers a unique opportunity to precisely delineate regional brain iron deposition [19, 20]. Serum biomarkers such as S100B (indicative of astroglial injury) and NSE (reflecting neuronal injury) have not been systematically integrated with neuroimaging to provide peripheral correlates of brain insult [21, 22].The utilization of these sophisticated methods in TM is still in its infancy.

We hypothesized that TM patients develop region-specific iron deposition within deep gray matter nuclei, which correlates with disrupted resting-state functional connectivity and cognitive dysfunction. To test this hypothesis, our study integrates QSM-based iron quantification, rs-fMRI functional connectivity analysis, neuropsychological screening (MMSE/MoCA), and serum S100B/NSE levels in human TM patients. Complementary murine models of graded cerebral iron overload, sampled across multiple time points, provide causal validation of the imaging and clinical findings. This multimodal, translational approach aims to identify early, subclinical neurological alterations attributable to iron-mediated toxicity in TM, thereby informing future monitoring or neuroprotective strategies (Fig. 1).

Fig. 1.

Technical route diagram

Materials and methods

Study participants

We prospectively recruited adult patients with genetically confirmed TM at the First Affiliated Hospital of Guangxi Medical University from September 2022 to September 2025; partial clinical variables were retrospectively retrieved from electronic medical records.TM diagnosis was validated via genetic sequencing identifying homozygous or compound heterozygous mutations in the α- or β-globin genes.Eligible patients were aged ≥ 16 years and had not undergone hematopoietic stem cell transplantation (HSCT) or gene therapy.Age- and education-matched HCs were randomly recruited from local community volunteers with no thalassemia mutations, hematological diseases, or chronic systemic illnesses.All participants were right-handed (Edinburgh Handedness Inventory) and excluded if they had a history of neurological/psychiatric disease, substance use disorders, or MRI contraindications (e.g., claustrophobia, implanted cardiac pacemakers).Subjects exhibited normal motor, cognitive, and communicative function sufficient to complete neuropsychological testing and MRI. Patients were classified as TDT or NTDT according to the Chinese Expert Consensus Guidelines for Clinical Blood Transfusion in patients with TDT.

This study was approved by the Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (Approval No. 2025-E0221). Written informed consent was obtained from all participants in accordance with the Declaration of Helsinki.

Exclusion criteria

Individuals failing to meet all inclusion criteria or presenting with severe end-organ dysfunction (hepatic, renal, or heart failure) were excluded from subsequent analyses.

Clinical data collection

Comprehensive demographic and clinical information was recorded, including age, sex, educational level, genotype, transfusion history, iron chelation regimen, serum ferritin (SF), hemoglobin, liver and renal function, hepatosplenomegaly, splenectomy status, liver iron concentration (LIC), cardiac T2* MRI, and the presence of extramedullary hematopoiesis. SF concentrations were measured in blood samples collected on the day of MRI acquisition. All patients underwent MRI no less than 7 days after their most recent transfusion to eliminate acute post-transfusion fluctuations in hemodynamics and iron biomarkers.For chelation therapy, the last dose of any oral or parenteral iron chelator was taken no less than 24 h before the MRI session, and the exact time of the last dose was recorded for each participant.

Neuroimaging acquisition and processing

All participants underwent brain MRI and QSM using a 3.0T Prisma scanner (Siemens AG, Germany) fitted with a 64-channel phased-array head coil.Standardized imaging protocols were applied.

Structural and functional imaging

High-resolution 3D T1-weighted images and rs-fMRI data were acquired under resting conditions. Image quality was independently reviewed by two senior neuroradiologists blinded to group status. Imaging parameters are summarized in Supplementary Table 1.

Preprocessing and analysis

Functional data were preprocessed using MATLAB 2017b with SPM12 and RESTplus v1.27 toolboxes. Standard steps included motion correction, spatial normalization, and smoothing. Strict quality control was applied: participants exhibiting head motion exceeding 2.0 mm maximum translation or 2.0°maximum rotation, or with a mean framewise displacement (FD) > 0.5 mm, were excluded from further analysis. All functional images were visually inspected for artifacts, signal dropout, and coverage errors prior to preprocessing. Following quality assurance, amplitude of low-frequency fluctuations (ALFF), fractional ALFF (fALFF), regional homogeneity (ReHo), and degree centrality (DC) metrics were calculated. Between-group comparisons (HCs, NTDT, TDT) were performed using analysis of covariance (ANCOVA) controlling for age and sex, with Gaussian random field (GRF) correction at voxel P < 0.05 and cluster P < 0.05.

Quantitative susceptibility mapping (QSM)

Three-dimensional multi-echo gradient echo sequences were obtained for QSM reconstruction. Data were processed using MATLAB 2020b, SPM12, and the STI Suite 3.0 (Duke University). Regions of interest (ROIs) included the substantia nigra, red nucleus, thalamus, striatum, and globus pallidus. ROIs were manually segmented on normalized MNI-space images using MRIcro by two independent neuroradiologists masked to clinical group assignments.Inter-rater reliability was assessed using the intraclass correlation coefficient (ICC) based on the mean susceptibility values extracted from each ROI; the ICC values exceeded 0.90 across all ROIs, indicating excellent inter-rater reliability. For each ROI, mean susceptibility values were extracted from triplicate measurements by a radiologist to further ensure reproducibility.

Neurocognitive assessment

Neuropsychological testing included the Mini-Mental State Examination (MMSE), Montreal Cognitive Assessment (MoCA), Hamilton Anxiety Scale (HAMA), and Hamilton Depression Scale (HAMD). Assessments were administered by trained psychologists who were blinded to clinical data. Total MMSE scores < 27 or total MoCA scores < 26 indicated probable cognitive dysfunction.

Animal model of iron overload

Forty-eight healthy C57BL/6 mice (24 males, 24 females; 17–25 g) were obtained from Guangdong Vital River Laboratory Animal Technology Co. Ltd. Mice were randomly allocated to the normal control (NC) or iron-overload (IO) groups. The IO group received weekly intraperitoneal injection of iron dextran at low (0.5 g/kg), medium (0.75 g/kg), or high (1.0 g/kg) doses; the NC group received equivalent saline volumes. At weeks 8, 10, and 16, tissues from the liver, spleen, heart, lung, kidney, and brain were harvested for Prussian blue staining and total brain iron quantification. For each time point (8, 10, and 16 weeks), there were 4 mice per group (NC, low-dose IO, medium-dose IO, high-dose IO), resulting in a total of 48 mice. Animal procedures were approved by the Animal Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (Approval No. 2025-D0311). Investigators were blinded to group allocation during analysis.

Measurement of S100B and NSE

Blood samples were collected on the same day as the rs-fMRI scan. Serum concentrations of S100B and NSE were quantified using commercially available ELISA kits (Elascience, China) following the manufacturer’s protocols. All assays were performed in duplicate, and intra- and inter-assay coefficients of variation were maintained below 10%.Chronic hemolysis in TM patients may release erythrocyte-derived NSE and artificially raise serum NSE concentrations. All samples were visually screened for hemolysis, and overtly hemolyzed specimens were discarded prior to testing. Nonetheless, subclinical hemolysis may still influence NSE concentrations, and this should be considered when interpreting the biomarker data.

Statistical analysis

All analyses were performed using SPSS 26.0 and GraphPad Prism 9.0. Normality was assessed using the Shapiro-Wilk test; continuous variables were compared using the Student’s t-test, one-way ANOVA, and categorical variables using chi-square test. For inter-group comparisons, ANCOVA was generally employed; pairwise comparisons between groups utilized the Bonferroni correction; For the comparison of inter-group differences in rs-fMRI, cognitive scale, and ROI results, ANCOVA was applied to adjust for age and years of education as covariates. Exploratory correlation analyses between SF, QSM values, and cognitive subdomains were performed using Pearson coefficients. To control for false positives in these multiple, secondary correlations, Bonferroni correction was applied, with an adjusted P < 0.05 considered significant. For all comparisons, Cohen’s d (two-group comparisons) and partial η² (ANOVA/ANCOVA) were reported as effect sizes, alongside 95% confidence intervals (CIs). The primary endpoint was between-group differences (HCs vs. NTDT vs. TDT) in regional QSM susceptibility (deep gray matter ROIs); key secondary endpoints were cognitive scores (MMSE/MoCA) and rs-fMRI metrics (ALFF, fALFF, ReHo, DC); Exploratory endpoints included serum S100B/NSE and murine brain iron quantification. All tests were two-tailed, with P < 0.05 considered statistically significant.

This was an exploratory study without a formal a priori sample size calculation; sample size was based on feasibility and prior similar studies in thalassemia. Consequently, while sufficient for detecting moderate-to-large effects, our study may be underpowered for small effect sizes and subgroup analyses. All results should be interpreted as hypothesis-generating.

Results

Clinical and demographic characteristics

A total of 88 TM patients were included, comprising 42 NTDT and 46 TDT cases. The distribution of genotypes was comparable between groups (α-TM: 64–35%; β-TM: 36–65%). The mean ages were 32.7 ± 10.3 years in NTDT and 23.9 ± 8.5 years in TDT. Educational levels were similar across groups (mean ≈ 14.6 years). SF was significantly higher in TDT (median 2649.5 ng/mL) than in NTDT (median 873.5 ng/mL), while hemoglobin levels were lower (80.0 g/L vs. 85.6 g/L). The two cohorts also differed in hepatic and cardiac iron burden: 32% of TDT patients had LIC > 15 mg Fe/g dw and 22% had cardiac T2* < 10 ms, compared with 14% and 2%, respectively, in NTDT. Most TDT patients (96%) received regular chelation therapy versus 36% in NTDT. These baseline findings confirm the biological gradient between transfusion dependence and systemic iron overload (Table 1).

Table 1.

Clinical and Demographical Data of Thalassemia Patients

	NTDT group	TDT group
	(n = 42)	(n = 46)
Age (years)	32.7 ± 10.3	23.9 ± 8.5
Female(%)	20(48)	23(50)
Education (years)	14.2 ± 2.0	15.0 ± 1.9
α-TM(%)	27(64)	16(35)
β-TM(%)	15(36)	30(65)
SF (ng/ml)	873.5(1168.2)	2649.5(1998.8)
HGB (g/L)	85.6(20.1)	80.0(21.8)
WBC (×109/L)	7.3(4.1)	7.1(5.9)
PLT (×109/L)	298.4(392.6)	464.6(415.6)
Ret (×1012/L)	0.25(0.178)	0.21 (0.264)
Hepatomegaly (%)	29 (69)	38(83)
Splenomegaly(%)	27(64)	23(50)
Splenectomy (%)	13(31)	22(48)
Extramedullary hematopoiesis (%)	13(31)	14(30)
LIC (%)
LIC < 3	13(31)	10(22)
3 ≤ LIC < 7	17(40)	7 (15)
7 < LIC ≤ 15	6 (14)	14(30)
LIC > 15	6 (14)	15(32)
Cardiac MRI T2* (%)
T2* > 20	41(98)	26(57)
15 ≤ T2* < 20	0 (0)	5 (11)
10 ≤ T2* < 14	0 (0)	5 (11)
T2* < 10	1 (2)	10(22)
Iron chelation therapy (%)
Regular	15(36)	44(96)
Irregular	27(64)	2(4)

Age and education are represented as the mean ± SD, while SF, HGB, WBC, PLT, and Ret are expressed as the median (IQR); all other variables are indicated as n (%) percentages are rounded to the nearest digit and thus may not add up to 100.

TM, thalassaemia; NTDT, non-transfusion-dependent; TDT, transfusion-dependent; SF, serum ferritin; HGB, haemoglobin; WBC, white blood cells; PLT, platelets; RET, reticulocytes; LIC, liver iron concentration.

Cognitive function and affective status

Neurocognitive performance showed a graded impairment from HCs > NTDT > TDT The mean MMSE scores were 29.78 ± 0.52, 27.98 ± 2.18, and 26.91 ± 1.67 in HCs, NTDT, and TDT, respectively (P < 0.05). Subdomain analysis revealed deficits in orientation, immediate memory, attention/calculation, and recall in both TM subgroups relative to HCs, with TDT patients showing additional impairment in attention/calculation and recall compared with NTDT. The MoCA showed a similar trend (29.58 ± 1.13, 24.43 ± 3.52, and 24.76 ± 2.57 for HCs, NTDT, and TDT; P < 0.05), confirming broad cognitive involvement across visuospatial, executive, and language domains. Anxiety was more frequent in the TM groups (P = 0.01), whereas depression was rare, observed in only one NTDT participant. Overall, cognitive screening demonstrated consistent decline in attention and memory domains that paralleled rs-fMRI findings in frontal and temporal regions, with minimal mood confounding (Table 2).

Table 2.

Cognitive performances of thalassaemia patients and HCs

	HCs (n = 79)	NTDT (n = 42)	TDT (n = 46)	P*	P**	P***
Age (years)	29 ± 6.7	33 ± 10.3	24 ± 8.5	0.082	< 0.01	< 0.01
Education (years)	15 ± 1.9	14.2 ± 2.0	15.0 ± 1.9	1.00	1.00	1.00
SF (ng/ml)	/	873.5(1168.2)	2649.5(1998.8)	/	/	< 0.05
HGB (g/l)	/	85.5(20.1)	80.0(21.8)	/	/	< 0.05
MMSE	29.78 ± 0.52	27.98 ± 2.18	26.91 ± 1.67	< 0.01	< 0.01	< 0.01
Orientation	10 ± 0	9.67 ± 1.28	9.91 ± 0.29	< 0.05	1.000	0.244
Immediate Memory	2.97 ± 0.16	2.71 ± 0.51	2.83 ± 0.38	< 0.01	0.061	0.381
Attention and Calculation	4.85 ± 0.46	4.4 ± 1.08	3.37 ± 1.16	< 0.01	< 0.01	< 0.01
Recall	2.95 ± 0.27	2.4 ± 0.70	1.91 ± 0.81	< 0.01	< 0.01	< 0.01
Naming	2 ± 0	2 ± 0	2 ± 0	1.00	1.00	1.00
Repetition	1 ± 0	0.98 ± 0.15	1 ± 0	0.32	1.00	0.45
Reading	1 ± 0	0.95 ± 0.22	1 ± 0	0.066	1.00	0.12
Command	3 ± 0	2.98 ± 0.15	3 ± 0	0.324	1.00	0.45
Expression	1 ± 0	0.98 ± 0.15	1 ± 0	0.324	1.00	0.45
Drawing	1 ± 0	0.9 ± 0.30	0.89 ± 0.32	0.078	< 0.05	1.00
MoCA	29.58 ± 1.13	24.43 ± 3.52	24.76 ± 2.57	< 0.01	< 0.01	1.00
Visuospatial and Executive Function	4.99 ± 0.11	4.1 ± 1.34	4.02 ± 1.11	< 0.01	< 0.01	1.00
Naming	2.99 ± 0.11	2.81 ± 0.40	2.85 ± 0.36	< 0.05	< 0.05	1.00
Attention	5.99 ± 0.11	5.17 ± 0.96	5.2 ± 1.11	< 0.01	< 0.01	1.00
Language	3 ± 0	2.4 ± 0.80	2.89 ± 0.32	< 0.01	0.53	< 0.01
Abstraction	2 ± 0	1.74 ± 0.54	1.89 ± 0.32	< 0.01	0.201	0.076
Delayed Recall	4.66 ± 0.89	2.24 ± 1.28	1.91 ± 1.19	< 0.01	< 0.01	0.487
Orientation	6 ± 0	5.98 ± 0.15	6 ± 0	0.324	1.00	1.00
HAMA	0.91 ± 2.52	4.48 ± 6.11	2.87 ± 3.10	< 0.01	< 0.05	0.161
HAMD	0.46 ± 1.92	3.1 ± 4.456	1.74 ± 2.54	< 0.01	0.056	0.092
P* HCs vs. NTDT; P** HCs vs. TDT; P*** NTDT vs. TDT;

Age, Education, MMSE, MoCA, HAMA and HAMD are presented as mean ± SD. SF and HGB are presented as medians (IQR). Inter-group comparisons were performed using one-way ANOVA with Bonferroni post-hoc tests.

HCs, healthy controls; NTDT, non-transfusion-dependent; TDT, transfusion-dependent; SF, serum ferritin; HGB, haemoglobin; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment. HAMA, Hamilton Anxiety Scale; HAMD, Hamilton Depression Scale.

Quantitative susceptibility mapping findings

To eliminate the interference of confounding factors such as age and years of education on brain iron deposition indicators, this study employed ANCOVA to compare QSM brain iron levels in deep nuclei across the three subject groups, with multiple pairwise comparisons performed using the Bonferroni method. Multivariate overall test results demonstrated that, after adjustment for demographic confounding factors, there were statistically significant differences in overall brain iron deposition characteristics among the three groups (Wilks’ λ = 0.569, F = 4.102, P < 0.001). ANCOVA results indicated statistically significant differences in brain iron levels in the left and right red nuclei, left and right substantia nigra, right subthalamic nucleus, and medial part of the left globus pallidus across the groups (F = 3.294 ~ 13.305, P = 0.001 ~ 0.040); No statistically significant differences were observed in other brain regions (P > 0.05) (Supplementary Table 2). Pairwise comparisons based on adjusted marginal means revealed that brain iron levels in the right subthalamic nucleus were significantly higher in the NTDT and TDT groups compared to the HCs; Brain iron levels in the medial part of the left globus pallidus were highest in the TDT, followed by the NTDT, and lowest in the HCs, with all pairwise comparisons showing statistically significant differences. The covariate analysis indicated that age and years of education did not significantly interfere with iron deposition in most brain regions, demonstrating the effectiveness of the correction strategy employed in this study and confirming the robustness and reliability of the intergroup difference results (Table 3).

Table 3.

Comparison of susceptibility values in various brain regions between three groups

ROI	HCs (n = 79)	NTDT(n = 42)	TDT(n = 46)	P*	P**	P***
Left red nucleus	0.039 ± 0.002	0.045 ± 0.003	0.027 ± 0.003	0.39	< 0.05	< 0.01
Right red nucleus	0.055 ± 0.003	0.065 ± 0.004	0.047 ± 0.004	0.054	0.059	< 0.01
Left substantia nigra	0.088 ± 0.004	0.099 ± 0.005	0.076 ± 0.005	< 0.05	< 0.05	< 0.05
Right substantia nigra	0.101 ± 0.004	0.115 ± 0.005	0.086 ± 0.005	< 0.05	< 0.05	< 0.01
Left thalamic bottom nucleus	0.022 ± 0.005	0.034 ± 0.007	0.032 ± 0.006	0.269	0.798	1
Right thalamic bottom nucleus	0.043 ± 0.005	0.062 ± 0.007	0.058 ± 0.007	< 0.05	0.494	0.788
Left striatum	0.024 ± 0.001	0.025 ± 0.002	0.022 ± 0.002	1	0.16	0.213
Right striatum	0.025 ± 0.001	0.028 ± 0.002	0.023 ± 0.002	0.201	0.95	< 0.05
Left lateral globus pallidus	0.080 ± 0.003	0.080 ± 0.004	0.071 ± 0.004	1	0.16	0.213
Right lateral globus pallidus	0.096 ± 0.003	0.096 ± 0.005	0.087 ± 0.005	1	0.296	0.338
Left medial globus pallidus	0.016 ± 0.003	0.032 ± 0.004	0.042 ± 0.004	< 0.05	< 0.01	0.182
Right medial globus pallidus	0.060 ± 0.003	0.065 ± 0.005	0.069 ± 0.005	0.976	0.275	1

All numerical values represent the ANCOVA marginal means ± standard errors, adjusted for age and years of education; Pairwise comparisons among the three groups were performed using the Bonferroni multiple comparison correction.P* : HCs vs. NTDT; P**: HCs vs. TDT; P***: NTDT vs. TDT; Measurement data are described using mean ± SD; Inter-group comparisons were performed using one-way ANOVA with Bonferroni post-hoc test; Susceptibility values are in units of ppm; HCs: healthy controls; NTDT: Non-transfusion-dependent༛TDT༚transfusion dependent

Resting-state fMRI findings

Rs-fMRI revealed significant alterations in spontaneous neural activity across cognition-, motor-, and emotion-related brain regions among groups (GRF-corrected P < 0.05).

In ALFF analyses, the left middle frontal gyrus and cerebellum showed the most prominent group effects (F = 46.7). Compared with HCs, both TDT and NTDT patients exhibited increased ALFF in the left cerebellum (t = 10.3) and middle frontal gyrus (t = 9.60). Relative to NTDT, TDT patients displayed higher ALFF in the left cerebellum (t = 4.20) and right inferior temporal (t = 4.68) and right inferior frontal gyrus (t = 3.92), accompanied by reduced ALFF in the right precuneus (t = −5.30) and inferior parietal lobule (t = −3.87).

fALFF analysis showed that the TM groups had less activity in the left angular gyrus (t = −6.94) and left caudate nucleus (t = −6.40) compared to the HCs. fALFF levels were higher in the left caudate nucleus (t = 5.19), the left middle temporal gyrus (t = 3.77), the right superior occipital gyrus (t = 3.55), and the right inferior frontal gyrus (t = 4.02) in TDT compared to NTDT. Both groups of patients had higher degree centrality (DC) measurements in the left middle frontal (t = 8.83) and left inferior temporal gyrus (t = 6.79). TDT had higher DC in the left hippocampus (t = 5.79) and lower DC in the right precuneus (t = −5.78) compared to NTDT. In NTDT, ReHo analysis showed better coordination in the left cerebellum (t = 6.74) and right middle frontal gyrus (t = 6.23), but in TDT, ReHo dropped in the right supplementary motor area (SMA) (t = −9.03) compared to HCs (Fig. 2, Supplementary Table 3).

Fig. 2.

Schematic diagram of brain regions with differences between groups in various indicators of rf-MRI. A: Schematic diagrams of brain regions showing differences in ALFF between TDT vs HCs, NTDT vs HCs, and TDT vs NTDT groups;B: Schematic diagrams of brain regions showing differences in fALFF between TDT vs HCs, NTDT vs HCs, and TDT vs NTDT groups; C: Schematic diagrams of brain regions showing differences in DC between TDT vs HC, NTDT vs HCs, and TDT vs NTDT groups; D: Schematic diagrams of brain regions showing differences in ReHo between TDT vs HCs, NTDT vs HCs, and TDT vs NTDT groups. The color gradient indicates T-value distribution, with blue to red spectrum corresponding to increasing T-value trends. In the brain map, blue areas represent brain regions showing decreased intergroup T-values, while red areas indicate increased T-values

Together, these findings suggest distributed cortical-subcortical dysregulation in TM, particularly within the frontal-limbic and cerebellar networks, consistent with subtle cognitive-emotional dysfunction. The differences were statistically significant (as determined by GRF check, voxel P < 0.05, cluster P < 0.05).

Correlation examination

In all TM patients, MMSE and MoCA scores were negatively correlated with SF levels (r = − 0.236, 95% CI: −0.4243 to − 0.0278; P = 0.028) but not with hemoglobin levels (P > 0.05) (Fig. 3a, b). Subdomain analyses revealed negative correlations between SF and MMSE attention/calculation (r =–0.22 [95% CI:−0.411 ~ −0.011], P < 0.05) as well as recall (r =–0.23 [95% CI:−0.419 ~ −0.021], P < 0.05) (Fig. 3c, d). These correlations suggest that systemic iron overload contributes to mild cognitive decline, warranting further longitudinal validation. Significant negative correlations were observed between MMSE total scores and QSM values in the substantia nigra, red nucleus, and striatum (r ≈ − 0.27 to − 0.43; P < 0.05). Within cognitive subdomains, higher iron levels in the striatum were linked to immediate memory, command, expression, and drawing (Supplementary Table 4). Regional iron burden, particularly within basal ganglia structures, therefore aligns with the observed cognitive and rs-fMRI alterations, supporting a mechanistic link between iron deposition and neurofunctional decline.

Fig. 3.

Correlation Examination

We further conducted correlation analyses to quantify linear associations between QSM-derived cerebral iron measurements and systemic clinical indicators. Susceptibility values reflecting iron deposition in the right subthalamic nucleus demonstrated a significant positive correlation with serum ferritin (SF) (r = 0.293, P < 0.05). By contrast, iron burden within the medial segment of the left globus pallidus was inversely correlated with cardiac T2* relaxation time (r = − 0.376, P < 0.01). Additional significant correlations were identified between reticulocyte levels and striatal susceptibility, as well as between platelet counts and iron accumulation in the left globus pallidus. No statistically significant linear relationships were observed between hemoglobin concentrations and regional QSM susceptibility across all examined subcortical nuclei. The inverse correlation between cardiac T2* and pallidal iron load corroborates the existence of systemic–central iron crosstalk, implying convergent iron deposition mechanisms under sustained severe systemic iron overload (Supplementary Table 5).

Supportive findings

Animal model of iron overload

In the murine iron-overload model, systemic iron deposition was evident in visceral organs by week 8, while brain iron accumulation remained limited to the choroid plexus. By week 16, progressive iron deposition correlated with dosage, with the high-dose group showing markedly elevated total brain iron concentration (230.4 ± 112.3 µmol/L) compared with all other groups (P < 0.01). Prussian blue staining confirmed a dose- and time-dependent increase in choroid plexus iron, whereas no significant parenchymal deposition was detected microscopically. These results corroborate human QSM findings, suggesting that cerebral iron accumulation begins at barrier interfaces before potential parenchymal involvement (Fig. 4).

Fig. 4.

Results of Prussian blue staining of mouse brains in each experimental group at 10w and 16W. A: time at 10W; B: time at 16W; NC: Normal control; L-IO; Low iron overload; M-IO; Middle iron overload; H-IO High iron. overload

Serum S100B and NSE levels

Serum S100B and NSE exhibited a stepwise elevation from HCs to NTDT to TDT. Mean serum S100B concentrations were 67.4 ± 5.8 pg/mL (HCs), 114.9 ± 5.8 pg/mL (NTDT), and 129.2 ± 22.8 pg/mL (TDT; one-way ANOVA, P = 0.032), while NSE levels were 6.64 ± 2.91, 8.51 ± 5.76, and 10.29 ± 8.12 ng/mL (P = 0.030). The increase was most pronounced in TDT versus HCs (P < 0.05), indicating progressive astrocytic and neuronal injury corresponding to transfusion dependence and systemic iron load (Fig. 5).

Fig. 5.

Schematic comparison of expression levels of S100B and NSE in the three groups *P <0.05

Discussion

This study utilizing multimodal technologies (rs-fMRI, QSM, neurocognitive tests, serum biomarkers, and animal models), systematically elucidated for the first time the functional network remodeling and cognitive impairment associated with brain iron deposition in TM patients. TDT exhibited more severe impairment compared to NTDT, providing novel multidimensional evidence for the pathogenesis and clinical management of subclinical neurological damage in TM patients. The core findings can be summarized into three key aspects: abnormal resting-state brain function (particularly in the left middle frontal gyrus and cerebellum), regional increases in brain iron content (red nucleus, substantia nigra, right subthalamic nucleus, and left medial globus pallidus), and stepwise elevations in serum S100B and NSE levels. These findings collectively indicate that iron homeostasis disruption and chronic anemia synergistically drive subclinical neurological damage, which is more pronounced in TDT patients [23–26].

The QSM findings in this study demonstrated significantly elevated magnetic susceptibility in the red nucleus, substantia nigra, right subthalamic nucleus, and left medial globus pallidus—regions that serve as critical nodes for motor-cognitive function integration. This regional iron deposition pattern aligns with prior observations by Long et al. and Akhlaghpoor et al. [27, 28], who observed reduced T2 values in the basal ganglia, thalamus, and red nucleus in patients with β-thalassemia, confirming the universality and specificity of cerebral iron deposition in TM patients. The study revealed that iron deposition was not uniformly distributed throughout the brain but predominantly affected the red nucleus, substantia nigra, globus pallidus, and subthalamic nucleus. This selective distribution pattern is likely closely associated with their inherent high blood flow, high metabolic activity, and distinct patterns of local iron transporter expression, providing crucial insights into the targeted mechanisms underlying cerebral iron deposition in TM patients [19, 26, 29, 30].

Rs-fMRI findings revealed significantly elevated ALFF values in the left middle frontal gyrus and cerebellum, decreased fALFF values in the left angular gyrus and caudate nucleus, and increased ReHo values in the right middle frontal gyrus and cerebellum. This “local inhibition with distal compensation” functional network pattern suggests that patients with TM are on the verge of neural network decompensation, rendering them incapable of maintaining normal neurofunctional homeostasis [17, 31–33].

Correspondingly, cognitive test results demonstrated that TM patients scored significantly lower on the MMSE and MoCA compared to healthy individuals, with the most pronounced impairments in the attention, calculation, and delayed recall domains. The degree of impairment exhibited a dose-response relationship where TDT was more severe than NTDT, consistent with the severity of cerebral iron deposition [34, 35]. The impairments in attention and computation (MMSE: TDT 3.37 ± 1.16, HCs 4.85 ± 0.46, P < 0.01) were chiefly linked to functional irregularities in the frontal-cerebellar pathway, as demonstrated by elevated ALFF and ReHo in the left middle frontal gyrus and cerebellum. Deficiencies in delayed recall and immediate memory (MoCA: TDT 1.91 ± 1.19, HCs 4.66 ± 0.89, P < 0.01) were linked to disruptions in the default mode network (DMN): TDT patients exhibited increased DC values in the hippocampus and reduced ReHo in the left cingulate gyrus—both essential components of the DMN—suggesting that impaired DMN integrity may play a role in the observed memory decline. Furthermore, declines in visuospatial and executive capabilities (MoCA: TDT 4.02 ± 1.11, HCs 4.99 ± 0.11, P < 0.01) correlated with reduced ALFF in the posterior parietal and precuneus areas, which are critical components of both the DMN and the dorsal attention network.

The association between cerebral iron deposition and abnormalities in functional brain imaging, coupled with cognitive decline, further confirms the presence of brain damage in patients with TM. These associations collectively indicate that cerebral iron accumulation does not cause uniform global impairment but instead selectively impacts particular iron-sensitive nuclei (such as the globus pallidus, red nucleus, and substantia nigra) and their associated cortico-subcortical pathways, thereby specifically hindering attention, memory, and executive functions.These results offer imaging-derived proof to inform forthcoming targeted neuropsychological evaluations utilizing more sophisticated test batteries for certain cognitive areas.It is crucial to recognize that although the MMSE and MoCA are useful instruments for cognitive evaluation and can effectively indicate general cognitive deficits, they fall short in accurately assessing advanced brain functioning [34, 36–39]. Future research should incorporate additional assessments to refine the cognitive impairment evaluation framework.

Correlation analyses further revealed a negative correlation between SF levels and the attention and recall subscales, whereas hemoglobin concentration showed no significant association. These findings suggest that hypoxia is not the sole contributor to cognitive impairment in TM patients, and neurotoxic effects induced by iron overload may play a more critical driving role [39–41].

To further elucidate the spatiotemporal evolution patterns and underlying mechanisms of brain iron deposition, this study provides crucial biological evidence through an animal model. In the mouse iron overload model, iron deposition first emerged at the choroid plexus—a critical interface of the blood-brain barrier (BBB)—and progressively worsened with increasing systemic iron load [42, 43]. Although no definitive intraparenchymal iron deposition was histologically observed at 16 weeks, total brain iron concentration had already risen significantly. This evolutionary pattern suggests that under chronic iron overload, the BBB initially restricts but cannot completely prevent iron entry into the brain [44, 45]. Iron likely accumulates first at the barrier interface (choroid plexus) and subsequently diffuses progressively into the parenchyma as exposure duration extends [46]. These animal experimental findings provide biological support for iron entry into the central nervous system and its delayed accumulation, implying that brain iron deposition in TM patients may follow a “barrier interface accumulation → deep nucleus progression” pathway, worsening progressively with increasing systemic iron load [30, 42, 47]. Additionally, the insensitivity of Prussian blue staining to low-level parenchymal iron further highlights the unique advantages of advanced imaging techniques such as QSM in early detection of brain iron deposition, offering reliable technical support for clinical identification of subclinical neurological damage [48–50].

Serum S100B and NSE serve as auxiliary biomarkers for neural injury, further strengthening the evidence chain regarding iron deposition-related neurological damage in the brain and reflecting the progressive astrocytic and neuronal injury observed in TM patients [22, 51–53]. Detection results demonstrate a stepwise increase in their levels from healthy controls to NTDT and then to TDT: elevated S100B indicates BBB dysfunction and astrocytic activation, while elevated NSE reflects compromised neuronal membrane integrity, showing a synergistic correlation with the severity of cerebral dysfunction and iron deposition [54]. It is important to emphasize that S100B and NSE remain investigational biomarkers whose levels are influenced by multiple factors; thus, they cannot serve as direct clinical replacement indicators. Their interpretation should be integrated with imaging studies and cognitive assessments to avoid misinterpretation based on a single marker.

Despite the significant findings of this study, several limitations remain that require addressing in future research. First, this single-center study with a moderate sample size limits the generalizability of our findings. Subsequent multicenter, large-sample studies are needed to validate the conclusions. Second, as this was a cross-sectional study lacking longitudinal follow-up data, the causal relationship between iron load threshold and cognitive decline could not be established; long-term follow-ups are necessary to clarify their dynamic association. Third, cognitive assessment relied solely on the MMSE/MoCA screening tools without incorporating more comprehensive neuropsychological tests, potentially underestimating the detection rate of mild cognitive impairment and compromising the comprehensiveness of cognitive assessment.

This study is of significant clinical relevance, providing novel insights and guidance for the management of patients with TM. Our results demonstrate that patients with TM exhibit detectable alterations in brain function in the absence of overt neurological symptoms, and that iron burden markers (SF, cardiac T2*), reticulocytes, MMSE, and MoCA may serve as potential predictors for future brain injury risk stratification. These findings provide preliminary data for predictor selection and may facilitate the development of future risk prediction models. We propose integrating routine neurocognitive screening (MMSE, MoCA) and resting-state fMRI into long-term follow-up protocols for high-risk TM individuals, particularly TDT patients. This would serve as a valuable adjunct to existing cardiac and hepatic iron monitoring, enabling early detection and intervention of subclinical neurological impairment. The present study demonstrates that cognitive dysfunction and neurological abnormalities become increasingly pronounced with the severity of systemic iron deposition. Accordingly, further longitudinal studies are warranted to elucidate the threshold of systemic iron burden in relation to brain injury, thereby providing a basis for clinical neurocognitive monitoring.

Conclusions

TM patients present quantifiable cognitive deficits and cerebral functional remodeling linked to focal brain iron accumulation. TDT individuals are especially vulnerable to subclinical brain parenchymal damage induced by systemic iron overload. Collectively, multimodal assessments combining neuroimaging, cognitive testing, peripheral biomarkers, and animal validation enable early identification of thalassemia-related neurological complications to guide timely neuroprotective interventions.

The color gradient indicates T-value distribution, with blue to red spectrum corresponding to increasing T-value trends. In the brain map, blue areas represent brain regions showing decreased intergroup T-values, while red areas indicate increased T-values.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

JXS and RRL designed the study; JXS, CT, and YFL collected and analyzed the data; JXS, MW, and ZBW drafted the manuscript; PP is responsible for image quality control analysis; YMH, CLW, and ZPG cared for patients; RRL and YRL conceived of the study and reviewed and revised the manuscript. All authors reviewed the manuscript.

Funding

The research was supported by Young Leader Talent Training Program of Guangxi Medical University (No. 202302), Advanced Innovation Teams and Xinghu Scholars Program of Guangxi Medical University, Joint Project on Regional High-Incidence Diseases Research of Guangxi Natural Science Foundation under Grant (No. 2023GXNSFAA026293), NHC Key Laboratory of Thalassemia Medicine and Guangxi Key Laboratory of Thalassemia Research.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (Approval No. 2025-E0221). Written informed consent was obtained from all participants in accordance with the Declaration of Helsinki. Animal procedures were approved by the Animal Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (Approval No. 2025-D0311).

Consent for publication

All the authors have agreed to publish.

Competing interests

The authors declare no competing interests.