1. Introduction
Autoimmune diseases are characterized by immune dysregulation and aberrant activation, leading to the misidentification of self-antigens and subsequent damage to cells, tissues, or organs1. These processes stem from lost immune tolerance and autoantibody production, which result in the erroneous destruction of host components1. Autoimmune diseases can be classified into systemic autoimmune diseases that affect multiple tissues, organs, or systems (e.g., systemic lupus erythematosus [SLE] and rheumatoid arthritis [RA]) and organ-specific autoimmune diseases that target a particular organ or tissue (e.g., type 1 diabetes mellitus [T1DM] and multiple sclerosis [MS]). Their multifaceted etiology involves environmental, microbial, hormonal, and genetic factors, but the exact underlying mechanisms remain unclear1. These diseases are globally prevalent and detrimental to the physical and mental well-being of individuals as patients can experience complications, psychological distress, and disability and often require lifelong treatment2. Notably, most autoimmune diseases can be clinically diagnosed, and several genetic, immunological, molecular, and clinical studies have been conducted to elucidate their mechanisms; however, the pathogenesis and pathophysiology of some diseases remain obscure.

Animal models, particularly mice, are invaluable in studying autoimmune diseases. Several mouse models have been developed, including spontaneous, induced, and genetically engineered models; the last group can be further divided into transgenic and knockout variants. Despite the mammalian kinship between mice and humans, there are significant differences in their immune systems, including innate immune molecules, responses to challenges, and immune system structure and function3. These interspecies differences in immune functioning complicate the clinical translation of findings of studies conducted using traditional mouse model. Humanized mouse models bridge this gap by more accurately mimicking the human immune system, thereby helping researchers address human-specific questions within a murine framework.

Humanized mice are characterized by the engraftment of functional human cells, tissues, or immune systems or by the expression of human transgenes4. These models serve as preclinical tools for studying human-specific pathogenesis and pathophysiology and are becoming increasingly vital for research on infectious diseases, cancer, regenerative medicine, transplantation, hematology, allergies, immunity, and drug safety4. Human immune components are reconstituted by engrafting immunodeficient mice with peripheral blood mononuclear cells (PBMCs), hematopoietic stem cells (HSCs), or various human tissues via multiple injection routes, including intravenous, intraperitoneal, intrafemoral, intrahepatic, intrasplenic, intraventricular, or intraarticular injection, or surgical transplantation under the renal capsule, in the mammary fat pad, or subcutaneous tissue4.

The choice between PBMC and HSC engraftment depends on the specific research objective. As they produce mature, patient-specific lymphocytes, PBMC-based models can efficiently replicate the effector phase of autoimmune diseases and can be used to test therapeutic antibody responses5. However, the risk of graft-versus-host disease (GVHD) limits their use in studies requiring long-term observation6. In contrast, HSC-based models permit the study of de novo human immune development and long-term pathogenesis of chronic conditions such as SLE or T1DM with minimal risk of GVHD, although they require more time for immune reconstitution, and often require additional human cytokines or transgenic human leukocyte antigen (HLA) expression for optimal function7.

Several mouse strains have been used in the development of humanized mouse models, including nude mice, severe combined immunodeficient (SCID) mice, recombination-activating gene-1–deficient or gene-2−deficient (Rag1−/− and Rag2−/−) mice, non-obese diabetic (NOD)/SCID mice, and mice lacking a functional interleukin-2 (IL-2) receptor common γ-chain, such as NOD.Cg-PrkdcscidIl2rgtm1Wjl (NSG), NOD.Cg-PrkdcscidIl2rgtm1Sug (NOG), BALB/c-Rag1−/−Il2rg−/− and BALB/c-Rag2−/−Il2rg−/− (both BRG), and NOD.Cg-Rag1tm1MomIl2rgtm1Wjl (NRG) mice8. Additionally, BRG-SIRPαh/m (BRGS), NRG-Tg(HLA-DRA/HLA-DRB10401, HLA-A2.1) (DRAGA), NRG-Tg(HLA-DRA/HLA-DRB10401) (DRAG), BRGS-Tg(thymic-stromal-cell-derived lymphopoietin) (BRGST), NSG-KitW41/W41 (NSGW), and BRGS-CSF1h/hIL3/CSF2h/hThpoh/h (MISTRG) mice have been utilized9.

In this review, we focus on humanized mice engrafted with functional human cells, tissues, or other components and examine models of common systemic and organ-specific autoimmune diseases, including neurological and non-neurological conditions. We also discuss the challenges and potential solutions associated with the use of humanized mouse models to investigate human-specific immune mechanisms. The general workflow for establishing humanized mouse models and their broad applications in autoimmune disease research are illustrated in Figure 1.

2. Systemic autoimmune diseases in humanized mice
2.1. SLE
SLE is a multifaceted autoimmune disorder that affects multiple organs, including the heart, kidneys, and nervous system10. SLE is characterized by immune dysregulation, loss of self-tolerance, and the production of diverse autoantibodies that form immune complexes10. SLE onset is correlated with genetic predisposition, environmental triggers, and hormonal influences10. However, its etiology is much more complex and remains incompletely understood as yet. Conventional mouse models often fail to replicate the human pathophysiological landscape, making humanized models essential for translating murine findings to clinical applications.

A pioneering study by Duchosal et al. in 1990 elucidated the role of peripheral blood leukocytes (PBLs) in SLE pathogenesis by intraperitoneally injecting PBLs from patients with SLE into SCID mice11. This strategy led to the detection of circulating human antinuclear antibodies and deposition of human IgG and mouse C3 in the kidneys, indicating that the transferred PBLs altered serological and immunological functions in the recipient mice. Thus, humanized mouse models can be used as a valuable in vivo system for SLE research.

In another study, Andrade et al. engrafted PBMCs from patients with SLE into BALB/c-Rag2−/−Il2rg−/− mice12, who subsequently developed various autoantibodies such as anti-double-stranded DNA antibodies, antinuclear antibodies (ANAs), and anticardiolipin antibodies in the peripheral blood. Notably, these mice developed severe lupus nephritis, characterized by proteinuria, glomerular necrosis, and renal infiltration of human cells. The model’s ability to develop multiple human-like clinical features underscores its utility in studying the pathophysiology of and therapies for autoimmune diseases.

Recently, Ratliff et al. reported that the expression of the DNA-binding protein ARID3a in hematopoietic stem/progenitor cells (HSPCs) correlates with SLE severity in NSG mice13. They intravenously transplanted CD34+ HSPCs from patients with SLE into NSG mice and detected ANAs in the serum. Mice engrafted with HSPCs expressing high levels of ARID3a+ generated a greater number of antibody-producing cells and antibodies, highlighting ARID3a’s pivotal role in SLE pathogenesis.

Gunawan et al. developed a humanized SLE mouse model by intrahepatically injecting fetal CD34+ HSCs into NSG mice preconditioned with pristane14. This resulted in the generation of various ANAs and a plethora of human pro-inflammatory cytokines in the mice. The mice developed lupus nephritis, presenting with glomerular enlargement with mesangial hyperplasia, increased glomerular cellularity, leukocyte infiltration, and human IgG and IgM deposition in the glomeruli, along with proteinuria. Additionally, pulmonary inflammation, marked by serosal and subpleural inflammation, fibrosis, and mononuclear cell infiltration, was observed. This humanized mouse model closely mirrors several immunological and pathological changes observed in patients with SLE, representing a promising tool for in vivo studies of human-specific SLE mechanisms.

2.2. Systemic sclerosis (SSc)
SSc is a complex autoimmune disease that is characterized by vasculopathy, immune activation, and multiorgan fibrosis15. The development of SSc mouse models has significantly advanced our comprehension of its pathogenesis, as has been the case for other autoimmune conditions15. These encompass sclerodermatous GVHD models; reactive oxygen species (ROS)-induced models; and models triggered by anti-DNA topoisomerase I antibody, angiotensin II, bleomycin, hypochlorous acid, and genetic modifications16. In this review, we focus on humanized mouse models for SSc, which more accurately emulate the human immune milieu.

Luchetti et al. previously introduced a novel skin-humanized mouse model of SSc17. They developed skin constructs using keratinocytes and fibroblasts derived from the skin of patients with SSc. These cells were engrafted onto SCID mice. The regenerated skin exhibited collagen deposition and fibroblast activation, along with elevated Ha-Ras, p-ERK1/2, and COL1A2 expression and increased ROS production. However, these SSc-like characteristics were transient, suggesting that the sustained presence of these manifestations depend on autoantibodies.

This hypothesis was further corroborated by the induction of SSc features in healthy skin engrafted with IgG from patients with SSc18. Bioengineered skin containing keratinocytes and fibroblasts from healthy donors was engrafted onto SCID mice and subsequently injected with IgG from patients with SSc. The treated grafts displayed dermal collagen accumulation, fibrosis, reduced vasculature, increased Ha-Ras and p-ERK1/2 signaling, enhanced ROS production, and COL1A2 upregulation. This humanized mouse model effectively replicates numerous histological and pathological changes observed in SSc in humans, acting as a valuable in vivo system to be used in disease progression studies.

Yue et al. established another humanized mouse model for SSc by injecting patient-derived PBMCs into BALB/c-Rag2−/−Il2rg−/− immunodeficient mice19. This approach resulted in the production ad release of autoantibodies such as ANAs in the serum and CD20+ B cell infiltration around the pulmonary vessels, bronchi, alveoli, kidneys, and muscle tissues. This previous study underscored the indispensable role of both T and B cells in the pathogenesis of PBMC-induced SSc in mice, offering a robust platform for the in vivo investigation of individual immune cell pathogenicity in SSc20.

Similarly, Park et al. developed a humanized SSc mouse model by intravenously injecting PBMCs from patients with SSc into NSG mice21. This model exhibited dermal thickening, increased dermal cytokine levels, heightened expression of fibrosis- and vasculopathy-associated factors, and endothelial cell activation in the skin. Additionally, IL-4, interferon-γ (IFN-γ), and Foxp3 were highly expressed in the skin and lungs, and an increased number of Th17 cells were observed in the blood along with evidence of Th17 cell infiltration. This model captures the key features of SSc in humans, highlighting the potential of patient-derived PBMCs in disease modeling.

To elucidate the role of plasmacytoid dendritic cells (pDCs) in SSc pathogenesis, Ross et al. intravenously injected pDCs from healthy donors into NOD/SCID mice followed by the application of imiquimod cream22. This treatment resulted in the expression of type I IFN, leukocyte recruitment, skin thickening, and development of inflammatory lesions. Subcutaneous bleomycin injections in mice engrafted with human pDCs led to the loss of subcutaneous fat, increased collagen formation, skin thickening, and elevated MX1 and pSTAT1 protein expression, indicating type I IFN signaling activation. These findings underscore the significant contribution of human pDCs to the development of skin fibrosis and inflammation in SSc.

2.3. RA
RA is a chronic systemic autoimmune disease that presents with synovitis, autoantibody production, and osteochondral destruction that often results in progressive disability23. Mouse models have been instrumental in understanding RA pathogenesis, including those induced by collagen, adjuvant, streptococcal cell wall, cartilage oligomeric matrix protein, proteoglycan, pristane, and glucose-6-phosphate isomerase23. Growing evidence has implicated specific HLA class II alleles in RA susceptibility23. For instance, transgenic mice expressing human HLA-DR4 or HLA-DR1 alleles develop chronic inflammatory arthritis upon immunization with type II collagen24. Furthermore, humanized mouse models have been developed to more accurately recapitulate the immunopathogenesis of RA25.

Emerging evidence has linked Epstein–Barr virus (EBV) infection to the development of RA16. Kuwana et al. induced arthritis in NOG mice engrafted with CD34+ HSCs and infected with EBV26. These mice exhibited synovial proliferation, inflammatory cell infiltration, and the presence of multinucleated giant cells and EBV-infected cells near the affected joints. This model effectively mimicked the key histological features of RA in humans, including bone destruction and edema, thereby highlighting the role of EBV in RA pathogenesis. Similarly, Nagasawa et al. demonstrated that EBV infection in humanized NOG mice led to bone erosion characterized by the differentiation of human osteoclasts within the bone marrow. This highlights the importance of this model as a valuable tool for studying EBV-driven bone resorption27.

Alternative models have focused on different induction methods and mouse strains. Misharin et al. developed an arthritis model by intrahepatically injecting human CD34+ HSCs into NSG mice, followed by the intraarticular administration of complete Freund’s adjuvant28. These mice secreted immunoglobulins, including IgG, and exhibited joint swelling, impaired function, immune cell infiltration, and bone destruction, all indicative of arthritis. Etanercept, a tumor necrosis factor (TNF) inhibitor, alleviated arthritis in this model, demonstrating the model’s utility for therapeutic investigation.

Tissue transplantation models using SCID mice have been instrumental in studying localized joint dynamics. Connolly et al. highlighted the pro-inflammatory role of acute-phase serum amyloid A in synovial tissues29. Human RA synovial fibroblasts invaded and damaged the transplanted cartilage, and they were present in the spleen, kidneys, and lymph nodes, revealing their migration route and suggesting their role in arthritis spread. Davis et al. complemented these findings by demonstrating that the SCID–human chimera can sustain the activation and recruitment of human T cells within the synovial graft30. Collectively, these diverse humanized models offer a robust framework for investigating the complex immunopathology of RA.

2.4. Sjögren’s syndrome (SjS)
SjS, a chronic systemic autoimmune disease of unknown origin and is characterized by the production of autoantibodies, activation of autoreactive B and T cells, and a type I IFN signature31. The disease predominantly targets exocrine glands, such as the salivary and lacrimal glands, but it is not uncommon for patients to present with systemic inflammation32. Clinical manifestations include fatigue; arthralgia; dryness of the eyes and mouth; connective tissue disorders; and potential dysregulation of the respiratory, nervous, and vascular systems31,33. Despite the development of various spontaneous, transgenic, knockout, and induction-based murine models, humanized models that accurately reflect human disease dynamics remain notably scarce34.

Young et al. developed a humanized SjS mouse model by intraperitoneally injecting PBMCs from patients with SjS into NSG mice35. Human inflammatory cytokines, including IFN-γ, IL-17, IL-6, and TNF-α, were detected in the serum of injected mice. Furthermore, human-derived CD4+ and CD8+ T cells and inflammation were observed in the lacrimal and salivary glands. These findings are similar to the pathological and histological features of human SjS, characterized by CD4+ T cell infiltration and pro-inflammatory cytokine production32. This innovative mouse model could prove instrumental for the investigation of the pathogenesis of SjS and developing therapeutic strategies.

The humanized mouse models for systemic autoimmune diseases, including their transplanted components and major findings, are summarized in Table 1.

3. Neurologic autoimmune diseases in humanized mice
3.1. MS
MS is characterized by chronic inflammation, demyelination, axonal loss, astrogliosis, and neurodegeneration within the central nervous system (CNS)36. This autoimmune disease involves aberrant T cell responses to myelin antigens, such as proteolipid protein (PLP) and myelin basic protein (MBP)37. Although the etiology and pathogenesis of this disease are not fully understood, MS is suspected to be associated with viral infections, vitamin D deficiency, oxidative stress, mitochondrial dysfunction, energy imbalance, and genetic predisposition36. Several mouse models have been instrumental in examining the immunopathological mechanisms underlying MS, including models developed through exposure to myelin proteins such as MBP, myelin oligodendrocyte glycoprotein (MOG), and PLP38. In addition, humanized transgenic mice expressing multiple human HLA alleles and mice engrafted with human components have been widely used to better mimic human MS.

In a seminal study in 1999, Madsen et al. developed a humanized MS mouse model by presenting an MBP peptide to T cells in specific transgenic mice39. These mice were engineered to express human HLA-DR2, a T cell receptor, and the human CD4 receptor. After being injected with an MBP peptide, the mice exhibited clinical signs of MS, including tail weakness, incontinence, ataxic gait, impaired coordination and balance, bradykinesia, paralysis, and spasticity. Demyelinating lesions were observed in the CNS, along with neutrophil, macrophage, and lymphocyte infiltration. This study sheds further light on the roles of HLA-DR2 and MBP-specific T cells in MS pathogenesis.

To further understand these cellular mechanisms, Zayoud et al. used a humanized NSG model to create a subclinical model of encephalomyelitis40. They intraperitoneally injected DCs and PBMCs from healthy donors into NSG mice, which were then challenged with myelin antigens. Consequently, human leukocytes and splenocytes as well as MOG-specific human IgG were detected in the mice. Inflammatory infiltration in the CNS; accumulation of human CD3+, CD4+, and CD8+ T cells in the parenchyma and meninges; and activated microglial aggregation in the cerebellum were also observed. Although the mice did not present with typical MS symptoms, this model could be valuable for studying human-specific treatments in MS.

The contribution of humoral immunity has also been a focus of investigation. Pedotti et al. found that passive transfer of IgG from patients with MS exacerbated experimental autoimmune encephalomyelitis (EAE) in SJL/J mice41. Following immunization with PLP and patient-derived IgG injection, the mice exhibited more severe clinical manifestations. These included presence of human IgG in the serum and gray matter, inflammatory lesions, white matter demyelination, and lymphocyte and macrophage infiltration in the CNS. This study highlighted the significant role of IgG antibodies in MS pathobiology.

Khare et al. discovered that exposure to antibodies from patients with MS worsened EAE in transgenic humanized mice42. They transferred IgG antibodies from patients with MS into C57BL/6 mice expressing human Fc-gamma receptors (FcγRs) and immunized these mice with MOG. The clinical manifestations of these mice were exacerbated after transfer. This aggravation depended on MOG recognition by antibodies, required the presence of Fc-FcγRs, and was associated with FcγRIIA+ macrophage infiltration. This previous study revealed the importance of FcγRs and MOG-specific antibodies in the etiopathogenesis of MS.

In addition to autoimmune triggers, viral factors for MS have been explored using humanized mice models. Firouzi et al. conducted a study on MS-associated retrovirus in humanized SCID mice to emulate the immune condition in humans with MS43. They injected lymphocytes from healthy donors and MS-associated retrovirus into SCID mice, observing for brain hemorrhage and severe neurological symptoms. This study demonstrated the utility of humanized SCID mice in studying the immunological mechanisms of MS-associated retrovirus in vivo.

Zdimerova et al. examined the roles of EBV and HLA-DR15 in MS pathogenesis by intrahepatically injecting CD34+ hematopoietic progenitor cells from human fetal liver into transgenic HLA-A2 NSG mice infected with EBV44. They observed T cell activation and CD8+ T cell expansion in mice immunized with cells from HLA-DR15–positive donors, and also noted high EBV viral loads, suggesting a poor ability to control EBV. These results indicated that EBV and HLA-DR15 promote T cell activation and impair EBV control in mice.

Humanized mice also served as vital platforms for pharmacological research. Schlöder et al. investigated the effect of dimethyl fumarate (DMF) treatment on the sensitivity of T cell to suppression by regulatory T cells (Tregs) in patients with MS45. They intraperitoneally injected PBMCs from patients with MS into Rag2−/−Il2rg−/− mice, with some mice additionally receiving Tregs from healthy donors. Decreased human immune cell infiltration was observed after Treg injection. Notably, PBMCs derived from DMF-treated patients exhibited a decreased CD4/CD8 ratio, and they were more susceptible to Treg-mediated suppression. This study highlighted the ability of humanized mouse models to mimic human immune dynamics, permitting the evaluations of the effect of therapeutic agents such as DMF on effector T cell responses.

3.2. Anti-N-methyl-D-aspartate receptor (NMDAR) encephalitis
Anti-NMDAR encephalitis is an immune-mediated disorder that is characterized by neuropsychiatric presentations and autoantibodies against the GluN1 subunit of NMDAR46. This interaction between autoantibodies and NMDAR disrupts the normal association of NMDAR with ephrin type B receptor 2 (EphB2), leading to NMDAR internalization and a subsequent reduction in receptor expression at the synapse46. The ensuing NMDAR hypofunction is implicated in the diverse pathophysiological alterations and clinical features observed in patients46. Multiple mouse models have been established to explore the immunopathogenesis of anti-NMDAR encephalitis, with a particular focus on humanized models that more accurately emulate the immune environment in humans.

In 2015, Planagumà et al. developed a mouse model by infusing cerebrospinal fluid from patients with anti-NMDAR encephalitis into the lateral ventricle of C57BL/6J mice47. This led to the emergence of memory impairment, depressive-like behaviors, human NMDAR-targeting antibodies in the hippocampus, and reductions in NMDAR density and protein levels. Notably, these symptoms resolved over time, coinciding with a decrease in antibody levels and a recovery of NMDAR expression. This study offered valuable insights into the cellular, synaptic, and network-level pathophysiology of anti-NMDAR encephalitis.

Given the critical roles of NMDARs in axonal and dendritic growth, neuronal survival, and glutamatergic synaptic transmission, particularly during neurodevelopment, Jurek et al. administered recombinant monoclonal NR1 antibodies to pregnant C57BL/6J mice to assess its impact on fetal brain development48. The antibodies bound to fetal brain synapses, resulting in reduced NMDAR density and altered electrophysiological properties. The offspring exhibited premature death, impaired neuroreflexes, hyperactivity, reduced anxiety, altered blood pH, and lower body weight, along with smaller cerebellar, midbrain, and brainstem sizes in aged individuals. These findings indicate that antibodies transferred across the placenta can adversely affect the developing nervous system.

García-Serra et al. developed a model for placental antibody transfer by injecting pregnant C57BL/6J mice with patient-derived IgG49. They observed that IgG bound to NMDARs, resulting in decreased receptor density and cortical layer thinning. The offspring displayed increased dendritic arborization, reduced spine density, microglial activation, and delays in innate reflexes and eye opening. Additionally, they developed depressive-like behaviors and deficits in nest building, motor coordination, memory, and hippocampal plasticity, demonstrating the detrimental effects of NMDAR antibodies on neurodevelopment.

Taraschenko et al. explored the pathogenic potential of anti-NMDAR antibodies in inducing seizures by infusing cerebrospinal fluid (CSF) or IgG from patients with anti-NMDAR encephalitis into C57BL/6 mice, leading to the induction of electrographic seizures50. These results suggested that autoantibodies play a direct role in seizure genesis in anti-NMDAR encephalitis and provided a model for investigating therapies for autoimmune seizure.

Carceles-Cordon et al. infused patient-derived CSF into the ventricles of C57BL/6J mice to assess its effects51. The mice exhibited a pronounced reduction in the prepulse inhibition of the acoustic startle reflex, indicating the presence of psychotic-like changes, progressive memory deficits, and reduced synaptic NMDAR density. Interestingly, the synaptic and cell surface levels of D1R were reduced, whereas D2R levels were increased. These phenomena resolved following the cessation of CSF infusion, highlighting the model’s ability to replicate clinical features of and NMDAR alterations in encephalitis.

Wright et al. infused IgG from patients with anti-NMDAR encephalitis into the left ventricles of C57BL/6 mice and subsequently administered pentylenetetrazol, which can induce convulsions at subthreshold concentrations52. The mice experienced convulsions, exhibited high seizure scores, and presented with epileptic spike events and human IgG deposition in the hippocampus. Seizure severity was correlated with IgG levels, reflecting the proconvulsive effects of patient-derived IgG and providing insights on the underlying seizure susceptibility mechanisms.

Therapeutic interventions have also been explored using these models. Planagumà et al. demonstrated that coinfusion of ephrin B2 with CSF from patients with anti-NMDAR encephalitis could prevent the emergence of memory deficits and depressive-like behaviors53. Mice infused with CSF alone developed depressive-like behaviors and memory deficits, along with hippocampal NMDAR antibody deposition, reduced synaptic NMDAR and EphB2 density, and impaired long-term synaptic plasticity. Importantly, coinfusion with ephrin B2 prevented these processes, and increased EphB2 and NMDAR density were observed.

Our group recently developed a humanized mouse model by engrafting BALB/c Rag2−/−Il2rg−/−SirpαNODFlk2−/− mice with PBMCs from patients with anti-NMDAR encephalitis54. Anti-GluN1 autoantibodies and compromised blood–brain barrier (BBB) integrity were detected in these mice, along with hyperactive locomotor phenotypes, cognitive deficits, and anxiety- and depressive-like behaviors. We identified IL-1β as a key pathogenic factor mediating BBB disruption, and targeting IL-1β alleviated BBB damage and neuropsychiatric behaviors. This study provided a clinically relevant humanized mouse model of anti-NMDAR encephalitis and revealed the intrinsic pathogenic property of the patients’ lymphocytes.

3.3. Myasthenia gravis (MG)
MG is an autoimmune disorder of the neuromuscular junction and is characterized by antibodies targeting the acetylcholine receptor (AChR), muscle-specific kinase (MuSK), and other postsynaptic muscle endplate molecules55. These antibodies induced a reduction in the number of AChRs, impairing nerve impulse transmission to muscles and manifesting as voluntary muscle weakness and fatigability55. Thyroid disorders are frequently associated with AChR-seropositive MG, and thymic abnormalities such as thymoma or follicular hyperplasia are often observed in patients56. To date, several murine models have been established to mimic these pathologies, including models based on AChR, MuSK, and low-density lipoprotein receptor-related protein 457.

Martino et al. developed a humanized MG mouse model by intraperitoneally injecting SCID mice with PBLs from patients58. These mice exhibited the presence of human anti-AChR IgG at the muscle endplates and reduced postsynaptic neuromuscular junctions, recapitulating key immunological features of MG and providing insights into anti-AChR antibody formation.

Toyka et al. injected patient-derived immunoglobulin into BDF1 mice to investigate its neuromuscular effects59. The mice exhibited decreased endplate potential amplitudes and AChR density at the neuromuscular junctions, along with decremental responses to repetitive nerve stimulation, although these changes were reversed by administering neostigmine. This study confirmed that passive transfer of immunoglobulin from patients could reproduce MG features in mice.

The pivotal role of the thymus was further explored by Schönbeck et al., who transplanted patient-derived thymic tissue under the kidney capsule in SCID mice60. The deposition of human anti-AChR antibodies at skeletal muscle endplates suggested autoantibody targeting in mice, highlighting the role of the thymus in MG onset. Building on this study, Sudres et al. used NSG mice transplanted with patient thymic tissue to evaluate the therapeutic potential of mesenchymal stem cells (MSCs)61. The mice exhibited high mortality and movement disorders associated with AChR depletion, both of which were ameliorated following MSC treatment, thereby validating the model for preclinical therapeutic testing.

Research has also expanded into non-AChR variants. Verschuuren et al. injected IgG from AChR antibody-negative patients into BKTO mice and challenged them with tubocurarine, which resulted in reduced diaphragm contractility and AChR density62. They then transferred IgG from patients with MuSK-associated MG into C57BL/6J mice, who then presented with muscle weakness, weight loss, and reduced postsynaptic AChR density, revealing the presence of non-AChR antibodies in patients and the pathogenicity of MuSK antibodies.

To expand the research on MuSK-associated MG, Huijbers et al. used NOD/SCID mice to demonstrate the therapeutic efficacy of efgartigimod against MuSK IgG4-induced weakness63. In addition to motor symptoms, Sabre et al. identified cognitive deficits and arrested weight gain in MuSK-injected mice, suggesting CNS involvement64. Mechanistically, Kawakami et al. investigated whether MuSK IgG impedes the binding of collagenic tail subunit (ColQ) to MuSK by injecting patient IgG into C57BL/6J mice65. They observed reduced AChR, ColQ, MuSK, and acetylcholinesterase (AChE) expression; smaller neuromuscular junctions; and reduced numbers of ColQ per AChR, suggesting that MuSK IgG disrupts ColQ-tailed AChE binding.

Klooster et al. injected IgG from patients with MuSK-associated MG into NOD/SCID mice, observing weight loss, muscle weakness, and reduced compound muscle action potential, indicating neuromuscular junction transmission disorders66. They also observed reduced AChR density, acetylcholine receptor sensitivity, and acetylcholine release disturbance, confirming the pathogenicity of IgG from patients with MuSK-associated MG and highlighting the presence of neuromuscular junction disorders in MuSK MG.

3.4. Rasmussen’s encephalitis (RE)
RE is a progressive inflammatory brain disorder characterized by intractable epilepsy, cognitive decline, and hemispheric brain atrophy67. The condition predominantly affects one cerebral hemisphere, but the precise etiology and pathogenesis remain elusive67. Neuropathological and immunological investigations suggest that RE is an immune-mediated disease, likely propelled by T cell responses to specific antigenic epitopes, with possible additional roles for autoantibodies67. Despite clinical studies, the development of robust experimental models to investigate the pathogenesis and immunological and pathological changes of RE and identify potential treatments is limited.

Recently, Kebir et al. developed an experimental mouse model exhibiting clinical and pathological features similar to that of human RE68. They intraperitoneally injected PBMCs from patients with RE into NSG mice, observing the development of astrogliosis and seizures of cortical origin. Human T lymphocytes expressing IFN-γ and granzyme B were detected in the mouse brain, indicating a T cell-mediated inflammatory response. Furthermore, the infiltration of human leukocytes and lymphocytes, microglial activation, and focal neuronal injury were noted in the brain tissue. This mouse model provides a valuable tool for elucidating the pathophysiological mechanisms of RE and developing targeted therapies.

The characteristics of humanized models for neurological autoimmune diseases are summarized in Table 2.

4. Common non-neurological organ-specific autoimmune diseases in humanized mice
4.1. T1DM
T1DM is an autoimmune disorder defined by immune-mediated destruction of insulin-secreting pancreatic β-cells, resulting in insulin deficiency, hyperglycemia, and ketosis69. Patients can experience a spectrum of complications as well as multiorgan damage, including chronic dysfunction of the eyes, kidneys, heart, blood vessels, and nerves69. The etiology of T1DM is multifactorial, involving genetic predisposition, immune dysregulation, and environmental factors69. Although various murine models, including induced, spontaneous, and transgenic strains, have been instrumental in T1DM research70, humanized mouse models have emerged as superior tools because of the significant physiological and immunological disparities between murine and human islets and immune systems.

Zhao et al. created a humanized diabetic mouse model by intraperitoneally injecting splenic mononuclear cells from diabetic NOD mice and PBMCs from patients with T1DM into NSG mice71. This led to the development of humanized immune system in the recipient mice, characterized by insulitis, leukocyte infiltration in the islets, impaired glucose tolerance, β-cell destruction, and weight loss. The researchers also identified human T cell migration to pancreatic islets mediated by the interaction between chemokine stromal cell-derived factor 1 and its receptor, highlighting the potential of this axis being used as a therapeutic target.

Humanized mice can also serve as critical platforms for studying islet graft rejection. King et al. demonstrated that when NSG mice with streptozotocin (STZ)-induced diabetes were transplanted with human islets followed by PBMC injection, it led to the transition from transient normoglycemia to hyperglycemia and a loss of human C-peptide, indicating acute graft rejection72. Although STZ is a standard tool for β-cell ablation, its efficacy can be inconsistent because of potential β-cell recovery.

To address this, Brehm et al. developed the NRG–Akita model by backcrossing the Ins2Akita mutation onto the NRG background73. The Ins2Akita mutation causes spontaneous diabetes via insulin misfolding and unfolded protein response. However, subsequent engraftment with HSCs led to hyperglycemia, human CD45+ cell infiltration, and β-cell destruction, indicative of islet rejection. The investigators linked these effects to insulin protein misfolding and subsequent β-cell apoptosis, offering novel insights into islet transplantation rejection and potential immunopathogenesis.

Jacobson et al. reported that human islets were not rejected in BRG mice, although the mechanisms behind this discrepancy remain unclear74. Refinement of humanized models to more accurately replicate the rejection processes could facilitate the development of human-specific therapies to mitigate graft rejection. Emerging cell-based approaches, such as regulatory cells or bone marrow-derived MSCs, display promise in prolonging graft survival75.

Several studies have dissected the roles of specific T cell subsets. Whitfield-Larry et al. utilized NSG-HLA-A2 transgenic mice engrafted with PBMCs from patients with T1DM76. They demonstrated that autoantigen-specific CD8+ T cells were the primary drivers of insulitis and IFN-γ production upon antigen challenge. Viehman et al. injected NSG-DR4 mice with PBMCs or HLA-DR4–restricted CD4+ T cells, revealing higher human CD45+ cell counts in mice receiving PBMCs77. Insulitis and reduced insulin production were observed in PBMC-injected mice, along with CD4+ T cell-induced infiltration into the islets and hepatic and pancreatic acinar infiltration. Unger et al. cloned islet-specific CD8+ T cells from a patient with T1DM and injected them into the pancreata of NSG mice, leading to T cell infiltration in the islets, islet disruption, and reduced insulin production, confirming the migration and cytotoxicity of autoreactive CD8+ T cells78.

These NSG-HLA transgenic mice did not exhibit hyperglycemia, suggesting a potential requirement for both CD4+ and CD8+ T cells in diabetes development; however, Tan et al. developed humanized mice by engrafting HLA-DQ8+ human fetal thymus and CD34+ cells into HLA-DQ8 transgenic mice79. Immunization with human CD4+ T cells and peptides led to hyperglycemia, pancreatic islet destruction, and human CD3+ T cell infiltration, providing evidence that diabetes can be induced in mice with T cells and offering a useful tool for T1DM pathogenesis research.

4.2. Inflammatory bowel disease (IBD)
IBD, encompassing Crohn’s disease (CD) and ulcerative colitis (UC), are chronic inflammatory conditions of the gastrointestinal tract mediated by immune system dysregulation80. UC primarily manifests with inflammatory lesions in the rectum and colon, whereas CD is characterized by discontinuous, transmural inflammation throughout the gastrointestinal tract, often with granuloma formation80. Clinical symptoms include abdominal pain, diarrhea, fatigue, and weight loss, along with additional extraintestinal manifestations and complications such as malnutrition, bowel obstruction, and various secondary conditions80. Despite extensive investigation, the precise etiology of IBD remains elusive. However, it is widely attributed to aberrant immune responses in genetically susceptible individuals, triggered by environmental factors, intestinal microbiota, and mucosal barrier dysfunction80. Clinical studies and investigations of mouse models, including spontaneous, chemically induced, genetic, and humanized models, have been conducted to further our understanding of IBD.

Weigmann et al. developed a humanized colitis mouse model by intraperitoneally injecting PBMCs from allergic donors into NSG mice, followed by allergen challenges. This led to the detection of allergen-specific human IgE in the serum81. The mice also exhibited gut inflammation, mucosal hypertrophy, wall thickening, human CD45+ cell infiltration in the colon, and lymphocyte and neutrophil infiltration in the mucosa. Notably, colitis was mitigated by the administration of anti-human IgE antibodies, highlighting the model’s utility for studying the immunological mechanisms and potential treatments of allergic diseases.

In another approach, Nolte et al. chemically induced IBD in human PBMC-engrafted NSG mice82. NSG mice were intravenously injected with PBMCs from patients with UC and challenged with oxazolone. The mice produced human IgG and IgE and presented with diarrhea, weight loss, and inflammatory colonic pathology. Evidence of human CD45+ leukocyte targeting, epithelial disruption, edema, and inflammatory cell infiltration was observed in the colonic lamina propria and mucosa. This model provided a valuable tool for studying leukocyte migration to inflammatory lesions and potential therapeutic strategies based on lymphocyte infiltration in UC.

Goettel et al. developed a chemically induced humanized colitis mouse model by injecting CD34+ HSCs from healthy donors into NSG mice expressing human HLA-DQ8 and lacking murine major histocompatibility complex (MHC) II83. These mice were then treated with 2,4,6-trinitrobenzenesulfonic acid (TNBS), leading to weight loss and colitis. A separate model using NSG mice with human CD4+ T cells and TNBS exhibited weight loss, goblet cell loss, edema, fibrosis, and transmural inflammation84. CD4+ T cell infiltration was detected in the lamina propria along with human cytokines such as TNF-α, IL2, IL4, and IL17A, suggesting that human CD4+ T cells plays a role in the pathogenesis of TNBS-induced colitis in NSG mice.

However, the use of chemicals in the development of humanized IBD mouse models presents certain limitations. Chemicals can induce severe inflammation, potentially obscuring subtle changes in the gut. Additionally, these models often lack a fully functional human innate immune system, which is crucial for accurately mimicking IBD in patients.

Table 3 provides an overview of the models developed for non-neurological organ-specific conditions.

5. Conclusion and future perspectives
Although traditional mouse models are pivotal in autoimmune research, interspecies microenvironmental differences limit their translational utility. Humanized mouse models bridge this gap by mimicking the human immune environment more accurately, making them invaluable in research on the pathogenesis, pathology, immunology, and pathophysiology of autoimmune diseases and development of therapeutic strategies for them. This review summarizes various methodologies used to establish humanized mouse models across several autoimmune diseases, highlighting the heterogeneous clinical and molecular manifestations. These variations underscore the necessity to meticulously select recipient mouse strains and engraftment techniques tailored to specific research objectives. Continuous efforts should be directed toward refining these methods and mouse strains to more comprehensively emulate the complex physiological changes observed in patients. In this context, the integration of multiomics technologies represents a critical frontier for validating model fidelity. Such high-resolution molecular validation will ensure that the pathogenic pathways and cell-state transitions observed in humanized mice truly represent human disease biology, thereby enhancing the predictive value of these models in drug discovery.

Despite the promising potential of humanized mice, several challenges persist. Current efforts focus on optimizing human component engraftment and improving the fidelity of human immune responses7,85. To enhance reconstruction and minimize graft rejection, researchers are aiming to bolster both innate and adaptive immunity while eliminating residual murine innate cells8. Techniques such as pre-engraftment irradiation and macrophage depletion (e.g., via Cl2MDP-encapsulated liposomes) have improved engraftment survival86. The development of novel gene knockout strains of immunodeficient mice is essential for minimizing residual murine cells and enhancing the engraftment of human cells. However, a critical balance must be maintained, as excessive reduction of murine cells can adversely affect the host mice87.

A multitude of mouse strains and engraftment methodologies are available to be used as humanized mouse models to study autoimmune diseases. Fortunately, standardized guidelines for the development of humanized mice have been developed, assisting researchers in selecting the most suitable recipient mouse strain and engraftment technique86.

One challenge in the selection of T cells following human stem cell engraftment is the absence of HLA class I and II molecules in mice, restricting the ability of these models to fully replicate the human immune context. This issue can be partially circumvented by using transgenic mice that express human HLA molecules. Nevertheless, the range of histocompatibility alleles that can be effectively expressed in mice remains limited88. Moreover, the coexistence of both murine MHC class II molecules and human HLA molecules complicates analysis of their respective functions in transgenic humanized mice. To address this, mice lacking endogenous MHC class II expression have been engineered to more accurately represent the human immune milieu89.

Human PBMC infusion is a prevalent method used to generate humanized mouse models90. Nevertheless, this engraftment process can precipitate GVHD in mice, potentially compromising the model’s validity. Using mice with targeted disruption of the IL-2 receptor common γ-chain mitigates this issue, thereby enhancing the engraftment of human cells23. Despite this, infusion of human cells into immunodeficient mice often results in suboptimal reconstitution and survival of human B cells and natural killer (NK) cells. This limitation can be partially overcome by administering recombinant human B lymphocyte stimulator protein or human IL-15 and Flt3l vectors, which stimulate the expansion of human B and NK cells, respectively91.

The procurement of cells and tissues from patients and healthy donors, although invaluable, is restricted by ethical considerations and limited availability. Induced pluripotent stem cell (iPSC) technology offers a solution by enabling the ex vivo amplification of human HSCs, acting as an ethical and renewable resource92. Moreover, HSCs can be serially transplanted from a recipient mouse to another mouse, perpetuating the humanized environment93. In addition to addressing resource constraints, researchers can investigate how a patient’s unique genetic background influences disease progression by generating humanized mice using iPSC-derived HSCs from specific individuals. This approach would enable personalized drug screening, indicating a significant shift from generalized research to tailored therapeutic strategies.

Humanized mice often exhibit impaired humoral immune responses because of the absence or disarray of secondary lymphoid structures that are critical for class switching and affinity maturation after immunization. The introduction of lymphoid tissue inducer cells can ameliorate this limitation94. Additionally, the absence of specific human cytokines in these mice hinders the engraftment, differentiation, and maturation of HSCs into functional immune cells. Supplementing mice with plasmids encoding human cytokines can address this but this approach might inadvertently promote the proliferation of mouse cells at the expense of engrafted human cells95.

Some existing mouse models are characterized by a short lifespan and limited observational timeframes for disease progression; this makes the development of sophisticated humanized mouse models imperative. These constraints also necessitate the creation of models that can more enduringly and accurately represent the human immune system and disease trajectories.

Another aspect that needs improvement is the integration of the human microbiome–immune axis. Recently, germ-free humanized mouse models have been developed to study human-specific pathogens96. Given the growing evidence that gut microbiota trigger and modulate autoimmune dysregulation, constructing models that combine a humanized immune system with a humanized gut microbiota via fecal microbiota transplantation from patients with SLE, MS, or RA is a key future direction. Such models would provide a sophisticated environment to elucidate how specific human microbial metabolites or molecular mimicry modulate human T cell polarization and autoantibody production, offering a more holistic understanding of the gene–environment interactions that drive systemic autoimmunity.

In summary, humanized mice, which partially reconstruct the human immune system and emulate its microenvironment, can prove to be potent investigative tools in autoimmune disease research. As these models evolve to more closely mirror human immune conditions, our understanding of autoimmune diseases will increasingly improve. Although current models do not fully replicate the human immune system, they are sufficient to explore significant questions concerning the etiology, progression, and therapeutics of autoimmune diseases. The relevance of humanized mouse models will persist, and they are poised to yield novel insights into autoimmune pathologies.