Natalizumab exacerbates astrocytopathy in NMOSD via blockade of endothelial VCAM1–astrocytic integrin α4 interaction

Significance

While natalizumab is an effective therapy for multiple sclerosis (MS), it paradoxically worsens neuromyelitis optica spectrum disorder (NMOSD), a distinct autoimmune astrocytopathy mediated by AQP4-IgG. Using a dual-disease mouse model, we demonstrate that natalizumab exacerbates NMOSD-like astrocytopathy by blocking endothelial VCAM1–astrocytic integrin α4 interactions, providing a mechanistic explanation for clinical observations of MS therapy-induced NMOSD worsening. Crucially, our findings establish that evaluating astrocyte impact should be mandatory when considering MS therapies for NMOSD. We further demonstrate that direct astrocyte protection strategies can effectively counteract these adverse effects. These insights bridge fundamental neuroimmunology with clinical practice, offering critical guidance for precision treatment in NMOSD.

Abstract

Neuromyelitis optica spectrum disorder (NMOSD) is a rare autoimmune inflammatory disorder of the central nervous system (CNS) that shares clinical features with multiple sclerosis (MS) but typically manifests with more severe symptoms. The presence of pathogenic IgG autoantibodies targeting aquaporin-4 (AQP4) channels on astrocytes serves as a highly specific biomarker that distinguishes NMOSD from MS. Unlike MS, NMOSD is characterized by profound astrocytic destruction and exhibits a distinct response to therapies. Notably, disease-modifying therapies (DMTs) effective in MS, including natalizumab, interferon-β, and fingolimod, not only fail to benefit NMOSD patients but may also exacerbate disease progression. The precise molecular mechanisms underlying this immunomodulator-induced exacerbation, however, remain not yet fully elucidated. Here, we demonstrate that natalizumab alleviated experimental autoimmune encephalomyelitis (EAE) while exacerbating the autoimmune astrocytopathy in an “EAE-NMOSD” mouse model, a phenomenon associated with a reduction in actively proliferating astrocytes. Through molecular and signaling pathway analyses, we identify that endothelial-derived vascular cell adhesion molecule 1 (VCAM1) activates astrocytes via integrin α4 signaling, thereby mitigating astrocytopathy in NMOSD-like mice. Furthermore, astrocyte-specific integrin α4 deficiency exacerbates astrocytopathy, and notably, natalizumab-induced disease exacerbation does not occur in integrin α4-conditional knockout (CKO) mice. Finally, pharmacological activation of astrocytes rescues natalizumab-induced damage and ameliorates demyelination in NMOSD-like mice. Collectively, our findings provide mechanistic gaps regarding the clinical phenomenon underlying natalizumab-induced NMOSD exacerbation and suggest astrocyte-targeted therapeutic strategies as a potential intervention for NMOSD.

Neuromyelitis optica spectrum disorder (NMOSD) is a severe, idiopathic demyelinating disease of the central nervous system (CNS) that primarily manifests as relapsing episodes of optic neuritis and transverse myelitis (1, 2). For over a century, NMOSD was frequently misclassified as a severe variant of multiple sclerosis (MS) (1, 3). A major breakthrough occurred in 2004 when aquaporin-4 (AQP4)-specific autoantibodies (AQP4-IgG) were identified in NMOSD patients, distinguishing it from MS and other demyelinating disorders (4). Advances in clinical characterization, neuroimaging, and immunopathogenesis research have since refined NMOSD diagnosis and treatment strategies (5–7), establishing it as a distinct clinical entity with its own diagnostic criteria (7). Unlike MS, NMOSD patients rarely have a chronic progressive course (8), and initial attacks often have devastating consequences (9), leading to worse long-term outcomes compared to MS patients.

NMOSD is primarily an autoimmune astrocytopathy driven by pathogenic AQP4-IgG, which specifically targets AQP4, a water channel localized to perivascular astrocyte endfeet (4, 10). The pathology features in NMOSD were AQP4 loss and astrocyte lysis, followed by demyelination and neuronal damage (1). In contrast, astrocytes in MS active lesions are termed “reactive astrogliosis” and exhibit increased glial fibrillary acidic protein (GFAP) expression (11). Indeed, cerebrospinal fluid (CSF) GFAP levels, a marker of astrocytic damage, are significantly elevated during NMOSD relapses compared to MS (12). In addition, serum GFAP levels strongly correlate with CSF GFAP levels (13), and both parameters are associated with NMOSD disease activity and severity (13–16).

Although NMOSD and MS share overlapping clinical manifestations, their therapeutic responses differ substantially. Early studies revealed that several MS disease-modifying therapies (DMTs), including natalizumab (17–20), interferon-β (21–24), and fingolimod (25, 26), are ineffective and even increase ARR in patients with NMOSD. However, the mechanisms underlying this treatment-induced disease exacerbation remain unclear. Current therapeutic therapies, which rely on broad immunosuppression or monoclonal antibodies (27–31), do not selectively target astrocytes. This limitation persists despite growing evidence that NMOSD is fundamentally an astrocytopathy, where astrocytic dysfunction represents the initial cellular pathology.

In this study, we aimed to elucidate the mechanistic gaps regarding the clinical phenomenon underlying natalizumab-induced NMOSD exacerbation, focusing on potential cell-specific attack mechanisms. Our findings indicate that natalizumab significantly alleviates EAE disease symptoms, whereas it exacerbates autoimmune astrocytopathy in an “EAE-NMOSD” mouse model. This worsening correlates with a reduction in actively proliferating astrocytes. In vitro, conditioned media from endothelial cells suppressed astrocyte proliferation in the presence of natalizumab, whereas direct application of natalizumab alone had no effect. Further analysis demonstrated that vascular cell adhesion molecule 1 (VCAM1) derived from endothelial cells activates astrocytic integrin α4, thereby promoting astrocyte proliferation and alleviating astrocytopathy in NMOSD-like mice. Conditional deletion of integrin α4 from astrocytes led to decreased numbers of astrocytes and worsened NMOSD-like astrocytopathy. Notably, pharmacological activation of integrin α4 downstream signaling rescued natalizumab-induced astrocytic damage and ameliorated demyelination in NMOSD-like mice, suggesting a potential therapeutic strategy for NMOSD.

Results

Natalizumab Ameliorates EAE But Aggravates NMOSD-Like Astrocytopathy.

Clinical evidence indicates that natalizumab increases annualized relapse rates (ARR) in NMOSD patients (17–20). To experimentally validate this clinical observation, we established an acute NMOSD-like astrocytopathy model by stereotactically injecting AQP4-IgG and human complement into the mouse striatum, which effectively recapitulates the astrocyte loss characteristic of human NMOSD pathology (32, 33). Specificity of the astrocytopathy in our NMOSD-like model was confirmed through systematic comparison of five experimental conditions: control IgG alone, AQP4-IgG alone, human complement alone, control IgG plus human complement, and AQP4-IgG plus human complement. While all control groups displayed only minor mechanical injury at the injection site, mice receiving AQP4-IgG together with human complement exhibited hallmark NMOSD-like pathology, including profound loss of AQP4 and extensive astrocyte damage (SI Appendix, Fig. S1). Using this validated model, we next assessed the effect of natalizumab. Mice were administered natalizumab or an isotype control (Human IgG4) immediately after the NMOSD-like model induction. Control mice received equivalent concentrations of control IgG and human complement (Fig. 1A). Notably, mice treated with natalizumab demonstrated more extensive loss of both GFAP and AQP4 compared to isotype controls (Human IgG4) (Fig. 1 B–D), indicating exacerbated astrocyte damage. These mice also exhibited more extensive loss of MBP, suggesting enhanced demyelination (SI Appendix, Fig. S2 A and B). To evaluate CNS inflammation, a critical feature of NMOSD pathogenesis, we performed immunostaining of Iba1 and CD45. natalizumab treatment did not significantly affect the Iba1⁺ area compared with controls (SI Appendix, Fig. S2 C and D). Analysis of CD45 immunostaining revealed a nonsignificant trend toward an increase in both the CD45⁺ area and the number of CD45⁺ cells after natalizumab administration (SI Appendix, Fig. S2 E and F). Together, these findings demonstrate that natalizumab worsens autoimmune astrocytopathy in an NMOSD-like mouse model.

Fig. 1.

Natalizumab ameliorates EAE but aggravates NMOSD-like astrocytopathy. (A) Schematic diagram of AQP4-IgG purification and concentration, and the experimental timeline for NMOSD-like model induction. Control mice were injected with control-IgG (commercial IgG derived from human serum) and human complement. Created with Biorender.com. (B) Representative immunofluorescent staining of GFAP/AQP4 in mouse brain sections from each experimental group. White dashed lines demarcate lesion areas showing GFAP and AQP4 loss. (C) Quantification of GFAP-negative area in each experimental group (Control mice: n = 7 mice; Human IgG4: n = 6 mice; NTZ: n = 6 mice). (D) Quantification of AQP4-negative area in each experimental group. (Control mice: n = 7 mice; Human IgG4: n = 6 mice; NTZ: n = 6 mice). (E) Schematic of natalizumab injections and the experimental timeline of the “EAE-NMOSD” model. Created with Biorender.com. (F) Clinical score of EAE mice treated with isotype control (Human IgG4) and natalizumab (NTZ) (n = 12 mice per group). (G) Incidence, (H) maximum clinical scores, and (I) day of onset analyses of clinical course shown in (F). Every circle represents a single mouse. Data are from one representative experiment of two independent EAE studies, and both experiments demonstrated consistent results. (J) Representative immunofluorescent staining of GFAP in the brain sections from Human IgG4 or NTZ mice 7 d post-NMOSD-like model induction. White dashed lines demarcate lesion areas showing GFAP loss. (K) Quantification of GFAP-negative area in Human IgG4 versus NTZ mice (n = 8 mice per group). (L) Representative immunofluorescent staining of AQP4 in the brain sections from Human IgG4 or NTZ mice 7 d post-NMOSD-like model induction. White dashed lines demarcate lesion areas showing AQP4 loss. (M) Quantification of AQP4-negative area in Human IgG4 versus NTZ mice (n = 8 mice per group). Pathology extent is expressed as the ratio of the AQP4/GFAP-negative area (α) to the reference hemicoronal section area (β). Data are presented as the means ± SEM. One-way ANOVA for (C and D), two-way ANOVA for (F), and unpaired two-tailed t test for (H, I, K, and M). **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

Considering natalizumab is an approved effective therapy for MS (34, 35), we investigated its divergent effects, ameliorating experimental autoimmune encephalomyelitis (EAE) while exacerbating an experimental model of NMOSD. Using an “EAE-NMOSD” model, we first induced EAE by immunizing mice with the myelin oligodendrocyte glycoprotein (MOG)_35-55 peptide. Two days later, animals were treated with natalizumab or an isotype control. This dose was selected to pharmacologically recapitulate the exposure achieved with the standard clinical dose in MS (300 mg monthly). At the peak of EAE disease activity (day 24 postimmunization), an acute NMOSD-like model was induced via intracranial injection of AQP4-IgG and complement (Fig. 1E). Serum natalizumab concentration measured just prior to this induction was 20.83 µg/ml, closely matching the reported mean average steady-state trough concentrations (23 to 29 µg/ml) in treated MS patients (Tysabri PI 2006), confirming that functionally relevant drug levels were maintained at this critical later time point (SI Appendix, Fig. S3).

Natalizumab selectively inhibits the interaction between integrin α4 (expressed on leukocytes) and VCAM1, effectively blocking leukocyte migration into inflamed tissues and attenuating inflammatory demyelination in the CNS (34, 36). Consistent with its known mechanism, natalizumab treatment significantly ameliorated EAE manifestations compared to isotype controls, as demonstrated by lower mean clinical score, reduced disease incidence, and decreased maximum scores (Fig. 1 F–H). However, disease onset did not differ significantly between natalizumab-treated and human IgG4 mice (Fig. 1I). Furthermore, natalizumab administration markedly decreased demyelination (SI Appendix, Fig. S4 A and B), inflammatory cell infiltration (SI Appendix, Fig. S4 C and D), and pathology (SI Appendix, Fig. S4 E–H) in the spinal cord. Critically, to confirm that the subsequent exacerbation of NMOSD-like astrocytopathy occurred under continuous pharmacologic activity of natalizumab, we analyzed the expression of ITGA in CNS tissue. In natalizumab-treated animals, we observed a significant reduction in brain ITGA4 protein levels (SI Appendix, Fig. S5 A and B) and a marked decrease in VCAM1-ITGA4 colocalization on astrocytes (SI Appendix, Fig. S5 A and C), providing direct evidence of sustained pathway blockade at the time of NMOSD-like model induction. Within this same model, natalizumab exerted a contrasting effect on NMOSD-like pathology. Immunofluorescence analysis showed that natalizumab treatment exacerbated GFAP and AQP4 loss (Fig. 1 J–M) and a more severe reduction in MBP signal (SI Appendix, Fig. S6 A and B), indicating aggravated astrocytopathy and demyelination. Furthermore, natalizumab treatment exhibits more severe neuroinflammation in “EAE-NMOSD” mice relative to control animals (SI Appendix, Fig. S6 C–G). These contrasting findings reveal that while natalizumab effectively ameliorates EAE, it paradoxically exacerbates astrocytopathy, myelin loss, and neuroinflammation in the CNS of “EAE-NMOSD” mice, revealing a pathophysiological divergence mediated by integrin α4 blockade.

Endothelial Cells Mediate the Inhibition of Astrocyte Proliferation by Natalizumab.

Given the central role of astrocytes in NMOSD pathogenesis, we focus on astrocyte activation status in our model. Quantitative analysis revealed that natalizumab treatment significantly reduced both the number of GFAP⁺Ki67⁺ proliferating astrocytes and overall GFAP fluorescence intensity, indicating suppression of astrocyte activation and proliferation in vivo (Fig. 2 A–C). These results suggest that natalizumab specifically modulates astrocyte status in NMOSD-like models. To determine whether this effect reflected direct astrocyte modulation, we conducted complementary in vitro studies using primary astrocyte cultures. Surprisingly, the CCK8 assay showed no significant impact of natalizumab on astrocyte proliferation (Fig. 2D). This finding was corroborated by real-time cell analysis (RTCA) demonstrating unaltered proliferation in natalizumab-treated cultures (Fig. 2 E and F), and further confirmed through EdU incorporation assays showing comparable proliferation rates between treated and control astrocytes (Fig. 2 G and H). These contrasting results suggest natalizumab’s effects on astrocyte activation in NMOSD are likely indirect.

Fig. 2.

Endothelial cells mediate the inhibition of astrocyte proliferation by natalizumab. (A) Representative images of GFAP and Ki67 in the brains of Human IgG4 and NTZ mice. Arrowheads indicate GFAP/Ki67 colocalization. (B) Relative fluorescence intensity of GFAP⁺ cells in the brains of Human IgG4 and NTZ mice (n = 6 mice per group). (C) Number of GFAP⁺Ki67⁺ cells in the brains of Human IgG4 and NTZ mice (n = 6 mice per group). (D) Relative absorbance (450 nm) in primary astrocytes treated with different concentrations of natalizumab in fresh medium (n = 3 independent experiments). (E and F) Primary astrocyte cell index curves and quantified area under the curve (AUC) for treatment with different concentrations of natalizumab in fresh medium (n = 3 independent experiments). Real-time cell analysis (RTCA): Label-free technology for continuous, real-time monitoring of cell proliferation and viability. (G) EdU incorporation assays of primary astrocyte proliferation following direct natalizumab exposure. (H) Percentage of EdU⁺ astrocytes from (G) (n = 4 independent experiments). (I) Relative absorbance (450 nm) in primary astrocytes treated with different concentrations of natalizumab in ECM (n = 3 independent experiments). (J and K) Primary astrocyte cell index curves and quantified area under the curve (AUC) for treatment with different concentrations of natalizumab in ECM (n = 3 independent experiments). (L) EdU incorporation assays of primary astrocyte proliferation following indirect natalizumab exposure. (M) Percentage of EdU⁺ astrocytes from (L) (n = 4 independent experiments). Data are presented as the means ± SEM. Unpaired two-tailed t test for (B, C, H, and M), and one-way ANOVA for (D, F, I, and K). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

Since natalizumab’s primary mechanism involves blocking integrin α4–mediated leukocyte–endothelial interactions, we next investigated whether its effects on astrocytes might occur through endothelial intermediaries. Strikingly, when astrocytes were cultured in endothelial cell-conditioned medium (ECM), natalizumab significantly suppressed astrocyte proliferation, as confirmed by CCK8 (Fig. 2I), RTCA (Fig. 2 J and K), and EdU incorporation assays (Fig. 2 L and M). These results establish that natalizumab indirectly modulates astrocyte proliferation through endothelial cell-dependent mechanisms, rather than through direct action on astrocytes themselves.

Endothelial Cell-Derived VCAM1 Activates Astrocytic Integrin α4 and Promotes Proliferation.

To investigate the role of endothelial cell-derived factors in astrocyte proliferation, we established an ECM-stimulated astrocyte model (Fig. 3A). ELISA analysis confirmed the presence of soluble VCAM1 (sVCAM1) in the ECM (Fig. 3B), a key endothelial cell adhesion molecule known to regulate cellular functions through its primary receptor, integrin α4β1 (VLA-4) (37, 38). Transcriptomic analysis revealed a significant upregulation of integrin subunits Itga4 and Itgb1 in ECM-treated astrocytes compared to controls (Fig. 3C). This finding was further validated by RT-qPCR, which showed elevated expression of Vcam1 along with both Itga4 and Itgb1, key targets of natalizumab (Fig. 3D). At the protein level, western blot analysis demonstrated markedly increased integrin α4 expression following ECM treatment (Fig. 3 E and F). Collectively, these results indicate that ECM not only induces a marked increase in VCAM1 expression but also significantly upregulates the integrin α4β1 at both transcriptional and translational levels in astrocytes.

Fig. 3.

Endothelial cell-derived VCAM1 activates astrocytic integrin α4 and promotes proliferation. (A) Schematic of ECM stimulating astrocytes. Created with Biorender.com. (B) sVCAM-1 concentration in control medium and ECM (n = 3 independent experiments). (C) Heatmap of differentially expressed genes (DEGs) in primary astrocytes cultured with control medium (Control) versus endothelial-conditioned medium (ECM), derived from bulk RNA-seq analysis. Upregulated genes are denoted in red and downregulated genes are indicated in blue. The color scale represents Z-score normalized expression. (D) qRT-PCR analysis of Vcam1 and integrin-related genes in primary astrocytes cultured with control medium or ECM (n = 3 independent experiments). (E and F) Immunoblot analysis of ITGA4 protein levels in primary astrocytes treated with control medium or ECM (n = 3 independent experiments). (G) sVCAM-1 concentrations in endothelial cell supernatants measured by ELISA after treatment with isotype control (Rat IgG1) and α-VCAM1 blocking antibody (n = 3 independent experiments). (H) qRT-PCR analysis of Vcam1 and integrin-related genes in primary astrocytes under different conditions (n = 3 independent experiments). (I and J) Immunoblot analysis of ITGA4 protein levels in primary astrocytes under different conditions (n = 3 independent experiments). (K and L) Primary astrocyte cell index curves and quantified area under the curve (AUC) for different treatments (n = 3 independent experiments). (M) EdU incorporation assays of primary astrocyte proliferation following different exposures. (N) Percentage of EdU⁺ astrocytes from (M) (n = 4 independent experiments). Supernatant from bEnd.3 cell cultures were designated as ECM, while that from astrocyte cultures served as the control medium. Data are presented as the means ± SEM. Unpaired two-tailed t test for (B, F, G, and J), two-way ANOVA for (D and H), and one-way ANOVA for (L and N). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

To determine whether VCAM1 in the ECM drives astrocyte proliferation, we applied a VCAM1-neutralizing antibody (α-VCAM1) to block its activity. First, we quantified sVCAM1 levels in culture supernatants using ELISA, which revealed a significant reduction following antibody treatment (Fig. 3G). Next, we examined the effect of VCAM1 inhibition on the integrin family. As expected, RT-qPCR analysis showed that Vcam1 suppression led to a marked decrease in Itga4 and Itgb1 expression in astrocytes (Fig. 3H). Western blot analysis further confirmed a significant reduction in integrin α4 protein levels (Fig. 3 I and J). Additionally, we observed that ECM significantly enhanced astrocyte proliferation compared to control medium, while VCAM1 neutralization substantially reduced astrocyte proliferation (Fig. 3 K–N). Collectively, these findings indicate that endothelial cell-derived VCAM1 activates astrocytic integrin α4, thereby promoting astrocyte proliferation in vitro.

Endothelial Cell-Specific VCAM1 Overexpression Increases Astrocyte Numbers and Alleviates Astrocytopathy in NMOSD-Like Mice.

To investigate the functional significance of endothelial cell-derived VCAM1 in NMOSD pathogenesis, we generated mice with striatal endothelial cell-specific VCAM1 overexpression (VCAM1-OE) and established the NMOSD-like astrocytopathy model through in situ injection of AQP4-IgG and human complement (Fig. 4A). Immunofluorescence colocalization analysis confirmed successful VCAM1 upregulation specifically in endothelial cells (Fig. 4 B and C). Strikingly, VCAM1-OE mice exhibited significantly increased GFAP⁺Ki67⁺ cell numbers (Fig. 4 D and E), demonstrating enhanced astrocyte proliferation. Furthermore, these mice showed marked protection against astrocytopathy, with reduced loss of both AQP4 and GFAP immunoreactivity compared to controls (Fig. 4 F–I), establishing that endothelial VCAM1 is a key modulator of astrocyte survival in the NMOSD-like model.

Fig. 4.

Endothelial cell-specific VCAM1 overexpression increases astrocyte numbers and alleviates astrocytopathy in NMOSD-like mice. (A) Schematic of lentiviral injections and the experimental timeline for the NMOSD-like model. Created with Biorender.com. (B) Immunostaining of CD31 and VCAM1 in the brains of Vector and VCAM1-OE mice. Arrowheads indicate EGFP/CD31/VCAM1 colocalization. (C) Number of CD31⁺VCAM1⁺ cells in the brains of Vector and VCAM1-OE mice (n = 8 mice per group). (D) Representative immunofluorescent staining of GFAP and Ki67 in the brains of Vector or VCAM1-OE mice. Arrowheads indicate GFAP/Ki67 colocalization. (E) Number of GFAP⁺Ki67⁺ cells in the brains of Vector and VCAM1-OE mice (n = 8 mice per group). (F) Representative immunofluorescent staining of GFAP in mouse brain sections from each experimental group. White dashed lines demarcate lesion areas showing GFAP loss. (G) Quantification of GFAP-negative area in Vector versus VCAM1-OE mice (n = 7 mice per group). (H) Representative immunofluorescent staining of AQP4 in mouse brain sections from each experimental group. White dashed lines demarcate lesion areas showing AQP4 loss. (I) Quantification of AQP4-negative area in Vector versus VCAM1-OE mice (n = 7 mice per group). Pathology extent is expressed as the ratio of the AQP4/GFAP-negative area (α) to the reference hemicoronal section area (β). Data are presented as the means ± SEM. Unpaired two-tailed t test for (C, E, G, and I). *P < 0.05, **P < 0.01, ****P < 0.0001.

To further validate these findings, we performed prophylactic anti-VCAM1 antibody administration in NMOSD-like mice (SI Appendix, Fig. S7A). Effective VCAM1 depletion was confirmed by reduced CD31⁺VCAM1⁺ cells in the brain following antibody treatment (SI Appendix, Fig. S7 B and C). This intervention produced opposite effects to VCAM1 overexpression. Anti-VCAM1 treatment significantly decreased GFAP⁺Ki67⁺ cell numbers (SI Appendix, Fig. S7 D and E) and exacerbated both GFAP and AQP4 loss (SI Appendix, Fig. S7 F–I). These complementary gain- and loss-of-function experiments demonstrate that endothelial VCAM1 critically regulates astrocyte homeostasis in NMOSD-like mice, with VCAM1 signaling promoting astrocyte proliferation while protecting against characteristic astrocytopathy.

Conditional Deficiency of Integrin α4 in Astrocytes Enhances Astrocytopathy.

To determine whether astrocytic integrin α4 mediates these VCAM1-dependent effects, we generated mice with a floxed Itga4 allele with Aldh1l1-CreER^T2 mice to achieve astrocyte-specific deletion of Itga4 (Fig. 5A). Genotyping of the resulting offspring confirmed the presence of the expected floxed Itga4 and Cre alleles (SI Appendix, Fig. S8A). Following tamoxifen induction, RT-qPCR and western blot analysis confirmed a significant reduction in ITGA4 expression in astrocytes after knockout (Fig. 5 B–D), which was further validated by immunofluorescence costaining (Fig. 5 E and F and SI Appendix, Fig. S8 B and C). As expected, the number of GFAP⁺Ki67⁺ cells significantly decreased, indicating that the astrocyte-specific deletion of Itga4 impaired astrocyte proliferation and activation (Fig. 5 G–I). We next examined the impact of Itga4 deficiency on NMOSD-like pathology. Astrocyte-specific deletion of Itga4 led to a marked worsening of NMOSD-like astrocytopathy, with a significant increase in the area of GFAP and AQP4 loss (Fig. 5 J and K). Additionally, the area of MBP loss was also significantly larger following the astrocyte-specific deletion of Itga4 (SI Appendix, Fig. S9 A and D). Immunofluorescence analysis in the NMOSD-like model revealed that astrocyte-specific deletion of Itga4 was associated with an increased Iba1⁺ area compared with controls (SI Appendix, Fig. S9 B and E). A similar upward trend in CD45⁺ area and the number of CD45⁺ cells was observed, although this change did not reach statistical significance (SI Appendix, Fig. S9 C, F, and G). Strikingly, when NMOSD-like injury was induced in these astrocyte-specific Itga4 deletion mice, the exacerbating effect of natalizumab was abolished (Fig. 5 L and M), indicating that the detrimental action of natalizumab in NMOSD-like pathology depends on astrocyte-expressed integrin α4. Together, these results underscore the critical role for astrocyte proliferation, mediated by integrin α4 signaling, plays a critical role in the process of acute autoimmune astrocytopathy in the NMOSD-like model.

Fig. 5.

Conditional deficiency of integrin α4 in astrocytes exacerbates astrocytopathy. (A) Schematic overview of generating Itga4 CKO mice. Created with Biorender.com. (B) qRT-PCR analysis of Itga4 from the brains of Itga4^fl/fl and Itga4 CKO mice (n = 6 mice per group). (C and D) Immunoblot analysis of ITGA4 protein level from the brains of Itga4^fl/fl and Itga4 CKO mice (n = 6 per group). (E) Representative immunofluorescent staining of GFAP and ITGA4 in the brains of Itga4^fl/fl and Itga4 CKO mice. Arrowheads indicate GFAP/ITGA4 colocalization. (F) Number of GFAP⁺ITGA4⁺ cells in the brains of Itga4^fl/fl and Itga4 CKO mice (n = 6 mice per group). (G) Representative immunofluorescent staining of GFAP and Ki67 in the brains of Itga4^fl/fl and Itga4 CKO mice. Arrowheads indicate GFAP/Ki67 colocalization. (H) Relative fluorescence intensity of GFAP⁺ cells in the brains of Itga4^fl/fl and Itga4 CKO mice (n = 8 mice per group). (I) Number of GFAP⁺Ki67⁺ cells in the brains of Itga4^fl/fl and Itga4 CKO mice (n = 8 mice per group). (J) Representative immunofluorescent staining of GFAP/AQP4 in the brain sections from Itga4^fl/fl and Itga4 CKO mice 7 d post-NMOSD-like model induction. White dashed lines demarcate lesion areas showing GFAP and AQP4 loss. (K) Quantification of GFAP/AQP4-negative area in Itga4^fl/fl versus Itga4 CKO mice (n = 8 mice per group). (L) Representative immunofluorescent staining of GFAP/AQP4 in the brain sections from Itga4 CKO mice treated with isotype control (Human IgG4) or natalizumab (NTZ) at 7 d post-NMOSD-like induction. White dashed lines demarcate lesion areas showing GFAP and AQP4 loss. (M) Quantification of GFAP/AQP4-negative area in Human IgG4 or NTZ treatment Itga4 CKO mice (n = 6 mice per group). Pathology extent is expressed as the ratio of the AQP4/GFAP-negative area (α) to the reference hemicoronal section area (β). Data are presented as the means ± SEM. Unpaired two-tailed t test for (B, D, F, H, I, K, and M). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

Pharmacological Activation of the Integrin Downstream Signaling Ameliorates CNS Pathology in NMOSD-Like Mice.

We next evaluated whether activating integrin signaling pharmacologically could mitigate NMOSD-like progression. To this end, we employed 740Y-P, an agonist of the PI3K-AKT pathway that activates downstream integrin signaling (39). After establishing the NMOSD-like model, we coadministered natalizumab with 740Y-P. Notably, this combination treatment significantly increased the number of GFAP⁺Ki67⁺ cells, demonstrating enhanced astrocyte proliferation and activation (Fig. 6 A–C). Moreover, compared with natalizumab plus PBS, 740Y-P plus natalizumab substantially reduced the areas of GFAP and AQP4 loss, indicating an amelioration of acute astrocytopathy (Fig. 6 D–G). Myelin damage was also alleviated, as shown by a decrease in the area of MBP loss (SI Appendix, Fig. S10 A and B). We next examined the effects of integrin signaling activation on CNS inflammation in the NMOSD-like model. Cotreatment with 740Y-P and natalizumab did not significantly alter the Iba1⁺ area compared to natalizumab alone (SI Appendix, Fig. S10 C and D). While analysis of CD45⁺ cells revealed a nonsignificant trend toward an increased immunopositive area, it significantly reduced the absolute number of CD45⁺ cells (SI Appendix, Fig. S10 E and F). These results indicate that pharmacological potentiation of integrin signaling alleviates NMOSD-like injury, highlighting a potential therapeutic strategy for mitigating disease severity.

Fig. 6.

Pharmacological activation of the integrin downstream signaling ameliorates CNS pathology in NMOSD-like mice. (A) Representative immunofluorescent staining of GFAP and Ki67 from mouse brains after the indicated treatment. Arrowheads indicate GFAP/Ki67 colocalization. (B) Relative fluorescence intensity of GFAP in the brains of the indicated groups (Human IgG4 + PBS: n = 6 mice; NTZ + PBS: n = 7 mice; Human IgG4 + 740Y-P: n = 7 mice; NTZ + 740Y-P: n = 7 mice). 740Y-P: 740YPDGFR. (C) Quantification of Ki67⁺ astrocyte numbers in the region around the injection site (Human IgG4 + PBS: n = 6 mice; NTZ + PBS: n = 7 mice; Human IgG4 + 740Y-P: n = 7 mice; NTZ + 740Y-P: n = 7 mice). (D) Representative immunofluorescent staining of GFAP in mouse brain sections from each experimental group. White dashed lines demarcate lesion areas showing GFAP loss. (E) Quantification of GFAP-negative area in each experimental group (Human IgG4 + PBS: n = 6 mice; NTZ + PBS: n = 6 mice; Human IgG4 + 740Y-P: n = 7 mice; NTZ + 740Y-P: n = 7 mice). (F) Representative immunofluorescent staining of AQP4 in mouse brain sections from each experimental group. White dashed lines demarcate lesion areas showing AQP4 loss. (G) Quantification of AQP4-negative area in each experimental group (Human IgG4 + PBS: n = 6 mice; NTZ + PBS: n = 6 mice; Human IgG4 + 740Y-P: n = 7 mice; NTZ + 740Y-P: n = 7 mice). Pathology extent is expressed as the ratio of the AQP4/GFAP-negative area (α) to the reference hemicoronal section area (β). Data are presented as the means ± SEM. One-way ANOVA for (B, C, E, and G). *P < 0.05, **P < 0.01, ****P < 0.001.

In summary, our study demonstrates that natalizumab exacerbates NMOSD-like astrocytopathy by disrupting the interaction between endothelial cell VCAM1 and astrocyte integrin α4, whereas activation of integrin signaling promotes astrocyte proliferation and mitigates the adverse effects of natalizumab on NMOSD-like mice. These findings underscore the critical role of the VCAM1/integrin α4 axis and may pave the way for astrocyte-targeted therapeutic strategies in NMOSD.

Discussion

Natalizumab, a well-established therapy for MS, has long been associated with clinical exacerbation of NMOSD (17–20). However, the molecular mechanisms and signaling pathways driving this adverse effect remain poorly understood. In this study, we demonstrate that natalizumab exacerbates NMOSD-like astrocytopathy by inhibiting astrocyte proliferation, a process mediated by endothelial-derived VCAM1 and astrocytic integrin α4 communication. Furthermore, we show that astrocyte-specific deletion of integrin α4 reduces astrocyte proliferation and worsens NMOSD-like pathology, rendering the condition unresponsive to natalizumab. Importantly, pharmacological activation of astrocytes mitigates astrocytopathy in NMOSD-like mice, indicating that therapies that target astrocytes may offer protective benefits in autoimmune demyelination diseases.

Despite the clinical manifestations overlapping between MS and NMOSD, the accumulation of biological, clinical, and neuropathological research, particularly with the advent of more sensitive and specific diagnoses (6), reveals fundamental distinctions, notably in astrocyte destruction and therapeutic responses. Unlike MS, the initial damage in the NMOSD lesions can already be severe and permanent neurologic disability. Therefore, early and accurate serological diagnosis is crucial in NMOSD patients, as misdiagnosis can delay appropriate treatment and lead to inappropriate therapy with natalizumab, beta-interferon, and fingolimod. Natalizumab, the first monoclonal antibody approved for relapsing MS, is a humanized recombinant IgG4 that prevents leukocyte trafficking into the CNS by blocking integrin α4-mediated leukocyte–endothelial interactions, thereby altering peripheral lymphocyte subsets and reducing inflammatory demyelination (34–36, 40). Notably, it reduces B cells in the CSF while increasing their numbers in the periphery (41). This raises the possibility that natalizumab-induced expansion of peripheral plasma cells could elevate circulating AQP4-IgG levels, given the established correlation between AQP4-IgG titers and disease activity in NMOSD (42). Such an increase in AQP4-IgG titers may partly explain why natalizumab, despite being detectable in the CSF (43, 44) and effective in MS patients, is paradoxically associated with massive astrocyte destruction (17) and extensive brain lesions (18) in NMOSD patients. Beyond this AQP4-IgG-mediated mechanism, our findings reveal another direct, target cell-specific mechanism that explains this clinical paradox. Consistent with the central role of astrocytes as autoimmune targets in NMOSD patients, we confirmed the efficacy of natalizumab in EAE (Fig. 1 F–I) and revealed its detrimental effects on astrocytes in NMOSD-like mice (Fig. 1 B–D). Specifically, natalizumab exacerbates astrocyte damage by disrupting the interaction between endothelial-derived VCAM1 and astrocytic integrin α4 (Figs. 4 and 5). Activation of downstream integrin α4 signaling with the PI3K-AKT pathway agonist 740Y-P enhances astrocyte activation and significantly reduces the brain lesions aggravated by natalizumab in an experimental mouse model of NMOSD (Fig. 6). Interestingly, this phenomenon of astrocyte response in 740Y-P-treated NMOSD-like mice is somewhat surprising because aberrant astrocyte activation is frequently observed in MS/EAE. However, a key difference between MS/EAE and NMOSD-like astrocytopathy/NMOSD is the pattern of astrocyte destruction. This may explain, in part, the widespread astrocytic loss observed in the brains of natalizumab-treated NMOSD patients. Our work, therefore, extends the conceptual framework beyond altered immune surveillance, identifying an astrocyte-centric mechanism that explains the opposing outcomes of integrin α4 inhibition in MS versus NMOSD.

In recent years, several new monoclonal antibody therapies targeting complement (30), B cells (27–29), or IL-6 receptors (31) have been approved for treating AQP4-IgG-seropositive NMOSD. However, these therapies do not specifically target astrocytes, which are the specific cell type affected by the pathogenic AQP4-IgG antibody in NMOSD pathogenesis (4, 10). In contrast, our findings demonstrate that modulating astrocyte activation via the VCAM1/integrin α4 signaling axis attenuates disease pathology in an experimental model of NMOSD (Fig. 6). Thus, this astrocyte-specific target may serve as an avenue for therapeutic intervention in NMOSD and provide a theoretical basis for the development and preclinical assessment that must focus on a specific cell type.

Astrocyte reactivity is a complex and multifaceted process (45), with varying roles in neurological disorders. While some evidence suggests that astrocytes can limit disease progression, particularly during acute inflammation, as seen in EAE (46–48), other studies indicate that their selective ablation during chronic inflammation can alleviate disease pathology (49, 50). However, astrocyte-associated pathologies primarily emerge during the chronic phases of lesion development, and have typically involved nonselective rather than astrocyte-specific targeting in response to diverse insults (51). In NMOSD, lesion formation is initiated by pathogenic AQP4-specific antibodies, which target AQP4 on astrocytes, leading to acute and abrupt perivascular astrocyte destruction. In this study, through modulation of astrocyte activation, we demonstrate a critical role for acute astrocyte responses in the pathogenesis of NMOSD-like mice.

Astrocytes are established as the primary cellular target in NMOSD, where AQP4-IgG binding triggers complement-dependent cytotoxicity and astrocyte loss (4, 10, 52). This paradigm of direct astrocytic injury is extended by our results, which reveal that beyond the initial insult, the communication between astrocytes and the endothelium influences the disease progression. We delineate a signaling axis and identify the VCAM1–integrin α4 interaction as a critical mediator of this crosstalk. Notably, our data show that natalizumab, by blocking this axis, unexpectedly worsens NMOSD-like pathology. This aggravation occurs through the inhibition of astrocyte proliferative response, which amplifies the loss of astrocytes following AQP4-IgG and complement attack. This mechanistic insight aligns with and expands the therapeutic principle established by Tradtrantip et al., wherein neutralization of pathogenic AQP4-IgG prevented disease (33). While their strategy blocks the initial injury trigger, our work points to a complementary, cytoprotective approach, enhancing intrinsic astrocyte proliferation. By modulating the integrin α4-dependent response, we propose a strategy to promote astrocyte activation and a promising therapeutic avenue. Thus, our study bridges the established concepts of astrocyte as target and targeted intervention as strategy, uncovering the VCAM1-integrin α4 pathway as a lever for such intervention. This not only provides a coherent explanation for the paradoxical effect of natalizumab in NMOSD but also positions the modulation of astrocytes as a therapy.

A more refined understanding of the tissue microenvironment’s influence on autoimmunity may elucidate critical mechanisms and identify novel therapeutic targets. Increasing evidence suggests that astrocytes engage in extensive interactions with both CNS-resident and CNS-infiltrating cells (53–56). Astrocytes are a central component of the glia limitans, with their perivascular end-feet forming a barrier that encloses pericytes, endothelial cells, and the basal lamina, thus creating a protective barrier (57). Given the proximity of astrocyte end-feet to endothelial cells, it is unsurprising that astrocytes are poised to respond to endothelial signals. Our study demonstrates that endothelial-derived VCAM1 interacts with astrocyte integrin α4, activating downstream signaling pathways that promote astrocyte proliferation and attenuating disease pathology in an experimental model of NMOSD (Fig. 6 and SI Appendix, Fig. S10). These findings position the VCAM1–integrin α4 axis as a microenvironment-dependent pathway that regulates astrocytes and may offer a target for modulating disease progression in NMOSD.

Our study employed a well-established passive transfer astrocytopathy model in experimental animals involving direct intracerebral administration of AQP4-IgG and human complement (32), which has only partially summarized the histopathological characteristics of NMOSD. We selected this approach precisely to isolate astrocyte responses from the complexities of concomitant adaptive immune activation. The work of Bradl et al., which demonstrated that AQP4-IgG exacerbates T cell–mediated EAE to generate NMOSD-like lesions, was pivotal in establishing the pathogenicity of AQP4-IgG within an inflammatory milieu and in differentiating NMOSD from MS (58). While that model is indispensable for studying the interplay between humoral and cellular immunity, our passive transfer astrocytopathy model intentionally bypasses this interplay to define the primary astrocytic injury cascade. Furthermore, our observations align with and extend the findings of Marignier et al., who demonstrated secondary oligodendrocyte damage following the initial astrocyte injury in similar passive transfer settings (59). By focusing on the acute phase, we aimed to identify the immediate signaling pathways activated within astrocytes at the early stages of injury.

While recent studies have improved our understanding of B cell–mediated immune tolerance to AQP4, active immunization models remain constrained. For instance, immunization with AQP4_201–220 linear peptide in B cell–deficient mice has limited pathology and fails to generate pathogenic AQP4-IgG antibodies in wild-type animals, a key hallmark of NMOSD (60). Furthermore, such models do not fully replicate the widespread astrocyte damage observed in NMOSD patients. Our passive transfer model is based on direct intracerebral injection of AQP4-IgG and human complement. This approach induces rapid astrocyte loss that mimics acute NMOSD-like pathology, thereby enabling focused investigation of the primary astrocytopathy and cellular interactions following AQP4-IgG and complement-dependent injury. The model reveals specific astrocyte responses to autoimmune attack that are difficult to access through AQP4_201–220 linear peptide immunization. Although the acute and transient nature of our NMOSD-like mouse model does not permit tracking of the long-term fate of lesions in an actual therapeutic experiment, it effectively clarifies the astrocyte reactions to AQP4-IgG and complement-dependent cytotoxicity. We recognize the value of other models, yet our model provides a reasonable representation of autoantibody-mediated NMOSD-like astrocytopathy. This model has allowed us to identify a signaling axis in which astrocytes, via integrin α4, respond to endothelial VCAM1 and contribute to NMOSD-like progression. Our findings highlight the therapeutic potential of modulating astrocyte proliferation via the integrin α4 signal in acute NMOSD. Future studies incorporating active immunization or combined passive transfer models will be essential to investigate how peripheral immune mechanisms interact with the VCAM1–integrin α4 pathway identified here. Collectively, our study establishes that natalizumab exacerbates NMOSD-like pathology by disrupting endothelial VCAM1–astrocytic integrin α4 signaling, which leads to suppressed astrocyte proliferation. Astrocyte-specific integrin α4 ablation aggravates astrocytopathy and renders natalizumab ineffective, while pharmacological astrocyte activation mitigates disease severity, highlighting the potential of targeting astrocyte proliferation as a valuable strategy for cell-targeted interventions in NMOSD.

Materials and Methods

All mouse experiments were conducted in accordance with the Regulations for the Administration of Affairs Concerning Experimental Animals (China) and were approved by the Shaanxi Normal University Animal Ethics Committee. This study integrated a multifaceted experimental approach combining bioinformatic investigations with biochemical and in vivo analyses. Pathogenic AQP4-IgG autoantibodies were affinity-purified from plasma exchange (PLEX) fluids obtained from seropositive NMOSD patients. For mechanistic studies, we established two relevant mouse models, a clinically relevant NMOSD-like astrocytopathy model and an “EAE-NMOSD” model. Astrocytopathy and neuroinflammation were assessed through immunofluorescence staining and quantitative imaging. A combination of genetic and pharmacological approaches, including endothelial-specific VCAM1 overexpression and astrocyte-specific Itga4 knockout mice, the use of a neutralizing α-VCAM1 antibody, and targeted agonists, to examine the VCAM1–integrin α4 axis in astrocytes. Primary astrocyte proliferation was quantified using the xCELLigence RTCA system, CCK8, and EDU assays. Detailed materials and methods are provided in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

RNAseq FASTAQ files generated during this study are available at the NCBI Sequence Read Archive (http://www.ncbi.nlm.nih.gov/sra/) under accession ID: PRJNA871055 (61). All other data are included in the manuscript and/or SI Appendix.