In vivo electroporation induces AQP4 overexpression in skeletal muscle. The M23 isoform of AQP4 exists in the plasma membrane as homotetrameric and heterotetrameric units, which aggregate to form intramembranous particles (IMPs) and high-order orthogonal arrays of particles (OAPs) (29, 30). To transfect the tibialis anterior muscle of each mouse, we used in vivo electroporation with a plasmid encoding the M23 isoform of mouse AQP4 (pAQP4) and an empty plasmid vector control (pEmpty). The distribution of AQP4 in muscle fibers after electroporation was visualized by immunostaining. The muscle from naive mice showed weak and diffuse distribution of AQP4 immunoreactivity and no Myc-DDK staining. The muscle from the mice that received pEmpty showed weak AQP4 immunoreactivity but strong Myc-DDK staining. Electroporation with pAQP4 increased the number and fluorescent intensity of AQP4- and Myc-DDK–costained muscle fibers with punctate distribution of immunoreactivity suggestive of OAP lattices (P < 0.001; Figure 1, B and D). Transfection of HEK293 cells with pAQP4 followed by immunostaining confirmed that AQP4 was expressed on the membrane of the transfected cells with an OAP lattice–like structure (Supplemental Figure 1; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.161003DS1). H&E staining revealed prominent necrotic inflammation in the muscle following electroporation as characterized by marked immune cell infiltration (P < 0.001; Figure 1, C and E).

In vivo electroporation triggers AQP4 overexpression in skeletal muscle. (A) Experimental design. Mice were pretreated with CFA and PTx. Then animals received in vivo electroporation of pAQP4 or pEmpty at the left tibialis anterior muscle. Electroporation was performed at days 0, 14, and 28. Animals were culled at day 42. (B) Coimmunostaining for AQP4 and Myc-DDK in skeletal muscle. Nuclei were counterstained with DAPI. (C) H&E staining for skeletal muscle. Images are representative of the longitudinal section of the tibialis anterior muscle from 5 mice per group. Insets are higher-magnification photomicrographs showing immune cell infiltration. (D) Percentage area of AQP4 and Myc-DDK colocalization in the muscle. (E) Percentage area of necrotic inflammation in the muscle. (F) Western blot analysis of protein from skeletal muscle cell lysate separated in native form by BN-PAGE, and in denatured form by SDS-PAGE. Top: BN-PAGE shows the expression of fusion proteins consisting of Myc-DDK and AQP4 OAPs, IMPs, and tetramers in the muscle after electroporation with pAQP4. Bottom: SDS-PAGE shows the expression of a fusion protein (60 kDa) consisting of Myc-DDK and AQP4 M23 monomers in the muscle after electroporation with pAQP4. Data are mean ± SEM; n = 5 per group. ***P < 0.001, 1-way ANOVA with post hoc Tukey’s test. Scale bar: 50 μm. Original magnification, ×400 (insets).

To characterize AQP4 expression in tibialis anterior muscle after electroporation, we performed Blue native PAGE (BN-PAGE) to separate the native forms of AQP4 tetramers, IMPs and OAPs. Immunoblot analysis revealed that electroporation with pAQP4 profoundly increased the expression of M23 tetramers, IMPs, and OAPs in the muscle compared with naive and pEmpty controls (Figure 1F, top). Next, we resolved denatured proteins from the muscle by SDS-PAGE. Immunoblot analysis revealed that electroporation with pAQP4 robustly increased the expression of a 60 kDa fusion protein, which represents AQP4 monomers together with Myc-DDK, in the muscle compared with naive and pEmpty controls (Figure 1F, bottom). These findings indicate that electroporation with pAQP4 induces the overexpression of AQP4 M23 isoform and triggers inflammation in mouse skeletal muscle.

AQP4 immunization generates circulating AQP4 autoantibodies. We have shown that cell-based indirect immunofluorescence assay using transfected HEK293 cells expressing human AQP4 on the cell membrane is sensitive and specific for the detection of AQP4-IgG in patients with NMOSD (31). To detect AQP4 autoantibodies in mouse serum after AQP4 immunization, we performed the same assay but used transfected HEK293 cells expressing the mouse AQP4 M23 isoform. Signal was visualized by mouse IgG–specific fluorescent-conjugated secondary antibody. In the positive control using commercial anti-AQP4 antibody to stain the transfected cells, we found positive AQP4 immunoreactivity on the membrane of the cells. In the negative controls, with no commercial anti-AQP4 antibody added and without transfection of the cells, no AQP4 immunoreactivity was observed (Figure 2A, top). Moreover, no AQP4 immunoreactivity was detected when the serum of naive, pEmpty(C/P+), and pAQP4(C/P–) mice was used to stain the cells. However, positive AQP4 immunoreactivity was observed when the serum of pAQP4(C/P+) mice was used (Figure 2A, bottom). These results indicate that the serum of pAQP4(C/P+) mice contains autoantibodies that recognize the M23 isoform of mouse AQP4, including its conformational epitopes.

AQP4 immunization generates AQP4 autoantibodies. (A) Detection of AQP4 autoantibodies in mouse serum by cell-based indirect immunofluorescence assay. HEK293 cells were transfected with a plasmid encoding the mouse AQP4 M23 isoform. AQP4 autoantibodies in the serum were visualized by fluorescence-conjugated secondary antibody specific for mouse IgG. Top left: Immunostaining with commercial anti-AQP4 antibody revealed a discontinuous pattern of AQP4 staining on the cell membrane (positive control). Top middle and right: No AQP4 immunoreactivity was observed when commercial anti-AQP4 antibody was absent or HEK293 cells were not transfected (negative controls). Bottom: Immunostaining of transfected HEK293 cells using the serum of naive, pEmpty(C/P+), pAQP4(C/P–), and pAQP4(C/P+) mice. Nuclei were counterstained with DAPI. Images are representative of 8 mice per group. Scale bar: 50 μm. Original magnification, ×400 (insets). (B) Titer of AQP4 autoantibodies was measured by ELISA using serial dilution of serum from 1:10 to 1:10,000. (C) Concentration of AQP4 autoantibodies was determined using serum diluted at 1:1,000. (D) Spinal cord sections of WT and Aqp4-deficient (Aqp4–/–) mice were immunostained using the serum of pEmpty(C/P+) and pAQP4(C/P+) mice. Images are representative of 3 mice per group. Data are mean ± SEM; n = 3 per group. ***P < 0.001, 1-way ANOVA with post hoc Tukey’s test. Scale bar: 50 μm.

To determine the titer of AQP4 autoantibodies, the serum of mice in different groups was analyzed using competitive ELISA. No AQP4 autoantibodies were detected in the serum of naive and pEmpty mice. A low titer of AQP4 autoantibodies was found in the serum of pAQP4(C/P–) mice. By contrast, the serum of pAQP4(C/P+) mice displayed a higher titer of AQP4 autoantibodies (Figure 2B). Using serum diluted in 1:1,000, the concentration of AQP4 autoantibodies in pAQP4(C/P+) mice was significantly higher than that in naive, pEmpty(C/P+), and pAQP4(C/P–) mice (P < 0.001; Figure 2C). These results support that AQP4 immunization via in vivo electroporation triggers the production of AQP4 autoantibodies in mouse circulation.

AQP4 is predominantly located on the astrocytic foot processes at the BBB (32). Next, we tested whether the serum from AQP4-immunized mice could immunostain the spinal cord sections that were obtained from WT and Aqp4-deficient (Aqp4–/–) mice. There was no detectable immunoreactivity when the serum of pEmpty(C/P+) mice was used to stain the spinal cord sections from WT mice and Aqp4–/– mice. However, when the serum of pAQP4(C/P+) mice was used, positive staining was found at the blood vessels of the spinal cord from WT mice, but not at that from Aqp4–/– mice (Figure 2D). These results further confirm that AQP4-immunized mice contain circulating AQP4 autoantibodies.

AQP4 immunization induces motor impairments. Clinical signs of encephalomyelitis were assessed by experimental autoimmune encephalomyelitis score. From day 0 to day 42, pAQP4(C/P+) mice did not display motor weakness compared with naive, pEmpty(C/P+), and pAQP4(C/P–) mice (score 0, data not shown). At day 42, we examined whether AQP4 immunization led to motor impairments using a beam walking test. pAQP4(C/P+) mice took a longer time to cross a 1.2 × 80 cm (width × length) beam than naive, pEmpty(C/P+), and pAQP4(C/P–) mice (P < 0.001; Figure 3A). During the beam walking test, pAQP4(C/P+) mice slipped more often than control mice (P < 0.05; Figure 3B). Similar findings were observed when the test was performed with a narrower 0.6 × 80 cm beam. pAQP4(C/P+) mice took a longer time (P < 0.001; Figure 3C) and slipped more often (P < 0.01; Figure 3D) than controls. These results indicate that AQP4-immunized mice displayed motor impairments.

AQP4 immunization induces motor impairments. (A and B) Beam walking test measuring time taken by a mouse to cross and number of paw slips while crossing a 1.2 × 80 cm (width × length) beam. (C and D) Time taken to cross and number of paw slips while crossing a 0.6 × 80 cm beam. Data are mean ± SEM; n = 8 per group. *P < 0.05, **P < 0.01, ***P < 0.001, 1-way ANOVA with post hoc Tukey’s test.

Circulating AQP4 autoantibodies infiltrate the spinal cord. Without passing through the BBB, circulating AQP4-IgGs in the peripheral blood do not cause CNS damage and acute attack in NMOSD (24). To disrupt the BBB in this model, mice were injected with CFA and pertussis toxin (PTx) before electroporation as described previously (27, 33). At day 42, immunostaining for the tight junction protein ZO-1 revealed a continuous staining pattern in the spinal cord blood vessels of naive and pAQP4(C/P–) mice, but a discontinuous pattern in those of pEmpty(C/P+) and pAQP4(C/P+) mice (Figure 4A). These results indicate that CFA and PTx injections can disrupt the BBB.

IgG infiltration and astrocytopathy after AQP4 immunization. (A) Immunostaining for ZO-1 in the spinal cord of naive, pEmpty(C/P+), pAQP4(C/P–), and pAQP4(C/P+) mice. Insets are higher-magnification photomicrographs showing the pattern of ZO-1 staining in blood vessels. Original magnification, ×400 (insets). (B) Immunostaining for mouse IgG in the spinal cord of naive, pEmpty(C/P+), pAQP4(C/P–), and pAQP4(C/P+) mice. Dotted line demarcates the area of mouse IgG immunoreactivity. (C) Coimmunostaining for AQP4 and GFAP in the spinal cord of naive, pEmpty(C/P+), pAQP4(C/P–), and pAQP4(C/P+) mice. (D) Quantification of mouse IgG infiltration. (E and F) Quantification of AQP4 and GFAP immunofluorescence intensities. Images are representative photomicrographs showing the ventrolateral white matter of cervical spinal cord cross sections from 5 mice per group. Nuclei were counterstained with DAPI. Data are mean ± SEM; n = 5 per group. ***P < 0.001, 1-way ANOVA with post hoc Tukey’s test. Scale bars: 50 μm.

Next, we investigated whether circulating AQP4 autoantibodies generated by AQP4 immunization infiltrated the spinal cord after BBB disruption. Mouse IgG immunoreactivity was found in the spinal cord parenchyma of pAQP4(C/P+) mice, but not in that of controls (Figure 4, B and D). Coimmunostaining confirmed that these mouse IgGs targeted AQP4 in the spinal cord (Supplemental Figure 2). These findings confirm that circulating AQP4 autoantibodies infiltrate the spinal cord through a disrupted BBB.

AQP4 immunization induces astrocytopathy. A hallmark histopathological feature of NMOSD lesions is astrocytopathy characterized by prominent loss of AQP4 and GFAP (34). We assessed whether AQP4 immunization induces astrocytopathy in our model. Coimmunostaining revealed a profound decrease in AQP4 and GFAP levels in the spinal cord of pAQP4(C/P+) mice compared with naive, pEmpty(C/P+), and pAQP4(C/P–) mice (Figure 4C). Merging of the images showed that the loss of AQP4 colocalized with the loss of GFAP immunoreactivity (Figure 4C). Quantification of immunofluorescence intensity confirmed a significant reduction in AQP4 and GFAP levels in pAQP4(C/P+) mice (P < 0.001; Figure 4, E and F). Similar results were observed in the optic nerve of mice in different groups (Supplemental Figure 3, A, C, and D). These data suggest that AQP4 autoantibodies generated by AQP4 immunization cause astrocytopathy in mouse CNS.

AQP4 immunization induces demyelination and axonal loss. Next, we assessed the effect of AQP4 immunization on demyelination, which is another histopathological feature of NMOSD lesions (34). The spinal cord of pAQP4(C/P+) mice displayed patchy loss of Olig2 (oligodendrocyte marker; Figure 5A) and myelin basic protein (MBP; myelin marker; Figure 5B) immunoreactivities, while this loss was not observed in naive, pEmpty(C/P+), and pAQP4(C/P–) mice. Quantification of immunofluorescence intensity showed significant decrease in Olig2 (P < 0.001; Figure 5E) and MBP (P < 0.001; Figure 5F) levels in pAQP4(C/P+) mice. Luxol fast blue staining confirmed that demyelination occurred in the spinal cord of pAQP4(C/P+) mice, but not in that of controls (Supplemental Figure 4A).

Oligodendrocyte loss, demyelination, and axonal loss after AQP4 immunization. (A–C) Immunostaining for Olig2 (A), MBP (B), and NF-H (C) in the spinal cord of naive, pEmpty(C/P+), pAQP4(C/P–), and pAQP4(C/P+) mice. (D) Merged images of MBP and NF-H coimmunostaining. (E and F) Quantification of Olig2 and MBP immunofluorescence intensities. (G) Quantification of the number of NF-H spots. Images are representative photomicrographs showing the ventrolateral white matter of cervical spinal cord cross sections from 5 mice per group. Nuclei were counterstained with DAPI. Data are mean ± SEM; n = 5 per group. **P < 0.01, ***P < 0.001, 1-way ANOVA with post hoc Tukey’s test. Scale bar: 50 μm.

We further examined axonal loss after AQP4 immunization using neurofilament-heavy (NF-H) as an axonal marker. NF-H–positive spots in the spinal cord of pAQP4(C/P+) mice were less dense than those in controls (Figure 5C). Counting of NF-H spots confirmed significant axonal loss in pAQP4(C/P+) mice (P < 0.001; Figure 5G). Coimmunofluorescence of MBP and NF-H revealed an association of demyelination with axonal loss (Figure 5D). These findings were also observed in the optic nerve of mice in different groups (Supplemental Figure 3, B, E, and F). Fluorescent Nissl staining did not show any loss of motoneurons in the ventral horn and interneurons around the central canal in the spinal cord of pAQP4(C/P+) mice (Supplemental Figure 4, B and C). Taken together, these results indicate that AQP4 immunization leads to demyelination and axonal loss.

AQP4 immunization activates microglia. Prominent microglia activation has been found in the lesions of patients with NMOSD and previous experimental models of NMOSD (10, 22, 23, 28, 35). In the present model, we examined whether astrocytopathy, demyelination, and axonal loss coincide with microglia activation. We used Iba-1 as a monocyte marker that labels both brain-derived microglia and infiltrated macrophages. The spinal cord of pAQP4(C/P+) mice displayed a marked increase in Iba-1 immunoreactivity compared with naive, pEmpty(C/P+), and pAQP4(C/P–) mice (Figure 6A). Quantification of immunofluorescence intensity confirmed a significant increase in Iba-1 level in pAQP4(C/P+) mice (P < 0.001; Figure 6B). Coimmunostaining of Iba-1 and GFAP revealed that activated microglia were in close proximity to astrocytes (Figure 6C).

Microglia activation and increase in proinflammatory cytokine levels after AQP4 immunization. (A) Immunostaining for Iba-1 in the spinal cord of naive, pEmpty(C/P+), pAQP4(C/P–), and pAQP4(C/P+) mice. Scale bar: 50 μm. Original magnification, ×400 (inset). (B) Quantification of Iba-1 immunofluorescence intensity. (C) Coimmunostaining for Iba-1 and GFAP. (D–F) ELISA of TNF-α, IL-1β, and IL-6 levels in the spinal cord homogenate of naive, pEmpty(C/P+), pAQP4(C/P–), and pAQP4(C/P+) mice. Images are representative photomicrographs showing the ventrolateral white matter of cervical spinal cord cross sections from 5 mice per group. Nuclei were counterstained with DAPI. Data are mean ± SEM; n = 5 per group (B), n = 3 per group (D–F). ***P < 0.001, 1-way ANOVA with post hoc Tukey’s test. Scale bar: 50 μm.

Activated microglia release excessive proinflammatory cytokines, including TNF-α, IL-1β, and IL-6, to mediate neuroinflammation (36). To assess the level of these cytokines in our model, ELISA was performed in the spinal cord homogenate of mice in different groups. The level of TNF-α, IL-1β, and IL-6 in pAQP4(C/P+) mice was significantly higher than that in naive, pEmpty(C/P+), and pAQP4(C/P–) mice (P < 0.001; Figure 6, D–F). These results suggest a role of microglia activation in the development of NMOSD-like lesions in AQP4-immunized mice.

Complement activation and immune cell infiltration in AQP4-immunized mice. Next, we assessed whether AQP4 immunization induces complement complex deposition and immune cell infiltration. C5b-9 (terminal complement complex marker) immunoreactivity was observed in the spinal cord of pAQP4(C/P+) mice, but not in that of naive, pEmpty(C/P+), and pAQP4(C/P–) mice (Supplemental Figure 5A). Scattered Ly6G-immunoreactive (neutrophil marker) and Siglec-F–immunoreactive (eosinophil marker) cells were found in the spinal cord of pAQP4(C/P+) mice, but not in that of controls (Supplemental Figure 5, B and C).

CD68-immunoreactive cells, which represented both brain-derived and infiltrated macrophages, were observed in the spinal cord of pAQP4(C/P+) mice, but not in that of controls (Supplemental Figure 5D). However, sparse F4/80-immunoreactive cells, which represented infiltrated macrophages, were found in the spinal cord of pAQP4(C/P+) mice (Supplemental Figure 5E). These results suggest that only a few circulating macrophages infiltrated the spinal cord parenchyma.

H&E staining confirmed prominent immune cell infiltration in the spinal cord of pAQP4(C/P+) mice (Supplemental Figure 5F). We did not find any immunoreactivity for CD49b (natural killer cells), CD19 (B cells), and CD4 (Th cells) in the spinal cord of mice in all groups (data not shown).

Aqp4–/– mice do not display NMOSD-like disease activity after AQP4 immunization. To confirm that this experimental disease model is mediated by an autoimmune reaction to AQP4, we performed AQP4 immunization in Aqp4–/– mice and WT littermate controls. Beam walking test revealed that WT pAQP4(C/P+) mice took a longer time to cross a 1.2 × 80 cm (width × length) beam than Aqp4–/– pAQP4(C/P+) mice (P < 0.001; Figure 7A). WT pAQP4(C/P+) mice slipped more often than Aqp4–/– pAQP4(C/P+) mice (P < 0.001; Figure 7B). Similar findings were observed when the test was performed with a 0.6 × 80 cm beam (Figure 7, C and D). Next, we examined whether Aqp4–/– mice displayed NMOSD-like pathologies after AQP4 immunization. As expected, the spinal cord of Aqp4–/– pAQP4(C/P+) mice did not show any AQP4 immunoreactivity. Importantly, there was a significant difference in GFAP immunoreactivity between the spinal cords of Aqp4–/– pAQP4(C/P+) and WT pAQP4(C/P+) mice (P < 0.001; Figure 7, E and F). Moreover, the spinal cord of Aqp4–/– pAQP4(C/P+) mice displayed profound reduction in Iba-1 immunoreactivity compared with that of WT pAQP4(C/P+) mice (P < 0.001; Figure 7, G and H). There was no patchy loss of MBP immunoreactivity and decrease in the number of NF-H spots in Aqp4–/– pAQP4(C/P+) mice compared with WT pAQP4(C/P+) mice (P < 0.001; Figure 7, I and J). These results indicate that the motor impairments and NMOSD-like pathologies observed in this model are driven by an autoimmune response against AQP4, hence they are not observed in immunized Aqp4–/– mice.

AQP4 immunization does not induce motor impairments and spinal cord pathologies in Aqp4-deficient mice. WT or Aqp4-deficient (Aqp4–/–) mice received in vivo electroporation of pAQP4 at the left tibialis anterior muscle. Electroporation was performed at days 0, 14, and 28. Animals were culled at day 42. (A and B) Beam walking test measuring time taken by a mouse to cross and number of paw slips while crossing a 1.2 × 80 cm (width × length) beam. (C and D) Time taken to cross and number of paw slips while crossing a 0.6 × 80 cm beam. (E) Immunostaining for AQP4 and GFAP in the spinal cord of WT pAQP4(C/P+) and Aqp4–/– pAQP4(C/P+) mice. (F) Quantification of AQP4 and GFAP immunofluorescence intensities. (G) Immunostaining for Iba-1 in the spinal cord of WT pAQP4(C/P+) and Aqp4–/– pAQP4(C/P+) mice. (H) Quantification of Iba-1 immunofluorescence intensity. (I) Immunostaining for MBP and NF-H in the spinal cord of WT pAQP4(C/P+) and Aqp4–/– pAQP4(C/P+) mice. (J) Quantification of MBP immunofluorescence intensity and number of NF-H spots. Images are representative photomicrographs showing the ventrolateral white matter of cervical spinal cord cross sections from 6 mice per group. Data are mean ± SEM; n = 6 per group. **P < 0.01, ***P < 0.001, Student’s 2-tailed t test. Scale bars: 50 μm.

AQP4 immunization induces the expansion of splenic Tfh, Th1, Th17, memory B, and plasma cells. Tfh cells drive GC responses with B cell proliferation, differentiation, isotype switching, somatic hypermutation, and affinity maturation (13). In patients with NMOSD, the frequency of circulating Tfh cells has been found to be correlated with the disease activity (15–18). Consistently, we did not find any difference in the circulating Tfh cell frequency between healthy controls and patients with NMOSD in remission receiving immunosuppressive therapy (Supplemental Table 1 and Supplemental Figure 6). In the present animal model, we observed increases in the frequency of Tfh cells (CD4+CXCR5+PD-1+, P < 0.001; Supplemental Figure 7, A and D), Th1 cells (CD4+IFN-γ+, P < 0.001; Supplemental Figure 7, B and E), Th17 cells (CD4+IL-17A+, P < 0.001; Supplemental Figure 7, C and F), memory B cells (CD19+CD80+, P < 0.05; Supplemental Figure 8, A and C), and plasma cells (CD19–CD138+TACI+, P < 0.05; Supplemental Figure 8, B and D) in the spleen of pAQP4(C/P+) mice compared with naive, pEmpty(C/P+), and pAQP4(C/P–) mice.

Anti–ICOS-L antibody ameliorates disease activity in AQP4-immunized mice. Tfh cells require a sustained ICOS/ICOS-L signaling to maintain their phenotype (37, 38). Abrogation of the signaling with antibodies targeting ICOS or ICOS-L has been shown to deplete Tfh cells and suppress GC responses (37, 38). To examine whether Tfh cells influence B cell responses, we depleted Tfh cells in pAQP4(C/P+) mice using anti–ICOS-L antibody (Figure 8A). We found that anti–ICOS-L antibody–treated pAQP4(C/P+) mice had fewer splenic Tfh cells than isotype control antibody–treated pAQP4(C/P+) mice (CD4+CXCR5+PD-1+, P < 0.001; Figure 8B). Moreover, the frequency of splenic Th1 cells (CD4+IFN-γ+, P < 0.01; Figure 8C), Th17 cells (CD4+IL-17A+, P < 0.001; Figure 8D), memory B cells (CD19+CD80+, P < 0.01; Figure 8E), and plasma cells (CD19–CD138+TACI+, P < 0.01; Figure 8F) in anti–ICOS-L antibody–treated pAQP4(C/P+) mice was significantly lower than that in controls.

ICOS-L blockade depletes Tfh cells and suppresses Th1, Th17, memory B, and plasma cell responses. (A) Experimental design. Beginning at day 0, pAQP4(C/P+) mice were given 150 μg anti–ICOS-L or isotype control antibody (i.p.) 3 times a week. Animals were culled at day 42. (B) Representative flow cytometry plots of splenic CXCR5+PD-1+ Tfh cells following isotype control or anti–ICOS-L treatment, pre-gated on CD4+ T cells. (C) Representative flow cytometry plots of splenic CD4+IFN-γ+ Th1 cells following isotype control or anti–ICOS-L treatment. (D) Representative flow cytometry plots of splenic CD4+IL-17A+ Th17 cells following isotype control or anti–ICOS-L treatment. (E) Representative flow cytometry plots of splenic CD19+CD80+ memory B cells following isotype control or anti–ICOS-L treatment. (F) Representative flow cytometry plots of splenic CD138+TACI+ plasma cells following isotype control or anti–ICOS-L treatment, pre-gated on CD19– cells. Flow data are quantified for all groups. Data are mean ± SEM; n = 4–5 per group. **P < 0.01, ***P < 0.001, Student’s 2-tailed t test.

Next, we investigated whether anti–ICOS-L antibody treatment ameliorates NMOSD-like disease activity in AQP4-immunized mice. Beam walking test revealed that anti–ICOS-L antibody–treated pAQP4(C/P+) mice took a shorter time to cross the 1.2 × 80 cm and 0.6 × 80 cm beams than isotype control antibody–treated pAQP4(C/P+) mice (P < 0.05; Figure 9, A and B). During the walking test on the 0.6 × 80 cm beam, anti–ICOS-L antibody–treated mice slipped less often than control (P < 0.05; Figure 9B). Consistent with improved motor functions, there was less AQP4 and GFAP loss in the spinal cord of anti–ICOS-L antibody–treated pAQP4(C/P+) mice compared with controls (Figure 9C). Quantification of immunofluorescence intensity confirmed that these reductions were significant (P < 0.05; Figure 9E). Furthermore, anti–ICOS-L antibody treatment prevented loss of MBP immunoreactivity and NF-H spots in the spinal cord of pAQP4(C/P+) mice compared with controls (Figure 9D). Quantification of MBP immunofluorescence intensity and number of NF-H spots confirmed that these observations were significant (MBP, P < 0.01; NF-H, P < 0.05; Figure 9F). Taken together, these results suggest that Tfh cell depletion via the abrogation of ICOS/ICOS-L signaling prevents AQP4 immunization–induced motor impairments and NMOSD-like pathologies.

Tfh cell depletion ameliorates NMOSD disease activity. Beginning at day 0, pAQP4(C/P+) mice were given 150 μg anti–ICOS-L or isotype control antibody (i.p.) 3 times a week. Animals were culled at day 42. (A and B) Beam walking test measuring time taken by a mouse to cross and number of paw slips while crossing a 1.2 × 80 cm beam (A) and a 0.6 × 80 cm beam (B). (C) Coimmunostaining for AQP4 and GFAP in the spinal cord of pAQP4(C/P+) mice following isotype control or anti–ICOS-L treatment. (D) Coimmunostaining for MBP and NF-H in the spinal cord of pAQP4(C/P+) mice following isotype control or anti–ICOS-L treatment. (E) Quantification of AQP4 and GFAP immunofluorescence intensities. (F) Quantification of MBP immunofluorescence intensity and number of NF-H spots. Images are representative photomicrographs showing the ventrolateral white matter of cervical spinal cord cross sections from 3 mice per group. Nuclei were counterstained with DAPI. Data are mean ± SEM; n = 3 per group. *P < 0.05, **P < 0.01, Student’s 2-tailed t test. Scale bars: 50 μm.