Demographic data
Totally 1,774 subjects were initially enrolled, including 1,082 HCs, 290 AQP4 antibody positive neuromyelitis optica spectrum disorders (AQP4 + NMOSD) and 402 multiple sclerosis (MS) cases. Fourteen HCs, 6 AQP4 + NMOSD and 6 MS were excluded due to a history of other CNS disease. Three HCs, 6 AQP4 + NMOSD and 5 MS were excluded due to poor image quality. Finally, 1,734 subjects, including 1,065 HCs (age = 45 [31, 54], median [interquartile range, IQR] female percentage = 566/1,065), 278 AQP4 + NMOSD (age = 43 [31, 53] years; female percentage = 256/278) and 391 MS (age = 34 [27, 42] years, female percentage = 264/391) were included in this study (Table 1).
Brain and spinal cord spatiotemporal atrophy subtypes in AQP4 + NMOSD and MS
Three AQP4 + NMOSD atrophy subtypes were identified (Figs. 1 and 2 and Additional file 1: Fig. S1): (1) cortical atrophy subtype (NMOSD-C, n = 87, 31.3%), with gradual atrophy of cortical, subcortical and cerebellar GM, spinal cord and brainstem across all stages; (2) spinal cord atrophy subtype (NMOSD-SC, n = 58, 20.9%), with gradual atrophy of the spinal cord, brainstem, subcortical, cerebellar and cortical GM across all stages; and (3) cerebellar atrophy subtype (NMOSD-CE, n = 26, 9.4%), with gradual atrophy of cerebellar GM and WM, and brainstem in early stages, and subcortical and cortical GM, cerebral WM and spinal cord in late stages. Additionally, 107 (38.5%) AQP4 + NMOSD were “normal-appearing” (NMOSD-NA).
SuStaIn subtypes of AQP4 + NMOSD and MS. a, Ternary plot showing the probability of each individual to be classified in a subtype. Dots are labeled by final subtype classification. b, Averaged z-score mappings of brain and spinal cord volumes for each disease and its subtypes. The positive Z-score, indicating atrophy compared to healthy controls, is used for visualization. c, Brain and spinal cord regional mapping of differences (T value) between one subtype and all other subtypes using OLS linear models with adjustment for SuStaIn stage. AQP4, aquaporin 4; NMOSD, neuromyelitis optica spectrum disorders; MS, relapsing–remitting multiple sclerosis; NMOSD-C, cortical atrophy leading subtype of AQP4 antibody positive (AQP4 +) NMOSD; NMOSD-SC, spinal cord atrophy leading subtype of AQP4 + NMOSD; NMOSD-CE, cerebellar atrophy leading subtype of AQP4 + NMOSD; MS-C, cortical atrophy leading subtype of MS; MS-SC, spinal cord atrophy subtype of MS; MS-DGM, deep gray matter atrophy subtype of MS
SuStaIn stages of AQP4 + NMOSD and MS subtypes. Progression of each subtype through SuStaIn stages. Each image is a mean of individuals classified for the listed stages. Here stages from 1–20 are displayed, comprising a majority of subtypeable cases (224 out of 268 for MS and 168 out of 171 for AQP4 + NMOSD). AQP4, aquaporin 4; NMOSD, neuromyelitis optica spectrum disorders; MS, relapsing–remitting multiple sclerosis; NMOSD-C, cortical atrophy leading subtype of AQP4 antibody positive (AQP4 +) NMOSD; NMOSD-SC, spinal cord atrophy leading subtype of AQP4 + NMOSD; NMOSD-CE, cerebellar atrophy leading subtype of AQP4 + NMOSD; MS-C, cortical atrophy leading subtype of MS; MS-SC, spinal cord atrophy subtype of MS; MS-DGM, deep gray matter atrophy subtype of MS
Three MS atrophy subtypes were also identified (Figs. 1 and 2 and Additional file 1: Fig. S1): (1) MS-C subtype (n = 72, 18.4%), with gradual atrophy of cortical and subcortical GM, cerebral WM and brainstem in early stages, and spinal cord, and cerebellum in late stages; (2) MS-SC subtype (n = 115, 29.4%), with gradual atrophy of the spinal cord, brainstem, subcortical GM, cerebral and cerebellar WM, cerebral and cerebellar GM across the stages; and (3) Deep gray matter atrophy subtype (MS-DGM, n = 81, 20.7%), with gradual atrophy of subcortical GM, cerebral WM and GM, brainstem, cerebellar WM and spinal cord across all stages. Additionally, 123 (31.4%) MS cases were “normal-appearing” (MS-NA).
Subtype stability and stage progression of AQP4 + NMOSD and MS subtypes
In AQP4 + NMOSD (Fig. 3c), 20 (20/28, 71.4%) individuals exhibited the same subtype at both baseline and follow-up or progressed from NA to a subtype. Disease stability after excluding individuals classified as NA at baseline and follow-up was found in 80% of cases (12/15).
Clinical and MRI characteristics, stability and stage association of AQP4 + NMOSD and MS atrophy subtypes. a Clinical characteristics of atrophy subtypes. b MRI characteristics of atrophy subtypes. c Subtype stability and stage progression of NMOSD and MS subtypes using longitudinal scans. d Clinical and MRI associations of atrophy stage. AQP4, aquaporin 4; NMOSD, neuromyelitis optica spectrum disorders; MS, multiple sclerosis; NMOSD-C, cortical atrophy leading subtype of AQP4 antibody positive (AQP4 +) NMOSD; NMOSD-SC, spinal cord atrophy leading subtype of AQP4 + NMOSD; NMOSD-NA, “normal-appearing” AQP4 + NMOSD; NMOSD-CE, cerebellar atrophy leading subtype of AQP4 + NMOSD; MS-NA, “normal-appearing” MS; MS-C, cortical atrophy leading subtype of MS; MS-SC, spinal cord atrophy subtype of MS; MS-DGM, deep gray matter atrophy subtype of MS; WMH, white matter hyperintensity; CVLT, California Verbal Learning Test; Brief BVMT, Visuospatial Memory Test-Revised; PASAT, Paced Auditory Serial Addition Test; SDMT, Symbol Digit Modalities Test; COWAT, Controlled Oral Word Association Test; EDSS, Expanded Disability Status Scale
In MS (Fig. 3c), 30 (30/33, 90.9%) individuals exhibited the same subtype at both baseline and follow-up or progressed from NA to a subtype. Disease stability after excluding individuals classified as NA at baseline and follow-up was found in 84.2% of cases (16/19).
Clinical and MRI features of AQP4 + NMOSD and MS atrophy subtypes
In AQP4 + NMOSD (Fig. 3a and b), compared with NMOSD-NA, NMOSD-C had higher EDSS score, lower COWAT score, larger choroid plexus volume, and lower FAs of cerebral WM and brainstem. Compared with NMOSD-NA, NMOSD-SC had a higher number of relapses, lower BVMT score, larger choroid plexus volume, and lower cerebral WM-FA. Compared with NMOSD-NA, NMOSD-CE only had lower brainstem FA. In between atrophy subtype comparisons, NMOSD-C showed lower COWAT score than NMOSD-SC and NMOSD-CE. NMOSD-SC showed lower BVMT score than NMOSD-C. NMOSD-CE had higher cerebellar GM-fALFF than NMOSD-SC. Details are found in Additional file 1: Supplementary Results. Additional analyses by stratifying patients according to their primary clinical syndromes were presented in Additional file 1: Fig. S6. Preliminary subgroup analyses showed similar trends with the main findings when the NMOSD patients were stratified by optic neuritis and myelitis, indicating weak associations between these clinical syndromes and specific atrophy subtypes.
In MS (Fig. 3a and b), compared with MS-NA, MS-C had lower PASAT and SDMT scores, and larger choroid plexus volume. Compared with MS-NA, MS-SC had a higher number of relapses, higher EDSS score, lower PASAT score, longer disease duration, larger choroid plexus volume, and lower FAs of cerebral WM, cerebellar WM and brainstem. Compared with MS-NA, MS-DGM had higher EDSS score, lower PASAT score, longer disease duration, larger choroid plexus volume, and lower cerebral and cerebellar WM-FAs. In between atrophy subtype comparisons, MS-SC had a higher number of relapses, and lower cerebral and cerebellar WM-FAs compared with MS-C and MS-DGM, while MS-C had lower SDMT score than MS-SC. Details are found in Additional file 1: Supplementary Results.
Stage associations in AQP4 + NMOSD and MS atrophy subtypes
In AQP4 + NMOSD, no differences in stages were observed among NMOSD-C, NMOSD-SC and NMOSD-CE. Atrophy stage in NMOSD-C was correlated with EDSS score (R = 0.25, p = 0.0031, pFDR = 0.022), number of relapses (R = 0.18, p = 0.047, pFDR = 0.21) and choroid plexus volume (R = 0.20, p = 0.0072, pFDR = 0.041) (Fig. 3d).
In MS, no differences in stages were observed among MS-C, MS-SC and MS-DGM. Atrophy stage in MS-C was correlated with age (R = 0.20, p = 0.014, pFDR = 0.39), disease duration (R = 0.45, p < 0.0001, pFDR < 0.0001), relapse (R = 0.35, p = 0.0026, pFDR = 0.0095), WMH volume (R = 0.24, p = 0.0027, pFDR = = 0.0095) and choroid plexus volume (R = 0.43, p < 0.0001, pFDR < 0.0001). Atrophy stage in MS-SC was correlated with disease duration (R = 0.16, p = 0.014, pFDR = 0.039), EDSS score (R = 0.15, p = 0.022, pFDR = 0.057), PASAT score (R = −0.21, p = 0.032, pFDR = 0.074), SDMT score (R = −0.38, p = 0.035, pFDR = 0.079), WMH volume (R = 0.17, p = 0.0063, pFDR = 0.020) and choroid plexus volume (R = 0.23, p = 0.00036, pFDR = 0.0014). Atrophy stage in MS-DGM was correlated with disease duration (R = 0.18, p = 0.026, pFDR = 0.065), EDSS score (R = 0.16, p = 0.041, pFDR = 0.088), PASAT score (R = −0.25, p = 0.014, pFDR = 0.039), SDMT score (R = −0.48, p = 0.0044, pFDR = 0.015), WMH volume (R = 0.29, p = 0.00020, pFDR = 0.00085) and choroid plexus volume (R = 0.36, p < 0.0001, pFDR < 0.0001) (Fig. 3d).
Disability worsening and relapse of AQP4 + NMOSD and MS atrophy subtypes
Here, we reported the step-wise backward Cox regression findings (final model by “autoReg” package in R, see Table 2 for details). In AQP4 + NMOSD, NMOSD-CE showed relatively reduced EDSS progression (Hazard Ratio [HR] = 0.11, 95%CI [0.01, 0.99], p = 0.049) and relapse (HR = 0.13 [0.03, 0.69], p = 0.017). In MS, a late stage had a slightly increased risk of disease phenotype conversion from relapsing–remitting MS to SPMS (HR = 1.03, [1.00, 1.06], p = 0.045). No association of disease subtypes, stages, age or sex was observed for the follow-up EDSS worsening or relapse of MS. A summary of AQP4 + NMOSD and MS atrophy subtypes were provided in Fig. 4.
A theoretical model summarizing brain and spinal cord atrophy subtypes in AQP4 + NMOSD (a) and MS (b). Atrophy varies along the axis of disability, relapse, cognition decline, age and disease duration (vertical axis in the diagram) in different AQP4 + NMOSD and MS subtypes. Atrophy varies along a spatiotemporal dimension (horizontal axis in the diagram), such that an individual can be described by their fit along one of at least three trajectories. The text indicates the clinical characteristics of each subtype. The text in bold reflects major clinical differences between subtypes, while normal text reflects MR-related characteristics that differentiate subtypes from normal-appearing individuals. AQP4, aquaporin 4; NMOSD, neuromyelitis optica spectrum disorders; MS, multiple sclerosis; NMOSD-C, cortical atrophy leading subtype of AQP4 antibody positive (AQP4 +) NMOSD; NMOSD-SC, spinal cord atrophy leading subtype of AQP4 + NMOSD; NMOSD-CE, cerebellar atrophy leading subtype of AQP4 + NMOSD; MS-C, cortical atrophy leading subtype of MS; MS-SC, spinal cord atrophy subtype of MS; MS-DGM, deep gray matter atrophy subtype of MS; WM, white matter
Treatment response to DMT among AQP4 + NMOSD and MS atrophy subtypes
For response to DMT regarding relapse, in AQP4 + NMOSD (Fig. 5), response rates were 78.6% (11/14) for NMOSD-NA, 38.5% (10/26) for NMOSD-C, 27.3% (3/11) for NMOSD-SC and 60.0% (3/5) for NMOSD-CE. NMOSD-C (p = 0.015, pFDR = 0.046) and NMOSD-SC (p = 0.010, pFDR = 0.046) had lower response rates compared with NMOSD-NA. In MS, response rates were 78.6% (11/14) for MS-NA, 77.8% (14/18) for MS-C, 37.9% (11/29) for MS-SC and 42.9% (9/21) for MS-DGM. MS-SC had a lower response rate than MS-NA (p = 0.013, pFDR = 0.038) and MS-C (p = 0.0078, pFDR = 0.038). MS-DGM had a lower response rate compared with MS-NA (p = 0.037, pFDR = 0.055) and MS-C (p = 0.027, pFDR = 0.054).
Treatment response to DMT regarding the disease relapse and physical disability worsening (EDSS worsening) among AQP4 + NMOSD and MS atrophy subtypes. AQP4, aquaporin 4; NMOSD, neuromyelitis optica spectrum disorders; MS, multiple sclerosis; NMOSD-NA, “normal-appearing” AQP4 antibody positive (AQP4 +) NMOSD; NMOSD-C, cortical atrophy leading subtype of AQP4 + NMOSD; NMOSD-SC, spinal cord atrophy leading subtype of AQP4 + NMOSD; NMOSD-CE, cerebellar atrophy leading subtype of AQP4 + NMOSD; MS-NA, “normal-appearing” MS; MS-C, cortical atrophy leading subtype of MS; MS-SC, spinal cord atrophy subtype of MS; MS-DGM, deep gray matter atrophy subtype of MS; DMT, disease-modifying therapy; EDSS, Expanded Disability Status Scale
For response to DMT regarding EDSS worsening, in AQP4 + NMOSD (Fig. 5), response rates were 85.7% (12/14) for NMOSD-NA, 84.6% (22/26) for NMOSD-C, 54.5% (6/11) for NMOSD-SC and 100% (5/5) for NMOSD-CE. No statistical difference among AQP4 + NMOSD subtypes was observed. In MS, response rates were 85% (17/20) for MS-NA, 85.7% (18/21) for MS-C, 53.3% (32/60) for MS-SC and 73.3% (22/30) for MS-DGM. MS-SC had a lower response rate than MS-NA (p = 0.012, pFDR = 0.035) and MS-C (p = 0.0086, pFDR = 0.035).