All imaging was conducted on a 256 slice Discovery CT750 HD CT Scanner (GE Healthcare, Waukesha, WI). Chest CT scans with axial technique 5-mm slice thickness were evaluated on a Picture Archiving and Communication System (PACS) workstation (Neusoft Co., China). All images were reviewed on a PACS, using a mediastinal window setting (level, 50 HU; width, 350 HU).

Whole blood samples (100 μl) were stained in TruCount™ tubes with anti-CD45RA, CD3, CD4, CD31, and CD62L antibodies (Biolegend) after red blood cell lysis. The CytExpert software (Beckman Coulter) was used for analysis.

After flow cytometric analysis, blood samples were centrifuged to collect the serum.

Serum IL-6 was measured using a human IL-6 Quantikine ELISA Kit according to the manufacturer’s instructions (R&D systems D6050).

Data are presented as mean (standard deviation), median (interquartile range), or frequency (%), as appropriate. Independent-samples t test or Mann–Whitney U test was used to analyze differences between NMOSD and HC, with respect to scalar and nonparametric parameters. The chi-squared test was used to examine the differences of sex and thymus shapes between groups. Linear regressions were performed to examine the relationships of the frequency and absolute counts of naive CD4⁺ T-cell subpopulations with age and sex. Statistical analysis of thymic predominant sides, volume, density, and thymic scores was performed after disaggregating patients by age, sex, and body mass index (BMI). Mediation effects were analyzed by SPSS Statistics 26.0 plug-in PROCESS V3.4. P values less than 0.05 were considered indicative of statistical significance.

Flow cytometry was used to investigate the naive CD4⁺ T cell subpopulations in 44 NMOSD patients (mean age, 42.8 years; range, 15–75) and 21 HC. A representative analysis is shown in Figure 1 . Compared with HC, NMOSD patients displayed significantly higher fraction of peripheral blood CD45RA⁺ CD62L⁺ naive CD4⁺ T cells (95% CI, −13.811 to −1.487; P = 0.016; Table 2 ) which showed an age-dependent decrease (P < 0.001; r = −0.423; Figure 2A ). The absolute numbers of CD45RA⁺ CD62L⁺ naive CD4⁺ T cells in NMOSD patients was also significantly higher than that in HC (95% CI, −1735.694 to −284.597; P = 0.007; Table 2 ) and showed a declining trend with increase in age, although not statistically significant (P = 0.074; r = −0.225; Figure 2B ). No significant difference was observed in naive CD4 T cells subpopulations between anti-AQP4-positive and anti-AQP4-negative patients.

CD31^{+ thymic} naive CD4⁺ T cells is a subpopulation of CD45RA ⁺ CD62L ⁺ naive CD4⁺ T cells and are regarded as direct marker of thymic output function (13, 20).

Because the fraction and absolute count of CD31^{+ thymic} naïve CD4⁺ T cells declines with increase in age (17), linear regressions were performed to examine the relationship of the frequency and absolute count of CD31^{+ thymic} naïve CD4⁺ T cells with age and sex. Correlation analysis showed an age-dependent decrease in the frequency of CD31^{+ thymic} naive CD4⁺ T cells (P < 0.0001; r = −0.549; Figure 2C ) and absolute numbers of CD31^{+ thymic} naive CD4⁺ T cells (P = 0.001; r = −0.416; Figure 2D ). NMOSD patients show significantly higher absolute numbers of CD31^{+ thymic} naïve CD4⁺ T cells (95% CI, -1285.531 – -121.624; P = 0.019; Table 2 ) than HC, whereas no changes were observed in the frequency of CD31^{+ thymic} naive CD4⁺ T cells. The frequency and absolute numbers of CD31^−central naive CD4⁺ T cells showed a significant positive correlation with age (P < 0.0001, r = 0.548; P = 0.011, r = 0.317; Figures 2E, F ). In addition, NMOSD patients displayed significantly higher absolute numbers of CD31^−central naive CD4⁺ T cells than HC (95% CI, −598.362 to −14.722; P = 0.04; Table 2 ).

We conclude that despite the absolute numbers of CD31^{+ thymic} naive CD4⁺ T cells and absolute numbers of CD31^-central naive CD4⁺ T cells showing opposite age-related trends, the absolute numbers of CD31^{+ thymic} naive CD4⁺ T cells and CD31^-central naive CD4⁺ T cells in patients seemed to increased proportionally, which may explain the lack of observed difference in the proportions. It is hard to draw conclusions about the NMOSD thymic output naive CD4 T cells changes, because only the absolute numbers were altered. In the absence of other parameters of thymic function, it may be controversial to assess thymus activity by thymic output naïve T cells alone.

Serum samples of 44 patients and 21 HC were collected to detect the level of IL-6. Notably, IL-6 levels in the serum of NMOSD patients were significantly higher than those in HC (P < 0.001; Figure 3A ). Moreover, partial correlation analysis after controlling for age and sex revealed a marked positive association between CD45RA⁺ CD62L⁺ naive CD4⁺ T cells proportion and IL-6 level (P = 0.048; r = 0.307; Figure 3B ), and the count of CD45RA⁺ CD62L⁺ naive CD4⁺ T cells increases with IL-6 level significantly (P = 0.008; r = 0.406; Figure 3C ). The subpopulation of CD45RA ⁺ CD62L ⁺ naive CD4⁺ T cells—CD31^{+ thymic} naive CD4⁺ T cells count increases with IL-6 significantly (P = 0.007; r = 0.408; Figure 3D ). We observed no significant correlation of CD31^−central naive CD4⁺ T cells with IL-6 level (data not shown).

The thymus gland has a unique morphology with a bilobed configuration, which can exhibit a pyramidal, triangular, arrowhead, or trapezoid shape on imaging. Therefore, morphological assessment of thymus requires dedicated methods specific for thymus (21). Parameters of thymus were measured as previously described ( Figure 4A ) (22, 23). Morphological evaluation was performed by an experienced technician who was blinded to the study using a thymic scoring system with a four-point scale (0-3) according to the proportion of fatty and soft tissues (24). CT density was measured by setting an oval region of interest covering the maximum area of the thymus gland with soft tissue density, excluding the surrounding mediastinal fatty tissue (22, 24). Predominant sides of thymus, density, and volume were analyzed blindly by two experienced radiologists. Linear and ordinal regressions were performed to examine the relationships of thymic predominant sides, volume, density, and thymic scores with age, sex, body mass index (BMI), and disease duration. For partial correlation coefficient, partial correlation analysis was performed after controlling for age or BMI.

Compared with HC, the trapezoid shape of the thymus gland was significantly more frequently observed in NMOSD patients (50% vs 32.3%; P = 0.044; Table 3 ). Representative axial CT images of the thymus with a different shape in NMOSD and HC were displayed in Figure 4B . The thymus of NMOSD patients exhibited distinct morphological changes characterized by significant increase in the left length, left thickness, right thickness, and transverse diameter of thymus ( Table 4 ). However, there were no significant between-group differences with respect to right length and anteroposterior diameter ( Table 4 ).

Partial correlation analysis showed a positive correlation between thymic predominant sides and BMI ( Figures 5A–D ). Given the significant positive correlation between BMI and age (r = 0.278, P = 0.002), partial correlation coefficient of these thymic predominant sides was calculated according to age and BMI. Compared with age, BMI showed a direct association with thymic predominant sides. In the mediation analysis, BMI was found to fully mediate the association between increase in thymic predominant sides with age ( Figures 6A–C ) and to partially mediate the association between density decline with age ( Figure 6D ). No significant correlation was observed between sex and thymic predominant sides.

Ordinal regressions were performed to examine the relationship of thymic scores with age, sex, and BMI. In thymus scoring, 25% NMOSD patients showed complete fatty replacement (score 0), and 50% NMOSD patients showed predominantly fatty thymus (score 1), whereas the proportion of HC group was 17% and 28%, respectively (P = 0.001, Figure 4C ). Thymic scores showed an inverse correlation with age (95% CI, 1.075–1.325; P = 0.001) and BMI (95% CI, 1.123–1.964; P = 0.006). We observed no significant correlation of thymic scores with sex.

Next, we compared the thymic density between the two groups after disaggregating by age, sex, and BMI. Age (P < 0.0001; r = −0.637; Figure 5E ) and BMI (95% CI, −4.109 to −1.271; P < 0.001; Table 4 ) showed a significant association with decline in thymus density. Thymic density in NMOSD patients was significantly lesser than that in the HC group (95% CI, 6.509–24.693; P = 0.001; Table 4 ). Moreover, partial correlation analysis after controlling for age and BMI revealed a marked association between thymic density and disease duration (P = 0.037; r = −0.278; Figure 5F ), but no significant correlation between thymic predominant sides and disease duration (data not shown). There was no significant correlation of thymus density with sex.

The increased thymic fat and decreased thymic density in NMOSD patients indicate exacerbation of age-dependent thymus involution as compared with HC, which explains the age-related declining trend of thymic function.

There was no significant difference between anti-AQP4-positive and anti–AQP4-negative patients, with respect to thymic predominant sides and thymic density or thymic score.

Owing to the difficulty in distinguishing thymus fat from mediastinum, use of CT to measure thymus volume is not straightforward. Especially in patients with complete fatty involution, precise delineation of the thymic borders is mostly impossible. We are not sure whether the volume of the thymus has definitively decreased (data not shown).

Fourteen NMOSD patients underwent chest CT scan and flow cytometric analysis of naive CD4 T cells subpopulations simultaneously. We assessed the association of thymic density, score, and predominant sides with naïve CD4⁺ T- cell subpopulation in these patients. As expected, the proportions and count of CD31^+thymic naive CD4⁺ T cells increased with thymic density and score ( Figures 7A, B, D, E ), since increase in thymic fat and decreased thymic density indicate the trend of decline in thymic function. The decline in CD31^+thymic naive CD4⁺ T cells was associated with the decrease in thymic function, which can be regarded as a direct marker of thymic output (17). In contrast, the proportions of CD31^-central naive CD4⁺ T cells declined with thymic density and score ( Figures 7C, F ). There was no significant correlation of CD31^-central naive CD4⁺ T cell count with thymic density or score. As the count of CD31^-central naive CD4⁺ T cells remains stable over time, this implies a peripheral regulation independent of thymic activity (18). We observed no significant correlation of CD45RA⁺ CD62L⁺ naive CD4⁺ T cells with thymic density or score.

The involution of thymus, characterized by disruption of thymic architecture and increase in adipocytes, contributes to the decrease in naïve T cell output. In thymic predominant sides, left thickness showed a significant association with the proportions and count of CD31^{+ thymic} naive CD4⁺ T cells and the proportions of CD31^−central naive CD4⁺ T cells ( Figures 8A–C ). Right thickness and transverse diameter showed a significant association with CD31^+thymic naive CD4⁺ T cell count ( Figures 8D, E ). There was no significant correlation of naive CD4⁺ T cells with thymic left length, right length, and anteroposterior diameter.

In the mediation analysis, both the thymic density and naive CD4⁺ T cells showed an inverse correlation age. In addition, the thymic density was found to fully mediate the relation between age and naïve CD4⁺ T cells ( Figures 7 and 8 ).