The generation of GILZ-deficient mice has been previously described (21, 22), as has the generation of Lyn-deficient C57Bl/6 mice (18, 20). All GILZ-deficient mice are male, since GILZ is X-linked and renders male mice sterile (22) and female GILZ-deficient mice cannot be bred. Mice were housed in specific pathogen-free conditions. We generated GILZ-deficient mice on a Lyn-deficient background (GILZ/Lyn^-/-) and compared them to wildtype (WT), GILZ knockout (GILZ^-/-), and Lyn knockout (Lyn^-/-) mice. WT and GILZ^-/- mice were co-housed as littermate controls, and Lyn^-/- and GILZ/Lyn^-/- animals were co-housed as littermate controls. All animals were housed in identical conditions in adjacent cages within our facility. Animals in experimental groups were carefully age and sex matched, but with ages pooled into <150 days or >200 days old so as to adequately power our statistical analyses. All procedures were approved by the Monash Medical Centre Animal Ethics Committee.

Mouse spleens were harvested and spleen weights were measured in milligrams (mg). Curves were fit to a one-phase decay model with robust regression using GraphPad Prism Software 7.0b. The age spans of mice chosen to display in Figure 1B were 70-84 days (10-12 weeks) and 245-315 days (35-45 weeks) so as to ensure adequate sample size and no significant differences in age that may alter the interpretation of the results. The statistics for the age groups are shown in Table 1 .

A Kruskal Wallis test followed by Dunn’s correction for multiple comparisons found no significant differences between any age groups in the 70-84 day old cohort. In the 245-315 day cohort, there was a significant difference, (GILZ^-/- mice significantly were older than GILZ/Lyn^-/- mice, P=0.03), and no other significant differences were detected between ages of mice across genotypes.

Serum was isolated from whole blood obtained by cardiac puncture following CO₂ asphyxiation. Concentrations of BAFF, IFNγ, IL-10, IL-1α, IL-1β, IL-18, IL-12p70, IL-6, IL-23A and IL-17A, were measured in serum by Luminex as described in the manufacturer’s protocol (Customized mouse 11-plex ProcartaPlex Kit, Invitrogen).

Serum samples were used at 1:200 dilution for the identification of antibodies to extractable nuclear antigens (ENA) by flow cytometry. The FIDIS Connective Profile MX 117 kit (Theradiag) was used as per the manufacturer’s instructions with the variation of the use of a PE-conjugated anti-mouse IgG F(ab’)2 secondary antibody (eBioscience) for flow cytometry (23).

Kidneys from 8 to 45-week-old male mice were frozen in Optimal Cutting Temperature (OCT) compound (Tissue Tek). Frozen sections were stained with PE-conjugated polyclonal anti-mouse IgG F(ab’)2 secondary antibody (eBioscience) and FITC-conjugated goat IgG fraction to detect mouse complement component C3 (MP Biomedicals). Formalin fixed sections were stained with Haematoxylin and Eosin or periodic acid-Schiff (PAS) stain. Glomerular segmental necrosis was defined as acellular areas of PAS positive staining (24).

Glomerular damage was scored in a blinded manner for segmental necrosis (shown as percentage of glomeruli affected) and crescent formation as described (25). IgG and C3 immunofluorescence staining were given an intensity score between 0–3, and an average of 20 glomeruli were scored for each mouse.

To assess GILZ expression in healthy subjects, we mined the publicly available datasets GSE123698 and GSE69832 for the probes corresponding to GILZ (gene name TSC22D3), as shown in Supplemental Figure 2 (26, 27). To measure gene expression in patients with SLE, we mined the dataset GSE88884, which contains gene expression data from PBMC from n = 60 healthy subjects and n = 1,760 SLE patients enrolled in the ILLUMINATE-1 and ILLUMINATE-2 tabalumab phase III clinical trials (28, 29), taken at baseline. We extracted the data for the probes TC0X001262.hg.1:1, TC0X001262.hg.1:2 and TC0X001262.hg.1:3 that identified GILZ (gene name TSC22D3). Relative expression values for GILZ were determined by averaging the three probe set values for each subject, without batch correction, as with previous analyses. The strength of associations between GILZ and cytokine expression were determined based on Spearman’s rank order correlation. Clinical data were kindly provided by Dr Robert Hoffman of Eli Lilly.

All analyses and data visualization were performed using GraphPad Prism Software 7.0b and R software (version 4.0.2). Continuous and categorical data are presented as mean (standard deviation (SD)) or median (interquartile range) (IQR) and frequencies (percentage), respectively, according to data distribution. Choice of parametric/non-parametric test was guided by the assessment of continuous data distribution, as examined by Shapiro-Wilk and Kolmogorov-Smirnov tests. Correlation between variables was analyzed using Spearman’s rank correlation test. Difference in continuous variable between 2 or more than 2 groups were examined using Student’s unpaired two-tailed t-test with Welch’s correction, or Kruskal-Wallis test followed by Dunn’s multiple comparison test, respectively, where appropriate. A P value of < 0.05 was deemed statistically significant.

Splenomegaly occurs in some patients with active SLE and has also been described in both Lyn^-/- and GILZ^-/- mice (30). In Lyn ^-/- mice, splenomegaly is driven by the accumulation of plasma cells and Mac-1⁺ lymphoblasts, from as early as 12 weeks of age (31). GILZ-deficient mice develop splenomegaly as they age and isotype-switched plasma cells in spleen are increased (12). We therefore examined whether Lyn and GILZ synergized to regulate the processes that cause splenomegaly. We observed that Lyn-deficient mice developed splenomegaly at an early age, more so when GILZ was deficient, and spleen size progressively increased with aging ( Figures 1A, B ). While a Kruskal-Wallis test followed by Dunn’s correction for multiple comparisons did not detect a statistically significant difference between Lyn^-/- and GILZ/Lyn^-/- mice of 35-45 weeks of age when all genotypes were included in the analysis ( Figure 1B ), a Student’s t-test between these two genotypes showed a significant difference (P=0.0077). We did not observe splenomegaly in GILZ-deficient mice ( Figures 1A, B ), in contrast to our previous report (12), which is likely due to a change in housing conditions between these studies, as microbiota and cohousing may affect the phenotypes (32). A detailed analysis of the microorganisms detected within sentinel animals in the same room in which our colonies were housed in shown in Supplemental Figure 1 , although we did not test whether the changes impacted on the development of an autoimmune phenotype in the GILZ^-/- strain. However, splenomegaly observed in Lyn-deficient mice was hastened in the absence of GILZ, and was worsened over time ( Figures 1A, B ). This detailed kinetic study of the degree of splenomegaly confirms that GILZ deficiency is permissive of the development of autoimmunity triggered by underlying genetic factors.

GILZ has recently been reported to decrease in murine macrophages as a function of the mild inflammatory state that develops with age (33). We previously showed that Lyn^-/- mice expressed lower GILZ, like patients with SLE, and this deficiency worsened with age (12). In the Lyn^-/- model, the development of autoimmunity is exacerbated over time and severity increases with age (34). However, in wildtype mice (12), and in peripheral blood mononuclear cells from healthy subjects (26, 27) and patients with SLE (35), we detected no meaningful change in GILZ expression that occurred over the lifespan ( Supplemental Figure 2 ). Thus, the exacerbation of the Lyn^-/- phenotype with age, that occurred as a result of GILZ deficiency, was likely due to factors other than an age-related effect.

Morbidity in the Lyn^-/- model is associated with the progressive development of renal injury. To determine GILZ whether deficiency worsens glomerulonephritis in the Lyn^-/- model, we compared the development of glomerulonephritis in Lyn^-/- mice with that in mice lacking both Lyn and GILZ ( Figure 2 ). We observed that aged GILZ/Lyn^-/- mice had more severe glomerulonephritis compared to Lyn-deficient mice, featuring significantly increased segmental necrosis and glomerular crescents on PAS staining of kidney sections ( Figure 2A with representative images shown in Figure 2B ), and reduced kidney size ( Figure 2C ). Therefore, kidney damage in Lyn^-/- mice was exacerbated by GILZ deficiency.

We next explored potential explanations for the heightened autoimmune phenotype observed in Lyn-deficient mice when GILZ was also deficient. The hyperactive B cells in Lyn ^-/- mice (20) and loss of B cell quiescence and tolerance we previously reported in GILZ^-/- mice (12) each result in the development of lupus-like autoimmunity. As a result, both strains produce ANA, and antibodies to ENA (dsDNA, histone, Smith (SM) and U1RNP) (12, 18). We sought to determine whether GILZ deficiency worsened the lupus phenotype of Lyn^-/- mice via exacerbation of autoantibody-mediated autoimmunity. While both Lyn^-/- and GILZ/Lyn^-/- mice expressed autoantibodies at a young age ( Figure 3A ), anti-dsDNA antibodies were present at significantly higher concentrations in the aged GILZ/Lyn^-/- mice compared to Lyn^-/- mice (>200 days) ( Figure 3B ); no other autoantibodies were significantly affected in aged GILZ/Lyn^-/- mice. Thus, while GILZ deficiency did not alter overall serum autoantibody production in Lyn^-/- mice in this study, it did significantly impact on an autoantibody specifically associated with nephritis in human SLE.

Previous studies report that Lyn-deficient mice develop glomerular immune complex deposition as early as 6–8 weeks of age and cumulative damage to kidneys subsequently worsens with ageing (18, 31). We previously reported that GILZ^-/- mice also develop mild immune complex-mediated glomerulonephritis as they age (12). In keeping with the generally equivalent levels of autoantibodies in Lyn^-/- and GILZ/Lyn^-/- mice, but in contrast to the higher levels of nephritis-associated anti-dsDNA antibodies, we detected negligible differences in glomerular immune complex deposition between mice of these two strains. While young mice (< 150 days old) of both Lyn^-/- and GILZ/Lyn^-/- strains showed deposition of C3 and IgG in the glomeruli, there was no significant difference in C3 deposition between the two genotypes ( Figure 4A and quantified in Figure 4B ). In aged mice, immunofluorescence staining of kidneys for C3 and IgG showed immune complex deposition in both Lyn^-/- and GILZ/Lyn^-/- mice, with no significant difference in the extent of C3 deposition observed ( Figure 4C ). Rather, there was a trend toward reduced IgG staining in glomeruli of aged GILZ/Lyn^-/- mice ( Figure 4C ). These observations suggest that GILZ deficiency worsened autoimmune disease outcomes in the Lyn^-/- model via processes other than through effects on the degree of autoantibody immune complex deposition in kidneys.

To further investigate the basis of the exacerbation of organ damage in GILZ/Lyn^-/- mice, we tested whether GILZ deficiency altered cytokine production in Lyn^-/- mice. Multiple cytokines were elevated in serum of young GILZ/Lyn^-/- mice, with substantial variance between individual mice as observed in human SLE. BAFF, IFNγ, IL-10, IL-1α, IL-1β, IL-6 and IL-23A all trended non-significantly numerically higher in young GILZ/Lyn^-/- mice than in Lyn^-/- littermate controls, and elevations in IL-18, IL-12 and IL-17A were significant ( Figure 5A ). In aged mice, in which the autoimmune phenotype was well established, these early significant differences were all attenuated ( Figure 5B ).

To assess whether low GILZ expression is permissive of cytokine expression in human SLE, we mined the dataset GSE88884, which contains microarray data from PBMC from n =1,760 SLE patients. It is important to note that this study was limited to mRNA levels in PBMC rather than absolute amounts of circulating cytokines. Relationships with GILZ expression were strongest with IL23A, where a negative correlation was present in SLE patients ( Figure 6A ).

GILZ is a glucocorticoid-induced gene, and we assessed whether GILZ expression and its associations with cytokines were altered by glucocorticoid exposure in SLE patients. Healthy subjects (n = 60) and SLE patients not taking glucocorticoids (n = 460) had normally distributed and similar levels of expression of GILZ, with median, mean (± standard deviation) of 7.769, 7.727 (± 0.113) and 7.750, 7.749 (± 0.156) respectively. GILZ expression in PBMC from SLE patients taking glucocorticoids (n = 1293) was median and mean of 7.771 and 7.795 (± 0.211), and was significantly higher than in patients not taking glucocorticoids (P=0.002; Figure 6B ). Since increased GILZ expression by glucocorticoids treatment could potentially obscure associations, we analyzed the subset of 460 patients not taking glucocorticoids. The level of IL23A was moderately negatively correlated to the expression of GILZ in both subsets of patients taking and not taking glucocorticoids ( Figures 6C, D ). Other cytokines including BAFF and IL18 were negatively correlated with GILZ expression in non-glucocorticoid-using patients, although the relationship between GILZ and BAFF did not reach significance (with P = 0.06) ( Figure 6C ).