Metabolomic profiling of patients with lupus nephritis reveals unique metabolites that are modulated through type I interferon inhibition by anifrolumab treatment in a phase 2 trial

Abstract

Objective: Patients with systemic lupus erythematosus (SLE) commonly develop lupus nephritis (LN), the most frequent severe organ manifestation of this systemic autoimmune disease. Using serum and urine samples from a phase 2 trial in LN, we investigated that how the LN metabolome is modulated in response to type I interferon receptor blockade with anifrolumab, an approved treatment for moderate to severe SLE.

Methods: Patients in TULIP-LN (NCT02547922) received standard therapy plus intravenous anifrolumab or placebo. Untargeted metabolomics analysis was performed on serum and urine samples from 128 and 119 patients, respectively. Metabolites impacted by anifrolumab and their associations with clinical, serological and kidney measures of LN disease activity were examined. An in vitro model was used to validate the impact of anifrolumab on metabolite-induced inflammation.

Results: In serum, indoxyl sulfate (IS) and cytosine were the metabolites most modulated by anifrolumab, while baseline levels correlated with measures of kidney damage. In urine, uracil and cytosine were the most modulated by anifrolumab and levels of these metabolites were associated with serological markers of disease activity. Anifrolumab reduced markers of vascular dysfunction and inflammation upregulated by IS in in vitro models. Baseline urine uracil levels predicted response to anifrolumab intensive regimen at Week 52.

Conclusion: Our findings establish a connection between type I interferon signalling, pyrimidine metabolism and uremic toxins in patients with LN and support the evaluation of urine uracil as a potential biomarker of response to anifrolumab.

Keywords: Autoimmune Diseases; Biological Therapy; Lupus Nephritis.

Figure 1. Metabolomic characterisation of response to anifrolumab in serum. (A) QQ plot illustrating the treatment–time interaction term from linear mixed-effects model of serum metabolites impacted by anifrolumab IR versus placebo from baseline to Week 24. Longitudinal trajectories of (B) IS and (C) cytosine levels for patients at Weeks 0, 12, 24 and 52 receiving anifrolumab BR (dark blue; baseline, n=41, 37 36 and 32), anifrolumab IR (light blue; baseline, n=44, 44, 44 and 32) or placebo (grey; baseline, n=43, 39, 38 and 37), shown as mean change from baseline (±SEM). (D) Composite uremic toxin score (IS, asymmetric dimethylarginine, homocitrulline, phenylacetylglutamine, urate and p-cresol sulfate) at Weeks 12, 24 and 52 for patients receiving anifrolumab BR (dark blue), anifrolumab IR (light blue) or placebo (grey), shown as uremic toxin SingScore (±SEM), a metric of metabolite set enrichment analysis. (E) IS and (F) cytosine levels at Weeks 12, 24 and 52 in CRR responders and non-responders, per treatment group, shown as mean change from baseline (±SEM). BR, basic regimen; CRR, complete renal response; IR, intensive regimen; IS, indoxyl sulfate.

Figure 2. Serum IS and cytosine correlated with baseline clinical measures of LN disease. (A) Spearman correlation of baseline serum IS and cytosine levels with measures of kidney damage and of LN disease. Comparison of normalised expression of serum IS (upper panels) and cytosine (lower panels) according to the various grades of kidney biopsy scoring shown as log₁₀ peak area: (B) endocapillary hypercellularity; (C) glomerular leucocyte infiltration; (D) globally and segmentally sclerotic glomeruli and (E) wire loops/hyaline thrombi. P values were determined using one-way ANOVA and represent the distribution as a whole. 21-gene type I IFNGS; logfold change in blood 21-gene type I interferon gene signature; ANOVA, analysis of variance; dsDNA, double-stranded DNA; eGFR, estimated glomerular filtration rate; IS, indoxyl sulfate; LN, lupus nephritis; NIH, National Institutes of Health; SLEDAI-2K, systemic lupus erythematosus disease activity index 2000; UPCR, urine protein–creatinine ratio.

Figure 3. Impact of IS treatment on injury and inflammation markers in an in vitro model of HGMECs. In vitro assessment of the impact of IS doses (200 μM, 1 mM and 2 mM) on the expression of injury markers. (A) VCAM1, ICAM1 and inflammatory markers and (B) CXCL2, IL-6 and CXCL8, shown as fold gene expression. (C) Impact of anifrolumab (1 μg/mL or IgG controls) and LPS (1 μg/mL) treatment on IS (2 mM) induced expression of VCAM1, IL-6 and IFI27, shown as fold gene expression. Experiments included LPS pretreatment (100 ng/mL or 1 μg/mL) of cells to mimic an inflammatory environment. Statistical analysis was performed using two-way ANOVA with Dunnett’s post hoc test (A and B) and a one-way ANOVA with Bonferroni post hoc test (C). **p<0.01; ***p<0.001 and ****p<0.0001. Ani, anifrolumab; ANOVA, analysis of variance; Ctrl, IgG control; HGMECs, human glomerular microvascular endothelial cells; IS, indoxyl sulfate; LPS, lipopolysaccharide.

Figure 4. Metabolomic characterisation of response to anifrolumab in urine. (A) QQ plot illustrating the treatment–time interaction term from linear mixed-effects model of untargeted urine metabolites. (B) Longitudinal trajectories of urine cytosine and uracil levels for patients at Weeks 0, 12, 24 and 52 receiving anifrolumab BR (dark blue; baseline, n=38, 34, 34 and 30), anifrolumab IR (light blue; baseline, n=43, 43, 42 and 40) or placebo (grey; baseline, n=38, 36, 35 and 35), shown as mean change from baseline (±SEM). (C) Comparison of urine levels of uracil (left panel) and cytosine (right panel) at baseline in patients from TULIP-LN (blue; n=119) compared with healthy donors (green; n=40), shown as log₁₀ peak area. p<0.001 as determined by Wilcox test. (D) Spearman correlation of baseline urine cytosine and uracil levels with measures of kidney damage and LN disease. 21-gene type I IFNGS, logfold change in blood 21-gene type I interferon gene signature; BR, basic regimen; dsDNA, double-stranded DNA; eGFR, estimated glomerular filtration rate; IR, intensified regimen; LN, lupus nephritis; NIH, National Institutes of Health; SLEDAI-2K, systemic lupus erythematosus disease activity index 2000; UPCR, urine protein–creatinine ratio.

Figure 5. Identification of metabolomic predictors of response to anifrolumab. Forest plots showing log odds and 95% CIs for (A) aCRR and (B) CRR at Week 52 following anifrolumab IR treatment. High and low metabolite levels were determined by splitting at the median, and positive log odds favour anifrolumab IR. Adjusted p values are shown in the tables. (C) Predicted probability of achieving aCRR at Week 52 according to baseline urine uracil levels in anifrolumab IR and placebo groups. (D) Predictive performance (receiver operating characteristic curve) assessing the specificity and sensitivity of urine uracil of achieving aCRR at Week 52 of anifrolumab IR treatment. aCRR, alternative CRR; AUC, area under the curve; CRR, complete renal response; IR, intensive regimen.