Discussion
We have identified several significant and mechanistically interesting changes in the urine proteome in participants in the BLISS-LN trial. To our knowledge, this is the first study of the LN urine proteome in the context of a successful LN clinical trial, and likewise the first study to examine the effect that a specific medication—belimumab, in this case—may have on urine proteomic patterns in LN. With respect to complete renal response, we have confirmed the findings of urine proteomic analyses in at least two other independent studies2 4 and thereby present additional evidence that the development of a multidimensional urinary biomarker to predict disease response in LN may be feasible. Furthermore, given the identified urinary signature associated with belimumab therapy, this work shows that treatment with specific medications can be detected in the urine proteome, and this may inform drug monitoring strategies.
With respect to renal response, longitudinal reduction in urinary CD163 abundance was strongly associated with complete response (per the definition used in the Methods) in the BLISS-LN trial regardless of treatment arm. Urinary CD163 is a macrophage marker which correlates with histological activity index on renal biopsy and has been previously associated with renal response to LN therapy in at least two independent studies, namely our prior findings in the Accelerating Medicines Partnership in Rheumatoid Arthritis/SLE4 and also those of Mejia-Vilet et al.2 As shown in figure 1, reduction in activated leucocyte cell adhesion molecule (ALCAM) and interferon beta (IFNb) were also seen in renal responders versus non-responders. Urinary ALCAM is increased in patients with SLE with nephritis, and increased concentrations of such have been associated with increased activity index on renal biopsy10 and an increased risk of decline in estimated glomerular filtration rate (eGFR).11 With respect to IFNb, this would also appear to play a role in lupus pathophysiology. A single-cell RNA sequencing analysis of renal biopsies in LN demonstrated that type I interferon response signatures were characteristic of LN and associated with failure to respond to immunosuppressive therapy.12 Anifrolumab, a type I interferon receptor antagonist, was recently approved for the treatment of SLE by the US Food and Drug Administration, though patients with severe LN were excluded from the preceding clinical trials.13 14 The phase II clinical trial of anifrolumab for LN failed to meet its primary endpoint, but a subset receiving high-intensity anifrolumab did have numeric benefit.15
These changes in the urine proteome with complete renal response were identified irrespective of belimumab therapy. We did not identify an obvious unique urinary signature predictive of response to belimumab specifically, though our small sample size limited any robust analysis along these lines. Nevertheless, these findings provide further evidence that dynamic urinary biomarkers may be used to monitor response to LN treatment in clinical practice and clinical trials.
With respect to the urinary proteomic signature of belimumab therapy, significant reduction in urinary CD23 by week 24 was most characteristic of belimumab therapy versus standard of care. CD23 is the low-affinity receptor for IgE on B cells and also regulates IgE synthesis. CD23+ B cell-mediated antigen presentation of IgE-antigen complexes has been implicated in the enhancement of antibody and CD4+ T-cell responses to said antigens.16 Anti-dsDNA IgE is common in SLE and is associated with active LN and worsened disease activity overall.17 Thus, modulation of this immune enhancement pathway might be an important mechanism of belimumab therapy. We note, however, that reduction in CD23 was not directly associated with response in this analysis, though this could be a reflection of our relatively small sample size. Further study is needed to determine whether modulation of CD23-mediated pathways might be directly implicated in LN treatment response.
It is important to note that the breadth of belimumab’s effect on the urine proteome may not be fully captured in this analysis due to the timing of urine sample collection in relation to renal biopsy and treatment initiation. In BLISS-LN, renal biopsy occurred up to 6 months before randomisation (week 0), and induction therapy with non-belimumab immunosuppression could occur up to 2 months before randomisation. In our group’s experience, urinary proteomic profiles shift significantly within the first 3 months after any treatment initiation.4 18 Thus, because of the lag time between biopsy and/or induction and week 0, initial proinflammatory urinary signals were possibly already blunted by other immunosuppressive therapy, including corticosteroids, by the time belimumab was added. This aspect of the BLISS-LN trial design also limited our ability to correlate urinary proteomic changes with renal biopsy features, as week 0 and renal biopsy were not timed to one another. As mentioned previously, urinary CD163 and ALCAM are already known to associate with renal biopsy features.2–4 10 11
There were several other limitations to this study. Only a relatively small subset of participants in the BLISS-LN trial was included in this analysis due to the need for complete data from multiple time points. We have presented a descriptive, exploratory analysis of urine proteomic changes—larger cohorts and independent datasets are needed in order to develop predictive, multidimensional (rather than single protein) biomarkers to better prognosticate and monitor disease activity. Indeed, it is inherently challenging to determine whether any single urinary protein is superior to proteinuria alone in predicting treatment response, as proteinuria itself is considered the ‘gold standard’ in current definitions of response. Additionally, individual urinary proteins are co-correlated with total proteinuria (as would be expected), though it has previously been demonstrated that total proteinuria only explains 40% of the variability in urinary CD163 and its association with treatment response.2 Within our group, additional studies are under way exploring these co-correlations and early prediction of treatment response using individual or multidimensional urinary biomarkers. Ultimately, long-term studies focused on eGFR preservation are needed to establish the utility of these biomarkers.19
Another limitation of this work is that the original outcome assignments in BLISS-LN could not be obtained, though in this case we constructed a definition of complete renal response in line with other clinical trials in LN. Finally, longitudinal renal function data—specifically, eGFR at baseline and over time—was also not available and could not be obtained. A follow-up analysis of BLISS-LN patients demonstrated that belimumab can slow eGFR decline in patients with LN as compared with standard therapy alone,20 but as these data were not available to us, we were unable to correlate urine proteomic changes to eGFR trends. Further research with larger cohorts of patients is needed to better understand how to use urine proteomic trends to predict long-term renal outcomes.