Faculty Bibliography

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Mothers giving birth to children with manifestations of neonatal lupus (NL) represent a unique population at risk for the development of clinically evident pathologic autoimmunity since many are asymptomatic and only become aware of anti-SSA/Ro positivity (anti-Ro+) based on heart block in their fetus. Accordingly, we hypothesized that the microbiome in saliva is associated with the development of autoreactivity and in some cases the progression in health status from benign to overt clinical disease including Sjögren's syndrome (SS) and systemic lupus erythematosus (SLE). The study comprised a clinical spectrum of anti-Ro+ mothers, all of whom gave birth to a child with NL: 9 were asymptomatic or had an undifferentiated autoimmune disease (Asym/UAS) and 16 fulfilled criteria for SS and/or SLE. Microbial diversity was reduced across all levels from kingdom to species for the anti-Ro+ mothers vs healthy controls; however, there were no significant differences between Asym/UAS and SS/SLE mothers. Relative abundance of Proteobacteria and more specifically class Betaproteobacteria decreased with clinical severity (healthy controls < Asym/UAS < SS/SLE). These ordered differences were maintained through the taxonomic hierarchy to three genera (Lautropia, Comamonas, and Neisseria) and species within these genera (L. mirabilis, N. flavescens and N. oralis). Biometric analysis comparing von Willebrand Factor domains present in human Ro60 with L. mirabilis proteins support the hypothesis of molecular mimicry. These data position the microbiome in the development of anti-Ro reactivity and subsequent clinical spectrum of disease.

Background/Purpose : The EULAR/ACR 2019 Classification Criteria for SLE have been validated in an international cohort of 696 SLE patients and 574 non-SLE patients with a sensitivity of 96.1% and a specificity of 93.4%. We comparatively evaluated the performance characteristics of the SLE classification systems in subsets of the validation cohort with regard to gender, race/ethnicity, and disease duration. Methods : 21 SLE expert centers from 16 countries submitted up to 100 SLE cases and 100 SLE mimicking controls each, using a standardized form without knowledge of the new criteria system to form the validation cohort. Cases and control diagnosis (SLE or not SLE) were independently verified by 3 SLE experts. The EULAR/ACR 2019 classifi-cation criteria validation cohort consisted of female (n=1,098) and male (n=172) patients; Asian (n=118), Black (n=68), Hispanic (n=124) and White (n=941) patients; and patients with an SLE duration of less than 1 year (n=34), 1-3 years (n=196), 3-5 years (n=157), and 5 or more years (n=879). Sensitivity and specificity with 95% confidence intervals (CI) were estimated for the EULAR-ACR 2019 criteria, the SLICC 2012 criteria and the ACR 1997 criteria. Results : As shown in Table 1, most of the point estimates for sensitivity and specificity in subsets lay within the 95% confidence intervals of the sensitivity and specificity of the EULAR/ACR 2019 criteria validation. In particular, sensitivity and specificity for all race/ethnicity groups were within the confidence intervals or even higher. Formally, the sensitivity was slightly lower for male patients, corresponding to a higher specificity, but the male 95% confidence intervals (0.86-0.98 for sensitivity, 0.90-0.99 for specificity) overlapped with those of the full cohort. Sensitivity appeared independent of disease duration at least from year 1 on, with all 95% confidence intervals overlapping (for the first year after diagnosis 0.52-1.00 for sensitivity, 0.69-0.97 for specificity). Conclusion : The point estimates of sensitivity and specificity suggest that the EULAR/ACR 2019 SLE classification criteria perform well in diverse race/ethnicity groups, in men and in early disease. These results now need to be independently validated in larger groups of African American/Black, Asian, and Hispanic patients, male patients and in early disease

Background/Purpose : Early treatment initiation in rheumatoid arthritis (RA) is fundamental to avoid chronic joint destruction and disability. Despite remarkable advances in RA therapeutics, oral methotrexate (MTX) remains the anchor drug and mainstay of treatment worldwide. However, MTX bioavailability has a wide inter-individual variability and >50% of patients with moderate or severe RA show no or suboptimal improvement in their symptoms in response to MTX. The reasons for these disparities in treatment response remain unclear. Prior studies have shown that the biotransformation of MTX is altered in germ-free and microbiome-depleted mice, prompting us to hypothesize that inter-individual differences in the human gut microbiome could impact drug bioavailability and thus clinical efficacy. We sought to determine differences in the microbiome of drug-naive, new onset RA (NORA) patients that could predict response to MTX therapy. Methods : We enrolled 27 drug-naive, NORA patients priori to MTX initiation (test cohort), and classified them as either MTX-responders (MTX-R; 39% of the cohort) or non-responders (MTX-NR; 61%) based on a stringent definition of clinical response (delta improvement of DAS28 >1.8 by month 4). We performed 16S rRNA gene and Shotgun Metagenomic sequencing on the baseline gut microbiomes of these NORA patients and confirmed the results in an independent validation cohort (n=31). NMR and LC-MS were performed in ex vivo incubations to measure the capacity of each NORA microbiome to metabolize MTX. Results : Our analysis revealed significant associations between the abundance of gut bacterial taxa and future MTX response. Patients that responded to therapy had significantly lower microbial diversity (p< 0.05). A significant difference in overall gut microbial community structure was also observed between groups (Bray-Curtis distance; PERMANOVA < 0.05). At the class level, we observed statistically higher abundance of Clostridia and lower abundance of Bacteroidia in MTX-NR (p< 0.05; q< 0.2). Furthermore, the baseline metagenome separated most MTX-R from MTX-NR (PCoA; PERMANOVA p< 0.05). We identified 8 microbial modules and 23 pathways, whose abundance significantly differed between groups (p< 0.05, q< 0.2), including genes related with purine and MTX metabolism, indicating a major difference in metabolic and biosynthetic potential between the microbiome of MTX-R and MTX-NR patients. Machine learning techniques were applied to this metagenomic data, resulting in a robust model based on bacterial gene abundance that accurately predicted response to MTX in an independent cohort. Finally, MTX available levels remaining after ex vivo incubation with distal gut samples from pre-treatment RA patients significantly correlated with the magnitude of future clinical response, suggesting a direct effect of the gut microbiome on MTX bioavailability and response to therapy. Conclusion : Together, these results provide the first step towards predicting response to oral MTX in NORA patients and support the utility of the gut microbiome as a prognostic tool and perhaps even as a target for manipulation in the treatment of rheumatic and autoimmune disease

Background/Purpose : The heterogeneity of the clinical manifestations of systemic lupus erythematosus (SLE) and lack of a diagnostic test make SLE difficult for epidemiologists to study. The Centers for Disease Control and Prevention (CDC) supported five population-based SLE surveillance registries, using harmonized methodology, to better estimate incidence and prevalence of SLE in diverse areas in the United States (US). Leveraging these data, we performed a meta-analysis to estimate the general prevalence of SLE in the US. Methods : The CDC registries were established in Michigan, Georgia, California, New York and the through the Indian Health Service (IHS). All registries used the 1997 revised ACR classification criteria for SLE as their primary case definition, and the surveillance time periods ranged from 2003 to 2009. Age-standardized prevalence was stratified by sex and race/ethnicity from the state-based registries; the American Indian/Alaska Native (AI/AN) estimate was based on the IHS registry that covered multiple states. For pooling data across the four sites with data on different racial/ethnic groups, we used Cochran ' s Q and I2 statistic to test for heterogeneity across sites. Due to significant heterogeneity, we used a random effects model to calculate pooled prevalence, which allows for more variation across sites. We then extrapolated to the 2017 Census population data according to sex and race-stratified groups, including data from the IHS registry, and summed the stratum-specific estimates to provide a total population estimate of SLE cases in the US. Results : The registries contributed 5,417 classified cases of SLE from a mix of urban and rural areas. From the metaanalysis of the four state-based registries, the overall prevalence was 72.8 (95%CI 65.3, 81.0) per 100,000 population. The prevalence among females was about 9 times higher than males (128.7 vs 14.6). In the meta-analysis, prevalence was highest among black females (230.9, 95%CI 178.2, 299.2), followed by Hispanic females (120.7, 95%CI 84.0, 173.4), white females (84.7, 95%CI 68.4, 104.8) and Asian/Pacific Islander females (84.4, 95%CI 48.3, 147.4). Among males, prevalence followed a similar pattern with the highest rates among black males (26.7, 95%CI 19.6, 36.4) followed by Hispanic males (18.0, 95%CI 15.6, 20.8), Asian/Pacific Islander males (11.2, 95%CI 5.7, 21.9), and white males (8.9, 95%CI 8.0, 10.1). The AI/AN prevalence estimates, which were not included in the meta-analysis, had the highest rates of SLE for both females (271, 95%CI 238, 307) and males (54, 95%CI 36, 77). Applying our sexand race-specific prevalence estimates to the corresponding population denominators from 2017 Census data, we estimated that 198,677 persons (179,186 females and 19,491 males) in the US fulfill ACR SLE classification criteria, Table 1. Conclusion : Using estimates from a coordinated network of population-based SLE registries, a more accurate prevalence estimate for the US was obtained. Our methods did not capture undiagnosed, incomplete, or other forms of lupus such as cutaneous lupus. Other case definitions may yield different results

Background/Purpose : Lupus nephritis (LN) complicates up to 60% of patients with systemic lupus erythematosus (SLE) and carries a high morbidity and mortality. The definitive diagnosis is based on kidney biopsy. This is invasive and not always readily available, thus delaying treatment. Sometimes multiple biopsies are required over the course of the disease. Importantly, while renal pathology is accurate at describing the morphology of renal disease, the underlying biology and molecular pathways are not thoroughly assessed. Urine proteomics is a non-invasive strategy that may provide insights regarding ongoing renal disease. Methods : One thousand proteins were quantified (RayBiotech Kiloplex assay) on a total of 112 longitudinal urine samples from 32 SLE patients with active LN and 7 healthy controls (HC) enrolled in the Accelerating Medicines Partnership (AMP). All patients underwent treatment as directed by their own physicians. Differentially excreted proteins at baseline (SLE vs HC, proliferative vs membranous LN, responders vs non responders) were identified using a linearmodel with moderated t statistic. Response to treatment was defined based on proteinuria at 1 year as complete (< 0.5g/24h) or partial (50% reduction but >0.5/24h). In the longitudinal analysis, a mixed model was employed to identify markers associated with proteinuria. Pathway enrichment analysis was performed using the genes coding for the differentially excreted analytes using Ingenuity Pathway Analysis (IPA) and other publicly available pathway libraries. Results : There were 186 proteins increased in SLE patients (Fig. 1). The most enriched pathway was TNFa (p< 0.001). We found 74 differentially excreted proteins comparing proliferative and pure membranous LN. CD4, MCP-1, MIP-1a, RANTES, IL-16, and IL-7, markers involved in CD4 T cell and monocyte biology, were enriched in proliferative disease. A few targets were exclusively identified in either class (i.e. CD4 in proliferative nephritis). We used a longitudinal model to identify specific urine proteins associated with worse proteinuria as a marker of severity. Proteinu-ria was associated with 105 markers (FDR < 0.05), the strongest association being CD163 (p = 10-9), a phagocyte marker. IPA implicated several pathways involving fibrosis, acute phase response, LPS/IL1, RXR, ICOS signaling and macrophage/fibroblasts (Fig. 2). Next, we identified 27 differentially excreted proteins in non-responders. IPA revealed that tretinoin, GM-CSF, TNF, and IL1 were among the top upstream regulators (Fig. 3). Conclusion : There is an inflammatory signature in the urine of patients with LN implicating monocyte and TNFa pathways. These signatures are associated with proliferative disease, worse proteinuria, and non-response to treatment. Of note, TNFa is involved in LN and has therapeutic potential. In phase 1 of AMP, monocytes were the main urine cell type identified by singe cell RNA sequencing in patients with LN. These results suggest that urine proteomics might identify and infer active pathological mechanisms in LN, paving the way for a more personalized approach to treatment. Further work in Phase 2 of AMP is being pursued to validate and extend these findings

Background/Purpose : The impact of renal injury in lupus nephritis (LN) is widespread with consequences to resident cells in other tissue beds, even non-lesional, non-sun exposed skin. Faithful reflection of a relevant renal tissue pathway in a more readily accessible compartment would allow for less invasive diagnostic alternatives. While ongoing studies are exploiting single cell RNA sequencing to link phenotype to biotype and identify cell specific pathways in the kidney, this study was initiated to address the hypothesis that these pathways may be reflected in uninvolved skin which is more likely to be serially biopsied. Methods : Single cell RNAseq was performed on cell suspensions prepared from ~2 mm punch biopsies of nonlesional, non-sun-exposed skin from the buttocks of 5 healthy controls, 4 SLE patients without LN and 18 SLE patients with proteinuria (with skin biopsies obtained within 24 hrs of the kidney biopsy). Histology revealed Class III ( n=6 ), Class III/V or IV/V mixed ( n=11 ), Class V ( n=1 ), and nephrosclerosis ( n=1 ). Dissociation of cryostored skin biopsies with collagenase and trypsin enzymes was followed by scRNA-seq using the 10x Genomics platform using V2 and V3 reagents. Results : We obtained 8,019 and 17,655 high-quality scRNA-seq profiles from single cell suspensions of control and SLE non-lesional, non-sun-exposed skin, respectively. A graph-based clustering method was applied and identified major clusters of cells as visualized by t-distributed stochastic neighbor embedding (tSNE). Differential gene expression analysis guided by established markers revealed these cell clusters as keratinocyte (KC), one smooth muscle cell cluster (SMC), fibroblast (FB), melanocyte (MEL), vascular endothelial cells (VEC), lymphatic endothelial cells (LEC), macrophages-dendritic cells (MAC-DC), T cells (TC) and sweat gland cells (SGC) (Figure 1A). Ranking cells by abundance, the result of the SLE skin cells was KC >FB >VEC >LEC, SMC, MAC-DC, TC, MEL and SGC. Overall, samples processed using the recent V3 single cell reagent kit showed higher genes and transcript captures compared to V2. However, these samples also captured more mitochondrial transcripts (Figure 1B). An analysis of gene expression changes in KC, SMC, and VSC from the LN patients versus controls demonstrated overexpression of interferon stimulated genes. However, the degree of interferon response varied in these cell types with KCs (basal KC, p=0.00312 and hair follicle KC, p=0.000012) showing the highest response followed by VECs (p=0.0043) and SMCs (p=0.0068). In addition to the interferon response signature, VECs from the LN patients also showed upregulation of MHC-II genes such as HLA-DRB5 and HLA-DRB1, suggesting increased antigen presentation capacity (Figure 1C). Conclusion : scRNA-seq identifies major skin cell types and further clustering identifies rarer cell populations. KCs, SMCs, and VECs from the skin of LN patients reveal diverse IFN response states and additionally VECs also show higher antigen presentation potential. The V3 upgrade of 10x Genomics single cell reagents capture more genes and UMIs per cell, but also higher mitochondrial content compared to the V2 version

Background/Purpose : Osteoporosis and bone fractures are a frequent cause of morbidity in systemic lupus erythematosus (SLE), and are felt to be related both to disease activity and glucocorticoid (GC) exposure. Dual energy X-ray absorptiometry (DEXA) is the standard tool to assess bone density, but it does not measure bone quality or strength and is not a robust predictor of fractures in SLE. Clinical 3T MRI scans have been shown to assess information about bone not captured by DEXA. This study aims to evaluate factors associated with bone structure measured by DEXA and MRI in an SLE cohort. Methods : DEXAs were performed on 31 women with SLE and 3T MRI of the non-dominant hip were performed on 29 of these cases. Results were associated with multiple demographic, clinical and laboratory measures. MRI parameters measured included trabecular plate width (PW), trabecular plate to rod ratio (PRR), plate volume fraction (PVF), rod volume fraction (RVF), trabecular bone thickness (Tb.Th), trabecular spacing (Tb.Sp) and trabecular network area (TNA). DEXA BMD was measured, and osteoporosis (OP) was defined as hip, spine or femoral neck Z score < -2.0 in premenopausal women, and T score < -2.5 in others, and low bone density (LBD) as Z score < -2.0 in premenopausal women and T score < -1.0 in others. Results : By DEXA, 8/31 (25.8%) had OP and 12 (38.7%) had LBD. History of lymphopenia (75.0% vs. 31.8%, p=0.049) and lower concurrent HCQ dose (340 vs. 400 mg, p=0.006) associated with DEXA OP, while older age (48.3 vs. 36.3 y, p=0.024) associated with LBD. Higher ESR was inversely correlated with favorable bone structure (PW r(22) = -.49, p=0.025, PRR rs = -.51, p=0.018, PVF rs = -.51, p=0.018, RVF rs = .51, p=0.018, Tb.Th rs = -.58, p=0.005, Tb.Sp rs = .44, p=0.046, TNA rs = -.50, p=0.022). Higher CRP was likewise inversely correlated with favorable bone structure (PW r(20) = -.61, p=0.004, PRR rs = -.57, p=0.009, PVF rs = -.57, p=0.009, RVF rs =.57, p=0.009, Tb.Th rs = -.56, p=.011, Tb.Sp rs =.67, p=0.001, TNA rs = -.64, p=0.002). A history of lupus nephritis was associated with unfavorable bone structure (PW 705.3 vs. 833.3 mum, p=0.048, PRR 6.6 vs. 8.1, p=0.024, PVF 0.83 vs. 0.89, p=0.024, RVF 0.17 vs. 0.11, p=0.024, Tb.Th 178.1 vs. 193.4 mm, p=0.012, Tb.Sp 358.6 vs. 296.5 mm, p=0.056, TNA 0.41 vs. 0.54 (1/mm), p=0.009). ESR, CRP and history of lupus nephritis were not significantly associated with DEXA hip BMD, OP or LBD. MRI parameters for favorable bone structure were inversely correlated with DEXA hip BMD (PW r(28) = -.47, p=0.011, Tb.Th rs = -.53, p=0.003) and BMI (PW r(28) = -.54, p=0.003, TbTh rs = -.72, p< 0.001, TNA rs = -.44, p=0.017). Conclusion : Higher ESR and CRP and a history of lupus nephritis associated with MRI parameters of unfavorable bone structure, but did not associate with DEXA abnormalities in SLE patients. MRI may be a more sensitive tool than DEXA to measure inflammatory effects on bone and potentially cumulative dose of steroid exposure. There were inverse correlations of MRI parameters with traditional osteoporosis risk factors and BMD measures on DEXA, and it is possible that each tool evaluates different aspects of bone health. Further evaluation of MRI screening for fracture risk in SLE and GC exposed individuals is warranted to better quantify risk and guide treatment

Background/Purpose: Both plasmacytoid dendritic cells (pDCs) and switched memory B cells (SMBCs) are considered to be key effector cells in systemic lupus erythematosus. It seems likely that within these classical cell lineages, additional diversity of function will exist that will contribute to disease pathogenesis. To explore this question, we performed single-cell RNA sequencing in pDCs and SMBCs from SLE patients and controls to assess gene expression patterns and cellular sub-groupings within these lineages. Methods : pDCs and SMBCs from SLE patients (n=10) and Healthy controls (n=5) were purified by magnetic separation. For deep sequencing, we used the Fluidigm C1 HT system with 800 capture site chips to capture single cells. Single cell capture was verified by direct visualization using the Array Scan system, allowing us to remove empty wells and wells with multiple cells. After quality control and adaptor trimming, the data was analyzed using SeqGeq software. pDCs and SMBCs were clustered using UMAP and pseudo-time analysis was performed using the Monocle program. Type I IFN activity in SLE plasma was measured using reporter cell assay. Results : A total of 2774 pDCs and 2578 SMBCs from SLE and healthy controls passed the quality control and were used for further analysis. In pDCs, we observed unique clusters for patients with high interferon, low interferon, and controls, indicating that the IFN response is a major determinant of overall gene expression patterns in SLE patient pDCs. IFN signature in pDCs correlated with circulating type I IFN activity in the SLE patients measured at the same time. Other genes upregulated in pDCs included the type I interferon regulator AXL and MACC1. The SMBCs were heterogeneous in patients and controls, and in contrast to the pDCs, the overall clustering pattern was independent of the IFN score. SMBC clusters were predominantly defined by genes indicating cellular activation or proliferation such as HLA-DRs and CREB1, or genes associated with nucleic acid processing such as DNASE1 and SNORD3B-1. Conclusion : We find distinct clusters of cells defined transcriptionally within the pDC and SMBC lineages, and the transcripts which define these subgroups differ between cell lineages. Type I IFN induced transcripts are important to pDC diversity, while in SMBCs transcripts related to cellular activation and nucleic acid processing are critical markers of transcriptional heterogeneity

Background/Purpose : Given the heterogeneity of systemic lupus erythematosus (SLE), the effect of any intervention is expected to vary. The ability to identify those most and least likely to benefit from a treatment would improve the interpretability of trial outcomes and advance medical care. Conventional subgroup analyses suffer from low power, can encompass only a few variables at a time, and require a priori specification of cut-points for continuous variables. We explored the utility of a machine learning-based algorithm for discovering in a SLE trial the subgroups in which adding experimental therapy to standard of care considerably enhances or diminishes response compared to placebo (PBO). Methods : A two-step virtual twin (VT) method was applied to combined data from the BLISS-52 (N=865) and BLISS-76 (N=819) trials. A random forest algorithm was first used to estimate for each patient, given baseline characteristics, the probabilities of SRI-4 response to belimumab and PBO. A regression tree was then constructed to partition the study population into distinct subgroups and identify those in which the estimated difference in these response probabilities is much greater or smaller than the treatment effect in the overall population. Two separate VT analyses were conducted of the 10 mg/kg and 1 mg/kg belimumab doses compared to PBO. Cross-validation was used to assess the method ' s performance. Results : In the combined BLISS trials, response rates to the primary endpoint (SRI-4) were 51% in those receiving 10 mg/kg belimumab, 46% (1 mg/kg), and 39% (PBO). VT analysis of 10 mg/kg vs. PBO found a 23% belimumab response advantage over PBO in patients with SLEDAI >= 7 & steroid dose >= 4 mg/d & low C4 & no BILAG A at baseline , vs 12% in the total population. In contrast, the estimated response difference in those entering with SLEDAI < 7 & normal C4 was 5% lower on 10 mg/kg than PBO. In analysis of 1 mg/kg vs. PBO, two subgroups showed enhanced belimumab effect: SLEDAI >= 8 & steroid dose >= 19 mg/d and SLEDAI >= 8 & steroid dose < 19 mg/d & BLyS >= 1.9 ng/mL ; average estimated between-treatment response differences were 18% and 14%, respectively, compared to 7% in the overall population. But in patients with SLEDAI < 8 & steroid dose < 16 mg/d & age < 43 , the 1 mg/kg belimumab response rate was estimated to be 7% lower. Cross-validation indicated the accuracy of the VT method to identify subgroups exceeded 70%. Conclusion : Enhanced belimumab response was associated with low C4 and higher disease activity, steroid dose, and BLyS levels, as in prior studies. However, the VT method identified alternative cutpoints for continuous variables and additional features predicting non-response. SLEDAI >= 7 or 8 was most predictive of response to treatment. Thus, lower response difference is identified in patients who are potentially too ill (BILAG A severity) or not ill enough (minimal disease criteria) to benefit from adding belimumab. The 1 mg/kg belimumab effect was enhanced only in those on high baseline steroid doses. The VT and other machine learning techniques are promising for subgroup discovery in SLE trials as more sophisticated biomarkers, especially potent but less common indicators, become available