19 patients diagnosed before age 18 who fulfilled the American College of Rheu- matology criteria for the classification of SLE [13] and age-matched healthy volunteers were included into the study. Patients and healthy volunteers were recruited locally by word of mouth. Demographic characteristics, disease activity and renal disease status, and immunomodulatory medications of all patients are listed in Table 1. Control group demographic data other than age (range 11.8 - 18) was not available. Peripheral blood, and clinical and laboratory data were collected for each patient and disease activity determined according to the modified SLEDAI Index 2000 [14] . Disease activity scores of five or greater were considered representing active, scores of less than five inactive disease [15] [16] . Sample procurement from jSLE033-122 was between 042006 and 062012; from CIIT- 1001-20 between 082015 and 012016. Serial samples obtained at times of high or

^aTherapeutic regimens including prednisone > 2 mg/kg/day and/or mycophenolatemofetil (MMF), cyclophosphamide (Cyc), or azathioprine (AZT) were considered immunosuppressive. HCQ = hydroxychloroquine. ^bNo SLEDAI available; clinically no disease activity, 2 reflects reduced complement levels.

low disease activity were available from three patients. PBMC were isolated as described [3] . Paired urine and plasma samples were collected from an additional 5 jSLE patients (Table 1). All activities were carried out in accordance with the Declaration of Helsinki and approved by local Institutional Review Boards. All subjects provided informed written consent.

PBMC staining and analysis were as described [3] . Unless otherwise specified, mAbs were from BD Pharmingen and included anti-CD3-Alexa Fluor 700 (UC- HT1), anti-CD4-APC (RPA-T4), anti-CD8-PacificBlue (RPA-T8), anti-CD11c- FITC (BU15, Thermo Scientific), anti-CD14-APC (RMO52, Beckman Coulter), anti-CD16-PacificBlue (3G8), anti-CD19-APC (HIB19) and anti-CD20-FITC (L27), anti-CD56-APC (N901, Beckman Coulter), and anti-NKG2D-PE (1D11; [2] ). Anti-NKG2DL mAb were anti-MICA/B-PE (6D4), anti-ULBP1-PE (170818, R & D Systems), anti-ULBP2/5/6-PE (165903; R & D Systems), anti- ULBP3-PE (166510; R & D Systems), and anti-ULBP4-PE (1H1; [12] ). In some cases, viable cell numbers were limiting resulting in partial data sets.

Monocytes, enriched from either patient or healthy donor PBMC using classic plastic adherence, were incubated with RMPI-1640/10% FBS supplemented with LPS (10 μg/ml; Sigma) or GM-CFS (20 ng/ml; R & D Systems), or with medium alone, for 24 hours and examined for NKG2DL expression by flow cytometry.

Soluble MICA (sMICA), soluble MICB (sMICB), and soluble ULBP1 (sULBP1) in jSLE patient plasma were determined using ELISA kits (Human MICA and Human MICB; Abcam; Human ULBP-1, R & D Systems) according to manufacturers’ instructions.

Populations were compared using the two-sample t-test and significance assigned where p < 0.05. Contribution of more than one sample by some subjects was accounted for using multivariate logistic generalized estimating equations (GEE). The correlation between NKG2D⁺CD4 T cell frequencies and soluble NKG2DL was estimated from linear regression.

Our previous observation of lupus-associated NKG2D⁺CD4 T cell population expansions was made in patients with the juvenile-onset form of this disease [3] . For consistency, we maintained this disease focus enrolling 19 jSLE patients and 20 healthy age-matched control (HC) donors. Flow cytometry of PBMC samples (altogether 22, since each two serial samples were available from three patients; Table 1) confirmed increased frequencies of NKG2D⁺CD4 T cells among the jSLE cohort compared to HC (10.82% ± 8.72% versus 3.64% ± 2.34; p ≤ 0.00001) with larger proportions of these cells in inactive compared to active disease (15.99% ± 10.48% versus 6.12% ± 3.47; p = 0.0007).

In healthy individuals, peripheral blood B cells provide the autoantigens and NKG2DL necessary for NKG2D⁺CD4 T cell proliferation [3] . Screening for NK- G2DL expression in jSLE thus first focused on the peripheral B cell compartment. Lupus patients, adult and juvenile-onset alike, are lymphopenic and, com- pared to healthy individuals, have reduced absolute B cell counts [17] . Accommodating these contractions and the typically small volumes of pediatric blood draws we started out with an exploratory flow cytometry analysis that defined B cells based only on CD19 and CD20 coexpression without consideration of additional functionally relevant markers. NKG2DL were tested as a third parameter using mAbs to MICA/B, ULBP1, ULBP2/5/6, ULBP3, and ULBP4 (Supplementary Figure S1 displays examples of gating strategy and primary data). CD19⁺CD20⁺ B cells were present in all but one (jSLE114) of the 22 jSLEPBMC samples with absolute counts lower than those described for healthy children (219.8 ± 282.4 cells/μl; [17] [18] ). Relative proportions of CD19⁺CD20⁺ cells were larger than those of the control donors (mean 13.6% ± 11.3 versus 7.6% ± 4.7; p = 0.0187) but independent of disease activity (Figure 1(a)).

NKG2DL expression profiles of jSLE B cells displayed pronounced individual variability and, overall, were similar to those recorded with HC (Figure 1(b) and Figure 1(c); [19] [20] ). In both cohorts, ULBP ligands (except ULBP4) were more prevalent than MICA/B. B cells expressing at least one, and most of the time two or more NKG2DL were present in all samples, albeit in ~50% of jSLE and HC samples their frequencies were low (<10% of total CD19⁺CD20⁺ cells). 10% or more NKG2DL-positive B cells were detected in 11 jSLE patients and 7 HC. In most samples, two or more NKG2DL were expressed at similar frequencies suggesting coordinate expression of more than one ligand by a given cell (Figure 1(b)). Indeed, co-staining for ULBP1 and ULBP2/5/6 confirmed co-ex- pression of these ligands on CD19⁺CD20⁺ B cells in three additional healthy donor PBMC samples (data not shown). There were no correlations between frequencies of NKG2DL-positive B cells and frequencies of NKG2D⁺CD4 T cells (data not shown). Thus although, as with normal B cells, jSLE B cells, if NKG- 2DL-positive, may well contribute to NKG2D⁺CD4 T cell proliferation, it seems unlikely that the jSLEB cell compartment alone provides the NKG2DL abundance presumed necessary for the extensive proliferative expansions of NKG2D⁺CD4 T cell populations that occur in jSLE [3] .

NKG2DL are present on normal monocytes presumably contributing to regulatory crosstalk with NKG2D⁺ lymphocytes [19] [20] [21] [22] . With no evidence for aberrant NKG2DL presence among lupus B cells we thus considered myelomonocytic cells as possible NKG2DL source for NKG2D⁺CD4 T cell activation and expanded the flow cytometry-based screen to lupus monocytes. All but one (jSLE044; excluded due to limiting cell numbers) patient and control PBMC samples were tested. Monocytes were classified as classical, intermediate, and non- classical subsets based on expression of CD14 and CD16 (Supplementary Figure S2; [23] ). Proportions of classical (CD14^brightCD16⁻) monocytes were decreased, and those of intermediate (CD14^brightCD16⁺) increased in jSLE compared to control samples; non-classical (CD14^dimCD16⁺) monocytes were similar in both groups (Figures 2(a)-(c); [24] ). None of these changes correlated with disease

activity scores (Figures 2(a)-(c)).

As with jSLEB cells, NKG2DL profiles of patient monocytes resembled the nor- mal expression pattern with substantial variability among individuals and at least one, and frequently two or more ligands expressed (Figure 3(a) and Figure 3(b)). Unlike with B cells, however, all patient monocyte populations differed from controls in the extent to which NKG2DL, and in particular MICA/B and ULBP1, were expressed (Figures 3(a)-(e)). jSLE monocyte populations contained, in part significantly, lower frequencies of ULBP1 positive cells than their corresponding

controls. MICA/B positive cells were less frequent among the classical CD14^brightCD16⁻ and intermediate CD14^brightCD16⁺ monocyte subsets compared to controls. Cells expressing ULB2/5/6 or ULBP3 were generally rare but slightly more frequent among jSLE CD14^dimCD16⁺ monocytes. Thus, unlike with B cells, jSLE monocytes displayed disease-associated NKG2DL phenotypes although none correlated with disease activityor kidney involvement (data not shown). Due to limiting sample sizes testing of ULBP4 expression was sporadic.

Lupus antigen presenting cells have functional impairments such as an inability to induce the immuneregulatory ligands PD-L1 and CD80 [25] [26] . The apparent lack of ULBP1 and/or MICA/B expression on most jSLE monocytes cells may reflect a similar defect. Hence, adherence-enriched jSLE and control monocytes were cultured in the presence of stimuli known to induce NKG2DL in normal myeloid cells and monitored for surface NKG2DL expression over time [19] [20] [21] . GM-CSF or LPS treatment resulted in induction of MICA/B and ULBP1 in normal controls but had no effect with jSLE monocytes (Supplementary Figure S3; [19] [20] [21] ). ELISA of culture supernatants for soluble NKG2DL

(sMICA, sMICB, sULBP1) was negative suggesting that the lack of surface NKG- 2DL on the jSLE monocytes was not due to enhanced ligand shedding [27] . Thus altogether, as with lupus B cells, lupus monocytes unlikely provide NKG2DL driving NKG2D⁺CD4 T cell proliferation. To ensure comprehensiveness, we also examined patient and control T cell and NK cell compartments for NKG2DL expression. Rare (<5%) NKG2DL-positive CD4 and CD8 T cells were detected in both cohorts with no prevalence for one or the other. NK cells (defined based on expression of CD56 and/or CD16) were NKG2DL-negative (data not shown).

Continuing the search for NKG2DL positive environments we resorted to screening for presence of soluble NKG2DL in matched jSLE patient plasma as surrogate readout for NKG2DL expression elsewhere [3] [12] . Although our earlier stu- dy found no significant association between frequencies of NKG2D⁺CD4 T cells and soluble MICA (sMICA) plasma concentrations we revisited this issue and extended the analysis to other soluble NKG2DL. Plasma samples matching all 22 patient PBMC specimens were screened by ELISA for sMICA, sMICB, and sULBP1. 14 of these samples were positive for sMICA yet concentrations generally low (picogram range) and unaligned with NKG2D⁺CD4 T cell frequencies (data not shown). sMICB was detectable in all 22 plasma samples at concentrations that varied widely (range: 199 pg/ml - 30.3 ng/ml) among samples and displayed a significant positive trend relationship with NKG2D⁺CD4 T cell proportions (Figure 4(a)). sMICB plasma values also correlated negatively with disease activity scores suggesting clinical relevance (Figure 4(b)). ELISA for sULBP1 was negative throughout. Although sMICB plasma values were independent of presence or absence of overt renal disease we considered kidney as the most likely site of aberrant NKG2DL expression and putative origin of NKG2DL shedding [28] . Because of lack of access to kidney biopsies we examined patient urine instead for the presence of soluble NKG2DL (sMICA, sMICB, and sULBP1). Urine specimens matching the 22 patient and 20 control PBMC and plasma were not available. We thus examined urine and paired plasma from an additional five jSLE patients (Table 1) and age-matched controls by sMICA and sMICB ELISA. All patient―but none of the HC―derived samples were positive for sMICB with concentrations in the two specimen types well aligned (Figure 4(c)). sMICA was detected in all jSLE urine samples and in three of the paired plasmas (Figure 4(c)). Extending these observations to adult lupus, paired urine and plasma specimens from three SLE patients also contained low concentrations of sMICA and abundant sMICB. ELISA for sULBP1 was negative throughout. Filtration of plasma sMICA/B in the kidney is unlikely as control urine samples from patients with MICA/B expressing tumors and abundant plasma sMICA/B were negative for sNKG2DL (data not shown). Altogether, these results point towards the kidney as the site of aberrant NKG2DL expression in jSLE with both cell surface and soluble ligands possibly driving the NKG2D⁺CD4 T cell population expansions typical for this disease.