The human glomerular endothelial cells are potent pro-inflammatory contributors in an in vitro model of lupus nephritis

Juvenile-onset lupus nephritis (LN) affects up to 80% of juvenile-onset systemic lupus erythematosus patients (JSLE). As the exact role of human renal glomerular endothelial cells (GEnCs) in LN has not been fully elucidated, the aim of this study was to investigate their involvement in LN. Conditionally immortalised human GEnCs (ciGEnCs) were treated with pro-inflammatory cytokines known to be involved in LN pathogenesis and also with LPS. Secretion and surface expression of pro-inflammatory proteins was quantified via ELISA and flow cytometry. NF-κΒ and STAT-1 activation was investigated via immunofluorescence. Serum samples from JSLE patients and from healthy controls were used to treat ciGEnCs to determine via qRT-PCR potential changes in the mRNA levels of pro-inflammatory genes. Our results identified TNF-α, IL-1β, IL-13, IFN-γ and LPS as robust in vitro stimuli of ciGEnCs. Each of them led to significantly increased production of different pro-inflammatory proteins, including; IL-6, IL-10, MCP-1, sVCAM-1, MIP-1α, IP-10, GM-CSF, M-CSF, TNF-α, IFN-γ, VCAM-1, ICAM-1, PD-L1 and ICOS-L. TNF-α and IL-1β were shown to activate NF-κB, whilst IFN-γ activated STAT-1. JSLE patient serum promoted IL-6 and IL-1β mRNA expression. In conclusion, our in vitro model provides evidence that human GEnCs play a pivotal role in LN-associated inflammatory process.

As the ciGEnCs were shown to mainly produce and secrete only two (MCP-1 and sVCAM-1) out of the seven identified novel urinary biomarkers for LN 20 , it was hypothesised that the role of GEnCs may be related to the production and local secretion of other pro-inflammatory mediators. To address this hypothesis, changes in secretion levels of pro-inflammatory cytokines (TNF-α, IFN-γ, IL-6 and IL-10), chemokines (macrophage inflammatory protein-1 alpha (MIP-1α), IFN-γ-induced protein-10 (IP-10) and IL-8) as well as the blood cell growth factors (granulocyte-macrophage and macrophage colony-stimulating factor (GM-CSF and M-CSF)) were examined via Luminex ELISA.
Due to lack of IL-6-and VEGF-induced changes in the mRNA expression levels of MCP-1, VCAM-1, IL-6 and IL-8 compared to untreated ciGEnCs (data not shown), ciGEnC pro-inflammatory protein secretion in IL-6-and VEGF-treated conditioned media was not tested in ELISA assays (the 96-well plate capacity of the Luminex assay also limited the amount of conditioned media tested to those that were expected to exhibit the most meaningful changes in protein secretion compared to the untreated ciGEnCs). However, IL-6 and VEGF were included in the combined cytokine treatments (All).
www.nature.com/scientificreports www.nature.com/scientificreports/ 24-hour serum treatments induce changes in the secretion levels of IL-6. The 24 h JSLE serum treatments were able to promote an increase in the mRNA expression of IL-1β and IL-6. For this reason, we also investigated the effect of serum treatments in IL-1β and IL-6 secretion by the ciGEnCs. IL-1β and IL-6 in serum samples were also tested to ensure whether potential presence of these cytokines in the ciGEnC conditioned www.nature.com/scientificreports www.nature.com/scientificreports/ media could be attributed to ciGEnC production. IL-1β presence in serum samples was relatively low (frequently below the level of detection for the assay) and did not differ among HCs, active and inactive LN patients (Fig. 4a). Furthermore, when ciGEnCs were treated with 5% sera from LN patients, IL-1β levels secreted by the cells were below the level of detection for the assay (except for 1 active disease sera sample) (Fig. 4b).

Pro-inflammatory protein release is occurring through NF-κB and STAT1 induction in ciGEnCs.
In order to determine whether the effects observed in ciGEnCs following cytokine and LPS treatments are www.nature.com/scientificreports www.nature.com/scientificreports/ mediated through the NF-kB pathway, we initially used an immunoblot assay to determine nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor, alpha (Iκ-Bα) degradation, as Iκ-Bα is the cytoplasmic inhibitor of NF-κΒ that prevents its nuclear translocation 22 . At 30 min, it was mainly IL-1β and, to a lesser extent, TNF-α, which were able to induce Iκ-Bα degradation (Fig. 7a).
Nuclear translocation of NF-kB and of signal transducer and activator of transcription (STAT)-1 and -2 (STAT-1 and STAT-2) was tested via immunofluorescence to determine pathway activation. TNF-α and IL-1β promoted NF-κB nuclear translocation at 30 min with IL-1β displaying a more robust effect than TNF-α, whereas LPS only induced low levels of NF-κB activation (Fig. 7b). IFN-γ induced STAT-1 nuclear translocation at 30 min (Fig. 7c) while it induced only moderate to low STAT-2 activation (Fig. 7d). www.nature.com/scientificreports www.nature.com/scientificreports/ Discussion LN is one of the main and most significant complications of JSLE. LN affects the majority of JSLE patients (approx. 80%) [5][6][7] , and can cause severe renal damage 3 and GFB impairment 4 . This study aimed to investigate the potential role of the human glomerular endothelium in LN inflammation using a cytokine-based in vitro model of LN.
The ciGEnCs were treated with cytokines previously found by our laboratory and by other studies to be elevated in plasma and/or serum samples of JSLE patients with LN (IFN-α 23 , IFN-γ 24 , TNF-α 24,25 , IL-1β 24 , IL-6 24 , IL-13 24 and VEGF 24,26 ) or shown by other studies to be expressed in the glomerular renal tissue of JSLE or adult patients with LN (IFN-γ 27 ), or to have certain polymorphisms associated with JSLE (IL-1β 27,28 ). The potential effect of bacterial endotoxin inflammation was also examined via 1 μg/ml LPS treatment, which, in murine models of LN and in human endotoxemia, is known to play a critical role in LN exacerbation after a bacterial infection [28][29][30][31] .
Our findings indicate that in line with in vitro models using HUVECs 17,32,33 and HBMECs 34-36 , upon activation with TNF-α, IL-1β, IL-13, IFN-γ and with LPS, ciGEnCs significantly increase the production of key pro-inflammatory proteins. These included: cytokines, chemokines, blood cell growth factors, adhesion molecules and T-cell co-stimulatory molecules. All proteins demonstrated in this study to be upregulated in human ciGEnCs have been previously shown to be involved in human LN or other types of human renal disease and are summarised on Table 1.
Single TNF-α and IL-1β treatments upregulated MIP-1α and GM-CSF and combined TNF-α and IL-1β treatment significantly increased MCP-1 secretion. LPS also had a prominent effect in MIP-1α and GM-CSF secretion. MCP-1 promotes infiltration by inflammatory dendritic cells in LN 37 and MIP-1α can activate and attract human granulocytes, macrophages and monocytes 38 whereas GM-CSF promotes the maturation and differentiation of macrophages, neutrophils, eosinophils and basophils 39 . IL-1β and LPS also upregulated IL-6 and TNF-α IFN-γ increased the secretion of IP-10, IL-10 and TNF-α, as did LPS. IP-10 mediates T-cell accumulation to the inflamed tissues 42 whereas IL-10 is involved in autoantibody production 43 . The combined cytokine treatment, which could be mainly attributed to TNF-α, IL-1β and IFN-γ led to increased amounts of secreted M-CSF as the other cytokines had minimal effects individually. When the combined effect of the prototypical pro-inflammatory cytokines -TNF-α and IL-1β-was specifically tested, the ciGEnCs were found to significantly upregulate M-CSF secretion compared to the untreated ciGEnCs. M-CSF promotes the proliferation, differentiation, and survival of monocytes and macrophages 44 . LPS had a prominent effect on secretion of IFN-γ which is involved in anti-viral immunity 45 . TNF-α and IL-1β appeared to increase secretion of IL-8, a potent neutrophil chemokine 46 although not to a level of statistical significance tested. These data suggest that GEnCs secrete mediators into the environment that induce the recruitment, activation and maturation of inflammatory leukocytes leading to perpetuation of the inflammatory response.
Two important adhesion molecules were found to be upregulated by the ciGEnCs following pro-inflammatory stimulation; VCAM-1 and ICAM-1. VCAM-1 is involved in the binding of lymphocytes, monocytes, eosinophils, and basophils but not neutrophils to the vascular endothelium 47 whereas ICAM-1 promotes all leukocytes' binding to the endothelium 48 . In this study, IL-13 had an effect exclusively on soluble and surface VCAM-1 expression with the latter being promoted by IL-13 in combination with TNF-α. ICAM-1 expression was significantly increased by TNF-α, IL-1β and LPS. 24 h ICAM-1 upregulation by TNF-α, IL-1β and LPS has also been confirmed in HUVECs 49,50 . Similar to the human ciGEnCs and in contrast to other types of human endothelial cells (carotid, coronary), lack of 24 h IL-1β-and LPS-mediated upregulation of surface VCAM-1 has been previously observed in HUVECs and human subclavian endothelial cells 50 .
The differential cytokine regulation of surface VCAM-1 and ICAM-1 expression could imply that certain pro-inflammatory stimuli could lead to preferential intra-renal accumulation of certain immune cell populations. Indeed, our findings from the neutrophil adhesion assay indicated that combinatorial TNF-α and IL-13 treatment of ciGEnCs had an effect similar to that of TNF-α alone and led to ciGEnC adhesion of similar numbers of neutrophils, which were significantly higher than those adhering to untreated or IL-13-treated ciGEnCs. Thus, ICAM-1 that was found in this study to be induced by TNF-α but not IL-13 (IL-13 had an effect only on VCAM-1 expression), could be the adhesion molecule predominantly responsible for neutrophil binding on ciGEnCs. This assumption is further corroborated by evidence in the literature that, in contrast to VCAM-1, ICAM-1 can promote binding of neutrophils on the endothelium 48 .
In this study an attempt was made to assess the expression of E/P-selectin by the ciGEnCs in response to our pro-inflammatory model however no expression was seen. A recent study demonstrated that GEnCs do not require E/P-selectin for neutrophil binding due to an increased dwell time by cells at this location 51 . This has been confirmed both in mouse models using lupus-prone mice and in human renal biopsies from 108 patients with primary renal disease and allograft recipients [52][53][54][55] .
PD-L1 surface expression that was upregulated at 24 h in ciGEnCs by IFN-γ treatment, can prevent the activation of CD-8 + T-cells 21 and inhibit the autoreactive T-cell function 56 . However, after 24 h, IL-1β significantly increased surface expression of the positive CD4 + T-cell co-stimulatory molecule, ICOS-L 57 . Therefore, T-cell-driven kidney inflammation could be promoted by CD4 + instead of CD8 + T-cells.
Incubation of ciGEnCs for 4 h and 24 h reflects acute phase inflammation 58 rather than chronic inflammation. The 24 h cytokine and LPS stimulation did not significantly induce apoptotic or necrotic cell death, indicating that during acute inflammation human ciGEnCs maintain their integrity and acquire a pro-inflammatory phenotype IL-10 Plasma IL-10 has been found to be significantly elevated in JSLE patients with active disease 73 .
M-CSF Serum and urine M-CSF levels have been shown to be a reliable and sensitive biomarkers for LN 74,75 .
GM-CSF Increased mRNA levels of GM-CSF are expressed by in vitro cultured human renal tubular cells derived from interstitial fibrotic kidneys compared to tubular cells derived from non-fibrotic kidneys 76 .

MIP-1α
Together with MCP-1, MIP-1α promotes macrophage recruitment and activation in the kidneys of patients with crescentic GN 77 .
TNF-α TNF-α gene polymorphisms leading to excessive serum TNF-α production can predispose patients to LN development 79 .

IFN-γ
Significantly increased IFN-γ positive immune-histochemical staining has been observed in biopsies from juvenile-onset LN patients compared to HCs 27 .

PD-L1
In JSLE, decreased PD-L1 expression by antigen-presenting cells has been associated with active disease 80 .
ICOS-L ICOS-L plasma levels are increased in patients with active SLE compared to those with inactive SLE 81 . www.nature.com/scientificreports www.nature.com/scientificreports/ due to cell activation and not due to cell death. Downstream activation of the central pro-inflammatory transcription factor NF-κB primarily occurred following IL-1β and TNF-α stimulation and to a lesser extent by LPS and was demonstrated by nuclear NF-κB translocation as well as by Iκ-Bα degradation. In contrast, IFN-γ was found to activate downstream STAT-1 signalling but not STAT-2 which is normally activated by IFN-α.
NF-κB shRNA knock-down in human microvascular endothelial cells (HMECs) has been shown to lead to strong suppression of the expression of TNF-α-induced genes such as MCP-1, M-CSF, GM-CSF, IL-6 and IL-8 59 . Furthermore, STAT-1 siRNA knock-down in IFN-γ-treated HUVECs has been shown to downregulate secretion of IP-10 and other T-cell related chemokines 60 providing supporting evidence that these pathways are activated in endothelial cells. Thus, it could be hypothesised that in human GEnCs similar mechanisms of action could be occurring.
The 24 h serum treatments did not demonstrate significant changes in the mRNA expression levels of most pro-inflammatory genes tested. However, IL-6 and IL-1β mRNA levels were increased by the JSLE sera. In addition, IL-6 secretion by the ciGEnCs underwent a pronounced increase after both active and inactive LN serum treatments. On the contrary, IL-1β secretion by ciGEnCs was not affected by either of the serum treatments. This finding suggests potential lack of inflammasome and caspase-1 activation in ciGEnCs following serum treatments which could lead to a reduction in the release of active IL-1β within the microenvironment.
The limited amount of blood samples due to rarity of the disease as well as the small blood sample volumes obtained from paediatric patients did not allow for further investigation and detection of the JSLE serum factors inducing the expression IL-6 and IL-1β production. A potential candidate, however, could be autoantibodies present in the sera samples. Indeed, previous research studies using affinity purified IgG antibodies from anti-endothelial cell antibody (AECA)-positive SLE patients to treat HUVECs in vitro, demonstrated increased HUVEC-derived secretion of IL-1 61 and IL-6 62 . Similar findings have also been observed in scleroderma and systemic vasculitis patients 62,63 .
The lack of strong responses induced by the JSLE sera may be due to the immunosuppressant regimens followed by patients. Increasing the serum doses used in these experiments may be required to demonstrate an effect; however, the 5% treatments (limited due to sample volumes from paediatric patients) did not suffice to induce a strong in vitro ciGEnC pro-inflammatory response.
Furthermore, the serum treatments could have had a more pronounced effect on ciGEnCs in an in vitro model in which ciGEnCs would be co-cultured with human podocytes and/or mesangial cells. Lack of crosstalk between native glomerular renal cells could be affecting the results obtained from the serum treatments. The development of a co-culture model, however, was beyond the scope of this study, the aim of which was to investigate pro-inflammatory properties of the human ciGEnCs, a type of human endothelial cells that has been little studied.
There are inherent limitations with using a cell line to generate a model of human disease. However, ciGEnCs retain all the morphological and functional characteristics of the primary GEnCs and express at levels comparable to those of primary GEnCs, the majority of endothelial cell-specific markers 64 . These characteristics render the ciGEnCs into a reliable tool for the study of human GEnCs. Future work could focus on validating these findings using primary cells or in an in vivo murine model however this was beyond the scope of this study.
In conclusion, our data imply that the glomerular endothelium could have a pivotal role in LN inflammation (Fig. 8). As demonstrated by our in vitro model, the GEnCs are not just passive bystanders but actively respond to the renal pro-inflammatory environment created in kidney disease.

Materials and Methods
Materials. Recombinant 64 . The ciGEnCs were cultured at 33 °C (with 5% CO 2 ) in endothelial growth medium 2-microvascular (EGM-2 MV Bulletkit, CC-3202, Lonza) containing 5% Fetal Bovine Serum (FBS) and growth factors as supplied (VEGF was excluded as it has been utilised for treatments). Once cells reached 50-70% confluence they were thermoswitched at 37 °C (with 5% CO 2 ) for seven days until complete differentiation was achieved and confirmed via light microscopy.
For all experiments, a minimum of three consecutive ciGEnC passages was used for reproducibility purposes. Consecutive passages between 29 and 35 were used, as the researchers who developed and donated the ciGEnCs had previously shown that ciGEnCs can retain their primary GEnC-specific morphological and functional characteristics up to passage 41 64 . All changes in mRNA expression or protein expression, secretion and production levels were compared to those of the untreated ciGEnCs. The fully differentiated ciGEnCs were incubated at 37 °C for 30 mins, 4 h and 24 h with 10 ng/ml (a concentration previously used to induce pro-inflammatory changes in endothelial cells in vitro [65][66][67] )-of seven human recombinant cytokines (IFN-α, IFN-γ, TNF-α, IL-1β, IL-6, IL-13, VEGF), individually and in combination (All) (cytokine concentration response assays were also performed and are presented on Supplementary Fig. 4) and with 1 μg/ml of LPS. (2019) 9:8348 | https://doi.org/10.1038/s41598-019-44868-y www.nature.com/scientificreports www.nature.com/scientificreports/ Fully differentiated ciGEnCs were treated for 4 h or 24 h with 5% (5% sera concentration matching the 5% FBS concentration in standard media used for ciGEnC culture) serum samples in medium without FBS, from JSLE patients with active LN (renal BILAG: A/B, N = 10) and inactive LN (renal BILAG: D/E, N = 10) and with 5% sera from age-matched healthy controls (N = 10). Following completion of treatments, qRT-PCR was performed as above to assess changes in the mRNA expression levels between ciGEnC groups treated with HC and LN sera. Patient and HC information are presented on Table 2. To test the effect of serum samples on IL-6 and IL-1β secretion by the ciGEnCs, an additional of N = 6 of active LN, inactive LN and HC serum samples were collected and used to treat the ciGEnCs for 24 h, at a 5% concentration (patient and HC information presented on Table 3). qRt-pCR. RNA was extracted using Trizol-chloroform and in-column DNase digestion (Qiagen RNeasy Mini Kit). RNA concentrations and purity were determined using Nanodrop (ND-1000 Spectrophotometer, Thermo Fisher Scientific). 200 ng RNA was transcribed into cDNA using the AffinityScript multi-temp cDNA synthesis kit (Agilent Technologies, Cheshire, UK) as per manufacturer's instructions. qRT-PCR was performed using primers described in Supplementary Table 2 with the Brilliant III Ultra-fast SYBR QPCR mastermix kit (Agilent Technologies) as per manufacturer's instructions. mRNA expression for target genes was normalised using the mean (for one housekeeping (HK) gene) or the geometric mean (for more than one HK genes) of the following HK genes: beta-actin (ACTB), TATA-binding protein (TBP), beta-tubulin (TUBB). The ΔΔCt value was calculated as follows:  Under the highly inflammatory environment created within the glomeruli by the presence of TNF-α, IL-1β, IL-13 and IFN-γ but also by the potential presence of LPS, the human GEnCs will be activated to produce, secrete and express a variety of pro-inflammatory cytokines, chemokines, blood cell growth factors and adhesion molecules that will further exacerbate the renal disease by either promoting the infiltration of immune cells within the glomeruli or by activating their neighbouring renal cells, the podocytes and mesangial cells.
www.nature.com/scientificreports www.nature.com/scientificreports/ for 1 h (room temperature). Protein detection was achieved using enhanced chemiluminescence (ECL) reagents (Li-COR). Li-COR software was used for protein band densitometry. Due to β-actin and Iκ-Bα having a similar molecular weight, the samples were loaded and were run twice onto two different gels, which were blocked for either β-actin or Iκ-Bα (full-length images are presented on Supplementary Fig. 3).
Neutrophil adhesion assay. ciGEnCs were stimulated with 10 ng/ml of TNF-α and IL-13, alone or in combination, for 24 h (untreated ciGEnCs were included). Whole-blood from adult healthy controls was used to isolate human neutrophils via HetaSep (StemCell)-aided separation of white blood cells from red blood cells  www.nature.com/scientificreports www.nature.com/scientificreports/ and subsequent Histopaque (Sigma-Aldrich) centrifugation of white blood cells. Neutrophils were re-suspended in RPMI (+10% FBS) at a concentration of 3 × 10 6 /ml. ciGEnCs were then washed in PBS and following PBS removal 1 ml of neutrophil suspension was added onto each well. Cells were then incubated for 90 min at 37 °C. Following 90 min incubation, conditioned media containing non-adherent neutrophils were collected and neutrophils were counted using a cell counter. The cells remaining on the plate were washed in PBS and images were taken at 20x magnification. 5 random images per well were collected and the number of neutrophils adherent to the ciGEnC monolayer was counted using ImageJ software. Immunofluorescence (IF). Transcription factor activation and nuclear translocation was assessed using IF. Following 30 min incubation, cells were fixed in 4% formaldehyde for 30 mins and permeabilised using 0.4% Triton-X-100 for 10 mins. Non-specific binding was blocked using PBS-1% BSA for 1 h (room temperature). Antibodies against human NF-κΒ, STAT-1 and STAT-2 were added overnight, at 4 °C. ciGEnCs were washed three times with PBS and incubated with fluorescence-conjugated secondary antibodies and 1 μg/ml of DAPI (4′,6-diamidino-2-phenylindole), for 2 h at room temperature. Cells were visualised immediately, using the EVOS FLoid Cell Imaging Station (ThermoFisher Scientific). Image analysis. Image analysis to determine nuclear translocation of the signal in response to cytokine or LPS treatment was performed using Fiji (ImageJ) and particularly the Coloc2 algorithm 68 . For every group of treatments, the Pearson's r-values for every single cell, representing the degree of nuclear colocalisation between DAPI and A488 or A568, were obtained. Statistical analysis. Statistical analysis of all data was performed using GraphPad Prism 4.0 software. Data are expressed as median values[range]. Multiple comparisons were made using Kruskal-Wallis or Friedman non-parametric test with Dunn's post-hoc test for three or more treatment groups and Mann-Whitney non-parametric test for two treatment groups. Statistical significance was set a priori at a p value < 0.05. For the IF assay, the r-values from each treatment group were used to perform non-parametric Kruskal-Wallis statistical test (with Dunn's post-hoc test) for three or more treatment groups, or Mann-Whitney tests for two treatment groups. For the nuclear colocalisation to be statistically significant, the r-values should range from 0.7 to 1 (0.7 indicated by dashed line in graphs).  Table 3. Serum treatments for changes in ciGEnC IL-1β and IL-6 secretion. Age, gender, ethnicity, renal BILAG scores and medications for JSLE patients; age, gender and ethnicity for paediatric healthy controls (N = 6/group).