Human brain endothelial cells endeavor to immunoregulate CD8 T cells via PD-1 ligand expression in multiple sclerosis

CNS tissue was obtained from surgical resections performed for the treatment of non-tumor related intractable epilepsy as previously described [28]. Consent and ethical approval were given prior to surgery (BH 07.001). Human brain endothelial cells (HBECs) were grown in M199 medium (Invitrogen, Burlington, ON, Canada) supplemented with 10% fetal bovine serum, 20% normal human serum, endothelial cell growth supplement (5 μg/ml) and insulin-selenium-transferin premix on 0.5% gelatin-coated tissue culture plates (all reagents from Sigma, Oakville, ON, Canada).

A written informed consent was obtained from healthy donors in accordance with the local ethical committee (HD 07.002 and BH 07.001). Peripheral blood mononuclear cells (PBMCs) were obtained by Ficoll density gradient as previously described [29]. CD8 or CD4 T cells were positively isolated from PBMCs using either CD8 or CD4 beads respectively (MACS, Miltenyi Biotec, Auburn, CA, USA) according to the manufacturer's instructions; purity assessed by flow cytometry was typically > 95%.

Total RNA was extracted and transcribed into cDNA as previously described [17, 30]. Relative mRNA expression was determined by quantitative real-time PCR (qPCR) using primers and TaqMan FAM-labeled MGB probes for PD-L1 and PD-L2 and ribosomal 18S (VIC-labeled probe, used as an endogenous control) obtained from Applied Biosystems (Foster City, CA, USA) according to manufacturer's instructions and as previously described [17, 30].

Cells were stained for surface and/or intracellular molecules as previously described [17, 31], acquired on a LSRII (BD Biosciences, Mississauga, ON, Canada) and analyzed with FlowJo software (Treestar, Ashland, OR, USA). Mouse monoclonal antibodies directed at human protein and conjugated to biotin, fluoroscein isothiocyanate (FITC), Alexa Fluor^® 700, phycoerythrin, Pacific Blue, or allophycocyanin were used. Surface stainings targeted: PD-L1 (eBioscience, San Diego, CA, USA), PD-L2 (eBioscience), HLA-ABC (Biolegend, San Diego, CA, USA), CD4 and CD8 (BD Biosciences). Intracellular stainings targeted: granzyme B (Caltag, Buckingham, UK) and IFN-γ (BD Biosciences). Appropriate isotype controls were used for all stainings. Δmedian fluorescence intensity (ΔMFI) was calculated by subtracting the fluorescence of the isotype from that of the stain.

Migration assays were performed in a modified Boyden chamber as previously described [28]. HBECs plated on Boyden chambers (Collaborative Biomedical Products, Bedford, MA) were stimulated with IFN-γ (200 U/ml) and TNF (200 U/ml) for 24 hours and then treated either with isotype control antibodies or blocking antibodies specific for PD-L1 (10 μg/ml, eBioscience) and/or PD-L2 (10 μg/ml, eBioscience) for one hour at 37°C. CD8 or CD4 T cells that had been exposed to plate-bound anti-CD3 (0.9 μg/ml, clone OKT3, purified in house) and anti-CD28 antibodies (1 μg/ml, BD Biosciences) for 72 hours were then added to the upper chamber (1 × 10⁶ cells per Boyden chamber) and allowed to migrate for 24 hours across HBECs. FITC-labeled BSA (50 μg/ml; Invitrogen) was concurrently added to the upper chamber and 50 μl samples were harvested from the upper and lower chambers at different time points and the fluorescence intensity in these samples was measured using a Synergy4 Biotek microplate reader. The diffusion rate of the FITC-BSA, a measure of the permeability, was expressed as a percentage and calculated as followed: [(BSA lower chamber)/(BSA upper chamber)] × 100. After migration, cells from the lower and upper chambers were collected, counted, and stained for different markers.

HBECs were plated (5 × 10⁵ cells per well in a 24-well plate), and after 3 days when reaching confluence, stimulated with IFN-γ (200 U/ml) and TNF (200 U/ml) for 24 hours. HBECs were washed three times to remove these inflammatory cytokines. An isotype control antibody or blocking antibodies specific for PD-L1 (10 μg/ml) and/or PD-L2 (10 μg/ml) were added one hour at 37°C to allow them to bind to their cognate ligands. Isolated alloreactive human CD8 T cells were labeled with CFSE as previously described [29] and then subsequently added (2 × 10⁵ cells per well) to HBECs without removing the blocking antibodies and in the presence of anti-CD3 (0.18 μg/ml) and anti-CD28 (1 μg/ml) antibodies. Blocking antibodies and anti-CD3 and anti-CD28 were left in the wells for the entire co-culture. After a 6 day co-culture, CD8 T cells were collected and stained for Live/dead fixable Aqua dead cell stain kit (Invitrogen) to exclude dead cells and stained for CD8, granzyme B, and IFN-γ for flow cytometry assessment.

Post-mortem brain sections from tissue donors without CNS disease and patients diagnosed clinically and confirmed by neuropathological examination as having MS were obtained from the NeuroResource tissue bank, UCL Institute of Neurology, London, U.K. Tissues were donated to the tissue bank with informed consent following ethical review by the London Research Ethics Committee, UK. This study was approved by the CHUM Ethical Committee (HD 07.002). Snap-frozen coded sections (~1 cm² and 10 μm thick) were cut from blocks of normal control and MS brain tissues. Sections cut before and immediately after the ones used for the immunofluorescence studies were stained with oil red O and hematoxylin, and scored as previously described [32] (Table 1). Sections were air-dried, fixed in cold acetone for 10 min, and blocked for non-specific binding for 1 hour with 10% donkey (for PD-L2 detection) or goat serum (for PD-L1 detection). Primary antibodies targeting PD-L1 (25 μg/ml, Biolegend) or PD-L2 (2 μg/ml, RD Systems, Burlington, ON, Canada) was incubated 1 hour at room temperature and then overnight at 4°C. Sections were then washed with PBS and incubated for 40 minutes with appropriate secondary antibodies: Alexa Fluor^® 488-conjugated goat-anti-mouse for PD-L1 and Alexa Fluor^® 488-conjugated donkey-anti-goat for PD-L2. Sections were then incubated at room temperature for 1 hour with antibodies targeting cell specific markers for endothelial cells (rabbit-anti-human-caveolin-1, Santa Cruz Biotechnology, Santa Cruz, CA, USA) followed by 40 minutes with secondary antibody (Rhodamine-conjugated goat-anti-rabbit, Jackson Immunoresearch, West Grove, PA, USA). Finally, sections were incubated with a nuclear stain TO-PRO^®-3 iodide (Invitrogen), treated with Sudan Black and mounted as previously described [17]. Controls were concurrently carried out on adjacent sections using appropriate primary isotype controls at the same concentrations. Slides were observed using a SP5 Leica confocal microscope. Confocal images were acquired simultaneously in different channels throughout 4-8 μm z-stack every 0.2-0.5 μm. We validated staining specificity by lack of signal only when the corresponding laser was turned off but not when others were still on. Several fields (> 5) containing blood vessels were taken randomly on each section and used for quantification of positive cells. Moreover, we confirmed the absence of bleed-through by re-examining selected sections using sequential scanning.

Statistical analyses were performed using PRISM Graphpad™ software (La Jolla, CA, USA) and included Student's t-test; P-values < 0.05 were considered significant.

We evaluated whether HBECs express detectable levels of PD-L1 and/or PD-L2. Primary cultures of HBECs were either left untreated or activated with pro-inflammatory cytokines IFN-γ, TNF, or IFN-γ+TNF to mimic the pro-inflammatory environment typically observed in the CNS of MS patients. HBECs expressed very low PD-L1 but detectable PD-L2 mRNA levels under basal conditions as assessed by qPCR (Figure 1A). IFN-γ+TNF treatment robustly increased those levels (Figure 1A), almost reaching statistical significance for PD-L1 expression (n = 4 donors, untreated vs. IFN-γ+TNF p = 0.075). Detection of PD-L1 and PD-L2 proteins by flow cytometry allowed quantification of both percentages of HBECs expressing these molecules and intensity of such expression (ΔMFI); typical flow cytometry detection is shown (Figure 1B). HBECs under basal conditions expressed very low/undetectable levels of PD-L1 protein (Figure 1B: 1.3%), while PD-L2 protein was already expressed by the majority of cells reaching 78.5% (Figure 1B). In response to different cytokine treatments tested, the proportion of HBECs expressing PD-L1 significantly increased reaching over 96%, especially in response to IFN-γ and IFN-γ+TNF (Figure 1B) (mean n = 4, PD-L1+ cells: IFN-γ: 98.5 ± 1.2% and IFN-γ+TNF 99.4 ± 0.3%; ** p < 0.003 compared to untreated), while TNF (Figure 1B) had a more modest impact (mean n = 4, PD-L1+ cells: 54.1 ± 15.6%, p = 0.065 compared to untreated). All cytokine treatments tested also boosted the proportion of HBECs expressing PD-L2 reaching over 96% (mean n = 4, PD-L2+ cells: IFN-γ: 97.2 ± 1.4%; IFN-γ+TNF: 98.9 ± 0.6%; TNF: 97.5 ± 2.4%). Moreover, cytokine treatments led to not only increased proportions of HBECs expressing PD-L1 or PD-L2 but also elevated intensity as shown by ΔMFI; IFN-γ+TNF having the more potent impact for PD-L1 levels (Figure 1C). We also observed an upregulation of MHC-class I molecules (HLA-ABC, Figure 1B) on HBECs upon cytokine treatment. In agreement with our flow cytometry results, PD-L1 was undetectable by immunocytochemistry on untreated HBECs but reached detectable levels after IFN-γ+TNF treatment, while PD-L2 was detectable both under basal conditions and following pro-inflammatory treatments (data not shown).

We elected to address whether the expression of PD-L1 and PD-L2 by HBECs influences their capacity to regulate the migration of CD8 and CD4 T cells into the CNS. We used a well-established in vitro model of the BBB in which HBECs are seeded in the upper compartment of a Boyden chamber [33], inflamed, and then incubated with either anti-PD-L1 and anti-PD-L2 blocking antibodies or isotype control antibodies. CD8 or CD4 T cells that have been stimulated for 3 days with anti-CD3+anti-CD28 to maximally increase the expression of PD-1 were added to the Boyden chamber. We have previously shown that although on average only 14% of ex vivo CD8 T cells expressed PD-1, this proportion reached 64% after such a stimulation [17]. We observed significantly greater numbers of CD8 T cells migrating through the in vitro BBB when blocking antibodies targeting PD-L1+PD-L2 were added compared to the isotype control (Figure 2A). Blocking only one ligand PD-L1 or PD-L2 had a more modest impact on the number of migrated CD8 T cells (data not shown). In parallel, we performed a permeability assay using BSA-FITC as a permeability tracer and observed an identical diffusion of BSA-FITC for the isotype control and the blocking antibodies conditions (Figure 2B). These results demonstrate that the elevated CD8 T cells transmigrating through the in vitro BBB in the presence of anti-PD-L1+anti-PD-L2 blocking antibodies were not due to a general disruption of the brain endothelial cell monolayer. Similarly, blocking these ligands led to an increased number of CD4 T cells migrating through our in vitro BBB (Figure 2C-D). Therefore, PD-L1 and PD-L2 expressed by HBECs contribute to dampening T cell migration through the barrier created by these specialized cells.

Previous studies have shown the capacity of PD-L1 and PD-L2 expressing endothelial cells, especially human umbilical vein endothelial cells (HUVECs) and mouse cardiac endothelial cells, to inhibit T cell responses [15, 20‐23]. However, whether human brain endothelial cells could modulate CD8 T cells responses has not been previously determined. Alloreactive human CD8 T cells were labeled with CFSE and then added to inflamed HBECs cultures that have been pre-incubated with either isotype control antibodies or anti-PD-L1 and anti-PD-L2 blocking antibodies. Proliferation (CFSE low cells) and production of IFN-γ and granzyme B were analyzed by flow cytometry. As shown for 2 different donors in Figure 3, blocking PD-L1 and PD-L2 had no consistent effect on CD8 T cell proliferation (Figure 3A, B). On the other hand, a modest but significant increase in the percentage of IFN-γ producing CD8 T cells (Figure 3A, C) was observed for all donors when PD-L1 and PD-L2 were blocked compared to the isotype control. The percentage of granzyme B producing CD8 T cells (Figure 3A, D) was significantly elevated when both ligands were blocked compared to the isotype control. Blocking only PD-L1 or only PD-L2 led to a partial increase of IFN-γ and granzyme B production in comparison to the blockage of both proteins (data not shown). Our results demonstrate that PD-L1 and PD-L2 expressed by HBECs were sufficient to significantly diminish the effector functions (cytokines and lytic enzyme) of human CD8 T cells.

To assess whether endothelial cells in the CNS of MS patients express PD-L1 and/or PD-L2, we performed immunohistochemistry on post-mortem brain tissues obtained from normal controls and MS patients (see description in Table 1). MS lesions were characterized using oil red 0 (ORO) and hematoxylin scoring as being acute, containing numerous phagocytic macrophages that had recently engulfed lipid-containing debris, or subacute, containing demyelinated areas but demonstrating less recent myelin destruction. Brain sections were stained for PD-L1 or PD-L2 and caveolin-1, a specific marker for endothelial cells, or appropriate isotype controls. Six to ten fields (at 630×, each field covering 0.0625 mm²) per section containing caveolin-1+ blood vessels were selected randomly (3 sections from controls and 7 sections MS lesions) and thoroughly analyzed to determine the percentage of blood vessels positive for PD-L1 or PD-L2, and representative fields are illustrated (Figures 4, 5). As shown in our earlier study [17], no or very low expression of PD-L1 was observed in the CNS of normal controls (Figure 4A-C). However, as we have previously reported, an elevated expression of PD-L1 was observed on astrocytes and microglia/macrophages in MS lesions, but no co-localization was found between PD-L1+ cells and caveolin-1+ cells (Figure 4E-G and 4I-K).

In contrast to PD-L1, PD-L2 was easily detected in normal control brain sections, and was co-localized with caveolin-1+ cells (Figure 5A-C, M). Whereas all caveolin-1+ cells were positive for PD-L2 in normal control sections, PD-L2 was only expressed by a subset of endothelial cells in MS sections (Figure 5E-G, I-K, N). Quantification of blood vessels identified by caveolin-1 labeling demonstrated that all blood vessels in normal tissues expressed PD-L2 but only 50% expressed PD-L2 in MS lesions (Figure 5O). The reduced PD-L2 expression was observed on vessels of all diameters and regardless whether these lesions were acute, sub-acute, or chronic. However, diminished PD-L2 expression was localized within areas of demyelination and not outside MS lesions as assessed by Sudan black staining [17]. PD-L2 labeling was either easily detectable or absent (Figure 5). PD-L2+ but caveolin-negative cells with a morphology suggestive of infiltrating leukocytes were observed around some blood vessels in MS lesions, whereas outside lesions and in normal control sections these cells were not seen.