Characterization of basigin monoclonal antibodies for receptor-mediated drug delivery to the brain

The brain uptake of biotherapeutics for brain diseases is hindered by the blood–brain barrier (BBB). The BBB selectively regulates the transport of large molecules into the brain and thereby maintains brain homeostasis. Receptor-mediated transcytosis (RMT) is one mechanism to deliver essential proteins into the brain parenchyma. Receptors expressed in the brain endothelial cells have been explored to ferry therapeutic antibodies across the BBB in bifunctional antibody formats. In this study, we generated and characterized monoclonal antibodies (mAbs) binding to the basigin receptor, which recently has been proposed as a target for RMT across the BBB. Antibody binding properties such as affinity have been demonstrated to be important factors for transcytosis capability and efficiency. Nevertheless, studies of basigin mAb properties' effect on RMT are limited. Here we characterize different basigin mAbs for their ability to associate with and subsequently internalize human brain endothelial cells. The mAbs were profiled to determine whether receptor binding epitope and affinity affected receptor-mediated uptake efficiency. By competitive epitope binning studies, basigin mAbs were categorized into five epitope bins. mAbs from three of the epitope bins demonstrated properties required for RMT candidates judged by binding characteristics and their superior level of internalization in human brain endothelial cells.

www.nature.com/scientificreports/ receptor is also highly attractive due to its low expression in neurons 8 . This is favorable since the mode of action of many brain therapeutic antibodies requires an extracellular target engagement (e.g., amyloid-β, α-synuclein, and tau aggregates), where uptake into neurons is unwarranted. The basigin receptor was first identified in mouse B16 melanoma cells and is part of the immunoglobin (Ig) superfamily 9,10 . By comparing mouse brain endothelial cells to liver and lung endothelial cells, basigin is 14 fold enriched in brain endothelial cells based on microarray expression analysis 11 . Moreover, basigin expression was found to be upregulated under ischemic conditions as well as in multiple sclerosis and Alzheimer's disease patients [12][13][14][15] . Basigin is a highly glycosylated type 1 transmembrane protein and a co-receptor for the lactate transporter, monocarboxylate transporter 1, present at the BBB 16 . Basigin has four isoforms where the most abundant is the isoform 2 with two extracellular Ig-like domains with intra-domain disulfide bonds 17 . Basigin has a highly conserved single transmembrane region and a short cytoplasmic tail 18,19 . The extracellular domain contains three N-glycosylation sites 20 . Basigin exists in different glycosylated forms -highly glycosylated and lowly glycosylated around 50 and 38 kDa, respectively [21][22][23] . The glycosylation has been suggested to be essential for basigin's function and localization to the plasma membrane 22,24 . The subcellular sorting and transcytosis of the basigin receptor in brain endothelial cells have not been well-explored, and most published data is based on trafficking in MDCK cells or non-polarized HeLa cells. After thoroughly studying the TfR1 as a transcytosis-shuttle, it appears likely that for an antibody targeting the basigin receptor via RMT, receptor binding characteristics, such as affinity, valency, pH-dependency, and epitope, would also be crucial for improved brain exposure. It has been reported extensively, that sorting of the antibodies to lysosomes affects the BBB-crossing ability of the antibodies. For TfR1 antibodies, lysosomal localization was suggested to be affected by binding affinity 25,26 , bivalent target interaction (i.e., avidity binding) 27 , and pH-dependency of receptor binding 28 . Hence binding epitope, affinity, and the functional impact of antibody binding may well influence the behavior of the target receptor. Consequently, receptor internalization, subcellular trafficking, and signaling might be affected differently across a diverse panel of antibodies.
In this study, we isolated a number of mouse monoclonal antibodies (mAbs) against the basigin receptor to select a set of candidates for basigin-mediated delivery across the BBB. The first step in our characterization was to define bins of antibodies with overlapping basigin epitopes. Subsequently, representatives from the different bins were subjected to further characterization of affinity, cellular binding and localization, and the ability to internalize. Here we showed a comprehensive characterization of a large panel of basigin antibodies and selection of quality lead candidates potentially to be used in the engineering of BBB crossing antibody constructs.

Results
Generation of novel basigin monoclonal antibodies. Anti-basigin mAbs were generated by standard hybridoma technology. Mice were immunized with recombinant forms of human basigin, corresponding to the extracellular domain of the receptor (basigin-ECD). Hybridomas were screened by enzyme-linked immunosorbent assay (ELISA) for binding to human basigin-ECD and cross-reactivity to rat, mouse, and porcine basigin-ECD. Thirty-two hybridomas with high titers were selected for V-gene recovery and sequencing. A few of the hybridomas were non-clonal, yielding multiple light chain (LC) and heavy chain (HC) sequences. By including all HC/LC combinations from clonal and non-clonal hybridomas, a final set of 54 mAbs were produced recombinantly as human IgG1/κ (G1m3) chimeras for further characterization (see methods in supplementary materials).
Epitope binning of basigin monoclonal antibodies. Initially, the 54 basigin mAbs were tested for binding to immobilized human basigin-ECD using biolayer interferometry (BLI), which reduced the number of mAbs to 21 positive binders. Alignment of heavy chain variable domain (V H ) sequences was performed to explore the diversity of the basigin antibody panel. The alignment was used to generate a phylogram for the V H sequences (Fig. 1). The individually V H complementarity-determining region 3 (CDR-H3) families, defined by an identical length and more than 80% amino acid sequences identity, should identify antibodies from the same V-D-J recombination lineage. The CDRs were designated according to the IMGT annotation scheme. LC sequences were not included in the phylogenetic tree. However, the LC sequences were almost identical within the CDR-H3 families. The 21 aligned mAbs were representing nine different CDR-H3 families, subsequently grouped into five epitope bins. In-tandem BLI was used to identify the epitope bins (Fig. 2). Antibodies from the same CDR-H3 family are expected to bind the same epitopes, i.e., belong to the same bin. Here antibodies that compete for antigen binding in a cross-blocking matrix are assigned to the same epitope bin. The biosensors were loaded with recombinant basigin-ECD and sequentially introduced to a saturating mAb (mAb1) and a blocking mAb (mAb2) (Fig. 2a). Examples of blocking profiles from a blocking and a non-blocking antibody pair are depicted in Fig. 2b. The epitope binning data was summarized in the matrix showing the Pearson correlation coefficients between the antibodies blocking profiles, which were calculated based on normalized binning data (Fig. 2c). Antibodies with high positive Pearson correlation coefficient (above 0.9) were placed into the same bin (marked in red). The mAbs in light red indicate a positive Pearson correlation coefficient between 0 and 0.9. Green indicates antibody pairs with non-overlapping epitope regions. All antibodies were run in a selfblocking setup as controls (boxed in the matrix with bold lines). The binning experiment resulted in the mapping of five distinct epitope bins (A, B, C, D, and AD) (rightmost column in Fig. 2c). Basigin mAb #85 showed a unique profile with cross-blocking of antibodies in both bin A and D, indicating that bin A and D epitopes are either in structural proximity or neighboring peptide stretches. Basigin mAbs with a baseline drift after loading above 5% of total binding were only used as the blocking antibody (mAb2) to limit the chance for bin misinterpretation. Two bin B representatives were excluded as saturating mAb based on their fast off-rates (#42 and #59). However, basigin mAb#42 and #59 were in the same CDR-H3 family as the other bin B mAbs, indicating that they were correctly assigned to bin B despite its lack of verification in both assay directions. Similarly, only Scientific RepoRtS | (2020) 10:14582 | https://doi.org/10.1038/s41598-020-71286-2 www.nature.com/scientificreports/ one (#81) of the seven bin C mAbs (#81, #82, #83, #73, #74, #76, and #75) could be confirmed in both assay orientations. As expected, antibodies from the same CDR-H3 families were mapped to the same epitope bind in the BLI assay (last column in Fig. 1). Overall, the alignment showed the diversity of the antibody panel with nine different CDR-H3 families grouped into five different epitope bins by the BLI assay. Bin A is represented by two CDR-H3 families, whereas bin B is comprised of a single CDR-H3 family. Bin C covers three CDR-H3 families, and bind D is represented by two antibodies from different CDR-H3 families. As could be expected, the single candidate representing bin AD belongs to a unique antibody lineage in the panel.
Affinity determination by surface plasmon resonance. The 21 candidates identified in the binning experiments as positive binders were selected for kinetic profiling of their binding to recombinant basigin-ECD by surface plasmon resonance (SPR) (Fig. 3). A pattern between epitope bin and target affinity was observed by comparing the binding kinetics measured by SPR (Table 1). Bin A mAbs have the highest affinities, followed by bin B, whereas most of bin C mAbs were low in affinities. Basigin mAb#85 in bin AD showed strong binding with a K D of 0.4 nM together with bin A antibodies that all have a high affinity in the low nanomolar range (0.3-4 nM). The basigin antibodies in bin B covered K D values ranging from 7 to 40 nM. The bin B mAbs originate from the same CDR-H3 family, but the CDR-H3 sequences did not contribute to the differences in affinity observed within epitope bin B. Bin C mAb#81 and #82 had binding affinities ranging from K D values of 10 to 20 nM, whereas mAb#73-76 and #83 had poor affinities with K D values above > 100 nM. The K D values of antibodies in bin D were calculated as 6 nM (mAb#79) and 20 nM (mAb#80). A slower dissociation of mAb#79 caused this difference. All bin A mAbs, mAb#85 (bin AD), and mAb#79 (bin D) showed fast association rates and slow dissociation rates. The dissociation rates were faster for bin B but showed similar association rates as the other epitope bins. Some of the kinetic constants were difficult to determine due to a poor data fit (Chi2 values in the last column in Table 1). The candidate set was dominated by antibodies with fast association kinetics, which posed a general challenge for proper kinetics analysis. The sensorgrams for basigin mAb#73-76 illustrated the combined fast association and dissociation rates observed for these antibodies, which made determining proper kinetic constants technically difficult (Fig. 3e). The fast on-rate kinetics observed for several candidates may contribute to the unspecific binding phase observed as an association "hump" in the sensorgrams (e.g., mAb#83 and mAb#73-76) and resulting in a corresponding negative response in the dissociation phase (Fig. 3e). The effect was particularly pronounced for mAb#75. For candidates with K D values higher than 40 nM, ligand saturation was not reached as expected in the analyte titration from 0-600 nM. The binding stoichiometry, when calculated based on the observed maximal response (R max ) and the theoretical R max for a 2-binding site ligand, was 0.7 or less for the higher K D antibodies.
Screening of cell association to human brain endothelial cells using flow cytometry. Since the BLI and SPR assays were done with a recombinant basigin-ECD, it was also important to screen all the basigin mAbs for binding to the cell surface-expressed basigin receptor. The evaluation of cell association was by flow cytometry using the human brain endothelial cell line, hCMEC/D3, which has endogenous expression of the basigin receptor. All the bin A, B, D, and AD mAbs were positive for binding the hCMEC/D3 cells with up to   www.nature.com/scientificreports/ Immunofluorescence staining of hCMEC/D3 cells with different basigin monoclonal antibodies. Sixteen representative mAbs from the different epitope bins were further characterized by immunofluorescence for their ability to interact with the basigin receptor in fixed hCMEC/D3 cells (Fig. 5). Bin A basigin mAbs generally bind the hCMEC/D3 cells with mAb#52 showing the strongest signal. Bin B antibodies yielded both strong, weak, and no binding signal on fixed hCMEC/D3 cells. Surprisingly, bin B mAb#69, #71, #42, and #59 were among the antibodies showing weak or no binding to the fixed cells even though these candidates showed strong binding to the hCMEC/D3 cells by flow cytometry analysis (Fig. 4). For bin C, where none of the antibodies associated with the living cells, mAb#82 showed differentiated, robust signal on fixed hCMEC/D3 cells. Bin D mAb#80 bound strongly to the cells, whereas mAb#79 only resulted in weak staining. The immunofluorescence staining of fixed cells can be used for examination of the subcellular binding pattern of the antibodies. Immunofluorescence staining with mAb#80 (bin D) seemed to be detecting a perinuclear localization of the basigin receptor (XZ projection pictures in Fig. 5). In contrast, bin A mAb#52 seemed to detect basigin receptors closer to the cell surface. Basigin mAb#41 (bin B) behaved differently compared to the rest of the antibodies displaying an actin-like staining pattern. In summary, the immunofluorescence staining revealed that some basigin mAbs were more influenced by fixation of the epitope than others and that different sub-populations of the receptor with different local concentrations and epitope representation may exist.
Internalization of selected basigin monoclonal antibodies. Based on our results, five bin representative basigin antibodies, mAb#52 (bin A, K D 3 nM), mAb#43 (bin B, K D 20 nM), mAb#82 (bind C, K D 20 nM), mAb#80 (bind D, K D 20 nM), and mAb#85 (bin AD, K D 0.4 nM), were selected and examined for their ability to support and trail receptor internalization in the hCMEC/D3 cells (Fig. 6). The antibodies were selected based on robust immunofluorescence staining of the hCMEC/D3 cells. The internalization was determined using highcontent screening microscopy with 43 images per condition in three independent experiments and plotted as the percentage increase in spot intensity per cell compared to the control after acid-removal of surface-bound mAbs. The effect of the acid-removal was tested on surface stained cells with and without acid stripping, which resulted in a decrease in spot intensity in acid-treated, indicating that the spots quantified are mainly internalized mAbs (supplementary Figure S3). Additionally, representative confocal z-stack images were manually taken of the different time points (Fig. 6a). A time-dependent increase in basigin mAb internalization was observed www.nature.com/scientificreports/ for mAb#52, #43, #82, and #80, whereas high-affinity mAb#85 reached the same level of basigin-mediated internalization already at 10 min and did not change from 10 to 30 min (Fig. 6b). Basigin mAb#80 resulted in a much higher spot intensity compared to the other antibodies, which was not surprising since mAb#80 associated strongly with the hCMEC/D3 cells in the immunofluorescence staining (Fig. 5). However, the acid stripping was not efficiently removing mAb#80 from the cell surface, and some of the signal could, therefore, be from surfacebound mAbs. All the tested mAbs appear to allow for receptor internalization, making them relevant candidates as engineering partners to facilitate basigin-mediated transcytosis of therapeutic proteins across the BBB.

Discussion
Zuchero et al. 1 showed that targeting basigin increased the uptake of antibodies across the BBB in vivo. However, they did not provide details on their basigin mAb epitopes. Many studies have demonstrated that antibodies can be transported across the BBB, reaching the brain parenchyma using RMT 29 . However, in most cases, the antibodies used for these experiments are only selected based on their affinity and capability to bind the endogenous *indicates that the pictures were processed differently compared to the rest due to avoid over-exposed images.
Here, the antibodies were selected by their epitope, which gave a broader panel of antibodies with possibly more different biological characteristics. Based on binding to recombinant basigin-ECD by BLI, 21 mAbs out of the 54 were selected for further characterization. The mAbs were initially differentiated based on their epitope binding region, affinity, and ability to bind the basigin receptor on living cells. For comparison, the data for the 21 mAbs are summarized in Table 2. The alignment of the 21 mAb V H sequences revealed nine CDR-H3 families, which were group into five epitope bins. Since it is known that affinity to the receptor can affect the transcytosis ability of the antibodies targeting the BBB 25,26 , the binding affinity of the selected 21 mAbs was determined.
The flow cytometry analysis of the hCMEC/D3 cells showed that some basigin mAbs that did not associate with the endogenous receptor expressed in the hCMEC/D3 cells still bind basigin-ECD in the BLI analysis (Figs. 2  and 4). The BLI experiment and the flow cytometry analysis of the basigin mAbs allowed for both monovalent and bivalent binding to basigin. The bivalent interaction was dependent on the density of the basigin on the surface of the cells or biosensors, which could cause the differences between the positive binders in the epitope binning and flow cytometry analysis. The bin C mAbs did not bind the cells in the flow cytometry analysis at this concentration, and a higher concentration may be needed for cellular binding of the low-affinity bin C mAbs (K D values > 100 nM).
Out of the 21 basigin mAbs, 16 were selected based on their epitope bin and affinity for the candidate selection step. The 16 mAbs represented all the epitope bins and had K D values below 40 nM. They were tested for immunofluorescence staining of hCMEC/D3 cells to determine whether their epitope and/or affinity was important for their ability to detect the cellular basigin receptors. Overall, the differences in interaction with the basigin within each epitope bin were not explained by their differences in affinity. However, to thoroughly www.nature.com/scientificreports/ investigate the affinity dependency, affinity engineering of representative basigin mAbs would be needed to exclude the influence of individual and local epitopes. The basigin mAbs detection of the basigin receptor was also independent of their epitope, as an intense immunofluorescence staining was observed across the epitope bins. The contradictory results obtained with mAb#82 in the flow cytometry analysis and immunofluorescence staining could be explained by the fixation step, which could modulate the epitope making it more susceptible. In contrast, basigin mAb#42 did not reveal any immunofluorescence signal in the staining of cells, indicating that the antibody binds a conformational epitope, which cell fixation disrupts. The basigin receptor exists in different glycosylated forms. It has been reported that the cellular localization of basigin is dependent on its glycosylation state. The lowly glycosylated basigin is suggested to be localized primarily in the endoplasmic reticulum, whereas the highly glycosylated basigin is proposed to be at the cell surface 22 . The glycosylation sites were conserved across species, indicating a functional or structurally important role of the glycan modifications 20 . Notably, the N-glycosylation is shown to be different between tissues and species 30 . The basigin receptors can form dimers, but whether the glycosylation affects the oligomerization is still debated. Glycosylation has been reported to enhance basigin receptor dimerization [21][22][23] , whereas other studies found that glycosylation was not essential 31,32 . The different glycosylated forms and dimerization patterns of basigin could explain the different abilities of the anti-basigin mAbs to detect membrane-associated basigin and their subcellular staining pattern. The extracellular domain of human basigin was subjected to enzymatic deglycosylation ( Figure S4). The deglycosylation resulted in a shift in molecular weight shown by Western blotting. However, the deglycosylation did not affect the detection of human basigin using basigin mAb#52 (bin A) and mAb#80 (bin D) as these binds both the highly and lowly glycosylated forms.
Internalization is a crucial step in the selection process to identify mAbs capable of crossing the BBB. Five basigin mAbs, one representative for each epitope bin, were selected for internalization as part of the functional characterization. The internalization seemed to be increased from 10 to 30 min for all mAbs but mAb#85. The basigin-mediated internalization of the high-affinity mAb#85 did not change over time, indicating that mAb#85 has reached saturation already at 10 min. Basigin mAb #43, #82, and #80 had all K D values of 20 nM, but their uptake levels were quite different. The difference could be explained by their epitope or receptor oligomer preference. mAb#80 resulted in a high internalization signal compared to the others, but the acid stripping of the hCMEC/D3 cell surface did not efficiently remove the mAb#80 surface binding, suggesting that mAb#80 were sticking to the cell surface. Also, the very strong interaction in the immunofluorescence staining could indicate that mAb#80 interacts with a specific subset of basigin receptors or that it exerts some unspecific interactions. The antibodies interact specifically with basigin shown by a substantial decrease in Western blotting band intensity after siRNA knockdown of basigin compared to samples were scramble siRNA was added ( Figure S5). The subcellular localization of the internalized basigin mAbs was hard to interpret due to shrinkage of the cytoplasm after acid treatment.  www.nature.com/scientificreports/ We have generated a diverse panel of basigin mAbs binding to different epitope regions. In terms of the transcytosis capability of antibodies, a lot of the studies conclude only based on one epitope on a given receptor. It is important to explore the BBB crossing based on both affinity and epitope to be able to select lead candidates for drug delivery. Even though a previous study demonstrated that there was no correlation between epitope and function of basigin mAbs in T cells, this needs to be further analyzed in brain endothelial cells with our repertoire of antibodies 33 .
It remains to be confirmed if these basigin mAbs can transcytose with the receptor through the brain endothelial cells. The hCMEC/D3 cells are not optimal for testing transport across the cell layer due to the low tightness of the barrier model 34 . For such experiments, BBB in vitro models based on primary brain endothelial cells or human induced pluripotent stem cells are tighter and more suitable 35 . The basigin mAbs have cross-reactivity to at least two species (human and pig) out of the four tested (human, pig, rat, and mouse). The cross-reactivity eases test of the mAbs in preclinical models and progression of the lead candidates into clinical studies. The porcine cross-reactivity is convenient for in vitro modeling of the transcytosis using the robust porcine BBB model as well as a good human pharmacokinetic prediction using pigs for the in vivo experiments 36,37 .
Our work resulted in an excellent and broad panel of anti-basigin mAbs. If mAb#80 (bin D) is primarily binding the basigin receptor retained on the cell surface, it would not be relevant for basigin-mediated BBB delivery. Bin C mAbs cannot be excluded as possible candidates, but based on our studies, this concentration revealed weak cellular association and low uptake of mAb#82. In conclusion, bin A, B, and AD mAbs seemed to be promising candidates for further investigation based on their basigin-mediated internalization. The thorough characterization of the antibodies reported here provides valuable information for the further preclinical in vitro and in vivo characterization of the antibody repertoire and basigin-based BBB transportation.

Methods
Epitope binning using biolayer interferometry. The epitope binning assay was performed at Octet RED384 (Pall Fortebio) in 384-tilted well plates in a tandem setup. All samples were prepared fresh in 1 × PBS-P + buffer-10x (GE Healthcare Life Sciences, cat#88,995,084), which also was used as an assay buffer. First, the Dip and Read streptavidin (SA) biosensors for kinetics (Pall Fortebio, cat#18-5,019) were dipped in assay buffer for 200 s to record a baseline step. Next, 0.5 µg/ml of the biotin-tagged basigin-ECD was loaded on the biosensors for 200 s. After 200 s wash in buffer, the basigin-ECD-coated biosensors were dipped in 100 nM saturating basigin mAb (mAb1) for 1,200 s to reach saturation. The mAb1-basigin coated biosensors were washed for 200 s in buffer and then moved to wells containing an array of blocking basigin mAbs (mAb2) for 1,200 s. Complete self-blocking was ensured with all the basigin mAbs. Basigin mAbs with a baseline drift > 5% were only used as a blocking antibody. The data analysis was performed by using ForteBio Data Analysis software (version 10.0, Fortebio, Fremont, CA, USA), and epitope binning matrices were exported to Microsoft Excel. Data normalization was conducted by division of the mAb2 signal by the reference signal (mAb2 only) and multiplied with 100. The Pearson correlation coefficients were calculated by rows, and columns were ordered according to their correlation. The highest correlations defined as Pearson's correlation coefficient above 0.9 (red), weak correlation as between 0 and 0.9 (light red), and negative correlation below 0 (green).
Surface plasmon resonance analysis of basigin monoclonal antibodies. Binding affinity analysis of basigin mAbs to basigin-ECD was performed on a Biacore S200 and T200 (GE Healthcare Life Sciences) using a kinetic capture setup on a CM4 chip. The CM4 surface was immobilized with a monoclonal mouse anti-human IgG (Fc) antibody (GE Healthcare Life Sciences, cat#BR100839) using amine coupling (GE Healthcare Life Sciences, cat#BR100050). The chip surface was initially activated with a 1:1 (v/v) mixture of 75 mg/ml 1-ethyl-3-(3dimethylaminopropyl) carbodiimide hydrochloride (EDC) and 11.5 mg/ml N-hydroxysuccinimide (NHS). The anti-human IgG antibody was diluted in 10 mM sodium acetate to 25 µg/ml and immobilized at a flow rate of 10 µl/min for 7 min. The activated carboxylate groups were blocked by the injection of 1 M ethanolamine hydrochloride-NaOH, pH 8.5. During immobilization, HBS-EP + (GE Healthcare Life Sciences, cat#BR100188) was used as a running buffer. The basigin mAbs were adjusted to a concentration of 1 µg/ml in running buffer and captured on anti-human Fc mAb immobilized sensor followed by injection of basigin-ECD for 200 s and buffer for 300-600 s (association and dissociation, respectively) at a flow rate at 30 µl/min at concentrations ranging from 600-0 nM. No basigin antibody was captured in the reference flow channel (FC1). At the end of each cycle, the sensor surface was regenerated in 3 M MgCl 2 for 30 s. 1 × HBS-P + (GE Healthcare Life Sciences, cat#BR-1003-68) with 1 mg/ml bovine serum albumin (BSA) (Rockland, cat#BSA-30) was used as running buffer for the kinetic analyses. Fitting of reference (FC1) subtracted data was performed using the Biacore S200 and T200 evaluation software (version 1.0, GE Healthcare Life Sciences, Uppsala, Sweden) using the pre-defined Langmuir 1:1 interaction model. The association rate constant (k a ), dissociation rate constant (k d ), and flow rateindependent component (tc) were fitted global, R max fitted local, and bulk response (RI) fitted constant. For some of the kinetics analyses, the high concentration of basigin-ECD was excluded due to binding to the reference. www.nature.com/scientificreports/ cytometry buffer, cells were incubated with 2.5 µg/ml secondary goat anti-human IgG Alexa Fluor 647 (Jackson ImmunoResearch, cat#109-605-008) for 30 min at 4 °C. The cells were fixed with BD Cytofix/CytoPerm (BD Biosciences, cat#554,714) for 15 min on ice. As negative controls for the staining procedure, cells were either incubated with secondary goat anti-human IgG Alexa Fluor 647, Live/dead fixable violet dead cell stain, or with buffer only (unstained). Since all the basigin mAbs are the same isotype, the basigin mAbs that did not bind the cells worked as Isotype controls. The samples were acquired using FACSVerse flow cytometry system (BD Biosciences, San Jose, CA, USA) with the BD FACSuite software calibrated with FACSuite CS&T research beads (BD Biosciences, cat#650,622), and analyzed using FlowJo (version 10, Tree Star, Inc., Ashland, OR, USA). The cells were gated based on the forward and side scatters to remove cell debris with an additional viability gate. Pulse geometry gating was used to remove doublets, and the final flow cytometry analysis was based on 50,000 events collected in the single-cell gate for each sample (Supplementary Figure S1). The setup was run three times, and the median fluorescence intensities were log-transformed and plotted in GraphPad Prism 8.0 (GraphPad Software, Inc, CA, USA). The fold change was back-transformed in order to describe the mean differences between samples.
Internalization assay. hCMEC/D3 cells were seeded in 96 well plates (Perkin Elmer, cat#3,904) coated with rat tail collagen type 1 at a cell density of 15,000 cells per well. The cells were incubated with 1 µg/ml of the selected basigin mAbs for 10 and 30 min. The cell surface was acid stripped with 0.2 M acetic acid/0.5 M NaCl in PBS for 4 min on ice to remove surface-bound antibodies (as described in 38 ) and subsequently washed in PBS and fixed in 4% PFA for 10 min at room temperature. After permeabilization with PBS 0.1% Triton X-100 (Sigma-Aldrich, cat#T9284) for 10 min, the internalized basigin antibodies were detected by 1 h incubation at room temperature with Alexa Fluor 568 conjugated goat anti-human secondary antibody (2 µg/ml) (Invitrogen, cat#A21090) diluted in PBS with 2% BSA. Nuclei were stained with Hoechst34580 (2 µg/ml) (Invitrogen, cat#H21486) diluted in H 2 O for 10 min. After being washed, the cells were analyzed in PBS. Cells only treated with secondary antibody were used as a negative control. All samples were in triplicates, and the experiment was performed three times. Quantification of basigin mAb internalization was done using Cellomics Arrayscan VTI setup (Thermo Fisher Scientific) equipped with an Orca-ER camera (Hamamatsu) using a Zeiss 20 × LD Plan-Neofluar (0.4 numerical aperture (NA)) objective and analyzed with the Spot Detector BioApplication. The algorithm was set to exclude dead cells based on nuclear size and fluorescence intensity. The intensity of the spots with fluorescence intensity greater than the pre-defined background level was quantified and normalized to the cell number per field. No region of interest was applied in this experiment. The three repeated experiments were normalized to the negative control and plotted as internalization signal above control levels in GraphPad Prism 8.0 (GraphPad Software, Inc, CA, USA).