Synergy between B cell receptor/antigen uptake and MHCII peptide editing relies on HLA-DO tuning

B cell receptors and surface-displayed peptide/MHCII complexes constitute two key components of the B-cell machinery to sense signals and communicate with other cell types during antigen-triggered activation. However, critical pathways synergizing antigen-BCR interaction and antigenic peptide-MHCII presentation remain elusive. Here, we report the discovery of factors involved in establishing such synergy. We applied a single-cell measure coupled with super-resolution microscopy to investigate the integrated function of two lysosomal regulators for peptide loading, HLA-DM and HLA-DO. In model cell lines and human tonsillar B cells, we found that tunable DM/DO stoichiometry governs DMfree activity for exchange of placeholder CLIP peptides with high affinity MHCII ligands. Compared to their naïve counterparts, memory B cells with less DMfree concentrate a higher proportion of CLIP/MHCII in lysosomal compartments. Upon activation mediated by high affinity BCR, DO tuning is synchronized with antigen internalization and rapidly potentiates DMfree activity to optimize antigen presentation for T-cell recruitment.

concomitantly downregulated during B cell engagement with antigen 19,30 . As an initial step towards dissection of this correlation in a physiological setting, we first used a panel of related, simplified model cell lines 12 , generated by sequential transfection of MHCII null /DM null /DO null TxB hybridoma T2 cells with the key molecules of interest: a single MHCII allele, HLA-DR4 (DRB1*04:01), then DM and then DO. CLIP peptides associated with DR4 were prominent in T2DR4 cells (DM null /DO null ), whereas expression of DM in T2DR4DM cells (DO null ) resulted in their near-complete removal, indicating sufficient catalysis by DM and available (extracelluar and endogenous) sources of high affinity peptides to replace CLIP (Fig. 1a). Consistent with the inhibitory role of DO on DM activity, the expression of DO in T2DR4DMDO cells significantly impaired CLIP removal. Unexpectedly, although over 99% of T2DR4DMDO cells expressed comparable levels of DM/DO, half of these DO + cells were unable to limit DM catalysis of CLIP removal (Fig. 1a).
To investigate inter-cell differences in DM/DO effects on CLIP removal, we evaluated T2DR4DMDO cells individually using index single-cell fluorescence-activated cell sorting (iFACS). Representative single cells with significantly different levels of total CLIP (CLIP tot ) showed almost identical levels of DM tot , measured as the index fluorescence intensity (FI DM ) (Fig. 1b). Differences in DR tot or DO tot among these cells were also not significant (Fig. 1c), unless we compared the average index FI by dividing hundreds of single cells into CLIP lo and CLIP hi groups ( Supplementary Fig. 1a). Importantly, we found that a new parameter DM free /CLIP freq revealed the influence of DM/DO stoichiometry on CLIP removal. DM free (=DM tot − DM DO-associated = FI DM /FSC − FI DO /FSC) represents the density of free DM, where DO-associated DM was estimated by DO tot , as almost all DO in vivo is associated with DM 25,31,32 . CLIP freq (=CLIP tot /DR tot = FI CLIP /FI DR ) reflects the frequency of CLIP-loaded DR (designated by "CLIP" hereafter). We found that DM free /CLIP freq was inversely related to DO tot level and reflected the efficacy of DM editing (Fig. 1d).
As DM free /CLIP freq is based on single-cell measurements, we generated clonal lines expanded from individually sorted single T2DR4DMDO cells ( Supplementary Fig. 1b) and tested the feasibility of using DM free /CLIP freq to reveal the correlation between DM/DO regulation and CLIP release. We calculated DM free /CLIP freq per line based on the mean FI of each protein of each clonal line ( Supplementary Fig. 1c). Similar to the iFACS results with single cells, clonal lines isolated from the CLIP lo group with lower levels of mean FI DO had higher DM free /CLIP freq per line than those from the CLIP hi group (Fig. 1e). We also cultured 10 single clonal lines under no selection pressure for DO expression for a week and tested the associated outcome of DO downregulation indicated by DM free /CLIP freq . The decrease of DO expression due to a lack of selection resulted in an increase of DM free /CLIP freq per line (Fig. 1f). A similar inverse correlation was also observed when another 6 single clonal lines were cultured with selection for increased DO expression ( Supplementary Fig. 1d). Unexpectedly, the mean FI CLIP per line was not coordinately decreased when the mean FI DO of the corresponding clonal line diminished (Fig. 1g), due to the emerging heterogeneity in cells expanded from single clones ( Fig. 1h; Supplementary Fig. 1e). This result implies the necessity of single-cell measurements in the study of DM/DO function even when the variation of DM/DO is as small as within a single clonal line, and indicates DM free /CLIP freq rather than the actual CLIP level 14,33,34 as an efficient measure of the physiological outcome of DM/DO regulation. Do regulation tunes DM free to facilitate cLip int release and limit cLip srf . Consistent with flow cytometric studies, three-dimensional structured illumination microscopy (3D-SIM) detected very low levels of CLIP in almost half of T2DR4DMDO cells (Fig. 2a). 3D-SIM also showed high levels of surface CLIP (CLIP srf ) peptides on CLIP hi cells. As DM primarily resides in the intracellular MHCII-containing compartments (MIIC) 25,26 , the engagement of CLIP srf by DM becomes spatially limited. This raised a question as to whether there was any difference in protein localization in CLIP lo and CLIP hi cells that affected CLIP removal by DM free .
To address this question, which is essential to the understanding of the physiological meaning of DM free / CLIP freq , we compared localization of DM and DO in two representative single clonal lines 1C3 (CLIP lo ) and 2D7 (CLIP hi ), each of which had minimum clonal variation in cell cultures with appropriate selection for DO expression. 3D-SIM analysis of both 1C3 and 2D7 cells showed that, by staining, most DM and DO molecules were in close proximity (<200 nm apart) to lysosomal-associated membrane protein LAMP1 ( Fig. 2b; Supplementary Movies 1,2), but not to early endosome antigen EEA1 or late endosome Ras-related protein RAB7 ( Supplementary  Fig. 2a). The surface-surface co-localization extension built in the Imaris v9.4 software (ImarisXT, Bitplane Inc, http://bitplane.com) enabled calculation of the percentage of DM and DO that co-localized (<80 nm in xy plane and <160 nm in z axis) with each other or with tested endosomal markers ( Supplementary Fig. 2b). This super resolution measurement revealed that, in both 1C3 and 2D7 cells, the majority of DO co-localized with LAMP1 and DM (Fig. 2c,d), confirming the lysosomal destination of DO, whose intracellular trafficking requires DM association 25,31,32 . Varied DM/DO ratios in DO + 1C3 and 2D7 or DO null T2DR4DM cells (Fig. 2e) had minimal influence on the lysosomal localization of DM (Fig. 2f); however, a significantly lower percentage of DM co-localized with DO in 1C3 than in 2D7 cells (Fig. 2g). These results demonstrate unchanged localization of lysosomal DM free in cells with varied levels of DM/DO, although in CLIP hi clones, less DM free contributed to the lower DM free /CLIP freq .
We also evaluated the distribution of CLIP in DM + /DO + 2D7 vs DM null /DO null T2DR4 cells to determine any differences between the two cell lines. Super-resolution SIM allowed separation of CLIP srf and CLIP int , and further (n = 3) of CLIP srf (line) and CLIP tot (filled) in T2, 2D7 or T2DR4 cells. MFI of CLIP int = MFI of CLIP tot -MFI of CLIP srf . (j,k) Comparisons of %CLIP or %DR co-localized with the indicated marker in 2D7 and T2DR4 cells (see also Supplementary Fig. 3). N cell (same for j,k): 10, 11, 4, 7, 30, 37. (l) The functional correlation between varied levels of DM/DO and their regulation of CLIP removal. Transparency is related to protein levels per cell. Data are represented as mean ± SEM. ns: non-significant, p > 0.05; *p < 0.05, ****p < 0.0001.   ; Fig. 2i), which is mostly co-localized with LAMP1 (Fig. 2j). We also noted the presence of slightly more DR co-localized with LAMP1 in 2D7 versus T2DR4 cells (Fig. 2k), but not to the degree that is likely to fully explain the increased proportion of CLIP int . Rather, increased CLIP int percentage in lysosomal compartments is likely attributable to DM/DO expression in 2D7 cells. Indeed, the presence of slightly more DR co-localized with LAMP1 in 2D7 versus T2DR4 cells (Fig. 2k) also likely reflects a functional outcome of DM/DO expression, as the presence of these chaperones may enhance the ability of DR to survive lysosomal pH [35][36][37] and/ or increase its residence time in the compartment.
The study of model cell lines demonstrates that both DM/DO expression and their stoichiometry influence the efficacy of CLIP removal and the intracellular accumulation (residence time) of CLIP LAMP1 . DM free /CLIP freq is a sensitive measure for DM free at the LAMP1 + intracellular checkpoint that catalyzes CLIP int removal and limits CLIP srf (Fig. 2l). Lowering DO promotes the increase of DM free /CLIP freq , which represents the effectiveness of DM editing.
Memory B cells possess low DM free /cLip freq and high cLip LAMP1 /cLip tot . We next applied DM free / CLIP freq in the comparison of primary B cell sub-populations with varied levels of DM/DO. Human tonsillar B cells were sorted by FACS into three canonical subsets: naïve (IgD + /CD38 −/lo ), memory (IgD − /CD38 −/lo ), and GC (IgD -/CD38 hi ) ( Fig. 3a; Supplementary Fig. 4a) as previously defined 18,19,21 . Bulk analysis of fluorescently labeled B cells by flow cytometry revealed the concomitant decrease of mean FIs representing levels of DO (most dramatically), DM and CLIP in GC B cells, whereas differences between naïve and memory cells were minimal ( Fig. 3b; Supplementary Fig. 4b). Using iFACS to analyze individual cells, we found higher DM free /CLIP freq on average in GC ( Fig. 3c; Supplementary Fig. 4c), indicating higher efficacy of CLIP removal by DM catalysis, which is likely necessary for these cells to receive help from T FH cells. In contrast, the significantly lower mean DM free /CLIP freq in memory than the other two subtypes of B cells implied a higher demand for potentiation of DM free -catalyzed CLIP removal by memory B cells upon activation.
Like the localization analysis in model cell lines, we also determined the site(s) of DM free in the tonsillar B cell subsets. We analyzed DM and DO by SIM and found DM was co-localized with LAMP1 ( Fig. 3d; Supplementary  Fig. 4d and Movies 6-8). The percentage of DM co-localized with various endosomal markers measured by 3D-SIM showed no difference between non-GC (naïve and memory) and activated GC B cells (Fig. 3e). Very similar to the results observed in T2 cell lines, the majority of DO co-localized with DM in non-GC DO + B cell sub-populations (Fig. 3f). However, a significantly higher percentage of DM in memory than naive cells co-localized with DO ( Fig. 3g), indicating that less DM free at LAMP1 + compartments resulted in lower DM free / CLIP freq in memory cells (Fig. 3c). In addition, the quantitative SIM analysis enabled further differentiation of memory from naive cells by CLIP int distribution ( Fig. 3h; Supplementary Fig. 5 and Movies 9,10). Memory cells with the lowest average DM free /CLIP freq showed significantly higher percentage of CLIP (and slightly higher percentage of DR) co-localized with LAMP1 compared with naïve and GC B cells (Fig. 3i), despite indistinguishable levels of CLIP tot or DR tot between memory and naïve B cells (Fig. 3b).
Collectively, these results agree with the model suggested by the analysis of T2 cell lines and indicate that the net intracellular accumulation of CLIP (CLIP LAMP1 /CLIP tot ) inversely correlates with DM free /CLIP freq in primary B cells (Fig. 3j). Non-GC B cells, especially memory (vs naïve) cells, with lower DM free /CLIP freq and higher levels of CLIP-loaded MHCII in DM free + /LAMP1 + compartments, must experience stronger potentiation to efficiently replace CLIP with high affinity MHCII ligands during antigen engagement at pre-GC or GC phases.

Do downregulation potentiates DM free activity in memory B cells.
To evaluate the potentiation of effective editing by active DM free in non-GC B cells, we sought to activate both naïve and memory cells in vitro. After BCR engagement with antigen and T FH cell-B cell interaction at the T/B border, GCs are formed within 24 h 3,38 , indicating that antigen processing and presentation involving DO downregulation occur within hours, as observed previously 30 . In a scheme that ensured in vitro activation of resting B cells mimicking a T-dependent pathway, we cross-linked BCR with anti-IgG and anti-IgM antibodies along with triggering CD40-mediated signaling, using recombinant CD40 ligand (CD40L) 39 . The total level of an early activation marker, CD69, in both non-GC (naïve and memory) but not GC B cells (already activated), significantly increased as early as 2 h post-activation (Fig. 4a). 6 h later, we observed increased levels of another activation marker, the co-stimulatory molecule, CD86 in both naïve and memory cells ( Supplementary Fig. 6a), confirming successful in vitro activation.
We next compared naïve and memory B cells that had been cultured with/without stimuli for only 2 h, when there was negligible false positive auto-fluorescence associated with dead cells emerging (e.g., >4 h) from the in vitro culturing (Supplementary Fig. 6b). At 2 h, moderate and significant decreases of DO levels appeared in naïve and memory cells, respectively, whereas changes in levels of DM, CLIP or DR did not reach significance using bulk analysis (Fig. 4b). Using single-cell measurement, we found significantly increased DM free /CLIP freq in both naïve and memory cells after stimulation (Fig. 4c), and the percentage of increase appeared to be similar in both view of the CLIP channel (red) and the calculated LAMP1/CLIP overlapping surface (yellow). (i) Comparisons of %CLIP or %DR co-localized with the indicated endosomal marker in different subtypes of B cells (see also Supplementary Fig. 5). N cell : 24, 21, 26, 19, 18, 16, 52, 60, 40. (j) An inverse correlation between CLIP LAMP1 / CLIP tot and DM free /CLIP freq among tonsil B cell sub-populations. Data are represented as mean ± SEM. ns: nonsignificant, p > 0.05; *p < 0.05, **p < 0.01; ***p < 0.001; ****p < 0.0001. www.nature.com/scientificreports www.nature.com/scientificreports/ types of B cells (Fig. 4d). In addition, CLIP co-localized with LAMP1 also significantly decreased by a similar percentage in both types of cells (Fig. 4e,f). The activation of naïve and memory cells via cross-linking of BCR drives similar % changes in DM free /CLIP freq and CLIP LAMP1 /CLIP tot , demonstrating the capability of potentiation for CLIP removal in both non-GC subsets. However, the more dramatic changes of DO levels in memory B cells (but not other related protein levels) confirmed their greater dependence on DO regulation to generate greater amount of active DM free and to drive such potentiation. The difference between affinity matured memory cells and their naïve counterparts implies a critical role of DO downregulation during BCR-mediated activation.

Prompt DO regulation is synchronized with high affinity BCR-antigen uptake. To test whether
there is any correlation between antigen-BCR and DO regulation, we examined the functional outcome during antigen-specific activation in human B cell lines, whose immunoglobulin G (IgG) BCRs have affinities that differ by over 2 orders of magnitude for the autoantigen glutamic acid decarboxylase (GAD65): DPA (<1 nM) and DPD (>100 nM) 40 . Flow cytometry detected DM, DO, DR and CLIP in both B cell lines (Fig. 5a).
In an assay where we tested the internalization of BCRs after GAD65-binding 41 , we first limited endocytosis by staining cells with GAD65 on ice and confirmed GAD65-binding on the surface (GAD65 srf ) of both cell lines (Fig. 5b). We then let cells incubate at 37 °C to allow internalization and measured the loss of GAD65 srf or surface human IgG (hIgG srf ). There was a parallel, time-dependent decrease of GAD65 srf and hIgG srf observed in DPA but not DPD cells (Fig. 5c). The internalization was fast, as the majority of protein loss on the surface of DPA cells occurred within 20 min (Fig. 5d). Notably, the background-subtracted GAD65 srf level in DPA cells decreased to a level similar to that of DPD cells (Fig. 5e), suggesting that there is a limit to the degree of internalization and internalization is BCR affinity-dependent. Of note, BCRs of both DPA and DPD cells are internalized after cross-linking by polyclonal antibodies, which overcomes BCR affinity differences ( Supplementary Fig. 7a).
Within the same time frame of GAD65-hIgG internalization, we observed a significant decrease in the level of DO in DPA compared with DPD cells, whose DO level had negligible changes (Fig. 5f). Very importantly, this response in DPA cells was fast, occurring within 20 min, when the proliferation of cell lines had minimal contributions to protein expression. Consistent with this, we observed nearly unchanged levels of DM, CLIP or DR (Fig. 5f), or even CD69 ( Supplementary Fig. 7b) in both cell lines. Similarly, DO levels decrease by 15%, within 20 min of the in vitro activation of human tonsil-derived resting B cells (particularly memory), while the level of CD69 is still unchanged (Supplementary Fig. 7c). Using single-cell measurements, we further found a significant increase of DM free /CLIP freq in DPA but not in DPD cells (Fig. 5g), demonstrating a substantially boosted efficacy for DM free -catalyzed CLIP removal following the quick internalization of antigen-BCR complexes and downregulation of DO in cells that expressed appreciably higher affinity BCRs for the GAD65 antigen.

Discussion
Essential role of DM/DO stoichiometry. The requirement of DO downregulation for B cell activation and entry into GC for proliferation, affinity maturation and selection [18][19][20][21] indicates highly regulated peptide loading machinery inside B cells to optimize T-B interaction for humoral responses. Studies using DO deletion or DO ectopic expression have shown different effects on the functional outcome of B cell presentation of various antigens 15,16,32,42,43 , including both autoimmune and anti-viral responses [44][45][46] . However, the physiological tuning of DO levels and associated changes in DM/DO stoichiometry is more nuanced than a process that alters the regulatory machinery either by completely removing DO or by switching on DO activity.
In this study, we developed a model system to compare B cell lines with a range of DM/DO ratios at steady state. At the single cell level, we were able to capture the functional outcomes of subtle changes in DM/DO stoichiometry and found that the presence of DO does not always completely block DM-catalyzed removal of MHCII-associated CLIP (e.g., cell line 1C3 with a decent level of DO has almost no CLIP). This unexpected finding provides evidence for the essential role of local DM/DO stoichiometry in the CLIP removal process. Importantly, the single cell measurements enabled the discovery of a key parameter, DM free /CLIP freq , which best illustrated this relationship both experimentally and theoretically (detailed below). physiological relevance of DM free /cLip freq to the effectiveness of DM editing. The peptide loading of MHCII involves MHCII synthesis, antigen processing, complex regulations and catalytic reactions. The relevant proteins traffic throughout multiple intracellular compartments and shuttle between these compartments and the plasma membrane 1 . Our single cell analysis has taken various factors into account while focusing on elucidation of the impact of DM/DO expression and their stoichiometry on the efficacy of CLIP removal. Mathematically, DM free /CLIP freq is the slope of a line connecting the origin with each dot that represents an individually analyzed cell on the DM free vs CLIP freq plot and a line can be drawn to represent the mean +/− SEM (Fig. 6a-d). Each comparison reveals the efficiency of active DM free , which scans peptide/MHCII complexes and catalyzes the exchange of bound CLIP (representative of low affinity peptides) for high affinity ligands: the higher DM free /CLIP freq is, the more effective DM editing is. From an enzymatic perspective, a small amount of active DM free molecules should catalyze the replacement of most CLIP (or low affinity) peptides given enough time. However, the varied effectiveness of DM editing, reflected by different DM free /CLIP freq of different B cells, can be with GAD65 at 37 °C for 20 min and fixed/permeabilized before flow cytometric analysis. 1 indicates no change. n = 5. (g) Comparisons of DM free /CLIP freq determined by iFACS of individual DPA and DPD cells that were incubated without stimulus (no stimulus) or pre-loaded with GAD65 and then incubated with GAD65 ( + GAD65) at 37 °C for 20 min. n = 3. Data at each condition include a 96-well plate of single cells. Data are represented as mean ± SEM. ns: non-significant, p > 0.05; *p < 0.05, **p < 0.01. (2019) 9:13877 | https://doi.org/10.1038/s41598-019-50455-y www.nature.com/scientificreports www.nature.com/scientificreports/ understood if one considers both temporal and spatial parameters in these cells. DM editing can reach a limit if the accumulation rate of CLIP surpasses the speed of DM-catalyzed CLIP removal. Reciprocally, this limit may increase the residence time of CLIP/MHCII in MIIC. This is in line with our observation that the 2D7 clonal line or non-GC B cells with low DM free /CLIP freq maintain an appreciable proportion of CLIP/MHCII in LAMP1 + compartments (CLIP LAMP1 /CLIP tot ) at steady state. The spatial parameter contributes to variations in DM free /CLIP freq by mainly affecting the distribution of CLIP, as 3D-SIM analysis detected minimal DM/DO outside of LAMP1 + compartments at steady state and nearly all DO signals were in close proximity to DM signals. This supports previous conclusions that both DM and DO are lysosomal residents and DO requires chaperoning by DM 25,31,32 . Therefore, in contrast to some in vitro suggestions 47 , DO is unlikely to act separately from DM on peptide exchange in vivo. This perspective is consistent with the requirement for DO downregulation rather than upregulation when antigen presentation by B cells is in high demand in the pre-GC response (that is the gate-keeper for GC entry) and in the GC reaction [18][19][20][21] . However, it remains possible that, prior to down-regulation, DO association enhances the specific activity of DM enzymatic function in acidic MIIC through structural stabilization and reciprocal chaperoning, as suggested 12,36 .
Our single cell measurements showed the highest DM free /CLIP freq in GC B cells (Fig. 6b), reflecting the importance of effective DM editing during B cell affinity maturation in GC. The noticeable presence of single cells with particularly low DM free /CLIP freq appearing in the GC population from each tested donor probably reflect differentiated cells with significantly low to zero amounts of total DM (leading to low to zero amounts of DM free ). This possibly reflects shortened DM half-life in the absence of DO chaperoning.

Importance of tunable DO levels in memory B cells.
Analyses using DM free /CLIP freq revealed that local DM/DO stoichiometry, tuned by DO levels, controls the threshold of effective DM editing. The tuning function of DO likely leads to the significantly higher levels of CLIP that survive DM editing and are detectable in non-GC (DM/DO ratio lo ) B cells compared to GC B cells (DM/DO ratio hi-∞ ), consistent with previous observations 19 .
The importance of DO tuning is further revealed by our single cell analysis of memory B cells. Memory B cells isolated from human tonsil tend to possess low DM free /CLIP freq , and high CLIP LAMP1 /CLIP tot , despite lower CLIP srf on memory compared to naïve cells 19,21 and comparable levels of CLIP tot between the two non-GC populations. The elevated percent of MIIC-localized CLIP/MHCII in memory B cells may result from both quiescent DM editing and BCR ligation-induced intracellular accumulation of CLIP/MHCII complexes during prior antigen exposure, as suggested 48 . The apparently quiescent level of DM editing is a result of limited DM activity free from DO inhibition/chaperoning in MIIC rather than insufficient amount of DM tot . Therefore, memory B cells are poised to be potent APC with more peptide-receptive MHCII and optimized active DM free to exchange high affinity ligands efficiently upon antigen re-exposure. Our discovery that DO downregulation is synchronized with BCR affinity-dependent antigen internalization (within 20 min) supports our hypothesis that memory B cells expressing affinity matured BCRs rapidly potentiate DM editing by fine-tuning DM/DO stoichiometry. This response is faster than the time required for transcriptional upregulation of DM or DR that might also contribute to DM editing and peptide loading. The BCR affinity-dependent DO tuning provides insight into mechanisms to optimize memory B cells as APC and to their efficient recruitment into GC.
The important role of DO tuning is also observed in the upregulation of DO in memory B cells (vs GC B cells). It is a tempting idea that, within GCs, B cells (including GC-dependent memory B cells) with relatively lower affinity BCR lack sufficient antigen-internalization as a driver for continuous DO suppression. This could lead to their becoming memory B precursors that restore DO expression to control DM editing and allow bi-directional chaperoning. Another hypothetical model for DO restoration in extra-follicular memory B cells is that the daughter, memory-prone cells inherit the majority of DO molecules as a result of asymmetric division of activated B cells 49,50 with the other daughter cell inheriting little due to its GC B cell fate. This is in line with our observation that there is a great amount of DO left even though DO level decreases during in vitro stimulation (see Fig. 4b).
2-synchronized-affinity-checkpoint model. Our investigation of DO function reveals its role in the synchrony of the aforementioned steps 1 and 2, involving affinity checks for ligand-receptor interaction along the antigen presentation pathway in B cells (Fig. 6e). The surface BCR affinity checkpoint at Step 1 sets a high bar for the specificity of antigen that is recognized and efficiently internalized by high affinity BCRs. The intracellular affinity checkpoint at Step 2 is synchronized with antigen internalization by DO down-regulation, which potentiates DM editing to ensure efficient peptide exchange for high affinity antigenic MHCII ligands. These ligands become locally available in MIIC likely due to the delivery and protection of antigen from degradation by high affinity BCR 51 and the intersection of the intracellular trafficking paths of BCR-bound antigen and nascent MHCII 1,50 . Further, in findings that imply facilitated hand-off of Ig-associated antigen for MHCII binding, we have previously shown that cross-linking of surface BCR increases the subsequent co-precipitation of Ig and free DM 52 . As DO-associated DM does not co-precipitate with Ig, this is likely another benefit of synchronized antigen binding and DO downregulation.
Several lines of evidence imply potential mechanisms underlying rapidly responsive, BCR affinity-dependent DO tuning. Although no other study demonstrates a direct correlation between BCR affinity and downstream effects on regulation of the peptide loading machinery, it has been suggested that the receptor (including BCR) signaling on APCs can lead to acidification of late-lysosomal compartments 53,54 . Therefore, it is possible that high affinity BCR not only favors antigen internalization and protection of antigen from proteolysis before delivery into MIIC 51 , but also triggers signals that drive significant acidification of MIIC. The resultant acid-activated proteases ensure the generation of MHCII ligands from BCR-protected antigen and likely facilitate the destruction of acid-labile DO 12,30 . Indeed, the combined effects of acid and proteases appear to significantly accelerate DO reduction kinetics; in the results reported here, we observed a rapid change in DO levels in activated B cells (0.3-2 h), whereas our in vitro data on DO stability in acidic pH alone showed a timeframe of 1-3 days for DO denaturation 12 .
Because of this fast DO tuning, the established synchrony between antigen-BCR internalization and DM editing ensures quick reciprocal feedback during T-B interaction. This model applies to many situations when B cell antigen presentation and recruitment of T cell help is pivotal, especially at pre-GC and GC phases. In those phases, the fate of a B cell largely relies on BCR affinity and the surface density of antigenic peptide/MHCII complexes 2,3 , which is governed by intracellular DM editing. Our model predicts that antigen presentation by Scientific RepoRtS | (2019) 9:13877 | https://doi.org/10.1038/s41598-019-50455-y www.nature.com/scientificreports www.nature.com/scientificreports/ affinity-matured memory B cells is more effective for recruiting T FH help and driving humoral immunity, as antigen internalization by high affinity BCR promptly potentiates DM editing. This peptide selection results in long-lasting surface peptide/MHCII complexes to support pre-GC interactions that seed and sustain GC reactions. This is particularly important at the time of initial B cell activation by antigen, when local cognate T cell help may not be immediately available.
Differences in DO alleles, albeit subtle, affect DO abundance and activity in humans and mice 44 , and peptidomes differ between DO + and DO null APCs 15,42,55 . As a result of these differences and variations in other parameters that guide B cell responses in different individuals, pre-GC activation of naïve/memory B cells and subsequent GC entry or GC-independent B cell activation may lead to significantly different immune responses that are either beneficial or detrimental (Fig. 6e). In the extreme, DO downregulation by DO-knockout or upregulation by DO ectopic expression may potentiate or attenuate DM editing, respectively, and lead to protective or deficient humoral responses or break self tolerance, as demonstrated in several mouse studies [44][45][46] . However, the basic goal of fine-tuning DM/DO stoichiometry in B cells is to facilitate coordination of antigen-BCR binding and binding of immunodominant peptides from that antigen to MHCII at the two synchronized affinity checkpoints. This system likely evolved to help maintain the immune balance between host defense and self-tolerance. Understanding these underlying mechanisms may inform future strategies to manage humoral immunity for therapeutic benefits.

Human tonsil B cell isolation. Deidentified tonsil tissues were obtained under a Stanford Institutional
Review Board approved protocol (IRB #38079). Tonsil tissues that were removed in a tonsillectomy procedure were minced to small sample pieces. These samples were washed with RPMI 1640 media (without supplements, Thermo Fisher Scientific) and passed through a 70 μm stainless steel cell strainer (BD Falcon) using a glass plunger to create a single cell suspension in 50 ml centrifuge tubes. Cells were then pelleted by centrifugation at 1500 rpm, room temperature (RT) for 5 min, resuspended in 30 ml RPMI 1640, and incubated at RT in a T150 Flask to remove tissue debris. Following 30-60 min of incubation, the cells were gently washed with RPMI 1640, passed through another 70 μm cell strainer into a new 50 ml centrifuge tube, and pelleted by centrifugation at 1500 rpm, RT for 5 min. The cell pellet was resuspended in 2-3 ml PBS + 5% heat inactivated FBS and added to a pre-washed nylon wool fiber column (Polysciences). The column with cells was incubated at 37 °C for 1 h, followed by 2-3 washes with PBS + 5% FBS to remove cells that were not attached to the nylon wool fiber (including T cells). Nylon wool adherent cells, the majority of which are B cells, were released by plunging and collected into a 50 ml centrifuge tube. After washes with PBS + 5% FBS and centrifugation at 1200 rpm for 10 min, cells were counted and resuspended in an appropriate amount of freezing media (RPMI-1640 containing 10% DMSO, 10% FBS, 1% penicillin/streptomycin, and 1% glutamine) for cryopreservation. www.nature.com/scientificreports www.nature.com/scientificreports/ BioLenggend). Antibodies used in the study of BCR-GAD65 internalization included Alexa Fluor 488-conjugated goat anti-human IgG (Thermo Fisher Scientific), monoclonal mouse anti-GAD65 (Millipore Sigma), monoclonal mouse anti-GAD2, (ImmunoGen), and Alexa Fluor 488-conjugated highly cross-adsorbed goat anti-mouse IgG(H + L) (Thermo Fisher Scientific). To generate directly-conjugated CLIP-, DR-, DM-specific antibodies that are not commercial available, CerCLIP.1, L243, Map.DM1 were purified from mouse ascites using protein G beads (GE Healthcare), and then labeled with corresponding Alexa Fluor dye carboxylic acid, succinimidyl ester (Thermo Fisher Scientific), followed by size exclusion chromatography using a Superdex 200 gel filtration column (GE Healthcare) to remove extra free dye.
Single cell cloning and culturing of T2-derived lines. 3 million T2DR4DMDO cells were first labeled with LIVE/DEAD Fixable Dead Cell Stain (Thermo Fisher Scientific) in phosphate buffered saline (PBS) on ice for 30 min, and then washed and resuspended in 0.3 ml PBS + 1% bovine serum albumin (BSA) + 2 mM ethylenediaminetetraacetic acid (EDTA) for surface staining of DR4-associated CLIP. Cells were stained with FITC-CerCLIP.1 (1:5 dilution) on ice for 30 min and washed three times with 0.5 ml PBS + 1% BSA + 2 mM EDTA before resuspension in 0.5 ml PBS + 1% BSA + 2 mM EDTA for single cell sorting. Single cell fluorescence-activated cell sorting (FACS) was performed using a BD FACSAria II (BD Biosciences) at the Stanford Shared FACS Facility. Each cell (CLIP hi or CLIP lo ) was sorted into a well containing 100 μl complete IMDM with GlutaMax, 10% FBS, and 1% penicillin/streptomycin in a 96-well round bottom plate. Another 100 μl of complete IMDM GlutaMAX media were added into each well of the 96-well plate before the incubation of the plate at 37 °C. The top ~150 μl of media was carefully replaced with fresh media every 7 days until a clump of cells was observed at the bottom of the plate. Cells were then transferred to bigger flat-bottom wells of a 48-well plate with doubled or tripled volumes of fresh media + 1 mg/ml G418 (Geneticin) + 1 μg/ml puromycin + 100 μg/ ml zeocin (Thermo Fisher Scientific) to maintain a selection pressure for the expression of DR4, DM and DO in sorted T2DR4DMDO clones. Under the appropriate selection condition, many single clonal lines maintained a consistent expression level of total DR4, DM or DO protein during several weeks of downstream scaling up and culturing. We also observed that the expression levels of these proteins could decrease if the selection reagent was removed from the culture.

Isolation of tonsillar B cell subsets. Human tonsillar B cells were thawed on ice and washed twice with
PBS + 5% FBS. The majority of dead cells and debris were separated from live B cells by density gradient centrifugation using Lymphocyte Separation Medium (MP Biomedicals). B cells were washed again, counted and resuspended in PBS at a density of 10×10 6 cells/ml. Cells were stained with LIVE/DEAD Fixable Dead Cell Stain for 30 min. Following the live/dead staining, the cells were washed with PBS and resuspended in phenol red free RPMI (RPMI 1640 Medium, no glutamine, no phenol red, Thermo Fisher Scientific) supplemented with 3% FBS. Subsequently, the B cells were stained on ice with a cocktail of antibodies including: anti-CD3, anti-CD11c, anti-CD14, anti-CD16, anti-CD19, anti-CD38, and anti-IgD antibodies. After 30-40 min of surface staining, cells were washed in phenol red free RPMI and passed through a 70 μm filter before loading onto the cell sorter. FACS was performed using a FACSAria II at the Stanford Shared FACS Facility using a 70 μM nozzle. B cells (CD3−, CD14−, CD16−, CD11c−, CD19+) were sorted into naïve (IgD+, CD38−), memory (IgD−, CD38−) and GC (IgD−, CD38+) cell sub-populations, as previously described 18,19 . Sorted cells, resuspended in RPMI supplemented with 3% FBS, were kept on ice. Purity of sorted B cell subsets was confirmed by reapplying a sample onto the cell sorter right after sorting.
Activation of tonsillar B cell subsets. Naïve, memory and GC B cells were incubated at a density of 10 6 cells/ml with or without an activation cocktail in IMDM medium supplemented with 10% FBS, 1% Penicillin/ Streptomycin, and 1% glutamine at 37 °C for various time periods prior to analysis by flow cytometry to determine total protein levels, by iFACS to calculate DM free /CLIP freq , or by 3D-SIM to quantify protein co-localization. The activation cocktail contained 20 ng/ml of IL-4 (Peprotech), 20 ng/ml of IL-2 (BioLegend), 200 ng/ml of MEGACD40L (human CD4 ligand, Enzo Biochem Inc), 5 μg/ml of Goat F(ab')2 anti-human IgG-UNLB mouse adsorbed and 5 μg/ml Goat F(ab')2 anti-human IgM-UNLB (Southern Biotech).
Crosslinking of BCR on DP cell lines. DPA or DPD cells were incubated at a density of 2 × 10 6 cells/ml with or without 5 μg/ml of Goat F(ab')2 anti-human IgG in complete IMDM GlutaMAX medium supplemented with 10% FBS, 1% penicillin/streptomycin, and 1% glutamine at 37 °C for an indicated time length. At different time point (i.e., 0, 20 min, 1 h, and 2 h), 0.2 million cells were collected by centrifugation at 400 x g, 4 °C for 5 min and incubated at a density of 10 × 10 6 cells/ml with 10 μg/ml Alexa Fluor 488-conjugated goat anti-human IgG in PBS + 1% BSA on ice for 30 min to stain surface BCR. After two washes with PBS + 1% BSA, cells were analyzed on a FACSCalibur flow cytometer (BD Biosciences).
Stimulation of DP cell lines by GAD65. DPA or DPD cells were incubated at a density of 10 × 10 6 cells/ ml with or without 2 μg/ml GAD65 (Abcam) in PBS + 1% BSA on ice for 1 h to load GAD65 onto BCR while BCR internalization is inhibited. GAD65-loaded or non-loaded cells were washed with PBS + 1% BSA, pelleted by centrifugation at 400 x g, 4 °C for 5 min, resuspended at a density of 2 × 10 6 cells/ml in complete IMDM GlutaMAX medium containing 5μg/ml GAD65 or no antigen, and then incubated at 37 °C for an indicated time length. At 0, 20 min, 1 h, and 2 h, 0.6 million cells were collected by centrifugation at 400 x g, 4 °C for 5 min and used in three different assays: 1) 0.2 million cells were used for iFACS analysis (see below). 2) 0.2 million cells were incubated at a density of 10 × 10 6 cells/ml with 10 μg/ml Alexa Fluor 488-conjugated goat anti-human IgG in PBS + 1% BSA on ice for 30 min to stain surface BCR. 3) 0.2 million cells were incubated at a density of 10 × 10 6 cells/ml with 10 μg/ml anti-GAD65 antibody on ice for 30 min to stain surface GAD65, followed (2019) 9:13877 | https://doi.org/10.1038/s41598-019-50455-y www.nature.com/scientificreports www.nature.com/scientificreports/ by two washes with PBS + 1% BSA and the secondary staining with 10 μg/ml Alexa Fluor 488-conjugated goat anti-mouse IgG(H + L) in PBS + 1% BSA on ice for 30 min. Antibody-labeled cells in 2) and 3) were analyzed on a BD FACSCalibur flow cytometer.
Bulk analysis of total MHC proteins by flow cytometry. Cells were fixed/permeabilized using the Cytofix/Cytoperm reagent (100 µl per million cells, BD Pharmingen) at RT for 20 min, and washed twice with 1x Permwash buffer (BD Pharmingen) prior to immunofluorescent antibody labeling. Washed cells were pelleted by centrifugation at 400 × g, 4 °C for 5 min and resuspended in 1x Permwash buffer containing fluorophore-conjugated DR-, DM-DO-, and CLIP-specific antibodies (~1 µg of each antibody per million cells per 100 µl buffer) and incubated on ice for 1 h. Antibody-labeled cells were washed twice with 1x Permwash buffer and then resuspended in PBS + 1% BSA for analysis on a flow cytometer, either FACSCalibur or LSRII (BD Biosciences). Flow cytometric data were analyzed using FlowJo software (Tree Star, Inc.).  www.nature.com/scientificreports www.nature.com/scientificreports/ permeabilized cells were exposed to the high intensity white light from the Metal Halide Light Source (MME-250, 250 W type, TRI Vision, www.triv.co.kr) for 2 h on ice to completely photo bleach residual fluorophores (validated by flow cytometry and 3D-SIM). Fixed/permeabilized cells were then stained with antibodies specific for an endosomal marker and selected proteins in 1x Permwash buffer + 50 mM glycine using the same approach as used above for flow cytometric analysis. The excess of antibodies in the staining solution were depleted by extensive washes: 2-3 washes using 1x Permwash + 50 mM glycine followed by 2-3 washes using PBS + 1% BSA + 50 mM glycine. Glycine was added to inhibit non-specific binding. Alex Fluor dyes were used due to their brightness and photostability. There was minimal difference when detecting endosomal localization of MHC proteins by a direct staining (i.e., using Alexa Fluor dye directly conjugated LAMP1, EEA1 or RAB7) versus an indirect staining (i.e., using rabbit anti-LAMP1, EEA1 or RAB7, followed by secondary Alexa Fluor dye conjugated anti-rabbit antibodies). In experiments that generated data for quantification of surface-surface co-localization, the direct staining was selected to minimize signal to noise ratios. Alexa Fluor 488 was used for endosomal markers, Alexa Fluor 568 was used for CLIP or DO, and Alexa Fluor 647 was used for DR or DM. Antibody-labeled cells were resuspended in PBS + 1% BSA + 50 mM glycine at a density of 10 6 cells/ml and 100 μl transferred to an empty Shandon single cytofunnel with white filter cards (Thermo Scientific Fisher) that had been pre-treated by washing with 1 ml PBS followed by cytospinning. The funnel with cell samples was then assembled with a high precision cover glass (24 × 50 mm, 170 ± 5 μm, No. 1.5 H, MARIENFELD) and a Shandon cytology stainless steel slide clip (Thermo Scientific Fisher) to cytospin the cells onto the coverslip and form a monolayer cell pellet. The cytospin was performed using a cytocentrifuge at slow acceleration mode 500 rpm for 5 min. After disassembling, the coverslip was air-dried briefly and the spot where the monolayer cell pellet formed was covered by 14 μl of SlowFade Gold Antifade mountant with DAPI (Thermo Fisher Scientific). The coverslip was then inverted and mounted over a pre-cleaned Gold Seal plain micro slide (25 × 75 mm, Thermo Fisher Scientific) and the edge was sealed by nail polish.

Determination of intracellular
3D-SIM data acquisition and reconstruction. Applied Precision N = 1.514 immersion oil (GE Healthcare) was applied on the coverslip. The micro slide was inverted and placed in position over the objective lens (U-PLANAPO 100×, N.A. 1.40) on the OMX V4 microscope platform with the sample and coverslip side facing down. Images were acquired at SIM mode using Applied Precision DeltaVision OMX imaging system (GE Healthcare) with four lasers (100 mW: 405 nm, 488 nm, 568 nm; 300 mW 642 nm MONET) and 3 emCCD detection cameras. The wavelengths for emission filters are 435/31 nm for the DAPI or equivalent dye channel, 528/48 nm for the Alexa Fluor 488 or equivalent dye channel, 609/37 nm for the Alexa Fluor 568 or equivalent dye channel, and 683/40 nm for the Alex Fluor 647 or equivalent dye channel (detection using the same camera as for the DAPI channel). Signal (maximum) to noise (minimum) ratio at each channel was optimized (>10 fold), and the laser power (%T) was set to ≤10% and exposure below ≤80 ms to minimize photobleaching. The size of T2 and tonsillar B cells are ~5-10 μm and ~2-5 μm, respectively. A 512 × 512 pixels (0.08 μm/pixel) XY microscopic field of interest, or more accurately, a 3D space in between the coverslip and the micro slide that contains the monolayer cell specimen was scanned along Z-axis at 0.125 μm steps. ~50-80 scanned optical sections of the 3D space containing the specimen yielded a stack of ~50-80 2D XY images. The stack of images was then reconstructed into a volumetric 3D image using softWoRx 7.0.0 image processing software (GE Healthcare). OMX structured illumination (SI) reconstruction was performed using channel-specific optical transfer function (OTF) with channel-specific wiener filters: 0.0030 for the DAPI channel, 0.0010 for the Alexa Fluor 488 channel, 0.0010 for the Alexa Fluor 568 channel, and 0.003 for the Alex Fluor 647 channel. OMX images were corrected for chromatic aberration and aligned using the blue-green-red (BGR) image source drawer.
Calculation of surface-surface co-localization using 3D-SIM data. The SI reconstructed 3D image was viewed and analyzed using Imaris v9.4 image analysis software (ImarisXT, Bitplane Inc, http://bitplane.com). For each source channel, a volumetric surface connecting voxels (0.08 um x 0.08 μm x 0.08 μm) that contain fluorescence signals associated with the target protein was generated to simulate the localization of the protein.
To minimize noise, the algorithm allowed smoothing by 0.08 μm "Surfaces Area Detail Level" and background subtraction by 0.16 μm "Diameter of largest Sphere that fits into the Object". The filter used to classify surface was "Number of Voxel above" 5. With volumetric surfaces available at two source channels, an overlapping surface was then calculated using the "surface-surface colocalization" Xtension, a METLAB-based algorithm built as an extension patch for Imaris. This Xtension generated a new channel for the overlapping surface that includes voxels shared by the two source channel surfaces. The volumetric summation of voxels within the overlapping surface divided by the volumetric summation of voxels within one surface at a source channel was used to estimate the percent of the protein represented by the source channel that is co-localized with the other protein.
Statistical analysis. All statistical analyses in this study compared a single variable between two groups of samples. Therefore, a two-tailed t-test was used to determine the statistical significance of the difference, with significance defined as P < 0.05. Welch's t-test was used if the two groups of samples had unequal variances or unequal sample sizes. The paired samples t-test was used if there were matched pairs of samples in the two groups. All statistics were performed with GraphPad Prism, using the built-in analysis.