The liver-secreted protein FGL1 restrains a subset of innate-like B cell responses via the receptor TACI

SUMMARY

The liver-secreted protein FGL1 is a ligand of the co-inhibitory receptor LAG3 on T cells; however, Fgl1^−/− mice exhibit autoimmune features distint from those of Lag3^−/− mice. Here we examined whether FGL1 acts via receptors beyond LAG3 to regulate autoimmunity. Recombinant FGL1 administration reduced autoimmune symptoms in B6/lpr lupus-like mice. This was associated with diminished B cell numbers and reduced antigen-specific IgM responses. Genome-wide surface proteome screening identified the TNFR family member TACI, a receptor for B cell activating receptors BAFF and APRIL, as a ligand for FGL1. TACI bound FGL1 via an N-terminal site not required for BAFF and APRIL interaction. The impact of FGL1 administration on B cell numbers and autoimmune phenotypes was lost in Taci^−/− mice. Mechanistically, FGL1 promotes TACI internalization, thereby regulating receptor availability for activating ligands. Thus, FGL1 regulates an innate subset of B cells via modulating TACI availability, with implications for autoimmunity and inflammation.

eTOC Blurb

Negative regulatory ligands have been underexplored for B cells. Su et al identify FGL1 as a new inhibitory ligand for a TACI-mediated pathway on a subset of B cells, exploring the biochemistry of the FGL1/TACI interaction, its function, and potential application in lupus-like autoimmune disease.

Graphical Abstract

INTRODUCTION

Active immune responses are tightly regulated by various inhibitory pathways back to baseline.¹ These regulatory mechanisms are well documented and clinically targeted in T cells for disease treatments.² However, these inhibitory pathways are less characterized in B cell-mediated immunity. Initial B cell activation upon antigen encounter via the B cell receptor (BCR) is both positively (e.g. CD19, CD21, CD81) and negatively (e.g. CD22, Siglec-G, FcγRIIb) regulated.³ These costimulatory and coinhibitory ligands can co-localize with the BCR, with mechanisms such as scaffolding functions and kinase recruitment to promote activation,^4,5 or recruiting phosphatases via ITIM domains to inhibit activation.^6,7 BCR ligation by specific antigens further triggers B cell activation, proliferation and plasma cell differentiation, as well as the induction of survival programs. These events are largely regulated by pathways within the tumor necrosis factor (TNF)/TNF receptor (TNFR) superfamily members (TNFRSF), including proteins such as B cell activating factor (BAFF, TNFSF13B) and a proliferation inducing ligand (APRIL, TNFSF13).⁸

Transmembrane activator and calcium-modulator and cyclophilin ligand (TACI, TNFRSF13B) is a TNFRSF receptor expressed on B cells.^9,10 BAFF and APRIL are stimulatory ligands for TACI, binding the receptor within its two cysteine rich domains (CRDs).^8,11,12 Regulators for the BAFF and APRIL pathways involve soluble decoy receptors.^13,14 These ligands promote B cell functions such as isotype switching and T cell-independent (TI) antibody production.^6,9 While B cells normally require T cells (T cell-dependent, TD) to generate a high-affinity antibody response, the TI antibody response is critical for immunity against antigens with repetitive structures such as bacterial polysaccharides.¹⁵ TACI may also control negative aspects of B cell function. Mice deficient in TACI (Taci−/−, Tnfrsf13b−/− mice) have an expanded B cell compartment across all mature B cell subsets and eventually develop autoimmunity;^16,17 however, this phenotype may be a result of elevated serum BAFF levels,¹⁸ which support higher numbers of B cells.

The TACI receptor-ligand pathway has clinical significance. An anti-BAFF antagonistic antibody (Benlysta) is approved by FDA for treating systemic lupus erythematosus (SLE), and anti-APRIL Sibeprenlimab is approved for primary IgA nephropathy (IgAN).^19,20 TACI-Ig (atacicept) is a decoy receptor fusion protein that can block the interaction of both BAFF and APRIL with their receptors.²¹ TACI-Ig has efficacy in mouse models of rheumatoid arthritis and SLE.^22,23 TACI-Ig is approved in 2021 for SLE in China and various clinical programs are currently ongoing in IgA nephropathy, lupus nephritis and other autoimmune diseases.^24–26 These data collectively suggest the biological and clinical importance of TACI in B cell regulation and autoimmune diseases.

FGL1 serves as a major T cell coinhibitory ligand for lymphocyte-activation gene 3 (LAG-3).^34–39 FGL1, a liver-secreted protein and member of the fibrinogen family, supports hepatocyte growth and metabolism and is upregulated in conditions of liver injury and acute inflammation.^27–33 While characterizing the Fgl1^−/− mice, we noticed that aged Fgl1^−/− mice spontaneously developed autoimmune symptoms such as increased autoantibody production.³⁴ Moreover, the aged autoimmune phenotypes of Fgl1^−/− and Lag3^−/− mice do not completely match,^{34,40,37,38,41,42} leading to the hypothesis that FGL1 may regulate B cell-mediated immunity and autoimmunity via a receptor beyond LAG-3. Through a genome-wide surface proteome screening,^9,10 we identified fibrinogen-like protein 1 (FGL1) as an inhibitory ligand for TACI to regulate B cell immunity.

RESULTS

FGL1 inhibits a subset of B cells in B6/lpr lupus-like mice

To determine whether FGL1 could directly inhibit autoimmunity and immunopathology, we infused recombinant immunoglobulin Fc-tagged FGL1 (Ig-FGL1) into B6.MRL-Fas^lpr/J (B6/lpr) mice, which are homozygous for the lymphoproliferation spontaneous mutation (Fas^lpr). These mice develop SLE-like autoimmune symptoms due to lack of lymphoid cell death.^43,44 Female B6/lpr mice around 8 weeks of age were injected intraperitoneally (i.p.) with 200 μg of Ig-FGL1 or control Ig weekly for 8 weeks. At 8 months, lymph node (LN) areas were smaller after FGL1 injection compared to control Ig (Figure 1A, S1A). FGL1-treated mice also had fewer numbers of live cells in LN at 10 months (Figure 1B). Since swollen LNs are a hallmark of B6/lpr mice,⁴⁵ this observation supports an inhibitory function of FGL1 on autoreactive lymphoid cells.

Figure 1. FGL1 treatment decreases B cell subpopulations in autoimmune mice.

A. Female B6/lpr mice were injected intraperitoneally (i.p.) weekly for 8 weeks with recombinant Ig-FGL1 (n=30) or control Ig (n=27) starting around 8 weeks of age. Areas of all enlarged lymph nodes (LN) were measured for each mouse by a caliper at 8 months and are shown as mm².

B. Mice were sacrificed and total number of live cells from LN of each mouse were counted via Vi-CELL at 10 months for Ig-FGL1 (n=24) or control Ig (n=25).

C. Live cells from LN of mice (same as B) were stained by a panel of over 30 mAbs and subjected to mass cytometry analysis. The data were analyzed by Louvain clustering algorithm.

D. Frequency of LN B cell cluster 9 shown as % of total CD45⁺ hematopoietic cells comparing control Ig versus Ig-FGL1.

E. Heatmap of normalized marker expression across 22 different Louvain clusters from mass cytometry analysis.

F. Frequency of B cells in peripheral blood at 10 months from the same mice as in B profiled by flow cytometry.

G. Frequency of T cell subpopulations from the same mice as in F.

Data are presented as mean + SEM of three pooled experiments. p value was calculated by unpaired two-tailed Student’s t-test comparing control Ig versus Ig-FGL1 (A, B, D, F). *p < 0.05, ** p < 0.01.

We performed high-dimensional, single-cell profiling of immune cells via mass cytometry (CyTOF) in the LN. The panel consisted of over 30 metal-conjugated antibodies spanning a wide array of lineage and functional markers to profile CD45⁺ cells. Data were analyzed with the Louvain algorithm, a community-based, unsupervised clustering algorithm.^46,47 A total of 22 subsets in LN were identified, visualized in two-dimensional space with uniform manifold approximation and projection (UMAP) (Figure 1C), and assigned to B, T, myeloid, or NK cell subsets depending on lineage marker expression. Clusters were numbered by frequency starting from 1, such that cluster 1 is the most abundant and cluster 22 the least abundant within all CD45⁺ cells. Interestingly, the only subset that significantly changed after FGL1 treatment was B cell cluster 9 (CD19⁺ TACI⁺ CD38⁺ Tbet⁺ MHCII⁺ IgA⁺ PDL2⁺ Ki67⁺), which decreased after FGL1 administration (Figure 1D–E, S1B). FGL1 treatment partially decreased this subset of B cells, but not to the level of wildtype (WT) healthy mice (Figure S1C). We also profiled spleens of FGL1-treated versus control Ig-treated B6/lpr mice and identified 19 clusters (Figure S2A–B). One B cell subset, splenic cluster 8 (CD19⁺ B220⁺ CD38⁺ MHCII⁺ BAFFR⁺ IgA⁺ PDL2⁺), was significantly decreased after FGL1 treatment and expressed markers statistically similar to those of lymph node cluster 9 (Figure S2B–D). In contrast, there were no statistical differences in clusters of LAG-3 expressing B cells (clusters 1 and 11 in lymph nodes, clusters 4 and 18 in spleen) and T cells (cluster 18 in lymph nodes, cluster 12 in spleen) between FGL1 and control treatment (Figure 1D–E, S2B–C). This finding indicates that FGL1 may selectively affect a subset of B cells in this mouse model through mechanisms beyond LAG-3 modulation.

Additional flow cytometry analysis of peripheral blood from B6/lpr mice revealed an overall similar pattern of B cell reduction but not other immune cells. Specifically in the blood, FGL1 treatment resulted in decreased B cell populations compared to control (Figure 1F), while other cells, such as T cells, remained unchanged (Figure 1G, S2E). Further phenotyping demonstrated that immature B cells and T cell subsets, such as CD4⁺, CD8⁺, Treg, or naïve or effector T cells, were also similar to control mice (Figure 1G, S2F–G). Activated or exhausted T cells were also unchanged (Figure S2F). There was a trend but statistically insignificant reduction in levels of autoantibodies against double stranded DNA in the serum of FGL1-treated mice (Figure S2H). In summary, our results indicate that FGL1 may predominately inhibit B but not T cells from spleen, LN and blood in a lupus model. It is also evident that FGL1 inhibits a divisional but not whole B cell response in B6/lpr mice.

FGL1 inhibits a subset of the B cell response in vivo

To dissect FGL1-mediated inhibition of B cell responses, we utilized several well-studied hapten immunization mouse models.^48–51 Haptens are small chemicals that cannot cross-link B cell receptors and do not recruit T cell help, so they are unable to stimulate antibody responses by themselves. However, haptens coupled to a large protein elicits anti-hapten antibodies, with T cell participation dependent on the type of protein carrier.⁵² Upon immunization with these hapten-conjugated proteins, either WT littermates or Fgl1−/− were tested for the responses of different antibody isotypes to the haptens. Antibody responses to 4-hydroxy-3-nitrophenylacetyl (NP) conjugated to chicken gamma globulin (CGG) were indifferent in Fgl1−/− versus WT at multiple time points (Figure 2A), including memory B cell responses (Figure S3A). Similar results were also obtained by immunization with 2,4,6-trinitrophenyl (TNP) conjugated to keyhole limpet hemocyanin (KLH) (Figure 2B, S3B). Because NP-CGG and TNP-KLH elicit T cell-dependent (TD) B cell responses,⁵² our results indicate that FGL1 may not affect TD B cell responses.

Figure 2. FGL1 inhibits a subset of antigen-specific antibody production and B cell numbers.

A. Humoral response to NP-CGG immunization in Fgl1 wild type (WT, n=5) vs Fgl1−/− (n=5) mice shown alongside pre-immunization controls. NP-CGG in alum was injected ip and serum was collected at 7 and 14 days post-injection. NP-specific antibody responses were measured by a specific sandwich ELISA.

B. Humoral response to TNP-KLH immunization in Fgl1 WT (n=6) vs Fgl1−/− (n=6) mice shown alongside pre-immunization controls. TNP-KLH in alum was injected ip and serum was collected at 7 and 14 days post-injection. TNP-specific antibody responses were measured by a specific sandwich ELISA.

C. Humoral response to NP-Ficoll immunization in Fgl1 WT (n=5) vs. Fgl1−/− (n=5) mice shown alongside pre-immunization controls. NP-Ficoll was injected ip and serum was collected at 7 days post-injection. NP-specific antibodies were measured by a specific sandwich ELISA in C, E, F.

D. ELISPOT was performed on purified B cells from spleen (left panel) and bone marrow (right panel) on day 7 from mice in C and presented as number (#) of anti-NP specific IgM antibody-secreting cells (ASC) per 10⁶ cells. Pre-immunization controls are included (n=5 for WT and Fgl1−/−). Mice were immunized as in C.

E. Effect of FGL1 overexpression on NP-specific IgM antibody production. Fgl1−/− mice (n=4) were injected iv with a FGL1-encoding plasmid (hydrodynamic injection) to overexpress FGL1 in vivo. This plasmid encodes a murine Fgl1 gene tagged to an engineered LALA-PG deactivating Fc mutation (Igmut-FGL1). Control plasmid (n=4) contains the LALA-PG mutation without FGL1 (Igmut). Fgl1−/− mice were subjected to hydrodynamic injection of the plasmids three days prior to NP-Ficoll immunization. Serum from each mouse was collected at 7 days post-NP-Ficoll injection.

F. LAG-3 blockade on the production of anti-NP-specific IgM antibody. WT female B6 mice (n=5) were injected ip with anti-LAG-3 mAb (clone C9B7W) or control Ig (n=5). One day later, mice were immunized with NP-Ficoll. Serum from each mouse was collected at 7 days post-NP-Ficoll injection.

Data are presented as the mean + SD, one-way ANOVA followed by Benjamini, Krieger, and Yekutieli’s multiple comparison (C-D) and unpaired two-tailed Student’s t-test in (E-F). Data are representative of at least two independent experiments. *p < 0.05, ** p < 0.01, *** p<0.001, **** p<0.0001, ns, not significant.

In contrast to NP-CGG and TNP-KLH, immunization with NP-Ficoll, a TI antigen, in Fgl1^−/− mice increased serum antigen-specific IgM responses in Fgl1^−/− mice compared to WT mice (Figure 2C). Other antibody isotypes, such as IgGs, were poorly elicited by this immunization compared to pre-immunization controls. Numbers of purified antigen-specific B cells increased in both the spleen and bone marrow of Fgl1^−/− mice (Figure 2D). These findings were further validated by overexpressing recombinant FGL1 in vivo by hydrodynamic injection of Ig-FGL1-encoding plasmid in Fgl1^−/− mice, which showed a decrease in antibody responses compared to control Ig (Figure 2E). The effect of Fc gamma receptor (FcγR) binding and Fc-mediated B cell depletion was excluded through incorporation of the L234A, L235A, and P329G (LALA-PG) mutations in Ig-Fc plasmid. These mutations are one of the most commonly used for disrupting FcγR binding and complement activity.^53,54 Thus, our results support that FGL1 is not a general B cell inhibitor but instead inhibits TI responses.

FGL1-mediated B cell inhibition is independent of LAG-3

LAG-3 is an inhibitory receptor for FGL1 on T cell-mediated responses to antigens.³⁴ To exclude a possible role of LAG-3 in FGL1-mediated inhibition in vivo, we treated WT B6 mice with an anti-LAG-3 mAb which can block the FGL1-LAG-3 interaction.^34,55 As shown in Figure 2F, anti-LAG-3 mAb had no effect on anti-NP IgM responses, demonstrating that FGL1-mediated B cell inhibition was independent of LAG-3. Moreover, aged Lag3^−/− mice did not develop autoimmune phenotypes such as autoantibody production,^40,41,56 suggesting the existence of an additional receptor mediating the anti-autoimmune effects of FGL1. Altogether, our findings support an inhibitory function of FGL1 on a B cell response through receptor(s) other than LAG-3.

FGL1 is a bona fide ligand for B cell receptor TACI

To determine receptor(s) of FGL1 on B cells, we utilized the human Genome-Scale Receptor Array (GSRA) technology, a high-throughput, semi-automatic protein-protein interaction discovery platform developed in our laboratory.^34,57 We screened Ig-FGL1 binding to over 6,000 human transmembrane proteins which are individually overexpressed on 293T cells. The binding of Ig-FGL1 was visualized by fluorescence using labeled anti-Fc mAb and a laser scanning macro-confocal fluorescence plate reader. Our screening revealed specific binding of FGL1 to TACI, a receptor with B cell functions and autoimmune relevance (Figure 3A). The binding of FGL1 to TACI was specific, as FGL1 did not interact with B cell maturation antigen (BCMA), the closest TACI relative (Figure 3A). The FGL1/TACI interaction could also be validated by flow cytometry using Ig-FGL1 recombinant protein on human TACI-transfected 293T cells (Figure 3B). Moreover, FGL2, a homolog of FGL1, did not exhibit binding to TACI, further confirming the interaction specificity (Figure 3A). Through domain deletion studies, we found that the C-terminal fibrinogen-like domain (FD) of FGL1, and not the N-terminal coil-coil domain (CCD), was responsible for binding TACI (Figure 3C). The FGL1/TACI interaction was also found in mouse (although weaker) and appears to be conserved between species, as mouse Ig-FGL1 cross-reacted to human TACI and vice versa (Figure S4A). This interaction could also be demonstrated using recombinant TACI-Ig to bind FGL1 fused to an artificial transmembrane domain (FGL1-TM) (Figure S4B).³⁴ BAFF-TM served as a positive control, binding to TACI-Ig and BCMA-Ig, while BCMA-Ig did not bind FGL1-TM (Figure S4B). Finally, the affinity/avidity of the FGL1/TACI interaction was measured via Octet bio-layer interferometry. Curve-fitting for TACI revealed a calculated K_D value of 40 nM under a heterogeneous ligand model (Figure 3D).

Figure 3. Identification and validation of FGL1 binding to TACI.

A. Identification of the human FGL1/TACI interaction in the GSRA system. 293T cells were engineered via stable transduction of adaptors to facilitate membrane expression (see methods). These 293T.2A cells were transfected with human genes mentioned on the x-axis and incubated with recombinant fusion proteins as indicated on the y-axis. Ig-FGL2 was included as a negative control, and Ig-BAFF and Ig-APRIL proteins served as positive controls for TACI binding.

B. Flow cytometry plots of TACI⁺ 293T transfectants stained with recombinant Ig-FGL1. Staining with Ig shown as a negative control (shaded histogram).

C. Binding of recombinant proteins with full-length FGL1, FGL1 coil-coil domain (CCD), or FGL1 fibrinogen-like domain (FD) to TACI⁺ 293T cells in the GSRA system.

D. Detection of FGL1/TACI binding. Starting with the highest response curve at 10 μg/ml, two-fold serial dilutions of recombinant FGL1 were incubated with immobilized TACI-Ig, as shown by Octet sensorgram depicting direct protein-protein interaction.

Genome screening and validation are representative of at least two independent experiments.

Mapping of the FGL1/TACI interaction

We further performed competition binding experiments to determine the topological relationship of FGL1 to TACI as well as the other two TACI ligands, BAFF and APRIL. A saturating dose (50 μg/mL) of FGL1 was pre-incubated with TACI⁺ 293T cells, followed by titration of either BAFF or APRIL to determine potential competition between the ligands. FGL1 did not affect BAFF or APRIL binding to TACI⁺ 293T cells over a wide range of concentrations (Figure 4A), hinting that FGL1 binds to a distinct site on TACI that does not overlap with other ligands.

Figure 4. FGL1 binding epitope on TACI is distinct from that of BAFF or APRIL.

A. Competition experiments of human FGL1 binding to human TACI with BAFF or APRIL. TACI⁺ 293T transfectants were incubated with a saturating dose of recombinant Ig-FGL1 or control Ig (50 μg/mL) for 30 minutes at 4°C prior to incubation with Flag-BAFF protein (40 μg/mL, 4-fold serial dilution, upper panel) or Flag-APRIL (40 μg/mL, 4-fold serial dilution, lower panel). Upon extensive washing, the cells were analyzed by flow cytometry and the results were shown as % binding. The control was isotype-matched Ig. Samples were run in duplicate. Data are representative of two independent experiments and presented as the mean + SD.

B. Characterization of the TACI binding domain. Plasmids encoding mutants with domain deletions and arginine point mutations of human TACI and human BCMA were transfected into 293T cells, incubated with Ig-FGL1 (5 μg/mL), and visualized by flow cytometry. Ig-BAFF protein (20 ng/mL) was included as a positive control. All receptors were expressed as glycolipid (GPI)-anchored proteins, and surface expression was confirmed with a mAb recognizing the GPI anchor. Data are presented as the mean + SEM, one-way ANOVA followed by Dunnett’s multiple comparison. The experiment was done once in this format, and twice more as titrations, with similar results. *p < 0.05, ** p < 0.01, *** p<0.001, **** p<0.0001, ns, not significant.

To determine the binding domain of TACI with FGL1, sequential receptor deletion constructs were generated and expressed as glycolipid (GPI)-anchored proteins in transfected 293T cells.⁵⁸ The expression of all the truncated TACI receptors was confirmed by staining with an anti-GPI anchor antibody. eGFP⁺ cells after co-transfection of eGFP plasmid were gated to determine the MFI of Ig-FGL1 and Ig-BAFF binding to the human and mouse TACI receptors.⁵⁸ The Ig-BAFF binding to the mock transfectants was utilized as a control; Ig-BAFF was found to interact with amino acid (aa) 31-70 and aa 67-118 corresponding to both CRDs of TACI (Figure 4B). BAFF binding appeared to have higher affinity than FGL1, as lower concentrations of BAFF reached a similar or higher MFI (Figure 4B).

The human TACI binding site for FGL1 was located largely in its N-terminal aa 1-30 region (Figure 4B). This region was not required for BAFF binding, which was consistent with the lack of competition observed in Figure 4A. This N-terminal region currently has no function ascribed to it. As shown, the first 30 amino acids are necessary and sufficient for human (h) FGL1 binding (Figure 4B). Of interest, region aa 1-30 contained a cluster of positively charged arginine residues (the first five at positions 6, 8, 9, 12, and 14), suggesting that electrostatic interactions may be important for binding. This possibility was tested by site-directed mutagenesis to replace these residues (RRR6/8/9, RR12/14) with neutral alanine (AAA6/8/9, AA12/14) or opposing negatively charged glutamate (EEE6/8/9, EE12/14) amino acids. FGL1 but not BAFF binding to TACI was lost upon these mutations, with contributions from RR12/14 and RRR6/8/9. However, shorter constructs containing all or some of these residues exhibited weaker FGL1 binding (aa 1-16) or no binding at all (aa 1-10, 1-13) (Figure 4B). Thus, FGL1 binding to TACI may be dependent on basic residues in the context of its natural N-terminal sequence. Of note, BCMA, which lacks N-terminal basic residues, did not bind FGL1 (Figure 4B), supporting possible contribution of these basic residues to FGL1 binding. Some N-terminal arginines were also present at the FGL1/TACI binding interface by Alphafold2 multimer modeling (Figure S5A). In contrast, other arginines more distal from the N-terminal region (ex. R20, 32) were not at the binding interface. While mouse TACI does not have a homologous N-terminal sequence as human TACI, it does nonetheless contain charged amino acid residues (KRK8/16/17) within its own N-terminal region.⁵⁹ Mutation of RK16/17 to LG (i.e to the corresponding amino acids in the human TACI sequence) did not cause a decrease in human FGL1 binding (Figure S5B). As in the human system, mouse BCMA also lacks these basic N-terminal residues – it did not bind hFGL1 but did interact with BAFF (Figure S5B).

The DxL motif within the conserved cysteine rich domains (CRD) are crucial for BAFF binding.⁶⁰ Human TACI has two such domains (CRD1 and CRD2). To support our receptor deletion data, we mutated the DxL motif to AxA in the CRD. DxL mutation for both CRDs in human TACI completely ablated BAFF binding, as expected (Figure S5C). In accordance with published literature, CRD2 DxL was more important than the CRD1 DxL for BAFF binding.^60,12 The two CRDs partially compensated for one another, as shown by progressive loss in binding from single to double DxL mutants (Figure S5C). Unlike BAFF, FGL1 binding was mostly retained even in the double DxL mutant (Figure S5C), suggesting a different mode of binding than BAFF. FGL1 binding was unaffected in the CRD1 DxL mutant but slightly decreased in CRD2 DxL mutant compared to WT TACI. There was no difference between CRD2 and double DxL mutants for FGL1 binding, suggesting that the small loss of FGL1 binding in the double mutant was predominately due to the CRD2 DxL mutation. In summary, our data elucidates the binding interaction between FGL1/TACI and confirms the distinct interactions of TACI with FGL1 versus BAFF.

TACI is the major receptor for FGL1-mediated B cell inhibition in vivo

We sought to determine whether TACI is responsible for FGL1’s inhibitory effect on B cells in vivo. Taci−/− mice were immunized with NP-Ficoll. IgM antibody responses to NP, as well as NP-specific ASCs in lymph nodes and spleens were measured at day 7 post challenge. Antibody responses to NP were inhibited significantly in Taci−/− mice compared with Taci+/+ mice (Figure S6A–B), similarly to FGL1 overexpression in Fgl1−/− mice (Figure 2E). Importantly, we observed enhanced NP-specific IgM and numbers of splenic antigen-specific IgM ASCs in Fgl1−/− mice upon NP-Ficoll immunization, but this difference was lost in the absence of TACI (Figure 5A–B). Day 7 absorbance values were above background, or levels of preimmunization controls. Finally, we injected recombinant Ig-FGL1 protein into Fgl1−/− mice post NP-Ficoll inoculation. NP-specific IgM levels were suppressed by Ig-FGL1 compared to control Ig in Taci+/+ Fgl1−/− mice but not in double Taci−/− Fgl1−/− mice (Figure 5C), showing that FGL1 protein-mediated B cell inhibition required TACI. FGL1 treatment also reduced the number of antigen-specific IgM ASCs in TACI-sufficient but not TACI-deficient mice (Figure 5D). Together, these experiments support TACI as the major receptor mediating FGL1’s inhibitory function on B cell responses.

Figure 5. FGL1 inhibits antigen-specific B cell numbers and antibody production via TACI.

A. Taci−/− Fgl1−/− mice ( n=5) and littermate Taci−/− mice with wild type Fgl1 (Taci−/− Fgl1+/+, n=5) were immunized with NP-Ficoll. NP-specific IgM antibody responses in sera were determined by a specific sandwich ELISA. WT (n=5) and Fgl1−/− (n=5) mice along with pre-immunization serum shown as controls.

B. The number of NP-specific IgM antibody secreting cells (ASC) were also evaluated by ELISPOT with purified B cells from spleen of mice in A alongside pre-immunization controls.

C. Ig-FGL1 vs. control Ig was injected one day prior to NP-Ficoll administration in Fgl1−/− mice and Taci−/− Fgl1−/− mice (n=5/group). These proteins were injected again at Day 2 and Day 4 post administration. Sandwich ELISA performed as in A.

D. ELISPOT for Ig-FGL1 vs. control Ig in Fgl1−/− mice and Taci−/− Fgl1−/− mice was performed as in B.

Data are presented as the mean + SD, one-way ANOVA followed by Benjamini, Krieger, and Yekutieli’s multiple comparison (A-D). Data are representative of at least two independent experiments. *p < 0.05, ** p < 0.01, *** p<0.001, **** p<0.0001, ns, not significant.

We next uncovered the role of the FGL1/TACI interaction in autoimmunity, the initial motivation for our study. Recombinant Ig-FGL1 or control Ig was given in the aforementioned manner to female littermate-controlled Taci−/− B6/lpr mice. There was no effect on LN sizes or the number of LN live cells after FGL1 administration in these mice (Figure 6A–B). We also performed CyTOF on LN and uncovered 23 clusters by the Louvain algorithm (Figure 6C). In the absence of TACI, no LN B cell clusters were affected by FGL1 treatment. Instead, we observed a decrease in cluster 8 (CD8⁺ Tbet⁺ Ly6C⁺), a T cell subset, in LN after FGL1 injection (Figure 6D–E). We confirmed these mice were indeed Taci−/− by the lack of TACI expression in the heatmap (Figure 6E). Profiling of blood by flow cytometry showed no difference in general immune populations in the blood (Figure 6F, S7A), or classic T and B cell populations, including Tregs, effector/naïve/memory T cells, or transitional B cells (Figure 6G, Figure S7B). Activated or exhausted T cell levels were also unchanged (Figure S7C). These findings, compared to Figure 1 and Figure S2, indicate that the effect of FGL1 on B cells is largely dependent on the presence of TACI.

Figure 6. FGL1 treatment does not decrease B cell subpopulations in Taci−/− autoimmune mice.

A. Female Taci−/− B6/lpr mice were injected with recombinant Ig-FGL1 (n=25) or control Ig (n=25) in the same manner as in Figure 1A. Areas of all enlarged lymph nodes (LN) were measured for each mouse by a caliper at 8 months and are shown as mm².

B. Mice were sacrificed and total number of live cells from LN of each mouse were counted via Vi-CELL at 10 months for Ig-FGL1 (n=20) or control Ig (n=19).

C. Live cells from LN of mice (same as B) were stained by a panel of over 30 mAbs and subjected to mass cytometry analysis. The data were analyzed by Louvain clustering algorithm.

D. Frequency of LN T cell cluster 8 shown as % of total CD45⁺ hematopoietic cells comparing control Ig versus Ig-FGL1 in Taci−/− B6/lpr mice.

E. Heatmap of normalized marker expression across 23 different Louvain clusters from mass cytometry analysis.

F. Frequency of B cells in peripheral blood at 10 months from mice in B profiled by flow cytometry.

G. Frequency of T cell subpopulations from the same mice as in F.

Data are presented as mean + SEM of three pooled experiments. p value was calculated by unpaired two-tailed Student’s t-test comparing control Ig vs. Ig-FGL1 (A, B, E). *p < 0.05, ns not significant.

To see if any of the FGL1-affected LN subsets significantly resembled one another between Taci+/+ and Taci−/− autoimmune mice, we generated correlation matrices and tested for similarities with Spearman’s correlation test (Figure S7D–F). Taci+/+ B cell cluster 9 (CD19⁺ TACI⁺ CD38⁺ Tbet⁺MHCII⁺IgA⁺PDL2⁺ Ki67⁺), which was decreased after FGL1 administration, was significantly similar to Taci−/− B cell cluster 5 (B220⁺ CD38⁺ Tbet⁺ MHCII⁺ CD80⁺ PDL2⁺ CD86⁺), which was not affected by FGL1 treatment (Figure S7D). Taci−/− T cell cluster 8 (CD8⁺ Tbet⁺ Ly6C⁺) was significantly similar to Taci+/+ T cell cluster 10 (CD8⁺ Tbet⁺ Ly6C⁺) (Figure S7E), though T cell frequencies were not altered by FGL1 in WT B6/lpr mice. These data reveal a dominant effect of FGL1 specifically regulating a TACI⁺ B cell subset.

FGL1 triggers TACI internalization from the cell surface

While our results support an inhibitory function of FGL1 on B cell responses via TACI, engagement of TACI by its other ligands, APRIL and BAFF, has been shown to positively promote TI antibody production.⁹ One possible interpretation is that FGL1 binding to TACI leads to its internalization, preventing the access of other ligands. This form of regulation has been well documented for a number of other receptors, such as CTLA-4 or BCR.^61,62 To test this, we incubated TACI-transduced 293T cells with Ig-FGL1 or control Ig. TACI was then co-stained with different cell markers (ex. plasma membrane, endosome, or lysosome) and subsequently imaged with confocal microscopy to trace and quantitate TACI colocalization. We found that FGL1 treatment decreased levels of TACI on the plasma membrane compared with control Ig (Figure 7A). Furthermore, FGL1 incubation increased TACI colocalization with EEA1, an endosomal marker (Figure 7B), but not with LAMP1, a lysosomal marker (Figure 7C), suggesting that FGL1 leads to TACI endocytosis but not degradation. TACI also co-localized with internalized FGL1, hinting at a FGL1-dependent process (Figure S8A). These findings indicate that FGL1 could internalize TACI, leading to removal of TACI from the cell surface as a potential inhibitory mechanism.

Figure 7. Induction of TACI internalization by FGL1 engagement.

A-C. Representative images of confocal microscope (upper panels) and quantifications (lower panels) of TACI colocalization with a plasma membrane marker (A), endosomal marker (B), or lysosomal marker (C) in TACI⁺ 293T cells. These cells were incubated with 1 μg/ml Ig-FGL1 or control Ig (Ig). At least 50 cells were quantified in each group and each cell was shown as a dot. Scale bars, 5 μm.

D. Schematic of TACI internalization assay on purified B cells. FGL1 was incubated with anti-TACI fluorescently labeled antibodies together with purified mouse splenic B cells. Anti-Fc receptor antibody was also included to block non-specific Fc/FcR interaction. Cells were incubated at 37°C to facilitate internalization, and the incubation of cells at 4°C was used as control. After extensive washing, glycine-HCl buffer at pH 3.5 was added for 1 min at room temperature, which disrupts extracellular interactions. Cells were then washed and stained with live dead fixable dye before flow cytometry analysis.

E. As described in D, indicated concentrations of FGL1 were added to the culture for 1 hour at 37°C or 4°C. The culture wells without TACI antibodies were used as negative controls. The results were presented as % TACI⁺ of live B cells. Samples were run in duplicate.

Data are presented as the mean + SD, unpaired two-tailed Student’s t-test (A-C), one-way ANOVA followed by Benjamini, Krieger, and Yekutieli’s multiple comparison (E). Data are representative of at least two independent experiments. *p < 0.05, ** p < 0.01, *** p<0.001, **** p<0.0001, ns, not significant.

We further analyzed TACI internalization on primary mouse B cells using a flow cytometry-based assay (Figure 7D). In this procedure, a transient cell surface wash with glycine-hydrochloric acid buffer before flow cytometry analysis removed surface staining and chemically optimized intracellular detection.⁶³ A dose-dependent, increased TACI internalization was observed after a 1 hour incubation of FGL1 with an anti-TACI antibody at 37° C on purified mouse B cells (Figure 7E, S8B). Lower temperatures are also known to inhibit or decelerate receptor internalization.⁶⁴ After 4°C incubation, we observed a dramatic reduction of intracellular TACI staining, with a small amount of residual FGL1-mediated signal (Figure 7E), suggesting a biologically relevant internalization process. We compared B220⁺ and B220⁻ cells with the whole live splenocyte population and found that internalization occurred predominately in B cells, with a small amount of internalization in B220⁻ subpopulations (Figure S8C). We confirmed that the assay was specific by performing it on Taci−/− splenocytes, which did not show internalization at the maximal dose of FGL1 in sharp contrast to Taci+/+ (Figure S8D). The effect was independent of LAG-3, as Lag3−/− B cells internalized TACI identically to WT (Figure S8D). Altogether, our data suggest FGL1-mediated TACI internalization as a possible mechanism for its inhibitory function of B cells.

DISCUSSION

We showed herein that recombinant FGL1 administration in symptomatic B6/lpr lupus-like mice reduced spontaneous lymphadenopathy and lymph node live cell numbers. FGL1 specifically decreased a subset of TACI⁺ B cells in pathologic lymph nodes whereas T cell-mediated autoimmunity was less affected. We identified FGL1 as an inhibitory ligand of TACI. Interestingly, FGL1 bound a discrete, N-terminal region of TACI, a binding site independent of its two stimulatory ligands, BAFF and APRIL. FGL1 was not a general inhibitor of antibody responses, but instead specifically reduced a TACI-mediated subset of B cell responses. Upon binding to FGL1, TACI internalization was promoted. Our findings uncover a negative regulatory pathway on B cells with implications for autoimmunity and other diseases with TACI involvement.

FGL1 is an inhibitory ligand for LAG-3 on T cells in tumor models.^34–39,65 LAG-3 can also be induced on B cells.^66,67 However, immunization with NP-Ficoll and other haptens induce LAG-3⁻, not LAG-3⁺, antigen-specific ASCs, suggesting that LAG-3 is not relevant in TI responses.⁶⁸ Thus, the limited expression of LAG-3 in these in vivo models may not contribute to the preferential B cell effect. In our WT B6/lpr lupus model, FGL1 decreased a B cell subset, but not any subpopulations of T cells or LAG-3⁺ subsets. However, our data cannot exclude the contribution of FGL1/LAG-3 in other models of autoimmune diseases, particularly those with more dominant T cell responses.

TACI engagement mainly affects TI antibody responses.^16,69 Our current findings also implicate FGL1 in suppressing TI antibody responses. FGL1 may preferentially affect more innate-like B cells, such as extrafollicular or marginal zone splenic B cells. We observed a reduction of one B cell subset in LNs and spleens of autoimmune mice through mass cytometry and a more general B cell reduction in blood via flow cytometry. We suspect that different FGL1-responsive subsets in blood may be uncovered via a higher-dimensional technique than flow cytometry. The TI antibody response may partially contribute to pathogenesis in the B6/lpr autoimmune model, as recombinant FGL1 inhibited LN sizes and cell number in the B6/lpr mouse model. Though there was an inhibitory trend, FGL1 did not significantly affect autoantibody production, implicating that autoantibodies may be predominately TD and not TI responses in this model.

We uncovered that the modality of FGL1 binding to the N-terminal region of human TACI is different and independent from BAFF and APRIL binding. Unlike BAFF, FGL1 still interacted with TACI mutated in the DxL motif. While the N-terminal arginine residues are necessary for FGL1 binding, they must be presented in a larger context than just the basic sequence. The N-terminal region of some TNFR family proteins (e.g. TNFR, Fas) contain a proximal ligand assembly domain (PLAD), which facilitates receptor self-association before ligand binding.⁷⁰ While a PLAD domain is not yet defined for TACI, FGL1 may affect the pre-ligand state of TACI or TACI self-interactions. FGL1/TACI could also elicit better function in human given the stronger interaction compared to mouse. After binding, our data points to receptor internalization as one potential mechanism for FGL1-mediated TACI inhibitory function, though other mechanisms may coexist.

Our results reveal a predominant role of the FGL1/TACI interaction on B cells in autoimmune mice. No B cell subsets were altered after FGL1 treatment in Taci−/− autoimmune mice, which may be due to lack of TACI expression. While T cells were largely unaffected by FGL1 treatment in WT B6/lpr mice, one subset of T cells was decreased after FGL1 treatment in Taci−/− B6/lpr mice. Thus, we conclude that the primary effect of FGL1/TACI is on B cells, and in the absence of TACI, secondary effects on other cell types through FGL1 can be observed. The TACI⁺ B cell subset affected by FGL1 (cluster 9, CD19⁺ TACI⁺ CD38⁺ Tbet⁺ MHCII⁺ IgA⁺ PDL2⁺ Ki67⁺) has a somewhat similar phenotype (other than a lack of CD11b/CD11c) as atypical B cells (ABCs), a B cell lineage associated with immunization, infection, and autoimmunity.⁷¹ In summary, our study highlights FGL1 as a unique pathway controlling a subset of B cell responses via TACI, revealing the discovery, function, mechanism, and therapeutic implications of this interaction.

Limitations of the Study

Our study identifies a functional interaction between FGL1 and TACI on B cells with implications for autoimmune disease. TACI can be expressed on other cell types, such as T cells, upon activation. It is possible that FGL1/TACI on T cells and FGL1/LAG3 on B cells may have some roles in other models. On B cells, it is premature to conclude that the effect of FGL1/TACI is entirely on TI antibody responses, or that the FGL1/TACI interaction would inhibit all TI antibody responses. Although B6/lpr mice have uncontrolled lymphoid cell proliferation due to lack of Fas-mediated death, the relative contribution of T cells vs. B cells in the pathogenesis is still debatable. Finally, the role of FGL1/TACI in other autoimmune diseases, as well as in infection, requires further investigation.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Lieping Chen (lieping.chen@yale.edu).

Materials availability

All materials generated or analyzed during this study are included in this article and its supplementary information files. Please contact the lead contact for unique material requests. Any material that can be shared will be released via a material transfer agreement for non-commercial usage.

STAR METHODS

Experimental Model and Study Participant Details

Mice

Mice of both sexes ranging from 6-12 weeks of age were housed in specific pathogen-free conditions in a barrier facility and utilized for experiments. Same-sex littermates were randomly assigned across experimental conditions. Fgl1−/− mice were generated as previously described.³⁴ Lag3−/− mice were gifted from the Vignali lab.⁷² mice were gifted from the Bram lab.⁶⁹ Fgl1−/− Taci−/− mice were generated by breeding the aforementioned mice together. F₁ mice were crossed with Taci−/− mice and then intercrossed to generate the desired strain along with the corresponding Fgl1 WT Taci−/− littermate control. Female B6.MRL-Fas^lpr/J (B6/lpr) (#000482) were purchased from the Jackson Laboratory (JAX). Taci−/− B6/lpr mice were generated by breeding Taci−/− mice with B6/lpr mice. The Yale University Institutional Animal Care & Use Committee approved all mouse procedures, which are in accordance with NIH guidelines.

Method Details

Plasmids, recombinant proteins, and antibodies

Human cDNA library constructs and FGL1-coding plasmids were produced or purchased as previously described.³⁴ Receptor domain deletions were expressed as glycolipid (GPI)-anchored proteins and cloned as previously described.⁷³ Genes containing DxL mutations of TACI were synthesized using GeneArt (Thermo Fisher). Recombinant Ig-FGL1 or Flag-FGL1 was generated in house. Ig-tagged proteins were purified from transfected 293T cells using Protein A affinity column (Cytiva), while Flag-tagged proteins were purified using anti-Flag M2 affinity gel (Sigma).³⁴ Antibodies used in flow cytometry are shown in the key resource table. Anti-LAG3 (clone C9B7W) was gifted from the Vignali lab.⁷⁴

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Goat anti-human IgG (polyclonal) Alexa Fluor 647	Thermo Fisher	Cat# A-21445; RRID: AB_2535862
Goat anti-human IgG-PE (polyclonal)	Southern Biotech	Cat# 2040-09; RRID: AB_2795648
Goat anti-rat Ig, Mouse ads-PE (polyclonal)	Southern Biotech	Cat#: 3010-09; RRID: AB_2795805
Mouse anti-human IgG Fc PE (clone HP6001)	Southern Biotech	Cat# 9054-09; RRID: AB_2796628
Rat anti-DYKDDDDK Tag BV421 (clone L5)	Biolegend	Cat# 637321; RRID: AB_2750051
Rat anti-DYKDDDDK Tag (clone L5)	Biolegend	Cat# 637321; RRID: AB_2750061
anti-TRAILR3 (clone LEIA)	In house	In house
SBA Clonotyping System-HRP	Southern Biotech	Cat# 5300-05
SBA Clonotyping System-C57BL/6-AP	Southern Biotech	Cat# 5300-04B
TruStain FcX Rat anti-mouse CD16/CD32 (clone 93)	Biolegend	Cat# 101319; RRID: AB_1574973
Rat anti-mouse CD16/CD32 (clone 2.4G2)	BD	Cat# 553142; RRID: AB_394656
Anti-mouse CD3 PE/Cy7 (clone 17A2)	Biolegend	Cat# 100220; RRID: AB_1732057
Rat anti-mouse TACI BV421 (clone 8F10)	BD	Cat# 742840; RRID: AB_2741091
Rat anti-mouse IgM Alexa Fluor 647 (clone RMM-1)	Biolegend	Cat# 406525; RRID: AB_2563478
Rat anti-mouse CD45 Alexa Fluor 700 (clone 30-F11)	Biolegend	Cat# 103127; RRID: AB_493714
Rat anti-mouse IgD Alexa Fluor 488 (clone 11-26c.2a)	Biolegend	Cat# 405717; RRID: AB_10730618
Rat anti-mouse CD19 BB700 (clone 1D3)	BD	Cat# 566412; RRID: AB_2744315
Rat anti-mouse CD62L BV711 (clone MEL-14)	BD	Cat# 568286
Rat anti-mouse Foxp3 Alexa Fluor 647 (clone MF-14)	Biolegend	Cat# 126407; RRID: AB_1089115
Rat anti-mouse CD8 BB700 (clone 53-6.7)	BD	Cat# 566409; RRID: AB_2744467
Rat anti-mouse Ly-6G BV711 (clone 1A8)	BD	Cat# 563979; RRID: AB_2738520
Rat anti-mouse CD45R/B220 BV480 (clone RA3-6B2)	BD	Cat# 565631; RRID: AB_2739311
Hamster anti-mouse CD11c PE (clone N418)	Biolegend	Cat# 117307; RRID: AB_313776
Rat anti-mouse F4/80 PE-CF594 (clone T45-2342)	BD	Cat# 565613; RRID: AB_2734770
Rat anti-mouse Ly-6C Alexa Fluor 488 (clone HK1.4)	Biolegend	Cat# 128021; RRID: AB_10640820
Rat anti-mouse CD11b BB700 (clone M1/70)	BD	Cat# 566417; RRID: AB_2744272
Hamster anti-mouse PD1 BV480 (clone J43)	BD	Cat# 746784; RRID: AB_2744041
Rat anti-mouse CD4 Alexa Fluor 700 (clone RM4-5)	Biolegend	Cat# 100536; RRID: AB_493701
Hamster anti-mouse CD69 PE/Dazzle 594 (clone H1.2F3)	Biolegend	Cat# 104535; RRID: AB_2565582
Rat anti-mouse/human CD44 Alexa Fluor 488 (clone IM7)	Biolegend	Cat# 103015; RRID: AB_493678
Mouse anti-mouse NK1.1 Alexa Fluor 647 (clone PK136)	Biolegend	Cat# 108719; RRID: AB_2132713
Rat anti-mouse CD93 BV711 (clone AA4.1)	BD	Cat# 740750; RRID: AB_2740418
Rat anti-mouse CD23 BV480 (clone B3B4)	BD	Cat# 746329; RRID: AB_2743652
Rat anti-mouse/human CD45R/B220 Alexa Fluor 700 (clone RA3-6B2)	Biolegend	Cat# 103231; RRID: AB_493716
Rat anti-mouse LAG3 PE/Cyanine7 (clone C9B7W)	Biolegend	Cat# 125225; RRID: AB_2715763
Rat anti-human TACI APC (clone 1A1)	Biolegend	Cat# 311911; RRID: AB_2565423
Mouse anti-human CD147 (clone 306202)	Biolegend	Cat# 306202; RRID: AB_314586
Anti-mouse LAMP1 (clone REA777)	Miltenyi Biotec	Cat# 130-111-506
Mouse anti-EEA1 (clone MA5-31575)	Thermo Fisher	Cat# MA5-31575; RRID: AB_2787202
Goat anti-mouse IgG Alexa Fluor 594 (polyclonal)	Biolegend	Cat# 405326; RRID: AB_2563308
Rat anti-mouse CD45 89Y (clone 30-F11)	Standard Biotools Inc	Cat# 3089005B
Hamster anti-mouse CD3e 152Sm (clone 145-2C11)	Standard Biotools Inc	Cat# 3152004B
Rat anti-Foxp3 158Gd (clone FJK-16s)	Standard Biotools Inc	Cat# 91H032158
Rat anti-mouse PD1 159Tb (clone 29F.1A12)	Standard Biotools Inc	Cat# 3159024B
Mouse anti-human/mouse Tbet 160Gd (clone 4B10)	Standard Biotools Inc	Cat# 3160010B
Mouse anti-human/mouse Ki-67 161Dy (clone B56)	Standard Biotools Inc	Cat# 3161007B
Rat anti-mouse CD62L 164Dy (clone MEL-14)	Standard Biotools Inc	Cat# 3164003B
Mouse anti-mouse NK1.1 170Er (clone PK136)	Standard Biotools Inc	Cat# 3170002B
Rat anti-mouse CD11b 172Yb (clone M1/70)	Standard Biotools Inc	Cat# 3172012B
Rat anti-mouse I-A/I-E (clone M5/114.15.2)	Biolegend	Cat# 107602; RRID: AB_313316
Rat anti-mouse/human CD44 (clone IM7)	Biolegend	Cat# 103002; RRID: AB_312952
Rat anti-mouse Ly-6G (clone 1A8)	Biolegend	Cat# 127637; RRID: AB_2563784
Rat anti-mouse CD38 (clone 90)	Biolegend	Cat# 102702; RRID: AB_312923
Rat anti-mouse CD21/35 (clone 7G6)	BD	Cat# 553817; RRID: AB_395069
Rat anti-mouse CD268 (clone 7H22-E16)	Biolegend	Cat# 134102; RRID: AB_1626123
Hamster anti-mouse CD80 (clone 16-10A1)	BD	Cat# 104702; RRID: AB_313123
Hamster anti-mouse CD11c (clone N418)	Biolegend	Cat# 117302; RRID: AB_313770
Rat anti-mouse CD8 (clone 53-6.7)	Biolegend	Cat# 100702; RRID: AB_312740
Rat anti-mouse CD19 (clone 6D5)	Biolegend	Cat# 115502; RRID: AB_313636
Rat anti-mouse CD4 (clone RM4-5)	Biolegend	Cat# 100506; RRID: AB_312708
Rat anti-mouse CD23 (clone B3B4)	Biolegend	Cat# 101602; RRID: AB_312827
Rat anti-mouse Ly-6C (clone HK1.4)	Biolegend	Cat# 128002; RRID: AB_1134213
Rat anti-mouse CXCR5 (clone L138D7)	Biolegend	Cat# 145502; RRID: AB_2561955
Rat anti-mouse/human GL7 (clone GL7)	Biolegend	Cat# 144602; RRID: AB_2561547
Rat anti-mouse IgM (clone RMM-1)	Biolegend	Cat# 406502; RRID: AB_315051
Rat anti-mouse PD-L2 (clone TY25)	Biolegend	Cat# 107202; RRID: AB_345250
Rat anti-mouse IgA (clone RMA-1)	Biolegend	Cat# 407002; RRID: AB_315076
Rat anti-mouse CD93 (clone AA4.1)	Biolegend	Cat# 94229
Rat anti-mouse CD86 (clone GL-1)	Biolegend	Cat# 105002; RRID: AB_313144
Rat anti-mouse TACI (clone 8F10)	Biolegend	Cat# 93238
Rat anti-mouse CD138 (clone 281-2)	Biolegend	Cat# 142502; RRID: AB_10959655
Rat anti-mouse LAG-3 (clone C9B7W)	Biolegend	Cat# 125202; RRID: AB_961187
Hamster anti-mouse CD69 (clone H1.2F3)	Biolegend	Cat# 104502; RRID: AB_313105
Rat anti-mouse F4/80 (clone BM8)	Biolegend	Cat# 123102; RRID: AB_893504
Rat anti-mouse IgD (clone 11-26c.2a)	Biolegend	Cat# 405702; RRID: AB_315024
Rat anti-mouse/human CD45R/B220 (clone RA3-6B2)	Biolegend	Cat# 103202; RRID: AB_312986
Bacterial and virus strains
GeneCopoeia GCI-L3 chemically competent cells	GeneCopoeia	Cat# CC001
Chemicals, peptides, and recombinant proteins
Cell-ID Intercalator-Ir—125 μM	Standard BioTools	Cat# 201192A
Cell-ID Cisplatin	Standard BioTools	Cat# 201064
Lipofectamine 3000	Themo Fisher	Cat# L3000008
DRAQ7	Biolegend	Cat# 424001
Live/Dead Fixable Near-IR Dead Cell Stain Kit	Thermo Fisher	Cat# L34975
DAPI	Sigma	Cat# D9542-1MG
NP-AECM-Ficoll	Biosearch Technologies	Cat# F-1420-10
NP-CGG	Biosearch Technologies	Cat# N-5055C-5
NP-BSA	Biosearch Technologies	Cat# N-5050H-10
TNP-KLH	Biosearch Technologies	Cat# T-5060-5
TNP-BSA	Biosearch Technologies	Cat# T-5050-10
Imject Alum Adjuvant	Thermo Fisher	Cat# 77161
Flag-BAFF	ACROBiosystems	Cat# BAF-H52D4-100ug
Flag-APRIL	In house	In house
Flag-FGL1	Wang et al., 2019	N/A
Ig-FGL1	Wang et al., 2019	N/A
TACI-Ig	R&D Systems	Cat# 174-TC-050
Critical commercial assays
1-Step Ultra TMB-ELISA Substrate Solution	Thermo Fisher	Cat# 34029
AP Conjugate Substrate Kit	Bio-Rad	Cat# 1706432
Octet Protein A Biosensor	Sartorius	Cat# 18-5010
Octet Kinetics Buffer 10X	Sartorius	Cat# 18-1105
Ebioscience Foxp3/Transcription Factor Staining Buffer Set	Thermo Fisher	Cat# 00-5523-00
Attune Performance Tracking Beads	Thermo Fisher	Cat# 4449754
ACK Lysing Buffer	Thermo Fisher	Cat# #A10492
EasySep Mouse B Cell Isolation Kit	Stemcell Technologies	Cat# 19854
HiTrap Protein A HP	Cytiva	Cat# 17040203
Anti-Flag M2 Affinity Gel	Sigma	Cat# A2220-1ML
EQ Four Element Calibration Beads	Standard BioTools	Cat# 201078
Maxpar Cell Acquisition Solution	Standard BioTools	Cat# 201240
Maxpar Cell Staining Buffer	Standard BioTools	Cat# 201068
Maxpar Fix and Perm Buffer	Standard BioTools	Cat# 201067
Maxpar X8 Multimetal Labeling Kit	Standard BioTools	Cat# 201300
Maxpar MCP9 Antibody Labeling Kit, 106Cd	Standard BioTools	Cat# 201106A
Maxpar MCP9 Antibody Labeling Kit, 112Cd	Standard BioTools	Cat# 201112A
Experimental models: Cell lines
293T.2A	Wang et al., 2019	N/A
Anti-LAG3 (C9B7W) Hybridoma	Workman et al., 2002	N/A
Anti-TRAILR3 (LEIA) Hybridoma	Bossen et al., 2006	N/A
Experimental models: Organisms/strains
Mouse: B6.MRL-Faslpr/J	The Jackson Laboratory	JAX: 000482
Mouse: FGL1-KO	Wang et al., 2019	N/A
Mouse: LAG3-KO	Workman et al., 2003	N/A
Mouse: TACI-KO	Von Bulow et al., 2001	N/A
Mouse: FGL1-KO TACI-KO	This paper	N/A
Mouse: TACI-KO B6.MRL-Faslpr/J	This paper	N/A
Recombinant DNA
cDNA library	Wang et al., 2019	N/A
FGL1 plasmids	Wang et al., 2019	N/A
Software and algorithms
CyTOF Normalization	N/A	https://github.com/nolanlab/bead-normalization
OMIQ	Dotmatics	https://www.omiq.ai/
Alphafold2	Jumper et al., 2021	N/A
Colabfold	Google Colaboratory	https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb
PyMOL	Pymol Schrödinger, LLC	https://pymol.org/
Octet analysis software	PALL	https://www.fortebio.com/octet-software.html
FlowJo X 10.7.1	Flowjo, LLC	https://www.flowjo.com/
Prism 10.0	Graphpad	https://www.graphpad.com/
Python 3.8.6	Python Software Foundation	https://www.python.org
Scanpy 1.9.2	Wolf et al., 2018	https://scanpy.readthedocs.io/en/stable/
ImageJ	ImageJ	https://imagej.net/ij/

Genome-scale receptor array screening system

The genome-scale receptor array screening system was developed previously.³⁴ In brief, 200 μL of cDNA library plasmids (4 μg/mL) were diluted in OptiMEM and added to 384 well VWR plates. A robotic liquid handling system (PlateMate, ThermoFisher) transferred 2 μL of the diluted library to 1536 well Greiner plates for storage at −80°C. Using the Multidrop Combi (Thermo Fisher), 2 μL of diluted Lipofectamine 3000 (Thermo Fisher) was dispensed to the 1536 well plate. Engineered 293T-2A cells expressing DAP10, DAP12, FcRγ, and CD3Z was then added at 2000 cells/well in 4 μL DMEM (10% FBS). Transient transfection occurred for 18 hours at 37°C, after which time 2 ng of Ig-FGL1, along with 3 ng of anti-Fc secondary antibody, was dispensed and incubated for 8 hours. Applied Biosystems 8200 cellular detection system (CDS) read the plates; CDS 8200 software was utilized for analysis. Internal positive controls were human Fc receptors. Similar experiments were performed to measure the interaction of BAFF, APRIL and FGL1 to TACI or BCMA.

Binding and competition experiments

Octet Red (Sartorius) was utilized to measure kinetics of direct protein-protein interactions. Protein A biosensors (Sartorius) were loaded with 10 μg/mL of TACI-Ig (R&D Systems) in 200 μL 1x Kinetics buffer (Sartorius). Sensors were briefly equilibrated in 1X Kinetics buffer before loading with a two-fold serial dilution starting at 10 μg/mL of Flag-FGL1. Protein association, disassociation, and K_d were analyzed using the Octet software.

Cells were washed, counted, and stained in PBS with Near IR fixable live/dead dye (Invitrogen) for 15 min at room temperature before proceeding to staining for flow cytometry (Attune NxT, Themo Fisher). The flow cytometer was calibrated with Attune performance tracking beads (Thermo Fisher) prior to every run. For binding experiments, Ig-FGL1 (1 μg/mL) was added to TACI-transfected 293T cells for 30 min at 4°C. After washing in FACS buffer (2% FBS in PBS), cells were stained with Alexa Fluor 647 anti-human IgG (1 μg/mL, Thermo Fisher) for 30 min at 4°C, followed by washing and analysis via flow cytometry. For binding to receptor mutants, plasmids coding for GPI-anchored domain deleted constructs (2.5 μg) were co-transfected using polyethyleneimide (PEI) in 293T cells (in 4 mL DMEM 10% FCS, in 6 cm diameter plates) along with 0.2 μg of an eGFP-coding plasmid, according to ⁵⁸. Cells were stained in triplicate for 20 min on ice in 50 μl of PBS, 5% FBS, 2.5 mM EDTA, 1 IU/ml of Liquemin (Roche Pharma) with either i) 5 μg/ml of Ig-hFGL1 (amino acids 23-312 of hFGL1), ii) 20 ng/ml of Ig-hBAFF (amino acid residues 136-285 of hBAFF), iii) control with no ligand, iv) 5 μg/ml of rat IgG2a anti-TRAILR3 antibody, followed by appropriate secondary reagents (PE-coupled anti-human (Southern Biotech, 1/500) for conditions i), ii) and iii) or PE-coupled anti-rat (Southern Biotech, 1/200) for condition iv)). Cells were finally resuspended in 100 μl PBS, 5% FBS, 2.5 mM EDTA containing the live/dead stain DRAQ7 (Biolegend, 1/2500) and analyzed on a CytoflexS B2-R3-V4-Y4 (Beckman-Coulter). Specific geometric mean fluorescence intensity of ligand binding on single live eGFP-positive cells (i.e. cells with fluorescence intensity of eGFP between 10⁵ and 10⁶, compared to negative cells at 10³) was obtained by subtracting background of condition iii). For competition experiments, a saturating dose (50 μg/mL) of Ig-FGL1 was added for 30 min at 4°C prior to the addition of Flag-BAFF (ACROBiosystems) or Flag-APRIL (prepared in house). After washing with FACS buffer, cells were then stained with anti-Flag BV421 (Biolegend) for 30 min at 4°C, washed again, and run on flow cytometry. Data was analyzed with FlowJo software (v10.7.1).

Flow cytometry

Immune cells in blood were profiled by flow cytometry. Red blood cells were lysed with FACS lysing solution (BD). All cells were filtered with a mesh strainer before running on the flow cytometer. Antibodies for immune cell profiling (ex. Ly6G, B220, CD11c, F4-80, MHCII, Ly6C, CD11b, CD44, CD3, CD62L, Foxp3, CD4, NK1.1, CD69, PD1, CD8, TACI, LAG3, CD45, IgD, CD19, CD93, CD23, IgM, CD69, Biolegend or BD) are shown in the Key Resource Table, following staining protocols similar to aforementioned binding experiments. Foxp3 transcription factor staining buffer set (Thermo Fisher) was utilized for intracellular staining.

Immunizations and FGL1 administration

Serum was collected from all mice before immunization. Mice were immunized ip with 50 μg of NP-AminoEthylCarboxyMethyl-FICOLL (NP-Ficoll, Biosearch Technologies) in PBS. Serum was collected at Day 7 and measured by ELISA. High binding 96-well plates (Corning) were coated with 1 μg/ml NP-conjugated bovine serum albumin (NP₂₇-BSA, Biosearch Technologies) at 4°C overnight. Plates were blocked with 10% FBS in PBS for 1 hour at 37°C. Sera were diluted in blocking buffer and incubated for 2 hours at 37°C. Bound antibody was detected via SBA clonotyping system-HRP (Southern Biotech). Following washes with PBST, 1-Step Ultra TMB-ELISA substrate solution (Thermo Scientific) was added. After color development, the reaction was stopped with acid and absorbance was measured at 450 nm.

To test for LAG-3 independence, mice were injected ip one day prior with 200 μg of anti-LAG3 or control Ig. Endotoxin-free plasmids of Ig-FGL1 containing the LALA-PG mutation (L234A/L235A/ P329G)⁷⁵ to avoid FcR binding was overexpressed by iv hydrodynamic injection in PBS according to the following protocol⁷⁶ three days before the administration of NP-Ficoll. 200 μg of Ig- human FGL1 was also injected ip into Fgl1−/− and double Fgl1−/− Taci−/− mice at D1, D2, and D4, with 50 ug NP-Ficoll injected ip at D0. For TD studies, mice were immunized ip with 100 μg of 4-Hydroxy-3-nitrophenylacetyl-Chicken Gamma Globulin (NP-CGG Ratio 20-29, Biosearch Technologies), or 37.5 μg of 2,4,6-Trinitrophenyl Keyhole Limpet Hemocyanin (TNP-KLH, Biosearch Technologies), mixed 1:1 with Imject alum adjuvant (Thermo Fisher). Serum was collected at 7 and 14 days post immunization and measured for Ig levels by ELISA. To investigate memory responses, mice were rechallenged 78 days post injection with 10 μg of NP-CGG in PBS or 50 days post injection with 18.75 μg of TNP-KLH in PBS ip. Serum was collected 7 days post injection and analyzed as above.

Female autoimmune mice at 8 weeks of age were injected ip with 200 μg Ig- human FGL1 weekly for 8 weeks, then rested for 4 months. Lymph node areas over the entire body were measured with a caliper on mice anesthetized with isoflurane at 8 months. Serum was collected retroorbitally every two months. Lymph node or splenic cells were counted on Vi-cell XR (Beckman Coulter) at 10 months, the time of sacrifice, and profiled via mass cytometry. Blood was profiled via flow cytometry.

Enzyme-linked immunospot (ELISPOT)

MultiScreenHTS IP filter plates (Millipore Sigma) were coated with 1 μg/ml NP-BSA (Biosearch Technologies) in PBS overnight at 4°C. ELISPOT was conducted on negatively selected purified mouse B cells (EasySep Mouse Cell Isolation Kit, Stemcell) from spleen and mouse femur and tibia bone marrow. Spleen was manually processed by grinding with glass slides, filtering, and ACK lysis (Thermo Fisher) followed by B cell purification. For bone marrow, bones were first immersed in 70% ethanol for 1 minute for sterilization, followed by 3 successive washes with PBS. Bones were cut with sterile forceps and scissors, then flushed with a 23-gauge needle attached to a 10cc syringe filled with ice cold 2% FBS in PBS. Bones were flushed until they became white. ACK lysis buffer (Thermo Fisher) was added, followed by washing, filtration, and B cell purification. Plates were blocked with 10% FBS in PBS for 30 min at 37°C. 0.5 million purified B cells were added to each well in RPMI, 2% FBS and cultured overnight in an incubator at 37°C, 5% CO₂. Plates were washed with PBST (0.05% Tween-20 in PBS) twice, and AP-conjugated antibodies against IgM heavy chain (Southern Biotech) at a 1/1000 dilution were incubated overnight at 4°C. The following day plates were washed with PBST four times, and 100 μL of AP substrate (Bio-Rad) was added to the plate and incubated at room temperature following manufacturer’s instructions. Plates were quenched with distilled water, and wells were counted on an AID ELISPOT reader.

Confocal microscopy

TACI-transduced HEK293T cells were stained with APC-conjugated anti-TACI antibody (Biolegend) for 15 mins at RT. Cells were then incubated with 1 μg/ml Ig- human FGL1 or control Ig at 37°C for 15 mins (staining for endosome colocalization) or 30 mins (staining for lysosome colocalization). Next, cells were fixed with 4% PFA and permeabilized with 0.1% saponin. Cells were further stained with anti-CD147 (Biolegend, dilution 1/100), anti-EEA1 (Thermo Fisher Scientific, dilution 1/100) or anti-LAMP1 (Miltenyi, dilution 1/50) primary antibody at 4°C overnight. AF594-conjugated goat anti-mouse IgG secondary antibody (Biolegend, dilution 1/500) was then stained for 2 hours at RT. DAPI (Sigma Aldrich, 1 μg/ml) was stained for cell nucleus. Data were acquired on a Leica Stellaris 8 confocal microscope using the lightening mode and analyzed using a macro in ImageJ. Data were blinded when scoring.

Internalization assay

An optimized internalization assay was utilized.⁶³ 0.5 million purified B cells (Stemcell) or splenocytes were added to a 96 well plate, and samples were washed and stained with DPBS, 2% FBS. FGL1 was then added to cells at varying concentrations in a 2X master mix, followed by an antibody cocktail with anti-mouse CD16/CD32 Fc block (BD) and 0.2 μg/ml (final concentration) of anti-TACI-BV421 (Biolegend). Cells were incubated at 37°C for 1 hour to examine receptor internalization in the presence and absence of FGL1, with additional samples incubated at 4°C as an internalization control. Cells were washed, and 50 μL of 0.1M glycine pH 3.5 was added for 1 min at room temperature to remove surface staining. Cells were washed again and stained with Near IR fixable live/dead dye (Invitrogen, 1/2000) in PBS for 10 minutes at room temperature. After another wash step, cells were stained with a B cell lineage marker (B220-BV480, Biolegend, 1/200) at 4°C for 15 min, washed again, and run on flow cytometry. Debris, doublets and dead cells were excluded from the analysis using FlowJo software (v10.7.1). Cells without antibodies were included as a negative control, along with cells from Lag3−/− and Taci−/− mice to examine receptor dependence and assay specificity.

Mass cytometry and data analysis

Panel of metal-conjugated antibodies utilized for this study are in the Key Resource Table. Antibodies not commercially available were labeled in house using purified antibodies with a Maxpar X8 metal labeling kit (Standard BioTools Inc) for lanthanide metals and Maxpar MCP9 labeling kit (Standard BioTools Inc) for cadmium metals according to manufacturer’s instructions. Lymph nodes were manually processed by grinding with glass slides, filtering, ACK lysis, and filtering again, with wash steps in between. 5 million cells were stained per sample. Cell suspensions were filtered through a 70-μm cell strainer and washed with Cell Staining Buffer (Fluidigm). Before antibody staining, cells were first incubated with anti-mouse CD16/CD32 Fc block (BD) for 10 minutes at room temperature to block Fc receptor engagement. An antibody cocktail with metal-labeled antibodies against extracellular targets was added for 30 min on ice. Cells were then washed and labeled with 12.5 μM Cell-IDTM Cisplatin (Standard BioTools Inc) to gate out dead cells. Labeling was quenched with pure FBS, after which cells were washed and fixed with Foxp3 Fixation/Permeabilization Concentrate and Diluent (Invitrogen) for 30 min on ice. Cells were washed and stained with a metal-labeled antibody cocktail against intracellular targets in Permeabilization Buffer (Invitrogen) for 30 min on ice, followed by another wash in Permeabilization Buffer and second wash in Cell Staining Buffer. 125 μM of Cell-ID Intercalator-Ir (Standard BioTools Inc), a DNA intercalator, was incubated at 4°C overnight, covered in foil to protect from light. Cells were washed, resuspended in calibration beads (Standard BioTools Inc) prepared in a 1:10 dilution in ddH₂0, and filtered in 35 μm cell strainer cap tubes before acquisition on CyTOF Helios (Standard BioTools Inc).

FCS files were normalized based on the beads in the sample at acquisition and gated based on non-beads, DNA, live/dead, and CD45⁺ in OMIQ software (Dotmatics). For each file, if the total number of cells exceeded 20,000, 20,000 cells were randomly sampled and exported as CSV files for downstream analyses in Python. UMAP visualization and Louvain clustering were performed using a Python library called Scanpy with default parameter settings.⁷⁷

Quantification and Statistical Analysis

Structure predictions were created using Alphafold2 multimer prediction in the ColabFold version 1.5.2 notebooks with mmeqs2.^78–80 The top-ranked model out of the 5 trained models from AlphaFold prediction was used for modeling in PyMol version 2.5.0. In the prediction of TACI:FGL1, the sequences of full-length FGL1 and 1-70 of TACI were used. Confidence metric using per-residue pLDDT scores are stored in the B-factor column of the .pdb files.

Prism 10.0 software (GraphPad) was utilized to generate graphs. Schematics were generated using BioRender.com. Statistical analyses were calculated utilizing Prism and Microsoft Excel, using the statistical method outlined in the Figure legend. For feature heatmaps, the mean expression levels of each marker (row) were z-score scaled to show the relative activity of that marker across all the Louvain clusters. Spearman’s correlation test was used to test for similarity, with r>0.7 and p<0.05 used to determine significance. Student’s t-test or one-way ANOVA followed by adjustments for multiple comparisons were utilized depending on whether the data compared multiple groups or not. p-values were reported as follows: ns, not significant, *, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001.

HIGHLIGHTS.

• FGL1 is an inhibitory ligand for B cells via TACI

• FGL1 binds at a N-terminal site distinct from TACI’s known ligands

• FGL1 suppresses a subset of B cell responses and induces TACI internalization

• FGL1 ameliorates autoimmune symptoms in B6/lpr lupus mouse model via TACI

Data and code availability

Data supporting the findings of this study are available from the corresponding authors upon reasonable request. This paper does not report original code. Additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.