Curbing Autoimmunity: A New Fab Fragment Targeting CD40‐CD40L Halts B‐Cell Activation and Differentiation

ABSTRACT

Dysregulation of the CD40‐CD40L axis is implicated in autoimmune diseases. Early clinical trials targeting CD40L with antibodies failed due to Fc‐mediated side effects. To address this, we developed an anti‐CD40L Fab fragment, Fab20, designed to block B‐cell activation. Fab20 was evaluated for its binding properties, CD40‐CD40L inhibition, and effects on human B‐cell activation and differentiation using immunoassays, cryo‐electron microscopy, flow cytometry, and cell cultures. Fab20 binds CD40L with a dissociation constant of 70 nM. Structural analysis revealed a “propeller‐like” structure consisting of three Fabs binding to the CD40L trimer, sterically blocking parts of the CD40 binding site. Fab20 effectively inhibited B‐cell activation, maintaining naïve B cells in their inactive state, and suppressed antibody (IgG) production over 14 days. Fab20 represents a promising novel therapeutic approach for treating autoimmune diseases driven by CD40‐CD40L dysregulation. Its mechanism of action, coupled with the absence of Fc‐mediated effects, suggests a favorable safety profile.

The CD40‐CD40L pathway is often dysregulated in B‐cell‐driven autoimmune diseases, making the development of CD40L inhibitors a promising therapeutic approach in managing autoimmune responses. We developed an inhibitory anti‐CD40L Fab fragment, Fab20, and demonstrated that Fab20 effectively suppresses B‐cell proliferation, differentiation, and antibody production in vitro. Made with Biorender.

Abbreviations

APCs

antigen presenting cells

BAFF

B‐cell activating factor

BLI

biolayer interferometry

BSA

bovine serum albumin

BV

Brilliant Violet

CD40L

CD40 Ligand

Cps

counts per second

Cryo‐EM

Cryo‐electron microscopy

DMSO

dimethyl sulfoxide

EDTA

ethylenediaminetetraacetic acid

FBS

fetal bovine serum

FSC

forward scatter

HEK293F

human embryonic kidney 293 Freestyle

HSA

human serum albumin

Ig

immunoglobulin

IL

interleukin

mAbs

monoclonal antibodies

MFI

median fluorescence intensity

MS

multiple sclerosis

NFκB

nuclear factor kappa‐light‐chain‐enhancer of activated B cells

NHDFs

Norman human dermal fibroblast

NHS

N‐hydroxysuccinimide ester

nIR

near infrared

O/N

overnight

PBMCs

peripheral blood mononuclear cells

PBS

phosphate‐buffered saline

PEG

polyethylene glycol

RA

rheumatoid arthritis

Rh

recombinant human

RT

room temperature

SD

standard deviation

SLE

systemic lupus erythematosus

SS

Sjögren's syndrome

SSC

Side scatter

TBS

tris‐buffered saline

TNFR

tumor necrosis factor receptor

TRAF

tumor necrosis factor receptor‐associated factor

TRIFMA

time‐resolved immunofluorometric assay

Tw

Tween20

1. Introduction

The adaptive immune response relies on coordinated interactions between different cell types. Antigen‐presenting cells (APCs) capture pathogens and present antigenic peptides on MHC molecules to helper T cells to activate them. The interaction between activated T cells and cognate B cells, along with cytokines from T cells, drives B‐cell activation. This leads to clonal expansion and differentiation of B cells [1, 2]. This activation is regulated by complex signaling systems, involving an array of co‐stimulatory and co‐inhibitory receptors. These receptors introduce an important signal for cell activation or inhibition and are crucial for maintaining central and peripheral tolerance [3, 4]. The interaction between CD40 and CD40 ligand (CD40L, also known as CD154) plays a significant role in B‐cell activation and differentiation [4, 5]. CD40 is a co‐stimulatory receptor on B cells and other APCs. It is a 48 kDa type I transmembrane protein and a member of the tumor necrosis factor‐receptor (TNFR) superfamily [6]. CD40 signaling is essential in inflammatory responses and plays an important role in T‐dependent B‐cell activation, proliferation, and differentiation [5, 7, 8]. The membrane molecule CD40L efficiently binds to CD40. CD40L is a trimeric type II transmembrane protein primarily expressed on activated T cells but is also present on platelets and in soluble form [6, 9, 10].

Upon CD40‐CD40L interaction, CD40 signaling activates downstream pathways involving TNFR‐associated factors (TRAFs), especially nuclear factor kappa‐light‐chain‐enhancer of activated B cells (NFκB) signaling. In B cells, this ensures a commitment to clonally expand the antigen‐specific response and the development of plasma and memory B cells. The CD40‐CD40L pathway is pivotal for maintaining the immune system's ability to uphold homeostasis and respond to foreign pathogens, for example, by inducing B‐cell antibody production [5, 11].

An unbalanced immune system can lead to disease, and CD40‐CD40L interaction is a hallmark in the pathology of several inflammatory and autoimmune diseases. In autoimmune diseases such as systemic lupus erythematosus (SLE) [12, 13, 14, 15], Sjögren's syndrome (SS) [16, 17], rheumatoid arthritis (RA) [18], and multiple sclerosis (MS) [19, 20], dysregulation of the CD40‐CD40L axis has been identified as a critical factor in loss of tolerance. CD40‐CD40L dysregulation results in excessive immune activation and inflammation, which subsequently leads to tissue damage and, in severe cases, organ failure, as seen in SLE patients with lupus nephritis. CD40 and CD40L are often overexpressed in autoimmune diseases, further driving the disease with excessive B‐cell activation and production of pro‐inflammatory cytokines and autoantibodies [15, 18, 19].

Given the role of CD40‐CD40L in autoimmunity, CD40L inhibition has been a promising therapeutic target since the beginning of the 21st century. The emergence of biological treatments using antibodies led to clinical trials utilizing anti‐CD40L antibodies, which demonstrated significant effects in diseases such as SLE [21, 22, 23, 24]. However, these first‐generation anti‐CD40L antibodies (Ruplizumab and Toralizumab) were discontinued due to side effects mediated by interactions with Fc receptors on platelets, leading to activation and aggregation of platelets and, thus, the occurrence of thromboembolic events [25, 26, 27, 28]. In recent years, new anti‐CD40L antibodies with modified Fc domains that do not induce immune activation have emerged in clinical trials; these include Letolizumab, Frexalimab, Tegoprubart, and TNX‐1500 (Ruplizumab with a modified Fc‐region), all considered “Fc‐silent” [29, 30, 31, 32, 33, 34]. An alternative approach to avoid the Fc‐mediated side effects and potentially enhance drug tissue penetration is to develop inhibitors smaller than full‐size antibodies. Here, Fab fragments (approximately 50 kDa), lacking the Fc domain and three times smaller than IgG molecules, show great therapeutic potential. To date, only one Fab fragment targeting CD40L, Dapirolizumab pegol, has been developed and is currently in phase 3 clinical trials for treating SLE [35]. Dapirolizumab pegol is a polyethylene glycosylated (PEGylated) Fab fragment and has demonstrated excellent safety while inhibiting the CD40‐CD40L interaction and ameliorating SLE symptoms, first in Cynomolgus monkeys and later in patients [35, 36]. The addition of PEG significantly increases the half‐life, while also increasing the size of the drug by 40% (20 kDa increase) [37]. Adding PEG could decrease biological activity and tissue penetration, thereby reducing drug bioavailability. Furthermore, even though Fc‐mediated toxicity has been addressed, repeated administration of PEGylated drugs may induce anti‐PEG antibody formation, further complicating long‐term use [38, 39].

In an autoimmune disease like SLE, treatment regimens with corticosteroids and general immunosuppression remain a cornerstone. This general, untargeted immunosuppression causes severe side effects, which underlines the need to develop targeted treatment options to improve patient outcomes and quality of life [40]. The development of monoclonal antibodies and antibody‐based therapies, which have revolutionized treatment within cancer and some autoimmune diseases, has not been as profound a success as anticipated for SLE. Only two antibody‐based drugs have been approved for the treatment of SLE up until this day: Belimumab (anti‐B‐cell activating factor [BAFF]) [41, 42] and Anifrolumab (anti‐interferon α receptor) [43]. Improving SLE treatment strategies may involve moving away from full‐size antibodies and toward molecules such as Fab fragments, which are smaller, less immunogenic, and easily engineered and modified [44].

In this study, we describe the development of a novel non‐PEGylated Fab fragment targeting human CD40L and demonstrate its efficient inhibitory effects on B‐cell activation. The goal was to inhibit the CD40‐CD40L pathway to prevent autoreactive B‐cell activation and differentiation and ultimately avoid the production and secretion of autoantibodies. We identified an antibody with a potent inhibitory capacity for preventing CD40L from binding to its receptor, CD40. From this antibody, we generated a Fab fragment, Fab20. This novel Fab fragment, with strong inhibitory effects on CD40‐CD40L interaction, prevented primary human B‐cell activation and differentiation in vitro. We hope that Fab20 might contribute to the treatment of patients suffering from autoimmune diseases such as SLE, SS, RA, and MS.

2. Results

2.1. A Mouse Anti‐Human CD40L Antibody and Recombinant Fab Fragment Binds CD40L and Inhibits CD40‐CD40L Interaction

A total of 30 mouse anti‐human CD40L mAbs were generated in‐house by immunizing NMRI mice with rhCD40L and producing hybridomas for mAb production, as described in Pedersen et al. [45]. The mAbs were tested for binding to rhCD40L coated in microtiter wells and binding to membrane‐bound CD40L by flow cytometry [45]. The candidates were tested for their ability to inhibit CD40‐CD40L interaction, either in a setup with CD40‐Fc added to rhCD40L coated in microtiter wells or CD40‐Fc added to CD40L on activated T cells (Figure S1). Mouse anti‐human CD40L mAb candidate number 20, mAb20, proved good binding to CD40L on stimulated PBMCs and did not bind unstimulated PBMCs, which lack CD40L (Figure 1A–C). MAb20 also demonstrated superior inhibitory capacity for preventing CD40‐CD40L interaction (Figure 1D–F), as well as being able to reduce the percentage of CD40‐Fc binding of CD40L on stimulated T cells (in a PBMC pool) by 90.22 ± 3.21 % (from 36.34 ± 7.27% to 3.37 ± 0.89%, mean ± SD of three individual assays).

FIGURE 1.

Binding and inhibitory capacity of mouse anti‐human CD40L monoclonal antibody, mAb20. (A) Immunoassay (TRIFMA) of mAb20 binding to rhCD40L coated in microtiter wells. As a negative control, wells were coated with HSA. The y‐axis corresponds to the europium signal given as counts per second (cps). All samples are run in doublets and blanked. (B) Flow cytometry showing mAb20 binding to membrane‐bound CD40L expressed on stimulated PBMCs (mainly T cells), detected by a goat anti‐mouse IgG PE‐labeled antibody. The negative control used murine IgG as the primary antibody, while the positive control was a commercially available mouse anti‐human CD40L PE‐labeled antibody. The y‐axis gives the percent CD40L‐positive cells. (C) Scatter plot of the use of mouse anti‐human CD40L mAb20 on stimulated and unstimulated PBMCs. The y‐axis shows forward scatter (FSC), and the x‐axis is the PE signal. (D) Immunoassay demonstrates the inhibitory capacity of mAb20. A competition assay where CD40‐Fc binds to rhCD40L in microtiter wells. The CD40‐Fc was mixed with the mAb20 or no admixture. The y‐axis equals the CD40‐Fc binding signal (cps). (E) Flow cytometric assessment of the ability of mouse anti‐human CD40L mAb20 to inhibit biotin‐CD40‐Fc binding to CD40L on stimulated PBMCs. The negative control consisted of adding Streptavidin‐BV421 only, giving the background signal. The y‐axis equals the percentage of cells bound by CD40‐Fc (BV421 signal). (F) Histogram illustrating the inhibition of mAb20 seen in (E). Blue histogram density curve corresponds to the signal with biotin‐CD40‐Fc only. The red histogram density curve corresponds to the signal of CD40‐Fc in the presence of mouse anti‐human CD40L mAb. All Figure 1 graphs and plots are representative experiments from three individual experiments (n = 3).

2.2. A Fab Fragment of a Mouse Anti‐Human CD40L Antibody Binds CD40L and Inhibits CD40‐CD40L Interaction

When targeting CD40L, Fc receptors have previously been shown to induce severe side effects [25, 26, 27, 28]. Thus, the prime inhibitory candidate, mAb20, was redesigned into a recombinant Fab fragment, Fab20 (Figure 2A), thereby removing the Fc receptor interaction.

FIGURE 2.

Production of recombinant Fab anti‐human CD40L. (A) Schematic of Fab20. The heavy chain was expressed with a Sortase‐tag for site‐specific labeling purposes as well as a His‐tag for purification purposes. (B) A flow cytometric titration assay for binding biotin‐Fab20 (titrating Fab20 from 10–0.08 µg/mL, corresponding to 190–1.5 nM) to NHDF‐CD40L or NHDF cells. Streptavidin‐BV421 was used to detect the bound biotinylated Fab20, and the y‐axis gives the percentage of CD40L‐positive cells. (C) Flow cytometric assessment of the capacity of Fab20 to inhibit the binding of biotin‐CD40‐Fc to NHDF with CD40L expression. The y‐axis corresponds to the percentage of cells bound by CD40‐Fc. Titration of the Fab20 (7–0.1 µg/mL, corresponding to 133–1.9 nM) demonstrates a concentration‐dependent inhibition of CD40‐CD40L interaction (IC50 is 0.63 µg/mL, corresponding to 12.01 nM). The red line corresponds to the signal of Streptavidin‐BV421 only (negative control). (D) Histograms of the Fab20 titration inhibition assay illustrated in (C). “CD40‐Fc only” is shown as the blue histogram. Green to red represents increasing concentrations of Fab20, red being the highest concentration used in the assay. All assay graphs and plots in Figures 2C–D are representative of three individual experiments (n = 3).

We subsequently assessed whether the Fab fragment of mouse anti‐human CD40L mAb20 retained binding and inhibitory abilities comparable to the full‐size antibody. Fab20 demonstrated binding to membrane‐bound CD40L (Figure 2B). Additionally, Fab20 proved to have kept its excellent inhibitory capacity by inhibiting the interaction between CD40‐Fc and CD40L. Fab20 exhibited a concentration‐dependent inhibitory activity (IC50 = 0.63 µg/mL [12.01 nM], Figure 2C,D), where the highest concentration of 7 µg/mL reduced the CD40‐CD40L interaction by 98.12 ± 0.66% (from 60.95 ± 6.11% to 1.14 ± 0.47%, mean ± SD, n = 3).

To further characterize the Fab fragment, the binding strength was tested by BLI to estimate the binding affinity (K_D) of Fab20 (Figure 3A). Since CD40L is a trimer, a 1:1 interaction between the Fab anti‐CD40L and rhCD40L could not be expected in this setup. Thus, the EVILFIT fitting software was used to fit BLI binding curves for K_D estimation, considering the possibility of heterogeneous binding (Figure S3). The EVILFIT software estimated three different types of interaction between CD40L and Fab20. The least abundant type of interaction between Fab20 and CD40L demonstrated a relatively weak binding with a K_D of 2.81 µM; however, this only contributed to approximately 27% of the interaction (Figure S3C). The two remaining types of binding proved to have a much stronger binding with K_D values of 70 and 71 nM, also being the most abundant types of interaction observed between Fab20 and rhCD40L (Figure S3D,E).

FIGURE 3.

Binding kinetics and structural analysis of Fab anti‐CD40L's binding to rhCD40L. (A) BLI binding curves of Fab20 binding to decreasing concentrations of rhCD40L (six different concentrations). The x‐axis is given as response (nm), and the y‐axis as time. The association phase corresponds to the first 400 s, followed by a 1200 s dissociation phase. Data are from one experiment. (B) A cartoon model and cryo‐EM density map (3.9 Å resolution (gold standard Fourier shell correlation [GSFSC]) of the “propeller shape” structure of three Fab fragments bound to human CD40L. Below are shown representative 2D classes. (C) Cryo‐EM focus map (3.4 Å resolution [GSFSC]) and corresponding model focusing on the binding of one Fab to one hCD40L monomer. (D) Zoom in on the important and well‐resolved interactions between Fab20 and the CD40L monomer. (E) Models of CD40 binding to CD40L compared with the Fab20 described in this report [46]. Clashing residues of Fab20 and CD40 that hinder CD40‐CD40L interaction are highlighted in red.

2.3. Structural Analysis of Fab Anti‐CD40L Interaction With CD40L

Cryo‐EM analysis revealed a propeller‐like structure that, despite some preferred orientation of the particles in the ice, could be reliably resolved to a resolution of 3.9 Å (Figure 3B, EMPIAR‐13315). Similar propeller‐like shapes could also be seen directly in the motion‐corrected, CTF‐fitted, and denoised images (Figure S4), confirming that three Fab20 molecules can bind the CD40L trimer simultaneously.

By exploiting the observed three‐fold (C3) symmetry of this propeller‐like structure, we made a symmetry‐expanded cryo‐EM focus map compiling the information from all three fab fragments bound to the CD40L trimer and reached an improved resolution of 3.4 Å (GSFSC), with local resolutions down to ∼2.5 Å (Figure 3C) (EMDB‐52634). The focus map consisted of part of the Fab fragment and one monomer of the CD40L. Based on this cryo‐EM focus map, we built a structural model of the interaction between Fab20 and the CD40L monomer (Figure 3C; Figure S4E) (PDB: 9I5N). The quality of the map allowed confident positioning of the backbones of CD40L and Fab20 molecules, as well as some of the important side chains, facilitating the binding.

Our structural analysis of Fab20 binding to CD40L revealed that the interaction interface between the Fab20 light chain (LC1) and the CD40L is larger than that of the heavy chain (HC). These interactions primarily target a loop on the side of CD40L (residues 196‐202) (Figure 3C,D). In this region, we observe a clear hydrophobic interaction primarily forming around the extruding CD40L residue Phe 201 by the Fab20 LC residues Leu 94, Phe 96, and HC residue Trp 52 (Figure 3D bottom right). The data also support the formation of electrostatic interactions between CD40L residue Glu 202 and a metal ion (probably Magnesium) coordinated by Fab20 LC residues Asp 30, Asp 31, Asp 92, as well as between the CD40L residue Lys 196 and Fab20 Lc residue Glu 50 (Figure 3D top right). Outside the described area, our model suggests an additional five hydrogen bonds (four formed from HC, and one from LC) as well as two salt bridges between Fab20 and CD40L. The summarized structural analysis of Fab20 binding to CD40L (as well as comparison with other published structures) is illustrated in Movie 1. Interestingly, Fab20 binds to a unique region of CD40L compared with the previously published Fab fragment of Ruplizumab (PDB: 1I9R; Figure S9A) and the predicted binding of the Fab fragment of Tegoprubart, Frexalimab, and Dapirolizumab (predictions made by AlphaFold; Figure S9B–D).

Comparing our Fab20 bound structure with the binding of CD40 (PDB: 3QD6), we can see clashes between the heavy chain of Fab20 and CD40 N‐terminal residues 26–32 and 40 (colored red in Figure 3E). Additionally, an electrostatic repulsion between the light chain of Fab20 and CD40 in the region of LC‐Thr27, LC‐Thr69, and CD40‐Glu114 might also help prevent simultaneous binding of Fab20 and CD40. Thus, even though Fab20 does not fully occupy the CD40 binding region, these clashes might be sufficient to perturb CD40‐CD40L binding while Fab20 is present (Figure 3E) [46].

2.4. A Fab Fragment of a Mouse Anti‐Human CD40L Antibody Prevents Human B‐Cell Proliferation in Vitro

The CD40‐CD40L pathway is an important target for ameliorating disease phenotype and symptoms in patients suffering from autoimmune diseases. The effect of CD40‐CD40L inhibition should prevent the activation and proliferation of autoreactive B cells that ultimately may result in the production of autoantibodies. After characterizing the binding abilities and kinetics as well as the inhibitory capacity of Fab20, mainly focusing on the effects on CD40L and the T cell perspective, we aimed to investigate the potential of Fab20 in preventing B‐cell proliferation.

This was done by setting up cocultures using NHDF‐CD40L cells and adding cytokines necessary for human B‐cell activation, here being IL‐2, IL‐21, BAFF, and a pulse of IL‐4. Naïve human B cells were subsequently added to the culture system (Figure 4A). The coculture system was either left without further addition of reagents (untreated control) or received Fab20 (or Fab isotype control, both at 3.5 µg/mL corresponding to 66.5 nM), aiming to inhibit CD40‐CD40L signaling and downstream activation. On days 4, 7, 10, and 14, B cells were harvested and counted to examine their proliferative abilities under either of the conditions. In the untreated control and the Fab isotype control, B cells proliferated significantly (Figure 4B,C). Fab20 completely blocked the ability of naïve B cells to proliferate in this setup, and due to the missing CD40L‐signal, fewer B cells were present in the wells. The cell count with the reseeding multiplication factor is given in Figure 4B. The cell counts, without the reseeding multiplication factor, are shown in Figure S8A–D. At all timepoints (days 4, 7, 10, and 14), a statistically significant difference in human B‐cell numbers was found when comparing Fab20‐treated B cells to both untreated control and Fab isotype control‐treated B cells (Tukey's multiple comparisons test). At all timepoints, significantly fewer B cells were present in wells receiving Fab20. No significant difference in cell numbers was observed when comparing untreated control with Fab isotype control‐treated B cells. The adjusted p‐values of each comparison are shown in Table 1.

FIGURE 4.

Naïve human B cell:NHDF‐CD40L coculture assay. (A) Graphical illustration of the B cell:NHDF‐CD40L coculture assay with relevant days of treatment and the readouts following the treatments (right side). B cells were seeded on top of CD40L‐expressing NHDFs. B cells were either left “untreated” as a control or received Fab20 or Fab isotype control. On days 4, 7, 10, and 14, cells from each treatment were harvested, counted, and cryopreserved for flow cytometry, and culture supernatants were saved. After the 14‐day assay, B cells were analyzed by flow cytometry, and levels of antibodies/IgG in culture supernatants were analyzed by TRIFMA. (B) B‐cell proliferation assay. B‐cell counts from days 4, 7, 10, and 14. Counts are from three individual wells of each treatment at each timepoint (n = 3), given below the x‐axis. Values given as mean ± SD, and the y‐axis equals the cell count per well (log‐scale). Tukey's Multiple Comparisons test was used for statistical analysis, *p < 0.05, ***p < 0.01. (C) Representative phase‐contrast light‐microscopy images (10× objective with 10× ocular) of wells containing B cells and CD40L‐NHDFs on day 14, either left untreated or treated with a Fab isotype control or Fab20 (at 3.5 µg/mL/66.5 nM). The scale bar is inserted in the lower right corner.

TABLE 1.

B‐cell counts statistics.

	Day 4	Day 7	Day 10	Day 14
Untreated vs. Fab20	0.0003	<0.0001	<0.0001	<0.0001
Fab isotype control vs. Fab20	0.0218	<0.0001	<0.0001	<0.0001
Untreated vs. Fab isotype control	0.1824	0.9886	0.2550	0.1472

Note: Adjusted p‐values from Tukey's multiple comparisons test on B cell counts from B cell: NHDF‐CD40L coculture assay on days 4, 7, 10, and 14.

Fab20 blockade of human B‐cell proliferation was also observed in the microscope, where significantly fewer B cells were present in the wells treated with Fab20. This effect was evident from day 4 to day 14 (Figure 4C; Figure S8E). No differences were observed between the untreated control and Fab isotype control wells.

Fab20 displayed a concentration‐dependent effect (Figure S6A–D) with the highest concentration of 3.5 µg/mL (66.5 nM) resulting in the most optimal inhibitory effect. This aligns well with the estimated dissociation constant of 70 nM (Figure 3A). The concentration‐dependent effect was clear when examining B‐cell proliferation, looking at cell counts and microscopy images. Additionally, a similar effect of Fab20 was observed in a less homogenous population of human pan B cells, where Fab20 demonstrated its impact on all human B cell subsets, including antigen‐experienced B‐cell subsets and not only the naïve B cells (Figure S7A,B).

2.5. A Fab Fragment of a Mouse Anti‐Human CD40L Antibody Inhibits Human B‐Cell Differentiation in Vitro

In addition to assessing B‐cell proliferation, the effect of Fab20 on human B‐cell activation and differentiation markers was also investigated by flow cytometry. Naïve human B cells at baseline (day 0) were characterized by being CD95/CD38 double negative (resting B cells). Upon B‐cell activation, CD95 is quickly upregulated, whereas CD38 is gradually upregulated predominantly on plasmablasts and plasma cells. In control cocultures, B cells rapidly upregulated CD95 (98.8 ± 0.46 % positive cells on day 4, Figure 5A,C). In contrast, CD38 was slowly upregulated over time with 28.3 ± 2.56% CD95⁺CD38⁺ B cells (plasmablasts and plasma cells) on day 14.

FIGURE 5.

Flow cytometric assessment of human B‐cell activation and differentiation. (A) Representative density plots of B‐cell activation and differentiation on days 0, 4, 7, 10, and 14; human B cells from untreated control wells, Fab20‐containing wells, or Fab isotype control wells. The y‐axis shows the CD95 signal. The x‐axis shows the CD38 signal. Three gates were created based on CD95 and CD38 status. CD95⁻CD38⁻ cells (double negative) were labeled as “resting B cells”. CD95⁺CD38⁻ B cells were labeled as “Activated B cells”, and CD95⁺CD38⁺ (double positive) B cells were labeled as a subset resembling plasmablasts and plasma cells, “PB+PC”. Naïve B cells were CD95⁻CD38⁻ (resting B cells). Density plots representative of three experiments (n = 3). (B) Representative density plots of IgD (x‐axis) and CD27 (y‐axis) status of B cells on days 0, 4, 7, 10, and 14. IgD⁺CD27⁻ B cells were classified as naïve B cells, and 97.1 % of B cells from baseline were IgD⁺CD27⁻. Upon activation, B cells downregulate IgD and become IgD⁻CD27⁻ double negative B cells. CD27 (IgD⁻CD27⁺) was used as a marker for B‐cell differentiation into memory B cells and plasmablast subsets. Density plots representative of three experiments (n = 3). (C) CD95 MFI on B cells demonstrates how CD95 cannot be detected on the surface of naïve B cells (day 0) but is quickly upregulated upon stimulation. Fab20 completely inhibits CD95 upregulation. N = 3, bar graphs with whiskers equals mean ± SD. (D) IgD MFI on B cells from the 14‐day assay. Day 0 B cells present with high IgD levels on the surface, but the expression is decreased to almost zero upon stimulation. Fab20 prevents IgD levels from decreasing, thus keeping B cells in a more naïve state. N = 3, bar graphs with whiskers equals mean ± SD. (E) Immunoassay quantifying total IgG levels in supernatants from B cell cocultures on days 4, 7, 10, and 14 (n = 4 from each timepoint, mean ± SD). On day 14, B‐cell differentiation caused untreated control and Fab isotype control B cells to produce high IgG levels. Fab20 completely prevented IgG production, also caused by the few surviving B cells on day 14. Unpaired t‐tests of day 14 IgG levels, *p < 0.05, ***p < 0.01.

Fab20 inhibited naïve B‐cell activation and differentiation. This, combined with the lack of proliferation in these cells and their continued lack of CD40‐CD40L stimulation, caused B cells to die, meaning at day 14, only a few B cells, mainly resting B cells (CD95‐CD38⁻, Figure 5A), were left for flow cytometric analysis. This effect was also seen when adding Fab20 to the coculture, where a concentration‐dependent response of Fab20 was observed (Figure S6E). Here, a gradual increase in CD95‐expression can be seen in day 4 cultures when decreasing Fab20 concentration. Moreover, Fab20 demonstrated very similar effects on human pan B cells, suggesting that the impact of Fab20 could influence multiple human B cell subsets, including antigen‐experienced and memory B cells, and not only naïve B cells (Figure S7C).

IgD is downregulated during B‐cell differentiation from naïve B cells into other B‐cell subsets. On the contrary, CD27 is not present on naïve B cells but is upregulated when B cells differentiate into memory B cells and plasma cells. Upon B‐cell activation, the IgD downregulation results in a subset of IgD⁻CD27⁻ double negative B cells, which may undergo class‐switching [47]. The naïve human B cells from day 0 (baseline) were indeed IgD⁺CD27⁻ (97.1 %). On day 4, 90.8 ± 0.85% were no longer IgD‐positive (IgD⁻CD27⁻ double negative B cells) in the untreated control group, and on day 14, 29.2 ± 1.63% were positive for CD27 and remained IgD negative (Figure 5B). Fab20 prevented this by keeping cells IgD‐positive, meaning that on day 4, only 5.57 ± 0.87% of B cells were no longer IgD‐positive (IgD⁻CD27⁻). Notably, very few B cells were present after treatment with Fab20, thereby complicating the analysis and interpretation of gating percentages. Thus, looking at IgD levels (median fluorescence intensity (MFI), Figure 5D) also emphasizes how Fab20 inhibits human B‐cell differentiation, causing these B cells to maintain a high level of IgD (IgD MFI = (1.9 ± 0.05)$\times$10⁴). In contrast, IgD was almost undetectable on B cells in the control groups (untreated = (0.02 ±0.002)$\times$10⁴).

This effect was also observed in human pan B cells (Figure S7D); however, the baseline levels of IgD and CD27 differed from naïve B cells, that is, 9.19% of pan B cells were IgD⁻CD27⁺ compared with 0.054% for the naïve B cell baseline. Here, Fab20 appeared to affect all subsets, resulting in a mean of 87.5 ± 2.59% IgD⁺CD27⁻ (naïve B cells) present in the Fab20‐treated wells at all time points (day 4–14).

2.6. The Fab Anti‐Human CD40L Inhibits Antibody Production From Human B Cells

Yet another aspect of B‐cell function and differentiation was examined by measuring antibody production. The levels of total IgG were measured in the supernatants from the coculture on days 4, 7, 10, and 14. This revealed that the naïve human B cells in the coculture first needed to proliferate and differentiate before beginning to produce significant amounts of IgG (Figure 5E). Fab20 completely prevented B cells from producing IgG, and only very low levels of IgG were measured in day 14 supernatants treated with Fab20. IgG secretion was inhibited by 99.97 ± 0.03% compared with the untreated control and by 99.94 ± 0.08% compared with the isotype control. The same effect of Fab20 was observed on the IgG secretion in pan B‐cell cocultures, demonstrating that it also effectively blocked antibody production by memory B cells (data not shown). Thus, different subsets of human effector B cells can be targeted successfully by treatment with Fab20.

3. Discussion

The initial studies of first‐line CD40L antibodies demonstrated pronounced effects for treating autoimmune diseases. However, clinical trials were halted due to Fc‐mediated side effects [25, 26, 27, 28]. Despite the setbacks, the CD40‐CD40L pathway is still a highly relevant target when aiming to treat autoimmune diseases, but Fc‐mediated side effects must be considered and avoided.

Therefore, we have developed a new Fab fragment of an anti‐CD40L antibody capable of inhibiting the CD40‐CD40L pathway. The monovalent Fab anti‐CD40L lacks the Fc domain, making it only one‐third the size of a full‐size antibody. This enables better tissue penetration, potentially increasing drug efficacy by offering access to cryptic epitopes while avoiding Fc‐mediated immune activation [44, 48, 49]. However, avoiding the Fc does result in a significantly decreased half‐life, which can be considered both an advantage and a disadvantage. It could be advantageous since a shorter half‐life enables a more dynamic control of CD40‐CD40L signaling, allowing it to rapidly turn CD40 signaling on and off. However, a shorter half‐life necessitates more frequent dosing to maintain therapeutic efficacy.

Here, we demonstrate how, among 30 different monoclonal antibodies targeting CD40L, we identified an antibody with superior inhibitory capacity for preventing CD40L from binding its receptor, CD40. We demonstrated this by using recombinant proteins and by showing how the anti‐CD40L antibody could prevent CD40 from binding to the membrane‐bound form of CD40L. To avoid Fc‐mediated side effects, potentially decrease immunogenicity, and improve tissue penetration, we redesigned the anti‐CD40L monoclonal antibody into a recombinant Fab fragment. We confirmed that Fab20 retained its binding and inhibitory capacity compared with the full‐size antibody, testing on recombinant CD40L coated in microtiter wells and on cells expressing CD40L. To further understand how Fab20 interacted with CD40L, analysis based on BLI and cryo‐EM was performed. These assays showed that Fab20 has a high affinity for CD40L with a K_D of 70 nM.

The cryo‐EM analysis revealed that Fab20 binds in a threefold stoichiometry to a unique site of CD40L compared with a previously reported CD40L binder. This binding to CD40L was facilitated by a mix of hydrophobic and ionic interactions, primarily dominated by the LC of Fab20 and with fewer interactions of the Fab HC. This is supported by the observation that the exchange of LC1 (in Fab20) to LC2, while keeping the HC constant, prevented binding. The binding was initially modelled using AlphaFold 3. Interestingly, the highest‐scoring AlphaFold model (which was also closest to the cryo‐EM model) identified some of the experimentally observed interactions, but not all (Figure S9E). The same hydrophobic CD40L Phe201 interaction, as observed in the cryo‐EM data, was also found in the predicted model. However, in the prediction, Fab20 was twisted by almost 180° around the Phe201, leading to a much smaller total interaction compared with the actual cryo‐EM informed model. This incorrect prediction by AlphaFold might be due to the metal ion interaction, which was not present in any of the predicted models. Even though AlphaFold failed to predict the orientation between Fab20 and CD40L, the region of interaction was largely correct.

The human CD40L is naturally N‐glycosylated on residue Asn240 of each monomer [50]. However, the recombinant CD40L used here was not glycosylated. By overlaying the glycosylated model with our model, we noticed that the glycosylation site chain is close to two asparagine residues in the Fab20 LC (Asp 30 and 28), which might form additional stabilizing interactions with the endogenous N‐glycosylated CD40L. Notably, the N‐glycosylation on Asn 240 in CD40L does not appear to sterically hinder Fab20 binding, which is also in line with the in vitro results where the Fab binds to the endogenous N‐glycosylated CD40L present on the various cell types we used.

A limitation of our study is the lack of comparison of Fab20 with other CD40L‐targeting therapeutics. However, we compared the binding regions of our Fab20 with other anti‐CD40L antibodies, including the first‐generation anti‐CD40L antibody, Ruplizumab (Fab region identical to the second‐generation anti‐CD40L antibody TNX‐1500), as well as the second‐generation anti‐CD40L therapeutics: Tegoprubart, Frexalimab, and Dapirolizumab. Interestingly, the region where Fab20 binds CD40L is unique in comparison with Ruplizumab [24] (crystal structure in PDB: 1I9R, Figure S9A) and likely also Tegoprubart, Frexalimab, and Dapirolizumab (binding predicted using AlphaFold) (Figure S9B–D). Consistent with these predictions, the amino acid sequence of Fab20 diverges substantially from those of other CD40L‐targeting therapeutics, supporting the conclusion that Fab20 engages a distinct epitope on CD40L. In general, Ruplizumab and Tegoprubart interact with CD40L mainly through polar contacts, while Frexalimab primarily interacts through electrostatic contacts (Figure S9A–C). In comparison, the interaction between Fab20 and CD40L is dominated more by hydrophobic interactions (Figure 4). Furthermore, structural superposition of Fab20 with Ruplizumab or Tegoprubart reveals no spatial overlap between the respective Fab fragments, suggesting that these constructs could bind CD40L simultaneously. In contrast, Frexalimab exhibits partial steric overlap with Fab20, indicating that concurrent binding is unlikely.

Comparison with the crystal structure of CD40 binding to CD40L [46] revealed some steric and electrostatic clashes between bound Fab20 and CD40, but not full competition for the same binding region. This steric and electrostatic clashing is likely enough to inhibit the CD40 binding, as observed in the in vitro data.

Finally, we demonstrated that Fab20 prevented B‐cell activation and differentiation in vitro. The effect of Fab20 was visible not only when using naïve human B cells but was also observed in pan B cells. Fab20 prevented B‐cell proliferation, which was evident by counting and assessing the cells by microscopy throughout the 14‐day assay. Additionally, B‐cell activation was inhibited, as demonstrated by the complete lack of CD95 upregulation and IgD downregulation on naïve human B cells treated with Fab20. Moreover, class‐switching and differentiation were also prevented by Fab20 treatment, evident by cells being retained as IgD‐positive while remaining negative for CD38 and CD27, both markers for B‐cell differentiation. A further assessment of the inhibitory effects of Fab20 on human B‐cell differentiation was shown by measuring the B cells’ ability to produce and secrete IgG. Here, Fab20 completely prevented IgG production and secretion from B cells during the 14‐day assay. Together, these results strongly suggest that this newly developed Fab anti‐CD40L can bind CD40L with high affinity and efficiently prevent CD40‐CD40L signaling, thus inhibiting T‐cell‐mediated B‐cell proliferation, activation, and differentiation. We envisage that Fab20 could serve as a new treatment option for autoimmune diseases where the CD40‐CD40L axis has been proven critical for developing symptoms, such as SLE, being one such disease.

Dapirolizumab, a single pegylated Fab’ anti‐CD40L, is in phase 3 clinical trials on SLE patients. The affinity of Dapirolizumab was reported with a remarkably low K_D, alleged to be 7.9 pM [51]. However, the setup of Dapirolizumab was not tested in a monovalent manner but by coupling the Fab’ fragment to an anti‐Fab antibody, thereby reinstating the bivalent nature of Fab into resembling an antibody. Thus, a comparison of the affinity of Dapirolizumab and Fab20 tested in the present report cannot be made. However, the dissociation constant of 70 nM we find for our monovalent Fab anti‐CD40L fragment is considered among high‐affinity binders, enabling efficient competition with the CD40 interaction.

As described above, a comparison of the Fab20 cryo‐EM structure and the AlphaFold 3‐predicted Dapirolizumab‐CD40L complex suggests that they bind different regions of CD40L (Figure S9D). Fab20 presented in this paper interacts with individual subunits in the CD40L trimer through both the heavy and the light chain, whereas Dapirolizumab is predicted to bind CD40L at the most extracellular part and does not overlap with Fab20 presented here (Figure S9D). The fact that the interface between CD40L and Dapirolizumab is predicted to be significantly different than Fab20 presented here suggests that these Fab fragments could both benefit the field of treatment strategies for patients suffering from autoimmune diseases.

The potential for our recombinant Fab anti‐CD40L (Fab20) as a future treatment option for patients with autoimmune disorders must be considered further. Given the in vitro nature of the assays in this study, Fab20 requires in vivo validation to confirm its inhibitory efficacy and safety profile. Additionally, the potential activation of compensatory signaling pathways should be carefully assessed. In future studies, it could be prudent to test Fab20 in a murine model with a humanized CD40‐CD40L axis in an autoimmune setting. However, to prolong the half‐life of Fab20 in vivo, different options will be explored either by PEGylation or by the addition of an albumin‐binding entity. Furthermore, additional studies are needed to evaluate the stability of Fab20, particularly concerning aggregation and its pH‐ and thermostability. These are all critical considerations for future large‐scale production. Finally, it would be relevant to consider other effects of CD40‐CD40L blockade in a therapeutic setting. Possibilities could be attenuated T‐cell proliferation and activation, consistent with impaired dendritic cell licensing and reduced costimulatory capacity in human systems [6], and as CD40 signaling indirectly regulates CD28‐dependent costimulation and IL‐12 production, its inhibition may limit sustained T‐cell expansion rather than initial TCR triggering [11]. In a low‐inflammation setting, this suboptimal priming may bias responding T cells toward hyporesponsiveness or tolerance [52]. Thus, CD40‐CD40L blockage is likely to impact not only B cells.

In conclusion, this newly developed Fab anti‐CD40L, Fab20, has demonstrated strong inhibitory effects on preventing CD40‐CD40L interaction and inhibiting downstream signaling, thereby preventing B‐cell proliferation, activation, and differentiation in vitro. We believe that with further testing, Fab20 could offer a new opportunity to treat patients suffering from autoimmune diseases with CD40‐CD40L axis dysregulation, especially SLE, MS, and RA.

4. Materials and Methods

4.1. Production and Design of Monoclonal Mouse Anti‐Human CD40L Antibodies and Mouse Anti‐Human CD40L Fab Fragment

Recombinant human CD40L (rhCD40L, residues: 113‐261 [53] + N‐terminal His‐tag) was expressed and purified in BL21(DE3) Escherichia coli. RhCD40L was used to immunize NMRI mice to fabricate hybridomas producing mouse anti‐human CD40L monoclonal antibodies (mAbs), as described in detail in Pedersen et al. [45]. In brief, NMRI mice were immunized subcutaneously three times at 14‐day intervals with 20 µg rhCD40L, followed by one round of intravenous immunization. Spleen cells were fused with myeloma cells to create hybridomas. Each hybridoma clone was tested for mAb production in an enzyme‐linked immunosorbent assay (ELISA). We initially obtained 30 different mouse anti‐human CD40L mAbs. The antigen‐binding domains of one mAb, mAb20, showing promising inhibitory capacity, were sequenced (service provided by Genscript). To obtain a recombinant Fab fragment, the variable heavy chain sequence was added to a murine constant heavy chain region 1 sequence, while the light chain variable region was combined with a murine constant kappa light chain region. Each construct was integrated into the pcDNA3.4 expression vector. In the heavy chain (HC), a C‐terminal Sortase‐tag followed by a His‐tag was included in the sequence. The Fab fragment was subsequently expressed in HEK293 Freestyle cells (HEK293F). HEK293F cells were incubated in FreeStyle 293 Expression medium (Gibco, 12338018). Fab was purified using Ni Sepharose Excel beads (Cytiva, 17371201) to capture the His‐tagged Fab. Following the wash of the beads, 300 mM imidazole in TBS was used to elute Fab from the beads.

From the sequencing data, one heavy chain sequence and two different light chain (LC) sequences (LC 1 and LC 2) were obtained. Two different Fab fragments were thus produced: the heavy chain in combination with either LC1 or LC2. Initial tests were made to determine which Fab (HC+LC1 or HC+LC2) had specificity toward CD40L (data not shown). The Fab with specificity against CD40L, anti‐CD40L Fab, is referred to as Fab20 (HC+LC1), while the nonbinding Fab (HC+LC2) was used in assays as an isotype control, since it holds the same HC sequence but another light chain. The nonbinding Fab was chosen as an ideal isotype control to control for heavy‐chain‐mediated effects, while at the same time, it also serves as a negative control for any implications of the expression system and purification procedure.

4.2. Binding of Anti‐Human CD40L Antibodies and Fab Fragment to CD40L

Assessment of the binding capacity of monoclonal mouse anti‐human CD40L antibodies to rhCD40L using a time‐resolved immunofluorometric assay (TRIFMA), the candidate anti‐CD40L mAbs were tested for binding to rhCD40L coated in microtiter wells, as described previously [45].

4.2.1. Flow Cytometry—Binding

The candidate anti‐CD40L mAbs were also tested for binding to CD40L expressed on activated T cells in a human peripheral blood mononuclear cell (PBMC) pool, also previously described [45].

Following the production of the anti‐CD40L Fab fragment, Fab20 was tested for its binding capacity to membrane‐bound CD40L. A Normal Human Dermal Fibroblast (NHDF) cell line expressing human CD40L (NHDF‐CD40L) was used. The NHDF cell line (CellSystems, Lot #03410) had been immortalized by transgenic insertion of human telomerase and subsequently transduced with a lentiviral vector inducing expression of the human CD40L gene, as described by Rovsing et al. [54]. NHDF‐CD40L was used to test the binding of the mouse anti‐human CD40L Fab fragment.

The Fab fragments were biotinylated using Biotin N‐hydroxysuccinimide ester (NHS) (Sigma‐Aldrich, H1759). Biotin coupling was performed by dialyzing Fab against phosphate‐buffered saline (PBS, pH 8.5) and then adding Biotin‐NHS (167 µg/mg protein) for incubation for 4 h at room temperature (RT). Biotinylated Fab was then dialyzed back into TBS over three rounds. NHDF‐CD40L cells were incubated with 1 % heat‐inactivated goat serum for 15 min to block Fc‐receptors. Next, cells were stained for flow cytometry in V‐bottomed microtiter plates using the biotinylated Fab20 at 10, 2, 0.4, and 0.08 µg/mL in flow staining buffer (PBS/2 % fetal bovine serum [FBS]). Cells were incubated for 30 min on ice, subsequently followed by the addition of 1 µg/mL Streptavidin‐Brilliant Violet 421 (BV421) (BD Biosciences, 563259) for detection of the biotinylated Fab and a Live/Dead Fixable Near‐IR (nIR) Dead Cell stain kit (ThermoScientific, L34976) diluted 1:2000. In the second incubation step, cells were left on ice and kept dark for 30 min. As a negative control, NHDFs not expressing CD40L were also stained for flow cytometry under the same conditions.

For data acquisition, the NovoCyte Quanteon 4025 flow cytometer equipped with four lasers (405, 488, 561, and 637 nm) and 25 fluorescence detectors (Agilent, San Clara, CA, USA) was used. Analysis was performed in FlowJo v10.10.0 software (BD Life Sciences). The gating strategy included live single cells.

4.2.2. Bio‐Layer Interferometry of Anti‐CD40L Fab Interaction With CD40L

Bio‐layer interferometry (BLI) using the Octet RED96 equipment (ForteBio) was conducted to examine the affinity of Fab20 to rhCD40L. Analysis of BLI data was performed in Prism (GraphPad Prism 10.0.0); however, since CD40L is a trimeric protein and the potential binding of Fab20 to CD40L could not be expected to be a 1:1 interaction in this setup, data were also analyzed using the MATLAB 2012a platform (Mathworks) using the fitting tool EVILFIT version 3 software created by Peter Schuck as previously described [55, 56, 57].

Black, flat‐bottomed 96‐well plates (Greiner) were used for binding kinetic experiments. Biotinylated Fab20 was diluted to 2.5 µg/mL in binding buffer (tris‐buffered saline [TBS], 0.1 mg/mL bovine serum albumin [BSA], 0.01 % Tween 20 [Tw]). In the plate, rhCD40L was prepared as a 2‐fold dilution series from 96 to 3 nM. The diluted biotinylated Fab20 was immobilized onto streptavidin‐coated sensors (OCTET Streptavidin [SA] Biosensors, Sartorius) up to 0.50 nm response. Before measuring binding, the baseline response was recorded. To assess the association, Fab20‐coated sensors were dipped into the rhCD40L dilutions for 400 s, and the sensor tips were subsequently submerged into wells only containing buffer to obtain the dissociation step (1,200 s). Last, sensors were subjected to three regeneration and cleaning cycles of 5 s, cycling between wells with a 500 mM phosphoric acid solution and wells with buffer only. During all steps, the plate was positioned on an orbital shaker (shaker speed of 700–1000 rpm).

The data were fitted using the EVILFIT fitting tool, using rhCD40L concentrations spanning from 0.15 to 5 µg/mL (3–96 nM). Input values for the association phase were fitted as t = association start + 1 s until t = association end + 1200 s. The limit of K _D values was set to an interval of 10⁻⁹ to 10⁻⁴ M, and k_off values in the interval from 10⁻⁷ to 10⁻¹ s⁻¹. The distribution P (k_a, K_A) was calculated using the discretization of the equation:

$$R_{\mathrm{total}}={\int }_{K_A\min }^{K_A\max }{\int }_{k_a\min }^{k_a\max }R\left(k_a,K_A,C_{\mathrm{analyte}}\right)P\left(k_a,K_A\right)d{k}_a\left.d{K}_A\right).$$

In a logarithmic grid of (k_a ,i, K_A ,i) values, with 15 and 18 grid points distributed on each axis. A global fit to association and dissociation curves for each analyte concentration was performed. Tikhonov regularization was used to determine the simplest distribution consistent with data (confidence level = 0.95) as described by Zhao et al. [58]. To estimate the contribution to the total signal from individual binding populations, the marked populations were integrated and normalized to the total signal from integrating the full grid of K _D and k _off values. The kinetics from individual populations were averaged from the signal in the marked regions (Figure 3C–E).

4.3. Structural Determination of Anti‐CD40L Fab Fragments’ Interaction With Recombinant Human CD40L

4.3.1. Cryo‐EM Sample Preparation

RhCD40L and Fab20 were mixed and briefly incubated at RT in buffer (1× PBS, 1 mM MgCl₂) in a 1:3 molar ratio with final concentrations of 2.47 and 7.42 µM, respectively. For imaging, we used ProtoChips Au‐FLAT 1.2/1.3 300 mesh grids, which were glow‐discharged for 45 s at 15 mA in a PELCO easiGlow immediately before sample deposition. 3 µL of the incubated sample was applied to the grid, while the sample application chamber was kept at 100% humidity and 4°C. The grid was then blotted with a manually calibrated stopping distance onto a double layer of Whatman no. 1 filter paper using a 4 s delay after application, 6 s of blot time, and 0 s of delay after blotting before plunging into liquid ethane at –184 °C. The blotting and plunge freezing were performed with Leica GP2.

4.3.2. Cryo‐EM Data Collection

The data were acquired at 300 keV on a Titan Krios G3i instrument (Thermo Fisher Scientific) equipped with a K3 camera (Gatan/Ametek) and energy filter operated in the energy‐filtered transmission electron microscopy mode using a slit width of 20 eV. The data were collected over a defocus range of −0.5 to −2.0 µm with a targeted dose of 60 e ⁻ Å⁻². Automated data collection was performed with EPU, and the data were saved as gain‐normalized compressed TIFF files (K3) or MRC files (K2) with pixel sizes of 0.645 Å px⁻¹. From this, 4307 movies were collected.

4.3.3. Cryo‐EM Single‐Particle Analysis and Data Processing

The data were processed using cryoSPARC v. 4.6.0 to apply motion correction, contrast transfer function (CTF) fitting, and initial particle picking [59]. The initial particles (1,059,752 particles [p]) were extracted and cropped to a pixel size of ∼5.4 Å px⁻¹, then sorted into 50 2D classes, of which 31 were used for ab initio reconstruction of five models. The best model was used for nonuniform refinement [60], applying C3 symmetry, followed by local refinement (without any symmetry applied), reaching the Nyquist resolution. This revealed a propeller‐like density map that resembled three Fab20s bound to CD40L. This density map was used to create fifty templates for template picking. The new template picked particles (1,500,369 p) were sorted by heterogeneous refinement into one propeller‐like class (using the density map from above) and four junk maps (generated from a previous ab‐initio reconstruction). The particles in the propeller class (401,792 p) were passed through a 3‐class ab‐initio reconstruction, and the best class (117,239 p) was now nonuniformly refined. Analyzing this by 50‐class 2D classification revealed that some nonoptimal particles remained in the particle stack, and by selecting 35 of the 2D classes and running a 3‐class ab‐initio job, provided a particle stack (48,153 p) of suitable particles, which was nonuniformly refined.

The selected particle stack of the template picked suitable particles was then used for Topaz training on a subset of micrographs (501 mics.) [61]. Particle picking of the Topaz trained system, extraction (456,44 p; box 300 px of 1.297 Å px⁻¹), and 2D classification (422,447 p) were followed by nonuniform, local, and heterogeneous refinement. The heterogeneous refinement was performed by feeding modified maps resembling the structure with 0, 1, 2, or 3 arms. After heterogeneous refinement, the particle stack resembling three Fab fragments simultaneously bound to CD40L (103,883 p) was used for final ab initio modelling and nonuniform refinement, without symmetry imposed and by applying C3 symmetry. A final local refinement was performed for the nonsymmetry‐imposed particles, yielding a map of the full structure. For the C3‐symmetry‐imposed particles, C3 symmetry expansion was performed, followed by local refinement (with C1 symmetry) using a focus mask covering only the CD40L and one Fab20. Together, this led to the formation of the two density maps presented here (termed Full [3.9 Å GSFSC] and Focus [3.4 Å GSFSC]). These two local refinement maps were sharpened using the deepEMhancer tool, which was used during model building, but not for validation [62]. The detailed parameters for the cryo‐EM density maps used here are noted in Table 2.

TABLE 2.

Cryo‐EM data collection, refinement, and validation statistics.

	#Full map	#Focus map (EMDB‐52634) (PDB 9I5N)
Data collection and processing
Magnification	130.000	130.000
Voltage (kV)	300	300
Electron exposure (e–/Å2)	60	60
Defocus range (µm)	−2.0 to −0.8	−2.0 to −0.8
Pixel size (Å)	1.294	1.294
Symmetry imposed	no	C3
Initial particle images (no.)	456,444 (topaz)	456,444 (topaz)
Final particle images (no.)	103,883	311,649 (sym. exp.)
Map resolution (Å) Fourier shell correlation threshold: 0.143	3.89	3.43
Map resolution range (Å)	3.279−56.756	3.006–31.329
Micrographs (curated)	4290	4290
Refinement
Initial model used		AlphaFold prediction (Fab)/1I9R (CD40L)
Model resolution (Å)
Fourier shell correlation threshold (0.143/0.5)	N/A	3.4/3.3
Model composition
Nonhydrogen atoms	N/A	5599
Protein residues	N/A	368
B factors (Å2)
Protein	N/A	62.85
R.m.s. deviations
Bond lengths (Å)	N/A	0.009 (0)
Bond angles (°)	N/A	1.496 (19)
Validation
MolProbity score	N/A	2.23
Clashscore	N/A	5.59
Poor rotamers (%)	N/A	2.87
Ramachandran plot
Favored (%)	N/A	92.54
Allowed (%)	N/A	6.91
Disallowed (%)	N/A	0.55

The original atomic model of rhCD40L and part of Fab20 was generated using AlphaFold 3 [63]. The model was created by inserting the sequence of rhCD40L and adding information about this being a trimer, followed by adding the sequence for the potential binding partner, Fab20. AlphaFold 3 predicted models of the potential interaction between the structures. Five models were predicted, with the first predicted model being the most likely. AlphaFold did not correctly predict the interaction, so the models were adapted to fit the cryo‐EM data. This was done in the ChimeraX software package [64] (version 1.7), also using ISOLDE [65] (version 1.7), followed by real‐space refinement with the PHENIX software [66] (version 1.21_5207). For the final refinement in PHENIX, we restricted the secondary structure, and the refinement and final map to model comparison (parameters can be found in Table 2) was performed by comprehensive validation (cryo‐EM) in the PHENIX software with the sharpened local refined focus map and two half maps (not the deepEMhancer map) [66]. Root mean square deviation (RMSD) was calculated in ChimeraX.

4.4. Inhibition of the CD40‐CD40L Interaction by Anti‐CD40L Antibodies and Fab Fragment

4.4.1. Inhibitory Capacity of Anti‐Human CD40L Antibodies and Fab Fragment

4.4.1.1. Time‐Resolved Immunofluorometric Assay

The ability of anti‐CD40L mAb to inhibit CD40:CD40L interaction was first tested in TRIFMA by coating NUNC Maxisorp 96‐well plates with 100 µL of 1 µg/mL of rhCD40L in PBS overnight (O/N) at room temperature (RT). Residual binding sites were blocked with 1 mg/mL of human serum albumin (HSA) for 1 h. Resembling CD40, a construct of rhCD40 (extracellular domain, res: 21‐121 [67]) added onto an Fc domain, was used for testing inhibition of CD40‐CD40L interaction. A mix of 1 µg/mL of rhCD40‐Fc and 50 ng/mL of anti‐CD40L mAbs in TBS/Tw was added to each well. Samples were incubated for 2 h at RT and subsequently washed three times in TBS/Tw before adding 0.2 µg/mL of biotinylated rabbit anti‐human IgG (DAKO, A0424) to each well for detection of bound rhCD40‐Fc. After 2 h of incubation and another wash, 100 µL Streptavidin‐Europium (Perkin Elmer, 1244‐360) diluted 1:1000 in TBS/Tw/0.25 µM ethylenediaminetetraacetic acid (EDTA) was added and incubated 1 h. Next, 125 µL Enhancement buffer (AMPLIQON, AMPQ99800.1000) was added, after another plate wash. Plates were read by time‐resolved fluorometry using the VICTOR5 Plate Reader. The amount of europium detected in the wells is given as counts per second.

4.4.1.2. Flow Cytometry—Inhibition

Anti‐CD40L mAb inhibition of CD40:CD40L interaction was also tested on stimulated PBMCs. PBMCs were stimulated with PMA and Ionomycin, as previously described [45]. Human CD40‐Fc was biotinylated with NHS‐biotin (biotin‐CD40‐Fc) as described above and mixed with Streptavidin‐BV421. Cells were stained in 100 µL with the addition of 0.5 µg/mL of the Streptavidin‐BV421 human biotin‐CD40‐Fc mix, either alone or in combination with 0.4 µg/mL of mAbs, that is, the mAb20 we describe in this report, as well as other anti‐CD40L mAbs we described previously [45]. The BV421 signal was considered equivalent to CD40‐Fc binding to CD40L, while a decrease in signal was interpreted as mAb inhibition. A negative control in the form of only adding Streptavidin‐BV421 to the cells was included, while the positive control of full CD40‐Fc signal was considered as the sample with the addition of only BV421‐labeled CD40‐Fc. Furthermore, a live/dead nIR marker was also included.

From the results of testing the inhibitory capacity of all 30 mouse anti‐human CD40L mAbs, the mAb with superior inhibitory capacity was chosen and redesigned as a Fab fragment. This Fab fragment, Fab20, was, as mentioned in Section 4.2.1, tested for its ability to bind membrane‐bound CD40L. Similarly, using NHDF and NHDF‐CD40L cells, the inhibitory capacity of Fab20 was tested. NHDF and NHDF‐CD40L were stained in 100 µL with the addition of 2 µg/mL of Streptavidin‐BV421‐labeled biotin‐CD40‐Fc complex, either alone or with the addition of a threefold dilution series of Fab20 (7–0.1 µg/mL). Dead cells were labeled using a Live/Dead nIR marker. As a negative control, cells were stained with Streptavidin‐BV421 only. NHDF cells (‐ CD40L) were also stained as an additional negative control.

Samples were run as described in Section 4.2.1. The gating strategy included live single cells (Figure S2).

4.4.2. Inhibition of B‐Cell Proliferation and Differentiation in a Human Coculture System by the Anti‐CD40L Fab Fragment

To examine the effect of Fab20 on CD40L‐mediated stimulation of B cells, a naïve human B cell/NHDF‐CD40L coculture system was set up as described in Rovsing et al. [54]. The rationale of the coculture system for B‐cell activation and differentiation modified into a human system was based on the murine coculture system described in Nojima et al. [68] and Kuraoka et al. [69].

4.4.2.1. Purification of Naïve Human B Cells

Naïve human B cells were purified by negative selection from PBMCs using a human Naïve B cell isolation kit II (Miltenyi, 130‐091‐150) following the instructions given by the manufacturer. PBMCs from healthy blood donors had, before B‐cell purification, been purified from a buffy coat by density‐gradient centrifugation using Ficoll‐Paque PLUS (Cytiva, 17144002). PBMCs were incubated with human Fc block (clone: Fc1, BD Biosciences, 564220) for 5 min on ice before adding the biotinylated antibody cocktail from the B‐cell isolation kit for labeling all non‐naïve B cells. Cells were incubated for 30 min on ice, centrifuged (200g, 10 min, 4°C), and resuspended in buffer (PBS, 2% FBS, 2 mM EDTA). Anti‐biotin magnetic beads were then added to the cells for a 20 min incubation step before adding the cocktail to an LS column (Miltenyi, 130‐042‐401). The flow‐through from the column was collected, as this consisted of the naïve B cells of interest. Cells were cryopreserved in freezing medium (90% FBS, 10% dimethyl sulfoxide [DMSO]) until use.

4.4.2.2. Assessment of Anti‐CD40L Fab Fragment Inhibition in NHDF/Naïve Human B‐Cell Coculture

A coculture system of NHDF‐CD40L and naïve human B cells was used to test the ability of the Fab20 to inhibit B‐cell proliferation and differentiation. The setup was as described in Rovsing et al. [54]. The NHDFs were selected for this culture system based on their slow growth, eliminating the need for irradiation. In brief, one day before initiating the coculture, NHDF‐CD40L cells were seeded into a 12‐well plate, 35,000 cells per well, and incubated O/N at 37°C, 5% CO₂. The following day, the culture medium was removed, and 15,000 naïve human B cells were seeded on top of the NHDF‐CD40L cells along with 10 ng/mL rh Interleukin‐4 (IL‐4) (PeproTech, 200‐04‐20UG), rhIL‐21 (PeproTech, 200‐21‐10UG), and rhBAFF (PeproTech, AF‐310‐13‐20UG), and 50 ng/mL of rhIL‐2 (PeproTech, 200‐02‐100UG) for proper B‐cell activation. The cells were either left untreated or with the addition of 3.5 µg/mL of Fab20. A Fab fragment (HC‐LC2) holding the same heavy chain, but a different light chain, was included as an isotype control (Fab isotype), as described above.

On day 4, cells from 1 well of each treatment were harvested, counted, and cryo‐preserved for flow cytometry. Cells were counted using the Cellometer K2 cell counter (Nexcelom Bioscience) following the manufacturer's instructions. The supernatants from the wells were harvested and kept at −20°C for immunoassays. In the remaining coculture wells, half of the medium was carefully removed and replaced with fresh medium containing IL‐21, BAFF, and IL‐2, but no longer supplemented with IL‐4. Fab20 or Fab isotype was also added to their respective wells.

On day 7, another well from each treatment was harvested and preserved as on day 4. The remaining B cells in the other wells were also harvested, counted, and reseeded onto new NHDF‐CD40L cells, which had been seeded the day before. This was done to avoid NHDF‐CD40L cells becoming too confluent, since this would affect the CD40L levels in the wells. The number of B cells was restored to the starting point of 15,000 cells (this was impossible for Fab20 treatment since too few cells survived until day 7).

On days 10 and 14, coculture wells were handled as on day 4. Representative phase contrast light microscopy images (10$\times$ objective and 10$\times$ eyepiece) were obtained for each treatment on all harvest days.

4.4.2.3. Effect of Anti‐CD40L Fab on Human B‐Cell Proliferation

Human B cells from each treatment (untreated control, Fab20, or Fab isotype control) were counted on days 4, 7, 10, and 14. Since B cells were reseeded and reduced to starting numbers of 15,000 on day 7, the counts from days 10 and 14 were influenced by this, so for these counts, the B‐cell division factor from day 7 was multiplied by the day 7 cell count numbers, given as:

$$\mathrm{Day}7\mathrm{cell}\mathrm{count}\times \left(\frac{\mathrm{Day}10\mathrm{or}14\mathrm{count}}{15,000}\right).$$

4.4.2.4. Effect of Anti‐CD40L Fab on Human B‐Cell Differentiation

On the same days, human B cells were counted; cells from one well from each treatment were cryopreserved and saved for flow cytometric assessment of human B‐cell surface markers. B cells from all timepoints and a sample of naïve B cells kept from day 0 for baseline levels were thawed quickly and resuspended in staining buffer (PBS/2% FBS/1 mM EDTA) in a 96‐well V‐bottomed plate. Before staining cells, Fc receptors were blocked using Fc‐block for 10 min on ice.

Next, the cells were stained for the following markers in a 50/50 mix of staining buffer and BD Horizon Brilliant stain buffer (BD Biosciences, 566349): anti‐CD19 FITC (BD, 555412), anti‐CD20 BV421 (BD, 562873), anti‐CD27 BV786 (BD, 563327), anti‐CD38 APC (BD, 555462), anti‐CD95 PE (BD, 561976), anti‐IgD PE‐Cy7 (BD, 561314), all diluted 1:250, and live/dead eFluor‐780 (Invitrogen, 65‐0865‐14) diluted 1:2000. The cells were stained for 30 min in the dark on ice. Cells were washed twice by spinning (200g, 4°C, 5 min) and resuspended in staining buffer.

Samples were analyzed as described in Section 4.2.1, and the gating strategy for this experiment consisted of excluding debris, followed by including single cells (on forward and side scatter [FSC + SSC]). Next, dead cells were excluded, and the presence of B cells was confirmed by gating on CD19/CD20 double‐positive cells. B‐cell differentiation was assessed by examining cell surface levels of IgD, CD27, CD38, and CD95 (Figure S5) [47].

4.4.2.5. Effect of Anti‐CD40L Fab on B‐Cell Immunoglobulin (IgG) Secretion

Supernatants from cocultures were examined for levels of IgG produced by human B cells under the different treatments. Levels of total IgG (IgG1, IgG2, IgG3, and IgG4) were used as a measure of antibody‐secreting B cells, which were caused by their differentiation into cells resembling plasma blasts and plasma cells.

The levels of IgG were analyzed by TRIFMA. For this, wells of microtiter plates (FluoroNunc) were coated O/N (4°C) with 0.3 µg/mL of donkey anti‐human IgG (Jackson ImmunoResearch, 709‐005‐098) in PBS. Residual binding sites were blocked with 1% BSA in TBS for 1 h. The wells were then washed in TBS/Tw, and supernatants diluted 1:50 or 1:500 in TBS/Tw/0.1% BSA were added to the wells in doublets and incubated 1 h at RT. A seven‐point threefold standard dilution curve of normal human IgG (Octapharma, 478393) starting at 100 ng/mL was included, as well as a blank (TBS/Tw/0.1% BSA) and three quality controls with known concentrations of human IgG. After incubation and another plate wash, 0.1 µg/mL of biotinylated goat anti‐human IgG (H+L) (Southernbiotech, 2016‐08), diluted in TBS/Tw/0.1 % BSA, was added to the wells for an additional 1 h incubation step. After washing the wells, 100 µL Streptavidin‐Eu³⁺ was added to the wells for 1 h incubation, and after a final plate wash, 125 µL of Enhancement solution was added before the plate was read as described in the TRIFMA section.

4.5. Statistical Analysis

GraphPad Prism 10.0.2 (GraphPad Software) was used for statistical data analysis.

Normality tests were used to determine if the data followed a Gaussian distribution. Statistical analysis consisted of parametric tests (paired/unpaired t‐tests) if a Gaussian distribution was observed. When applicable, if data did not meet Gaussian distribution criteria even after log transformation, nonparametric tests were applied (Wilcoxon signed rank test [paired] or Mann–Whitney test [unpaired]). Additionally, Tukey's multiple comparisons test was used to compare multiple means between groups (nonparametric). If applied, graph whiskers represent standard deviation (SD). p‐values ≤ 0.05 were considered statistically significant (α = 0.05).

For IC50 values, concentrations were log‐transformed, and nonlinear regression analysis was performed.

Author Contributions—CRediT

K.P. performed all experiments and data acquisition regarding TRIFMA, protein production, flow cytometry, BLI, T‐cell stimulation, and B‐cell‐NHDF cocultures. K.P. also performed all data analysis and statistics regarding the above‐mentioned methods and drafted the original manuscript. K.G. produced NHDFs, developed the B‐cell‐NHDF coculture system, and assisted with B‐cell flow cytometric analysis. E.L.K. led all studies regarding cryo‐EM. E.L.K., T.S., and E.S.A. performed cryo‐EM data acquisition and analysis as well as model fitting, validation, and analysis. J.V. and L.C. assisted with cryo‐EM and BLI. A.G.H. purified monoclonal antibodies and Fab fragments. Y.P. developed the monoclonal antibody through mouse immunization and hybridoma technology. A.B.R. produced the NHDFs and developed the B‐cell‐NHDF coculture system. K.P., N.S.L., A.T., S.E.D., and S.T. contributed to the conceptualization and design of the study. All authors contributed to reading and reviewing the final manuscript.

Funding

The work was funded by the Danish National Research Foundation (Grant DNRF135).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1: eji70158‐sup‐0001‐VideoS1.mp4.

Supporting File 2: eji70158‐sup‐0002‐SuppMat.docx.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.