Research ArticleAutoimmunity

A CD40L-targeting protein reduces autoantibodies and improves disease activity in patients with autoimmunity

See allHide authors and affiliations

Science Translational Medicine  24 Apr 2019:
Vol. 11, Issue 489, eaar6584
DOI: 10.1126/scitranslmed.aar6584

Curbing CD40 signaling

The CD40 axis is a major collaborative mechanism of B and T cell responses. Previous attempts to disrupt this pathway to treat autoimmune disease led to adverse thrombotic events due to engagement of Fc receptors and platelet expression of CD40 ligand (CD40L). To avoid this issue, Karnell et al. designed a nonantibody scaffold protein, VIB4920, which blocks human CD40L. VIB4920 inhibits B cell activation but does not induce platelet aggregation in vitro. VIB4920 administration resulted in blunted responses to immunization in healthy people, and people with rheumatoid arthritis experienced reduced disease activity. No thrombotic side effects were encountered in either clinical trial. This next-generation therapeutic has the potential to be widely used to treat various autoimmune diseases.

Abstract

The CD40/CD40L axis plays a central role in the generation of humoral immune responses and is an attractive target for treating autoimmune diseases in the clinic. Here, we report the generation and clinical results of a CD40L binding protein, VIB4920, which lacks an Fc domain, therefore avoiding platelet-related safety issues observed with earlier monoclonal antibody therapeutics that targeted CD40L. VIB4920 blocked downstream CD40 signaling events, resulting in inhibition of human B cell activation and plasma cell differentiation, and did not induce platelet aggregation in preclinical studies. In a phase 1 study in healthy volunteers, VIB4920 suppressed antigen-specific IgG in a dose-dependent fashion after priming and boosting with the T-dependent antigen, KLH. Furthermore, VIB4920 significantly reduced circulating Ki67+ dividing B cells, class-switched memory B cells, and a plasma cell gene signature after immunization. In a phase 1b proof-of-concept study in patients with rheumatoid arthritis, VIB4920 significantly decreased disease activity, achieving low disease activity or clinical remission in more than 50% of patients in the two higher-dose groups. Dose-dependent decreases in rheumatoid factor autoantibodies and Vectra DA biomarker score provide additional evidence that VIB4920 effectively blocked the CD40/CD40L pathway. VIB4920 demonstrated a good overall safety profile in both clinical studies. Together, these data demonstrate the potential of VIB4920 to significantly affect autoimmune disease and humoral immune activation and to support further evaluation of this molecule in inflammatory conditions.

INTRODUCTION

The CD40/CD40L pathway plays a critical role in driving humoral immune responses and has been implicated in the pathogenesis of several autoimmune diseases. CD40 is constitutively expressed on a variety of antigen-presenting cells—including dendritic cells, macrophages, and B cells (1)—but can also be expressed on nonhematopoietic cells. Expression of the ligand, CD40L (also known CD154), is highly regulated and predominately expressed on activated CD4+ T cells (2). CD40/CD40L interactions between B cells and activated T cells are essential for mounting effective humoral responses to T-dependent antigens (35). The CD40/CD40L axis drives B cell expansion, differentiation, and isotype switching in vitro (69). In vivo, CD40 signaling is required for germinal center (GC) formation, somatic hypermutation, and the generation of memory B cells and long-lived plasma cells (1013). CD40 or CD40L defects in humans lead to X-linked hyperimmunoglobulin syndrome, a disease characterized by impaired isotype class switching that manifests as high levels of serum immunoglobulin M (IgM) with low to no detectable IgG, IgA, or IgE and increased susceptibility to infections (1416).

Clinical trials with compounds directed against CD40L have demonstrated the potential benefits of targeting the CD40 pathway in autoimmune diseases such as lupus nephritis (17) and immune thrombocytopenic purpura (18). However, these programs were halted because of adverse thromboembolic events related to platelet aggregation caused by cross-linking of anti-CD40L antibodies cobound to platelet CD40L and FcγRIIa on adjacent platelets. Mouse models support a role for FcγRIIa in anti-CD40L–induced thrombocytopenia. In mice transgenic for human FcγRIIa, anti-CD40L monoclonal antibody (mAb) induced shock and thrombocytopenia (19). This effect was not observed in either wild-type mice or transgenic mice injected with an aglycosylated version of the antibody, unable to engage FcγR.

To avoid complications associated with mAb-based targeting of CD40L, here, we have generated a CD40L-specific nonantibody scaffold using an engineered Tn3 protein (18, 20). We demonstrate that VIB4920, a bivalent Tn3 molecule fused to human serum albumin (HSA), binds CD40L and prevents interaction with CD40 receptor. VIB4920 was evaluated preclinically and in phase 1 studies with healthy volunteers and patients with rheumatoid arthritis (RA).

RESULTS

Development and characterization of VIB4920

Isolation and optimization of CD40L-specific Tn3 proteins. Tn3 is a small protein scaffold, about 90 amino acids in length, that has immunoglobulin-like folds, including loops structurally analogous to antibody complementarity-determining regions (CDRs), that can be randomized to select for specific binding properties (20). CD40L-specific Tn3 clones were screened for their ability to biochemically inhibit binding of human CD40L to CD40. A panel of clones inhibited binding of CD40L to CD40, with seven of them exhibiting median inhibitory concentration (IC50) values below 1 μM (Fig. 1A). The two most potent inhibitors (Tn3 proteins 309 and 311) were further evaluated in a cell-based reporter assay and dose-dependently inhibited CD40L-induced nuclear factor κB (NF-κB) reporter gene expression (Fig. 1B), highlighting the ability of these proteins to functionally inhibit CD40/CD40L signaling.

Fig. 1 Generation and biochemical characterization of VIB4920, a human Tn3 specific for CD40L.

(A) Inhibition of CD40/CD40L interactions was measured by Proteon. The average of duplicate wells is shown. (B) Human embryonic kidney (HEK) 293 cells expressing CD40 and an NF-κB reporter were stimulated with recombinant CD40L overnight in the presence of anti-CD40L Tn3 proteins. The percent inhibition of luciferase activity is shown. Data represent the mean of duplicate wells. One of two independent studies is shown. (C) Human peripheral blood mononuclear cells (PBMCs) were stimulated with recombinant CD40L and expression of CD86 on CD19+ cells was assessed by flow cytometry. (D) Inhibition of CD40/CD40L interactions by enzyme-linked immunosorbent assay (ELISA). Data represent the mean of duplicate wells. (E) Binding of monovalent Tn3 protein 342 to tumor necrosis factor (TNF) superfamily members was evaluated by ELISA. (F) Proposed structure of VIB4920 based on crystallization of 342 Tn3 [Protein Data Bank (PDB) ID 6BRB] and published crystal structure of HSA (PDB ID 1AO6). The linker is shown in blue. (G) Cartoon presentation of CD40/CD40L and 342/CD40L structures aligned through a common CD40L molecule. CD40L is shown in green, Tn3 is shown in magenta, and CD40 receptor is shown in cyan. OD, optical density.

Simultaneous binding to multiple targets, as occurs in the case of bivalent antibodies, can result in markedly increased avidity. To explore the impact of bivalency on the potency of CD40L-specific Tn3 proteins, we linked two copies of identical Tn3 modules (309-309) via a flexible Gly4Ser-containing spacer to form a tandem bivalent fusion protein. Compared to the monovalent Tn3, there was a nearly 1000-fold improvement in the potency of the bivalent construct in inhibiting up-regulation of the activation marker CD86 on human CD19+ B cells stimulated through CD40 (Fig. 1C). In addition, clone 309 was also affinity matured through random mutagenesis in the variable CDR-like loop regions, generating clone 342, with further improved binding affinity for CD40L (309, 190 nM; 342, 1.4 nM) and about 300-fold improvement in potency of inhibition of CD40/CD40L interactions (Fig. 1D). Clone 342 was evaluated for specificity and found to selectively bind to CD40L and not to other members of the tumor necrosis factor superfamily (Fig. 1E).

Standard Tn3 molecules would be expected to exhibit very rapid clearance from circulation when administered systemically because of their small size. To improve the pharmacokinetic (PK) properties of the protein, the CD40L-specific Tn3 tandem was fused to serum albumin (21, 22). Fusion of a bivalent mouse surrogate CD40L-specific Tn3 protein, M13-M13, to mouse serum albumin (MSA) resulted in a 65-fold increase in serum half-life and 345-fold decrease in clearance compared to the original Tn3 molecule (table S1). On the basis of these observations, we generated VIB4920, composed of tandem 342 CD40L-specific Tn3 proteins for optimal potency, fused to HSA for improved half-life (Fig. 1F).

To better understand the molecular nature of the interaction between CD40L and VIB4920, we performed crystallography studies. CD40L-specific Tn3 (342) and soluble CD40L (sCD40L) proteins were expressed, purified, and cocrystallized, and the structure was determined at 2.8 Å resolution (fig. S1A). The interface with CD40L is composed of amino acids mostly from the second modified loop of the Tn3 (fig. S1B). Initial characterization of the molecule indicated that the 342 Tn3 was able to block the interaction between CD40L and CD40. Superimposing the structure of the CD40/CD40L complex with that of 342/CD40L showed that 342 and CD40 share a common binding site on CD40L (Fig. 1G). Hence, VIB4920 competes with CD40 and prevents its association with CD40L.

Preclinical characterization of VIB4920

VIB4920 blocks activation and differentiation of human B cells and does not induce platelet aggregation in vitro. CD40 signaling has been extensively characterized and involves the activation of a variety of different pathways and transcription factors, including NF-κB (23), which can promote B cell activation, proliferation, and differentiation (24). Stimulation of an NF-κB luciferase reporter cell line that expresses human CD40 with recombinant human CD40L or CD40L-expressing cells induced NF-κB activation, which was dose-dependently blocked by VIB4920 (IC50, 0.899 nM; Fig. 2A).

Fig. 2 VIB4920 inhibits CD40 signaling and activation of human B cells and does not induce platelet aggregation in ex vivo studies.

(A) CD40+ HEK293 cells were stimulated with CD40L, and NF-κB luciferase activity was assessed. Shown is the percent inhibition of the luciferase signal. Data represent the mean of duplicate wells from one of two independent studies. (B) Human PBMCs were stimulated with recombinant MEGACD40L, and the percentage of CD19+/CD86+ cells was measured by flow cytometry. Data represent the mean of duplicate wells from one of three independent studies. (C) Human B cells were stimulated with interleukin-21 (IL-21) and MEGACD40L. The dotted line represents adenosine 5′-triphosphate (ATP) concentrations from unstimulated cells. Data shown are the means and SD of triplicate wells and are representative of two independent experiments. (D and E) Human B cells were left unstimulated (nil) or were stimulated with IL-21, anti-IgM, and MEGACD40L, and plasma cell number was quantified. 5c8 is an anti-CD40L mAb used as a positive control for inhibition of CD40L in this assay. (D) The percentage of IgD CD38hi plasma cells with the indicated molecules at 1 nM. (E) Plasma cell (PC) number at various concentrations as indicated. Data shown are the means and SD of triplicate wells and are representative of two independent experiments. (F and G) Washed human platelets were incubated with preformed immune complexes of anti-CD40L and MEGACD40L, and aggregation was measured. Percent aggregation is shown. (F) Where indicated, platelets were preincubated with anti-CD32a antibody (IV.3) before the addition of immune complex. Adenosine diphosphate (ADP) is a positive control for platelet aggregation. Data are representative of two independent experiments. ***P < 0.001, ****P < 0.0001, by two-tailed unpaired Student’s t test.

CD86 is rapidly up-regulated on B cells after activation, including activation through CD40 (25). VIB4920 fully prevented CD40L-mediated up-regulation of CD86 by primary human B cells (Fig. 2B). To evaluate the ability of VIB4920 to inhibit B cell proliferative responses, we stimulated human B cells for 3 days with recombinant CD40L and IL-21. VIB4920 completely blocked CD40L-dependent expansion of primary human B cells under these conditions (Fig. 2C). After cell division, B cells receiving CD40 signals in combination with IL-21 and B cell receptor stimulation differentiate into plasma cells (26). More specifically, after 7 days of culture with recombinant CD40L, IL-21, and anti-IgM, roughly 50% of cells detected in culture were characterized as IgD CD38hi plasma cells (Fig. 2D). VIB4920 blocked plasma cell differentiation in vitro with complete inhibition of plasma cell generation at the highest concentration tested (Fig. 2E). The CD40L-targeted Tn3 molecule was comparable to an antibody-based inhibitor (5c8) at suppressing primary human B cell activation, proliferation, and plasma cell differentiation (Fig. 2).

The central role of CD40L in promoting T cell–dependent immune responses has been well characterized (9, 27). Therefore, a T cell–dependent immunization model was used to evaluate the ability of the Tn3-MSA fusion protein to block humoral immune responses in vivo. Because of insufficient sequence homology between human and murine CD40L, a CD40L-specific mouse surrogate Tn3, M31, was used for these studies. Mice were inoculated with sheep red blood cells (SRBCs) and then treated daily with anti-CD40L Tn3 protein on days 9 to 13 after immunization. The immune response in treated animals was assessed on day 14 by quantitating splenic and lymph node GC B cells by flow cytometry. As expected, immunization with SRBCs in control-treated mice led to an expansion of GC B cell frequency (fig. S2A). A dose-dependent reduction of GC B cell frequency was observed in mice treated with the CD40L-specific Tn3-MSA fusion protein (fig. S2A). At a dose of 30 mg/kg, the CD40L-specific Tn3-MSA fusion protein induced complete suppression of GC formation, as assessed by the near absence of GC B cells in the spleen and lymph nodes, equivalent to control nonimmunized mice. In addition, anti-SRBC IgG levels mirrored that of the GC B cell response, with reductions in SRBC-specific immunoglobulin titers at higher doses of anti-CD40L Tn3 (fig. S2B).

Anti-CD40L–directed mAbs have failed in clinical trials because of safety concerns, largely due to thromboembolic complications related to cross-linking CD40L on the cell surface of platelets. To confirm that VIB4920, which lacks an Fc domain, does not induce platelet aggregation, we evaluated its impact on washed human platelets in vitro. As previously described, when precomplexed with sCD40L, anti-CD40L mAb (human IgG1) showed a marked ability to induce platelet aggregation (Fig. 2, F and G). The response was rapid, with mAb-sCD40L immune complexes inducing 80% of the platelets to aggregate within 8 min. Preincubation of platelets with an antibody that blocks FcγRIIa (mAb IV.3) prevented mAb-immune complex-mediated aggregation, consistent with an essential role for Fc receptors in this response (Fig. 2F). By contrast, at several concentrations tested, VIB4920 showed no propensity to induce platelet aggregation in this assay (Fig. 2G). These data suggest that the absence of an Fc region in Tn3 constructs could reduce the risk of platelet aggregation and thromboembolic events that have been observed with therapeutic anti-CD40L antibodies in the clinic.

Early clinical experience

VIB4920 was evaluated in a single ascending dose (SAD) study in healthy volunteers (fig. S3A) and in a multiple ascending dose (MAD) study in patients with moderate to severe active RA (fig. S4). The main demographic and clinical characteristics of the study population for the MAD study at baseline are shown in Table 1.

Table 1 RA cohort demographics and clinical characteristics.

DAS28-CRP, disease activity score 28-joint count–C-reactive protein; CDAI, clinical disease activity index; ESR, erythrocyte sedimentation rate; RF, rheumatoid factor; ACPA, anti–citrullinated peptide antibodies; DMARDs, disease-modifying antirheumatic drugs.

View this table:

PK and PD after intravenous administration of VIB4920. The PK profile of VIB4920 in both healthy volunteers and patients with RA was linear with increasing exposure in a dose-proportional manner (Fig. 3, A and B). After a single intravenous dose of 3 to 3000 mg, the mean terminal half-life of the molecule was 8 days with a half-life up to 10.1 ± 1.87 days at the highest dose. In the MAD study in subjects with RA, VIB4920 demonstrated a similar PK profile with a mean half-life of 9.27 ± 1.47 days.

Fig. 3 VIB4920 demonstrates a favorable PK profile in healthy volunteers and patients with RA.

(A and B) Circulating concentrations of VIB4920 were determined by ELISA at the indicated time points in (A) the SAD study of healthy volunteers and (B) the MAD study in RA. The dotted line represents the lower limit of sensitivity for the assay. Error bars represent SD of the mean, which was not calculated for groups with n = 2 subjects. (C) sCD40L concentrations were assessed in the SAD study of healthy volunteers at the indicated time points by ELISA. The dotted line represents the lower limit of detection for this assay. LLOQ, lower limit of quantification.

sCD40L is an 18-kDa trimer that is detected at low levels in healthy donors and increased in the circulation of patients with autoimmune disease (28, 29). Measurement of sCD40L levels after VIB4920 administration represents a potential measure of target engagement because sCD40L bound to VIB4920 could be retained and accumulate in circulation. As expected, there was a dose-dependent increase in total sCD40L in the plasma after a single administration of VIB4920 (Fig. 3C), suggesting target engagement. The time to reach the maximum total sCD40L in the plasma increased from 11.5 to 84 days as the dose increased from 3 to 3000 mg, indicating that target engagement may be maintained for a longer duration in the highest-dose group.

Safety: VIB4920 is well tolerated in healthy volunteers and patients with RA. Overall, VIB4920 was well tolerated with a balanced distribution of treatment emergent adverse events (TEAEs) observed between placebo and the active dose groups in both clinical studies (Table 2 and tables S2 and S3). There were no infusion-related reactions, severe infections, or deaths. The only treatment emergent serious adverse event reported in the SAD study was a tibia fracture in the placebo group. The most common TEAEs reported in the MAD study in RA were diarrhea, hyperhidrosis, upper respiratory tract infection, and urinary tract infection, each occurring in three patients (7.1%). One grade 4 serious adverse event of encephalitis was reported in the 1500-mg VIB4920 dose group, occurring after six doses of study drug. No etiological infectious agent was identified, and the event was considered unrelated to VIB4920. Several months after discontinuing VIB4920, similar symptoms recurred, and the patient was subsequently diagnosed with metastatic melanoma of the brain (table S3). No clinically relevant coagulation or platelet function abnormalities were observed after treatment with VIB4920 in either human study.

Table 2 VIB4920 demonstrates an acceptable safety profile in patients with RA.

View this table:

VIB4920 inhibits T cell–dependent antibody response in humans. To evaluate the ability of VIB4920 to influence humoral immune responses in healthy participants, a T cell–dependent antibody response was induced by immunization with keyhole limpet hemocyanin (KLH). Subjects received two subcutaneous KLH immunizations: The first immunization was administered 14 days before dosing with either VIB4920 or placebo (visit day −14), and the second was given 15 days after dosing (visit day 15) (fig. S3B). In placebo-treated subjects, 1 week after the secondary immunization with KLH (visit day 22), an increase in anti-KLH IgG titers was detected, with peak levels of IgG observed on visit day 29 (Fig. 4A). Consistent with previous reports, the secondary anti-KLH response was dominated by IgG, with a modest increase in KLH-specific IgM detected after boosting (Fig. 4B) (30). Anti-KLH titers were not affected by VIB4920 at the low doses. However, starting with the 300-mg dose, the secondary response to KLH was reduced in a dose-dependent manner. We observed a 78 and 86% reduction in anti-KLH IgG compared to placebo at day 43 in the 1000- and 3000-mg cohorts, respectively (Fig. 4, A and C). In the highest-dose group, seven of eight subjects had undetectable titers of anti-KLH IgG at day 43, suggesting near-complete suppression of the humoral immune response by VIB4920.

Fig. 4 VIB4920 inhibits B cell proliferation and T cell–dependent antibody response in healthy human participants in a dose-dependent manner.

Healthy volunteers were immunized with KLH 14 days before treatment with placebo or VIB4920 and were reimmunized 15 days after dosing. Three- and 10-mg dose groups only contained n = 2 subjects and were therefore not included in the anti-KLH titer plots. (A) Anti-KLH IgG and (B) anti-KLH IgM titers were assessed by ELISA at various time points. (C) Dose-response model for inhibition of anti-KLH IgG at day 43. (D and E) The frequency of (D) proliferating B cells (Ki67+ CD19+) or (E) class-switched memory B cells (Ki67+ CD19+ CD27+ IgD) in circulation was quantified by flow cytometry at various time points as indicated in volunteers receiving either placebo or the highest dose of VIB4920. (F) Plasma cell signature score in whole blood is shown, as evaluated by TaqMan polymerase chain reaction. Mean and SE expression values for placebo and high-dose VIB4920 groups are shown. *P < 0.05 and **P < 0.01 versus placebo, by Mann-Whitney U test.

In placebo-treated subjects, secondary immunization with KLH induced B cell proliferation, indicated by an increase in the frequency of Ki67+ CD19+ B cells in circulation on visit day 22, 7 days after rechallenge (Fig. 4D). Before boosting, the baseline frequency of proliferating B cells was reduced compared to the placebo cohort in subjects that received high-dose VIB4920, consistent with the mechanism of action of the molecule. In cohorts receiving high-dose VIB4920, the B cell proliferative response 7 days after reimmunization was significantly impaired, demonstrated by the lack of increase in Ki67+ B cells in the 3000-mg cohort, where proliferating B cells instead fell by as much as 50% below baseline (Fig. 4D). Further phenotyping revealed that the greatest impact of VIB4920 on proliferating B cells was within the IgD CD27+ isotype-switched memory population (Fig. 4E). These data are in line with the T cell–dependent antibody response results, which demonstrate suppression of IgG production in response to secondary challenge.

Immunization also induced a marked increase in a previously validated plasma cell gene signature score (31) detected in whole blood of subjects within the placebo cohort 1 week after rechallenge with KLH, which returned to baseline by day 29 (Fig. 4F). In the highest-dose cohort (3000 mg), there was a significantly reduced plasma cell gene signature score in peripheral blood compared to placebo before boosting with KLH and no increase in plasma cell gene signature score after reimmunization (Fig. 4F). These data highlight the mechanism of action of VIB4920 and demonstrate its potent ability to suppress B cell activation and plasma cell differentiation.

High-dose VIB4920 reduces the frequency of anti-drug antibodies. Anti-drug antibodies (ADAs) were detected in most of the patients receiving low doses of VIB4920 in both the single-dose (Fig. 5A) and the multiple-dose (Fig. 5B) studies. In contrast, the frequency of ADAs was substantially reduced at higher-dose levels of VIB4920. In the single-dose study, although 18 of 20 subjects had detectable ADA titers in the 3- to 100-mg dose cohorts, ADAs were only detected in one of eight subjects in the 3000-mg dose group (Fig. 5A). In the multiple-dose study, more than 30% of patients with RA treated with 75 and 500 mg of VIB4920 developed ADAs (Fig. 5B). Two of the three subjects with ADA in the 75-mg dose group developed ADA during the treatment phase, whereas in the 500-mg dose group, ADAs became detectable in all three subjects after treatment (Fig. 5C). ADAs were only detected in 1 of 12 subjects (8.3%) in the 1000-mg dose group and were not detected in any subject receiving 1500 mg of VIB4920 (Fig. 5B). The observation that VIB4920 effectively suppressed the ADA response at higher doses supports the immunomodulatory capacity of the molecule.

Fig. 5 VIB4920 demonstrates dose-dependent reductions in ADA.

(A to C) The presence of ADAs was determined by ELISA. (A) Each subject from the SAD study with healthy volunteers is depicted by an individual line. Subjects with high ADAs (>480 median titer; top dotted line) are indicated with a magenta line; subjects with low ADAs (<480 median titer) are indicated with a dark blue line; subjects with undetectable ADAs are noted with a light blue line. The lower dotted line represents lower titer limit, below which samples are considered negative. (B and C) ADA data from the MAD study in RA. (B) The percentage of subjects with a positive ADA titer at any postdose time point during the study. (C) ADA titer over time in subjects with detectable ADA. LOD, limit of detection.

VIB4920 reduces disease activity as well as immunological and inflammatory biomarkers in patients with RA. In addition to safety and tolerability, evidence of clinical benefit with VIB4920 in patients with RA was also evaluated. Key end points in the MAD study in RA measured at week 12 included change in disease activity (DAS28-CRP) and biomarkers such as RF autoantibodies, serum CRP, and Vectra DA score.

DAS28-CRP is a composite clinical disease activity score used in RA that takes into account the number of swollen and tender joints, CRP levels, and a patient global health assessment. VIB4920 significantly reduced disease activity quantified by DAS28-CRP score in patients with RA at higher doses (Fig. 6A). At week 12 (day 85), the adjusted mean change from the baseline of DAS28-CRP (SE) was −2.3 (0.3) in VIB4920 1500-mg group, −2.2 (0.3) in VIB4920 1000-mg group, −1.2 (0.3) in VIB4920 500-mg group, 0.1 (0.4) in VIB4920 75-mg group, and −1.0 (0.3) in the placebo group (Fig. 6A). The effect of VIB4920 on DAS28-CRP was rapid, with reductions evident by day 15, after only a single dose of drug. The reduction of disease activity as compared with placebo was both clinically and statistically meaningful in the groups receiving the highest two doses of VIB4920: The adjusted mean difference compared to placebo at week 12 (SE) for 1500- and 1000-mg VIB4920 groups was −1.4 (0.4) and −1.2 (0.4) with P values of 0.002 and 0.006, respectively. Using a linear dose-response model, a statistically significant dose response was demonstrated for DAS28-CRP (P < 0.001; fig. S5A). The significant result was mainly driven by the 1000- and 1500-mg groups, with the 500- and 75-mg dose groups showing little to no benefit over placebo (fig. S5A). In terms of individual clinical response, 75% of patients in the 1500-mg group and 50% of patients in 1000-mg dose group achieved a DAS28-CRP score of 3.2 or less at week 12, indicating that they were in low disease activity or clinical remission at the primary end point (table S4).

Fig. 6 VIB4920 reduces disease index scores and autoantibodies in patients with RA.

Disease assessments as indicated were determined from RA donors given placebo or VIB4920 at the indicated visit days. BSL, baseline. (A) DAS28-CRP, (B) CDAI, (C) patient global assessment, (D) physician global assessment, (E) RF autoantibodies, and (F) Vectra disease activity (DA) score. (A to F) Shown is the change in each measure from baseline (means and SE). Data were analyzed using a mixed model for repeated measures analysis with corresponding baseline result included as a covariate.

RFs are a family of autoantibodies produced against the Fc portion of IgG, which are elevated in RA and are associated with poor prognosis (32, 33). VIB4920 significantly reduced RF titers at 500-, 1000-, and 1500-mg dose levels (Fig. 6E). Reductions in RF titers from baseline were evident in response to VIB4920 as early as day 29, with high-dose VIB4920 reducing RF titers by about 50% by day 85. A significant Emax dose response was demonstrated for change from baseline in RF at day 85 (P < 0.001; fig. S5B).

Vectra DA is a commercially available and validated biomarker panel that measures 12 biomarkers (adhesion molecules, growth factors, cytokines, matrix metalloproteinases, skeletal proteins, hormones, and acute phase proteins) and combines these parameters into a single score to assess the key mechanisms and pathways that drive RA disease activity (34). At day 85, the Vectra DA multibiomarker score was reduced significantly by VIB4920 at the 1000- and 1500-mg dose levels (P = 0.018 and P = 0.001, respectively; Fig. 6F). These clinical and biomarker efficacy results were highly consistent across other end points evaluated in this trial (including CDAI, tender and swollen joint counts, patient’s and physician’s global assessment, and serum CRP level; Fig. 6, B to D, and fig. S6), supporting 1000 and 1500 mg as clinically efficacious doses in this study.

DISCUSSION

The centrality of the CD40 pathway in humoral immunity has long made it an attractive target for the treatment of autoimmune diseases and transplant rejection. However, previous approaches with mAbs targeting CD40L were halted because of thromboembolic events observed in the clinical trials (17, 35, 36). Here, we describe the generation and characterization of VIB4920, a protein scaffold with specificity for CD40L, which did not induce platelet aggregation in vitro. By blocking the CD40 binding interface on CD40L, VIB4920 prevented the association between ligand and receptor and represents an approach to suppress CD40L-mediated signaling events with a non-mAb scaffold protein.

Early clinical studies of anti-CD40L blockade highlight the potential of this approach to modulate key disease parameters in autoimmune patients. In a phase 2 trial in lupus nephritis, the anti-CD40L mAb BG9588 improved hematuria and reduced anti–double-stranded DNA (dsDNA) antibody titers by almost 40% after 1 month (17). In addition, in a subpopulation of these patients, even a short course of BG9588 resulted in a 50 to 90% reduction in the frequency of total circulating IgG-secreting cells, with anti-dsDNA–secreting cells (both IgM and IgG) reduced below the level of detection (37). In line with these findings, blockade of CD40/CD40L interactions with VIB4920 resulted in a substantial reduction in B cell proliferation, plasma cell gene signature, and antibody production in a recall response to a T cell–dependent antigen. Higher doses of VIB4920 also reduced the incidence of ADAs in both healthy volunteers and patients with RA, further demonstrating the ability of VIB4920 to suppress humoral immune activation in humans. Higher doses of VIB4920 also decreased RF titers in active RA, demonstrating a clinically relevant effect on pathogenic antibodies in patients with active autoimmune disease.

In addition to demonstrating proof of mechanism, VIB4920 also achieved a proof of clinical concept in RA, a prototypic autoimmune disease, by significantly improving disease activity with the two highest-dose levels tested. The improvement was rapid (the separation versus placebo was evident as early as 2 weeks for clinical activity score DAS28-CRP) and robust at week 12, the time of primary evaluation of efficacy. The mean reduction in DAS28 score versus placebo with VIB4920 was similar to the treatment effects seen with other compounds licensed for the treatment of RA in a methotrexate (MTX)–refractory population. Moreover, in our study, the population included more refractory patients who failed one or multiple biological agents and may have represented a more severe subset of RA. The consistency of improvement across a variety of clinical and laboratory outcome measures further supports the potential clinical efficacy of VIB4920. However, the small number of autoimmune patients evaluated to date represents a limitation of the current studies with VIB4920. Future larger clinical trials are needed to further explore the potential of this molecule to affect disease parameters in settings of autoimmunity.

Thromboembolic complications were observed after treatment with several unique anti-CD40L mAbs (17, 35, 36). The mechanism of this toxicity has been definitively linked to the Fc portion of anti-CD40L mAbs. Our approach to targeting the CD40/CD40L pathway relies on a Tn3 protein scaffold that lacks an Fc region, which would not be predicted to induce adverse platelet responses. Ex vivo studies with washed human platelets demonstrate that VIB4920 does not induce aggregation. It remains a theoretical possibility, however, that ADAs in complex with VIB4920 could induce platelet aggregation in vivo. However, there were no platelet/coagulation abnormalities observed in the phase 1 studies in healthy volunteers and patients with RA, and in subjects with positive ADA titers, no thromboembolic complications have been observed to date. VIB4920 demonstrated a favorable safety profile in both healthy volunteers and in RA subjects. Administration of VIB4920 was not associated with an increased incidence of infections, and comparable TEAEs were observed between the placebo and VIB4920 cohorts.

In summary, we have generated a CD40L-targeted protein scaffold with the potential to substantially modulate humoral immune activation. VIB4920 represents an alternative to mAb-based targeting of CD40L, which does not induce platelet aggregation in vitro and demonstrates a favorable safety profile in early clinical evaluation. Substantial reductions in DAS28-CRP and several other clinical biomarkers in patients with RA suggest that effective blockade of CD40L by VIB4920 may provide therapeutic benefit in autoimmune patients.

MATERIALS AND METHODS

Study design

The objective of this study was to evaluate the impact of CD40/CD40L blockade using a Tn3-based inhibitor, VIB4920, in both preclinical and clinical settings. In vitro studies were performed with both cell lines and primary human cells to determine the impact of VIB4920 on CD40 signaling, B cell expansion, and plasma cell differentiation. The effect of Tn3-based CD40L inhibition on humoral immune responses was assessed in a preclinical mouse SRBC immunization model. For preclinical studies, human blood from healthy donors was collected after informed consent as approved by MedImmune’s Institutional Review Board (no. 00004341). The number of independent experiments and the number of replicates per experiment are indicated in the figure legends where appropriate. VIB4920 was also evaluated in two phase 1 clinical studies (NCT02151110 and NCT02780388) and characterized for its ability to block humoral immune responses to immunization in healthy volunteers and for its impact on biomarkers and disease activity in RA subjects. Clinical studies were conducted in a randomized, blinded fashion as detailed in the clinical study design sections below. Primary data are shown in data file S1.

Tn3 generation

The generation of Tn3 phage display libraries, selection of mouse and human CD40L-specific variants, and construction and expression of multivalent Tn3 constructs were performed as previously described (20, 28, 29).

NF-κB reporter assay

HEK293 cells expressing an NF-κB luciferase reporter (Panomics) were engineered to stably express human full-length CD40R. Cells were seeded at a density of 5 × 104 cells per well in a 96-well poly-d-lysine–coated plates (BD Biosciences) and stimulated with MEGACD40L recombinant protein (1.5 μg/ml; Enzo Biosciences) or CD40L-overexpressing D1.1 Jurkat subclone (American Type Culture Collection) cells for 16 to 24 hours in the presence or absence of control or CD40L-specific Tn3s at indicated concentrations. Luminescence was detected using the Bright-Glo Luciferase Assay System (Promega) on a SpectraMax M5 plate reader (Molecular Devices).

CD86 up-regulation assay

PBMCs were isolated from CPT tubes (BD Biosciences) after centrifugation. PBMCs (2.5 × 105 to 5.0 × 105 cells per well) were stimulated in a 96-well round-bottom plate with recombinant MEGACD40L (100 ng/ml; Enzo Biosciences) for 16 to 18 hours in the presence of CD40L-specific Tn3s or mAb (clone 5c8) as indicated. Flow cytometry was used to evaluate CD86 expression on CD19+ B cells. The following antibodies were used: CD86 (clone 2331, BD Pharmingen) and CD19 (clone HIB19, BD Pharmingen).

Human B cell assay

PBMCs were isolated as described above. Total B cells were negatively selected using MACS cell separation technology (Miltenyi Biotec), which routinely yielded greater than 95% purity. Purified peripheral blood B cells were cultured at a density of 0.5 × 105 to 1.0 × 105 cells per well in 96-well round-bottom plates. Culture medium for B cell experiments was RPMI 1640 (Invitrogen) supplemented with 10% fetal calf serum, penicillin-streptomycin [penicillin (100 U/ml ) and streptomycin (100 μg/ml)], 2-mercaptoethanol (55 μM), l-glutamine (2 mM), and Hepes (5 mM). At initiation of culture, B cells were stimulated with a combination of IL-21 (33 ng/ml; PeproTech Inc.) and MEGACD40L (1.5 nM; Enzo Biosciences) with or without anti-IgM F(ab′)2 (5.0 μg/ml; Jackson ImmunoResearch Laboratories). B cell expansion was quantified by measuring adenosine 5′-triphosphate on day 3 or 4 of culture using the CellTiter-Glo Luminescent Assay (Promega), according to the manufacturer’s instructions. Plasma cell differentiation was quantified on day 7 by flow cytometry. Cells were acquired for a fixed amount of time, and plasma cells were defined as CD19+ IgD CD38hi cells.

Murine SRBC immunization model

Balb/c mice (the Jackson Laboratories) were immunized on day 0 with 0.2 ml of SRBC (Colorado Serum Company) by intraperitoneal injection. Control (30 mg/kg) or CD40L-specific Tn3s (up to 30 mg/kg) were administered daily from days 9 to 13 (intravenously). The frequency of GC B cells in the spleen was quantified on day 14 by flow cytometry. GC B cells were defined as CD19+B220+ Fas+PNA+ B cells.

Platelet aggregation assay

Human blood was collected from healthy donors into Acid Citrate Dextrose (ACD) Solution B tubes. After centrifugation, platelet-rich plasma was transferred into a polypropylene tube and incubated for 10 min with apyrase (2 U/ml) to prevent platelet activation during processing. Platelets were pelleted and resuspended in modified Tyrode’s buffer [137 mM NaCl, 2.7 mM KCl, 1 mM MgCl2, 5.6 mM dextrose, 3.3 mM NaH2PO4, 20 mM Hepes, and 0.1% bovine serum albumin (BSA) (pH 7.4)].

Immune complexes were generated by mixing the mAb (h5c8 or negative control antibody) or anti-CD40L Tn3 with hCD40L (293 Cell Source) for 5 min at room temperature. In some experiments, platelets were preincubated with anti-CD32a antibody (IV.3, MedImmune).

Phase 1a subjects and study design

A phase 1, randomized, blinded, placebo-controlled study was conducted in healthy adults aged 18 to 49. Subjects were randomized in a 2:1 ratio for the first two cohorts and a 4:1 ratio for all remaining cohorts. T cell–dependent antibody response was induced in subjects by administering two separate immunizations of 1 mg of KLH subcutaneously. The first KLH immunization was administered during the screening period, 14 days before dosing with either VIB4920 or placebo, and the second KLH immunization was administered on day 15 after dosing with either VIB4920 or placebo.

Phase 1b patients and study design

A phase 1b randomized, blinded, placebo-controlled study was conducted in patients aged 18 to 70 years old diagnosed with RA according to European League Against Rheumatism/American College of Rheumatology criteria (38) for at least 6 months before entering the study. Subjects had moderate to severe activity as defined by a DAS28-CRP score of at least 3.2 at screening and at least four swollen and four tender joints at screening and randomization. Patients were positive for either RF (RF-IgM, ≥14 U/ml) or anti–citrullinated peptide antibodies at screening. Patients received MTX at a dose of 7.5 to 25 mg/week or, in case of MTX intolerance, a different conventional disease-modifying antirheumatic drugs, started at least 12 weeks and at a stable dose for at least 6 weeks before screening. Previous treatment with biological agents (except rituximab or other B cell–depleting agents) given for RA was accepted, provided that proper washout was performed before randomization in our study. Patients were treated with placebo (0.9% saline, n = 15) or VIB4920 (75 mg, n = 8; 500 mg, n = 10; 1000 mg, n = 12; or 1500 mg, n = 12) given by intravenous infusion every other week for 12 weeks, followed by 12 weeks of posttreatment observation. Fifty-three patients completed 12 weeks of treatment; two patients (one in 75-mg VIB4920 group and one in 1500-mg VIB4920 group) discontinued treatment due to adverse events. One patient in the placebo group withdrew informed consent, and one patient in the 75-mg VIB4920 group was lost to follow up. Measurements for Vectra DA score were performed by Crescendo Bioscience, and RF autoantibody measurements were performed by Covance Central Laboratories Services.

PK assay for VIB4920

VIB4920 in human K2EDTA plasma was measured using a validated sandwich ELISA method in which wash steps with 1× phosphate-buffered saline (PBS)/0.1% Tween 20 (PBST) followed each incubation to remove unbound components. Briefly, Nunc microtiter plates were coated overnight at 2° to 8°C with anti-VIB4920 mouse mAb (1 μg/ml; MedImmune). Standards, quality controls (QCs), and samples containing VIB4920 were diluted to the method minimum required dilution (MRD) of 1:50 in 0.5% BSA/PBST before plate addition. After a 2-hour incubation, biotin-labeled anti-VIB4920 rat antibody (1 μg/ml; MedImmune) was added to the plate and incubated for 1 hour. The binding complex was visualized with successive incubations of streptavidin-linked horseradish peroxidase (HRP; GE Healthcare) and SureBlue tetramethylbenzidine (TMB) peroxidase substrate (KPL Inc.). Color development was stopped with 0.2 M sulfuric acid before analysis at 450 nm on a microplate reader. The quantitative range was 0.05 to 1.60 μg/ml; samples measuring above the quantitative range were diluted with pooled K2EDTA plasma to bring the concentration within the measurable range of the method.

Quantitation of sCD40L

Total sCD40L (free sCD40L and sCD40L bound to VIB4920) in human K2EDTA plasma was measured using a human sCD40L Platinum ELISA kit (eBioscience) that had been modified to meet program needs. Briefly, standards, QCs, and samples containing sCD40L were diluted 1:50 in assay diluent containing 0.5% BSA/PBST and VIB4920. Wash steps with PBST followed each incubation to remove unbound components. The diluted samples were added to a plate precoated with anti-sCD40L antibody and incubated for 1.5 hours. HRP-conjugated anti-human sCD40L was then added to bind to sCD40L captured by the coat antibody. The binding complex was visualized with successive additions of TMB peroxidase substrate and stop solution (phosphoric acid) before analysis at 450 and 540 nm on a microplate reader. The quantitative range was 6.25 to 400.00 ng/ml; samples measuring above the quantitative range were diluted with pooled K2EDTA plasma to bring the concentration within the measurable range of the method.

Measurement of ADA

The presence of ADAs to VIB4920 in human K2EDTA plasma was determined using a validated sandwich ELISA method in which wash steps with PBST followed each incubation to remove unbound components. Briefly, QCs and samples were diluted 1:60 in assay diluent containing 0.5% BSA/PBST, added to a washed Pierce Protein G coated plate (Thermo Fisher Scientific), and incubated for 2 hours. Overnight incubation of biotin-labeled VIB4920 (1 μg/ml) prepared in assay diluent specifically detected ADA to VIB4920. The binding complex was visualized with successive incubations of streptavidin-linked HRP (GE Healthcare) and SureBlue TMB peroxidase substrate (KPL Inc.). Color development was stopped with 0.2 M sulfuric acid before analysis at 450 nm on a microplate reader. Each sample was subject to a three-tier process where the sample response was first compared to a statistically determined cutoff optical density value, at or above which a sample was considered potentially positive and below which the sample was determined negative for ADA. The potentially positive samples were subjected to a second competition evaluation in the presence of excess VIB4920; samples with a percent inhibition at or above the statistically determined confirmatory cut point were defined as confirmed positive and taken into a titer evaluation. Samples below the confirmatory cut point were considered negative for ADA to VIB4920. Titered samples were serially diluted in pooled human K2EDTA plasma to below the screening cutoff, and the titer result was reported as the reciprocal of the highest dilution at which the sample measured positive before measuring negative.

Assessment of anti-KLH antibodies

Anti-KLH antibody in human serum was measured using a validated sandwich ELISA method in which wash steps with PBST followed each incubation to remove unbound component. Briefly, Nunc microtiter plates were coated overnight at 2° to 8°C with KLH (3 μg/ml; IMMUCOTHEL, biosyn Arzneimittel GmbH) prepared in PBS.Standards and QCs are composed of a mixture of anti-KLH antibodies of varying isotype and affinity (AstraZeneca). Samples containing anti-KLH IgG antibodies were diluted to the method MRD of 1:250 in 0.5% BSA/PBST before plate addition. For IgM, standards and QCs were prepared in 2% serum pool, with samples diluted 1:50 in 0.5% BSA/PBST before plate addition. After a 2-hour incubation, HRP-conjugated mouse anti-human IgG or anti-human IgM (Invitrogen) was added to the plate and incubated for 1 hour. The binding complex was visualized with successive additions of TMB peroxidase substrate and stop solution (0.2 M sulfuric acid) before analysis at 450 nm on a microplate reader. The quantitative range was 163.3 to 10,000 ng/ml for IgG and 7.8 to 500 μg/ml for IgM; samples measuring above the quantitative range were diluted with serum to bring the concentration within the measurable range of the method.

Flow cytometry in phase 1a

Blood was collected in Cyto-Chex blood collection tubes (Streck), shipped to Covance Central Laboratory Services, and tested by flow cytometry using a validated method. Briefly, cells were stained with fluorochrome-labeled antibodies to CD45 (clone HI30), CD19 (clone HIB19), IgD (clone IA6-2), CD27 (clone M-T271), and CD38 (clone HIT2; all from BD Biosciences) to identify B cell populations. Cells were subsequently treated with FACS Permeabilizing Solution 2 (Becton Dickinson) and stained for intracellular Ki67 (clone Ki67, BioLegend) expression to measure proliferating cells.

Plasma cell signature

Plasma cell gene signature was determined as previously described (31). ΔΔCt values were calculated using the mean of two reference genes (β-actin and glyceraldehyde-3-phosphate dehydrogenase) and each patient’s baseline expression as controls. Fold change values were determined by calculating 2−ΔΔCt. Plasma cell panel included the following genes: IGHA1, IGJ, IGKC, IGKV4-1, and TNFRSF17.

Statistical analysis

Two-tailed unpaired Student’s t tests were used to evaluate the impact of treatment on primary human B cell expansion, plasma differentiation, and the GC B cell response in the SRBC model (Fig. 2, C and E, and fig. S2). Mann-Whitney U test was used to compare VIB4920 versus placebo at multiple time points for gene signature score (Fig. 4F). Statistical tests and plots were performed using GraphPad Prism software. The dose response for change from baseline in DAS28-CRP and RF at day 85 was analyzed using Multiple Comparison Procedure-Modelling (MCP-Mod) approach (39), including corresponding baseline as a covariate, with three prespecified candidate models for the dose response (linear, Emax, and a Hill-Emax model). The testing of dose-response signal was adjusted for multiplicity to control family-wise error rate at 0.10 level. The final model was selected among those indicated as significant based on the Akaike Information Criteria. Change from baseline in DAS28-CRP, RF, Vectra DA, CDAI, tender joint count, swollen joint count, patient’s and physician’s global assessment, and serum CRP were analyzed using a mixed model for repeated measures analysis with corresponding baseline result included as a covariate.

SUPPLEMENTARY MATERIALS

stm.sciencemag.org/cgi/content/full/11/489/eaar6584/DC1

Materials and Methods

Fig. S1. Structural characterization of CD40L-Tn3 molecule and interactions.

Fig. S2. Mouse surrogate Tn3 shows potent neutralizing activity in vivo in response to immunization.

Fig. S3. Study design for phase 1a study to evaluate safety of VIB4920.

Fig. S4. Phase 1b study design to evaluate VIB4920 in patients with RA.

Fig. S5. Dose-dependent reduction in DAS28-CRP and RF autoantibodies in RA in response to VIB4920.

Fig. S6. Impact of VIB4920 on tender/swollen joint counts and CRP in phase 1b study in patients with RA.

Table S1. Fusion with serum albumin greatly improves half-life of Tn3 molecule.

Table S2. VIB4920 demonstrates a good safety profile in healthy volunteers.

Table S3. VIB4920 demonstrates an acceptable safety profile in patients with RA.

Table S4. DAS28 categories at day 85.

Data file S1. Primary data.

REFERENCES AND NOTES

Acknowledgments: We would like to thank L. Carter, A. J. Coyle, K. Folliot, A. Godwood, L. Grinberg, F. Huang, G. Illei, C. Morehouse, O. Sanduja, M. Sheehan, D. Sinibaldi, and L. Wang from MedImmune. Funding: This project was solely funded by MedImmune LLC. Author contributions: M.B. and T.T. were involved in the generation and molecular characterization of the Tn3 proteins. V.O. designed and ran the crystallography studies. S.D. was the research lead for the project and was responsible for preclinical studies supported by J.L.K., X.X., and I.Y. M.A. was the clinical lead for the program. Statistical analysis of clinical data was performed by L.W. and R.M. M.G., L.Y., and J.L. generated and interpreted PK, PD, and ADA data for the clinical studies. S.C.E. performed analysis of phase 1 flow cytometry data. K.S. was the translational lead for the project with support from M.d.l.R. in generation of plasma cell gene signature data for phase 1 study. U.M.-L. served as advisor for the study design and interpretation of data for the phase 1b study. D.H. supported study design and clinical oversight for the phase 1b study. Work was performed under the guidance/supervision of S.D., R.E., R.H., and J.D. J.L.K. and M.A. wrote the manuscript with input from all authors. Competing interests: All authors were or currently are full-time employees and shareholders of MedImmune/AstraZeneca. J.L.K., L.W., M.G., L.Y., R.E., and J.D. are employees and stakeholders of Viela Bio. This work is related to patent number WO2013055745 filed on 10 October 2012. Data and materials availability: All data related to this study are in the paper or the Supplementary Materials. Atomic coordinates and corresponding structure factors have been deposited with Protein Data Bank with accession code 6BRB.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article