Building a potent TREM2 agonistic, biparatopic, common light chain antibody

Abstract

Triggering Receptor Expressed on Myeloid cells 2 (TREM2) plays an important role in microglial function and has been genetically linked to Alzheimer's disease. Activation of TREM2 signaling may contribute to protection against neurotoxic effects of amyloid. Numerous TREM2 activating antibodies have been shown to modulate downstream microglial functions to different degrees, with mixed results in preclinical models and in the clinic. We sought to generate an effectorless agonistic antibody that acted solely through TREM2 engagement with sufficient potency to activate TREM2 in the brain. Our approach focused on building a multivalent biparatopic TREM2 antibody that could mimic the higher order clustering induced by native polyanionic ligands of TREM2. We describe our screening strategy and findings that led to the discovery of a potential therapeutic molecule composed of antibodies selected for optimal affinity, binding epitopes, and geometry. The most productive antibody pair was selected from a common light chain yeast-display library, which required multiple rounds of affinity maturation. Lead antibody candidates were converted into asymmetric tetravalent bispecifics via controlled Fab-arm exchange and subsequently screened in signaling assays. The most productive antibody pair was reengineered into a symmetric tetravalent format, increasing potency and simplifying development. This molecule exhibited higher efficacy and potency in signaling assays than other antibody formats tested and elicited TREM2-mediated chemokine responses in vivo. Our results demonstrate a biparatopic strategy for producing a high potency TREM2 agonistic antibody with low effector function that can modulate TREM2 signaling in vitro and brain pharmacodynamic responses in vivo.

Keywords: Agonistic antibodies; Alzheimer’s disease; TREM2; biparatopic antibodies; bispecific antibodies; common light chain; microglia; neuroinflammation.

Figure 1.

TREM2 tool antibodies activity in various formats. (a) BMDM viability assay with TREM2 mAbs and isotype control in solution (5 day incubation, 0.2 nM CSF-1 with 100 nM antibodies or media only), (b) same mAbs platebound in BMDM viability assay. Red dashed line indicates viability of cells with isotype control in both assays. (c) Alternate antibody formats evaluated include asymmetric bispecific (AsBs), tetravalent (Tv), symmetric tetravalent bispecific (SymTvbs) with Efab domain substitution indicated in light gray, and asymmetric tetravalent bispecific (AsTvBs). Bispecifics were generated by combining antibodies 3 or 4 with antibody 10. (d) BMDM viability assay of alternate formats compared to soluble mAbs (48 h incubation, no CSF-1 with antibodies at 100 nM concentration or media only). Red dashed line indicates viability of cells with isotype control. (e) Schematic of NanoBiT huTrem2/hudap12/husyk assay developed to enable high-throughput screening of antibodies for their ability to trigger DAP12 phosphorylation, resulting in an interaction with SYK, in a human system. (f) NanoBiT pDAP12 assay validation with soluble and crosslinked antibodies at 50 nM, with and without huTREM2 expression. (g) Dose-response curves of mAbs and alternate formats in NanoBiT pDAP12 assay, normalized to Emax of plate control AsTvBs 4/10 @ 200 nM. (h) Schematic of hypothetical binding mode of a tetravalent bispecific with TREM2 shown in cyan and gray, TREM2 biparatopic antibody shown in black and blue, and cell membrane shown in pink.

Figure 2.

(a) Common light chain (cLC) antibody campaign overview. Red dots indicate number of CDRs involved in affinity maturation at each stage. (b) Model of membrane-bound TREM2 ECD including IgV (resi 19–134, green) from PDB 5ELI and stalk (resi 135–174, orange line) with sheddase site between His157 and Ser158 indicated by dashed line. (c) Family tree of select antibodies from cLC campaign. Generation 1 = antibodies from screen; subsequent generations: 2 = H1/H2 shuffle; 3 = H3 mutagenesis, 4: H1/H2/H3 mutagenesis using walking oligos. Monovalent binding affinity (K_D) reported for antibodies with measurable value as determined by surface plasmon resonance analysis of antibodies using single-cycle analysis and kinetic parameters determined using steady state or 1:1 kinetics model.

Figure 3.

Screening of cLC antibodies in asymmetric tetravalent format to identify optimal pairing for agonism. (a) Workflow for identifying productive antibody pairs encompassing multiple rounds of affinity maturation, controlled-Fab arm exchange (cFAE) to generate bispecific pairs (mutations – A: F405L, B: K409R), and activity screening, followed by reengineering into a symmetric tetravalent bispecific format with one heavy and light chain sequence. (b) Antibodies combined into asymmetric tetravalent format from the first two generations. (c, d) NanoBiT assay data for combinations of first- and second-generation antibodies, normalized to plate control of AsTvBs 4/10 at 250 nM. (e) NanoBiT and (f) THP1 pSYK assay data for the most productive combinations: the fourth-generation stalk binder CAD7409 combined with IgV binder CAD3358 family members from the 2^nd-4^th generations, normalized to plate control of AsTvBs 4/10 at 83.3 nM or 200 nM, respectively.

Figure 4.

Characterization of the lead antibodies combined in SymTvBs format. (a) SymTvBs reformatting of lead pair 7409/7268 into two antibody orientations. (b, c) Dose-response curves in NanoBiT and THP1 assays, 200 nM nM AsTvBs 4/10 used for normalization. (d) Viability in hTREM2 BMDMs, derived from human TREM2 bac transgenic mice, with antibodies tested at 20 and 1 nM, normalized to 200 nM AsTvBs. (e, f) Activity in THP1 pSYK AlphaLISA and aggregation assay, normalized to 31.6 nM undefined 4/10. (g) THP1 pSYK specific for pTyr525/526 by AlphaLISA with 20 nM antibody with or without 0.5 mg/mL liposomes (*p < 0.05; **p < 0.01).

Figure 5.

Pharmacodynamic characterization of the lead bispecific antibody SymTvBs 7409/7268 in a human TERM2 Bac transgenic mouse model. Mice 8–12 weeks old were IV dosed with 25, 50, and 100 mpk antibody or PBS and taken down after 24 hours to determine acute responses by (a) sTREM2 MSD and (b-e) FirePLex immunoassay. Antibody levels in the (f) cerebellum and (g) serum were determined by hIgG MSD. Cerebellum concentrations were 380 pM, 690 pM, and 1.54 nM, all exceeding the EC50 for activity in the THP1 pSYK MSD and AlphaLISA assays, 150 and 190 pM, respectively (* p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001).

Figure 6.

Crystal structures of Fabs 7268 and 7411 in complex with TREM2. (a) Side view of the Fab 7268 (heavy chain, magenta; light chain, turquoise) bound to MBP-TREM2 IgV (gray and green). (b) Close-up view of the Fab 7268/TREM2 IgV binding interface, highlighting key residues within 4 Å of each other. (c) Side view of the Fab 7411 (heavy chain, purple; light chain, marine) in complex with the TREM2 stalk peptide (residues 131–148, green). (d) Detailed view of the Fab 7411/TREM2 stalk peptide interface, illustrating the TREM2 stalk wrapping around the CDR H3 residues of Fab 7411.