Engineering a protease-stable, oral single-domain antibody to inhibit IL-23 signaling

Abstract

Interleukin (IL)-23 is a validated therapeutic target in inflammatory bowel disease. While antibodies targeting IL23 demonstrate clinical efficacy, they face challenges such as high costs, safety risks, and the necessity of parenteral administration. Here, we present a workflow to simultaneously enhance the affinity and protease stability of an inhibitory anti-IL23R VHH for oral use. Cocrystal structure analysis reveals that the anti-IL23R VHH employs both CDR and framework residues to achieve picomolar affinity for IL23R. The engineered VHH remains stable for over 8 h in intestinal fluid and 24 h in fecal samples. Oral administration of this VHH achieves deep pathway inhibition in a murine colitis model. Furthermore, a single pill provides sustained IL23R inhibition in nonhuman primate blood for over 24 h. With high potency, gut stability, high production yield, and favorable drug-like properties, oral VHHs offer a promising approach for inflammatory bowel diseases.

Keywords: antibody; immunology; oral delivery; protein engineering.

Fig. 1.

Engineering and characterization of a high-affinity, protease-stable inhibitory anti-IL23R VHH. (A) Three anti-IL23R VHHs (1D11, 2A8, 2F1) block the interaction of IL23R with IL23 as determined by the ELISA (n = 5 replicates). (B) SDS-Page analysis of VHHs treated with trypsin, chymotrypsin, elastase, or pancreatin for 15 or 60 min at 37 °C reveals varying levels of protease stability with 2A8 exhibiting the best initial stability. L: ladder, Pr: preincubation, Ez: enzyme only. (C) Engineering three mutations into 2A8 resulted in significantly improved stability to trypsin (R27F, V97P, Top panel) and chymotrypsin (S100bQ, Bottom panel) as observed by SDS-Page. (D) Diagram of key sites with the parental 2A8 clone. Mapped sites of proteolysis are highlighted by triangles (filled for the primary site; open for the secondary site). Mutations introduced to 2A8 are listed as single residues in black (protease stability) or gray (affinity and stability). Sequences corresponding to CDRH1, CDRH2, and CDRH3 are underlined. (E) Affinity and protease stability analysis of a panel of VHHs with combinatorial mutations identified a lead VHH with improved affinity and protease stability (star). (F) Diagram of v129 sequence. Sequence differences from the parental clone as highlighted in bold. Sequences corresponding to CDRH1, CDRH2, and CDRH3 are underlined. (G) SPR analysis of lead VHH (n = 3 replicates). (H and I) The lead VHH exhibited stability in mouse small intestinal fluid (H) or pancreatin (I) for over 8 h, while the parent VHH was rapidly degraded (n = 3 replicates). (J) Fecal samples were harvested from the mice gavaged with the lead VHH at 4 h, and further incubated in vitro at 37 °C for the indicated time. After homogenization, active VHH in the supernatant was analyzed (n = 3 mice). (K) The lead VHH exhibited prolonged stability in human (n = 2 donors) fecal homogenate supernatant over a 24 h window. For the sample at time = 0 h were quenched right after starting the incubation. Data represented as mean ± s.d. (L) Parental and lead VHH exhibit a dose-dependent inhibition of the IL23–IL23R interaction by the ELISA (n = 5 replicates, mean ± s.d.). (M) Treatment of PBMCs with IL-23 induces IL17F production and the lead VHH can inhibit this effect in a dose-dependent manner (n = 4 donors).

Fig. 2.

Structural basis for inhibition of IL23R by VHH. (A) The VHH (gold) binds to domain 1 (D1) of IL23R (domains shown in shades of gray and labeled), interacting primarily via its CDR2 and CDR3 (IL23R interface colored green and teal, respectively). (B) IL23p19 (pink) in the 5MZV structure recognizes a substantially overlapping region of D1 (binding site, colored purple). (C and D) “Open book” representation of the interaction surfaces on IL23R (Left) and the VHH (C, Right) or IL23p19 (D, Right). The interacting surface of D1 is colored as in a, and the VHH is colored to mirror that (CDR1 = purple, CDR2 = green, CDR3 = teal, interacting regions outside of the CDRs = gray). The interacting region of IL23p19 is colored purple. (E) Interface of the complex, with IL23R D1 represented as a semitransparent gray surface and the VHH as a ribbon diagram, each with interfacial residues shown as sticks. The amino terminal (N-term) region of D1 is labeled. Hydrogen bonds and residue numbers are omitted for clarity.

Fig. 3.

Pharmacokinetics of anti-IL23R VHH after oral gavage in healthy and colitis mice. (A) C57BL/6 mice were orally gavaged with parental and v134 VHH at 20 mpk and were housed in metabolic cage individually. Feces samples and urine samples were collected, and at the end of experiment, the contents were harvested from the colon and cecum. VHHs were isolated from samples and analyzed by the ELISA (n = 4). (B–F), Healthy or DSS-treated human IL23R exon 3 knock-in mice were orally gavaged with VHH v134, taken down at the indicated time, and active VHH in serum (B), urine (C), mesenteric lymph node (mLN, D), Peyer’s patch (E), and colon (F) were quantified by the ELISA (n = 3 to 4, mean ± s.e.m.). (G) Representative images of immunohistochemistry staining of tissues from healthy or DSS-treated human IL23R exon 3 knock-in mice at the indicated time point after oral gavage with VHH v134. Goblet cells (red arrowhead) and Paneth cells (yellow arrowhead) are indicated (n = 3 to 4). (Scale bar, 200 µm.)

Fig. 4.

Oral VHH demonstrates strong PD in two mouse models of colitis. (A and B) Human IL23R exon 3 knock-in mice received 3% DSS in drinking water from day −9 to day −2, then mice were treated with VHH v134 or vehicle control orally at −18, −3, and 0 h, or anti-IL23R IgG or control IgG intraperitoneally at −3 h. At time 0, human IL23 or PBS was injected intravenously, and tissue samples were harvested at 3 h. Without (open circle) or with (closed circle) IL23 treatment. VHH treatment led to robust inhibition of Il22 and Il17f in colon, ileum, or mesenteric lymph nodes (A, n = 7 to 8), and in mLN, Peyer’s patches, proximal, middle, and distal colon and rectum (B, n = 4 to 6). (C and D) Anti-CD40 model. Human IL23R exon 3 knock-in Rag2 knock-out mice received low concentration (1.5%) of DSS in drinking water from day −4 to day 3, and anti-CD40 or PBS was injected intravenously on day 0. VHH v134 or vehicle control was given orally 3 times a day from day 2 to day 5, or anti-IL23R IgG or control IgG was given on day 3 and day 5. The tissue samples were harvested on day 6. Without (open symbols) or with (closed symbols) anti-CD40 treatment. (C) Colon weight/length ratio showing normalization in the group treated with VHH v134 or anti-IL23R IgG (n = 5 to 9). (D) VHH treatment led to robust inhibition of Il22, Il17f, Tnf, and Lcn2 gene expression in the colon tissue (n = 5 to 9). Data represented as mean ± s.e.m. with plots showing the value of individual mice. Statistical significance was calculated for individual mice using a two-tailed unpaired t test (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).

Fig. 5.

Pharmacokinetics and pharmacodynamics of orally delivered VHH in cynomolgus monkeys. Cynomolgus monkeys received oral VHH v134 containing or control pills on days 0, 3, 4, 5, and 6. Blood samples were collected for PK and PD analysis at the indicated time points. Animals were taken down at 8 h after the last dosing. (A) Enteric coated capsule containing mini-tablets of spray-dried VHH. (B) Serum PK analysis after the first dose reveals sustained exposure over 24 h (n = 4). (C) Representative images of immunohistochemistry staining of VHH in ascending colon tissues from the animal at 8 h after the last oral dose with VHH v134. (D and E) Receptor occupancy assay. Whole blood was stimulated with hIL23 ex vivo, and phosphorylated STAT3 was analyzed by FACS. (D) Representative FACS plot showing pSTAT3+ population after ex vivo stimulation with hIL23. Samples from predosed animal (Left) and 8 h postdose animal (Right). (E) Receptor occupancy assay showed sustained target engagement (>75%) over a 24 h window that was maintained upon once-daily dosing for 4 d. The percentage of pSTAT3 was normalized to the value at the predose (n = 4). Data represented as mean ± s.e.m.