Intestinal mucosal barrier repair and immune regulation with an AI-developed gut-restricted PHD inhibitor

Fu, Yanyun; Ding, Xiao; Zhang, Man; Feng, Chunlei; Yan, Ziqi; Wang, Feng; Xu, Jianyu; Lin, Xiaoxia; Ding, Xiaoyu; Wang, Ling; Fan, Yaya; Li, Taotao; Yin, Yushu; Liang, Xing; Xu, Chenxi; Chen, Shan; Pulous, Fadi E.; Gennert, David; Pun, Frank W.; Kamya, Petrina; Ren, Feng; Aliper, Alex; Zhavoronkov, Alex

doi:10.1038/s41587-024-02503-w

Download PDF

Brief Communication
Open access
Published: 11 December 2024

Intestinal mucosal barrier repair and immune regulation with an AI-developed gut-restricted PHD inhibitor

Yanyun Fu^1,2,3,4,
Xiao Ding^1,2,4,
Man Zhang^1,2,3,
Chunlei Feng^1,2,
Ziqi Yan^1,2,
Feng Wang^1,2,
Jianyu Xu^1,2,
Xiaoxia Lin^1,2,
Xiaoyu Ding^1,2,
Ling Wang^1,2,
Yaya Fan^1,2,
Taotao Li^1,2,
Yushu Yin^1,2,
Xing Liang^1,2,
Chenxi Xu^1,2,
Shan Chen^1,2,
Fadi E. Pulous^1,5,
David Gennert ORCID: orcid.org/0000-0002-8668-3601^1,5,
Frank W. Pun ORCID: orcid.org/0000-0001-8801-6645^1,3,
Petrina Kamya^1,6,
Feng Ren ORCID: orcid.org/0000-0001-9157-9182^1,2,3,4,
Alex Aliper^1,4 &
…
Alex Zhavoronkov ORCID: orcid.org/0000-0001-7067-8966^1,2,3,5

Nature Biotechnology (2024)Cite this article

20k Accesses
12 Citations
168 Altmetric
Metrics details

Subjects

Abstract

Hypoxia-inducible factor prolyl hydroxylase (PHD) inhibitors have been approved for treating renal anemia yet have failed clinical testing for inflammatory bowel disease because of a lack of efficacy. Here we used a multimodel multimodal generative artificial intelligence platform to design an orally gut-restricted selective PHD1 and PHD2 inhibitor that exhibits favorable safety and pharmacokinetic profiles in preclinical studies. ISM012-042 restores intestinal barrier function and alleviates gut inflammation in multiple experimental colitis models.

Artificial intelligence guided discovery of a barrier-protective therapy in inflammatory bowel disease

Article Open access 12 July 2021

Mechanistic insights into gut microbe derived siderophores and PHD2 interactions with implications for HIF-1α stabilization

Article Open access 07 January 2025

Oral administration of human carbonic anhydrase I suppresses colitis in a murine inflammatory bowel disease model

Article Open access 26 October 2022

Main

Interventions for inflammatory bowel disease (IBD) remain a pressing clinical need as 40% of persons with moderate-to-severe disease lack adequate treatments^1,2. IBD is a progressive, chronic condition characterized by inflammation and epithelial damage within the gastrointestinal tract^3,4. Histologic mucosal remission strongly correlates with improved long-term clinical and patient-reported outcomes, although the current standard-of-care (SOC) anti-inflammatory treatments do not directly promote mucosal healing^5,6. Therapies targeting recently reported autoimmune components of IBD are challenging to design given the complex mechanisms at play³. However, targeted therapies that promote epithelial repair and reduce gut inflammation may be effective, pragmatic alternatives. The prolyl hydroxylase 1 and 2 (PHD1 and PHD2)–hypoxia-inducible factor 1α (HIF1α) axis was identified and prioritized by PandaOmics^7,8,9,10,11, a commercially available artificial intelligence (AI) platform that sources omics and published data for target identification, as a therapeutic target for IBD (Supplementary Fig. 1a). The PHD genes (PHD1–PHD3) are oxygen sensors and promote HIF1α or HIF2α degradation under oxygen-rich conditions. Intestinal PHD1– or PHD2–HIF1α signaling has been implicated in epithelial barrier function through its transcriptional regulation of mucus production, tight-junction formation, immunosuppressive metabolite generation and wound healing. Additionally, intestinal PHDs have crucial roles in immune cell recruitment, proinflammatory cytokine production and colonic adaptive immune cell reprogramming¹². The colon has a steep oxygen gradient from the anaerobic lumen across the epithelium to the richly vascularized subepithelial mucosa. Thus, a disrupted oxygen gradient in IBD indicates dysregulated PHD activity^12,13,14. Notably, despite proof of concept in preclinical colitis models, existing PHD inhibitors have failed clinical trials for IBD. This may be because systemic PHD inhibition is associated with safety risks such as cardiovascular and tumorigenic side effects and current PHD inhibitors such as 2-oxoglutarate (2-OG) analogs or Fe(II) chelators lack specificity, thus increasing the risk of off-target adverse effects. The recent, high-profile failure of GB004 (ref. ¹⁵), an Fe(II)-chelating PHD inhibitor, in phase 2 testing (NCT04556383) reflects this therapeutic bottleneck. Potential on-target toxicities of PHD inhibition are, therefore, well studied and can be assessed by measuring erythropoietin (EPO) and vascular endothelial growth factor (VEGF) levels as useful, established proxies for clinical evaluation.

To address these challenges, we designed a potent PHD inhibitor, ISM012-042, for the treatment of IBD, which exhibits gut-restricted pharmacokinetic (PK) properties, excellent activity and specificity against PHDs. The commercially available generative AI platform Chemistry42 (ref. ¹⁶) and its generative chemistry, Alchemistry binding free energy estimate and ADMET (absorption, distribution, metabolism, excretion and toxicity) profiling modules were used during hit generation, potency optimization and lead optimization processes (Fig. 1a). On the basis of reported complex structures of Takeda-17, JPHM-2-167 and molidustat with PHD2, the privileged fragment (benzonitrile group) and six key pharmacophore points were set to initiate fragment growth in a Chemistry42 structure-based drug design module¹⁶. Compound 1 was generated and identified on the basis of novelty and synthetic accessibility, demonstrating good hydroxylase inhibitory activity against PHD2 with a half-maximal inhibitory concentration (IC₅₀) of 4 nM (ref. ¹⁷). Structure–activity relationship optimization was used to fine-tune candidate compound properties and we then evaluated the potency of the generated compounds using Alchemistry to compute binding free energy estimates for PHD2–ligand complexes (Supplementary Fig. 1b) and predicted properties such as solubility and permeability using ADMET profiling (Supplementary Fig. 1c), further prioritizing low solubility and high clearance to minimize systemic exposure while retaining its moderate permeability for cell penetration (Supplementary Table 1); this procedure led to the selection of ISM012-042 (Supplementary Information 1).

**Fig. 1: Development of a specific and gut-restricted potent PHD inhibitor, ISM012-042.**

In the crystal structure of PHD2 in complex with ISM012-042, two adjacent heteroatoms (heterocyclic N and carbonyl O) of ISM012-042 form a bidentate coordination with the Mn(II) ion, together with H374, H313, D315 and H₂O (Fig. 1b). ISM012-042 exhibited a distinct binding mechanism from 2-OG by inducing a protein conformational change to create an expanded inner subpocket (Fig. 1c). The cyanopyridine group pushes away R383 and Y303, while the cyano N forms a hydrogen bond with N331. As a result, ISM012-042 inhibited PHD1 and PHD2 hydroxylase activity with IC₅₀ values of 1.9 and 2.5 nM, respectively. (Fig. 1d,e). ISM012-042 inhibited PHD3 (Supplementary Fig. 2a), which has been reported to have a protective role during colitis¹⁸, although this function is independent of its enzymatic activity¹⁹.

Next, we evaluated ISM012-042 and GB004 activity across multiple extracellular iron levels in enzymatic and HIF1α–HiBit cellular bioluminescence assays. Whereas GB004’s iron-chelating activity was ablated upon the addition of increasing iron concentrations, ISM012-042’s inhibition of PHD1 was unchanged in an enzymatic assay with IC₅₀ values < 2 nM (Supplementary Fig. 2b,c). Furthermore, ISM012-042 stabilized HIF1α with a half-maximal effective concentration (EC₅₀) of 3.0–3.5 μM across all conditions (Fig. 1f), whereas GB004’s activity was impaired at iron concentrations > 10 μM (Fig. 1g). ISM012-042’s stabilization of HIF1α protein was confirmed by western blot (Supplementary Fig. 2d,e). We also evaluated the target specificity of ISM012-042 and GB004 by measuring the effect on other Fe(II)-dependent and 2-OG-dependent ___domain-containing histone demethylases (Supplementary Table 2). ISM012-042 did not inhibit any of the tested enzymes up to 10 μM, whereas GB004 inhibited these enzymes with IC₅₀ values that were comparable to the PHD1 IC₅₀ (less than 4.4-fold difference), suggesting that ISM012-042 has lower off-target risk concerns.

ISM012-042’s barrier-protective function was tested by measuring the transepithelial electrical resistance (TEER) of Caco-2 monolayers treated with dextran sodium sulfate (DSS), a known barrier-disrupting agent. Caco-2 cells treated with 2.5 μM ISM012-042 were protected from DSS-induced barrier disruption (Fig. 1h), likely because of retention of the tight-junction protein zonula occludens 1 (ZO1) at cell–cell contacts (Supplementary Fig. 2f). To evaluate ISM012-042’s anti-inflammatory function, interleukin 12 (IL-12) and tumor necrosis factor (TNF) mRNA was measured following ISM012-042 treatment of lipopolysaccharide (LPS)-induced murine bone marrow-derived dendritic cells (BMDCs). ISM012-042 dose-dependently reduced IL-12p35, the subunit of the bioactive IL-12, and TNF expression as effectively as the existing PHD inhibitor, roxadustat (Fig. 1i). These data demonstrate that ISM012-042 restores intestinal barrier function and reduces proinflammatory cytokine production in vitro.

ISM012-042’s gut-restricted distribution was confirmed in healthy and oxazolone-induced colitis mice. We observed a 67-fold enrichment in the colon relative to the plasma (area under the curve (AUC)_0−t = 0.3 h nmol ml⁻¹) of healthy mice orally dosed with 10 mg kg⁻¹ ISM012-042 with a similar colon–plasma exposure ratio observed in colitis mice (Fig. 1j,k and Supplementary Table 3), indicating that this distribution was independent of the intestinal barrier integrity or colitis severity. We recapitulated this gut-specific distribution in rats (Supplementary Table 3), which were administered 30 mg 100 μCi⁻¹ kg⁻¹ ¹⁴C-labeled ISM012-042 to trace the distribution kinetics of our compound over 48 h (Fig. 1l). The ¹⁴C-labeled ISM012-042-derived radioactivity was distributed into the small intestine, stomach and cecum as early as 0.5 h after dosing and peaked at 4 h, with most of the compound excreted by 24 h. Excretion was enriched in feces (~85% dose) with roughly 5% of total radioactivity (TRA) in the bile.

Preclinical experiments demonstrated a low drug–drug interaction risk and a favorable safety profile of ISM012-042, which we summarize in Supplementary Table 1. Oral administration of ISM012-042 for 28 days was well tolerated as measured by viability, clinical signs, body weight and clinical pathology, with no adverse findings in rats (Supplementary Fig. 3 and Supplementary Tables 4 and 5) up to 1,000 mg kg⁻¹ day⁻¹. ISM012-042 was not mutagenic in the in vitro genetic toxicity studies and was negative for chromosomal damage in an in vivo rat micronucleus assay. These collective data reiterate the favorable safety profile of ISM012-042.

Next, we investigated the in vivo efficacy of ISM012-042 in the 2,4,6-trinitrobenzene sulfonic acid (TNBS)-induced colitis model²⁰. ISM012-042 (1, 3 or 10 mg kg⁻¹) or the current SOC intervention, mesalamine (100 mg kg⁻¹), was given once daily starting 1 day before TNBS induction (Fig. 2a). ISM012-042 treatment at 3 and 10 mg kg⁻¹ attenuated colitis progression as effectively as 100 mg kg⁻¹ mesalamine as measured by the disease activity index (DAI) (Fig. 2a and Supplementary Tables 6 and 7). A dose-dependent increase in HIF1α staining intensity was observed along the luminal aspect of colon tissues of IBD mice at the experimental endpoint (Fig. 2b). ISM012-042 (3 and 10 mg kg⁻¹) reduced the body weight loss, shortening of colon length and colon weight (Supplementary Fig. 4a–c). The high colon-to-plasma exposure ratio at 4 h after the final dose further confirmed our compound’s gut-restrictive property in this model (Supplementary Fig. 4d). ISM012-042-treated colonic tissue appeared healthier with lower histopathology scores than colonic tissue from vehicle-treated mice (Supplementary Fig. 4e,f). ISM012-042 treatment dose-dependently increased HIF1α transcription of barrier-protective genes trefoil factor 3 (Tff3, also known as Itf), tight-junction protein 1 (Tjp1) and 5′ nucleotidase (Nt5e, also known as Cd73) (Fig. 2c). Treatment of mice with active TNBS-induced colitis with ISM012-042 also promoted remission from colitis symptoms (Fig. 2d). ISM012-042-treated animals displayed significant weight gain and amelioration of disease severity 48 h after treatment (Supplementary Fig. 5a and Fig. 2d). We also observed dose-dependent improvements in colonic gross pathology at the study’s endpoint (Supplementary Fig. 5b,c). An in vivo permeability assay demonstrated that oral ISM012-042 reduced plasma levels of fluorescein isothiocyanate (FITC)–dextran relative to vehicle controls, indicating the functional restoration of the intestinal barrier (Fig. 2e). We, thus, conclude that ISM012-042 alleviates murine colitis by restoring intestinal barrier function.

**Fig. 2: ISM012-042 resolves colitis, restores intestinal barrier function and reprograms the colonic immune environment.**

To test whether ISM012-042 regulates the colonic immune system, we quantified immune cells from the colonic lamina propria by flow cytometry following treatment with ISM012-042. Strikingly, ISM012-042 treatment reduced the proportions of monocytes, neutrophils and proinflammatory CD3⁺ T cell subsets compared to vehicle controls (Fig. 2f and Supplementary Fig. 5d). We performed transcriptomic analysis of colon tissue from ISM012-042 and vehicle-treated mice and observed upregulation of IL-17 signaling upon TNBS challenge, which was reversed by ISM012-042 (3 mg kg⁻¹) treatment on day 2 after colitis induction (Supplementary Fig. 4g). The reduction in IL-17 signaling by ISM012-042 was validated by proteomics analysis of treated colonic mucosa (Supplementary Fig. 4h). Furthermore, we analyzed fecal excrement for lipocalin 2 (LCN2) protein, a downstream target of IL-17 signaling. We found that it was dose-dependently reduced as early as 24 h after treatment (Supplementary Fig. 5e). Fecal calprotectin levels were similarly reduced at this time point. Given this correlation, we selected LCN2 as a clinical pharmacodynamic marker in an ongoing phase 1 trial (NCT06012578) for ISM012-042. Plasma levels of proinflammatory cytokines IL-6 and TNF were significantly reduced by treatment with 3 and 10 mg kg⁻¹ ISM012-042, indicative of ISM012-042’s systemic anti-inflammatory activity to restore healthy colonic immune cell programs (Supplementary Fig. 4i).

Next, we used the oxazolone-induced IBD model, which mimics the T helper cell type 2 response observed in human disease, to evaluate the efficacy of ISM012-042 (Fig. 2g and Supplementary Tables 8 and 9). Prophylactic treatment of ISM012-042 dose-dependently prevented body weight loss (Supplementary Fig. 6a), improved DAI scoring (Fig. 2g), improved colon density (Supplementary Fig. 6b,c) and reduced fecal LCN2 protein relative to vehicle controls (Fig. 2j). Cytokine and chemokine analysis of colon tissue from ISM012-042 and vehicle-treated mice revealed that ISM012-042 treatment reduced proinflammatory cytokines keratinocyte chemoattractant, monocyte chemoattractant protein 1, macrophage inflammatory proteins 1α and 2, TNF and IL-12p70 relative to control mice. Type 2 immunity-related cytokines IL-33 and IL-5 were also reduced in a dose-dependent manner (Fig. 2h). Importantly, circulating EPO and VEGF levels, key downstream targets of noncolonic HIF2α, were unchanged. In contrast, circulating IL-1β levels were decreased (Fig. 2i and Supplementary Fig. 6d). Thus, we conclude that ISM012-042 is safer than existing PHD inhibitors as it exhibited low systemic inhibition of PHDs. We also performed a head-to-head preventive efficacy comparison of ISM012-042 and GB004 using the oxazolone model. Although GB004 treatment led to a trend toward colitis alleviation as assessed by body weight loss and the endpoint colonic histopathology, the observations were not statistically significant because of high individual variation. In contrast, ISM012-042 significantly alleviated colitis in this model (Supplementary Fig. 6e,f). We hypothesize that the variable efficacy observed in the GB004 group may be explained by individual variations in iron concentration. To further verify its value in therapeutic intervention, we also tested initiation of the treatment regimen when the oxazolone-induced colitis model reached maximal disease severity on day 2. ISM012-042 treatment similarly promoted disease remission, restored intestinal permeability and improved colon gross pathology (Fig. 2k–m and Supplementary Fig. 6g,h).

In summary, we report the identification, development and preclinical assessment of a small molecule, ISM012-042, for treating IBD. Our generative AI-powered Chemistry42 drug design platform produced this gut-restricted and safe PHD-specific inhibitor with potently preventative and therapeutic anticolitis activity in multiple clinically relevant murine colitis models. The generative chemistry module identified a novel PHD inhibitor candidate scaffold, the Alchemistry module optimized its potency by refining the molecular structure and the ADMET profiling module modeled the physiochemical and biological properties that led us to select an effective and safe gut-restrictive molecule. The overarching generative chemistry techniques used are tunable and translatable to other molecular design tasks that may be of immediate value to the biomedical research community.

ISM012-042 attenuated colitis by restoring intestinal barrier function and alleviating intestinal inflammation. Unlike previous PHD inhibitors tested for IBD, ISM012-042 inhibited PHDs independently of extracellular iron levels and exerted anticolitis activity at markedly lower doses than SOC treatment mesalamine, thus mitigating known risks of prolonged mesalamine administration²¹. Following the validation of molecular generation conditions and biological mechanisms in multiple preclinical models, ISM012-042 received regulatory approval to start a phase 1 clinical trial (NCT06012578) with the first human volunteers dosed before December 7, 2023.

Methods

Synthesis of ISM012-042

The synthesis of ISM012-042 is laid out in detail in Supplementary Information 1.

In vitro experiments

Crystallization and data analysis

Recombinant PHD2_181–426 protein was expressed in Escherichia coli and purified as previously described²². Protein sample was mixed with crystallization solution composed of 0.2 M NH₄ phosphate monobasic, 0.1 M Tris pH 8.5 and 50% MPD using the sitting drop vapor diffusion method. The crystal structure of PHD2 (P181–E407) cocrystallized with MnCl₂, ISM012-042 and 5% PEG smear appeared after 7 days at 20 °C. A crystal with the proper size was isolated and quickly cooled in liquid nitrogen for data collection. The crystals producing high-quality diffraction spots were chosen for full dataset collection at the Shanghai Synchrotron Radiation Facility synchrotron source. The diffraction data were indexed and integrated with XDS (version January 31, 2020) and scaled by aimless (CCP4 7.1). The structure was solved by molecular replacement with Phaser (CCP4 7.1). Refinement was carried out through multiple rounds with Coot (version 0.9.5) and refmac5 (CCP4 7.1). Crystallographic parameters and data collection statistics are described in Supplementary Table 10.

ADMET profiling

Metabolic stability in liver microsomes

ISM012-042 at 1.00 µM was incubated with CD1 mouse, SD rat, beagle dog, cynomolgus monkey and human liver microsomes with a reduced nicotinamide adenine dinucleotide phosphate (NADPH)-regenerating system in a 37.0 °C water bath for up to 60 min. The samples at 0, 5, 15, 30, 45 and 60 min were analyzed using liquid chromatography coupled with tandem mass spectrometry (LC–MS/MS; SCIEX API400 MS instrument coupled to a Waters ACQUITY ultrahigh-performance LC (UPLC) instrument). The residual percentage and the clearance rate of ISM012-042 were calculated from the measured concentration.

Permeability assay

Caco-2 cells at passage number 42 were seeded on 96-well transport inserts and cultured for 21 days before being used. ISM012-042 was dosed bidirectionally at 1.00 µM with or without 10.0 µM GF120918, a potent inhibitor of efflux transporters such as P-glycoprotein and breast cancer resistance protein. The plate was incubated for 2 h at 37.0 °C with 5.0% CO₂ at saturated humidity without shaking. Samples were taken at 0 and 120 min after incubation and analyzed by LC–MS/MS. The apparent permeability (P_app, cm s⁻¹) was calculated as follows, VR is the solution volume in the receiver chamber, CR is the final concentration of the test compound in the receiver chamber and C₀ is the initial concentration of the test compound in the donor chamber:

$${P_{\rm{app}}}=\frac{\rm{VR}}{\rm{Area}\times {Time}}\times \frac{\left[{\rm{drug}}\right]{\rm{receiver}}}{\left[{\rm{drug}}\right]{\rm{initial},{donor}}}=\frac{\rm{VR}}{\rm{Area}\times {Time}}\times \frac{\rm{CR}}{C_0}$$

Cytochrome P450 inhibition

Pooled human liver microsomes were incubated in the presence of ISM012-042 (at 0, 0.000300, 0.00100, 0.00300, 0.0100, 0.0300, 0.100, 0.300 and 1.00 µM), NADPH and a selective substrate for each cytochrome P450 isoform (midazolam and testosterone) at 37 °C in a water bath. Reactions were terminated by adding 200 µl of stop solution. Formation of the selective metabolite from its substrate was determined by LC–MS/MS and reported as enzyme activities (pmol min⁻¹ mg⁻¹ protein) of each cytochrome P450 isoform. IC₅₀ values were determined using a three-parameter or four-parameter logistic equation.

PK analysis of ISM012-042

All animal experiments were conducted by WuXi AppTec. All animals received a single oral dose after fasting overnight with free access to water and food was provided at 4 h after dosing. The blood samples were collected in K₂ EDTA-coated tubes at 0.25, 0.5, 1, 2, 4, 6 and 8 h after dosing. Each group was killed by CO₂ inhalation at 1, 2, 4 and 8 h after dosing and the colon tissue was collected. The dissected colon was washed with saline and homogenized by adding five volumes (w/v, 1:5) of precooled methanol and PBS (v/v, 1:2). The plasma and colon homogenate concentrations were determined by protein precipitation and LC–MS/MS detection (SCIEX Triple Quad 6500+ or 7500 MS instrument coupled to a Waters ACQUITY UPLC I-Class PLUS system). The PK parameters of plasma and colon were then calculated using noncompartmental analysis (Phoenix WinNonlin software, version 8.3.5, Certara). The protocol and any amendments or procedures involving the care or use of animals on this study were reviewed and approved (assurance identification numbers PK01-001-2019v1.0 and NJ-20220531-Rats) by the WuXi AppTec Institutional Animal Care and Use Committee (IACUC) before the initiation of such procedures.

Tissue distribution analysis of ¹⁴C-labeled ISM012-042

Sprague–Dawley (SD) rats were divided into seven groups (n = 3 per sex per time point) and administered a single oral dose of ¹⁴C-labeled ISM012-042 at 30 mg 100 μCi⁻¹ kg⁻¹. Four groups were anesthetized by isoflurane inhalation at 0.5, 4, 8 and 24 h after dosing. The heart, lung, liver, spleen, kidney, whole brain, small intestine, cecum, colon, rectum, stomach wall, skeletal muscle, body fat, bone marrow, reproductive organs (testis and epididymis for males; uterus and ovaries for females) and blood samples were collected. Three groups were anesthetized at 2, 12 and 48 h after dosing, after which the small intestine, cecum, colon and rectum were collected. A total of 12 SD rats (six males and six females) were administered a single oral dose of ¹⁴C-labeled ISM012-042 at 30 mg 100 µCi⁻¹ kg⁻¹. Urine, feces and cage wash up to 168 h after dosing were collected from six intact rats (three males and three females) for mass balance; bile, urine, feces and cage wash up to 72 h after dosing were collected from six bile-duct-cannulated rats (three males and three females) for biliary excretion. The TRA of plasma, bile, urine and cage wash samples was directly analyzed by liquid scintillation counting (LSC; PerkinElmer Instruments). Blood samples, fecal samples homogenized in 50% isopropanol in water (v/v), skeletal muscle, body fat, bone marrow and any other tissues were homogenized as needed and mixed with lytic agent SOLVABLE, heated in a water bath and analyzed by LSC. The blood samples were mixed with a hydrogen peroxide solution in water (7:93, v/v) before cocktail scintillation. Colon samples were homogenized on wet ice using homogenizing buffer (15 mmol L⁻¹ phosphate buffer and methanol, 2:1, v/v) in a ratio of 1:5 (1 g of colon with 5 ml of buffer). The TRA values determined in bile, urine, feces and cage wash were used to calculate the recovery percentage of the administered dose.

AlphaScreen PHD hydroxylase assay

Compound screening assays were performed at Pharmaron. The final concentrations of the reaction for the respective PHD isoforms included recombinant PHD1 (Active Motif, 81064), PHD2 (Active Motif, 81065) or PHD3 (Active Motif, 81033), 2 μM 2-OG for PHD1 and PHD2 or 10 μM 2-OG for PHD3, biotinylated CODD peptide (Sangon; 3:1 ratio of enzyme to CODD peptide), l-ascorbic acid, 10 μM Fe(II) and inhibitors with 2% DMSO. Preincubated donor–acceptor bead mix (AlphaScreen streptavidin-conjugated donor and protein A-conjugated acceptor beads; PerkinElmer) with anti-HIF1α hydroxy-P546 antibody (Cell Signaling, 3434) was then added to the reaction mixture for 30 min at room temperature. The reaction was quenched with 5 μl of 30 mM EDTA. The luminescence signal was measured using an Envision (Perkin Elmer) plate reader. The data were analyzed using GraphPad Prism 8.0. For the examination of Fe(II)-dependent activities, the final Fe(II) concentration in the reaction was changed to the indicated level.

Enzymatic assay for FIH hydroxylase activity and JMJD2D, JMJD2E, JMJD1B and UTX demethylase activity

Protocols for the in vitro HTRF hydroxylase assay²³ and the AlphaLISA histone demethylase assays²⁴ were reported previously. For the FIH (abcam, ab86916) hydroxylase assay, the final Fe(II) reaction concentration was 5 μM. For the lysine demethylase assays, recombinant JMJD2D (Active Motif, 31459), JMJD2E (BPS Bioscience, 50118), JMJD1B (Active Motif, 31429) or UTX (Active Motif, 31460) was incubated with Fe(II) at 2 μM.

Cell lines

Caco-2 with HiBiT-tagged HIF1α was constructed by inserting the HiBiT coding sequence (Wuxi AppTec) at the C terminus of the endogenous HIF1A gene in Caco-2 cells (American Type Culture Collection, HTB-37) using clustered regularly interspaced short palindromic repeats and Cas9 and cultured in Eagle’s minimum essential medium containing 20% FBS and 1% penicillin–streptomycin at 37 °C with 5% CO₂.

Mycoplasma testing

All cell lines were checked for Mycoplasma by WuXi AppTec and Shanghai Genechem.

Cell culture conditions

All cell lines were cultured at 37 °C in a humidified atmosphere of 95% air and 5% CO₂.

Detection of HiBiT tag

Caco-2 HIF1α–HiBiT cells were seeded in a 384-well plate (5.5 × 10³ cells per well for 48 h). The cells were treated with compound (10 μM, ten doses, in duplicate for ISM012-042 or GB004, freshly prepared before added to the medium) for 6 h. Cells were loaded with 0, 2, 10 or 50 μM Fe(II) sulfate in the medium for 20 min before DMSO or compound addition. HIF1α–HiBiT abundance in Caco-2 HIF1α–HiBiT cells was measured after 6 h through incubation with the Nano-Glo HiBiT lytic detection system and luminescence was measured using Envision. The HIF1α–HiBiT abundance in each well was normalized to the abundance of DMSO as a function of relative luminescence units (RLU). Abundance was measured in independent triplicates. After normalization, RLU values were used for regression analysis and EC₅₀ curve fitting using GraphPad Prism.

TEER assay

Caco-2 cells were seeded at 5.0 × 10⁴ cells per well within a 24-well transwell plate with 0.4-µm-pore polyester membrane inserts. The medium was replaced every 2 or 3 days. The TEER values of Caco-2 cell monolayers were monitored using a Millipore Millicell ERS-2 voltohmmeter once daily before monolayer establishment. When the TEER values reached a plateau (about 14 days after seeding) and were stable for the following 3–4 days, an efficient barrier function was considered established and ready for compound treatment. The monolayers were treated with ISM012-042 (1 or 2.5 μM) or DMSO in six replicates for 24 h before 3% DSS (w/v) treatment in both compartments of the transwell membrane for another 24 h along with the compound treatment. TEER values were obtained at 0, 6, 24, 48 and 96 h after the addition of DSS.

Immunoblotting

Confluent Caco-2 cells were incubated for 6 h in a medium supplemented with inhibitor, roxadustat positive control or DMSO. Then, 20 μM MG-132 (MedChemExpress, HY-13259) was added to Caco-2 cells for the last 4 h of a 6-h treatment with PHD inhibitors or DMSO. The hydroxy-HIF1α and total HIF1α proteins in the whole-cell lysates were assessed by western blotting and detected using anti-HIF1α (P564) (Cell Signaling Technology, 3434; 1:1,000) and anti-HIF1α (Cell Signaling Technology, 36169S; 1:1,000) antibodies with glyceraldehyde 3-phosphate dehydrogenase (GAPDH) as a loading control (Merck, MAB374; 1:2,000). GAPDH band intensities were used to normalize total HIF1α or hydroxylated signals. The blots were visualized with an ImageQuant LAS 4000 camera (GE Healthcare).

Immunofluorescence cell staining of tight-junction protein

Briefly, Caco-2 cells were seeded on 14-mm-diameter coverslips precoated with 0.1% gelatin at 1 × 10⁶ cells per well within a 12-well cell culture plate and pretreated with 3 μM ISM012-042 or DMSO for 1 h before tight-junction damage by 3% DSS (MP Biomedical, 216011050). After 24-h DSS treatment, cells were washed with cold PBS, fixed in methanol (prechilled at −20 °C) for 3 min and then incubated with rabbit polyclonal anti-ZO1 (D6L1E) rabbit monoclonal antibody (Cell Signaling Technology, 13663; 1:100) for 1 h. Cells were incubated with FITC-labeled secondary antibodies (AffiniPure goat anti-rabbit IgG (H + L), Jackson, 111-095-003; 1:50) for 1 h. DAPI (Invitrogen, D3571) was used to label DNA in nuclei. Slides were mounted on glass slides using the Prolong Gold antifade reagent (Invitrogen, P36970) and visualized using a confocal fluorescence microscope (Leica TCS SP8 MP Multiphoton Microscope, Leica Microsystem). Each figure (×68 magnification) represents eight fields captured for each coverslip. The mean fluorescence intensity of each figure was analyzed using ImageJ version 1.54g (National Institutes of Health).

BMDC culture and qPCR assay

Bone marrow cells were flushed from the femurs and tibiae of C57BL/6J mice with 0.5% BSA and 2 mM EDTA pH 8.0 in 1× PBS, passed through a nylon mesh and treated with ACK lysing buffer (Life Technologies, A10492-01) for 2 min. The cells were then washed and cultured in nonadherent Petri dishes with DMEM/F12 supplemented with GlutaMAX, 10% heat-shocked FBS (Gibco, A56697-01), 1% penicillin–streptomycin (Life Technologies, 15140122), 1% sodium pyruvate (Life Technologies, 11360070), 1% HEPES (Life Technologies, 15630106), 1% MEM nonessential amino acids (Life Technologies, 11140050) and 20 ng ml⁻¹ granulocyte–macrophage colony-stimulating factor (GM-CSF; StemCell Technologies, 78017). The medium was replaced on days 3 and 5 with recombinant mouse GM-CSF to induce differentiation. By day 5, over 90% of the cells expressed dendritic cell-specific markers, as confirmed by fluorescence-activated cell sorting (FACS). On day 6, DCs were replated into new 24-well plates and treated with ISM012-042 (1, 3 or 10 µM) or roxadustat (10 µM). Control and treated BMDCs were also stimulated with LPS (100 ng ml⁻¹) for 24 h. Subsequently, RNA extraction was performed using a Tiangen RNA extraction kit (DP761), followed by reverse transcription (RT) with PrimeScript RT Master Mix (Takara, RR036A). RT–qPCR was conducted using PowerUp SYBR green mix (Applied Biosystem, A25742) on a QuantStudio 6 Flex real-time PCR system (Applied Biosystems), with mRNA levels normalized to Gapdh using the 2^−ΔΔCt method. All reactions were performed in triplicate. Primer sequences are provided in Supplementary Table 11.

Animal studies

General toxicity

Ten SD rats per sex per group (purchased from Vital River Laboratory Animal Technology) were used in 28-day repeat-dose general toxicity studies. The rats were 7–9 weeks old and the body weights ranged from 165 to 202 g for females and 300 to 350 g for males at initiation of dosing. During the study, the rats had ad libitum access to food (Beijing/Tianjin Keao Xieli Feed, 22073213 and 22113213) and water. The rats were housed in a temperature-controlled (20–24 °C) with 40–70% humidity and a 12-h light–dark cycle. ISM012-042 was suspended in 0.5% (v/v) Tween 80 with 0.5% (w/v) carboxymethyl cellulose (CMC)-32 solution at concentrations to provide a 0, 100, 300 or 1,000 mg kg⁻¹ dose in a 10 ml kg⁻¹ dose volume by oral gavage daily for 28 days. The toxicity assessment covered viability, clinical signs, body weights, food consumption, ophthalmologic examinations, clinical pathology (hematology, coagulation parameters, serum chemistry and urinalysis), gross necropsy observations, organ weight and histopathology. The protocol and any amendments or procedures involving the care or use of animals on this study were reviewed and approved (assurance identification number SZ20221201-Rats-A) by the WuXi AppTec IACUC before the initiation of such procedures.

Mice and chemically induced colitis

For the TNBS colitis model, female BALB/c mice aged 8 weeks (body weight: ~19–21 g) were purchased from Lingchang Biotech and were bred in specific pathogen-free individually ventilated cages in a temperature-controlled (20 ± 2 °C) room with 40–70% humidity and a 12-h light–dark cycle. Chow pellets and tap water were available ad libitum. The mice were acclimated at the animal facility for at least 3 days before the experiments. On day 0, mice were anesthetized and received a 100-μl rectal instillation of either vehicle (50% ethanol) or TNBS (2% TNBS in 50% ethanol). Daily treatment with compound or vehicle (0.5% CMC–Na plus 2% Tween 80) by oral gavage occurred on days −1 (prevention mode) or day 2 (therapeutic mode) through day 7.

For the oxazolone colitis model, female BALB/c mice aged 8 weeks (body weight: ~18–20 g) were purchased from Charles River Laboratories and bred in specific pathogen-free individually ventilated cages. On day −5, mice were presensitized with 200 μl of 2% oxazolone dissolved in a 4:1 mixture of acetone and olive oil on a 4-cm² field of the shaved back, followed by intracolonic injection of 100 μl of either oxazolone (1.5% oxazolone dissolved in 50% ethanol) or vehicle (50% ethanol) on day 0. Before intracolonic injection, mice were anesthetized with 0.3 ml of avertin (1.25%) and maintained in the head-down position for 2 min following intracolonic administration. All experimental protocols were approved (assurance identification numbers IM01-005-2021v1.0 and IM01-SH003-2023v1.1) by Wuxi AppTec IACUC.

Evaluation of colitis severity

Body weight and DAI soring were recorded daily to assess colitis severity. DAI score was the sum of the weight loss, stool and bleeding subscores. A blinded scoring system was used to determine colitis severity. The DAI scorer, blinded to the group information and animal identifier, was responsible for the stool consistency and bleeding evaluations. DAI scoring standards are presented in Supplementary Table 12.

FOB test

If no blood was visible with the naked eye, the fecal occult blood (FOB) test was performed with an FOB test strip (the improved pyramidon method). The FOB score (0–2) was interpreted as follows: 0, no color appeared after 2 min; 1, a dimmed color appeared after 1–2 min; 2, a deep color appeared after 1–2 min.

Intestinal permeability assay

FITC–dextran (molecular weight, 4,000; 600 mg kg⁻¹; Sigma-Aldrich, 46944) was administered by oral gavage after 4–6 h of fasting. Then, 4 h after dosing with FITC–dextran, mice were anesthetized and blood was sampled in K₃ EDTA-coated tubes. Plasma was isolated by centrifugation at 2,000g for 10 min at 4 °C. FITC–dextran concentrations in plasma were analyzed in duplicate using a spectrophotometer with an excitation wavelength of 485 nm and emission wavelength of 535 nm. Baseline blood plasma fluorescence was determined in mice after oral gavage with water and subtracted from fluorescence obtained after FITC–dextran gavage. FITC–dextran concentrations were determined from standard curves generated by serial dilutions of FITC–dextran.

RNA extraction and real-time qPCR

Whole-colon tissue from each mouse was snap-frozen in liquid nitrogen and stored at −80 °C until analysis. Samples were homogenized with a Qiagen TissueLyser LT in TRI reagent and total RNA was purified using the RNeasy mini kit (Qiagen). RNA was converted into complementary DNA using the FastKing RT kit (Tiangen) according to the manufacturer’s instructions. qPCR was performed on a QuantStudio 7 Flex real-time PCR system (Applied Biosystems) using SYBR green master mix (Applied Biosystems). The relative expression was normalized to Gapdh using the 2^−ΔΔCt method. Primers sequences are listed in Supplementary Table 11.

Flow cytometry

Lamina propria lymphocytes were isolated from the intestines of TNBS mouse models and subjected to antibody staining and flow cytometry as previously described²⁵. Cell suspensions were washed with 1× PBS and stained for live–dead markers and surface antibodies on ice for 30 min, followed by a wash with FACS buffer (0.5% BSA and 2 mM EDTA in 1× PBS). Cells were stimulated with a cell stimulation cocktail plus transporter inhibitors for 4 h before staining to detect cytokines. Intracellular staining was performed using an intracellular labeling kit, with cells fixed, permeabilized and then stained with antibodies in 1× perm washing buffer for 30 min at 4 °C. Intracellular antibodies included interferon-γ, IL-17A, forkhead box P3, TNF and retinoic acid-related orphan receptor-γt, while all other antibodies were used for surface staining. All antibodies were used at dilution of 1:50. Sample collection and data analysis were performed using a BD Fortessa flow cytometry analyzer and FlowJo software (version 10). Details of antibodies and reagents are provided in Supplementary Table 13.

Lymphocytes were initially selected on the basis of forward scatter (FSC) and side scatter (SSC) and doublets were excluded using FSC-A/FSC-W and SSC-A/SSC-H plots. Dead cells were excluded by gating on the FVS620^low population. CD45⁺ cells were then selected to identify pan-leukocytes.

Histopathology

The colon was removed and its length was measured when it was in a relaxed position without stretching; it was weighed after feces removal. The colon was opened longitudinally, washed thoroughly with PBS, rolled from the proximal to distal end, neutralized and fixed in paraformaldehyde, followed by staining with hematoxylin and eosin. Slides were scored by pathologists blinded to the animal identifier. The histological features that were assessed are listed in Supplementary Table 14. A total histological severity score, ranging from 0 to 11, was obtained by summing the eight-item scores.

HIF1α immunohistochemistry and quantification

Colon Swiss rolls embedded in formalin-fixed paraffin-embedded tissue block were sectioned into 4-μm-thick slides. The slides were deparaffinized, rehydrated and incubated in epitope retrieval solution 1 (pH 6) for 20 min at 100 °C. The primary antibody to HIF1α clone E1V6A (1:400; Cell Signaling Technologies, #48085) was incubated for 30 min at room temperature followed by bond polymer refine detection (Leica Biosystems, DS9800). Slides were digitized using Aperio VERSA 8 (Leica), and image acquisition was performed using Aperio ImageScope version 12.4 (Leica). Eight nonoverlapping fields of view per each colon Swiss-roll section were analyzed by HALO software (Indica Labs). For each colon Swiss-roll section, the percentage of HIF1α and nucleus double-stained counts in total cell counts (12,000–14,000 cells counted from eight fields of view) were quantitated for the mucosa layer at the midsegment to the distal end of the colon roll at ×200 magnification.

Cytokine and chemokine analysis

Whole-colon or half-colon tissue was cut longitudinally and homogenized and colonic cytokines were measured using the V-PLEX proinflammatory panel one mouse kit (Meso Scale Discovery, K15048D) and V-PLEX cytokine panel one mouse kit (Meso Scale Discovery, K15245D). Plasma EPO, VEGFA, IL-1β, IL-6 and TNF were measured using customized U-PLEX biomarker group 2 assays (Meso Scale Discovery, K152ADM).

Fecal biomarker measurement

Mice were placed in clean polycarbonate cages without bedding until they defecated or for 1 h, whichever came first. The samples were homogenized in radioimmunoprecipitation assay buffer or other lysis buffer containing protease inhibitor with a 1:5 (w/v) ratio after weighing followed by sonication and centrifugation at 13,800g for 10 min at 4 °C. Total protein concentration of supernatant from fecal samples was determined using a BCA protein assay kit (Thermo Fisher Scientific, 23225) for the final normalization. Calprotectin and LCN2 were detected using the respective ELISA kits (Elabscience, E-EL-M1143c; Beyotime, PN757).

RNA-seq data processing

Around 100 mg of colon tissue was collected from sham or TNBS-induced colitis mouse models treated with vehicle or ISM012-042 on day 2 after colitis induction. The mRNA isolation and library preparation for next-generation sequencing were conducted described above. Libraries were sequenced using the Illumina Novaseq. Raw sequencing data obtained were processed using the DRAGEN (version 4.0.3) pipeline, which aligned the reads against the National Center for Biotechnology Information (NCBI) RefSeq mouse genome (mm10) to generate count data. The quality of sequencing and alignments was assessed using MultiQC (version 1.12)²⁶. Differential gene expression analysis was performed using the DESeq2 package (version 1.34.0) with R (version 4.1.2)²⁷. Subsequently, the log₂ fold changes derived were used for gene set enrichment analysis conducted with the fgsea package (version 1.20.0) with R (version 4.1.2)²⁸ and the P value was adjusted using the Benjamini–Hochberg correction method. The enrichment analysis used gene sets from the Kyoto Encyclopedia of Genes and Genomes pathway database²⁹.

Quantitative MS

Colonic mucosa obtained from TNBS-induced colitis models treated with GB004, ISM012-042 or vehicle were homogenized and assayed by tandem mass tag MS, as described previously³⁰ and analyzed using the MASCOT engine (Matrix Science, version 2.2) embedded in Proteome Discoverer 1.4 software against the UniProt mouse database (17,097 sequences; downloaded on October 4, 2022) for identification and quantitation. Differentially expressed proteins were identified using a two-sided t-test in R (version 4.1.2).

Statistics

Experiments were repeated at least three times, with one representative dataset shown. Data are presented as the mean ± s.d. Statistical analyses were performed using GraphPad Prism versions 8.0 and 10.1. A P value < 0.05 was considered statistically significant. The Shapiro–Wilk test or Kolmogorov–Smirnov test was used to test the normality of data distribution. A two-sided, unpaired Mann–Whitney U-test was used for comparisons between two groups. One-way or two-way analyses of variance (ANOVAs) with Tukey’s or Dunnett’s multiple-comparisons tests were used with the P value adjusted for multiple comparisons as indicated in the figure legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

All data generated in this study underlying all reported figures are provided in the main text and Supplementary Information. The RNA-seq dataset generated from this study was deposited to the NCBI Gene Expression Omnibus under accession number GSE280622. Illumina sequencing reads were aligned to the GRCm38.p6 (mm10) genome using a RefSeq annotation gff file retrieved from the NCBI on 6 July 2022 (GCF_000001635.26). The quantitative MS proteomics data searched against the UniProt mouse database (17,097 sequences; downloaded on October 4, 2022) were deposited to the ProteomeXchange Consortium through the PRIDE partner repository with dataset identifier PXD057486. Atomic coordinates and structure factors for the cocrystal structure were deposited to the PDB under accession number 8Y0V. Source data are provided with this paper.

Code availability

Chemistry42 is an industry-grade commercial software platform available at http://chemistry42.com. All computer code used in this study is available from GitHub (https://github.com/xcxtx8/ISM012-042-PHDi).

References

GBD 2017 Inflammatory Bowel Disease Collaborators. The global, regional, and national burden of inflammatory bowel disease in 195 countries and territories, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet Gastroenterol. Hepatol. 5, 17–30 (2020).
Gibble, T. H. et al. Identification of inadequate responders to advanced therapy among commercially-insured adult patients with Crohn’s disease and ulcerative colitis in the United States. BMC Gastroenterol. 23, 63 (2023).
Article CAS PubMed PubMed Central Google Scholar
Rudbaek, J. J. et al. Deciphering the different phases of preclinical inflammatory bowel disease. Nat. Rev. Gastroenterol. Hepatol. 21, 86–100 (2024).
Article PubMed Google Scholar
Kaser, A., Zeissig, S. & Blumberg, R. S. Inflammatory bowel disease. Annu. Rev. Immunol. 28, 573–621 (2010).
Article CAS PubMed PubMed Central Google Scholar
Turner, D. et al. STRIDE-II: an update on the Selecting Therapeutic Targets in Inflammatory Bowel Disease (STRIDE) initiative of the International Organization for the Study of IBD (IOIBD): determining therapeutic goals for treat-to-target strategies in IBD. Gastroenterology 160, 1570–1583 (2021).
Article CAS PubMed Google Scholar
Bryant, R. V., Winer, S., SPL, T. & Riddell, R. H. Systematic review: histological remission in inflammatory bowel disease. Is ‘complete’ remission the new treatment paradigm? An IOIBD initiative. J. Crohns Colitis 8, 1582–1597 (2014).
Article CAS PubMed Google Scholar
Ren, F. et al. A small-molecule TNIK inhibitor targets fibrosis in preclinical and clinical models. Nat. Biotechnol. https://doi.org/10.1038/s41587-024-02143-0 (2024).
Article PubMed Google Scholar
Pun, F. W., Ozerov, I. V. & Zhavoronkov, A. AI-powered therapeutic target discovery. Trends Pharmacol. Sci. 44, 561–572 (2023).
Article CAS PubMed Google Scholar
Pun, F. W. et al. Identification of therapeutic targets for amyotrophic lateral sclerosis using PandaOmics—an AI-enabled biological target discovery platform. Front. Aging Neurosci. 14, 914017 (2022).
Article CAS PubMed PubMed Central Google Scholar
Pun, F. W. et al. Hallmarks of aging-based dual-purpose disease and age-associated targets predicted using PandaOmics AI-powered discovery engine. Aging (Albany NY) 14, 2475–2506 (2022).
Article CAS PubMed Google Scholar
Olsen, A. et al. Identification of dual-purpose therapeutic targets implicated in aging and glioblastoma multiforme using PandaOmics—an AI-enabled biological target discovery platform. Aging (Albany NY) 15, 2863–2876 (2023).
CAS PubMed Google Scholar
Lun, J., Zhang, H., Guo, J., Yu, M. & Fang, J. Hypoxia inducible factor prolyl hydroxylases in inflammatory bowel disease. Front. Pharmacol. 14, 1045997 (2023).
Article CAS PubMed PubMed Central Google Scholar
Van Welden, S., Selfridge, A. C. & Hindryckx, P. Intestinal hypoxia and hypoxia-induced signalling as therapeutic targets for IBD. Nat. Rev. Gastroenterol. Hepatol. 14, 596–611 (2017).
Article PubMed Google Scholar
Singhal, R. & Shah, Y. M. Oxygen battle in the gut: hypoxia and hypoxia-inducible factors in metabolic and inflammatory responses in the intestine. J. Biol. Chem. 295, 10493–10505 (2020).
Article CAS PubMed PubMed Central Google Scholar
Danese, S. et al. Randomised clinical trial: a phase 1b study of GB004, an oral HIF-1α stabiliser, for treatment of ulcerative colitis. Aliment. Pharmacol. Ther. 55, 401–411 (2022).
Article CAS PubMed PubMed Central Google Scholar
Ivanenkov, Y. A. et al. Chemistry42: an AI-driven platform for molecular design and optimization. J. Chem. Inf. Model. 63, 695–701 (2023).
Article CAS PubMed PubMed Central Google Scholar
Xu, J. et al. Discovery of novel and potent prolyl hydroxylase ___domain-containing protein (PHD) inhibitors for the treatment of anemia. J. Med. Chem. 67, 1393–1405 (2024).
Article CAS PubMed Google Scholar
Chen, Y. et al. PHD3 stabilizes the tight junction protein occludin and protects intestinal epithelial barrier function. J. Biol. Chem. 290, 20580–20589 (2015).
Article CAS PubMed PubMed Central Google Scholar
Xu, Y. M. et al. Prolyl hydroxylase 3 controls the intestine goblet cell generation through stabilizing ATOH1. Cell Death Differ. 27, 2131–2142 (2020).
Article CAS PubMed PubMed Central Google Scholar
Wirtz, S., Neufert, C., Weigmann, B. & Neurath, M. F. Chemically induced mouse models of intestinal inflammation. Nat. Protoc. 2, 541–546 (2007).
Article CAS PubMed Google Scholar
Sehgal, P., Colombel, J. F., Aboubakr, A. & Narula, N. Systematic review: safety of mesalazine in ulcerative colitis. Aliment. Pharmacol. Ther. 47, 1597–1609 (2018).
Article CAS PubMed Google Scholar
Chowdhury, R. et al. Structural basis for binding of hypoxia-Inducible factor to the oxygen-sensing prolyl hydroxylases. Structure 17, 981–989 (2009).
Article CAS PubMed Google Scholar
Koivunen, P., Hirsilä, M., Günzler, V., Kivirikko, K. I. & Myllyharju, J. Catalytic properties of the asparaginyl hydroxylase (FIH) in the oxygen sensing pathway are distinct from those of its prolyl 4-hydroxylases. J. Biol. Chem. 279, 9899–9904 (2004).
Article CAS PubMed Google Scholar
Rose, N. R. et al. Plant growth regulator daminozide is a selective inhibitor of human KDM2/7 histone demethylases. J. Med. Chem. 55, 6639–6643 (2012).
Article CAS PubMed PubMed Central Google Scholar
Kim, E., Tran, M., Sun, Y. & Huh, J. R. Isolation and analyses of lamina propria lymphocytes from mouse intestines. STAR Protoc. 3, 101366 (2022).
Article CAS PubMed PubMed Central Google Scholar
Ewels, P., Magnusson, M., Lundin, S. & Käller, M. MultiQC: summarize analysis results for multiple tools and samples in a single report. Bioinformatics 32, 3047–3048 (2016).
Article CAS PubMed PubMed Central Google Scholar
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).
Article PubMed PubMed Central Google Scholar
Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl Acad. Sci. USA 102, 15545–15550 (2005).
Article CAS PubMed PubMed Central Google Scholar
Kanehisa, M., Furumichi, M., Tanabe, M., Sato, Y. & Morishima, K. KEGG: new perspectives on genomes, pathways, diseases and drugs. Nucleic Acids Res. 45, D353–d361 (2017).
Article CAS PubMed Google Scholar
Zhu, Y. et al. Proteomic analysis of solid pseudopapillary tumor of the pancreas reveals dysfunction of the endoplasmic reticulum protein processing pathway. Mol. Cell. Proteomics 13, 2593–2603 (2014).
Article CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

We thank L. Qin, Q. Meng, C. Zhu and R. Yan for their contributions to this study. We also thank the contract research organization, WuXi AppTec, for performing the in vitro and in vivo experiments.

Author information

Authors and Affiliations

Insilico Medicine US, Inc., Boston, MA, USA
Yanyun Fu, Xiao Ding, Man Zhang, Chunlei Feng, Ziqi Yan, Feng Wang, Jianyu Xu, Xiaoxia Lin, Xiaoyu Ding, Ling Wang, Yaya Fan, Taotao Li, Yushu Yin, Xing Liang, Chenxi Xu, Shan Chen, Fadi E. Pulous, David Gennert, Frank W. Pun, Petrina Kamya, Feng Ren, Alex Aliper & Alex Zhavoronkov
Insilico Medicine Shanghai, Ltd., Shanghai, China
Yanyun Fu, Xiao Ding, Man Zhang, Chunlei Feng, Ziqi Yan, Feng Wang, Jianyu Xu, Xiaoxia Lin, Xiaoyu Ding, Ling Wang, Yaya Fan, Taotao Li, Yushu Yin, Xing Liang, Chenxi Xu, Shan Chen, Feng Ren & Alex Zhavoronkov
Insilico Medicine Hong Kong, Ltd., Hong Kong, China
Yanyun Fu, Man Zhang, Frank W. Pun, Feng Ren & Alex Zhavoronkov
Insilico Medicine AI, Ltd., Abu Dhabi, UAE
Yanyun Fu, Xiao Ding, Feng Ren & Alex Aliper
Insilico Medicine US, Inc., New York, NY, USA
Fadi E. Pulous, David Gennert & Alex Zhavoronkov
Insilico Medicine Canada, Inc., Montréal, Québec, Canada
Petrina Kamya

Authors

Yanyun Fu
View author publications
Search author on:PubMed Google Scholar
Xiao Ding
View author publications
Search author on:PubMed Google Scholar
Man Zhang
View author publications
Search author on:PubMed Google Scholar
Chunlei Feng
View author publications
Search author on:PubMed Google Scholar
Ziqi Yan
View author publications
Search author on:PubMed Google Scholar
Feng Wang
View author publications
Search author on:PubMed Google Scholar
Jianyu Xu
View author publications
Search author on:PubMed Google Scholar
Xiaoxia Lin
View author publications
Search author on:PubMed Google Scholar
Xiaoyu Ding
View author publications
Search author on:PubMed Google Scholar
Ling Wang
View author publications
Search author on:PubMed Google Scholar
Yaya Fan
View author publications
Search author on:PubMed Google Scholar
Taotao Li
View author publications
Search author on:PubMed Google Scholar
Yushu Yin
View author publications
Search author on:PubMed Google Scholar
Xing Liang
View author publications
Search author on:PubMed Google Scholar
Chenxi Xu
View author publications
Search author on:PubMed Google Scholar
Shan Chen
View author publications
Search author on:PubMed Google Scholar
Fadi E. Pulous
View author publications
Search author on:PubMed Google Scholar
David Gennert
View author publications
Search author on:PubMed Google Scholar
Frank W. Pun
View author publications
Search author on:PubMed Google Scholar
Petrina Kamya
View author publications
Search author on:PubMed Google Scholar
Feng Ren
View author publications
Search author on:PubMed Google Scholar
Alex Aliper
View author publications
Search author on:PubMed Google Scholar
Alex Zhavoronkov
View author publications
Search author on:PubMed Google Scholar

Contributions

A.Z. and F.R. conceptualized the study, analyzed the data and wrote the paper. M.Z., Y. Fu, Xiao Ding, C.F., F.W., Z.Y., J.X., Xiaoxia Lin, Xiaoyu Ding, L.W., Y. Fan, T.L., Y.Y., Xing Liang, C.X. and S.C. intellectually or manually contributed to the experimental design for assays carried out in this study. F.W.P. and A.A. analyzed the data included in this study. All authors discussed the results and commented on the paper.

Corresponding author

Correspondence to Alex Zhavoronkov.

Ethics declarations

Competing interests

Insilico Medicine is a company developing an AI-based end-to-end integrated pipeline for drug discovery and development that is engaged in drug discovery programs for aging, fibrosis and oncology. Y. Fu, Xiao Ding, C.F., Z.Y., F.W., M.Z., J.X., Xiaoxia Lin, Xiaoyu Ding, L.W., Y. Fan, T.L., Y.Y., Xing Liang, C.X., S.C., F.E.P., D.G., F.W.P., P.K., F.R., A.A. and A.Z. are affiliated with Insilico Medicine. Xiao Ding, J.X. and F.R. are inventors of patent WO2023072257A1 for PHD inhibitor molecules assigned to Insilico Medicine.

Peer review

Peer review information

Nature Biotechnology thanks the anonymous reviewers for their contribution to the peer review of this work.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Inclusion and Ethics Statement Study authors have fulfilled criteria for authorship as outlined by the Nature Portfolio Group and specific contributions are listed in the author contributions section. Research roles were agreed among collaborators before carrying out the experiments of this study. The research acknowledges local and external contributions to this study. All experiments were conducted following extensive training and, where appropriate, following approval by ethical bodies.

Supplementary information

Supplementary Information

The synthesis and characterization of ISM012-042 and radiolabeled ISM012-042. Supplementary Figs. 1–6, Tables 1–3 and unprocessed blots.

Reporting Summary

Supplementary Tables

Supplemental Tables 4–14.

Supplementary Data 1

Statistical source data for supplementary figures.

Source data

Source Data Figs. 1 and 2

Raw data with statistical analysis with clearly named tabs for each main figure.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Fu, Y., Ding, X., Zhang, M. et al. Intestinal mucosal barrier repair and immune regulation with an AI-developed gut-restricted PHD inhibitor. Nat Biotechnol (2024). https://doi.org/10.1038/s41587-024-02503-w

Download citation

Received: 03 April 2024
Accepted: 12 November 2024
Published: 11 December 2024
DOI: https://doi.org/10.1038/s41587-024-02503-w

This article is cited by

AI/ML methodologies and the future-will they be successful in designing the next generation of new chemical entities?
- Rachelle J. Bienstock
Journal of Cheminformatics (2025)
Oral ENPP1 inhibitor designed using generative AI as next generation STING modulator for solid tumors
- Congying Pu
- Hui Cui
- Alex Zhavoronkov
Nature Communications (2025)
Four ways to power-up AI for drug discovery
- Anthony King
Nature (2025)
Comprehensive computational strategies for multi-target drug discovery in inflammatory bowel disease utilizing bioactive compounds
- Pardis Mansouri
- Pegah Mansouri
- Esmaeil Behmard
Scientific Reports (2025)