Efficient in vivo pharmacological inhibition of ΔFOSB, an AP-1 transcription factor, in brain

Abstract

ΔFOSB, an unusually stable member of the AP-1 family of transcription factors, mediates long-term maladaptations that play a key role in the pathogenesis of drug addiction, cognitive decline, dyskinesias, and several other chronic neurological and psychiatric conditions. We have recently identified that 2-phenoxybenzenesulfonic acid-containing compounds disrupt the binding of ΔFOSB to DNA in vitro in cell-based assays, and one such compound, JPC0661, disrupts ΔFOSB binding to genomic DNA in vivo in mouse brain with partial efficiency. JPC0661 binds to a groove outside of the DNA-binding cleft of the ΔFOSB/JUND bZIP heterodimer in a co-crystal structure. Here, we generated a panel of analogs of JPC0661 to establish structure-activity relationships and improve its in vivo efficacy by replacing its amino-pyrazolone cap moiety with various substituents. We show that one such analog, YL0441, disrupts the binding of ΔFOSB to DNA in vitro and in vivo, and suppresses ΔFOSB-function in cell-based assays. Importantly, infusion of YL0441 into the hippocampus of APP mice (a mouse model for Alzheimer’s disease neuropathology) leads to virtually complete loss of ΔFOSB bound to genomic DNA as detected by CUT&RUN sequencing. Our findings corroborate that the binding/release of AP1 transcription factors to DNA can be controlled via small molecules in vivo, even by analogs of a compound that binds to a groove outside of the DNA-binding cleft, and that our lead can be optimized via medicinal chemistry to yield a much more efficacious inhibitor of ΔFOSB function in vivo. These findings define a strategy to design small-molecule inhibitors for other AP-1 and AP-1-related transcription factors, in particular, those involved in neuropsychiatric and neurological disorders.

INTRODUCTION

ΔFOSB, an AP-1 transcription factor, has emerged as a promising drug target for neurological and psychiatric disorders, including drug addiction, cognitive decline, and dyskinesias. The ΔFOSB protein accumulates in specific regions of the brain in response to a range of chronic (but not acute) stimuli.¹ For example, ΔFOSB protein levels rise in the striatum in response to drugs of abuse and mediate increased drug seeking and intake.^1,2 Likewise, ΔFOSB protein levels rise in the striatum in response to antipsychotic medications or to L-DOPA, which are used to treat a variety of psychotic disorders and Parkinson’s disease, respectively, and ΔFOSB is thought to play a key role in mediating the debilitating involuntary movements that accompany these treatments.^3,4 Very high levels of ΔFOSB protein are also found in the hippocampus of individuals with epilepsy or with Alzheimer’s Disease (AD) examined postmortem, and the magnitude of expression corresponds to the severity of cognitive impairment.^5–7

The ΔFOSB protein, which is formed through alternative splicing of the FosB gene transcript and lacks the C-terminal 101 residues of full-length FOSB, is uniquely stable with a half-life of ~7 days in vivo in brain, while FOSB and all other FOS family proteins are much less stable with a half-life of several hours at most.^2,8–10 ΔFOSB stably alters gene expression by binding to AP-1 DNA consensus sites (AP-1 site; 5’-TGA C/G TCA-3’) in the genome at specific regulatory sites, and mediates a variety of long-term neural and behavioral adaptations.¹ ΔFOSB binds to promoter regions of select target genes, where it can decrease or increase gene expression, but most ΔFOSB binds to AP-1 sites at putative enhancer elements distributed throughout intergenic regions and gene bodies, where it impacts target gene expression from these distal locations.^1,11,12 The ΔFOSB protein (237 residues) contains an intrinsically disordered N-terminal region (~150 residues), which likely recruits various cofactors, while the C-terminal basic region/leucine zipper (bZIP) forms a forceps-shaped molecule upon assembly with another AP-1 bZIP partner that generates a DNA-binding cleft located at the fulcrum of the forceps.^13–15 In brain, the most common dimerization partner of ΔFOSB is thought to be JUND.^16,17 Dimerization of ΔFOSB with its partner via their respective leucine zippers enables the DNA-binding motifs (a helical region with basic residues in the bZIP domain) to insert into the major groove of DNA.^13,14 DNA binding is further controlled by two cysteine residues found N-terminal in the DNA-binding motifs in the bZIP domains (ΔFOSB Cys172 and JUND Cys285 in human) that form a redox switch.^14,18 Under oxidizing conditions and in the absence of DNA, ΔFOSB Cys172 and JUND Cys285 form a disulfide bond that kinks ΔFOSB (but not JUND), malforming the DNA binding site so that it can no longer bind DNA efficiently.^14,18,19 ΔFOSB homomers form in vitro as well and bind to AP-1-consensus DNA sites selectively, but they may not be as sensitive to redox control, and their role in vivo is not known.¹⁵ Despite much recent progress, in particular in terms of structure-function relationships, the exact molecular mechanisms of ΔFOSB action remain unclear, especially the exact impact on the expression of each target gene. Furthermore, despite clear evidence that ΔFOSB plays a critical role in the pathogenesis of several neurological and psychiatric disorders, its broader therapeutic value as a drug target has yet to be established.

The availability of small-molecule compounds that control ΔFOSB action in vivo, would be enormously useful to interrogate ΔFOSB’s molecular mechanisms, elucidate its roles in different disease pathologies and its utility as a therapeutic target. We previously identified JPC0661 as an inhibitor of ΔFOSB action in vitro and in vivo. JPC0661 is a small molecule that contains as a hallmark, a central 2-phenoxybenzenesulfonic acid moiety.¹² JPC0661 inhibits the binding of the ΔFOSB/JUND heterodimer to a TAMRA-labeled AP-1 oligonucleotide with an IC₅₀ of about 11 μM and the ΔFOSB homomer with an IC₅₀ of about 12 μM.¹² In cell-based reporter assays, JPC0661 inhibits transcription of a luciferase reporter gene with the AP-1 consensus sequence built into its promoter region with an IC₅₀ less than about 1 μM.¹² In a crystal structure of ΔFOSB/JUND bZIP+JPC0661, JPC0661 binds in a groove outside of the DNA-binding cleft where it is poised to impact DNA-binding and/or release (Fig. 1a).¹² Two molecules of JPC0661 (Lig1 and Lig2) are accommodated in a deep hydrophobic groove of the ΔFOSB subunit formed by the side chains of Lys171, Arg175, Glu178, Leu179, Arg182, and Leu183, with the positively charged side chains Lys171 and Arg175 at one end of the groove and Arg182 at the opposite end (Fig. 1b).¹² Importantly, the sulfonic acid moiety of JPC0661 Lig1 recruits and rearranges ΔFOSB Lys171 and Arg175, residues that bind to the phosphate-backbone of DNA in the ΔFOSB/JUND+DNA complex in crystal structures (PDB ID: 5VPE; 5VPF) and are presumably critical for ΔFOSB binding to the phosphate-backbone of DNA, while the sulfonic acid moiety of JPC0661 Lig2 interacts with ΔFOSB Arg182 near the ΔFOSB/JUND interface (Fig. 1b).^12,14

Figure 1.

Binding site of the lead compound JPC0661 on the ΔFOSB/JUND bZIP heterodimer. A) 1.7 Å crystal structure of the ΔFOSB/JUND bZIP in complex with JPC0661 (PDB ID: 9OC3). The redox-switch cysteine residues, ΔFOSB C¹⁷² and JUND C²⁸⁵, are highlighted as yellow spheres, ΔFOSB is colored lilac, and JUND is shown in light cyan. Their respective DNA-binding motifs (ΔFOSB N¹⁶⁵–R¹⁷³ and JUND N²⁷⁸–R²⁸⁶) are depicted in dark lilac and dark cyan, respectively. The JPC0661 molecules (Lig1 and Lig2) are rendered with carbon atoms in light and dark green, respectively; oxygen atoms are in red, nitrogen in blue, and sulfur in yellow. B) Overlap of the JPC0661 binding site with residues involved in DNA binding in the ΔFOSB/JUND bZIP mapped onto the ΔFOSB/JUND bZIP+DNA structure (PDB ID: 5VPE). ΔFOSB is shown in lilac and JUND in light cyan. The DNA-binding regions and DNA are shown in tan, the JPC0661-binding region is in dark green, and the overlapping area between the JPC0661-binding and DNA-binding regions (i.e., residues that contact JPC0661 and DNA) is depicted in light green; the redox switch residues, ΔFOSB C¹⁷² and JUND C²⁸⁵, are marked in yellow and also fall within the DNA-binding region. Prominent residues are labeled: ΔFOSB (K¹⁷¹, C¹⁷², R¹⁷⁵, R¹⁷⁶, E¹⁷⁸, L¹⁷⁹, R¹⁸², L¹⁸³) and JUND (C²⁸⁵, R²⁸⁶, K²⁸⁹, L²⁹⁰, E²⁹⁷).

JPC0661 is a promising lead compound because it is minimally toxic in Neuro 2A and human neural progenitor cells.¹² Furthermore, in-vivo administration of JPC0661 for three days into the dorsal hippocampus of APP mice, a mouse model of Alzheimer’s disease neuropathology, reduces the number of ΔFOSB-bound AP-1 sites by ~60% in CUT&RUN studies.¹² These APP mice express highly elevated levels of ΔFOSB protein in the hippocampus, and they exhibit spontaneous non-convulsive seizure activity similar to that observed in individuals with AD.^5,6 However, despite the encouraging results using JPC0661 as an inhibitor of ΔFOSB action, this compound has several drawbacks: it displays only moderate activity in vivo (~60% loss of ΔFOSB DNA-bound sites), there are two molecules binding side-by-side to the ΔFOSB bZIP subunit in the crystal structure, and the compound rearranges side chains lining the druggable groove suggesting an induced-fit binding mechanism. Thus, while JPC0661 demonstrates proof-of-principle that the action of ΔFOSB (and, by extension, other AP-1 transcription factors) can be pharmacologically regulated in vivo, it is unknown whether it serves as a lead compound that can be further optimized using targeted medicinal chemistry.

Here, we demonstrate that JPC0661, a first-generation inhibitor of ΔFOSB, can be optimized using targeted medicinal chemistry, revealing further insights into the structure-activity relationships that govern compound activity. The resultant analogs display varying abilities to disrupt DNA binding of ΔFOSB/JUND and ΔFOSB/ΔFOSB biochemically and to work as inhibitors of ΔFOSB-mediated transcription in cell-based luciferase reporter assays. Our best-optimized compound, YL0441, shows an IC₅₀ < ~0.1 μM in cell-based reporter assays. Importantly, in CUT&RUN-seq experiments, YL0441 decreases the number of ΔFOSB-bound AP-1 peaks by ~94% in the hippocampus of APP mice. Furthermore, YL0441 possesses good pharmacological properties, including low cellular toxicity and high metabolic stability. Finally, our biochemical studies reveal that YL0441 does not prevent or promote closure of the redox switch, suggesting that its inhibitory effect on DNA-binding, like JPC0661, which binds outside of the DNA-binding cleft, works via a mechanism independent of the redox switch. Our study establishes that 2-phenoxybenzenesulfonic acid-containing compounds can be chemically optimized to yield in vivo inhibitors that are highly efficacious in disrupting the biological function of ΔFOSB, i.e., the association with DNA.

RESULTS

Strategy to improve JPC0661 as an inhibitor of ΔFOSB/JUND and ΔFOSB/ΔFOSB

We embarked on a medicinal chemistry campaign to optimize JPC0661 for cell-based assays and in vivo animal studies (Fig. 2a and Fig. 2b).

Figure 2.

Chemical analogs of JPC0661. A) Hit-to-Lead optimization strategy based on the starting lead compound JPC0661 with chemical moieties named. B) Substituents used to probe the role of the phenoxy group and amino-pyrazolone moiety in JPC0661.

JPC0661 is composed of a central phenyl ring with attached phenoxy, sulfonic acid, and amino-pyrazolone groups. In the JPC0661 co-crystal structure with ΔFOSB/JUND, the phenoxy groups of Lig1 and Lig2 are buried deep into a hydrophobic groove on ΔFOSB.¹² The sulfonic acid group of Lig1 forms a critical salt-bridge and captures the DNA-binding residues ΔFOSB Lys171 and Arg175, while additional stabilization is provided by π–π stacking and a water-mediated hydrogen-bonding network; the sulfonic acid group of Lig2 interacts with ΔFOSB Arg182.¹² Guided by structure-based design principles, we focused primarily on replacing the amino-pyrazolone moiety using a ring-opening approach (Fig. 2a) and exploring alternative substituents, such as cinnamamido, benzamido, pyrrolyl, and triazolyl groups, while leaving the phenoxy and sulfonic acid group anchors intact (Fig. 2b). These substitutions were selected based on predicted binding interactions from the ΔFOSB/JUND bZIP+JPC0661 co-crystal structure: amido groups (e.g., cinnamamido, benzamido) were expected to enhance hydrogen-bonding interactions at the rim of the cleft, while the heteroaryl caps (e.g., pyrrolyl, triazolyl) could engage in π-π -stacking with nearby aromatic residues. In parallel, a fluoro-group was introduced to leverage for future PET tracer design, either by replacing the phenoxy group in select compounds (QB0360, QB0361, QB0363, QB0365) or by incorporating fluorinated aryl substituents into the cap region (e.g., 4-F, 4-CF₃) to also modulate electronic properties and lipophilicity (QB0301, QB0309) (Fig. 2b). A fluorinated aryl substituent is the precursor for the reaction to introduce a phenoxy group. Thus, we included these fluorinated compounds in our study to assess how critical a phenoxy moiety is for compound activity and to establish SARs, while simultaneously providing potential fluoro-imaging ligands should these compounds prove active. See Supplemental Material for the synthetic routes and experimental procedures generating these molecules. A combination of ¹H NMR, ¹³C NMR, high-resolution mass spectrometry, and HPLC analyses was used to validate the chemical structures of all 22 newly synthesized compounds (see compendium in Table S1) and confirm that the purities were satisfactory for biological studies.

Analogs of JPC0661 disrupt ΔFOSB/JUND and ΔFOSB/ΔFOSB binding to DNA

To assess the activity of our analogs, we used a series of biochemical and cell-based assays. We first measured fluorescence polarization (FP)-based dose response curves (FP-DRCs), testing the ability of each compound to disrupt the binding of purified full-length ΔFOSB/JUND heterodimers or ΔFOSB/ΔFOSB homomers to a TAMRA-labeled oligonucleotide carrying an AP-1-consensus site (5’-TGA C/G TCA-3’).^12,19,20 The parent compound JPC0661 inhibited DNA binding to ΔFOSB/JUND and ΔFOSB/ΔFOSB in the micromolar range with IC₅₀ 8.9 μM (95% CI 5.7–13.9 μM) and 14.8 μM (95% CI 10.7–20.4 μM), respectively (Fig. 3a), as we recently reported.¹² Of the 22 compounds, six compounds were considered active with an IC₅₀ lower than ~100 μM, YL0441, QB0309, QB0348, QB0311, QB0343 and QB0326 (Fig. 3a; Suppl. Fig. S1). Of these active compounds, four looked particularly promising: YL0441 (IC₅₀ 13.7 μM and 12.3 μM), QB0309 (IC₅₀ 19.1 μM and 30.6 μM), QB0348 (IC₅₀ 8.2 μM and 9.4 μM), and QB0311 (IC₅₀ 2.9 μM and 10.2 μM) for ΔFOSB/JUND and ΔFOSB/ΔFOSB, respectively (Fig. 3a; Suppl. Fig. S1). In addition to the phenoxy and sulfonic acid groups, YL0441, QB0309, QB0311, and QB0348 contain an acrylamide linker to the central phenyl ring, attached to furan, 4-(trifluoromethyl)phenyl, 3,4-dihydroxyphenyl, and 4-biphenyl moieties, respectively, instead of the amino-pyrazolone group attached to the central phenyl ring. Importantly, YL0441, QB0309, QB0348, and QB0311 each disrupted DNA binding to both ΔFOSB dimer types (ΔFOSB/JUND heterodimers and ΔFOSB/ΔFOSB homomers) to roughly similar extents in FP assays (Fig. 3a). This is important because it indicates the potential of these compounds to inhibit ΔFOSB in different protein complexes in vivo, regardless of its dimerization partner, which is consistent with JPC0661 binding only to ΔFOSB but not to JUND in the crystal structure. Twelve compounds had very low activity or were inactive, while four compounds (QB0292, QB0293, QB0294, and QB0295) gave ambiguous results because they altered the fluorescence of the TAMRA-labeled oligonucleotide in absence of purified protein, thus interfering with the assay (Fig. 3b; Suppl. Fig. S1). Notably, QB0365 exhibited inconsistent performance in our initial assays, suggesting that its activity requires further validation. Thus, in total four compounds out of the 22-analog panel disrupted DNA-binding biochemically in the low micromolar range.

Figure 3.

Inhibitory effect of 22 analogs of JPC0661 on DNA binding to ΔFOSB/JUND heterodimers and ΔFOSB/ΔFOSB homomers in fluorescence polarization dose-response curves (FP-DRCs). A) FP-DRCs of a TAMRA-labeled oligonucleotide containing the AP-1 consensus motif (TMR-cdk5) binding to protein in the presence of varying amounts of compound. In these assays, 25 nM TMR-cdk5 was incubated with either 280 nM ΔFOSB/JUND full-length protein, 320 nM ΔFOSB full-length protein (●), or no protein (○), along with increasing concentrations of compound (0–200 μM). Standard control conditions representing 100% inhibition (i.e., the TMR-cdk5 DNA alone, no protein; ♦; n = 16) and 0% inhibition (i.e., ΔFOSB/JUND protein bound to the TMR-cdk5 DNA; ■; n = 16) were also included. The data were fitted with a three-parameter logistic function constraining the bottom of the curve to the TMR-cdk5 oligo alone control (solid lines), yielding estimates for the IC₅₀ value that represents the ability of a compound to disrupt protein:DNA-binding. Error bars represent the standard error of the mean (SEM) from four replicate measurements per data point. IC₅₀ values are presented with 95% confidence intervals (CI) for representative plots. B) List of the compounds from the 22 analog panel whose effect on DNA binding was limited or which interfered with the assay by interacting directly with the TAMRA-labeled oligonucleotide (“cmpd interferes”), yielding ambiguous results. These compounds were excluded from further investigation.

Next, we tested our panel of compounds for their ability to regulate ΔFOSB-driven gene expression in a cell-based reporter assay.^12,19,20 We used a luciferase reporter gene under control of AP-1-responsive elements stably integrated into HEK293 cells (AP-1-luc HEK293 cells) to monitor the effect of compounds on the expression of luciferase (Fig. 4a; Suppl. Fig. S2). To induce ΔFOSB protein accumulation, these cells were first serum-starved (0.5% serum, 24 h) and then stimulated with high-serum conditions (20% serum, 24 h), as described previously.^12,19 Our starting compound, JPC0661, inhibited AP-1-transcription factor-mediated expression of the luciferase reporter gene with an IC₅₀ of ~0.2 μM in these assays (Fig. 4a), somewhat better than we had observed previously (IC₅₀ of ~1–3 μM).¹² Of the six compounds that were most active in the biochemical FP-DRC assay (which we set as a necessary prerequisite to consider a compound for follow-up), three were also robustly active in the reporter assay: YL0441 (IC₅₀ 0.10 μM, 95% CI 0.01 – 1.85 μM), QB0309 (IC₅₀ 0.64 μM, 95% CI 0.14 – 2.84 μM) and QB0348 (IC₅₀ 0.63 μM, 95% CI 0.05 – 7.55 μM) (Fig. 4a). Compounds QB0311, QB0343, and QB0326 were minimally inhibiting or not at all (Fig. 4a). Several compounds showed significant inhibitory effect on gene expression with IC₅₀ values less than ~1 μM in our reporter assay, despite being minimally active or inconclusive in the biochemical FP-DRC assay (e.g., QB0349, QB0354, QB0301, QB0294) (Suppl. Fig. S2). These discrepancies highlight the experimental gap between in vitro assays with purified AP-1 transcription factors in isolation and cellular assays where the ΔFOSB protein (and any other endogenous AP-1 transcription factors) are induced via serum stimulation and the complete endogenous transcription machinery is present.

Figure 4.

Analogs of JPC0661 regulate expression of an AP-1-driven reporter gene in cell-based assays. A) Effects of compounds (0.003–100 μM) on AP-1-luciferase reporter activity evaluated in AP-1-luc HEK293 cells exposed to a serum shock to induce endogenous ΔFOSB. Dose-dependent inhibition of the expression of the AP-1 reporter by a compound was quantified by measuring changes in luciferase signal, expressed as relative fluorescence units (RFU). Each compound was tested in at least two independent experiments (n = 4 wells per experiment nominally), yielding a minimum of 6–8 wells per compound concentration. The data were normalized to the luciferase signal from blank wells (i.e., no compound) within each experiment and then combined (n = 8 wells). The data were fitted to a three-parameter logistic model to determine IC₅₀ values and the associated 95% confidence intervals (CI) where possible. Data points are presented as mean ± SEM. B) Effect of serially diluted compounds (0.003–100 μM) on the viability of AP-1-luc HEK293 cells assessed using the CellTiter-Glo viability assay (2-hour incubation). Cell viability was normalized to the control containing 0.5% (v/v) DMSO in the absence of compound (n = 7 wells). Data points represent the mean ± SEM, with a total of 10–12 wells analyzed per concentration. A red dotted line indicates the average of the data points for 0.5% (v/v) DMSO alone (i.e., in the absence of compound). C) Effect of YL0441 on the viability of Neuro 2a cells assessed using the CellTiter-Glo cell viability assay after 72-hour incubation at 0, 12.5, 25, and 50 μM compound concentration. Viability was normalized to a DMSO control that contained 0.5% (v/v) DMSO and no compound (n = 8 wells). Data are shown as mean ± SEM, with 12 wells analyzed per compound concentration. D) The metabolic stability of YL0441 versus JPC0661 was assessed in mouse liver microsomes by analyzing compound depletion over time with and without NADPH by LC/MS/MS.

To assess the toxicity of the compounds under conditions like the reporter assay, we tested cell viability as a function of increasing compound concentration (0.003–100 μM) with the CellTiter-Glo viability assay in AP-1-luc HEK293 cells (Fig. 4b; Suppl. Fig. S3). Eleven out of the 22 analogs tested at a concentration of 31.5 μM showed no overt decrease in cell viability, among them YL0441, one of the compounds that inhibited ΔFOSB function in both our biochemical assay as well as in our cellular assay (Fig. 4b). Overall, our series of 2-phenoxybenzenesulfonic acid analogs appears generally well tolerated in cells, serving either as active compounds or inactive control compounds for future in vivo studies (Fig. 4b; Suppl. Fig. S3).

Given its favorable drug-like chemical properties, following Lipinski’s Rule of Five (e.g., MW=385 Da, cLogP=2.15) with an acceptable CNS MPO (Central Nervous System Multiparameter Optimization) score of 4.8, as well as promising biochemical and cell-based activity, YL0441 was selected for further studies. To evaluate the toxicity of YL0441 prior to in vivo administration studies, we assessed its impact on the viability of Neuro 2A cells over 3 days up to a maximum dose of 50 μM, finding that YL0441 was well tolerated in these cells (Fig. 4c). To assess the metabolic stability of YL0441, we used a mouse liver microsomal stability assay and found that YL0441 showed no apparent loss of compound within the 60-minute timeframe tested, similar to our starting compound JPC0661 (Fig. 4d). Thus, leveraging a panel of 22 novel analogs, we successfully generated a new lead compound, YL0441, which inhibits AP-1-driven gene expression effectively in a cell-based reporter assay and possesses low toxicity and favorable metabolic stability (summarized in Supplemental Material Table S1).

YL0441 efficiently disrupts ΔFOSB binding to genomic DNA in brain in vivo

To test whether YL0441 disrupts the binding of endogenous ΔFOSB to genomic DNA in vivo, we infused YL0441 or vehicle directly into the hippocampus of APP mice and then assessed the amount of ΔFOSB bound to DNA in the presence or absence of compound. YL0441 or vehicle was delivered for three days via osmotic minipumps connected to a cannula implanted into the right dorsal hippocampus of 2.5-month-old APP transgenic mice. At this age, hippocampal ΔFOSB levels are robustly and consistently elevated in these mice.⁶ CUT&RUN-sequencing was then performed on the dorsal two-thirds of hippocampal tissue that received the infusion. The contralateral (left) hemibrain, which did not receive any infusion, was fixed and immuno-stained for ΔFOSB using an anti-ΔFOSB antibody (Cell Signaling, #D3S8R) to confirm elevated ΔFOSB levels in the hippocampus of these APP mice (Suppl. Fig. S4).

ΔFOSB-bound chromatin was profiled using the anti-ΔFOSB antibody #D3S8R and the number of ΔFOSB-bound DNA peaks quantified as described previously.^11,12 Importantly, this anti-ΔFOSB antibody does not bind detectably to brain slices from FosB/ΔFosB knockout mice,²¹ so that the peaks pulled down represent ΔFOSB bound to genomic DNA, and not any other AP-1 transcription factors or off-target proteins bound to DNA.¹¹ This enabled us to examine the in vivo impact of YL0441, directly focusing on ΔFOSB alone. Infusion of YL0441 resulted in the dramatic loss in the total number of ΔFOSB binding peaks: whereas animals treated with vehicle displayed 9,642 peaks, YL0441-treated animals retained only 540 peaks, a 94% reduction in total binding events (Fig. 5a). Furthermore, the average peak width decreased from ~286 bp (vehicle-treated mice) to ~254 bp (YL0441-treated mice), suggesting a widespread loss of ΔFOSB-chromatin engagement and reduced ΔFOSB binding (Fig. 5b). These findings indicate that YL0441 interferes with ΔFOSB binding to DNA and/or promotes its release from chromatin in vivo.

Figure 5.

YL0441 potently disrupts ΔFOSB–chromatin engagement in the dorsal hippocampus of APP mice. YL0441 (‘Treatment’) or vehicle (‘Control’) was infused via a cannula unilaterally into the dorsal hippocampus of APP mice for 3 days using osmotic minipumps. CUT&RUN-seq was performed on nuclei isolated from the dorsal hippocampal tissue of each individual animal treated with either vehicle or YL0441 to determine the genome-wide binding of ΔFOSB to genomic DNA in vivo. A) Total number of high-confidence ΔFOSB-bound peaks detected in vehicle-infused (‘Control’) versus YL0441-infused (‘Treatment’) mice. B) Average peak width (bp) of the ΔFOSB-bound peak sets in A). Data in B) are shown as mean ± SEM; **p < 0.01. No sex differences were observed for A) and B). C) Examination of the ΔFOSB-bound peaks under vehicle and YL0441 conditions reveals a widespread genomic distribution over promoter, intronic, and intergenic regions in the dorsal hippocampus, with similar patterns seen for both ‘Treatment’ and ‘Control’ conditions. Stacked-bar charts depict the different classes of genomic features occupied by ΔFOSB (UCSC annotations), including the regions promoter-proximal (≤ 1 kb) and promoter-distal (1–5 kb) with respect to the translation start site (TSS), the 5′ UTR/first exon, intragenic (internal exons + introns) regions, the 3′ UTR/downstream region, and distal intergenic regions. Values are expressed as percentages of the total peaks in each condition (‘Control’ vs ‘Treatment’). D) Total number of unique ΔFOSB-bound genes in vehicle-infused vs. YL0441-infused APP mice shows an approximately 94% reduction in ΔFOSB binding after YL0441 treatment. E) ChIPseeker heatmaps showing the peak distribution of ΔFOSB binding intensities centered around the TSSs in APP mice under vehicle- (left) and YL0441- (right) conditions. YL0441 results in a dramatic loss of ΔFOSB binding near TSS sites at gene promoters. Note: each bar represents one ΔFOSB CUT&RUN peak plotted relative to its genomic distance to the nearest annotated TSS along the X-axis. The bar’s length shows the span of that peak (in bp). Negative values indicate upstream locations from the TSS (i.e., ‘promoter’), positive values indicate downstream locations (i.e., gene body/5′ UTR), with the TSS defined at x=0. Peaks are stacked from top to bottom vertically, sorted by the distance to the TSS, so that the top of the stack lists genes with the most ΔFOSB-peaks bound within −3000 to +3000 bp from the TSS, and the bottom, the least. F) Number of unique genes with ΔFOSB peaks bound to promoter regions, i.e., falling within −3000 to +3000 bp of the TSS (left) compared to those bound to distal intergenic regions (right). Venn diagram analysis reveals that the predominant effect of YL0441 (gold) is to globally deplete peaks seen under vehicle-treated conditions (light blue), regardless of their location within the chromatin. G) Homer motif analysis showing that the vast majority of ΔFOSB-bound peaks (in vehicle-infused animals) are bound to DNA sequences that contain the AP-1 consensus (TGA C/G TCA). H) Example of ΔFOSB binding at the promoter region of Sirt2 promoter (Genome browser snapshot of chr7: position), illustrating the dramatic loss of ΔFOSB occupancy upon YL0441 exposure. The tracks are shown for samples from the dorsal hippocampus of APP mice infused with vehicle (top track in sky blue) or with YL0441 (middle track in gold-green) using an anti-ΔFOSB -antibody for CUT&RUN-seq vs. using an IgG -antibody as a control for non-specific binding (bottom track in grey; subtracted background). I) Example of ΔFOSB binding at a distal intergenic/putative enhancer region of an upstream intergenic enhancer (12.5 kb) at the Camk2a locus (Genome browser snapshot of chr18: position), demonstrating a dramatic loss of ΔFOSB-bound sites in response to YL0441-treatment. Track conventions as described in H).

Genomic annotation revealed that, in vehicle-treated mice, ΔFOSB peaks were broadly distributed across promoters (34%), exons (10%), introns (35%), and distal intergenic regions (21%) (Fig. 5c). After treatment with YL0441, the collection of residual peaks displayed a similar distribution (Fig. 5c). The ΔFOSB-binding peaks at promoters of target genes (i.e., located <300 kb from a transcription start site (TSS)) mapped to ~ 2,897 unique genes in vehicle-treated mice, which decreased to ~165 genes in YL0441-treated mice (Fig. 5d). ChIPseeker heatmaps showed that under normal conditions (i.e., vehicle-treated mice) ΔFOSB-peaks clustered tightly in a vertical column centered on 0 bp (i.e., the TSS of genes), indicating ΔFOSB binding at promoters. Many fewer binding sites were observed at ±1–3 kb distance from the TSSs (either 5’-upstream or 3’-downstream), indicating a much smaller number of ΔFOSB peaks that bound to such proximal regulatory sites (Fig. 5e). More distal regulatory peaks, located far from TSSs, were also lost, indicating a universal decrease in ΔFOSB-binding sites on a broad scale as a result of YL0441-treatment. The remaining few ΔFOSB peaks clustered predominantly within ~500 bp of TSSs, suggesting a small cohort of high-affinity sites that YL0441 does not fully displace (Fig. 5e). Venn diagram analysis shows that, of the 2,897 unique genes associated with promoter peaks (i.e., 2,738+159), only 165 genes (159+6) remained after treatment with YL0441 i.e., a 95% loss (2,738/2,897) (Fig. 5f). Likewise, of the 1,786 unique genes associated with distal intergenic peaks (i.e., 1,627+159), only 166 genes (159+7) were found after treatment with YL0441, i.e., a 91% loss (1,627/1,786) (Fig. 5f). Importantly, we found no evidence that YL0441 treatment induced significant ΔFOSB binding at novel sites not occupied under control conditions.

De novo motif analysis confirmed that the large majority of ΔFOSB peaks from vehicle-treated mice (9,642 ΔFOSB peaks) were enriched for the canonical AP-1 binding motif (TGA C/G TCA), consistent with specific ΔFOSB occupancy (Fig. 5g). Analysis of single representative loci from canonical ΔFOSB targets revealed large-scale YL0441-induced depletion of ΔFOSB peaks, for instance, at the Sirt2 and Camk2a loci – two known in vivo targets of ΔFOSB ² – in IGV Browser snapshots (Fig. 5h and 5i). Collectively, these data indicate that YL0441 dramatically reduces the abundance and intensity of ΔFOSB–DNA interactions in vivo, eliminating 94% of chromatin contacts within the 72 hours of compound administration.

Insight into the mechanism of YL0441 action

To gain insight into the molecular mechanism of YL0441 action, we probed its effect on the redox switch within ΔFOSB/JUND heterodimers: ΔFOSB Cys172/JUND Cys285. Closure of the redox switch (e.g., in response to oxidative stress) induces a dramatic conformational change of ΔFOSB, introducing a kink in the helix at the hinge formed by ΔFOSB Arg176-Arg177 that allows ΔFOSB Cys172 to approach and form a disulfide bond with JUND Cys285. This conformational change strongly inhibits DNA binding by distorting the DNA-binding cleft.^14,18,19 To determine whether binding of YL0441 affects the ability of the bZIP helices to splay open and insert their DNA-binding motifs into the major groove of DNA, we tested whether YL0441 alters the overall flexibility of the ΔFOSB bZIP α-helix, using closure of the redox switch as a readout. To this end, recombinant ΔFOSB/JUND bZIP (containing the sole cysteine residues ΔFOSB Cys172 and JUND Cys285) was oxidized with diamide (100 μM) in the presence of excess compound (500 μM), and the ability of the redox switch to close and generate a disulfide-bonded heterodimer was monitored. We compared YL0441, JPC0661 (our starting lead), and the inactive compound QB0365 in this biochemical assay. As a control, the ΔFOSB/JUND bZIP was treated with diamide alone (no compound) to close the redox switch, and with the compound Z21599131480, which covalently modifies ΔFOSB Cys172, rendering the redox switch constitutively open.¹⁹ None of the three compounds (YL0441, JPC0661, and QB0365) prevented closure of the redox switch in ΔFOSB, as assessed with non-reducing SDS-PAGE (Fig. 6a). We next tested whether these compounds could instead promote oxidation of the ΔFOSB/JUND bZIP directly, thereby disrupting DNA binding. We found that YL0441, JPC0661, and QB0365 did not oxidize the protein to induce closure of the redox switch (Fig. 6b). Taken together, YL0441 disrupts ΔFOSB binding to DNA using a mechanism that does not involve the redox switch, consistent with the JPC0661 binding site residing in a druggable groove outside of the DNA-binding cleft.

Figure 6.

Molecular insight into YL0441 action. A) JPC0661, YL0441, and QB0365 do not prevent closure of the redox switch, unlike the cysteine-targeting Z2159931480. ΔFOSB/JUND bZIP was incubated with compounds (0.5 mM) with 100 μM diamide (‘ox’, oxidized) or without diamide (‘red’, reduced) and assessed by SDS-PAGE with or without reducing agent in the loading buffer. B) JPC0661, YL0441, and QB0365 do not promote closure of the redox switch by oxidizing the protein. ΔFOSB/JUND bZIP protein was incubated with the compounds (0.5 mM) with no diamide (‘red). Protein alone (no compound) was also incubated with diamide (‘ox’; 100 μM) as a control. Samples were then assessed by SDS-PAGE (with or without reducing agent in the loading buffer). In A) and B) ‘cntrl’ denotes the ΔFOSB/JUND bZIP protein in absence of compound; ‘M’ indicates the molecular markers (kDa).

DISCUSSION

Here, we demonstrate that the lead compound JPC0661, belonging to the 2-phenoxybenzenesulfonic acid-containing class of compounds, can be chemically optimized as inhibitors of ΔFOSB function in vitro and in vivo. Our most potent analog, YL0441, has an IC₅₀ of 0.1 μM in AP-1 reporter assays and causes an almost complete loss of ΔFOSB-bound sites to genomic DNA in brain in vivo. This in vivo activity represents a dramatic improvement over JPC0661, which caused only a ~60% loss of ΔFOSB-bound sites,¹² and we thus consider YL0441 a highly efficacious molecule to disrupt the biological activity of ΔFOSB. YL0441 is metabolically stable and displays no detectable cellular toxicity, making it a promising starting point for further development and a useful tool to assess the therapeutic potential of ΔFOSB in animal models for an array of different neuropsychiatric disorders. Furthermore, biochemical and cellular screening of our panel of analogs allows us to establish preliminary structure-activity relationships (SAR), enabling the design of further optimized chemical probes and diagnostic tools suitable for a range of in vivo studies.

Novel compounds targeting ΔFOSB transcription factor action

Despite their clear involvement in many human diseases, relatively modest progress has been made in targeting the AP-1 transcription factor family pharmacologically.^22–24 This family is composed of many members in addition to ΔFOSB and FOSB, including FOS, FOSL1, FOSL2, JUN, JUNB, and JUND, and is related to other bZIP transcription factors such as the ATF family and CREB family.²⁵ These proteins form heterodimers, and in some cases homomers, via their DNA-binding bZIP domains in a mix-and-match strategy.²⁶ Importantly, the different AP-1 transcription factor subunits and complexes each have unique biological functions.^25,27 Thus, compounds that discriminate between the different AP1 subunits and bind selectively to a particular AP-1 subunit or an AP-1 complex would be tremendously valuable. However, the DNA binding cleft buried deep in the fulcrum of bZIP dimer forceps, an obvious site for drug discovery, is highly conserved across AP-1-subunits,¹⁹ making the design of selective compounds challenging. As an alternative strategy, we have shown that the ΔFOSB redox switch residue, Cys172, another attractive site for drug discovery, can be covalently targeted by small molecules, resulting in altered ΔFOSB-driven gene expression in cell-based assays.¹⁹ However, although compounds covalently modifying cysteine residues are gaining acceptance as a drug discovery strategy and receiving FDA approval, such compounds are challenging to develop due to their chemical reactivity and increased potential for off-target effects, metabolic instability, and/or toxicity.^28,29 Our discovery of a druggable groove within the ΔFOSB subunit – located outside of the DNA-binding cleft and exhibiting little sequence conservation among AP-1 family members – provides a fundamentally new strategy for targeting this transcription factor family.¹² Building on JPC0661, which binds to this groove, we designed and tested a panel of 2-phenoxybenzenesulfonic acid-containing analogs to further explore and validate this lead compound.

Our medicinal chemistry effort produced the improved inhibitor YL0441, which potently inhibits ΔFOSB function in vivo and establishes preliminary structure-activity relationships. The sulfonic acid moiety is indispensable for activity, and the set of molecules appears to engage the DNA-interacting residues ΔFOSB Lys171 and Arg175 as well as ΔFOSB Arg182 at either side of the groove in the crystal structure of ΔFOSB/JUND bZIP+JPC0661.¹² The phenoxy group anchors the scaffold deep in this hydrophobic groove, establishing the core pharmacophore. Therefore, our campaign focused on replacing the amino-pyrazolone cap of JPC0661, which confers only modest efficacy. Substitution with cinnamamido or heteroaryl substituents (e.g., biphenyl or furan) resulted in a marked increase in potency and cellular activity, culminating in the excellent inhibitory profile observed for YL0441. It is important to note that, while the biochemical IC₅₀ values of YL0441 and our starting compound JPC0661 are very similar in FP-DRCs (~10–15 μM for ΔFOSB/JUND and ΔFOSB), in the cell-based reporter assay, YL0441 is ~50–100-fold more active in the cell-based reporter assay (IC₅₀ <0.1 μM) and leads to a near complete loss of ΔFOSB-bound peaks to genomic DNA in vivo compared to only 60% for JPC0661. These findings underscore the importance of using multiple orthogonal assays to identify compounds that will have the greatest in vivo efficacy. The biochemical FP-assay, while highly useful in early stages of chemical probe development, relies on recombinant proteins that likely do not fully recapitulate the properties of ΔFOSB in a cellular environment. It also does not capture the additional effects and complexities introduced by binding partners, such as co-factors and AP-1 consensus sites presented in chromatin rather than as short oligonucleotides. In addition to YL0441, two more compounds (QB0309 and QB0348) also exhibited strong activity, though with moderate cellular toxicity, identifying them as possible secondary leads. QB0311, another compound with a large inhibitory effect in biochemical assays, showed weak cellular efficacy and exhibited significant toxicity, indicating pharmacokinetic (ADMET) limitations. Of note, several fluorinated analogs (e.g., QB0360 and QB0361), though not measurably active, also did not decrease cell viability, so that these compounds might hold translational interest for their potential development as PET ligands for in vivo imaging if their activity can be improved. Collectively, our findings establish 2-phenoxybenzenesulfonic acid-containing compounds as a highly promising scaffold for rational medicinal chemistry-driven optimization.

ΔFOSB and other AP-1 transcription factors as drug targets

The ability of YL0441 to pharmacologically reduce the occupancy of ΔFOSB to genomic DNA so effectively in vivo is striking. Three-day infusion of YL0441 reduced ΔFOSB-binding sites from 9,642 peaks (vehicle) to 540 peaks (compound-treated) in CUT&RUN-seq by a global release of transcription factor from all AP-1-motifs (i.e., at promoter regions, gene bodies, and distal intergenic regions) and not by a selective pruning of individual loci. While the number of 9,642 ΔFOSB-binding peaks experimentally determined in vehicle-treated tissue may seem large, it likely represents only a fraction of the possible AP-1 motif landscape. In the human genome for instance there are ~260,000 predicted potential binding sites,³⁰ however in practice, only a small subset of these motifs is experimentally identifiable at a time. Inaccessibility due to chromatin, as well as cooperative or competitive interactions among different AP-1 transcription factors, can gate access to these motifs,^25,31 while ChIP-seq/CUT&RUN experiments inevitably under-sample very transient or low-occupancy protein:DNA interactions. Our studies, using dorsal hippocampal-derived tissue from APP mice treated with vehicle or YL0441, fall within the same range as a recent study that detected 8,199 ΔFOSB AP-1 peaks in nucleus accumbens of saline-treated mice (rising to 11,843 ΔFOSB AP-1 peaks after chronic cocaine exposure).¹¹ The near-total loss of ΔFOSB bound to genomic DNA in vivo, together with the significant narrowing of the residual peak widths, indicates that YL0441 potently reduces ΔFOSB chromatin occupancy and does not permit ΔFOSB to be redirected to ectopic, non-related AP-1 motifs. Thus, while these data do not distinguish whether the compound prevents initial DNA engagement or accelerates dissociation of pre-bound complexes, the preservation of canonical AP-1 motif enrichment among the few remaining peaks suggests that the compound induces a drastic loss of occupancy rather than retargeting to other sites in genomic DNA.

Our studies also highlight the utility of YL0441 to probe the role and mechanism of ΔFOSB as an epigenetic factor in vivo, an important transcription factor involved in the molecular pathology of many neuropsychiatric and chronic neurological disorders. AP-1 transcription factors, like ΔFOSB, likely act as context-dependent pioneer transcription factors that remodel and/or maintain chromatin accessibility to prepare target genes for transcription,^31,32 and assemble into large protein complexes at enhancer and promoter sites.³³ But how tightly ΔFOSB binds to genomic DNA is not known. Recent work suggests that AP-1 transcription factors bind dynamically, taking turns to occupy the same AP-1 sites in genomic DNA (‘AP-1 hotspots’),^31,34 where they also likely recruit other bZIP-containing partners generating a portfolio of non-redundant DNA-binding transcription factor complexes that recognize both overlapping as well as unique genomic loci.^31,35,36 Therefore, it is even more impactful that YL0441 disrupts virtually all ΔFOSB-bound AP-1-sites. This is an important finding because ΔFOSB, like other AP-1 transcription factors, can work either as an activator or repressor of gene expression depending on the exact target gene and cellular context, likely because it can recruit different bZIP dimerization partners or different co-factor proteins.^31,35,36 Because JPC0661 appears to engage only ΔFOSB not JUND (based on our crystal structure), other compounds or optimized analogs could be developed to target a specific AP-1 subunit (e.g., ΔFOSB) selectively over other AP-1 transcription factors and act with high efficacy. Such AP-1-subunit-specific compounds would be powerful tools to experimentally dissect the effects of inhibiting a particular AP-1 transcription factor in vivo (regardless of its particular dimerization partner), for instance, to probe the expression of particular gene targets, such as those mediating long-lasting maladaptations underlying drug addiction, cognitive decline, and dyskinesias.

It is of paramount importance to identify and validate a set of compounds that disrupt the binding of ΔFOSB to AP-1 sites in vivo pharmacologically so that key disease-driving genes can be identified and their relationship to the different disease pathologies delineated. Genetic studies demonstrate that ΔFOSB has unique and controlling effects on the pathogenesis of several neuropsychiatric disorders. For example, genetically inhibiting elevated ΔFOSB in the striatum in response to drugs of abuse reduces their rewarding effects,^1,2; decreasing ΔFOSB that accumulates in response to chronic L-DOPA used to treat Parkinson’s disease patients reduces abnormal involuntary movements (AIMs) while leaving the antiparkinsonian action of L-DOPA undiminished;⁴; genetic inhibition of accumulated ΔFOSB in hippocampus in mouse models of Alzheimer’s disease reverses cognitive deficits in AD mice.^5,6 YL0441 provides a clean chemogenetic tool to probe ΔFOSB-dependent transcriptional programs in vivo and to dissect how persistent AP-1 signaling shapes gene networks that converge on disease-relevant phenotypes. YL0441 also sets a quantitative benchmark for future ΔFOSB antagonists and raises the possibility of leveraging AP-1 displacement therapeutically in disorders where pathological ΔFOSB accumulation is a driver of maladaptive transcriptional memory.

METHODS

Constructs

Recombinant proteins overexpressed in Sf9 insect cells:

(a) full-length mouse ΔFOSB (a FOSB splice variant; UniProt ID: P13346, residues Phe2–Glu237) and (b) full-length mouse JUND (UniProt ID: J04509), residues Glu2–Tyr341. All constructs were cloned into the pFastBac1 vector and contained an N-terminal (His)₆-tag (sequence MGHHHHHH).

Recombinant proteins overexpressed in E. coli:

(a) mouse ΔFOSB bZIP domain (residues Glu153-K219); and (b) mouse JUND bZIP domain (residues Gln260–Val326 from UniProt J04509). Note, the amino acid sequences for these ΔFOSB bZIP and JUND bZIP domain constructs are identical between mouse and human, but the JunD residue numbering is offset by 6 in the human JUND sequence (Gln266–V332) compared to mouse; human numbering is used throughout this study. All plasmids were sequence-verified before use. Both constructs were cloned into the pET21a-NESG vector and contained an N-terminal (His)₆-tag with a TEV protease cleavage site (sequence MGHHHHHHENLYFQS).

Reagents and biological resources

Cell lines.

AP-1 luciferase reporter HEK293 cells (AP-1-luc HEK293): BPS Bioscience, San Diego, CA; Cat. #60405; Neuro2A cell line: ATCC; Manassas, Virginia; catalog# CCL-131.

Oligonucleotides.

The 19-mer cdk5 duplex oligonucleotide (‘cdk5 oligo’), derived from the AP-1 site of the cyclin-dependent kinase 5 promoter (5′-CGTCGGTGACTCAAAACAC-3′; AP-1 site underlined), was used in various assays. The version used in fluorescence experiments (‘TMR-cdk5’) was prepared by annealing equimolar amounts of complementary strands labeled at their 5′ ends with TAMRA (Sigma-Aldrich) by heating to 95 °C for 2.5 minutes, followed by gradual cooling with a rate of approximately 1 °C/min to room temperature. Annealed duplexes were stored at −20 °C in annealing buffer (10 mM Tris, pH 8.0; 50 mM NaCl) at a concentration of 50 μM.

Chemical Synthesis

The detailed synthetic methods, synthetic routes, and experimental procedures to generate new analogs are provided as Supporting Information in the Supplemental Material. The acrylamide-derived aryl sulfonic acid analogs were synthesized starting from commercially available 2-fluoro-5-nitrobenzenesulfonyl chloride to generate phenyl 5-nitro-2-phenoxybenzene sulfonate for sulfonic acid protection. The subsequent reduction of nitro group of phenyl 5-nitro-2-phenoxybenzenesulfonate through Pd-catalyzed hydrogenation provided phenyl 5-amino-2-phenoxybenzenesulfonate, followed by the coupling with acrylic acid derivatives to afford the desired amide coupling products. The final cleavage of the phenyl group from sulfonate ester intermediates through the hydrolysis using 2 N aqueous NaOH in a mixed dioxane solution, followed by acidification with 2 N aqueous HCl to adjust the pH to approximately 3–4, provided the desired acrylamide-derived aryl sulfonic acid analogs. The chemical structures and purity of all newly synthesized compounds have been characterized using ¹H NMR, ¹³C NMR, high-resolution mass spectrometry, and HPLC analyses for structural validation and quality control.

Purification of wild-type ΔFOSB/JUND heterodimers and ΔFOSB homomers.

Mouse ΔFOSB/JUND heterodimers and ΔFOSB/ΔFOSB homomers were produced in Sf9 insect cells using the Bac-to-Bac expression system (Invitrogen), following protocols previously reported.^12,13,19,20 Briefly, high-titer baculoviruses (~1×10⁸ pfu/mL) were used to infect Sf9 cells at ~1.5×10⁶ cells/mL. Heterodimers were obtained by co-infection with N(His)6-ΔFOSB and N(His)6-JUND viruses at MOIs of 1.0–1.5 and 1.0–3.0, respectively. For ΔFOSB/ΔFOSB homomers, the same cell density and an MOI of 1.0–1.5 were used. Infections were carried out in 6 L SF900 cultures for 72–84 hours at 28 °C, shaking at 145 rpm. Cells were harvested (15 min, 3000 rpm, 4 °C), resuspended in PBS, flash-frozen in liquid nitrogen, and stored at −80 °C.

For protein purification, cells from 6 L cultures were resuspended in 300 mL lysis buffer (25 mM Tris pH 8.0, 0.2% [v/v] Triton X-100, 1 mM TCEP) supplemented with protease inhibitors (0.5 mM PMSF, 10 μg/mL pepstatin, 10 μg/mL leupeptin, and two tablets of cOmplete^™ EDTA-free Protease Inhibitor Cocktail, Roche). After incubation on ice for 30 min, cells were lysed by sonication. The lysate was treated with 300 mM NaCl, 5 mM MgCl₂, and 50 μg/mL DNase, incubated for 1 hour on ice, then supplemented with 0.5 M NaBr and 10 mM imidazole. Insoluble material was removed by centrifugation at 18,000 rpm for 30 min at 4 °C. For subsequent Ni-affinity purification, Ni-NTA resin (10 mL of a 50% slurry; Thermo Scientific), pre-equilibrated in buffer A (25 mM Tris pH 8.0, 1.0 M NaCl), was added to the clarified lysate and incubated for 3 hours at 4 °C. The resin was transferred to an empty column, and bound proteins were eluted using a gradient of buffer B (25 mM Tris pH 8.0, 1.0 M NaCl, 0.5 M imidazole). Fractions containing protein were combined, diluted with 25 mM Tris pH 9.0, 1 M NaCl to a final protein concentration of 0.075 mg/mL, then dialyzed overnight against 25 mM Tris pH 9.0, 300 mM NaCl, 1 mM DTT, 0.5 mM PMSF, followed by a 3 hours dialysis against 25 mM Tris pH 9.0, 75 mM NaCl, 1 mM DTT, 0.5 mM PMSF before anion exchange chromatography. The sample was loaded onto a Mono Q 5/50 GL column (Cytiva) equilibrated with buffer A (25 mM Tris pH 9.0, 75 mM NaCl, 1 mM DTT) and eluted with a linear 0–100% gradient of buffer B (25 mM Tris pH 9.0, 1 M NaCl, 1 mM DTT). Peak fractions were pooled, concentrated, and further purified by size-exclusion chromatography using a HiLoad 16/600 Superdex 200 16/600 size exclusion column (GE Healthcare) equilibrated in 20 mM Tris pH 8.0, 1 M NaCl. For ΔFOSB/JUND, SDS-PAGE was used to confirm co-purification of both components in a 1:1 complex. Final protein purity was assessed by SDS-PAGE. Purified ΔFOSB/JUND and ΔFOSB/ΔFOSB preparations were flash-frozen in aliquots (2–5 mg/mL) in 20 mM Tris pH 8.0, 1 M NaCl.

Preparation of ΔFOSB/JUND bZIP domains.

ΔFOSB and JunD bZIP domains were expressed in E. coli as previously described.^12,14,15,19 Briefly, 6 L cultures of E. coli Rosetta 2 (DE3) cells (Invitrogen) harboring the respective expression plasmids were grown in LB medium at 37 °C to an OD₆₀₀ of ~0.5, cooled to 16 °C, and induced with 0.5 mM IPTG overnight. Cells were harvested by centrifugation at 4 °C for 30 min at 4000 rpm, resuspended in 10 mL PBS per 1 L of culture, flash-frozen in liquid nitrogen, and stored at −80 °C until purification.

For protein purification, cell pellets were thawed on ice, resuspended in lysis buffer (20 mM Tris pH 8.0, 250 mM NaCl, 1 mM TCEP, 0.5 mM PMSF), and treated with 1.1 mg/mL lysozyme. After a 30-minute incubation on ice, cells were lysed by sonication. The lysate was supplemented with 5 mM MgCl₂, 1 M NaCl, and 30 μg/mL DNase I, and incubated for an additional 30 min on ice. Before Ni-affinity purification, 0.5 M NaBr and 20 mM imidazole were added, and the mixture clarified by centrifugation at 18,000 rpm for 30 min at 4 °C. Ni-NTA resin (10 mL of a 50% slurry; Thermo Scientific), pre-equilibrated in buffer A (20 mM Tris pH 8.0, 1 M NaCl, 500 mM NaBr), was added to the supernatant and incubated for 3 hours at 4 °C. The resin was transferred to a column, and bound protein eluted using a gradient of buffer B (20 mM Tris pH 8.0, 1 M NaCl, 500 mM NaBr, 500 mM imidazole). To remove the (His)₆-tag from the ΔFOSB and JunD bZIP domains, (His)₆-TEV protease was added to the recombinant protein fractions at 40 μg protease/mg target protein in digestion buffer (50 mM Tris-HCl pH 8.0, 0.5 M NaCl, 1 mM DTT, 1% (v/v) glycerol), followed by an overnight incubation at 4 °C. To remove ΔFOSB and JunD bZIP domains still carrying (His)₆-tags as well as the (His)₆-TEV protease, 1 M NaCl and 0.5 mL of Ni-NTA resin (50% slurry) equilibrated in digestion buffer, were added, and the sample incubated for 1 h at room temperature. The resin was pelleted by centrifugation at 900 rpm for 10 min, the proteins now lacking (His)₆-tags recovered in the supernatant, passed through a Bio-Spin^® column (Bio-Rad, #7326008), and the protein concentration determined using the Bio-Rad Protein Assay (Kit I, #5000002). For heterodimer formation, ΔFOSB and JUND bZIP domains were mixed at a 1:1 molar ratio. For both ΔFOSB/JUND bZIP heterodimers and ΔFOSB/ΔFOSB bZIP homomers, buffer was then exchanged by dialysis at room temperature against 20 mM HEPES pH 7.0, 500 mM NaCl. As a final step, the proteins were concentrated to ≤5 mg/mL and purified by size exclusion chromatography at 4 °C using a Superdex HiLoad 16/600 75 pg column (Cytiva) equilibrated in GF buffer (20 mM HEPES pH 7.0, 500 mM NaCl). Purity was confirmed by SDS-PAGE. Proteins were concentrated to ~8–10 mg/mL, flash-frozen in liquid nitrogen, and stored at −80 °C.

Fluorescence polarization dose-response curves (FP-DRCs) to test the impact of compounds on the binding of DNA to ΔFOSB/JUND and ΔFOSB/ΔFOSB.

The inhibitory effect of compounds on ΔFOSB/JUND or ΔFOSB/ΔFOSB binding to DNA was evaluated in 10-point fluorescence polarization dose response curve (FP-DRC) assays following established protocols.^12,19,20 Serial compound dilutions (0–200 μM) were prepared in 384-well plates (Corning #3676) using an Echo 550 Acoustic Liquid Handler (Beckman) or manually. ΔFOSB/JUND (280 nM monomer) or ΔFOSB (300–320 nM monomer) was added to wells along with the TMR-labeled ‘cdk5 oligo’ (TMR-cdk5) at a final concentration of 25 nM. Wells containing 25 nM TMR-cdk5 alone, representing 100% inhibition (i.e., no protein-DNA-binding), were used as the ‘positive control’, while wells containing 25 nM TMR-cdk5 plus protein (either 280 nM ΔFOSB/JUND or 320 nM ΔFOSB) without any compound representing 0% inhibition (i.e., full protein:DNA-binding) were used as the ‘negative control’. Buffer only was used as the ‘no-protein’ control. All wells, including controls, received DMSO at a consistent final concentration of 0.5% (v/v). FP-DRC assays with the compound concentration series were run in quadruplicate for each protein complex (ΔFOSB/JUND heterodimers in 20 mM HEPES pH 7.5, 150 mM NaCl, and ΔFOSB homomers in 20 mM HEPES pH 7.5, 50 mM NaCl) and in duplicate for TMR-cdk5 alone without protein. The wells containing the compound concentration series and TMR-cdk5 (but no protein) were used to detect any possible direct interference of the compound with the oligonucleotide. Each plate included, additionally, 16 positive control wells (TMR-cdk5 only, 100% inhibition) and 16 negative control wells (protein + TMR-cdk5, 0% inhibition). After 15 minutes of incubation at room temperature, the FP signal was measured using either a Synergy Neo2 (BioTek) or PHERAstar (BMG LabTech) plate reader (excitation 530–540 nm, emission 590 nm). Data were fitted with GraphPad Prism 6 using a 3-parameter logistic curve (‘log(inhibition) vs. response-variable slope fixing the bottom of the plot to the averaged value of the 100%-inhibition control wells on each plate), yielding IC₅₀, SEM, and 95% confidence interval (CI) values.

Cell-Based Reporter and Toxicity Assays in AP-1-luc HEK293 cells

An AP-1 reporter assay was used to assess how compounds modulate AP-1-driven transcriptional activity of a luciferase reporter gene. The experimental protocol followed established methods.^12,19 Briefly, an AP-1-luciferase reporter HEK293 cell line (BPS Bioscience, USA) was cultured at 37 °C and 5% CO₂ in growth medium 1B (BPS Bioscience) supplemented with 10% fetal bovine serum (FBS; Invitrogen) and 1% penicillin/streptomycin (Hyclone). Cells were plated at a density of 6.0×10⁴ cells per well in quadruplicate across two 48-well plates. Once the cell cultures reached ~70% confluence, the medium was switched to assay medium 1B containing 0.5% FBS to induce a 24-hour serum starvation. Compounds in a concentration series (0.003 −100 μM) or a vehicle control (0.5% (v/v) DMSO) were added to cells in parallel, followed by a 2-hour incubation. Subsequently, cells were exposed for 24 hours to medium containing 20% FBS (“serum-stimulated”) to further induce ΔFOSB protein accumulation. After treatment, cells were lysed and the luciferase activity was quantified using the Promega luciferase assay system on a BioTek Cytation 3 plate reader. Data from two independent experiments (typically n=4) were each normalized against blank wells (i.e., vehicle control treatment), combined, and then analyzed via nonlinear regression in GraphPad Prism v10 to determine IC₅₀ values with 95% confidence intervals (three-parameter logistic model). In parallel, cell viability assays were performed using AP1-luc-HEK293 cells and Neuro 2A cells. For AP1-luc-HEK293 cells, the viability assay was performed under conditions identical to the AP-1 reporter assay to evaluate compound-induced toxicity. After a two-hour incubation with compound or vehicle (0.5% (v/v) DMSO), the medium was exchanged for that containing 20% FBS (“serum-stimulated”), followed by 24 hours of incubation. For Neuro 2A cells, the assay was performed separately. Cells were treated with the compound or vehicle (0.5% (v/v) DMSO) for 72 hours. Cell viability was then determined using the CellTiter-Glo luminescent assay (Promega) according to the manufacturer’s protocol, and luminescence was measured using a BioTek Cytation 3 reader. Statistical analyses were performed in GraphPad Prism v.10.0 (GraphPad Software), and data are presented as means ± standard error of the mean (SEM).

Mouse liver microsomal stability assays

Compound stability was assessed in mouse liver microsomes (238.5 μL; 0.63 mg/mL) incubated with 1.5 μl compound (0.2 mM stock in 20% DMSO) at 37 °C in buffer containing 50 mM potassium phosphate (pH 7.4) in the presence or absence of 1 mM NADPH by Biodura-Sundia. The final volume per sample was 300 μL containing an end concentration of liver microsomes of 0.5 mg/mL, 1 μM compound, and 0.125% (v/v) DMSO concentration. Samples were collected at 0, 5, 15, 30, and 60 minutes (with NADPH) and at 0 and 60 minutes (without NADPH), and the reaction was quenched by the addition of acetonitrile. After centrifugation, an internal standard was added to the supernatants, which were analyzed by mass spectrometry (LS/MS/MS). Compound disappearance over time was used to calculate half-life (t½, in minutes), in vitro intrinsic clearance (CL_int, in μl/min/mg protein), and predicted hepatic clearance (CL_hep in μl/min/kg).

Administration of YL0441 in vivo

For in vivo studies, we used 2–3-month-old heterozygous transgenic male and female mice expressing human amyloid precursor protein (APP) carrying Swedish (K670N, M671L) and Indiana (V717F) mutations linked to familial Alzheimer’s disease (Line J20; MMRRC_034836-JAX; hAPP770 transgene). ³⁷ Using a cannula connected to an Alzet micro-osmotic pump (model 1003D), YL0441 was delivered unilaterally into the right dorsal hippocampus. Micro-osmotic pumps were each filled with vehicle (0.02% (v/v) DMSO/0.9% (v/v) saline) or YL0441 (50 μM compound in 0.2% DMSO/0.9% saline) per manufacturer’s instructions. The brain infusion cannula was connected to the pump using medical-grade polyvinyl chloride catheter tubing and primed overnight in 0.9% sterile saline before implantation. To permit the unilateral infusion of compound or vehicle, one filled micro-osmotic pump per mouse was implanted subcutaneously in the intrascapular region, while the tip of the cannula targeted stereotactic coordinates: dorsal-ventral (D/V), −2.0 mm; anterior-posterior (A/P), −2.1 mm; and medial-lateral (M/L), −1.2 mm, and the cannula base was affixed to the skull using Superglue. The lot of Alzet 1003D micro-osmotic pumps used delivered fluid at a rate of 0.96 μL/h, yielding a final drug delivery of 18.5 ng YL0441/h. YL0441 or vehicle was infused for three days into the hippocampus. Mice were then anesthetized with commercial euthanasia solution, transcardially perfused with ice-cold 0.9% saline solution, and the brains collected and hemisected. The left hemibrain was fixed in 4% paraformaldehyde and used for immunohistochemistry. The right hemibrain was flash-frozen on dry ice, and then later the hippocampus was isolated, and the dorsal two-thirds sub-dissected to use for CUT&RUN analysis. All procedures were approved by the Institutional Animal Care and Use Committees of Baylor College of Medicine under protocol AN-6943. Before sectioning on a freezing sliding microtome, the fixed hemibrains were cryoprotected in 30% sucrose in phosphate-buffered saline. Subsequently, serial coronal sections (30 μm thickness) were divided into 10 subseries, each containing every tenth section throughout the rostral-caudal extent of the brain. One subseries was used to stain for ΔFOSB applying the avidin-biotin diaminobenzidine (DAB) method, with as primary antibody, the rabbit anti-ΔFOSB antibody (1:5000; Cell Signaling, D3S8R), and as secondary antibody, the biotinylated goat anti-rabbit (1:200; Vector, BA-1000), with DAB as the chromogen. All samples were processed, stained, and imaged at the same time, using identical parameters and imaging settings. To quantify ΔFOSB protein levels in the tissue, an experimenter blinded to treatment evaluated and confirmed ΔFOSB induction in the transgenic mice. The ImageJ software was used to measure the mean gray value of the dentate gyrus granule cell layer, which was then normalized to the mean gray value of the stratum radiatum and averaged over two consecutive sections for each sample. For these studies, 11 APP mice (6 male and 5 female) were treated with YL0441, and 10 mice (6 male and 4 female) were treated with vehicle.

CUT&RUN-sequencing

Nuclei Isolation:

Nuclei were isolated from dorsal hippocampal tissue of APP mice, which had been infused with YL0441 or vehicle. Tissue was Dounce-homogenized on ice in lysis buffer (320 mM sucrose, 5 mM CaCl₂, 0.1 mM EDTA, 10 mM Tris-HCl pH 8.0, 1 mM DTT, 0.1% Triton X-100, 1.5 mM spermidine, protease inhibitors) using ~30–40 strokes per pestle. Homogenates were passed through a 40 μm strainer, transferred to ultracentrifuge tubes, layered over a 1.8 M sucrose cushion (10 mM Tris-HCl pH 8.0, 1 mM DTT, 1.5 mM spermidine, protease inhibitors), and centrifuged at ~85,000 × g (24,000 rpm; SW41Ti, rmax 13.2 cm) for 1.25 h at 4 °C to pellet nuclei.

CUT&RUN Assay:

CUT&RUN followed published protocols with minor adaptations.^11,12,18 Nuclei were resuspended in Wash Buffer (20 mM HEPES-NaOH, pH 7.5, 150 mM NaCl, 0.5 mM spermidine, 0.1% Triton X-100, 0.1% Tween-20, 0.1% BSA, with protease inhibitors). BioMag^®Plus Concanavalin A beads (BP531, Bang Laboratories) were activated twice in Binding Buffer (20 mM HEPES-KOH pH 7.9, 10 mM KCl, 1 mM CaCl₂, 1 mM MnCl₂), then incubated with nuclei for 10 min at room temperature. Bead-bound nuclei were incubated overnight at 4 °C in Antibody Buffer (Wash Buffer; ±2 mM EDTA) with anti-ΔFOSB (Cell Signaling #D3S8R; 1:50) or rabbit IgG (1:50). After washes, beads were incubated with pAG-MNase (EpiCypher) for 1 h at 4 °C. Cleavage was initiated in Digestion/Calcium Buffer (3.5 mM HEPES-NaOH pH 7.5, 10 mM CaCl₂, 0.1% Triton X-100, 0.1% Tween-20) for 5 min on a pre-chilled 4 °C block. Reactions were quenched with EDTA/EGTA stop buffer (170 mM NaCl, 20 mM EDTA, 4 mM EGTA, 0.1% Triton X-100, 0.1% Tween-20, 25 μg/mL RNase, 20 μg/mL glycogen). DNA fragments were eluted at 37 °C for 30 min with shaking, cleared (16,000 × g, 5 min, 4 °C), and ethanol-precipitated.

Library preparation and sequencing:

Libraries were prepared with NEBNext^® Ultra^™ II DNA Library Prep (New England BioLabs) and sequenced on an Illumina HiSeq 4000 using 2 × 150 bp paired-end reads to a depth of 40 million reads per library. Biological replicates: 10–11 per group.

Data Analysis:

FASTQ files were processed with the NGS-Data-Charmer pipeline (https://github.com/shenlab-sinai/NGS-Data-Charmer). Trim-Galore (v0.6.5) was used for adapter trimming, followed by secondary trimming with Cutadapt (v2.10). Reads were aligned to the mm10 genome using HISAT2 (v2.2.0), and duplicate reads were removed using the Picard (v3.0) ‘MarkDuplicates’ module. For visualization, de-duplicated BAMs were converted to bigWig using deepTools (v3.5.0) bamCoverage (--binSize 10 --normalizeUsing RPKM) and viewed in IGV (v2.12.2). Peaks were called with MACS2 (v2.2.6; -f BAMPE -q 0.01 --keep-dup) using IgG as background, and annotated with ChIPseeker (v1.22.1). Peak files were annotated using ChIPseeker (v1.22.1), and heatmaps were created for each group. No differences were seen between male and female mice; hence, all analyses combined the two sexes. The Homer software was used to perform Known Motif Enrichment analyses and motif discovery (v4.9).³⁸ The nuclei from mice from Cohort 1 and Cohort 2 were processed, sequenced separately, and the data were combined for final bioinformatic analyses.

Supplementary Material

• Synthetic routes and experimental procedures to generate new analogs for ΔFOSB inhibitors

• Supplemental Figures:

  • Fluorescence Polarization Dose Response Curves for all 22 compounds tested
  • Cell-based AP-1 reporter assays for the 16 compounds not shown in Fig. 3
  • Cell viability assays for the 16 compounds not shown in Fig. 4
  • Expression of ΔFOSB in the dentate gyrus of mice treated with vehicle or YL0441

• Summary Table of the in vitro compound testing for the panel of 22 analogs

ABBREVIATIONS

AD

Alzheimer’s disease

AP-1

Activator protein 1

APP

amyloid precursor protein

BSA

bovine serum albumin

bZIP

basic leucine zipper

CI

confidence interval

CL

clearance

DMSO

dimethyl sulfoxide

DOPA

L-3,4-dihydroxyphenylalanine

DRC

dose-response curve

DTT

dithiothreitol

D/V

dorsal/ventral

FBS

fetal bovine serum

FP

fluorescence polarization

GO

gene ontology

HEPES

4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid

IPTG

isopropyl-D-thiogalactoside

LB

Luria-Bertani

LC

liquid chromatography

MOI

multiplicity of infection

NTA

nickel-nitrilotriacetic acid complex

pAG-MNase

protein A-G micrococcal nuclease

PBS

phosphate-buffered saline

PDB

Protein Data Bank

pfu

plaque-forming unit

PMSF

phenylmethylsulfonyl fluoride

SEM

standard error of the mean

TAMRA

tetramethylrhodamine

TCEP

tris(2-carboxyethyl)phosphine

TEV

tobacco etch virus

TOF

time of flight

Tris

tris(hydroxymethyl)aminomethane

DATA AVAILABILITY

All CUT&RUN sequencing data generated in this study have been deposited in the NCBI Gene Expression Omnibus (GEO) under accession number GSE319820.