DMF (222180250, Across), MMF (AK114632, ARK Pharm), and reduced glutathione (B96824, Innochem) were obtained through the Innochem reagent procurement platform. Lipopolysaccharide (LPS) from Escherichia coli 055:B5 (L2880, Sigma) and glutathione monoethyl ester (G1404, Sigma) were purchased from Sigma‒Aldrich. The non-covalent Nrf2 activator KI696 was kindly supplied by Dr. D.L. Yin of Institute of Materia Medica. The antibodies used for western blotting are listed in Supplementary Table S1 (Supplementary Material). The plasmid pGL4.13-Nfkbiz-promoter and pGL4.13-Nfkbiz-3UTR were constructed by YouBio (Changsha, China). The insertion sequences in both plasmids are listed in Supplementary Table S2 (Supplementary Material). The plasmid pSV40-Nfkbiz-3TUR, which harbours the mouse Nfkbiz ORF and its 3′-UTR elements, was constructed in our laboratory. The plasmid pcDNA3.1-mNrf2-6 × His was purchased from Miaolingbio (Wuhan, China). The plasmid pRL-TK was kindly provided by the group of Dr. Z.W. Hu of Institute of Materia Medica. All plasmids were transfected into cells using Lipofectamine 3,000 reagent (L3000008, Thermo Fisher Scientific).

The RAW264.7 cells were kind gifts from Dr. T.T. Zhang of the Institute of Materia Medica. Cells were cultured in complete DMEM containing 10% FBS (SE100-B, Vistech) and penicillin‒streptomycin (P1400, Solarbio). HEK293-ARE reporter cells harbouring pGL4.37 constructs were cultured in complete DMEM supplemented with 300 μg/mL hygromycin B (B29290, Innochem). The cells were subcultured every 2–3 days.

BMDCs were prepared from 6 to 8-week-old wild-type or Nrf2^−/− C56BL/6 mice according to the CSH protocol with modifications (Weischenfeldt and Porse, 2008). Briefly, bone marrow cells were collected from the femurs and tibias of mice and then cultured in L929-conditioned medium (L-medium) at an initial density of 2 × 10⁶ cells/mL. An equal volume of fresh L-medium was added on day 3. On day 6, half of the culture medium was replaced with fresh L-medium. Mature BMDMs were harvested at day 8 and treated as designed.

Cells were treated as designed in 96-well plates. Then, 10 μL of MTT solution was added to each well and incubated at 37°C for 4 h. The culture medium was carefully discarded, followed by the addition of 150 μL of DMSO per well to dissolve the purple formazan crystals on a shaker. The absorbance was measured at a wavelength of 570 nm.

The culture medium was added to an equal volume of 1% (w/w) sulfanilamide solution (S108473-100 g, Aladdin) and incubated at room temperature for 5 min. Another equal volume of 0.1% (w/w) N-(1-naphthyl) ethylenediamine solution (N113103-5 g, Aladdin) was then added to the mixture and incubated for another 5 min at room temperature. The absorbance was measured at 525 nm. The concentration of nitrite in the samples was calculated according to the standard curve prepared in parallel using sodium nitrite.

Total RNA was isolated from cells or blood samples using the TransZol UP RNA extraction reagent (ET111-01, TransGen Biotech). The RNA was converted to cDNA using Hifair^® III first Strand cDNA Synthesis SuperMix (11141ES60, Yeasen). After reverse transcription, PCRs were performed in an ABI 7900HT Real-Time PCR System (Applied Biosystems) with Hieff UNICON^® Power qPCR SYBR Green Master Mix (11197ES08, Yeasen). The PCR protocol involved denaturation at 95°C for 10 s and combined annealing and extension at 60°C for 30 s over 40 cycles. The melting curve was then generated to ensure that the amplification in each reaction was specific. Actb was used as the internal control. Fold changes in RNA levels were calculated using the ΔΔCt method. The PCR primers are listed in Supplementary Table S3 (Supplementary Material).

Cells or tissue samples were washed with cold PBS twice and then lysed with cold radioimmunoprecipitation assay (RIPA) buffer (P0013B, Beyotime) supplemented with protease inhibitor cocktail (4693116001, Roche). Equal amounts (30 μg/lane) of denatured cell lysate were separated by SDS‒PAGE and then transferred to a polyvinylidene difluoride membrane (IPVH00010, Millipore). Target proteins on the membrane were probed with antibodies (see Reagents) and detected using the ImageQuant LAS 500 imager (GE Healthcare).

Sixto eight-week-old male C57BL/6J mice were randomly divided into 3 groups and administered 0.5% CMC-Na vehicle, 100 mg/kg DMF (suspended in 0.5% CMC-Na), or 100 mg/kg MMF (dissolved in 0.5% CMC-Na) twice a day for 1 week. Blood was then drawn from retro-orbital veins into heparinized tubes and diluted with RPMI-1640 medium supplemented with 100 U/mL heparin at a ratio of 1:1 (v/v). The diluted blood was seeded into 48-well plates and treated ex vivo with LPS for 6 h. Afterwards, the supernatant was collected for ELISA, and the blood cells were collected for qPCR analysis.

The levels of TNF-α and IL-6 in the samples were determined using the Mouse TNF ELISA Suit (555268, BD Biosciences) and Mouse IL-6 optEIA ELISA Suit (555240, BD Biosciences). The operations were conducted following the manufacturer’s instructions.

GSH and test compounds were diluted with buffer A (400 mM Tris-HCl, pH = 7.4) to 2 mM and 200 μM, respectively, and warmed at 37°C. Equal volumes of prewarmed GSH and the test compounds were mixed in a tube and incubated at 37°C. An aliquot of 100 μL of the reaction mixture was sampled every 2 min and immediately added to 100 μL of 2 mM DTNB prepared in buffer B (buffer A with 1 mM EDTA) to stop the reaction. The absorbance was measured at 412 nm.

Cell samples were harvested and lysed by the snap freezing-thawing method. After deproteinization, the supernatant was collected for the determination of GSH and GSSG using the GSH and GSSG assay kit (S0053, Beyotime). The precipitate, which contained denatured proteins, was dissolved in 0.1 M NaOH, and the protein concentration was measured using the Pierce™ BCA protein assay kit (23227, Thermo Scientific). Finally, the amount of intracellular GSH or GSSG was normalized using the corresponding protein concentration.

For the ARE-luciferase reporter assay, HEK293-ARE reporter cells were seeded into 96-well plates at a density of 5 × 10⁴ cells/mL and treated as indicated. Luciferase activity was detected using Steady-Glo^® Luciferase Assay Reagent (E2520, Promega) according to the manufacturer’s instructions.

For the dual-luciferase reporter assay, RAW264.7 cells were transfected with pGL4.13-Nfkbiz-promoter or pGL4.13-Nfkbiz-3UTR together with pRL-TK as the internal control for 24 h. Cells were then treated as indicated and harvested. The dual-luciferase activity was determined using the Duo-Lite Luciferase Assay Reagent (DD1205-01, Vazyme).

Chromatin for ChIP was prepared from RAW264.7 cells by fixing the cells in 1% formaldehyde for 10 min, followed by quenching with 125 mM glycine for 5 min. The cell pellet was resuspended and lysed with ChIP lysis buffer and sonicated to generate 200–1,000 bp DNA fragments with an ultrasonic sonicator. The antibodies used in this study for immunoprecipitation were rabbit monoclonal anti-STAT3 and anti-NF-κB antibodies. The precipitated DNA-chromatin products were purified, and the DNA levels were quantified by qPCR. The DNA levels are expressed as the percentage of input DNA.

Data are presented as the means ± S.E.M.s. The Shapiro‒Wilk normality test was used to assess whether data were normally distributed. Statistical analysis, as indicated in each figure legend, was performed using the GraphPad Prism 7.00 software (San Diego, CA, USA; RRID: SCR_000306). Differences at p < 0.05 were considered statistically significant.

We first confirmed the effects of DMF on macrophage activation by coincubation of RAW264.7 cells with DMF and LPS. The nitrite level in the culture medium was taken as an indicator of RAW264.7-cell activation. As previously reported (To et al., 2015), DMF concentration-dependently inhibited LPS-induced nitrite release after 24 h treatment in RAW264.7 cells without affecting cell viability (Figures 1A, B). In contrast, its metabolite, MMF, demonstrated only minor effects (Figure 1A). When the coincubation time was shortened to within 6 h, only DMF was able to inhibit nitrite production (Figure 1C). Since the half-life of DMF in human serum was estimated to be 0.37 h (approximately 22 min) (Litjens et al., 2004b), the DMF incubation time was adjusted to the physiologically relevant exposure time. As shown in Figure 1D, DMF significantly suppressed the release of nitrite in RAW264.7 cells even when the incubation time was shortened to 15 min. In addition, significant inhibitory effects were also observed when cells were briefly exposed to lower concentrations of DMF for multiple times (Figure 1E), which mimics the long-term applications of DMF in patients. The above data showed that DMF but not MMF exhibits rapid inhibition on macrophages activation, suggesting that DMF may have distinctive anti-inflammatory effects.

To explore the inhibitory effect of DMF on inflammation in greater detail, the transcription levels of proinflammatory genes, including Tnf, Il-1b, Il-6, and Nos2 (Figures 2A–D), were measured. Interestingly, short-term exposure to DMF significantly inhibited the transcription of Il-6 (Figure 2C) and Nos2 (Figure 2D) but not Tnf (Figure 2A) or Il-1b (Figure 2B). Considering that both Il-6 and Nos2 transcription belong to the secondary transcription program in the LPS/TLR4 pathway (Figure 2E) (Medzhitov and Horng, 2009; Ramirez-Carrozzi et al., 2009), the mRNA levels of transcription factors (TFs) that regulate the secondary transcription program, including Fosb, Nfkbid, Cebpd, Nfkbiz, Atf3, Junb, and Irf1(Figures 2F–L), were detected. Among these TFs, DMF inhibited the transcription of Nfkbiz mRNA to the greatest extent (Figure 2F). Additionally, western blotting confirmed that DMF inhibited the expression of IκBζ (the protein product of the Nfkbiz gene) in a dose-dependent manner (Figure 3A); however, the metabolite MMF did not show similar effects (Figure 3B). The inhibitory effect of DMF on IκBζ expression was also observed when the exposure time was shortened to 15 min (Figure 3C). Transient exposure to DMF multiple times also inhibited LPS-induced expression of IκBζ (Figure 3D). Importantly, in LPS-treated RAW264.7 cells with ectopic IκBζ expression, DMF failed to inhibit the transcription level of Nos2, further confirming that IκBζ is the downstream effector in the anti-inflammatory action of DMF (Figure 3E). Taken together, these data indicate that IκBζ may be a key target underlying the rapid anti-inflammatory effects of DMF.

If DMF indeed has a direct impact on immune cells in vivo, the anti-inflammatory effects of DMF should be distinct from those of its metabolite, MMF, in mice. The above results show that DMF, but not MMF, strongly inhibited LPS-induced macrophage activation in vitro. We next employed a whole-blood assay comparing the effects of DMF and MMF on inflammation in vivo to test the idea above (Figure 4A). Mice were orally administered the same dose of DMF or MMF for 7 consecutive days. Then, blood samples were collected, and their reactivity to LPS stimulation was profiled ex vivo. The whole-blood assay revealed that blood cells from mice treated with DMF or MMF produced decreased levels of TNFα and IL-6 after LPS stimulation (Figures 4B, C). The transcription of Tnf and Il6 mRNA was also decreased in both groups (Figures 4D, E). However, it should be noted that in comparison with MMF, DMF exerted a more profound suppressive effect on the production of IL-6 at both the protein and mRNA levels (Figures 4C, E). Notably, the transcription of Nos2 and Nfkbiz were both inhibited by oral DMF but not by oral MMF (Figures 4F, G), which were consistent with the in vitro results. These results revealed that the anti-inflammatory effects of DMF and MMF differ in mice, suggesting that MMF dose not account for all of the effects of DMF in vivo.

We next attempted to explore the mechanisms underlying DMF inhibition on LPS-induced IκBζ expression. Both DMF and MMF are prone to react with nucleophiles through Michael addition, based on their structures containing α,β-unsaturated carbonyls (Figure 5A). GSH is one of the most abundant nucleophiles within cells and plays an important role in regulating the proinflammatory activation of macrophages (Buchmuller-Rouiller et al., 1995; Muri and Kopf, 2021). However, under physiological conditions (pH 7.4°C and 37°C), DMF, but not MMF, extensively reacted with GSH within minutes (Figure 5A). Additionally, DMF quickly depleted intracellular GSH without increasing the oxidized GSSG (Figures 5B, C). We hypothesized that rapid GSH depletion by DMF might account for its quick inhibition of LPS-induced IκBζ expression. To test this idea, the GSH synthesis inhibitor buthionine sulfoximine (BSO) was applied to mimic the GSH-depleting effect of DMF. Notably, BSO suppressed the expression of IκBζ (Figure 5D). In addition, glutathione monoethyl ester (GSH-MEE), a membrane-permeable GSH derivative, was supplied to DMF-treated cells to restore the intracellular GSH pool. GSH-MEE supplementation increased nitrite production and restored IκBζ expression in DMF-treated cells (Figures 5E, F). Taken together, these results confirm the regulatory role of GSH in IκBζ expression and indicate that DMF may inhibit IκBζ expression through GSH depletion.

A As DMF has been characterized as a potent covalent Nrf2 activator, we wondered whether short-term exposure to DMF is enough to activate the Nrf2 pathway. Using the luciferase reporter assay, we observed that 12 h of exposure to DMF induced maximal (approximately 6-fold) luciferase signal enhancement in cells stably expressing the NRF2/ARE-responsive luciferase reporter gene (Figure 5A). Although the intensity of the luciferase signal decreased with shorter exposure times, 15 min of exposure to DMF still resulted in half-maximal (∼3-fold) signal enhancement (Figure 6A), indicating that short-term exposure to DMF is sufficient to activate the Nrf2 pathway. Activation of the Nrf2 pathway by short-time exposure to DMF was further confirmed by western blot, as indicated by the enhanced expression of HO-1 in cells (Figure 6B). However, short-term exposure to the metabolite MMF failed to activate the Nrf2 pathway, further suggesting that the agonistic effect of DMF on the Nrf2 pathway is one of the distinctive mechanisms by which it exerts anti-inflammatory effects.

Next, we investigated whether Nrf2 plays a role in the inhibitory effect of DMF on IκBζ expression. To test this notion, RAW264.7 cells were treated with KI696, a non-covalent Nrf2 activator that activates Nrf2 through a mechanism different from that of DMF (Davies et al., 2016). KI696 activated the expression of HO-1 in RAW264.7 cells and inhibited the expression of IκBζ in LPS-treated RAW264.7 cells (Figure 6C). Taken together, the above data suggest that Nrf2 activation is also a mechanism by which DMF inhibits IκBζ expression.

To investigate whether DMF is likely to regulate IκBζ expression beyond Nrf2 activation, the effects of DMF were examined in both Nrf2^+/+ and Nrf2^−/− BMDMs. The lack of functional Nrf2 in the Nrf2^−/− BMDMs was confirmed by the lower levels of Hmox1 and Nqo1 mRNA in both resting and LPS-activated states (Figures 7A, B). Consistent with the results in RAW264.7 cells, LPS stimulation significantly upregulated both the mRNA and protein levels of IκBζ in Nrf2^+/+ and Nrf2^−/− BMDMs (Figures 7C, D). In Nrf2^−/− BMDMs, there was a higher level of IκBζ than in Nrf2^+/+ BMDMs (Figures 7C, D), further validating the regulatory role of Nrf2 in IκBζ expression. Importantly, DMF also significantly suppressed IκBζ expression in the Nrf2^−/− BMDMs (Figures 7C, D), whereas the metabolite MMF exerted little impact on IκBζ expression in both the Nrf2^−/− and Nrf2^+/+ BMDMs (Figure 7D). The data above suggest that the mechanisms by which DMF inhibits IκBζ may involve pathways other than Nrf2.

The full induction of IκBζ expression requires both the activation of its promoter and the stabilization of its mRNA (Muromoto et al., 2019). To investigate the mechanism of DMF-regulated IκBζ inhibition, the promoter region of the Nfkbiz gene and the 3′-UTR elements that govern the stability of Nfkbiz mRNA were introduced into luciferase reporter constructs (Figure 8A). In the Nfkbiz-promoter-luciferase assay, DMF-treated cells showed a decreased luciferase signal upon LPS stimulation (Figure 8B). In the luciferase-3′-TUR assay, DMF failed to suppress the luciferase signal induced by LPS stimulation (Figure 8C). These results revealed that DMF suppresses the promoter activity of Nfkbiz but not the stability of its mRNA.

It has been reported that NF-κB and STAT3 govern Nfkbiz promoter activity (Muller et al., 2018; Muromoto et al., 2019). For NF-κB, DMF delayed the early phosphorylation of p65 after 5 min of LPS stimulation but did not affect later phosphorylation events (Figure 8D). DMF-treated cells exhibited delayed STAT3 phosphorylation, a hallmark of STAT3 protein activation, in response to LPS challenge (Figure 8E). Moreover, DMF also reduced the nuclear translocation of phosphorylated STAT3 (Figure 8F). To further investigate the role of DMF in regulating Nfkbiz expression, we performed ChIP‒qPCR assays using anti-STAT3 or anti-NF-κB Ab. Markedly increased binding to the Nfkbiz promoter was observed for STAT3 and NF-κB in the cell response to LPS treatment. However, DMF only significantly abrogated the binding of constitutive STAT3, but not NF-κB, to the IκB-ζ promoter region (Figure 8G).

Together, the results mentioned above indicate that STAT3 signalling is involved in DMF-mediated Nfkbiz transcription inhibition.