Abstract

   Citrus fruit has long been considered a healthy food, but its role and
   detailed mechanism in lifespan extension are not clear. Here, by using
   the nematode C. elegans, we identified that nomilin, a bitter-taste
   limoloid that is enriched in citrus, significantly extended the
   animals’ lifespan, healthspan, and toxin resistance. Further analyses
   indicate that this ageing inhibiting activity depended on the
   insulin-like pathway DAF-2/DAF-16 and nuclear hormone receptors
   NHR-8/DAF-12. Moreover, the human pregnane X receptor (hPXR) was
   identified as the mammalian counterpart of NHR-8/DAF-12 and X-ray
   crystallography showed that nomilin directly binds with hPXR. The hPXR
   mutations that prevented nomilin binding blocked the activity of
   nomilin both in mammalian cells and in C. elegans. Finally, dietary
   nomilin supplementation improved healthspan and lifespan in
   D-galactose- and doxorubicin-induced senescent mice as well as in male
   senescence accelerated mice prone 8 (SAMP8) mice, and induced a
   longevity gene signature similar to that of most longevity
   interventions in the liver of bile-duct-ligation male mice. Taken
   together, we identified that nomilin may extend lifespan and healthspan
   in animals via the activation of PXR mediated detoxification functions.

   Subject terms: Receptor pharmacology, Drug regulation, Preclinical
   research
     __________________________________________________________________

   The increase in detoxification gene expression is a common
   transcriptome marker for longevity interventions. Here, the authors
   show that nomilin extends lifespan and healthspan in animals through
   activation of PXR regulated detoxification functions.

Introduction

   Delaying the ageing process is one of the major aims of modern
   biomedical research. Manipulation of multiple signalling pathways or
   dietary restriction has been shown to extend the lifespan and
   healthspan in animal models, but these methods are either not practical
   or not satisfactory for application to the general
   population^[64]1–[65]6. Increasing evidence has shown that the
   expression of detoxification enzyme genes and resistance to toxins are
   increased in long-lived flies, worms and rodents^[66]7. Detoxification
   gene expression is increased in the liver and shows more resistance to
   hepatotoxins in long-lived Little mice, Ames dwarf mice, Snell dwarf
   mice, growth hormone receptor knockout mice^[67]8–[68]11, and dietary
   and methionine-restriction mice^[69]12–[70]15. This phenomenon is also
   observed in long-lived Caenorhabditis. elegans (C. elegans) and
   Drosophila melanogaster, which are also more resistant to
   xenobiotics^[71]16–[72]18. Recently, Tyshkovskiy et al. have shown that
   the expression of drug metabolism and detoxification genes, such as
   cytochrome P450 enzymes (CYPs) and glutathione-S-transferases (GSTs),
   was increased in the livers of mice that underwent 17 known
   lifespan-extending interventions, indicating that targeting
   detoxification may be a useful longevity intervention therapy^[73]19.

   The expression of xenobiotic detoxification enzyme genes is
   transcriptionally regulated by nuclear hormone receptors (NHR). In
   mammals, pregnane X receptor (PXR) is a major regulator of the
   expression of drug metabolism and xenobiotic detoxification
   genes^[74]20. In C. elegans, NHR8 and DAF-12 transcriptionally regulate
   the expression of these genes in order to excrete toxins. NHR-8 and
   DAF-12 are required for the longer lifespan and healthspan in C.
   elegans^[75]1,[76]21–[77]24, indicating that NHR-8 and DAF-12 are
   required for the longevity of C. elegans. Hence, we propose that the
   activation of the NHRs mediating detoxification gene expression may be
   a strategy for lifespan-extending intervention and ageing-related
   diseases.

   Nomilin is a naturally-occurring compound in citrus fruits such as
   lemons, grapefruits, oranges as well as in tangerine seed and
   peel^[78]25,[79]26. A number of studies have showed that nomilin may
   exert a variety of pharmacological properties including anti-cancer,
   anti-inflammatory, anti-obesity, anti-viral, anti-oxidant,
   immune-modulatory and neuro-protective effects^[80]25,[81]26. Here, we
   show that nomilin is a PXR agonist, and may extend lifespan and
   healthspan in C. elegans and mice via NHR-regulated detoxification
   functions, and induce the common transcriptome markers seen in the
   liver of mice in response to most lifespan-extending interventions.

Results

Nomilin extends lifespan in wild type (WT) C. elegans

   Recently, certain metabolites/small molecules have been shown to have a
   potentially useful ageing inhibiting ability in the nematode C.
   elegans, and have been reported to have a similar effect in
   mammals^[82]27–[83]30. Inspired by those findings, we searched for
   components present in oranges that have a longevity intervention
   effect, since orange extracts have been reported to extend lifespan and
   healthspan in C. elegans^[84]31,[85]32. Among many known components, we
   were particularly interested in nomilin, a limonoid enriched in citrus
   fruits^[86]26, because it has also been suggested to have certain
   health-promoting and disease-preventing properties^[87]33–[88]37.
   Surprisingly, we found that nomilin extended the lifespan of WT N2 C.
   elegans in a dose-dependent manner. Treatment with 25, 50 and 100 μM
   nomilin significantly increased the average lifespan by 9.4%, 24% and
   24%, respectively (Fig. [89]1a, & Supplementary Table [90]1). However,
   when the concentration was increased to 200 μM, nomilin showed the
   lower lifespan extending effects, implying that higher concentration of
   nomilin may have a side effect on C. elegans (Supplementary
   Fig. [91]S1a, Supplementary Table [92]S2). Then, we compared the
   lifespan-extending effects of nomilin and its analogue limonin. Limonin
   displayed less effects on the survival time when compared to nomilin
   (Supplementary Fig. [93]S1b, Supplementary Table [94]S2), indicating
   the structure specificity of nomilin. In addition, the accumulation of
   lipofuscin, a biomarker of senescence in C. elegans, was also
   significantly reduced under nomilin supplementation (Supplementary
   Fig. [95]S1c, d). Locomotion behaviours in aged adults (which have been
   commonly used to analyse the ageing-related health-span of C. elegans),
   such as body-bend, head-swing, and pharynx-pumping, were also
   significantly improved under nomilin treatment (Supplementary
   Fig. [96]S1e-h). Moreover, like many long-lived C. elegans models,
   nomilin-treated animals also showed increased resistance to heat and
   oxidative stress (Supplementary Fig. [97]S1i, j). Taken together, these
   data suggest that nomilin may delay the ageing process and extend
   lifespan and health span in C. elegans.

Fig. 1. Nomilin extends the lifespan of C. elegans via a
daf-2/daf-16-dependent pathway.

   [98]Fig. 1
   [99]Open in a new tab

   a, b Lifespan curves showing the lifespan-extending effects of nomilin
   at various concentrations on WT N2 C. elegans (a), daf-2(e1370) (b).
   Nomilin significantly extended the lifespan of WT, but not the daf-2,
   daf-16, and sir-2.1 animals. raga-1 partially blocked the lifespan
   extension effect of nomilin. c Representative fluorescent pictures
   showing the DAF-16::GFP nuclear localisation in control (upper) and
   nomilin-treated (bottom) animals. d The statistical data of (c,
   two-tailed unpaired Student’s t-test, n = 10/each, ***p < 0.001 vs.
   control group). e Lifespan curves of nomilin in daf-16 (mu86). f
   Quantification of mRNA levels of daf-16 downstream genes in C. elegans
   (two-tailed unpaired Student’s t-test, n = 3/each). g, h Lifespan
   curves of nomilin in sir-2.1(ok434) (g), and raga-1(ok386) mutants (h).
   All data were expressed as mean ± SEM. Detailed information is shown in
   Supplementary Table [100]S1.

   It has been reported that reduction of food intake or dead bacteria
   feeding may extend lifespan in C. elegans^[101]38,[102]39. To exclude
   the possibility that the lifespan-extending effects of nomilin were due
   to reduced food uptake, we performed three experiments. First, we found
   that nomilin supplementation did not affect the growth speed of E. coli
   OP50 (Supplementary Fig. [103]S1k), the lab food of C. elegans. Second,
   we assessed the lifespan of animals grown on heat-killed OP50, and
   found that dead bacteria extended the lifespan of worms and nomilin
   further increased the lifespan (Supplementary Fig. [104]S1l,
   Supplementary Table [105]S2). Third, a food-taxing experiment showed
   that C. elegans did not avoid the nomilin-supplemented bacteria lawn
   (Supplementary Fig. [106]S1m, n). These data suggest that the
   lifespan-extending effects of nomilin are not likely to result from the
   reduction of food-intake or suppression of bacteria growth. Moreover,
   because infertility may extend the worm lifespan, we checked the
   average brood size and the offspring number of worms and found that
   there were no significant differences between nomilin-treated and
   control animals (Supplementary Fig. [107]S1o, p). Thus, these results
   indicate that nomilin extends C. elegans lifespan and healthspan
   directly.

DAF-2 and DAF-16 are required for the extending lifespan effect of nomilin

   We then tested which specific signalling pathway plays a major role in
   nomilin-associated lifespan extension. The insulin/insulin-like growth
   factor signalling (IIS) pathway plays essential roles in longevity and
   the resistance of the body to various stressors, such as oxidative
   stress and xenobiotic stress, in C. elegans^[108]40,[109]41. We found
   that nomilin could not further extend the lifespan in C. elegans
   insulin-like peptide receptor mutant daf-2(-) (Fig. [110]1b). Moreover,
   nomilin supplementation significantly promoted nuclear translocation of
   DAF-16::GFP, a FOXO transcription factor downstream of DAF-2
   IIS^[111]42 (Fig. [112]1c, d), and daf-16(-) also completely blocked
   the lifespan extension effect of nomilin (Fig. [113]1e). To confirm
   that DAF-16 signalling is involved in nomilin function, the mRNA levels
   of DAF-16 downstream targets were tested. The results showed that the
   mRNA levels of sod-2, sod-3, clk-2 and lin-2 were increased by nomilin
   (Fig. [114]1f). Mutation of C. elegans histone deacetylase SIRT/sir-2.1
   (Fig. [115]1f) or mTORC1 signalling component RagA/B homologue raga-1
   (Fig. [116]1g), another two well-known ageing pathways, could not fully
   block the lifespan extension effect of nomilin. Interestingly, we found
   that nomilin did not enhance the dauer formation either in the WT or in
   the daf-2(e1370) mutant background (Supplementary Fig. [117]S1q, r). We
   thought that the reason normilin mainly affected longevity instead of
   dauer formation, possibly because it targeted the intestinal cells and
   affected the local IIS activity (Fig. [118]1c, d). It was consistent
   with the report that the intestinal IIS pathway mainly regulates
   longevity, but not the dauer formation process, while the neuronal IIS
   pathway does the opposite^[119]43. These data suggest that the
   lifespan-extending effects of nomilin in C. elegans mainly depend on
   the intestinal IIS pathway.

Nomilin activates detoxification enzymes and protects C. elegans from
multiple toxins

   To better understand the mechanistic role of nomilin in the lifespan
   extension process, we performed a literature search and found several
   studies that suggest that nomilin can enhance the expression of the
   detoxification genes glutathione S-transferase and quinone reductase
   that, which are responsible for the detoxification of
   quinone-containing compounds in the liver and intestine of
   rodents^[120]44,[121]45. To investigate whether these detoxification
   roles of nomilin are conserved in C. elegans, we examined the
   detoxification gene expression under nomilin supplementation in worms
   using real-time reverse transcription-polymerase chain reaction
   (RT-PCR). As expected, the expression of multiple detoxification genes
   was increased, such as phase I genes cyp35a3-5 and cyp37a1; phase II
   genes gst-4 and ugt44; and phase III genes gpg-3 and gpg-12/13
   (Fig. [122]2a). We further tested whether nomilin could protect C.
   elegans from toxins such as colchicine, chloroquine, paraquat and
   methyl mercury chloride (MeHgCl), which are involved in the ageing
   process and the pathogenesis of senile diseases^[123]46–[124]50.
   Animals survived better under nomilin treatment in a dose-dependent
   manner (Fig. [125]2b–e), indicating that, in addition to lifespan
   extension, nomilin also protects the worms from many toxins.

Fig. 2. Nomilin executes its ageing inhibiting and detoxification abilities
via nuclear hormone receptors nhr-8/daf-12 in C. elegans.

   [126]Fig. 2
   [127]Open in a new tab

   a Quantification of mRNA levels of detoxifying genes in C. elegans
   (two-tailed unpaired Student’s t-test, n = 3/each, each sample contains
   about 1000 worms). b–e Survival curves showing the protective effects
   of nomilin on worms with the indicated genotypes upon various chemical
   toxin treatments (two-way ANOVA test, n = 3/each, ***p < 0.001).
   Nomilin-treated C. elegans were more resistant to chloroquine (b),
   colchicine (c), paraquat (d), and MeHgCl (e) than wild type N2, but not
   nhr-8 (tm1800) or daf-12 (rh61rh411) animals. f–h lifespan curves
   showing the lifespan-extending effects of nomilin on WT, nhr-8 mutant
   (tm1800), and daf-12 mutant (rh61rh411) C. elegans. The detailed
   information is shown in Supplementary Table [128]S3. i Effects of
   nomilin on nuclear trans-localisation of nhr8::daf-16::GFP and
   daf12::daf-16::GFP worms. The worms were treated with 50 μM of nomilin
   from L1 to L4, and 10 animals were examined per condition. j Average
   number of cells with DAF-16::GFP nuclear localisation in nhr-8 and
   daf-12 mutants. All data were expressed as mean ± SEM, ***p < 0.001 vs.
   control group, n = 9 or 13 worms per group.

Nomilin extends lifespan and improves toxin resistance via nuclear hormone
receptors NHR-8 and DAF-12

   We then tried to identify the direct target of nomilin in C. elegans
   through which it exerts its lifespan extension and detoxification
   abilities. Although the InR/DAF-2 pathway is known to increase lifespan
   and toxin resistance^[129]51,[130]52, it is highly unlikely that DAF-2
   binds with nomilin directly, given its nature as an insulin-like
   peptide receptor^[131]41. Instead, nomilin may function via binding
   with certain nuclear hormone receptors (named NHR hereafter), a large
   family of proteins that can interact with small metabolites and
   regulate metabolism and other physiological functions. From the
   literature review, we found that two NHRs (NHR-8 and DAF-12) have been
   reported to play major roles in both lifespan extension and
   detoxification^[132]53,[133]54. We then tested the role of these two
   NHRs during nomilin treatment. Surprisingly, compared to N2 worms
   (Fig. [134]2f), we found that both the loss-of-function mutation of the
   daf-12 worm (rh61rh411) (which causes the loss of the capacity for
   ligand binding and DNA binding)^[135]55 and nhr-8 (tm1800) fully
   suppressed the lifespan extension effect of nomilin (Fig. [136]2g–h,
   Supplementary Table [137]S3). In addition, we found the detoxification
   effect of nomilin also depended either on NHR-8 or DAF-12 (or both, in
   the case of methyl mercury chloride) (Fig. [138]2b–e). To confirm that
   the action of nomilin was mediated by nhr-8 and daf-12, we crossed the
   muIs109 males with daf-12 (rh61rh411) and nhr-8 (tm1800) hermaphrodites
   and obtained a homozygous strain of Pdaf-16::gfp::nhr8 and
   Pdaf-16::gfp::daf-12 worms, who were then treated with nomilin. The
   results showed that nomilin did not promote the nuclear translocation
   of DAF-16 in daf-12 and nhr-8 mutant worms (Fig. [139]2i–j).

   Next, we investigated whether the downstream detoxification enzymes of
   daf-12 and nhr-8 are involved in the lifespan-extending effects of
   nomilin. Indeed, upregulation of most detoxification genes by nomilin
   (Fig. [140]2a) was blocked in nhr-8 and daf-12 mutant (Fig. [141]3a,
   b), indicating that these genes are the targets of nhr-8 and daf-12.
   Then, nomilin-activated genes gst-4, cyp35a3, pgp-3 and pgp-14 were
   knocked down using RNAi in N2 worms, who were then treated with
   nomilin. The results showed that the lifespan-extending effects of
   nomilin were attenuated in gst-4, cyp35a3 and pgp-3 knockdown worms
   when compared to those under RNAi treatment, while deficiency of pgp-14
   that was not regulated by nomilin (Fig. [142]2a) did not change the
   effects of nomilin (Fig. [143]3c–g, Supplementary Table [144]S4).
   Collectively, these data indicate that nomilin extends lifespan through
   the targeting of nuclear hormone receptors NHR-8 and DAF-12.

Fig. 3. Effects of nomilin on the detoxifying gene expression in nhr-8 and
daf-12 mutant C. elegans.

   [145]Fig. 3
   [146]Open in a new tab

   a, b Quantification of mRNA level of genes in nhr-8 (a) and daf-12 (b)
   mutant C. elegans. (two-tailed unpaired Student’s t-test, n = 3/each,
   each sample contains about 1000 worms). c–g Survival curve of nomilin
   in xenobiotic metabolism gene RNAi N2 C. elegans. Synchronised L1 worms
   were fed with E. coli (HT115) containing an empty control vector
   (L4440) until L4, then transferred to plates containing cyp35a3, gst-4,
   pgp-3 or pgp-14 RNAi HT115 with nomilin (50 μM) and DMSO (0.1%) as
   controls. All data were expressed as mean ± SEM. The detailed
   information is shown in Supplementary Table [147]S4.

IIS signalling is involved in detoxification in C. elegans

   The upregulation of xenobiotic detoxification genes is a common
   characteristic of long-lived flies, worms and rodents, some of which
   showed stronger resistance to xenobiotic stressors. It has been
   proposed that the resistance to a broad range of stressors, including
   heat, oxidative stress and xenobiotics, could be a longevity-assurance
   mechanism. In C. elegans, several lines of evidence have suggested that
   detoxification and longevity are coupled. Long-lived daf-2 mutants
   showed a similar transcriptomic signature of increased detoxification
   gene expression to flies and mice; however, resistance to toxins in
   daf-2 mutants has not been studied to date. Thus, we explored the
   detoxification functions of daf-2 and daf-16 mutants. Under challenge
   with paraquat or MeHgCl, daf-2 mutants were more resistant than WT
   worms. In contrast, short-lived daf-16 mutants were more sensitive
   (Supplementary Fig. [148]S2a, b), suggesting that IIS signalling may be
   involved in xenobiotic detoxification.

   To investigate the correlation between daf-2 and nhr-8/daf-12 in
   detoxification, we carried out RNAi against nhr-8 or daf-12 in daf-2
   worms (daf-2::nhr-8 RNAi, daf-2::daf-12 RNAi), and then challenged with
   paraquat or MeHgCl. The results showed that deficiency of nhr-8 and
   daf-12 diminished the detoxification ability of daf-2 worms
   (Supplementary Fig. [149]S2a, b), indicating that the detoxification
   ability of nhr-8/daf-12 and IIS longevity signalling may show crosstalk
   with the lifespan extension function.

Nomilin is a specific PXR agonist

   Next, we tried to identify whether there is a mammalian counterpart of
   C. elegans NHR-8/DAF-12. Previous reports have suggested that
   NHR-8/DAF-12 belong to the NR1 subfamily, a group of NHRs specifically
   functioning in xenobiotic metabolism^[150]7. Because of the limited
   sequence homology between mammalian and C. elegans NHRs, we applied an
   HEK293 cell-based reporter assay to test which mammalian NHRs could be
   activated by nomilin. Among nine well-known human NR1 subfamily NHRs
   (hPPARα, hPPARβ, hPPARγ, FXR, LXRα/β, NRF2, hCAR and hPXR), only hPXR
   could be effectively activated by nomilin (Fig. [151]4a, Supplementary
   Fig. [152]S3a–i). The activation effect of nomilin was similar to the
   known strong hPXR agonist rifampicin (Rif). Moreover, the nomilin
   analogue deacetylnomilin (which lacks an acetyl group at the hydroxyl
   residue on ^2C) displayed similar activation effects (or moderately
   stronger), while another analogue, limonin (which lacks the iconic
   heptatomic lactone ring), did not have any activity (Fig. [153]4a),
   suggesting that the heptatomic lactone ring of nomilin may be essential
   for binding to hPXR. In addition, the Time-resolved fluorescence
   resonance energy transfer (TR-FRET) assay showed that the binding
   between the labelled hPXR ligand and hPXR-LBD was significantly
   attenuated by nomilin in a dose-dependent manner (IC50: 5.8 μM and Kd:
   13.3 μM), more so than by potent PXR agonist T0901317 (IC50: 11.7 nM
   and Kd: 30.1 μM), while deacetylnomilin showed lower affinity (IC50:
   22.7 μM and Kd: 198.3 μM, Fig. [154]4b), suggesting that nomilin may
   compete with the labelled ligand to bind with hPXR. Together, these
   data suggest that nomilin is a specific agonistic ligand of hPXR, the
   potential mammalian ortholog of NHR-8/DAF-12.

Fig. 4. Nomilin is a PXR agonist and the crystal structure for
hPXR^LBD-NCOA1^676–700 bound to nomilin.

   [155]Fig. 4
   [156]Open in a new tab

   a Structure of nomilin and its analogues and hPXR reporter gene assay.
   The results represent three independent experiments. (One-way ANOVA
   test in deacetylnomilin and limonin; and Kruskal-Wallis test in
   nomilin, n = 3/each, the data were shown as means ± SEM. ***p < 0.001
   compared to the control group). b TR-FRET assay. The TR-FRET ratio
   (520/495) was calculated by subtracting the background. c Dimeric human
   PXRLBD-NCOA1676-700 (in blue and red for protomer A and B,
   respectively) and co-activator peptide (in orange) fusion protein. The
   nomilin molecules are shown as stick models and coloured by element. d
   The space-filling model of protomer A was sliced to show the binding
   pocket for nomilin (shown as stick model) and some residues closely
   interacting with nomilin. H407 was shown in its two alternative
   conformations. e The omit Fo-Fc electron density maps for nomilin in
   protomer A (upper) and B (lower) are shown as mesh models and contoured
   to 1.0 σ. f A close view of the binding site of protomer A. The nomilin
   is shown as stick and space-filling models, with the surrounding
   residues shown as a stick model. g The schematic diagram for the
   hPXR-nomilin interaction network. h A comparison between binding
   pockets of hPXR LBD in complex with nomilin and rifampicin. Both hPXRs
   are shown as cartoon models in blue and grey for nomilin-bound and
   rifampicin-bound structures, respectively. The nomilin and rifampicin
   are shown as stick models coloured by element (green-red for nomilin
   and grey-red for rifampicin). The structure model of the
   hPXR-rifampicin complex was generated with coordinates from PDB ID
   1SKX. i hPXR mutations change the effects of nomilin action. The
   plasmids were transfected into HEK293T cells, which were treated with
   nomilin or rifampicin for 24 h (two-tailed unpaired Student’s t-test,
   n = 3/each, the data were shown as means ± SEM, ***p < 0.001 vs.
   control group).

The crystal structure of the nomilin-hPXR complex indicates critical amino
acids for the binding affinity

   To further confirm the direct binding between nomilin and hPXR, and
   discover more information on its structure and activation mechanism, we
   expressed the ligand-binding domian (LBD) domain of hPXR (residues
   130–432) fused with residues 676–700 of nuclear receptor coactivator 1
   (NCOA1^676–700) at the C-terminal as a co-activator peptide. The
   purified hPXR^LBD-NCOA1^676–700 (hPXR chimera hereafter) was
   co-crystallised with nomilin and the structure of the protein-drug
   complex was determined at a resolution of 2.1 Å, allowing a detailed
   observation of the drug-target interaction (Fig. [157]4c and
   Supplementary Fig. [158]3j, k). The hPXR chimera forms a homodimer in
   the purification and crystal structure, as previously reported. The
   dimeric hPXR forms an interface via the β1' strand of the
   characteristic antiparallel β-sheet^[159]56–[160]58. A region of
   non-proteinaceous electron density was identified in the Fo-Fc map of
   the LBDs of each of the hPXR protomers, and the molecular framework of
   nomilin fit the electron density very well (Fig. [161]4d). The electron
   density in protomer A was refined with better continuity than that in
   protomer B (Fig. [162]4e). As shown in Fig. [163]4f, g, the binding of
   nomilin in the hPXR chimera was majorly mediated by the hydrogen bonds
   formed by two carbonyl oxygens with S247 and Q285, as well as the
   strong hydrophobic interaction among the backbone carbons of nomilin,
   and the cavity formed by a series of hydrophobic residues including
   M243, W299, I414 and M425 (Fig. [164]4g and Supplementary Fig. [165]3l,
   m). To exclude that nomilin was modified during the crystallization
   process, the hPXR/nomilin crystal was assayed by mass spectrometry. The
   result did not show different structure in the crystal.

   Two methionine residues, M243 and M425, closely contact both nomilin
   and rifampicin in their crystal structures, respectively^[166]59. In a
   functional assay of hPXR mutants, the binding of nomilin is abolished
   by the mutation of M425 and the function of rifampicin relies more on
   the M243 residue. From Fig. [167]4h, the biphenyl moiety in rifampicin
   interacts with M243 more closely than nomilin does, making rifampicin
   more sensitive to the local spatial variation introduced by the M243Q
   mutation (Supplementary Table [168]S5). In addition, one of the
   structural differences between hPXR bound with nomilin and rifampicin
   is the helix formed by amino acids 193–209, which is well-refined in
   our structure and previously reported structures of hPXR in complex
   with SR12813, clotrimazole, and hyperforin, but absent in the structure
   of hPXR-rifampicin complex (Supplementary
   Fig. [169]S4a)^[170]58,[171]60–[172]62. The structural superposition
   between hPXR bound with nomilin and rifampicin showed the spatial clash
   between the biphenyl moiety of rifampicin and the helix^193–209
   (Supplementary Fig. [173]S4b), suggesting a displacement of the helix
   in the rifampicin-bound state. These data showed the LBD conformational
   change between nomilin-activated and rifampicin-activated states, which
   results in different biological effects.

   To confirm the working model of hPXR, conservative and non-conservative
   mutations were made of the residues critical for nomilin binding, and a
   reporter gene assay was performed on these hPXR mutants. The
   transactivities of all mutant hPXR were lower than that of wild type
   hPXR, which may reflect the lower response of hPXR mutants to
   endogenous PXR agonists (Fig. [174]4i). M243A/Q did not affect the
   nomilin activity. However, S247A/R, M425A/Q and W299R completely
   blocked the activity of nomilin (Fig. [175]4i), indicating that these
   amino acid residues are critical for nomilin-dependent hPXR activation.

Mammalian PXR is a functional ortholog of NHR-8 /DAF-12

   Next, we attempted to verify whether mammalian PXR is a functional
   ortholog of NHR/DAF-12 that mediates lifespan extension and
   detoxification by nomilin. As expected, nhr-8 and daf-12 mutation
   shortened lifespan in both nomilin-treated and control animals
   (Fig. [176]5a–c, Supplementary Table [177]S6). We found that
   overexpression of WT hPXR could partially restore the lifespan effect
   of nomilin in nhr-8 or daf-12 mutant animals (Fig. [178]5a–c,
   Supplementary Table [179]S6), while hPXR^S247R (which blocks the
   binding between nomilin and hPXR) mutation only slightly restored the
   lifespan extension in nhr-8, and completely failed to restore the
   lifespan extension in daf-12 mutants under nomilin treatment
   (Fig. [180]5a–c, Supplementary Table [181]S6). The partial effect of
   hPXR to restore the lifespan extension effect of nomilin was possibly
   due to that mammalian PXR could not fully activate the C. elegans
   target genes. Moreover, to investigate whether hPXR could activate the
   target genes of NHR-8 and DAF-12, nomilin-treated hPXR transgenic nhr-8
   and daf-12 worms were used to test mRNA levels. The results showed that
   nomilin only increased gst-4, pgp-3 and pgp-13 mRNA levels in hPXR
   transgenic nhr-8 worms, and pgp-13 mRNA in hPXR transgenic daf-12 worms
   (Fig. [182]5d, e). These data indicate that hPXR could partially
   compensate for the function of NHR-8 and DAF-12 in mediating
   nomilin-dependent lifespan-extending effects in C. elegans, and the
   activation is dependent on the binding activity between nomilin and
   hPXR. It has been reported that PXR has multiple consensus DNA binding
   sequences including a relatively conserved 3' half-site A-G-T-T-C-A
   sequence^[183]63. Similarly, DAF-12 was also reported to have a similar
   3' half-site A-G-T-T/G-C-A/G DNA binding sequence^[184]64, and NHR-8
   and DAF-12 share significant homology in DNA- and ligand-binding
   domains (DBD; LBD), and have identical residues in the P-box, a motif
   in the first zinc finger that functions in DNA recognition^[185]65.
   Therefore, these may explain why hPXR could partially rescue the
   phenotypes of daf-12(-) mutant. Taken together, these data suggest that
   hPXR is an ortholog of NHR-8/DAF-12 and implicate nomilin in lifespan
   extension in mammals via the activation of hPXR.

Fig. 5. hPXR partially restores the lifespan-extending effects of nomilin in
nhr-8 or daf-12 mutants.

   [186]Fig. 5
   [187]Open in a new tab

   a–c lifespan curves showing the lifespan-extending effects of nomilin
   in C. elegans with indicated genotypes. Nomilin extended the lifespan
   in WT animals overexpressed with vector control (a left) or WT hPXR (a
   middle), but not hPXR^S247R (a hPXR mutant that blocked its binding
   with nomilin) (a right). Similarly, overexpression of WT hPXR (b, c
   middle), but not the vector control (b, c left) or hPXR^S247R (b, c
   right), enabled the maximal lifespan extension of nomilin in nhr-8 or
   daf-12 animals. The detailed data are shown in Supplementary
   Table [188]S6. d, e, The quantification of gene expression levels in
   hPXR transgenic nhr-8 (d) and daf-12 (e) mutant C. elegans. (Two-tailed
   unpaired Student’s t-test, n = 3/each, each sample contains about 1000
   worms. The data were shown as means ± SEM).

Nomilin improves healthspan of D-galactose induced early-senescence mice

   To test whether nomilin could improve healthspan in toxin-induced
   senescence, we used D-galactose to mimic the symptoms of human ageing
   in mice. D-galactose can be oxidized into hydrogen peroxide, which
   increases reactive oxygen species in cells, resulting in ageing of
   multiple organs^[189]66,[190]67. D-galactose induced liver inflammation
   and the inflammatory cells infiltrated into the liver tissues
   (Supplementary Fig. [191]S5a, b). The expression of inflammatory genes
   Tnfα, Il-β and Mcp-1 was induced, and anti-oxidation genes Ho-1, Nrf2
   and Sod-1 were suppressed in the liver of mice treated with D-galactose
   (Supplementary Fig. [192]S5c, d). In contrast, nomilin significantly
   reversed these changes (Supplementary Fig. [193]S5c, d).

   Age-related damage was also observed in the central nervous system.
   D-galactose increased apoptotic cells in the CA1, CA3 and dentate gyrus
   in the hippocampus of the mice, which may result in neurodegeneration,
   while nomilin treatment reduced the numbers of dead cells
   (Fig. [194]6a, b). Cognitive decline is correlated with the change of
   the hippocampus during ageing. Thus, we adopted 8-arm maze to assess
   cognitive functions of the mice. In both short-term and long-term
   memory tests, the mean exploration time of D-galactose-treated mice was
   increased, while the performance rate was significantly reduced
   compared to those of control mice. However, the mean exploration time
   and the performance rate were reversed by nomilin treatment
   (Fig. [195]6c, d, e, f). The lower mobility resulting from impaired
   balance, lower stability and extremity strength is an age-related
   change in elders reflecting the functional decline of organs. The motor
   slowing in aged people is also commonly related to the structural and
   functional alterations of the elder brain^[196]68. Next, we performed
   the pole test and beam balance test in D-galactose-induced
   early-senescence mice; the T-climbing time in the pole test, and
   passing time in the beam balance test were longer than those in control
   mice (Fig. [197]6g, h). Following nomilin treatment, the times of
   T-climbing and passing time were reduced, equivalent to those of
   control mice (Fig. [198]6g, h), suggesting that nomilin could improve
   motor deficits in toxin-induced senescence mice. Gait analysis is a
   sensitive method for evaluating motor functions. Next, we used the
   Catwalk gait analysis system to assess whether nomilin could improve
   D-galactose-induced movement disorders. The number of steps, print
   position of both right and left paws, base of support (BOS) in the hind
   paws, and step cycles of the fore limbs were increased in
   D-galactose-induced mice, while stride length in the hind limbs and
   swing speed in the fore limbs of D-galactose mice were reduced, which
   were all reversed by nomilin treatment (Fig. [199]6i–n).

Fig. 6. Effects of nomilin on apoptosis of the hippocampus, cognitive
capacity and neuromuscular functions in D-galactose-induced mice.

   [200]Fig. 6
   [201]Open in a new tab

   a apoptotic cells in CA1, CA3 and the dentate gyrus of the hippocampus.
   b The quantitation of apoptotic cells in (a). (n = 6–7 for Ctrl,
   n = 7–8 for D-gal, n = 9–10 for D-gal+NML) Exploration time (c,
   n = 7/group) and performance rate (d, n = 7/group) in short-term memory
   test. Exploration time (e, n = 7/group) and performance rate (f,
   n = 7/group) in long-term memory test. g T-climbing down in pole test
   (n = 20 for Ctrl and D-gal, n = 23 for D-gal+NML). h Passing time in
   beam balance test (n = 19 for Ctrl, n = 18 for D-gal, n = 21 for
   D-gal+NML). The number of steps (i, n = 6 for Ctrl, n = 7 for D-gal,
   n = 9 for D-gal+NML), the print position (j, n = 5–6 for Ctrl, n = 6–7
   for D-gal, n = 8 for D-gal+NML), base of support (k, n = 5-6 for Ctrl,
   n = 7 for D-gal, n = 9 for D-gal+NML), step cycles (l, n = 5–6 for
   Ctrl, n = 6–7 for D-gal, n = 9 for D-gal+NML), stride length (m, n = 6
   for Ctrl, n = 7 for D-gal, n = 9 for D-gal+NML) and swing speed (n,
   n = 6 for Ctrl, n = 7 for D-gal, n = 9 for D-gal+NML) in gait analysis.
   The mice were treated with D-galactose (125 mg/kg/day) and nomilin for
   7 weeks. Scale bar = 100 μm (a). The data were shown as mean ± SEM.
   p-values were determined by one-way ANOVA test (b–i). p-values were
   determined by two-way ANOVA test (j–n). ###p < 0.001 vs the control
   group; ***p < 0.001 vs the D-galactose group. BOS base of support, NML
   nomilin, LF left forelimb, RF right forelimb, LH left hindlimb, RH
   right hindlimb, Ctrl Control, D-gal D-galactose, NML nomilin.

   Then, we investigated whether nomilin activated mPXR downstream
   signalling in D-galactose-treated mice. The expression levels of mPXR
   downstream targets Cyp3a11/13, Cyp2d22, Cyp2e1, Cyp8b1, Cyp51, Gsta1/2,
   Tyw1 and Por in the liver were upregulated by nomilin (Supplementary
   Fig. [202]S6a), suggesting that it also activates mPXR in mice. To
   confirm that the mPXR target gene expression was increased by nomilin
   treatment, a Western blot was performed to assay the protein levels of
   Cyp3a11, Cyp51a1 and Gsta1. The results showed that the protein levels
   of Cyp3a11 and Gsta1 in the liver were increased by nomilin treatment
   (Supplementary Fig. [203]S6b, c), supporting that mPXR signalling was
   activated by nomilin. Taken together, the data suggest that nomilin may
   improve toxin-induced senescence, probably via the activation of
   detoxification function in mice.

   PXR and its downstream detoxifying enzymes are also expressed in the
   brain^[204]69–[205]71. Thus, we were curious to know whether nomilin
   can activate mPXR signalling in the hippocampus of D-galactose-treated
   mice. The analysis of mPXR target gene expression showed that the mRNA
   levels of Gsta1, Mdr3, Cyp8b1 and Cyp27a1 were reduced in the
   hippocampus, suggesting that the detoxification function was inhibited
   by D-galactose-treatment (Supplementary Fig. [206]S6d). Nomilin
   significantly increased the mRNA levels of Gsta1, Gsta2, Mdr3, Cyp8b1,
   Cyp27a1 and Cyp2d22 in the hippocampus (Supplementary Fig. [207]S6d).
   These data suggest that nomilin may also increase mPXR signalling in
   the brain of mice.

PXR deficiency diminishes healthspan-extending effects of nomilin in mice

   To confirm nomilin targeting of PXR, we analysed the
   healthspan-extending effects of nomilin in PXR knockout mice treated
   with D-galactose. The PXR^-/- mice exhibited longer passing time in the
   beam balance test (Fig. [208]7a) and T-climbing time in the pole test
   (Fig. [209]7b). The dwelling time on the rotating rod test was reduced
   (Fig. [210]7c), while the number of falls from the rotating rod was
   increased (Fig. [211]7d) due to D-galactose treatment, which are
   similar to that in wild type mice. In gait test, the hind limb stance
   width was increased, and the hind limb stride length was decreased in
   D-gal treated PXR^-/- mice, whereas the intervention of nomilin did not
   ameliorate gait instability in PXR^-/- mice (Fig. [212]7e, f). The
   proportion of number of entries and exploration time of D-gal-induced
   mice were decreased compared to control mice, indicating D-gal-induced
   PXR^-/- mice also have significant memory dysfunction, which is
   consistent with those in WT mice. However, memory dysfunction was not
   improved in the mice by nomilin treatment (Fig. [213]7g, h).
   D-galactose also resulted in neuron death in CA1 and CA3 of the
   hippocampus (Fig. [214]7i–l), and inflammatory infiltration in the
   liver of PXR^-/- mice (Fig. [215]7m, n). However, nomilin did not
   improve the T-climbing time, the passing time, the dwelling time and
   the number of falls, or reduce cell death in the hippocampus and
   inflammatory cell infiltration in the liver of D-galactose-treated
   PXR^-/- mice (Fig. [216]7a–n). These data suggest that nomilin could
   not improve the impaired motor mobility, neuron death and hepatic
   inflammation in D-galactose-induced senescent mPXR deficient mice,
   which further demonstrates that the healthspan-extending effects of
   nomilin occur via mPXR activation.

Fig. 7. Effects of nomilin on healthspan in D-galactose-treated PXR knockout
mice.

   [217]Fig. 7
   [218]Open in a new tab

   a Passing time in beam balance test (n = 8/group). b T-climbing down in
   pole test (n = 8/group). c Dwelling time on rotating rod test
   (n = 8/group). d Number of falls from the rotating rod (n = 8/group).
   Hind limb stride length (e, n = 8/group) and hind limb stance width (f,
   n = 8/group) in gait analysis. g Proportion of the number of times
   entering new arm in the Y maze (n = 8/group). h Proportion of time
   spent in exploring new arm in Y maze (n = 8/group). i, j Apoptotic
   cells and the quantification of apoptotic cells in CA1 of the
   hippocampus (n = 5/group). k, l Apoptotic cells and the quantification
   of apoptotic cells in CA3 of the hippocampus (n = 5/group). m H&E
   staining of liver sections. n Inflammatory infiltration area per mm^2
   liver sections (n = 5/group). The mice were treated with D-galactose
   (125 mg/kg/day) and nomilin for 7 weeks. The data were shown as
   mean ± SEM. p-values were determined by one-way ANOVA test (a–h, j, l,
   m). ###p < 0.001 vs the control group; ***p < 0.001 vs the D-galactose
   group.

Nomilin counteracts doxorubicin-induced senescence in mice

   The chemotherapeutic drug doxorubicin may induce accelerated ageing and
   other long-term health conditions in cancer survivors^[219]72–[220]74.
   This drug has been used to induce cellular and organ senescent in
   animal models^[221]73,[222]75. Therefore, we assayed whether nomilin
   could extend the lifespan and healthspan in doxorubicin-treated mice.
   In the lifespan assay, the mice were treated with both doxorubicin and
   nomilin. Strikingly, the mean lifespan of nomilin-treated mice was
   extended by 50.57% (Fig. [223]8a, Fig. S[224]7). In the healthspan
   experiments, doxorubicin increased the T-climbing time in the pole test
   and passing time in the beam balance test (Fig. [225]8b, c), whereas
   nomilin treatment reduced the times of T-climbing and passing time
   (Fig. [226]8b, c), suggesting that nomilin could improve physical
   conditions in doxorubicin-induced senescence mice. Next, we assayed
   whether liver function was also improved by nomilin. In agreement with
   previous reports, doxorubicin increased the inflammatory cell
   infiltration in the liver (Fig. [227]8d, e), serum levels of aspartate
   aminotransferase (AST) and alanine transaminase (ALT), indicators of
   liver damage (Fig. S[228]8a, b). Interestingly, nomilin counteracted
   the inflammatory cell infiltration in the liver and the increase of AST
   and ALT (Fig. [229]8d, e, Fig. S[230]8a, b) in doxorubicin-treated
   mice. Meanwhile, the expression of senescence-related secretory
   phenotypic genes Nlrp3, Tnfα, Il-6, Il-1β, Mcp1, p16^INK4A was
   increased in the liver of doxorubicin-induced mice, while nomilin
   treatment downregulated the mRNA levels of Nlrp3 and Il-6
   (Fig. [231]8f). Similarly, the expression levels of mPXR downstream
   genes Cyp3a11, Por, Gsta1/2 and Mdr3 in the liver were increased by
   nomilin intervention, indicating that the upregulation of
   detoxification by nomilin may protect mice from doxorubicin-induced
   damage (Fig. [232]8g). Doxorubicin also induced myocardial atrophy and
   collagen deposition, the markers of fibrosis, in the heart of mice
   (Fig. [233]8h, i). Nomilin treatment attenuated cardiomyopathy by
   reducing cardiac atrophy and fibrosis areas in the heart (Fig. [234]8h,
   i). Taken together, these results indicate that nomilin may improve
   hepatic, cardiac senescence via counteracting toxicity in
   doxorubicin-induced aged mice. The data further support that nomilin
   may have a detoxification function.

Fig. 8. Effects of nomilin on lifespan and healthspan in doxorubicin-induced
senescence mice.

   [235]Fig. 8
   [236]Open in a new tab

   a Survival curve of accelerated ageing mice. The mice were treated with
   doxorubicin (5 mg/kg, three times a week) and nomilin (50 mg/kg/day).
   (n = 12 for Dox, n = 11 for Dox+NML) b T-climbing down in pole test
   (n = 12 for Ctrl, n = 8 for Dox and Dox+NML). c Passing time in beam
   balance test (n = 12 for Ctrl, n = 8 for Dox and Dox+NML). d H&E
   staining of the liver sections. e Inflammatory infiltration area per
   mm^2 liver sections in (d), (n = 5/group). f The expression of
   senescence-related secretory phenotypic genes in the liver of
   doxorubicin-treated mice (n = 5/group). g The expression of PXR
   downstream genes in the liver of doxorubicin-treated mice. β-Actin was
   used as an internal control (n = 5/group). h Cardiac fibrosis induced
   by doxorubicin administration, determined by Masson’s trichrome
   staining. i The quantitative analysis of fibrosis area in (h),
   (n = 4/group). The mice were treated with doxorubicin (5 mg/kg, three
   times a week) for 2 weeks and nomilin (50 mg/kg/day for 4 weeks). The
   data were shown as mean ± SEM. p-values were determined by one-way
   ANOVA test (b, c, e–g, i). ^###p < 0.001 vs the control group;
   ***p < 0.001 vs the doxorubicin group. Dox doxorubicin, NML nomilin.

Nomilin extends healthspan in SAMP8 mice

   The senescence-accelerated mouse is an accelerated ageing model used in
   gerontological research because of its accelerated senescence and
   various spontaneous pathobiological phenotypes^[237]76,[238]77. Here,
   we chose senescence accelerated mice prone 8 (SAMP8) mouse as an ageing
   model and senescence-accelerated mouse resistant 1 (SAMR1) mouse as a
   normal control to investigate the effects of nomilin on accelerated
   senescence. SAMP8 mice exhibited significant motor impairment, as
   evidenced by a significant increase in T-climbing, passing time and the
   number of falls from the rod, and a decrease in the time dwelling on
   the rod compared to SAMR1 mice. However, these parameters in
   nomilin-treated SAMP8 mice were reversed (Fig. [239]9b–e). Previous
   studies have reported emotional disorders and memory deficits in SAMP8
   mice^[240]76,[241]78,[242]79. In this study, we evaluated anxiety-like
   behaviour using elevated plus maze and open field tests. Results showed
   that SAMP8 mice displayed significant anxiety-like behaviour, as
   indicated by a decrease in the percentage of time spent and the number
   of entries into the open arms compared to SAMR1 mice, which was
   consistent with previous results (Fig. [243]9f, g). In contrast, the
   nomilin intervention decreased the anxiety-like behaviour of SAMP8 mice
   (Fig. [244]9f, g). Similarly, the open field test showed that SAMP8
   mice exhibited less exploration of the central area compared to SAMR1
   mice, while nomilin-treated mice showed an increase tendency to explore
   the central region (Fig. [245]9h, i, j). The novel object recognition
   experiment was carried out to assess learning and memory abilities of
   the mice. SAMR1 mice showed a stronger interest in the new object than
   in the old object, while the recognition index of SAMP8 mice decreased.
   After nomilin intervention, the mice showed an increase tendency in
   their ability to recognize new objects (Fig. [246]9k). And Y-maze test
   showed that decrease tendency in the percentage of exploration time and
   number of entries into the new arm in SAMP8 mice when compared to those
   in SAMR1 mice, whereas nomilin-treated mice increased the number of
   entries into the new arm (Fig. [247]9l, m), suggesting that spatial
   memory impairment in SAMP8 mice was improved by nomilin treatment.
   Overall, these findings suggest that nomilin may improve age-related
   disorders such as motor impairments, anxiety-like behaviour, and memory
   deficits in SAMP8 mice.

Fig. 9. Effects of nomilin on healthspan in SAMP8 mice.

   [248]Fig. 9
   [249]Open in a new tab

   a Timeline for drug treatment and behaviour test. b T-climbing down in
   pole test (n = 8 for SAMP8, n = 9 for SAMR1 and SAMP8-NML). c Passing
   time in beam balance test (n = 8 for SAMP8, n = 10 for SAMR1 and
   SAMP8-NML). d Number of falls from rotating rod (n = 7 for SAMP8,
   n = 10 for SAMR1 and SAMP8-NML). e Dwelling time on rotating rod (n = 7
   for SAMP8, n = 10 for SAMR1 and SAMP8-NML). Proportion of times
   entering the open arm (f, n = 8 for SAMR1, n = 6 for SAMP8, n = 7 for
   SAMP8-NML) and Proportion of exploration time in the open arm (g, n = 8
   for SAMR1, n = 6 for SAMP8, n = 7 for SAMP8-NML) in elevated-plus maze.
   Entries in the centre (h, n = 7 for SAMP8, n = 8 for SAMR1 and
   SAMP8-NML) and time spent in centre (i, n = 7 for SAMP8, n = 8 for
   SAMR1 and SAMP8-NML) in open field. j Trajectory in open field. k
   Recognition index of mice in novel object recognition test (n = 6 for
   SAMP8, n = 7 for SAMR1 and SAMP8-NML). l Proportion of time exploring
   the new arm in Y maze (n = 6 for SAMP8, n = 7 for SAMR1 and SAMP8-NML).
   m Proportion of times entering the new arm in the Y maze (n = 6 for
   SAMP8, n = 7 for SAMR1 and SAMP8-NML). The data were shown as
   mean ± SEM. p-values were determined by one-way ANOVA test (b–i, k–m).
   ###p < 0.001 vs the SAMR1 group; ***p < 0.001 vs the SAMP8 group.

Nomilin activates mPXR and induces a longevity gene signature in mice

   Next, we investigated whether nomilin could protect the liver against
   damage from cholestatic hepatotoxicity through detoxifying toxic bile
   acids. We supplemented bile duct-ligated (BDL) mice with nomilin and
   measured their levels of liver damage, since PXR agonist
   pregnane-16α-carbonitrile (PCN) has been reported to relieve liver
   damage in this mouse model^[250]80,[251]81. Our histological assay
   confirmed that BDL induced severe diffused vacuolization, inflammatory
   cell infiltration and hepatic parenchymal necrosis in the mouse livers.
   Notably, similar to PCN (Fig. [252]10a), nomilin significantly
   ameliorated the inflammation, fibrosis and necrosis of the liver in WT
   BDL mice, but not in mPXR knockout BDL mice (Fig. [253]10a), indicating
   that nomilin does activate mPXR in vivo in mammals. Moreover, serum
   biochemical indices showed that nomilin decreased serum ALT and AST
   levels under BDL surgery, while nomilin showed no effect on either ALT
   or AST in normal control mice (Fig. [254]10b, Sham v.s. Sham + N),
   further confirming that nomilin may protect the liver from damage due
   to BDL injury (Fig. [255]10b, BDL v.s. BDL + N), without toxicity in
   mice.

Fig. 10. Nomilin protects BDL-induced liver cholestatic injury through mPXR
and upregulates longevity related genes in mice.

   [256]Fig. 10
   [257]Open in a new tab

   a Pictures showing H&E or Mason staining of liver sections in BDL mice
   with or without PCN and nomilin supplementation. Nomilin effectively
   attenuated the BDL-induced liver damage in WT (upper), but not in
   PXR^-/- mice (bottom). b Bar-graphs showing the levels of serum ALT and
   AST in sham or BDL mice (n = 7 for BDL + NML and BDL + PCN, n = 8 for
   BDL). The data are shown as mean ± SEM. p-values were determined by
   one-way ANOVA test. ***p < 0.001 vs BDL group. c, d The RNA-seq
   hierarchical clustering heatmap showing differentially expressed genes
   (c) and genes in the detoxification process (d) from control (BDL) and
   nomilin-treated (BDL + NML) mouse liver under BDL surgery. e, f A chart
   showing the downregulated (e) and up-regulated (f) differentially
   expressed genes of KEGG pathways compared with the control group in
   both nomilin-treated and long lifespan mice. g A chart showing the top
   20 up-regulated (blue) and downregulated (pink) molecular functions and
   biological processes (GO). h A diagram depicting the effects of nomilin
   on longevity through the activation of nuclear hormone receptors and
   detoxification signalling in C. elegans and mice.

   Gene expression analysis by transcriptome sequencing (RNA-seq) also
   confirmed that reported mPXR-induced genes^[258]20,[259]82 are
   upregulated in the liver of nomilin-treated mice. Specifically, among
   the 193 genes upregulated by nomilin, at least 27 genes were mPXR
   downstream targets identified by previous studies (Fig. [260]10c, d).
   These genes are involved in drug and toxin metabolism in the mouse
   liver, which may explain why the liver damage of BDL mice was
   significantly attenuated under nomilin treatment (Fig. [261]10d).

   It has been reported that most longevity interventions induce common
   gene expression signatures in the liver of mice, which could be used to
   predict the lifespan-extension effect of new candidate
   compounds^[262]19. For example, the transcript levels of genes coding
   for ribosomal proteins, oxidative phosphorylation, drug and xenobiotic
   metabolism-cytochrome P450 enzymes, glutathione metabolism,
   tricarboxylic acid cycle, amino acid metabolism, age-related
   neurodegenerative diseases, complement and coagulation cascades, fatty
   acid oxidation, steroid and retinol metabolism and the peroxisome
   proliferators-activated receptor (PPAR) pathway were upregulated in the
   liver of long-lived mice^[263]19. We further characterised the
   molecular function and biological processes enriched in the samples
   using gene annotation. Both Gene Set Enrichment Analyses (KEGG) terms
   and Gene Ontology (GO) revealed that many top gene categories and
   metabolic pathways were highly similar to those of most longevity
   interventions (Fig. [264]10e–g), indicating that nomilin may share a
   common molecular pathway for regulating longevity with most
   lifespan-extending interventions.

Discussion

   Various toxins in the environment are risk factors for human health,
   and are linked to many age-related diseases such as Alzheimer’s disease
   and Parkinson’s disease^[265]83,[266]84. The increase of detoxification
   gene expression is a common transcriptomic signature in long-lived
   worms, flies and rodents, suggesting that xenobiotic detoxification may
   be linked with longevity-promotion. Nuclear receptors have been
   identified as regulators of healthy ageing. In mammals, PXR is a major
   transcription factor for regulating the expression of phase I–III drug
   metabolising/xenobiotic detoxifying genes. Although many studies have
   shown that PXR may show cross-talk with longevity signalling^[267]20,
   it is unknown whether targeting PXR plays a role in ageing inhibition.
   In the present study, we demonstrated that a component present in
   citrus fruits, nomilin, is a PXR agonist and extends lifespan and
   healthspan. Meanwhile, nomilin activates the expression of many phase
   I–III enzymes and efflux transporters in mice and C. elegans, which is
   correlated with IIS signalling (Fig. [268]10h). These data are in line
   with the results of most longevity interventions, such as caloric
   restriction, which upregulates most phase-I/II enzymes and phase-III
   efflux transporters in the livers of male mice^[269]85, further
   supporting that detoxification may be a mechanism for
   longevity-promotion. Our results show that PXR may have physiological
   functions in ageing in addition to drug metabolism.

   Interestingly, citrus and grapefruit juices may contain a large amount
   of nomilin or its precursors, which could be hydrolysed in the liver
   and by the intestinal flora^[270]26; however, most of them would be
   removed in the “debittering” processing in the orange juice industry
   because of their bitter taste^[271]86. We showed that
   nomilin-treatments did not change the body weight and food consumption
   of the mice (Supplementary Fig. [272]S9a-h), indicating that nomilin
   may be a safe component. Therefore, our results suggest that a revisit
   of the “debittering” process may be needed given the potential
   beneficial function of nomilin. Notably, as a xenobiotic-sensor to
   protect the body from endo/xenobiotics by detoxification of toxins, PXR
   was originally characterised as a regulator of drug metabolism^[273]87.
   As a PXR agonist, nomilin may accelerate drug metabolism and attenuate
   the efficiency of therapy. Whether the consumption of
   nomilin-containing citrus fruits and juices change drug metabolism
   needs to be investigated.

   In conclusion, we found that nomilin extends the lifespan and
   healthspan in C. elegans and mice, and regulates the gene expression of
   detoxification enzymes through the activation of nuclear hormone
   receptors. The detoxification function of nomilin is probably linked to
   IIS longevity signalling. Our data suggest that targeting PXR maybe a
   feasible strategy for longevity and health promotion.

Methods

Chemicals

   Nomilin and limonin (purity > 99.8%) were obtained from Pusi Biotech
   (Chengdu, China). Nomilin identity was confirmed by a mass spectrometry
   assay (see Supplementary Methods). Chloroquine (Yuanye, Shanghai,
   China), colchicine (Yuanye, Shanghai, China), paraquat (Thermo Fisher
   Scientific, Waltham, USA), methylmercury chloride (MeHgCl, Dr.
   Ehrenstorfer GmbH, Augsburg, German), PCN (GLPBIO, Montclair, USA) and
   doxorubicin hydrochloride (Yuanye, Shanghai, China) were commercially
   available.

C. elegans strains and maintenance

   The following C. elegans strains were used in this study: N2: Wild-type
   Bristol isolate, CB1370: daf-2 (e1370), CF1038: daf-16 (mu86), MAH97:
   muIs109 [daf-16p::GFP::DAF-16 cDNA + odr-1p::RFP], VC199: sir-2.1
   (ok434), VC222: raga-1 (ok386), RW12220: pha-4
   (st12220[pha-4::TY1::EGFP::3xFLAG]), DR2281: daf-9 (m540), AA86: daf-12
   (rh61rh411). These were obtained from the CGC (Caenorhabditis Genetics
   Center), which is funded by the NIH National Center for Research
   Resources (NCRR). nhr-8 (tm1800) was obtained from National BioResource
   Project (Tokyo, Japan). The worms were cultured on nematode growth
   medium (NGM) agar plates seeded with live bacteria at 20 °C (E. coli,
   strain OP50) as food source, according to standard
   protocols^[274]88,[275]89.

Lifespan experiments

   Synchronised L1 worms were cultivated on standard NGM plates at 20 °C
   for about 3 days. Then, L4 adults were transferred to plates (30 worms
   per plate), fed with 0, 25, 50 100 μM nomilin and limonin or DMSO
   (0.1%) solvent control mixed with OP50, respectively. Worms were judged
   as dead when they did not respond to repeated prodding with a pick and
   had no pharynx pumping. Dead worms were counted daily. Worms that
   crawled off plates or bagging were excluded.

   Heat-killed OP50 were prepared with a 20× concentrate for 1 h at 75 °C,
   as previously described^[276]90.

   For RNA interference (RNAi) lifespan experiments, synchronised L1 worms
   were fed with E. coli (HT115) containing an empty control vector
   (L4440) until L4, then transferred to plates where they were fed with
   daf-12, nhr-8, cyp35a3, gst-4, pgp-3 or pgp-14 RNAi constructs with or
   without nomilin (50 μM), using L4440 and DMSO (0.1%) as controls on NGM
   containing isopropyl-beta-D-thiogalactopyranoside (IPTG, 1 mg/ml) and
   ampicillin (50 μg/ml). All RNAi constructs were obtained from the
   Ahringer RNAi library and grown at 37 °C overnight in LB containing
   ampicillin (50 μg/ml) after sequence verification.

DAF-16:GFP translocation experiments

   muIs109 [Pdaf-16::gfp::daf-16; Podr-1::rfp]^[277]91,
   Pdaf-16::gfp::daf-16; Podr-1::rfp; nhr-8 (1800), Pdaf-16::gfp::daf-16;
   and Podr-1::rfp; daf-12 (rh61rh411) worms were treated with nomilin
   from L1 to L4, anaesthetised with 100 mM NaN[3] and mounted on 2%
   agarose pads. The GFP fluorescent signals of DAF-16 localisation were
   examined in 10 animals per condition and captured by a confocal
   microscope (SP-8 Leica, Germany). Images were acquired with a digital
   camera. The number of GFP-positive nuclei of each worm was calculated.

Cross strategy

   To obtain a homozygous strain of Pdaf-16::gfp::daf-16; Podr-1::rfp;
   nhr-8 (1800) and Pdaf-16::gfp::daf-16; Podr-1::rfp; daf-12 (rh61rh411),
   muIs109 males were mated with daf-12 (rh61rh411) or nhr-8 (tm1800)
   hermaphrodites (P0). F1 worms carrying the RFP fluorescence were
   considered as cross-progeny and singled into 10 35 mm NGM plates. F2
   with all progeny carrying the RFP fluorescence were considered as
   muIs109 homozygous and singled for nhr-8 (tm1800) or daf-12 (rh61rh411)
   genotyping. nhr-8(-) homozygous worms were identified by PCR with the
   primers nhr-8 F (catttatacttctaaaccaacaattgt), nhr-8 D
   (ccggataatttcattgaaacttact), and nhr-8 R (ggtacatatcacaggttatcgaga).
   daf-12(-) homozygous worms were identified by sequencing using daf-12 F
   (attgtatttcagggtatcatggatc) and daf-12 R (ggtgataaatgtggctgttgatta).

Heat stress and oxidative stress experiments

   In stress resistance assays, the number of surviving worms was
   monitored following exposure to the indicated stressor. For heat shock
   experiments, L4 phase N2 worms were treated with nomilin for 10 days,
   and then the worms were placed at 35 °C for 12 h. Every 2 h the worms
   were observed for survival. The experiments were repeated three times.
   For oxidative stress resistance experiments, the L4 worms were placed
   on 0.05% H[2]O[2] NGM agar plates for 12 h at 20 °C. Every 2 h the
   worms were observed for survival.

Dauer induction assay

   The dauer induction by high-density growth was performed according a
   previously reported method^[278]92, except that 50 μl E. coli OP50 with
   10% DMSO or 50 μM nomilin was seeded on the egg white plates (about
   1.5–6 × 10^4 eggs/plate). The dauer induction in the daf-2(e1370)
   mutation was carried out as a standard protocol in WormBook^[279]93.
   Briefly, the worms were maintained at 15.0^oC on standard NGM plates,
   and allowed gravid adult hermaphrodites to lay eggs for several hours
   on 3.5 mm NGM plates that seeding with 100 μl E. coli OP50 (with 10%
   DMSO, or 50 μM NML) at 20.0 °C, then removed adults when about 100–200
   eggs were laid on the plates, and shifted the plates to a incubator for
   the dauer formation at 22.0/23.5 °C for 68–80 h. Then, 1 ml of 1%
   sodium dodecyl sulphate was added to the plates and incubated for
   30 min to count dauers (survivors).

Cell cultures and reporter assays

   HEK 293 T cells (CRL-11268, ATCC) were maintained in Dulbecco’s
   modified eagle medium with 10% foetal bovine serum (FBS, Hyclone,
   Logan, UT, USA). For reporter assays, the expression plasmid
   pSG5-hPXR/CYP3A4-Luc, pcDNA3.1-hCAR/CYP3A4-Luc, pCMXGal-hPPARα, β, γ
   LBD, LXRα/β LBD, and the Gal4 reporter vector MH1004-TK-Luc were
   cotransfected with pREP7. For transfection, each well contained 100 ng
   of total plasmids and 0.2 µl of FuGENEHD transfection reagent (Roche,
   Germany) for 24 h. Then, nomilin and hPXR, hCAR, PPARα, β, γ, LXRα, β,
   and FXR agonist (Rifampicin, CICTO, fenofibric acid, GW4064,
   pioglitazone T0901317 and GW4064, respectively, Sigma Aldrich, St.
   Louis, MO, USA) were added to fresh media and incubated for another
   24 h. The luciferase activity was measured using the Dual Luciferase
   Reporter Assay System (Promega, USA), and the transfection efficiencies
   were normalised according to Renilla luciferase activity.

TR-FRET assay

   Time-resolved fluorescence resonance energy transfer (TR-FRET) was
   performed using a LanthaScreen™ TR-FRET PXR (SXR) competitive binding
   assay kit according to the manufacturer’s protocol (PV4839, Invitrogen,
   Darmstadt, Germany). Briefly, labelled hPXR-ligand Fluormone™ PXR (SXR)
   Green (40 nM), hPXR-LBD-GST (5 nM), goat terbium-anti-GST antibody
   (5 nM) and nomilin, deacetylnomilin, limonin and/or T0901317 (10 μM)
   were incubated in assay buffer at room temperature for 2 h, and the 520
   and 495 nm fluorescence signal was assayed using a PerkinElmer EnVision
   Multilabel Reader. The 520/495 value was calculated by subtracting the
   background TR-FRET ratio. The dissociation constant (K[d]) was fitted
   into a one-site total binding saturation equation in GraphPad Prism
   software. IC[50] values were determined using log (inhibitor) vs.
   response - Variable slope model fit by GraphPad Prism software
   according to the previous method^[280]94. All experiments were
   performed in triplicate.

Protein expression and purification

   A cDNA encoding a PXR^LBD-NCOA1^676–700 fusion protein comprised of
   residues 130–432 of hPXR (Uniprot ID O75469-1) with its C-terminal
   linked to residues 676–700 of nuclear receptor coactivator 1 (Uniprot
   ID Q15788-1) and spaced with -Ser-Ser-Ser-Gly-Gly-Thr- was synthesised
   and cloned into a plasmid modified from pFastBac Dual (Invitrogen),
   with a C-terminal TEV protease recognition site and 6 × polyhistidine
   affinity tag. The recombinant baculovirus encoding the
   hPXR^LBD-NCOA1^676–700 fusion protein was generated and used to infect
   the Spodoptera frugiperda cell line Sf9 for overexpression. The Sf9
   cells were harvested 48–72 h after infection and collected by
   centrifugation (1500 g, 15 min, 20 °C). To lyse the cells, the pellets
   were re-suspended with ice-cooled buffer containing 150 mM NaCl, 20 mM
   HEPES, pH 7.5, 0.1 mg/ml DNase I, 2 mM MgCl[2], 1 mM TCEP and protease
   inhibitor cocktail, and subjected to sonication lysis. The cell lysate
   was clarified by centrifugation at 46,000 g for 45 min and the
   supernatant was subjected to immobilised metal affinity chromatography
   (IMAC) with Talon Metal Affinity Resin (Clontech). After the removal of
   the tag with tobacco etch virus protease (recombinant protein with
   His-tag), the imidazole in eluate was removed by dialysis against
   buffer containing 150 mM NaCl, 20 mM HEPES, pH 7.5, 10% (v/v) glycerol
   and 1 mM TCEP, and then subjected to IMAC to remove tobacco etch virus
   protease. The flow-through fraction containing the
   hPXR^LBD-NCOA1^676–700 fusion protein was further isolated using
   size-exclusion chromatography with a Superdex 200 Increase 10/300 GL
   column (GE Health Sciences), equilibrated in 150 mM NaCl, 20 mM HEPES
   and 1 mM TCEP. The peak fractions containing the hPXR^LBD-NCOA1^676–700
   fusion protein was pooled and supplemented with nomilin to a final
   concentration of 1 mM, followed by concentrating to ~7 mg/ml for
   crystallization trials.

Crystallization and X-ray data collection/processing

   The hPXR^LBD-NCOA1^676–700 was mixed with crystallization solution
   comprising 10% (v/v) 2-propanol, 100 mM imidazole/hydrochloric acid pH
   8.0 at an initial ratio of 1:1. Crystals were grown at 4 °C using the
   sitting drop vapour diffusion method and were cryoprotected by dipping
   in crystallization solution supplemented with 20% glycerol and then
   flash-freezing in liquid nitrogen. Diffraction data were collected at
   the Shanghai Synchrotron Radiation Facility (SSRF), beamlines BL18U1
   and BL19U1. The data were collected and processed with HKL2000^[281]95.
   The crystallographic parameters and data collection statistics are
   given in Supplementary Table [282]S3. The hPXR structure model with PDB
   ID 5X0R was used as the search model^[283]56, and the molecular
   replacement and initial model building were performed in
   Phenix^[284]96. Iterative cycles of refinement were carried out using
   PHENIX and Coot^[285]97. All structure graphs in this paper were
   produced using PyMOL (The PyMOL Molecular Graphics System, Version 1.9
   Schrödinger, LLC.) and LigPlot^+ (LigPlot^+ version v1.4.5)^[286]98.

Generation of hPXR point mutations

   hPXR mutations M243Q, M243A, S247R, S247A, W299R, M425Q and M425A were
   created using the pSG5-hPXR expression plasmid as a template, and
   PrimeSTAR DNA polymerase was used to amplify the DNA. The primers for
   PCR are:

   M243A (ATG → GCT):

   TGCCCCACGCTGCTGACATGTCAACCTACAT; ATGTCAGCAGCGTGGGGCAGCAGGGAGAAGAT.

   M243Q (ATG → CAA):

   TGCCCCACCAAGCTGACATGTCAACCTACAT; ATGTCAGCTTGGTGGGGCAGCAGGGAGAAGA.

   S247A (TCA → GCT)

   CTGACATGGCTACCTACATGTTCAAAGGCAT;

   ATGTAGGTAGCCATGTCAGCCATGTGGGGCA.

   S247R (TCA → AGA):

   CTGACATGAGAACCTACATGTTCAAAGGCAT;

   ATGTAGGTTCTCATGTCAGCCATGTGGGGCA.

   W299R (TGG → AGA):

   CTGGAACCAGAGAGTGTGGCCGGCTGTCCTA; CCACACTCTCTGGTTCCAGTCTCCGCGTTGA.

   M425A (ATG → GCT):

   TACGCCCCTCGCTCAGGAGTTGTTCGGCATCACA; AACTCCTGAGCGAGGGGCGTAGCAAAGGGGTGTA.

   M425Q (ATG → CAA):

   TACGCCCCTCCAACAGGAGTTGTTCGGCATCACA; AACAACTCCTGTTGGAGGGGCGTAGCAAAGGGGT.

   All mutations were confirmed by sequence analysis. The reporter gene
   assay was carried out as described above.

RNA extraction and real time RT-PCR

   For the gene expression assay, 50 μM nomilin-treated N2, daf-12(-) and
   nhr-8(-) or untreated L4 worms were subjected to quantitative real-time
   PCR. The total RNA was extracted from about 1000 worms using the TRIzol
   reagent (Sangon Biotech, Shanghai, China) according to the
   manufacturer’s instructions. The residual DNA was removed using gDNA
   wiper mix, and 1 μg of total RNA was reverse-transcribed to
   complementary DNA using HiScript II qRT SuperMix II (Sangon Biotech,
   Shanghai, China). Quantitative real time PCR was performed using the
   ABI StepOnePlus Real Time PCR system (Applied Biosystems, Foster City,
   CA, USA) using SYBR Green PCR Master Mix (Sangon Biotech, Shanghai,
   China). The results were analysed with β-actin as the internal control.
   Sequences for primers are listed in Supplementary Table [287]5.

   For mouse experiments, 10 mg of tissues were used to extract total RNA
   and quantitative PCR was performed as described above. The primer
   sequences are listed in Supplementary Table [288]6.

Western blot analysis

   For total protein extraction, the liver tissues were homogenised in
   sample buffer and boiled for 5 min. The samples were separated using
   10% SDS–PAGE, transferred to a PVDF membrane, and blocked with 5%
   bovine serum albumin at room temperature for 2 h. Then, the blots were
   incubated with polyantibodies against Gsta, Cyp3a11, Cyp51a1 and GAPDH
   (ProteinTech, Rosemont, USA) at 4 °C for 12 h. The membrane was washed
   and incubated with secondary antibody for 2 h at room temperature. The
   signals were detected and analysed using an Odyssey Two-Colour Infrared
   Imaging System (LI-COR Biosci- ences, Lincoln, NE, USA). GAPDH was
   assayed as a loading control.

Generation of transgenic worms

   Human PXR and the hPXR^S247R cDNA sequence were cloned and driven by
   Prpl-28 promoters and injected into the wild-type N2, nrh-8 (-) and
   daf-12 (-) mutant lines with the Pmyo-2::RFP co-injection marker under
   the IM 300 Microinjector (NARISHIGE, Japan). pSM delta vectors were
   injected into the three lines as the vehicle controls.

   The fluorescence-marked transgenic strains pSM; Pmyo-2::rfp, pSM;
   Pmyo-2::rfp; nhr-8(tm1800), Pmyo-2::rfp; daf-12 (rh61rh411),
   Prpl-28:hPXR; Pmyo::rfp, Prpl-28:hPXR; Pmyo::rfp; nhr-8 (tm1800),
   Prpl-28:hPXR; Pmyo::rfp; daf-12 (rh61rh411), Prpl-28:hPXR^S247R;
   Pmyo::rfp, Prpl-28: hPXR^S247R; Pmyo::rfp; nhr-8 (tm1800), Prpl-28:
   hPXR^S247R; and Pmyo::rfp; daf-12 (rh61rh411) were generated for
   life-span experiments with DMSO and nomilin treatment.

Detoxification assay

   Detoxification assays were performed in 12-well polystyrene tissue
   culture plates. Concentrated stock solutions of chloroquine,
   colchicine, paraquat and MeHgCl at 50 mM were prepared in complete S
   medium, and filtered through 0.22 µm nitrocellulose filters. The
   working solutions were diluted to the indicated concentrations using S
   buffer (chloroquine, colchicine and MeHgCl) or M9 buffer (paraquat).
   OP50 was added to the working solutions. Serial dilutions were made in
   bulk and aliquoted into individual wells (1 ml per well). The
   synchronous N2, nhr-8 (-) and daf-12 (-) worms were treated with
   nomilin at concentrations of 0, 6, 12, 25, 50 and 100 μM from the L1 to
   L4 stage. Then, the worms were transferred to S buffer or M9 buffer
   containing the same concentration of nomilin plus toxins (4 mM
   chloroquine and colchicine, 2 μM MeHgCl in S buffer, and 100 mM
   paraquat in M9 buffer) in 96-well plates, each containing five worms.
   The worms were cultured and monitored under a stereo microscope at the
   indicated time points.

   For the detoxification lifespan assay, adult (Day 10) N2, daf-2, daf-16
   and daf-2 with nhr-8 or daf-12 RNAi (from L4) were challenged with
   paraquat (100, 200 mM) in M9 or MeHgCl (1, 2 μM) in S buffer for 24 h.
   The death rate was recorded at 1, 2, 4, 6, 8 and 24 h. Meanwhile,
   nomilin (12, 50 μM)-treated N2 (from L4 to Day 10) were also picked
   into paraquat or MeHgCl buffer containing nomilin (12, 50 μM).

D-galactose induced senescence in mice

   Sixty 8-week-old female and male C57BL/6 mice were obtained from
   Shanghai Laboratory Animal Center of Chinese Academy of Science
   (Shanghai, China). All mice were housed under a 12 h light/dark cycle
   in a room with controlled temperature (22 ± 1 °C). All procedures were
   approved by the Experimental Animal Ethical Committee at Shanghai
   University of Traditional Chinese Medicine (PZSHUTCM191122007). After
   1 week of adaptive feeding, the mice were randomly divided into three
   groups (20 mice per group, half male and half female): control group
   (0.9% saline + normal diet), model group (125 mg/kg/day D-galactose +
   normal diet) and nomilin intervention group (125 mg/kg/day D-galactose
   + 50 mg/kg nomilin mixed into diet). Mice were injected subcutaneously
   with 0.9% saline or 125 mg/kg D-galactose daily for 7 weeks. Behaviour
   assessments were performed after 6 weeks of treatment, and the mice
   were sacrificed after anaesthetised with 20% urethane (Sinopharm
   Chemical Reagent Co., Shanghai, China) for further study.

   PXR null knockout mice (C57BL/6N-Nr1i2^em1Cya) were purchased from
   Cyagen Biosciences (Suzhou, Jiangsu, China). The detailed information
   on this mouse can be seen at
   [289]https://www.cyagen.com/cn/zh-cn/sperm-bank-live/18171. Twenty-four
   8–12-week-old female and male PXR^-/- mice were divided into the
   control group, D-galactose group and D-galactose+nomilin group (n = 8
   in each group). D-galactose induced senescence and nomilin treatment
   was performed as indicated above.

Doxorubicin induced accelerated ageing in mice

   To induce senescence in the model, male C57BL/6 mice (6–8 weeks) were
   intraperitoneally injected with doxorubicin at 5 mg/kg three times
   weekly for 2 weeks. The control group mice were administered an
   equivalent volume of saline. For lifespan experiments, the mice were
   containued to receive doxorubicin (5 mg/kg three times per week) and
   nomilin (50 mg/kg/day) until all mice died. For the heathspan
   experiments, the mice were given doxorubicin (5 mg/kg three times per
   week) for 2 weeks and nomilin (50 mg/kg/day) by oral gavage for
   4 weeks. The mice were anaesthetised with 20% urethane (Sinopharm
   Chemical Reagent Co., Shanghai, China), and blood and tissues were
   harvested and analysed.

SAMP8 mice

   Nineteen male SAMP8 mice and ten male SAMR1 mice were obtained from
   Beijing HFK Bioscience, China. SAMP8 mice were divided into two groups
   randomly according to their body weight. NML was mixed into food (at
   dose of 40 mg/100 g diet) to feed mice from 19-week-age old. And then
   these mice were taken to test their behaviour at the indicated timeline
   in Fig. [290]9a. Body weight and food intake were recorded every
   2 days. The animal protocols were approved by the Experimental Animal
   Ethical Committee at Shanghai University of Traditional Chinese
   Medicine (PZSHUTCM2212020004).

Behaviour assessments of mice

Pole test

   The pole test uses a device consisting of a wooden stick (diameter
   1 cm, height 52 cm) and a wooden ball (diameter 2.5 cm) at the top of
   the stick. The device was wrapped with medical tape to prevent mice
   from slipping. Mice were placed head down and hind paws were placed on
   the ball in order to record the time of climbing down the stick. In the
   6th week of treatment, each mouse was allowed to perform two trials and
   the average value was used for statistical purposes. In this process,
   T-climbing down would be recorded as 20 s when the mouse took >20 s to
   climb down to the cage from the pole.

Balance beam test

   To measure coordination and balance, a cylindrical wooden stick (1 cm
   in diameter and 50 cm in length) wrapped with medical tape was placed
   above the cage, and the cage was covered with a layer of bedding to
   prevent mice from being injured. The mice were placed on one side of
   the stick, and the time for the mice to reach the other end of the
   wooden stick was recorded (the front paws touching the edge of the cage
   was considered as successful arrival). Each mouse was given two trials
   and the interval time between trails was 30 min. When the mouse walked
   from one side of the stick to the other for >30 s, the passing time was
   recorded as 30 s.

Rotarod test

   The mice were placed on a rotarod apparatus (Shanghai Bio-will Co.,
   Ltd.) to examine motor dysfunction associated with neurological
   impairment. Before testing, the mice were put on the rod for 3 min at
   5 rpm to acclimate to the device. Two hours later, the rod speed was
   accelerated from 5 to 40 rpm and the mice was put on for 5 min. The
   time of the mice fell from the rod and the number of times it fell from
   the rod within 5 min were recorded. Two trials were performed per mouse
   with a 30 min interval between trials.

Gait analysis

   Gait analysis is used to assess neurological and neuromuscular
   functions by detecting walking parameters in freely moving
   mice^[291]99. Before the test, the mice were place in the testing room
   without interference and light for 1 h to adapt. Each mouse was put on
   the left initial terminal and trained to voluntarily walk along the
   glass track to the other side. Based on optical technology from the
   Catwalk Automated Gait Analysis System (Noldus Information Technology,
   Wageningen, Netherlands), three correct runs were recorded for each
   animal and the associated gait parameters were analysed using CatWalk
   XT version 10.6.

8-arm maze

   The 8-armed maze was used to evaluate the learning and memory ability
   of mice. The basic principle of this experiment is that controlling
   mouse explores the arms of maze driven by food. After a period of
   training, animals can remember the spatial position of food in the
   maze. The experiment used an 8-arm radial maze, each arm is 50 cm long,
   7 cm wide, and 11 cm high. In the center, there is a circular platform
   with a diameter of 25 cm, leading to eight arms. Mice fasted for 12 h
   before training, and each mouse was given 2–3 g food per day throughout
   the experiment to stay hungry. On the first day, about 10 mg of bread
   was placed at the end of each arm, and three mice were placed on the
   central plate. The door to each arm was opened and the mice were
   allowed to explore the maze for 10 min. The training was repeated the
   next day. Again, bread was placed at the end of each arm, but on the
   third and 4th day of training, only one mouse was placed in the maze
   for 5 min each time. After a day off, only two randomly selected arms
   had bread at the ends. The end of the test signal is that two pieces of
   bread have been eaten or the exploration time has reached 5 min. This
   training lasted for 5 days. On the 11th day, performance rate (P) and
   exploration time were recorded to reflect the short-term memory ability
   of mice^[292]100. The formula for calculating the performance rate is
   listed below, where n refers to the total number of times entering the
   arm, reference memory error (RME) refers to the number of visits to an
   arm without food, working memory error (WME) refers to the number of
   visits to a previously visited arm. After 3 days of rest, the test was
   repeated and the above indicators were recorded to show the long-term
   memory ability of the mice.
   [MATH: <mi mathvariant="normal">P</mi><mo>=</mo><mfrac><mrow><mi
   mathvariant="normal">n</mi><mo>−</mo><mrow><mo>(</mo><mrow><mi
   mathvariant="normal">RME</mi><mo>+</mo><mi
   mathvariant="normal">WME</mi></mrow><mo>)</mo></mrow></mrow><mrow><mi
   mathvariant="normal">n</mi></mrow></mfrac><mo>×</mo><mn>100</mn><mi>%</
   mi> :MATH]
   1

Elevated plus-maze test

   The elevated plus-maze consists of crisscrossing open and closed arms,
   of which the closed arm is surrounded by 15 cm high wall. The entire
   experimental setup is 1 m above the ground. The experiment examined the
   anxious behaviour of rodents due to their dislike of open field and
   heights. The mouse was placed in the central area of the maze, with its
   back to the experimenter and facing the open arm. And it was allowed to
   move freely through the maze for 5 min. The proportion of times the
   mouse entered the open arm and the proportion of time spent in the open
   arm were calculated.

Open field test

   The open field test was used to evaluate the inquiry behaviour and the
   tension of mice in an unfamiliar environment. Mice were placed on a
   square open field of 50 cm × 50 cm × 50 cm, and the time and number of
   mouse exploration in the central region were recorded to assess the
   anxiety behaviour.

Y-maze

   The Y-maze is a classical behavioural test used to measure spatial
   memory in mice^[293]101. The device consists of three arms of equal
   length (30 cm × 5 cm × 15 cm). The learning and memory ability of mice
   is displayed by detecting the number of times exploring the new arm and
   the total number of explorations. The gate of new arm was closed during
   the training period, and each mouse was placed facing the wall in the
   starting arm. Then it was allowed to explore the maze freely for 10 min
   and learn to remember the spatial position of the remaining two arms.
   After an hour break, the gate of the new arm was opened, and the mouse
   was placed in the starting arm facing the wall, and it was freely
   explored for 5 min. The entire experiment was recorded with a camera.
   And frequency enter to each arm and explore time in each arm was
   counted by the EthoVision XT analysis system. The memory ability is
   expressed as the proportion of frequency of entering to new arm and
   proportion of time exploration in new arm.

Novel objective recognition

   Novel object recognition was used to detect the cognitive abilities of
   mice based on their nature of being intensely curious about new
   targets. The device includes a cube field with a side length of 50 cm
   and three objects of different shapes and colours. During the
   acclimatization period, two identical objects A were placed in the
   field, each about 10 cm away from the wall. Then mouse was placed in
   the field with back to the object and at the same distance from both
   objects. EthoVision XT analysis system was used to record the
   exploration time of mice on each object (touching the object with the
   mouth or nose and approaching the object within a range of about 2–3 cm
   are considered to be exploring the object). After 10 min exploration,
   mouse was put back into its cage. And after 1 h of rest, the mouse was
   placed back in the field. At this point, one of the two objects in the
   field was replaced with a new object B. And, the recognitive index
   refers to the proportion of times exploring the new object.

Bile duct ligation (BDL) experiments

   All animal experiments were approved by the Ethical Committee of
   Shanghai University of Traditional Chinese Medicine (Approval number:
   PZSHUTCM190609001). Male C57BL/6 mice and PXR^-/- mice (9-week-old,
   body weight >25 g) underwent BDL surgery under 1% pentobarbital
   anaesthesia according to the previous description, using a sham group
   as a control^[294]102. BDL-surgery mice were divided into two groups
   (n = 9–10) 1 day after surgery and treated with nomilin (100 mg/kg),
   PCN (100 mg/kg) or vehicle (0.5% CMC-Na) orally for 2 weeks. At the end
   of the experiment, mice were anaesthetised with 20% urethane (Sinopharm
   Chemical Reagent Co., Shanghai, China) after overnight fasting. Heart
   blood samples were taken and serum was separated for ALT and AST
   analysis using an automatic biochemical analyser (Hitachi 7020, Japan).

RNA sequencing analysis

   The liver tissues of the mice were collected for RNA sequencing
   analysis. Total RNA from BDL and BDL + NML groups was isolated from
   mouse liver tissue using the TRIzol reagent according to the
   manufacturer’s protocol. RNA purity and quantification were evaluated
   using the NanoDrop 2000 spectrophotometer (Thermo Scientific, USA). RNA
   integrity was assessed using the Agilent 2100 Bioanalyzer (Agilent
   Technologies, Santa Clara, CA, USA). Then, the libraries were
   constructed using the TruSeq Stranded mRNA LT Sample Prep Kit
   (Illumina, San Diego, CA, USA) according to the manufacturer’s
   instructions. The transcriptome sequencing and analysis were conducted
   by OE Biotech Co., Ltd. (Shanghai, China).

   The libraries were sequenced on an Illumina HiSeq X Ten platform and
   150 bp paired-end reads were generated. Raw data in fastq format were
   firstly processed using Trimmomatic and the low-quality reads were
   removed to obtain the clean reads. Then, clean reads for each sample
   were retained for subsequent analyses. The clean reads were mapped to
   the Mus musculus genome (GRCm38.p6) using HISAT2. The fragments per
   kilobase million of each gene was calculated using Cufflinks, and the
   read counts of each gene were obtained using HTSeq-count. Differential
   expression analysis was performed using the DESeq (2012) R package.
   P-value < 0.05 and fold change > 1.5 were set as the thresholds for
   significantly differential expression. Hierarchical cluster analysis of
   differentially expressed genes (DEGs) was performed to demonstrate the
   expression pattern of genes in different groups and samples. GO
   enrichment and KEGG pathway enrichment analysis of DEGs were performed
   using R, based on the hypergeometric distribution.

Statistical analysis

   Results of lifespan experiments were analysed using Kaplan-Meier
   survival analysis and compared among groups, scoring for significance
   using the log-rank test. The results of survival values following
   stress conditions were analysed using Student’s t-test. For
   Supplementary Table [295]1, [296]2 and [297]4, the average of the mean
   lifespan, the minimum and the maximum lifespan of a set of independent
   experiments were calculated and expressed as mean ± SEM. SPSS was used
   for statistical analysis. A p-value of 0.05 or less was considered to
   be statistically significant.

Reporting summary

   Further information on research design is available in the [298]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [299]Supplementary Information^ (10.9MB, pdf)
   [300]Peer Review File^ (774.3KB, pdf)
   [301]Reporting Summary^ (5.6MB, pdf)

Acknowledgements