Abstract
Flavin containing monooxygenases (FMOs) are promiscuous enzymes known
for metabolizing a wide range of exogenous compounds. In C. elegans,
fmo-2 expression increases lifespan and healthspan downstream of
multiple longevity-promoting pathways through an unknown mechanism.
Here, we report that, beyond its classification as a xenobiotic enzyme,
fmo-2 expression leads to rewiring of endogenous metabolism principally
through changes in one carbon metabolism (OCM). These changes are
likely relevant, as we find that genetically modifying OCM enzyme
expression leads to alterations in longevity that interact with fmo-2
expression. Using computer modeling, we identify decreased methylation
as the major OCM flux modified by FMO-2 that is sufficient to
recapitulate its longevity benefits. We further find that tryptophan is
decreased in multiple mammalian FMO overexpression models and is a
validated substrate for FMO-2. Our resulting model connects a single
enzyme to two previously unconnected key metabolic pathways and
provides a framework for the metabolic interconnectivity of
longevity-promoting pathways such as dietary restriction. FMOs are
well-conserved enzymes that are also induced by lifespan-extending
interventions in mice, supporting a conserved and important role in
promoting health and longevity through metabolic remodeling.
Subject terms: Ageing, Metabolomics, Metabolomics
__________________________________________________________________
Flavin containing monooxygenase 2 (FMO-2) is known to increase lifespan
under dietary restriction through incompletely understood mechanisms.
Here the authors report that FMO-2 modifies tryptophan and methionine
metabolic pathways to enhance stress resistance and slow aging in C.
elegans.
Introduction
Flavin-containing monooxygenases (FMOs) are a family of enzymes that
oxygenate substrates with nucleophilic centers, such as nitrogen and
sulfur^[52]1. They were first discovered 50 years ago and have been
studied extensively under the context of xenobiotic and drug
metabolism^[53]1. FMOs bind to an FAD prosthetic group and interact
with an NADPH cofactor to oxygenate substrates^[54]2. The FMO protein
family is highly conserved both genetically and structurally from
bacteria to humans^[55]2,[56]3. Considering the conserved nature of
FMOs, it is plausible that they share an endogenous role in addition to
detoxifying xenobiotics.
Through a screen of genes downstream the hypoxia-inducible factor-1
(HIF-1), whose stabilization leads to lifespan extension in C. elegans,
flavin-containing monooxygenase-2 (fmo-2) was identified as necessary
for the longevity and health benefits of both hypoxia and dietary
restriction (DR)^[57]4. The fmo-2 gene is also sufficient to confer
these benefits on its own when overexpressed^[58]4. Recently, studies
also suggest potential endogenous role(s) for mammalian FMOs, where
changes in expression of multiple FMO proteins affect systemic
metabolism^[59]5–[60]10. Initial correlative reports also link FMOs to
the aging process, showing that Fmo genes are frequently induced in
long-lived mouse models, such as DR mice^[61]5,[62]6. However, the
mechanism(s) for how Fmos modulate endogenous metabolism and/or aging
in vivo is unclear, as is their potential to benefit health and
longevity in multiple species.
While frequently implicated in cancer cells, recent studies identify
one carbon metabolism (OCM) as a common downstream target of multiple
longevity pathways^[63]11–[64]14. OCM is an important intermediate
metabolic pathway and refers to a two-cycle metabolic network including
the folate cycle and the methionine cycle^[65]15. OCM takes nutrient
inputs, including glucose and vitamin B12, and utilizes them to
synthesize intermediates for metabolic processes involved in growth and
survival, including nucleotide metabolism, the transsulfuration and
transmethylation pathways, and lipid metabolism^[66]12,[67]13,[68]16.
In particular, suppressing expression of the methionine cycle gene
sams-1 by RNA-mediated interference (RNAi) extends the wild-type worm
lifespan, but fails to further extend the lifespan of the genetic DR
model eat-2 mutants^[69]17.
Kynurenine synthesis from tryptophan and subsequent metabolism is
another important metabolic pathway that can play a role in many
processes, including longevity regulation. Knocking out tryptophan
2,3-dioxygenase (TDO), which catalyzes the first and rate-limiting step
of this pathway, leads to lifespan extension in worms and
flies^[70]18,[71]19. Similarly, suppressing the kynurenine pathway by
knocking down kynureninase (kynu-1) in worms also increases
lifespan^[72]20. The kynurenine pathway competes for tryptophan with
the serotonergic biosynthesis pathway and produces nicotinamide adenine
dinucleotide (NAD) and other metabolites, including kynurenic acid and
picolinic acid^[73]21.
Given that (1) induction of Fmos correlates with increased longevity
across species, (2) nematode fmo-2 is necessary and sufficient to
improve health and longevity downstream of metabolic perturbations,
such as DR, and (3) loss of Fmo expression can modify aspects of
metabolism, we hypothesized that Fmos affect aging by modifying one or
more distinct metabolic processes.
In this work, we sought to determine the metabolic changes that occur
when the expression of nematode fmo-2 is perturbed to identify its
mechanism of longevity regulation. Our resulting data support a model
where fmo-2 oxygenates tryptophan, leading to alteration of OCM
components that confer longevity and healthspan benefits by reducing
flux through methylation processes.
Results
Fmo-2 alters one carbon metabolism
Based on the conserved enzymatic mechanism^[74]2,[75]3 and our
published data supporting a key role for nematode FMO-2 in regulating
stress resistance, healthspan and longevity^[76]4, we hypothesized that
FMO-2 may significantly alter endogenous metabolism in C. elegans. To
test if systemic metabolism was broadly altered by FMO-2, we used
untargeted metabolomics analysis (Supplementary Data [77]1) of three
strains with varying fmo-2 expression: the wild-type reference strain
(N2 Bristol), the fmo-2(ok2147) putative knockout strain (FMO-2 KO),
and our previously published long-lived fmo-2 overexpression (KAE9)
strain (FMO-2 OE). The resulting principal component analysis (PCA)
shows a substantial explained variance (65.3%) through principal
components (PC) 1 and 2 (Fig. [78]1a). Our untargeted metabolomics data
suggest a distinct difference in the metabolome between the three
strains, consistent with expression of nematode fmo-2 being sufficient
to modify endogenous metabolism (Fig. [79]1b).
Fig. 1. One carbon metabolism is altered by fmo-2 expression level.
[80]Fig. 1
[81]Open in a new tab
a Principal component analysis of untargeted LC-MS metabolomics data of
wild type, FMO-2 OE, and FMO-2 KO strains of C. elegans. b Heatmap of
untargeted LC-MS metabolomics data of the wild type, FMO-2 OE and FMO-2
KO. The red color indicates metabolites with a higher abundance, while
the green color indicates those with a lower abundance. c Pathway
enrichment analysis using untargeted LC-MS metabolomics data of wild
type and FMO-2 OE. The size and the color of the bubbles represent the
enrichment factor and the p-value respectively for the pathways. The
darker the color, the more significant the p-value. And larger the
radius of the bubble, the greater the enrichment score. d Comparison of
targeted metabolomics data of metabolites related to OCM between the
wild type (black), FMO-2 OE (blue), and FMO-2 KO (red), normalized to
the average of wild type intensity. pho4 = phosphate.
SAM = s-adenosylmethionine. * represents p < 0.05, ** represents
p < 0.01, and *** represents p < 0.001 using two-tailed Student’s
unpaired t-test. # represents p < 0.05 and ## represents p < 0.01 using
one-way ANOVA trend analysis. n = 7 biologically independent
experiments. For t-test (represented by *) p-value = 0.00854 (WT vs
FMO-2 OE, Homocysteine), 0.0063 (WT vs FMO-2 KO, Methionine), 0.0449
(WT vs FMO-2 OE, Pyridoxal 5′-pho4), 0.00144 (WT vs FMO-2 OE, SAM). For
trend analysis (represented by #), p-value = 0.0016 (Homocysteine),
0.0059 (Methionine), 0.0213 (Pyridoxal 5′-pho4), 0.0013 (SAM). In box
plots, the median is shown by the center line. The upper boundary of
the box represents the 75% interquartile range, while the lower
boundary represents the 25% interquartile range. a–c are generated
using MetaboAnalyst. Statistics for c and d are in Supplementary
Table [82]1 and [83]2, respectively No notes = Not significant.
Having established broadly that fmo-2 expression modifies metabolism,
we next asked what key metabolic aspects are modified. Using p-value
<0.05 as our significance threshold, we identified five metabolic
pathways that are significantly altered by the overexpression of fmo-2,
most of which are involved in amino acid metabolism (Fig. [84]1c,
Supplementary Table [85]1). Of the five pathways, we observed the most
significant enrichment in glycine, serine, and threonine metabolism
(Fig. [86]1c). Exogenous supplementation of glycine in worm diet is
reported to extend lifespan by remodeling the methionine cycle^[87]22,
a component of one carbon metabolism (OCM) and another significantly
enriched metabolic pathway from our analysis, cysteine and methionine
metabolism (Fig. [88]1c, Supplementary Table [89]1). Indeed, OCM is a
nexus of multiple metabolic pathways that are necessary for survival;
OCM is implicated in multiple longevity pathways, including dietary
restriction, insulin/IGF-1 signaling, and the metformin-induced
longevity response^[90]13,[91]16,[92]23. Due to its relevance in
multiple longevity pathways and the direct involvement of cysteine and
methionine metabolism within this metabolic network, we postulated that
fmo-2 regulates longevity through its interactions with OCM.
To test whether fmo-2 expression modifies OCM, we used targeted
metabolomics analysis on a panel of metabolites involved in OCM and
related pathways to determine whether their abundance levels were
altered following fmo-2 expression (Supplementary Data [93]3). We
hypothesized that the affected metabolites would have abundance levels
that correlate with fmo-2 expression level. To look for changes between
groups, we initially compared the metabolite levels between the wild
type and FMO-2 OE and also between the wild type and FMO-2 KO. In line
with our hypothesis that OCM is altered by fmo-2 expression, we
observed significant changes in abundance levels of homocysteine,
s-adenosylmethionine (SAM), cystathionine and pyridoxal 5′-phosphate in
FMO-2 OE, and of methionine in FMO-2 KO, when compared to the wild type
(Supplementary Table [94]2). To test whether fmo-2 expression levels
statistically modulate OCM metabolites in correlation with fmo-2
expression, we performed ANOVA trend analysis on the data from all
three strains. We identified a significant change in levels of
methionine, pyridoxal 5′-phosphate, homocysteine, and SAM in line with
changes in the expression of fmo-2 (Fig. [95]1d). Furthermore, betaine,
folic acid, serine, and cystathionine levels all show a trend with
fmo-2 expression, but they did not reach statistical significance below
a 0.05 p-value (Supplementary Fig. [96]1). Taken together, our results
are consistent with the hypothesis that the OCM pathway is modified by
fmo-2 expression.
One carbon metabolism interacts with fmo-2 to regulate stress resistance and
longevity
Having established that FMO-2 modifies endogenous metabolism broadly
and OCM specifically, we next hypothesized that these metabolic changes
are causal for longevity phenotypes. Previous studies identify
increased stress resistance as a common phenotype shared by multiple
long-lived organisms both within and between species^[97]24–[98]27 and
fmo-2 is known to be highly regulated at the transcript level by many
stresses^[99]28. To determine the functional interaction between fmo-2
and OCM, we used RNAi to knockdown the expression of genes involved in
OCM (Fig. [100]2a) and tested for their role in promoting or repressing
survival against the oxidative stressor paraquat. Of the eight genes
that we tested, the individual knockdown of four genes, sams-1, mel-32,
mtrr-1, and Y106G6E.4, exhibit altered stress resistance phenotypes for
the wild type and FMO-2 OE (Supplementary Fig. [101]2A–D), as assessed
using log-rank test with a cutoff threshold of p < 0.05 compared to the
empty vector (EV) controls. Two of these genes, sams-1, and mel-32,
show interaction with FMO-2 OE, as assessed by Cox regression analysis
with p-value cutoff of 0.01 (Supplementary Table [102]3). We note that
7 of 8 genes showed significant (p < 0.05) knockdown via RT-QPCR and
the 8th p = 0.066, with most genes knocked down by at least 70% by
their individual RNAi (Supplementary Fig. [103]3). Interestingly, while
sams-1 knockdown extends worm lifespan^[104]17, we find that knocking
down sams-1 decreases the stress resistance of the wild type and FMO-2
OE (Supplementary Fig. [105]2A), suggesting that the regulation of
lifespan and stress resistance are uncoupled in this instance, as have
been reported previously^[106]29. This result is similar to previous
work showing that sams-1 knockdown is detrimental to survival under
pathogen exposure^[107]30. We further find that sams-1 knockdown
interacts with FMO-2 OE, whereby it more severely decreases FMO-2 OE
paraquat resistance compared to that of the wild type (Supplementary
Fig. [108]2A). Conversely, we find that knocking down mel-32 increases
the resistance of both the wild type and FMO-2 OE, but it again shows
an interaction with FMO-2 OE, whereby FMO-2 OE has more modest
extension compared to the wild type (Supplementary Fig. [109]2B).
Fig. 2. Fmo-2 interacts with OCM genes to regulate lifespan.
[110]Fig. 2
[111]Open in a new tab
a Diagram of OCM network. Genes included here are labeled in blue and
genes not included are labeled in gray. Lifespan analysis comparing the
wild type, FMO-2 OE and FMO-2 KO on empty-vector (EV) and b alh-3 RNAi,
c atic-1 RNAi, d sams-1 RNAi, e mel-32 RNAi, f cth-2 RNAi, g mtrr-1
RNAi, and h Y106G6E.4 RNAi. Percent change in mean lifespan compared to
their respective EV controls are shown in the left bottom corner of the
figures. Black circle = wild type, blue diamond = FMO-2 OE, and red
triangle= FMO-2 KO. Solid line = EV and dotted line = RNAi. * denotes
significant change in mean lifespan at p < 0.05 using log-rank and #
denotes significant interaction between the RNAi of interest and fmo-2
genotype at p < 0.01 using Cox regression analysis. N.S. = not
significant. Statistics are in Table [112]1, Supplementary
Table [113]5, and Supplementary Data [114]5.
We also find that knocking down Y106G6E.4 and mtrr-1 increase the
stress resistance of both the wild type and FMO-2 OE without
significant interaction with FMO-2 OE (Supplementary Fig. [115]2C, D),
suggesting that the stress resistance conferred by the suppression of
these genes is independent of fmo-2. Knocking down atic-1, alh-3,
cbs-1, and cth-2 yield inconsistent results in our experiments
(Supplementary Table [116]4). Overall, our data show that 2/8 of the
genes that we tested interact with FMO-2 OE to modify paraquat
resistance. While these results do not definitively prove that FMO-2
acts through OCM, they are consistent with the hypothesis that OCM is a
regulator of stress resistance and that there is a genetic interaction
between fmo-2 and OCM in that regulation.
To test the interaction between fmo-2 and OCM more directly, we
performed lifespan experiments using RNAi knockdown of genes from our
paraquat resistance screen (Fig. [117]2b–h). We included FMO-2 KO in
the lifespan analysis to determine if the interactions that we identify
are dependent on fmo-2 expression. Multiple gene knockdowns show
altered lifespan phenotypes for the wild type, FMO-2 OE, and FMO-2 KO
that suggest interaction with FMO-2. Of the eight genes we tested,
knockdown of two genes, alh-3 and atic-1, suppress the lifespan
extension of FMO-2 OE without affecting the lifespan of the wild type
and FMO-2 KO (Fig. [118]2b, c). Similar to our paraquat stress
resistance experiments, we conclude this based on (1) log-rank test
showing changes only in FMO-2 OE lifespan with a cutoff threshold of
p < 0.05 compared to their respective empty vector (EV) controls, and
(2) cox regression analysis (Supplementary Table [119]5, Table [120]1)
showing a differential interaction with FMO-2 OE and FMO-2 KO, using a
cutoff threshold of p < 0.01. These results suggest that alh-3 and
atic-1 are required for fmo-2-mediated lifespan extension and are
consistent with previous reports that their expression levels are
upregulated in long-lived worms^[121]11,[122]31. alh-3 is upregulated
in eat-2 mutants and atic-1 is upregulated in both eat-2 and daf-2
mutants^[123]11,[124]31–[125]33. In addition, atic-1 is involved in the
conversion of 10-formyl-THF to
formyl-5-aminoimidazole-4-carboxamide-1-beta-4-ribofuranoside (FAICAR)
(Fig. [126]2a). FAICAR inhibits the synthesis of AICAR^[127]34, which
is a molecule reported to induce the phosphorylation and activation of
AMP-activated protein kinase (AMPK)^[128]35, whose activation extends
lifespan^[129]36. Therefore, suppressing atic-1 expression may reduce
FMO-2 OE lifespan by inhibiting AMPK activation. This potentiates the
possibility that AMPK may also be involved in the FMO-2-mediated
longevity regulation. In sum, it is plausible that these genes are
required for multiple longevity pathways, including FMO-2
overexpression.
Table 1.
Cox regression analysis of OCM/tryptophan metabolism genes and fmo-2
modified strains lifespan data
Experimental FMO-2 KO FMO-2 OE Interact with FMO-2 KO Interact with
FMO-2 OE
Haz. ratio p-value Haz. ratio p-value Haz. ratio p-value Haz. ratio
p-value Haz. ratio p-value
alh-3 0.974 0.715 1.050 0.510 0.480 <0.001 1.088 0.413 1.333 0.005
atic-1 0.923 0.269 1.051 0.501 0.511 <0.001 0.898 0.302 1.641 <0.001
sams-1 0.298 <0.001 1.003 0.972 0.359 <0.001 1.632 <0.001 2.652 <0.001
mel-32 1.098 0.246 1.110 0.191 0.505 <0.001 0.891 0.307 0.455 <0.001
cth-2 1.619 <0.001 1.122 0.072 0.491 <0.001 1.183 0.058 0.932 0.417
mtrr-1 1.184 0.025 1.152 0.058 0.490 <0.001 0.963 0.718 1.089 0.421
Y106G6E.4 0.949 0.470 1.053 0.485 0.481 <0.001 0.848 0.118 1.079 0.456
kmo-1 3.160 <0.001 1.389 <0.001 0.513 <0.001 0.678 <0.001 0.798 0.038
tdo-2 0.306 <0.001 1.388 <0.001 0.510 <0.001 1.464 0.001 1.830 <0.001
nkat-1 1.090 0.221 1.298 <0.001 0.512 <0.001 0.836 0.067 0.785 0.015
Formate 0.623 <0.001 1.320 0.005 0.403 <0.001 1.231 0.156 1.625 0.001
[130]Open in a new tab
Experimental = effect of experimental condition on worm lifespan; FMO-2
KO = effect of knocking out fmo-2 on worm lifespan; FMO-2 OE = effect
of overexpressing fmo-2 on worm lifespan; interact with FMO-2
KO = interaction between experimental condition and knocking out fmo-2
on worm lifespan; interact with FMO-2 OE = interaction between
experimental condition and overexpressing fmo-2 on worm lifespan;
Hazard ratio > 1 = decrease in lifespan; Hazard ratio < 1 = increase in
lifespan.
In contrast to alh-3 and atic-1, knockdown of sams-1 increases the
lifespan of the wild type, FMO-2 KO, and FMO-2 OE animals
(Fig. [131]2d). However, sams-1 knockdown shows interactions with both
FMO-2 KO and FMO-2 OE, whereby both strains show less extension under
sams-1 knockdown compared to the wild type, with FMO-2 OE showing a
larger interaction effect size (Table [132]1). This suggests that there
is a functional overlap between sams-1 and fmo-2 in regulating
longevity. Additionally, we find that knocking down mel-32 only
interacts with FMO-2 OE and extends its lifespan (Table [133]1 and
Fig. [134]2e). It is plausible that the metabolic alterations resulting
from increased fmo-2 expression are required for mel-32 gene
suppression to promote worm lifespan.
Knockdown of the remaining four genes, cbs-1, cth-2, mtrr-1, and
Y106G6E.4, do not show interaction with FMO-2 KO and FMO-2 OE
(Table [135]1). However, knocking down cth-2 decreases the lifespan of
all three strains, suggesting that this gene is generally required for
worm health and survival (Fig. [136]2f). Knocking down mtrr-1 and
Y106G6E.4 do not affect the lifespan of the worms in our experiments
(Fig. [137]2g, h). Knocking down cbs-1 yields inconsistent results in
our experiments (Supplementary Table [138]5).
In total, our data show that half (4/8) of the genes tested interact
with FMO-2 OE: two genes are required for FMO-2 OE lifespan extension
(alh-3 and atic-1), another gene interacts with FMO-2 OE and FMO-2 KO
(sams-1), placing it in the same functional pathway with FMO-2, and one
gene only extends the lifespan of FMO-2 OE when knocked down (mel-32).
Together, our lifespan data, combined with our metabolomics results,
support an interaction between fmo-2 and genes involved with OCM in
regulating worm lifespan.
Fmo-2 influences longevity by modulating the transmethylation pathway
Our data are consistent with a model where fmo-2 interacts with OCM to
regulate longevity and stress resistance. Previous studies identify
multiple pathways that affect longevity and are also involved in OCM,
including nucleotide metabolism, the transsulfuration pathway, and the
transmethylation pathway^[139]11,[140]16,[141]17. Some of these
pathways are also implicated in modifying longevity downstream of
dietary restriction in multiple animal models^[142]16,[143]17,[144]37,
making it likely that one or more of these pathways are in the same
functional pathway as fmo-2. However, the metabolic consequences of
fmo-2 expression on these pathways are not clear based on the changes
observed in our targeted metabolomics analysis alone, as the data only
show metabolic changes at a single time point and most of the
metabolites within OCM are intermediate metabolites. The stress
resistance and lifespan results further complicate interpretation as
some genes do not affect these phenotypes and some have effects that
are independent of fmo-2.
To help determine the biological relevance of the changes we observed
in the OCM network following fmo-2 expression, we applied a
computational model (Supplementary Data [145]6) to predict how enzyme
expression (Supplementary Table [146]6) changes may affect the output
fluxes of OCM. The model assumes a steady-state mass balance of fluxes
in the reactions illustrated in Fig. [147]3a. This simple model
includes eight reaction fluxes and five fluxes representing transport
of methionine (met), tetrahydrofolate (thf), s-adenosylmethionine
(sam), cysteine (cys), and 5,10-methylenetetrahydrofolate (5,10thf)
into and out of the folate cycle and the methionine cycle. The model
output fluxes represent important inputs for the nucleotide metabolism,
the transsulfuration pathway, and the transmethylation pathway, each of
which are reported to be important for influencing the aging
process^[148]11,[149]16,[150]17 and are potential targets for the
fmo-2-mediated longevity response. The stoichiometric coefficients for
the reaction and transport processes in this system are stored in the
matrix S (Supplementary Table [151]7), where under steady-state
conditions S*J = 0, where J is the vector of fluxes^[152]38,[153]39.
The entries in the vector J and matrix S are defined in Fig. [154]3a.
Vectors that satisfy the mass-balance relationship S*J = 0 are said to
belong to the nullspace of S. To predict how changes in the expression
of genes for the enzymes catalyzing the reactions in this network may
affect the output fluxes, we projected the gene expression data
(Supplementary Table [155]6) onto the nullspace of S (details provided
in the “Methods”). This projection predicts an inverse correlation
between fmo-2 expression and flux through methylation reactions, where
the methylation flux is predicted to be reduced in FMO-2 OE and
increased in FMO-2 KO compared to wild type (Fig. [156]3b,
Supplementary Fig. [157]3A–C). This analysis does not predict
correlative changes to flux through nucleotide metabolism or the
transsulfuration pathway.
Fig. 3. Methylation flux is altered following changes in fmo-2 expression.
[158]Fig. 3
[159]Open in a new tab
a Schematic of computational model. b Model predictions of output
metabolic fluxes. c SAM/SAH ratio of the wild type (WT), FMO-2 OE (OE),
and FMO-2 KO (KO). Black color = WT, blue color = OE, red color = KO.
** represents p < 0.01 using two-tailed unpaired Student’s t-test and
## represents p < 0.01 using one-way ANOVA trend analysis. For t-test,
p-value = 0.0061 (WT vs FMO-2 OE), and for trend analysis, p-value =
0.0074. In box plot, the median is shown by the center line. n = 7
biologically independent experiments. The upper boundary of the box
represents the 75% interquartile range, while the lower boundary
represents the 25% interquartile range.
Based on this analysis, we hypothesized that artificially decreasing
the flux through methylation should replicate FMO-2 OE longevity in the
wild type and FMO-2 KO strains, while not affecting the FMO-2 OE worms.
sams-1 encodes for s-adenosylmethionine synthase and is involved in the
conversion of methionine into s-adenosylmethionine (SAM). Suppression
of sams-1 has been previously shown to decrease SAM level^[160]40 and
increase longevity^[161]17. Moreover, while sams-1 has multiple
orthologs, previous work has shown that knocking down sams-1 is
sufficient to manipulate SAM levels in worms^[162]40. We find that
sams-1 RNAi recapitulates FMO-2 OE lifespan extension in the wild type
while interacting with FMO-2 OE, whose lifespan was significantly less
affected (Fig. [163]2d, Table [164]1). Our data are consistent with
previous studies showing that knockdown of sams-1 fails to further
extend the lifespan of the genetic DR model eat-2 mutants^[165]17,
thereby placing sams-1 knockdown in the same functional pathway as
FMO-2 OE.
To validate the model metabolically, we used the abundance level of SAM
and s-adenosylhomocysteine (SAH) from our targeted metabolomics
analysis to calculate the SAM/SAH ratio. The SAM/SAH ratio is used as a
biomarker for methylation potential, where a decrease in the ratio
suggests a hypomethylated state and an increase suggests a
hypermethylated state^[166]41,[167]42. Consistent with our
computational model prediction, we observed a significant reduction in
the SAM/SAH ratio for FMO-2 OE (hypomethylation) and a significant
trend in SAM/SAH ratio corresponding to fmo-2 expression by ANOVA trend
analysis. (Fig. [168]3c). In addition, we find that supplementing the
worm diet with 2 mM SAM results in a reduction of FMO-2 OE lifespan
(Supplementary Fig. [169]3D). Overall, our computational model
prediction and experimental data support the hypothesis that fmo-2
expression reduces flux through the transmethylation pathway, and that
this reduction extends worm lifespan.
Mammalian FMO metabolomics analysis reveals tryptophan as a substrate of
FMO-2
Our data thus far suggest a model where fmo-2 interacts with OCM to
modulate the aging process. However, since FMOs are promiscuous enzymes
that oxygenate many nucleophilic atoms, the mechanism by which fmo-2
induction leads to changes in OCM is not readily evident. FMOs are
known as xenobiotic metabolizing enzymes, with many known exogenous
targets and few known endogenous targets^[170]1. Despite extensive
knowledge on their enzymatic activity and recent data linking FMOs to
endogenous metabolism, no link between specific and systemic metabolism
has been made. We hypothesize that a limited number of FMO targets are
causal in FMO-2’s effects on OCM and, importantly, on the aging
process.
Due to the high degree of conservation of catalytic residues between
mouse FMOs and CeFMO-2 (Fig. [171]4a), we referred to our previously
published targeted metabolomics of mouse FMO overexpressing (OE) HepG2
cells to determine potential metabolic targets of FMO-2^[172]10. Using
this previously published dataset, our selection criteria for putative
substrates of FMO-2 included identifying metabolites that had decreased
abundance in at least three of the five FMO OE cell lines to pDEST
controls. We used this stringent criteria to identify the most
well-conserved targets of FMOs, given that no data exist for CeFMO-2
targets. Using this approach, we identified tryptophan and
phenylalanine as potential substrates of FMOs (Fig. [173]4b). To
determine if either of these are substrates of FMO-2, we measured the
enzymatic activity of isolated FMO-2 protein in the presence of varying
concentrations of tryptophan and phenylalanine. We find that FMO-2 is
active toward tryptophan at a reasonable K[m] and k[cat] (K[m]:
880 ± 430 µM; k[cat]: 9.7 ± 1.5 s^−1), suggesting that tryptophan is a
viable substrate of FMO-2 (Fig. [174]4c, Supplementary Table [175]9).
In comparison, rat and invertebrate TDO enzymes with tryptophan
demonstrate Km values of 221 µM and 276.5 µM, respectively, while
mammalian IDO1 and IDO2 have Km values of 19.1–74 µM and 45.9 mM,
respectively^[176]43, putting the Km value of CeFMO-2 within the range
of TDO and IDO proteins. FMO-2 was also active toward phenylalanine,
but enzymatic activity did not become apparent until 10 mM, suggesting
that phenylalanine is not likely a good endogenous substrate of FMO-2
(Fig. [177]4d). Since FMO-2 has no previously reported activity toward
tryptophan, we used LC-MS with 100, 250, and 500 µM tryptophan under
the same enzymatic conditions to determine the product of tryptophan
oxygenation. Our resulting data show increasing formation of
N-formylkynurenine in a concentration-dependent manner with increasing
tryptophan in each of the samples. This result suggests that
N-formylkynurenine is a product formed by FMO-2 activity toward
tryptophan (Fig. [178]4e). Consistent with these findings, we observe
that tryptophan level is significantly reduced in FMO-2 OE compared to
the wild type (Fig. [179]4f). In addition, we find that tryptophan
induces fmo-2 gene expression in vivo, consistent with our hypothesis
that tryptophan is a substrate of FMO-2 (Supplementary Fig. [180]4).
The kinetic parameters of FMO-2 toward NADPH, methimazole, and
tryptophan are summarized in Fig. [181]4g. We also tested additional
known substrates of mammalian FMOs, all of which were either poor
substrates (e.g., cysteine, phenylalanine, and TMA) or non-substrates
of FMO-2 (e.g., 2-heptanone). They are summarized with either the
concentration of substrate at which FMO-2 activity is first detected or
labeled not determined (N.D.) in Supplementary Table [182]9. While we
have observed that phenylalanine level is also significantly reduced in
FMO-2 OE compared to the wild type (Fig. [183]4h), we believe the
effect of FMO-2 on phenylalanine to be indirect as it is a poor
substrate in vitro (Fig. [184]4d).
Fig. 4. Mammalian FMO metabolomics analysis reveals the tryptophan/kynurenine
pathway as a target of FMO-2.
[185]Fig. 4
[186]Open in a new tab
a Conserved catalytic residues between CeFMO-2 and mFMO5 (indicated by
red arrows). b The level of phenylalanine and tryptophan present in
HepG2 cells expressing pDEST control vector, mFMO2, mFMO4, and mFMO5.
Black dot = vector control and blue triangle = FMO OE. * represents
p < 0.05 by two-tailed unpaired Student’s t-test. For phenylalanine,
p-value = 0.0084 (control vs mFMO2 OE), 0.00095 (control vs mFMO4 OE),
0.04 (control vs mFMO5 OE). For tryptophan, p-value = 0.0188 (control
vs mFMO2 OE), 0.000918 (control vs mFMO4 OE), 0.098 (control vs mFMO5
OE). Data is represented as mean values +/− SD, n = 3 biological
replicates. c, d The reaction rate by concentration for purified
CeFMO-2 enzyme toward tryptophan (n = 7 biologically independent
replicates) and phenylalanine (n = 2 biologically independent
replicates) at 30 °C. Data is represented as mean values +/− SD. e The
abundance of N-formylkynurenine based on LC-MS analysis of CeFMO-2
activity toward 100, 250, and 500 μM tryptophan at 30 °C. n = 3 (100 µM
and 250 µM Tryptophan) and n = 2 (500 µM Tryptophan) biologically
independent replicates. Data is represented as mean values +/− SD. f
Comparison of targeted metabolomics data of tryptophan between the wild
type (black), FMO-2 OE (blue), and FMO-2 KO (red), normalized to the
average of wild type intensity. n = 4 biologically independent
replicates and p-value = 0.028 (WT vs FMO-2 OE). * = p < 0.05 using
two-tailed unpaired Student’s t-test. g Summary table of
Michaelis-Menten parameters for CeFMO-2 cofactor and substrate. h
Comparison of targeted metabolomics data of phenylalanine between the
wild type (black), FMO-2 OE (blue), and FMO-2 KO (red), normalized to
the average of wild type intensity. n = 4 biologically independent
replicates and p-value = 0.0097 (WT vs FMO-2 OE). ** = p < 0.01 using
two-tailed unpaired Student’s t-test. i Lifespan assay comparing the
survival of the wild type, FMO-2 OE, and FMO-2 KO on control and 1 mM
formate supplementation conditions. Black circle = wild type, blue
diamond = FMO-2 OE, and red triangle = FMO-2 KO. Solid line = EV,
dotted line = formate. * denotes significant change in mean lifespan at
p < 0.05 using log-rank and # denotes significant interaction between
the condition of interest and fmo-2 genotype at p < 0.01 using Cox
regression analysis. N.S. = not significant. Statistics are in
Supplementary Table [187]10 and Supplementary Data [188]5. For box
plots in f and h, the median is shown by the center line. The upper
boundary of the box represents the 75% interquartile range, while the
lower boundary represents the 25% interquartile range.
Based on our initial data linking FMO-2 to OCM, it is important to note
that in addition to being a key process in the kynurenine pathway, the
conversion of tryptophan to N-formylkynurenine precedes the conversion
of N-formylkynurenine to kynurenine by formamidase, a process that
releases formate, which is also a carbon source for OCM^[189]44.
Formate can enter OCM through the folate cycle, thus providing a
connection between tryptophan metabolism, the kynurenine pathway, and
OCM. Indeed, we observe a significant interaction between formate
supplementation and FMO-2 OE, whereby supplementing the worm diet with
formate extends the lifespan of the wild-type without further extending
the lifespan of FMO-2 OE (Fig. [190]4i and Table [191]1). Based on this
knowledge, we hypothesize that the kynurenine pathway is a target of
FMO-2 that leads to changes in OCM. To test this hypothesis, we
assessed whether genes involved in tryptophan metabolism interact with
FMO-2 (Fig. [192]5a). Like our RNAi analyses of the OCM genes, we
observe (1) changes in the stress resistance and lifespan of the wild
type, FMO-2 OE, and FMO-2 KO worms following the knockdown of genes
involved in the kynurenine pathway and (2) interactions between these
genes and FMO-2 using the same parameters that we used for the
OCM-related genes (Fig. [193]5b–g, Tables [194]1 and Supplementary
Table [195]3). Here, we again observed a separation between the
regulation of stress resistance and lifespan under kmo-1 and tdo-2
knockdown. Knocking down kmo-1 increases the stress resistance of the
wild type and FMO-2 OE, but shows significant interaction with FMO-2
OE, whereby these worms are less affected by the RNAi compared to their
wild type counterparts (Fig. [196]5b, Supplementary Table [197]3).
Conversely, knocking down kmo-1 decreases the lifespan of the wild
type, FMO-2 OE, and FMO-2 KO (Fig. [198]5e). However, this knockdown
shows significant interaction with FMO-2 KO, whereby these worms are
less affected by the RNAi compared to their wild type counterparts
(Table [199]1). It is possible that knocking down kmo-1 increases flux
through KYNU-1, whose gene suppression extends lifespan^[200]20,
thereby reducing FMO-2 OE lifespan. Similar to atic-1 RNAi, this
finding potentiates that knocking down kynu-1 is in the same functional
pathway as overexpressing fmo-2. Knocking down tdo-2 decreases the
stress resistance of the wild type and FMO-2 OE but has a significant
interaction with FMO-2 OE, whereby these worms are more affected by the
RNAi compared to their wild type counterparts (Fig. [201]5c,
Supplementary Table [202]3). Conversely, knocking down tdo-2 extends
the lifespan of the wild type, FMO-2 OE, and FMO-2 KO (Fig. [203]5f).
Tdo-2 knockdown was previously reported to extend lifespan by
inhibiting tryptophan degradation and thereby improving the regulation
of proteotoxicity^[204]19. We find that tdo-2 knockdown interacts with
both FMO-2 KO and FMO-2 OE to modulate longevity, where both strains
show reduced lifespan extension compared to wild-type animals
(Table [205]1). Knocking down tdo-2, in addition to increasing
tryptophan level^[206]19, may increase fmo-2 activity as a compensatory
response to metabolize tryptophan, thereby resulting in lifespan
extension. Similar to sams-1, it is likely that there is a functional
overlap between tdo-2 and fmo-2 in regulating worm lifespan.
Fig. 5. Fmo-2 interacts with kynurenine metabolism to regulate stress
resistance and lifespan.
[207]Fig. 5
[208]Open in a new tab
a Diagram of the kynurenine pathway. 5 mM paraquat stress resistance
assay comparing the wild type and FMO-2 OE on empty-vector (EV) and b
kmo-1 RNAi, c tdo-2 RNAi, and d nkat-1 RNAi. Lifespan assay comparing
the survival of the wild type, FMO-2 OE, and FMO-2 KO on EV, and e
kmo-1 RNAi, f tdo-2 RNAi, and g nkat-1 RNAi. Percent change in mean
lifespan compared to their respective EV controls are shown in the left
bottom corner of the figures. Black circle = wild type, blue diamond =
FMO-2 OE, and red triangle = FMO-2 KO. Solid line = EV and dotted line
= RNAi. * denotes significant change in mean lifespan at p < 0.05 using
log-rank and # denotes significant interaction between the RNAi of
interest and fmo-2 genotype at p < 0.01 using Cox regression analysis.
N.S. = not significant. Statistics are in Supplementary Tables [209]3,
[210]4, [211]5 and Supplementary Data [212]4, [213]5.
Knocking down nkat-1 increases the stress resistance of the wild type
and FMO-2 OE but does not show a significant interaction with FMO-2 OE,
consistent with it not acting in the same functional pathway as FMO-2
OE to regulate stress resistance (Fig. [214]5d, Supplementary
Table [215]3). Knocking down nkat-1 only extends the lifespan of FMO-2
OE but does not show an interaction with it below our statistical
threshold (Fig. [216]5g, Table [217]1). nkat-1 is involved in the
conversion of kynurenine to kynurenic acid, which is a neuromodulator
that has effects on behavior and possibly lifespan^[218]45,[219]46. We
hypothesize that either increased kynurenic acid production could limit
lifespan slightly and suppressing kynurenic acid synthesis therefore
increase FMO-2 OE, or that suppressing kynurenic acid synthesis may
lead to increased flux in other parts of the pathway (e.g., NAD) that
may also be beneficial for lifespan under the metabolic changes
following fmo-2 overexpression.
Knocking down afmd-1 yields inconsistent results in our experiments
(Supplementary Table [220]5). In sum, our data support an interaction
between multiple genes involved in the kynurenine pathway and FMO-2.
Furthermore, our data again separate stress resistance and longevity,
and are consistent with the possibility that FMO-2 confers stress
resistance and longevity through separable mechanisms.
To prevent egg hatching, our stress resistance and lifespan experiments
use 5-fluorodeoxyuridine (FUdR), which is a compound that can interact
with a component of one-carbon metabolism, thymidylate synthase
(TYMS-1)^[221]47. We tested whether this drug influences FMO-2 OE
lifespan phenotype. We find that FMO-2 OE extends the worm lifespan
even in the absence of FUdR (Supplementary Fig. [222]5), suggesting
that FUdR does not have a major effect on FMO-2-mediated lifespan
extension and consistent with published data showing that FUdR is also
not required for decreased sams-1 to extend lifespan^[223]48. While
these results suggest that FUdR is not crucial for the effects of
FMO-2-mediated lifespan extension, they do not rule out some other
interaction within the pathways tested, and should be considered when
interpreting the data.
Discussion
For half a century, FMOs have been primarily classified as xenobiotic
enzymes. However, the mechanisms by which these enzymes affect
endogenous metabolism are still not well studied, with the exception of
human FMO1 (and FMO3) and mouse FMO1 converting hypotaurine to
taurine^[224]49. Based on our data, we propose a model where
overexpression of fmo-2 to levels similar those observed under hypoxia
and dietary restriction is sufficient to remodel metabolism in the
nematode C. elegans (Fig. [225]6). Here, we show that fmo-2, a
regulator of longevity that is critical for lifespan extension and
stress response under dietary restriction and hypoxia, interacts with
both tryptophan and one-carbon metabolism to confer longevity and
stress resistance benefits. We find that modulating the expression of a
single oxygenating protein can cause a multitude of metabolic and
physiological effects, similar to the activation of transcription
factors and kinases. Our results suggest a broader, more significant
role for FMO-2 than previously known.
Fig. 6. Proposed model.
Fig. 6
[226]Open in a new tab
In control conditions, there is very low fmo-2 expression, leading to
low levels of tryptophan metabolism/kynurenine production through
FMO-2, and maintaining normal flux through one carbon metabolism and
normal lifespan. When fmo-2 is induced, FMO-2 oxygenates tryptophan,
leading to increased kynurenine production and decreased methylation
output flux through OCM, thereby extending nematode lifespan. When
fmo-2 is absent, these metabolic changes do not occur, preventing an
extension in lifespan. The gray line represents decreased flux and the
blue line represents increased flux.
Our data are consistent with a model where the reduction of flux
through the methylation pathway leads to longevity benefits. By
projecting gene expression data to a stoichiometric model for OCM
metabolism, we predict that FMO overexpression results in a reduction
in methylation flux. This model-based prediction based on gene
expression data is experimentally validated, indicating that this
approach can be a powerful tool to simplify the understanding of
complex metabolic pathways and to study the biology of aging.
Perturbation in the SAM/SAH ratio by either the supplementation of
metformin or a mutation in sams-1 also extends worm
lifespan^[227]13,[228]17. While multiple studies report that methionine
restriction robustly extends lifespan across species, including worms,
flies, and mice^[229]13,[230]50,[231]51, others show that exogenous
supplementation of methionine is not detrimental to lifespan^[232]52.
In our study, we observe a level of methionine that shows an increasing
trend with increasing fmo-2 expression level (Fig. [233]1d). Although
SAM levels depend on methionine, there is precedent for SAM levels
being low even when methionine levels are high^[234]53. Two potential
reasons for these observations could be (1) FMO-2 blocks methionine to
SAM conversion, which would increase methionine, or (2) there could be
an increase in the fluxes of internal reactions within OCM. We
hypothesize that #2 is more likely, since methionine is not an in vitro
substrate for FMO-2 (Supplementary Table [235]9). Taken together, these
findings suggest that methionine utilization rather than methionine
abundance is a key factor that influences the aging process.
Although suppressing sams-1 expression phenocopies FMO-2 OE lifespan in
the wild type and FMO-2 KO, doing so reduces the stress resistance of
the worms against paraquat. This separation of lifespan and stress
resistance is occasionally observed under other long-lived
conditions^[236]54. It is unclear if simply reducing methylation is
sufficient to promote longevity benefits, or if this mechanism requires
suppression of specific methylation processes. It will be important for
future studies to determine how cells regulate different methylation
fluxes under sams-1 knockdown and decreased overall methylation. One
potential mechanism under this genetic condition could be that specific
methyltransferases that are essential for survival will have higher
affinity to methyl groups to outcompete other nonessential or
deleterious methyltransferases. It will also be interesting to test why
both FMO-2 OE and sams-1 RNAi extend lifespan, while causing opposite
effects on paraquat resistance. We hypothesize that the reduction of
SAM level in sams-1 RNAi is more severe than in FMO-2 OE, which could
cause more severe effects, such as a reduction in stress resistance.
This could explain some of the phenotypes observed in sams-1 RNAi
animals (e.g., delayed development^[237]48) that we do not observe in
FMO-2 OE^[238]4.
We note that while our data suggest methylation as the key downstream
effector of FMO-2, we have not excluded the possibility that the
transsulfuration pathway may also be involved in this mechanism. The
transsulfuration pathway is reported to be a necessary and sufficient
component of DR-mediated lifespan extension in flies^[239]16. It will
be interesting to determine the mechanistic relationship between the
transsulfuration and transmethylation pathways in regulating longevity.
We also note that the stress resistance and lifespan experiments use
FUdR. FUdR is a potential confounding influence, and although FMO-2 OE
mediated lifespan extension is independent of FUdR, its effect cannot
be fully excluded since it wasn’t studied in all of the conditions. We
also note that we considered whether our findings could result from an
artifact of an overexpression model of FMO-2. However, we believe this
is unlikely based on our previous work establishing FMO-2 as a
regulator of longevity downstream of DR^[240]4 and a more recently
published work reporting that OCM is altered by the DR genetic model
eat-2 mutant^[241]14. Similarly, we note that these experiments use
RNAi, and that while some of these enzymes are essential and null
mutants are not available, others are available and could prove useful
in future studies focusing on both OCM and the tryptophan pathway.
Our data also support an interaction between fmo-2 and tryptophan
metabolism to influence longevity. These findings are particularly
interesting because we identify a putative endogenous metabolic pathway
of FMOs in relation to the aging process. Based on cell line
metabolomics, enzyme kinetics, and LC-MS data, there are at least two
plausible mechanisms for how oxygenation of tryptophan by FMO-2 can
lead to the synthesis of N-formylkynurenine. First, FMOs across taxa
are known to dimerize and form higher-order oligomers^[242]55,[243]56.
Therefore, it is possible that FMO-2 dimerizes and dioxygenates
tryptophan forming N-formyl-kynurenine, which is then converted to
kynurenine by formamidase. Second, the same process could be involved
in subsequent oxygenation by FMO-2 monomers, but it is unknown how
stable a monooxygenated form of tryptophan would be within the cell,
making the first mechanism more likely. To our knowledge this is a
primary example of the dioxygenation of a substrate that could
potentially require dimerized FMOs. The mechanism of this reaction and
its potential requirement of dimerized FMOs will be a target of future
research. Furthermore, the dioxygenation of tryptophan by FMOs is
especially interesting considering only dioxygenases, such as tdo-2,
ido-1, and ido-2, have been shown to mediate the conversion of
tryptophan to N-formylkynurenine^[244]19. Regardless, our data
implicate tryptophan as a bona fide in vitro and likely in vivo
substrate of animal FMOs either through dioxygenating or
monooxygenating mechanisms. Although we tested multiple conventional
and unconventional FMO substrates, such as methimazole^[245]55–[246]57
and 2-heptanone^[247]58 (Supplementary Table [248]9), respectively,
much work remains to fully establish the general FMO-2 substrate
profile and how it compares to those of mammalian FMOs. We note that
the pathway analysis (Supplementary Table [249]1) also examined
tryptophan metabolism, but as a pathway it was not significantly
enriched. This could be due to one of at least 3 reasons: (1) while
there are 30 total metabolites within the tryptophan pathway, the
untargeted metabolomics assay was only able to cleanly identify 5 of
the 30 metabolites, two of which were significantly altered. Many
metabolites within this pathway are transient or present at very low
concentrations, requiring more specialized sample preparation or
additional targeted metabolomics methods to measure directly. Using our
general untargeted metabolomics method we were therefore unable to
measure much of the pathway. (2) It is possible that FMO-2’s effects on
tryptophan lead to compensatory responses elsewhere in the tryptophan
pathway that mask the effects, or (3) FMO-2 may also metabolize
tryptophan to compounds other than NFK that do not lead to changes
throughout the canonical tryptophan pathway. We favor explanation #1,
as the untargeted data were always meant to provide clues for further
analysis and do not have the same scope and statistical power for every
pathway. Unlike the pathway analysis, which considers multiple
metabolites for statistics, our initial identification of tryptophan
metabolism was based on a change of a single metabolite, tryptophan.
Our data support a model where the interaction between FMO-2 and
tryptophan metabolism directly or indirectly modulates the metabolite
profile of OCM, altering flux patterns that are consistent with our
computational model predictions and subsequent genetic analyses.
Further investigation is needed to understand the complete details of
the fmo-2-mediated connection between OCM and tryptophan in regulating
lifespan. Based on the knowledge we gained from this study and previous
work, we propose the following two mechanisms may occur together or
separately: (1) Oxygenation of tryptophan by FMO-2 alters OCM flux by
increasing formate levels as a direct link between tryptophan
metabolism and OCM. Formate is a single carbon-containing molecule that
can enter the folate cycle as a carbon source^[250]44, and we find that
formate addition extends WT lifespan without affecting FMO-2 OE.
Formate is generated as a byproduct when kynurenine is synthesized from
N-formylkynurenine by formamidase^[251]44. It is possible that
increased formate levels are causal in conferring stress resistance and
longevity benefits under metabolically stressful conditions, such as DR
or hypoxia. This hypothesis could explain why knocking down alh-3
reduces FMO-2 OE lifespan. Knockdown of alh-3 would prevent the
breakdown of 10-formyl-THF and synthesis of THF, which would then
prevent formate from entering OCM by being converted into
10-formyl-THF^[252]59–[253]61. Since we hypothesize that increased
formate enters OCM to affect worm lifespan, preventing the
incorporation of formate into OCM would result in an accumulation of
formate, a molecule that has been reported to be toxic in human
cells^[254]62. Thus, this would result in the reduction of lifespan
specifically for FMO-2 OE, as we have observed. (2) FMO-2 interacts
with the mechanistic target of rapamycin (mTOR). Dietary restriction
leads to inhibition of mTOR signaling, which is a central regulator of
lifespan and aging^[255]63. Interestingly, both DR- and
rapamycin-mediated mTOR inhibition induce the expression of FMOs. A
recent study shows that diaminodiphenyl sulfone (DDS) induces the
expression of fmo-2 and extends lifespan, but it does not further
extend lifespan in combination with rapamycin^[256]41. This finding is
consistent with the hypothesis that fmo-2 interacts with mTOR
inhibition to extend lifespan. We also show that fmo-2 interacts with
SAM and tryptophan metabolism, both of which are known to alter mTOR
activity^[257]64–[258]66. Thus, examination into the role of mTOR in
fmo-2-mediated lifespan extension is warranted.
Taken together, our study expands the role of FMO-2 from a xenobiotic
enzyme to a metabolic regulator of longevity that has global effects on
the metabolome in worms. In particular, the identification of OCM as a
target of FMO-2 has implications outside the aging field, considering
that OCM remodeling has been studied under the context of cancer
biology for more than 70 years^[259]67. Furthermore, through the
identification of tryptophan as a putative substrate for CeFMO-2, this
study highlights the conserved importance of FMOs in multiple contexts,
including aging and many diseases where OCM and/or the kynurenine
pathway play a role. These findings illustrate the potential for
therapeutic targets of these proteins for treating age-related diseases
and/or increasing longevity and healthspan. This exciting translational
potential for the conserved roles of FMOs will be a target for future
research.
Methods
Ethical approval or guidance was not required as local laws governing
the use of research animals do not apply to invertebrate models.
Strains and growth conditions
Standard C. elegans cultivation procedures were used as previously
described^[260]4,[261]68. Briefly, N2 wild type, KAE9
((eft-3p::fmo-2 + h2b::gfp + Cbr-unc-119(+)), and VC1668
(fmo-2(ok2147)) strains were maintained on solid nematode growth media
(NGM) using E. coli OP50 throughout life except where RNAi (E. coli
HT115) were used. Worms were transferred or picked gently using a
platinum wire. All experiments were conducted at 20 °C.
Metabolomics
OP50 bacteria were treated with 0.5% (v/v) paraformaldehyde (50980495,
Fisher Scientific) as previously described^[262]68. Briefly, overnight
bacterial culture was treated with 0.5% paraformaldehyde (PFA) and
incubated at 37 °C in a shaking incubator at 200 rpm for 1 h. The
treated bacteria were centrifuged and washed with LB five times to
remove residual PFA before seeding onto 100 mm NGM plates.
Approximately 500 eggs were put on these plates and grown until they
reached late L4 larval stage. The worms were washed off the plates with
10 mL of M9 buffer and were collected in 15 mL conical tubes. The worms
were pelleted using a clinical centrifuge for 1 min at 150 × g and the
supernatant was vacuum aspirated. The worms were washed once with 10 mL
of M9 buffer and then with 10 mL of 150 mM ammonium acetate (A637,
Fisher Scientific) to remove phosphates from M9, each time being
centrifuged and the supernatant being aspirated. After these washing
steps, the pellets were flash frozen in liquid nitrogen.
Metabolites were extracted from pellets by addition of 500 µL of
ice-cold 9:1 methanol: chloroform, followed immediately by probe
sonication for 30 s with a Branson 450 Sonicator. The resulting
homogenates were kept on ice for 5 min and were then centrifuged for
10 min at 4000 × g at 4 °C. Supernatant was then transferred to
autosampler vials for analysis. Hydrophilic interaction liquid
chromatography-electrospray ionization mass spectrometry
(HILIC-LC-ESI-MS) analysis was performed in negative ion mode using an
Agilent 1200 LC system coupled to an Agilent 6220 time-of-flight mass
spectrometer equipped with a Dual ESI source. Chromatography was
performed as previously described^[263]69,[264]70. Briefly, a
Phenomenex Luna NH2 column was used with dimensions of 150 mm × 1.0 mm
i.d., the flow rate was 0.07 mL/min, and the injection volume was
10 µL. Mobile phase composition was as follows: mobile phase A was
acetonitrile, and mobile phase B was 5 mM ammonium acetate in water
adjusted to pH 9.9 using ammonium hydroxide. The gradient was used as
follows: 20% to 100% B over 15 min, 3 min hold at 100% B, then return
to 20% B at 18.1 min and re-equilibrate for 12 min. MS source and data
acquisition parameters were as follows: drying gas temperature 350 °C,
drying gas flow 10 L/min, nebulizer pressure 20 psi, capillary voltage
3500 V, fragmentor voltage 175 V, MS scan range 50–1200 m/z, scan rate
1 spectrum/s, reference mass correction enabled using 1.25 µM HP-0921
reference compound (Agilent, Santa Clara, CA). Untargeted peak
detection and alignment was performed using XCMS^[265]71 using the
following parameters: peak finding method: centWave, maximum ppm
deviation: 30, min/max peak width: 10/60 s, signal/noise threshold: 6,
minimum m/z difference for peaks with overlapping retention time: 0.01,
alignment bandwidth 10 s m/z slice width for peak grouping 0.025;
remaining parameters were set at default values.
The resulting untargeted metabolomics data were analyzed using
Metaboanalyst 4.0 ([266]http://metaboanalyst.ca). Within Metaboanalyst,
the data were median normalized, adjusted using auto scaling, and were
then subjected to principal component analysis using default
parameters. Pathway analysis was performed using Metaboanalyst’s
functional analysis module, which is an enhanced implementation of the
original Mummichog pathway and network analysis algorithm. P-values and
t-scores of each MS peak data were calculated between the wild type and
FMO-2 OE (Supplementary Data [267]2). Mass tolerance was set to 10
parts per million (ppm) and mummichog algorithm p-value cutoff was set
to 0.05. Default parameters were used for other settings and the
analysis was done using the C. elegans pathway library. The mummichog
algorithm produces putative compound annotation for interpretation at
the pathway level. For individual features of interest, compound
identities assigned by the algorithm were validated by targeted
analysis using authentic standards as described below.
Targeted metabolomics analysis used the same LC-MS parameters as
untargeted, but data analysis was performed using Agilent MassHunter
Quantitative Analysis software. Metabolite identification was performed
by matching accurate mass and retention time with authentic standards
analyzed in-house on the same method. Statistical analysis for targeted
metabolomics data was done using Microsoft Excel (Microsoft 365)
following median normalization and log transformation.
Stress resistance assay
Paraquat (Methyl viologen dichloride hydrate, 856177, Sigma-Aldrich)
was used to induce oxidative stress. Worms were synchronized from eggs
on RNAi plates seeded with E. coli HT115 strain expressing dsRNAi for a
particular gene and at L4 stage 40 worms were transferred on RNAi-FUdR
(40690016, Bioworld) plates containing 5 mM paraquat. A minimum of two
plates per strain per condition were used per replicate experiment. As
described previously, worms were then scored every day and considered
dead when they did not move in response to prodding under a dissection
microscope^[268]4. Worms that crawled off the plate were not
considered, but ruptured worms were noted and considered.
Lifespans
Gravid adults were placed on NGM plates containing 1 mM
β-D-isothiogalactopyranoside (367931, Fisher Scientific), 25 μg/ml
carbenicillin (BP26485, Fisher Scientific), and the corresponding RNAi
clone from the Vidal or Ahringer RNAi library. After 3 h, the adults
were removed, and the eggs were allowed to develop at 20 °C until they
reached late L4/young adult stage. From here, 40 to 90 worms were
placed on each RNAi plate and transferred to fresh RNAi + FUdR plates
on day 1, day 2, day 4, and day 6 of adulthood. A minimum of two plates
per strain per condition were used per replicate experiment.
Experimental animals were scored every 2–3 days and considered dead
when they did not move in response to prodding under a dissection
microscope. Worms that crawled off the plate were not considered, but
ruptured worms were considered. A similar method was used for formate
(F0507, Sigma-Aldrich) and s-adenosylmethionine (PureBulk, Roseburg,
OR, USA) supplementation lifespan experiments, except either 1 mM
formate or 2 mM s-adenosylmethionine was added to the NGM plates
without IPTG. A similar method was also used for the non-FUdR lifespan
experiments, except FUdR was not added to the plates and worms were fed
E. coli OP50.
Cox regression methods
Supplementary Data [269]4 and [270]5 provide the results of Cox
proportional hazards regression models, which were run in Stata 14. The
model includes a categorical variable for strain, using Wild Type (N2)
as the base category, and including dummy variables for FMO-2 OE and/or
FMO-2 KO. It also includes a dummy variable for the individual RNAi
versus empty vector (EV) control. Variables of particular interest for
this paper are the interactions between the RNAi dummy and the fmo-2
mutant dummies, which capture the differential effect and interaction
of various RNAi on fmo-2 mutants versus control worms. To account for
multiple testing, only interactions with a p-value <0.01 were
considered to be significant.
Computational modeling
The computer model was generated by building a stoichiometric matrix S
(10 reactants by 13 reactions), accounting for all reactions shown in
Fig. [271]4a. A steady-state approximation was used, as shown in
Eq. [272]1. In Eq. [273]1, S is the stoichiometric matrix and J is a
vector of fluxes for each of the reactions.
[MATH: S⋅J=0 :MATH]
1
To obtain a biologically relevant solution, we projected the expression
data of genes involved in the reactions used in the model to the
nullspace of S by solving for Eq. [274]2. Single genes were used as
representative genes for each reaction to simplify the model. Gene
expressions related to input fluxes were assumed to be one for all
strains. Reactions used in the model and the relevant gene expression
data are shown in Supplementary Tables [275]6 and [276]7. In
Eq. [277]2, M is the nullspace of S, b is the vector of relative gene
expression data from the wild type, FMO-2 OE or FMO-2 KO that have been
normalized to the wild type, and x is a vector such that Sx is the
projection of b onto the column space of M, which gives us the vector
of reaction fluxes, J, within the nullspace of S. To account for data
variability, expression level with greater than 0.5x or less than 1.5x
fold changes were assumed to be equal to the wild type control.
Equation [278]2 was solved using the lsqminnorm function in MATLAB
2018a. The lsqminnorm function returns the minimum norm least-squares
solution to Mx = b by minimizing both the norm of M * x – b and the
norm of x.
[MATH: M⋅x=b :MATH]
2
The inner product of the resulting vector x and the nullspace matrix M
was obtained to calculate the reaction flux predictions resulting from
the gene expression projection as shown in Eq. [279]3. The calculated J
for FMO-2 OE and FMO-2 KO were normalized to that of the wild type to
obtain the relative fluxes.
[MATH: M⋅x=J :MATH]
3
Quantitative PCR
RNA was isolated from day 1 adult worms following three rounds of
freeze-thaw in liquid nitrogen using Invitrogen’s Trizol extraction
method and 1 µg of RNA was reverse transcribed to cDNA using
SuperScript™ II Reverse Transcriptase (18064071, Invitrogen,). Gene
expression levels were measured using 1 μg of cDNA and SYBR^TM Green
PCR Mastermix (A25742, Applied Biosystems) and primers at 10 μM
concentration. mRNA levels were normalized using previously published
housekeeping gene controls, tba-1 and pmp-3^[280]72. For the RNAi
validation, same protocol was used for RNA isolation and cDNA
preparation, and Y45F10D.4 was used as a reference gene^[281]73. List
of primers used are in Supplementary Table [282]12.
Enzyme kinetic assays
Oxygenation activity of FMO-2 was characterized using the method
previously described^[283]74. Briefly, oxygenation of substrates was
determined by spectrophotometrically following the consumption of NADPH
at 340 nm using the molar extinction coefficient 6.22 mM^−1 cm^−1.
Components of the assay buffer included 25 mM sodium phosphate buffer
(pH 8.5), 0.5 mM diethylenetriaminepentaacetic acid (DETAPAC), 0.5 mM
NADPH, and 0.04 µM FMO-2 with excess FAD. The final substrate
concentrations for tryptophan were 100, 250, 500, 750 µM and 1, 2.5, 5,
7.5, and 10 mM. The final substrate concentrations for MMI were 100,
300, and 600 µM and 1, 3, 5, 7, 10, and 30 mM. To determine the rate of
oxidation of NADPH by FMO, NADPH concentrations of 10, 30, 100, 300,
500, and 700 µM and 1 and 1.5 mM were used. Experiments were conducted
at 30 °C while shaking. Kinetic parameters (i.e., k[cat] and K[m]) were
determined by fitting plots of the rate of turnover vs the substrate
concentration to the Michaelis-Menten equation using GraphPad Prism
(version 9.1.0; GraphPad Software Inc., San Diego, CA). Purified FMO-2
protein was purchased from GenScript. NADPH (10107824001), FAD (F6625),
MMI, L-tryptophan (T0254), and all other substrates were purchased from
Sigma-Aldrich (St. Louis, MO). DETAPAC (AC114322500) and sodium
phosphate buffer (S374-500) were purchased from Fisher (Waltham, MA).
In vitro studies LC-MS
Analysis of samples from in vitro studies with purified FMO2 was
performed using LC-MS with untargeted feature detection. Samples
contained 100, 250, or 500 µM tryptophan in the same conditions as the
enzymatic assays with FMO-2 protein. 100 µL of conditioned media were
vortexed with 400 µL of 1:1:1 methanol:acetonitrile:acetone to
precipitate protein. The extract was centrifuged for 10 min at
16,000 × g and 200 µL of supernatant were transferred to a clean
autosampler vial with insert and dried under a stream of nitrogen gas.
The dried extract was reconstituted in 50 µL of 85/15
acetonitrile/water and analyzed by HILIC-TOF-MS on an Agilent 1290
Infinity II/Agilent 6545 QTOF. Chromatography was performed on a Waters
BEH Amide column (2.1 mm ID × 10 cm, 1.7 µm particle diameter) with
mobile phase prepared as described previously^[284]75 except that
mobile phase A contained 5% acetonitrile. Briefly, the flow rate was
0.3 mL/min, the column temperature 55 °C, and the gradient was as
follows: 0–0.70 min 100%B, 0.7–6.7 min 100–85%B, 6.7–8.7 min 85%B,
8.7–16 min 85–28%B, 16–16.7 min 28%B, 16.7–16.8, 28–0%B. Total run time
was 22 min. Ion polarity was positive, gas temp was 320 °C, drying gas
was 8 L/min, nebulizer was 35 psi, sheath gas temp and flow were 350 °C
and 11 L/min, capillary voltage 3500 V. The instrument was operated in
full scan mode at 2 spectra/s and a mass range of 50–1200 Da. Feature
detection and alignment were performed using XCMS (3.16). Potential
reaction products were detected by computationally examining the data
for features present in each sample set. Identification of potential
reaction products was performed using MS/MS data acquired from a pooled
sample.
Fmo-2 induction experiment
Synchronized L4 fmo-2p::mCherry transcriptional reporter strain animals
were incubated overnight in S-media enriched with OP50 with and without
1 mM tryptophan in a 96-well plate. Some worms were also incubated
overnight in S-media without OP50 as a positive control to induce fmo-2
expression under dietary restriction.
Microscopy
Images were acquired with the LASx software and Leica scope at ×6.3
magnification with more than 15 worms per treatment. Prior to imaging,
worms were anesthetized with 0.5 M sodium azide (NaN3) on a 2% agarose
pad. Fluorescence mean comparisons were calculated using ImageJ
software.
Statistical analyses
Log-rank test was used to derive p-value for lifespan and paraquat
survival assays using p < 0.05 cut-off threshold compared to EV
controls via OASIS 2^[285]76. Unpaired Student’s t-test was used to
derive p-values for targeted metabolomics data using p < 0.05 cut-off
threshold compared to the wild type. Using one-way ANOVA for trend
analysis, p-values were calculated for fmo-2 expression levels that
affect OCM metabolite levels. Unpaired Student’s t-test was used to
derive p-values for comparing the metabolomics data of HepG2 pDEST
control and FMO2, FMO4, and FMO5 OE cell lines using p < 0.05 cut-off
threshold.
Reporting summary
Further information on research design is available in the [286]Nature
Portfolio Reporting Summary linked to this article.
Supplementary information
[287]Supplementary Information^ (1.2MB, pdf)
[288]41467_2023_36181_MOESM2_ESM.pdf^ (3.7KB, pdf)
Description of Additional Supplementary Files
[289]Supplementary Data 1^ (81KB, xls)
[290]Supplementary Data 2^ (81KB, xls)
[291]Supplementary Data 3^ (17KB, xlsx)
[292]Supplementary Data 4^ (20.1KB, xlsx)
[293]Supplementary Data 5^ (28KB, xlsx)
[294]Supplementary Data 6^ (14KB, docx)
[295]Supplementary Data 7^ (865.1KB, xlsx)
[296]Reporting Summary^ (70.3KB, pdf)
Acknowledgements