of 16
nature medicine
https://doi.org/10.1038/s41591-023-02217-7
Artic�e
Gut microbial metabolism of 5-ASA
diminishes its clinical efficacy in
inflammatory bowel disease
In the format provided by the
authors and unedited
Supplementary
Table 1. Characteristics of 5
-
ASA users vs non
-
users in the I
BDMDB.
IBD users of 5
-
ASA were more likely to have ulcerative colitis. Overall, rates of hospitalization and history of bowel surgery
were also lower for 5
-
ASA users compared to non
-
users.
Non
-
users
Users
(n=34)
(n=45)
p
-
value
Age at
consent, mean (SD)
27.7 (17.4)
26.7 (17.1)
0.81
Age at diagnosis, mean (SD)
22.1 (14.2)
20.9 (11.0)
0.68
Male (%)
17 (50)
21 (47)
0.95
UC (%)
6 (18)
24 (53)
0.003
Race/ethnicity (%)
White
30 (88)
40 (89)
0.13
Other
4 (12)
5 (11)
Prior
bowel surgery (%)
0.11
Yes
7 (21)
4 (9)
Antibiotic use (%)
14 (41)
25 (56)
0.30
Bonded 5
-
ASA use (%)
a
0 (0)
4 (9)
0.20
Hospitalized during the study (%)
13 (38)
7 (15)
0.04
Dysbiosis
b
21 (62)
19 (42)
0.14
a
Bonded 5
-
ASA formulations include sulfasalazine, balsalazide, and olsalazine
b
As previously defined in the HMP2, Methods
22
Supplementary
Table 2. Significantly differentially abundant metabolomic features pre
-
and post
-
5
-
ASA
administration among new
-
users (two
-
sided paired Wilcoxon, FDR < 0.25).
24.7% (n=569) were
completely absent or at the lowest levels of detection prior to 5
-
ASA administration. 28.9% (n=666) of these
features had a fold change of 1000x or more, and 8.0% (n
=184) were highly abundant after starting 5
-
ASA
(median peak intensity > 10
6
; by comparison, median peak intensity of N
-
acetyl 5
-
ASA
--
a known inactive
microbial metabolite of 5
-
ASA
--
was 1.66x10
8
). Only 17 (0.56%) of these were annotated.
See
attached
tables
due to size constraints.
Supplementary
Table
3
. 5
-
ASA use is associated wit
h large
-
scale differences across individual gut
microbial species abundance in patients with IBD.
Significant associations (n=24, FDR q < 0.25) of 5
-
ASA use and microbial species
.
All models included each participant’s identifier as random effects and
simu
ltaneously adjusted for age, disease type, use of antibiotic, and dysbiosis status.
Species
Beta
P
FDR
Gemmiger
formicilis
-
0.00518
0.000119
0.007644138
Bifidobacterium
longum
0.024185
0.000221
0.009487238
Dielma
fastidiosa
0.005456
0.000372
0.011991481
Intestinimonas
butyriciproducens
-
0.0118
0.000763
0.015567799
Lachnospira
pectinoschiza
-
0.02995
0.000915
0.015567799
Faecalibacterium
prausnitzii
0.077469
0.000987
0.015567799
Fusicatenibacter
saccharivorans
-
0.02626
0.001075
0.015567799
Phascolarctobacterium
faecium
-
0.01404
0.001086
0.015567799
Lactobacillus
rogosae
-
0.00461
0.002316
0.029877308
Eubacterium
eligens
0.032305
0.003037
0.035614424
Butyricimonas
synergistica
-
0.00486
0.003329
0.035790148
Ruminococcus
bromii
-
0.02084
0.003658
0.036294803
Paraprevotella
xylaniphila
-
0.00815
0.004008
0.03693114
Turicimonas
muris
-
0.0055
0.006253
0.053774509
Anaerostipes
hadrus
0.021856
0.011292
0.091043147
Ruthenibacterium
lactatiformans
-
0.01123
0.013849
0.105092694
Parasutterella
excrementihominis
-
0.0171
0.015598
0.11178451
Anaerotignum
lactatifermentans
-
0.00775
0.018159
0.123290076
Dialister
invisus
-
0.01012
0.024624
0.158826559
Anaerotruncus
colihominis
-
0.00414
0.028712
0.173321541
Bilophila
wadsworthia
-
0.00699
0.030248
0.173321541
Akkermansia
muciniphila
-
0.0283
0.03201
0.173321541
Roseburia
inulinivorans
-
0.0179
0.032246
0.173321541
Proteobacteria
bacterium
CAG_139
-
0.01021
0.045073
0.232574975
Supplementary
Table
4
. M/z ratios and retention times for 5
-
ASA in the IBDMDB and PRISM cohorts.
Identity of N
-
acetyl 5
-
ASA in PRISM and for N
-
propionyl 5
-
ASA and N
-
butyryl 5
-
ASA in the IBDMDB and
PRISM cohorts were inferred using mass and retention time matchi
ng coupled with known mass shifts from
functional groups. As further confirmation of these two possible 5
-
ASA derivatives, we searched publicly
available MS
2
data from the Global Natural Product Social (GNPS) database (MassIVE MSV000084556)
to generate mol
ecular co
-
occurrence networks, inputting the known spectra of
N
-
acetyl 5
-
ASA
. In this
analysis, we similarly found two unlabeled nodes linked with
N
-
acetyl 5
-
ASA which matched the
m/z
of the
compounds we identified, as well as a third node with an
m/z
of 238.11 which may correspond to yet another
derivative of 5
-
ASA that has not been described, likely a 5
-
carbon acyl group plus 5
-
ASA given the
additional mass shift (
+
12) from butyryl CoA.
PRISM
IBDMDB
Compound
M/Z
Retention time
M/Z
Retention time
5
-
ASA
154.0493
3.98
154.0502
3.83
N
-
butyryl 5
-
ASA
224.0909
2.32
224.0922
2.24
N
-
acetyl 5
-
ASA
196.0599
2.97
196.0609
2.81
N
-
propionyl 5
-
ASA
210.0754
2.32
210.0764
2.52
Supplementary
Table
5
. Input sequences for
homology
-
based search for 5
-
ASA degrading gut
microbial enzymes
.
There were no hits at the metatranscriptomic level in the IBDMDB.
See
attached
tables
due to size constraints.
Supplementary
Table
6
. Gene context for candidate gene
clusters associated with 5
-
ASA
metabolism.
There were no obvious upstream or downstream promoters/regulators.
See
attached
tables
due to size constraints.
Supplementary
Table
7
. Data processing refinement statistics and quality of the
Fc
THL model.
Protein Characteristics
Wavelength
0.9792
Resolution range
73.46
-
1.89 (1.958
-
1.89)
Space group
P 1
Unit cell
75.23 81.46 160.19 95.61 100.45 108.27
Total reflections
500387 (50866)
Unique reflections
263887 (26397)
Multiplicity
1.9 (1.9)
Completeness (%)
94.02 (93.71)
Mean
I/sigma(I)
7.10 (0.81)
Wilson B
-
factor
39.07
R
-
merge
0.0697 (1.012)
R
-
meas
0.09856 (1.432)
R
-
pim
0.0697 (1.012)
CC1/2
0.99 (0.391)
CC*
0.997 (0.75)
Reflections used in
refinement
263708 (26331)
Reflections used for
R
-
free
13239 (1334)
R
-
work
0.1758 (0.3703)
R
-
free
0.2087 (0.3946)
CC(work)
0.964 (0.695)
CC(free)
0.952 (0.649)
Number of non
-
hydrogen atoms
25714
macromolecules
24521
ligands
0
solvent
1193
Protein residues
3364
RMS(bonds)
0.013
RMS(angles)
1.18
Ramachandran
favored (%)
96.87
Ramachandran
allowed (%)
3.07
Ramachandran
outliers (%)
0.06
Rotamer outliers
(%)
1.00
Clashscore
4.15
Average B
-
factor
50.50
macromolecules
50.58
solvent
48.76
Number of TLS
groups
34
Statistics for the
highest
-
resolution shell are shown in parentheses.
Supplementary
Table
8
. Presence of 5
-
ASA metabolizing acetyltransferases are associated with
steroid use in the HMP2.
Odds ratios with 95% CI are presented. Analyses were adjusted for age, sex,
disease type (CD vs. UC), host acetylation phenotype, and smoking status.
Feature
Genes
Absent
Present
C7H1G6
1 (referent)
2.81 (1.68
-
4.68)
R5CY66
1 (referent)
2.88 (1.66
-
5.00)
R6CZ24
1 (referent)
2.58 (1.40
-
4.77)
T5S060
1 (referent)
3.24 (1.63
-
6.42)
Host factors
Disease type
UC
CD
1 (referent)
3.11 (1.85
-
5.21)
Age, y
0.92 (0.89
-
0.94)
Supplementary
Table
9
. Characteristics of the study population selected from the SPARC IBD
cohort
. Compared to the IBDMDB, given that SPARC IBD is a cohort of adult patients, participants were
older. They were also more likely to have UC and were less likely to be male. The prevalence of
acetyltransferases that were linked with treatment failure was
similar in both cohorts.
Users
(n=208)
Age at consent, mean (SD)
45.5 (15.1)
Male (%)
88 (42.3%)
IBD subtype
UC (%)
140 (67.3)
CD (%)
67 (32.2)
IBD
-
U
a
(%)
1 (0.4)
Prevalence of acetyltransferases, %
C7H1G6
44.0
R5CY66
20.4
R6CZ24
7.2
T5S060
18.8
a
IBD
-
undifferentiated
Supplementary
Table
10
. Presence of 5
-
ASA inactivating acetyltransferases.
Gene prevalence (%)
varies across participants with IBD, but does not vary significantly by 5
-
ASA status
(GLMM p
>
0.07)
, arguing
for effect modification of the effects of 5
-
ASA on prevention of disease relapse.
A
mong participants without
IBD
prevalence rates of these enzymes
also appear similar.
5
-
ASA users
5
-
ASA nonusers
Non
-
IBD
UniRef90 ID
C7H1G6
47.8
52.3
67.0
R5CY66
17.4
20.3
43.7
R6CZ24
20.7
14.3
2
8.5
T5S060
14.1
9.2
12.3
Supplementary
Table 1
1
.
Amino acid s
equences used for gene synthesis.
See
attached
tables
due to size constraints.
Supplementary
Table 1
2
.
Gene s
equences used for heterologous expression.
See
attached
tables
due to size constraints.
Supplementary
Figure
1
. 5
-
ASA associated metabolites predictably rise and fall with initiation and
discontinuation of the medication.
Time series for a 10
-
year
-
old boy with newly diagnosed Crohn’s colitis
in the IBDMDB who had a stuttering disease course throughout the year
-
lon
g study. Initially started on
Pentasa
an
ethylcellulose
-
coated, controlled
-
release
of oral 5
-
ASA
he experienced a flare and was
switched to prednisone. Ultimately, he resumed 5
-
ASA
--
Lialda, a multimatrix formulation of 5
-
ASA
but
flared again and wa
s escalated to infliximab and methotrexate. With each of these starts and stops, 5
-
ASA
-
associated metabolomic features predictably sunk and surged, even with administration of two different
formulations. Abbreviations: HBI, Harvey
-
Bradshaw Index (metric of
disease activity).
Supplementary
Figure
2
. Fecal 5
-
ASA levels are concordant with self
-
reported 5
-
ASA use in
PRISM.
As in HMP2, there was high accuracy between self
-
reported use of 5
-
ASA with detection of fecal
5
-
ASA. We used the PRISM cohort, an inde
pendent cohort of patients with IBD with untargeted fecal
metabolomics, to validate our annotation of 5
-
ASA derivatives beyond N
-
acetyl 5
-
ASA.
Boxplots show
median and lower/upper quartiles; whiskers show inner fences.
N=120 indep
endent biological samples
.
A
A
u
s
e
A
A
m