PostScript

Table 1 Continued
Before PS adjustment After PS adjustment Before PS adjustment After PS adjustment

PPI Non-­PPI PPI (n=11 Non-­PPI PPI (n=10 H2RA PPI H2RA

Gut: first published as 10.1136/gutjnl-2021-325209 on 8 July 2021. Downloaded from http://gut.bmj.com/ on June 23, 2024 by guest. Protected by copyright.
Characteristic (%) (n=1 01 438) (n=3 53 266) SMD 741) (n=11 741) SMD 708) (n=1 19 493) SMD (n=5067) (n=5067) SMD

Values are presented as proportion of the patients (%).

*The applicated raw database was the National Health Insurance Service Common Data Model.

GORD, gastro-­oesophageal reflux disease; H2RA, histamine 2 receptor antagonist; PPI, proton pump inhibitor; PS, propensity matching; SMD, standard mean difference.

last day of the eradication treatment as (n=19) and RA group (n=76), using
the index date. While the authors have a three-­ pronged association frame-
mentioned in the discussion section To cite Patel J, Berezowski I, Janapala RN, et al. Gut work (figure 1, online supplemental
that they could not determine H. Pylori 2022;71:1039–1041. material).3 Metagenomic genes were
infection status by a laboratory test, it Received 10 June 2021 collapsed into metagenomic species
is important to know if both groups Accepted 2 July 2021 (MGS)3 4 and grouped into KEGG func-
were matched regarding clarithromycin Published Online First 14 July 2021 tional modules (figure 1A).3 Addition-
based triple therapy or bismuth-­b ased ally, the co-­a bundant metabolites were
Gut 2022;71:1039–1041. doi:10.1136/
quadruple therapy, as they might differ gutjnl-2021-325385 categorised into metabolite clusters
in efficacy considering 20% resistance using WGCNA framework (figure 1A). 3
rate to clarithromycin in Korea.2 The ORCID iDs The functional modules associated
Jigar Patel http://orcid.org/0000-0002-2379-9869
authors also mentioned a cumulative with clinical phenotypes (eg, types of
Ivan Berezowski http://orcid.org/0000-0003-4406-​
dose–response relationship between 7517 arthritis and levels of cytokines) were
increasing PPI duration and the risk of Rajesh Naidu Janapala http://orcid.org/0000-0001-​ further identified and the cross-­domain
gastric cancer development. Though it 9660-9512 associations between these modules
is true for PPI≥30 and≥180 days, the and metabolite clusters were assessed
same is not true for PPI≥365 days, as REFERENCES (figure 1B).3 Furthermore, the leave-­
the result was not statistically signifi- 1 Seo SI, Park CH, You SC, et al. Association between one-­ out analysis was performed to
cant (PPI≥365 days, HR 3.5, 95% CI proton pump inhibitor use and gastric cancer: a
determine the MGS that particularly
population-­b ased cohort study using two different
0.85 to 23.49, p value 0.14). types of nationwide databases in Korea. Gut contributed to the observed linkage
Summarily, the gastric cancer incidence 2021;70:2066–75. between functional modules and clin-
rate is higher in Korean men than women 2 Lee JH, Ahn JY, Choi KD, et al. Nationwide ical phenotypes (figure 1C).3
when we compare ≥40 years of age antibiotic resistance mapping of Helicobacter We found that gut microbial functionality
pylori in Korea: a prospective multicenter study.
groups.3 Hence, it is important to know in ascorbate degradation (KEGG module:
Helicobacter 2019;24:e12592.
if gender was equally distributed for each 3 Song M, Kang D, Yang JJ, et al. Age and sex M00550) was positively correlated
age range in table 1. Our concern is if the interactions in gastric cancer incidence and with the types of arthritis (healthy=0,
PPI group had more men in ≥40 years age mortality trends in Korea. Gastric Cancer OA=1, RA=2, pWilcox=2.15×10–4) and
group, it can skew the result in the favour 2015;18:580–9. the levels of proinflammatory cyto-
of PPI. kines TNF-α (tumour necrosis factor-α,
Jigar Patel ‍ ‍,1 Ivan Berezowski ‍ ‍,1
Hidden link in gut–joint axis: pWilcox=6.59×10–4) and IL-­ 6 (interleu-
Rajesh Naidu Janapala ‍ ‍,1 Ali Pourmand1 gut microbes promote kin-­6, pWilcox=1.12×10–3). Ascorbate
(vitamin C) was previously reported to
The George Washington University School of Medicine
and Health Sciences, Washington, District of Columbia,
rheumatoid arthritis at early prevent the development of inflammatory
USA stage by enhancing arthritis,5 possibly through facilitating
collagen synthesis, moderating autoim-
Correspondence to Dr Jigar Patel, The George ascorbate degradation mune responses and ameliorating inflam-
Washington University School of Medicine and Health
Sciences, Washington, DC 20037, USA; mation.6 Additionally, the patients with RA
​jigar.​r.p​ atel@​outlook.​com With great interest, we read the review are usually ascorbate deficient and require
Acknowledgements All named authors meet the article by Agus et al, which suggested high-­dose supplementation to maintain an
International Committee of Medical Journal Editors that gut microbiome alterations could acceptable plasma level of ascorbate.7 In
criteria for authorship for this article, take responsibility
affect metabolic homeostasis.1 More- this study, the functional module of ascor-
for the integrity of the work as a whole, and have given
their approval for this version to be published. over, gut microbiome alterations in bate degradation was observed to posi-
concert with metabolites perturbation tively correlate with the blood metabolite
Contributors All authors have made substantial
contributions to drafting the manuscript, revising it could contribute to the early develop- cluster MB02 (pWilcox=6.90×10–3), which
critically for important intellectual content and final ment of rheumatoid arthritis (RA). 2 We was represented by the level of palmitic
approval of the version to be submitted. thus conducted a three-­ pronged asso- acid (kME (eigengene-­based connectivity)
Funding The authors have not declared a specific ciation study 3 on multiomics datasets =0.911, kIN (intramodular connectivity)
grant for this research from any funding agency in the to detect the potential microbiome– =3.46, online supplemental table 1) that
public, commercial or not-­for-­profit sectors. metabolites–arthritis link. acts as a proinflammatory factor, upregu-
Competing interests None declared. We integrated multiomics datasets lating IL-­6 secretion by human chondro-
Patient consent for publication Not required. including gut metagenomics, clinical cytes and fibroblast-­like synovial cells in
Provenance and peer review Not commissioned; phenotypes and metabolites of blood inflammatory arthritis.8 Furthermore, we
internally peer reviewed. and knee-­ j oint synovial fluid from found that Escherichia coli and Strepto-
© Author(s) (or their employer(s)) 2022. No commercial 122 participants in the healthy group coccus bovis were the driving species for
re-­use. See rights and permissions. Published by BMJ. (n=27), osteoarthritis (OA) group the observed linkage between ascorbate
Gut May 2022 Vol 71 No 5 1041
 PostScript

A Metabolites Gut microbiome degradation9 and the arthritis types or

Blood metabolites (277) microbial genes (4M) the cytokines levels of TNF-α and IL-­ 6
SF metabolites (248)
Gene groups
(figure 1C). Subsequently, we grouped
KEGG orthology MGS framework patients with RA by four stages according

Gut: first published as 10.1136/gutjnl-2021-325209 on 8 July 2021. Downloaded from http://gut.bmj.com/ on June 23, 2024 by guest. Protected by copyright.
WGCNA
to the comprehensive scores in rheuma-
Blood metabolites SF metabolites Microbiome functional modules MGSs (553) toid diagnostic criteria,10 as RASI: 6–7,
clusters (43) clusters (31) KEGG (404)
RASII: 8, RASIII: 9 and RASIV: 10 (online
Phenotypes Phenotypes Phenotypes
Arthritis types TNF-α, TNF-αadj Arthritis types
supplemental table 2). We observed that
Arthritis typesadj IL-6, IL-6adj Arthritis typesadj both E. coli and S. bovis were prevalent
(27) (29) (12) (10) Phenotype association filtering (FDR < 0.1)
at RA stage I (RASI), while S. bovis was
B depleted after RASI or in the OA group.
Positive association
Phynotype

RA and OA metabotype
Metabolite clusters

Healthy metabotype
Phynotype Metabolite clusters
It suggested S. bovis mainly functioned at
the early stage of RA, while E. coli might
Arthritis types(Healthy, OA, RA) (Gender, Age adjusted)

Positive association
Negative association RA metabotype OA metabotype
Negative association
No significant association
No significant association

be crucial throughout the entire devel-

MB32: N2,N2-Dimethylguanosine

MB30: 4-Guanidinobutanoic acid

Arthritis types (Healthy, OA, RA)

MB29: Glycochenodeoxycholate
MB40: Hydroxyphenyllactic acid

MB31: 4-Methoxycinnamic acid

MB06: DL-Indole-3-lactic acid

MB22: Mannitol-1-phosphate

−0.4 0 0.4

opmental stages of RA and OA. Taken

MB11: 1-Methylnicotinamide
MB28: 7,8-Dihydrobiopterin
MB19: L-Pyroglutamic acid

−0.3 0 0.3

MSF29: 5-Methyl-5,6-Dihydrouracil

MSF16: DL-α-Hydroxybutyric acid

MB36: 2-Deoxy-D-ribose

MB32: N2,N2-Dimethylguanosine
SCCbg.adj.
MB42: Alpha-D-Glucose

MB05: 1-Methylhistidine
MB08: 2-Oxoadipic acid

MB29: Glycochenodeoxycholate

MB31: 4_Methoxycinnamic acid

MB04: Dihydrothymine

SCCbg.adj.

MSF25: p-Chlorophenylalanine

MSF23: 4-Imidazoleacetic acid

MB03: Pseudouridine

MB11: B-1-Methylnicotinamide
MSF04: Formylanthranilic acid
TNF-α (Gender, Age adjusted)
MB26: Arachidic acid
MB12: L-Tryptophan

MB06: DL-Indole-3-lactic.acid
MB02: Palmitic acid
MB17: L-Threonine

MB16: L-Isoleucine

MSF20: L-Pyroglutamic acid

IL-6 (Gender, Age adjusted)
MB21: Thymidine

MB18: D-Lyxose

MB24: L-Proline

MSF08: Phenyllactic acid

MB36: 2-Deoxy-D-ribose
MB34: Ser-Asp
MB33: Glu-Pro
MB38: Gly-Glu

together, we speculate that E. coli and S.

MB04: Dihydrothymine
MSF02: Hypoxanthine

MB35: Salicyluric acid

MSF09: L-Threonate
MSF11: Linoleic acid

MB12: L-Tryptophan
MB17: L-Threonine

MB21: Thymidine

MSF19: Cytosine
MSF30: Creatine
MSF22: Pro-Phe

MSF03: Pro-Asp
MSF17: Lys-Lys
MSF06: Pro-Tyr
MB38: Gly-Glu

MB37: Thr-Val
M00076:Dermatan sulfate degradation ++ * * + * +++++ + +

bovis could facilitate ascorbate degrada-

TNF-α
Glycosaminoglycan metabolism

IL-6
Glycan metabolism M00078:Heparan sulfate degradation * + * * +++ * * * + * * + * + * + +

tion and thus promote proinflammatory

M00357:Methanogenesis ** + * ** ** ** ** * + * ** * + + * * + * * *
Lipopolysaccharide metabolism M00866:KDO2−lipid A biosynthesis + * ** * * * * * * * + ** * ** * + + ** * ** + + * * * *

Cofactor and vitamin metabolism M00123:Biotin biosynthesis + * +++++++++ * ++++ ++ + Methane metabolism M00567:Methanogenesis ** * * ** ** ** ** ** * * + ** ** ** ** + ** ** ** ** ** +
Gut microbial functions

responses that facilitate the development

Histidine metabolism M00045:Histidine degradation * * * * + * + * * * + * * * ** ** + * *
M00356:Methanogenesis ** ** ** ** ** * + + * + + + + + * +
Gut microbial functions

ATP synthesis M00144:NADH:quinone oxidoreductase * * * ** + + + ** + + + * * ** * ++ ** * * ++ *

M00563:Methanogenesis ** ** ** ** * * + + * + + + + * + + +
Energy metabolism

of inflammatory arthritis.
M00568:Catechol ortho−cleavage + * * + + + * + + + + * + + +
Carbon fixation M00620:Incomplete reductive citrate cycle ** + + + + + + + * ++

Other carbohydrate metabolism M00741:Propanoyl−CoA metabolism * * * + * * ++ * + * ++ + * * * + + + M00551:Benzoate degradation + + + + + + + + + + * + + * + * + + + +

Aromatics degradation
+ ** + ** + ** + + + + * + +

Overall, we demonstrate that gut

Carbohydrate metabolism Central carbohydrate metabolism M00001:Glycolysis ++ +++++ * + + * M00545:Trans−cinnamate degradation

Other carbohydrate metabolism M00550:Ascorbate degradation ++ * * * + * * + * * + * ++ + M00878:Phenylacetate degradation * * + * * + * + + * + **

Arginine and proline metabolism M00879:Arginine succinyltransferase pathway + * * * * ++ * + * + + + + + + + + + + + + + +

microbiota could promote RA progres-

Other carbohydrate metabolism M00550:Ascorbate degradation *

Aromatics degradation M00545:Trans−cinnamate degradation + ** * ** * * * ** * ** + * + * * + * * ** ** + Cofactor and vitamin metabolism * * + + + + * + + + + + + *

M00577:Biotin biosynthesis

sion via enhancing ascorbate degrada-

C Correlation of M00550 with arthritis types Correlation of M00550 with TNF-α
tion and provide a potential approach
KOs not in module KOs in module KOs not in module KOs in module
1.00 1.00 to prevent the development of arthritis
Scaled density
Scaled density

0.75 0.75 Unclassified species

through interfering gut–joint axis. The
0.50
Escherichia coli 0.50
Streptococcus bovis
Escherichia coli results of this study could be prospected
in following contexts: First, our study
MGS00002
MGS00278

MGS00548
MGS00014

MGS00182
MGS00548

0.25 0.25

0.00 0.00
provides a reservoir of the potential
0.10 0.15 0.20 0.25 0.30
microbiome–metabolites–arthritis links
0.2 0.3 0.4 0.5
Spearman correlation coefficients Spearman correlation coefficients as a reference of gut–joint axis for future
studies. Second, the findings supplement
Correlation of M00550 with IL-6
MGS00014 MGS00002 MGS00278 KOs not in module KOs in module the potential mechanisms related to meta-
250 50 1.00
bolic perturbation through which gut
Scaled density

50

200 40 30 0.75 Unclassified species microbiome promotes arthritis.1 2 Third,

Escherichia coli
considering the inflammatory pathways
150 30
20 0.50 Streptococcus bovis
100 20
of arthritis were revisited in COVID-­19,11
MGS00002
MGS00014
MGS00278

10 0.25
50 10

0
H OA S1 S2 S3 S4
0
H OA S1 S2 S3 S4
0
H OA S1 S2 S3 S4
0.00 it deserves further investigations whether
0.05 0.10 0.15 0.20 0.25 microbiome–ascorbate–inflammation link
Spearman correlation coefficients of this study could contribute to the treat-
ment of COVID-­19.
Figure 1 Overview of the three-­pronged association framework integrating multiomics datasets.
(A) Metabolites are summarised as co-­abundance clusters, and microbial genes are grouped Yan Zhao ‍ ‍,1 Mingyue Cheng ‍ ‍,2
into KEGG modules and MGS, which are further filtered for statistically positive or negative Liang Zou ‍ ‍,1 Luxu Yin ‍ ‍,1 Chaofang Zhong,2
associations (based on Spearman correlation) with the clinical phenotypes. The association Yuguo Zha ‍ ‍,2 Xue Zhu ‍ ‍,2 Lei Zhang ‍ ‍,3
analyses were divided by using healthy, OA and RA samples for arthritis types and using OA and Kang Ning ‍ ‍,2 Jinxiang Han ‍ ‍1
1
RA samples for cytokine levels. The number in brackets represent the number of metabolites/ First Affiliated Hospital of Shandong First Medical
metabolite clusters/microbial genes/KEGG modules/MGS in each analytical module. (B) The University, Biomedical Sciences College & Shandong
Medicinal Biotechnology Centre, Shandong First
filtered features are further used for cross-­domain association analyses. For each analysis, the Medical University & Shandong Academy of Medical
left panel shows the significant associations (Mann-­Whitney U test FDR<0.1) between KEGG Sciences, Jinan, China
2
modules and clinical phenotypes, and colour indicates significantly positive association (red), Key Laboratory of Molecular Biophysics of the Ministry
significantly negative association (blue) or insignificant association (grey). The right panel shows of Education, Hubei Key Laboratory of Bioinformatics
and Molecular-­imaging, Center for Artificial Intelligence
the associations between KEGG modules and metabolite clusters, and the colour represents the Biology, Department of Bioinformatics and Systems
median Spearman correlation coefficient (SCC) of metabolite clusters with KEGG orthologies Biology, College of Life Science and Technology,
(KOs) in KEGG module minus those with KOs not in KEGG module. Mann-­Whitney U test FDRs Huazhong University of Science and Technology,
are denoted: +FDR<0.1; *FDR<0.01; **FDR<0.001. (C) The MGS that particularly contributed to Wuhan, China
3
the observed linkage between functional modules and clinical phenotypes. Three density plots: Microbiome-­X, National Institute of Health Data
Science of China & Institute for Medical Dataology,
Dashed line represents the median SCC of the phenotypes with KOs in M00550 (red) and all other Department of Biostatistics, School of Public Health,
KOs (blue). Density plot shows the median SCC of the phenotypes with KOs in M00550, when a Cheeloo College of Medicine, Shandong University,
given MGS (indicated by short vertical lines) has been excluded from the analysis. The bottom-­ Jinan, China
left dot plots show the mean±SEM of the top three driving MGS abundances among patients at Correspondence to Professor Lei Zhang, Shandong
each stage of disease development, with the four RA stages connected to display the variance. University, Jinan, China; ​zhanglei7@​sdu.​edu.​cn,
FDR, false discovery rate; IL-­6, interleukin-­6; MGS, metagenomic species; OA, osteoarthritis; RA, Professor Kang Ning, Huazhong University of Science
rheumatoid arthritis; TNF-α, tumour necrosis factor-α. and Technology, Wuhan, China;
​ningkang@​hust.​edu.​cn and Professor Jinxiang Han,

1042 Gut May 2022 Vol 71 No 5


Shandong First Medical University & Shandong
Academy of Medical Sciences, Jinan, China; j​ xhan@​
sdfmu.e​ du.​cn
Correction notice This article has been corrected
since it published Online First. The figure has been
replaced for clarity.
Contributors YZ and MC designed the study,
conducted the data analysis and wrote the manuscript.
YZ, MC, LZ, LY, CZ, YZ and XZ collected the samples,
conducted the experiments and participated in data
analysis. LZ, KN and JH supervised the study and
revised the manuscript.
Funding This study was funded by National
Natural Science Foundation of China (grant numbers
32071465, 82003766, 31871334, and 31671374),
Academic Promotion Project of Shandong First
Medical University (grant number 2019LJ001) and
Ministry of Science and Technology’s National Key
Research and Development Programme (grant number
2018YFC0910502).
Competing interests None declared.
Patient consent for publication Not required.
Ethics approval The study was approved by the
Ethics Committee of the First Affiliated Hospital of
Shandong First Medical University (No. 2017-­02).
Provenance and peer review Not commissioned;
externally peer reviewed.
Supplementary material This content has been
supplied by the author(s). It has not been vetted by
BMJ Publishing Group Limited (BMJ) and may not have
been peer-­reviewed. Any opinions or recommendations
discussed are solely those of the author(s) and are
not endorsed by BMJ. BMJ disclaims all liability and
responsibility arising from any reliance placed on the
content. Where the content includes any translated
material, BMJ does not warrant the accuracy and
reliability of the translations (including but not limited
to local regulations, clinical guidelines, terminology,
drug names and drug dosages), and is not responsible
for any error and/or omissions arising from translation
and adaptation or otherwise.
Open access This is an open access article distributed
in accordance with the Creative Commons Attribution
Non Commercial (CC BY-­NC 4.0) license, which permits
others to distribute, remix, adapt, build upon this work
non-­commercially, and license their derivative works on
different terms, provided the original work is properly
cited, appropriate credit is given, any changes made
indicated, and the use is non-­commercial. See: http://​
creativecommons.org/licenses/by-nc/4.0/.
© Author(s) (or their employer(s)) 2022. Re-­use
permitted under CC BY-­NC. No commercial re-­use. See
rights and permissions. Published by BMJ.
►► Additional supplemental material is published
online only. To view, please visit the journal online
(http://​dx.​doi.​org/​10.​1136/​gutjnl-​2021-​325209).
YZ and MC contributed equally.
To cite Zhao Y, Cheng M, Zou L, et al. Gut
2022;71:1041–1043.
Received 19 May 2021
Accepted 3 July 2021
Published Online First 8 July 2021
Gut May 2022 Vol 71 No 5
Gut 2022;71:1041–1043. doi:10.1136/
gutjnl-2021-325209
ORCID iDs
Yan Zhao http://orcid.org/0000-0002-2816-8951
Mingyue Cheng http://orcid.org/0000-0003-1243-5039
Liang Zou http://orcid.org/0000-0003-0899-2318
Luxu Yin http://orcid.org/0000-0003-2374-9965
Yuguo Zha http://orcid.org/0000-0003-3702-9416
Xue Zhu http://orcid.org/0000-0002-7337-6186
Lei Zhang http://orcid.org/0000-0003-4284-2604
Kang Ning http://orcid.org/0000-0003-3325-5387
Jinxiang Han http://orcid.org/0000-0001-9269-4553
References
1 Agus A, Clément K, Sokol H. Gut microbiota-­derived
metabolites as central regulators in metabolic
disorders. Gut 2021;70:1174–82.
2 Zaiss MM, Joyce Wu H-­J, Mauro D, et al. The gut-j­ oint
axis in rheumatoid arthritis. Nat Rev Rheumatol
2021;17:224–37.
3 Pedersen HK, Forslund SK, Gudmundsdottir V, et al.
A computational framework to integrate high-­
throughput ’-omics’ datasets for the identification
of potential mechanistic links. Nat Protoc
2018;13:2781–800.
4 Nielsen HB, Almeida M, Juncker AS, et al. Identification
and assembly of genomes and genetic elements
in complex metagenomic samples without using
reference genomes. Nat Biotechnol 2014;32:822–8.
5 Pattison DJ, Silman AJ, Goodson NJ, et al. Vitamin C
and the risk of developing inflammatory polyarthritis:
prospective nested case-­control study. Ann Rheum Dis
2004;63:843–7.
6 Carr AC, Maggini S. Vitamin C and immune function.
Nutrients 2017;9:1211.
7 Abrams E, Sandson J. Effect of ascorbic acid
on rheumatoid synovial fluid. Ann Rheum Dis
1964;23:295–9.
8 Frommer KW, Schäffler A, Rehart S, et al. Free fatty
acids: potential proinflammatory mediators in
rheumatic diseases. Ann Rheum Dis 2015;74:303–10.
9 Yew WS, Gerlt JA. Utilization of L-­ascorbate by
Escherichia coli K-­12: assignments of functions to
products of the yjf-­sga and yia-­sgb operons. J Bacteriol
2002;184:302–6.
10 Aletaha D, Neogi T, Silman AJ. Rheumatoid
arthritis classification criteria: an American College
of Rheumatology/European League against
rheumatism collaborative initiative. Arthritis Rheum
2010;2010:2569–81.
11 Schett G, Manger B, Simon D, et al. COVID-­19
revisiting inflammatory pathways of arthritis. Nat Rev
Rheumatol 2020;16:465–70.
OLGIM staging and proton
pump inhibitor use predict the
risk of gastric cancer
We read with great interest the recent
publication written by Lee et al1 in which
the authors conducted a prospective,
longitudinal and multicentre study to
evaluate the association between intes-
tinal metaplasia (IM) and gastric cancer
(GC). As expected, IM was found to be
a significant risk factor for GC (adjusted
HR (aHR) 5.46; 95% CI 1.51 to 19.0).
Nevertheless, operative link on gastric
IM (OLGIM) staging was more closely
correlated with the risk of GC (stage III/
PostScript
IV; aHR 20.7; 95% CI 5.04 to 85.6, and
stage II; aHR 7.34; 95% CI 1.60 to 33.7).
They concluded that risk stratification of
GC based on OLGIM staging was useful
Gut: first published as 10.1136/gutjnl-2021-325209 on 8 July 2021. Downloaded from http://gut.bmj.com/ on June 23, 2024 by guest. Protected by copyright.
for endoscopic surveillance.
We agree with their results that OLGIM
staging is the most important predictor
for GC, although there are many other
risk factors of GC. According to our find-
ings and those of others, long-­term use of
proton pump inhibitors (PPIs) is associ-
ated with an increased risk of GC, partic-
ularly in patients with advanced IM.2
Therefore, we hypothesise that combining
the information of PPI use with OLGIM
staging may improve the accuracy of GC
prediction.2–4
Using our endoscopic database, we esti-
mated the effect of multiple parameters,
such as OLGIM staging and PPI use, on the
prediction of GC. The OLGIM stage was
determined in line with the previous study;5
furthermore, we evaluated the IM distribu-
tion in the antrum and corpus from single
biopsy specimens from each site.6
In total, 1551 patients with gastric
mucosa biopsy (male, 51.3%; mean
age, 69.7 years) were selected from
the database of University of Tokyo
Hospital from 1998 to 2017. During a
mean follow-­up period of 6.3 years, 95
patients developed GC. Patient char-
acteristics are shown in table 1. We
developed two predictive models for
predicting posteradicated GC, with the
variables of age, sex and the following
factors: (1) OLGIM stage and (2)
OLGIM stage and PPI use. Furthermore,
we calculated the cumulative area under
the curve (AUC) during all observational
time points (table 2).
PPI use (aHR 3.09; 95% CI 2.02 to
4.74), IM distribution (antral IM (aHR
2.12; 95% CI 1.27 to 3.51), corpus IM
(aHR 3.98; 95% CI 2.40 to 6.58) and
OLGIM stage (stage I (aHR 2.43; 95%
CI 1.47 to 4.03), stage II (aHR 2.70;
95% CI 1.52 to 4.80), stage III (aHR
4.28; 95% CI 2.08 to 8.82) and stage
IV (aHR 7.17; 95% CI 2.41 to 21.32))
were associated with an increased risk
of GC, while aHR is greatest in patients
with advanced OLGIM staging.
The cumulative AUC of the model
with PPI use (0.74, 95% CI 0.66 to
0.82) was higher than that of the model
without PPI use (0.71, 95% CI 0.64 to
0.79).
Our findings suggest that OLGIM
staging, even if only scored by one biopsy
specimen from the antrum and corpus,
could accurately predict the risk of GC.
Nevertheless, information regarding PPI
use significantly improves the accuracy
1043
 BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Gut

1 Supplementary information
2

3 A hidden link in gut-joint axis: Gut microbes promote rheumatoid

4 arthritis at early stage by enhancing ascorbate degradation
5
6 Yan Zhao,1# Mingyue Cheng,2# Liang Zou,1 Luxu Yin,1 Chaofang Zhong,2 Yugo

7 Zha,2 Xue Zhu,2 Lei Zhang,3,4* Kang Ning,2* Jinxiang Han1*

8
1
9 Shandong Medicine and Health Key Laboratory of Rheumatism, Shandong Key

10 Laboratory of Rheumatic Disease and Translational Medicine, Department of

11 Rheumatology and Autoimmunology, Shandong Provincial Qianfoshan Hospital,

12 First Affiliated Hospital of Shandong First Medical University. Jinan 250014,

13 Shandong, China
2
14 Key Laboratory of Molecular Biophysics of the Ministry of Education, Hubei Key

15 Laboratory of Bioinformatics and Molecular-imaging, Center of AI Biology,

16 Department of Bioinformatics and Systems Biology, College of Life Science and

17 Technology, Huazhong University of Science and Technology, Wuhan 430074,

18 Hubei, China
3
19 Microbiome-X, National Institute of Health Data Science of China & Institute for

20 Medical Dataology, Cheeloo College of Medicine, Shandong University, Jinan

21 250014, Shandong, China

4
22 Department of Biostatistics, School of Public Health, Cheeloo College of Medicine,

23 Shandong University, Jinan 250014, Shandong, China

24

25 *Correspondence to Dr Jinxiang Han, E-mail: jxhan@sdfmu.edu.cn; Dr Kang Ning,

26 E-mail: ningkang@hust.edu.cn; Dr Lei Zhang, E-mail: zhanglei7@sdu.edu.cn.

27
#
28 These authors contributed equally to this work

Zhao Y, et al. Gut 2021;0:1–2. doi: 10.1136/gutjnl-2021-325209

 BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Gut

29 Sample description

30 A total of 122 fecal and 122 serum samples were collected from 122 outpatients from

31 the Shandong Provincial Qianfoshan Hospital (Jinan, Shandong, China). These

32 outpatients included 27 healthy individuals, 19 patients with osteoarthritis (OA), and

33 76 patients with rheumatoid arthritis (RA). Subsequently, the fecal samples were

34 sequenced and the serum samples were used to examine serum metabolites and

35 inflammatory cytokines. Serum inflammatory cytokines TNF-α and IL-6 were

36 quantified by the MESO SCALE DISCOVERY (MSD®) Quick Plex S600MM

37 multiplex assay. The cytokine levels of healthy individuals were extremely low and not

38 available. In addition, 95 knee-joint synovial fluid samples were collected from the RA

39 and OA patients to examine synovial fluid metabolites. Both serum and synovial fluid

40 metabolites were examined by UHPLC-MS/MS.

41 All of the participants were at fasting status during sample collection in the morning.

42 The participants were recruited in this study following the standards shown below:

43 1. Healthy individuals in good health condition with no gastrointestinal diseases, such

44 as diarrhea, constipation, and hematochezia, in the recent one month, no

45 hepatobiliary system diseases, no history of gastrointestinal tumors or inflammatory

46 diseases, no serious heart, liver, kidney, lung, brain or other organ disorders, no

47 infections, chronic diseases, or antibiotic treatment;

48 2. Healthy individuals had not taken any acid inhibitors, gastrointestinal motility drugs,

49 antibiotics, or living bacteria products such as yogurt in the recent one month;

50 3. Healthy individuals with no history or family history of mental illness, and no

51 history of gastrointestinal surgery;

52 4. RA/OA individuals with no other co-morbidity.

53 Metagenome sequencing and data processing

54 Whole-genome shot-gun sequencing of fecal samples were carried out on the Illumina

55 Hiseq X Ten. All samples were paired-end sequenced with a 150-bp read length. After

Zhao Y, et al. Gut 2021;0:1–2. doi: 10.1136/gutjnl-2021-325209

 BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Gut

56 quality control, the paired-end reads were assembled into contigs using MEGAHIT

57 (version 1.2.6)1 with the minimum contig length set at 500 bp. The open reading frames

58 (ORFs) were predicted from the assembled contigs using Prodigal (version 2.6.3)2 with

59 default parameters. The ORFs of <100 bp were removed. The ORFs were then clustered

60 to remove redundancy using Cd-hit (version 4.6.6)3 with a sequence identity threshold

61 set at 0.95 and the alignment coverage set at 0.9, which resulted in a catalog of

62 4,047,645 non-redundant genes. The non-redundant genes were then collapsed into

63 metagenomic species (MGS)4 5 and grouped into KEGG functional modules.4

64 Identification of MGS

65 High-quality reads were mapped to the catalog of non-redundant genes using Bowtie 2

66 (version 2.2.9)6 with default parameters. The abundance profile for each catalogue gene

67 was calculated as the sum of uniquely mapped sequence reads, using 19M sequence

68 reads per sample (downsized). The co-abundance clustering of the 4,047,645 genes was

69 performed using canopy algorithm (http://git.dworzynski.eu/mgs-canopy-algorithm),5

70 and 553 gene clusters that met the previously described criteria5 and contained more

71 than 700 genes were referred to as MGS. MGS present in at least 4 samples were used

72 for the following analysis. The abundance profiles of MGS were determined as the

73 medium gene abundance throughout the samples. MGS were taxonomically annotated

74 as described by Nielsen et al.5 and each MGS gene was annotated by sequence

75 similarity to NCBI bacterial genome (BLASTN, E-value < 0.001)

76 Annotation of KEGG modules

77 The catalog of the non-redundant genes was functionally annotated to KEGG database

78 (release 94.0) by KofamKOALA (version 1.3.0).7 8 The produced KEGG Orthologies

79 (KOs) were mapped to the KEGG modules annotation downloaded on August 1, 2020

80 from the KEGG BRITE database. KOs present in at least 4 samples were used for the

81 following analysis. The KO abundance profile was calculated by summing the

Zhao Y, et al. Gut 2021;0:1–2. doi: 10.1136/gutjnl-2021-325209

 BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Gut

82 abundances of genes that were annotated to each KO.

83 Clustering of co-abundant metabolites

84 Co-abundant metabolites in serum or synovial fluid were identified using the R package

85 WGCNA9. As recommended by Pedersen et al.,4 a signed network and biweighted mid-

86 correlation were used for clustering with the soft threshold β = 8 for both serum and

87 synovial fluid metabolites. The minimum cluster size was set as 3. Similar clusters were

88 subsequently merged if the biweight mid-correlation between the cluster’s eigen

89 vectors exceeded 0.8 for both serum and synovial fluid metabolites. The kIN of a

90 metabolite was calculated by summing connectivity with all other metabolites in the

91 given metabolite cluster. The kME was determined by the bicor-correlation between

92 the metabolite profile and module eigenvector. Both kIN and kME were used to

93 measure the intramodular hub-metabolite status.

94 Cross-domain association analyses

95 The clinical phenotypes, including types of arthritis (Healthy = 0, OA = 1, RA = 2) and

96 the levels of pro-inflammatory cytokines TNF-α and IL-6, were used in the association

97 analysis. TNF-α and IL-6 were selected based on their potentials to act as the

98 therapeutic targets for RA treatment.10 11 The associations between clinical phenotypes

99 and KEGG modules/metabolites clusters were determined through evaluating if the

100 Spearman correlations of the phenotype with the abundances of KOs/metabolites in the

101 given KEGG module/metabolite clusters were significantly higher or lower (Mann–

102 Whitney U-test FDR < 0.1) than with the abundances of all other KOs/metabolites. The

103 phenotypes adjusted by age and gender were also tested. Moreover, the union set of the

104 significant associations between KEGG modules and phenotypes/phenotypes adjusted

105 by age and gender, and the intersect set of the significant associations between

106 metabolites clusters and phenotypes/phenotypes adjusted by age and gender, were used

107 for the following association analysis. The associations between metabolite clusters and

Zhao Y, et al. Gut 2021;0:1–2. doi: 10.1136/gutjnl-2021-325209

 BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Gut

108 KEGG modules were determined through evaluating if the Spearman correlations of

109 the eigen vectors of the metabolite clusters with the abundances of KOs in the given

110 KEGG module were significantly higher or lower (Mann–Whitney U-test FDR < 0.1)

111 than with the abundances of all other KOs/metabolites.

112 Leave-one-out analysis

113 Leave-one-out analysis was used to identify the specific MGS driving the observed

114 associations between KEGG module M00550 and the clinical phenotypes, including

115 the types of arthritis or the levels of pro-inflammatory cytokines TNF-α and IL-6. The

116 calculation of the KO abundance was iterated excluding the genes from a different MGS,

117 in each iteration. The effect of a given MGS on a specified association was defined as

118 the change in median Spearman correlation coefficient between KOs and clinical

119 phenotypes when genes from the respective MGS were left out, as previously

120 described.4 12

121 Taxonomic identity of differentially present microbes across conditions

122 MetaPhlAn213 was used to generate species profiles. Species that were present in less

123 than 10% samples were excluded. Supplementary Figure 1 displays the union set of the

124 species (n=15) with significantly different abundances (Mann–Whitney U-test FDR <

125 0.05) between the healthy and RA groups or between the healthy and OA groups.

126

Zhao Y, et al. Gut 2021;0:1–2. doi: 10.1136/gutjnl-2021-325209

 BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Gut

127

128 Supplementary figure 1 Taxonomic identity of differentially present microbes across

129 conditions. Each row represents a species with significantly different abundances

130 (Mann–Whitney U-test FDR < 0.05) between the healthy and RA groups or between

131 the healthy and OA groups. Each column represents a sample from one of the groups

132 including the healthy, RAP1, RAP2, RAP3, RAP4, and OA groups. Color of each

133 heatmap unit represents the scaled abundance of a certain species in a specific sample.

134 Species are colored for significantly elevation (red) or depletion (green) in the arthritis

135 groups, in comparison with the healthy groups.

136 Data accession

137 Whole-genome shot-gun sequencing data are available in the Genome Sequence

138 Archive (GSA) section of National Genomics Data Center (project accession number

139 CRA004348) at https://bigd.big.ac.cn/gsa/browse/CRA004348.

Zhao Y, et al. Gut 2021;0:1–2. doi: 10.1136/gutjnl-2021-325209

 BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Gut

140 References
141 1. Li D, Luo R, Liu CM, et al. MEGAHIT v1.0: A fast and scalable metagenome
142 assembler driven by advanced methodologies and community practices.
143 Methods 2016;102:3–11.
144 2. Hyatt D, Chen GL, Locascio PF, et al. Prodigal: prokaryotic gene recognition and
145 translation initiation site identification. BMC Bioinformatics 2010;11:119.
146 3. Li W, Godzik A. Cd-hit: a fast program for clustering and comparing large sets of
147 protein or nucleotide sequences. Bioinformatics 2006;22:1658–9.
148 4. Pedersen HK, Forslund SK, Gudmundsdottir V, et al. A computational framework to
149 integrate high-throughput '-omics' datasets for the identification of potential
150 mechanistic links. Nat Protoc 2018;13:2781–800.
151 5. Nielsen HB, Almeida M, Juncker AS, et al. Identification and assembly of genomes
152 and genetic elements in complex metagenomic samples without using reference
153 genomes. Nat Biotechnol 2014;32:822–8.
154 6. Langmead B, Salzberg SL. Fast gapped-read alignment with Bowtie 2. Nat Methods
155 2012;9(4):357–9.
156 7. Kanehisa M, Goto S, Sato Y, et al. Data, information, knowledge and principle: back
157 to metabolism in KEGG. Nucleic Acids Res 2014;42:D199–205.
158 8. Aramaki T, Blanc-Mathieu R, Endo H, et al. KofamKOALA: KEGG Ortholog
159 assignment based on profile HMM and adaptive score threshold. Bioinformatics
160 2020;36(7):2251–52.
161 9. Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network
162 analysis. BMC Bioinformatics 2008;9:559.
163 10. Ceccarelli F, Massafra U, Perricone C, et al. Anti-TNF treatment response in
164 rheumatoid arthritis patients with moderate disease activity: a prospective
165 observational multicentre study (MODERATE). Clin Exp Rheumatol
166 2017;35(1):24–32.
167 11. Nakahara H, Nishimoto N. Anti-interleukin-6 receptor antibody therapy in
168 rheumatic diseases. Endocr Metab Immune Disord Drug Targets
169 2006;6(4):373–81.
170 12. Pedersen HK, Gudmundsdottir V, Nielsen HB, et al. Human gut microbes impact
171 host serum metabolome and insulin sensitivity. Nature 2016;535(7612):376–81.
172 13. Truong DT, Franzosa EA, Tickle TL, et al. MetaPhlAn2 for enhanced metagenomic
173 taxonomic profiling. Nat Methods 2015;12:902–3.

Zhao Y, et al. Gut 2021;0:1–2. doi: 10.1136/gutjnl-2021-325209

 BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Gut

1 Supplementary information
2

3 A hidden link in gut-joint axis: Gut microbes promote rheumatoid

4 arthritis at early stage by enhancing ascorbate degradation
5
6 Yan Zhao,1# Mingyue Cheng,2# Liang Zou,1 Luxu Yin,1 Chaofang Zhong,2 Yugo

7 Zha,2 Xue Zhu,2 Lei Zhang,3,4* Kang Ning,2* Jinxiang Han1*

8
1
9 Shandong Medicine and Health Key Laboratory of Rheumatism, Shandong Key

10 Laboratory of Rheumatic Disease and Translational Medicine, Department of

11 Rheumatology and Autoimmunology, Shandong Provincial Qianfoshan Hospital,

12 First Affiliated Hospital of Shandong First Medical University. Jinan 250014,

13 Shandong, China
2
14 Key Laboratory of Molecular Biophysics of the Ministry of Education, Hubei Key

15 Laboratory of Bioinformatics and Molecular-imaging, Center of AI Biology,

16 Department of Bioinformatics and Systems Biology, College of Life Science and

17 Technology, Huazhong University of Science and Technology, Wuhan 430074,

18 Hubei, China
3
19 Microbiome-X, National Institute of Health Data Science of China & Institute for

20 Medical Dataology, Cheeloo College of Medicine, Shandong University, Jinan

21 250014, Shandong, China

4
22 Department of Biostatistics, School of Public Health, Cheeloo College of Medicine,

23 Shandong University, Jinan 250014, Shandong, China

24

25 *Correspondence to Dr Jinxiang Han, E-mail: jxhan@sdfmu.edu.cn; Dr Kang Ning,

26 E-mail: ningkang@hust.edu.cn; Dr Lei Zhang, E-mail: zhanglei7@sdu.edu.cn.

27
#
28 These authors contributed equally to this work

Zhao Y, et al. Gut 2021;0:1–2. doi: 10.1136/gutjnl-2021-325209

 BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Gut

29 Sample description

30 A total of 122 fecal and 122 serum samples were collected from 122 outpatients from

31 the Shandong Provincial Qianfoshan Hospital (Jinan, Shandong, China). These

32 outpatients included 27 healthy individuals, 19 patients with osteoarthritis (OA), and

33 76 patients with rheumatoid arthritis (RA). Subsequently, the fecal samples were

34 sequenced and the serum samples were used to examine serum metabolites and

35 inflammatory cytokines. Serum inflammatory cytokines TNF-α and IL-6 were

36 quantified by the MESO SCALE DISCOVERY (MSD®) Quick Plex S600MM

37 multiplex assay. The cytokine levels of healthy individuals were extremely low and not

38 available. In addition, 95 knee-joint synovial fluid samples were collected from the RA

39 and OA patients to examine synovial fluid metabolites. Both serum and synovial fluid

40 metabolites were examined by UHPLC-MS/MS.

41 All of the participants were at fasting status during sample collection in the morning.

42 The participants were recruited in this study following the standards shown below:

43 1. Healthy individuals in good health condition with no gastrointestinal diseases, such

44 as diarrhea, constipation, and hematochezia, in the recent one month, no

45 hepatobiliary system diseases, no history of gastrointestinal tumors or inflammatory

46 diseases, no serious heart, liver, kidney, lung, brain or other organ disorders, no

47 infections, chronic diseases, or antibiotic treatment;

48 2. Healthy individuals had not taken any acid inhibitors, gastrointestinal motility drugs,

49 antibiotics, or living bacteria products such as yogurt in the recent one month;

50 3. Healthy individuals with no history or family history of mental illness, and no

51 history of gastrointestinal surgery;

52 4. RA/OA individuals with no other co-morbidity.

53 Metagenome sequencing and data processing

54 Whole-genome shot-gun sequencing of fecal samples were carried out on the Illumina

55 Hiseq X Ten. All samples were paired-end sequenced with a 150-bp read length. After

Zhao Y, et al. Gut 2021;0:1–2. doi: 10.1136/gutjnl-2021-325209

 BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Gut

56 quality control, the paired-end reads were assembled into contigs using MEGAHIT

57 (version 1.2.6)1 with the minimum contig length set at 500 bp. The open reading frames

58 (ORFs) were predicted from the assembled contigs using Prodigal (version 2.6.3)2 with

59 default parameters. The ORFs of <100 bp were removed. The ORFs were then clustered

60 to remove redundancy using Cd-hit (version 4.6.6)3 with a sequence identity threshold

61 set at 0.95 and the alignment coverage set at 0.9, which resulted in a catalog of

62 4,047,645 non-redundant genes. The non-redundant genes were then collapsed into

63 metagenomic species (MGS)4 5 and grouped into KEGG functional modules.4

64 Identification of MGS

65 High-quality reads were mapped to the catalog of non-redundant genes using Bowtie 2

66 (version 2.2.9)6 with default parameters. The abundance profile for each catalogue gene

67 was calculated as the sum of uniquely mapped sequence reads, using 19M sequence

68 reads per sample (downsized). The co-abundance clustering of the 4,047,645 genes was

69 performed using canopy algorithm (http://git.dworzynski.eu/mgs-canopy-algorithm),5

70 and 553 gene clusters that met the previously described criteria5 and contained more

71 than 700 genes were referred to as MGS. MGS present in at least 4 samples were used

72 for the following analysis. The abundance profiles of MGS were determined as the

73 medium gene abundance throughout the samples. MGS were taxonomically annotated

74 as described by Nielsen et al.5 and each MGS gene was annotated by sequence

75 similarity to NCBI bacterial genome (BLASTN, E-value < 0.001)

76 Annotation of KEGG modules

77 The catalog of the non-redundant genes was functionally annotated to KEGG database

78 (release 94.0) by KofamKOALA (version 1.3.0).7 8 The produced KEGG Orthologies

79 (KOs) were mapped to the KEGG modules annotation downloaded on August 1, 2020

80 from the KEGG BRITE database. KOs present in at least 4 samples were used for the

81 following analysis. The KO abundance profile was calculated by summing the

Zhao Y, et al. Gut 2021;0:1–2. doi: 10.1136/gutjnl-2021-325209

 BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Gut

82 abundances of genes that were annotated to each KO.

83 Clustering of co-abundant metabolites

84 Co-abundant metabolites in serum or synovial fluid were identified using the R package

85 WGCNA9. As recommended by Pedersen et al.,4 a signed network and biweighted mid-

86 correlation were used for clustering with the soft threshold β = 8 for both serum and

87 synovial fluid metabolites. The minimum cluster size was set as 3. Similar clusters were

88 subsequently merged if the biweight mid-correlation between the cluster’s eigen

89 vectors exceeded 0.8 for both serum and synovial fluid metabolites. The kIN of a

90 metabolite was calculated by summing connectivity with all other metabolites in the

91 given metabolite cluster. The kME was determined by the bicor-correlation between

92 the metabolite profile and module eigenvector. Both kIN and kME were used to

93 measure the intramodular hub-metabolite status.

94 Cross-domain association analyses

95 The clinical phenotypes, including types of arthritis (Healthy = 0, OA = 1, RA = 2) and

96 the levels of pro-inflammatory cytokines TNF-α and IL-6, were used in the association

97 analysis. TNF-α and IL-6 were selected based on their potentials to act as the

98 therapeutic targets for RA treatment.10 11 The associations between clinical phenotypes

99 and KEGG modules/metabolites clusters were determined through evaluating if the

100 Spearman correlations of the phenotype with the abundances of KOs/metabolites in the

101 given KEGG module/metabolite clusters were significantly higher or lower (Mann–

102 Whitney U-test FDR < 0.1) than with the abundances of all other KOs/metabolites. The

103 phenotypes adjusted by age and gender were also tested. Moreover, the union set of the

104 significant associations between KEGG modules and phenotypes/phenotypes adjusted

105 by age and gender, and the intersect set of the significant associations between

106 metabolites clusters and phenotypes/phenotypes adjusted by age and gender, were used

107 for the following association analysis. The associations between metabolite clusters and

Zhao Y, et al. Gut 2021;0:1–2. doi: 10.1136/gutjnl-2021-325209

 BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Gut

108 KEGG modules were determined through evaluating if the Spearman correlations of

109 the eigen vectors of the metabolite clusters with the abundances of KOs in the given

110 KEGG module were significantly higher or lower (Mann–Whitney U-test FDR < 0.1)

111 than with the abundances of all other KOs/metabolites.

112 Leave-one-out analysis

113 Leave-one-out analysis was used to identify the specific MGS driving the observed

114 associations between KEGG module M00550 and the clinical phenotypes, including

115 the types of arthritis or the levels of pro-inflammatory cytokines TNF-α and IL-6. The

116 calculation of the KO abundance was iterated excluding the genes from a different MGS,

117 in each iteration. The effect of a given MGS on a specified association was defined as

118 the change in median Spearman correlation coefficient between KOs and clinical

119 phenotypes when genes from the respective MGS were left out, as previously

120 described.4 12

121 Taxonomic identity of differentially present microbes across conditions

122 MetaPhlAn213 was used to generate species profiles. Species that were present in less

123 than 10% samples were excluded. Supplementary Figure 1 displays the union set of the

124 species (n=15) with significantly different abundances (Mann–Whitney U-test FDR <

125 0.05) between the healthy and RA groups or between the healthy and OA groups.

126

Zhao Y, et al. Gut 2021;0:1–2. doi: 10.1136/gutjnl-2021-325209

 BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Gut

127

128 Supplementary figure 1 Taxonomic identity of differentially present microbes across

129 conditions. Each row represents a species with significantly different abundances

130 (Mann–Whitney U-test FDR < 0.05) between the healthy and RA groups or between

131 the healthy and OA groups. Each column represents a sample from one of the groups

132 including the healthy, RAP1, RAP2, RAP3, RAP4, and OA groups. Color of each

133 heatmap unit represents the scaled abundance of a certain species in a specific sample.

134 Species are colored for significantly elevation (red) or depletion (green) in the arthritis

135 groups, in comparison with the healthy groups.

136 Data accession

137 Whole-genome shot-gun sequencing data are available in the Genome Sequence

138 Archive (GSA) section of National Genomics Data Center (project accession number

139 CRA004348) at https://bigd.big.ac.cn/gsa/browse/CRA004348.

Zhao Y, et al. Gut 2021;0:1–2. doi: 10.1136/gutjnl-2021-325209

 BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Gut

140 References
141 1. Li D, Luo R, Liu CM, et al. MEGAHIT v1.0: A fast and scalable metagenome
142 assembler driven by advanced methodologies and community practices.
143 Methods 2016;102:3–11.
144 2. Hyatt D, Chen GL, Locascio PF, et al. Prodigal: prokaryotic gene recognition and
145 translation initiation site identification. BMC Bioinformatics 2010;11:119.
146 3. Li W, Godzik A. Cd-hit: a fast program for clustering and comparing large sets of
147 protein or nucleotide sequences. Bioinformatics 2006;22:1658–9.
148 4. Pedersen HK, Forslund SK, Gudmundsdottir V, et al. A computational framework to
149 integrate high-throughput '-omics' datasets for the identification of potential
150 mechanistic links. Nat Protoc 2018;13:2781–800.
151 5. Nielsen HB, Almeida M, Juncker AS, et al. Identification and assembly of genomes
152 and genetic elements in complex metagenomic samples without using reference
153 genomes. Nat Biotechnol 2014;32:822–8.
154 6. Langmead B, Salzberg SL. Fast gapped-read alignment with Bowtie 2. Nat Methods
155 2012;9(4):357–9.
156 7. Kanehisa M, Goto S, Sato Y, et al. Data, information, knowledge and principle: back
157 to metabolism in KEGG. Nucleic Acids Res 2014;42:D199–205.
158 8. Aramaki T, Blanc-Mathieu R, Endo H, et al. KofamKOALA: KEGG Ortholog
159 assignment based on profile HMM and adaptive score threshold. Bioinformatics
160 2020;36(7):2251–52.
161 9. Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network
162 analysis. BMC Bioinformatics 2008;9:559.
163 10. Ceccarelli F, Massafra U, Perricone C, et al. Anti-TNF treatment response in
164 rheumatoid arthritis patients with moderate disease activity: a prospective
165 observational multicentre study (MODERATE). Clin Exp Rheumatol
166 2017;35(1):24–32.
167 11. Nakahara H, Nishimoto N. Anti-interleukin-6 receptor antibody therapy in
168 rheumatic diseases. Endocr Metab Immune Disord Drug Targets
169 2006;6(4):373–81.
170 12. Pedersen HK, Gudmundsdottir V, Nielsen HB, et al. Human gut microbes impact
171 host serum metabolome and insulin sensitivity. Nature 2016;535(7612):376–81.
172 13. Truong DT, Franzosa EA, Tickle TL, et al. MetaPhlAn2 for enhanced metagenomic
173 taxonomic profiling. Nat Methods 2015;12:902–3.

Zhao Y, et al. Gut 2021;0:1–2. doi: 10.1136/gutjnl-2021-325209