Nutrition & Diabetes www.nature.com/nutd
/\RY|£LE OPEN Check for updates
Multi-omics signatures in new-onset diabetes predict metabolic
response to dietary inulin: findings from an observational study
followed by an interventional trial
N. Ďáskova1 1 2
, I. M o d o s 2
, M. Krbcová 3
, M . K u z m a 4
, H. Pelantová4
, J . Hradecký5
, M . Heczková2
, M . Bratova2
, P. Vídeňská6
,
P. Šplíchalová7
, M . Králová8 1 2
, M. Heniková 3
, J . Potočková3
, A. Ouřadová3
, R. Landberg 9
, T. K ü h n 1 0 1 1 1 3
, M . Cahová 2 1 3 H
and
J. Gojda 3 1 3
© The Author(s) 2023
AIM: The metabolic performance of the gut microbiota contributes to the onset of type 2 diabetes. However, targeted dietary
interventions are limited by the highly variable inter-individual response. We hypothesized (1) that the composition of the complex
gut microbiome and metabolome (MIME) differ across metabolic spectra (lean-obese-diabetes); (2) that specific MIME patterns
could explain the differential responses to dietary inulin; and (3) that the response can be predicted based on baseline MIME
signature and clinical characteristics.
METHOD: Forty-nine patients with newly diagnosed pre/diabetes (DM), 66 metabolically healthy overweight/obese (OB), and 32
healthy lean (LH) volunteers were compared in a cross-sectional case-control study integrating clinical variables, dietary intake, gut
microbiome, and fecal/serum metabolomes (16S rRNA sequencing, metabolomics profiling). Subsequently, 27 D M were recruited
for a predictive study: 3 months of dietary inulin (10g/day) intervention.
RESULTS: MIME composition was different between groups. While the D M and LH groups represented opposite poles of the
abundance spectrum, OB was closer to DM. Inulin supplementation was associated with an overall improvement in glycemic
indices, though the response was very variable, with a shift in microbiome composition toward a more favorable profile and
increased serum butyric and propionic acid concentrations. The improved glycemic outcomes of inulin treatment were dependent
on better baseline glycemic status and variables related to the gut microbiota, including the abundance of certain bacterial taxa
(i.e., Blautia, Eubacterium halii group, Lachnoclostridium, Ruminiclostridium, Dialister, or Phascolarctobacterium), serum concentrations
of branched-chain amino acid derivatives and asparagine, and fecal concentrations of indole and several other volatile organic
compounds.
CONCLUSION: W e demonstrated that obesity is a stronger determinant of different MIME patterns than impaired glucose
metabolism. The large inter-individual variability in the metabolic effects of dietary inulin was explained by differences in baseline
glycemic status and MIME signatures. These could be further validated to personalize nutritional interventions in patients with
newly diagnosed diabetes.
Nutrition and Diabetes (2023)13:7; https://doi.org/10.1038/s41387-023-00235-5
Obesity and its associated metabolic diseases, including type 2
diabetes, currently represent o n e of the greatest challenges to
global health care [1]. Recently, it has been suggested that the
composition and performance of the gut microbiota contribute to
individual risks. The critical role of the gut microbiota in the
development of obesity was suggested by a seminal study by
Turnbaugh [2], followed by others confirming differences in
microbiota composition between lean and obese individuals [3,4].
Further research showed an association between the gut
microbiota and the development of type 2 diabetes [5-8], with
evidence of a specific gut microbiota signature characteristic of
prediabetes [9, 10]. However, while many studies suggest that
type 2 diabetes is associated with gut dysbiosis [11], results on the
composition and function of the microbiota are inconsistent and
sometimes contradictory. For example, a-diversity has been
reported to be significantly lower [6, 12, 13], not significantly
'First Faculty o f M e d i c i n e , Charles University, Prague, C z e c h Republic, i n s t i t u t e for Clinical a n d E x p e r i m e n t a l M e d i c i n e , Prague, C z e c h Republic, d e p a r t m e n t o f Internal M e d i c i n e ,
Kralovske V i n o h r a d y University Hospital a n d Third Faculty o f M e d i c i n e , Charles University, Prague, C z e c h Republic, i n s t i t u t e o f M i c r o b i o l o g y o f t h e CAS, Prague, C z e c h Republic.
5
F a c u l t y o f Forestry a n d W o o d Sciences, C z e c h University o f Life Sciences, Prague, C z e c h Republic. 6
M e n d e l University, D e p a r t m e n t o f C h e m i s t r y a n d Biochemistry, Brno, C z e c h
Republic. 7
R E C E T O X , Faculty o f Science Masaryk University, Brno, C z e c h Republic. 8
A m b i s University, D e p a r t m e n t o f E c o n o m i c s a n d M a n a g e m e n t , Prague, C z e c h Republic.
' D i v i s i o n o f F o o d a n d Nutrition Science, D e p a r t m e n t o f Biology a n d Biological E n g i n e e r i n g , C h a l m e r s University o f T e c h n o l o g y , G o t e b o r g , S w e d e n , ' " i n s t i t u t e o f G l o b a l F o o d
Security, Q u e e n ' s University Belfast, Belfast, UK. " H e i d e l b e r g Institute o f G l o b a l Health (HIGH), M e d i c a l Faculty a n d University Hospital, H e i d e l b e r g University, H e i d e l b e r g ,
G e r m a n y . 1 2
P r e s e n t address: Institute for Clinical a n d Experimental M e d i c i n e , Prague, C z e c h Republic. 1 3
T h e s e authors c o n t r i b u t e d equally: T. Kúhn, M . Cahová, J . Gojda.
H
e m a i l : m o n i k a . c a h o v a @ i k e m . c z
Received: 9 August 2022 Revised: 22 March 2023 Accepted: 6 A p r i l 2023
Published online: 21 April 2023
SPRINGER NATURE
N. Ďáskova et al.
reduced [14], or comparable to nondiabetic subjects in patients
with T2D [15,16]. Most studies report significant differences in the
composition of the gut microbiota between diseased a n d healthy
subjects [17], but they differ greatly with respect to specific taxa.
Some studies show that T2D is associated with an increased
Firmicutes/Bacteroidetes ratio [6, 13, 14, 18, 19], whereas others
report a significant increase [14, 18] or decrease [6, 13] in
Proteobacteria. A t the genus level, there are f e w dysregulated taxa
that have been consistently reported, i.e., an increase in
Streptococcus [9, 15, 20], Escherichia [15, 21, 22], Veillonella
[6, 21], Lactobacillus [13, 18, 23], a n d Collinsella [12, 15]; decrease
in Akkermansia [15, 18], Dialister [15, 19], Haemophilus [12, 15],
Roseburia [12,15], a n d Faecalibacterium [10,12,13], whereas many
others show changes in both directions [17]. Diet composition is a
known risk factor for the d e v e l o p m e n t of type 2 diabetes. In
addition to direct effects o n host physiology, diet plays an
important role in shaping the microbiome, thereby influencing its
metabolic program [24]. Therefore, dietary interventions focused
on modulating the composition and/or performance of the gut
microbiota appear to be a promising therapeutic target.
Supplementation with prebiotic supplements, a n d dietary fiber
in particular, is often r e c o m m e n d e d as a beneficial treatment for
non-communicable diseases, but controlled clinical trials indicate
pronounced differences in response to treatment, with considerable
personal variability [25]. The underlying causes are not yet
clear, but strong inter-individual differences in microbial response
to dietary fiber likely play a key role [26, 27]. Therefore, the
identification of the microbial taxa that mediate the beneficial
effects of dietary fiber may o p e n n e w avenues for individualized
treatment approaches [28]. In the present study, w e aimed to
determine (i) whether the composition of the complex gut
microbiome a n d metabolome (MIME) differ in lean healthy, obese
healthy, a n d obese diabetic drug-naive type 2 diabetic patients; (ii)
whether the effects of inulin o n glucose tolerance a n d insulin
sensitivity can be explained, at least in part, by the response of the
gut microbiota to inulin intervention; a n d (iii) whether this
response can be predicted from the initial MIME signature.
MATERIAL AND METHODS
The current study was performed within t h e framework o f t h e TRIEMA
project: Treatment o f Insulin Resistance b y Modification o f Gut Microbiota
(ClinicalTrials.gov Identifier: NCT03710850). The first study from the project
has been already published [24].
Study design and population
Observational study. Forty-nine newly diagnosed patients with pre/
diabetes ( D M : BMI >25, fasting glycemia > 5 . 6 m M , and/or 2hOGTT
glycemia >7.8mM), 66 metabolically healthy overweight/obese (OB: BMI
>25) a n d 32 lean healthy (LH: BMI <25) subjects were screened a n d
enrolled in t h e cross-sectional case-control study. A clinical visit was
scheduled after enrollment. Volunteers were examined after a 12-h
overnight fast; blood a n d urine samples were collected; a clinical
examination, bioimpedance analysis, a n d oral glucose tolerance test
(OGTT, 75 g glucose) were performed. A prospective 3-day dietary record
and stool samples were collected from each participant. Dietary records
and stool samples were obtained n o longer than a week after t h e
clinical visit.
Prospective study. Twenty-seven patients (DM) were then enrolled in a onearm,
non-controlled intervention study in which they were fed 10 g of inulin
daily for 3 months. The sample size determination for the intervention study
was calculated for the primary outcome, glucose disposal (GD). According t o
GD, standard deviations ranged from 1.8 t o 2.5 mg/kg/min in both insulinsensitive
and insulin-resistant individuals, with high insulin levels (i.e., 80 m i l /
m2
) showing less variability with SD u p t o 0.51 [29]. W e anticipate that
participants will respond individually t o the intervention, and w e will divide
them into tertiles (responders, neutral, and non-responders). If we consider a
difference between changes of 2 0 % (i.e., ~1.5 mg/kg/min) t o be significant t o
have 9 0 % power to detect a difference at the 0.05 alpha level, we must have
6 subjects in each group. To account for dropouts or incomplete data, we
aimed t o have at least 9 subjects in each group (i.e., responders vs. nonresponders).
Baseline and post-intervention examinations were identical t o
those described above. In addition, indirect calorimetry and a two-step
glucose clamp (10 and 8 0 m l U / m 2
BSA insulin dose) were performed [30],
Insulin sensitivity (IS) of adipose tissue was expressed as the change in nonesterified
fatty acids (NEFA) and plasma glycerol levels from baseline t o the
steady state of the first step of the clamp, whereas IS o f skeletal muscle was
expressed as space-corrected glucose infusion rate per kg fat-free mass (Mcor
mg/kg FFM/min) and metabolic clearance of glucose divided by steady-state
insulinemia (MCR/I, ml/kg FFM/min) at the steady state of the second step.
Detailed calculations are described in Supplementary Material. All participants
signed an informed consent before enrollment in each respective study. The
research protocol was approved by the Ethics Committee of University
Hospital Kralovske Vinohrady (EK-VP /26/0/2017) in accordance with the
Declaration of Helsinki. The study was registered under NCT03710850.
Gut microbiome analysis
DNA from stool samples was isolated using the QIAmp PowerFecal D N A Kit
(Qiagen, Hilden, Germany), and the V 4 region o f the bacterial 16 S rRNA
gene was amplified b y PCR. Sequencing was performed using the Miseq
reagent kit V2 with a MiSeq instrument (lllumina, Hayward, CA, USA). The
raw sequences were processed using a D A D A 2 A m p l i c o n Denoiser [31].
Short-chain fatty acids (SCFA) in plasma
SCFA were analyzed in plasma by LC-MS according t o a method described
before [32].
Volatile compounds (VOCs) analysis in feces
Volatile fingerprinting o f fecal samples was performed using an Agilent
7890B gas chromatograph (Santa Clara, California, USA) coupled t o a
Pegasus 4 D time-of-flight mass spectrometer (LECO, Geleen, The Netherlands).
Data acquisition and initial data processing were performed using
instrumental SW ChromaTOF by LECO.
NMR analyses
Serum samples (after protein precipitation) were measured o n a 600 M H z
Bruker Avance III spectrometer (Bruker BioSpin, Rheinstetten, Germany)
equipped with a 5 m m TCI cryogenic probe head. The concentrations of
individual metabolites, identified by comparison o f proton a n d carbon
chemical shift with t h e H M D B database, were expressed as P Q N [33]
normalized intensities of corresponding signals in C P M G spectra. The list of
quantified metabolites with corresponding ' H and 1 3
C chemical shifts is
given in Table S1. The representative 1
H N M R spectrum is shown in Fig. S1.
Statistics
The statistical analyses were performed using R software packages and inhouse
scripts [34]. T h e microbiome a n d VOCs data were treated as
compositional (proportions o f total read count in each sample or
proportion of the total area o f selected masses), and before all statistical
analyses, the data were transformed by centered log-ratio (clr) transformation
with a multiplicative simple replacement for handling zero values.
According t o their abundance and prevalence, the bacteria were classified
as "core microbial taxa" w h e n fulfilling t h e following conditions, i.e.
abundance o f > 0 . 1 % and prevalence o f > 7 5 % at least in o n e experimental
group. Other microbial taxa were classified as rare.
All methods are described in detail in Supplementary Material.
RESULTS
Observational study: clinical characteristics
The clinical characteristics of the study participants are shown in
Table 1. As expected, the groups differed in terms of glycemic
indices, insulin sensitivity, a n d beta cell function. Biomarkers of
lipid metabolism were significantly elevated in both the OB a n d
D M groups compared with LH.
Observational study: fecal microbiome composition
In all samples, w e f o u n d 44,332 amplicon sequence variants (ASVs)
and identified 13 phyla, 30 classes, 56 orders, 104 families, a n d 367
genera. Considering only the ASVs, all a-diversity indices were
SPRINGER NATURE Nutrition and Diabetes (2023)13:7
N. Ďáskova et al.
Table 1. G r o u p characteristics for lean (LH), o b e s e (OB) a n d persons w i t h pre/diabetes (DM).
LH DM OB K-W test DMCT
LH vs OB LH vs DM OB vs DM
General characteristics
Sex (F/M) 16/16 26/23 47/19
W e i g h t (kg) 74.8 [23.1] 99.5 [17.4] 87.2 [25.8] "0.05 "0.001 "0.001 "0.01
A g e (years) 30.9 [11.0] 58.3 [13.1] 51.3 [14.2] "0.05 "0.001 "0.001 n.s.
BMI (kg/m2
) 23.0 [4.0] 34.9 [9.1] 30.8 [6.6] "0.05 "0.001 "0.001 "0.05
W H R 0.8 [0.1] 1.0 [0.1] 0.9 [0.1] "0.05 "0.001 "0.001 "0.05
Body composition
Fat (kg) 14.2 [4.8] 39.5 [22.3] 32.9[14.7] "0.05 "0.001 "0.001 n.s.
F F M (kg) 56.5 [22.5] 61.3[14.8] 51.9[17.3] "0.05 n.s. n.s. "0.05
T B W (kg) 41.4 [16.5] 44.9[10.8] 38.0[12.6] '0.05 n.s. n.s. "0.05
Macronutrient intake
Total e n e r g y (kcal/day) 2101[1583] 2017 [879] 1777[555] n.s. N/A N/A N/A
Proteins (g/day) 81 [29] 82 [33] 72 [28.5] n.s. N/A N/A N/A
Lipids (g/day) 83 [49] 79 [40] 65 [35.5] "0.05 "0.05 n.s. n.s.
Carbohydrates (g/day) 232 [98] 207 [96] 197 [73.5] n.s. N/A N/A N/A
Dietary fiber (g/day) 18 [19] 16 [9] 15 [7.5] n.s. N/A N/A N/A
Glucose metabolism
Fasting g l u c o s e (mmol/l) 4.8 [0.3] 5.9 [0.8] 5.3 [0.6] "0.05 "0.001 "0.001 "0.001
2 h O G T T g l u c o s e (mmol/l) 5.7 [1.1] 8.9 [3.1] 6.4 [1.6] "0.05 n.s. "0.001 "0.001
A U C for O G T T g l u c o s e ( m m o l /
1 x 1 2 0 m i n " 1
)
254 [114] 499 [282] 239 [150] "0.05 n.s. "0.001 "0.001
A U C for O G T T insulin (mlU/l x 120 m i n - 1
) 3890[2707] 8948[6596] 6453[4122] "0.05 "0.01 "0.001 "0.05
Insulin (mlU/l) 4.0 [2.7] 15.9 [8.6] 9.5 [5.7] "0.05 "0.001 "0.001 "0.001
C-peptide (pmol/l) 233 [97] 769 [357] 5.3 [0.6] "0.05 "0.001 "0.001 "0.01
H b A l c (mmol/mol) 32 [2] 38 [7] 6.4 [1.6] "0.05 "0.001 "0.001 "0.001
M a t s u d a index 10.2 [6.4] 2.0 [1.7] 4.0 [3.4] "0.05 "0.01 "0.001 "0.001
Insulinogenic index 0.8 [0.7] 0.8 [1.0] 1.1 [1.0] "0.05 n.s. n.s. n.s.
Oral disposition index 6.7 [4.9] 1.9 [1.2] 4.9 [5.7] "0.001 n.s. "0.001 "0.001
Beta cell index 163 [134] 45 [25] 108 [145] "0.001 n.s. "0.001 "0.001
TyG i n d e x 0.51 [0.67] 1.54[0.59] 1.01 [0.60] "0.05 "0.001 "0.001 "0.01
Lipid metabolism
Total cholesterol (mmol/l) 4.30 [1.09] 5.01 [1.23] 5.15 [1.24] "0.05 "0.01 "0.05 n.s.
HDL-C (mmol/l) 1.67 [0.47] 1.26 [0.30] 1.39 [0.56] "0.05 "0.05 "0.001 n.s.
LDL-C (mmol/l) 2.37 [1.15] 3.05 [1.40] 3.06 [1.16] "0.05 "0.001 "0.05 n.s.
Triacylglycerols (mmol/l) 0.69 [0.52] 1.53 [0.93] 1.10 [0.71] "0.05 "0.001 "0.001 "0.05
Inflammatory markers
CRP (mg/l) 0.7 [0.9] 3.3 [4.5] 2.3 [4.0] "0.05 "0.001 "0.001 n.s.
Stool characteristics
p H in feces 7.26 [0.67] 7.04 [0.52] 7.27 [0.50] n.s. N/A N/A N/A
dry mass (%) 25.1 [8.9] 24.5 [9.9] 23.0 [6.9] n.s. N/A N/A N/A
Data were given as median [71],
AUC area under the curve during oral glucose tolerance test, BMI body mass index, CRP C-reactive protein, DMCT Dunn's multiple comparison test, FFM fat-free
mass, HDL-C high-density lipoprotein-cholesterol, HbAl glycated hemoglobin, K-W Kruskal-Wallis test, LDL-C low-density lipoprotein-cholesterol, N/A not
applicable, ns not significant, TyG index In (fasting triglyceride x fasting glucose)/2; TBW total body water, WHR waist-hip ratio. Insulinogenic index (AINS 0-30/
A G L U 0-30), ISI-M Matsuda-deFronzo index; oral disposition index (IGI*ISI); beta cell index (iAUCi n s u M n /iAUCg |u c o s e )*ISI.
*p < 0.05, **p < 0.01, ***p < 0.001.
significantly lower in O B a n d D M compared with LH, whereas no
differences were f o u n d between the D M a n d O B groups (Fig. S2).
W h e n ASVs were aggregated a n d classified at t h e genus level,
only the Shannon index remained significantly lower in O B and
D M c o m p a r e d with LH (Fig. S3).
At t h e p h y l u m level, t h e m i c r o b i o t a c o m p o s i t i o n was
d o m i n a t e d by Firmicutes a n d Bacteroidetes, f o l l o w e d by m u c h
less a b u n d a n t Actinobacteria, Proteobacteria, a n d Verrucomicrobia.
T h e m e d i a n a b u n d a n c e o f all other phyla was less than
0 . 0 1 % . There were n o significant differences in t h e
Nutrition and Diabetes (2023)13:7 SPRINGER NATURE
N. Ďáskova et al.
representation of individual phyla (Table S2). The separation of
individual samples at the genus level is visualized in Fig. 1A.
M u l t i v a r i a t e statistics revealed significant differences in
p-diversity (p< 0.001), a n d pairwise analysis confirmed significant
differences between OB vs. LH ( p <
0.001) a n d D M vs. LH
(p< 0.001), but not between D M a n d OB. Using univariable
analysis, w e identified 37 taxa that had significantly different
abundance a m o n g groups; 15 of t h e m met the criteria of "core"
microbiota, i.e., an abundance of >0.05% a n d a prevalence of
> 7 5 % in at least o n e group (Fig. 1B a n d Table S3), accounting
for 4 5 % of all core genera. Thirteen core genera were more
abundant in LH c o m p a r e d to the other t w o groups, while
E
•
G r o u p s
LH
O B
• M
B
Dim1 ( 1 0 . 8 % )
Erysipelotricha U C G - 0 0 3
L a c h n o s p i r a c e a e incertae sedis
Lachnospiraceae N D 3 0 0 7 g r o u p *
Bifidobacterium
Anaerostipes *
Fusicatenibacter
L a c h n o s p i r a c e a e _ u n a s s i g n e d (*)
[Eubacterium] hallii g r o u p *
Blautia (*)
D o r e a
Lachnospiraceae N K 4 A 1 3 6 g r o u p "
Faecalibacterium *
Christensenellaceae R - 7 g r o u p *
Pseudobutyrivibrio (*)
Lachnoclostridium (*)
0.5
Lachnospiraceae F C S 0 2 group
[Ruminococcus] gauvreauii group
Marvinbryantia *
Lachnospiraceae U C G - 0 0 8
Family XIII A D 3 0 1 1 g r o u p
Prevotella 7
Prevotellaceae
Tyzzerella 3
Catenibacterium (*)
Tyzzerella 4
Alloprevotella
Mitsuokella (*)
Megasphaera *
Fusobacterium (*)
B a c t e r o i d a l e s _ u n a s s i g n e d
Ruminococcaceae U C G - 0 0 4
Flavonifractor *
Desulfovibrio
Succinivibrio
Slackia
Megamonas
L H O B D M
Fig. 1 Fecal microbiome composition. A 2D PCA scores plot on
genera level after clr transformation. The explained variance of each
component is included in the axis labels. The large points represent
the centroids of each group. B Abundances of all significant genera
(FDR <0.1). Proportional data were used. Each cell then represents
the mean in each group for the corresponding genera. Rows were
z-scaled. Core genera are defined by the condition abundance
>0.05% and prevalence >75% at least in one group. Genera marked
by * are confirmed butyrate producers, and genera marked by (*) are
potential butyrate producers.
Pseudobutyrivibrio a n d Lachnoclostridium were enriched only in
DM. Confirmed butyrate producers, i.e., Anaerostipes, Eubacterium
halii, Faecalibacterium, Christensenellaceae R-7 group, were
more abundant in the core microbiota LH than in the core
microbiota O B or D M . Most of the taxa enriched in D M and/or
OB belong to the "non-core" taxa. A m o n g them, potentially
harmful genera were identified {Fusobacterium, Megasphera,
and Desulfovibrio). Significant positive correlations were found
between Fusobacterium abundance a n d C-peptide concentration
in all groups. The c o m m o n or unique taxa specific to the
groups are shown in Fig. S4.
The discrimination of the groups as a function of microbiome
composition was investigated using a machine learning approach
(LASSO regression model). This model, which has an accuracy of
5 1 % and a sensitivity of 6 6 % (LH), 5 0 % (OB), and 4 3 % (DM), does
not reliably classify LH, OB, and D M (Fig. S5). When w e grouped
OB and DM, the accuracy of the model increased to 7 5 % and the
sensitivity to 6 5 % (LH) (Fig. S6).
Observational study: fecal metabolome
In the fecal metabolome, w e identified 185 different VOCs. Within
this subset, 113 VOCs were of very low abundance ("0.1%), 54
VOCs each accounted for 0.1-1% of the total, 12 VOCs accounted
for 1-5% of the total, a n d six were very abundant (>5%). The
separation of individual samples is visualized in Fig. 2. Multivariable
statistics revealed significant differences in B-diversity
(p = 0.0017). The pairwise analysis confirmed significant differences
between the D M vs. LH groups (p<0.01) a n d OB vs. LH
(p < 0.05), but not between D M and OB.
Univariable analysis followed by effect size analysis revealed ten
VOCs with significantly different abundance between groups (FDR
; •
•
• •
•
i •
••
• •
T V " " " * *
•—
r
* : » V
•
•
— —
— •
•
• •
G r o u p s
LH
O B
D M
Dim1 ( 2 6 % )
B
L H O B D M
N o n a n o i c acid • i
B e nze ne a ceta Ide hyd e
2-Octanol 0.5
Propyl acetate
D e c a n e 0
Tetradecanal
t r a n s - O c i m e n e
H u m u l e n e
-0.5
H u m u l e n e
-0.5
Methyl pentanoate
• -1
Fig. 2 Fecal metabolome composition. A 2D PCA scores plot on
VOCs abundances after clr transformation. Only VOCs meeting
condition A U C x p > 0 . 1 % AUCt o tai,p are shown. The explained
variance of each component is included in the axis labels. The
large points represent the centroids of each group. B Abundances of
significant metabolites. Proportional data were used. Each cell then
represents the median in each group for the corresponding
metabolite. Rows were z-scaled.
SPRINGER NATURE Nutrition a n d Diabetes (2023)13:7
N. Ďáskova et al.
p < 0.1) (Fig. 2B and Table S4). Nonanoic acid was more abundant,
while all other compounds, including SCFA esters, were less
abundant in the OB and D M groups compared to LH. Only methyl
pentanoate showed an opposite pattern in the D M and OB groups
(DM>LH = OB) (Fig. S7). Nonanoic acid correlated positively with
the TyG index in all groups.
A LASSO model created for the classification of tested subjects
into three categories (LH vs OB vs DM) achieved only 5 2 %
accuracy and only 4 8 % (LH), 5 4 % (OB), and 5 3 % (DM) sensitivity
(Fig. S8). When we combined subjects from OB and D M into one
category, classification accuracy increased to 80.5%, but sensitivity
remained low at 5 2 % (Fig. S9).
Observational study: serum/plasma metabolome
To determine the composition of the serum metabolome, we used
an untargeted NMR approach and LC-MS analyzes that allows
accurate determination of SCFA concentration in plasma. In total,
we identified 35 quantified analytes by NMR and nine SCFAs by
LC-MS, only acetate/acetic acid was identified by both methods.
PERMANOVA analysis suggested the separation of the groups, and
subsequent pairwise tests revealed significant differences
(p< 0.001) in serum metabolome composition between all
compared pairs.
The univariable analysis identified 21 metabolites that were
significantly different in abundance between groups (Fig. 3B and
Table S5). Based o n the univariable analysis, we identified LH, OB,
and DM-specific groups of serum metabolites. For most metabolites,
the D M and LH groups represented the opposite poles of the
abundance spectra, with OB closer to the D M group. All three
groups differed in serum concentrations of intermediates of
saccharide metabolism (glucose, lactate, and mannose) and two
amino acids (AA) (glutamine, alanine). The concentration of seven
compounds, including three SCFA (propionic acid, succinic acid,
valeric acid), two A A (tyrosine, histidine), and glycerol was
comparable at OB and D M , but differed from LH. Six compounds,
including two branched-chain amino acid (BCAA) derivatives (2oxoisovalerate,
3-methyl-2-oxovalerate), 2-hydroxybutyrate, acetone,
2-propanol, and formic acid, presented a specific DMassociated
signature (Fig. S10).
A LASSO -model based on serum metabolome data was able to
classify unknown subjects into the categories LH, OB, or D M with
an accuracy of 7 4 % and a sensitivity of 9 0 % (LH), 7 2 % (OB), or 6 5 %
(DM) (Fig. S11). When we grouped subjects from OB and D M
groups together, model accuracy increased to 8 9 % and sensitivity
(LH) increased to 8 8 % (Fig. S12). None of the models selected
glucose as a key discriminant.
Observational study: integrative analysis
We further investigated whether a combination of all variables would
allow better classification between groups. With this integrated
LASSO model, an unknown subject could be assigned to one of the
three groups (LH, OB, and DM) with an accuracy of 7 7 % and a
sensitivity of 8 8 % (LH), 7 9 % (OB), and 6 6 % (DM), respectively. LASSO
coefficients included five variables from the microbiome dataset, one
variable from the fecal metabolome dataset, and nine variables from
the serum metabolome dataset (Fig. S13). When we constructed the
LASSO model only for two groups (LH vs. OB + DM), we were able to
classify an unknown subject with 9 1 % accuracy and 8 9 % sensitivity.
Ten microbes, five fecal VOCs, and 11 serum metabolites contributed
to the discrimination between groups (Fig. S14).
Finally, w e looked for a possible complex interaction between
different MIME components in individual groups. Figure 4 depicts
the positive and negative Spearman correlations among datasets
filtered by |p| > 0.5; these correlations unravel differences in
interaction networks within each group. In the LH group, we
observed a rich network among variables both within and outside
the datasets, whereas the complexity in OB and D M was much
lower.
Groups
Dim1 (13.6%)
B
Saccharide
metabolism
Amino
acids
i
0.5
0
- 0 . 5
- 1
LH
O B
D M
BCAA metabolism
intermediates
NAD+ regeneration
Lipid metabolism
G l u c o s e
M a n n o s e
Lactate
Pyruvate
Glutamine
Asparagine
Histidine
Glycine
Alanine
Tyrosine
3-Methyl-2-oxovalerate
2-Oxoisovalerate
Valeric acid
Formate
Propionic acid
Succinic acid
Acetone
| 2-Propanol
| 2-Hydroxybutyrate
Glycerol
I Ethanol
LH O B D M
Fig. 3 Serum metabolome composition. A 2D PCA scores plot. The
explained variance of each component is included in the axis labels.
The large points represent the centroids of each group.
B Abundances of significant metabolites. Each cell then represents
the median in each group for the corresponding metabolite. Rows
were z-scaled.
Prospective study: effect of inulin on omics signature
Twenty-seven newly diagnosed D M subjects participated in a threemonth,
single-arm, non-controlled intervention study in which they
were administered inulin (10g/day) without other antidiabetic
medications and/or lifestyle interventions. No clinically significant
adverse events occurred, and all subjects completed the study. The
inulin intervention was associated with a significant change in
microbiota composition (PERMANOVA p<
0.001) and a significant
decrease in a-diversity (Fig. 5A, B). At the phylum level, the
abundance of Bacteroidetes and Proteobacteria significantly
decreased, whereas the proportion of Actinobacteria and Verrucomicrobia
significantly increased (Table S6). Univariable analysis
revealed 28 taxa with significantly different abundance before and
after inulin treatment (Fig. 5C and Table S7). The abundance of 16
bacterial taxa (genera or higher taxonomic units), including
confirmed butyrate producers such as Faecalibacterium, Anaerostipes,
and Eubacterium halii or bacteria considered beneficial such as
Lactobacillus, Bifidobacterium, and Akkermansia, increased after
treatment. The abundance of 12 taxa, including Alistipes, Odoribacter,
or Bacteroides, decreased.
In serum and feces, inulin intake was not associated with a shift
in total metabolome composition, but using univariable analysis,
we identified several metabolites that were significantly different
Nutrition and Diabetes (2023)13:7 SPRINGER NATURE
N. Ďáskova et al.
A Positive correlations B Negative correlations
Fig. 4 Correlation chord diagrams between variables of different datasets. S p e a r m a n c o r r e l a t i o n s w e r e c a l c u l a t e d f o r e a c h g r o u p (LH, O B ,
D M ) separately. O n l y c o r r e l a t i o n s a m o n g variables f r o m d i f f e r e n t d a t a s e t s (clinical variables, m i c r o b i o m e , s e r u m , a n d fecal m e t a b o l o m e ) a n d
c h a r a c t e r i z e d b y |p| > 0.5 are p r e s e n t e d . Positive (A, C, E) a n d n e g a t i v e (B, D, F) c o r r e l a t i o n s are s h o w n separately. T h e c o l o r s o n t h e circuit c o d e
i n d i v i d u a l datasets, t h e c o l o r o f t h e e d g e s c o r r e s p o n d s t o o n e o f t h e d a t a s e t s t h a t are l i n k e d b y t h e e d g e . Blue: m i c r o b i o m e ; g r e e n : fecal
m e t a b o l o m e ; y e l l o w : clinical variables; violet: s e r u m m e t a b o l o m e .
SPRINGER NATURE Nutrition and Diabetes (2023)13:7
N. Ďáskova et al.
B
Q
CD
O b s e r v e d s p e c i e s
p < 0.001
13
co 1000
CD
1
A B
Timepoint
C h a o l
p < 0.001
A B
Timepoint
S h a n n o n
p = 0.016
1,
A B
Timepoint
InvSimpson
p = 0.057
Ii
rv
A B
Timepoint
Dim1 (14.6%)
*Anaerostipes
*[Eubacterium] hallii
Lactobacillus
Lachnospiraceae FCS020
*Megasphaera
*Lachnospiraceae ND3007
Actinomyces
Fusicatenibacter
(*) Blautia
Bifidobacterium
i Lachnospiraceae_unassigned
*Faecalibacterium
(*) Akkermansia
Dorea
'Coprococcus 1
*Collinsella
*Barnesiella
"Butyhcimonas
Sutterella
Lachnospiraceae NC2004
Bacteroides
Lachnospiraceae UCG-008
*Lachnospiraceae UCG-004
(*) Parabacteroides
(*) Clostridials
(*) Odoribacter
Alistipes
-1.0
Timepoint
-•- A
-•- B
-0.5 0.0
Cliffs delta
0.5 1.0
Fig. 5 Effect of inulin on fecal microbiome composition. A Alpha diversity calculated on rarefied ASV data, p value represents the result of
Wilcoxon test; B 2D PCA scores plot on genera level. The explained variance of each component is included in the axis labels. The large points
represent the centroids of each group. C Biomarker bacterial genera. Prior to all calculations, data were clr transformed. Biomarkers were
generated from univariable discriminant analysis (FDR >0.1), with effect size estimated by Cliff's delta with a 9 5 % confidence interval. A, time
point prior to intervention; B, time point post-intervention. Core genera (bold) are defined by the condition abundance >0.05% and
prevalence >75% at least in one group. Genera marked by * are confirmed butyrate producers, and genera marked by (*) are potential
butyrate producers. ASV, amplicon sequence variant.
Nutrition and Diabetes (2023)13:7 SPRINGER NATURE
N. Ďáskova et al.
W
CD
W
03
O
CD
E
C
W
CD
W
03
O
D
E
C
CD
cn
03
O
D
E
C
CD
cn
03
o
E
C
10.0
7.5
5.0
2.5
0.0
4
2
0
10.0
7.5
5.0
2.5
0.0
AUC OGTT glucose
I
-50 -30 -10 10 30
AUC OGTT insulin
-30 -10 10 30
Mcorr (FFM)
I
50
-50 -30 -10 10 30 50 70 90
C peptide
I
I
I
I
I
I
50
-50 -30 -10 10 30 50 70
100*(B-A)/A[%]
90
9-
6-
3-
0-
2hr OGTT glucose
I
-70 -50 -30 -10 10
6-
4-
2-
0Fasting
insulinemia
-70 -50 -30 -10 10 30 50 70 90
-30 -10 10
6-
4-
2-
0MCR/I
(FFM)
I
30
•
HbA1c
I
I
I
1
30
-70 -50 -30 -10 10 30 50 70 90
100 * (B-A)/A[%]
Fig. 6 Effect of inulin supplementation on selected markers of glucose metabolism. Data are expressed as the percentual change baseline
to post-intervention. A, time point prior to intervention; B, time point post-intervention. Dashed line, 0 % ; dotted lines, ±10% range.
before and after the intervention. In serum, the concentration of
butyric acid, propionic acid, and asparagine increased significantly,
whereas the concentration of glycerol and 2-propanol decreased
after inulin treatment (Fig. S15 and Table S8). In feces, three VOCs
were significantly different in abundance (p<
0.05) before and
after inulin treatment, including two propionic acid esters
(increased) and 1-hexanol (decreased) (Table S9). However, the
significance disappeared after multiple comparisons.
Prospective study: effect of inulin on glucose metabolism
Inulin intake affected markers of glucose tolerance and insulin
sensitivity, but the individual response was highly variable; we
observed positive, no, or negative changes for each of the
variables (Fig. 6 and Table S10). In the entire intervention group,
we observed a significant improvement in glucose tolerance
(120 min OGTT glucose) and a trend toward a reduction in AUC for
OGTT glucose and fasting glycemia. Skeletal muscle insulin
sensitivity, measured by glucose clamp and expressed as MCR/I
value, increased by more than 1 0 % after the intervention
compared with baseline in 14 subjects (from +11.4 to +62.4%),
whereas it did not change or decrease in 13 subjects (from +4.8 to
-48.7%). A similar distribution was observed for other indices of
insulin sensitivity (Mcorr corrected for FFM, A U C OGTT insulin, and
fasting insulinemia).
Prospective study: predictors of the metabolic effect of inulin
Because we replicated previous findings of large inter-individual
differences in metabolic responses to inulin, we sought to identify
predictors of these differences. To this end, we built linear
regression models for all glucose metabolism parameters studied
SPRINGER NATURE Nutrition and Diabetes (2023)13:7
N. Ďáskova et al.
Table 2. Predictors o f t h e inulin treatment effect o n glucose homeostasis parameters.
outcome predictor p val
ßx
ßy
p val ßy R2
A U C O G T T glucose Ruminidostridium - 4 1 . 6 0 0.015 -0.11 0.236 0.393
Lachnospiraceaejncertae sedis 40.37 0.015 - 0 . 1 8 0.044 0.249
Lachnodostridium 35.83 0.033 - 0 . 1 6 0.083 0.317
3-methyl-2-oxovalerate 37.78 0.018 - 0 . 1 8 0.044 0.326
alanine 36.28 0.024 - 0 . 1 9 0.040 0.287
ethanol -51.31 0.001 -0.21 0.008 0.501
2 h O G T T glucose A U C O G T T insulin 1.00 0.027 - 0 . 0 3 0.876 0.286
fasting insulinemia 0.91 0.037 - 0 . 1 0 0.536 0.181
H O M A INS 0.88 0.046 - 0 . 1 2 0.487 0.215
Eubacterium halii group 0.99 0.032 - 0 . 1 7 0.321 0.280
3-methyl-2-oxovalerate 1.08 0.012 - 0 . 0 9 0.585 0.287
3-hydroxyisobutyrate 0.98 0.024 - 0 . 1 4 0.392 0.278
2-oxoisocaproate 0.93 0.033 - 0 . 1 2 0.462 0.229
pyruvate 0.93 0.034 - 0 . 0 9 0.580 0.245
indole 1.33 0.002 -0.11 0.465 0.350
tridecanol 1.17 0.008 - 0 . 1 3 0.426 0.300
c-Dodecalactone 1.13 0.012 - 0 . 1 8 0.276 0.269
methyl h e p t e n o n e 1.05 0.020 - 0 . 1 6 0.360 0.217
2-undecanone 1.05 0.021 - 0 . 1 6 0.341 0.240
methyl butanal 1.02 0.024 -0.11 0.527 0.214
MCR/I (FFM) ISI (Matsuda) 0.01 0.005 - 0 . 6 7 0.008 0.313
A U C O G T T insulin -0.01 0.005 - 0 . 5 3 0.017 0.291
2 hr O G T T insulinemia -0.01 0.024 - 0 . 4 5 0.051 0.251
H O M A INS -0.01 0.027 - 0 . 5 7 0.029 0.208
fasting insulinemia -0.01 0.030 - 0 . 5 6 0.032 0.269
H O M A IR -0.01 0.036 - 0 . 5 5 0.035 0.205
IGI -0.01 0.045 - 0 . 4 0 0.080 0.189
Blautia -0.01 0.027 - 0 . 2 2 0.329 0.222
[Eubarterium] hallii group -0.01 0.030 - 0 . 1 6 0.451 0.180
asparagine 0.01 0.011 -0.31 0.121 0.230
A M c o r r (FFM) ISI (Matsuda) 1.02 0.001 - 0 . 5 3 0.025 0.474
A U C O G T T insulin - 0 . 9 0 0.002 -0.31 0.132 0.335
H O M A INS - 0 . 9 0 0.003 - 0 . 3 8 0.080 0.318
Fasting insulinemia - 0 . 8 8 0.003 - 0 . 3 7 0.090 0.326
H O M A IR - 0 . 8 6 0.006 - 0 . 3 8 0.091 0.243
IGI - 0 . 7 5 0.009 - 0 . 2 0 0.367 0.243
2 h O G T T insulinemia - 0 . 6 6 0.029 - 0 . 2 4 0.293 0.201
Dialister - 0 . 5 8 0.038 -0.01 0.979 0.172
Phascolardobacterium 0.55 0.048 0.08 0.705 0.203
asparagine 0.75 0.011 - 0 . 2 6 0.239 0.210
The data shown in the table are derived from the linear regression model described by the equation Y ( B )
- Y , A )
= p0 + p Y Y , A )
+ PxX, A )
+ e, where Y ( A )
stands for
outcome variable at time A; Y ( B )
stands for outcome variable at time B, B > A; X ( A )
stands for a standardized variable at time A representing in each model any
single clinical, metabolome or microbiome variable; px , p y are model coefficients; e stands for random error. Fecal metabolites were filtered by the condition Z
A U C X >0.1 % X A U C t o t a i across all samples; bacteria were filtered by the condition median abundance >0.1 % of the total X of bacteria across all samples.
HbAlC glycosylated hemoglobin, Mcorr glucose disposal space corrected and adjusted t o fat-free mass, MCR/I metabolic clearance rate of glucose space
corrected and adjusted to fat-free mass divided by steady-state insulinemia, OGTT oral glucose tolerance test, R2
proportion of variation in y explained by the
predictors obtained using bootstrapping (50 iterations).
as o u t c o m e variables, with all clinical or omics variables as
predictors; w e omitted variables with significant coefficients that
had high leverage (Figs. S16-S19). Despite t h e limitations of our
model, it showed several potentially interesting findings (summarized
in Table 2). For example, t h e effect o f inulin o n skeletal
muscle insulin sensitivity (Mcorr a n d MCR/I) could be predicted
from pre-intervention glycemic measures. In contrast, the MIME
predictors of the inulin effect were mostly not associated with preintervention
o u t c o m e variables. Change in A U C OGTT glucose was
negatively associated with an initial abundance of
Nutrition and Diabetes (2023)13:7 SPRINGER NATURE
N. Ďáskova et al.
Ruminiclostridium, whereas increases in Mcorr a n d MCR/I were
associated with higher initial serum asparagine (both parameters)
and lower Dialister (Mcorr) or Blautia and Eubacterium halii (MCR/I).
Initial serum concentrations of BCAA derivatives were positively
associated with increases in A U C and 2-hour OGTT glucose. All
results are summarized in Table 2.
DISCUSSION
Our main findings are: (i) obesity is the dominant factor
determining the MIME signature, whereas glycemic status has a
lesser additional influence; (ii) the metabolic response to inulin
supplementation in individuals with newly diagnosed prediabetes/diabetes
is highly variable but can be predicted, at least in
part, from baseline clinical characteristics and MIME signatures.
Indeed, more insulin-resistant individuals with poorer glycemic
indices and elevated circulating BCAA derivatives and fecal indole
and p-cresol are less likely to respond to inulin supplementation.
Observational study: gut microbiome and metabolome
Obesity is a prominent risk factor for the development of type 2
diabetes. Numerous studies have identified groups of bacterial
taxa that are enriched or depleted in obesity and type 2 diabetes,
and despite considerable heterogeneity in the results, some
common observations have been noted. First, type 2 diabetes is
associated with the depletion of potentially beneficial bacteria
rather than the presence of some dominant potentially harmful
bacteria. Second, the abundance of butyrate producers and the
functional potential for butyrate production is reduced in type 2
diabetes [10, 20, 35]. Third, the diversity of the microbiota is lower
in diseased individuals compared with healthy controls [6, 36].
Some of our results are consistent with the above, whereas
others are contradictory. In contrast to the results of Wu [10], the
change in the composition of the gut microbiota in our study was
not related to glycemic status but mainly to obesity. The dominant
butyrate producers, such as Faecalibacterium, Anaerostipes, Eubacterium
halii, or Blautia were significantly less abundant in the
microbiota of D M and OB, but w e did not detect lower SCFA
concentrations in either feces or serum. In contrast, MCFA,
nonanoic and decanoic acids were elevated in OB and DM. MCFA
can originate from dietary sources [35], but also from microbial or
yeast fermentation [37]. SCFA and MCFA have different immunomodulatory
properties; whereas SCFA attenuate inflammation,
MCFA have the opposite effect [35, 38, 39]. In addition, MCFA may
enhance intestinal permeation because of their physicochemical
properties as anionic surfactants [40]. Based on these findings, we
might suggest that it is not the lower level of SCFA but the
increased level of MCFA in the lumen that contributes to the
complications associated with obesity, such as impaired intestinal
barrier function or chronic low-grade inflammation.
Observational study: serum metabolome
The serum metabolome signature of obesity and diabetes
overlapped greatly in the study. Compared to lean subjects, both
the OB a n d D M signatures follow the same direction a n d differ
only in magnitude. The "adiposity signature," which is similar in
both OB and D M , includes SCFA (succinic and propionic acid
increased, while valeric acid decreased), aromatic A A tyrosine
(increased), and two other A A (histidine a n d asparagine,
decreased). The concentration shift of five other metabolites, i.e.,
intermediates of saccharide metabolism (glucose, lactate, and
mannose, increased) a n d A A glutamine a n d alanine, follows the
concordant direction to LH, but there is a significant difference
among all three groups. Six metabolites are specific for DM. This
signature consists of three BCAA derivatives, formic acid, 2hydroxybutyrate,
acetone, and 2-propanol.
Our findings are consistent with previously published observations
[41, 42]. Some signature metabolites could be attributed to
altered saccharide metabolism in obesity and diabetes, such as
glucose, mannose, and lactate. 2-propanol, acetone, and
2-hydroxybutyrate might be related to NADH/NAD+ redox
imbalance, which has been proposed as o n e of the features of
T2D [43].
Some other signature metabolites, i.e., SCFA and BCAA, are
located at the interface between the host and microbiota. SCFA in
serum have not previously been described as components of an
obesity-related serum signature, probably because of the analytical
difficulties associated with their determination in serum. They
are exclusively microbial products, some of which (circulating
butyric acid a n d propionic acid) have been associated with
beneficial effects [44]. Elevated circulating BCAAs have been
associated with insulin-resistance conditions such as obesity,
diabetes [45], a n d even cancer [46]. For mammals, BCAAs are
essential and must be supplied from external sources. Recent
research has deciphered the importance of the gut microbiota in
modulating the availability of many necessary compounds,
including BCAA, to the host [47].
Inulin intervention and the effects on microbiota composition
and performance
Three months of regular consumption of 10 g inulin/day was
associated with a significant shift in the composition of the
microbiota, characterized by a marked increase in potentially
beneficial bacteria, many of which are capable of butyrate
production [48]. At the same time, several bacterial taxa were
depleted, such as those associated with the fermentation of
proteins [49, 50]. This observation is largely consistent with
previously published reports [51, 52].
We did not detect a significant shift in the composition of the
fecal metabolome, although there was a non-significant trend
toward an increase in SCFA esters content. Participants were asked
not to change their dietary habits, and the only difference before
and after the intervention was the amount of inulin consumed.
This change could primarily increase the production of SCFA, but
these compounds are readily utilized by other microbes or
colonocytes at the site of their production, and only about 5 % of
SCFA are excreted in the feces [53]. A small fraction of SCFA from
the intestine may enter the bloodstream, and indeed w e observed
a significant increase in serum butyric and propionic acid
concentrations at the e n d of the intervention. Muller et al. [54]
have previously reported that it is not fecal but circulating SCFA,
particularly butyrate, that can provide a link between the gut
microbiota and whole-body insulin sensitivity. SCFA are ligands of
the G protein-coupled receptors GPR41 a n d GPR43, which are
expressed in many tissues, including adipose tissue a n d skeletal
muscle [55, 56]. Animal studies have shown that oral administration
of SCFA or intravenous infusion improves insulin sensitivity
[54].
Predicting the individual effect of an inulin intervention
The increasing understanding of the role of the microbiome in
host physiology opened new avenues for research focused o n the
possibility of predicting the outcome of a given intervention
based on the individual MIME setting. Clinically relevant results
have been obtained in cancer research, e.g., the success of therapy
with Anti-programmed Cell Death Protein-1 (PD—1) has been
shown to depend significantly on the baseline composition of the
patient's gut microbiota [57-60]. MIME has also been successfully
used to predict the response of IBD patients to a low FODMAP diet
[61] or anti-TNF treatment [62], the efficacy of synbiotic treatment
of gastrointestinal disease in children [63], or the prediction of the
clinical outcome of bariatric surgery [64]. The gut microbiota may
serve as a biomarker for selecting the most effective drugs for the
treatment of rheumatoid arthritis [65], and gut bacterial signatures
have even been described to characterize the diagnosis and
predict treatment outcomes in bipolar depression [66].
SPRINGER NATURE Nutrition a n d Diabetes (2023)13:7
N. Ďáskova et al.
Inulin-type dietary fiber is thought to alleviate several features
of metabolic syndrome; however, results from human studies are
inconsistent. A recent systematic review [67], which included 33
RCTs, showed that inulin intake (average 11 g/day) significantly
reduced blood glucose, total cholesterol, and TAG in individuals
with prediabetes and diabetes. However, a c o m m o n feature of all
included studies was the wide heterogeneity of individual
responses to treatment, making clear dietary recommendations
difficult. Therefore, w e sought to identify factors that would allow
a personalized assessment of the efficacy of inulin treatment. We
found that patients with a profile suggestive of less impaired
glucose homeostasis were likely to improve metabolically. In
addition, w e identified several other potential predictors that were
not dependent on pre-intervention glycemic indices, including
lower serum BCAA derivatives (3-methy-2-oxovalerate, 2-oxoisocaproate),
serum 3-hydroxyisobutyrate (product of NADH oxidation),
fecal indole, and/or various bacteria {Ruminiclostridium,
Lachnoclostridium, Eubacterium halii, etc.), which could allow a
more accurate prediction of inulin intervention outcomes. In the
prediabetes phase, patients are often advised to change their
lifestyle a n d diet. Despite initial adherence to advice, outcomes
may be highly variable, and patients w h o have failed despite their
best efforts may be demotivated to adhere to further recommendations.
The tool of predicting the individual appropriateness of a
particular intervention, in this case, the administration of inulin,
would help personalize treatment so that it has a higher chance of
success in potential responders and does not expose potential
non-responders to repeated failures.
Strengths and limitations of the study
There are several strengths of the study. First, the D M group
included only participants with newly diagnosed type 2 diabetes
prior and/or concomitant treatment, thus excluding confounding
effects of antidiabetic drugs on the effects of inulin. Second, we
did not rely solely on the measurement of fecal SCFA as the only
indicator of SCFA production in the colon, but used a highly
sensitive LC-MS method that allows its quantification in serum.
Third, w e evaluated the complex effects of the inulin intervention
using a multi-omics approach. Nevertheless, the study is limited
by several factors. First, w e were able to include only a limited
number of subjects, a n d the results were not validated in an
independent cohort. For this, the results were internally validated
by permutation tests. Second, the lean healthy subjects differed
from the OB or D M groups by age, because obesity and associated
comorbidities are more c o m m o n in older populations. A g e is one
of the external factors affecting microbiota composition, but this is
especially true for very young children or the elderly (over 70 years
of age). In adolescence a n d adulthood, the composition of the
microbiome is remarkably stable in terms of diversity indices, PCA
metrics, or representation of selected taxa [68-70]. Therefore, we
believe that the age difference in our population did not result in
a significant bias. Third, w e did not control dietary intake during
the prospective intervention study with inulin because w e did not
want to further burden participants a n d increase the risk of
dropping out of the study, but all participants were explicitly
asked to maintain their usual dietary habits. A n indirect measure
of adherence to the habitual diet may be the BMI of participants,
which did not change significantly during the intervention period.
Finally, the prospective study design was a single-arm, noncontrolled
intervention study, so the causality of the effect of
inulin on metabolic outcomes cannot be inferred. The small
number of participants in the prospective study did not allow us
to build more complex models to account for possible synergies
among predictors. Because the study aimed to explore predictors,
and w e found several novel biomarkers that predict response to
inulin treatment, these will need to be validated on a larger scale
in future studies.
In summary, w e showed that the gut microbiota and
metabolome profiles in OB and D M differed from those of lean
healthy individuals, whereas the differences between OB and D M
were less pronounced. We identified several omics-derived
biomarkers that may play a central role in the development of
obesity-associated metabolic changes. In patients with newly
diagnosed pre/diabetes, w e observed substantial inter-individual
variability in the effects of inulin on glucose homeostasis and
identified several predictors of treatment response. If replicated in
further studies with other populations, the identified predictors
could facilitate the estimation of inulin intervention outcomes,
paving the way for the concept of personalized dietary management
of early diabetes.
DATA AVAILABILITY
T h e s e q u e n c i n g data w e r e available at t h e S R A d a t a b a s e u n d e r t h e accession n u m b e r
PRJNA823864. T h e o t h e r datasets g e n e r a t e d a n d a n a l y z e d d u r i n g t h e current study
are available f r o m t h e c o r r e s p o n d i n g a u t h o r o n request.
REFERENCES
1. C h e n g HL, M e d l o w S, S t e i n b e c k K. T h e health c o n s e q u e n c e s o f obesity in y o u n g
a d u l t h o o d . Curr O b e s Rep. 2016;5:30-7.
2. T u r n b a u g h PJ, Ley RE, M a h o w a l d M A , M a g r i n i V, Mardis ER, G o r d o n Jl. A n obesityassociated
g u t m i c r o b i o m e w i t h increased capacity f o r energy harvest. Nature.
2006;444:1027-31.
3. Cornejo-Pareja I, M u n o z - G a r a c h A , C l e m e n t e - P o s t i g o M , T i n a h o n e s FJ. Importance
o f g u t m i c r o b i o t a in obesity. Eur J Clin Nutr. 2019;72:26-37.
4. L i m YY, L e e YS, O o i D S Q . E n g i n e e r i n g t h e g u t m i c r o b i o m e f o r t r e a t m e n t o f
obesity: a review o f current u n d e r s t a n d i n g a n d progress. B i o t e c h n o l J .
2020;15:e2000013.
5. Q i n J , Li Y, Cai Z, Li S, Z h u J , Z h a n g F, et al. A m e t a g e n o m e - w i d e association study
of g u t m i c r o b i o t a in t y p e 2 diabetes. Nature. 2012;490:55-60.
6. Nuli R, C a i J , Kadeer A , Z h a n g Y, M o h e m a i t i P. Integrative analysis t o w a r d different
g l u c o s e tolerance-related g u t m i c r o b i o t a a n d diet. Front E n d o c r i n o l .
2019;10:295.
7. Z h o u W, Sailani MR, C o n t r e p o i s K, Z h o u Y, A h a d i S, L e o p o l d SR, et al. L o n g i t u d i n a l
multi-omics o f host-microbe d y n a m i c s in prediabetes. Nature. 2019;569:663-71.
8. Forslund K, H i l d e b r a n d F, Nielsen T, Falony G, Le Chatelier E, S u n a g a w a S, et al.
D i s e n t a n g l i n g t y p e 2 diabetes a n d m e t f o r m i n t r e a t m e n t signatures in t h e h u m a n
gut m i c r o b i o t a . Nature. 2015;528:262-6.
9. Allin K H , Tremaroli V, Caesar R, J e n s e n B A H , D a m g a a r d M T F , Bahl M l , et a l .
A b e r r a n t intestinal m i c r o b i o t a in individuals with prediabetes. D i a b e t o l o g i a .
2018;61:810-20.
10. W u H, Tremaroli V, S c h m i d t C, L u n d q v i s t A, O l s s o n L M , Kramer M , et al. T h e g u t
m i c r o b i o t a in prediabetes a n d diabetes: a p o p u l a t i o n - b a s e d cross-sectional
study. Cell M e t a b . 2020;32:379-90.e3.
11. Tilg H, M o s c h e n AR. M i c r o b i o t a a n d diabetes: a n e v o l v i n g relationship. G u t .
2014;63:1513-21.
12. Z h a n g X, S h e n D, F a n g Z, Jie Z, Q i u X, Z h a n g C, et a l . H u m a n g u t m i c r o b i o t a
c h a n g e s reveal t h e p r o g r e s s i o n o f g l u c o s e i n t o l e r a n c e . P L o S O N E .
2 0 1 3 ; 8 : e 7 1 1 0 8 .
13. B h u t e SS, Suryavanshi M V , Joshi S M , Yajnik CS, S h o u c h e YS, G h a s k a d b i SS. G u t
microbial diversity assessment o f Indian type-2-diabetics reveals alterations in
Eubacteria, A r c h a e a , a n d Eukaryotes. Front M i c r o b i o l . 2017;8:214.
14. Z h a o L, L o u H, P e n g Y, C h e n S, Z h a n g Y, Li X. C o m p r e h e n s i v e relationships
b e t w e e n g u t m i c r o b i o m e a n d faecal m e t a b o l o m e in individuals with t y p e 2
diabetes a n d its c o m p l i c a t i o n s . Endocrine. 2019;66:526-37.
15. Z h o n g H, Ren H, Lu Y, Fang C, H o u G, Y a n g Z, et al. Distinct g u t m e t a g e n o m i c s
a n d m e t a p r o t e o m i c s signatures in prediabetics a n d treatment-naive t y p e 2
diabetics. E B i o M e d i c i n e . 2019;47:373-83.
16. W a n g L, Y u X, X u X, M i n g J , W a n g Z, G a o B, et al. T h e fecal m i c r o b i o t a is already
altered in n o r m o g l y c e m i c individuals w h o g o o n t o have t y p e 2 diabetes. Front
Cell Infect M i c r o b i o l . 2021;11:598672.
17. L e t c h u m a n a n G, A b d u l l a h N, Marlini M , B a h a r o m N, Lawley B, O m a r MR, et al. G u t
m i c r o b i o t a c o m p o s i t i o n in prediabetes a n d newly d i a g n o s e d t y p e 2 diabetes: a
systematic review o f observational studies. Front Cell Infect M i c r o b i o l .
2022;12:943427.
18. Gaike A H , Paul D, Bhute S, Dhotre DP, P a n d e P, Upadhyaya S, et al. T h e gut microbial
diversity o f newly diagnosed diabetics b u t n o t o f prediabetics is significantly different
f r o m that o f healthy nondiabetics. mSystems. 2020;5:e00578-19.
Nutrition a n d Diabetes (2023)13:7 SPRINGER NATURE
N. Ďáskova et al.
19. Li L, Li C, Lv M, H u Q, G u o L, X i o n g D. Correlation b e t w e e n alterations o f gut
m i c r o b i o t a a n d miR-122-5p expression in patients with t y p e 2 diabetes mellitus.
A n n Transl M e d . 2020;8:1481.
20. Karlsson FH, Tremaroli V, N o o k a e w I, B e r g s t r o m G, Behre CJ, Fagerberg B, et al.
Gut m e t a g e n o m e in European w o m e n with n o r m a l , i m p a i r e d a n d diabetic g l u cose
control. Nature. 2013;498:99-103.
21. Diener C, Reyes-Escogido ML, Jimenez-Ceja L M , Matus M, G o m e z - N a v a r r o C M ,
C h u N D , et al. Progressive shifts in t h e g u t m i c r o b i o m e reflect prediabetes a n d
diabetes d e v e l o p m e n t in a treatment-naive M e x i c a n cohort. Front E n d o c r i n o l .
2020;11:602326.
22. G h a e m i F, Fateh A, S e p a h y A A , Z a n g e n e h M, G h a n e i M , Siadat SD. Intestinal
m i c r o b i o t a c o m p o s i t i o n in Iranian diabetic, pre-diabetic a n d healthy individuals. J
Diabetes M e t a b Disord. 2020;19:1199-203.
23. C h e n PC, Chien YW, Y a n g SC. The alteration o f g u t m i c r o b i o t a in newly d i a g n o s e d
type 2 diabetic patients. Nutrition. 2019;63-64:51-6.
24. P r o c h a z k o v a M, Budinska E, K u z m a M, Pelantova H, Hradecky J , H e c z k o v a M , et al.
V e g a n diet is associated with favorable effects on t h e m e t a b o l i c p e r f o r m a n c e of
intestinal m i c r o b i o t a : a cross-sectional multi-omics study. Front Nutr.
2021;8:783302.
25. C o l a n t o n i o A G , W e r n e r SL, B r o w n M. The effects o f prebiotics a n d substances
with prebiotic properties on m e t a b o l i c a n d i n f l a m m a t o r y biomarkers in individuals
with t y p e 2 diabetes mellitus: a systematic review. J A c a d Nutr Diet.
2020;120:587-607.e2.
26. Davis L M , Martinez I, Walter J , G o i n C, Hutkins RW. B a r c o d e d p y r o s e q u e n c i n g
reveals that c o n s u m p t i o n o f galactooligosaccharides results in a highly specific
bifidogenic response in h u m a n s . PLoS ONE. 201 l;6:e25200.
27. Z h a o L, Z h a n g F, D i n g X, W u G, L a m YY, W a n g X, et al. G u t bacteria selectively
p r o m o t e d by dietary fibers alleviate t y p e 2 diabetes. Science. 2018;359:1151-6.
28. W a r e h a m NJ. Personalised prevention of t y p e 2 diabetes. D i a b e t o l o g i a .
2022;65:1796-1803.
29. Le DS, Brookshire T, Krakoff J , Bunt JC. Repeatability a n d reproducibility of the
hyperinsulinemic-euglycemic c l a m p a n d the tracer dilution t e c h n i q u e in a c o n trolled
inpatient setting. M e t a b o l i s m . 2009;58:304-10.
30. D e F r o n z o RA, T o b i n JD, A n d r e s R. G l u c o s e c l a m p t e c h n i q u e : a m e t h o d for
quantifying insulin secretion a n d resistance. A m J Physiol. 1979;237:E214-23.
31. Callahan BJ, M c M u r d i e PJ, Rosen MJ, Han A W , J o h n s o n AJ, H o l m e s SP. D A D A 2 :
high-resolution s a m p l e inference f r o m lllumina a m p l i c o n data. Nat M e t h o d s .
2016;13:581-3.
32. Han J, Lin K, Sequeira C, Borchers C H . A n isotope-labeled c h e m i c a l derivatization
m e t h o d for t h e quantitation o f short-chain fatty acids in h u m a n feces by liquid
c h r o m a t o g r a p h y - t a n d e m mass spectrometry. A n a l C h i m Acta. 2015;854:86-94.
33. Dieterle F, Ross A, Schlotterbeck G, S e n n H. Probabilistic q u o t i e n t n o r m a l i z a t i o n
as robust m e t h o d to a c c o u n t for dilution o f c o m p l e x biological mixtures.
A p p l i c a t i o n in 1H N M R m e t a b o n o m i c s . A n a l C h e m . 2006;78:4281-90.
34. R C o r e T e a m . R: a l a n g u a g e a n d e n v i r o n m e n t for statistical c o m p u t i n g . Vienna,
Austria: R F o u n d a t i o n for Statistical C o m p u t i n g . 2017.
35. S a m Q H , Ling H, Y e w WS, Tan Z, Ravikumar S, C h a n g M W , et al. The divergent
i m m u n o m o d u l a t o r y effects o f short chain fatty acids a n d m e d i u m chain fatty
acids. Int J M o l Sei. 2021;22:6453.
36. Le Chatelier E, Nielsen T, Qin J, Prifti E, H i l d e b r a n d F, Falony G, et al. Richness of
h u m a n g u t m i c r o b i o m e correlates w i t h m e t a b o l i c markers. Nature.
2013;500:541-6.
37. Nissen L, Samaei SP, Babini E, Gianotti A. G l u t e n free s o u r d o u g h b r e a d e n r i c h e d
with cricket flour for protein fortification: A n t i o x i d a n t i m p r o v e m e n t a n d Volatil
o m e characterization. F o o d C h e m . 2020;333:127410.
38. L u s c o m b e VB, Lucy D, Bataille CJR, Russell AJ, Greaves DR. 20 years an o r p h a n : is
G P R 8 4 a plausible m e d i u m - c h a i n fatty acid-sensing receptor? D N A Cell Biol.
2020;39:1926-37.
39. Saresella M, M a r v e n t a n o I, Barone M, La Rosa F, Piancone F, M e n d o z z i L, et al.
Alterations in circulating fatty acid are associated with g u t m i c r o b i o t a dysbiosis
a n d i n f l a m m a t i o n in m u l t i p l e sclerosis. Front I m m u n o l . 2020;11:1390.
40. Brayden DJ, Mäher S, Bahar B, Walsh E. S o d i u m caprate-induced increases in
intestinal permeability a n d epithelial d a m a g e are p r e v e n t e d by m i s o p r o s t o l . Eur J
P h a r m B i o p h a r m . 2015;94:194-206.
41. Halama A, S u l e i m a n N N , Kulinski M , Bettahi I, Hassoun S, A l k a s e m M, et al. The
m e t a b o l i c footprint o f c o m p r o m i s e d insulin sensitivity u n d e r fasting a n d
hyperinsulinemic-euglycemic c l a m p c o n d i t i o n s in an A r a b p o p u l a t i o n . Sei Rep.
2020;10:17164.
42. Gall WE, B e e b e K, L a w t o n KA, A d a m KP, Mitchell M W , Nakhle PJ, et al. alphahydroxybutyrate
is an early b i o m a r k e r o f insulin resistance a n d g l u c o s e intolera
n c e in a n o n d i a b e t i c p o p u l a t i o n . PLoS ONE. 2 0 1 0 ; 5 : e l 0 8 8 3 .
43. S o n g J, Y a n g X, Yan LJ. Role o f p s e u d o h y p o x i a in t h e p a t h o g e n e s i s o f t y p e 2
diabetes. H y p o x i a . 2019;7:33-40.
44. K u m a r J , Rani K, Datt C. M o l e c u l a r link b e t w e e n dietary fibre, gut m i c r o b i o t a a n d
health. M o l Biol Rep. 2020;47:6229-37.
45. N e w g a r d CB. Interplay b e t w e e n lipids a n d branched-chain a m i n o acids in
d e v e l o p m e n t of insulin resistance. Cell M e t a b . 2012;15:606-14.
46. Rossmeislova L, G o j d a J , S m o l k o v a K. Pancreatic cancer: branched-chain a m i n o
acids as putative key m e t a b o l i c regulators? Cancer Metastasis Rev.
2021;40:1115-39.
47. Gojda J, C a h o v a M. G u t m i c r o b i o t a as t h e link b e t w e e n elevated B C A A s e r u m
levels a n d insulin resistance. Biomolecules. 2021;11:1414.
48. Le Bastard Q, C h a p e l e t G, J a v a u d i n F, Lepelletier D, Batard E, Montassier E. The
effects of inulin o n g u t microbial c o m p o s i t i o n : a systematic review of e v i d e n c e
f r o m h u m a n studies. Eur J Clin M i c r o b i o l Infect Dis. 2020;39:403-13.
49. Diether NE, Willing BP. Microbial f e r m e n t a t i o n o f dietary p r o t e i n : an i m p o r t a n t
factor in diet(-)microbe(-)host interaction. M i c r o o r g a n i s m s . 2019;7:19.
50. Yao CK, M u i r JG, G i b s o n PR. Review article: insights into c o l o n i c protein ferm
e n t a t i o n , its m o d u l a t i o n a n d potential health implications. A l i m e n t P h a r m Ther.
2016;43:181-96.
51. V a n d e p u t t e D, Tito RY, V a n l e e u w e n R, Falony G, Raes J. Practical considerations
for large-scale gut m i c r o b i o m e studies. F E M S M i c r o b i o l Rev. 2017;41:S154-S67.
52. Healey G, M u r p h y R, Butts C, B r o u g h L, W h e l a n K, C o a d J . Habitual dietary fibre
intake influences g u t m i c r o b i o t a response t o an inulin-type fructan prebiotic: a
r a n d o m i s e d , d o u b l e - b l i n d , placebo-controlled, cross-over, h u m a n intervention
study. Br J Nutr. 2018;119:176-89.
53. R e c h k e m m e r G, R o n n a u K, v o n Engelhardt W. F e r m e n t a t i o n o f polysaccharides
a n d a b s o r p t i o n o f short chain fatty acids in t h e m a m m a l i a n h i n d g u t . C o m p
B i o c h e m Physiol A C o m p Physiol. 1988;90:563-8.
54. Muller M , H e r n a n d e z M A G , G o o s s e n s G H , Reijnders D, Holst JJ, J o c k e n JWE, et al.
Circulating but not faecal short-chain fatty acids are related t o insulin sensitivity,
lipolysis a n d GLP-1 concentrations in h u m a n s . Sei Rep. 2019;9:12515.
55. T a n g C, A h m e d K, Gille A, Lu S, G r o n e HJ, Tunaru S, et al. Loss of F F A 2 a n d FFA3
increases insulin secretion a n d i m p r o v e s g l u c o s e tolerance in t y p e 2 diabetes.
Nat M e d . 2015;21:173-7.
56. Koh A, De V a d d e r F, Kovatcheva-Datchary P, B a c k h e d F. F r o m dietary fiber to host
physiology: short-chain fatty acids as key bacterial metabolites. Cell.
2016;165:1332-45.
57. M a o J, W a n g D, L o n g J, Y a n g X, Lin J, S o n g Y, et al. Gut m i c r o b i o m e is associated
with the clinical response t o anti-PD-1 based i m m u n o t h e r a p y in hepatobiliary
cancers. J I m m u n o t h e r Cancer. 2021;9:e003334.
58. F a h r m a n n JF, Saini NY, Chia-Chi C, Irajizad E, Strati P, Nair R, et al. A polyaminecentric,
b l o o d - b a s e d metabolite panel predictive o f p o o r response t o CAR-T cell
therapy in large B cell l y m p h o m a . Cell Rep. M e d . 2022;3:100720.
59. Sannicolo S, Giaj Levra M , Le Gouellec A, A s p o r d C, Boccard J, C h a p e r o t L, et al.
Identification o f a predictive m e t a b o l i c signature of response t o i m m u n e
c h e c k p o i n t inhibitors in non-small cell lung cancer: METABO-ICI clinical study
protocol. Respir M e d Res. 2021 ;80:100845.
60. M c C u l l o c h JA, Davar D, Rodrigues RR, Badger JH, Fang JR, C o l e A M , et al.
Intestinal microbiota signatures o f clinical response a n d immune-related adverse
events in m e l a n o m a patients treated with anti-PD-1. Nat M e d . 2022;28:545-56.
61. Vervier K, M o s s S, K u m a r N, A d o u m A, Barne M, B r o w n e H, et al. T w o microbiota
subtypes identified in irritable b o w e l s y n d r o m e with distinct responses to the
low F O D M A P diet. Gut. 2022;71:1821-30.
62. Busquets D, Oliver L, A m o e d o J , Ramio-Pujol S, M a l a g o n M, Serrano M, et al. RAID
prediction: pilot study o f fecal microbial signature with capacity to predict
response to anti-TNF treatment. Inflamm B o w e l Dis. 2021;27:S63-S6.
63. Tierney BT, Versalovic J, Fasano A, Petrosino JF, C h u m p i t a z i BP, Mayer EA, et al.
Functional response to a microbial synbiotic in the gastrointestinal system of
children: a r a n d o m i z e d clinical trial. Pediatr Res. 2022. https://doi.org/10.1038/
S41390-022-02289-0.
64. Vaz M, Pereira SS, M o n t e i r o M P . M e t a b o l o m i c signatures after bariatric surgery - a
systematic review. Rev E n d o c r M e t a b Disord. 2022;23:503-19.
65. Wei M, C h u C Q . Prediction o f t r e a t m e n t response: p e r s o n a l i z e d m e d i c i n e in the
m a n a g e m e n t o f r h e u m a t o i d arthritis. Best Pr Res Clin R h e u m a t o l .
2022;36:101741.
66. Lai J , Li A, Jiang J , Yuan X, Z h a n g P, Xi C, et al. M e t a g e n o m i c analysis reveals gut
bacterial signatures for diagnosis a n d t r e a t m e n t o u t c o m e prediction in bipolar
depression. Psychiatry Res. 2022;307:114326.
67. Li L, Li P, X u L. Assessing the effects o f inulin-type fructan intake on b o d y weight,
b l o o d glucose, a n d lipid profile: a systematic review a n d meta-analysis o f rand
o m i z e d c o n t r o l l e d trials. F o o d Sei Nutr. 2021;9:4598-616.
68. O d a m a k i T, Kato K, Sugahara H, Hashikura N, Takahashi S, Xiao JZ, et al. Agerelated
changes in g u t m i c r o b i o t a c o m p o s i t i o n f r o m n e w b o r n t o centenarian: a
cross-sectional study. B M C M i c r o b i o l . 2016;16:90.
69. R o d r i g u e z J M , M u r p h y K, Stanton C, Ross RP, Kober O l , J u g e N, et al. The c o m position
o f the g u t m i c r o b i o t a t h r o u g h o u t life, w i t h an e m p h a s i s on early life.
M i c r o Ecol Health Dis. 2015;26:26050.
70. Salazar N, G o n z a l e z S, N o g a c k a A M , Rios-Covian D, A r b o l e y a S, G u e i m o n d e M,
et al. M i c r o b i o m e : effects o f a g e i n g a n d diet. Curr Issues M o l Biol. 2020;36:33-62.
SPRINGER NATURE Nutrition and Diabetes (2023)13:7
N. Ďáskova et al.
71. Feins EN, Ireland C, G a u v r e a u K, C h a v e z M , Callahan R, Jenkins KJ, et al. P u l m
o n a r y vein stenosis: a n a t o m i c considerations, surgical m a n a g e m e n t , a n d o u t c
o m e s . J Thorac Cardiovasc Surg. 2022;163:2198-207.e3.
ACKNOWLEDGEMENTS
This study w a s s u p p o r t e d by t h e Ministry o f Health o f t h e C z e c h Republic, grant NV-
18-01-00040 a n d b y t h e project National Institute for Research o f M e t a b o l i c a n d
Cardiovascular Diseases (Program EXCELES, Project N o . L X 2 2 N P O 5 1 0 4 ) — F u n d e d b y
t h e E u r o p e a n U n i o n — N e x t G e n e r a t i o n EU. Jan G o j d a is s u p p o r t e d by EFSD
m e n t o r s h i p p r o g r a m s u p p o r t e d by A s t r a Z e n e c a .
13
ADDITIONAL INFORMATION
Supplementary information T h e o n l i n e version contains s u p p l e m e n t a r y material
available at https://doi.org/10.1038/s41387-023-00235-5.
Correspondence a n d requests for materials s h o u l d b e a d d r e s s e d t o M . Cahovä.
Reprints and permission information is available at http://www.nature.com/
reprints
Publisher's note Springer Nature remains neutral with regard t o jurisdictional claims
in p u b l i s h e d m a p s a n d institutional affiliations.
AUTHOR CONTRIBUTIONS
N D , TK, M C , a n d JG w e r e i n v o l v e d in t h e study c o n c e p t a n d d e s i g n . N D , M P , M K , HP,
JH, M H e c , M B , PV, PS, M H e n , A O , JP, a n d RL participated in t h e acquisition o f data.
N D , IM, a n d M K w e r e i n v o l v e d in statistical analysis. N D , IM, M K , HP, RL, TK, M K , M C ,
a n d JG participated in interpretation o f data. N D , M C , T K a n d JG drafted t h e
manuscript. All authors participated in t h e critical revision o f t h e m a n u s c r i p t a n d
a p p r o v e d t h e final version o f t h e manuscript. T h e senior a n d c o r r e s p o n d i n g authors
(JG, TK, a n d M C ) take full responsibility f o r t h e w o r k and/or t h e c o n d u c t o f t h e study,
they h a d access t o t h e data a n d c o n t r o l l e d t h e decision t o publish.
CONFLICT OF INTEREST
T h e authors declare n o c o m p e t i n g interests.
O p e n Access This article is licensed u n d e r a Creative C o m m o n s
Attribution 4.0 International License, w h i c h permits use, sharing,
a d a p t a t i o n , distribution a n d r e p r o d u c t i o n in a n y m e d i u m or format, as long as y o u give
a p p r o p r i a t e credit t o t h e original author(s) a n d t h e source, p r o v i d e a link t o t h e Creative
C o m m o n s license, a n d indicate if c h a n g e s w e r e m a d e . T h e images or other t h i r d party
material in this article are i n c l u d e d in t h e article's Creative C o m m o n s license, unless
i n d i c a t e d o t h e r w i s e in a credit line t o t h e material. If material is n o t i n c l u d e d in t h e
article's Creative C o m m o n s license a n d your i n t e n d e d use is n o t p e r m i t t e d by statutory
regulation or exceeds t h e p e r m i t t e d use, y o u will n e e d t o o b t a i n p e r m i s s i o n directly
f r o m t h e c o p y r i g h t holder. T o v i e w a c o p y o f this license, visit http://
c r e a t i v e c o m m o n s . o r g / l i c e n ses/by/4.0/.
© T h e Author(s) 2 0 2 3
Nutrition and Diabetes (2023)13:7 SPRINGER NATURE